Mathematical Analysis- m

'MEASURE THEORY
(Including Banach Space, Hilbert Space, Spectral Theory)
[For Honours and Post-Graduate Students]

By
KESHA WA PRASAD GUPTA
B. Sc. HO;IS., [Link].
Department of IIfathematic!>,
P.P.N. College, Kanpur.

KRISHN A Prakashan Mandir

11. Shivaji Road. Meerut-250001 (U. P.) INDIA.
 Our Most Pop�lar Publications on Mathematics

A. R. Vasishtha J. N. Sharma & A. R. Vasishtha

1. Matrices. 2l. Functional Analysis.
2. Modern Algebra.
22. Linear Algebra.
3. Vector A lgebra .
23. Mathematical Analysis-II.
4. Theory of Numbers.
24. Real Analysis.
A. R. Vasisht ha & D. C. Agarwal
25. Vector Calculus.
5. Analytkal Solid Geometry
.I. N. Sharma & Dr. R. K. Gupta
6. Dynamics of a Particle. 26. Differential Equations.
7. Co-ordinate Solid Geometry 27. Differential Equations
(Fully Solved) (With Special Functions).
A. R. Vasishtha & Dr. R. K. GUpta 28. Mathematical Methods.
8. Integral Transforms. 29. Special Functions.

D. C. Agarwal K. P. Gupta

9. Advanced Integral Calculus. 30. Measure Theory (Analysis-III).

10. Tensor Calculus & Riemannian 31. Set Theory and Related Topics.

Geometry Dr. Mittal & Agarwal

Dr. Goyal & Gupta 32. Differential Geometry.
11. Statics. Dr. P. P. Gupta
12. Theory of Relativity. 33. Electro-Magnetic Theory.
Dr. Goyal & Sharma 34. Rigid Dynamics- I .

13. Mathematical Stati st ics. 35. Rigid Dynamics-U.

Dr. GUpta & Malik Dr. R. K. Gup ta

14. Calculus of Finite Differen ces and 36. Linear Programming.

Numerical Analysis. 37. Operations Research.

Dr. Shanti Swamp
Sharma, Gupta & Kumar
38. Fluid Dynamics.
i5. Spherical Astronomy.
39. Hydrodynamics.
J. N. Sharma
40. In t e gral E qua ti ons .

16. Advanced Differential Calculus.

Sharma, Gupta & Agarwal
17. Funct ions of a Complex Variable.
41. Linear Difference Equations.
18. Infinite Series & Products.
19. Metric Spaces (Analysis-I).
20. Topology.

MEASURE THEORY
Thirteenth E dition 1994-95
All rights reserved by the author.

Price Rs. 55'00 only.

Published by KRISHNA Prakashan Mandir, 11, Shivaji Road, Meerut-l (U.P.)

India and Printed at Fine Art Press, Meerut.
 PREFACE

How to define the measure of a set in a pure Mathematics is

a modern concept and it is needless to state that there is hardly
a book by any Indian author on this topic. Hence the necessity
of writing such a book under the caption, 'The Theory of
Lebesgue Measure and Integration' was taken because of the huge
and unbearable cost of foreign books on the subject.

The subject being very modern in nature, all needed defini­

tions have been given at the beginning of the chapters, and though
apparently the volume looks very thin but it covers the entire
course as laid down in various Indian Universities.
It is needless to emphasise that the style adopted in the book
is lucid, cleat, easy and clearly understandable to the students. A
good many solved and unsolved examples have been given in every
chapter so that students may have enough practice in the subject.
I am grateful to Dr. (Mrs.) P. Srivastava, Reader in Mathe­
matics, B. H. U., Dr. S. N. Lal and Sri Ram Bechan Ram
Lecturers in Maths., B. H. U., for their valuable guidance. I am
also thankful to my friends and well wishers Dr. M. P. Jaiswal,
Prof. R. G. Gupta, S. P. S. Bhadauria for their valuable sugges­
tions. Thanks are also due to the printers and publishers who
took pains in bringing out this work.
This book may have some defects and I shall thankfully
receive constructive suggestions for its improvement from all those
interested in studying the subject.
Kanpur K. P. GUPTA

PREFACE TO THE THIRTEENTH EDITION

The present edition has been thoroughly revised and enlarged.

This edition solves the purpose of Measure Theory and Functional
Analysis (Modern Analysis) both. All the chapters have also gone
Substantial changes. An attempt has been made to remove any
type of mistakes of previous edition.

K. P. GUPTA
 'l'f''f'!ff'flm'l fircrivr~ I
fet!' CltTT"~ ~~ ~~~ II
f~ f~~U( fiffet~~ fifU~ I
f~q"~mn~m ~s~ II

We adure you the guardian of the south-east

quarter and Ruler of the whole uni~'erse, eternal bliss
personified, the omnipresent and all pervading Brahma
manifest in the form of Vedas. We worship Lord Shiva
shining in His own glory, devoid of material attributes,
undifferentiated, desire less, all pervading consciousness,
having nothing to wrap about hifl1se/f except ether.

- Authors and Publishers

 Contents
Chapters Pages
1. Basic concepts of set 1
2. Basic set operations 6
3. Functions and sequences 11
4. Bounded. derived, open and closed sets on the real line 16
5. Countability of sets 21
6. Measure and outer measure 46
Ring of sets, a-ring of sets 46
Algebra of sets 47
a-algebra of sets. Finitely additive 47
Translation invaJ iant 49
Postulates for all ideal measure function 49
Measure space 50
Completeness of measure function 50
Caratheodory's postulate for outer measun: 50
Measurable set 51
Problems related to measure functions 51
Problems related to ring of sets, a-algebra 63
7. Lebesgue measure of a set 73
Measure of open and closed intervals 73
Measure of rectangle and a parallelopiped 74
Exterior and Interior measure 74
Some Theorems and solved problems 75
Equivalence of Lebesgue-exterior measure and Caratheodory
outer measure 95
Fundamental Theorems 99
Limiting sets 100
Covering in the sense of Vitali 106
Vitali's covering Theorem 107
Sets of the Type Fa and Gil 110
Borel measurable set 110
Unsolved problems III
8. Measurable functions 115
Measurable functions. Almo,l cverywerc lI5
Equivalent functions Characteristic function 115
Simple function. Limit supelior and limit inferior 116
Some Theorems and solved problems 117
Lebesgue measurable functions 118
Borel mensurability of functions 136
Little Wood's three principles 136
 ( ,i )

Chapters Pages
9. The Lebesgue integral of a function 140
Lebesgue integral of a function 140
Lebesgue integral of an unbounded function 144
First mean value Theorem 147
Countable additivity property of the integral 150
10. Theorem on convergence of sequences of measurable
Functions 176
Convergence in measure 176
F. Reisz's Theorem 178
D. F. Egor's Theorem 179
Lebesgue bounded convergence theorem 181
Lebesgue's dominated convergence theorem 183
Beppo-Levis's Theorem 185
Faton's Lemma 186
11. Absolute continuous functions. Indefinite integral and
Differentiation 194
Continuous function, Absolute continuous function 194
Indefinite integral. Dift'erentiable 195
Increasing and decreasing functions 195
Functions of bounded variation 195
Lebesgue point 196
Problems related to functions of bounded variation 197
Problems related to absolute continuous functions 205
Problems related to indefinite integral 214
Fundamental theorem of integral calculus 219
Problcms related to Lebesguc point of a function 220
12. V-Space 225
Conjugate number, LP-space 225
Norm of an elemcnt of LP-space 225
Convergence sequence. Cauchy sequence 226
Completeness of Ll'-space, Banach space 226
Some theorems 226
Holder's inequality 229
Minkowski's inequality 231
Schwarz's inequality 232
LP-space is a normcd linear space 237
Reisz Fisher theorem 238
Some theorems 241
13. Further theorems on Lebesgue Integration 246
Integration by parts 246
Second mean value theorem; 248
Relation between Lebesgue integral and Stieltjes integral 250
Cumulative distribution function 254
Lebesgue-St ieitjes integral 255
 ( vii )

Chapters Pages
14. The Weierstrass approximation theorem and semi-
continuous function 256
Bernstein polynomial 256
S. N. Bernstein's theorem 256
Weierstrass approximation theorem 260
Semi-continuous function 261
Theorem on semi-continuous functions 262
15. Signed measure 265
Signed measure. Positive and negative sets 266
Theorems on positive and negative sets 267
Hahn decomposition theorem 272
Singular measures 273
Jordon decomposition. Absolutely continuous measure function 274
Radon Nikodym theorem 274
Lebesgue decomposition theorem 277
16. Product measure 280
17. Fourier series 284
Periodic function. Trigonometric series 284
Finite discontinuity 284
Fourier series and Fourier Coefficients. Even and odd functions 285
Parsevel's identity for Fourier series " 292
Riemann--Lebesgue Theorem 294
Dirichlet's integral 295
Summation of series by arithmetic means 297
Fejer's integral 297
Summability of Fourier series. Fejer Theorem 298
18. Banach space 301
19. Hilbert Space 352
20. Finite Dimensional Spectral Theory 416
 SYMBOLS AND THEIR MEANING
Symbol Meaning
3 there exist or there exists
=> implies
is implied by
implies and is implied by
iff if and only if
& and
¥ for every
E belongs to
rf; does not belong to
:::> is a superset of
c ·is a subset of
o null set
N Set of natural numbers
R Set of real numbers
R+ set of positive real numbers
Q set of rational numbers
Def. Definition
Q+ set of positive rational numbers
s.t. such that
U Union
n intersection
is cardinally equivalent to
<an> sequence consisting of points a" a2. a3, a4, ...
<an: n E N> sequence consisting of points at, a2, 03, ...
e.g. for example
w.r.t. with respect to
an~a the sequence <all> converges in limit to a
!no (S) outer measure of a set S
me (8) Exterior measure of a set S
mi (8) Interior measure of a set
In (8) measure of a set S
L-integrable Integrablelin the sense of Lebesgue
R-integrable Integrable in the sense of Riemann
E (f(x) > a) {x E E : f(x) > a}
t f(x) dx
a.e.
Lebesgue integral orr(x) over E
almost everywhere
AI complement of A
Tx T(x)
 1
Basic Concepts of Set

The basic idea of a set is the origin point in the study of

Modern Algebra.
1'0. Der. A set is a collection of well distinct objects.
The objects of a set are called the elements or members of that
set and their membership is defined by certain conditions. The ele-
ments of a set can be anything: mango, pen, sun, moon, river etc.
Examples. (i) The set of interior points of a circle of unit
radius with centre at the origin.
(ii) The set of pen, ihkpot, pencil.
(iii) The set of Ram, Mohan, Shyam, Bhole Shanker.
(iv) The students of Banaras Hindu University.
(v) Collection of letters a, b, c, d, e forms a set.
Remark. (i) Set and aggregate both have the same mean-
ing.
(ii) The elements of a set must be distinguished from one
another.
1'1. Notation. S~ts are usually denoted by capital letters
A, D, C, D, E, P, Q, X, ...
and their elements are denoted by corresponding small letters
a, b, c, d, e, p, q, x, ...
It is not necessary that the elements of a set A are denoted by a.
We can also write the symbol b for the general element of the
set A.
If a is an element of a set A, then this fact is denoted by the
symbol a E A. The symbol E is used for anyone of the follow-
ing phrases:
(i) belong to. (ii) belongs to. (Iii) is in. (iv) are in.
"belongs to", determined from the context "a E A" is read as
"a belongs to A".
If a is not a member of A, then we shall write a rt. A
2 BASIC CONCEPfS OF SET

It is customary to put a vertical line "I" or. an oblique "I"

througt. a symbol to indicate the opposite meaning of the symbol.
If we write "Let x E X", then the meaning of this is
"Let x be an element of X".
Remark. (i) It should be noted that order is not preserved
in case of set, where as the Older is necessarily preserved in case
of a sequence. That is to say, each of the sets {I, 2, 3}, {3, 2, I},
{t, 3, 2} denotes the same set.
(ii) The repetition of an element does not change the nature
of a set, i.e., each of the sets {I, 2, 2, 3}, {I, 3, }, 2}, {I, 2, 2,3, 3, 3}
{t, 2, 3} denotes the same set.
1'2. Set of sets. It may happen that the elements of a set
are sets themselves. A set consisting of a number of sets is called
set of sets. Set of set,; is also expr.:ss·ed by saying "family of sets"
or "class of sets". In order to make a distinction between a set
and a family of sets, we use bold letters.
A, D, C, P, ...... in case or family of sets.
For example {A, B, C, D, E} is a family of sets.
1'3. Subset. If every element of A is also an element of a set
B, then A is called a subset of B. This relationship between A
and B is denoled by the sFmbo I
A C R and is read as "A is a subset of B".
or "A is contained in R".
"Let A C !l" is read as
.. Let A be a suhset of B".
Symbolically A is called a sub,et of B iff
any a E A ~ a E B.
Examples. (i) {a, {b}, {c}, {b, c} are subsets of the set
la, b, c}.
(ii) Evidently
any a E A ~ a E A.
Hence, by def. of subset,
A C A for every set A.
1'4. Super set. A C B is also expressed by writing R::> A
and is read as "B contains A" or "B is a superset of A".
1'5. Equality of sets. Two sets A and B are said to be equal
if every elemeut of A belongs to B and every element of R :l1ongs
to A. This relationship between A and B is denoted by the
symbol A= B.
[Link] CONCEPTS OF SE r 3

Sy,mbolically A=B iff

any x E A ~ x E B.
Alternately
A C B, B C A {:> A=B.
If A=B is not true, then we write A::;eB.
1'6. Proper subset. A is called a proper subset of B if A c B
and A::;eB and this relationship is denoted by the symbol ..4 ~ B
and is read as "A is a proper subset of B".
From this definition it follows that every element of If is an
element of Band B contains at least an element which does not
belong to A.
Examples. (i) {a, b} is a proper subset of la, b, c}.
(ii) Since A C A and A =A and therefore A is not a nroper
subset of A, i.e.,
A k A is not true for every set A
1'7. Finite set. A set consisting of a finite number of ele-
ments is called a finite set.
Example. {a, b}, {I, i, 3}, {a1' a2, aa, ...... , a.} are aU finite
sets.
1'8. Infinite set. A set consisting of an infinite number of
elements is called an infinite set e.g., N, Q, R, Z, ... are all
infinite sets.
1-9. Null set. A set containing no element is called a null
set and is denoted by symbol.p. It is also called empty set or
void set.
Examples. (i) {x: x:;t:x1=~.
For x::;ex is not possible for any element x.
(ii) {x E R: X2 < O}=.fo.
For the square of any real quantity ;;iii 0 and so x· < 0 is
impossible for any x E R.
Remark. There will be only one empty set and so we shall
always speak of the empty set instead of an empty set.
1'10. Power set. The family consisting of all subsets of a
set is called the power set ot'that set and is denoted by P, e.g.
{B: B C A} is the power set of A and so we write
P={B: B C A}.
Evidently the value of P varies with the value of A, i.e., P depends
upon A and so we denore the power set of A by the symbol P (A).
Finally,
P (A)={B : B C AJ.
 4 BASIC CONCEPTS OF SET

e.g. it A={l, 2, 3}, then

P (A)={ {I}. {2}, {3}, {I, 2}, {2, 3}, {3, I}, {I, 2, :}, q, }

Ex. 1. Prove that, C A for every set A.

Solution. Since, does not contain any element and therefore
the condition
any x E ' " x E A
is fulfilled and hence the result follows.
Ex. 2. If a set A contains n elements, then P (A) contains 2"
elements.
Solution. The total number of elements of P (A)
=nCO+nC1 ~-"C2+···+nc"
=(1 + I)", by Binomial expansion
=2".
1'11. Universal set. If all the sets under consideration are
subsets of a fixed set, then this fixed set is called universal set or
universe of discourse and is denoted by the symbol X or by U.
Singleton set. A setjconsisting of only one element is called
a singleton set e.g. each 6f the following sets is a singleton set
{a}, {b}, {I}, {2}.
1'12. Indexed set and Index set. Let At be a non-empty set
for each r in a set 6. In this ca~e the sets AI' A 2 , 04 3 , ...... , A.
are called indexed sets and the sets 6.={l, 2,3, ...... , n} is called
irdex set. Here the suffix r E II of Ar is called all index. Such a
family of sel." is denoted by
{Ar : r E 6.} or {Ar} r E 6.
Example. Let A 1 ={a, b, c}, A 2 =:c, d, /, m}, A3={m, n, q},
Ac={p, q, r, sl, A.;=(a, I, d, f}, 6={1, 2, 3, 4, 5}.
Here we find that 3 a non-emp y set At "f r E 6.. Hence 6
is called an index set and the sets AI' A!, A3 , AI' A; are c:.llied
indexed sets.
1'13. Heriditary property. A nOll-empty family A={Ar} of
sets is said to be heriditar y if
Ar C A., A, E A => Ar E A.
1'14. Definition. A family {A.} of sets i5 sa id to be
p.·rwise disjoint if
Ar n ..4.=, ¥ r, sEll S.t. r=l=s.
BASIC CONCEP rs OF ~ ET

Problems Set
1. Prove the following:
(i) A C t/J => A=t/J
(ii) A c B, B C C ,. A C C.
2. Are the sets ~, to} and {t/J} different?
3. If A={l, a}, then find P (A).
4. State whether each of the following statements is correct
or incorrect. Here S is any set and S::;6~.
(i) S E P (S), (ii) S C P (S), (iii) {S} E P (S).

Answer. 3. {t/J, A, {I}, {a} }

4. (i) correct. (ii) incorrect. (Iii) incorrect.
 2
Bas ic Set 0 perations

Introduction. From arithmetic we know that-

(0 a+b is defined as the sum of the members a and b.
(ii) a-b is defined as the difference of a and b.
(iii) ab is defined as the product of a and b.
(iv) ~ is defined as division of a and b,
provided b#O in the last-case.
In a similar fashion we will define the operations of union,
intersection, difference, product in case of sets.
Union. The union of two sets A and B, defined by A U B,
is defined as the set of those elements which either belong to A or
to B. Symbolically,
A U B={X : x € A or x € B}
A U B is usually read as 'A union B'. The following have
the same meaning: union, sum, logical sum, join.
The union of a finite number of sets
A10 AI.· .. · .. , A,. is denoted by
fl
Al U A2 U ............... U A,. or by U Ar
,=1
CD

Similarly Al U A. U .......... U A .. U ...... = U A, •

•
Also U Ar={x: x E Ar for some value of r, I ~ r ~ n}.
'=1
In this diagram the shaded portions repre-
sents A U B. Such type of diagram is called
Vend diagram.
Examples. (i) P the set of points
which lie on the line AB and Q that of
the line CD, then P U, Q is the set of those points which lie
either on AB or CD.
 ~IC SET OPlRATION3

(ii) If A={l, 2}, B={a, b, c, I}, then

A U B={l, 2, a, b, c, I},
{2, a, b, c, I}={ I, 2, a, b, c}.
Remarks. (i) Some writers write A+B in place of A U B.
(ii) The meaning of "for some element a E A"
or "for at least one element a in A"
is the same.
Intersection. The intersection of two sets A and B, denoted
by A n B, is defined as the set containing those elements which
" belong to A and B both. Symbolically,
A n B={x : x E A and x E B}
A n Breads" A intersection B".
The intersection of a finite number of sets AI' A2 ,· ........ A.. is
denoted by
,.
or by n Ar •
r-l

n
By def. n Ar ={x : x E Ar "f r S.t. 1 ~ r ~ n}.
r=l
Examples. (i) If A={a, b, c},
B={l, ?, a}, then A n B={a}.
(ii) If A={l, 2, 3}, B ={c, d, e}, then
An B=t/I.
In the diagram the shaded portion
represent A n B.
Remarks. (i) Some writers write A.B for A n B.
(ii) The following words have the same meaning:
(iii) The phrases ''for every r" and ''for all values of r" both
have the same meaning.
(iv) It is ea-;y to prove that
,A n B=B n A
AU B=B U A
(A U B) U C=A u (B U Cj
(A n B) n C=A n (B n C)
(v) From the definition it is quite obvious that
and 'X E A n B ~ x E A
and any x E A n B <.;> x E B
so that A nBC A, A nBC B.
8 BASIC SET OPERA TJONS

(vi) Evidently
any a E A => a E A U B.
and any a E B ~ a E A u B.
so that
A C A U B, B C A u B.
Observe the difference between (v) and (vi).
Disjoint sets. Any two sets A and B are said to be disjoint
sets if their intersection is ,p. More precisely A and B are disjoint
sets if A n B=t/I, e.g.
{l, 2}, {a, b} are disjoint sets.
Difference or sets. The difference of a set A w.r.t. a set Bis
a set which contains only those elements of A which do not
belong to B and is denoted by the symbol A-B. Symbolically
A-B={x: x E A and x f1. B}.
Similarly we define
B-A={x : x E B and x ft. A}.
Example. (i) If A={l, 2, 3, a, b}, B={a, b, c, d},
then A-B={l, 2, 3}, B-A={c, d}.
Complement of a set. The complement of a set A w.r.t. the
universal set X is defined as the set X -A and is denoted by X'.
Symbolically,
A'=X-A={X : x E X and x ft A}
It is also clear that
any x E A' ~ X ft A.
Remarks. (i) X'=rjl
For X--X=rP
(ii) ,p'=X. For ",'=X-,p=X
(iii) A U A'=X, A n A'=c!>
(iv) A-B=A n B', B-A=B n A'
For A-B={x: x E A, x ft B}
={x : x E A, x E B'}
=A n B'
Similarly B-A=B n A'
(v) (A,),=A
(vi) x E A U B => x E A or x E B
x rt. A U B => x rt. A and x (/. E,
Symmetric of difference sets. The symmetric difference of
two sets A and E, denoted by A 6 B, is defined as
A A R=(A--B) U (B-A)
8ASIC SET OPERATIONS 9
. -
Example. If A={l, 2, 3, 4}, B={3, 4, 5, 6} then
A-B={I, 2}, B-A={5, 6},
A 6 B=(A.-B) U (B-A)={1,2} U {5,6}
={l, 2, 5, 6}.
Distributive law. The union of sets is distributive w.r.t. the
intersection of sets and conversely, I.e.
(A U B) n C=(A n C) u (11 n C)
(A n B) u C=(A u C) n (11 u C).
De-Morgan's law. Let AIX be a non-empty set for each IX in
an index set 6.. Then
(i) ( u AIJ. )' = ( () ~IX' )
(I.E 6 «ELl
(ii) ( () ArJ. )' = U Aac'
(I.E 6 IXE6
For proof refer to author' book on "Set theory and Related
Topics."
Ordered pair. An element of the form (a, b) is called an
ordered pair. The elements a, b in the ordered pair (a, b) are
called the first element and the second element respectively.
Equality of ordered pairs. Let (a, b) and (c, d) bc any two
ordered pairs.
We define
(a, b)=(c, d) iff a=c, b=d
i.e. (a, b)=(c, d) <=> a=c, b=d.
Product of sets. Let A and B be any two given sets. The
product of A, B is defined as the set consisting of all ordered
pairs (a, b) where a E A, b E B and is denoted by the symbol
A x B. Symbolically,
AxB={(a, b): a E A, b E B}
A X 11 is read as "A cross .B" .
The product set A x B is sometimes called Direct product or
Cartesian product. The reason for the name Cartesian product
is that the set A X B can be plotted in a Cartesian plane.
Similarly
BXA={(b, a) : b E B. a E A}

rr
J'
_ cJ ___ ~{(L bl

_____.fb
n ..,.. .----- A - ------:::: )(
10 BASIC SET OPERATIONS

Evidently A x B::j:.B xA.

If either B or A equal to ~, then
AxB=t/J.
Example. If A={I, 2}, B={a, b, c}, then
AxB={(l, a), (I, b), (I, c), (2, a), (2, b), (2, c)}
AxA={I, 2}x{l, 2}
={( I, J), (I, 2), (2, 1), (2, 2)}.
From this example it is clear that if A contains 2 elements
and B contains 3 elements, then
A x B contains 2 X 3 = 6 elements.
Generalising this result, we find that if A contains m ele-
ments and B contains n elements, then A X B contains m x n
elements.
This fact implies that A x B is an infinite set if either A or
B or both are infinite sets.
Product sets in general. The product set of the sets, A, B, C
is denoted by A X B x C and is defined as
A·xBxC={(a, b, c) ; a E A, b E B, c E C}.
Generalising this result to a finite number of sets, we have
A1 xA2 x .. x An={(al' a2 , ••• , an) : ar E Ar ¥ r, s. t. l~r~n}
Examples. (i) The product R x R [or simply R2J is defined
as Euclidean plane (of two dimensions).
(ii) The product set R x R x R [or simply R3J is defined as
Euclidean space of three dimensions.
 3
Functions and Sequences

3'0. Function. Suppose X' and Yare two given sets. By

some given rule, if each element x E X corresponds to a unique
element Y E Y, then that rule is called a map of X into Y and is
denoted by f.
We write f: X... Y. This is read as "f maps X into Y" or . 'I
is a mapping of X into Y".
Mapping sometimes called operation or map or correspon-
dence or function.
Let an element y E Y be corresponded by an element x EX,
then Y is called the image of x and is denoted by f (x). Thus
y=/(x). "/(x) is read a'S 1 of x"
Here x is defined as the pre-image of y. The set X is defined
as the domain of the ftinction/ and Y is called co-domain of j,
I(X) is cabed the range or the image set off.
It is also expressed as
/(X)={/(x) : x E X}
Evidently /(X) C Y.
Remark. From the definition of a map it is clea.r that a map
./: X_ Y is said to be well-defined if
(i) any element x EX=> f(x) E Y
(ii) any element x EX=> a unique element/(x) € Y.
In other words 'no element of X can have more than one
I-image, i.e., given any x E X, we can not find two or more than
two elements Yl' Y2 E Y such that Yl-/(X), Y2 f(x).
(iii) Two or more than two elements of X may have the
same/ -image in Y. i.e.,
if Xl. x 2 • Xa, , .. E X, then,
/(x1)=/(X z) =/(Xa) = ...
Examples. (i) Let a map /: A-B be defined by the dia-
gram:
12 " FUNCtIoNS' AN[j SEQUENCES

8
f

Here a E A ~ 3 1,2 E B such that 1 =!(a)=2.

This ~ ! is not well-defined.
(iii) The map ! A~B defined by the following diagram is
well defined.

Aj~~f
f. ------..-=
b
_"; •

10---I

Here domain of f {a, b, c}

range of 1={2, I}
Co-domain of/=(2, 1, -I).
3'1. Onto and into mappings
Letf: X---,>-Ybe a map.
I is called an 'into' map if 3 at least one element in Y which
does not have a pre-image in X.
i.e.,fis an 'into' map if/(X) is a proper subset of Y.
i.e., if!(X) C Y.
/ is said to be an 'oIllO' map if !(X)= Y, i.e., if evcry element
in Y has a pre-image in X.
3-2. One-one and many-one mappings.
Let: X --;. Y be a map.
/ is called a one-olle map if
/(X1)=(X 2) ! Xl> X 2 E X ~ XI=X~,
01' equivalently
Xl :;CX2 : Xl> X 2 E X ~ f(xd:;tf(x 2}
i.e. if different elements In X have different images in Y.
. / is called a many-ol/e l1lap if dilfercnt clement s in X hav~ the
same i-image in Y i.e., if
j(x})=f(x") : Xl, x 2 E X ~ X'*X 2'
pUNCflONS AND SEQUENCES 13

Examples. (i) Let f: A ~B be a map defined by the dia-

gram:
e
j

Here f(A)={f(x) : x E A}
={x, y, z}=B.
Also /is one-one.
Finally, f is one-one onto map.
(i) The map f: A~ B defined by the follow ing diagram is
many-one into map.
A fj

3·3 Real niued map. A map is said to be real valued map

if a set is mapped into the set R by it.
That is to say, any map
f: A ~R is called a rea' valued map.
3·4. Set Function. A function is called a set Junction if its
domain consists of a family of sets
e.g., the map f: {A. : i E N}-+N
defined by the formula

is a set fu nctio n.
3·S. Real Valued Set Functi.n A function is called real
valued set function if its domain is a family of sc!s and its co·
domain is the set of real numbers R.
e.g., a map f: {Ar t
r E L;.. -? R
is real valued set function and this map is said to be finitely
additive if
 14 FUNCTIONS AND SEQUENCES

I ( U Ar
,=1
)= UI (
,=1
Ar )."

The map I is said to be couatable additive if

I ( ~1 Ar )= r~ I ( Ar )-
3'6. Extended real valued set function. An extended real
valued set function is a real valued set function which also takes
either of the value -{- 00, -- 00.
i.e., a map I: { _[-00,00] such that
Ar}r E 6.
f(Ar}=+oo or for some value of r.
-00
It may be noted that 00 and -00 both are not real numbers.
Step function. We divide the interval [a, b] by means of
points Xl' X 2, ...... , X" s.t. a=xo < Xl < X 2 <, ...... , < x,,=b.
A function/s.t./(x)=c, for X'-l < x <x, is called step function
where c, are constants ;=1, 2, ...... , n.
3'7. Sequence'. By a sequence we mean a map I whose
domain is N, That is to sap, a may I: N _ A
is called a sequence.
The value of the function I at n E N is denoted by
f<n) or by I •.
The sequence itself may be referred to by the symbol
</(n):n EN> or <fen) >
The set {I(n) : nE N} is defined as the range of the sequeace.
If this sequence is in X, then the sequence is said to be in X•
. This X may be R or the set of complex numbers or any other set
of numbers.
3 8. Convergent sequence. A sequence < aft > is said to
converge to a if
given E > 0, 3 no E N S.t. ¥ n ? no => I a-a" I < ~
A sequence < a" > is said to be convergent if
given 6> 0,3 no ~ N S.t. ¥ m ~ 110 =:> I am-am+fl I < ~ ¥ pEN.
3'9. Bounded sequence. A sequence < an > is said to be
bounded if
°
3 a number M > such that I a" I < M ¥ n I::. N.
It is easy to prove that
Convergence => boundedlles~.
3'10. Metric spa.:e. Let X7:-tP be any given space.
Let X, y, z E X be arbitrary
[Link]::TIONS AND ~EQUENCfS 15

A function P: XxX .... R

having (he properties listed below
(i) p (x, y) ;;;. 0
(ii) P (x, y)=O iff X= y
(iii) p (x, y)=P (y, x)
(iv) P (x, y)..Lp (y, z) ~ P C<, z) (triangle inequality)
is called a distance functioll or a metric for X.
3'11. Definition. A sequence < an > is said to be strictly
monotonic increasing if an < an+! ¥ n E N.
A sequence < an > is said to be strictly monotonic decreasino...
if an+! < a.,. >.f n E N.
A sequence' < an > is said to be monotonic increasing or
monotonic non-clccrcasing if an ,,;;; an+! ¥ 11 E N.
A s~quence < an > is said to be monoton'c decreasing or
monotonic non-increasing if an+! ~ an ¥ 11 E N.
A sequence < an > is said to be monotonic or monotone if it
is either monotonic increasing or monotonic decreasing. Notice
that a sequence will be either monotonic increasing or monotonic
decreasing.
3'12. Axiom of Choice. Cartesian product of a family of
non-empty sets is non-empty.
Axiom of choice may also be stat·~d as "there exists a choice
fu nction for any non-empty fami Iy of nOll-empty sets".
 4
Bounded Sets, Derived Sets,
Open Sets and Closed Sets
on the Real Line

Rea) Jine. The elements of the set R can be represented by

means of points on a straight line. The origin of this straight line
is taken to represent the number o. The real numbers x > 0 are
represented on the right of lero, where as the real numbers x< 0
are represented on the left of o. Each point of this straight line
represents a unique real number and each real number determines
a unique point on the line. For this reason the set R is referred
to as the real line or real axis. This is why, we use the words
point and number conversely.
Remarks. Throughout this chapter we shall take R as the uni-
versal set provided nothing is contradictory stated.
O.,en and closed intervals.
Let a, b E R be arbitrary.
We write
[s, b]={x E R : a <; x ~ b},
(a, b)={x E R: a < x < b},
and can these sets of real numbers by the names closed interval
and op~"inferval res pecti ve Iy.
Also we write
[a, b)={x : a ~ x < b}
(a, b]={x : a < x ~ b}
and call each of these sets 'of real numbers by the name semi-open
interval or semi-clo ..ed interval or half-open interval or half closed
interval.
Neigbbourbood of a point. Let ~ > 0 be any real number.
Let Xu be any point on the real line. Then the set
{x E R: I x -xo I < ~} is defined as thJ. 6-neighbourhood of
the point Xo. Thus,
. -neighbourhood of the point Xo
={x E R: I x -Xo I < ~}={x E R : Xo --4I < x < XO+Ii}
SETS ON TtiE REAL LINE 17

=(XO-E, xo+·)·
If A is a neighbourhood of x, then A-{x} is called a deleted
neighbourhood of x, i.e., the set (X-f, x+.)-{x} is called deleted
.-neighbourhood of x.
Example. The set of real numbers x satisfying the condition
3 < x < 7 is 2-neighbourhood of the point x=s.
The idea of open set and closed set is connected with the
idea of neighbourhood.
Open set. A set A c R is caned an open set if
given x e: A, 3 f > 0 such that the 6-neighbourhood of
x is a ·subs~t of A.
From this it is clear that every 6 -neighbourhood is a.n open
set but an open set is not necessarily an • - neighbourhood.
Example. (i) The set of all real numbers x such that
o< x < 2 is a I-neighbourhood of the point Xo= 1 as well
as this set is an open set.
(ii) The set pf all real numbers x such that 0 < x < 5 is an
open set but not an ,- neighbourhood for any t > O.
Continuity. A real valued function f is said to be continu-
ous at the point Xo if given E > 0, 3 8 '> 0 such that
I f(x)-/(xo) I < E whenever I x-x, I < 8.
If a function is continuous at each point x for which it is
defined, then it is called a continuous function.
Bounded linear set :
(i) Let A C R b! any given s~t. Then A is c![Link] linear set.
(ii) A real number b is called an upper bouocl for the set
A if
x~byxEA
(iii) An upper bound b of A is called the lU'1t upper bound
Or supremum of the sct A if
b ~ y for every upper bound y of A.
In brief. the supremum of A is written a~ sup (A). Also the
lea'>t upper bound is written as. I. u. b.
Example. b is the sl!premuln fllr all the reat numbers in
(a, h] and h+x is an upper bound for the set [a, hJ ¥ X e: R+.
(iv) A set A is called bounded abuvf: or bounded on tbe ri&bt
if it has an upper bound.
18 SETS ON THE REAL LINE

Examples. (I) The set of all re3J numbers in [a, b] is bounded

above.
(2) The set {x E R : x < 5} is bounded above.
(3) The set (X E Q+ : Xl < 5) is bound:!d above and its
upper bounds are 6, 7, 8, 1l/2, ... since vS ¢ Q.. Hence supre-
mum of the set does not exist.
(4) The set {I, I, i, t, ...... ,} is bounded above.
(v) A set A is said to be unbou .ded on the right if it is not
bounded on the right e.g. the set lx E R : x < 5/.
(vi) A real number a is called a lower bound for the set A if
aWi;x¥xEA
A lower bound m of A is called greatest lower bound or
talBlum of A if a ~ m for every lower bound a of A. In short the
in{~mum of.4 is written as inf (.4) and greatest lower bound is
Wrttten as g. 1. b.
A set A is called bounded on the left or bounded below if it has
a lowo:r b.)und.
An interval which has at least one end point as 00 or - «. is
an unbounded interval.
Examples. (1) The set {x E R : x > 5},
is bounded below.
(2) The set N={l. 2, 3, 4, ...... } is bounded below.
(3) The set {x E Q+ : x ~ 5} is [Link] on the left (i.e.
bounded below) and its greatest lower bound is 5.
(4) The set of real numbers [a, b] is bounded b.!low and its
infimum is a.
(vii) The set A is said to be unbaun:lej o:t tile lert if it is not
bounded on the left e.g. the set {x E R : x < 5}.
(viii) The set A is called bcnm!led if it is bounded both below
and above e.g. the set of all real numbers in [a, b] and (a, b).
Another definition. The set A is said to be bounded if 3 a
numb!r
l ..
m. > 0 E R such tbt I x I ~ m ¥ x E A.
EVidently A c [-m, mj=a finitei:ltervaL So we also define that
the set A ii bound ~d if it is a SU~l:::t of a finite interval.
(ix) The_ set A is caJled unbo1in~ed if it is not bounded.
Eumples. 0) A finite set is [Link] bounded.
(2) lhe set A={l, i, l, 1, ... } is bounded. For A C TO, 1 .
(3) An infinite set mayor may not b! bounded.
(4) Let A={ 5, 6, 7, 8, 9} be a subset of the set N of natural
ftUl!lbers. Then 5 is lower bound of A as 5 ~ x "rI- x E A.
Similarly I, 2, 3, 4 are also lower bounds of A. 5 is the greatest
SETS ON THE REAL LINE 19

lower bound of A. 9 is the upper bound of A as x ~ 9 ¥ x E A

Similarly 10, I J, 12, ... etc. are also upper bounds of A, wh.:reas
9 is the least upper bound of A.
(x) If A~tP, then we ado;:,! th~ convention,
sup A= -- 00, inf (A)= +00.
Limit point and derived set.
A point x is said to be a limit point of a set A if every nej!h~
bourhood G containing x contains a point of A other than x, I e.,
x is said to be a limit point of A if for every neighbourhood
G with x E G,
(G-{x}) n A#t/J.
The set of all limit points of a set A is called the derived set
0/ A and is denoted by D(A).
Evidently
x E A => x E D(A) or x fI. D(A).
e.g. the limit point of the set {J, j, 1, t, ... } is O.
Condensation point. A point x is said to be a condensation
point of a set A if every neighbourhood containing x contains an
infinite number of points of A.
In accordance with this definition a finite set has no conden-
sation point.
Closed set. A set A is said to bl! closed if every limiting
point of A belongs to the set A itself.
Symbolically, a set A is said fo be closed if
D(A) C A.
Examples. (I) {O, 1,1, I, 1, ...... }=A is a closed set. For
D(AJ={O} C A.
(2) A closed interval is a closed set.
For D«(a, bJ)=(a, b] C la, bJ.
(3) Every real number is a limit point so that ration'll limit
points belongs to the set Q. This =>
D(R)=R C R
D (Q)=R <t: Q
so that R i-; a closed set whereas Q is not a closed set.
Open set. A point x is called an interior point of the set A
if 3 a neighbourhood G containing x such that G C A.
From this definition it is clear that if x is an interior point
of A then x must belong to the set A. .
A set A is called an open set if every point of C is an interior
point of A.
20 SETS ON THE REAL LINE

Examples. (1) R is an open set.

(2) An open interval is an open set.
(3) Q is neither open nor closed.
[Link]. (i) An arbitrary union of open sets is open.
(ii) The intersection of a finite number of open sets is open,
whereas arbitrary intersection of open sets mayor may not be
open.
(iii) An arbitrary intersection of closed sets is closed.
(iv) A finite union of closed sets is closed whereas an arbi-
trary union of closed sets mayor may not be closed.
 5
Countability of Sets

Introduction. A question may arise on any two given sets

that they have the same number of elements or not. The answer
can be found by counting the number of elements of the set if the
sets are finite. On the other hand if the selS are infinite sets,
then the answer depends upon how one defines two sets to have
the same number of elements.
5'0. Definition. A set A is said to be cardinally equivalent
to a set B if 3 (at least one) one-one map from A onto B. This
fact is denoted by the symbol A-B.
This fact is also expressed by saying that
(i) A is numerically equivalent to B.
(ii) A is equivalent to B.
(iii) A is equipollent to B.
(iv) A and B have the same power.
Examples (i) Any two singleton sets are equivalent
(ii) Let N={l, 2, 3, ... }, E={2, 4, 6, ... }.
Define a map!: N-+E by the!ormulal(n)-2n.
Evidently I is one-one and onto. Hence N--E. This example
shows that the set N is equivalent to a proper subset of itself.
This is a striking property of infinite sets.
(iii)Let A={J. 2. 3, ... }. [Link]{alt a., a..... }.
Then Ar-B under the mapping n ~a,..
(iv) Let A=[2, 6], B=[a, bJ,
where [2, 6] denotes tlv! set 01 all real numbers in the closed inter-
,al [2, 6]. Similar meanillg lor [a, b].
Then [2, 6]-[0, b] under the mapping

x-a+ b-a
4- (x-2).
22 COUNTABILITY OF SETS

Remark. In order to find the mapping from [2, 6] to [a, b],

firstly find the equation of a straight line joining the point (2, a)
to (6, b) by the formula

Y-Y1=)'2- Yl (X-Xl) and then put y-f(x).

X 2 -X1
Theorem l. To prove that the relation A,...,B ill the family of
sets is an equivalence relation.
Proof. The given relation is
(i) reflexive: A-A for every set A.
For the identity map /04 : A-+A given by fA (x) .... x is one-one
onto.
(ii) Symmetric: A-B => B---A.
onto
For A-B => 3 a one-one map f: A ~--+ B.
onto
=> f- 1 : B ---+ A is also one-one.
:$ B-A.
(iii) transitive: A-B, B-C:o> A--C.
onto
For A-B, B-C .. 3 one-one mapsf: A ---+ B
onto
and g : B - - .. C.
onto
=- gf: A - - - C is also one-one,
.. A--C.
rt means that the given relation satisfies all the conditions of
an equivalence relation. Hence the result
!"t. Cardinal numbers. [Kanpur 71, 70; Utkal 5]
The relation A-B in the family of sets is an equivalence
relation. Hence the [Link] decomposes the family of sets into
di!;joint equivalence classes. Every equivalence class defines a
unique cardinal number. In other words the cardinal number of
a class of equivalence sets is any representative of the class. The
cardinal number of a set is sometimes called its power or potenc~'.
If P is any finite set consisting of p elements, then its cardi-
nal number is defined as p.
The cardinal number of ~ is defined as O.
A cardinal number corresponding to a finite set is called a
finite cardinal number. A cardinal number corresponding to an
COUN'fABILIrY OF ~ETS 23
infinite set is called a transfinite cardinal number. It follows from
these definitions that all the transfinite cardinal numbers a"C 'grea-
ter than any fi!1ite cardinal number.
Notation. The cardinal numbers of the sets
P, Q, R, S, ........ .
arc denoted by the corresponding small letters
P. q. r, s, ........ .
Also the cardinal number of a set A is denoted by I A I or
card A.
Then, by definition
/ P 1= card P=p, I Q /=card Q=q.
The cardinal numbers of the set N and the set of all real
numbers [0, l~ are denoted by a and c respectively.
Then . / N 1=8, I [0, l] /=c.
The symbol Xo (read as aleph nUll) is also used to denote I N /.
This symbol was originally used by Cantor.
If P""Q then by definition of cardinal number,
1 P/·=IQI·
5'2. Sum of cardinal numbers. Let P and Q be any two sets
with cardinality p and q S.t. their intersection is empty i e.,
I PI=p.1 Q I=q, P () Q=;
Then the sum of p and q is defined as
p+q=1 P u QI·
More generally, let P". be a cardinal number corresponding to
the set p~ ¥ CIt in an index set 1::,.. Further suppo'se that. .
Prx. ("\ P~=,p >tf Of, ~ € 6 s.t. 1X:;t=~. We define the sum of cardinal
numbers as
E p",= U P", J
-EDCltEI::,. •
Examples. (I) 1/ A={l, 2}, B={5, 6, 7), then
1A 1=2, I B 1=3 and A () B=~ so that
1A U B 1==1 A 1+1 B 1- 2+3-5.
Also A u B={I, 2, 5, 6, 7} and hence 1 A U .B 1-=5.
This is a verification of out definition 5'2.
(ii) Let A={a, b, c}, B:oo{a, p, q, r, s}.
Then A U B=(a, b, c} U {a, p, q, r, s}.
={a, b, c, p, q, r, s}.
... 1A u B 1=7.
24 COUNfABILITY OF SETS

Evidently 1A 1=3, 1B 1=5 and A n B::;t:~.

:. By definition 1A UBI =I=- I A 1+ I B I
i.e., 7 ;::3+ 5.
This example also verifies our definition of sum of cardinal
numbers and it also shows that An B=; is a necessary condition
for defining the rule 1AuB 1=1 A 1+1 B I.
S·3. Prodlict of cardinal numbers
Let P and Q be any sets s.t. I PI =p, 1 Q I=q.
We define the product of cardinal numbers p and q as
pxq= I PxQ I.
More generally, let Pa. be a non-empty set for each Gt in an
index set A and let pa.=1 Pa.. Then we define the product of
cardinal numbers as npa.=j x Pa.I.
where ilpa. stands for the Pl.P2' Ua· ....
and xPrJ. stands for the cartesian product.
Example. Let P={a, b}. Q={a. J, 2, 3}.
Then I P 1=2, I Q 1=4, pnQ,ptP.
By definition, 1 P x Q I =2 x 4=8.
Now I Px Q 1=1 {(a, a), (a, I), (a, 2), (a, 3),
(b, a), (b, 1), (b. 2), (b, 3)} I
=8.
This verifies our definition of product of cardinal numbers.
!§·4. Definition. A set A is called an infinite set if A is equi-
valent to a proper subset of itself.
In the language of notation, a set A is called an infinite set if
A -.I to a proper subset of A e.g. N, Q, R, C etc.
5'5. Definition. A set A is called a finite set if it is equiva-
lent to {I, 2, 3, ... ,n} for some value of n E N.
S·6. A set A is called a denumerable set if A ..... N.
e.g. tal' aa, a8,·· ... } is denumerable, since this set -- N
under the map an .... n. Moreover this is a sequence and therefore
a sequence is a/ways denumJrable.
A denumerable set is sometimes called enumerable or counta-
bly infinite .
•: A.....,N ~ 1A 1=1 N I => I A 1=8.
Also A.-..J N ~ A is denumerable.
Hence a denumerable set can also be defined as :
A set A is denumerable if I A 1=8.
COUNTABILITY OF SETS 25
We shall always use this very definition in future
5'7. Definition. Countable sets. A set is called (:ount4ble or
atmost countable if it is finite or denumerable.
Example. (i) A={I, 2, 3, 4}. Then A is finite so that, by
definition, A is countable.
(ii) Let A={a1 • a2 • aa,·"} so that A is denumerable. Hence
by definition A is countable.
Here observe tbe difference between coulitable and countably
infinite.
5 8. Definition. A set A is [Link] an uncountable set if A is
an infinite set and A is not cardinally equivalent to N.
e.g. Rand C are uncountable sets.
The following have the same meaning:
Non countable, non-denumerable, non-enumerable, uncoun-
table.
5'9. Definition. A set A is said to have the power of conti-
l1um if A r- [0, 1]. Also we say that the cardinality of A is (:.
Theorem 2. To prove that
I AxB 1=1 B 1+1
B 1+1 B I.. ·to I A I terms.
Proof. Let A and B be any two sets.
AxB={(x, y): x E A, Y E B}
= U {x, y) : )' E B}
xEA
Then AxB I 1=\ u {(x, y) : y E B}
x E A ... (1)
For a fixed x E A. consider the map
.f: 8~{(x. y) : )' E B} given by
.f (y)= (x, y) ~ )' E B.
Evidently.f is one-one onto.
Then B '" {(x: y) : y E B}
This ~ 1 B 1 =1 {(x, y) : y E B} I.
In this event (') shows that
I AxB 1=1 B 1+1 B 1+· .. to 1A 1terms.
Q.E.D.
Verification of the theorem 2. Let A={l, 2}, 8-(1,2, 3}.
Then 1 A 1=2, I B 1=3
I AxB 1=2.3=6
A ><B={(I, 1), (I, 2), (1, 3), (2, 1), (2,2), (2, 3)}
 26 COUNTABILITV OF SETS

If we take A1={(I, I), (I, 2), (I, 3)},

B1 =t(2, I), (2, 2), (2, 3)}.
then A x B=.A I U B t • Al () Bl =;.
that 1AxB 1=1 Al u Bl lie. 2.3=3+3.
Remark. Every property of addition and multiplication of
natural numbers is not true for cardinal numbers in general.
e.g. a+a=a does not imply a=O
a.a=8= La does not imply a= 1.
This proves that Cancellation law is not true for the operpa-
tions of addition and multiplication of cardinal numbers.
Ex. t. If fI. be OIly cardinal number, then
fI. E;; 1 A 1 ~ 01 => 1 A 1 -«.

Solution. Let« be any cardinal number.

Let II E;;: 1 A 1 <; fl.. To prove that 1 A 1 =cx.
Let 01=1 B I Then we have to prove that I A I -I B I. which
is [Link] to proving that A-B.
« EO;; I AI .. I B I .s;; I A I ... (1)
:0 B- to a subset of A or B-A
IAI E;at=> IAI E; IBI
=> A- to a subset of B or A-B. ...(2)
The statements (I) and (2) taken together imply the required
result.
Ex. 2. If () and {J are cardinal numbers such that fI. ~ f3 and
{J ~ ar, prove that at = 11. [Banaras 1969]
Solution. Let A and B be two sets s.t. 1 A 1 =a, 1 B 1 =-~.
Also let at ~ {J. (J ~ ac.
To prove that Cl""'~. we have to prove that: A I - 1 B I.
GlC;;B=> IAI ~ IBI
lO A -- to a subset of B or A-B ... (1)
~~Cl"IBI~IA:
~ B- to a subset of A or B-A. . .. (2)
The statements (I) and (2) imply the required result.
Theorem 3. A set A is countably infinite iff A can be put In
the form of the sequ[!nce tOI' O 2 , tl3 • .• 1 of tlte distinc: elements.
Proof. A set A is countably infinite iff
3 a one-one onto map f: N ~ A
i.e. iff (i) m.;f:n and m, n E N ~ f (m)-:;cf (n).
(ii) Given any element b E A, 3 an e1em'!nt a C N s.t. f(a) b 111

i.e., iff f[N]=A and f is one-one.

 COUNT ABILI j Y OF SETS 27

or iff given any element n € N. 3 Of! € A lI.t.j(n):::a n andf is

one-one.
Above means A={al. a2 • aa •... }'
where a10 02' 0a.· .. are distinct elements.
Theorem 3ft. Every subset of a cvuntable set is coufttable.
[Meerut 1987, 88]
Proof. Let A be a countable set. Then A is either finite or
enumerable.
(I) When A is finite.
Then every subset of A is also finite and so is countable.
(ii) When A is eoumerable.
Then A can be written as
A={x1, XI''''} ... (1)
Let B be a subset of A.
(i) If B=;. then evidently B is countable.
(ii) If BtF"', then B is expressible as
B={x. l , x n2 ,· .. }, where XIII E A.
Evidently B is enumerable, by virtue of (l).
Theorem 4. The set of all real numbers iI' the closed illferval
[0. 1] is not denumerable. [Banaras 1971 ; Kallpur 68]
Proof. Let A denote the set of an real numb.:rs in tne [Link]
interval lO, I J. SymbolicallY we write
A={x E R : 0 ~ X ~ l}=[O, 1).
To prove that A is uncollntable.
Suppose not. Then A is couhtable. This implies that every
element of A must appear in the sequence
x"' .. .of distinct elements. i.e.,
X10 X 2 , Xa,· ......
B={xlo x 2 , xa, .. ·} ... {I)
Write decimal expansion of those xl's as follows:
x 1=0'xu Xl 2 Xu xu .... · X 1m . .. ..
X~=O.X21 X 22 X 23 X~, ...... x:.,. ..... .

X.,,=0·Xm1 XtII~ X"'a X"" ..•..• XIII"",

where XII E {O, 1. 2, 3, 4, S, 6, 7, 8, 9} ¥ i and j and each deci-

mal contains an infinite number of non-zero elements.
We can write 1 and i in two ways as given below:
J 1=1'0000... ... ... .
1 1-0'5000 ....... ..
28 COUNTABILIrY OF SETS

and J t =0'99999 ...... =0'9 }

1 1=0'49999 ...... =0'49 ... (2)
But in the present case we shall write the decimal representa-
tion of the elements A in the form (2).
Construct a real number ~=O, ~l> ea.
~a ... ~". ... s.t.
if Xu = 5, write ~l =4
and if Xu :f.:5, write fl = 5
in general if x",,,,=5, write ,,,,=4
and if x",,,,;t:5, write f",=5.
In either case, it is clear that x".".;t:'''' ¥ m. . .. (3)
Since ~1It=4 or 5 ¥ m
:. eE [0, 1], i.e., ~ E A ... (4)
x",=~ ~ O.X"'1 X.I xma ...... =O'~l ~l! ~a"""
=> xmm=~", T m. Contrary to (3).
:. x",:;;c~ ¥ m.
In this event (I) shows that ~ e
A. Contrary to (4). Hence
the required result follows. Q.E.D.
Theorem S. To prove that the set R is uncountable.
[Ufkal 6S ; Puajab 62 ; Vikram 61]
Proof. Firstly we shall show that I R/=c.
We know that cardinality of the set of real numbers in the
open interval (0. b) is c, so that
I (a, b) /=c ¥ 0, b E R s.t. a < b.

This .. \ ~ )i=c.
(-r' ...(1)
The set R is the set of all real numbers lying in the open
interval (-00, (0), i.e., R=(-O<!>, co).
Define a mapj: (-~, ~ )~ (-00,00) by the formula
/(x)=tan x.
Evidently j is well defined. Also it is easy to verify that! is
one-one onto.

:. (-~, ~)- (-00, (0).

This I(
~ -~, ~ )=1 (-co, 00) I·
i.e. c=/ R I. in a~ordance with (I).
This ~ R is uncountable, Q.E.D.
COUNTABILITY OF SETS 29

Ex. 3. To ~ow that lor every real number x, the real numbers
in lhe semi-open interval [x, x+ I) lorm an uncountable set.
Solution. Let x E R be arbitrary. Define a map
I: [x, x+I)_[O, 1] by the formula.!(X)=X-x,
Evidently l(x)=x-x=O
l(x+l)=x+l-x=1
:. I is well defined.
f(XJ=/(X2 ) ; Xl> X 2 E rx, x+l) => X1-X=XI-x

This => I is one-one.

I is a continuous map => I is an 'onto' map.
:. [x, X+I) '" [0, 1)
This => I [x, x+l) 1=1 [0, I) ,.
But the set of all real numbers in [0, 1) form an uncounta-
ble set and hence the set of all real numbers [x, x+ I). Proved.
Theorem 6. (i) II Ai is countable infinite set, then
fI
U A. is countable infinite and hence deduce that n.a =a.
1=1

If Ai
- (ii) is countably infinite set lor i = 1,2, 3, ... ...... then

U Ai is counfably infinite and hence deduce that

8+8+8+ ...... +to 8 terms=a.

[Meerut 1972; Kurukshetra 69]
Or Union 01 countable collection 01 cOll'Jtable sets is countable.
[Meerut 1987, 86]
Proof. (i) Let A. be a countably infinite set.

Let A=U" A •
• =1

To prove that A is countably infinite.

Let the elements of A; be displayed as
Al : au a12 Q I8 ...... 0 1....... .
,42 : au at~ a 21 .... .. 0 2" ..... .
A8 : 0al aaJ 0 83 , ..... as" ...... .

Write B={au , 012'''' ... a 1" .. • ; 0 21 , a22' ...... ; 0"1' a2,,, .. ·}·
The S(~t B considered as the set of distinct elements is coun-
tably infinite. :. , B I =8. ...( 1)
30 COl!NTABILlTY OF SElS

If A, n A}= r/> for i;f=j, then evidently

I A I = I B, =a.
If Ai n A, t:.d> for i :;t:j, th'.!n A is equipollent
with some sub-
set of B. This:? I A I ~ ! B I -'-~a or I A I E;;;. a ... (2)
But Al C A and 1 Al I =8
:. a= I Al I ~ I A I or a ~ I A I ... (3)
Combin:ng (2) and (3), we get
II ..; I A I ~ a, which:? I A I = a.
In either ca<;e I A I =a This proves the required result.
Dedu<:tion. Further assume that
A, n A,=,p for i#j and ;,j=l, 2, ...... , n.
Then by what we have established it follows that I A I =8.

or IU At I=a,
1'=1
or ;
1=1
I A, I =8.
a+a, ...... + to n terms=a, or IZ.a=a.
or

(ii) Let AJ be a countably infinite set, and A=- U

.. A.
1_1

To prove that A is countably infinite.

Let the elements of A, be displayed as follows
Al : all 0 12 013·· .... 01.... • .. .
A~ : a2l 0 22 0 28 ...... a ....... ..

Aa : 03l 032 a33 · ..... 0." .... ..

First we shall assume that A, n A/=I/I for i::;t:j.

Divide the elements of A into blocks s.L mIll block contain m
elements and at} wili be in mtll block iff; +j=m + l. Within the
m'" block the second sufii'< j of Oi! increase from I to m. Thus A
can be expressed as
A={ou ; 0 21 , 0 12 ; 0 31 , 022' 013 ; ... ; 0"1>"" 0 1,,; .. }.
Clearl.t' A is expressible in the form of a sequence of distinct
elements and hence denumerable so that ! 11 =a.
Secondly we assume that A, n AFi=t/J for i .,cj.
In this case A is equipollent with some subset of N'
so that ! A I ~ I N I or I A I ~ a. . .. (4)
But Al C A and I Al I ~ 2.
COUNT ABILITY OF SETS 31

so that a= I Al IE;;/ A ' or a ~ I A / ... (5)

This => / A / =a, by (4) and (5).
This proves the required result.
Deduction. Further assume that
B. n A,=~ for i.:l=j.
Then from what ha'i been done it follows that

I A 1 -a::>· IUA.I=a => ;

I 1 '~1
/ A. 1 =.
=> a+a+ ... +... to a terms=a
This completes the proof.
. Theorem 7. To prove that N X N is countable,
or to prove that a2 =a.a=a.
Proof. To prove a.a=a
is the same as to prove I NxN 1= 1 N ,.
The elements of N x N may be displayed a'i follows:
(1.1) (1,2) (1,3) ...... (I,m) ...
(2, I) (2. 2) (2, 3) ...... (2. m) ...

(n, I) (n, 2) (n. 3) ...... (n, m) ...

then
If we write

(i)
.
At={(i. 1). (I. 2). (i,3) ....... } •

NxN= u A.
\ ,-I
(ii) A. nA,=q. for i#j
(iii) A, is a countably infinite "V i .

These facts lead to the conclu,ion that U

.. Ai is countabl~

infinite i.e., N )( N is countably infinite and hence

NxN ...... N .
This => / NxN /=/ N I
Ex. 4. Prove I P (A) 1=2 / A / ;f A is any finite set.
Solution. Let A be any finite set with cardinality n.
P (A) is the family of all subsets of A.
/ P (A) / =·co+ ..C'+ ..C3+ ........('"
=(1 + 1)"=2"=2 I A /
:. I peA) / =2 I A I .
1 heorem 8. (I) If A, is non-enumerable set ¥ .. e: :01.
32 COUNTABILIlY OF SETS

-
then U AI is non-enumerable and hence deduce that
1=1 ~ /

c+c+c+ ...... to a terms=c.

(U) If Ai is non-enumerable set for 1 <; i ~ n.
It
U AI is non-enumerable and hence deduce that
1=1 .

c+c+c+ ...... to n terms=c.

Proof. (i) Let AI be a non-enumerable set ¥ i E N so that

we can write I A, I =c ¥ i E N. Let A=U A•.

..
1=1

To prove that A is non-enumerabie.

FOr proving this we must show that 1A I=c.
1A, I=c shows that the set A, is numerically equivalent to
the set of real numbers in the interval [0, I). That is to say,
AI-[O. i)

Similarly A2 r- [ } . }+~)
1 I 1 1 1)
Aa"'" [ 2+22' 2+2i+V

i.e. A 1 -[0, I). A2 -I},}) Aa - [}, ~)

A __
..
[1- 2,,-1
.1_ ' 1- 2"
!.).
If we a~sume that A, n A, =~ for i :;t:.j, then above results
taken together imply that

U A • ..., U[1- 2-L ,1-~)

).1 1=1 2
or A-lO, I) or 1A 1=1 [0, J) \=c
or 1 A I=c.
If we assume that A, nA,:;i:t/> for i;#:j, then
00
u A. is equivalent with some subset of L0, J).
0-1
COUNfABILlTY OF SETS 33

IU I~
,-1
Ai c or 1 A 1 ~ C

But AICA and 1 AII=c.

1 ~ 1 A, or I A ~ c.
c=I Al
1 A 1 ~ c, 1 A 1 ~ c => c ~ / A / ~ c => I A 1=c.
In either case,' [ A I=c.
Deduction. Further assume that AinA,=t/> for i#j.

We have seen that!u Ai I=c, this => E / Ai /=c

i-I iEN
or c+c+c+ ...... to a terms=c.
(ii) Let Ai be a non-enumerable set for I ~ i ~ n.
Then we can write I Ai I=c for I E;;; i ~ n.
.
To prove that U Ai is non-enumerable.
;=1

1 Al I=c implies that Al is cardinally equivalent with the set

of real numbers in the semi-open interval [a}, O 2 ) where a10 a2 E R
and a1 < a2• This is expressed by writing AI-[Olo 4.!).
Here we shall suppose that Or < Or+!. where ar, Or+! E R
for r=l, 2,3, ... , n.
Thus we have
A~ ....... [0[, O2 ), A2 ,.." [a 2 • 0 3), A3 '" [oa,' a4 ), ... An ,..., [a... 0"+1)'

If we assume that· Ai n A,=q, for i#j, then the above state-

ments taken together imply that u

.. Ai -.J [01' Ontl).·
;=1

This => i~l Ai \=1 [01' 0"1-1) 1=c =>, /;91 I=c. Ai

If we assume that Ai n Ar;t=~ for ;;j::j, then

n
U Ai is equivalent with some sub,et of [ai' o"tl) so that
i=l

.. (1)

But Al
"
C U Ai and 1Al I=c.
4.=1
 34 COUNTABILlTY OF SETS

:. c= 1 Al I ~ li~ At I or c ~ \ ~lAt ... (2)

By (I) and (2), r UAi I=c,

J=1

:. In either case I UA, I'==c.

t·ol

From this the required result follows.

Deduction. (i) Further assume that
Ai n As-¢> for I#}.

Then I UAt I=c => IUAi I=c,

... 1 /==1

=> c+c+ ... +c+ ... to n terms=c,

This completes the proof.
(ii) To prove that ~+~=at for every finite cardinal a.
[Banaras 1971J
Solution. Let A and B be two disjoint infinite sets s.t.
I A 1=«-1 B I.
Then A U B is also an infinite set and so I A U B 1=1&
A n B=~ => I A U B 1=1 A 1+1 B 1
=> III = :x +1&. Proved.
Theorem 9. If afinlte set 0/ elements is added to an enume-
rable set, the reSUlting set is also enun!rable.
Or
to prove thlt n+a=l, n being any finite cardinal number.
Proof. Let A be an enumerable set. Let B be a finite set
with cardinality n. Then 1A I=a, I B I==n. Let A n B =.p.
To prove that B u A i'1 enumerable.
For this we must show that I B U A / =1 N I.
or n+a=a
Write B={blt b2 , ba.... , bn1
A, being an enumerable set, is expressible in the form of a seque-
nce. So we can write / A={an~l' an+2. anH. an+, ... }
Define a map f: N-B U -A by the formula
/(r)=l br ~f 1 <;; r <;; n
1Qr If r ~ n
Evidently r is one-one onto. Hence N "'" B U A.
This..=>_1 N /=/ B u A/=> a=n+a. Proved.
 CC::>UNTAB(LITY OF ,SETS 3S

Theorem 10. Prove n+«=~ 'V n E N, at being any infinite

cardinal number.
Or
If a finite set of elements is added to an infinite set, the power
of the set is unaffected.
Proof. Let A be any infinite set with power IX.
Let C be any finite set with cardinal number n.
Let A n C=rP.
We have to prove that 1A U C 1=1 A I.
For proving this we prove that cx+n=lX.
A is an infinite set => 3 B C A s.t. I B I==a.
Write A-B=P. Then AU C=B U PuC
or A U C=B U CUP. . .. (1)
B U C is enumerable, being a finite union of a finite set and
an enumerable set.
B U C -." N. Now 1B I=a => B,..., N, => N '" B.
B U C ,.., N, N '" B ~ B U C '" B.
Also, P '" P (by reflexivity).
Now, B U C r- Band P '" P.
Therefore B U CUP'" B U P
which => I B lJ CUP 1= 1B U P 1
=> 1B U CUP 1=1 A I : A-B=P
=> 1A U C 1=1 A 1 on using (I)
=> cx+n=« => n+GI=«.
Hence the proposition.
Theorem 11. Prove a+«=GI.
CIC being any transfinite cardinal number.

Or, If an enumerable set is added to an infinite set, the power

of the infinite set is unaffected.
Proof. Let A be any infinite set with cardinality" so that
I A 1=«. Let A n N=.p.
We have to prove that I A U N 1=1 A I.
If we show that «+a=2, ,the result will follow.
A is an infinite set => 3 B C A s.t. 1 B 1=a ..
We have A U N=(A-B) U B U N=(A-B) U (B U N)
or A U N=(A-B) U (B UN). . .. (1)
BUN, being a finite union of countably infinite sets, is
countably infinite.
:. BUN'""" N. But B -- N. .: B is enumerable.
The relation M ,..., N. where M and N are any two sets, is an
equivalence relation in the family of sets.
36 COUNT ABILIfV OF SETS

:. B --' N => N -." B (by symme try)

BUN ,.., N, N I"W B => BUN - B. (by transiti vity)
Now BUN - Band A-B - A-B. (by reflexivity)
Combi ning these tw:> we have,
(A--B) U (B U N) ,.... (A-B) U B
Using (1), we get A UN,.. , A.
This => I A U N 1==1 A ,.
=> lZ+a=~. Q.E.D.
Ex. 5. If we subtract an enumerable set from a non-enumer-
able so:t, then the remaining set is non-enumerable. [Agra 19661
Proof. Let A be a non-en umerab le set. Let B be enumerab1e
set. Let A-- B=P.
We claim P is non-en umerab le.
Suppos e the contrar y. Then P is an enume rable set.
Moreo ver A =:B U P. A, b~ing a finite union of counta ble
infinite sets, is an cnum~rable set_ A co:[Link]. For A is non-
enum~rable. Hence the require d result follows .
Ex. 6. If an enumerable set is subtracted from all enumerable
set, the rema;n;ng set .will be eltllmerab Ie.
Proof. Let A a'1d B b:>th be enume rable sets.
To prove that A - B is enum;! rable.
Let A -B=P . We have to prove that Pis counta bly infinite.
Supp0'ie the contrar y. Then P i'i non-en umerab le.
A=B U P. A is non-enumerable, being a finite union ,of a
non-en umerab le and an ~numerable set. A contrad iction. For A
is enumerable. Hence the require d result follows.
<!le.
Ex. 7. If (I. ;s any transfinite card;n~l nu-nber, then a
Proof. Let A h~ any infillite set with cardina lity It, i.e.,
1 A I ~7a:.
To prove thrtt a <; '%.
A is infinite set => 3 B cAs t. l B I=a
B C A => I B I ~ 1 A I => a ~ at.
Ex. 8. Prove that the set of all rationals is countable.
(Punjab 1969; Kurukshetra 69; Kanpur 68)
Proof. A rationa l numbe r is real numb~r expressible in the
form ~ where m and n are integers and n:;t:O.
n
We claim Q is counta ble.
COUNfABILITY OF SETS 37
Define a map I: Q+-+N X N by the formula

1(: )=(m, n)
where m and n are positive integers prime to each other. Evidently
1 is one-one. Q+ is equipollent with some subset of N x N.
Hence 'Q+ I ~ I NxN I = I N I =a
or I Q+ I ~ a. . .. (1)
Since N C Q+, :. I N I ~ I Q+ I, or 8 ~ I Q+ I ... (2)
From (1) and (2), 8 ~ I Q+ I ~ a, \\hich ~ I Q+ I =a,
Q+ is cardinally equivalent to Q- under the map T!_ -+~~
q q
:. I Q+ I = I Q- I =8
i Q I = I Q+ U Q- u to} I
=8+n+l
=(a+a)+1 :. associative law holds.
=a+l=a
:. I Q I =8, which:::> Q is enumerable and hence countable.
Ix. 9. Prove that the et 01 integers is countable.
Proof. To prove that Z is countable, it is enough to prove
that I Z I =8.
Write Z+={l, 2, 3, ...... }
Z-={-I,-2,-·3, ... }.
Then Z= Z+ U Z- U {O}.
Also Z+, Z-, to} arc disjoint sets
:. I Z I = I Z+ i + I Z- I + : to} I
= I Z+ I + I Z- I + I to} I ... (l)
Z+ "oJ N under the mapping n -+ n
Z- ,.... N under the mapping -n -- n
so that I Z+ I = I N I , I Z- I = I N I
or I Z+ I = I Z- I = I N I =8. . .. (2)
Substituting these values in (I), we get
I Z i =8+8+1=(a+a)+I=a+:=8.
:. : Z I =a. Q.E.D.
Ex. to. Prove thai a < c.
Solution. To prove that 8 < c.
We know I N 1 - a, R i =c, NCR.
I

NCR;} ! N I ~ R i :) 8 ~ C.
38 COUNTABt-LlTY OF SETS

N is not cardinally equivalent to R under any mappi·ng.

:. a:;t:c.
Now a:;t:c, a ~ c ~ a < c
Theorem 12. (Cantor's theorem) : To prove that
I A I <i I P (A) I for any set A.
Proof. Let A be any set.
To prove that I A , < I P (A) I
Write Bl = {{x} : x E A}.
Then (i) Bl C P (A)
and (ii) Bl -- A under the mapping {x} -to x.
(i) ~IB11~IP(A)1 ... (1)
(ii) ~ I Bl I = I A I ~ I A I ~ I P (A) I , on using (1)
Remains to prove that I A I ;l: I P (A) I . . .. (2)
Suppose not. Then I A I = I P (A) I which => A,.., peA)
so that 3 one-one map
onto
I: A ---l- P (A) ... (3)
Take B=={x E A : x ff. f (x)} ... (4)
x E A .. I (x) E P (A), by (3) .
.. f (x) C A, since P (A) is the family of all subsets
of A.
Also B C A, according to (4).
Thus, Be A,/(x) C A for any x E A.
This . 3 CIt E A. s.t.f(II.)=B.
CIt e: A. => 11. e: B or tx ft. B.
Consider the case in which 11. e: B. . .. (5)
Now, /It e: A. ell e: B
=> ell ft. I (a.), according to (4) .
• ell ft. B, Contrary to (5). .: I (tx)=B.
:. The possibility Ie e: B is ruled out.
Consider the second possibility in which tx fF. B. .. .(6)
Now, • e: A. => ell ft. B
... II ¢ I (tx) .: I (II)-B.
. . II e: B according to (4).
Contrary to (6).
:. The possibility CIt (/:. B is also ruled out.
 COUIIITABILITY OF SETS 39
It amounts to saying
~ E A does not imply at E B or at B. rt.
Again we get a contradiction.
It means that our initial assumption is wrong. This means
that (2) is true.
Remark. If I A I =n, (finite cardinal number) then the last
theorem implies that n < 2n.
[Kanpur 1910; Banaras 69]
": I P (A) I =2 1 A '=2 n •
Ex. 11. Let X be any set. Let A (X) be the family of all the
cizaracteristic fUllctions of X. 7 hell prove that the power set of X is
equil'alent to A (X). .
Solution. Let A (X) denotc the family of all the characteristic
functions of a sct X.
To prove that A (X) f"oJ P (X).
Let B C X be arbitrary. Let X8 denote the characteristic
function of B relative to X. Then
JI if x E B
X B: X ... {I, O} s.t. X B (x)= to if x (/:. B

Evidently X B is an arbitrary element of A (X) and B is an

arbitrary element of P (X).
Defined a mapf: A (X) - P (X) by requiring thatf(XB)=B.
This shows that f maps the characteristic func~ion of B relative
to X into the set B. Evidently f is onc-one onto.
A (X) '" P (X). Q.E.D.
Ex. 12. Prove that the set of algebraic numbers is countable.
[Punjab 65; Kanpur 72; Bhagalpur 68]
Proof. First we shall show that the set of polynomials
P (x)=aO+a1x+a2 x 2 + '" +amxm
with integral coefficients is enumerable. Let m be a fixed positive
integer. Let Pmn qenote a polynomial of degree m in which
I ao I + I a1 I a2 I + I + ... + I am I =n, ... (1)
where n is any positive integer.
Then Pm2 denotes a polynomial of degree m with coefficients
s.t.
+,
I 00 I + I a1 I I a2 I +... + I am I = 2.
40 COUNTABILITY OF SETS

Write Pm= U {Pm,. : n EN}.

Then P. is a countable union of polynomials Pm. of fixed
degree m in which the coefficients are subjected to the condition
(1). Since N is enumerable and therefore Pm is enumerable.
Write P=u {Pm: mEN}
P is enumerable, being a countable union of enumerable
sets. Evidently P is the set of aU polynomials with integral co-
efficients.
Secondly we shall show that the sct of algebraic numbers is
countably infinite.
00
Write P'= U {p" (x) : p,. (x)=O}
II-I '

where P.. (x) stands for AJlQiynomial of degree n with in,tegral

coefficients. From what has been done it follows that P' is enu-
merable.
An algebraic number is defined as the root a polynomial
P (x)=O with integral coefficients. Take
A,,={x : x is a, solution of p,. (x)=O}.
The A" is finite. SilWe a polynomial of degree n has at most
n roots.
Define A= U {Aa : EN}.
Evidently A is the set of algebraic numbers. A being a
countable union of countable sets, is countable. Since A is not
finite and hence A is enumerable.
This completes the proof
Ex. 13. To prove that the set 0.( a/l irrationals in any interval
is non~enumerable. [Raj. 1965; Agra [Link]. Statistics 65J
,Proof. We know that the set of aU real numbers (i.e. the set
of all rationals and irrationals) in any interval is uncountable. We
also know that the set of all rationals in .any interval is enume-
rable. Then the complement of the set of rationals relative to the
set of real numbers in any interval is uncountable, i.e., the set of
irrationals in that interval is uncountable. For complement of an
enumerable set w.r.t. a non-enumerable set is non-enumerable.
Ex, 14. Show that the set of a/l transcendental numbl?rs ill alH'
interval is non-enumera~/e. .
(Punjab 1965; Agra \[Link], Statistics 65J
COUNTABlLln OF SETS 41

Proof. We know that the set of all real numbers (i.e. the set
of all algebraic numbers and transcendental numbers) in any in·
terval is non·enumerable. We also know that the set of all alge-
braic numbers in any interval is enumerable.
Then the complement of the set of all algebraic numbers in
any interval relative to the sct of aU real numbers in that interval
is uncountable, i.e., the set of alI transccndentals ·in that interval
is uncountable. For the complement of enumerable set reia'jve
to a non-enumerable set is non-enumerable.
Ex. 15. Two enumerable sets are equivalent.
Proof. Let A and B ~e both enumerable sets.
To prove that A -- B.
By hypothesis, A r-- N, B "-' N.
Since the relation A """"' B is an equivalence relation in the
family of sets.
B -- N ~ N ""'"' B. (by symmetry)
By transitivity A,... N, N '""" B => A -- B.
Theorem 11. Prove that C.c=c. fBanaras 1967]
Proof. Let A be the set of all real numbers in the closed
interval [0, 1]. In symbol A=[O, IJ. Then I A I =c.
To prove that c.c=c, it is enoguh to show that
IAXA/=IAI·
Write B={(O, x) : x E A}.
Define a map I: A---'>- B by the requirement
1 (x) = (0, x) ..,. x E A
Evidently 1 is one-one onto so that A -- B.
I A i '-= / B I . ... (1)
Evidently B c A X A.
:. IB I ~ IAXA I
or I A I ~ I AxA I ... (2), on using (I>
Let x, yEA be arbitrary. Then x and y can be uniquelY
written as x=0·X1X S X3 •••
y=O·yly2Y.··.··.··.
in the form of infinite decimals which do not end wit:. zer~. That
is to say, this infillite decimal contains non-7ero digit'>.
We write i=0·4999, .... =049
instead of t=0'500
 42 COUNTABILITY OF SETS·

Define a map /1 : A x A~A by the formula

It (x, Y)=O·X 1Y\X2Y2XaYs .. ·· ..
Evidently /1 is one-One. This =>
lAxAI~IAI
Combining this with (2), we get
IA:~[AxAI~IA[
or IAxA!=IAj. Q.E.D.
Ex. 16. Find the power 0/ an aggregate 0/ numbers gil'en by
M
2m ' M and m being positive and integral. [Agra 1964J

Solutio~. Let A={~ : M, mEN}

To determined the power of A.
Write AM={~ :mE N} yo ME N
Elements of AM can be displayed as follows:
Al : ~ 1 }a ... }........
2 2 2
A2 : 2i 22 23
... 22ft ......
3 3 3 3
A3 : 2i .2i 28 ." 2ft ......

n n
23 ... 2ft ....

Obviously
(i) AT is enumerable ¥ rE N.
(ii) Ar n Ar'=r/J ¥ r, r' E N s.t. r-:pr',
...
(iii) A= U A r •
r--l
Being an enumerable union of enumerable sets, A is enumera-
ble and hence its cardinal number is a i.e., the power of the
given set is a. Ans.
Ex. 17. Show that the set 0/ points in the closed interval
[2, 4] and in the open-interval (J, 2) are cardinali} equivalent.
[Banaras 1970J
Solution. Consider the map
---r
/: [2, 41- [I, 2] s.t. f(x) = ( ,:-2 ) + 1.
f is continuous => / is onto.
 COUNTABJLlTY OF SETS 43

I'.()
f( Xl) =J~X2 ::> ~12-~+I-_X22-·2+1

=> X 1 =X 2 •
This says that! is one-one.
Thus f is one-one onto. Hence
[2,41 '" [I, 2J.
This => I r2, 4J I = I [l, 2J I . . .. (1 )
But I [1, 2J I = I (1,2) I . ... (2)
Combining (I)' and (2), we get the required result.
Theorem 12. (Dedekind~peirce). El'er)" infinite set is equivalent
to its proper subset. •
or Every infinite set can be put into a one-one correspondence wilh
a proper subset of itself.
Proof. Let S be an infinite set. Then S contains a denumer-
able subset, say,
Write S*=S-{al, a z, ... } and Sl=S-{a1}
Then SI is a proper subset of S. Also Sand S1 both are
infinite sets. Define a map f S-+SI S.t.
f (a,.)=a'+ 1 for n= I, 2, 3, ...
and /(s)=s ¥ s E S*.
Evidently / is one-one onto map Hence S is equivalent to
SI' It follows that S is equivalent to its proper subset.
Note. This fact was used by Dedekind to define infinite sets
as follows.
A set is iJ~rinite iff it call be put hlto a one-one correspondence
with a proper subset of itself.
Theorem 13. (GaliJeo's Paradox). Any denumerable set can be
put into a one-olle correspondence with a proper subset of itself.
Proof. Let A be a denumerable set. Then A is expressible as
A={at. a3 , aa,· .. }.
Write B=A-{al}={a2 , aa, ... }
Define a map f : A-+B s.t.
f(a,.)=o"+1 forn=l, 2, 3, ...
Evidently f is one-one and onto map.
Also B is a proper subset of A.
Hence the theorem.
44 COUNTABILl1Y OF SETS

Theorem 14. Every nOll-empty open set 011 real line is the
countable union of nan-empty disjoint open intervals.
[Meerut 1988,86]
Or A bounded set of lion-overlapping intervals is countable.
Proof. (i) Let G C R be an open set and x E G be arbitrary.
Then 3 an open interval I", with centre x such that x E I., C G.
Let C. denote union of all such open intervals. Then C.,= u tI",CG}.
Evidently COl is the largest open interval containing x such that
x E C., C G.
Moreover, each x EGis contained in /'" and each / .. is
contained in G. The suggests that G= U C"'.
We have either CIII n C,=t/> or C., n CY=l=~, ¥ x, y E G.
Consider the case CIII n Cy::j::.cp. Then C", U Cy is an open interval
containing both x and y such that
C.. C (CIII U Cy) C G, C, C (C", u C,) C G.
By definition of Ce; and C y , we have
(Co: u Cy) c C.. ' (CIII U Cy) C Cy.
Hence C",=C~ U Cy and Cy=CIII U ell, i.e. C.. =Cy
Thus, '" x, y E G, we have either Cm=C, or CIII n C,=r/I,
Also G=U Cm•
This declares tl-at G is the union of pair-wise disjoint non-
empty intervals.
It remains to prove that the intervals C", form a countable
collection.
(m Let G={G,.} be a collection of open subsets of R such
that (i) G,.:/ t/J. 'V n, Oi) G.. n Gm=t/> for Il ,em.
We know that each G •• contains an infinite number of rati-
onal and irrational numbers. Let K=lK,,}, where K" is the set of
all rational numb;!rs in Gn. D::fine a map
f: K - G such thatf(k,,)=G n •
. .since for each k .. E K, there is a unique open set G.. E G,
md conversely. :. f is one-one onto. Therefore K -- G.
But K considered as the set of rational numbers U KA , is
.::ountable. Since K C Q. Hence G is countable.
Theorem 15. ~r {G,.} is all)' cdlectioll ~ non-empty disjoinl
open sets ofR, then prOl'e that (G,,} is c(Juntable. {Kanpur 1979, 771
Proof. For proof, sec (ii) part (Of '[ h.:orcm 14
Problem 18. SIIolI' ,IIat a cOllllfahle set is a Boret set.
COUNTABILITY OF SETS 45

Solution. Let A be a countable set so that A is expressible as

A={a1 , a2 , a~, ... }
{ar}={x: x=ar }

= U ~ x : ar ~ x < an + ~}
"=1 ( n
=Countable inters!?Clion of closed sets.
Also A= U {or}
r EN
These facts prove that A can be
obtained by the formation of countable union and intersection
of closed sets and open sets and hence A is a Borel set.
5'10. Continum Hypothesis.
It states that there exists no set having its cardinal number
between a and c. But this hypothesis does not assert that there
exists no set with cardinal numb.;r greater than c. In this con-
nection Cantor had proved that the pow.:!r set peA) of any set A
has cardinal number greater than the cardinality of A. Conse-
quently the cardinality of P (R) is greater than c.
 6
Measure
And
Outer Measure

6'0. Def. Boolean ring. (or ring of sets). [Gujarat 70]

A non-empty family A of subsets of X is called a ring 'of sets
or B-ring if
any A, B E A ~ A-B E A, A U BE A.
That is to say, ring of sets is closed under the formation of
unions and differences.
Remark. (1) A n B=.A.-(A-B), A t:. B=(A--B) U (B - A),
where A t:. B stands for the symmetric difference of A and B.
These equations dec/are that A n B E A, A t:. B E A if A is a
ring of sets.
(2) ", E A. <·For .=A~A.
(d. DeC. (1-ring of sets. [Gujarat 70J
A non-empty family A of subsets of X is called a (1- ring of
sets if it is closed under the formation of differences and count-
ab'e unions, i.e., if
A, B E A ~ A - B E A
00

and A, E A ~ U Ai E A.
~~1

00 "" eo
Remark. n Ai=A- U (A-A.) where A= U Ai.
i=1 ,=1 i=1

This equality says that (1- ring is also closed under the for-
mation of countable intersection.
Examples of ring. (i) P (X), the power set of X, is a-ring
where X is any set.
(ii) If X is any set, then class of all finite subsets of X is a
ring. This class is (1- ring if X is finite.
MEASURE AND OUTER MEASURE 47
62.' Def. Algebra of sets (or Boolean Algebra or Field). A
family A of subsets of X is called an algebra of sets if
(I) A E A => A' E A where A'=X-A
(2) A, B E A .. A U B E A.
6'3. Def. a-algebra of sets (or a-field). A non-empty family
A of subsets of X is called a-algebra of sets or a-Boolean Algebra
of Borel Field if
(i) A E A => A' E A
GO

(ii) Ai E A => U At E A.

Remark. (1). Each algebra (a-algebra) A is a ring (a-ring)

because if A and B are in A, then A-BEA, since A--B=(A'u B)'.
(2) A ring A of subsets of X is an algebra in X iff X E A.
For if A is an algebra, then X = A' U A where A E A so that
X E A. Conversely if X E A, then A' E A ¥ A E A. For
A'=X -A so that A is an algebra.
6'4. Der. Semi-ring. A family A of subsets of X is called a
semi-ring if (1) tP E A and
(2) A, B E A=> A-B E A
(3) AI E A S.t A, n A1=f# for i-:f::j .. 3 A E A
GO

6·S. Det. Monotone class. A non-empty class of sets A is

called a mO:lOtone class if for every monotonic sequence < AI >
of sets in A, we have lim An E A.
Examples If A.,. C A.,. pi, then <An> is an increasing sequence.
00

For increasing sequellce, we write <An t >. If A= u An, then

n-l
we denote it as An t A. If An+-! C An, then < An > is decreasing
GO

sequence and we write <;A,. ~ > and if n An=A, then we write

ft=J

An.j, A If M is a monotone class, then AE M for every sequence

<An> in M.
Any a-ring is a monoton~ class. A monotone class is a a-ring
iff it is a ring.
, 6'6. DefinItion. Complete lattice. The set
R.=R U {-co, ~}=[-oo, co]
48 MEASURE AND OUTER MEASURE

is called Extended set of real numbers. Briefly R. is called com-

plete lattice.
6'7. Definition. Set Function.
Refer to D!finition (3'4), Chapter 3, on p~ge 13.
6·S. Definition. Extended real valued set function. Refer to
Definition (3'6), Chapter 3, on page 14.
6'9. Definition A fU!1ction (set function) is said to be finite
function (finite set function) if all its functional values are finite.
Def. Additive Class. [Kanpur Statistics 75]
Let Q be a space and A be a cla~s of subsets of ~:I. A is said
to be finitely additive class of sets if
(i) cf, E A
(ii) A, B E: A ~ A-B E: A, A U BE: A.
This definition is parallel to the definition of ring of sets.
Hence we can say that every ring of sets is finitely ajditive class
and vice-versa.
Def. Completely Additive Cla..s. (Kmpur Statistics 77, 75)
Let Q be a space. The class A is said to be a complete ly
additive class of sets if
(i) r/J E A
(ii) A E A '* A' E: A
(iii) If <An> is any sequence in A, then U An E: A. All
these condition<; prove that Q E A.
Hence every completely additive class might b~ a field or
a-field and vice-versa.
DeC. Completely Additive Set Function.
[Kanpur Statistics 77, 75]
An extended real valued set function I-' is said to be comple-
tely additiv~ set function if I-' : A....,. [ - 'X> , <Xl] s. t.
(i) A is completely additive class.
(ii) I-' (cp)=O.
(iii) Ther\.. e;<ists a sequent;e <All> of disjoint sets in A S.t.

I-' ( ~l Ai )= igl I-' (Ai).

Def. Continuou.. Function. Let J.I. be a measure function on

the ring A. Then J.I. is additive, positive and monotone. Now if
< E" > is an incre:lsing sequence, then <I'(E,,) > is also increasing
sequence. Moreov!r if E" l E and En, £ E: A, and ",(£,,) t tl-{E).
MEASURE AND OUTER MEASURE 49

We express this condition by using that ~ is continuous frell!

below. Similarly ~ is continuous from above if En ~ E.
"10. Definition. Let X be a set. Let A be a a-ring of seu
which are subsets of X•
.[Link] m be an extended real valued set function defined on tke
g-ring A.
(1) m is called non-negative if m (A) ;;;;, 0 V E A.
(2) m is called additive if
¥ A, B E A s.t. A n B=1> => m (A U B)=m (A)+m (B).
(3) m is called finitely additive if
A. E A for i=l, 2, ... , n S.t. A. n AI=1> for i=l=j =>

m( UA,
1=1
)= i: 1=1
m (A,)

(4) m is called Countably additive if

A, E A ¥ i E N s.t. A,nAI=1> for i:;f;j ~ m( UA,)=;m (A,)

i=l i=l

(5) m is monotone if
¥ A, B E A s.t. A C B => m (A) EO;; m (B)

(6) m is subtractive if
A, BEA s.t. A :J B, m(A) is finite => m(A-B)=m(A)-m(B)
(7) In is Countably subadditive if

A, E A ViE N => In ( '~1 A.) ~ '~l m (Ai)

(8) m is translation invariant if
A E A, x EX=> m (A+x)=m (A), where
A+x={y+x : yEA}
(9) m is regular if
AEA, E > 0 => 3 G, FE A S.t. m (G)=E~m(A) ~ m(F)+"
6'11. Postulates for an ideal measure function.
A measure is an extended real valued set function m defined
on g·ring A of subsets of X s.t.
(I) meA) ~ 0 ¥ A E A. In particular m (1))=0.
(2) A, B E A s.t. A C B ~ m (A) ~ m (B)
(3) A, B E A s.t. A n B=1> => meA U B)=m(A)+m(B).
(4) For an intervall,
m(l)=l (/)=length of the intervall.
so MEASURE AND OT,!!.ER MEASURB

(5) If <A,> is a sequence of pai rwi se d i sjoi n t sets in A, then

m (u )
i=1
A, = E
'=1
m (A,)

(6) m is translation invariant, i.e.,

m (A+x)=m (A) y. x€X,
where A+x={y+x: yEA}.

-
6'12. Definition. Measurable space. [Kanpur 1969]
If A is a subsets of X and m is m eas ure on A,
algebra of
then the tripl e (X, A, m) is called a measurable sp ace.
If m assumes only the values 0 and 00, then the m ea sure
space (X, A, m) is called degenerate.
Any set A E A is called a measurable set ( relative to the
measure m).
6'13. Definition. Let m be a measure function on a a-ring A
of subsets of X.
(1) Completeness of me�sure function. Th e measure function
m is said to be complete if
B e A s.t. m(B)=O, A C B =- A � A
i.e., if all the subsets of a set of me asu re zero are measurable. In
this case (X, A, m) is call'Jd a comple te measure space.

(2) An y set A€ A i9 said to have finite measure if m(A)< IX..,

(3) The measure of any set A e A is said to be C7-finite if
3 a sequence < A" E R: n EN> s.t.

(i) A C U A..
00

11=1
, (ii) m (A,,) < 00 \j n e N.

(4) The measure fu nct io n m is cal led finite or a-finite accor­

ding as the measure of every set is finite or a-finite.

(5) The measure function m is called totaUy finite (totally

a-finite) if
(i) X e A i.e. A is algebra of sets.
(ii) m(X) is finit,e, (�r a-finite).
6''(4. Herid �[Link] froperty. A n:1n-empty class S={Si} of
sets is said to. have. heriditary if
SI C S,. S� e S => SI E S.
6'15. Caratbeedory's postulate for outer measure.
[Kolbapur 70; Banaras 66,65]
o

Let an extended real valued set fu nction mo defined on 'U

MERSURE AND OU fER MEASURE 51
heriditarya-ring A={A.} have the following properties:
(i) mo(A,) ~ 0
and mo(Ar)=O if A,=p
(ii) Ar C A. <=> mo(A,) ~ mo(A.)
(iii) mo (u Ar)
r
<r
1:mo(Ar)

(iv) mo (A, u A.)=mo (A,)+mo (A.) provided A, n A,=rfo.

These properties are called Caratheodory postulates for outer
measure.
6'16. Definition. Measurable set.
[Banaras 71; Poona 70; Kolhapur 70]
Let m~ denote an outer measure function defined on a heri-
ditary a --ring A={Ai}.
Any set A E A is said to bc measurable w.r.t. the given outer
measure function mo if m~(T)=mo(A n T)+m.(A' n T) V TeA
or equivalently, mo(T) ;;;?; mo(A n n+miA' n T).
If we say that l.I. set A is measurable, then we shall write
mo(A) =m(A).
Remark. FO'r any set E,
E=A n
E+A' E. n
In accordance with Definition (6' 15), this gives
nlo(E)=mo(A n E+A' n E).
<: mo (A n E) +mo(A' n E).
i.e. 1I10(E) ~ t11 o(A n E)+mo(A' n E).
To prove
mo(E)'-~mo(A () E)+mo(A' n E),
it is enough to provc that
1I1o{E)?-lIl o(A n E)+mo(A' n E).
6'17. Elementary set. A plane set is said to be an elemen-
tary set if it is expressible as a finite union of pairwise disjoiat
rectangles.
Problems related to measure functions.
Theorem I. A a -additive set function is a continuous set
function. [Kanpur Statistics 77J
Proof. Let p. be a a -additive set function <A,,> be any
sequence of sets in the class A on which the a-additive set func-
tion is defined, where A is an algebra.
(i) Let <A" t > be an increasing sequence in the class A
so that Al C A2 C A3 C ......
52 MEASURE AND OUTER MIlASURE

-
Then U An E A.
,,=1
Let B1 =A1 , Dn=A .. -An_1 for n
_ 00
> 1.

Then D. n Dm=rP for n#m and U D,,= U An,

_1 ,,=1

k ~

Ak = n Dn and so I-' (A k )= E I-' (Dn)

w=1 "=1

... (1)

or I-' (lim An) = Lim I-'(A,,)

n .. oo
This => II- is continuous from below.
(ii) Let <A" '" > be a descreasing sequence, then
At :::> A2 ::> A3 :::> A, :l...... Then
co co
AI-An C A t -An+1 and A 1 - U· A,.= U (At-An) .
.. =1 .. =1

Hence, by part (i),

I-' ( Al-n~l An )= I-' [ "~1 (A 1 ·- An) ]

Lim
= n~oo
I-'{AI-An), by (1)
Lim
=p.{A1 ) ... I-'(A..)
Il-+oo

or "'(AI) --p.( U An
n=1
)=I,(A1 ) - Lim
n-+oo
I-'(An)

or I-' ( U An
,,=1
)= Lim
n-+ 00
p.(An)

or I-' Lim (,1)= Lim (A ).

n-oo nn n _; 00 I-' ..
This ~ II- is continuous from above. Hence po is a continuous
function.
MEASURE AND OUTBR MEASURE S:f

Theorem 2. Let X be a space 0/ at least two points and

Xo e X. For each A C X, we define

~
(A)=JO ~f
11 If xoeA
x,,.
A

then prove that ~ is an outer meQsure. [Kanpur Statistics 1976]

Proof. Let A, Be X be arbitrary. Then
(i) ,...(A);il O. This follows from the definition of p. In
particular ~(.p)=0 as Xo l/:. ~ Q' ",(.p)=0.
(ii) A C 8 => ",(A) ~ ",(8).
For A C B => (l).xo EB if Xu e A
(2).xo ft B if x. ft A
... I'(B)=l=,...(A) if Xo E A
& 1'(8)=0=I'(A) if Xu ft. A
=> I'(A)=I'(11) if Xo E A or
Again if x. e: B, then Xo e: A or Xo ,. A.
x.,.
A.

Hence if Xu e: B, then ",(A)= 1 or O.

This ... ",(A) ~ I'(E). For 0 < 1.
Finally A C B ~ I'(A) ~ p.(E).
(ii) To prove ",(A U 8)=I'(B)+I'(B) if A n B=.p.
We have Xu ¢ A U R or Xo E:: A U B
x. ,. A U B => Xu ~ A and x. ~ B.
Conscquenrly I'(A U B)=O => ",(A)=O=",(.8).
:. ,..(A U 8)=I>'(A)+I'(B).
In the second case,
Xu e A U .B => Xo E A and Xo ¢ B
or Xu e: B and Xu ¢. A,for A n B=rf..
:. I'(A U B)=1 => ,.(A)= 1,1'(8)=0
or I'(B) = I, 1'(04)=0•
. This => I'(A u B)=,...{A)+I'(B).
Hence in both cases, (iii) is proved
{iv) To prove p.( U A .. ) C;;; E 1'(14,,), where A" C X.
" "
We have x. ¢. U A. or Xo E U A..
If Xo ~ U A", then x. ~ An ¥ n
:. If 1'( U A.. )=O, then ,,(...4,,)=0 ¥ II so that

..
i: ~(A,,)=O+O+O+ ... +0=0

:. p,(U A .. )=1: ",(A,,).

" II

In the second case if Xo E U A", then x. e A

n.
for at least
one value of If or Xo E A" V- ft.
 MEASURE AND OUTER MEASURE

i.e., p{A )= 1 or p{An)= 1 ¥ n.

no
Hence if p( U An)= I, then
.
E p{An )=O+1+0+ ... +0=1
n

00
or l.: (J.(An) = 1 +l-f-} + ... + 1=00.
"""I

This ~ p (U An)=E P{An)

or p (U An)<E p.(An)
~ p(U An)~Ep(An)'
From (i), (ii), (iii) and (iv), it follows that p is an outer mea-
sure function.
Theorem 3. An outer Ineasure is monotonic and a - sub additive.
[Kanpur [Link]. Statistics 1976; Banaras 69]
Proof. Let A be a a --ring of sets which are subsets of X.
Let (X. A, mol be a (outer) measure space.
(i) To prove that mo is monotonic.
For this we have to show that
lno(A) IE;; lno(B) ~ mo(C) ::;:;; me(D) ~ ......
if A C Be C cD ..... .
Consider the case in which A C B.
Notice that B=(B -A) U A, (B-A) n A-t/l.
This::> mo (B)=lno (B-A)+mo (A) ~ mo (A).
For Ina (B-A) ;;;;;, 0
::> mo (B) ~ mo (A) ~ Ino (A) ~ Ino (B).
Similarly B C C, C C D
. ::> Ino (B) ~ mo (C). mo (C) ~ Ino (D).
Combining these results
mo (A) ~ 111" (8) ~ mo (C) E; Ino (D) <;; ......
if ACBCCCDC ......
(ii) To prove that 1110 is a --subadditive.
For this we have to show that

mo (U 1=1
AI) :s:;; .E mo (A.), where <
.=1
All > is a sequence of
sets in A.
8-1
Write Bn=An- U A,
i=l

Then
MEASURE AND OUTER MEASURE 55

00 00
This shows that => (1) U E,= U At
1=1 i=l

(2) E, n BI=~ for i~j.

(3) B. c A.. ¥ n.

(2) => mo (,~ Et)= ,~ mo (D,)

=> mo (U A,)== 'f

i_I t=l
1110 (E,) by (I)

00
~}; 1110 (A,) by (3)
,=1

mo (u AI) ~ rr:
• --1 1=1
mo (A,) .

Theorem 4. If T be a subset of the set of real numbers s.l.

0< ma (T) < 00and if p.(A)=m. (A n T) 'V A C R, then show
that p. is an outer measure (Caratheodory's outer measure).
Proof. Let A, B, C, T C R be arbitrary, but Tis fixed and
o< mo (T) < 00.

Also we define Ii real valued set function I' as follows:

p.(A)=mo (A n T).
To prove that p. is outer measure.
(1) p.(A) ~ O.
For mo (A n T);;' 0 .. I-'(A)=mo (A n T) ;;. 0
* p.(Al ~ O.
(2) f£(~)=O.
For A.=~ ... A () T=rp'" Ino (A n T)="1o (rp)=O
~ I-'(A)=mo (A n T):o=O ~ p.(A):za:O.

(3) ACE * peA) ~ p.(B)

For ACE ... A nrc B n T
n
* mo (A T) ~ mo CE T) n
c> p.(A) ~ p.(B).
(4) I'(A U B)=p.(A)+p.(B) if A n E=rp
For A n B= if> ~ (A n T) n (B n T)=,p
*mo [(A n T) U(B n T)]==mo (A n T>+mo (8 n 'f)
Q p.(A)+p.(B)=mo [(A U D) n TJ=p{A CJ B)
c> I'(A U B)=p(A)+I.L(B)
56 MEASURE AND OU'fER MEA~U1U!

(5) 1'( U A t,) <; E I'(An)

" /I

For I'(U An)=mo [(U An) n T)J=lIlo [U (Au n T)]

" " n
~ X mo (A n T)= X I'(An)
" ..
..• 1'( U An} E;; E p(A..)•
" "
From (1), (2), (3), (4) and (5), it follows that I' is outer
measure.
Problems related measurable sets.
Theorem 5. A set is outer measurable iff A' is outer measurable.
Proof. Let a set A be outer measurable so that for any set T,
mo (T)=mo (T n A)+mo (T n A')
=mo (T () A')+mo (T n A)
=mo (T n A')+mo [T n (A')'].
This proves that A' is outer measurable.
Conversely suppose that A' is outer measurable so that for
any set T,
mo (T)=mo (t n A')+mo [T n (A')']
=mo tT n A')+mo (T n A)
-mo (T n A)+mu (T n A').
This .. A is outer measurable. Proved.
Theorem 6. The union of two outer (mo) measurable sets is
outer measurable. [Meerut 1972; Banaras 65J
Proof. Let SI, S2 be any two outer measurable sets.
To prove that SI U Sa is outer measurable ..
Taking T as any set and applying the criterian of measurabi.
iity of SI
mo (T)=m o (T n Sl)+mO (T n SI')· ... (1)
Taking T n SI' for the set T and applying the criterian of
neasurability of S., we get
mo (T n Sl'),=mO (T n SI' n S2)+mO (T n SI' n S~').
Using the fact that (SI n S~)' -SI' U S~'.
we get
mo (T n S/)=rna (T n S/ n S2)+m O IT n (St U S3}·J. .. (2)
MEASURE AND OUTER MEASURE

Lastly taking T n (SI u S2) for the set T and applying the
criterian of IlIeaslilrability to SI. we get
n70 [Tn (SIU S2)]=m O [Tn(SIU S~) ns\]
+1»0 [Tn(Sl uSII) ns1']
Making use of the fact that
(SI U S2) n 8 1 =SI
(SI US2)nS1 ' =(SI n S l')U (S2nSl')
=¢U(S2 nSl')
-S2 nSl'
we have
mo [Tn(SI U S2)]=n70 (TnS1)+m O (TnS2 nS1')
using (2), we get
mo [Tn (SI u S2)]=mO(Tn SI)+mO (Tn SI')]
-mo [Tn(SlUSlI)']
using (1), we get
mo [Tn(SlUS2 )]=mO (T)-m o [TU(SIU S2)']
i.e .• mo [Tn(SIUS~)]+mo [Tn (SlU S2),J :::atno{T).
This proves that SI U S2 is outer measurable. Proved.
Problem 1. The intersection oj two outer measurable sets is
outer measurable.
Solation. Let AI' A2 be any two outer measurable sets.
To prove that Al nA 2 is outer measurable.
AI' A2 are outer measurable sets.
=;> A/, A2 ' arc outer measurable sets.

=> AI' U A/ is an outer measurable set.

~ (AI' U A 2 ')' is an outer measurable set.
~ (AI')' n (All') is outer measurable.
=> Al n A~ is outer measurable. Hence the result.
Probl(m 2. ~r SI' S2 are our;'/, measurable sets and S2 CS1'
thell show I haf SI- S2 is outer measurable set.
Solution. Let SI and S2 b~ outer measurable sets and let
S2 CSI
To prove that SI-S2 is outer measurable.
S1> S2 are outer measurable sets
=> SI. S2' are outer measurable
=> $, ns/ is outer measurable ... (*)
Sln..s~'=Sln(X -SJ
=SlnX- SlnS2=Sl-SZ
•
58 MEASURE AND OUTER MEASURB

:. SI- S2 is outer measurable, by (:It).

Problem 3. If SI and S2 are measurable sets and if S2 n81 -cp,
then m (S1US2)=m (SI)+m (S2)'
Solution. Let S10 S2 be measurable sets so that
mo (Sl)=m (SI)' mo (S2)=m (S2)'
and mo (S) US2 )=m (SIUSa), [Prove this as in Th. S]
Also suppose that SI n Sa=CP'
To prove that
m (SluSS)=m (81)+m (Sa)'
By the postulates of Caratheodory outer measure,
mo (SI U 8.)=mo (SI) +mo (Sa),
In accordance with initial assumption, this takes the form
m (SlU S2)=m (Sl)+m (Sa).
Alternative method. Taking S1US2 for the set Tand applying
the criterian of measurability to S1> we get
mo (SIU Sa)=mo [(SIU Sa) nSl ]+mO [SI'() (SIU S2)]
=mO(SI)+mO [(SI' nSl )u (St' nSz)]
=mO(SI)+m O [cpU S2n(¥-Sl)]
=m o (Sl)+mO (S2) j For Sl()Sa=rp -> S2eX-SI
=mo lSl)+mO (S2) 1 => S2n(X - SI)==8.
In accordance with initial assumption this takes the form
m (SlUS2)'=nI (SJ+m ('sa).
Theorem 7. If the outer measure of a set is zero, then the set
is measurable. [Banaras 1971]
Proof. Let A be any set such that mo (A)=O
To prove that A is outer measurable.
For any set T,
T=AnT+A'nT.
By Caratheodory postulates for outer measure,
mo (T)=mo (An T,A' T) n
t;;; mo ~AnT)+mo (TUA')
or n
nlo (T) E>;; mo (A T) +mo (TnA') .,.(1)
': TnAC:A
. mo (TnA) ~ nlo (A)-O
"
mo (TnA) <; 0
But mo (TnA) ;JP 0
:. 0 ~ mo (TnA) ~ 0
Consequently /)/" (TnA)-O " (2)
MEASURE AND OUTER MEASURE 59

Again TnA'cT
:. Ino (TnA') ~ mo (T)
using (2),
mo (TIJA')+mo (TnA) ~ Ino (T)+O
or Ino (TIJA')+mo (TnA) ~ nlo (T). ",(3)
Combining (I) and (3), we get
Ino (T)=m o (TnA' )+1110 (Tn A).
This proves thnt A is outer measurable.
Problem 4. Show that every subset of a set of measure zero is
measurable and its measure is equal to zero.
Solution. Let A be a set of measure zero so that A is measu-
rable and In (A)o~O. Let BCA be arbitrary.
To prove that B is mcasur"blc and In (B)=O.
BCA, m(A)=O => m(B) ~ m(A)=-O => m(B) .s;;; 0
But m( 8)~ 0 ¥ B.
m(B) ~ 0, !nCB) ~ 0 => m{B)=I.J.
Since any set of measure Zero is measurable and hence B is
measurable. Proved.
Problem~. Let A be a mo measurable subset of B,
< <'''', A C B, then
If W is any set with mo (W)
mo [WnBnA']=mo (WnB) - Ino (WIJA)
Solution. A is mo measurable =>
mo (T)=11I0 (TnA)+mo (Tn A ) ¥ T
Taking T= Wf"lB, we get
mo (WnB)=mo (WnBnA) tmo (WnBIJA')
or mo (WnB) - n1 0 (WnA)=mo ([Link]"lA')
r:r.;
ACB -=> AnB=AJ.
Problem 6. !f A is mo-measllrable subset with n1 0 (A) < 00
and B is allY set cOl/tail/jllg A, then
1110 (BnA )=nlg (B) - mo (A).
Sohlfiou. Suppose ACE and A is outer measurable so that
mo (T)=mo (TIJA)+m o lTnA') ¥ S.
Taking T=B, we get mo (R)=/n o (Bf'lA)+m o (BIJA')
01" 1110 (B)-mo (A)=1110 (RnA') as ACB ~ A f'lB=A.
Problem 7. Let A be a measurable set. If B is any set,
then /I/O) lA U R)+I11(j (AnB)=-lilo (A)+m o (B).
(Meerut 1972; Kanpur 70J
60 MEASURE AND OUTER MEASURE

Solution. Suppose A is measurable so that

m. (1)=mo (TnA) + mo (Tn A')
Taking T=B and T=AUB, we get two equations
mo (O)=mo (BnA)+m o (BnA'). ...(1)
"'0 (AUB)=mo [(AUB)nA]+m o [(AUB)nA']. ..(2)
Evidently (A U B) n A=A, and
(AUB) n A'=(A nA')u(BnA')=~ U (BnA')=BnA'
Putting these values in (2),
mo(AuB)=mo(A)+mo(BnA') ... (3)
(1)-(3) gives
mo(B)-mo(AUB)=mo(BnA)-mo(A)
or mo(B)+mo(A)=mo(AuB)+mo(BnA).
Der. Let (X, d) be a metric space and 14* be an outer mea-
sure on P(X) S.t.
p.*(AUB)=p.*(A)+p.*(B) whenever d\A, B) > O.
Then 14ft. is cailed outer measure.
Theorem 8. To show that an open set in a metric space is mea-
surable with respect to any outer measure. tBanaras 1965J
Proof. Let (.Y, d) be a metric space and G be an open sub-
set of X.
To show that G is outer measurable w.r.t. metric d.
Let ECGCX and G'=X-G.
lJcfiue EO!'-~{XEE: d(x, G') ~~}- ... (1)
From this it follo\\s that < E.. > is a monotonic sequence
hence Lim t .... exists.
Also Lim E.. =E.
For allY set TC X.
T=TnG-t-TnG'. ...(ft.)
By Caratheodory' postulates for outer measure.
ma(T)=m\l(TnG-t-1'nG') ~ mo(TnG)+mo(TnG')
i.e. moU') '- mo( . nG)+mo(TnG') ... (2)
Sct Tn G=e
Then T=,EU(TnG'):JE.. U(TnG'), by (1)
(110(T) ~ 1110 [Ly, U (TnG')]
:. ... (3)
Evidently d (En, TnG') > O.
This follows frum our construction (1).
Then 1110 lE.. U (TnG')]=m o (En)+m o (TnG')
Making n ..... oo.
MEASURE AND OllTER MEASURE 61

Lim rn [E" U (T n G')]= Lim rno (E.. )+mo (T

n-+oo
n G')
=mo (E)+rno (T n G')
=lno (T n G)+mo (T n G') ... (4)
Making ll-oo in (3) and using (4),
mo (T) ;;;;, rno (T n G)+rno (T n G').
Combining this with (2), we get
rno (T)'=rn o (T n G)+mo (T n G').
This => G is outer measurable. Proved.
Theorem 9. If < E,. > is monotonic non-increasing [Link]
of measurable sets (outer measuralyle sets), then the limit set
('~

E= n Ek is a measurable (outer measllrable) set and for every T

k~l

ojfinite outer measure,

Lim
rno (T n E)= rno (T n En).
n-co
Proof. (i) To "how that Lim rno (T
n-+oo
n E.. ) exists and

Lim rno (T n En)=m o (T n E).

n-+ 00
L':!t T be a set of finite measure.
Since El ::J E2 ::J £3 ::J ......
It follows that T n £1 ::J Tn E2 :J T n £3 ~ ..... .
Therefore 1110 (T n EI ) ~ Ino (T n E2 ) ~ Ino (T n E3)";;;.··.
Consequently < 1110 (TE,,): /lEN> is monotonic non-increa-
sing sequence. In brief we shall denote Al n A2 n Aa n ... , by
AIA2Aa ..... Similarly the set A U B U C is denoted by A+B+C.
Also every member of < rno (TEn) : n EN> is non-negative.
Hence this sequence has a limit so that Lim 1110 (TEn) exists_
[Link]
00
~n~C~=>EC~=>TnEcTn&
"=1
=> TE C TEn => rno(TE) ~ rno (TE,,)
Lim Lim
=> n--".>oo rno (TE) < n-+oo rno (TEn)
·
T ak mg Lim
n_---
rno (TEn)=},., we get rno (TE)
'IV
< .\ •• ·.(1)
62 MEASURE AND OUTER MEASURE

The set T can be broken as

T=TE+TE1' + E1TE2 ' +E2 TEa'+·· .. ··+&.-1 TE..' ...... (*)
Making use of Caratheodory postulates for outer measure,
Ino (T) ~ mo (TE)+m o (TE1')+m O (EITE2')-i-mo (F2TEs')
...... +mo (£"-1 TE,,') ... (2)
Since each En is outer measurable,
mo (T)=nto (TE.. )+mo (TEn')
Ino (T)=rn o (TE1)+mO (TEl') ... (3)
Taking E'II-l T for the set T and applying the condition of
meac;urability to En, we get
mo (EII - 1 T)=rn O (E.._1TE,,)+mo (En- 1 TE.. ·)
or rno (E.- 1 T)=rn O (EnT)+rn o (En_lTEn')
[For En-I::> En => En n En_1 =EnJ.
or rno (E"-lTE,/)=rno (TEn- 1)-rn o (TEn). ...(4)
Writing (2) with the help of (3) :md (4)
rno (T) ~ 1110 (TE)+[rn o (T)-mo (TE 1)J+[rnO (TEl)-rno(rE~)]
+[nto (TE2)--mO (TEa)]+· .... ·
=[mo (TEn-I) -rno (TEn)]
=m o (TE)+rno (T)-mo (TEn).
Taking limit as n-'t-oo,
mo (T) ~ mo (TE)+m o (T)- ,\
or ,\ ~ rno (TE). ... (5)
Combining (I) and (5), we get
A=rno (TE)
i.e. Lim
lila (TE,,)=mo (TE).
n-'?oo

(ii) To show that E co= n

,,=1
En is measurabl:::.

For any set T C .Y, we have

T=TE+TE' ... (6)
From which we get
rno (T) ~ mo (TE)+rn o (TE'). ...(7)
From (6), TE'=T-TE
using this in (*), we get
TE' =T' E1 ' + El TE2 ' +~ TEa' + ..... + E..~1 TEn'
This gives
mo (TE):::;;; mo (TE1')+mo (E1TE/)+m o (E2 TE3')+ ......
+ mo (t~'-lT£,.. )
MEASU!,E AND OUTER MEASURE 63

Writing this with the help of (3) and (4),

mo (TE') ~ [mo (T)-mo (TE 1 )]+[mO(TEl)-(mO (TE2)]
+[mo (TE2)-rno (TEs)]+···+ [rno (TE"-I)--m O (TE,.)l
=mo (T)- rno (TEn)
Making n--l>-rrJ,

mo (TE') ~ m (T) -- Lim rno (TEn)

n-oo
or rno (T) ~ rn (TE') + Lim rno (TEn).
n-oo
But Lim mo (TE,,)=m o erE)
or Ino (T). ~ rno (TE'H-1I10 (TE)
Combining (7) and (8), we get
rno (T)=mo (TE')+mo (TE)
This ~ E is outer measurable.
Problems related to ring of sets, a--algebra.
Problem 8. A Boolean ring B containing X (whole space) ;s an
algebra
Solution. Let B be a Boolean ring s.t. X E B.
To prove that B is an algebra we have to prove that
¥ A eX=> A' E B
A, X E B => X --A=A' E B, by def. of ring.
Problem 9. Intersection of two B-rings is a B-ring.
[Gujarat 1970]
Solution. Let 8 1 and B2 be two Brings.
To prove that 8 1 n B~ is a B-ring, we have to prove that
A, B E Bl n B~ => A-B, A U BE Bl n B2
A, B E Bl n B2
~ A, B E Bl and A, B E 8 2
=> A -- B, A U B E 8 1 and A -B, A U B E B2
~ A-B, A U BE 8 1 n B 2 •

Problem ]0. EI'ery algebra is a ring.

Solution. Let A be an algebra. To prove A is a ring, we have
to prove A, B E A=> (1) A -B E A. (2) A U Be A.
(2) follows from the definition of algebra.
A, B E A ~ A', B'. BE A, by def. of algebra.
~ A' U B E A ~ (A U B), E A
~ AN n B'=A n B'=A-B e A
64 MEASURE AND OUTER MEASURE

Theorem to. A mOll%ne ring j..s a a-ring.

Proof. Let A be a monotone ring. Let Ai E A ¥ i. By
assumption A is closed under the formation of finite unions and
n
so Al , Al U A 2 • ••••.• , U Ai etc. belong to A.
1=1

Write B,,-=.· U" Ai' Then Bn E A. Also < Bn > is an increll.-

t=1

00 00
sing sequence and U B,.= U An.
"=1 n=1

A is a monotone cIa&s => lim Bn E A

00
~ U An E A ~ A is a a-ring.
n=1

co 00
For lim Bn = U B .. =.JJ An.
n=1 n=1

Theorem 11. A a field is a monotone and monotone field is a

a-field.
Proof. Let A be a a-field and < An > be a monotonic sequ-
ence in A.
00
(i) If < A,. > is increasing sequence, then lim A.. = U A.. EA,
n=l

from the definition of a-field so that A is a monotone class.

r 00
(ii) If <An> is decreasing sequence, then 1m An=" A,.EA,
n-,; 00 1=1

by the property of a-field,

Hence, in either case, a-field is a monotone class.
Second part. Conversely let M be a monotone field. C-onsi-
der a sequence < Bn > in M.
Let &.= U BIr. if < Bn > is non-decreasing sequence.
k<.n
and F,,= n BIr. if < BII > is non-increasing sequence
k<.n
lim
lim En
Fn E M} because M IS
E M . a monotone cIass.
 MEASURE AND OUTER MEASURE 65

lim Bn t= u
11;;;::1
Bn=U [
.. U
k~n
8/:]= U En= lim E..
n

lim Btl J. = n
11~1
B,.= n
..
l n Bk]= n F,.=lim F..
k~n "
Now we can say that
00 'Y;
lim Bn t , lim B,. t E M ~ U BII tE M, n Bn ~ E M

=> M is a a-field.
Problem 10. Show that a ring is closed under formation of
symmetric difference and illlersections. [Gujarat 1970]
Solution. Recall that a non-empty family A of subsets of X
is called a ring of sets if
A, BE A ~ A--B E A, A U BE A
We want to show that
(i) A, B E A => A n B E A
Oi) A, B E A => A L':::. B E A
where 6. denotes the symmetric difference of two sets.
(i) A, BE A :::> A E A, A -B E A
=> A - (A -- B) E A
=> A n BE A For A---(A-B)=A n B.
(ii) A, BE A :::> A-B, B-A e A
=> (A-B) U (B--A) E A
=> A 6 B=(.4-B) U (B-A) E A
=> A L':::. BE A.
Theorem 12. If mo is an outer measure, then the class A of !no
measurable sets is a-algebra (or a-jield). [Kanpur Statistics 1977]
Proof. Let A be the class mo - measurable sets.
Let A, B E A be arbitrary, then A, B are outer mea<1urable
sets.
To prove that A is a a - algebra of sets, we have to show
that
(i) A' E A
(Ii) A U BE A
(iii) jf < A, > js a sequence of sets in A, then
 66 MEASURE AND OUTER MEASURE

(1) We have supposed that A is an outer measurable set so

tha' A' is outer measurable.
(2) We have seen that if A and B are outer measurable sets,
then A U B is an outer measurable set.
(3) Let < Ai > be a sequence of outer measurable sets.
00 n
Also let A= U Ai, B,.= U Ai.
;=1 i-I

To prove that A is outer measurable.

By (2), A1> A2 are outer measurable sets
=> Al U A 2) is outer measurable
=> (AI U A2 ) U .42 is outer measurable, by the
same reasoning.
Generalising this, we can say that any finite union of outer
measurable sets is outer measurable [Link] so Bft is outer measurable.
Consequently for any T,
rno (T)=rno (T n Bn)+rn o (T n B,.') ... (1)
00 ..
A= U Ai:J U A.=Bn
i-I i=1

This => A :J Bn => 8.' :J A' => A' C B,.'

=> Tn A' C Tn Bn' => rno (T n A') ~ rno (T n Bn')
Now (I) becomes
rno (T)=me (TnB,.)+rn o (TnB,.') ~ rno (TnB,.)+rn o (TnA')
or rno (T) ~ rno (T n B .. )+rno (T n A') ... (2)
•
Bn= U A. :J An. This => B.. :;) A. n 8n=A...
i=1

Thus An n B.. =A.. Similarly Bn n A.'=Bn_ 1 •

Since An is outer measurable and hence
rno (T)=rno (T n A.)+rno (T nAn').
Taking T n B. for T, this becomes
rno (T n B,,)=rno (T n B" n A,.) +rno (T n Bn n A.;)
=rno (T n An)+rn o (T n B .. _1 )
or rno (T n B,,)=rno (T n A,,)+rno (T n B n_ 1 ) ... (3)
Putting n=2, 3, 4, ...... , n -1, n in the equation (3),
rno (T n B 2 )=rnO (T n A 2 )+ rno (T n B l )
rno (T n Bs)=rno (1' n Aa)+rno (T n B 2 )
rno (T n B.)=rno (T n A4 )+rnO (T n Ba)
 MEASURE AND OUTER MEASURE 67

mo (T n B.'-!)=m o (T n An-l)+mO (T n B .. - 2)
mo (T n B,,)=mo (T n AnH-mo (T n Bn- I )
Upon addition, we obtain
n " ,,~
E mo (T
,-2
n B,.)= L
r~2
mo (T n A,)+ E mo (T
r-l
n Dr)

to
This ~ fIlo (T n D,.)= E Ino (T () A,)+nl o (T
r~-2
n Bt )

fa

E rno (T n A,).
= ,=1 For Dl=A).

.
:. nlo (T n D,,)= E rno (T
,=1
n A,). . .. (*)

Using this in equat!on (2),

n
rno (1') ~ r: mo (T n A,)+mo (T n A') ... (4)
.=1

A= U Ai => T n A= Tn ('U<=1 Ai
j~t
)= U .=1
(1' n A,)

z> rno (T n A)=rno (~1 (T n A,») ~! rno (T n A.)

00
==> }; nlo (T
,=1
n Ar) ~ rno (T n A).

Using this in (4)

rno (T) ;;?! rno (T n A)+rno (T n A'). . .. (5)
But T=T n A+T n A'
mo (T)=mo (T n A+T n A')
~ mo (T n A)+rno (T n A')
or mo (T) ~ rno (T n A)+mo (T il A'). ... (6)
Combining (5) and (6),
rno (T)=fn o ~ (T n A)+mo (T () A')
00
This ==> A is outer measurable ==> A= U A, E A.
1=1

Theorem 13. Let (X, A, m) be measurable space. Then

(i) A, D E A s.t. A C B => m (A) ~ m (D)
(ij) If < An > is a sequence of sets in A, then
 68 MEASURE AND OUTER MEASURE

m( .91 A. ) ~.~ 2' (A.). [Indore 1978J

Proof. Let (X, A, m) be a measurable space.

(i) Let A, B E A s.t. A C B.
To prove m (A) .s;;; m (B).
Notice that B=(B-A) U A, (B-A) n A=rp
This ~ m (B)=m (B-A)+m (A) ~ m (A)
Also m (B- A) ~ O.
~ m (B) ;;0. m (A) ==- m (A) ~ m (B).
(ii) Let < An > be a sequence of sets in A.
"-I
Write Bn=AII - - U A.
'=1

Then Bl=Alo B 2 =A 2 -A1 , Bs=Aa-(AI U A2),······

This shows that (1) B. n BI=t/I for i::j:j
00 00

(2) U A,= U B~
1=1 ;=1

(3) Bn C A" ¥ n.

(1) ==- m (~1 B, )= ,~ m (BI)

=> m [~1 Ai ]=;~ m (B , ), by (2).

00
~ L m(A.), by (3).
'=1

=> m [U
1=1
AI] ~ ;
1=1
m (A,),

Problem 11. Let A be an algebra of subsets of X and let

< An > be a [Link] of sets in A. Theil prol'e that there is a
sequence < Bn > of sets in A slIch that Bn n Bm=!,~for nz-:j:.11 alld
00 00
U B , = U A,.
1=1 ;=1

[Kanper Statistics 1976; Gujrat 1970J

Solution. Let A be an algebra of subsets of X so that
(i) A, B ~ A => A U B E A
(ii) A' E A.
 MEASURE AND OUTER MEASURE 69

(iii)To prove A-B E A

A, B E A => A, B' E A => A n B' E A
=> A () B'=A-B E A.
Let < A" > be a sequence of sets in A.
To prove that 3 a sequence < Bn > of sets in A.
00 00
s.t. B,. n Bm=,p for m:;t;n and U B.= U Ai.
>=1 i=1

By assumption, Ai E A for i= 1, 2, 3, 4, ......

11-1
Write Bn=A,. - U Ai'
1=1

Then
. 0 0 0 0
This shows that (I) U Bi= U Ai
i=1 i=l

(2) B .. n Bm=,p for m=l=n.

(3) Bn C All "I n.
(i) => Al U A2 E A ~ Al U A2 U Aa E A.
"-1
Generalising this, we see that U Ai E A.
i=1

n-l "-1
U Ai E A, An E A ~ An -- U Ai E A, by (iii)
<=1 ;=1

=> Bn E A
=> < B" > is a sequence of sets in A.
Also we have seen that < B" > has the properties (I) and
(2). Hence Proved.
Problem 12. If I' is completely a-additive set function on a
class E and if < En t > is any sequence of sets from the class E,
thell lim (I' £11)=1' (lim E,,). [Kanpur Statistics 1975. 76J
Solution. < Eit > is non-decreasing sequence and its limit
00
will be U Eft. This increasing sequence is converted into sequ-

ence of disjoint sets as :

Write Eo=rb, Bn =EII -En _ 1 for n=l, 2, 3, ...
00 00
Then we have U B.. = U E,,=lim En
"=1 II-I
70 MEASURE AND OUTER MEASURE

... lim &=

go
u
n=1
00
En= U (E.. -Bn-l)
10=1

...

Problem 13. Let", be a measure on an algebra A.

(i) If An C A"+I' An e A. 1 EO; n < <X1

00
U An E A, thell p.
and ~ ( ()()
U A" )
= lim p. (A,,).
~ n_~

ai) If A" :::> A"+l' A" E A, 1 ~ n < 00,

00
'" (A 1) < 00 and
n=1
n A" E A, then

'" ( ~1 An )= n~":'" p. (An).

Solution. Refer theorem 1.

Problem 14. If <Bt.> be a sequence of disjoint measurable
sets. then for any set A,

mo [ A n (U )]=.f
'=1
Ei
.=1
mo -(A U Ei ) [Meerut 1987J

Solution. Prove equation (*) as in Theorem 12 that

n
mo (r () Bn)= E !no (T () A,)
r=l

It
or mo (T n Bn)= E !no (T n E,) ... (4)
r=l

00 It
where A= U Eh Bn = U E,
0=1 1=1

(I.e. replace A, by E, in theorem 12)

 MEASURE AND OciTER MEASURE 71

Replacing T by A in (4), we get

mo [An C9tEi ) J=I~ mo(A nE.)

Exercises
1. Show that the additive set function defined over a ring
can not assume both the values +00 and -00. [Poona 1970]
2. What is a measure space? [Kanpur 1969]
3. State Caratheodory's criterian of measurability of a set.
[Banaras 1966]
4. State Caratheodory's criterian of measurability and
prove the union of two measurable sets is measurable.
LBanaras 1965]
5. (a) Define a a-filed and a monotone class of sets, show
t4at a monontone filed is a a-filed.
(b) Define continuty of a set function, and show that a
a-additive set function is continuous.
(c) If fL is a measure function, defined on a class E of sets.
If E is any sequence of sets from E for which p (E..) <00, then
p. (lim ~ub 1:...) :> lim sup pE,..
[Kanpur Statistics 1977]
6. <a) Let p.* be an outer measure defined on all subsets of
Q. Show that a necessary and sufficient condition for a set to
be p.*-measurable is that ¥ ACQ for which
/'*(A) < OC', p.*(A) ;;;:, p*(A n E) +p.*(A-E).
(b) Define outer and inner measures. If p.* is an outer
measure on Q and M is the class of all p.*-measurable sets then
show that M is a a-filed and p.* is a measure on M.
[Kanpur Statistics 1977]
7. Show that an open set in a metric space is measurable
w.r t. any metric outer measure. [Banaras 65]
g. A set E of real numbers is of measure zero if there exists
a sequence of intervals <.11/> such that Ee Lim sup I.. , and
E in < co, where i" is the length of lit. [Banaras 65]
9. Let po be a countably additive non-negative measure
defined on an algebra of subsets of a set X such that p.(X) < 00.
Show that if {A" : E N} is a contracting sequence of p.-measur-

able subsets of X, then po ( n~NA" ]=;~mN p (A,,).

f2 MEASURE AND OUTER MEASURE

Also give a counter example to show that the result is not

valid if f4 (X)=OC!>. [Banaras 64]
Hint. Refer problem 13.
10. Write a short note on the problem of measure in Eucli-
dean space. [Banaras 64]
11. <a) Define (i) ring, (ii) a-ring and (iii) monotone class.
Show that a ring is closed under the formation of symmetric
differences and intersections.
(b) Let A be an algebra of subsets of)( and let < A,. ....... be
a sequence of sets ill A. Then prove that there is a sequence
< B,. > of sets in A such that
B" n B... =rp for m#:n
00 co
and U Bi= U Ai
i=l 1=1 [Gujrat 70]
12. Intersecti:m tw~ B-rings (Boolean rings) is a B-ring.
of
[Gujrat 70]
13. The union of two rno measurable sets is mo measurable.
[Meerut 72]
14. (a) EXplain Caratheodory's concept of an outer measure
po* on a set X. Define p.*-measurubility of a subset A of X. If
p.*(A}=O. show that A is measurable.
(b) If <A.. > is a contracting sequence of po-measurable sets
and po(A I ) < 00, prove that
f' (lim A.. )=lim f4 (Aft)
[Banara~ 71]
(Give an example to show that the finiteness condition is
essential). [Banaras 71J
 7
Lebesgue Measure
of a Set

Introduction. The concept of measure is an extension of

concept of length. Measure of an interval in R is usually defined
as its length, the measure of a polygon in R2 is usually its area
and so on.
7·0. Measure of an open set and a closed set.
An open set is an enumerable union of non-overlapping
open intervals. The measure of an open set is defined as the
limiting sum of the lengths of its intervals. If the set G is contai-
ned in an interval (a, b), then the measure of G, denoted by
m (G), has the property that
m (G) ~ b-a.
We know that the complement of a closed set w.r.t. an open
set is open. Let a closed set F be contained in an open interval
(a, b). We define the measure of F as follows:
m(F)=b--a --m(F'), where F'=(a, b)-F.
7·1. Measure of an open interval.
The length of any open interval G is defined as its length and
is denoted by the symbol meG).
Examples: (i) m [(2, 5)]=5-2=3
(ii) m[(a, b)]=b-a.
Evidently meG) > o.
7·2. Measure of a closed interval.
Let La, bJ be the smallest closed interval containing a closed
set F. Then we define m(F)=b-a-m(F').
F' being the complement of F w.r.t. the interval.
To prove that the measure of a closed interval is its length.
Let A=[a, b]. Then A is a closed interval and hence a closed
set. [0, b] is the st:Jlallest closl1d interval containing A.
 74 LEBESGUE MEASURE OF A SEr

Complement of A relative to [a, b]=A'=cp. For A=[a. b].

By definition, m(A)=b-a-m(A')=b-a-m(q,)=b-a-O
=b-a
i.e. m([a. b])=h-a.
7'3. Measure of rectangle.
The area of an open rectangle R (a < x < b, c < y -< d) i.e.
(b-a) (d-c) is defined as the measure of R.
Thus m(R)=(b-a) (d-c).
The area of a closed rectangle R (a ~ x ~ h, c ~ y ~ d), i.e.
(b -a) (d -c) is defined as the measure of R.
Thus m(R)=(b-a) (d-c).
7'4. Measure of a parallelopiped.
The volume of an open parallelopied
V (a < x < b, c < y < d, 1< z < m)
is defined as the measure of V.
Thus m(V)=(b-a) (d-c) (m -I).
Similarly the measure of a closed parallelopiped
V' (a ~ x ~ b, c ~ y ~ d, I ~ z ~ m)
is defined as its volume.
Thus m{V')=(b--a) Cd-c) (m-/).
7'5. Exterior and interior measure. (Lebesgue measurable set)
[Kan~ur 75. 73. 68]
The Lebesgue exterior measure (or simply exterior measure)
of any set A, denoted by m,(A), is denned as follows:
m,(A)=inf. {meG) : G :J A, G is open} ..• (1)
From this it follows that meG) > me(A). Hence
given. > 0, 3 an open set G :J A s.t.
m(G) < me{A)+,. . .. (2)
If the set A is contained in an interval (a. b), then
o ~ m,,(A) ~ b-a.
Also Al C A2 =:> m.(A 1 ) <;; m.(A 2)
and m.(,) =0.
The Lebesgue Interior measure (or simply interior measure of
a set A contained in an interval (a, b), denoted by m.(A), defined
as follows:
m,(A)=sup {m(F) : F is closed, F C A}.
It can also be expressed as
[The equation (2) is of vital importance for further study].
LEBESGUE MEASURE OF A SET 75

m. (A)=b-a-m, (A'). . .. (2)

From the definition it is clear that
mi (A) ~ 0 and mi (</»=0.
A set A is said to be measurable (or Lebesgue measurable)i f
m. (A)=m_ (A). The common value of m. (A) and m. (A) is called
its measure and is denoted by m (A).
Thus mfJ (A) =mi (A)=m (A), if A is measurable.
Remark. g.l.b.=inf and l.u.b.=sup.
Theorem 1. Show that m. (A) ~ mi (A) for any set A.
[Agra 73]
Proof. Let any set A be contained in an interval (a, b). Then
by definition of exterior measure 3 open sets 01' G2 containing A
and A' respectively s.t.
m (Gl ) < m. (A)+. l
m (G 2 ) < m tl (A')+.r ... (1)
This proves that every point of the interval (a+" b-e) is an
interior point of an interval contained in G1 or O2, By Heine
Borel theorem, we can determine a finite set 0 from these intervals
s.t. 0 :::> (a +', b -() " .(2)
and m (Ol)+m (0 2) ~ m (0) ... (3)
(2) => m (0) ~ m [(a+., b-e)]=b-e-(a+e)
=> m (G) ~ b-a-2E. . .. (4)
By (1), m (Ol)+m (0 2) < m. (A)+111. (A')+2e.
Using (3),
m (G) ~ m (Gl)+m (G 2 ) < m. (A)+m, (A')+2e
or m (0) < m. (A)+m_ (A')+2 •.
Using (4), we get
or b-a-2f. < m, (A)+m, (A')+26
or b-a m, (A') < m. (,.,)+4&
or m, (A)=b- a-m. (A{) < m. (A)+4.
or m. (A) < m, (A>+4e.
Since e is arbitrary and hence making E~O, we get the requi-
red result m. (A) ~ m. (A).
Theorem 2. A sei .A is measurable (i.e. Lebesgue measurable)
iff its complement A' is measurable.
Proof. Let a set A be contained in an interval (a, b). Then
we know that
m. (A)=b-a-m. (A'). ... (1)
76 LEBESGUE MEASURE OF ~ SET

(i) Let A be measurable so that

mi (A)=m, (A)=m (A). . .. (2)
To prove that A' is measurable, we have to show that
m, (A') =mi (A'). . .. (3)
Writing (1) with the help of (2),
m (A)=b-a-m, (A')
or m. (A')=b-a-m (A). ...(4)
(1) is true ¥ A. Replacing A by A' in (1), we obtain
mi (A')=b-a-m, (A). For (A,),=A.
Using (2), we get
mt (A')=b-a-m (A). ...(5)
Since, R.H.S. of (4)=R.H.S. of (5)
Hence, L.H.S. of (4)=L.H.S. of (5)
i.e. m, (A')=tni (A').
This => A' is measurable.
(ii) Let A' be measurable so that
me (A')=mi (A/)=m (A'). ...(6)
To prove that A is measurable
By virtue of (I), we have
tnt (A)=b-a-m, (A')
and mi (A')=b-a-m, (A).
In view of (6), these equations become
mi (A)=b-a-m (A') ... (7)
m (A')=b-a-m. (A). ...(8)
Equating the two values of m (A')
b-a-m. (A)=b--a-m, (A).
or mi (A)=m. (A).
This => A is measurable. Proved.
Note. Let a measurable set A be contained in [a, b).
Then In; (A)=b-a-m. (A')
or m (A)=b-a-m (A')
or m (A)+m (A')=b-a.
For A is measurable => A' is measurable.
Theorem 2. (a) Evay bounded open set and bounded closed set
are measurable. lMeerut 1987, 86]
Proof. Let G be a bounded open set. Form G by means of
a finite number of disjoint closed intervals. Then the sum of the
lengths of these intervals is equal to the length (measure) of G.
Hence m, (G)=m (G) ... (1)
 LEBESGUE MEASURE OF A SET 77
Thus G can be obtained in an open set consisting of G itself,
and contains a closed set s.t. difference of measures of closed set
and open set is arbitrarily small. Consequently
m. (G)=m (G)=m. (G)
This ~ G is measurable.
Since complement of a measurable set is measurable. Also
complement of open set is closed set. This ~ Gf is measurable
and G is closed.
f

Theorem 3. . Let A and B be any two sets such that A e B.

Then (i) mi (A) ~ m. (B)
(U) m, (A) ~ m. (B).
Proof. Let A e B.
(i) To prove that mi (A) ~ m; (B).
By definition of interior measure,
Ini (A)=sup {m (F) : F is closed, Fe A}
lni (B)=sup {m (F) : F is closed, FeB}.
Let S be the set of numbers consistjng of measures of all
closed subsets of A. Similarly we define the set T for B. Then
m; (A)=sup S, mi (B)=sup T. . .. (1)
F is any closed set s.t. F e A => F CAe B. For A C B.
~ Fe B
:;. F is closed subset of B.
This shows that
any m (F) E S ~ m (F) E T.
Hence SeT. Consequently sup S ~ sup T.
For least upper bound, i.e. supremum of a subset of any set
can not exceed the least upper bound of the set itself.
sup S ~ sup T =? mi (A) <; nil. (B), by virtue of (I).
(i) To prove that m, (A) ~ m, (B).
By definition of exterior measure,
m, (A)=inf {m (G) : G is open, G ::J A}
m. (B)=inf {m (G) : G is open, G ::J BJ.
Let S be the set of numbers consisting of measures of all
open supersets of A. Similarly \\'e define the set T for B. Then
m. (A)=inf S, me (B)=inf T. . .. ()
G is any open set S.t. G ::J B =? G ::J B :> A. For B :> A.
=> G :J A.
~ G is open super set of A.
This prove, the fact that
and III (G) E T =-> 1J1 (G) E S.
78 LEBESGUE MEA~UPE OF A SET

Hence T C S. Consequently inf S <;; inf T.

For greatest lower bound of a set can not exceed greatest
lower bound of a subset of that set
inf S <;; inf 1 => m. (A) ~ m~ (B), according to (2).
Theorem 4. ,r
El and E% are measurable (in the sense of
Lebesgue), then El U E2 is measurable.
[Meerut 1988, 86; Kanpur M. Sc. P. 79]
Proof. LetE1: be a me:1surable set for k= 1,2 and
E=£1 U E2 •
To prove that E is measurable.
Consider an arbitrary num bel' f > O.
EI: is a meastJrable <;et => 3 a closed set F1: and an open set
G" S.t. F; C cI: C G"o

m (Gd-m (Fd < T"

Write F=F1 U F2 , G='G 1 U G2 •
Then F is a closed set and G is an open set so that F and G
are measurable sets.
2 2 2
F1, C Ek C GI • :::> U Fl. CUE/;. C U G,:.
/:~'l 1:=1 le·,1

=>FCEcG
=> FeE, E c G
=> mi (F) <;; 11'1, (E), mil (E) ~ 11'1 (G)
~ 11'1 (F) ~ 11'1 .. (E) ~ me (E) ~ 11'1 (G)
[For 11'1, (F)=m (F), m. (G)=m (G)
and ml (A) < mil (A) v A]
=> m. (E)-mt (E) ~ m (G)-- m(F). ...(1)
G -F=G n F'=A finite intersection of open sets
=An open set=A measurable set
m (G-F)=m (G)-·m (F). ...(2)
Similarly m (G/<-F,,)=m (G,,)-m (F,,).
2
Notice that G-F C 1: (G1.- -F1)
"-1
I
This => m (G - F) ~ 1: m (G,.-- F;-)
1<=1

2 •
.. m(G)-m (F) ~ E [m (G/.)-m (Fl')] < -2 .2=.
k-l
 LEBESGUE MEASURE OF A SEl

~ m (G) -In (F) < Ii

=> m. (E)-mi (E) ~ m (G)-m (F) <: 6, by (1)
::> m. (E)-mi (E) < e.
Making E-+O
me (£) -mi (E) ~ O. For E is arbitrary.
This => me (E) ~ mi (E).
But m. (E) ~ m, (E) is true ¥ E.
Hence m. (E)=m, (E).
This ~ E is measurable.
Problem t. If X and Yare measurable sets, then X n Y i~
measurable. What can you say about X U Y. [Kanpur 1975, 791
Solution. For the sake of convenience we write E1 =X, £2= Y
(i) To prove X U Y=E1 U E2 is measurable.
(prove this as in theorem 4)
(ii) X, Yare measurable
=> X' Y' are measurable, by theorem 2.
=> X' U y' is measurable by (i)
=> (X n Y)'=X' u Y'is measurable.
=> X n Y is measurable.
[For A is measurable ¢> A' is [Link].
Theorem 5. If G1 and G2 are any two open sets in[ a, b], then,
m (G1 )+111 (G 2 )=m (G 1 U G2 )+m (G 1 n G2 )
i.e., m (G 1 )+m (Gz)=m (G 1 +G2 )+m (G 1G 2 ) [Kanpur 1973]
Proof. Here m(Gl) = length of interval in G1
The length of the interval which is common in G1 n G2 is
equal 10 the length of the interval in G1 n G2 =G1G2• On the
other hand the lengths of common intervals in m (G1 ) +m G2
occurs twice. Consequently
m(G1 )+m(G2 )=m(G I +G 2 ) +m(G1G2).
Theorem 6. If El and E2 are subsets o/[a, bJ, then
m.(El)+m.(Ez) ~ m.(E1 U E2 )+I1I.(E1 n £2)
alld mi(£1)+mi(E2 } ~ mi(EI U E2)-miCE, n £2)
Or To prove that
(i) m*(E1 )+m*(£2) ~ m*(E1 U t~)+m*(I:;l n E2 )
(ii) m*(£!)+m*(£2) ~ m*(£1 U £2)+m*(EI n t z)
[Banaras 1968J
Proof. Let El and £~ be subsets vi closed interval [a, bJ. Fix
E > O. Choose open sets G1 and G2 s.t. El C G1> Ez C G2 ,
 ~o LEBESGUE MEASURE OE A SET

m(Gt ) < m.(E1 )+E/2 and m(G 2 ) < m'(Es)+(E/2).

Then m(G1)+m(G2 ) < m.,(E1 )+m..(E2 )+f
But m(G1)+mtG2)=m(Gl U G 2 )+m(Gl n G2 )
Therefore meG} u G2 )+m(Gl n G 2 ) < m.(E1)+m,,(E2 )+,
But Gl U G 2 and Gl n G2 are sets containing El U £2
and £1 n E2 respectively. Then the last gives
m,,(EI U E2)+m. (E) n E2 ) < m. (E1)+m.(E2)+"
Since" is arbitrary and hence making E-+O.
We obtain the first required result namely
miEI U E2)+m.(E I n E 2 ) ~ m.\E1)+m.(E2) ... (1)
This is true I.f E 1 , E2 C [a, b].
Replacing Ez, E2 by E 1 ', E2 ' respectively in (1), we obtain
m/E/ U E2')"" m"(£l' n £2') ~ m.(£t')+m,(E2 ') ••• (2)

where E1'=[a, b]-El' £2'=[a, b]-E2'

Applying De-Morgan's law in (2).
m~[(EI n E 2 )']+me [(El U E 2 )'] ~ m.(E/)+me(E/).
By definition of interior measure, this gives
[b-a-mi (E1 n E2)J+[b-a-mi (El U E2 )]
E;;; [b-a-mi(E1)]+[b-a- m;(E2 )]
Simplifying this we get
mi(El n E2 )+mi(El U E2) ~ mi(El)+m.(Es).
This proves the second required result.
For mi(A)=b-a-nle(A') ¥ A c [a, b]
and x < y => -x > - y. Proved.
Theorem 7. If El and £2 are measurable subsets of [a, b] then
prove that
m (E1 )+ m (E2)=m (El U Ez)+m (El n E2)'
[Meerut 19S8, 87; Kanpur 79, 88; Kolhapur 70; Banaras 69]
Solution. Let El and E2 be measurable subsets of[a, b) so
that
m.(E1)=m,(E1)=mr E1 ) }
m.(E2 )=mi(Ez)=m(E2 ) ... (*)

Firstly we shall prove that

(i) m"(£1)+m,,(E2)~m.(EI U E 2 )+m.(E'1 n £2)
(ii) mi(E1) .. mi(B2)e.l11i(F."t U E2)+mi(E1 n E2)·
Prove this as in Theorem 6.
Wr:ting (i) and (ii) with the help of (*)
(i)' II1(E 1 )+m(E2 ) ~ m(£1 lJ E2 )+m (E 1 n £2)
(ii)' IJl (E1 )+I11(E2 ) ~ m(EI U EzH-m(E l n E2 ).
 LEBESGUE MEASURE OF A SET 81

For finite union and intersection of measurable sets are

meac;urable so that El U E2 and El n E2 both as measurable and so
m. (El U Ea)=m, (El U Ea)=m (El U E2 )
me (El n E 2)=m, (El n E2)=m (E1 nEs).
Combining (i)' and (ii)' we get the required result.
Problem. If El and E9 are measurable subsets 0/[0, b], prove
that El U E 2, El n E2 and E1 -& are measurable and show also
that
(i) m(h.;)+m(E2)=m(El n E2 )+m (El U E2 )
(ii) m(E1 -E2 )=m (E1 )-m(E2) if E z C El
[Kanpur [Link]. 1979]
Solution. (i) Prove as in theorem 7.
(ii) Since difference of measurable sets is measurable
:. E1 -E2 measurable set
(E1 -E2 ) n E2 =rp
:. m [(E1 - E2 ) U E2J= m (E1 -E2 )+m(E2)
or m(E1 )=m (E1 - E2)+m(E2)
or m(El) - m(E2 )=m (E1 -E2)
Problem 2. If A, B, 'C, are p.-measurable sets, show that
p.(A)+p.(B)+ p.(C)=p. (A U B U C)+ p. (B n C)
+p. (C n A)+p. (A n B)-p. (A n B n C).
[Banarsas III, 1970]
Solation. Here we shall use the following lemma:
Lemma. If E1 , E2 are measurable subsets of [a, b], then
m(E1 )+m(E2 )=m (El U E2 )+m (E1 n E2 )·
Now we come to the proof of the problem.
Let A, B, C be measurable subsets of [a, b].
Since finite union and intersection of measurable sets are
measurable and hence A n B n C, A U B U C, A U B, A n C
etc. are all measurable.
Taking E]=A U 8, E2 =C in the lemma, we get
m (A U B)+m (C)=m (A U B U C)+m[(A U B) n C]
... (1)
Again, by the lemma.
m (A)+ m(B)=m (A U B)+m (A n B)
>r m (A)+m (B)-m(A n B)=m (A U B). ... (2)
Hence m (A U B) n C]
=m[(A n C) u (B n C)]
82 LEBESGUE MEASURE OF A SET

==m (A n C)+m (B n C)-m (A n C n B n C).

[This follows from (2)]
=m (A n C)+m (B n C)-m (A n B n C)
or m[(A U B) n C]=m (A n C)+m (B nC)-m (A n B n C) .
... (3)
Writing (1) with the help of (2) and (3)
m (A)+m (B)-m (A n B)+m (C)
=m (A U B U C)+m (A n C)+m (8 n C)-m(AnBn<:)
or m (A)+m (B)+m (C)=m (A U B U C)+m (B n C)
+m (C n A)+m (A n B)-m (A n B nC).
If A, B, C are ,,-measurable sets, then the las~ gives the
required result.
" (A)+ " (B)+" (C)
=="(.4 U B U C)+" (B n C)+I' (C n .4)+" (A n B)
- " (A n B n C).
Problem 3. Let E 1 , E2 , ... be sequence of sets of refl/ numbers,
prove that
m* (E1 +E2+"~) ~ m* (E1)+m* (E2 )+ ...
where m*(E) is the [Link] outer measure.
If in addition the sets E, are pairwise disjoint, i.e., no two of
them have common poinl8 then show that
m* (E1 +E2 + ... ) ~ m* (E1 )+m.(E2 )+ ...
where m* (E) denotes the inner measure of E. [Kanpur 1972]
SolutioD. We write m. (E) in place of m* (E) and m, (E) in
place of m* (E).
To prove that

(i) m. ( E Er
r-l
)

if E1 , E 2 , ... are pairwise disjoint.

To prove (i).
Given. > 0, 3 an open set Gn S.t.
E
E. C G,., m (G n ) < m. (E;.) + Tn
LEBESGUE MEASURE OF A SET 83

An arbitrary union of open sets is open '1nd therefore

-"
]; G, is an open set s. t.
'~1

... (1)

This follows from the fact that En C G", ¥ n.

(1) => m.( UE, )-E;;; m, (U G, .)=m(u G

r~1 r=1 '=1
r )'"
00
1: m(G,)
r_l

~ m. (,! E. ) ~ !,m (G,) <.!. [ m. (E.l+; ]

~ m(! ( r~l Fr ) < ,~ mo (E,)+I! ..

Since c is arbitrary and hence making 1-0, we get

m,( gl E, ) <; tl m~ (E,)

This is also expressed as

m, ( !1 Er ) <; r~lm. (E,l ... (*)

To prove (ii).
Suppose E1 , E2 , ... ••• are pairwise disjoint so that
en n En=t/> for n#=m.
Consequently
m. (Em n En)=O=m. (E". n En) if m;#:n. . .. (2)
We know that
m. (E1)+m. (~) E:; m, (El U EA)+m, (El n £2)'
In accordance with (2), this becomes
m. (E1)+m, (£2) ~ mj (£1 U E2 )
84 LEBESGUIi MEASURE OF A sn
By induction, it follows that

m, (Ql Er);. !1 mi (Er) ... (3}

n being a positive integer.

00 ,.
E/idently U Er :::> U E.,
,=-1 r=-l

and so m, (u
,_1
Er);;. m, (u
,-1
Er ') ~i'1'=1
m, (E,.), by (3)

or m, (gl Er ) ~ !lm, (E,).

Making n__ , we get m, (Ql Er ') ~ £1 m, (Er)·

This is also expressed as m, (X E,.) ~ ~ m,
'1'-1 '1'=1
(E,).

This completes the proof.

Problem. Let < E, > be a sequence of measurable sets. Then
prove that

m(~l' E, ) <; ,! m (Ei ). [Meerut 1999. 87]

Solution. E. is measurable set

=> m (E,)=m. (E,) ... (2)
00
and U is measurable and so
'1'-1

... (3)

Now prove equation (*) of the above problem (3),

I.e., m. (u '1'-1
Er) <; E
'1'-1
m. (E,). . .. (*)

Putting (2) and (3) in (*), we get the required result.

LBBIISGUB MEASURE OF A SET 8S

Problem 4. If M is a measurable set, then prove that for any

set E.
m, (E)=m. (EM)+m. (E-EM). [Kanpur 1972]
Sollltion. We know that
(l~ -me (E 1 +£2) ~ m. (E1)+m. (E2)
(2) m. (E1 )+m. (E2) ~ m. (E1 +E2)+m. (E1E2)
By virtue of (I), m. (E) ~ m. (E-EM)+m. (EM). . .. (3)
By (2), me (E)+m. (M) ~ m. (EM)+m. (E+M)
;;;;. m. (EM)+m. (E-EM+M)
[as E+M ::> E-EM+M]
~ m. (EM)+m. (E-EM)+m, (M)
[as m. (E1 +E2) ~ m. (E1)+m, (EJ if El n E 2=c;6].
But m. (M)=m. (M)=m (M) as M is measurable
m. (E)+m (M) ;;;;. m. (E4l)+m. (E-EM)+m (M)
or m. (E),;;;;' m. (EM)+m. (E-EM). ...(4)
Combining (3) and (4),
m. (E)=m. (£ -EM)+1Y.. (EM). Proved.
Theorem 8. Prove that a set E of real numbers is measurable
iffgiven e > 0, there exists ~, el' e2' S.t. E=t+e1-e2, and
m. (el ) < e, m. (e 2) < Ii and E is the sum of a finite number of
intervals. [Kanpur 1973]
Proof I. Let E be a measurable set and E C R.
To prove t'ha' E=~+el-e2 where m. (el) < t, m. (e 2) <-
e
and is the sum of a finite numb!r of intervals.
By assumption 3 open'set G s.t. G ::> E and
m (G) < m (E)+ •. .. _(1)
Write ...(2)
and ~ is a union of enough intervals of G. Also take
el=G-~ . .. (3)
and m. (el ) < t. ... (4)
m. (e 2 )=m (G --E)=m (G)-m (E) < c, by (1)
:. m. (e2) < e.
- [Since G --E is measurable]
(2)-(3) => e2-el=G-E--(G-~)=~--E
=> E=~+el-e2' ... (6)
From (4), (5) and (6), the required result follows.
86 LEBE~GUE MEASURE OF A SET

II. Let E=1;+e 1 -e2 with m, (e 1 ) < e, m. (e 2) < e, E C R.

To prove that E is a .measurable set. Let E C [a, bJ.
Since E=(!;+e1 )-eZ'
Hence m. (E)=m, (f+e 1 )-m. (e a)
~ m. (1;+e l ) ~ m. (f)+m, (e l ) < m (~)+4!
or m. (E) < m (~)+E. ; .. (7)
E=~+11-12 ~ [a, bJ---E=[a, bJ-f+/2 - /1
~ E =~'+1.,-11::;' m. (E')=m. (~'+/2)-m. (11)
o m. (E') < m, (~'+/2)-e < m. (1;'+/2) ~ m.(~')+m (/2)
=> m. (E') < m (f)+E as E is measurable and so is 1;'.
o [a, bJ-m. (E') > [a, b]--m (~')-4!
~ mi (E) > m (~) -41 as m, (E)=[a, b]-me (E')
and-_ - m. (E)-4 < m m < m, (E)+e, by (7).
=> m, (E)-fi < mi (E)+fi
~ m. (E) ~ m, (E) on making E~O.
But m. (F) ~ m, (E). Hence m. (E)=m, (E).
This::> E is a measurable set.
Problem 5. Is the Lebesgue measure of the union of two mea-
surable sets equal to the sum of their measures? [Kanpur 1975]
Solution. Since m (A lJ B)=m (A)+m (B) if A n B=r/>.
Ans. Yes, if their intersection is null set.
Problem 6. Construct a non-dense perfect set in the interval
[0, IJ whose measure is i. '[Kanpur 1968]
Solution. Consider the closed interval [0, IJ. Let ,\ E [O~ 1].
Delete a portion of length I" from the interval [0, I] s.t.
11 +2/2 +22 / 2 + ... +2,,-1 1,,=>.-- n~-2
Stage Number of intervals dropped
1st 1
2nd 2
3rd 22

nth 2"-1 2,,-1 I"

The sum of the lengths of the intervals dropped upto nth
stage =/1 +2/2 +22 / 3 +", +2"-1 I"
=.\- n~2' by assumption.
The limiting sum of the dropped intervals
LEBESGUE MEASURE OF A SE r 87

= nl:oo l A- n~21=~-0=A'
Let A denote the set of end points of the interval {III} with
their limiting points, then A is a non-dense perfect set. Also
m(A)=m [0, I]-length of dropped intervals=l-A.
If we take A=i, then m (A) = l-i=i.
The set A formed as above is a non-dense perfect set with
measure 1.
Theorem 9. 'Any set A is measurable iff an open set G con-
taining A and a closed set H contained in A can be so determined
that I G I - I H I < 6, where I G I stands for length of open inter-
vals in G. '
Proof. Definition of exterior measure => given. > 0, we
can find an open set G :J A s.t. I G I < m, (A)+e/2. ...(1)
Similarly we can find a closed set H S.t.
m. (A) < I HI +e/2 or - I HI < (E/2)-m, (A) ... (2)
(1)+(2) gives I G I - I HI < e+m. (A)-m. (A). ...(3)
I. Let the set A be measurable so that m.(A)=mi (A).
To prove 1G 1- 1if I < e. ...(4)
Now (3) => I G 1- I HI < e+O. Hence the result.
II. Let (4) be true. To prove A a'l measurable.
The definitions of exterior and interior measure show that
m. (A) ~ 1 Gland mi (A) ~ I HI.
SO that m. (A)-m; (A) < 1(;1 - I HI < e, by (4).
But E is arbitrary and so making e~O,
m. (A)-m; (A) < 0 or m. (A) < m, (A).
But m. (A) ~ n1i (A) is true ¥ A.
Hence m. (A)=m; (A) so that A is measurable.
Problem 9. A subset E of [a, b) is measurablt> ~ffgiven e>O,
:I open sets G1 , G2 S.t. G1 :J E, O2 => E', m (G1 n G2 ) < e.
Solution. Let E C [a, b]. The definition of exterior measure
sllg~ests that 3 op(!n sets G1JE, OJ :J E' s.t. m(G1 ) <me (E)+6/2,
m (G 2) < m~ {£,)+./2.
This => m (G1)+m (G~) < me (E)+m. (E )+e.
But m (OI)+m {G 2 )=m (G l U G2 )+m (01 n O2 ),
Hence m (G 1 U G2 )+m (G 1 n G2 ) < m. (E)+m, (£')+e .. (I)
 88 LEBESGUE MEASURE OF A SET

G1 J E, G2 ::> E' ~ G1 UG 2 ::> EUE'=[a, b] => G1 UG 2 ::> [a, b]

=> G1 UG 2 =[a, b] as G1 U G. c [a, b]
=> m (G 1 U G2 )=b-a.
Now (1) => b-a+m (G 1 n G2) < m. (E)+m, (£')+_ ... (2)
I. Let E be measurable so that mil (E)=mi (E). . .. (3)
To prove m (G 1 n G2 ) < 6.

We know that mi (E)=b-a-m. (E')

or m. (E)+m ll (E')=b-a, by (3).
Now (2) becomes b- a+m (01 n G2 ) < b-a+c.
This ~ m (G 1 n G2 ) < IS.
II. Let m (G1 n Gs), < E To prove E is measurable.
G1 ::> E => m. (E) ~ m (Gl ). Similarly m. (E') ~ m (G 2 )
Hence mll(E)+m.(E');:;,:;; m(G1 ) +m(G2)=m(G1 U G2)+m(G1 nG2)
or mil (E)+mll (E') ~ b-a+6.
This ~ m. (E) ~ b-a-m. (E'H'E=m. (E)+ •.
Making £_0, we get m. (E) ~ m, (E).
But m. (E) ;;;;, m, (E).
This => m. (E)=mi (E) s> E is measurable.
Theorem 11). If A ill a countable set then prove that
m.(A)=O.
Proof. Let A be an enumerable (countable) set.
To prove m, (A)=O.
By assumption, A is expressible as A={xft : n EN}.
Let each point x" be enclosed by an open interval of length
_/2 in such a manner that these intervals form a family of
ft

disjoint open intervals. The measure of this family of sets is

- E (J)
..1!.1 _/2"= I -(t) =f.
This => A can be contained in an open interval of length _.
This Q m, (A) EO; _.
But II! is arbitrary and hence making 5 .... 0, we get mil (A) ~ O.
But m. (A) ;;iii 0 ¥ A. Hence m. (A)=O.
Corollary l. Since the set Q of rational numbers is counta-
ble and so m. (Q)=O. Also m. (N)=O=m. (I).
Corollary 2. Since the length of the interval (0, 1)
or [0, 1] is 1 so that m. «0, ]»=m ([0, l])=l:;t:O.
LEBESGUE MEASURE OF A SET 89

Hence [0, 1] is not countable.

Corollary 3. The converse of Theorem lO is not always true.
Example. Outer measure of Cantor's set is zero but it is
uncountable set. Also Cantor's set is itself measurable.
Hence any set with outer measure different from zero is un-
countable.
Theorem 11. lfme (A)=O, then A is measurable.
Or, to prove that a linear set 0" Lebesgue exterior measure zero is
Lebesgue measurable. [Banaras 1966]
Proof. Let m. (A)=O. To prove A is measurable.
We know that m. (A) ;:;at mi (A) ~ 0 Y A.
This => 0 ;;;, m, (A) ;:;at 0 => rni (A)=O.
Thus me (A)=O=m, (A) or m. (A)=m. (A), this =>
A is measurable and its measure is zero.
Ex. An enumerable set is measurable and its measure is zero.
[Kanpur 1976 ; Nagpur 64 ; Banaras 69]
Solution. Prove as Theorem lO that me (A)=O.
This => A is measurable and its measure is zero.
Problem 7. The difference of two measurable sets is measur-
able. [Meerut 1986J
Solution, Let El and E2 be measurable sets.
To prove that EI-E2 is a measurable set
EI-E2=El n E 2 '·
E2 is measurable => Ez' is measurable
~ El n E 2 ' is measurable
~ EI-E2 is measurable.
Problem 8. If a measurable set G1 contains another measura-
ble set G2 , then show that G1 -G2 is measurable alld its measure is
1I1(G1 ) -m(G2 )·
Solution G~ is measurable => G2 ' is measurable -
=> G1 n G2 ' is measurable
=> G1 -G2 =G 1 n G?' is measurable
Remains to prove that m (G 1 - G 2 )=m (G 1 )-m (G 2 ).
Since (G 1 -G2 ) n G2 =rp. Hence
111 [(G 1 -G2 )+G1 J=m (G 1 -G 2 )+m (G 2 )
or In (G1)=m (G 1 -GZ)+m (G 2 )
or In (G 1 }-m (G 2 )=m (G 1 -G2 ).
Problem 9. A singleton set is a meas!lrable set and its measure
is zero.
90 LEBESGUE MEASURE OF A SET

Proof. Let A={a} be a singleton set.

To prove that A is a measurable set.
Since every singleton set is a closed set and a closed set is a
measurable set. Hence A is measurable. Furthermore A can be
regarded as closed interval [a, a] which contains only the point a.
By definition of measure m ([a, a])=a -a=O.
m (A)=O. '
Problem 10. Sh~w that the Lebesgue measure of the following
set is zero
{x E R : 0 <: x < 1 and x has a decimal expansion not using
the digit 7}. [Banaras III, 1970]
An alternate statement. Prove that the measure of the set
of points in the interval (0, 1) representing numbers whose expan-
sion as infinite decimals do not contain some particular digit
(say 7) is zero. [Agra 1973, 65]
Solution. Divioe the open interval (0, 1) into 10 equal parts.
Delete the interior of the 8th part. Again subdivide the remaining
9 intervals, each in 10 equal parts al\d delete the interior of 8th
parts each of them. This process will be continued indefinitely.
The points, which remain, constiute the required set. We denote
this set by E. Then E={x E R : 0 < x < I and x has a decimal
expansion not using the digit 7}.
(0, l)-E is the complementary set E'. The set E' consists of
an enumerable number of disjoint open intervals. Here E' is an
open set. An open set is always measurable whose measure is
the sum of measures of intervals which form the open set.
The number of intervals deleted and their lengths can be
written as
No. of subdivision No. of intervals Length of each
'deleted deleted interval
1st 1 ]/10
2nd 9 1{102
3rd 92 1{103

nth 1{1O"
LBBSGUE MEASURE OF A SBT 91

1 9/102 1 9/102 10
= 10 +[='9710=10 + -J(Hf=10=1
m (E')=1
m (EU E')=m (0, 1)= 1-
or m (E)+m (E')=1. For EnE'=q,
°
or m (E)=l-m (E')=I--1=0 or m (E)-=O.
Therefore measure of the required set is zero.
Remark. The end points of the deleted intervals with their
limiting points from a non-dense perfect set.
Problem 11. Let x be a number in the interval (0, 1) written
in the scale of 10 as x=[Link] ... x" ... where none of the x's is 5.
Find the content (or measure) of the set so defined
Solution. Divide the interval (0, 1) into 10 equal parts.
Delete the interior of the tth part. Again subdivide the remaining
9 intervals each in 10 equal parts and delete the interior of 6th
part in each of them. Tt,.is process will be continued indefinitely.
The points, which remain, constitute the required set. We denote
this set by E. Then
E={x€R: O<x<1 and x has a decimal expansion not
using the digit 5}.
«( , 1)-£ is complementary ~et E'. The set £' consists of
an enumerable number of disjoint open intervals. Here E' is an
open set. An open set is always measurable whose measure is the
sum of measure of intervals which form the open set.
The number of intervals deleted and their lengths can be
written as
No. of subdivision No. of intervals Length of each
deleted. deleted interval
15t 1 1/10
2nd 9 1/102
~rd 9CI 1/103

11th ]/10"

1 9 92 S,,-l
m (E')=IO+ 102 +103+"'+ 10"-+'"
1 9/102 _1 9/102
=10+1-9/10-10+./10 =1
m (E')=I.
92 LW3E GUE MEASURE OF A SEf

m (EuE')=m (0, 1)=1-0=1, EnE'=~

m (E)+m (E')= 1
or m (E)=l-m (E')=1-1=0
Measure of the required set is zero.
Problem 12. Find the measure of the set {x.} in' (0, 1) where
x,. expressed as an infinite decimal does not contain the digits
5 and 7.
Solution. Divide the interval (0, 1) into equal parts. Delete
the interiors of 6th and 8th parts. Again subdivide the remaining
intervals, each in 10 equal parts and delete the interiors of 6th
and 8th parts in each of them. This process will be continued
indefinitely. The points, which remain, constitute the required
set. We denote this set by E. Then
E={xER : O<x<l and x has decimal expansion not using
the digits 5 and 7}
(0, 1)-E is the complementary set E'. The set E' consists of
an enumerable number of disjoint open intervals Here E' is an
open set. An open set is always measurable, whose measure is
the sum of measure of intervals which form the open set.
The number of intervals deleted and their lengths can be
written as
No. of subdivision No. of intervals Length of each
deleted deleted interval
1st 2 1/10
2nd 2-8 1/102
3rd 2'8 2 1/103

1/10"

, 2 2-8 2-8 2 _ 2-8,,-1

m (E )=[6+ IOZ+)]3 +-_.+ ~ +-_.
2/10 2/10
=1~871()=2/10 =1
m (E')=1.
But m (EUE')=m (0,1)=1-0, and E~E'=rp
m (E)+m (E')=I-O=1
or m(E)=l-m(E')=I-l=O
Measure of the required set is O. Ans.
LEBESGUE MEASUEE OF A SET 93

76. Definition. A relation which holds except in a set of

measure zero is said to hold almost everywhere. The phrase
"almost everywhere" is, in short, written as 'a.e'.
Remark. The phrase "almost everywhere" is a particular
case of "everwhere."
7'7. Cantor's tennary set.
[Vikram 1969; Banaras 66, 64; Meerut 1990, 88, 86]

1r-~F1~-rt::.=::::::=='1 M
0121 2!!
iii 3 9 9

Take the closed interval [0, 1). Divide the closed interval
[0, 1) into three equal parts. Take away the interior of the middle
part, i.e., the open interval (1/3, 2/3). This is the first stage of
our construction.
Again subdivide the remaining two closed intervals [0, 1J
and [2/3, 1], each in three equal parts. Delete the interiors of
the middle parts i.e. the open intervals (3~" 3~ ) and ( ~ , ~ )-
This is the second stage of our construction. Similarly at ptll
I
stage, 2p - 1 open intervals are removed of length 3:11 each. The
points, which remain, constitute Cantor's tennary set. We denote
this set by E-[O, I] - E is the complementary set £' w.r.t. [0, 1]
and consists of an enumerable mutually disjoint open intervals ..
(i) Hencl;! E' is an open set. An open set is always measura-
ble whose measure in the sum measures of intervals which form
the set £'
1 2 2J 2"-1
m(E')=3+ji+ 33+······Jri + ...
=i [l +2/3 +(2/3)2+ ... ]
1 1
=3' 1-2/3 =1.
• m(E')=I
m (E)+m (£')=1
or m (£)=1-1=0.
This zero is the measure of the set E.
[Kanpur 1989; Meerut 88, 86; BaDaras 6"]
94 LEBESGUE MEASURE OF A SET

(ii) Since g' is open and hence E is closed.

This ~ D (E) C E.
Also here we get ECD (E).
so that E=D(E).
This ~ E is a perfect set.
(iii) Let the points of E be represented by infinite decimals
(in the scale of 3).
Let O.olo2aa ... in the scale of 3) be a point of E.
S' 0 a1 a2 + a3 a.
Ince .a,02aa···=3+F 33-+"'+3"'"
Since we are expanding decimals in the scale of 3. Hence the
a1, a2> 0a, ... can not take the value 3 or greater than 3. Hence the
only possibility is that 0 1 , a2 , aa, ... can take the values 0, 1,2. At
the first step all the points with a1 = 1 are removed. Similarly
second step removes all the remaining points with 02= 1. Simi-
larly at nth stage all the points with 0" as 1 are taken away. It
means that a1 , a 2 , ... can not take the value 1. Thus we see that if
o a1a2a3' .. is an element of E, then
a.. =O, 2 ¥ n E N.
The points 1/3 and 2/3 are taken to be represented by
00222 ... and 0'2000 ...
(iv) To prove tbat the set E is uncountable. [Meerut 1990, 86]
Suppose not.
Then the points of E form a countable set. It means that
every point in E must appear in the sequence.

of distinct points.
i.e., E={x...: n E NJ. . .. (1)
Writing decimal expansion of these x/s,
Xl =O'Xll X 12 XU" ,Xltn
X2=0'X 21 X 22 X 23 · "X1'"

X,,=O'X"l X"2 X"3" . X"", . ..

where each X,p=O or 2 ¥ rand ¥ h. . .. (2)
Construct real number
~=0'fl~2~3"
s.t. if x ...",=O, write ~",=2
LEBESGUE MEASURE OF A SET 95

and if x".".==2, write ~=O.

In either case x..¢~". ¥ m.
This => x",¢~ ¥ m =>~ it E ... (3) according to (I).
": f",=O or 2 ¥ m.
This => t e: E ... (4), according to (2).
Evidently (4) contradicts (3). Hence the required result
follows.
Tbeorem 12. Show that the Lebesgue exterior measure ;s a
Caratheodory outer measure. [Banaras 1967]
Proor. To show that Lebesgue exterior measure is Caratheo-
dory outer measure, it is enough to show that
(i) m. (E) ;;. O.
(ii) El C E z => m, (E1 ) <; m. (E2 ).
(iii) For any finite or infinite sequence of [Link]
EIO E 2 ...... , E,........

" .
m. (u E k ) <; E m, (Ek )

(iv) m. {El U Ez)==m. (El)+m. (E2) jf E'l. n E.==",.

By definition, m. (E)=inf {m (G) : G is open, G :J E}.
(i) From this it is quite clear that m. (E) ;;. O.
For m (G) ;;. O.
(ii) Let El C E 2•
To prove m. (E1) <; m. (EJ.
For proof refer Theorem 3 (ii), Page 76.
(iii) To establish the result (iii), it is enough to show that

m" (E) < ~. m.(E

00

1'-1
k ). where E= U E 1,.
00

t-l

Case 1. When the series Em. (EI:) is divergent. Since every

covering of E is a covering of E l , E 2 , : and therefore
m. (E) ~ 00.
Case 2. When the series E m. (EI:) is convergent. Let the
/I

family {ItI : j= I, 2, ...... } of open intervals be open covering of

Ek y. k. Given E > 0,

Here the totality of these open intervals is at most enumera-

ble. Evidently this is open covering of E.
96 LEBESGUE MEASURE 'OF A SET

00
= E m, (Ek
k-l
)+_.
00
Thus m, (E) ~ ~ m. (E.,)+E.
k~l

Making E-+O, we get the required result

00
m. (E) ~ E m. (E,,)
k=l

(iv) Let El n E 2 =q, so that d (EI , E 2 ) > O.

d being the metric on X. Let d (EIO E 2 )=3 > o.
Also let E=E1 U E 2 •
To prove that m. (E)=m. (El)+m. (E2 ).
Let {I., : k=l, 2, ...... } be an open covering to E S.t.
d {I,,) < a, d (lk) represents the diameter of I".
Case I. When either m, (El ) or m2 (E2 ) is finite.
Let T={Ik' : k= 1,2, ..... } be a sub-family of {II: : kEN}
00
s.t. U /,.' covers E1 •
/.=1

Similarly suppose that S={I/: k=l, 2, ...... } is a sub-family

00
of(l,,: k=l, 2, ... } s.t. U I/:- covers E 2 , where
1.:_1

k E N}={I/, 12' la'" .. ... , 11' , / 2', 13- , ...... }. Obviously Tis
{lk :

an open covering of E. and has no points in E2 • Similarly S is

an open covering of E2 and has no points in E1• Then
, 00 00
m.(El)+m,(E2)~ E m(llc')+ 1J m(ll:)
11-1 "~l

00 00
= ~ m (llC' U [,:)= 1: m (l/l:)=m. (E)
/:=1 k=I'

i.e., m. (E1 )+me (E2 ) ~ m. (E).

By (iii), just proved,
m. (E1)+m, (E2 ) ;;.. m, (E1 U E2 )=m. (E)
i.e. m. (E1)+m. ;;. m. (E). ...(4)
Combining (3) and (4), we get the required result.
LEBESGUE MEASURE OF A SET 97

Theorem 13. The uniOll of a finite number of measurable sets

is measurable.
Proof. Let E1 , E2 •••••. , En be measurable sets and let
II
E=U E".
k=1

To prove that E is measurable.

Consider an arbitrary fi > O..
E~ is a measurable set => 3 a closed set P,: and an open set
G" S.t. F" C E" eGa,
m (G,..}-m (F,.) < nF. ... (1)
n n
Write F=u F,,, G= U G.
k=1 k=1

Being a finite union of closed sets, F is closed.

Similar arguments prove that G is an open set.
,. II R

F" C E,.. C G" => U F" C U Ek C U G"

"=1 k~l "=1
=> Fe E C G.
=> FeE, E c G.
=> m, (F) ~ mt (E), m. (E) ~ m. (G)
(Refer Theorem 3)
=> m(F)=mi(F) ~ mi (E), m. (E) ~ m. (G)=m(G)
[For Closed sets and open sets both ore measurable sets]
=> m(F) ~ m, (E) ~ m. (E) ~ m (G)
=> m. (E)-m, (E) ~ m (G) -11'1 (F). . .. (2)'.
G-F=G n F'. For G n F'=Grt(X-F)=GnX--GnF=G-F.
=A finite intersection of open sets
=An open set=A measurable set
G={G-F) U F and (G -F) n F=r/>.
:. m (G)=m (G-F)+m (F).
From which, m (G-F)=m (G)-m (F)
Similarly, III (G,. -F, )=m (G,) In (Ft ).
Notice that
n "
G Feu (G,. -FJ. This => III (G -f) ~ E III (G k - Fd
k-l k=l
98 LEBESGUE [Link] OF A SET

~ m (G)-m (F) .;;; E• [m (Gt)-m (Ft )] < •

-.n=., by (1)
11=1 n

.. m (G)-m (F) <: E.

Using this in (2),

m_ (E)-m, (E) C;; m (G)-m (F) < fi

or m. (E)-m, (E) < •.

But. is arbitrary. Hence it can be made as small as we please.
Making ..... 0 in the last inequality.
m,(E)-m,(E) E;; O.
This ~ m,CE) .;;; ml(E)
But m_(E) ~ m,(E) is always true.
Combining the last two, m.(E)=mi(E)
This proves that E is a measurable set.
Tbeorem 14. The intersection of a finite number of measur-
able sets is measurable.
Proof. Let E1• Ea ....... ' En be measurable sets.

To prove that u" E1, is measurable.

11=1

We know that:
The union of finite number of measurable sets is measurable.
Let all the sets E/: be contained in an interval fl. Then
Et'=complement of Ek w.r.t. fl.

By De-Morgan's law
11=1
UE,,' = [ .=1
U Ek]' ... (1)

Ele is measurable .. its complement E'/: is measurable.

~ u"
11=1
e~' is measurable.

~ [ "'=1
U E1:]' is measurable.
" is measurable.
onE"
"-1
LEBESGUE MEASURE OF A SET 99
Theorem 15. First Funclamental Theorem. If E1 , E2, •••
are disjoint measurable sets and
E=E1 +E2••••••
00
then E is measurable and m (E)= E m (EI»'
1>=1

[Meerut 1990; Kanpur [Link]. P .. l98S; Banaras 68]

Proof. Let E J , E 2 , ...... be disjoint measurable sets so that
m. (Ek)=m, (Ek)-m (E k ) yo k
E, n EI=cfi for i='l=j.

Let all the sets be contained in an interval (a, b] then

or ... (1)

• •
Let S,.= E 1-1" then m(S.)= Em (E,). For ErnE.=cfi for r='l=s
r~l r=l

00
Then S" C 1: Er=E, i.e., S. C E so that E' C Sfa'.
r=1

E' C S,,' ~ m. (E') ~ m. (S,.')=m (S.')

=> m. (E') ~ m (S,/)
[For S. is measurable~ by theorem 13 and so S,,'
is measurable].
=> m. (E') ~ m (S,,')=b-a-m (Sn)

=b-a- "E m (E,,)

r=1

~ m. (E') ~ " m (Er)

b-a- E
r=1

n
=> E m (Er) ~ b-a-m. (E')=m, (E),
r=1
100 LEBESGUL MEASUF.E OF A !'oET

:0
• m (E)
m, (E) ;;. E
7'=1

00
Making n-+oo; we obtain m. (E) ~ E m (Er)
7'=1

=> m, (E) ;;;;. E m (Er) ~ m. (E) ... (3), On using (I)

7'=1

.. m, (E) ;;. m. (E).

But m, (E) ~ m. (E) is always true.
:. m. (E)=m. (E).
This ., E is measurable and
00
.. m (E) ~ E m (Ek ) ~ m (E), by (3).
~_1

This concludes the problem.

Theorem 16. Second Fundamental Theorem. If E 1 , E 2 ,···are
00
meosuroh/e sets, then the set U Er is measurable.
7'=1

00
Proof. E= U E~ is a measurable set.
k_l

Let

By De-Morgan's law, F'=[ nE,; ]= UE

k-l ~=l
t'

Ek is measurable ::> its complement Ek ' is measurable.

00
=> 1: E,r' is measurable
1:=1

.. F' is measurable => F is measurable.

Lialtinl Sets.
Theorem 17. Suppose that < E" > is a mOl/otonic increasing
 LEBESGUE MEASURE OF A SET 101
~equence, i.e., each set of the sequence contains the following one.
00
If E is their limiting sum, i.e. E= U £.,., then
11=1

(E) Lim E}
m \. = n_oom ( n. [Gujrat 70]
An Alternate Statement. Define a set function on a ring R.
If a set function
~ on a ring R is countable additive and there
exist A and a sequence < .An > e R such that A, C .If'+! and
00
.4.= U ...4., prove that 9> (An) ~ rP (A) as n~a::>.
ra=1

[Kanpur [Link]. P. 1910]

Proof. Let < E" > be a monotonic increasing sequence of
measurable sets, so that
00
El C E2 C £3 C···· .. Also let E= U £1"
lII=1

An enumerable union of measurable sets is measurable. Also

difference of two measurable sets is a measurable set. Hence E
and E,+l-E,. are measurable sets. Notice that
£=E1 +(E2 - E1 )+( E3 - £2)+'" +(E,-E,-I)+ ..... .
00
i.e. £=E1 + ,=1
~ (E"+l-E,,)

is a disjoint union of measurable sets. Therefore

00
m (E)=m (El) + ,,_1
~ m (Er+l-Er) ... (J)

" (m (Er+l)-m (E,)]

m (E)=m (El )+ 1:
or ,=1

or ... (2)" .
00
For 2-' [m (Er+J--m (Er)]
,.=1

-=Lim [{m (E?)-1J1 (E1)}+{m (Ea)-m (E2 )}

n-+oo
+ ... +{m (En)-m (Ell-1m
102 LEBESGUE MEASU~E OF A. SET

Lim
From (2), m (B)= n_oom (En).
Ex. Show that if < Eft > is sequence of measurable sets all
contained in a set offinite measure and Lim En exists, then Lim Eft
is measurable and m (Lim E,,)=Lim m (Eft). [Banaras 1967J
Soilltion. Let < En > be a sequence of measurable sets.
Then this sequence will be either monotonic increasing or mono-
tonic decreasing. To be particular at this stage, we further sup-
pose that the sequence is monotonic increasing so that
E1 C E2 C E8 c ..... .
00
Let E= U Ek • Now prove as in the abOYe theorem 17 that
.10-1

m ,lim
\n_oo
En )=lim
n_oo
In (E).
/I

Theorem 18. If E 1 , Ez, ...... are measurable sets and

00
n E~ and if El ::> E2 :) ... , then
E= r_1

Lim m (Eft), z.e.

m (E)= n~;rJ . m (Lim Eta ) = Lim m (En).
n_ 00 n_oo
Or
Let p. be a countably additive non-negative measure defined on a
a-algebra of subsets of a set X such that I' (X) < 00. ~lrow that if
{An: n E N} is a contracting sequence of I'-measurable subsets
of X, then

I' [ n ] Lim
n E N An == nE N J.I. (An).
[Banaras 1964J
Proof. Let < Eft > be a contracting (monotonic decreasing)
sequence of measurable sets so that
~I ::> E2 ::> Ea ::> ......

00 Lim
Also let E= n En. Tken E=
,,-I "_00 E.
LEBESGUE MEASURE OE A SET 103

i.e. ... (1)

is a disjoint union of measurable sets. Since an enumerable union

and intersection of measurable sets are measurable. Also difference
of two measurable sets is measurable. Now (1) is expressible as
00
m (E1)=m (E)+ l: m (Er-Er+1 ). ...(2)
r=1

00
or m (E1)=m (E)+ E [m (Er)-m (Er +J )]
r=l

"-1
or E [m (Er)-m (Er+l)]
r=l

or ... (3)

= n-+oo
Lim [m (El)-m (Eq)]+[m (E2
~
)-riz (E.)1
+ ... T[m (E"-l)-m (ErI ) ]

= Lim [m (E 1 ) - m (En)]
[Link]
(3) can also be written as
Lim
m (E1)=m (E)+m (E1)- n'+oo m (En)

Lim
or m (E)= III (En).
n~OO

Theorem 19. The set of all measurable (i.e. Lebesgue measu-

rable) sets is a--algebra (or a-field).
Proof Let A be the set of all measurable sets.
Let A, B E A be arbitrary so that A and B are measurable
sets.
To prove that A is a -a-algebra, we have to show that
(i) A' E A
104 LEBESGUE MEASURE OF A SET

(f'i) AU.:; .2 A
(iii) if < En > is a sequence of measurable sets in A
00

then u E, E A.
~=1

Let all the sets belongiilg to A be contained in an interval (a, b).

(i) Let A be measurable so that m. (A)=m, (A)=m (A) .... (1)
To prove that the complement A' of A is measurable.
We know that m, (A)=b-a-m. (A'). ...(2)
This is true for every A.
Replacing A by A',
m, (A/)=b-a-m. (A) ... (3)
writing (2) and (:S) with the help of (I).
m (A)=b-a-m. (A')
mi (A')=b-a-m (A)
From which, m. (A/)=b-a-m (A)=m, (A')
or m. (A')=m, (A').
This => A' is measurable.
(ii) Let E 1 , E2 E A so that El and £2 are measurable sets and
let E=E1 U E 2•
To prove that E E A, we have to show that E is measurable.
Consider an arbitrary number c > O.
Ek is a measurable set => :I a closed set F" and an open set
Gle e s.t. F.,. C E" C Gk ,
E
m (G,,)-m (FTe) < T'
Write
Then F is a closed set and G is an open set so that F and G
are measurable sets,
2 I I
FTe C Ei C G" => U Fit
.=1
e i=1
u Ek e II_I
u G7r,

~ Fe Ee G.
=> FeE, E e G•
.. m, (F) ~ m, (E), m, (E) ~ In (G)
o m (F) ~ m, (E) ~ m. (6) ~ m (G)
[For m, (F)=m (F), m. (G)=m (G)
and m, (A) <; m. (A) ¥ A]
=> m. (E)-m. (E) ~ m (G)-m (F). •.. (4)
LEEr~Gt'E MEASURE OF A SEt lOS

2
This ~ m (G-- F) E;', l: m (G,,-Fre )
k=l

=> m (G) -m (F) ~ k:l

2
[11'1 (Gre)- 11'1 (Fk ) ]
e
< 2.2=.
=> m (G)- m (F) < ,
=> m. (E)-m. (E) ~ m (G)-m (F) < I. by (4)
=> m, (E)-m. (E) < E.

MakjngE~O,

m. (E)-I1I. (£) ~ O. For I is arbitrary.

This => m. (E) ~ In. (E).
But m. (E) ~ m. (E) is true for any set E
=> m. (E)=m. (E)
=> E is measurable.
(iii) Let < En > be a sequence of measurable sets and let
00
E= l- En .
.. =1

To prove that £ is a measurable set.

For the sake of convenience we suppose Ek nE/:'=0 for k:l=k'.
Here the operation E stands for u.
We know that

or

00
or - m. (E) ~ ~ m (E,,). . .. (5)
.=1

Let •
S.. = 1.' Ek •
1c~1

El and E. are measurable sets. El U E. is measurable by (ii),

=> S.. is measurable
106 LEBI:SGUe MEASURB OF A SET

~ its complement S ..' is also measurable.

m (S..)=m ( 1:
Jt~l
EII)= i:
k=1
m (Ek )

i.e.
.
m (S..)= 1: m (Et ). . .. (6)
11=1

S.. = 1:" EkCE i.e. Sn C E so that E' C S,.'.

1=1

E' C S ..' ~ m~ (E') ~

me (S,.')=m (S,/)
~ m~ (E') .s;; m (S,,')=b-a-m (S.. )
~ m. (E') ~ b-a-m (Sn)
=- m (5..) .s;; b-a-m. (E')=m. (E)
=- m (S.. ) ~ mi (E)
II
~ nt. (E) ~ m (S..)= }; m (EA:)
.=1

~ 111. (E) ~ E " m (Ek ).

11=1

.Making n-'?oo, .
00
m. (E) ~ L: m (Ek ) ;;;;. me (E), on using (5).
. . 1:=1

or mi (E) ~ m~ (E).
But m. (E) ~ m. (E) is true for any set.
Consequently m. (E)=m. (E).
This ~ E is measurable.
This concludes the problem.
7'8. Definition. Convering in the sense of Vitali. [Banaras IV 70]
Let E ·be a set and M be family of closed intervals such that
none of them is a singleton set. The set E is said to be covered
by the family M in the .sense of Vitali if -
[Link] x E E and ¥ ~ > 0 ~ 3 a closed interval I E M [Link]/,
In (I) < 6.

That is to say, E is said to be covered oy the family M in

the sense of Vitali if every point of the set E is contained in
some small closed intervals belong to M.
LEBESGUE MEASURE OF A SET 107

Theorem 20. Vitali's Covering Theorem. ~f a bounded set E

is covered by a family M of closed intervals in the sense of Vitali
then it is possible to find a countable (finite or enumerable sub-
family of closed intervals {/k} of M such that
l",n/.. =0 for m=l=n and m. [E-E I,:J=O.
I,

[Punjab 1967; Banaras IV 70J

Proof. Let E be a bounded set which is covered by a family
M of closed intervals in the sense of Vitali so that
x E E, e > 0 ~ 3 a closed interval I!. E M s.t.
m (/10) < e ...... (*).
To prove that 3 a pairwise disjoint sub-family {/k} of M with
the property m. [E - E Ie] = 0,
Ie

where the symbol E stands for union.

Choose an interval (a, b) S.t. E C (a, b). Form a family Mo
of closed intervals from M s.t.
M o={I,: EM: II, C (a, b)}.
From the construction it is clear that Mo C AI and every
member of Mo is contained in (a, b).
Also Mo covers £ in the sense of Vitali.
Pick an interval II E M.
If E ell' then the theorem is proved.
For m. (£-/1 )=0, by virtue of (*).
In the contrary case pick another interval 12 E Mo
s.t. 12n/I=~
If E c 12 , then the theorem follows.

For m. ( E - 1:
k-l
I" )=0, by virtue of (*).

III the contrary case, pick an interval 13 E Mo S.t. I". n 1.. =~ for
m=l=n and tn, n= I, 2, 3.
Repeating this process Il times, we have a family consisting
of pairwise disjoint closed intervals 110 12 , ....... I".

If £ C E" Ir, the proof is complete.

,=1

In the contrary case, E <t: ...• Ir so that

r-1
108 LEBESGUE MEASURE OF A SET

.(1)

,.
In this case, Sl!t F,.= E I r , G,,=(a, h)--F,..
r~l

Now we consider those intervals of M which are subsets of

the open set 0,.. By virtue of (1) such intervals exist and their
length ~ m «a, b».
Let k .. denote the least upper bound of the lengths of these
intervals and let 1,.+1 be one of them for which

m (1.+1) >
k
i
Of course k" > O. For no interval of the family is a singleton set
Therefore I ..+1 nI.. =rp for r=l, 2, ......... , n-l. If the process of
constructing the intervals II' 12 ........... ends after a finite numl'er
of steps, then the theorem is proved.
If this process docs not come to an end after a finite number
of steps, then we are left with a sequence < I. > of pairwise
disjoint closed intervals. Now we shall show that this sequence
also satisfies the given condition, i.e.,

me [E- ~
k=l
h ]=0. . .. (2)

For proving this, we construct a closed interval D/: ¥ k s.t.

(i) the length of Dk is 5 times the length of I,,, i.e.
m (D 1,)=5m (lk)'
(ii) the middle point of Dk coincides with the middle point
of II;.
By initial assumption, every element of the sequence < In>
is contained in (a, b) so that

III (1~1 I. ) ~ 111 «a. b». ...(3)

Consequently III (~l Die )=sm( tl lit; ) ~ m «a, b»

=finite number.
LEBESGUE MEASURE OF A SET 109

... m( ; D.. )
7<-1
< 00 •

E- -
In order to prove (2), we shall prove that

-
E I .. C E Die ¥ i. . .. (4)

-
.=1 7.:=,

Let x E E - E I", then x e= G, ¥ i.,

• 1t=1

X E G" Gi is open => i a closed interval I E Mo s.t.

x E Ie Gi •
However, it is impossible that ICC ... . .. (5)
This follows from the fact that
m (I) ~ k tt ~ 2m (I'Hl) ¥ II.
This is impossible. For m (1,,)-+0, by (3).
Consequently the relation (5) is not satisfied for some nand
hence the relation
I n F..=I=. ... (6)
holds. Next we assume that n is a least positive integer s.t. (6)
holds. Therefore n > i.
For 1 n Fi=rP and Fl C F2 C Fa···
From the definition of n, it follows that
1 n F..- 1 =;·
We can now deduce two results from this. The first is
I n I.=I=if. . .. (7)
and the second j., , C G"- l
and m (I) ~ k"-1 < 2m (I.. ). . .. (8)
From (7) and (8), we find that I C DII •

-
As a result of which ICE D/: so that x E
Iccd
-
X Dk
1 ~t

Hence (2) follows.

Remark. This proof was originally given by the famous
mathematician Banach.
Under the hypothesis of the prl!vi<.'us Theorem 19, we can
prove the following theorem:
Theorem 21. For every, > 0, 3 a fillire sub-srstem '1' 12 ""
I .. of pairwise di.~;oil/l elored il/tena/s of the system :\,1 S.t.
110' LABESGUE MEASURE OF ~ SET

m. [E - i: It J<
t=1
E.

7'9. Definition. Set of the type Fa. A set E is said to be of

the type Fa if it is expressible as a union of an enumerable num-
ber of closed sets Flu i.e.
GO

E= U Ft.
k=l

7'10. Definition. Set of the type G,. A set E i'i said to be

of the type Ga if it is expressible as an intersection of an enume-
rable number of open sets G/;> i.e.
00

E= n G".
k=1

Evidently complement of Fa is G6 and conversely.

7'11. Definition. Borel Set. A set E is called a Borel set
if it can be obtained from closed and open sots by using a finite
or an enumerable number of union and intersection operations.
Example. The sets of the type Fa and G6 are Borel sets.
7'12. Definition. Borel measorable.
A bounded Borel set is said to be Borel measurable.
Most writers refer to all Borel sets as being Borel measurable.
Theorem 22.' The sets of the type Fa and G. are measurable
sets.
Proof. Let E be a set of the type Fa. Then E is expressible
as

-
E= EFt.
/1=1
where FI: is a closed set.

Fir. is a closed, set ~ Ft is a measurable set.

Also an enumerable union of measurable sets is measurable.
00
Hence U F;" i.e. E is measurable
"=1
(ii) Let E be a set of the type Ga so that E can be expressed
as
LEBESGUE MEASURE OF A SET 111

Gk is a open set * Glo is a measurable set.

Also an enumerable intersection of measurable sets is measur-
00
able. Hence . U G1" i.e. E is measurable.
110=1

Tbeorem 23. A Borel measurable set is Lebesgue measurable .

.£Meerut 1987, 86J
Proof. T~et a set E be Borel measurable so that E is a boun-
ded Borel set. To prove that E is Lebesgue measurable.
E is expressible as E=(U Gk ) n (n Fk )
k k

where Gk is an open set and Flo is a closed set.

E is bounded ~ all Gk and aU F_ are bounded. Also G1, is
open and F" is closed.
~ Gk and F1, both are Lebesgue measur~ble
sets.
~ U G1" n Fr, are Lebesgue measurable sets.
k k

For a countable union or intersection of

measurable sets is measurable.
~ (U Gk ) n (n Flo) is also Lebesgue measurable
Ie Ie

by the same reasoning.

=:> E is Lebesgue measurable.
Exercises
1. Define a set function on a ring R. If a set function.p on
a ring R is countably additive and there exist A and a sequence
< An > E R such that Ar C Ar+1 and
00
A= U A,., prove that .p (A .. ) -+ c/> (A) as n_oo •
..=1

[Kanpur M. See P. 1980]

2. If E1 and E2 are measurable subsets of [a, b], prove that
E1 U E2 , El n £2 and E1-E2 are measurable and show also that
. (i) m (E1)+m (E2 )=m (E1 U E 2 )+m (E1 n E 2 )
(ii) m (E1}-m (E2 )=m (E1 -E2 ) if E2 c E1 •
[Kaapur M. Sc. P. 1979]
3. Define the Leb~sgue measure of an open set of real
numbers. Prove that an enumerable set is measurable and its
measure is zero. Show further that a continuous function defined
in a cl03ed interval is measurable. [Kanpur 19731
112 LEBE~GUE MEASURE OF A SET

4. Define outer and inner measures. If 1-'''' is an outer

measure on Q and M is the class of all ",·-measurable sets, then
show that M ill a _a-field and fL'" is a measure on M.
[Kanpur Statistics 1977]
S. Define measurable sets in the sense of Labesgue giving
two examples.
If X and Yare measurable sets, show that Xn Y is measur e.
able. What can you say about X U Y? LKanpur 1975]
6. (i) Give an example of a set which is not measurable in
the Lebesgue sense. Justify your example.
Oi) Is the Lebesgue measure of the union of two measura-
ble sets equal to the sum of their measures? [Kanpur 1974]
7. (a) Define the Lebesgue measure of an open set of real
numbers. If 0 1 and O2 are any two open sets then prove that
m(01+ 0 ;l)+m(Oh 02)=m(01)+m(02)' where m(O) dendtes the
Lebes'!ue measure of 0 1 ,
(b) Prove that a set E of real numbers is measurable iff,
given E > 0, EI set~, el' e2 s.t. E=~+el-e2 and ~ is the sum ofa
finite number of intervals, and m. (eJ < E, me (e2) < c.
[Kanpur 1973]
8. (a) Let E 1 , E 2, ... , be a sequence of sets of real numbers,
Prove that
m* (E1 +E2 + .. ,) ~ m* (E1)+IJ-.* (E2 )+, .. ,
where m'" (E) is the Lebesgue outer measure.
If in addition the sets E, are pairwise disjoint, i.e., no two of
them have common points, then prove tha t
m* (E1 +E2 +, .. ) ~ m", (E1)+m", (E2)-+-:"
where m* (E) denotes the jnner measure of E.
(b) If M is measurable set, then prove that for any set E,
m'" (E)=m""(EM)+m'" (E-£.\4). lKanpur 1972)
9. Define Lebesgue measure of a set. Construct a non-dense
perfect set in the interval (0, 1) whose measure is 1/2.
lKanpur 1968J
10. Define the Lebesgue exterior measure m"'(A) of a subset
A of R. For any sequence < A.. > of subset of R, prove that
m"'( U An) ~ I. 11/* (A .. ). [Banaras Ill, 1971J
,
 LeBESGUE MEASURE OF A SET 113

11. (a) If A, B, Care I'-measurable sets, show that

peA) +I'( B) +1-'(C) =,..( A U B U e) +1'(Bn e) + l'(en.4)
+1'(.4 n B)-I-'(AnBnC).
(b) Show that the Lebesgue measure of the following set is
zero:
{x E R : 0 < x < 1 and x has a decimal expansion not
using the digit 7}. [Banaras III, 1970]
12. Define measures of open and closed sets and prove that
m(E1 )+m(E;)=m(E)UE2 H-m(E1 nE2)
where E) and E t are both open or both closed sets.
Show that a countable set is measurable and is of measure
zero.
Prove that an outer measure is monotonic and cr-subadditive.
[Danans 1%9]
13. Prove that
Ii) m*(E1 uEz)+m*(E1nE2 ) ~ m*(E1)+m*(Ez)
(ii) m.(EI UE2 )+m.(EI nE2 ) ~ m*(EI )+m.(E2).
00

.-1
Show that if E 1 • E2 ... are measurable and disjoint, then U Ell is

measurable and m (U
t=I
Ek)= Um(E
Ie-I
k ). [Danaras 1968]

14 Construct a set which is not measurable in the sense of

Lebesgue. [Danaras 1968]
15. (a) Prove that Lebes~ue exterior measure is a Caratheo-
dory outer measure.
(b) Show that if < E.. > is a sequence of measurable sets
all contained in a set of finite measure and lim Eft exists, then
lim E,. is measurable and
m (lim E,,)=lim In (E.. )
where the limit on the right hand side of tile equltlity exists.
[Banaras 1%7]
16. (a) State Caratheodory's criterion of measurability of
a set. Prove that a linear set of Lebesgue exterior measure zero
is Lebesgue measurable.
(b) DGscribe the construction of Cantor's ternary set and
find [Link] Lebesgue measure. (Baoaras 1966]
114 LBBESGUE MEAWRE OF A SET

17. (a) Establish the equivalence of Caratheodory's and

Lebesgue Criterion of Lebesgue measurability of a bounded sub-
set of the real line.
(b) Show that the Lebesgue measure of the Cantor sub-
set of the unit intt:rval is zero. [Banaras 1964]
18. If,.,. is a measure on a ring Rand < E,. > is an increa-
sing sequence of sets in R for which lim E.. E R, prove that
.
,.,. (lim E.. )=lim ,.,. (En). [Gujrat 1970J
,. n
 8
Measurable Functions

8'0. Definition. Measurable function.

[Poona 1970; Meerut 71; Kolhapur 69;
Kanpur 1983,74,73, [Link]. P. 75,76; Gujarat 70]
An extended real valued function/. defined over a measura-
ble set E, is said to be measurable (in the sense of Lebesgue) if
the set {x E E :/(x) > al ...(1)
is measurable for every extended real number a.
The set {x E E: / (x) > a} is denoted by the symbol
E(/> a).
Remark (1) The condition (1) of the above definition can be
replaced by any of the three conditions of theorem 4, page 118.
Remark (2) The measure of the set E (/ > a) may be finite
or infinite.
Remark (3) If E=R, then the set E (/ > a) becomes an
open set.
8'1. Definition. Almost everywhere.
A relation, which holds except ID a set of ,measure zero, is
said to hold almost everywhere. That is to say, a property P is
said to hold almost everywhere on the set E if 3 Eo C E S.t.
(i) the property P holds 'V x e E-Eo
(ii) m'(E.,'=O.
The following have the same meaning
(i) almost everywhere on the set E
(ii) for almost all points of E.
The phrase almost everywhere, is denoted by the symbol
a. e. or p. p.
S·2. Definition. Equivalent functions.
Two functions/and g defined on the same set E, are said to
be eqUivalent if m [E (f:,cg)]=O.
That is to say, functions / and g defined on the same set E,
are said to be equivalent if 3 A, BeE S.t.
E=A U B,J=g on A,J:;t:g on B, m (B)=O.
 116 MEASURABLE FUNCTIONS

.·3. De6aition. Characteristic function. [Meerut 1990]

Let A be a subset of a set E. The characteristic function KA
of A is defined as :
I if x E A
KA(X)= { 0 if x E E. . A
The function Kix) is measurable iff A is measurable. Hence
existence of a non-measurable set implies the existence of a non-
measurable function. The function K,.(x) is also called indicator
flUlCtJo. of A .
..... Deftnition Simple Function. [Kanpur Statistics 1977, 75]
A function is said to be simple function jf the range of the
function is finite. For example, characterist;c function.
, A real valued function ~ is called simple if it is measurable
and assumes only a finite number of values.
If ~ is simple and has finite number of values Ill> 112, ..... «ta,
then", is expressible a'l
~(x)=
.
E _, KAi
'=1
where A.={x : ~(X)="I}'
The sum, product alld difference of two simple functions are
simple.
8'5. Limit Superior aad Limit Inferior
Let < Ar : r EN> be a sequence of sets. The supremum
00
of this sequence, denoted by set U A, and the limit superior of
r~l

__ __ ,,",,00

this sequence, denoted by Lim A" is defined as lim A.= n U A...

n=1 tn._II

00
The iufimum of the sequence, is defined as the set n A.. and
..=1

lilllit ioferior of the sequence is denoted by Lim A•• and is defined

00 00
as Lim A.. = unA...
a-I _::::IB

If Lim A.. = Lim A". then we sar that the limit of the sequeace
exists and the common value is dC!noted by lim A...
Thus Lim A.=Lim A,,=Lim All if Lim A .. exists.
 MEASURABLE FUNCTIONS 117

Monotone Sequence. A sequence of sets <A.> is said to be

non-decreasing (expanding) if A,. C A,,+! V n and is denoted by
< A. t >.
A sequence of sets <A.> is said to be non-increasing (con-
tracting) if A.. +! C A,. ¥ n and is denoted by <A. ~ >. In either
case the sequence is said to be a monotonic sequence.
Theorem. Every monotone sequence of sets is convergent.
Proof. Let <A.. > be a non-decreasing sequence, then

By def. lim inf A.. = U

n=1
[n Am]= UA,.
Ift=.. 11=1
... (1)

lim sup A.. = n01 [In~" A. J ... (2)

00
But we see that for any non-decreasing sequence, U Aft is in-
,,=11
dependent of k, so we may take k= I. lienee, by (2),

Lim sup A,,= n [u J= U

R=l m-.l
A",
m~l
A".= U A...
n-l

00
Hence, lim sup An= U A.. =lim inf A..
.. =1

00
or lim sup A.. =lim inf A .. = lim A .. = U An.
n=1

Hence monotonic non-decreasin~ sequence is convergent.

Similarly we can also prove that monotonic non-increasing sequ-
ence is convergent.
Theorem 1. Let f be a measurable function defined over a
00
measurable set E,: ¥ kEN and E = U E/...
1<=1

Then f is measurable Ol'er E

Proof. Let a be any real number. Let (be a mea~urable
function over a measurable set Ek V kEN so that
Ek (f > a) is measurab;~ ¥ kEN.
118 MEASURABLE FUNCTIONS

Let

To prove that f is measurable over E, we have to show that

E (f > a) is measurable.
Being an enumerable union of measurable sets, E is measurable.

E (f> a)=( UE
k=l
1: ) (f:> a)= U [E
k=l
t (f > a)]

=An enumerable union of measurable sets

=A measurable set.
E (f > a) is a measurable set.
Theorem 2. Iff is a constant function over a measurable set E,
then f is measurable over E. 0r
If f(x)=c (constant) for evpry vaiue of x E E, then f is measu-
rable over E, E being a measurable set.
ProoC. Letfbe a constant function defined over a measurable
set E so thatf(x)=c ¥ x E E. Let a be any real number.
To prove thatfis measuTable over E, we have to show that
E (f > a) is a measurable set. Evidently
E (f > a)=JE .if c > a
l4> If c ~ a
In both cases, E (f > a) is measurable set.
For E and tP both are mea<;urable sets.
Theorem 3. Iff is a measurable function over a measurable set
E, then f is also [Link] over any set ACE, A being a measu-
rable set.
ProoC. Let a be any real number. Let f be a measurable
function defined. over a measurble set E so that E (f > a) is a
measurable set. Let ACE be any arbitrary measurable set.
To prove thatfis measurable over A, it is enough to show
that A (f > a) is a measurable set.
A (f > a)=[ E (f E > a)] n A
= Intersection of two measurable sets
=Measurable set.
:. A (f > a) is a measurable set. .
Lsbesgue measurable Cunctions.
Theorem 4. To gi.'e four equivalent definition for the Lebesgue
measure of a real valued function. Prove their equivalence.
(Kanpur 1968; JSanara~ 69, 66]
MEASURABLE FUNCTIONS 119

Proof. Let f be an extended real valued function defined over

a measurable set E. Thenf is Lebesgue measurable (or simply
measurable) function jff anyone of the following is measurable:
(i) E (f > 0), (ii) E(f~o), (iii) E (f < 0),
(iv) E (f ~ a),
where 0 is any real number and E (f > a) stands for the set
(x E E :f(x) > o}.
Similarly we define the other three sets.
To prove that these four [Link] are equivalent.
(i) ~ (ii).

,For E(f~ a)= ~1 [( Ef> a- ~)]

an arbitrary intersection of [Link] sets is measurable.
(ii) ~ (i).

For £ (I > 0)= ~1 E( f ~ a+ ~

[ )]
and an arbitrary union of measurable sets is measurable.
(ii) <> (iii).
For E (f < a)=[E (I ~ a)J'
and a set is measurable iff its complement is. Here complement
of every set is taken w.r.t. the set E.
(iii) ~ (iv).

For £(/~ a)= n£(f< ~+~)

n=1 n
and an arbi~rary intersection of measurable sets is measurable.
(iv) ~ (iii).

For E(f < a)= ~1 [E ( f ~ a - ! )]

and an arbitrary uni"n of measurable sets is measurable.
This completes the proof.
00 ,
Remark. 1.1. {XEE :f(x);;tGt}= U {x E E :f(x) > «-(lin)}
n_1

2. E(/<<<)=£-[£(/><<)]
120 MEASUR~BLE PUNC nONS

00
3. E(f~ IX)= n {x E E:f(x) < cx+(ljn)}
,,=1

4. E(f> «)=E-[E(f< «)]

00
5. {x E £ :/(x)=oo} n {x E E: f(x) > n}
11=1

00
6. {x E E:f(x)=-oo}= n {x E E:f(x)
, _1
< -n}.

Remark 2. An extended real valued function f is said to be

measurable (Lebesgue measurable) if its domain is measurable
and if it satisfies one of the first four statements.
Remark 3. / is said to be Lebesgue measurable over E if
(v) E (/=a) is measurable for every extended real number a.
To prove that the first four statements imply the statement (v).
(ii) and (iv). (v) ¥ a E R.

For £(/=a)=[E(f< a)] n [(Ef;;;' a)]

and a finite intersection of measurable sets is measurable.
(ii) ~ (v) for a=oo.
00
For E(/=oo)= n [£(/~ n)].
,,=1

Similarly (iv) => (v) for a= - 00.

Finally, (ii) and (iv) ~ (v) for every extended real number a.
This proves the r~quired result.
Theorem 5 Let/ be a measurable function defined orer a
measurable set E. Let c be any real number. Show that
c/,/+c, I /I,P, }- (if/ vanishes nowhere Oil E) are measur-
able functions. [Meerut 1990, 88; Kanpur 85, 71; Gujrat 70J
Proof. Letf be a measurable function defined over a measur-
able set E. Let c be any real constant.
/ is measurable over E ~ E (f >
< a) both are
a) and E (f
measurable set 'tf a E: R.
(i) To prove that cf is measurable over E. [Mi:erut 1972J
If c=O, then cf is constant and hence measurable.
Consider the case in which c:;t:O.
 MEASURABLE FUNC I IONS 121

£( I > :) if e > 0
E (el> a)= {
E( I < :) if e < O.
Both the sets on R.H.S. are measurable.
Hence E (ef > a) is measurable and so is el.
Therefore el is measurable VeE R.
(ii) To prove that I f I is measurable. [Indore 78J
E ( III < a)=[E (I < a) U [E (j > -a)] if a ~ O.
Also a finite union of two measurable sets is a measurable
set. E is given to be measurable. Therefore the measurability
of E ( II I < a) follows.
Consequently III is measurable over E. [Kaopur 83J
For I x I < a => -a < x < a.
(iii) To prove that c+1 is a measurable function.
E(c+l> a)=E(/> a--e)
=E (I> b), on setting b=a-c:,
= measu rable set.
Hence the result (iii) follows.
(iv) To prove that 12 is a measurable function.
E (f2 < a)=E ( III < ya) if a ~ 0
But E( If I < va)=[£(/< va)] u [E(f> -va)J.
: . .E(f2 < a)=[E(/< v'a)J U [EU> -ya)] if a ~ O.
Hence the measurability o~~~onseqUentlY
(2 is measurable over F. . -- _·;i ,'- '0 .
(v) Let / vanish no\\ here on E so that f (x):;i:O V x E E.
H ence fIeXists.
.

E ( / > 0) ifa=O
(
1 )
ffE(/>O)]n[E(/<l/a)
E / > a --= [[E(f< O)J n lE(/< I/a]) U
if a;;;.. 0

[E(/>O)] ifa<O.
Also finite union and intersection of measurable sets are
measurable. Hence E ( ; > a ) is measurable and so is ; .
Remark. By the theorem, just proved
.r is measurable => cl is measurable function VeE R
:;. ~I is measurable function, on taking c= - I.
122 MEA SURA BLE FUNCTIONS

Theorem 6. Let f and g be measurable functions defined over

a measurable set E. Show that f+g, f - g, fg ore measurable fune.
tions over E. [Meerut 1986 ; Banaras 69, 67 ; Kanpur 79, 77, 71]
Proof. Let f and g be measurable functions defined over a
measurable set E.
To prove that f+g, f-g, fg are measurable over E.
Firstly we shall establish the following lemma:
Lemma. If f and g are measurable functions defined over a
measurable set E, then the set E (f > g) is measurable.
[Kanpur 76, 7S, 72]
The proof of the lemma starts.
Let f> g, then by the corollary to the axiom of Archimedes
there exists a rational number r S.t. f> r > g. Therefore
E (f > g)= U [[E (f > r)] n [E (g < r)]]
reQ
=An enumerable union of measurable sets
=measurable set
Hence the lemma.
For (i) Q is an enumerable set
(ii) fand g are measurable => E (f > r) and E (g <'r)
are measurable set.
(iii) Intersection of t\\O [Link] sets is a measurable set.
Now the proof of the main theorem starts,
g is measurable :> cg is measurable >.f c E R
=- a +cg is measurable >.f a, c E R
=> a-g is measurable on taking c= -1.
I. To prove that f+g is measurable over E. [Gujarat'70]
Let a be any real number.
E(f+g > a)=E( f> a-g)
= measurable set, by the lemma just proved.
For f and a -g both are measurable functions over E.
Thus E (f+g > a) is a measurable set and hence f+g is a
measurable function.
II. To prove that f-g is measurable.
f and g are measurable over E => f and - g are measurable
over E
q f-g is measurable over E .
•: f+ (-g) =f-g.
MEASURABLE FUNCTIONS 123

III. To prove that fg is measurable over E.

f and g are measurable functions over E
=> f +g, f - g are measurable functions over E
=> (f+g)2, (f-g)! are measurable functions over E
. ~ (ftg)2_(f _g)2 is a measurable function over E

=> ! [(f - g)2-(f-g)2] is a measurable function over E

=> fg is a measurable function over E.

Alternate method. To prove that f+g is measurable.
Since for any real number «, 3 a rational number r s.t.
f(x) < r < t1.-g (x)
whenever f(x)+g (x) < t1.
{x E E: f(x)+g (x) < «}
=U ,. [(x E E: f(x) < r} n {x E E: g (xi < «-r}l.
This union is taken over a set of rational numbers. It follows
that f+g is measurable. Similarly we can prove that f - g is also
measurable.
Theorem 7. If f and g are measurable functions defined over
a measurable set E and if g vanishes nowhere on the set E, tlren the

quotient function .~ is measurable Ol'er E. [Kanpur 1979, 77J

Proof. Let f and g be measurable functions defined over a
measurable set E. Also let g vanish nowhere on E so that
g (x):;eO '" x E E. Then ~
g
exists.

I. To prove that fg is measurable. lGujarat 1970J

f and g are measurable functions
=> f+g. f--g both are measurable functions
~ (f +g)2 and (f - g)2 both are measurable functions
=> (f+g)L·(f g)2 is a measurable function
=> 1 [(f+g)Z-(f-g)2] is a measurable function
=- fg is a measurable function.
II. To prove that..! is measurable over E.
g
Let a be any real number.
 124 MEASURABLE FUNCTIONS

E(g>O) ifa=O
(E (g > 0)] () (E (g < I(a)] if a>O
[E (g > O)]U[(E (g < O)]()[E Ig < I/o)]]
ifa<O.
Also finite union and intersection of [Link] sets are
mea~urable: Hence E ( ; > a ) is measurable in every case.

Therefore) is measurable over E.

g

III. To prove that f is measurable over 1:;.

g

Evidently Lg =f. J.-.

g
, ~g is measurable function, by II

= product of two measurable functions.

=measurahle function, by J.
Hence the result (Ill).
Theorem 8. Pr0l1e that the space of measurable functions is
closed under the usual operations of Arithmetic. [Bbagalpur 1968J
Proof. The fundamental operations of Arithmetic are:
addition. subtraction, multiplication, division.
Let f, g be arbitrary members of the space of measurable
functions.

If we show thatf+g, f-g, fg and f (if g vanishes nowhere

g
on E) are mea~urahle functions, the result wili follow. Of course
the functions f and g are measurable over a measurable set E.
Here write the proof of Theorems 6 and 7.
Ex. 1. !If and g are real valued measurable functions on a
measurable space X, prr l'e that the f U g, f () g are measurable
functions. [Gujarat 1970J
Solution. Let c he any real constant. Letl and g be real
[Link] measurable functions defined on a measurable set X so that
X ( f > c) and X (g > c) are measurable sets.
(i) To prove that f U g is measurable.
X (f U g> c)=[X (f > c)] U [X (g > c)]
=finite union of measurable sets
=measurable set.
MEASURARLE FUNenONS 125

This proves the result (i).

(ii) To show that f n g is measurable
X ( rn g > c)=[X (f > c)] () [X (g > e))
=finite intersection of mea!>urable sets
=measurable set.
Thi~ => the result (ii).
Ex 1. Show that the set of all measurable functions form a
vector space over R. [KaDpur 1914]
Solutio.. Let E be a m~asurable set. Let V be the set of all
measuruble functions defined over E. Then
Y={[Link] E+R is a measurable function}.
To prove that Y (R) is a vector space.
Firstly we shall prove the following lemmas.
[Link] 1. Iff is measurable, then af is measurable, a being
any real number. Also let c be any real number.
if a > 0

if a < 0
Both the sets on the R.H.S. are measurable. Hence E (af> c)
is measurable so that af is measurable.
Lemma 2. Iff and g are measurable, then f+g is also
measurable
E(f>g)= U [E (! > r > g)], Q=set or rational numbers
reQ
U [E (f > r)] n [E (g < r))
rEQ
=An enumerable union of measurable sets
=measurable sd.
g is me:\surable => ag is measurable => -g is measurable
=> c - g is measurable.
Now E {f+g > a)=E ( r < a -g)=measutable set
For E ( f> g) is measurable.
Now we come to the main problem.
Let!. g, hEY be arbitrary. Then a, b, c E R,
R being the set of real numbers.
(i) (Y, +) is abelian group
l. !+g E Y, by Lemma 2.
2. 3 0 E V s.t.f+O=O+f=/.
126 MEASUI<ABLE FUNCTIONS

Here 0 is zero function so that 0 (X)=O >,f x.

It means that 0 is a special case of a constant function and
so it is measurable. Also
(/+0) (x)-f(x) + 0 (x) =1 (x) +0=1 (x), this. 1+0-1.
3. -f E V s.t. -1+1=0, by Lemma 1.
4. f+g=g+/as (R, +) is an abelian group.
5. 1+ (g+h)=(f+g)+h as (R, +) is an abelian group.
(ii) al E V, by Lemma 1.
6. (a+b)/=a/+b f
7. a (/¢-g)=a/+ag.
8. J.I=f.
(6), (7). (8) follow from the fact that (R, +, .) is a field.
From what has been shown it follows that V (R) is a vector space.
Theorem 9. Let < /" > be a sequence 01 measurable /unc-
tions defined over a measurable set E. Then sup {f1'/2'/3 ... } and
inl { J;. h. la ....... } are measurable over E. [Kanpur 1986, 76]
Proof. Let < In > be a sequence of measurable functions
defined over a measurable set E so that
E ( I .. > a) is a measurable set ¥ n, a being any real number.
Write g(x)=sup {/,,(x) : n=l, 2, 3, ...... }
h (x) =inf f/n(x): n=l, 2, 3, ...... }
hex) can be defined in terms of supremum as follows:
h(x)=-sup {-f,,(x): n=l, 2, 3, ...... }.
To prove that g and h are measurable over E, it is enough
to show that g is measurable over E.
c
E (g > a)= U [E (fn
,.z:;l
> a»)

= ""U {x E E :f,. (x) > a}

.=1
= An enumerable union of measurable sets
=measurable set
Theorem 10. Let < III > be a sequence of [Link] func-
tions defined over a measurable set E. Show that
sup {J;'/2""'f,.}, in/{fh!2"'" !.},
Lim!", Lim!n, are measurable Gver E. Henc~ show that Lim /t. is
measurablover E if Lim!n exists.
[Kanpur 1986, 76, 72, 70; Banaras 197., 68, 65]
 MEASURABLE FUNCTIONS 127

Proof. Let:< In > be a sequence of measurable functions

defined over a measurable set E.
Step I. To prove that sup {Ir : 1 EO; r ~ n} and
inf { J.. : ] :s:;;; r < n} are measurable over E.
Define
M (x)=sup {/r (x) : 1 EO; r < n}

= U" IT (x)
r-l

and m (x)=inf {/r (x) : 1 r n} < <

m (x) can also be defined in terms of supremum as follows:
m (x)=-sup {-Ir (x): 1 ~ r n}. <
If we show that M (x) is measurable over E, the re&ult will
follow.
For proving this, it is enough to prove that E (M > a) is a
measurable set, a being any real number.
In is a measurable function over E => E (f,. > a) is a measur-
able set V n.
Evidently

E[M> a)= U" E[(fr > a)]

=A finite union of measurable sets

=A measurable set.
This proves that E (M > a) is a measurable set.
Step II. To prove that (Tm In and Lim fn are measurable
functions over the set E. rBanaras 1965]
D.!fine
Mk (x)=sup { f" (x)},
n;Pk
111,. (X)= inf {I" (x)}.
n;pk
Then

Lim fn (x)= inf {Mk (x)}

k~l
Lim I .. (x)= sup {mk (x)}.
-- k ~ 1
By case (i), M,. (x) and mk(x) both are measurable functions
~ [Link] over E V kEN and ag lin, by case (i),
 128 ME,\SUR,\BLE FUNCTIONS

inf {Mk(X)}, SUp {mk(x)}

k;;;tl k:?1

are measurable over E, i.e. lim/.. , Lim f. both ate measurable

over E.
Step. III. Let Lim / .. exist:
[Meerut 1971; Kolhapur 70]
To show that Lim /. is measurable over E.
By case (ii), Llm"j., Lim/.. both are measurable over E.

By hypothesis,
Limf.=Lim /.=Limf,•.

From what has been done, it follows that Lim /. is measur·

able over E.
Theorem 11. A continuous/unction defined over a. measurable
set E is measurable. [Kanpur 76; Meerut 71.; Kolbapur 69]
Or, Show that a continuous function defined in a closed interval is
measurable. [Kanpar 1987J
Proof. Let / be a continuous function defined over a mea·
surable set E.
To prove that / is a measurable function over E.
Let a· be any real number.
We claim E (f;;;' a) is a closed set.
Let A=E (/ ~ a). . .. (1)
To prove that A is closed, it is enough to prove that
D (A) C A. . .. (2)
Let Xo E D(A) be arbitrary, then Xo is a limit point of A so
that 3 a sequence <x,.> whose elements x" E A
s.t. lim
n-+oo x,.=xo·
/ is continuous at xo, X,.-X o ~ /(x") ~ /(x o). . .. (3)
This follows from Heine's definition of continuity
By (t), x" E A => ft x,.) ~ a => lim j (X,.) ~ a
n-+oo
=> /(x.) ~ a, according to (3),
=> Xd E A, by (I).
 MIlASURABLE FUNCTIONS 129

Finally, any Xo E D(A) ~ Xo E A.

This ~ D(A) c A
=> A is closed
~ A is measurabl.
=> E( I ~ a) is measurable
=> I is a measurable function over the set E.
Theorem II (a). Let I be a function delined on a measurable
set E. Then f is measurable iff lor any open set G C R,
1-1 (G) is a measurable set.
Proof. Suppose a function f is defined on a measurable set
E. Let G C R be an open set, where R=set of real numbers.
I. Suppose I is measurable on E.
Aim. /-1 (G) is a measurable set.
Since any open subset G of R is expressible as a countable
union of disjoint open intervals. Hence G is expressible as
00
G= U In, where 1.. = (a,,, bit)
n=1

/-1 (G)= U {x E E: I(x)

. E , ft }

But/(x) E I,,=(a n , b.. ) ~ /(x)E (a .. , b.)

~ an </(x) < b"
:. 1-1 (G)= U (E (I > aft) n E ( / < bn )]
,. ... (1)
I is measurable => E ( / < an) and E (I < b..)
both are measurable sets.
Also intersection of measurable sets is measurable. Now
R.H.S. of (I) is measurable.
Hence 1-1 (G) is measurable.
II. Conversely suppose that/-1 (G) is measurable.
Aim. f is measurable.
Take G=(a, 00), where a > O.
Then I-I (G)=(x E E : I(x) E (a, oo)}
~{x E E: a </(x) < oo}
=E«(>a)
or I-I (G)=E (/ > a)
Also 1-1 (G) is measurable.
Hence E (f > a) is measurable.
Consequently lis measurable on E.
 128 MEASURABLE FUNCTIONS

are measurable over E, i.e. Limf.. , Lim f,. both ate measurable
over E.
Step. III. Let Lim f .. exist:
[Meerut 1971; Kolhapur 70]
To show that Lim f. is measurable over E.
By case (ii), Llin-f., Limf.. both are measurable over E.

By hypothesis,
Limf.=Lim f.=Limf, •.

From what has been done, it follows that Lim f. is measur-

able over E.
Theorem 11. A continuous function defined over a measurable
set E is measurable. [Kanpur 76; Meerut 72; Kolbapur 69]
Or, Show that a continuous function defined in a closed interval is
measurable. [Kanpar 1987j
Proof. Let f be a continuous function defined over a mea-
surable set E.
To prove that f is a measurable function over E.
Let a be any real number.
We claim E (f ~ a) is a closed set.
Let A=E (f ~ a). ...(1)
To prove that A is closed, it is enough to prove that
D (A) C A. ...(2)
Let Xo E D(A) be arbitrary, then Xo is a limit point of A so
that 3 a sequence <x.. > .whose elements x .. E A
s.t. Jim
n-+IX) x .. =xo•
f is continuous at x o, x .. -Xo ~ f(x,,) ~ /(x o)' ... (3)
This follows from Heine's definition of continuity
By (1), x .. e A => Ax..) ~ a => lim j (x .. ) ~a
n-+oo
=> /(x.) ~ a, according to (3),
=> x" E A, by (I).
 MEASURABLE FUNCTIONS 129

Finally, any E D(A) => Xo E A.

Xo
This => D(A) c A
=> A is closed
=> A is measurabl.
=> E( f ~ a) is measurable
=> / is a measurable function over the set E.
Theorem II (a). Let f be a function defined on a measurable
set E. Then f is measurable iff for any open set G C R,
f- 1 (G) is a measurable set.
Proof. Suppose a function f is defined on a measurable set
E. Let G C R be an open set, where R=set of real numbers.
I. Suppose / is measurable on E.
Aim. /-1 (G) is a measurable set.
Since any open subset G of R is expressible as a countable
union of disjoint open intervals. Hence G is expressible as
00
G= U I", where 1,,= (a,,, b,.)
,.=1

..
f- 1 (G)= U {x E E: f(x) E ,ft }

Butf(x) E I,.=(a", h,,) => f(x)E (a,., btl)

=> an < lex) < b..
:. f- 1 (G)= U [E (/ > a.. ) n E (f < bn )]
" ... (1)
f is measurable => E ( f < an) and E (f < b,.)
both are measurable sets.
Also intersection of measurable sets is measurable. Now
R.H.S. of (I) is measurable.
Hence f- I (G) is measurable.
II. Conversely suppose thatf-I (G) is measurable.
Aim. f is measurable.
Take G=(a, (0), where a > O.
Then /-1 (G)={x E E :I(x) E (a, oo)}
--={x E E: a < f{x) < oo}
=£«(>a)
or [-1 (G)=E (/ > a)
Also f- 1 (G) is measurable.
Hence E ( / > a) is measurable.
Consequently / is measurable on E.
130 MIlO\SURABLIL FUNCTIONS

Remark. The converse of Theorem 11 is not ture. That is

to say, a measurable function [Link] not be continuous.
Example. Consider a function f: R ... {O, I} defined by
{ lifO ~ x <; :
f(x)= 0 otherwise.
Evidently I is measurable but not continuous. The point
x=O is a point of discontinuity. Here R=set of reals.
Theorem 12. If < f. > is a sequence of measurable functions
then show that Limf. is measurable. [Meerut 1986]
0, To prove thqt the limit 01 a convergent sequenr.e 01 meas.-
urable lunctions is measurable. [K8n.,ur M. Se. P. 1986)
Proof. Let < Ita' be a sequence of measurable functions.
To prove that Lim f. is measurable.
t. < I. > is a monotonic sequence and hence it is either
monotonic increa'!ling or monotonic decreasing.
If < /.1 > is a monotonic increasing sequence, then
Lim f .. = sup {f.}.
n ~ 1
If < I. > is a monotonic decreasing sequence, then
·Limffl= inf {lit}.
n~1
Hence if we prove that
sup {/1,f•• ...... } and inf{/l,h, ... · .. }
are measurable. then the problem will be p-roved.
Let g(x)=sup {fIJ., ...... } ... inf {h. fs •...... }.
00
Then £(g > a)= U [£(f. > a)]=An enumerable union of
"=1 measurable sets.
Hence E(g > a) is mea'iurable and so g is measurable.
hex) is expressible as
h(x)=--sup {-I.(x} : n=l, 2, ...... }.
This .. h(x) is also mea'!lurable as g is measurable.
This completes the proof.
Ex. Let < I. > be a sequ:mce 01 real valued measurable
functions. If lim
n-+-IX)
1.=1 eXists, then show that I is also measurable.
[Kanpur 1975]
Solution. Here write the proof of above Theorem 12.
 MEASURABLF. FUNCTIONS 131

Tbeor,em 13. III is a meafurable function, defined on a measurable

set E and if g and f are equivalent functions, then g is a measurable
function on E.
An Alternate Statement. Iff is a measurable function and if
f=g almost every where, then g i~ measurable.
Proof. Let E be a measurab'.e set and let a be any real num-
ber. Also let fbe a measurable function on E so that E (f > a)
is a measuPabte set. Further suppose that f and g are equivalent
functions on E so that
f=g a.e.· on E ... (1)
To prove that g is measurable over E, we have to show that
E (g > a) is a measurable set.
(I) c> 3 ACE S.t. m (A)=O,
f(x)=g (x) V- x E E-A=B (say).
~ f:#g on A. f=g on B, m (A)=O,
A n B=0, E=A U B.
m (A) =0 ~ A is measurable.
E and A are measurable sets => B=E -A is measurable.
f is measurable on E, BeE is measurable
~ f is measurable over B. Refer theorem 3, page 118.
=> A (f > a) is a measurable set
=> B (g > a) is a measurable set.
For f=g on B c> B (f > a)=B (g > a)
A (g > a)={x E A: g{x) > a} C A.
Also m(A)=O.
This c> A (g > a) is measurable.
For every subset of a set of measure zero is measurable.
E (g > a)=[A (g > a)] U [B (g > a)]. For E=AuB.
= union of two measurable sets.
=measurable set.
:. E (g > a) is a measurable set.
Ex. Answer the following questions in < Yes' or 'No'. Is a
function equal to a measurable function almost everywhere itself
measurable? [Kaapur 1974]
Solution. Yes.
(Refer Theorem 13).
Theorem 14. To show thar the characteristic function of a set
E is measurable iff E is a measurable set.
[1\1eerut 1990, 88; Kolbapur 69; Banaras 69; Gujarat 10]
132 MEASURABLE FUNCTIONS

Or The set E and its characteristic function both are measur-

able or both non-measurable.
Proof. Let E be a subset of a measurable set A. The charac-
teristic function KII of E is defined as
{I if x E E
KE(x)= 0 if· x E A-E
We shall show that
(i) if KE is measurable, then E is measurable.
(ii) if E is measurable, then KE ~s measurable.
(i) Let KE be measurable on A so that A (K F. > 0) is a
measurable set. .
But E=A (KI> > 0).
Hence E is measurable.
(ii) Let E be measurable and a be any real number. Then
r
0 if a ~ 1
A (KE > a)= I E If 0 ~ a < I
lAir a < 0
Every set on the R.H.S. is measurable.
Hence A(KE>a) is measurable and so KE is measurable on A
Theorem IS. Iff and g be measurable real valued functions
defined on X and F tf(x), g(x)]=h(x), x E X
be real and continuous function on the Euclidean plane RI, show that
h is measurable. LKanpilr [Link]. 1984, 81; Kanpur 69J
Proof. Let a be any real number. Let
, G={(u, v) : F (u, 11) > a}.
Then G is an open subset of R2 !oo that G is expressible as
00
G= U In,
R~l

where <./,,> is a sequence of open interval,> s t.

/,,=:(11, }.) : an < II < bn , en < I' <: dll}.
Evidently {x EX: f(x) E (a", b,.)}
={x E X: an <f(x) < b,.}
==(x E X :f(x) > a.} n {x E X :f~x) < b,,}
=An intersection or two measurable sets.
[For f is measurable]
= measurable set.
 MEASURABLE FUNCTIONS

:. {X EX: I (X) E (a". bA)} is a measurable set.

Similarly measurability of g over X implies that
{x EX: g(x) E (cn , d.)} is a measurable set.
Then it follows that
{x EX: F(f(x). g(x» E T.\
{x E X :f(x) E (a", b..)}n{x EX: g{x) E (c.. , d,.)}
is measurable set. Now
=(x EX: hex) > a}={x EX: F (/(x), g(x» E G}

= -
U {x € X: F (/(x), g(x» E f .. }
11-1

=measurable set.
Consequently h is measurable on X.
Ex. 3. A/unction f is measurable iff the set {x :/(x) < r} is
measurable /~r every rational number r. [Kolbapur 1970J
Solutiou. Suppose a ,function / is defined over a measurable
set E. Let r be an arbitrary rational number.
(i) Suppose / is measurable over £ so that E (I < c) is a
measurable set, c being any real number.
To prove that {x : f(x) < r} is measurable.
E (I < c) is measurable y. c E R
::> in particular, E (f < c) is measurable ¥ CEQ. For QCR.
~ £ (f < r) is measurable for r E Q.
~ {x: /(x) < r)={x E E: f(x) < r} is measurable.
~ {x :/(x) < r} is measurable.
(ii) Conversely suppose that {x : f(x) < r} is measurable.
To prove that f is measurable.
Observe that
£(/< c)={x E E:f(x) < c}
= U {x E E: f{x} < r < c}
rE.Q
== U {x E E ;f(x) < r, r E Q}
r<c
=cnumerable union of measurable sets.
, [For Q is enumerable]
=measurablc set.
:. E ( / < c) is measurable and so / is measurable over E.
Ex. 4. Iff is a continuous function and g is a measurable /unc-
tion, {hen prove that the compositefunctionf 0 g is measurable.
134 MEASURABbE FUNCTIONS

Solution. Let G be an open set and g be a measurable

function. Let A={x : g(x) E G}.
To prove A is a measurable set. . .. (1)
G is an open set ::;. G is expressible in the form

..
G=U·/.. , where 1.. =(0", b.. ).

Hence A={x : g(x) E G}=U {x: g(x) E I .. }

..
.4= U [{x: g(x) E (a... b..)}= U {x: < <
ot
.. .. 0 .. g(x) b..}

= U [{x: g(x) > a,,}n{x : g(x) <

.. bIt}]

=enumerable union of measurable sets

=measurable set. Hence the result (1).
{x :f(g(x» < a}={x : g(x) E G} ... (2)
where G={y : fey) < a}
f is continuous => G={y : fey) < a} is open
=> {x : g(x) E G} is measurable, by (1)
=> fog is measurable, by (2).
Theorem 16. (E. Botel). Suppose a measurable function f
defined on a closed interval [a, b], is finite almost everywhere on
the sume inrerval. For all numbers. a • • > 0, there exists a conti-
nuousfunction", [a, b] such that
m [(E If-; I ~ a)] < E. [Paojab 1969)
Proof. Letfbe a measurable function defined on [a, b] S.t.
f is finite almost everywhere ~ [a, b].
Case I. Letfbe bounded.
Then 3 a number k > 0 s.t. If(x} I < k ... (1) ¥ X E [a. b].
Let a, • > 0 be arbitrary numbers. Fixing C1, 6 by ~hoosir,~ a
positive integer m so large that
(kIm) <: a.
Write
where r=l-m. 2-m, ...... m-I
m -1 . )
aDd E",=E (m k ~f < k •
Evidently E,nEI J~ for i¢j.
. where i,j=l-m, 2-m, ....... m.
MEASURABLE FUNCTIONS llS

[a, b]=
.
1: E,
,_l-m

Select a closed interval F, C E, s.t.

m (E,) < m (Er )+(./2m).
Take

Then m ([0, b])-m (F) < •. ... (2)

Depne a function ,p on F by writing
rk
,p (x)=-
m
¥ x E F" r=l-m, 2-m, ...... m. J

Then ,p is constant on each closed set Fr.

Also F,()Fr'=0 for r¢r'.
Conseqvently ,p is continuous on Fr.
I ~ (x) !=I~ I=~k ~ k.
:. I ~ (x) I ~ k.
Finally, :f x E F r , then
I/(X)-~ (x) \=1 k-:'(=~I m-rl ~ ~ < d.
. Ilex) -~ (x) I < 0 "" x E F,.
Now we shall state a lemma without proof.
Lemma. Let F be a closed set contained in the closed
interval [a, b]. If the function ,p (x) is defined and continuous
on the set F, then it is possible to find out a function r/J(x} on
[a, b) s.t.
(i) 1\1 (x) is continuous
(iiI 1\1 (x)=,p (x) V x E F
(iii) max. I 1\1 (x) !=max. 1,p (x) I.
By the lemma, just stated, it is possible to find a fUnction
I\I(x) defined on [a, b) having all the properties stated in the
lemma. Moreover
E (1/-'" ! ~ a) c [a, b]--F.
Hence m [E( 1/-1\1 I;;;;. 0)] ~ m ([a, bJ-F) < t, by (2)
i.e. m [E ('/_.1\1) ~ 0] < E.
Thus we have proved the theorem for a bounded functlon/.
Case II. Let/be unbounded.
Let us suppose that /(x) is unbounded. Then we can find
out a bounded function g such that
m.[(E/#g)] < i·
 MEASURABLE FUNCTIONS
136

Applying Case I, just proved, to the bounded function g, we

are able to find a continuous function cjI s.t.
m [ECI g-cjll ;;;;. 11)] < f'
From this we can deduce that
E ( If-cjll ;? 11) C E (f:Fg)+E ( I g-cjll ~ 0).
This =>
m [E ( If-cjll ;;;;. 0)] ~ m [E (f:Fg»)
+m [E( I g-.p I~ 0)]
E E
<2"+2"=.'
m[E(If-cjll ;;;;'0)] < •.
The function .p satisfie'i all the requirements.
Hence the theorem is also true in case of unbounded function.
8'6. Borel measurability fllnctions.
A functionf is said to be Borel measurable if the set
{x :/(x) < a} ¥ a E R
is a Borel set. (I) Ev~ry Borel measurable function is (Lebesgue)
measurable function. F,o'r every Borel measurable set is a
Lebesgue measurable set,.
(2) If I is Borel measurable function and A is Bord set,
then 1-1 (A) is Borel measurable.
(3) The product of two Borel measurable functions is a
Borel measurable function.
8'7. Little Wood's three principles.
In connection with the theory of functions of a real variable,
J.E. Little Wood says. "The extent of knowledge required is
nothing like so great as is sometimes supposed. There are three
principles, roughly expressible as follows:
1. Every (measurable) set is nearly a finite union of
intervals; 2. every (measurable) function is nearly continuous;
3. every convergent sequence of (measurable) functions is nearly
uniformly continuous."
Problem. Show that the function I defilled on R by
(x+5, x < -1
I(x)= -< 2 , -1 ~ x < 0
Lxi , x ~ 0
is a measurable function.
Soilltion. Let a be any real number. Then a < 0 and
R (I ~ a) .. x+5 ~ a => 'x ~ a-5.
 MEASUREBLE FUNCTIONS ]37
:. .R. (I ~ a)=[ -00, a-5].
I n view of this, we get
R (/~ a)=j(-OO, a-5]- 5U] ·{OJ
( - 00,
if a < 0
if a=O
(-00, a--5] U [O,ya] if 0 < a < 2
(-oc, a-5] U [-1, ya] !f 2 =E;; a ~ 4
(-00, Va], I[Link]"" a
Since each set on R.H.S. is measurable.
:. R (I ~ a) is measurable for every real value of a. Hence
I is measurable, .
EXERCISES
1. IfI is a measurable function, then II I is also measurable.
[Kanpur 83J
2. Let I and g be measurable real valued functions defined on
X and F be real and continuous on R, if
h (x)=F (/(x), g (x»,
prove that" is measurable. [Kanpur [Link]. P. 1981]
3. If/ and g are measurable functions on [a, b], prove that
/ +g, / - g, /g are also measurable functions. If g (x).,i:O
when a ~ x ~ b, prove that f-.
g
is also measurable.

[Kanpur [Link]. P. 1979]

4. If I, g are measurable functions on [a, b], show that functions
l+g,lg are measurable. If g(x).;JCO (a ~ x ~ b), show that
Ilg is also measurable. lKanpur 1977J
5. Define mea~urable function. If / and g are finite and measu-
rable functions defined on a set E, prove that E (/ > g) is
measurable. Show Ihnt the functions/+g and/g are measu-
rable. lK'<lnpur J.976J
6. If <In> is a sequence of measurable functIOns on [a, bJ
such that the sequence </.. > is bounded for evt;ry x in
[a, bJ, then prove that the functions sup /r., inf /n, lim )",
" n

lim/.. are all measurable on [a, b]. [Kanpur 1976]

7. Define measurable functions. Show that the sum and product
of n (;;;;. 2) measurable functions are also measurable.
Let </.. > be a sequence of real valued measurable functions.
If -lim In .../ exists, then f is also measurable. lKanp.u 1975J
n-oo
 138 MEASURABLE FUNCTIONS

8. Define 'measurable functions'. Show that the set of aU

measurable functions form a vector space over R. Show also
that the limit of pointwise convergent sequence of measura-
ble functions is measurable. [Kanpur 1974]
9. Define' measurable function'. If I and g are finite and measu-
rable functions defined on a measurable set E, prove that the
set E (I> g) is measurable. Show further that the functions
I+g andfg are measurable. [Kanpur 1973J
10. Define measurable function. If <In'> is a ~equence of mea-
surable functions, prove that
lim lim
n_oo
I .. and - - I..
n-+oo
are measurable. Show also that a continuous function defined
in a closed interval in measurable. [Kanpur 1972]
11. Define a measurable function. Prove that if I(x) and g(x) are
finite and measurable, then the functions I (x)±g(x) and
/(x).g(x) are measurable. [Kanpur 1911]
12. (a) If g : R_R is continuous and I: R_R is measurable,
show that gol is measurable.
(b) Let <An> be a sequence of Lebesgue measurable
subsets of R such that

00 00
and B= n U A...
1"=1 fI=k

Show that B is measurable and that p.(B)=O.

[Banaras III 1970]
13. Let I be an extended real valued func~ion whose domain is
measurable. Prove the,equivalence of the following statements:
(i) For eaeh real number a, the set {x : f(x»a} is measurable.
(ii) For each real number a, the set {x : f(x)~a} is measurable.
Deduce that if/and g are measurable functions, then/+g
and Ig are also measurable.
Show that characteristic function of a selt E is measurable iff
E is a measurable set. [Banaras 69J
14. Construct a set which is not measurable in the sense of
Lebesgue.
 ME \SURABLE FUNCTIONS 139

Show that if < In > is a sequence of measurable functions,

then -Lim/n , Lim I,. are measurable. (Banaras 68J
15. Prove that if the functions I and g are finite measurable, then
l+g,1 -g andfg are mea'iurable functions. [Banaras 67J
16. If < fn > is a sequeGce of measurable functions show that
Lim sup In, Lim inff" are measurable. [Banaras 67]
17. (a) Define a measurable function on a set X, and show that
the characteristic function of a measuraple set is 3t.
measurable function.
(b) If I and g are real valued measurable functions on a
measurable space X, prove thatf+g,fg,f U g and! n g
are measurable. [Gujrat 70J
18. (a) Iff and g be measurable functions and c be a constant,
prove that functions cf and f+ g arc measurable.
(b) A real valued function wluch is continuous in an open
interval is measurable. lMeerut 1972]
19. When is a function said to be measurable?
Prove that the limit of a convergent sequence of measurable
functions is measurable. [Meerut 71]
20. Prove that the following condition is necessary and sufiiclent
that the function be measurable:
tire set {x: I(x) ~ r}
is measurable for every rational number r. [Kolbapur 69J
21. A functionf is measurable ifr the set
{x :f(x) < r}
is measurable for every rational number. [Kolhapur 70J
 9
The Lebesgue Integral of a
Function

Introduction. Lebesgue's definition of an integral is more

general in comparison to Riemann's. It enables us to integrate
those functions for which Riemann's method fails.
That is to say, if a function is integrable in the sense of Rie-
mann in an integral (0, b), then it is also ~ebesgue integrable
over (a, b) but not conversely, i.e. if a function is Lebesgue inte-
grable over an interval (a, b), then it is not necessarily Riemann
integrable over (a, b).
The phrase "Integrable in the sense of Lebesgue" is written
as "L-integrable."
From now onwards, we adopt the following conventions:

r f(x) dx=Lebesgue integral of/ex) over E.

,8

(R) J
E lex) dx=Riemann integral of/ex) over E.
9'0. To define the Lebesgue Integral of a function.
[Meerut 1987; Kanpur 86,72, 70J
First Method. Sets in two dimensional world are ref.:rred to
as plane sets. In two dimensional world, an interval is taken to
be a rectangle (in particular a space) and the area of the rect:lOgle
represent'> the measure of the rectangle.
Let lex) be an txtended real valued function defined over a
measurable set E.
We define
Q U: E)=-, {(x, y) : x E E, 0 ~ y ~ Ax)}.
Do (/, E)={(x, y) : x E I::, 0 ~ y < /lx)}.
The function/(x) is said to be Lebesgue \ Integrable over E if
{J (/, E) is measurable and the Lebesgue integral of lex) over E
is dd'ined as the plane measure of D (/, E), i.e.
l,. /(x) dx=m [.0 (/, E)].
THE LEBESGUE INTEGRAL OF A FU:-ICTION 141

Second Method.
Case (I). When f(x) is bounded over E.
[Kanpur 1983,32, 77, 76,70 ; Indore 78]
Let I(x) be a bounded measurable function defined over a
measurable set E.
We can take 0: </(x) < fJ ¥ x E E.
Divide the closed interval [ex, ~l by means of points
Ot=Yo < Yl <: Yz < ... < Y.. =.B,
and introduce the sets
ek=E [h ~ I{x) < YtHl.
Evidently,
fl-I
E= U el:, f'k n ek'=.p for k:j:k'.
10=0

Since I(x) is measurable over E and hence the sets el: are
measurable. Therefore, we have
11-1
m(E)= , meek)'
10=0

11-1 11-1
Define S= L: YI< m(e,.), S= l} YAH m(eIJ·
7:=0 k=O

Then sand S are respectively called Lower and upper Lebes-

gue sums corresponding to this mode of sub-division.
Now we shall prove a lemma.
Lemma. To prove that s <; S for any mode of sub-division.
Let jlo and So be Lebesgue sums corresponding a to particular
mode of sub-division. We shall first show that if we add a new
point y of sub-division and again compute sums sand S, then
So ~ s, S E;;; So

i.e. lower sum does not decrease and upper sum does not incre-
ase, when new sub-division points are added.
Let YI <; Y ~ YHI'
We see that for k=t=i the sets ei; and semi-open y
intervals [Yb YI:H) come into new method of sub-
division as well as in old. -Y
The set e, is replaced by two disjoint sets ePI and ePI, where
e/(II=£ (Yi ~ I(x) < y)
e,(2)=E (y ~ /(x) < Yl+l)'
Hence 'l1(e,)=m (l';(J')+m (l'/2».
142 THE LEbESGUE INTEGRAl. OF A FUNClION

The semi-open intervals [)\, )',.1) is replaced by two disjoint

semi-open intervals [Y;, y,
and [y, Yi+I)'
Thus,
i-I n-l
s= I' YI; m(ek)+Y' m(ep»+y.m(e/ 2»+ 1: Yk.m(e,,)
k=O 1:-i+1

'-1 n-l
~ 1: YI .. m(e;,)+Yt.m(eP»+Ytm(ePi)+ 1: y" meek)
k=O k_i+l

1\-1
= E Yk.m(e,J=so
beO

s ~ So I e., So ~ s.
By making parallel arguments we can show that S ~ So
This proves the result (I).
Going back to the actual lemma continued above, we
proceed as follows to prove it.
Let (Sh Sl) and (S2' S2) be Lebesgue sums corresponding to
any two modes of sub-division of the segment [ae, tl), say I and II.
Set up a method III, the method III consists of division points of
the method I and II. Here we also suppose that the method II
consists of division points in addition to the method I. They by
what we have established, it follows that
SI ~ S2' SI ~ Sa, S2 ~ Sa,
S2 E;;; SI' Sa EO; SI' Sa"':::;; S2'
so that SI < S2 ~ Sa, Sa ~ S2 ~ SI .:.(2)
where Sa and Sa are Lebesgue sums corresponding to the method
III.
Since Sa ~ Sa.
In this ev~nt (2) takes the form
$1 ~ S2 ~ Sa ~ Sa ~ S2 ~ SI' .. (3)
From this the required result follows, i.e.,
S ~ 5 for any mode of sub-division.
Q.E.D. (lemma)
For (3) it is quite clear that the sequence <.s/<> is bounded
abovc and hcnce sup. {s,,} exists
k~l
r-- -1--1--1-1-1-1"--1-'
o ~ ~ ~ U V ~ ~ ~
THE LEBESGUE INTlGRRL OF A FUNCTION 143

It is also clear trom (3) that the sequence < S" : kEN> is
bounded below and hence info {S.~} exists.
k~l
Write u=sup {.flo}, V=inf {Sk}
k ~ 1 k~l
Then U ~ St, V eo;; SI.
We claim U V. <
Suppose not. Then U> V.
For any. > 0, we have
U I!
- "2 < 810
V'
'2'E
~
S l'

Now,
i.e. 8 1 -S1 > U-V-E. ... (4)
By assumption, U> V so that we can take
U-V> 2,.
Then (4) reduces to 8 1 -S1 > •
Making • -+ 0, 8 1 - S1 ~ O.
A contradiction. For SI ~ SI·
Hence our initial assumption is wrong.
This ~ U~ V.
Thus S1 <; U ~ V ~ S1
From which
"-1
i.e, 0< V-U < k=O
E (Y"+1-YI,).m(ek)'

Taking A=max (YkH -Yk), we get

"-1
o~ V-U ~ E A.m(ek)
k=O
or o ~ v-u ~·A,m(E). ... (5)
Since .\ can be made arbitrarily small and therefore making
A-+O in (5), we get
o eo;; V---U ~ 0
i.c. (in limit)
V=U
Definition. The common value of the members U and V is
called Lebesgue integral of ftx) over E, i e.
t !(x) dx=common value of the memb:!rs U and V.
144 THE LEBESGUE INTEGRAL OF A FUNCTION

If we take E=(o, b), then the following symbol is used:

J: f(x) ix=common value of raembers U and V.

Thus we see that a bounded [Link] function is Lebesgue-
integrable. .
Case II. When the function f(x) is unbounded.
[Kanpur 76, 75, 73, 71, 70 ; Banaras 67, 71]
Let f(x) be an unbounded measurable function defined over
a measurable set E. Suppose, without loosing generality, that
I(x) ~ O.
For if the result is applicable to positive functions, it is equa-
lly applicable to negative functions and therefore it is so by
addition in general case.
Let mEN .be arbitrary. Define a function [/(x)].. as
follows:

{
when f(x) ~ m
f(x)
[/(x)]m= when /(x) > m.
m
Now we shall show that [/(x)] .. is bounded and measurable
over E. Evidently [f(x)Jm=min {I(x), m},
rough upper bound of [f(x)Jm=m
rough lower bound of [f(x)]m=O.
This ,hows that [f(x)]", is bounded over E.
Let a be any real positive number.

E{[f(x)Jm > a}={ E[f~ > 0] !~: ~:

f(x) is measurable over E implies the measurability of E [/(x»o],
o is also a measurable set.
From what has been done, it follows that
E {[f(x)Jm > a} is measurable and hence
[J~x)]m is measurable over E.
The fUIlction [f(x)]m being bounded and mt'l\slIrable, is
L-integrable over E.
Notice that
[f(x)], ~ [ftx)\+l ¥ r E N
and therefore, by Theorem 4,

• J
E
[f(x)J, dx ~ J
E
[f(x)]r+l dx .

This proves that the sequence

<t r f(x)1, dx : r EN>
THE LEBESGUE INTEGRAL OF A FUNCTION 145

is monotonic non-decreasing and hence

Lim
m·..o,.oo
J [/(x)Jm dx exists
I>
and we write

f/
(x) dx= Lim
E m-..o,.oo E
f
[f(x)]", dx. . .. (1)
If the second member of (I) is finite, then we say that func-
tion/ (x) is L-infegrable or summable over E. Thus we see that
every non-negetIve unbounded measurable function is L-integrable
but only functions baving finite integrals are called summable.
In order to define the integral of any unbounded measurable
function we proceed as follows:
/(x) ~ 0 for x E El
and j(x) < 0 for x eEl'
Let E=...oEI U E2 , then El n El=0.
By countable additivity property of the integral, i.e. Theorem 3.

fI>
l (x)]", dx~oJ
~
r/(x)]", dX+[ ~ [J~x)]", dx. ...(2)
Making m-+oo, we get

I/6
(x) dX=J
El
lex) dX+[ / (x) dx.
£2

If J£1
/(x) dx=oo, J E,
/(x) dx= -00, then

f f(x) dx=(l-oo)+(-oo). whi~h is meaningless.

The integral of a function has a meaning if and only if one
of the integrals on the R. H.S. of (2) is finite.
Definition. Let f be a measurable function on [a, bJ. If both
fiand/- are L-integrable on (a, b), then we say that/is L-inte-
grable. In this case we write / E L (a, b).
Thcorme I. ~r tile ordinate set Q (/, E) is measurable then
/ (x) is measllrable over the measurable sel, E.
Proof. Let D «(.
E) be Lebesgue-measurable, E being a
measurable set.
r
To prove that (x) is measurable over E, is is enough to
prove that E (j(x) >0) is measurable, a being any real positive
number.
Let mEN be arbitrary.
Deline t.. . -tx E E: r,"<) > a}=E (f.> a)
146 THE LEBESGUE INTEGRAL OF A FUNCTION

Ea+l/m={X E E :f(x) > a +1}=E ( f > a+k)

and 1=( a, a+k)
Evidently
and therefore
Ea X I ::J EOH/.,. X I.
Also EJ X I ~ 0 n (R" X /) ::> E"+llm X I ... (1)
where Rn=Rx Rx ...... (n factors).
o is known to be measurable. R" is always measurable. An
interval is always measurable and hene e I is measurable. From
-this it follows that Q n (Rn x I) is measurable so that
m. [.on (R"x/)]=m e [f.'n (R"xln..=m LOn (Rnxl)) ... (2)
~
From (1)
m, (Ea X /) ;;;. m; [SJ n (R" x I)]
and m.[Dn (R"x J)] ~ m. (Eo+l/rn X I).
Making use of (2), we get
m. (Eo x I) ~ m [Un (R" x I)]
and m [D n (RI> x l)] ~ m, (E,,+1/m X I).
Combining the last two inequalities, we have
m, (EI> x I) ~ m [D n (R"E I)) ;;;. m. (E"tl,,,, x /)
From which
m, (Fax I) ~
m. (E"aH;mx/)
i.e. tn.. (Eo) Xm,(I) ? ·m. (EO+1Im) X m. (I)
i.e. m, (Eo) X m (I) ~ me (EaH,m) X m (I)
or m, (Eo) ~ m. (EaH/m)·
Making m-+-oo and observing that
Lim",-+o:> m, (Ea-tl/"')= m. (Lim,..-+- 00 E O +1I"')
we get, m. (Eo) ~ me (Eo). ...(3}
But me (Eo) ~ m, (E.) is true lor any set E". . .. (4)
Combining (3) and (4), we get m; (Ed)=m e (Eo).
This proves that Eo is measurable, i.e.,
E (f > a) is measurable. Hence proved.
Ex.1. The Lebesguf! integral of a bounded measurable Junc-
tionf(x) defined over a measarable set E is the common limit of the
upper and lower Lebesgue sums s alld S.
Sol. To prove that
THE LEBE("GUE INTEGRAL OF A FUNCTION 147

IE f (x) dx=s=S. [in the limit]

We know that

s ~ Lf (x) dx ~ S.

Taking limit as max. (YH1- y;.) ~ 0, we get

s ~ tf (x) dx ~ S=s

ff E (x) dx=S=s [in limit]

First Mean Value Theorem 2. If a bounded measurable func-

tion f satisfies the inequality
II ~ I (x) <,; ~ on the measurable set E,
then rJI..m (E) <,; fI
Ii (x) dx ~ ~.m (E).

Proof. let I be a bounded measurable function defined over

a measurable set E. Let I (x) satisfy the inequality.
lit <, I (x) ~ ~ ¥ X E E.

To prove that a..m(E) <,; tf (x) dx <; ~.m (E).

Let n E N be such that
IX <'/(x) <;; fJ
takes the form
1 1
IX-if <I (x) <,~+;;- on the set E.

Taking at -.!....=a,
n
()+!...=b, we get
It

a < f (x) < b l,f X E E.

Divide the interval [a, bJ by means of points
(l=Y. < Yl < Y2 < ...... < y.=b.
Let e,,=E (Yk ~/(x) < YH1)'
"-I
Evidently E = U e", e,: n ek = 0 for k::j:k'.
~-o

Since I (x) is measurable over E and hence the measurability

of the sets ek •
A-I
'J.'hen m (E)= 1: m ~e.. ).
k=l
 )48 THE LEBESGUE Il'iITEGRAL OF A FUNCTION

If a <; Yk ~ h on the set e k , then

a.m (el') ~ Yk m (e.) ~ h.m (et).
Summing the result f6r the whole set E, we get
"-1 "-1 "-I
1: a m (ek) EO; 1: Yt m (ell) I!i',; 1.' b m (ek)
k_\ 1;=0 i~O
."
"-I
I.e. a m (E) <; 1: Ytm (ei) ~ h m (E)
10=0

Making max. (Yk+l-h) ~ 0, we get

a m (E) ~ J (x) dx .;;
E'
h m (Lj

or (IX-~)m(E) E;; tJ(X)dX <;;(fH-,n m(E)

Making n-+oo, we get the required result.
IX m (E) ~ Ie J (x) dx < fJ.m (E).
Ex. 2. If the Junction J is constant on the measurable set E,
say,f (x)=c, then

J .f
E
(-¥) dx=c. m (E).

Sol. Let.f be real valued function defined over a measur-

able set E s.t . .f (x)=c (constant) V x E E.
To prove that

IE
f (x) dx=c.m (E).
f(x)=c 'V x E E ~ c <;. f(x) ~ c V x E E.
Using first mean value Theorem (2),
c.m (E) ~ t f(x) dx < c.m, (E).
This ~ t f(x) dx=c.m (E).

Ex. 3. Let f be a bounded measurable functioll defilled o\'er a

measurable set E S.t. m (£)=0.
SholV that ff
Ii (x) dx=O.
Sol. Let f be a boundc~ measurable function defined over a
measurable set E s.t. m (£)=0.
We claim In f(;) dx=O.
THE LEBESGUE INTEGRAL Of A FUNCTION 149

We can take at <;; f(x) ~ f) on the set E.

Using first mean value Theorem, we get
«.m (E) EO;; Is f(x) dx ~ p.m (E)
Using the fact that m(E)=O, we get

o <; IE f(x) dx=O

or L. f(x) dx=O.
This completes the proof.
Ex. 4. Let f be a bounded measurable function defined over
a measurable set E s.t. (i) f(x) ~ 0
(ii) IE f(x) dx=O.
Prove that f(x) =0 almost every~here in E, i.e., f(x) is equiva-
lent to zero function on E. (Kanpur 1989; Bbagalpur 68)
Sol. f be a bounded measurable function defined over
mt:asurable set E s.t.
(i) f (x) ;;;J 0 ¥ x E E

(ii) IE f(x) dx=O.

To prove that f(x)=O almost everywhere in E.
Let Eo=E (f(x)=O),
E.=E (n~i < f(x) <;; ~) a.;. x e N.
M being an upper bound of/(x) in E.
Aim. f(x)=O ¥ x on E.
and m (E-Eo)=O.
00
Then E= U E., EknE,,'=(lj for k::j:k'.
"",-0

Siuce f is measurable over E and hence the measurability of

the sets En. it means that

",(I)

M
We have nt- . <f(x) on the set E. ¥ n E N.
Using lIrst mean value theorem, we get
All'
n+ m(En) < IE {(x) dx ~ I
tI B
f dx=~O
150 THE LBBESGUE INTEGRAL OF FUNCTION

M
Finally, n+ l·m(E,,) ~ 0 V nEN
or m(£,,) ~ 0 vnEN
But m(E,,) ;ils 0 V neN is true for any set Erl •
Combining the last two inequalities, we get
m(E,,)=O V nE N
00
Thus => E m(En) =0.
,,=1

Making use of (I), we get m(E - Eo) ;. O.

Thus we have shown that
f(x) = 0 V x E Eo C E s.t. m (E-Eo)=O.
This proves that
f(x)=O a.e. (almost everywhere) on the set E.
Hence the result.
Remark 1. Show that iff is integrable and

J: ftt) dt=O V x E [a, b],

then /-0 a.e. in (a, b).
(Meerut 1972 ; Banaras 68)
Hint. Write the so¥ution of Ex. 4, Page 149.
Remark 2. Let f(x~= -1 V x E (0, i),
f(x) = 1 ¥ x E (i, 1).
Evidently f{x) is defined over (0, I)

J f(x) dx= _J1 /2 f(x) dX+Jl

l
o 0 1/2
f(x) dx

__ _ J1 /2 dx+Jl dx
o 1/2
=-i+i=O.
Thus, J: f(x)=o,
but f(x)-#O for any x E (0, I).
This proves that the integral of the function vanishes
although the function is not equivalent to zero. Contrary to the
result, just proved.
Theorem 3. CountabiJity additive property of tbe integral (in
ease of a bounded function).
Let f be a bounded measurable function defined over a measur-
able set E and let E be the union of pairwise disjoil1t sets Et. Show
THE LEBESGUE INTEGRAL OF FUNCTION 151

that J [(x)dx= E fE [(x) dx,

£. "=1 '"
00
where E= U Ek S.t. Ek nE,c'=0 [or k¥=k'.
11:=1

[Indore 78]
Proof. Let[ be a bounded measurable function defined over
a measurable set E. Let
00
E U Ek s.t, E",nE,,'={lj for ki=k'.
"=1

To show that f [(x) dx= E IE f(x) dx.

E k_l k

We can take « < [(x) < fJ on the set E, Divide the closed
interval [oe, f.l] by means of points
Cl=Yo < Yl < Y2 < ... < y .. =/l
and introduce the sets e,,'=E (y" ~ [(x) < YA:+l)'

"-I
Evidently E= U e" s.t. ek n e,,'=0 for k=l:-k'.
l=O

Since [ is 'measurable and hence the measurability of the

fI-l
sets e/c, so that m (E)= E m (e",).
k=O

Let E=E1 UE2 , E1 nE2 =0.

e,PI=E1 (Yk ~ [(x) < YkH),
e.(2)=E2 (Y" <
[(x) < YII:+l)'
E 1 nE2 =0 :0> ek(llnek(2)=0.
Then m(e/t)=m (ek(l)+m (e,,(2».
We have
"-1 "-1 "-I
I; Y" m (e,,) = E Yk.m (e,(l1)+ E Yrc.m (e..!I').
k=O k=O k=O

Making max. CYk+l -Yk)~O, we have

Ie J
f(x) dx= El f(x) dx+ JE2 [(x) dx. " .(1)
This shows that the result is true in case
E= E 1 UE2 s.t. El()E2~-0.
152 THB LEBESGUE INTEGRAL OF A HJNCTION

Let the result be valid in case

II-I
E= U E~ S.t. E_nEk'=r/> for k#;k'.
1t=1

Then we have J £
f(x) dx= II-I
1:
11:=1
fE f(x) dx.
k

Consider the case, when

..
E= U Ek
A:~I
S.t. Et n El '=-=t/> for k::j:.k'.

Hence E can also be expH~ssed as

E= ( "-1
U E.~
)
U E...
k=1

Then by what we have establised it follows that

fB f{x) dx= f(x) dx+ I IE. f(x) dx
II-I
U Ell
I,~I

= If
,=1
IE ~
f(x) dx+JE
n
f(X) dx

E
= /'-1 IE' II
f(x) dx.
This shows that t~e required result is true in case

"
E= U E.~; S.t. Ek()E/=t/I for k:;t=k'.
k~1

This means that the required rc:su/t is true for k= J, 2, 3, ...

n. By mathematical induction, it can be shown that the result is
true for any possible value of k.
Consider the case in which
00
E= U E1• S.t. [Link] '=<!> for kl=k'.
1._1

Here E can also be expressed as

THF LlHlI!SGUE INTEGRAL OF A FUNCTION 153

Taking Rn+t= ..
U
*=,. j-l
Fie. we get E= (..U E" ) U RD+!.
k--l
,

From what has been done it follows that

L rex) dx
,
fl(X) Jx+JR lit-!
J(x) dx

or I t;;
,f(x) dx~ - ':=1 ire: eJ:
f(x) dx+JR
fltl
f(x) dx. . .. (2)

Since f(x) is bounded on the set E and therefore we can take

Cl ~ lex) ~ (J on the set R II +1'
On using the first mean value theorem, we get
•• m (R .. t l ) ~ JR f(x) dx ~ (J.m (R n+1)' ..• (3)
'<f-I
00
for k,#k'.
/, ~." 11

00

111 ~ R"/-l) 1: 111 (F.)

"-"tl
lim 11m ....
III (R", 1)= 2' III (L',J
n-+oo n-+oo " lit I

-,0 (hy general principle of convergence)

lim
til (H" II)=~()'
1/->-00 ".(4)
Taking limit as II-+~"" in lJ) and observing (4), we get
c(.O ~ lim ~R lex) lix , {J.O
11-00 "II'

() -<'"
-::::, )illl J '.
R I(.\) , /'-<"')
.\ "'" t ,

I .
fI :. 00 "'1

or lim R ./(x) dx 0,
f/--:>CXi "H ... (5)
Making 1/-+ 00 in (2) and observing (5), we get

J~.
I(x) dx - Ii III
1/ ~ 00 Ie 1
"f
~; /:' f( \) dx+O
.~

r. J
"_I "'k
L' /(x) dx. Q.E.D.
154 THE LEBESGUE INTLGRAL OF A FUl'CTlON

Ex. If A and B are disjoint measurable subsets 0/ [a, b] and

if/is bounded function Lebesgue integrable on [a, b], then

(Meerut 1912)
Sol. For the sake of convenience, we take,
E1=A, E2 =B, E1 U E 2 =E.
Then E1> E2 C [a, b] and E1 n E2 =<$.
Now we have to show that

IE [dx+ fE/dx + fEz [dx.

Now write the proof of the above theorem 3, Page 150, upto
equation (1).
Theorem 3b. If/is a bounded [unction in L [a, b], and if
a < c < b, then/e L [a, c],/e L [c, b]

and
(Kanpur MSc. P. 1978)
Proof. Suppose / is a bounded function in [a, b], then
A={x E [a, b] : [(x) > ex} is a measurable set.
Let E1={x E [a, c] : lex) > Cl}=[a, c) n A
=Intersection of measurable sets.
=measurable set
Hence £1 is measurable so that / is measurable [a, c]. Then
f is bounded and measurable on [a, c] so that f is Lebesgue integ-
rable on [a, c], i.e., f E L [a, c). Similarly we can prove that
[E L [c, b]. Remains to prove that

Lf=JE/+JE2f
where £=(a, b), E.=(a, c), £2=(C, b).
Then E=E, U E 2 , E1 n E2 =1>·
Let ek(l)=E1 (h ~/IX) < YHl),

=£2 (Yk ~ [(x)

e/c(2 1 < Yk+l)
El n E2=", => ek(l) n e,p)=cp.
Then m (e/:)=m (e1P»+m(e k (2».
This ~
THE LEBESGUE INTEGRAL OF A FUNCTION 155

Making max. (YTe+l - Yk)--+O, we have

f 11
j(x) dx=!
E~
f(x) dx+J E2
f(x) dx.

Theorem 4. Let f and g be any two bounded mcasnrable func-

tions definerJ on a measurable ~et E. SholV that

(i) J" f(x) dx ~ L g (x) dx if f{x) ~ g.(X)

(ii) t [f(x)+g(x)] dx-= L f{x) dx+ L, g(x) dx

(Bhagalpur 66)

(iii) II>
CJ(x) dx=c I
E
f(x) dx, c being any constant

Proof. Let I and g be any two bounded measurable functions,

defined over a measurable set E. We can take.
at </(x) < {3 on the set E.
Divide the closed interval [IX, til by means of points
«=Yo < Yl < Yz < '" < Yn={3
and introduce the sets
ek=E (Yt <; /(x} < Yk+l)'
·1-1
Then E= U ek s.t. et n e,:=q, for k=/;k'.
"=0
Since I(x) is measurable over £ and hence the sets e~ are
measurable, so that
"-1
m (£)= E m (e,).
k~O

Step (i). Let f(x) ~ g(x).

To prove that f E
f(x) dx ~ f [;
g(x) dx.

Clearly Y_ < I(x) ~ g(x) ¥ X E ell.

From which y" ~ g\x) on the set elt.

By first mean value theorem 2,
y,:.m (e~) < f elt
g(x) dx.
Summing the result for the whole set E,
156 THE LEBESGUE IN I EGEAL OF A FUNCTION

Using countable additivity property of the integral, we have

Making max. (YHI--Yk)-)oO, we get

f E
(x) dx c::;;; IE
g(x) dx.

Step n. To prove that

Ie [f(x)+g(x)] dx=L f(x) dx+t g(x)dx.

(Kanpur 73, [Link]. P. 75)
Let s be the lower Lebesgue sum relative to f(x) with the
scale Yo, YI, y~, ... ,y.. Let s' be the lower Lebesgue sum relative to
f(x)+a with the scale yo+a, YI+a, Y2+a, ... ,y,.+a, a is any real
positiv lumber,
.. I
s= ~')'k.11l (Ct)
" ·0

ra-l
S· =-. ~' (y*+a) m (Ck)
/:'-"'11

A-I "-I
-= J.' )'k.1II (t'k) -I-a }.' 11/ (c,,)
I..-U

" 1
}.. rlc.1II (cJ I a.1II (1:")
I. II

-=:-s \-a.11I (1:').

Making max. (Yk+ 1 .l't)-·"O, we get

f E
(f-j a) tis -"" J b
tJx + (1.111 (J~') ... ( 1)
By cOllntubh: additivity propl'rty of the integral,

II:
( (-I g) dx' -
", f U+g)
"
1.;
t
-II (
,
k
lix .. , (~)
THE LEBESGUE INT'flGRAL OF A FUNCTION 157

~ ~: Ie k
(Yk+g) dx.

In accordance with (I), this gives

L (I+g) dx ~ ~~: (fek g dx I h 111 (e/;) ]

Finally L. (f-f-g) dx ~ Ie g dx+ J: y".m (e l .). ... (3)

If we replace Yk hy Yld-I on the R.H.S. of (3), we get

In (/+,1;'). dx < L ,I;' dx+ :~:J Yk+l m (e/J. ...(4)

Combining (3) and (4), we get

n-l
+ E
k~O
YHI m(el..)·

Making max. (YHI - yd-+ 0, we get

fI:.
gdx+J Idx
E E l - .
~ f (/+g) dx ~J gd~+I E
fdx.

This ~ J I:.
(f+g) d'(=j . dx+f dx.
~
f g
.. .(5)
By mathematical induction, it can he shown that

J(j~ +/~+ ... +1;.) (h=-= I ~

fl dx +J 1-.
f~ dx j .•.

+L (" dx
where < fn(x) : 11 EN> is a sequence of bounded measurable
functions defined over E. I Kanpur [Link]. P. 1985]
Step (iii). Let c he any constant.

To prove that L c(.\) dx=c L /(x) dx.

lIndorc 1978]
158 THE LEBESGUE INTEGRAL OF A FUNCTION

Let s be lower Lebesgue sum relative to [(x) with the scale

Yo, YI' Y2' ... , Yn. Let c > O.
Let s' be the lower Lebesgue sum relative to cf (x) with the
scale CYo, ('YI' CY2' ... , cy". Evi<lently
"-1
s' = E cy,.. m(eJ;)
k=O

"-1
=c X Yk m (ed=cs
k=O
i.e., s'=cs.
Making max (h-u- Yk)-70, we get

f
E
cf dx=c ff
[';
dx.
If c=O. the result is obvious.
Ex. If f and g are bounded measurable functions 01/ a set E.
offinite measure prove that .

t (af+bg)=a Ie f+b t g.
[Kolbapur
, 69 ; Banaras 69]
Solution. Let.f and g be bounded measurable functions
defined over a set E of finite measure.
First we shall establish the following results:

(i) f
E [f{x) +g (x)] dx = S f(x) dx+
B t g(x) dx

(ii) t cf (x) dx=c Jlex) dx, c bemg any constant.

For proof. see Theurem 4, just proved.
Our main aim is to show that

t (af+bg)=a SE /+h t g
a and b being constants.
f and g are bounded measurable functions
=> af+bg is also bounded measurable.
Now Ie [a/(x)-1-hg(x)] d~=In alex) dx+Lbg(x) dx, by (i)
=0 IE f{x) dx+b IE g(x) dx, by (ii).
Proved.
THE LEBESGUE INTEGRAL OF A FUNCTION 159

Ex. Iff and g are integrable (boundC!d or unbounded) functions

over a measurable set E (of finite or infinite measure), then prove
that f+g is also integrabla over E and that

L f+g=t f+fE g.
[Kanpur [Link]. 7.3, [Link]. P. 75]
Solution. Here prove the equation (5) of the above
Theorem 4.
Theorem 5. If two bounded measurable functions f and g are
equivalent on a measurable set E, i.e.,
if f=g a.e. Of! E,
then they have the same integral, i.e., t f(x) dx= Is g(x) dx.
or, The bounded measurable functions which are equal almost
everywhere have the same integral. Also prove that the converse of
this theorem is not true.
Proof. Letf(x) and g(x) be bound.:::d measurable functions
defined over a measurable set E. Let
f(x)=g(x) a.e. on the set E.
To prove that

J fix) dx=J
,. E
g(X) dx.

Our assumptiol1 implies that

f(x)=g(x) ¥ x E E except in a set of measure zero.
This fact is expressed by saying that
f(x)=g(x) ¥ x E £1 C £ s t. m (£ -E1 )=0.
Clearly E=(8-El) U E1 , (E-E1) () El=0.
By countable additivity property of the integral,

/,. (f-g) dX=!(E_E1 ) (f-g) dx+JEl (f-g) dx. . .. (1)

f(x)=g(x) ¥ x E El => IEl (f-g) dx=JEl :f-I) dx.

... (2)
160 THE LEBFSGUE INTEGRAL OF A FUNe flON

m(E- £1)=0 ~ [(£-£1) C[-g) dx=O. . .. (3)

Writing (I) with the help of (2) and (3), we get
t «(-g) dx=<H-O

01 t (f--g) dx=O

or f/
F.
dx+/
6
-g dx=O

or IE f dx-J£ g dx=O
or L,fdx-J,.gdX.

The converse of this theorem is not true.

i e., if f
£ .f(x) dx= IE g(x) dx, then f(x):;t=g (x) a.e. on the set £

i.-e., if t ([ .g) dx=O, then (x)-g(x):;i:O a.e. on E,

i.e., in particular if ff E dx=O, then /(x):;t:O a.e. on E.

We need only give an example in favour of this statement.

Define a function (in the closed interval r--I, I] st.
/(X)={-1 when x ~ 0
+1 when x < o.

Then [1 (x) dx= [1 (x) dx+ I: .f(x) dx

or II
-1
/(x) dx=jO
-I
dx+fl -dx
0

=[ II -( x I =1-1=0.
Finally fl
-1
(dx)=O

but (x):;eO a.e. in the integral [--I , J 1.

Thus we see that the integral of (vani<;hes but the function
is not equivalent to zero.
Theorem 6. What is Lchesguc integral (~/f(x) in 01/ interval?
Show that

if (R) f: I(x) dx ('[Link], so docs

iRE. LEBESGUE INTEGRAL OF A FU"IC;nON 161

(L) 1: 1 (x) dx (and they arc equal), but not conversely.

[Bannras 65; Kolhapur 1 0 ; Mt'erut 90, 88, 81; Kanpur90, 87]
Proof I. For the proof of the fir"t part of the theorem, refer
first method of Theorem 1.

To prove that if (R) J: f (x) dr. exists, then

I: f (x) dx exists and

f:f(X) dx=(R) J:f(X) dx.

(R) 1: f exists ~f is continuous and bounded in
[a, bJ
f is continuous => f is measurable.
Now f is bOllnded and measurable and so it is L-integrable

over [a, bJ. Hence (L) f:f(X) dx exists.

Divide the interval [a. bJ by [Link] of points
Yo, YI' Y2'·· .. ·····• Y .. s.t.
a=yo< Yl<Yt<· .. <:y.. =b
and introduce the sets
e;,={x E [a, b] : y., ~ f (x) < }":tl}·
Let m." Mk be lower and upper bounds off (x) for the poinh
lying in the interval [.vb Yktl]'
"-I
Lower Riemann WIU= L m\'(Yk+l - rk)
k -0

"-I
Upper Riemann sum = ].' Mk(htl'- y,J
k~O

rex)
i;;. R·int,;grahle over [a, b]. Hence in the limiting
case whch /'t1~Yb both Riemann sums tend to a definite limit

"R) f~ f(x) dx.

L-intcgral of /{y) over la, b],- J:f(X) dx

= 'EI jY1d1 f(x) dx the integral is additivt:,
k~O Yk
162 THE LEBESGUE INTEGRAL OF A FUNCTION

Using first mean value theorem,

I
< ell f(x) dx < M~.m(ek)
m/:.m(ek)

i.e., mil (Yl<+l-Yk) < JYk+lf(x) dx <; M" (YHI -·Yk)·

Yk
Summing this for the whole interval [a, b], we get

Making max. (Yk+l - Yk)~O, we get

(R) I:f(X) dx ~ /:f(X) dx <; (R) I: f(x) dx

which ~ l:f(X) dx=(R) I:f(X) dx.

II. The converse of this tbeorem is not true,
,·,e., if I: f(x) dx exists, then it is not necessary that

(R) I: f(...;) dx exists.

We need only give an example of a function which is integra-

ble in the sense of Lebestue but not in the sense of Riemann!.
[Meerut 88; Kanpur [Link]. P. 1982]
Letf\x) be a function of x defined over an interval [0, 1] s.t.
f(x) = 1 when x is irrational
and /(x)=O when x is rational.
[Kanpur 74; Banaras 69]
Clearly f(x) is discontinuous. Points of discontinuity form
measurable sets. Every sub·inteual of (0, 1) contains rationals
as well as irrationals. For all values of r and for all modes of
sub-division, M,=l, m,=O .
• -1 .-1
Lower Riemann sum= l. m, (y,+:-y,)= }; O.(Y'+l-Y,)=O.
, r=O r:::::110

a-l .-1
Upper Riemann sum= 1: M'.(Y'H -y,)= E 1.(Y,+1 -y,)=1.
,-0 ,=0
Upper Riemann sum~Lower Riemann sum.
This => f (x) is not R-integrable over (0, I).
THE LEBESGUE INfEGRAL OF A FUNCTION 163

This ~ (R) J: f{x) dx does not exist.

Let El and E2 be the sets of points of irrationals and ratio-
nals in [0, 1] respectively. Let E=E1 UEa•
Evidently El n EI = flJ .
DU, E)=Q(f, E1>+U(f, E~). .
D( f, E 1 )= {(x, Y) : x EEl' 0 <;; Y <; Ax)}
={(x, Y) : x EEl' 0 ~ Y ~ I}
={x: x E E1}x{y: Y € [0, I]}
m DU, E1)=1 X 1=1.
Again D( I. E2 )=f(x, y} : x E E2 ,O " Y ~ f(x)}
={(x, y): x E..E~, 0 ~ Y t;;; O}=E.x{O}
m D(f, E2)=m(E2)xm({O})=m(E2 )xO=O.

tf(X) dx=mU(f, E)=mU(f, E1)+m!l (f, E.)=I+O=1.

This proves that f(x) is [Link] over [0, 1],

i.e. f: f (x) j;e exists.

Finally, tf(X) dx exists but (R) J: f(x) dx does not exist.

Ex. Prove that a function integrable over a closed interval
I=[a, b) in the sense of Riemann is also integrable over I in the
sense of the Lebesgue. Discuss the integrability in the Lebesgue
sense of the fwzction f over [0, I] defined by
{
I if X is irrational
f(x)= 0 if x is rational. (Banaras 1969]
Solution. Prove this as in Theorem 6.
Ex. Let a function f be defined on the interval (0, I) as
follows:
is irrational
I(X)={~
X
X is Irrational.
Is I integrable ill Riemann unse? Is If integrable in the
Lebesgue sense '1 lKanpur 74J
Solution. Here write I I part of Theorem 6. Here notice
that
U ([, E1)={(x, y) : x EEl, 0 ~ y ~ I(x)}
={x: x € E 1}x {y: 0 ~ .v ~ O}
=E1xlO}.
164 THE LEBESGUE INTEGRAL OF A Fl1NCTION

Hence mQ(/, E1 )=m (Et).m({O})=O as m({O})=O,

m[Q(J, E2 )]=m(E2).m([O, 1])=1 x 1=.::1.
Ex, 5. Let p > 0, q> 0, then show that

(i) J:;:,-:q dX=~ "-/+q +p~fq --p-l-)q +, ..

and deduce that

(it) LOg2~-'-1--}+~-}.+ _,
~ 1 J I
(iii) 4 =1-- 3 +5---=r+ ,.,
Solution. We have
x P- 1( 1+XO)-l = x P -) (1 _.. xtl.!- x'q _x3Q +,.\4Q - .,)

00
= XP-l L' (x2I1Q ._ X(2,,+1)Q)
'.=0
, 00
= 1: [X2"~l-p-l_X(2"+1)G-t-P-ll
,,=0

Integration vie Ids the .result

.. /

Jo ~V-l
1 dx= ~ [ X2 .. q~::, _ _ _~~~~~q+p 11
1T x q .. =0 '[Link] r P (2 + I) q +p ~(j=O

III
= ____ +____ - __
1 .L
P p',-q 1'+2:.1 p+3q I '"

J l XoP-l

o 1 +Xfl
1 ! I
dt=-·- .--- -- + .-~-
p p+q p +i..q
-.-1. -+.,.
.v +3'{ , .. (l)
Hence the result (i).
Put p=q = I and observe that

J1 1' . dx= [ log (! +x) 11

I = Jog 2,
o I +x Jo
we get log 2= I·· i + t -- t +'" Hence the rc,ult (ii).
Taking p= I, 1j=2 in (I) [Link] obscuing that

}11 I
Q +X2 dx= [ tan- J x J1 r.:
n ='0j',
 THE LEBESGUE INTEGRAL OF A FUNCTION 165

we get

Henc~ the result (iii).

Ex.,
6 . .~.
P rove that th e fiunctlOll sin x . b
IS /lot Le esgue integr-
x
able over [0, 00[. [Banarlls 1971]

·
So1Ut 100. T 0 Sh ow t hat th e f unctIon
. - [Link] not L -mtegra
' bIe.
x

-x
If' we show t hat [Link] not summabl e over [0,00[, we can
conclude the result.

For this we must show that J OO

o
I sin x ,
!-I- - - !
X
00.

Consider the integral

I"'"
o
sin ~ dx=
X
£
r- 1
Jr.
(r-ll"
I~}..!l
X
X J dx

{y+ <!-..:J..tt1.1 d
-E" Jft' I siny+(r-I)1t
r~l 0 y

~ " J" I
1.:
r~l 0
sin...ll+(r--l)1t)
rn
I if)'

I ~.,
== 2'
n
- I siny I d)'= 1.:.. _.
2
r~ 1 rIT. 0 r·' I rIt

Lim
n_oo
In-lsin~.J dx ~ ~
0 x 1t
E !r
.-1
=~x :JO=co
7t

i.e., 1 L~~~J
00
o x
dx= 00.

Ex. .Give an example of a function which is not integrable in

the sens~ of Lebesgue. [Kanpur 74, 75]
Solution. Write the solution of the above solved Ex. 6.
Ex. 7. Prove thai the function

f(x)=-dd-
x
(X2 sin x·~)=2X sin~-~
xx
cos ~
x
is not L-integrable over [0, 1]. [Banaras 67; Kanpur 88]
I'. • 1 2 I
Solution. Let J(x)=2x sm Xi-x cos,Xi'
166 THE LEBESGUE INTEGRAL OF A FUNCTION

The function 2x sin

x~ is bounded and continuous over
[0, 1]. Continuity of 2x sin ~ implies that it is measurable over
[0, 1.]. Being bounded and measurable over [0, I]. 2x sin ~
x
is
L-integrable over [0, 1]. To show thatf(x) is not L-integrable
over [0, 1], it suffices to show that - ~
x
cos!... is not summable
x
over [0, J].
For ~ llis we must show that

J:~\ cos ~I dx=~.

Let a,,={(2n+l)rc}-I/1, b,={(2n- U rc}-I/.,

a
•
E; x <.: btl ~ !a" ~ !.. ;;.!.. ~ ..!.. ~ .!... ~--
x b,. b.'" Xli a,,1
<?"

, 1
.. (2n-A) rc x <; (2n+i>
~ "I 1t

~. 1 \ 1
,.,. I cos Xl ;;t 2"

It
Va x
1-
Q
I ~
I dx=,l;
cOS"2
X
I
J ,,-I
Ibnan -1 IcOS"l1 II dx
X X

~ t./JbOnn -21x
ft_l
• b,,· 1 · 6n+ 1
dx= E llog -2=4- 1: log 6--1=00
.....1 an n=1 n--

o
00
Jll Icos 1I dx=oo.
oX
- "21
X I

This proves the required result.

Ex. Answer the following questions in 'Yes' or 'No'.
(vii) Do the Lebesgue integrable functions over (0, 1) form
a vector space 'l [Kanpur 74]
(viii) Is a function equal to Q measurable function almost
everywhere itself measurable? [Kanpur 74]
(Ix) Is the Lebesgue measure of the union of two measurable
sets equal to the sum of their measureJ ? [Kanpur 74J
(x) Is a function, which is differentiable over (0, 1), Lebesgue
integrable over (0, 1) '1 [Kanpur 74]
THE LEBE5GU 3 INTEGRAL OF A FUNCTION 167

(xi) Is a continuous function over (0, I) Lebesgue integrable?

[Kanpur 75]
Solution. (vii) yes, (viii) yes, (ix) no, (x) yes, (xi) yes.
Ex. 8. Show that if
f(X)=! (0 < x E:;; 1), f(0)=19.
thenf is not integrable on [0, I]. [Kanpur 1979, 77]
Solution. Let n be a natural number. Evidently f is an
unbounded function. Recall that
[f{x)]
n
=1
f{x) !fO ~f(x) ~ n
n Iff(x) > n.
Here we have [f(x)].=! if!
x n
~x <1
[f(x)] < n => X-1 ~ n ::;. n-1 ~ X
n

[f(X)]n=n ifO\.< x <~

[f(O)].= 19.
The value of[ f(x)]n at x=o does not affect our calculations.

I /
I [f]" dx=Jl t1 n dx+Jl
o
~ dx
0 lin X

=[ nx lIn +[ log 1:n X

1
--log -+I=I+log n
n
Lim
n-+oo
fl
0
[f]. dx= I + Lim log n.
n-+ 00
Since the value of R.H.S. does not exist, hence f is not
Lebesgue integrable over [0, I].
Ex. 9. Show that the function f defined by
t
f(x)=x l /3 'O < x < I,
and f(O) =0
is Lebesgue integrable over [0, 1].
Solution. Let n be a real number. Recall that
[Ji
(X)]
"
={
f(x) !f 0 ~ f(x) ~ n
n Iff(x) > n
Here we have
[rex)] =-h.
xn n~ ~ x ~ 1
168 THE LEBESGUE INTEGRAL OF A FUNCTION

l For /(x) ~ n~ -}-p <; n ;;> ~,3 ~ X ]

[ j(X) ] It =ll if 0 < x < I/'~'l

[j(O) ] .. =0

1[f] o
1
II
dX=JIJ;
In x
3 ~3
I
dX+jl/n3
0
Il dx

3 ( X 2j3 1 ( ) l/n 3
=- ) +n x
2 ljn 3 0

__} (1--!..)+~- 2-
-2
3 !.n
n2 n a- Z

Lim
n-+oo
JI [I]
0 2
,.
dx=~.
This ~ I is Lebesgue integrable over [0, 11-

Remark. Although (R) J: I(x) dX.=~.

Theorem 1. Suppose I is measurable on a !l/f!a:,urable set E.
To prove thai f is integrable iff II I is integrable and that

ISEfl ~ LIll· [Kanpur 73J

Proof t Let f be a measurable function on II. measurable
set E. Suppof>e f is Lebesgue integrable over E. Then its positive
and negative parts 1+ and /_ are also Lebesgue integrable over E.
:But 1/1=/++1-. This'means that If!' is [Link] integrabl.:
over E.
II. Letf be a measurable function on a measurable set E S.t.
I!I is Lebesgue [Link] over E, then f 1f I <
B
00.

Since 0 ~f+(x) ~ If(x) IV x E E.

Hence t.f+(x) dx ~ t I/(x) I dx < 00

or f £
I+(x) dx < 00, that:;::> . f+ is Lebesgue integrable:

Similarly we can prove that 1- is Lebe~gue integrable.

But f=l... -1-· Hence f is Lebesgue integrabk on E.
1 HE LEBESGUE INTEGRAL OF A fUNCTION 16'1

Ifl. Letf(x) ~O on £1 and.1~x) <0 on E~. Theil E 1 n£2=0

and E=£1 U E 2 • By countable properly of tho;: integrai

J Ifi dx=fE If I dX+J~- If I dx

E 1 .~! ... (1)

and J,fdx=[Link]+J fdX=Jr I/[Link] dx

E £1 £2 E1 1 °2

This ~

I I
E ldxi<:l
iJ
I ,1:. 1
I
olfidx+1 If r
E I/i d '; "'''\E
I, 2 i J 1
: J
1(1+ £-111=
~
JEli
O'j

:::>IJIE fll ~ Jill.

E
This completes the proof.
Theorem 8, Let I be a bnunded measurable lunction deji'ned
over a measurable set £. If
Thl'.I1 , E f \ ~ E I fl.
0
I'
[Meerut 1972, 71; Kolhapur 70; Banaras 71j
Proof. Hert' write III Part of the about Theorem 7.
Theorem 9a. Let f and g be unhounded measurable fUllctiuns
defined over measurable set E. Show that

(I) JEfdX= 1!1 fEkldx where E=I~l E .. Sot,

n £/=0 for k::j=k'. Ek
[Meerut 1975, 72, 71J
Proof. Let f and g be unbounded measurable fUllctions defi-
ned over a Tll('asurable set £.
Suppost', without losing generality, that/(x) ~ 0, g (X) ~ o.
r or if the result is applicable to positive functions, it is equally
applicable to negative ';unctions therefore it is so by addit;on in
general case.
Let mEN be arbitrary. Define [!tx)]m and [g (x)] ... as
follows:
[!(X)]m={f(X) when/ex) ~ til
m when/ex) > 111

and then we define J

E
(f) dx= Lim
m~oo
I 1:-
0
[f]
In
dx.
Similarly define

f E
g (x) dx= Lim
m·,.oo
fE
(g (xli", dx.
170 THE LEBESGUE INTEGRAL of A "UNCTION

00
Step (i) Let E= E E. s.t. Ek n Ek'={lJ for k�k'.
It=l

To prove that J E
f{x) dx= jJ IE f(x) dx.

E
1:=1 "

[f (x)]", is bounded and measurable over

Since and there­
fore, the integral of[f(x)]", is additive, i.e .•

I E
rf(x)]", dx E
.=1
J.:o�k [f(x)]", dx

i:
� 1<-1 II IE [f{x)�111 dx
Making In-+«> and then n .... «>. we get

J E f(x) dx ;a '£1 IEk f(x) dx. . .. (1)

Again

IE [f(x)]", dx= ,,! lEk [f(x)]111 dx � ; 17;< f(x) dx

.... 1 Lk

.: [f(x)].. <f(x)

i.e.,
I E [/(x)�.. dx�� ! JEt f{x) fix.

Making m ... OCt we get

I E dx <
f t1 IE" f dx. ...(2)

Combining (1) and (2), we ge t the requird result.

lEI dx= k�J Ek f dx. . .. ( 3)

Similar Problem. Let X be measurable space in which M is

the O-ring 01 measurable sets and I" is the measure. Suppose f is
measurable and non-negative on X and for A E M,

define. t/>(A) = L f dl", then 1> is countably additive.

(Kanpur 1986]
Solution. Prove as above that

fEI dl'== E, I 1 Ele /"1" ...(4)

THE LEBESGUb INTEGRAL OF A FUNCTION
171

But ~(A)= I L dll, E= ~l E k= 1~ Ek

:. if> (E)= I~l r/>(E k) or rp ( k~ Ek )=~! ¢(E k)

This proves that cP is countably additive.

Theorem 9b. Let I, g be non-negatIve valued functions on la, b].
If I, gEL [a, b], thenl+g E L [a, b] and

Ie (/+g)=JE I+/ E g.
[Kan{}ur 1972, 73; M. Sc. P. 75]
Proof. To prove that

IE (/+g) dx=fE Idx+JE g d.".

Let 1\1 (x)=/(x)+g (x).
Now we have to show that

IE q, (x) dX=/Ef(X) dx+fE g(x)dx.

Case I. When I(x) ? 0, g (x) ~ 0 so that 4 (x) ~ O.
Firstly we shall show that
[,p (x)]", ~ l/(x)1",+[g (x)]", ~ [1\1 (x)hlll. . .. (3)
Let 1 (x) EO;; m, g (x) < m ; then
[J(x)]",=/(x), [g (x)]",=g (x)
[1\1 (x)]m EO;; '" (x)
=f(x)+g (x)
=l f (x)]",+[g (x)] ..
i.e. [1/1 (x)]", < [f(x)] ... +lg (x)]".. ...(4)
If either! (x) or g (x) is greater than m, then 4 (x) > m so
that
r
[1/' (x).)",=m < f(x)]",+[g (x)lm
i.e. [I), (x)]", < [f(x ]",+ [g (x)]"' ...(5)
From (4) and (5), it follows that
[1jI (x)]", <
[hx)]",+[g (x)]",
... (6)
is always true.
Exactly in a similar way, we can show that
[!(X)]m+[g (x)]", ~ [tjI (X)]2"" ... (7)
Combining (6) and (7), we get the result (3).
Integrating (3).and observing that
 172 THE LEBESGUE INTJ:GRAL or A rU?-:CTION

J {~f(x )]... +!g

E (X)Jm} dx= IE [1(x)1". dx+ JE [g (.0:) 1m dx,
we ~et IE!Y (x)Jm dx ~ IE [f(X)]", dx+ IE [g (x)]", dx
< JE [[Link](x)h'n dx
Maldng m-7OC,

IE 4{x) dx ~ fE/ex) dx+ IE g (x) dx ~ IE y; (x) dx

or L~ '/1 (X) I
dx= Efpq dx+t g (x) dx.
Case U. When j(x) ~ O. g (x) < 0, q, (x) ~ O.
Then Ij.(x)-H - g(x»=f(x).
Now the resuIt follows from case 1, since-g(x) > 0.
Case III. When [(x) ~ 0, g(x) < 0, tI:(x) < O.
Then (-1/1 (x»+f(x)==( --g (x».
E v~ry function in this equation is > O. Now the result
follows from caSt: I.
R.:mark. Iff is an unbounded measurable function defined
ova (l lllc~surable set E, then it can be easily shown that

(i) iJ~J(X) dx !~ t: Ij(x) I dx [Kanpur J991J

(ii) f. (I (x)
J 1:.
dx=c r
,L
.J(X) dx

(iii) If ((x) ~ 0, Il:.J(x) dx=O, thenf(x)=O

almost everywhere i fI E.
(i.) Ifj{x)= go (x) a e. on the set E, then

fEf(X) dx= f£ g(x) dx.

I::XtRCJS[S
1. Iff (x) i~ bounded IDca:;urable fundion on [a, b], show
thatJ(x) is L.:-bc:sgu..: integrable over [a, b].
[Klinpur [Link]. P. 1983, 82J
2. Given an example of a functIOn which is integrable in
the sense of Lebe~gue but nol in tbe ~ense of Riemann.
I Kanpur [Link]. P. 1982J
3. Iff is a bounded function in L [a, b] and if a < c < b,
prove that
JE L La, C],fE L [e, b]
and J: f'- [f+ J: f· LKanpur [Link]. P. 1978j
THE LEBESGUE INfEGRt\L OF A. fUNCTI0N J73

4. Iff is a bounded measurable function on [a, b], show

that f is L -i ntegrable on [a, b j. [Kanpur 1977, 76]
5. Explain how the Lebesgue intt:;gral of a bClunded function
is extended to unbounded functions. Show that if
!(x)=-,-ll/x,O < x ~ I
1 0, x=O
then f is not integrable in the Lebesgue sense over [0, 11.
~Kanpur J976]
6. Define the Lebesgue integral of ~n unbounded measura·
ble function over a measurable set. Gjve an [Link] of a func-
tion which is not integrable in the Lebesgue sense.
lKanpur 1975J
7. Iff and g are integrable functions over a measur::..blc set
E, th';n prove that f+g is also integrablt: over E and tha.t

J (f-rg)=J
E E
f+f E g
lKallpur [Link]. Pre. 7S, :"[Link]. Flnal 73]
8. Give an example of a function which is not Lebesgue
integrable. Justify four answer. [Kanpur 197'4]
9. Define the Lebesgue integral of atl unbounded measura-
ble function over!l measurable sec E. Prove that f is integraLle
iff If I is integrable and that

:if z: f~f z: If:

d r x-oSill
L e t ji(X ) ~= dxL · X21 l' In t hcopen Intcrva
. I (0 , )I . SI
,1OW

thilt f is not integrable in the Lebesgue sense over (0, I).

lKanpur 1913J
10. (a) Define the Lebesgue integral of a bounded mcasura-
bk function f in a bounded [Link] se~ E. proving the
existence of the integral.
(b) rftwo fllnctionsj:lnd g are Lehesgue integrable over a
~et E, then prove that f i-g is also integrable and

t U+g)=L.r+t g·
(c) Let f be integrahle over a set E and suppose tl:at £ is
the union (sum) of a count&ble family of pairwise disjoint
measurable sets E.; i.e., E=-=l::d-E2+ ..... .
Prove that
f /=1' JE
r n .. "
r. ~ Kaupllr 1972!
174 THE LEBESGUE INTEGRAL OF A FUNCTION

11 Explain how the Lebesgue integral of a bounded function

is extended to unb:>unded functions. Prove that the function
l(x)=2x sin ~-~x cos.!....
x x
is not integrable in the Lebesgue's sense over (0, 1).
[Kanpur 1971]
12. Prove that a bounded measurable function on [a, b] is
Lebesgue integrable over (a, b]. [Kanpnr 1970]
13. Iff is a non-negative measurable function defined over
a set E of measure > 0 and
tldX=O
then/(x) is a e. zero. [Bhagalpur 1968]
14. D.!fine the Lebesgue integral of an unbounded function.
on [a, bJ. Show that if
f(x)=~ (0 < x ~ 1)'/(0)=19,
then/is not integrable on [0, 1]. [Kanpur 1970]
IS. Define the Lebesgue integral of a function I, with
respect to the measure over a given set. [Kanpur 1969]
16. Explain the full definition of the Lebesgue integral of a
",-measurable function/: X~R. Point out the case in which the
integral is not defined. Define L' (X, ,.,.). [Banaras 1971]
Hint. Case II of § 9-0.
17. Let (X~ M, ,.,.) be a measure space and I: X-+R be
,,-measurable and non-negative. Prove that
I" fdp., ME M
is countably additive on M. [Banaras 1971]
Hint. Here write Theorem 9a Page 166.
18. Decide whether or not each of the following two func-
tions is in L' (R, m), and give reasons for your decisions:
. sin oX
(1) 1(0)= I, I(x)=----"- for x::;cO
x
(ii) g is even and
J2-,,-1 for x E [2k, 2k+1l
g(x}=1_2-k - 1 for x E [2k+ I, 2k+2]
where kEN. I [Ban~ras 1971]
Hint. For first part, see Ex. 6, Page 16~.
19. Show that I: [0, 1]~R is Riemann integrable iff the
discontinuities off form a set of Lebesgue meflsure zero.
[Banaras 1970]
THE LEBeSGUE INTEGRH OF A FUNCTION 175

20. Iff=g a.e., show that

L f=,t g
f [Link] g being bounded measurable functions on a set E of finite
measure. [Banaras 1969]
21. Prove that if f(x) =0 at every point of Cantor's ten nary
set andf(x)=p in each of the complementary intervals of length
3-1', then

J: f<x) dx
exists in the Lebesgue sense and is equal to 3. [Banaras 1968]
22. Define Lebesgue integral of an unbounded function.
Show that the function
f(x)=~ ~)
dx (X2 sin X2
is not integrable in the sen!.e of Lebesgue over (0, 1).
[Banaras 1967]
23. Give an example of a bounded real function on [0, I]
which is L'!besgue integrable but not Riemann integrable. Also
give an example of a function f on [0, 00] such that f is not
Lebesgue integrable over (0, 00] though the improper Riemann
integral off exists and is finite. [Banaras 1966]
24. Write a note comparing Riemann's definition of an
integral with Lebesgue's. [Banaras 1965]
25. Define the Lebesgue integrable of a measurable non-
negative function f over a measurable set E. [Meerut 1970]
26. Iff is bounded and integrable in the Riemann sense in
a closed interval [a. b) then prove that f is measurable and integra-
ble in the Lebesgue sense and the Riemann integral off over [a, b]
is equal to the Lebesgue integral over [a, b]. [Meerut 1970]
27. Iffis Lcb,:<;gue integrable on [a, b], and
J: /(1) dt=O V x E [a, b],
prove thatf(t)=O almost everywhere in [a, b]. [Ml'erut 1972]
 10
Theorems on Convergence
of Seq uences of
Measurable Functions

Introduction. Convergence in measure is essentially weaker

than convergence almost everywhere. There exist sequences of
measurable (even continuous) functions which are divergent at
every point, but which are convergent in measure.
10'0. Der. Convergence in mean. A sequence < f" > of
Lebesgue integrable functions is said to converge in mean to a
function I if
Lim
n-+oo
J 1/.. --/1
E
dx=O.
10'1. Definition. Convergence in measure. [Indore 1978)
Let < In > be a sequence of measurable function defined
over a measurable set E. Let/be a measurable functicn defined
over E S.t.
(i) /(x) < 00 a.e. on the set E.
(ii) Lim m[E(I/n-fl ~E)]=O ¥ E > O.
n-+oo
Then the sequence < f" > is said to converge in measure to the
function!
1O· 2. Definition. Pointwise convergence.
Let < fn > be a sequence of measurable function defined
over a measurable set E. If there exists a measurable functionf
over E S.t.
Lim
(,,(x) = lex) 'V x E E,
n--+ CI,
then we say that the sequence < fn > converges pointwj',~ to
fon E.
1Q. 3. Definition. Convergence almost c\'erywhere. [Indore 1Q78!
Let ,/ In / be a sequence of measurable functions defined
 THEOREMS ON CONVERGENCE OF SEQUENCES 177

over a measurable set E. If 3 a measurable function f over E and

a lIet A S.t. (i) m (A)=O
lim I.n (X)=J\x
(1'\') ,rr) \f X E E-A
n-+oo
(i.e., <fn > converges pointwise to f on E-A)
then we say that the sequence < f .. > converges to I almost
everywhere on E.
10'4. Definition. Uniform Convergence. A sequencl:: < I .. >
of measurable functions defined over a measurable set E is said to
converge uniformly almost everywhere to a functionf if :I ACE
s.t. m(A)=O and given E > 0, 3 no E N s.t. ¥ n ;;;::: no and
V x E E--A => If..(x)-/(x) 1< e.
Theorem 1. Let < fn > be a sequence of i1;ltegrable functions
which converge zn mean to a lunction f, then f ..-I in measare.
[Indore ,1970]
Proof. Let < f,. > be a sequence of integrable functions
defined over a measurable set E. Let En=E (1/.. -1 I ~ 8) where
8 > O. Then 1/,,(x)-/(x) I ~ 8 y X E En.

This => fEn I/n(x)-j(X) 1 dx ~ m(En).8 ... (1)

Since/R~{in mean on E and hence
lim
n-+oo
I Ifn(x)-(x) I dx=O ... (2)

E. C E => fEn IIn-I I ~ IE 11..- II ... (3)

(2) and (3) ~ IE" I/n-/I=O as n-+OO.

In this event (I) => m(En }.& ~ 0 ::;> m(E,,) EO;; 0 as n-H>O.
lim
But m(En) ~ 0, V Ell' Hence m(En) =0.
n--jooo
This => m [E (/I. -II ~ 8)]=Oas 11_00 Y a> O.
=> 1,,-1 is measure.
Theorem 2. Let < f,. > be a sequence of measurable lunctions
Oil a measurable set E an d I,'f' n_oo
lim I.• (x) =J,I')
(X a.e. on E .
thell show that I is measurable 011 E.
Proof. Let A={x E E: lim In (x):;i: I(x)}. Then b} def.
n-+oo
of almost everywhere, m(A)=O. Define, g..(x)=/n(x) if E A
and g(x)=O if x E A.
178 THEOREMS ON CONVERGENCE OF SEQUENCES

This =>
lim gn =lim/n if x ,. A ~ g(x)=/(x) if x ,. A
and g(x)=O if x E A.
The above facts prove that < g,. > converges pointwise to g
on E. Also gn is mea'iurable V n. Hence limit function, i.e., g is
mea'iurable. Consequently I i'i measurable.
Theorem 3. (F. Riesz). Let <: f,. > be a sequence ollunc-
tions which converges in m'!asure to the lunction Ion a measurable
jet E. Then there exists a subsequence which also converges to the
functionl almost everywhere. [Meerut 1986j
Proof. Let <a" > be a monotonic decreasing sequence of
positive terms sach that lim Uf! =0. Let En;: be a convergent series
00
of positive terms so that E nk=O.
1:=1

< In > converges if} measure to I

lim
n -+00

... ( I)
It means that corresponding to the seq!lence < Uk > we can
construct a st:quence < n. > satisfying the condition (1), <Ink>
is a sequence of < I .. >. Now we want to show that
lim
1.1«x)=flx) a.e. on E.
[Link]

Write .. (2)

From this it follows that < A" > is a monotonic decreasing

sequence of m:!a~urable sets. This declares that
lim m(A,,)=m(B) .. (3). (Refer Theorem 18. Chapter 7, page 102).
00
It follows from (I) and (2) that m(A .. ) < 2;,,=nn/c. Consequently
\ .
lim m(An) =0 .. (4) as En" is convergent. This => m(B)=O, by (3).
Let y E E -B b~ arbitrary. Then y f!. B
y ,. B => y t Ano for some natural number no
=> y If. E( lin:, II ~ Ok) 'V- k ~ nO
=> IIn~(") -/lY) ! < Ok ,?L k ~ no
lim
=> k_oo 1,.··(y)=/(Y) V Y E E-B.
THEOREMS ON CONVERGIlNCE OF SEQUENCES 179

Also m(B)=O.
lim
Hence k -+00 / ..t=/ a.e. on E.

Theorem 4. (D.F. Egorff). If a sequence 0/ measurable/unctions

is convergent almost everywhere on a measurable set E, 0/ finite
measure, then/or every 6>0,3 a measnrable set EoeE s.t.
(i) m(Eo)< 8, (ii) the sequence is uniformly convergent on
E---Eo· [Meerut 1988,90]
Proof. Let < f. > be a sequence of measurable functions
defined on a measurable set E s.t. < fra > converges almost
everywhere to a finite measurable function / on E. Let
lim/..{x)=/(x) V x E A. Then, by def. of a.e., m (E-..4)=O.
Since every function is measurable on a set of measure zero and

E::' = Ii [E (Ift{X) -f(x) I <l (11m) )] and E"'= ]1 E:'.

f,.-+/ on E ~ .lim m
n~ao
(E - E"')=O.
..
:. Given« > 0, 3 positive integer no depending on .. and m
s.t. m (E-E~.. ) < e/2m
v ••• (1)

Define Eo= U (E --E:

m=l g
)
Being countable union of measurable sets, !?u is measurable.
Also Eo C E

m (Eo)=m [ U (E-E"0m )] ~ E m (E-E"')

'"= 1 110
<
m=1
E, by (1).

This ¢ m(Eo)< 6.
This completes the first part of the promblem

E-Eo=E- U (OE -Ernno )=E n

"'=1
I
(u E"')
... =1 no

Thus l.,f n ~ rio and x E E -Eo ~ x E E~

=> If.. (x) - f(x) I < 11m.
Hence < f" > converges uniformly t%n £-Eo·
Theorem 5. (D.F. Egorff). If a sequence of measurable func-
tions converges to a/inite limit almost ef'erywhere ill a measurable
180 THEOREMS ON CONVERGENCE OF SEQUENCES

set E, then, given 8, we can lind a set 01 measure greater than

m (E) -/l in which the sequence converges uniformly.
Proof. Let < In > be a sequence of measurable functions
defined over a measurable set E. Let this sequence converge to a
finite measurable function! alntost everywhere on E. Let 8 > O.
To prove that :i ~ measurable set ACE S.t.
(i) meA) > m (E)-8
Oi) < In > converges uniformly on A.
Consider a convergent series ~7)k of positive terms and a
sequence < an > of positive terms converging to zero s.t.
171 > 172 > 173 > ........ .

00
Write An (17)= E E (1/k--f1 :? 0),0> O.
k='TI

Then m (An (17)1=0 as n- oo • ...(1) For In-f.

By virtue of (J). corresponding to every positive integer k,
we can find another positive inte~er 7)k s.t.
m [(ARk) (0)] < Tlk.
00
Now we cho.))e a positive integer ko S.t. E 7). < 8
i=ko
vo
Set B= . 2-' An, ("I)' Then m (8) < ~.
l=ko
Write A=E-·B. Then E=A U B, A n B=0.
rn (£)=111 {A)+m (8)
or m (A)=m (£) -m (B)
> m (E) -8. For m (B) < 8 ~ -m (8) > -8.
:. m (A) > m (£)-8.
Rem'lins to ShOW that
Itt (x)-"/{x) uniformly on A.
Let. > 0 be arbitrary. Then we can find an integer k S.t.
k ;;;;t ko, 0" < c
ailY x E A ~ y ~ B. For A=E--B.
~ x f'I. Ani (a,)
~ .Y f'I. E ( Ilk I I ~ ai) for k;;;-:",
~ /It(x) fix) I < 0, < ~ k ~ ni
~ /ft{x) - /Ix) !< E ¥ k ~ n" x E A
~ In "'/uniformly on A.
Let g be a function s.t./and g are equivalent on £ so that
m [E ( 1--1= g)]=O. . .. (2)
THEOREMS ON CONVERGENCE OF SEQUENCES 181

To prove that the sequence < In > converges in measure to

the function g.
For this we have to show that
[E ( I/,.-g I ~ 6)J=0. . .. (3)
.. Evidently E(
In

Iln-g I ~ E) C E(f~g)+E( lin-II;;'.)

This =>
m [E ( I I .. -g I ';J 6)J ~ m [E (l;i=g)]+m [E ( Il,.-I I ~ .)].
Using (1) and (2), we get
m [E ( 1 /..-g I ~ 6)J ~ 0+0=0
I.e. m [E ( I/n-g I ~ 6)J ~ O. ,... (4)
But measure is a non-negative quantity so that
m [E( Iln-g I;;" 6)] ~ O. . .. (5)
Combining (4) and (5), we get the required equation (3).
Proved.
Theorem 6. Lebesgue bounded convergence theorem.
Let < In > be a sequence 01 bounded measurable lunctions
defined over a measurable set E s.t.
I/.. (x) I < M >.f n E Nand>.; x E E.
Let < /., > converge ill measure to a measurable junction I on
the set E, M being a positive cOllstant.
Then Lim
n~oo
f E
j,,(x) dx=J
E
j{x) dx.
[Meerut 1986; Kolhapur 76 ; Kanpur 85, 70, 73)
. Proof. Let < /.. > be a ~equence of bounded measurable
functions defined over a measurable set E S.t.
Iln(x) 1 < M>,f n E Nand >,f x E E. ...(1)
Then In is integrable over E >.; n E N, M being a positive
constant. Let < In > converge in measure to a measurable
function Ion E so that
Lim m[E( If" II ~E)]=O,
IH'OO ... (2)

for every. > O.

To prove that
Lim
1l~ 00
f E
In (x) dx=J !(x) dx.
E " .(3)
From (I) and (2), I/(x) I < M. . .. (4)
Suppose 8 > 0 is arbitrary and set
A.=E( lin-II ;;" 0), Bn=E( 1/.. -/1 < 0).
Then E=A" U B.. and An n B,,=(lj.
182 THEOREMS ON GONVERGENCE OF SEQUENCES

The definition of convergence in measure ensures that

Lim m (An)-O.
n-oo . .. (5)
This follows from (2).
By countable additivity property of the integral

t lin-II dX=JAn 1/11-/1 dX+JBn lin-II dx.

We have lin-II <: 8 ¥ X E BII •
. .. (6)

Using first mean value theorem.

IBn I/.. ~/I dx < 8.m (Btl) E;;; 3.m (E)

or IBn I In-I I dx < 3.m (E). •• (7)

Take an arbitrary E > 0 and choose 8 such that
e
8.m (£) < "2
In this event (7) takes the form

IB.. l/n-/1 dx < ; ... (8)

Now 8 is fixed
lin-II ~ II.. I + III < M+M=2M
or lin-II < 2M.
Using first mean value theorem, we get
fl/n-/I dx < 2M.m (An) ... (9)
(5) =>
Given. > 0, :I no E N S.t. n ~ 110 => I m (A .. ) I < 4~
~ m (A .. ) •
< 4M ... (10)
:. m (An) ~ O.
From (9) and (10),

I I fn -I I dx < ; ~
A.. ¥ n 110
... (11)
Writing (6) with the help of (8) and (II), we get
t 1/.. -1 I dx < ; + ; =~ ¥ n ;;. no

or L, I I,,-{ I dx < 6 ¥ n ~ 110

THEORFMS ON CONVERGENCE OF SEQUENCES 183

It I. t
(f.. -f) dx E;; /!n-f I dx <•
or It (fn-/) dx I< fi "'I n ;a, no·
This ~

Lim
n~oo
J (J.
I> If
-f) dx=O

or Lim
n-+oo
I I>
fn (x) dx- J I>
f(x) dx=O

or Lim
n-+oo
J /.. (x) dx=
I>
II>
f{x) dx.
Hence the result.
Ex. State and prove the theorem of bounded convergence for
Lebesgue integrals. Show that this theorem is not true lor Riemann
integrals. [Kanpur 1973]
Tbeorem 7. Tbe Dominated Convergence Theorem.
Let < fn : n EN> be a sequence of measurable functions
defined over a measurable set E S.t.
I fn (x) I < '" (x) yo x E E and "'I n E N
where", is integrable over E and the sequence < fn > converges in
measure to a measurable I on E.
Then Lim I
J f{x) dx= n-oo f. (x) dx.
I> fi
[Kanpur 1991, 89, 87; Punjab 69; Banaras 65; Meerut 88, 90J
Proof. Let < In > be a sequence of measurable functions
defined over a measurable set E. Let", be integrable over E s.t.
Ifn (x) I <.CiJ (x) "'I x E E and "'I n E N. . .. (1)
This ~ I. (x) is bounded V- n.
Also fn (x) is given to be measurable.
:. < In > is a sequence of bounded measurable functi~ns
and hence these are L-integrable.
Let the sequence < In> converge in mea!>ure to a mcac;ur-
able function I on E so that
Lim m [E
n-+oo
(I f,n -f I ~ &)J=O
... (2)
for every e > o.
To prove that Lim
n-+oo
J In (x) dx= I
B I>
/(x) dx.
 184 THEOREMS ON CONVPRGENCE OF SEQUENCf:S

From (1) and (2),

I/(X) 1 < IjI (x) ... (3)
Moreover, tfo(x) is integrable over E.
Then (3) shows that I(x) is integrable over E.
It means that (In - f) is integrable over E.
Let ~ > 0 be arbitrary and suppose that
An=E (i In-I I ~ 8), Bn=E (I In -II < S).
Then E=An U Bn, An n Bn=0.
The definition (10-1) of convergence in measure ensures that
Lim
n~oo
meA..)=0 . .. (4)
By countable additivity property of the integral,

t; lin-II dX=!An I/.. --fl dX+!BIJ IJ~-/I dx ... (5)

We have I f,,-fl < 8 ¥ X E Bn.
Using first mean value theorem, we get

fBn I In-I I dx < ~.m(B.. ) ~ 8.m(E)

i.e., JIBn I .. -II dx < a.m(E) ... (6)

Take e > 0 and choose 8 such that
E
8.11I(E) < "2
Then (6) reduces t.

f Bn 1 In - I I dx < ~ .. ,(7)
Now 8 is fixed,
or I f~--/I ~ 1fn 1+11 I < q,+1\i=2tjJ
lin-II < ~I\I V X E E.
On integrating, we get

IAn I In-f I dx < ~ fAn q,(x) dx ... (8)

-. 1f..I<y,¥.tEE
-: I\i(x) ~ O.

This ~ IJAn I\i(x) dx \= JA~ tjJ(x) dx.

In this event (8) takes the form
JA.. I I .. -f I dx < i IJAn IjI(X) dx r ... (9)
(4) ~
Given 'I} > 0,3 no E N S.t. n ~ no => I meAn) I < TJ
THEOREMS ON CONVERGENCE OF SEQUENCES 185

=> m(A,..} < 1/

• In (An) ~ 0
Making use of absolute continuity of the integral, we have

IJA" 4(x) dx I< ;

[For this refer theorem (11) of chapter (11) page 210]
whenver m(AII) < lJ for n ;? no'
Using this in (9), we get

fAn I f" - f I dx < ; ... (10)

Writing (5) with the help of (7) and (10), we get

t I f,,-f I dx < ~+; =~

or L, Ifn-fl dx < € for 11 ~ 110

II (fn-f) dx i~ t 1.1;,-fl dx < e

or It (fn-f) dx: < E ¥ n ~ no

This =>

n~
Lim
00
I (f~-t)
E' .
dx=O

or Lim
11-+00
I f,
f;.(x) dx--J f(x) dx=O
E

or Lim
Il-'J>o 00
I E
/;,(x) dx=J f(x) dx.
E
This concludes the proof. Q.E.D.
Ex. State and prove Lebesgue's dominated convergence
theorem. [Banaras 1965, 67. IV 70]
Beppo-Levi's Theorem 8. [Link] < In > be a non-decreasing
sequence of integrable jimctions defined over a measurable set E.
L e1 Lim j'.. b
e 'lI1tegra ble ol'er~. E ~ tlell
'T'I
n -+ (Xl
Lim J'
n -+ CX) E
J,,(x) dx= r
Lim I ..(x) dx.
"' I: tl-'J>o 00
[Meerut 1988, 87; Banaras 69, IV 70; Kanpur 1'14,82,69]
Proof. Let < in > be a non-decreasing sequence of integr-
able functions defined over a measurable set E.
L et Lim f," b
e 'mtegra bl e over E.
11-+00
186 THEOREMS ON CONVERGENCE OF SEQUENCES

To prove that Lim

n->oo B
In(x) dx= J
Lim I .. (x) dx.
B n_oo
f
Since .< In > is a non-decreasing sequence and hence
11 ~/2 ~/3 ~ ..... .
This ~ 11 ~ In ¥ n ~ 1.. - 11 ~ 0 ~ I\In ~ 0 ¥ n,
on taking ~ .. =/n-fl'
Moreover < In > is a sequence of integrable functions
implies that <,p.. > is a sequence of integrable functions.
Finally <o/n> is a sequence of non-negative integrable functions.
Applying this to the Lebesgue's bounded convergence
theorem,
lim f
n .... oo B
.f.
'!In dX= I Bn-oo
lim .1.
'f'n dX

or lim
n ..... oo
J (/. -fJ dx=I
B B n-oo
lim (f.. -/l) dx

or lim
11_00
f In J /1 f n-+oo
E
dx-
E
dx=
E
lim In dx- J 11 d'(
E

or r
n_OO,E
I n_oo'
Lim I ..(x) dx=.
B
lim (,,(x) dx.
Hence the result.
Remark. The theorem 8 is also known as "Lebesgue's mono-
tone convergence theorem".
Ex. State and prove Lebesgue monotone convergence theorem.
[Meerut 1988; Kanpur [Link]. P. 81]
9. Fatou's lemma. Let < f" > be a sequence of non-negative
integrable lunctions delined over a measurable set E S.t.
(l') [Link],,,= ( a.e. on F
n-+cc .

(ii) Lim.
n_oo ml I,,(x) dx <
I 00.

Then
IE .
(x) d" ~ Lim.
n-+oo 1111 E (n(X) dx.
~
J
[!\leet'uf 198.8, 87; Kanpur 81; Bhagalpur 68;
Banaras 64; InQore 78]
Proof. Let < fn > be a sequence of non-negative integrable
fu nctions defined over a measurable set E s. t.
(J') f = [Link] f f,n a e on E
Il-+ 00 .
THEOREMS ON CONVERGENCE OF SEQUENCES 187

(ii) Lim infI. fn(x) dx < co.

n-+-::t:> Ii

To prove that r
Lim infI fn(x) dx.
n_CO
£
f(x) dx
£
~
Define gl"(X)= inf {fn(x)}=inf {fn(x) : n ~ k}
n~k
Then gn(x) ~ fn(x) ¥ n
and therefore t g .. (x) dx ~ \ £ fn(x) dx.

Consequently,
Lim
n-co F.
I
gn(X) dx ~ Lim infI fn(x) dx
n-+ co Ii ... (1)
Since < gn > is an increasing sequence of non-negative
integrable functions and hence by Beppo-Levi's Theorem 8,
Lim
n_cx>
I Ii
gn(X) dx= I F.
Lim gn(x) dx
n-+ 00
=1 n-oo
Lim inf fn(x) dx

=J £ f(x) dx

i.e. Lim
n-+co Ii
I
gn(X) dx= J f(x) dx.
E

Using this in (1), we get

J E
,r(X) dx ~ Lim inf
n-oo
I £
fn(x) dx.
Hence the result.
Ex. State and prOl'e Fatou's lemmaIor integration.
[8bagaJpur 1969 ; 8anaras 64]
Problem 1. State the [Link]'s dominated convergence
theorem and use if to e~'aluate the following integral:
Lim f,,(x) dx J1
n-'>- 00 0
n3 / 2 x
'
where fn(x) = 1 +n2x~' 0 ~ x ~ 1, n= 1,2, 3,... ... ,
[Kanpur 1976, 74, 72]
Solution. For the statement, see Theorem 7.
1 n3 / 2 X2 1
,fn(x)=-. ---:-
x l+n ,. -=·I.(x)
2x J• ~ x '1'. ,
say

Then fn(X) ~ Ij.(x). Also 4(;\) is integrable in [0, 11·

188 THEOREMS ON CONVERGENCE OF SEQUENCES

Hence, by Lebesgue dominated convergence theorem,

lim Jlo fn(x) dx= r

·0
lim fn(x) dx
... (1)

= I: 11:: (r+~~2- )
-= 1: n~~ (In)( .~ :X 112
2 ) dx

=Jlo O. (·~2)
U+x
dx=fl 0 dx=O.
0 ADS.
Problem 2. If rI. > 0, prove that

[Link]+00 J"0(1- nx- )" X'- 1 dx= 10 00

e-rxCl - I dx
where the integrals are taken in the Lebesgue sense.

Solution.
Then fn(x)
Let fn(x) = ( 1-:
~ .p (x) V n where Ij!=e-ID
r x a- 1 • IX

XCI-I.
> 0.
[Kanpur 1973]

Notice that lim (1-

n
11-+ ex>
X)" =e-X.
r} (x) is Lebesgue integrable. Hence, by Lebesgue dominated
convergence theorem,
lIm
Il-oo
J" f,.(x) dx= JOO
0 0
lim
11_00
f,.(x) dx

i.e. lim
Il_ex>
J" (1--::')"
0 n
x"- dx= Joo e- l
0
W X~-1 dx.
Problem 3. Show that the theorem of bounded cotlv.!rgence
applies to .(,.(x) = 1 ::2X2for °< x < J. lKaDpur 1972]
nx
Solution. fn(x) = 1 +n 2x 2
1 ' I
=-1-- '~---1---------2 '-. ~ t
nx +I1X [\/(nx) -V'(IIX)] +2
Take M=i. Then I f,,(x) I ~ M.
Hence, bv Lebesgue convergence theorem,
lim 11
f,,(x) dx=
Il-+ ex> 0
lim f,,(x) dx. Jl
0 11-+00 ..• (1)

L.H.S. of (J)=lim JIo 1+11\ : IlX

1 log (l +n2x2),
dx=lim -2
1/
(form~)
00
THEOREM<; ON CONVERGENCE OF SEQUENCES 189

=Iim Uf(I +~~:~].2nx2 =Iim ~- =0

2 1+ 112X2
R.H.S. Cf(I)=J:lim CI;;X2) dx=I:n~:[M(i7n~+X2)JdX
=[1o (0). (0" x
-.l~ ,
21 dx=fl 0 dx=O. 0

:. L.H.S.=O==R.H.S.
This:;. L.H S,=R.H.S.
This verifies Lebesgue bounded convergence theorem.
Theorem 10.' Let f be a non-negative fLlllction which is integra-
ble over a measurable E. Then given £>0, 3 '>0 s.t. V AcE with

m (A) < S, we have Jf < c.

A [Meerut 1986]
Proof. Let f ~0 ... (1)
defined on a measurable set E. Let ACE S.t.
m (A) < S ... (2)
Case I. When f is bounded on E.
Then 3 M > 0 S.t. I f (x) I~ M ...(3)
This ~ lex) ~ M asf~ 0
Using first mean value theorem, we get
f
.A
f < M.m (A) < M 3, by (2)

or
J .f < M.S
A ... (4)
Given E > 0, we determinc 1) s.L.
i
15 < M or M3 < E.

Then (4) give~ Lf < e

Case II. Whcn f is unbounded on E. We define a measura-

ble function .I;, on £ as follows:
r
. " C\
')={ f(x)
Il
if.f(x)
if/ex) >
<n II

Observe that .I;. (x) ~ f"tl (x)

and 0 /~ (x) < 11. <
It means that < f .. (x) > is a sequence of bounded measura-
ble function defined on E. We write
190 THEOREMS ON CONVERGENCE OF ~E9UENCES

J I=Lim f f,.
E n~oo ...(5)
Here we have In ~ I
and therefore J.4 In ~ LI
or L (I-.f..) ~ 0 ,.,(6)
This ~ given. > 0, EI no E N s.t. n ~ no.

It (I-In) \< E'

or L(f-In») <.' vn~no

or L(I-i < no ) E'
... (7)

Since .rno ~ no and so f.4 f ~ llo

nil meA)

Using (2), JI
.4 no
< llo·8

Taking llo~ < .-, we get

f f <~.
.4 no ...(8)
(7) +(8) gives

L[(r-fnJ+InJ -< E'+."=e, say

or, LI<E
Thus in both cases we have proved that

L/< E.

Problem 4, II -1 <: < 0 a

I: (0, and 1] -. R be defined
by ji) Ixo , 0 < x ~ 1
\X =l
o , x=o
then ver~fy Beppo Levi' s th~orem.

Solution. For each n, we define

In : (0, 1] -> R S.t.
f,,(x) ={xa !f lin ~ x ~ I
o If 0 ~ X < 1/11
THEOREMS ON CONVERGENCE OF SEQUENCES 191

observe that < In > is an increasing sequence of functions S.t.

In-I
and I: In= a~l (1- n~+I) "f nEN
or Io/n
I
~ a+l
1
"f nE N
By Monotone convergence theorem,

J: I dx exists and

Io/=n--:r:> [1o/n=a.f.-T
I Lt I

Note: fis 1,InbQunded and it is not R-integrable over [0, 1]

alth~ugh (R) I: I exists.
EXERCISES
1. State and prove Lebesgue's monotone convergence
theorem. [Kanpur [Link]. P .. 1984,82]
2. If E E M and < In > is a sequence of non-negative
measurable functions such that
lim inff,.(x)=/(x), x E E, prow that
n-+XJ _

f E
Idp. ~ lim infI I" dr"
n-+ co E
[Kanpur [Link]. P. 1981]

3. State and prove Lebesgue's dominated convergence

theorem, [Kanpur 1983]
and use it to evaluate the following integral:
lim 1
J j~{x) dx, where
n_ co 0
n3 / 2 x
In(x)= l+n t \-2' 0 ~ x ~ t, n=l, 2, 3,
[Kanpur 1976]
4. State the Lebesgue's dominated convergence theolem
and usc it to evaluate the following integral:
lim fl n3 '2 x
(,,(x) dx where In (X) = 1 + 2 -2' 0 ~ X ~ 1.
11-+ co 0- n x
[Kanpur 1974]
5. (a) State and prove the theorem of bounded convergence
for Lebesgue integrals. Show that this theorem is not true for
Riemann integrals.
192 THEOREMS ON CONVERGENCE OF SEQUENCES

(b) If ~ > 0, prove that

lim t n(l_~)n xa-1 dX=J"'"
e-1JIj X,,-l dx,
11-+00 1.
n 0
where the integrals are taken in the Lebesgue sense.
[Kanpur 1973]
6. (a) Let < ,f,. > be a sequence of Lebesgue integrable
functions in a set E such that
lim
fn(x) ~ 0 for x E E and fn(x)=f(x)
n-;..oo

almost everywhere in E. Prove that J E

f ~ n-+oo
lim f
E
fn.
Give an example of a sequence < fn in- > for which strict
equality holds in the above result.
Hint. See Theorem 9, Page 186.
(b) State and prove Lebesgue's theorerp of dominated con-
vergence. Show that the theorem of bounded convergence
applies to f, (x) nx
n =1+n20
and the theorem of dominated cqnvergence to
g,. (X)=11 3 / 2 x/(1 +112X2) for 0 E;;; x ~ 1. [Kanpur 1972]
7. Let < fn > be a sequence of measurable functions de-
fined on a set E of finite measure, and suppose that there is a
real number M such that I f,,(x) I .;:; ; M for all n and all x.
lim .
If f(x) = fn(x) for each x In E, then prove that
11-00

I E
f- lim
-11_ 00
J J, E n' [Kanpur 1970]
8. Define the Lebesgue-integral of a functionf, with respect
to the measure Over a given set. Let < fn > be a sequence of
measurable functions defined on the p.-rneasurable space E wch
that °
~ fl (x) ~ f2 (x) E;;; ... ... (x E E).
lim .
If f,,(x)-;..f(x), then show that
11-;" 00

f fn dp. J.f d/-', (n.-~CXj).

<IE
-?
l:.
[Kanpur 1969]

Hint. Refer Theorem 9.

9. If f: R ... R is Lebesgue summable and for
n=l, 2, 3, ....... ..
E.={x E R : If~x) I > n}
 THEOREMS ON CONVERGENCE OF SEQUENCES 193

00
show that the series ~' p. (En) converges. [Banaras III 701
n=1

10. State and prove Beppo-Levi's theorem. [Banaras IV 70]

11. State and prove Lebesgue's dominated ~onvergence
theorem. [Banaras 1965, IV 70)
12. Let < In > be a sequence of real valued measurable
functions on (-00, 00) and -let
00
1: 1/" (x) I=~ (x) E V (-00, 00).
n=1

Prove that

f: L~) In (x) dX=n~ S: I. (x) dx

provided that either side of the equality exists. [Banaras IV 70]

Hint. r/> (x) E V (-00, (0) => ~ (x) is measurablo and
~ (x) is L-intep·able on
( - 00, (0).
13. Let < I .. > be a monotonic non-decreasing sequence of
integrable functions over E and let Lim
n-»oo
In i,
prove that

f 1=E
lim
n-+oo
JB
1ft.
[Banaras 69]
14. State and prove the theorem of dominated convergence.
[Banaras 67]
15. Let fbe a (finite) rea) valued Lebesgue measurable
function on [0, 1] and let
E .. ={x E [0, I]: n-l <:'f(x) < n}
for n=O, ±I, ±2, ...... , show that

1f 1

o
dp. <: 0<", iff
00
1.'
_00
I nip. (Eft) < 00. [Baaaras 66]
 11
Absolute Continuous Functions,
Indefinite Integral and
Differentiation

11'0. Continuous function.

A real [Link] function [(x) is said to be continuous at Xo if
given f > 0,38 > 0 s.t. I f(x) -f(xo} i < Ii

whenever I x'-:"'xo I < 8.

[Baoaras 65,67, IV 70]
11'1. Absolute continuous function. [Meerut 87 ; Kolhapur 70]
Let f be a finite real vilued function defined over a closed
interval [a. b] s.t.

given f > 0, 3 8 > 0 s.t./1' [f(b r )'- f(a r )]

, r=1
I< E

whenever E (br - ar ) < 8, where

r-l

01 < bi E;;;; O2 < b2 ~ ... «;; an < bn •

Without altering the sense of definition, we replace the

condition Ir~l [f(br} --[(or)] I< E by th~ stronger condition.

to

X
r=1
I f(b r) -flor) I < e.

Then the function f(x) is said to be absolutely continuous in

the interval [a, b).
An absolutely continuous function is continuous. Since we can
take the above sum consisting of only one term.
ABSOLUTE CONTINUOUS FONCTIONS, INDEFINITE INTEGRAU 195

11'2. Indefinite Integral.

The fundamental theorem of integral calculus states that
integration and differentiation are reverse processes of one
another.
Let f(x) be inte~rable over [a, b] and let

F(X)=[ f(t) dt+c

¥ x E [a, b].
where c is any constant.
Then the function F(x) is defined as an indefinite integral (or
simply integral of f(x». The term indefinite refers to the variable
x-the upper limit of the integral.
11'3. Differentiable.
If f(x) is a continuous function, then
F(X)=[ f(t) dt+F(a)

is said to be differentiable, F(a) being any finite constani.

it '4. Def. Monotonic functions.
Suppose a function f is defined on a closed intervai [a, b].
Let x. y E [a, b].
f is said to be strictly increasing function if
x > y => f(x) > f(y)
f is said to be monotonic non-decreasing if
x > y => f(x) ;;;;J f(y).
f is said to be monotonic non-increasing if
x > Y => flX) ~ Ji.y).
f is said to be strictly decrea<;ing function if
x > y => /(x) < Jiy).
Increasing and decreac;ing functions are called monotonic
functions. Monotonic functions are a/ways regarded as finite valued
functions.
11'5. Definition: Function of bounded Variation.
[Indore 1978 ; Banaras 67]
A real valued functionf is said to be of bounded variation in
a closed interval [a, b] iff it can be expressed as
f(x)=¢ (x) -1\I(x) '>of x E [a, b)
where -;6(x) and ~(x) are monotonic functions. --....
Alternatively a function of bounded variation is defined as
follows:
Suppose a real valued function is defined on a closed interval
[a, b]. The interval [a, b] is divided by means of points
a=xo < Xl < x 2 < ...... <: x .. =b
"-1
Define v= k=O
1: I !(xk+l)-flXJ;.) I·
196 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

The least upper bound of the set of all possible sums V is

,.. • b
called the total variation of I and is denoted by V ( I)
o

b b
i.e., V (f)=sup V. If V=finite, then
o 11

I is called a function of finite variation or function of bounded

variation. We also say that I has linite variation in [a, b].
11'6. Der. I..ipschitz condition.
A function I is said to satisfy Lip~hitz condition if :I
positive constant M s.t.
I/(x)--I(y) I ~ M I x-y i.
11'7. Del. Lebesgue point. If
lim 1 J,"+I'
h_O II", I I(t)-I(x) I dt=O, then x is said to be a
Lebesgue point of the function l(t).
U·8. Fundamental Theorem or Integral CalcuJus. [Meerut 1990]
If I<x) is continuous and if
F(X)=J"" l(t) dt+F(a)
... (1)
then the theorem states that
F'(x)=/(x). . .. (2)
That is to say differentiation and integra~jon are reverse
processes.
By elementary theory of fntegral Calculus, (2) is proved as
follows:
To prove (2).

u:
From (1),
F(x+h)-.F(X)=I~" I(t) dt+F(a)- l(t) dt+F(a) ]

= f: tll
I(t) dt+ 1: l(t) d(

= J",+II
e l(t) dt.

F(x+h)-F(x) _ I(x) = l.. J.+Ii

l(t) dt-/(x) -.!... ("+" dt
h h ,. h J..
1 f",+11
= II Ie [f(t) -/(x)] dt
ABSOLU rE CONTINUOUS FUNCtIONS, INDEFINITE INTEGRALS 197

!F(X+!)-F(X) -f(x) \=m' \ [+"[f(t)--/(X)] dtl

~1-hIIJ:t"lf(t)-f(X}
- I dt. ... (3)
Since f(x) is continuous and therefore
given, > 0, 3 3 > 0 s.t. I f(t) - f(x) ! < e
whenever I t-x I < ~.
Using this in (3), we get
IF(x+hl=-F(X) -f(x} \< I ~ I J: t
" [Link], IhI< 8

or IF(X+~)-F(X) -f(x} \ < e, I h I < ~.

Making h-,;O and consequently .......0,
lim F(x+h)-F(x) _ /(x) =0 a.e.
h_O h
lim F (x+h)-F(x)
or
h-,;O I,
lex) a.e.
dF (x)
or (ix-=f(x) a.e.
or F'(x}=/(x) a.e. Hence the result.
Problems related to functions of bounded variation.
Theorem 1. A monotonic/unction on [a, b] has finite variation
on [a, b]. That is to say, a monotonic function defined on closed
interval [a, b] is of bounded variation. [Indore 1978]
Proof. It is enough to prove the theorem for an increasing
function. Let f be an increasing function defined on the interval
[a, b] so thatf(x,} <; f(xr+l) for Xr <; xr+l' Divide the closed
interval [a, bJ by means of points
a=xo < Xl < XI < ...... < xn=h.

For f(x r +1 ) -/(xr ) ;;:a, 0

=[ f(x l ) -/(xo)] +[ f(x z) - ftXl)]+'" ... +
[ft x ..) - f\X"-l)]
--f(xn} -f(xv} = /(b) --/(a) = finite number.
[For f is monotonic;? ftb) and f(a) are finite numbers]
Thus V=finite quantity.
This proves that the total variation- is cOJ!stant and is inde-
198 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFlNITE INTEGRALS

pendent of the mode of sub-division. Hence f is of boullded

variation. .
Theorem 2. Iff is of bouneed variation on [a, b], then
V=P+N and P-N=j(b)--f(Q),
where V, P, N, respectively denote total, positive and negative
vari:ltions off on [a, b].
Proof. Let
II-I
V= X I f(xr+t)-f(x r ) I.
r=l

Hence the closed interval [a, b] i:i divi~ed by means of points

a=xo < Xl < Xl < ...... < xn=b.
Let p be the sum of those differences f(xr+.)-f(x,) which
are positive. -n that the sum of those diff~rences which are
negative. Evidently
/
v=p+n, f(b)-f(a)=p -no
From which we get
v+f(b)-f(a)=2p
v-f(b)+f(a)=2n
i.e. v=2p+ f(a) ---/(b) ... (1)
v=2n+ f(b)-f(a) ... (2)
Set P=sup p, N=sup n, V=sup v
where we take the suprema over all possible sub-division of [a, b).
Taking suprema in (1) and (2),
V=2P+f(a)--/(b) .,.(3)
V 2N+f~b)--fta) ... (4)
upon addition, 2V=2P+2N
or V=P+N
(3) ... (4) gives
0=2 (P- N)+2 [f(a)- f(b)]
or f(b)-fta)=P-N.
Theorem 3. To prove that a function f is of bounded variation
if and only if it is expressible as a difference (If two monotonic
functions both non-increasing or both non-decreasing.
'. [Meerut 1988; Banaras 69]
Proof. Let a function f be defined and finite on a closed
interval [a, b] so that f (a) and.f (b) are finite numbers. We shall
show that ,
(i) iff is of bounded variation, then it is representable as a
difference of two monotonic increasing functions.
 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 199:

(ii) iff=g-h, where g and h both are monotonic increasing

functions, then f is of bounded variation.
(iii) Let f be of bounded variation. Divide the interval [a, b]
by means of points
a=xO<x1 <x2 <···<xn =b.

Let V= !: / f(Xr+J-f(X r )/

Let p be the sum of those differences f(x r +1)-- f(xr) which are
positive, - n that the sum of those differences which are negative.
Evidently v=p+n, feb) - f(a)=p-n.
From which we get v+f(b)-f(a)=2p
v-f(b)+f(a)=2n
v=2p+f(a)- feb) ]
i.e. v=2n+ feb) - f(a) ... (1)
Set P=sup p, N=sup 11, V=sup v,
where the suprema is taken over all pos~ ible subdivisions of [0, b].
Taking suprema in (I), we obtain
V-2P+f(a)-flb) ... (2)
V=2N+ftb)--f(a) ... (3)
Further we suppose that Vex), P(x), N(x) respectively denote
total variation, posItive and negative variations off in the inter-
val [a, x], where x :E;; b. With the help of (2) and (3),
V=2P(x)+f(a)-f(x) ... (4)
V=2N(x)+!(x) --f(a) ... (5)
(4)- (5) gives
0=2 [P(x)-Nlx)]+2 [f(a)-f(x)]
or .f(x)=P(x) - N(x)+ f(a).
Taking f(a)+P(x)=P'(x), we get
f(x)=P'(x)-N(x). . .. (6)
It is easy to verify that P(x) and N(x) both are monotonic
inc(c:lsing functions.
P(x) is an increasing function implies that P(x) is also
increasing function.
Now the required result at once follows from (6).
(ii) Letf=g-h, where g and h both are increasing functions.
For any mode of sub-divison of [a, b],
 200 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

-1
V (/)= E I j(x'+l)-/(x,.) I
,.=0
where a=xo<x1 <x2 <.·.<xn =b
I I(Xr -tl)-/(x,,)I=1 [g(x~+1)-h(xr+J] - [g(xr)- hex,)] I
= I [g(X r+1)-g(x,.)]-[h(xr+l)-h(x,)] i
~ I g(X'+I)-g(X,) 1+ Ih(x,+1)-Iz(X,.) I
=[g(x,+1)-g(x,)] + [h(X"+l) -h(x,.)].
For g and h both are monotonic increasing
functions.

... n-l
E If(xr+J-I(x,,)
r-O
I <;;
11-1
E
r~O
[g(X'+l)-g (x,)]

=g(b) - g(a) +h(b) - h(a)

=a finite number.
[For 1 is finite ,valued ~ g and h both are finite valued].
... V (f) <; a finite number.
This. 1 is of boundld variation.
Theorem 4. The su~, difference and product of two lunctions
1 and g 01bounded variations are lunctions 01 bounded variations.
Hence show that L is 01 bounded variation if
g .
I g (x) I ;;;at 0 > 0 ¥ x. [Kanpur [Link]. P. 1988]
Proof. Let/and g be functions of bounded variations on
the closed interval [a, b]. Divide this interval by means of points
a=xO<x1 <xl<'" <xn=b
eo
Denote the total variation of1 in [a, b] by V (/). Then, by
a
assumption
eo
V (I) < finite number}

~
•.• (1)

(g) < finite number

a

(i) Letl+g=s.
To prove that s is ~f bounded variation in [a, b].
ABSOLlJ. E CONTINUOUS FUNCTIONS, INDEFINIE INTEGRALS 201

I S(XHl)-S(X~) 1=1 [ftXHl)+g(X'+l)]-[!(X/c)+g(X k )] I·

=1 [!(X,:+1)-!(Xk)]+[g(X,,+1) -g(X~)] I
~!!(XHl)-!(Xk) 1+1 g(Xk+I)-g(X k) ,

ft-l ft-l
:. 1: I S(Sk+l)- S(S/.) 1 ~ }; I!(X k +1)-!(Xk) I
k=O k~O

71-1
+ J: I g(XHl) - g(X,,)
;:=0

b b b
or V (s)
a
:< V
a
(! )+ V (g)
a

b
or V (S) < 00, by virtue of (I).
a

This ~ S is of bounded variation in [a, b].

(ii) Let d=!-g.
I d(x"+1)-d(Xk) I = I [f(hl·l)-g(XHl)]-[!(x .. )-g(x/.:)l I
= I [!(XHl)-!(X/.:)]-[g(x/.: 41 )-g(xt)l I
~ 1 !(Xhl) - !(x.. ) I + I g(XHl)- g(Xi) I·

... "-1
}; 1 d(XHl) --d(x,,)
k=D

n-l
+ 1:
,,~O
I g(x,:+1) - g(x/.:) I
b b b
or V (d) ~ V (!)+ ., (g} < 00, by virtue of (1).
a a n

a
This ~ V (d) < ex.
u

=> d is of bounded variation in [a, b].

(iii) Let p(x)=!(x) g(x).
To prove that the function p is of bounded variation in [a, bJ.
Write A=sup { I lex) I : x E [a, b]}
B=sup { 1 g(x) I : x E [a, b]}.
I P(XHl) - p(xd 1
= I f(XH1) g( Xl:+l) - !(x,t) g(x,J I
202 ABSOLUTE CONTINUOUS FUNC110NS INDEFlNI1E INTEGRALS

=1 [(XI<+I) g(XHl)-[(Xk) g(XHl)+[(Xk) g(XT'+l)-f(Xk) g(Xk) I

=1 g(X kH ) [[(XkH) -[(XT,)]+[(Xk) [g(Xk+l)-g(X,,)] I
..;;;;1 g(XkH) I . I [(XHl)--[(X T,) 1+1[(Xk) I . I g(XHJ-g(Xk) I
~ B 1/(XT,+l)-/(Xk) 1+ A I g(Xk+l)-g(X k) I
"-1 ft-l
:. E I p (Xk+l)- p(Xk) I ~ BEl [(Xk+l)-[(Xk) I
10=0 k=O

b b b
or
a
Y (p) ~ B V ([)+A V (g)
G a
< 00, by virtue of (I).

b
This => V (p) < qo
a

=> p is of bounded variation in [a, b].

(iv) Let g(x) ~ (1 > 0 1.1 x E (a, b]
Set h=!.
g
To prove that h is of bounded variation in [a, b].
I
11=- , g(x) ~
g
0 > 0 ¥ x E [a, b]

I 1
gx) ~ (1
=> h(x)=-( - > 0 y. x E [a, b].

I, h(xl'+l)-
.
h (x k) /'=1' _ I_ - _1_
g(XHl) g (x,)
(=/' g(Xk)-g(Xkl-l)
g(x) g(XHl)
I'

~ ;o •I g(xk ) -g(XHl) I

:. "-1 I
E h(x',H) -hex,,')
110=0
I
~ 2I
f1
n-l
k=O
1: I g(x;;+t) -h(x,J I
b ,b
or -Y (h) ~ '2 Y (g)
a C1 a
< 00

b
or V (h) < 00.
Il

Accordingly h is of bounded variation.

By case (iii), Jh=J. !..=
g gJ is of bounded variafion in [a , b] •
ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INIEGRALS 203

Theorem 5. Let f be a function of bounded variation in [a, b].

Iff is continuous at X=X o, then the function
0-

'It (x)=V(f)
a

is €ontinuous at x=xo'
Proof. Let a < Xo < b. Divide the closed interval [xo, b] by
means of points
Xo < Xl < xs< .... ····.< x,.=b.
Vfe can find an • >0 s.t.
b b
V <f) > V (f)-. . .. ( I)
Xo "
f is continuous at x=xo ::;. I f(XI)-f(xo) I< e, u,henever
I Xl-xu 1< 8.
--...
From (1),

n-l b
E : RXHI-/(Xk) I +2E=
= 1,=1 V(/)+2e

b b
Thus V(n <V(/)+2E.
Xo Xl

b b Xl
But V (f)=V(f)+ V (f)
Xo

... b
V (f)+ V (/>
Xo
Xl b
< V (/)+2£

Xl
or V (f) « 2e
Xa ... (2)
Xl Xo
1t (XI )-1t (Xo) = V (/)- V (/)
II II
204 AESOLUTE CONTINUOUS FUNCTIONS JNL LFINITE INTEGRALS

XJ b " C
= V (f). For V= V+ V so tha~
a c a

II c II
V-V=V
a a

Xo
or n(xl ) -n(xo) < 26
or I ~(Xl) - 7t(xo) I < 2e wheneverI Xl -xo I < 8
Generalising this,
I n(xk)-7t(XO) I <2E whenever I Xk-XO I < 8.
This => n(x) is continuous at x=xo.
Tberoem 6. An integral is afunction of bounded l'ariation.
An Alternate Statement. If f(x) is a function of integrable on
[a, b] then prove that the function F(x) = J: 1 (u) du, has finite
variation on [a. b). [Indore 1978J
Proof. Let F(x) be an indefinite integral of f (x) defined
over (a, b).
To prove that F(x) is a function of bounded variation in the
open interval (a, b). --
Our assumption implies thatj(K) is integrable over (a, b) and
F(x) = [ fIt) df+F(o) ¥ x E (0, b), ... (1)
F(a) being any finite constant.
Any function/(x) i:; expressible in the form
/(x)=f+(x)-f-(x), ... (2)
where (+(x = J/ x) ¥ X ~ 0
. ») 0 vx<O
and f-(x)=I "t x < a.
-f (x)
0 1-
V x ~ 0
Since f (x) is integ'able over (a, b) and therefore integrating
(2). we get
[ /(t) dt= J: /+(r) dt--J: I-(t) dt V x E (a, b)
Using (I). we get
F(X)-F(a)=[ J+(f) dt-"-i:f-(t) dl "t x E (u, b)

or F(X)=F(!l)+J:f+(t) dt-[f-(t) dt
ABSOLUTE CONflNUOUS FUNCTIONS, INDEFINITE INTEGRALS 205

Taking 1: f+(t) dt=FI(x)

and J: f-(t) dt=F2 (x),

we get F(x)=F(a)+FI(x)-F2(X) ... (3)
V- x E (a, b).
Since the integrand of the integrals
J: f+(/) dl and 1: f-(t) dl
is non-negative and is therefore non-decreasing function of x, i.e.,
FI(x) and FI(x) are montonic non·decreasing functions of x.
Consequently F(a)+FI(x)=Fa(x) (say)
and F2(x) are monotonic non-decreasing functions of x. Then
(3) becomes F(x)=F3(X)-Flx) 'V x E (a, b).
This shows that F(x) is a function of bounded variation in
(a, b).
Problems related to absolute continuous (unctions.
Theorem 7. If I (x) and g(x) are absolutely continuous func-
tions, then their sum, difference and product are also absolutely
continuous (unctions. Further if g(x) does not vanish for any x,
then the quotient B(X»
gx
is a/so absolutely continuous.
Proof Letf(x) and g(x) be absolutely continuous functions
over the closed interval ra, b] so that
given • > 0, 3 8 > 0 s.t •.
n n
X I fib,·)-f\aJ I < e and k I g(bk)-g(a,,) I < E
"=1 11=1

whenever "
l,' (b;;-aJ < 8
~I

V- aI' bI , a2> b 2, ...... , an, b,. s.t.

a 1 < bI ~ a~ < b2 ~ ...... ~ an < b,..
Step (i). To prove that /(x)±g(x) is absolutely continuous
over [a, b].
n
E I f(b,:)±g(b,J -l f(akLf::g(ak)] I
11=-1
206 ABSOLUTE CONTINUOUS FUNCTIONS, IN~EFINiTE INTBJRALS

Finally

From t~ t~required result follows.

Step (ii). To prove that f(x) g(x) is absolutely continuous
over [a, b).
n
1: If(b l.) g(b,)-f(ai') g(ak) I
1:".1

n n
< 1: I RbI:)
lc=]
I . I g(blc) -g(ak) I + 1: I g(ak) I I !(b~) -feat) It
t-l

We know that
absolute continuity => continuity
~ bounded ness
~ f(x) and g(x) are bounded in [a, b]
=> f(x) Ml' g(x) <
M 2 • Of x E [a, b],<
Ml and M2 are upper bounds of f(x)
and g(x) respectively in [a, b].
In this event (I) takes the form

Setting E (I Ml I + I M21 )=~', we get

n
E
1:~1
I Rbi) g(b,J -flak) g(a/;) I < .'.
From this the required result follows.
Step (iii). Let g(x) vanish now here in [d, b].
so that :I a > 0 S.t. I g(x) I ;:. 0 >of x E [a, b].

To prove that !Ix) is absolutely continuous over [a, b].

gt X)
 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINlTE INl EGRALS 207

k=1
i: 1_1___I _1 =
g(ilk ) g(ak)
1
i; I g(bt)-g(at) I
k=l' g(b k ) (a k ) I
<=-
a2

Setting ~=«;', we get

a

n 1 I
~l g(b k ) - i(a~)
1 \ < e'.

This proves that _(1 is absolutely continuou over [a, bJ.

g X)

By step (ii), /(x). _1_ = f(Jf~ is absolutely continuous over

. ~x) ~~ J
[a, bJ. This concludes the problem.
Ex. Iff, g : [0, 1]-+ k are absolutely continuous functions,
prove that f +g and fg are ,absolutely continuous functions.
[Banaras IV 1970]
Solution. For solution refer to proof of Theorem 7, steps (i)
and (ii).
Theorem 8. Let f(x} be tln absolutely continuous function on
the closed interval [a, b]. Let afunction F(y) satisfy the Lipschitz
condition in the ugment [A, B} C [a, b). The composite function
F[f(x») is absolutely continuous in [A, B].
Proof. Let /(x) be an absolutely continuous function on the
closed interval [a, bI then/{x) is a~solutely continuous on
[A, BJ era, b] so that ¥ > 0, 3 S > 0 s.t.
I
E

E f(bk)-/(a~)
11=1
J< 6 ... (1)

¥ numbers 01' bl , ...... , an, b n S.t.

A=al<bl:(a2<b2~ ..... ~n<,bn=B.
n
and 1.' (blt-ak)<i5.
k-l -

_Also let F(y) satisfy Lipschitz condition in [4, B) so that 3 'I > 0
s.t.
I F(Blt)-F(Alt) I :( 'rJ I B,,-A. ,.
Then It~: IF(,f(blt ) -F(/{ak» \ :(\~l \f(bk)-f(ak ) I
<'rJe, on using (l)
208 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

Taking'rj €=fi', we obtain

.£11 F(f(b1c»-F(f(ak» 1< 15', wheneverTc~ (bTc-aTc) < 8

for a finite sequence of pairwise disjoint intervals < (all, bk) >.
This proves that F(f(x» is absolutely continuous in [A, B].
Theorem 9. Every absolutely continuous function is of bounded
variation. [Meerut 1988, 90]
Proof. Letf be an absolutely continuous function on a closed
interval [a, b] so that we can select a 3 > 0 S.t.

~
1c~1
\ f(h k) - ftat) \
,
< 1 whenever E (bk-a1c) < 8
110=1

for aU numbers a1> bI , a 2, bz, ...... a,., b,.

where a=a1 < bi ~ (12 < b2 ~ ...... ~ an < bn=b.
Again divide the closed interval [a, b] by means of points
a=co -< C1 < C2 < ..... c =b
no
in no parts s.t. CkH -CI: < 8.
Consequently for any subdivision of [CTc, CkH]

E I f(XHl)-f(x.) I <;;; 1 where Xt+1> XI E [CTc, CI:+1]

•
i.e. V~~+1 (f) ~ 1.
It follows that
f ...... +V C"o
Vba (f )=V C1 (f)+V c. ()+ (f)
Co C1 C"O_l

~ 1+1+ ...... =n o

i.e. ~! ( f) < 00.

Consequently f is of bounded variation.

Theorem 10. If a function f is absolutely continuous is an inter-
val and iff' (x)=O almost everywhere, thenfis constant.
[Meerut 1986; Banaras 71, 68; Punjab 65, 68; Kol~apur 70]
Proof. Suppose f is absolutely continuous in an interval
[a, b] and f'(x)=O a.e. on [a, b].
 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 209

To prove thatfis cW)nstant in [a, b], we have to show that

f(c) /(a) for any c E [., b].
Let E be a set of those points for which f'(x) =0.
Set E~[a, c].
any x E E => !'(x)=O
=> 3 7], h > 0 s.t. If(x+~~ - f(x) I < 'I

=> If(x+h)-f(x) I< 'I h.

Hence corresponding to each x E E, 3 a small closed interval
[x, x+hJ contained in [a, c] s.t.
If(x+h)-f(x) I < 7] h. . .. (1)
It means that the intervals [x, x+h] cover E in Vitali's sense.
Hence we can find a finite number of pairwise disjoint closed
intervals lk s.t.
/l=(X I, YI], i 2=[X2, Yz].···:·· .. ·, I .. =[x... Yn]
which cover all of E except for a set of measure zero less than a.
where a is pre-assigned number> [Link] to each E in
the definition of absolutely continuity of f If we suppose
X k ~ Xk+l' then

a=yo <;; Xl < Yt xl! < Y2< <...... <

X .. < y" X ..+1=C -<
and k~O I (Xk+ 1 -Ylr) I < I), l~O k(x m ) - /(Yk) 1 < E.

For f is absolutely continuous.

Notice that

,~ k (y,.) --I(x,) I" .:~: (y, - x,). ,,,,,oni;ng:o (I).

< 7j (b-a)

i.e. iI/f(Y/.:) -f(x k ) I< 'I (b -a) ... (2)

Now I/(C)-f(a j
.
1=1 ;,: [f(x +1)-f(Yl:)l+ E[f(Yk)-f(xt)]1
k=G
k
k..o

",~ If{X.H) -/(y.) 1+ !. kLY·l-f{X.) I

< .+'Ij (b-a)
210 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE--INTEGRALS

:. I/(c)-J(a) I < 6+17 (b-a).

But 6 and 'TJ are arbitrary. Hence they can be made arbitrary
small as we please
:. IJ(c) --J(a) 1=0
~ J(c)=Jta). Proved.
Theorem 11. Let a Junction J be integrable over a measurable
set E and let < E It > be a seqilence oj subsets oj E such that
Lim m(E,,)=O ....
k-+oo

Then k-H~) f
Lim E.J(X) dx=O

or equivoimtly IIE~J(X) dX! < E, k ~ ko, whenel'er m(E,,)< o.

(This property of integral is known as absolute continuity of
the integral).
Proof. Let /(x) be.,tategrable over a measurable set E and
let <EIt> be a sequence of subJets of E S.t.
Urn
k -+f>o m(Ek)=O. • •. (1)

,Lim f
To prove that k_do '£ J<x) dx=O.
; Ie

Case I. ~henJ(x) is
Throughout the discussion we shall suppose that
E" C E ¥ kEN.
Slncej(x) is bounded over E and hence we can take
« ~ J(x) ~ (3 on the set Ell.
Using the first mean value theorem, we get

Making k-+oo and observing (1) we get

Lim f
«.0 ~ k-oo EIcJ(x) ~x ~ ~.O

or Lim f
o ~ k-+oo E" J(x) dx ~ 0

This ~ k~:fE".f(X) dx=O

ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 211

Case II. When f(x) is unbounded.

Suppose, without losing generality, thatf(x) ~ O. For if
the result is applicable to positive functions, it is equally applica-
ble to negative functions and therefore it is so by addition in
general case.
Let mEN be arbitrary. Define [f(x)]", as follows:

[/(x)]",= {f{X) when /(x) ~ m

m whenf(x) > m.
Then [.r(x)]", is L-integrable and we define

JIi
.r(x) dx= Lim
[Link]
f Ii
L.r(X)].n dx. . .. (2)

Observe that [.r(x)] ... < f(x)

and therefore JEk [f(x)]", dx <; fE./(X) dx

or f E" {f(x) -[ f(x)]m} dx ~ O.

... (3)
This Q

given E' > 0, 3 ng E N S.t. m ~ ".

lJEk {f(x)-[f(x)Jm} dX\ < Ii'

~ JE. [ f(x)-[ f{x)]", dx 1 < ,f according to (3,1

=- JEJf(x)-[ f{X)ln.l tlx < Ii', ... (4)

in particular_

Since

and therefore JEJ lex) ]". dx < n•. m(E k ). ...(5)

(1) ~ giv<!n 8 > 0, 3 kg E N s.t. k ~ k. =- I m(E,.) I < 3
.. M(Et ) < 3 ... (6)
-: m(E'J.):::O= O.

In view of (6), (S) becomes

JEJ f(x) 1. dx ~ n. m(E,.) < ,...&

212 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

or JEt [ f(x) ]no dx < no 8 for k~ko· ... (7)

Taking noo < ~', we get

JEt [f(x)] no dX < ~" for k ~ ko ... (8)

Adding (4) and (8), we get

fEk {JiX)-[ f(x) lnJ dx+IEt [f(X) ]nodx < ,'+,'.

Setting e=E'+e, we get JEtf(X) dx < II
.. I(x) ~ o.
:. JEk If(x) I dx < e for k ~ /r o• _.. (9)

Now, IIEk/(X) dX\ ~ JEk If(x) I dx < 6

or IJEkf(X) dx I< E for k ~ kJ whenever m(E~) < 8

This => I
k~: E~ f(x) dx=O. Q.E.D. ,
Theorem 12. An integral is a continuous fUllction.
Proof. Let F(x) be an indefinite integral of f{x) defined over
the open interval (a, b) so thatf(x) is integrable over (a, b) and

F(x) = [itt) dt+F(a) ... (*) ¥ x e (a, b)

F(a) being any finite constant.

Firstly we shall establish a lemma:
Lemma. Let f(x} be integrable over a measurable set E and
Lim
let <En'> be a sequencf! ofsub,ets of E S.t. k m(Ek)=O. ·~oo

Then Lim
k-+OQ fEft f(x) dx""=-O·

The pro"r of the lem n 1 start~. For proofreferto Theorem 11.

Q.E.D. (lemma)
Going back to the actual theorem, continued above, we
pwtced as follows to prove it.
 ABWLUTE CONTINUOUS FUNCTIONS, INDEFINiTE INlEGRALS .213

Let x E (a, b) be arbitrary. Let Xl' X~ E (x-i), x+3).

Set E1<:=[x1 , xz].
Then m(Ek)=m ([Xl> x 2]) < 3.
By the lemma, just proved,

IfE f(/) dl I <

I, f!
for k ~ k.
whenever meek) < 3.
In present case this takes the form

if;: fl/) dll <• whenever m ([Xl> XI)) < 3.

In view of (*) this becomes
I F(x2)-FtXI) I < e whenever I Xa-XI I < 8.
Finally,
given • > 0,38> 0 s.t. I F(X2)-F(XI) I < •
whenever I XZ-xi I< 8.
This proves that F(x) is a continuous function.
Theorem 13. An indefinite integral is an absolutely cOfl/inuous
function. [B.n.V. 1967]
Proor. Let F(x) be an indefinite Lebe~gue integral of f(x)
defined over [a, b] so that
F(x) = r: fu) dt + F(a)
... (.)
¥ x E [a, b]
and lex) is integrable
over [a, b].
F(a) being any constant
To prove that F(x) is absolutely continuous over [a, b].
Now we prove a lemma.
Lemma. Let f(x) be integrable over a measurable set E and
lim
let < £1'> be a sequence of subsets of E S.t. k _00 m(E1<;)=O.

Then kLim
_00
I
E 1: f(x) dx=O.

The proof of the lemma starts.

The proof of this lemma is given in Theorem 11.
Going back to the actual theorem continued above, we pro-
ceed as follows to prove it. By the lemma, just proved
214 AESOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

\fEk f(x) dx I< G,

k ~ ko.
whenever m(E~) < (j.
Having this in mind, let us prove the the,orem.
Let aI' hI, a2 , b2 , . " ••• an, [Link] 2n real numbers s.t.
0 1 < b1 <.; O 2 < b2 ~ ...... ~ a. < bn •

" n
Let m(Ek )= E m(b"
• _1
0,)= ~ (b.-a.)
'~1
< S, where l) > O.

Then \ 1: fb,a. f(t) dt

0=1
< E

i.e. I~ [F(bi)-F(Ot)]/<E
. 0_1

according to (*)
We have the final result as

given 6 > 0,38 > 0 s·t.1 i

.=1
[F(b.)- F(a,)] 1< e

n
whenever 1,'
i_t
(bi-o,) < ~

for all values of 01> b1 , a 2 , b2 • ...... , an, b•• s.t.

a1 < b1 ~ a 2 < b 2 ~ ...... ...;; On < 5...
This proves that F(x) is absolutely continuous function.
Problems related to indefinite integral.
Theorem 14. Every absolutely continuous function is an inde-
finite integral of its own derivative. [Banaras 1967]
Or To' show that F(x) is an indefinite integral if F(x) is absolu-
tely continuous. [Banaras 1965]
Proof. Let f(x) be an absolutely continuous function ill a
closed interval [a, b] so that

f' (x) and J: f' (I) dt

exists finitely ¥ x E [a, bJ.
Let F(x) be an indefinite integral of rex), so that
F(x)=/(a) + 1: f (1) dr, x E [0, b]
... (1)
To prove that F{x)~/(x).
ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 215

By Theorem 13, F(x) is absolutely continuous in [a, b].

Also (I) ~ F' (x)=1' (x) a.e.

From which fx [F(x)-f(x)]=O.

Upon integration, F(x)--f(x)=c (const.) ... (2)

Tak,in g x=a in (1), F(a) =f(a) + J: I'(t) dt=f(a)

i.e. F(a)=f(a)
or F(x)-f(x)=O for x=a.
Subjecting (2) to this condition, we obtain O=c.
Then (2) is reduced to F(x) - f(x)=O a.e.
i.e. . F(x)=f(x) a.e.
In this event (l) proves that f(x) is indefinite integral of its
own derivative.
Theorem 15. A necessary and sufficient condition that a func-
tion should be an indefinite integral is that it should be absolutely
continuous. [Meerut 1986 ; Punjab 66; Banaras 67, 71J
Proof. (i) Let F(x) be an indefinite Lebesgue integral of
(x) defined over [a, b] so that

F(X)=[ [(I) dt+f(a)

... (*)
. ¥ X E. [a, b],
and f(x) is integrable over [a, b],
F(a) being any constant.
To prove that F(x) is absolutely continuous over [a, b1.
For proof refer to Theorem 13.
(ii) Let f(x) be an absolutely continuous function in a
closed interval [a, b] so that
III

I' (x) and J" f' (f) dt

exist finitely V x E [a, b).
Let F(x) be an indefinite integral of I (x) so that

F(x)=f(a)+J"'" f' (f) dt

where x € [a, b].
 216 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

To prove that F(x) f(x).

For proof refer to Theorem 14.
Theorem 16. Let F(x) be an indefinite integral of a bounded
measurable function f(x). Then
F' (x)=ftx) a.e.
An alternate statement. If f is a bounded and measurable
function on [a, b] and
Ftx) = 1: f(t) dt+F(a)
then prove that F' (x) .... f(x) a.e. in [a, b].
[Kanpur 199O,Meerut 1988,87, 86 ; Punjab 67, 65 ; Banaras 67]
Proor. Let F(x) be an indefinite integrvl of a bounded
measurable function f(x) so that f(x) is integrable and
F(X)=[ f(t) dt+F(a),
... ( I)
F(a) being any finite constant.
To prove that F'(x)=f(x) a.e.
By hypothesis f(t) is bounded and therefore we can write
I f(t) I <; "'I,
where M is an upper boudd of flt).
From (1),
F(x+hl-·F(X)= ! U:+Tl [(t) dt-:-F(a)- J: /(t) dl- F\a) ]

== ! [J:+~ f(l) dt ]
Taking modulus of both sides
IF(X+h~-F(X)[=m I J:H f(t) dr i
1 JIII+I>
~ ThT I f(tJ I . I dt I
OJ

AI JIII+I> AI
~ThT 01 Idt/=ThT·1hl=M.
:. IF(X+~-.F(X)1 ~ M
... (2)
F(x+h)-F(x) I .
and h -+ F (x) a.e. as.h~O.

Let h take a [Link] of values tending to 0, i.e. It-.O as

n~oo.
 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFiNITE INTEGRALS 217

Let C E [a, bJ. By Lebesgue bounded convergence theorem,

fao I'1m [F(X+h>-F(X)]d X= J'1m JC F(x+h)-F(x)d
----- x
h a h

or J: F'(x) dx=lim (} J: F(x+h) dx- ~ J: F(x) dx 1

=Iim u- [:: F(I) dt+} J: F(x) dx 1
on putting x +hr=- t

=lim ~ U:H F(t) dt+ J:-Ih F(t) dt ]

. as jb.. F(x) dx= Ib F~t) dt
,a

=lim lll.
rl J~h F(I) dt-ll
1 J~h
a F(t) dt
]
=F(c)-F(a)

= 1: I(x) dx, by (1).

[See Remark written, just below]
or Ja" [F' (x)-/(x)] dx=O '" c E [a, b]

This => F'(x)--J(x)=O ~ F'(x)=/(x) a.e.

Remark, If we write g'(t)=Ft1). then

li~ ~ [+'- F(I) df=lim } 1:+'- [~l get) ] dt

I [
=Iim /, g (x) Jc-c+'-
-I'
- 1m j?'(c-l-h)-gi
II
C _ ' ( ) - 1:'( )
- )--g c -C' C ,
Theorem 17. III: [a, b] .... R is all increojillg lUi cticn (non-
decreasing). then
I I : [a. b] .... R is L-integrable and
I: I '(x) dx ~I (b)-I (a).
[Punjab 65, 67; Banaras 69, 71J
Proof. Consider a sequence < In > of non-negative func-
tions I. : [a, bJ~R defined as
In(x)=n [ I (x+}) ---/(x) 1'" x E [a, b].
218 ABSOLUTE CONTINUOUS FUNCTIONS, INDEfiNITE IN1EGKALS

Since I ..(x) is an increasing function and hence it is L-inte-

grable. The sequence < In > converges to I (x) a.e. I

Lim .
i.e. f'(x) = I .. (x) a.e.
n-+oo

Applying Fatou's lemma. we get

I b
..
f'(x) dx ~ Lim inf
n-+ 00
fibQ
I ..(x) dX} ... (1)
Extend the definition of the function I by setting
l(x)-4(b) If b <: x ~ b+l

Since the integral

.J: I .. (x) dx is monotonic function of n and therefore it can

be taken in the sense of Riemann. Consequently
Lim inf
n-+oo
1 b
0
I,.(x) dx

=n~: [1: I(x+~)-/(X)] dx

infn

=n~: inf n
[J: I lx+~ ) dx--J: I(x) dx ]
= Lim inf II [f~-tl/" Itt) dt--JI I\x) dx ].
n-+ 00 Ja -1'" a

. x+-=t
on puttIng 1 fi
. t h erst
m integral
n
Lim
= ,,-+ 00
inf II [jt+l/" IlX) dx+
0+1."
Ia
li
I(x) dx 1
Lim
== n_oo ipf n [fD+l/"
b
I(x) dx+ Ia
a+l/"
I(x) dx ]

= Lim inf n [Jb+l/n I\x) dx - 1'1+1/" I(x) dx ]

n-+oo b /J

[1
[Link] n - [/\b)-'Aa)]
= n-J>oo n
]
by remark of Theorem 16. P. 217.
=/tb) - I( '1).
Finally,
Lim inf Jil I,,(x) dx=/tb)-/~a).
n-+ 00 /J
..
ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 219

Using this in (I), we obtain

J: rex) dx :S;" f(b)--f(a).

Proved.
Theorem 18. Fundamental theorem of integral calculus :
If f is an integrable function its indefinite integral F(x) has
almost everywhere a finite differential coefficient
or Iff is an integrable function on [a, bJ and if
F(x) = 1: F(t) dt+F(a).
then F'(x)=f(x) for almost everywhere on la, b].
[Banaras 1967]
Proof. Letfbe an integrable function on [a, b] and let
F(X)=jlDa f(t) dt+F(a),
... ( I)
1'(0) being any constant.
To prove that F (x)=f(x) a.e. 011 [a, b).
Without loss of generality, we may suppose that f(x) ~ 0 tv x. Let
m be a positive integer and define a function [f (Xj]". as follows
If(X)Jm={f~X) [Link](x) ~ m
m Iff~x) > m
Evidently [f(x)l", ~f(x) in every case
so that Ax) - [f~x)]m ~ O.
Consequently 1: l f(l)-[/(I)].,.] dt=G",(x) __ (2)
is an increasing function of x. As a result of which Gm{x) has
non-negative differential coefficient.
Since [fV)]". is bounded and measurable and so
~x 1: [f(t)Jm dt=[f(x)]". ... (3), by theorem 16.

By (2), J: f(t) dt=G".(x) +J: [f(l)],,, dt,

F(x)-F(a)":'Gm(x)+ J: [f(t)l", dl, by (I).
Differentiating w.r.t. x,
i.e. F'(x)--O=G'.,.(x)+[f(x)]m, by (3)
~ [f(x)].,. as G'".(x);;;;r o.
:. FJx) ~ [f\x)]m a.e.
But m i, arbitrary and hence m~ OJ, give~
':F'(x)'~ f(Jc} ... (4)
 220 ABSOLUIE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

But 1: f(x) dx=F(b)-F(a), by (1)

Also J: r(x) dx=F(b)-F(a).

The last two equations show that

J: r(x) dx= J: f(x) dx

or I: [r(x)-f(x)] dx=O.
Also F'(x)-f(x) ~ 0 a.e. by (4).
Then we must have
F'(x) - f(x)=O a.e. or F'(x}-f(x) a.e.
Problems related to Lebesgue point of a function.
Theorem 19. Let x be a Lebesgue point of afunctionf(t).
Then show that the indefinite integral

F(x) =f(a) + 1: f(t) dt

is aijj'erelltiable at cach point x and r(x)=f(x).
Proof. Let x be a Lebesgue point of the functionf(t), so
that
Lim I J",+h
h-+O 11", I f(t)-f(x) I dt=O ... (1)

Also let F(x)=fw)+J"'o /(t) dt

... (2)
To prove that F'(x) =f(x).
F (x) =Lim F(x+h) -F(x)
Since h ~O It
Hence if we show that
Lim \ F(x+h~ -F(x) _ (x)
11--+ } " .
1=0, ... (3)
the result will follow.
From (2).

F(x+h)--F(x)= ",+A
a f(f) dt- I'"0 /(t) dt
J
=.j ",+A j. t) dt+ JOII f(t) dt= "' 1·+" /\t) dt
ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS 221

:. !I~±h)h =- ~i~) = h~ I"'+" :II

f(t) dt
••• (4)
1 J,"+II f(x) dt=h-1 f(x)"J,"+h
Also 11:11 I
dt=h f(x) [1'"+"
t '" =f(x)

or 1 JGI+"
II 81 f(x) dt=f(x)

or 1 JGI+"
f(x)=h 81 f(x) dt . (5)
Subtracting (5) from (4),
IF (X+~-F (x) -- f(x) \=1 ~f:+h [[(t) - f(x)ldt \
1 £+TI •
~ h-I., 'f(t)-f(x)! dt.
Making h-+O and using (I), we get

,1-+0
I
~im F(x+h~--F(x) -f(x) ~ 0
h
I ~ ... (6)
But Lim
h ...O,
I~(x+h)-F(x)
hi?'
-f(x) I ~ 0
... (7)
Since modulus of any quantity is always non-negative.
Combining (6) and (7), we get the required result.
Theorem 20. Every point of continuity of a summable function
f(t) is a Lebesgue point off(t).
Proof. Letf(t) be sum mabie over the closed interval [a, b].
Also letftt) be continuous. at t=xl •
To prove that Xl is a Lebesgue point of the functionf(t).
Recall thatftt) is summable over [a, b] if

Lb
f(t) dt=a finite quantity
•

f(t) is continuous at f=X l implies

given e > 0, there exists 8 > 0 such that
'f(t) -/(.'"1) , < t whenever I t - Xl I< 3.
From which we get

IXl+h
Xl
\
I fll)--/tXl) I dt < •
jX2+h
Xl
dt=f,h

1 fXl+h
or II Xl I f(l) -f(x!) I dl < E
... (2)
whenever I h I< 3
 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE INTEGRALS

i.e.

Combining the last two inequalities.

Lim I.!-.JXI+h
h.-+-O 'h Xl
I f(t)-f(x l ) I dt 1=0
i.e. Lim ~ IXI+h I f(t)--f(x l ) I dt=O
h-+O h Xl
showing thereby Xl is a ~ebesgue point off(t).
EXERCISES
1. What is an absolutely continuous function? If I is abs~.
lutely continuous on [a, b] and
f '(x)=O a.e. prove that I is constant.
[Banaras 1971; Meerut 72]
2. (a) Let f be an increasing real valued function on [a, b].
Show that f is Lebesgue measurable and that
J: f' (x) dx ~ feb) -f(a)
Letf: [a, b] _ R be such that
(b)
I ./(y) - f\x) I <;; k (y-x)
whenever a ~ X < Y ~ b.
Prove that
J: f' (x) dx=/(b) -/(0) . [Baaaras 1971]
3. Prove that if f is absolutely continuous, then f'(x) exists
almo~t everywhere. [Baoaras 1971J
4. Show that the function f defined on lO, 1] by
Ii x>={X cos (1tX/2) for 0 < ~ EI; 1
(0 for x=o
is continuous but not of bounded variation on [0, 1].
[Baaaras [Link]. Final III 1970]
ABSOLUl E CONTINUOUS fUNCTIONS, INDEFINITE INTEGRALS 223

5. Define 'absolute continuity' of a function

I: [0, I ]-+R.
If/, g: [0, l]-+R are absolutely continuous, prove that/+g
and Ig are absolutely continuons [Banaras [Link]. Final IV 1970]
6. Prove that a monotone function is differentiable 'almost
everywhere'. [Banaras [Link]. Final IV 1970]
7. Let [a, b] and [c, d] be compact intervals of the real line.
Let/b~ a [Link] summable real valued function on [c, d].

Defining F(x) = f: .r (U) ~ x ~ d)

du (c
prove that Foi/J is .absolutely continuous, .p being an absoulte
continuous function on [a, b] into [c, 4J.
[Banaras M~Sc. Final IV 1970]
Hint. Refer Theorem 8, Page 207.
8. Prove that if.r (x) is absolutely continuous in an interval
and I (x)=O almost everywhere, thenl (x) is constant.
Let E be a measurable set and let E" be the linear set which
is the-section of E by the ordinate x from the origin. Then show
that E~ is measurable for almost all x and

" (E)= f m (E~) dx,

where'f' denotes the plane measure. [Banaras 1968]
9. Define (a) function of bounded variation and (b) abso-
lutely continuous function.
Prove that a necessary and suffi;;ient condition that a function
should be an indefinite integral is that it should be absolutely
continuous.
10. Show that if
F(x)=F(a) + J:.r (t) dt
then F(x)=I(x) almost everywhere. [Banaras 1967]
Hint. See Theorem 16, Page 216
ll. Define absolute continuity for a real function of a real
variable. Show that F (x) is an indefinite integral if F is absolutely
continuous. [Bansras 1965]
n. State, without proofs, the following theorems: and
explain all the concepts occurring in the statements:
(i) Fubini's Theorem,
(ii) Radon = Nikodym Theorem. [Banaras 1964}
224 ABSOLUTE CONTINUOUS FUNCTIONS, INDEFINITE IN rEGRALS

13. Explain what do you understand by f(x} is an absolutely

continuous function on an interval (a, b]. If f is integrable on
[a, hI, show that

F(x) = I: f(t} dt, a < x< h

is absolutely continuous on a given interval (a, b].
Give an example of a function which is continuous but not
absolutely continuous on a given interval. [Kanpur 1970]
14. State and prove Vitali's covering theorem and use it to
prove that an increasing real valued function on the interval
[a, b1 is differentiable almost everywhere and that

r. f'(x) dx ~ f(b)-f (a). [Punjab 1967]

 12
LP-Spaces

12'0. Conjugate number.

Let p, q be any two positive real numbers S.t.
(i) P > I (ii) ~ +.l= 1.
P q
Then q is called conjugate to p.
If p=2, then q=2, so that 2 is self-conjugate number.
12'1. V-Space.
By LV (a, b) we mean a class of functions I(x) S.t.
(i) J<x) is meltslIrable over (a, b).
(ii) / f /P is L-integrable over (a, b) for P > 0,
i.e. J: I f II' dx < 00 for P > O.
Sometimes we denote the class of such functions by the
symbol LP in case the mention of interval is not necessary.
Then [1 or simply L denote<; the class of measurable func-
tion I(x) which are also L-integrable.
Examples. (i) If I(x), g(x) E LV (a, h), then
j(x)±g(x) E U (a, b),
where (a, b) is finite or infinite.
(ii) If I(x) E U' (a, b), I(x) E U (a, b), then
I(x) E U (a, b) if p < r < q.
(iii) Every bounded function bebngs to Lv (a, b) >.f p,
where (a, b) is a finite interval.
(iv) The spaces LV (a, b) and Lv (a, b) are said to be conju-
gate to each ot her if ~
p
+ ~q = I.-
12'2. Norm of an clement of LP-Spaee.
Let /(x) E /.1' (a, b) be arbitr~ry.
The norm of (V':), denoted by /I f lip, is defined as follows:

ilfl!~-(J: Ifl pdPf 1

•
 226

If p=J, then we write 11/111=11/11=1: III dx,

where (a, b) is finite or infinite interval.
12'3. [Link].
'Let I(x), g(x) E LP (a, b) be arbitrary.
We define d (I, g)=11 I -g II,,·
Then d is called distance function or metric on L'P (a, b),
d (f, g) denotes the distance between the functions/(x) and g(x).
12'4. Convergent sequence.
Let < In > be a sequence of functions belonging to £P-
space. The sequence is said to converge in mean with index p, if
given II> 0,3 no E N s.t. m, n ~ no => 111m (x)-In (x) lip < •
or if f I I,,(x)-/m(x) IP dx~O as m-+oo and n~oo.
In this ca'le the sequence is called meanlundamental sequence
Or Cauchy sequence.
The sequ~nce is said to converge in mean with index p to a
function I(x) if
given I! > 03 no. eN
s.t. n ~ no => If/-I,. lip < 6
or if II 1-/n'IP dx~(} as n_oo.
12·S. Definition. Cauch; sequence.
Let < lro > be a sequence of functions belonging to a L'P
space, This sequence is said to be a Fundamental sequence or
Cauchy sequence if
given II> 0, 3, no E N s.t. m, n ;iii no :0> iif.. -/.... \lp < •.
Later on we shall show that every convergent sequence is a.
Cauchy sequence. (Refer Theorem 13, Page 241).
11'6. Definition. Completeness of L'P·space.
An LP-space is said to be complete if every Cauchy sequence
in tne space is convergent at some point of the space,' i.e. for
every Cauchy sequence < I .. > in the space, there is an element
lin the space s.t. /.,-+1.
A complete normed linear space is called Banach space.
Theorem I. II IE Lp and g ~ I, then g E V,.
Proof. Let IE Lp (a, b), so that
(i) I is measurable over (a, b),
(ii) I/I P is integrable i.e., J: I/P'dx < 00.

Also let g(X) <; Ji.x).

[!>·SPACES 221

To prove that g E U' (a, b), it suffices to show that

(iii) g is measurable over (a, b)
(iv) I g 121 is integrable over (a, b), i.e.,
I: !g II' dx < 00.

Let at be any real positive number.

{x E (a, b) : g(x) > at}={X E (a, b) : CIt < g(x)}
={x E (a, b) : at < g(x) <; I(x)}
={x E (a, b) : CIt < I(x)}
=measurable set, according to (i)
Finally, {x E (a, b) : g(x) > ot}=rneasurable set.
This => g(x) is measurable over (a, b). Hence the result (iii).
•• g(x) ~ I(x)

. J:gig ;" dx ~ • If) If 121 dx < 00, by (ii)

or
J" I II' dx <g 00.

This proves the result (iv).

)t Theorem 2. lIfE L'P (a, b), p > 1 ; then IE L (a, b).
Proof. Let f E L" (a, b), p > I.
To prove that IE L (a, b), it is enough to prove that

(i) {is measurable over (a, b)

(ii) 1 I I is integrable over (a, b).
By hypothesis,
(iii) I is measurable over (a. b).
(iv) I 1/1' is integrable over (a, b).
Let E=(a, b).
Clearly (iii) => 0).
Let A=E( III ~ I), B=E( III < 1).
Then E=A U B, A n B=0.
By countable additivity property of the integral,

L, !I I ~x LII I +L, II I
= dx dx ... (1)
I ft x ) I < 1 'ff x E B.
Using first mean value theorem.
, I/(x) I dx < m(B)=a finite quantity.
B \
I:his proves that I/(x) I is integrable over B. . .• (2)
228
.: f(x) ~ 1 ¥ x E A.
:. / f / ~ / f III on the set A. For p > 1.
Integrating, we get
LIf I dx ~ LI Fill dx < 00, according to (iv)

i.e. LlfldX<oo.

This proves that I f I is inte1rable over A. . .. (3)

From (I), (2) and (3), the result (ii) follows.
Theorem 3. If fE 0', g E L'P, thenf+g E P.
[Meerut 1988]
Proof. Let f, g E LP (a, b) so that
(i) f, g are measurable over (a, b)
(ii) I f ill, I g III are integrable over (a, b), i.e.

I: Ifll> dx < co, J: I gill dx < co.

To prove that f+g E P (a, b), it suffices to prove that

(iii) f+g is measurable over (a, b).
(iv) If+g III is integrable over (a, b), i.e.,
J: If+g IP dx < 00.

Evidently (i) ~ (iii).

Let £=(a, b).
Let A=E( If I '" / g I), B=E( if I > I g I)·
Then A n B=0, E=A U B.
By countable additivity property of the integral

L. I fi-g ill d~=-:'L I f+g ill dx+ t /f+g /11 dx ... (1)
I f +g II' ~~ ( I f I + I g I )Il
~ ( I g I + / g / )1l=2 P / g ip on the set A
Integrating, we get

LI f+g III .Ix ~ 2p LI Kill dx

< 2 p x oo=oc, according to (ii)

i.e. L If+gjI'dx < 00
... (2)
Again / f + g /1' ~ ( I j I + I g / )7'
<" 21' I r I'. on the set B.
LP·SPACES 229
Consequently

I. l/+g I" dx < 2" Is If I" dx < 22'xoo=00, by (ii)

i.e. Js lf+gl2'dX<oo. . .. (3)

In view of (2) and (3) : (I) shows that
t I/+g 12' < 00 or J: If+g I" dx < 00.
Hence the result (iv).
Tbeorem 4. Holder's inequalities.
Let lEU (a, b), g E L<l (a, b), p > l.
Let U (a, b) and L'l (a, b) be conjugate to each other.
Show thatfg E L (a, b) and II/g 1/</11.. " g 1/'1. [Meerut 1986]
An Alternate Stat~ment. lip and q are non 'negatil'e extended
I I
real numbers such that -+-=
p ,q
I

and ifIE LP, g E L'l, then f.g E V and

J/fg I ~ (Jlfl fill (II g I Clr

P
q

with equality only when dl:EJiJ

a.e. [Banaras i969]
Proof. Letf E LP (a, b), g E U (a, b), p > l.
Let £1' (a, b) and La (a, b) be conjugate to one another
1 I
so that - + - = I.
p q
To prove that (i) Ig E L (a, b).
(ii) 1/ fg II ~ /I f 11,,·11 g II"
Let 0 < at < I. Consider the function
.p(x)=x«- (XX for O<x< 00.
It is ea~y to see that ljJ(x) is maximum for X= I. It means
that ljJ(x) ~4(1) for O<x<oc..
i.e., XX- aX ~ I~--at
i.e., x%-17x ~ I-at.
A
Putting x = B ' A>O, B>O, we get

i..)« -at (~) /' ) -a.

(B or A«B-z-« ~_./ 1-.
B ~, 'B~
or A«Bl-Ql-rI. A ~ (l-':'rI.) B
or [Link]-« ~ Aat+B (I-rl.) ... (1)
230 UJ-SPACES

This inequality has been proved under the condition A > 0,

B > 0, but (1) also holds when A and B or both equal to zero.

Setting ac=!...
p
so that p > I. c: at < I)
1 1
Therefore ac+-=
q
1, or I-cx=-
q
A B
Then (1) beeomes AlfP Bl/v ~
""'p
-+-
q ... (2)
If either f(x) or g(x) is equivalent to zero, then the theorem
is trivial. Excluding this trivial case we assert that both the
integral I J: If I" dx and J: 1g Iv dx
are stric.t1y positive. Then each of the following functions has a
meaning.
• f(x) g(x)
Wnte F(x) = II f II;' G(x)= Wg~
If we set Al/"=I F(x) I, Bl1a=1 G(x) I,
then (2) reduces to
I F(x) G(x) I ~ If(x!1'+1 G(x) Iv
p q ... (3)
Integrating it,

J"I F(x) G(x) I dx ~ p~ J" I F(x) IV dx+l.q J" I G(x) I" dx /

o a a

=!...~" "Ifl"dx +C blgllldx

p
a
J If 11> dx J Jig I" dx
a (I a

=-
1 J: If
II
111 dx
+-
1 I: II
\g Iv dx 1 1
=-+-=1
PIG I f Iv dx q J 1g Iq dx P
a q

or J: 1 F(x) G(x) I dx <1

or
J:II \/(x) g(x) 1 dx
f II" II iY; . - - ..; ]
.. (4)

or IIfg II ~ IIfllpII, g IIv. ...(5)

This proves the second required result.
f E L" (a, b), g E La (0, b) implies
 UJ-SPACES 231

(iii) I and g are measurable over (a, b).

(iv) II 121 and I g Iq are integrable over (a, b) i.e.

J: I lip dx < 00, J:

1 g Iq dx < 00,

i.e. U: II 121 dx fP < 00, U: I g Iq dx ) l/q < 00

i.e. " I II, < 00, II g IIq < 00.

Using this in (5), we find that Il/g II < 00
This => fg is integrable over (a, b).
(iii) => I(x) g(x) is measurable over (a, b).
The last two statements taken together imply the result (i).
Theorem 5. Minkowski's inequality. Iff. g E LP (a, b) where
p> 1, then
11/+g I, ~ II 111,+11 g 1121'
(Burdblwan 1990,Meerut 1988, 87 ; Banaras 70, 68; Indore 78J
Proof. Let/, g E £P (a, b), p > 1.
To show that 0) I+g E Lp (a, b)
(ii) 11/+g II, < 111/1.+11 gil,·
Let q be a number conjugate to p, then L +!.
p q
= 1.
I, g E £P (a, b) .,. I+g E LP (a, b) Refer Theorem 3.
and J, g E LP (a, b) => I+g E LV (a, b)
• (/+g)plq E Lq (a, b).
Now 1 E £P (a, b), (f+g)Plq E U (a, b).
On applying Holder's inequality to these functions, we get
I: If 1·1 J~ g I plq dx ~ (J: III" dx rIP U: 'f+g I(P/q)q dx )"'1
or f:[Link]+glp,qdX.~ O:I/PJdx f'p u: I/+g Iv dxf/'l
... (1)
Since /, g E LV (a, b) and therefore interchanging f and g in
the last inequality, we get
J: I g I·I[+g Iv/ q dx ~ (J: I g Iv dxf'V U: If+g IV dx/,q
... (2)
Adding (I) and (2), we get
J: If ,., [+ g IV;q dx +J: I g 1·lf+g Iv/q dx
 232 OJ-SPACES

~ [(J"OI/IJ>dx )1/11+ (Jbo,gl1l dX )1/11] ((bJ.I/+gI1ldx )1111

... (3)
..• p~+~=l
q
:• l+I!.._p
q .
So tong as I/+g I'=I/+g 1.I/+g 111«
1

~ (1/1+lg 1).I/+g 111111

=1/1.I/+g 1"}I1+1 g '.I/+g Ipla.
Integrating, we get
J: II+gIPdx~ J: 1/1.I/+gllllqdx+J: Igl·l/fglPlqdx
On using (3), we get

J: I I+g P' dx " [0: I I I" dx rIP

+0: 1 I" dx tv J(J: if+g dx r
g IP ll

Dividing by (J: 11+ g Ip dx til and observing that

1 1
1--=-, we get
q ,p

0: I/+g!, dx r'p~(J: 1/111 dx Y'' +0: I g I" dx f'P

or II/+g II, ~ 11/11,,+ II gil,,·
Ex. State and prove Minkowski's inequality lor V'-spaces.
[Banaras [Link]. Final IV 1970]
Theorem 6. Schwarz's inequality. II f, g E L2 (a, b); then
Ig eLand IIlg 112 ~ 11/1/2 II g 112'
Proof. Suppose I, g E LI (a, b) so that
(i) f and g are measurable over (a, b)
Oi) 1I 12, I g 12 are integrable over (a, b), a.e.,
J: I j(x) dx < f: g(x) dx <
12 00, I 12 00.

This => J: I j(x) I~ dx+ I: 1g(x) Ii dx < 00.

. .. (1)
To prove that
(iii)Ig E L(a, b)
(iv) IIlg II ~ 11/112' II g 112
(i) => fg is measurable over (a, b) ... (2)
O'-SPACES 233

Suppose x E (a, b) is arbitrary.

Then obviously ( I f(x) /-1 g(x) I )2 ~ O.
This gives 2 I f(x) 1.1 g(x) I Et;; I f(x) 12+1 g(x) 12

or n f(x) g(x) I dx ~! U: 'f(x) 12 dx+ J: I g(x) /2 dx ]

By virtue of (1), this ~ I:, f(x) g(x) I dx < 00

This => 'f(x) g(x) , is integrable over (a, b) ... (3)

From (2) and (3), the result (iii) follows.
Let at be any real number. Then
(II If' + I g I )2 ~ 0
or at 2 If 1
2 +211 I fl· I g I + I g 12 ~ 0
or oe 2 J: If 12 dx+2:x J: Ifg I dx+j: I g 12 dx ~ 0
or a 2p+aq+r ~ 0 ... (4)

where p=J: [Link] dx, q=2 J: Ifg I dx, r=J: Ig 12 dx

If p=O, i.e, if J: If /2 ~x=o then If 12 =0

or /(x)=O a.e. on (a, b)

In this case both the sides of the inequality namely

are Zero.
I: Ifg I dx ~ U: 'f 12 fa 0: I dx
Excluding this trivial case we suppose that p:;toO.
g /2 dx t2
Set f1.=~q in (4). Then ~!:+ ( ;pq) q+r ;.. 0
q2 q2
or ----+r ~ 0 or -. q2+4pr ~ 0
4p 2 p ' c?'

or q ~ 2pl'2 r1/2.
Putting the values of p, q, r we get i

2 J: IlgldX~2(J: Ifl t U: Igl~dx fa 2 dx

2

or J: I I ~ (1: If t2 0: / t2
fg dx 12 dx g 12 dx
.(5)
This is also expressed by writing
IIjg II ~ II I liz II gl!z. Hence the result (iy).
234

Remark. This theorem IS also expressed by saying that:

Let f and g be measurable functions on [a, b]. If f and g are
square integrable, then fg E L [a, b] and

[Banaras 1971]
Solution. First write the proof of the previous Theorem 6.

Then 1: 1 fg I dx ~(J:
Since If 12-P and het'lce the last gives
If 12 dx r 0: /2
I g 12 dx r~

J: I/g 1 dx ~ (J:r dxtlO: g2 dXY'2 ... (6)

But IJ: fg dx I~ I: I fg I dx

~[ jbf = dx ]I/B[Jb.. 154 dx ]1/1 ,

{J
by (6)

... IIJ .. fg dx
b
I L
~ b12 dx )1/2[Jb g2 dx ]1/2 .
[
(J
Proved .
Ex. t. IflE L2 [0, I]. show that

lJo1f(x) dx I~ [j1 0 I f(x) 12 dx )1/2 . [Banaras 1971]

Solution. If f.g E V (0, 1], then by Schwarz's inequality,
we get IIfg II ~ 11/.1'2 II g 112

or 1: I fg 1dx ~ U: If 12 dx rIll: I g I' d.~J/I

Taking g(x) = 1 V x, we get

J: ! f I dx ~ [1: II 12 dx r /2

[For I: Igl2 dx =(t dX=1]

or IJ: f dx I ~ J: I f I dx ~ U: I f II dx Tla

or If: I~ U: I f r'2.
f dx 12 dx
Ex. 2. Iff and g are Jquare integrable in the Lehesge sense,
then prove that 1+ g is also square integrable in the Lebesgue sense
Qnd
[Meer ut 1971J
 235

Solution. f and g are square integrable in the Lebesgue

sense so that
f, g E L2 [a, b].
To prove (I +g) E L2 [a, b]
and li/+g 112 ~ 11l112+11 gilM
By Schwarz's inequality,
f, geL2[a, b] ~ fgE L[a, b] ..• (1)
Alsof, gEU[a, b) ~ p, g2EL[a, b] ... (2)
(1) and (2) :::> P+g2+2/gEL[a, b] ~ (/+g)2EL[a, b]
~ l+gE Uta, b].
This proves the first required result
I1+ g 12 ~ ( I f I + I g 1)2
= 1/1 2 + IgI2+2Ifl·lgl·
= I/I!+/g 12 +21lg 1
. J: I f+g 12 ~ J: If !I+ C I g i2+2 1: Ilg I .
By Schwarz's inequality, this gives

f: I f+ g 12 ~ J: I11 + J: I
1 g 12
+2[1: If 12 T/2 U: I g 12 J/ 2

=[0: I t2 +0: I \2 fa r
fl2 g

Taking positive square root

(J: I/+gI2t2~(j: 1/12 f2+(J: I t2 g,a

or II 1+ g 112 ~ 'I f II~+II g 112·

Tbeorme 7. Let < fn > be a St quence of integrable functiuns
which converge in mean to a IUllction f 1 hell the sequence con-
verges in mea,mre to f
Proof. See theorem I, chapter 10, page 177.
Theorme 8. Let I (x) ~ 0 be an integrable lunction a e. 011 a
measurable set E.
Then! E I (x) dx=O iff f (x)=O a.e. on E.
The proof of this theorem is left .as an exercise for students,
236 LP-SPACES

Therome 9. Show that (U', d) is a metric space.

[Kolhapur 1968J
Proof. Let In, f,,, E LP (a, b) be arbitrary.

We define d (In, f,,,)= II In-1m 111'=0: lin-I", Ip dx t P

We have to prove that (£1', d) is a metric space.

(i) :. I in -1m (x) I ~ 0

(x)

... 0: I In f.,. I p dx t P
;;;. 0
i.e. II/n--Im II p=d (In.!m) ~ O.
(ii) d (I, g)=O <=> II I -g II p=O
*S: I I-g 121 dx=O
<=> 11- g 121 =0
a.e., By Theorem 8.
-=- 1/- g 1 =0
a.e., <=> I =g
11 a.e.
• d (.f. g)=O ifl=g a.e.
(iii) d(/, g)=d(g,/) For II--gi = Ig-II
(iv) IIf-g 11,,=
II (f--h)+(h-g) II" where!, g, hEU'.
On applying Minkowski's inequality, we get
IIf-gll,,= II (f-h)+(It--g) II"
E;;; II f-/H,,+ II h-g lip
=d (f, h)+d (h, g)
Finally, d (j, g) <
d (I. h)+d (h, g).
From what has been done the required result follows.
Theorem 10. An U'-space is a linear space.
Proof. Let I, gE V' (a, b) and cE R be arbitr:!.!'y.
To prove that D'-space is a linear space, we have to show
that
(1) f. gEU' (a, b) => f+ gE L (a, b)
(2) fE LP (a, b) and cER ::} cfE LV (a, b)
For proving (I), See theorem ..>.
(3) LeI fED' (a, b) and cER. Theil
fis measurable over (a; b) ... (4)

J: I flP dx< 00

(4) => cfis measurable over (a, b). For c is a constant.

 237

J: I clip dx=1 ell' J: I/IP dx < ! c Ip 00=00, bv (5)

fb
i.e. L I ({ Ip dx < 00

Also cl is shown to be measurable over (a, b). These state-

ments imply that c/E LP (a, b). Hence the result.
Theorem n. An LP-space is a normed linear space.
[Meerut 1990 ; Kolhapur 70]
Proof. Throughout the discussion we shall suppose that
I, g E LP (a, b) and c E R.
To prove that U' (a, b) is a normed linear space, we have to
prove that
l. I+g E U (a, b)
2. cl E DJ (a, b)
3. II/il" ~ 0
In particular nfllp-=O iff/=O a.e.
4. IIclll,,=1 [Link]/i"
5. U+Kilp~lI/llv+lIglip
I. To prove
/. g E V' (a, b) ~ (, K arc measurable functions on (a, b)

and (f: I/I tl' < (f:

P =, I gIPf'J' <: 00

~ I ~~g is measurable on (a, b)

and (J: !I+g \" fP < 00

Q f+g E LP (a, b)

II. IE L" (a, /1) =;. (f"I 11') <

f is measurable and
b
L:f
\ 1/1'
00

~ dis mca<;urable and (J: I Cfi/'fJP « ..,

=> .IE V' (0, b).

III. Since II i d () and ~o ( J: I f'!/yiJ> ;;;.. 0; i.e., lI,n, ~ ()

Next we want to prllv.! II f ii,,=O .:> l=u.
It i" obvious that 1=--00 ~ !i I ! p=O.
Again if r ~ 0 and II f'I ,,-~O, then (=0.
(Refer solved prohlem 4, ch'lptcr 9, page 149).
238 L1'-SPACES

IV. II cIW=O: I cliP dXf'P =1 c I ( Illp dXY'1' !:

=1 c 1·llfl'"
V. For proving this, see Theorem 5, page 231.
This concludes the problem.
Theorem 12. ([Link] Theorem).
I. The necessary and sufficient condition lor a sequence
< j" "> ollunctions In E LP (a, b) to converge in mean to a lunc-
tion I E LP (a, b) is that
"/m--/.. 111' .... 0 as m ..... oo , n-+oo. [Banaras 1971]
Or An U-space is complete space. [Burdhw8n 1990. Meerut 1988]
Proof. Let <: In > be a sequence of functions belonging to
L" (a, b). Let this sequence converge in mean to a function
f E Lp (a, b) so that
given. < 0, 3 no E N s.t. m ~ 110 => II/- /m II,. < • ... (1)
To prove that Lim
m-+oo
"In -1m 11,,=0, it is enough to show
n-+oo
given E > 0, i 110 E N s.t. m. n ~ no => \I 1m --In III' < E.

For m, n ~ no, we have

I'ln- 1m d,,=lIln-I+I-lm II..
~"f,.-I"II+"I-I... III', by Minkowski's inequality.
<6+e=2e.
Fmally, 111m -In jI,,<2e V- m, n~no
From this the required result follows.
II. Conversely suppose that <.f~ > is a sequence of func-
tions belonging to U(a, b) s.t.
given e>O, i noEN s.t. m, n~l1o => Ii J.n -In 11,,<. ... (2)
To prove that the givcn sequence convcrges in mean to a
function fE LP (a, b).
I
Set e=2k and no=nk.
1
Then III.,. -In III' < 2k ¥ m, n ~ nk'
Or we can take 1/ fm--fn" I, < e for m>nk' ... (3)
we have 111<n~<n3 ..... <nt ......
Set gk----fnk.
By '" hat we have established it foIlows that
 239

1
-=1
2"

00
i.e. E II g,_+I-gl·lIp < 1.
k=l

This proves that the series E II gHl- g" II" is convergent.

k

... (4)

with the conventipn that

r
g(x)=oo if the series on r.h.s. of (5) is divergent.

(1: I g I" dx r" =mL~: U: [

Making use of Minkowski's inequality, we get
I gl I +k~11 gHl -gk I rl"

IIg",,~ Lim (ligl/lp+XIIgHI-gkllp]

m~oo k-l

00
= Ell g'+l-g"
II gll'p+ . ,,=1 II"
< II gl /I,,+I=A finite quantity.
i.e. II g 1/" < A finite quantity.
This proves that gE LP (a, b).
Let E be the set of those points x for which g(x)=oo.
D.:fine l(x)=O, xEE

and x$Ebut xE(a, b)

If x(/. E, then

I(x)= Lim [ gl + E
m.
(g, H --g,J
]
m ~oo k-l

= Lim g", ()
x ..
m-+oo
Lim
1 (x)= m ....oo g",(x), XE (a, b) s.t. x$. E·
Also 1 (x)=O, XEE s.t. xE (a, b).
Lim
It follows that I(x)= g",(x) a.e on (a, b) (5)
In ... ex:. •••
240
. Lim
ThIS ~
111-+ 0 0
If -- g", 1=0 a,e. in the set (a, b).

00
Evidently g", (X)=gl+ l: (gHl--g,.)
k=l

00
I g", I ~ I gll+ L
k=l
I gHl-- gk I
00
< I gl I + E k=l
I gHl-glc I=g.
:. I gIll I ~ g >of m.
Making m~oo and observing (5), we get I f I < g.
Also If-g", I ~ If I + I g", I ~ gt-g=2g.
If-g", I ~ 2g.
Thus we have shown that 3 g(x) E LP (a, b) s.t.
I f- gm I ~ 2g ¥ m

and Lim
m-+oo I .f- g", 1-0
- a.e·. . (a,
In b)

On applying Lebesgue's Dominated Convergence Theorem,

we get
Lim
111-+ 00
f~ If-g",llIdx=f b Lim
m-? 00
If -gm I" dx= Jba [Link]=O

U:
Q a

or m~: I f--gm I" dx J/l'=O

Lim
or
m-+oo
"f . gIll 1/,,=0
'
or ,;~r: II f- -/;.", II" - O. (,"'" (x)=j~", (X)

or Lim II.f' --J,In II,,-~' 0 . (·n u"ing (3).

111-+00

This proves the required result.

Remark. This theorem can 'also be stated in the foIlowing
manner.
(i) The LII-Space is complete. [Meerut 1988, 87; Punjab 691
(ii) The LII-Space is a Banach spa!.".
(iii) A convergence sequence <: fn . ~ in Lt'-space ha~. a limit
in LII-Space.
(iv) Every Cauchy sequence "(II> in the f.!·-Spacc converges
to a limit function r E: [}'-Spacc.
 24(

Ex. The LP-spaces are complete. [Meerut 88, 87; Punjab 69]
Solutio.. To prove the problem, we have to prove every
Cauchy sequence < fn > in LP (a, b) is convergent. Then given
E > 0.3 no E N s,t. m, n ~ no => IIf", --fn II" < E.

Now write proof of" II part of Theorem 12 .

•
Theorem 13. Suppose < f .. > is a sequence of functions
belonging to an LP-space. 1/ this sequence is convergent, then it is
a Cauchy sequence. [Meerut 1990]
Of A sequence 0/ functions belonging to an LP-space is a
Cauchy sequence if this sequence has its limit in the space.
Proof. Let the sequence < /n > of functions belong to an £1'-
space S.t. this is a convergent sequence so that it must converge
to a functionf E LP-space (say) i.e., 3f E £P (a, b) S.t.

Lim fn=f a.e.

n~oo ..( I)
To prove that < {" > is a Cauchy sequence.
(1) ~ given E > 0, there exists a positive integer no S.t.
E
>.;f /I ~ no:;:> 11/- f" II II < '2 ¥ m, n ~ nil

IIfm -In II ,,=[1 (." -/+f f" II p

,< 11/", -/11 + 11/-/"," p. hy Minkowsk:'s inequality
~ ,
<:2+2=£'
11/."-/,, Ii " < E ' " In, n ~ no
wt"dch proves that the sequl:nce < fll > is a Cauchy sequence.
Ex. If < /.. > is a Cauchy sequence, then it has a limit.
Hint Refer to first part of Theorem 12, the result£ at ollce
follows.
Theorem 14. A sequence <. /~ :> 0/ functions belonging to an
LP-space has at most olle limit.
Prooi. If po~slhle let /n~/, rr.~g.
Then the identity .f g=/ /n +f,.-g gives
ii/ g II ~ :: / -Iii Ii + II I .. - g n . . .. (1)
This follows from Minkowski in~quality.
Since r.,
-.-/:;:> !If,,-/il =0 as n~oo
Ift-+g ~ II/.. -g II =0 as ,,--,00.
Now (I) shows that IIf- g II ~ 0
242

But IIf-g II ~ 0 is always true.

Combining the last two,
IIf-gil=o
This => f==g a.e. Proved.
Approximation by continuous functions.
Theorem 15. Iff is a bounded measurable function on [a, b]
thenf.r a given ~ > 0, a a continuous function g on [a, b] s.t.
IIf-g 112 < •.
Proof. Suppose f is a bounded mea~urable function over the
closed interval [a, b] and let

F(X)=S: f(t)dt,x E [a;b] ... (1)

Then F' (x)=f(x) a.e. on [a, b]. ... (2)
Let M be an upper bound of I/(x) 1in [a, b].
Then. I f(x) I ~ M V x E [a, b].
If x, x+h E [a, b], then
F(x+h) -F(x)= [+11 f(t) dt - 1: f(t) dt

[+II f(t) dt+ 1: f(/) dl= [+II J(t) dl

=

F(x+h)-F(x)= j
Gl+1I
:. f(t) dt01
... (3)

From which,

1 F(x+h)--F(x) i=/ (+TIf(t) dt I ~ J:+h If(l) I dt

~ J dt=[Link]+II
01 M

1 F(xth)-f~x) 1 ~ Mh.
Taking h < 8 and Mh < El , we get
I F(x+h)-F(x) I < tl' whenever h < 8.
This => F(x) is continuous in [a, b].
For each natural number n, we define
Gn(x)=n r
J
Gl+(l/"'
,
/(t) dt, x E [a, b]
.•• (4)
 V'-SPACES
243

= n[ F (X +~) - F( x) ]
G,.(x)=n [ F(X+! )-F(X)] ... (5)
Also F(x) is continuous in [a, b].
Consequently G.(x) is continuous in [a, b].
From (5),

Lim ~
/
Lim
F (X+n!-.) --F(x)
n-+oo u,,(x)= n-+oo - - - lIn - - -

_ Lim F(x+h) -F(.~'Le=F'(X) = r.(x) by (2)

-h-+O h J~ ,

Lim
Gn(x)=/(x) a.e. on [a, b]
n-?oo ... (6)

From (4), I G.. (x) I = I n j"'t1'" .f(t) dl \ ~ n 1.,+1/11 1/(1) I dt

! \ (l: (;C

",+l/.. 11M
~ 11M J dt=- =M.
n
...
x

I G,,(x) I ~ M ¥ n E N and V x E [a, bJ

I G,,-/12 ~ (I Gn I + III )2 ~ (M+M)2=4M2
G•• -/12 ~ 4M2, y. n.
By virtue of (6),
Lim Lim I Gn-11 2 -
- 0
G..(x)-/(x)=O or .
n-~ 0Ci n->oo
Hence, by Lebesgue's bounded corivergence theorem,

Lim J: I G.. -/12 x= 1: Lim I G.. -/12 dx= J: 0 dx=O

:. Lim fb I Gil-liZ dx=O
n-+oo CI

r Lim
n ... oo
(fb
CI
I Gra -/12 dx )1 12=0.
This '* Lim II G.. -/1I2=0
'* given > 0, 3 positive integer no S.t.
fi

V n ~ 110 => II/-G.. 11 ~ < 11

244
~ II 1-- Gnu II :I < •
~ II/-g II 2 < f, Gnu=g·
Thus 3 a continuous function
1 IIHl/no
g(x)=- /(1) dt. x E (a, b)
no •
s.t. II/-gll:a<. lI- .>0.
This completes the proof.
EXERCISES
1. <a) State and prove Schwarz inequality for integrals.
(b) IfI e L2 [0, 1], show that

II:/(X) dx I ~ u: I/(x) 12 dXr~

[Banaras 1971)
2. (a) Let < I.> E LI [0, 1] and IE L2 [0, 1] such that
In-./in the n1!an. Show that there exists a subsequence <1",,>
of < I.> such that I -+ I a.e. on [0, 1].
nk
(b) Let I, g E V (-co, 00) and define

h(x)oo: J~oo /(x - y) g(y) dy ( - 00 < x< 00).

Prove that he V (-00, co). [Banaras IV 1970]

3. (a> Stat~ and pro\'~ Minkowski's inequality, for
D'-Spaces.
(b) Let I, g E V [0, J] and define

F(xl= J: l(t) dl, G(x)= J: get) dt where x E [0, 1].

Prove that J: Fg dX=[ FG t --I: IG dx

and cst:lhli~h that the integral .. appearing i \ the formula do cxi<;t
and are finite [Baoaras IV 19701
4. Show that
(i) x-I E D' for p < 2 in (0, 1).
(ii) ~ -I (1+log I x I r- I ill V, in (0, (0) but nm in L'. in
(0, 00). for [Link] value of p other than 2. IBaullr.... 1,iil
O'·SPACES 245

5. LetIJt.!2' ...... E £P. Then prove that a necessary and

sufficient condition that
11/,.-/11,,-+0
as n-:;oo, ·is that 1If..-/m 11,,_0
as m and n-+oo, where
II ~ /lp= (11 rp(x) Ip dX) 1/1'
Show also that the limit function I is uniquely determined by
the sequence of functions <I,,>, except for a set of measure zero,
and that there exists a sequence < f > such that lim f f
nr nr
almost everywhere. [Banaras 1967J
 13
Further Theorems on
Lebesgue Integration

Integration by parts. The formula of integration by parts in

the Lebf sgue theory is the same as the ordinary one i.e.,

J:fg dx=[(f fdx ) g J: -J: [~~. f 1fdx dx.

Theorem 1. If f(x) is an indefinite integral and if G(x) is al1
indefinite integral of g(x) , then

J: lex) g(x) dx=-~[ f(x) G (x) J: -

J:f'(X) G(x) dx.
Proof. We know that the product of two absolutely conti-
nuous functions is an abt;olutely continuoas function. Let G(x)
be an indefinite inte:ral of g(x) for x E [n, bJ so that

G(x)= 1: get) dt + C ... (l), C being constant

l is an indefinite integral ~ f is absolutely continuous.

(1) => G(x) is absolutely continuous on [a, b] and g(x) is
Lebesgue integrable in [a, b].
Since f and G both are absolutely continuous and hence
f(x) G(x) is absolutely continuous. Consequently f(x) G(x) is
Lebesgue integrable. (Refer to Theorem 14, Chapter 11, page
214).

Hence 1: f(x) G(x) dx exists.

J(t) is indefinite integral ~ f(t) is absolutely continuous

=> J(t) is indefinite integral of its derivative ret).
=> f'(t) is integrable => f'(t) G(t) is integrable.
.. I: r (t) G(t) dt exists.
fUkTHBlt THEOREMS ON LEBESGUE INtEGRA nON 2~7

Thus we have shown that the integrals

J: f(t) G(t) dt and 1: ['(t) G(t) dt both exist.
From (1), G' (x)=g(x) a.e. on [a. bJ.
I: ;x [f(x) G(x)J dX=[ f(x) G(x) 1: ... (2)
But ~x [f(X) G(x) l=f' (x) G(x)+/(x) G'(x)
=f'(x) G(x)+f(x) g(x) .
. Using this in (2),
J: [f~(x) G(x)+f(x) g(x)] dX=[ f(X)G(X)l

or J: f'(x) G(x) dx+ r: [(x) g(x) dX=[ f(x) G(x) 1:

or J:[(X) g(x) dX=[ f(x) G(x) ]: - J:f'(X) G(x)dx.

Ex. Let f, g both integrable over [0, I]. Also let f. g both
be measurable over [0, I] and

F(x) = J: f(t) dt, G(x) = J: g(t) dt, where xE [0, J].

Prove that 1: Fg dX=[ FG J: -I: fG dx

and established that the integrals appearing in the formula do exist
and are finite. [Banaras IV 1970J
Solution.

1: ~x (FG] dX=[ FG 1: ... (1)

But ~x [ FG ]=F'G+FG'=--fG+Fg.

:. ~x [FG ]=fG+Fg. ...(2)

For F(x) = 1: f(t) :;. F (x)=[(x) a.e.

and G(x)= [ get) dt => G'(x)=g(x) a.e.

From (I) and (2), J: (fG+Fg) dX=[ FG J:

or r
,0
fG dX+Jl Fg dX=[FG]l
0 0
248 FURTHER THEOREMS ON LEBESGUE INTEGRATION

or
,
r0
Fg dX=[FG r_ Jl
.0 0
fG dx.
... (3)
Complete the proof of existence of the integra! appearing in
(3) by the help of Theorem 1.
I
Theorem 2. First mean value theorem. Suppose f .(.t) i~ bozJzded
and g(x) ;;a 0 in a measurable set E and are [Link] over E;
then 3 a number" between the bounds ofI(x) in E s.l.

I I(x) g(x) dX=A IE g(x) dx.

Ii

Proof. Since f(x) is bounded in E and so it will have its g.l.b.

(greatest lower bound) and l.u.b. (least upper bound) Gt and (3
respectively. Then IX ~ I(x) E; (3.
This => a g(x) ~ f(x) g(x) ~ f3.g(x) as g ~ 0

- IX f Ii g(x) dx ~ f Ii f(x) g(x) dx ~ (3 IE g(x) dx

J

[asf, g are integrable => fg is integrable]

=> f~.fg.
dX=A
..
Jrg dx
... (1)
where q. ~ i\ ~ 13.

Corollalt'y. To prove t f g dx=/(~) Ie g dx.

f is integrable => f is continuous => 3 x, y E E S.t.

f(x)=rz,J(Y)=fJ and/(x) takes every value
between a and (J
=> 3 , E E s.t. f(~)=A where IX ~ ). ~ p.
:. t fg dx=f(~) t g dx, by (I).

Theorem 3. Second mean value theore... If f( x) is integrable

over (a, b), and g(x) is positive, bounded, and non-increasing, then

f/(X) g(x) dx=g (a+O)f:f(X) dx

where ~ is some number between a and b.
Proof. Fix s > 0 S.t.
e < g (a+O) -g (b -0).
This => 3 a point Xl s.t.
g(a+O)-g(~) < e
fURTHER TREOREMS ON LEBE~GUE INTEGRATION 249
Similarly there are points x 2 , xa," ... , x" s.t.
g(xr - 1 +O)-g(x) < f, (x r - 1 < x < x.)
~ E (x > x r ),
so long as g(x'_l+O)-g(b-O) > E. Otherwise we take x,,=b.
Thus the point b is reached after a finite number of steps,
since the variation of g(x) in each interval (Xr-h x.) is at least E.
Write ifJ(x)=g(x.+O) in each interval Xr ~ x < Xr+I' Then
o :::;;; Iji(x) -g(x) < f!, except possibly at the points

I: ~(x) f(x) dx= l~l g(Xr+O) J ::+1 f(x) dx.

Suppose F(x) = J: f(T) dt.

... (1)
Also suppose that In and M are lower and upper bounds of
F(x) respectively. Then, by virtue of Abel's lemma,

mg (a+O) ~ J: Iji(x) f(x) dx ~ Mg(a+O).

. .. (2)
But

\J: ifJ(x) f(x) dx- J: g(x) f(x) dx I~ IS f: '/(x) , dx.

This tends to ° as E_O.

Hence Limo !Jb ljif dX-Jb fg dx ',1=0

Ii-+ I a a

or Lim Jb ~f dx= Lim JO fg dx ... (3)

.-+0 a e~ 0 a

Making ~-)-O in (2) and observing (3), we get

lIIg(a+O) ~ Jag(X)f(x) dx
iJ
~ Mg(o+O)

or In :::;;; g(a+O)
J Jba g(x) fIx) dx ~ M.

But an indefinite integral is a continuous function and hence

F(x) is a continuous functi on. Consequently F(x) takes every
value between In and M so that at x=~, we have

1
F(e)= g a+O) J"' g(x) f(x)
a dx ... (4)
250 FURTHER. THEOREMS ON LIiBESGUE [Link]

But F(x) = 1: f(t) dt ... (5)

J'o I
f(t) dt=g(Q+O) Jba g(X) leX) dx by (4) & (5).

or f: f(x) g(x) dx=g (a +0) J: f(x) dx ..

where a<f<b.
This completes the proof.
Remark. If g(x) is positive and non-decreasing, then the
corresponding formula is
J: f(x) g(x) dx=g(b-O) r: f(x) dx,
where a<~<b.
Introduction. Stieltjes integral. Lebesgue Integral is a
generalisation of Riemann integral. Another generalisation of
Riemann integral is Stielties integral.
4. Definition. Stieltjes integral.
Let f and g be finite functions on the closed interval [a, b],
[Link] [a, b] into parts by means of points.
a=xO<x1 <X2 < ...... <x.. =b.
Choose a point eTc, in [Xk' x~+l] and form the
"-1
sum 11= L f(Ek) [g(XHl) -g(xk)].
1=0

If the sum a tends to a finite limit I as

y=max (Xt+t-X1J .... 0,
independent of both the method of subdivision and the choice of
the point ~k' The limit / is called the stieltjes integral of tbe func-
tionf w.r.t. the function g and is denoted by

.f: (x) dg(x) or by (S) I: lex) dg(x),

where S denotes that the integral is taken in the sense of Stieltjes.
From the definition of Stieltjes integral it is clear that Rie-
mann integral is a special case of Stieltjes integral, obtained by
putting g(x)=x.
Theorem S. Relation between L~!lesglJe 'integral and Stieltjes
integral. If f is a continous function and g is absulutely continu-
ous function both 011 [a, b], ,hell
fURTHER THEOREMS ON LEBESGUE INTEGRATION 251

(S) Ib
a
j<x) dg(x)==(L) Jb f(x)
~
g'(x) dx.

Proof. Evidently both the integrals exist. Divide the closed

interval [a, b] by means of points
a=xJ <xl <x 2 < ... . ... <xn=b.
Choose a point fie in the interval (x., x1'+l)'
"-1
Form the sum (1= E f('C,d [g(Xi.:-tI)- g(Xk))'
k=O

Write '\=max. (XHI-Xk).

.
Th(;n, by definition Lim a= (S)
.\-~O 1°aI(x) dg(x) ... (1)

g(x) is absolutely continuous implies

(L) fXkH g'(x) dx=g(X k11 )-g(Xk)

x.
a-(L) 1: f(x) g'(x) dx

= k:/(ek) [g(Xk-tI)-g(x_)l- i~
~
:n (~) fX
x::+1 f(x) g' (x) dx

= E~f(~k). (L)
k=o
fXk +1 g'(x) dx- "i,1 (L)
Xk k~O
I Xk+1 f(X) g'(x) dx
Xk

If we denote the oscillation of .r(x) in (x", Xk~ 1) by Wk, then

the last becomes

'I a--(L) JO (x) g' (x) dx

v
\=\ n,e (L)
1e=0
JXk+1 Wk
X"
g'(x} dx I
~ ~; I(L) e:f- t Irk g'(x) dx \

=< "-1
I~O 11'". (L) JXx:-1-1 I g'(x) I dx

~ "-I
lJ «.(L) IX k+l I g'C~) I dx, where ot=malC. (w.).
k=l' x"
252 FURTHER THEOREMS ON LHIESGUE INTEGRATION

=0: (L) J: I g'(X) I dx.

Finally, I a-(L) J:f(X) g'(x) dx i E;; ~ (L) I: I g'(x) I dx

Making .\~O, consequently at~O, we obtain

~~~ Ia-(L) 1: f{x) g'(x) dx I~ 0

But .\~O
Lim Ia--(L) Jb,/(x) g'(x) dx / ;;.;;t 0

Therefore ~~~ I (1- (L) 1: f(x) g'(x) dx /=0

This ~ Lim
1 a=(L) JO f(x) g'(x) dx
"1.-+0 ..

=> (S) J: f(x) dg (x)=(L) J: fiX) g'(x) dx by (I).

Proved.
Theorem 6. Helly's second theorem. Let f(x) be a continuous
function defined 011 the interval (a, b) and < gn(x) > be a sequence
of functions which converges to a finite fUllctkm g(X) at erery point
of [a, b]. If

V/) (gn) < kl 'V n, then Lim Jb f(x) dg .. (x)= Jb f(x) dg (x).
a n~ 00 IJ Q

Proof. Divide the interval [a, b] by means of points xo, Xl"',

X.. s.t. a=xo < Xl < ...... < xn=b.

b n_ 1 [ [h
Then!" (gm) = to g ... (Xk+l)-g ... (xd· But ~ (gn) < kl "'f n.

.. ,(I)

... (2)

as lim gm=g.
b
or V (g) ~ kl ... (3)
II
fURl HER THEOREMS ON LEBESGUE INTEGRATION 253

J f(X) dg (X) = }..'

b

a
11-1

k=o
fX t-ll
X,:
f(X) dg (X) = "-I J~
l..' • I:H
k=O X"

[ I(x) -I(x,,) +/(x,,) ] dg (X).

Since I is continuous and hence continuous in [x •• X::l-11·

SO we can take l/(x) ·j(Xk) i < (e/3k 1) \I XE[X •• Xlc+l] ... (4)

:. Ilfb
a
I dg II ~ i;1 JXk Ll I I(x) - I(Xk) I . I dg (x) I
1c=0 Xk

~ 3-~~ . kl +l' 1;:-11 i f(Xk) I . I dg (x) I by (2)

or II: r I~ ; + :~: !
dl! f(Xk) J Ig(XkU) -g (Xk)
This can al<;o be expressed as

\r rdg I< ~-3 + '~:l , f(x~)

&
a k=O
g(Xkll) --g (Xt)
,
I
I
... (5)

where O<8~1.
Similarly we can ~et

:',b f dg",
I E;;; 68'
i' + "-I
;,; i f(Xt l I \ i
i gm (X'·H) ... (6)
I a I.~O : i
where
Lim
Since m.'>oo g." =- g
Henl'e 3 po~iti\'c in:egcr 170 'i.t. III ~ "0 :>

\ .'~l f(."(t) [g,., (Xk.I)- g.,. (Xk)]

k~O
-"if I(Xt)
tc~o
(g(X;, 1 t> g(x) 1;I < ~I--
J

In this event (5) and (6) taken together imply that

lI~
I a f dgm
faa r dg < I Ii
254 FVRrHER THEOREMS ON [Link] INTEGRATION

Thi<;::> Lim
In_CO
I b
a
I(x) dgm (X)=Jb I(x) dg (x)
a

~ n_oo
Lim Jba I{x) dgn (X)=J b I(x) dg (x).
a

7. Definition. Cumulative distribution function. Let B be a

family of Borel sets, each of which is a subset of R. A measure II-
defined 011 B is called Borel measure on the real line. If II- is finite
Borel measure, therl we associate to it a function F known as its
cumulath'c distribution function by setting
F(x)=f1o (-00, x] ¥ x E R ... ( I)
It is easy to verify that F is a real valued monotone increas-
ing function
Also ~, (a, b]=F(b)-F(a). ...(2)
For F(b)-F(o)=p (-00, D]_.p (-00, aJ=f' (a, b].

Since (a, b]=

,,=1
n (a, b+~]
n

Hence it (a, b]= Lim f' ( a,

. n-+oo
b+J..]
n
[Refer to Theorem 17, Chapter 7, page 100]

Consequently F(b)= Lim F (b+~)=F(b+).

n_co n
This proves, by definition, that F(x) is continuous on the
right.
Furthermore p(b}= Lim p (b- ~,b ]
n_oo n

= n~: [ F(b) -F (b- !)], by (2)

=F(b) - Lim F (b-

n_oo n
!.)
=F(b)-F(b-).

If p{b} were zero, i.e., Borel measure of a singleton set {b}

is zero, then
F(b)=F(b- ,.
This shows that F(x) is continuous on the left. Finally we
have shown that
F(b)=F(b+ )=F(b-).
This shows that F(x) is ccntinuous if p{b} =0
00
Since n (- ,-nl=0, m(0)=0.
n~l
FURTHER THEOREMS ON LEBESGUE INTEGRATION 255

Lim Lim
Therefore F(n)=O or F(x) =0.
n -+--00 x-+-oo
Thus we have shown that 'if po is Borel measure on the real
line, then the cumulative distribution function F is a monotone
increasing bounded function which is continuous on the right
hand and
Lim
F(x) =0.
x-.-oo
Definition. Lebesgue-Stieltjeg integral.
Let f be a non-negative Borel measurable function and F a
monotone increasing function which is continuous on the right.
Then the Lebesgue-Stieltjes integral off w.r.t. F is defined as
f f dF=f f dpo,
where po is a finite Borel measure and F its cumulative distribu-
tion function. :
Let f be any Borel measurable function, then we define
f+(X)~-{ f(x) if f(x) ~ 0
o if f(x) < 0
J 0 if f(x) ~ 0
f-(x) = 1 -f(x) if f(x) <. 0
Then f(x)=f+ (x) -f- (X).
In this case J f dF= J f+ dF- Jf- dF ... (3)
The fuuction f is said to be integrable in the sense of Lebes-
gue-StieItjes if both the integrals on r.h.s. of (3) are finite.
 14
The Weierstrass
Approxi mation Theorems
and
Sem i-continuous Functions

14'0. Definition.
A Polynomial function (or poiynomial) is a function P which
is expressible as
P(x)=OO+OlX+ 02XZ+ ..... +On X,. ,
where n is some non-negative integer and
°
0 ,°1, 02' .... .. 0" are constants.
14'1. Definition.
Let f b\! a finite function defined on the closed interval
rO, ;J. The polynomial

Bn (x)= l: Cx'
k~o I..
(l_-X),,-k j ('!...)
n
i~ called Bern.,tein polynomial of degree n for function f. Here
("
. I~ . I coe ffi'
t h ~ B'Inomla Clcnt k' , n ! k -, .
k _ • (n- ).
Theorem t. (S.N. Bernstein's Theorem). If a jUlIction f(x) is
continuous all the segment lO, I], thell Bn (x) -+ f(x) uniformly
lI'.r.l. x as n~oo.
Proof. The following proof is due to Bernstein. Here we
shaH make use of Bcrnste in polynomial Bn (x) defined as
Bn (x)= £" C
TO
Xl: (I-x)n-!.
•
f (
-k) ... (1 )
1,0 1 11

where t
II! (.~ =.
- - _.- - •
k:(n-k)! k

The polynomial B" (x) is calIed Bernstein polynomial of

degr\!e !1 for .I~x).
n
Lemma. To prole 1: C (k [Link].(i - X)"-I: ~ 4'
"... ~
 SEMI CONTINUOUS FUNCTIONS 257

Let a, b be arbitrary rOO,1 numbers. By Binomial theorem.

.. n
]J C ak bn-k=(a+b)n ... (2)
<=0 k

Differentiating partially this w.e.t. a we obtain

" n
E [Link]"-[Link]-k=n (a+b)lt-l
1:=0 k

On simplification this gives

n "
E [Link].b"-"=na (a+b) ..-l. ... (3)
k=o"
Differentiating: this w.r.t.
,
a.
n It
E C k 2.a H b'H=n (a+b),,-l+n (If-I) a (a+b)n-%
k~o k

or
. "
E C.k 2a"b"-k=na (a+b),,-l+ n (n--1) a2 (a+b),,-t
.
... (4)
k=O k

Throught the discussion we shall suppose that x E [0, I].

Set a=x, b=l-x so that a+h=l.
Then (2), (3) and (4) yield respectively
n n
E C x' (1- x)n-k= 1 .. (5)
k~o k

It n
~' [Link] (l_x)n-k=nx ... (5')
k=o k

n 'N

E C.k 2 .xk. (I--x),,-I:=nx+n (n -1) x1!=nx [I +n:c-x]

k=o k

~r
" ..
E C.k 2 .x!'.(I-x)"-k =nx (l+nx-x)
k=o"
Multiplying (5), (5'), (5") by n 2x 2 • - 2nx, 1 respt'ch vely ,[Link]
then adding,
" It
~' C (k-nx)%.xk.(l-x)"-k=nx(l-x) ... (6)
"=0 "
Since (2x-I)1 ~ 0 or 4x1-4x ~ 1 or x (I-x) ~ 1/4
or nx (L -x) E;; (n/4). Now (6) is reduced to
258 SEMI CONTINUOUS FUNCTIONS

n
..0 .
It It
X (k-nx)'I. ext (l-x)"-· ~ - ... (7)
4
Hence the lemma.
Since fis coDtinuou~ and therefore
given , > 0, 3 8 > 0 s.t. I f(x) - fey) I < «
whenever I x-y I < ~; x, y e (0, I).
Write M=ma'C. { l1(x) I : x E [0, I]}.
Multiplying (5) by f(x) and then subtracting (I) from this we
obtain ¥ n E N,

fix)--BIt (x)= i: l- f(x)-f (~)]

.~ n
ex· (l-x),,-t II

=£' +L'. ... (8)

where £' is the sum over those values of k s.t.l ! -x !<~, where

E" is the sum over those values of k s.t./ ~ -x ~ 8.

.n
I
Divide 'all the ntlmbers k=O. 1.2, ... ·...• n into two parts A
and B according to the rules .

k E Aul: --xl < 8

and k E B if I: -x I~ 8.
It means that A and B correspond to ];' and £. respectively.

If k E A, then /f (-! )--f(x) , < t.

I E' I =/ E' [f(X) - f ( : )] ~ x. (l-x)"-t

"~ ];' If(~)-f (,n!!-.-)II C

•
x· (I - x),,-ar

< •
t.E' C JK" (l--x),,-I:=c, by (5) .
•
... I E' 1<<< ... (9)
.k E B ~ 'n
Jk --x II ~ 8 ~ f k - nx/
-lis ~ 1
SEMI CONTINUOUS FUNCTIONS 259
(k-nx)2
=> - n2~' ~ 1. For k, n, x all are real.

If k E B, then

I );"1 ~ E" If(X)--f ( : ) \ ~ x~ (l-x),,-t

~ EW [ I f{x) I +k (~ )IJ f x' (I -x)"'"

~
""
2ME" (k-nx)2
2.2
c· XIt (1- X)ft-I:
J
nOli

Finally I E· I <; 2~2-' ... (10)

Writing (8) with the help of (9) and (10), we get
M
I f(x)-B .. (x) I < E+ 2nF' ... (11)
M
Choose no E N S.t. no > 2.8;-'
M M
Then II ~ no =- n ~n ~ "D > 271f. => n > 2E81

M M
=> 6 > 2n/)2 => 2nJiI- < E

M
=- 2n~i-+E < 2E.
Now (II) becomes
I.f<x)- B .. (X) I, < 2E ... (12)
B.. (x) cOl1vergl!:s uniformly to I(x) w.r.t. x E [0, 1) as n-HIO.
Proved.
Problem. If a function f(x) is continuous on [0, 1], then
(or every _ > 0, there is a polynomial P (x) s.t. I /(x)-P(x) f < •
yo iC E (0, 1].

Solution. Here write proof of Theorem 1. In the end of

the proof put B,.(x)=P(x).
260 SEMI CONTINUOUS FUNCTIONS

Problem 2. For ellery continuous/unction/(x) on (0, I), there

is a sequence 0/ polynomials < Pn(x» S.t. < Pn{x) > converges
uni/crmly to J(x) w.,..t. x E LO, 1] as n~tX).
Solution. Here write proof of theorem I. In the end of the
proof put B,.(x)=Pn(x).

Problem 3. Prove that E

.. C.x~.(t _x)n-k (k- nx)1 ~ (n/4).
1>:=0 I>:

Solution. Here write the proof of equation (7) Theorem 1.

Now we shall prove a very important theorem known as
Weierstrass approximation theorem of real analysis.
Theorem 2. (Weierstrass Theorem). Let lex) be a continuous
/ullction defined on the closed interval [a, b J. For every 6 > 0,
there exists a polynomial P(x) S.t.,
I/(x) -P(x) 1< 6 Y x E fa b].
,Thi.' theorem can aho be rephrased as :
For each continuous Junction, there exists a sequence <Pn(x»
0/ polYllomiais s. t. < P.. > COnl'erges uniformly toftx) w.r.t.
X E fl, bl as n~oo.
Proof. Firstly we shall establish the following lemma:
Lemma. For every continuous functionJ(x) defined on the
closed interval [0, I], there exists a sequence of polynomials
< P .. (x) : n EN> s.t. < p,.(x) >
converges uniformly to lex) w.r.t. x E [0, I] as n~OO.
The proof of the lemma starts
[For proof refer to Theorem I].
This completes the proof of the lemma. Going back to the
proof of the actual theorem, suppose that / is a continuous func-
tion on [a, b]. If [a, b]=[O, I), then theorem is proved.
Now we suppose (a, b]:;e[O, 1].
Define g(y)=/[a+(b-a) y]
Then g(O) =/(a) , g(l)=/tb).
Also/is [Link] on La, b].
This leads to the conclusion that g(y) is continuous on (0. 1].
Hence, by the lemma, just proved, :I a polynomial Q(y) s.t.
I g(y)-Q(y) ! < 6, ..... (1), 0 ~ y ~ 1.
If x E [a, bJ [Link] y E [0, I], then set'
x a
y= -_._-- so that
boa
SEMI CONTiNUOUS FUNCTIONS '61

g(y)~g (:=:)=1 [a+(b-a) (~=:)]=f(X).

Then we have .
j!(X)-Q (:=:) I< e, a ~ x ~ b, by (l2).
The polynomial Q (~=:) .sati!>fies our requirements.
Theory and problems related to semi-continuous runctio~s.
14'2. Definition. Limit superior and Limit inferior. Let a
function f be defined on a set E and let Xu be a limit point of E.
Consider an arbitrary real number ~ > O. Write
M,,(xu)=sup {[(x)}, m.,(xo)= inf {f(x)}.
x E (xo 8, x o +3) n E.
The quantity M~(xo) does not increase and M~(xo) does not
decrease with decreasing 8. Hence the (finite or infinite) limits
Lim Lim
M(xo) = 8_0 M.,(x o), m(xo)=8_0 m.,(xo)
exi!,t. M(xo) is called limit superior and m(xu) is called limit
inferior. We also write
Lim
M(xo) = [(x),
x-+xo
If Xo E E, then we have the relation
m(x,,) E;;, I(xo> C; M(xo)·
Notice that the function may not be defined at Xo itself.
14'3. Definition. Semi· continuous function A function I,
defined on a closed intervallO, Ij, is said to be lower semi·continu-
ous or upper semi-continuous at the point Xo according as
Lim Lim r .
- - (x)-/(xo) or J(X)=j x I
X-Xo' x-+xu 0

Iff is continuous at the point x~, then it will be both upper

and lower continuous.
F·or I"IS contmuous at Xo => Ji1,x)= Lim ji(x)= Lim r-
- -- J{x)
X-Xo X--'o
262' SEMI CONTINUOUS FUN'!TIONS

Conversely if1 is lower and upper both continuous at the

point Xo then 1 is continuous at XII provided 1 is a finite function.
Then lower semi-continuity of a fmlction at a point Xo is
equivalent to upper semi-continuity of the function - I at the
same point xo'
14'4. Definition. A function I is said to be lower semi
continuous on the segment [a, b) if it is defined and lower semi
continuous at each point of this seginent. In a similar fashion
we definf upper semi-continuity of a function on a segmen~
[a, b].
Theorem 3. Let a lunction I be defined in a closed interval
[a, b] and let Xo E [a, b]. Also let < x" > be an arbitrary sequence
01 points in [a, bJ s.t. < x" > converges to xI) Then a necessary
anc1 sufficient condition that 1 be lower semi continuous at Xo is that
Lim
I(x.) , n_oo
- I(x,,)

Proof.
Lim
X,,=X o => 3 a subsequence < X ; kEN> s.t.
n-+oo n"
Lim Lim
k-oo l(x/1,) = /1.-+ oc I(x,,) ... (1)
Making use of the result :
The limit superior of the function I at a point Xu is the great-
est of the numbers which are limits of the sequence of the for~
l(x1),f(X 2),f(xa), ...
where x" E [a, hI and x,,~Xo
We see that Lim I(x) ~ Lim /(x )
x ... Xo n_ CI) 11/:

. . Lim Lim
Usmg (1), thIS becomes '.-- f(x) ::;;;; - . (x,,)
" - Xo fI-+CIJ ... (2)

(i) Let f be lower semi-continuous at ;r 0 so that

Lim
f(xo) = ":::':;Yo f(x). . .. (3)
Lim
To prove thatf(xo) <; n-+ch I(x .. ).

In view of (3), (2) is reduced to I(:xo) ~ .!:i'' .1 I(x,,)

n-+ 00
Lim
(Ii) Convi:rsely suppose that/(xo) ~ - - IlXn )
n~ 0Cl ... (4)
 SEMI CONIINUOUS FUNCTIONS 263

To prove that f is lower semi continuous at Xe.

From (2), Lim f(x) E;; Lim f(x ff,)
x~xo n~oo

But x .. -xo ~ f(x,.)- f(xo)

Lim
:. - - /(x) ~ f(xo) ..• (5)
X-Xo
Combining (4) and (5)
Urn- ji(X ) = ji( xo)'
-
X-Xo
This :a> f is lower semi continuous at Xo.
Theorem 4. Let a function f be gefilled on a closed interval
[a, b], Xo E [a, b] and f(x o) >- 00. A necessary and sufficient
condition that I be lower semi continuous at x, is that 'to every
I(xo} > A, there corresponds a 0 > 0 such that I(x) > A
provided I x-xo I < b, X E [a, b].
Proof. Suppose a functionf is defined on a closed interval
[a, b] and Xo E [a, bJ. Also suppose thatf(xo} > - co.
(i) Let f be lower semi continuous so that
Lim
3 8> 0 s.t. f(x o) = - - f(x)
)C -+Xo ..• ( I)
Lim Lim '
where f(x)=m(xo) = --~ m.l (x..)
x ~Xo x-)oxo ... (2)
with m.,(xo)=inf {I(x) : x E (xo-8, xo+o) n [a, b]}.
Also let f(x o' > A.
To prove that 3 XE [a, b] s.t.f(x) > A
whenever I x -Xo I < O.
From (J) and (2), it follows that 3
x E (xo-'S, x o+8)nla, b]s.t. m~(xo) > A .•. (3)
Since f(x} ~ ma(xo) for I x - Xo I < 8.
On using (3),j(x) > A for I x-xo I < 8.
Conversely, suppose that/(xo) > A.
Now we can find a 8 > 0 corresponding to this A s.t.
ma(xo) ~ A, which in turn implies that m(xo) ;;;;. A.
Increasing A and taking the limit as
A - /(xo) , we obtain m(xo) d: I(xo)'
But f(x o) ~ m(xo) is always true.
 264 SEMI CONTINUOUS FUNCTIONS

Combining the last two inequalities, we get

Lim
f(x o}= m(xo) = - f(x o}·
X-+Xo

This proves that/ex) is lower semi contin~ous at Xo·

Remark. If (x.) is a finite number, then the above theorem

can also be restated in the following form:
A necessary and sufficient condition that /tx) be lower semi-
continuous at the point Xo is that to every to> 0, 3 a number ~>O
s.t. f(x o) - E < f(x) provided
I x-xo I < Ii and x E (a, b].
The theorem gives a relation between semi continuity and
continuity.
Theorem S. Suppose that the function f and g are defined on
the segment [a, b] and also suppOle that f and g both are lower semi
continuous at x=rxo. If the sumi+g is defined on [a, bJ, then it is
lower ~emi continuous at xO.
Solutiod. We may suppose thatf(xo) +g(xo) > - 00, hut such
of the numbersf(xo} and g(x o) is distinct from -d:). Consider an
arbitrary number A s.t. /(x..,}+g(x o) > A.
Then there exist numbers Band C s.t.
f(x o} > B, g(xll ) > C so that B+C > A.
Also f(x) and g(x) both are semi continuous at xu. Therefore
a a number ~ > 0 s.t.
!(x»B, g(x»C where xe[a, b) and I x-xo I < 8.
lRefer to Theorem 4J
For this x,
j(x) > B, g(x) > C .. f(x)+g(x) > B+C > A
=> f(x)+g(x) > A.
This proves that f{x) + g(x) is lower semi continuous at xo.
(Refer to Theorem 4]
Theorem 6. A necessary and sufficient condition that a fUIlClioll
f defined on the segment F- [a, b] be lower semi cofltinuous 011 E is
that the set E (f ~ c:) be closed for arbitary real c.
Proof. Let a function/be defined on the segment £=(a, b1.
Also suppose that c is any real number.
(i) Let f be lower semi continuous on E. Let A ~E (f ~ c).
To ~rove that 4 is closed.
 SEMI CONrI~UOLJ~, HIN('noNS 265

If the points of sequence <x,,> belong to A and if x,,~xo'

then lower sl'mi continuity off implies that
f (xo ) ~ -Lim.)
--- J~xn . [Refer to Theorem 3]
I1~OO

Xn E A ~ f (x,,) E; C

~n~oo'
Lim r( )
X,.
c'
..... C

Lim
=> ,(xo) ~ --. f(xn}
I1~OO
~ C

=> f (xu) ~ C => xoEA.

Thus x,,-Xo E A.
This => limit point of the set
{xn: 11=1, 2, 3, ...... } C A belongs to A
=> A is closed.
(ii) Conversely suppo<;e that A is a closed s~t.
To prove that f is lower semi continuous on E.
·
W rIte B= - - Ii(x).
Lim
x-:>xo
If possible, let us suppose that f (xo) > B. . .. ( I)
Our assumption implies that 3 a sequence < Xn'> in E S.t.
X"-+Xo'
Ifwe suppose that c is any number lying between B andJ(xo).
then B<c<f(xu) ... (2)
Thenf(x,.)<c for sufficiently large II
i.e. x"E A.
Then Xo E A, for A is. closed. Consequently f (xo) E;;; c con·
trary to (2). Hence (I) is impossible. Consequently,
Lim .
f (xo) ~ B= x-+xo f (x)
i.e. f (xo) <;; X~~o f{x)
This => f is semi continuous at XO'
 15
Singed Measure

Introduction. We have seen in the preceding chapter that a

measure is a non-negative function. If a [Link] is allowed to
take both positive and negative values, then this idea leads to the
consideration of signed measure which is defined as follows:
IS·9. Definition Signed measure. [Kanpur Statics 1976]
Let A be a a-algebra of sub-sets of X. An extended real
valued set function
v: A-+[ - co, co]
is called a signed measure on a measurable space (X, A) if it
satisfies the following conditions:
(i) v takes atmost one of the values - 00 or + 00.
(ii) l'(0)=O
(iii) v is countably additive, i.e.,

v( U A,
1-1
)= r:
'=1
v(A,)

for any scquen~e < A, > of disjoint measurable sets.

Froin this definition it is clear that a measure is a particular
ca'ie of a signtd measure. That is to say, every measure on A is
also a signed measure. The converse of this is not true in
general, i.e., signed measur.! is not a measure in general.
A singed measure v is said to be finite if
v (A) < co \f A E A.
15'1. Definition. Positive and negative sets. A subset E of
X is said to be positive relative to a signed measure v defined on
a measurable l-pace (X. A) if
(i) E is measurable.
(ii) ¥ ACE s.t. A is measurable =:t v (A) ~ O.
Notice that every measurable sub!;et of a positive set is a
positive set.
A subset E of X is said to be a negative set relative to a
signed measure v defined on a measurable space (X. A) if.
SIGNED MEASURE 267

(i) E is measurable.
(ii) y AC E S.t. A is measurable *
v(A) ~ O.
A set E( C X) is said to be a null set relative to a !>igned mea-
sure v if E b both positive and negative.
15'2. Distinction between a set of measure zero and a null set.
A measurable set E is a set of measure zero iff every measurable
subset of it has v measure Zero.
The measure of every null set is zero.
A set of measure zero may be a union of 1wo measurable sets
whose measures are not zero but are negative of each other.
Similarly we can explain difference between a set of posith'e
measure and a positive set. Note that 0, the null set, is a positive
sub:>et of X.
Theorem 1. A union of any coul/fable collection of positive
subsets of X is positive.
Proof. Let <. An> be a sequence of positive sets in X. Let
00
A= U Ai. Also let B be any measurable subset of A.
i=l

To prove that A is a positive set.

Write Bn=B nAn' n A'''-ln .... n A'J V nEN
Here dashes denote complement 0 respectIve sets w.r.t. X,
i.e., A,,'=X -·A n .
We know that if a set is measurable then its complement is
also measurabk. (\ Iso a countable inters~ction of measurable sets
is measurable. These facts lead to the conclusion that Bn is a
measurable 'iub~et of tIll' positive set All and hence v(B.) ;;;;a 0, by
definition of po~itlve set. From the con:.tIUction of B .. it is clear
that sets Btl are disjoint. Moreover
Of) 00
B,.= U Bn. Hence v(B)= U v(R,.) ~ O.
,.=J ,_1
Thus we have shown that
(1) A is a measurable set.
For An is a positive set
=> An is a measurable set

LJ
.. =1
A,. is measurable => A is measurable .

(2) ¥ BCA s.t. B is measurable set => v,B) ;;;;a O.

268 SIGNED MIlASURIl

By definition, this proves that A is a positive set

Similarly we can prove the following theorem:
Theorem 2. EI'ery subset of a 11egatire set is negative set
and a countable union of negati1'e sets is a negative set.
Theorem 3. If E is a measurable set with !inite negative mea-
sure, i.e. if - 00 < v(£)<O, then E contains a negative set A with
the properly v(A)<O.
Proof. If £ itself is a negative set then we may take A=£
and the t}leorem is proved.
Now consider the case in which E itself is not a negative fet.
Then E must contain a subset of positive measure.
This =? 3 El C E and a smallest positive integcr "1 S.t.
I'(El)~(J/nl)'
Evidently E=(E-E) u £1 and (E- £1) n E1=0.
Hencc V(E)=I'(E -£1) +1'(£1) ... (1)
or 1'( E-E1):=· v(E) - v(El~ ... (2)
Sim:e \.( E) is finite and hence (l) dcclarc~ that v(E E 1 ) and
I"(E1 ) both are finite.
Since I'(E) <0. Hence v(£) is a negative finite number. Then
(2) shows that r(E- E 1 )<0.
If E _. El is a negative set then we may take A=E--E1 •
Otherwise E-El contains subsets ofpositivc measure. Let /1 2 be a
smallest positive number S.t. there is a measurable set E~cE·· E1
with the property that I'(El »(I,'1/2)
E=[E --FlU E z1u(E1 UEt ), (E·-EtU E2 )n(l;;lUE2 )=0.
\.(E) = I'(E - FlU E 2 )+ I"(Etl+ I'(E.,)
+
=--o\'(E- El U £2) I'(E 1 U E~)
or 1'(E-E1 u E.!)=I'(E)-I'(E1)-l'iEt )<O.
For I'(E)<O,I'(E,»O for r=I,2.
Thus 1'(E-·-E1UEJ<-O.
This shows that E - E1 U E.: is a set of negative measure. If
E • £1 U E.: is a negative set, then we may take A=- E-:-EI U E l .
Otherwise we repeat the above process.
Continuing this process, we shall get either a negative subset
< 0 or a sequence < nk : k\ EN> of pO'iitive
A of E S.t. I'(A)
interers and a sequence < E" > of distinct me~surable ~ets s.t.
!- <
n.
v(Ek ) <: 0')
SIGNeD MEASURE 1.69

In the latter case, suppose that

A=( E- U Ei) ,_~J

.. (3)

A .. u.. ual, from this equation we can deduce that

v(E)=v(A)+v ( ~1 Ei )
00
=v(A)+ E l'(E~)
i==l

>v(A)+

... (4)

Since v assumes atmost one of values - co, and co, and veE)
is finite, therefore, (4) says that v( A) is finite and the series
00 1 .
E - IS convergent.
1'~1 n_
00
From (4), ~'(A)<v(E)- E
"~l n,.
=finite negative-finite negative
=a finite negative number.
... (A)<O .
(3) says that A is a measurable set.
For an enumerable u1}ion of m.!llsurable set~ is measurable,
and difference of two measurable sets is measurable.
It remains to show tlrat A is a negative set.
ror this let B C A be any arbitrary measurable set.
00 ;-1
BCA=E- U EkCE- U Eu
t=l ~~1

i-I
i.e. BCE- U £1'
k~l
270 SIGNED MEASURE

The choice of positive integer shows that

I
v(8) ~ --
11,-1 ... (5)
This is true "'if i E N.
Making ni--'>- 00 in (5), we get
vCR) < O.
Thus we have shown that A is a measurable set s.t. (v) A <0
and "'if RCA S.t. R is a measurable set => vCR) ~ O.
This proves that A is a negative set wIth v(A)<O.
This concludes the problem.
Tbeorem 4. If E i~ a measurable set oj finite positive measure,
i.e.,O < v(E)<oo, then E contains a positive set A wilh v(A»O.
Proof. If E itself is a positive set, then we may take A=E,
and the theorem is proved. '
Now consider the case in which E itself is not a positive set.
Then E must contain subsets of negative measure.
This => :I a measurable set El C E and a smallest positive
. 1
IOteger n 1 S.t. v( E1 ) < -- --- .
n1
Evidently E=(E-El) U El and (E-E1) n E1 =0·
Hence v(£)=v(E-EIH-v(EI) ... (1)
or v(E-El)=v(E)- v(El)' ... (2)
Since veE) is finite and hence (1) declares that veE -EJ and
veE!) both are finite. Given that veE) >0. Hence vee) is positive.
Also, by assumption, v(EI) is negative. Hence (2) shows that
vee-Ell is finite.
If E -El is a positive set, then we may take A=E--E1 •
Otherwise E - £1 contains subsets of negative measure. Let n2 be a
smallest positive integer S.t. there is a measurable set E 2 CE--£1
with the property that I'(E2) < -- n}- .
2
(E-El U E l ) n (E1 U E2 )=0
and E=(E-·El U E~) U (E1 U E2 ).
This => v(E)=v(E-E~ U E~)+v(E1 U E'IJ
=> v(E)-v(El)-V (E2 )=v (E-El ~ E2 )
=> v(E-·El U £2) < o.
For vee) > 0 and veE,) < 0 so that
-veEr) > 0 for r= 1,2.
SIGNED MEASURE 271

This proves that E-El U E2 is a set of positive measure. If

E-El U E2 i<; a positive set, then we may take A=E-El U E 2•
Otherwise we repeat above process.
Continuing this process, we shall get either a positive subset
A of E S.t. v(A)> 0 or a sequence < nt : kEN> of positive
integers and a sequence < E:, > of disjoint measurable sets s.t.
-00 < v(E!.)
.
< - n"-I
In the later case, suppose that
00
A=E- U E~. ... (3)

As usual, from this equation we can obtain

... (4)

00 •
or \'(E, < v(A) _ E 1
~,=I nlc
Since v assumes atmost one of values - 00 and 00, and veE)
is. finite. Therefore, (4) says that veAl is finite and the series
00 1 .
E - IS convergent.
II-I n~

00
Fr.>m (4), v( .4) > v( £) + l,'
t=l n"
=Sum of two finite positIve numbers
=positive number.
Hence veAl > O.
(3) says that A is a measurable set. For an enumerable
unior. of measurable sets is measurable, and difference of two
measurable sets is measurable.
Remains to show that A is a positive set.
For ,this Jet B C A be arbitrary S.t. B is measurable.
00 i-I
Then B C A = E- U E~ C E - U Et
.-1 .=1

I-I
or BeE - U E t
1:=1
272 SIGNED MEASUR,E

The choice of positive integers shows that

I
v(B) ~ - .
11, -1 ••• (5)

This is true viE N.

Making ni-~ 00 in (5), we get v(B) ~ O.
Thus we have seen that A is a measurable set S.t. veAl > ()
and "t B C A S.t. B is measurable ~ veAl ~ O.
Con~equcntly A is a positive 'iet with veAl > O.
This concludes the problem.
Theorem 5. (Hahn Decomposition Theorem). Suppose r is a
signed measure on a measurable space (X, A). Then :I a positil' . set
P and a negatii'e set Qs.t. P n Q=0, X=P U Q A being
a-a/~ebra of subsets of X. rBurdbw an 199.0, Punjab 1967•. 65 j
Proof. Let A be a-algebra of subsets of X. Let v be a signed
measure on a measurable space (X, A). Since v assumes atmost
one of the values - co and + 00. Without loss of generality we
can suppose that v does not take - co. Consider the family B of
all negative subsets of X and let
A=inf [I'(E) : E E B}.
Th en:J a sequence <En> In . B s.t. Lim v( En) \
-n.
n--* CO
B is a family of negative sets
::;. <Eft> is a ~equence of negative sets
oc
=> U E t is a negative set, by
i=l

Theorem 2.
=> Q is a negative set on taking
00
Q= U E,.
;=1

17hus Q is a negative subset of X. Then, [Link] to (1),

I'(Q) ~.\. Next we consider the subset Q-E.. of Q.
Q=(Q- En) U En
v(Q)=v(Q-En}+v(En ) <
v(En)
or v(Q) <; v(Em) V n E N.
En E B V n E N.
In view of (I). these two facts prove that v(Q) ~ .\.
SIGNED MEASURE 273

Thus we have shown that

v(Q) ~ >.. and v(Q) ~ .\.
This ~ v(Q)=1.
::>.\>-00
Next our aim is to show that P=X-Q is a positive subset
ofX.
Suppose the contrary. Then P is not positive and so P is
negative. Hence by definition. for every measurable set E C p.
vee) < O. Now E is a measurable subset of X with negative
measure. Making use of the result :
If E is a measurable set of finite negative measure. i.e.,
- 0 0 < veE) < 0, then E contains a set A with yeA) < 0,

we obtain a negative set ACE s.t. YeA) < O.

Since A and Q both are disjoint negative subsets of X and
their union A U Q is also negative. Refer to Theorem 2.
Consequently
v(A U Q) ~ .\, by virtue of (1).
But ,\ ~ v(A U Q)'= v(A) + v(Q) = v(A)+ >...
or .\ ..; v(A)+.\
or veAl ;;. O.
Contrary to the fact that YeA) < O.
Hence our assumption is wrong, i.e.," P is not positive" is
wrong. Therefore P is positive.
Thus P=X - Q is positive and Q is negative.
:. X=p U Q, P n Q=0·
A decomposition of a measurable space X into t~ subsets P and
Q s.t. X=P U Q, P n Q=0, where P and Q are positive and
negative sets respectively relative to. the signed measure v, is called
Hahn decomposition for the signed measure v. Also P and Q are
called positive and negative components of X.
Remark. The Hahn decomposition is not unique. If {Po Q}
is a Hahn decomposition for v, then we may define two signed
measures ~+ and v- s.t. y=v+-v- by setting
v+(E)=v(E n P)
v-(E)=-v(E n
Q).
IS·3. Singular measures. Two measures VI and V2 are said
to be mutually singular (in symbol "I .L v2 ) if 3 a measurable set
A C Xs.t.
v1(A)=0=Vi(X -A).
For example, the measures v" and v-, defined as above, are
mutuaIly singular.
274 SIGNED MEASURE

15'4. Definition. Jordan decomposition. [Punjab 67,65]

Let v be a signed measure on a measurable space (X, A). By
Jordan decomposition of v, we mean a pair (VI' Va) of mutually
singular measures VI and VB satisfying the condition V=V1-V I • For
example a pair of measures (v+, v-) defined as above, is a Jordan
decomposition of v.
15'5. Definition. Absolutely continuous measure function.
Let v and ~ be measure functions defined on a space (X, A).
The measure v is said to be absolutely continuous w.r.t. p. iff
p.(A)=O ¥ A E A ~ v(A)=O and is denoted by V ~ "..
In case ~ is a-finite, the converse is also true.
Theorem 6. (Radon Nikodyn Theorem). Let (X, A,,...) be a
a-!inite· measure space. LeI v'be a measure defined on A s.l. v is
absolutely. continuous w.r.t. p.. TlJen there exists a non-negative
measurable function f s.t.
V(E)=isfdf' ¥ Ee A.
The function f is uniqne in the sense that if g is any measurable
functioll with this property, then g(x)-=f(x) almost everywhere in X
w.r t. ".. {BuJ'dbw,an,1990;,Kanpur Statistics 1977, 76;
Punjab 66, 65; Banaras 64]
Proof. To establis,h the existence of the functionf, let us
assume that p. is finite.
~ is finite *v -fJ'" is a signed measure on A for each rational
number «.
Let T denote the set of all non-negative rational number.
For every t e T, let (Ph Q,) be a Hahn decomposilion for the
signed measure V-Ip.. If at ;;. t, then Q, is negative for v -11."..
For if E is a measurable subset of Q" then
v(E)-Glf' (E) <; v(E)-t.
For E is measurable subset of Q" at ;;. t
=- v(E)-fJp.(E) ~ v(E) -lp.(E) ~ 0
since - I I <; -to
[Link]> v(E)-atp.(E) ~ 0
~ (v-Gl~) (E) EO; O.
Similarly P, is positivt: for each signed measure V-att if at~t.
Write P= n {P, : t e T}.
Being an intersection of countable collection of measurable
sets, Pis mea-surable. Let E be any mea-surable subset of P. Then
SIGNED MEASURE 275
E c p=n P, => EC()P,CP, Y 1 e T
~ ECP, Y 1 E T
.. (v-tl') E :> O. For P, is positive
=> v(E)-tjJ(E) :> 0
~ veE) ;;;. II'(E) => t I' (E) <; veE) Y t E T
~ I'(E)=O or v(E)- 00.

Hence I'(E) > 0 => v(E)=oo.

Now we define a non-negative function
I: X-[-oo, CX)] by writing
f(x)=inf{t e T: x E Q,} Y x'E Z.
Here we adopt the usual convention that the infimum of any
empty collection of real numbers is 00.
Therefore I(x) = 0 Y x E P.
To prove that / is measurable.
Let r E Rand q E Q be arbitrary.
Suppose S.= U {Q, : y t < q}
= Union of measurable set
=measurable set.
Now definition of I declares that
{x E X :/(x) Et;; r}=() {Sf: q e Q s.t. q > r}
=intersection of measurable sets
. measurable set.
Thus {x EX: /(x) c; r} is measurable and so I is measur-
able.
Next our aim is to show that
V(E)-[Link]' Y EE A.

Let us now suppose that f'(Enp)-O.

If I'(E()P) > 0, then v(Enp)=O.

Also I" / I I
djA ;;;. E() P djA= co. This => "I dl'= 00.
[For /(x)=O Y x E P, hence/: E ... [ -00, 00] is not integrable]
Hence V(E)=ff/df'
This establishes the equality in case I' (E n P) > O.
Absolute continuity of v w.r.t. I' implies that v(Enp)-O.
Select a positive integer n and set .
Et:={X E E: ~~1 < lex) <; ~ J.
276 SIGNED MEASURE

For each rational number q > !:.,

n
we have
E1c C{x EX :f(x) ~ n}CS.= n{Q,: y t < q}.
Also Q, is a negative set for the signed measure v-ql' and
Elt i~ a sub~et of Q,. Therefore E" is a negative set.
(Refer Theorem 2).
This => (v-ql') (Et ) ~ 0
=> v(Ek ) ~ qp.(Ek )
k
=> v(Ek) ~ -n '" (E,,). .... (1)
On the other hand, since
k--ll
E" C { x E X :f<.x) > 11 f c Pet-t,/ft

and hence [ v -- (k~ 1) '" ] (E,,) ~ O. For P(k-t,/ft is positive,

This gives v(Ek ) ~ ( k ~ 1) 1'(Ek ). • •• (2)
From (I) and (2),
k-l) ",(E
(-n Ie ) ~ v(E/,) ~
k
Ii ",(E~). . .. (J)
Since E is a disjoint u"ion of sets En P and E".
Hence ",(Enp)='=v(Enp).
It follows from coulltahle additive prop .• y of integral that

ff Ii
d",= ~ froLt f
11=0
d"" v(E) = E v (Et )
k-l

By definition of E ••
k-l k
- ~f(x)~-on E".
n "
Applying first mean [Link] theorem,
n
(k-l) f
IA(E,,) ~ Etf(X) dx ~ k
n",(E".). . .. (4)
From (3) and (4),
( k---n° 1) ",(E lt) <f E/: f dx - v(Ek) ~
J k
n y(E.). .
Summing this over k.
- ~ I' (E) ~
n
I
l>
f(x) dx-v(E) <; ~ ",(E).
n -
Since n is arbitrary and hence making 11 ..... 00,

0= f£ f(x) dx-v(E) ~ 0,
SIGNED MEASURE . 277

or f
£ f(x) dx=v{E).
This completes the eXistence proof.
Second part. To prove the almost uniqueness of the function
J: X-+[ -00, 00].
Let g: X .[ -<X>, co] be any measurable function satisfying
the condition l'(E)== f,. g d". ¥ E e: A.
For each n e: N, let
A .. ={x E X :J(x)-J(x) ~ J/n} E A
and Bn={x e: x: g(x)-/(x) ~ lIn} E A.
By definition of An, I(x) -g(x) ;;t JIn on A...
Applying first mean value theorem, we get
fA .. (I-g) d". ~ ~ I£(A II )

or fA .. Id".-fAn gdfl ~ ~"'(A.)

or v(A..)-v(A,,) ~ (lIn) fl(A.)
or (lIn) ".(A .. ) ~ 0 or I£(An) ~ O.
But I£(A,.) ;;;;. 0 is always true
:. ".(An) =0.
Similarly we also have I'(B,.)=O.

Let C={x e: X : J(x) #g(x)} = U (A,.UB .. ). -

n_l

Then p(C)=1: p(A,.)+}J ,,(B..)==1: 0+1: 0=0.

Then p(C)=O.
" .. ."
This => f-g a.e., on X w.r.t. p.
Exactly in a similar way the theorem can be proved in case ~
is a - infinite.
Remark. The functionJ: X-+[ -00,00], defined as above,
is called Radon Nikodom derivativ~ of the measure v w.r.t. 1£. In
symbol, f= dv.
dp.
Theorem 7. (Lebesgue Decomposition Theorem).
[Punjab 69, 66. 65]
Lft (X, A. ".) be a a-finite measure space end v a.-finite
measure defined on A. Then 3 two uniquely determined measllres
\I. and 1'1 s.t. V= VO+V1 • Va 1. "., VI « ",.
278 SIGNED MEASURE

Proof. >'=I-'+v.
p. and v are G-finite implies that>. is a-finite. Evidently
" < ~ and v <c >..
p. <C ). stands for "p. is absolutely continuous w.r.t. )..".
By Radon Nikodym theorem, we are able to find non-nega-
tive functions/, g: X~[ -00, 00] s.t.
P(E)=J fdA, v(E)=f g d>. "I E E A.
'E E
Let A={x EX: f(x) > OJ, B={x EX: f(x)=O}.
Then X ··~UB, A()B=0. Furthermore p.(B)=LfdA=O:
Defir ~ two functions Vo, VI: A-.[oo, 00] by requiring that
Vo (E)=v (EnB), VI (E)=v (EnA) ¥ EEA.

Then Vo and VI are measures on A and satisfy the condition

V=VO+Vl'
v. (A)=v(AnB)=v (0)=0, or Vo (A)=O.
Thus p(B)=O=vo (A)=vo (X -B)
i.e. p.(B)=O==vo (X -B)
This ~ Vo is mutually singular to p.
;;:> vol-I-'·
To show that VI ~ p..

For this let EEA be arbitrary s.t. I-'(E)=O.

Then Isf dA=p(E)=O

or fE/ d)..=O. Also /(x) ~0 ¥ xE E.

This ~ f=O a.e. on E relative to ~.
Since f>O on A()E and hence
VI (E)==v (EnA) by definition of VJ
~" (EnA)=O.
:. VI (E) ~ O. But VI (E) ~ o.
Combining these two results, VI (E)=O.
Thus we have shown that ,.(£)=0 => VI (E)=-O
Th~ .~<~
To prove the uniqueness of Vo and VI'
To show that v. and VI are unique, let vo' and VI' be measures
s.t. 1I=v. +Vl' and has the same property as that of the measures
"0 aru1"vl respectively. Then v=VO+v1 and V= vo' +Vl' are the
two Lebesgue de.::ompositions ofv. Then Vo -VO'=Vl'-V1, Agaip
SIGNED MEASURE 279
v1' -VI is absolutely continuous and Vo- vo' is singular relative to
v. We have vo"'v.', V1==V1'.
Hence Vo and VI are unique.
, Problem 1. Let A and B be two measurable sets and V a signed
measure s.t. A C B and I V (B) I < co. Then show that
I v (A) I < co. .
Solutiou. Since B==(B-A)+A, (B-A) n A=ti.
Hence v (8)-=v (B-A)+v (A).
Hence if v (B-A) and v (A) both are w, then v (D) is w. If
v(B-A) or veAl is GO, then again v(B) is co. Given I v (B) I <co.
i.e., v(8) is finite. Hence v(B) will be finite if both v(B- A) or
v(A) both are finite. Hence I v(A) I GO.
This completes the proof.
~mark. The decomposition V= VO+v1 is called Lebesgue
decomposition of a-finite measure v w.r.t. a-finite measure 1£.
EXERCISES
1. State and prove Hahn decomposition theorem and the lQl"dan
Decomposition theorem far signed measures. [PunJab 1970]
2. State and prove Lebesgue decomposition theorem.
[Punjab 1969, 66]
3. State and prove Radon Nikodym Theorem. [Punjah 19661
 16
Product Measure

Introduction. Recall that

Xx Y={(x. y) : xE X, yE Y}.
Xx Y is called cartesian product of X and Y.
16'0. Definition. If ACX, Be Y, then
AxBCXx Y.
A x B is defined as rectangle and the components A, Bare
known as its sides.
Theorem 1. If (X, A) and (Y, B) be two measurable spaces
and if A E A, .dE B, then the rectangle A X B is a measurable set.
16'1. Definition. Let (X, A) and (Y, B) be two mca'.>urable
spaces.
Let xEX, YEY and ECXx Y.
Write E.,={y : (x, y)E E}.
E,,={x : (.~, Y)EE}.
Then E. is called section of E determined by x. Since this
section is determined by an element x and therefore Ez is called
X-section. Every section of a measurable set is measurable.
Therefore E.. , E, are measurable sets.
16'2. Definition. Let (X, A) and (Y, B) be two measurable
spaces. Let
xEX,yEY, EeXxY
be ar!'itrary.
Letfbe measurable function on E.
Write f.(y)=f(x, y),
fy(x)-/(x, y).
Thenf,(x) and f ..(y) are functions over the sets Eyand E.. res-
pectively. J. is called X-se~tion of J. Since every section of a
PRODUCT MEASURE 281

measurable function is measurable and !herefore j.. andly are

measurable functions.
16'3. Definition. Product measure. [Punjab 1967J
Suppose (Y, A, 1') and (X, B, v) are a-finite measure spaces.
Let EcXx Y.
Consider a real valued set function 7>,
i.e. 1t : Xx Y-R

l'.t. 1t(E)=J v(E.,) dJlo=f f,(E,,) dv

and 1t(A X B)=p.(A).v(B)

where AEA, BEB.
Then 1t is a measure function and is referred to as product of
Jlo and v.
Also we write :t=Jlo xv.
16'4. Definition. Double Integral.
IfI is Lebesgue integrable over a measurable set E, then

IE JI(x, y) dx dy
is called iterated integral and f f(x, y) d7C is called double integral.

If A is a measurable subset of E, then we define

fJ 6 l(x, y) dx dy= fJ I(x, y) XA(X, y) dx dy.

2. Fubini's Theorem. Suppose E=[a, b] x [c, dJ. Also
suppo'ie that I is lebesgue integrable over E. Then

tI lex, y) dx dy= 1: U: I(x, y) dy] dx

J fI(x, y) d~ dy= J: [J: I(x. y) 1dy
6 dx
(Banaras IV 1970, 67; Puajitb 67, 66J
Ex. I. State Fubini's theurem and explain why

J: 0: (;~;;a dX} dy# J: U: (~:~;:)2 dy}dX.

[Banaras 1970, 67J

where (x, y) E..~ [0, 11.

282 PRODUCT MEASURE

..·

...·

..·

Thus we see that

J: {J:[(X, y) dX} dy=-~#:

. . J: U: [(x, y) dy } dx

or J: {J:/(X, dX} J: {J:

y) dy# I(x, y) dy} dx.
The reason for this is that [ is not Lebesgue integrable Over
[0, 1] x [0, 1].
EXERCISES
1. State Fubini's theorem. Discuss its applicability to the
function
. x2_y2
f(x, y) (X2+y2)2'
X2+yl#O,
Xc [0, 11. y6 [0, 1]. [Banaras 1967J
PRODUCT MEASURE 283
2. State Fubini's theorem and explain why
Jl{Jl (X2+y2)2
o 0
y
X2_ lll
dx
}
dy=j:.
Jl {II
0
X II _
0 (Xl:
2
+y2)2
y
dy
}
dx
[Banaras IV 1970]
3. Define ptoduct measure. State and prove Fubini's
theorem. [Punjab 1967]
4. Develop the ~oncept of product measure to the state
needed to prove Fuhini's theorem and prove the latter theorem.
[Punjab 1966]
 17
Fourier Series

Introduction. A trigometrical series is a series of the form

00 +
...
1J (On cos nx+bn sin nx) ... (1)
tI~l

where the coefficients 0 0 , 01 • •••. .. 0'1, b1 , b2 , ... bn are independ~nt

of x. The problem of representing a given series by a series of
the form (1) was first discovered by Fourier in a problem of the
conduction of heat.
17'0. Periodic function. A function f(x) is said to be periodic
if f(x +1) =f(x) V x.
If I is the smallest positive number s.t.
f\x+/)=flX) ¥ x,
then the fUnction /(x) is called periodic function with period I.
Examples: (i) Each of the functions sin x, con x, sec x.'
cosec x is a periodic function wIth period.2n.
(ii) tan x, cot x both are periodic functions with period n.
(iii) sin nx, cos nx are periodic functions with period 2r./n.

( 2n) .
For sin n x+jJ =SIJl nx 'V x.
(iv) The sum of a number cf periodic functions is II. periodic
00
function e.g. 1: (an cos nx+ bn sin nx) is a periodic function with
...1

period 21t.
17'1. Finite discontinuit). A function f(x) is said to be
continuous at Xo if
Lim Lim
" .... 0 f(xo+h}= h~O f(xo-h)=/(xo}·
"If both the limits/(xu -0) and/(xo+O) exist, but are not
equal, then we say thatf(x) has a finite discontinuity at XU' The
F,Jurier series off(x) converges to i [/(xo+O)+f(xo-O)].
FOURIER SERIES 285

17'2. Even and odd functions. A functionf(x) is said to be

an even function if f( -X)-f(x) ¥ x.
For examole ~os x, Xl, x 3 sin x, x 2 cos x all are even func-
tions.
A functionf(x) is said to be an odd function if
f( ---x}= - f(x) 1.;1 x.
For example sin x, X2 sin x, x 3 cos x all are odd functions.
The geometric characteristic of the graph of an even function
f(x) is that the graph is symmetric W.I.t. y-axis and for the graph
of an odd function, we have symmetry about the origin.
17'3. Ortht)gonal functions. The functions fl' f2,'" ..... .in
L[ -ft,TC] are said to be orthogonal w.r.t. the intervall-1t, 'It] if

l. f".(x)f.(x) d~ =0 for m::j;n.

The functions arc said to be orthonormal if

I·
-w
f".(x)f.. {x) dx= 1°
1
if m:;l:n
)'f m=n
.
Th e functIOn I sin nx cos nx are orthonormal functions.
- - - ----
V2TC' V'" ' y1t
17·4. Trigono'metric polynomial. A polynomial of the form

Tn(X)::"~2°-+ 1· (a.. cos nx+b" sin IlX)

"=0
is called trigonometrical polynomial:
17·5. Fourier series. Suppose a fU,Detion f is Lebesgue integ-
rable over the interval [c, c+2lt] and
a
f(x)= f II
+,,':1 [an COS nx+b. sin nx] ... (1)

is lIniformly convergent in [c, c+21tl. Then

I
0 .. -=--
Ie+2tr J(x) cos nx dx, b.=- J
J c+2" f(x) sin nx dx,
,,7t r 1t c ..
(n=O, 1,2, ... ).
The series (1) is called Fourier series of !(x).
The constants
a.. , bn are called Fourier constants
17'6. We shall use the following result in the subsequent
articles.
286 FOURIER SERIES

Let m and n be any two integers.

(I') fe-fa"
c
[
cos nx ]C+21t
sin nx dx= ---n- c: -0 ifn:;t:O·

Similarly
f 0CHtf cos nx dx=O if n:;t:O.

(ii) J:t~. sin mx sin nx dx=i J:+ 21r

[CdS (m-n) x-cos (m+n)x]dx

=i [Sin m-n
(m-n)x sin (m+n)x
m-t-n
]c+lI
c
1t

=0 if m:;en.
(',,',') fC+2.
c: cos mx cos nx dx=i JC+2.
c [cos (m+n)~+cos (m-n)x]d.:t
=0 ifm:;t:n.
(iv)
.
jC+21t sin 2 nx dx= i
1<+2#
, '
(I-cos 2nx) dx=rc if n:;t:O.

Similarly 1 C+2.
• cos! nx dx= f .
IC+21t
(l +cos 2nx) dx
if n#O.
f
=1t

(v) f"+ C 2Ir

sin mx cos nx dx=i
.+rtf
C [sin (m+n)x+sin (m -n)x] dx
=0 ifm#n.
Here c takes any real value.
a
Theorem 1. If a functionf(x) = 2°
00
+n-l •
E (an cos nx+bll sm nx)

is Lebesgue integrable and uniformly convergent in [c, c+21t].

Then a.. =I- fC+~. lex) cos nx dx, bn= fC+ 2•
f(x) sin nx dx,
~ " "
(n=O, 1, 2, 3 ...... )
a
Proof. Let f(x) = 2° + ..001:_1 (a.. cos nx+b.. sin nx) ... (1)

In'egrating (I), '

1C+:.
c
a
f(x} dx=f·[c+27t-c]

+ :'1
00
a.. [ Jc+2tr cos ItX dx+ bn 1C+I.]
C 0 sin nx dx
=7[ao (all other integrals vanish)
FOURIER SERIES 287

or 00=- t
1t
r"+I. f(x) dx
0 ... (2)
Again, for any fixed positive integer m, we mUltiply (1) by
cos mx and integrate. Then

IC+II.
II [(x) cos mx dx= t le+I" cos
Q
II niX dx

TEa"
0') [0+2. cos mx cos nx• dx+ 00E b" JO+1. cos mxsin nX dx
_1 0 fI-l /I

Again the int~grals on the right vanish except one, namely

J cos II+2tr
1/ mx cos mx dx, (n=m)·
1/+2.
Hence 1 II lex) cos mxdx==a",.
/0+2. cos l mx dX=rt am

... 0",= -
I Ic+ttr
n D
/(x) cos mx dx • . .. (3)
Similarly, if we multiply (I) by sin mx and integrate, then we
can easily obhin bm = I-
rt
1 0

0
+1 • !(x) sin mx dx .

1 O+2• flx)
By (2) and (3). am =- , r
cos mx dx, (m=O, 1.2, ... )
1t •

This completes the proof.

Deduction: (i) Taking c=O, we get
I
Q,.=-
ft
J2' !(x) cos nx dx, for n=O, ], 2, 3,
0
o'

1 JlItr lex) sm
b,.=- . nx dx, for n=l, 2, 3....
ft • .

(ii) Taking c= -71, we get

a.. =-
1
1t
f' _"
"x) cos nx dx for n=O, 1,2, 3, ...

b,.= ~ I' f(.t) sin nx dx, for n=r I, 2, 3, ..

'/I: _"

Furthermore
00
f(x) = (0./2) + E (r" cos nx+h,. sin nx)
... 1
288 FOURIER SERIES

= 2~ [" fey) ~v+() In)'!l C. fey) COS ny COS nx

+sin ny sin nx) ~v

or f(x)= 2~ [ . f(y) dy+(l/lr) ,,~ /:,. fey) cos n(x-y) dy,

where
Remark. Remember the result of deduction.
Remark. If f(x) is given, then we can compute the coeffi·
cients associated with f(x) with the help of Theorem 1 and thus
Fourier series of functionf(x) is determined .. Now the question
arises that the Fourier series will converge and that will represent
f(x) or not.
A very wide class of functions possess Fourier series and that
do converge and represent f(x). If a function defined in the
interval [c, c+2n:] possesses Taylor series expansion, then it will
certainly possess a Fourier series representing it. If a function
does not posses Taylor series expansion, then it will not possess
a Fourier series representing it.
Theorem 2. If a function I(x) ;s uniformly convergent in the
interval [ -I, I] and

f(X)=~+E [ a" cos (~;X) +b" "in (n;x) ]

then a.=+ (, f(x) co's (n;x) dx, bn = +flex) sin (n;x) dx,
where (n=O, 1,2, ......... ).
Proof. The proof is exactly similar to Theorem I.
Remark. If I(x) is [Link] in the interval (c, c+21t)

and if I(x)=~ +"~1 [0,. cos (n;x) -tb.. sin (n~x)]

then
where
0 ..

n=O, 1,2,3, ..... .

+
=+1:+2" f(x) cos (n;x).dX' b,.= ]:+•• f(x) sin (n;x) dx
Deductions. For half range series:

(a) 00
If f(x)=:r+ .~1
fIC
o. cos (nnx)
-,- ,when I(x) E L [0, /)
 FOURIER SERIES 289

then all=~ 1: f (X) cos (n;x) dx, (n=O, 1,2.......... )

(b) Iff(x) = ,.~ h" sin (n;x) andf(x) E L [0, /],

then b"=7 /:f(X) sin (n~x) dX, (n=l, 2,3•....... )

Remember the above results.

Tbeorem 3. a
If S.,.(x)=/+ ':1
"" (a,. cos nx+b,. sin nx),

then Sm(X)=! (" fU+x) sin (~-tJ2!. dt

TtJ-... 2sm(tf2)
where f(x) has a period of length 2~.
Proof. Tt(a" cos nx+b,. sin nx)

=cos nx 1:"./(U) cos nu du+sin nx [" feu) sin nu du

=J:." feu) cos n(u-x) du

Also ao=!
7t
J"
-ow
feu) du

a m
Hence S",(X)=2....!!+ E (a" cos nx+b,. sin nx)
,.=1

=2-
I
"It
I"
_"
feu) du+ E -1 . . I'"
.. =1 "It _II'
f(u) cos n(u-x) du

=- ]J"
"It -"
feu) [)-2+ III cos n(u -x)
1:
,.=1
] du

=!J"
Tt _w
f(u) [Sin2 (~+l) (U-X)] du
I (u-x)
SIn
Putting u-x=t, we get
S",(x)=! ("-.. sin ~m+t) t .f(t+x} dt
"It L,,-o: 2 sm (6t)
Since the length of interval [ -1S -x, r.-x] is 2x and hence
We can replace this interval by another interval of length 2"1t
namely [ ~7t, "It]. Thus
1 J"
sin (m+l) t
S",(x)=;t _., 2 sin (it) f(t+x) dt
Hence the result.
290 FOURIER SERIES

Problem 1. Find Fourier series for f{x) = X2 , 0 < X < 21f.

Solution. f(x)=~-t:.~o (a .. cos nx+bn sin nx) ... (1)

a.=-I ICH. f{x) cos nx dx==- 11 2•

f(x) cos nx dx,
rr: 0 rr: 0
on taking c=O
=!rr: J~"0 X2 co.. nx dx=~rr: l{~ sin nx}2. - ~ J211
non 0
x sin nx dX]

= __ ~ [{_ X
mt
C~}2.
n 0 0
4: cos 2nlt
+J2" ~ dxJ= nrr: tl

or a.. =~ if n,rO.
n
ao=-
I f2. 1 fll1l
f(x)dx=- X2
1 87;3 8
dX=:'-3-=3- fl.l!.
lt 0 r.: 0 •

b.. =-1 J2" xl! sin nx dx=-1 [{ ---xl! cos IlX}2.

lt~ lt II 0

+ii2J2.
0 x cos I/X dx ]

1 [ - 4n
=- -
l cos 2nrc+"22 ( x sin nx )2. 2 1 sin nx dx1 --2
2•
rr: n n no 0 ~

=! [ __ 4rr:2 _ ~_ ( __ !) (cos nx)2"J= __ 4Tt.

It n n2 11 0 n
Putting the values in (1),
Ji\x ~,[4"2 cos nx--
l ) =8r.-2+.. 4rr:.]
sm IIX
6 tI-l n n Ans.
Problem 2. Expandf(x)=x, 0 < x < 2 in a half range (a)
sine series, (b) cosine series.

Solution. (aJ Let f(x) = !l b,. sin (~;x) .•. (1)

where b,,=; /:f(X) "in ("_;X) dx. Here 1=2.

b"=2 • X -2 dx-= [f. -x (co> 2

211 . (Ilrr:X)
SIR
"rr:~) ·n:f.
2 l2

2 • cos (n7tx)
+mc T dx ] JI
FOURIER SERIES 291

4 cos nlt+-.-
= [ -- 2 2 (. sm -n1tX)2] = -4- (-1)".
nrc nrc nrc 2, nrc

Hence (I) becomes X= - --

1: . (nrcx)
4 (- I)" sm -2
"=1 nrc

(b) Let
a
I(x)=f+
.
:1a"
01)
cos (nrcx)
-,- ... (2) Here 1=2.

a,,= f- t I(x) cos (n~x) dX=~ f: I(x) cos (n;x) dx

= r x cos (nltx) dx= (2X sin n1tX)2 _~ II sin ~ dIC
J8 2 nit 2 . nrc, 2
4 ( nrcx)! 4
=0+ T2 cos -2 =""""iZ
r
LCOS nrc-I]
nlt 4) nit

=-iz
n
[( -I)" --I] jf
Tt
n:;t:O.

:. an = ~.~ jf n= 1, 3,
nit
5, 7, ...... and a,.=O if n=2, 4 .... .

ao =1- J: J{x) dx=~ r: x dx=2.

Now (2) becomes, f(x) = I +}; 22 [( -I)" -I] cos -2
01) 4 nltx
"=1 n 7t
or f(x)=l-~[~
n: P
cos (r.X)+~ cos (3nX)
2.:z 2

+~ cos(5~X)+ ...... ]
Problem 3. Expand f(x)==sin x, 0 < x < x in a Fourier
cosine series.

.
Solution. We have a ':-'1 a. cos (nrex)
I(x)=t+ -,-
,

where a..=~ f~ Itx) cos (";X, dx. Here /=rc.

a.. _3 I-
rc 0
sin x cos I'IX dx
... (1)

=!Jit
n: ,
[sin (I-n) x+sin (l+n) xl dx
=_! [COS (I-II) X + cos (n+ 1)
rc 1-n n+1
xr if n:;Cl.
292 FOURIER SERIES

-~[l~n {COS (l~n) 7t-l}+ n~l {COS (n+l) 1t-I}]

2'
(nil-I) 7t [J +cos mc] if n;;i:J.

By (I), a1=-2J" sin x cos x dx=-1t1 J" sin 2x dx

1t 0 0

=l [-cos;x 1: =0.
'lo=~J'Ir
lto
sin x dx=~1t (-cos x)'Ir0 =~1t by (I).
Putting the values in (I).
4 aD 1
f{x)=2-+ 0 - E (2 1)
1t ,,~l n - 7t
[1 +cos nlt] cos nx

. 2 4[ 2 1 -l cos2x+ 1_ cos4x+ 1_ cos6x+ ... ]

or smx=;-; L 4ll 1 62 1
17'7. The L2-theory of Fourier series. A functionf is said
to be square integrable over [0, 27t] when / is measurable and
J:tr /2 dx < 00. In this case we also write f2 E L [0, 21t] or
fELl [0, 2n']. The space of square integrable functions is
denoted by the symbol L2 [0, 2n].
Theorem 3. Parsevel's identity for Fourier series.

Let the Fourier series '!2J!+ E(a" cos nx+b" sin nx)
"=1
of f(x) converge uniformly to f{x) at every point of the interval
(0, 21t) and Iet/{x} E L (0, 21t). Then

a
Proof. Letf(x)=t+ ,,:'1 (a.. cos nx tb.. sin "x)
00
._.(1)

Multiplying (1) by f<x) and then integrating. we get

 FOURIER SERIES 293

l°·W [f(X)]2 dX=-2°

a J2W f(x}
..~
00
dx+ E [ a. J:~ f' x) cos nx dx
+b. J:-It X) sin nx dx J
Making use of the fact that
1
a ll =- J2. lex) cos nx dx, b.=J2W f(x) sin nx dx,
n 0 0 .

we get J:-
o
[[(x)]· dx-=(ao/2).7Wo+ 00
E [t7o"l+nb,,'].
11-1

Problem 4. To prove tltat for all positive integers m.

a,1 +
2 n-l
(a"lI+b,,2) ~ ~I
E -I
J'
[/(x)]1 dx

where all and bll are the Fourier coef/icientl of I(x). Ilx) being
piecewise continuous in (-I, I).

Solution. We have/(x)=~+ !1 [a" cos (n;x)

+b" sin (~~)] ... (1)

where a,,=~ f, f(x) cos (n7~ ) dx 1
b"=rIJ'_,I(x) sm. (mcx)
... (2)
--r- dx.

'fake S,"(x)-~- + !1 [all cos (y)+b n sin {n~x)]

... (3)
Evidently [[_-S",]2 ~ o. Hence JI [f-SrnlH dx ~ O.
-I

or [,JI dx+ t,SIII dx ;;;a. 2

2 dx f/ Sil ... (4)
Multiplying (3) by f and integrating, we get in virtue of
(2), as
294 fOURIER SERIES

or ... (5)

Squaring (3) and then integrating,

I~, S.- dx=!!!i.2/+ !ix t, [0,,2 cos2 (n~x) +b..2sin (~~~

2 )] dx

+2 Etil
,,-1
I' [
-I
all an cos (nnx
-,- ) +ao b" Sill
. \nnt-) J dx

... (6)

For fl cos (n~x ) dx==O= f~, sin \ n~x ) dx

and f, cos 2 r~X) dX=1 f, [1 +cos (2'~"~)] dx=~I·-O""1 eh

Putting the value from (5), (6) in (4) and simplifying, we
obtain ~

or 1
-I J'-I
P dx;;. a2° + /1=1
2
L' (a,,2 +htJ2)
Iii
... (7)

Hence the result.

Deduction. Taking lim it as m .. 00 in (7). we get lkssch.
inequality namely

Tlleorem 4. Riemann-LebEsgue theorem. !If E:: L [-1t, nj

and if < an > and < bn > are FOl/ri('r c()[Link].~ of lJ functior
lim lim
f , then n-?oo 0 11 =0, n-+oo b,,-=O.
FOURIER SERIES 295
Proof. Fix Ii > O. Define an unbounded measurable func-
tion g on [ -n, 1t] s.t.

J:" 1 f{x)--g(x) I dx < j!. . .. (1)

Then g E 1..2 [ -1t, n] so that
<:lO
, E (A ..=+Bn 2 ) < 00 ... (2)
».=1

where A,,=-
1 J"_11' 7t
g(x) cos nx dx

B,,=! J" g (x) sin If X dx

-'It 1t
Lim
(2) :=> E A,,2 < 00 ::> A,,=O
n-+oo
::> 3 no E N s.t. I An 1 < e/2 ¥ n ~ no· ••• (3)
For any n,
lan-An I =11 1t
(II'
1-11'
rf(x)-g(x)] cos nx dx
~- 7t~J"_" If--g I dx <!.
7t
112
11.

Thus I a,,--A.. ! < (6/2)

a"=(a.• -A.. )+A,, ... (4)
E e
I a" I <. I a" --An I + I A .. I < T+2" =E It 11 ~"o
[This follows from (3) and (4)].

This

Exactly in a similar fashion we can show Lim b =0.

n- 00 It
Theorem 5. Establish Dirichlet's integt;,al i.e., iff E L (0, 2,,)
and if S" deflates the nth partial slim of the Fourier series ofivc).
thell
1
S", (x)~==-2
J'It sill (n+l) u
---. - '(-I .')- [nf(x+u)+f(x-u)] duo
7t 0 Sill nU
[Banars 1961, 63]
U m
Proof. Sm (x)= ~-+ 1: (a" cos flX-t btl sin I1X). . _.( I)
- "~l
296 FOURIER SERIES

with Q,.=!KJo(2" f(t)

nt dt, b,.=! 2• f(t} sin nt dt
COS
n 0
1
Putting these values in (1),

1t Sm=--K1 Ill..
0
f(t) [ i+ m1:
n-l
(cos nt cos nx+ sin nt sin nx) ] dt

or [Link]=J:' f(t} [~ +n~l cos n (t-x) ]dt

Put t-x=u. Then

1t Sm= J2"-« f(x+u) [I-2+ m

~ cos nu ] dll
-~ n-l

. "! sin (m+i) u

Smce j+ 1: cos nu= - 2 . (i) Dm(u), say.
,.~l SIn u

Hem:e 1t,Sm=f_r"II-1I f(xtU) Dm (ll) du.

. .. (2)
Evidently Dm (u) is a periodic function of period 2/1 and feu)
is given to be periodic fUI)ction of period 21t. Hence the interval
[-x, 21t-x] can be replaced by another interval of length 2 ..
namely [0, 21tJ. Thus

Tt Sm= J•2" l(x+u) Dm (u) du= I"0 f(x+u) Dm (u) du

+J:" f(x+ u) Dm (u) du

Putting u= -t in the second integral,
71: Sm=I' f(u+x) Dm (u) du+
o
r-
J-1I"
2
.. flx--t) Dm (-t) (-dt)

-= 1: f(u+x) Dm (u) du+ r: I(x -I) D", (t) dt

as Dm (/)=Dm (- 1)

=[ f(x+u) Dm (u) du+[ I(x-t) Dm (I) dt, by the

same reasoning discussed above

. . 1: [/(x+IIJ+/(x-U) Dm (1I) du

or SIII=l
1t
J*' [f(x+u)+f(x-u)]
0
Dm (u) du
FOURIER SERIES 297

or S,,= ~
11:
J" [!(x+u)+!(x-u)] Dn (u) duo
0

17'7. Summation of series by arithmetic means. [Burdbwan 1990]

Let E an be a series of real numbers with partial sums

If l: an is not convergent, i.e., if lim S" does not tend to a limit,

then it is sometimes possible to associate with the series a 'sum'
in a less direct way. This simplest method is known as 'Summa-
tion by arithmetic means.'
.
W rIte a,,= Sl +S2+n "+S,,
- .
Then a .. is the arithmetic mean of the partial sums. If S" -+ S,
then also (1,,- S. But an may tend to a limit even though lim Sf!
does not exist.
A series for which a" tend to It limit, is said to be summable
by arithmetic means, or by (C, 1). Hence C stands for Cesaro's
fur first order.
Examples (i) E( - 1)" is summable (C, 1) to the sum i.
(ii) I+O-I+I+O-l+ ............ is !>ummablc (C, I) to the
sum 2/3.
Theorem 6. Fejer's integral. Establish rh,' illtegral
-_ 1
0,. ~2--
nn
1-
0
sin 2 (.nll)
--'~(l
Sin 211)
,
• lflX+U)+!(X-U)] duo
.

Proof. Let S,. denote the 11th partial sum of the Fourier
series of f. Then

S,,=!
R
j" 0
l.!(x+u) +/(x-u)j Dn (U) dx.
... (3)

Wh ere l. ~~ sin-(11+.) U
D,. (U)=2+ ~ cos mu .... - -,-- - --
f)bl 2 Sill (iU)
[Prove this as in Theorem 5].

Write

then, by (3), a .. _1-.

nn
J" [/(x+u)+f(x-II)) [
0
l·
",_1
Em (u) ]dU
298

nut 1: Dm (u)= 1 sin ('!z+i)!~ = sin2 (/~'!.!.3l.

m",l m~l 2 sm (iu) 2 Sll1~ (~U)
. 1 sin 2 (nuI2)
Taking Kn (u) =2- . 2( ) , we get
,t Sill tu
r1n=~ f1t £!(x+u)+f(x-u)] Kn (u) du
n 0 ... (4~

or r1n =-21- Jao [ f(x+u)+f(x-u) ] sin 2(nuI2)

12 duo
2
n!; 0 SID (U )

This is known as Fejer's integral.

Remark. It was Fejer who found that the operation of'
"summation by arithmetic means" is applicable to the Fourier,
series
178. Sumtnability of Fourier series.
Here we shall discuss the case, for what values of x the
Fourier seri,!s of f e L [- Jr, Jr] is (C, 1) summable to f(x) ? i.e.,
lim
for what x e [-Jr, n], a.. (x)=f(x).
[Link]
1 ,.
Here all (x)=- E Sm (x). We have se~n that
n m=l

(1n (x)=.!n I'"a [!(x+t)+f(x- I)] K,. (t) dt

... (4)
[This is equation (4) of Theorem 6]
tn a gpecial case if l= I, then Sn(x)=1 V Il so that
all (x)=-=(nln) = 1. Now (4) gives

1=-J 2 ft
Kn(t)dt or /lx)=-
2
f(;'(.)K.. (t)dt (5)
jft
no .. ·
1
1t 0

(4) - (5) gives c. (x) l(x);..+..;2 J1t0 ({(X+ U +fl .\ --

'--~2'-- - t) -·f(x) Kn (t) df

lim lim.
If
n .... oo
a" (X)=/(x), i.e.,
n-oo [a .. (x)-j(x)]=O
2
then -;- JIt [!(X+l)
0----2+f(x -I) --J~x)
'... --- 1 K.. (1)=0 as Il~OO.

Theorem 7. 1// is continuous and Ie L t --tt, ttl. then

~CL+ £ (a .. cos nx+bn Sin I/x)=f(x), (c, I).

:2 .. _1
FOURIER SERII:.. S· 299 1
Proof. We have seen that
1 ~ 'It [f(x+t)+I(x-t)] Dn (t) dt
Sn (x)=-
1t 0

a,. (x)=~ ; Sm(X)=-~ Jft [f(x+t)+/(x-~t)] [ EDm (t) ] dl

n m~l nr. " m=l"
1 ·n J 11 sin (m+~) t
.>Jut - X Dm (/)-- E -~--'---t-"
n 111_1 n 111-1 2 f>in ( "2 )
= sin:! (n+.n/ =Kn (I)
2n sin 2 ~ '2 )
1 11
Hence u,. (x)= - J l/(x+t)-t-/(x--l)] Kn (t) dt
... (I)
1t 0

In a special case if 1=:= 1, then S .. (x)= 1 ¥ n so that

011 (X)= (njn) = 1. Then (1) ~
2 11
1=-j Kn(t)dt
It 0 ... (2)
2
or I(x)= 1: 10" I(x) K" (f) dl ... (3)
(1)-(3) gives

I CI" (x)-fvc) I"'" ! 11: [J(x+tH f(x .-. t)-2/(.t)] K .. (I) dr\
< !- lir
1: 0
[I /(.t+t)-/(x) 1+1 (X--'I) -f(x) I ] I K" (t) I dl ... ( 4',
f is continuO\JI; ~ I l(x1) - f(xz) I -< 6
whenever I X 1 -X2 I <; s.
1 ft
:. I CI" (x) -f(x) I< ~
It
J0 «+1) I Kn (t) I dl
or I on-f I <. 24
1T
J"
0
I K .. (1) I dt
. (5)
Here ! Xl - x 2 I < a ~ I x±t-x l < 0~ It I< S
~ 0 '" t < o.
Also 0 < 8 <: 7:.

From (2) and (5), 1 0 ,. -II < 1C 2'

2e 1C
=.
or I an (x) -/(x) ! < e ... (6)
¥ nand 0 < t -< 8.
300 FOtJRIER SERIES

t
If t ;;;a 8, then Kn (I) " 211 sinl -(8j2Y

· In\!J"
•• 0
[}i(x+t)-/(x)+!(x-t)-!(x)]KtI{t)dt II
~ i 1: [f(x+t)-/{x} I+I/(x-t)-f(x) IJ· ~2n~ si!2 (8/2) dl
Now (4) declares that
IOn (x)-f(x) I < '1 ... (7)

.By (6) and (7), lim Ott (x)-/(x), (C, 1).

n-+oo
 18
Banach Space

18'1. Vector Space or (Linear Space). Let (F,+, .) be a field.

Then a non-empty set V together with two binary operations
vector addition and sca lar multiplication is called vector space
o¥er tne field F if the following conditions are satisfied:
. '(i) (V, +) is an abelian group, i.e.,
+ ),
(1) V is closed w.r.t. ( i.e.,
• 11- U, vE V=> u+v E V.
(2) A<fdition is associative in V, i.e.
(Il+V)+W=u+(v+w) 'V- u, v, wE V.
(3) There exists a unique vector 0 called zero vector in V s.t.
u+O=O+u=u 'V- U E V.
(4) Given any uE V,:I a unique vector -u E V S.t.
u+( --u)=( -u)+u=O.
(5) Addition is commutative, i.e.
u+v=v+u 11- u, VE V.
(ii) 'I E V, a E F => au E V (scalar multipltication law).
This law must satisfy the following conditions:
(6) a(u+v)=au+av.
(7) (a+b) ll=au+bu.
(8) (ab) u=a (bu)
(9) For the unity element I EF
lu=u
where II, V, W E Vand a, b E F.

Instead of saying "V is a vector space over a field F", we

always say "V(F) is a vector space."
The elements of V are called vectors and clements of Fare
called scalars.
When F=R the field of real numbers, then VCR) is called real
vectors space. Similaqy V(C) is called cpmplex vector space.
302 BANACH SPACE

18'2. Subspace. Let V(F) be a vector space. A non-empty

subset Wof V is called a subspace of V if W itself is a vector
space over F w.r 1. the operations of vector addition and scalar
multiplication in V.
The spaces to} and V(F) are called trivial subspaces or
improper sub<;paces of V, others are said to be DOD-trivial or proper
subspaces of V.
18'3. Linear sum or sum. Let U and W be subspaces of a
vector space V(f). The sum of U and W is denoted by U+ Wand
is defined as U+W={u+w: U E U, w E W}.
18'4. Direct sum. A vector space V (F) is said to be the
direct sum of its subspaces U ahd W if every v E V is uniquely
expressible as v=u+w where U E U, w E W.
We denote direct sum of U and W by UEB W. Evidently
V=UEBW';' V=U+W.
Note that V=UEBW ~ V=U+W, Un W={O}.
In this case U and Ware called complementary subspaces
of V.
If U n W={Ol, then U and Ware called disjoint spaces.
18'5. Quotient space. Let W be a subspace of a vector space
V (F). Let v E V be arbitrary. Then
v+ W={v+w : w E W}
is called a coset of Win V.
Denote the collection of all cosets of W in V by VI W. Then
VIW={v+W: v e V}
V/W is a vector space over F w.r.t. vector addition and scalar
mullipiication as follows:
(u+ W)+(v+ W)=(u+v)+ w
a(u+W)=au i-W \f u, v E V and a E F.
This vector space V; IV is called Quotient space or Factor
space of V by W.
18'6. Basis or Hamel basis. Any subset S of a vector space
V{F) is said to be basis for V if S is linearly independent and
vectors of S generate V i.e. L(S)= V.
The idea of base or basis is due to German mathematiciar
Hamel and so basis is also called Hamel basis.
There exist more than one basis for the same vector spac1
V{F) but the humber of elements in each basis is the same.
The number of elements in a basis of a finite dimension
 BANACH SPACE 303

vector space is called dimension of V. In brief dimension V is

denoted by dim V.
1S·7. Linear transformation. Any map T: X(K)-ioY(K) is
called linear transformation or linear map or homomorphism if
T(Oll+bl')=aTu+bTv V- a, b E K and ll, v E V.
A linear map T: X(K)~X(K)
is called linellr operation on X.
A linear map T: X(K)-ioK
is called linear funcHonal on X.
A linear map T: X(K)~Y(K)
is called isomorphism if T is one-one and onto.
The sec of all linear maps from X into X is denoted by
L(X, Y)
and is a vector space over K w.r.t. vector addition and scalar
multiplication. i.e.,
(Tl +T2) (1I)=T1u+T2u V T 1• T2 E L(X, Y)
(aTl ) (u)=aT1u ¥ a E K
where 1I E X.

A special case of L(X, Y), i.e., when Y= R. The clements of

L(X, R) ar~ called linear functionals and is denoted by X*.
IS·8. Partially orllered set. A relation ~ defined in a set
A with prorerties li~,ted bleow
(i) refl ;xive, x ~ x
(ii) antisymmetric. x ~ y, y ~ x '"" x=y
(iii) transitive, x ~ y, y ~ Z:::? X ~ z
¥ x, y, Z E A
is called a partial order in A. In this case the set A is called a
partially ordered set and is denoted by (A, ~) or simply by A.
Any two elements x, y in a partially ordered set A are said
to be e()mpar~ble if either x ~ )' or y ~ x.
If ev..:ry pair of elements of a partially ordered set is compar-
Llble, then A is called chain or linearly ordcrcil or totally ordered
set.
]8'9. Minimal and maximal clement... Let (A, ~) be a
'p<[Link];aily ordered set. An clement a E A is called a maximal
clement of A if
x ~ a for every comparable element x E A.
An clement b E A is called minimal element of A jf h ~ x
• for every comparable element x in A.
304 BANACH SPACE

18'10. Supremum. Infimum. See def. in Chapter 4.

18'11. Zorn's Lemma. If S is a partially ordered set in
which every totally ordered subset has an upper bound, then S
has a maximal element.
18'12. Normed Jinear space. Let V(K) be a vet or space.
For any u E V, we define a function (norm function)
II II: V~R s.t. ¥ x, Y E V,
we have (n I ) II x II ~ 0, (n 2 ) II x II =0 if x=O
(na) 1\ (XX II = I ex I . II x II , O l E K
(n 4 ) II x+y II <;; II x II + /I y /I •
The pair (V, II II) is called normed linear space. Here K=C
or R. The real number II x II is called norm of vector x.
18'13. Bounded map. Let X and Y be normed linear spaces.
A linear map T: X-Y is said to bounded if there is an M s.t.
/I Tx II ~ M II x !I ¥ x E: X, M > O.
Theorem 1. Let N be a normed linear space. Let d be J
function on N defined as d(x, y)= 1\ x-y 1\ ¥ x, YEN. Then dis
a metric on N
Proof. (I) x, YEN => x-y E N.
Since II x-y II ;a. 0, by (n I ). Hence d(x, y) ~ O.
(2) d(x, Y)= II X-Y II = II (-1) (y-x) II = I (-1) I .11 y-X II
= II y-x U =d(y, x)
:. d(x, y)=d(y, x)
+
(3) d(x, z)= /I x-zlI=lI{x-y)+(y-z) II ~ II x-y II lIy-zll·
by (114) =d(x, y)+d (y, z)
or d(x, z) ~ d(x, y)+d(y, z).
Thus d is a metric on N.
Remark. Thus we have seen that a normed linear space can
be treated as a metric space. Thus we have given a topological
structure to normed linear space. Hence we can define closed
sphere, limiting point etc. in a normed linear space.
18'14. Banach space. [Kanpur 1984]. Any sequence <Xn>
in a normcd linear space (N, II II) is said tv converge to Xo E N,
i.e., xn-+xo if given E > 0, 3 positive integer no S.t. n ;;? no
~ II Xn - Xo II < E. It follows that
lim 0 'f
n-+oo II xn-xo II = I xn~xo·
A sequence <XII> in N is called Caucby sequeD~e if given
• > 0, 11 positive integer 110 s.t.
m, n ;;> no ~ II X",-x,. II < E.
BANACH SPACE 305

A normed linear space is said to be complete if every Cauchy

sequence in N is convergent in N.
A normed linear space which is complete as a metric space,
is called a Banach Space or B space. [Kanpur [Link]. F. 1983]
Examples. (i) The vector space C [0, I] with the norm
II x II =max { I XCI) /, t E [0, I]}
is a Banch space.
(ii) The vector sp8.G,e C. of convergent sequence wiih the
norm II x II =sup { I x-I : n=I, 2... } [Kanpur 1989]
is a Banch space
(iii) The vector space 11" p ~ I of sequ,ences for which

.! I x.. \p < 00 with the norm II x II =[ !) x. 11' J/P

is a Banach space.:
Theorem 2. To show that (LP, d) is metric space.
[KoJhapur 1968]
Proof. See Theorem 9, chapter Il.
IS·1S. Continuity. Let Nand M be norilled linear spaces.
A map f: N~ M is said to be continuous at Xo E N iff given any
• > 0, a 8 > 0 s.t. II X-~o II < 8 => IIf(x)-f(x.) II < t. )
The map f is said to be continuous at every point of N iff it
is continuous at each point of N.
We also say that f is continuous at Xo if for every sequence
< x" > E N converging to xo. the sequence < f(x~) ~ E M
conver~ea tof(xo) E M, i.e., if
x" .. x o =:;,.'f(x..) -+ I(xo).
IS·16. Bounded map. A linear map T: N-M is said to ~
bounded map if a a real number k ~ 0 s.t.
II T (x) II <; k II x II 1.;1 x E N.
18·17. Isomorphism. Let Nand M be normed linear spaces
and T: N~M be a linear map. Then T is called topological
isomorphism if
(i) T is one-one and onto
(ii) T and T-l both are continuous maps.
IS· IS. Isometric isomorphism. L~t Nand M be Donned linear
spaces. A linear map T: N-M is called isometric isomorphism
if T is on e-one
and 11 T(x) II = II x U ¥ x E N.
306 BANA-CH SPACE

18'19. Grapb of a function. [Kanpur 1980)

Let/: X- Y be map. Then graph ofa function lis defined
as a set consisting of elements of the from (x. /(x» where x EX.
Thu,> graph/={(x,f(x» : x F: X} C Xx Y
as/Ix) E Y.
Symbolically we write graphj=/"
18'20. Closec! linear map. Let N and lV' be normed linear
spaces. A linear map T: N • N' is said to be closed map if
< x n > eN. xn .... x. T(x.)-y
~x E N, where y= T(x).
18'21. Some elem~ntary defil1itions. (iJ Let (X, d) be'a
metric space and A C X. A point Xo E X is called limit point or
limiting point or cluster point or accumulation point if every
open sphere S~ (xo) contains points of A other than xo, i.e.,
(Sr (xo)-{xol) n A~ fl)
S,(xo) = open sphere of radius r and centre Xo
={x EX: d(x. Xo) < r}.
(ii) The set of all limit points of A is denoted by D (A) and
is called derived set of A.
(iii) A sequence < Xn > in a metric space (X. d) is called a
Cauchy sequeuce if given c > 0, 3 positive integer no S.t.
m, n ~ 110 ~ d(x,., xm) < f.

A Cauchy sequence of real numbers is a convergent sequence.

(iv) A point x of a metric space (X. d) is called an interior
point of a set A C Y if 3 ope 1 sphere S, (x) s.t. S, (x) C A. '
The set of all interior points of A is denoted by A' or by set
(AO). AU is always an open set AO C A. Also
A is open do A=Ao.
(v) Closure Let (X, dJ be a metric space and A C X. Then
closure of A is denoted by Aand i" defined a3
A = n {F: F J A, F is closed, F C X}
A is closed and A C A
'A is closed -<* A=A
(vi) Adherent or contract point. A point x E X is called
adherent point of A if every neighbourhood of x contains a point
of A. Thus it is clear that if x is adherent point of A, then
(i) x is a limit point
or (ii) x is not a limit point of A but x E A.
 BANACH SPACE 307
(vii) If x is not a limit point of A, then x is called isolated
point of A.
(viii) Den' e set. A set A is called somewhere dense set if
int (::t)=(At:;t:0
i.e. if closure of A contains some open sets.
Dense Subset. Let X be a normed linear space. A subset A
of X is called dense subset of X if every Cauchy sequence in A
converges to a point in ]{.
4 is called nowhere dense set or non-dense if (.4t=0.
The space X is said to be Separable if 3 ACX st. A is count-
able s.t. A=X.
(iv) Baire's Category Theory. Let (X, d) be a metric space
and A C X.
The set A is called of first. category if it can be expressed as
a countable union of non-dense sets. The set A is called of the
second category if it is not of the first category.
Baire's Category Theorem. A complete space is a set of
second category.
Result. A closed subspace of a complete metric space is
complete.
18'22. Jl"unctional Conjugate Space. Let N be a normed
linear space over the field K of '>calars, where K=C or R represents
respectiveiy the set of complex numbers or real numbers.
Then any continuous linear mapf:N~K is called functional.
The set of all such functionals is denoted by N*. The set N* is
called conjugate space (or adjoint space or dual space) of N. The
set (N*)*=N** is called second conjugate space of N.
[Kanpur [Link]. F. 1983]
Convergence. Let Nand N' be norm~d linear spaces. Let
B(N, N/) denote normed linear space of all bounded linear maps
Tn: N ... N'.
(i) Strong Convergence. The :.equence <T.. > of B(N, N')
s
is said to converge strongly to T, written a; Tn - - + T if
¥ u E N, Tn (u)~T(u)
i.e., if given f > 0, :I no E N s.t.
II Tn (u) - T (u) 1\ < Ii ¥ n ~ no·
(ii) Uniform Convergence. The sequence Tn E B(N, N/) is
said to converge uniformly t~ T
if given fi > 0,:1 no EN s.t.1\ Tn--TII <_ ¥ n;;;;t no·
308 BANo\CH SPACE

(iii) Weak Convergence. < T" > E B (N, N') is said to

. w
converge weakly, written as T .. -~ T
if ¢[T. (u)] _ I,6[T (u)]
y 1,6 E N*, dual space of Nand ¥ u E N.
It can be proved easily that uniform convergence 0 strong
convergence .. weak conveJgence.
18'23. Projection. Def. A projection on a Banach space
B is a linear map S.t. Ea=E and E is continuous map.
Tbeorem 1. Let N be a normeo linear space and x, yEN.
Then I II x II - II y II I ~ II x-y n . [Kanpur [Link]. F. 1987.89]
Proof. Evidently
!I xII = II (x-y)+yU ~ Ilx-yll + Ilyll
. as II u + v II EO; II u II + II v II
or IIxll-lIy!I~lIx-YH ... (1)
Interchanging x and y in (I)
II y II -- II x n<; II }' -x II = II ( - J) (x - y) II
=1-11 .lIx-YII·= nX-YIi
or II y n - ux II <: II x- y II as UIIU II = I GI 1 • II u II
or - ( II x II - "y II ) ~ II x -- y II
or IIxll-IIYJl~--lIx-YIl ... (2)
By (t)and (2), -llx-JyIl ~ IIxll-IIYII E;; Ux-YII
or III x II - II ~ 11\ <; II x - y II
[For -4 ~ a ~ 4 ~ I a 1 <; 4]
Theore~ 2. Let N be a normed linear space witil norm II II.
Then the mBp/: N-+R S.I.f(x)= II x II is continuous, ,hal is, the
norm II lion N is a continuous map. [Kanpur 1987; Meerut 87j
Proof. Let <XfI > be a sequence in N s.t.
lim
n_ 'Xl x.=xo ... (1 ,

Given/(x)= II x II ¥ x E N.
Then I/(x,,)-/(xu) 1 = I II x,,11 - II Xo II I
~ II X,,- Xu II , by Theorem I.
-+0 as n-+ oo , by (I).
:. I I(x,,) -/(xo} 1 <: 0 as n __ «>
But I I(x,,) -/(xo) Il( 0 is always true.
:. 1Ilx.)- Ilxo} I = 0 as n .. 00.

This lim /(XfI)--f(XO)

0
n_ou ... 12)
By (1) and (2), we see that
x,,· ... x• .. I(x,.)-+/txo)
 BANACH SPACE 309

This, by definition, proves that! is a continuous map.

Lemma. If a ;;, 0, b ~ 0
J I
and -+-=I,I<p<oo
p q
aP b fl
then ab ~ -+-
"" p q
.
Proof. If 0=0 or b=O, the lemma is obvious.
So Jet us suppose that a > 0, b :> O.
Define !(t)=(I-.l)+At-t A ••• (1)

for 0 < A< 1.

Then !'(t)=>'-A!).-l=A (l-t A- 1 ) ... (2)
Also 0 < A < I ;> I-A> 0
f'(t) = -A (A-1) t).-2
or .f'(t)='\ (I >.) tA-I. ...(2a)
Now /'(t)=O gives>. (1- 1).-])=0 or t=l as A < 1
At t=I.0 <,\ < I,J:(t)=positive. by (2a).
This proves that the function has minimum value at t= 1.
Consequently !(i) ~!(I)=O.
or !(t}?;J0 or l-A+At-I).;;;'O .. (3)
Write ,\== IIp. This is justifiable step as Jfp < 1 and also
A<1.
Then (3) gives
I t
1-- -+- _t 1 /P ?;J 0
P P
or q
~+~
P
_tl/'P -.
<?
0
, as
I 1
-+-=1
p q
Put I=apb-<l Then
1 'Pb-_
_ + _a_ Il
_ (apb-fl)l/p ;;;;t 0
q p
Multiplying both sides by bf ,
btl aP
-+-
q p
-ahtl-f{P ?;J 0

[ But .!..+!...=IQ·i+l=q:> q- [=1

p q p P
J
bqa P
or -+-
q p
-ab?;J 0
aP btl
or ab ~
"'" p
-+_.
q
This completes the proof of the lemma.
310 BANACH SPACI:

Theorem 3. It states that if X=(Xt> X2 , •••••• , x,,) and Y===(Yl'

Y., ... , y,,), 1 < p < 00, then

"
or
'_1.'1 I x,y, I ~ II x II" II Y I/ll

where the scalars x" Yi are real or complex.

Proor. If x=o or y=O, then the inequality is obvious. So
let us suppose that x¢O, ncO. Firstly we shall prove the follow-
ing l e m m a . '
(H ,re state and prove the above lemma). Now we come to
the proof of the main theorem. By the lemma,
&.. , . a,p+b.
q
aw,~
p
-q

Put a.=1Xjj;'
I x. I b
.= -IIYI u.-I .
y,

Then I x. 1·1 y. I .... !.. ~ +!_ ..!.1!.t

II x II. Ii Y Ii. '<: P II x I\.p q II Y Hv"
11 n "
E I x.y, I E I x, Ip E I Yi III
or
/-1 <; ~ i_I +!-. i_1

1\ X II" II Y IIIl P II x I"P q TJ~-n;"-

But n x II.=-{ '~ll X,IPF'P .. II x 111'21= i~ IXII"

In view of this the last gives,

or

Theorem 4. Minkowskl's Ineqoality. It states that if x-(Xtt

XI' ... , x,,), Y=(Yh Ya, ... , )'n) and I < P < eo, then
I: x+Y II" < II x 1,+11 Y ill'
BANACH SPACE 311

i.e., [ ,2:'" I x,+y, Ip ]1/f1 ~~ { E" I X, jP }lll> { Ely,

" I' }1/"
'-1 .=1 '-1
whue Xl, Y. are scalars (real or complex).
Proor. Step I. State and prove the Theorem 3.

Step II. II x+y 11,= { 1:" I x.+y, II' }I/II

1_1

..
or II x+y 11,11= '_E1 I x'+Y. IP=E
, I x,+y, I . I x,+y, IP-l

Applying Step I to R.H.S.,

\I x+y Il.p ~ p; I x, IP}I/p {E I x.+y, l(p/v).9}l/'.
• •
.
+{E I [Link]}I/P C1: I x,+y,1
, (11/9) "J l /1

:. II x+y 11"i>
~ II X /I, {II x+y /I,P}'j<1+11 y III' {II x+y II.'P/II
=(11 x 1/,,+11 y /I,,) (II x+y 1/.)1'/11
Dividing by nx+y /I.P/Q and noting that

P - q~= 1 as 1+E-=p,
q
we <>et
-
:. II x+y III' ~ II X /1,+11 y II,,·
Problem 1. ShoH! that the /inear spaces R" and en of all n-
tuples x=(xtt X 2 • .... x .. ) of real and complex numbers are Banach
spaces under the norm

[Meerut 1908]

(These spaces are called Euclidean and unitary spaces respectively).

Solution. (i) !I x " ~ O. For each I x, I ~ 0
(ii) II x 11=0 ~ x=O.
312 BANACH SPACE

For II x 1/=0 <:> { ~ 1 xd 2 }i/l =0 {;:> E 1 Xi 12=:0

t_l ' i

<:> Xt=O ¥ i <:> X-(X 1, Xli' ... , x,,)=CJ.

(iii) II x+ y II .;; " x II + II y II •

This follows from Minkowsi inequality for p=2. Hence CIt
and R" are normed lineal" spaces. These spaces are metric spaces
w.r.t. the metric d defined by d (x, Y)=II x _.y II
To prove the Completeness of en (or R"). Suppose < x" >
is a Cauchy sequence in Crt or ~". Since each x" is an n tuple of
complex {or real) numbers and hence
x m=(Xm l' X mz , .. , x m.. ).
This -> Xm~ is kth co-ordinate of X m . By definition of Cauchy
sequence given
E > 0,3 no € N s.t. m, I ~ no => II X m - XI II < 6

=> II Xm - XI 112 < t

.
2 -> E 1 Xm,-Xu 12 < .2
>=1
=> I Xml-X" pI < ~2 ¥ i ~ 1 Xm,--XU I < 6 ¥ i.
This proves that the sequence < Xmi> is a Cauchy sequence
\f i. But C (or R) is complete. Hence every sequence < Xml >
. z, In
converges to a pomt . C (R)
or so t h at Lim Xml=Z' ¥ l.. m-.C>o
Consequently every Cauchy sequence < x,,> converges to a point
z=(zt> Z2' .•• , z .. ) E en (or R"I. Hence Cn and R" are complete
spaces and consequently they are Banach spaces.
Similar Problem. Define a normed linear space. Prove /fl the
linear space of aU sequences
x={x1 , X 2 , .•• , x" ... }, of scalars such that
00
E I Xn IV < 00 with norm defined by
_1

is a complete normed linear ~pacL. [Kanpur 1988, 85J

Tbeorem 5. Let M be a closed linear subspace of a nornled
linear space N. If norm of a coset x + M in the quotient space is
defined as -
"x+M II=inf{" x+m II: m E M}
then NIM is a normed linear space. Further if N is a Banach space,
then so is NIM. [Kanpur [Link]. F. 198Z, 81, 79 ; Meerut 82 ;
Buodelkbaod 83, 82 ; Madras 79; Amritsar 82J
BANACH SPACE 313

Proo[ Let x, yEN and m, nt' E M be arbitrary. We know

that NjM is a linear space w.r.t. the operations
(x+M)+(y+M)=(:c+ y)+M (I)
and ~ (x+M)=lXx+M ... (2)
where IX is a scalar.
M is the zero element of Nj M. ...(3)
Aho x+M=M <=> x E M. ...(4)
Now we verify all the postulates for a norm.
(i) II x+MII ~ O.
For" x+m II ~ 0 and every set of non-negative real numbers
is bounded below and so inf { II x+m II: m E M} exists and is
non-negative.
(ii) To prove II x+M II =0 <=> x+ M=M=zero element of
NIM. Observe thatx+M=M => x E M.
This suggests that
IIx+MI\=inf{l!x-lmll:x,mE M~
=inf { II y II : y=x+m E M}
=inf { II y II : y E M)=O.
[Being l)ubspace, M cnnt~ ins zero vector whose norm is real
number 0].
Finally x+ M=M => II x+M /1=0. ...(5)
Conversely suppose that II x+M 11=0.
:. inf{1I x+m II: Tn E MII=-O.
This => 3 a sequence <mn> E M where n= I, 2, .. S.t.
Lim
n-+ 00
"x+m n II =0.
Lim
Therefore 171,,= -x.
11-+00
But M is closed and <mn> is a convergent sequence in M
converging to -x. -x E M.
This => x E Mas M i~ slI:bspace
Thus" x+M II =0 => X EM=> x +-M=M. .. (6)
From (5) and (6), the result (ii) follows.
(iii) To prove II (x+M)+(y+M) II ~ 11 x+ Mil + II y+M II.
By (\), II (x+M)+<y+ M) II = II (x+y)-t-M"
=inf{ II x+y+m II: m EM}
=inf{ II x+y+m+m' II: m,m' E M}
[For m+m' EM=> m, m' E M as Mis subspace]
=inf { II (x+m)+(y+m') II : m, 1/1' E M}
~inf { II x+m II + II y+m' II : m, m' c: ,\1}
[For II x+ y II ~ U.'( II + II y II in NJ
314 BANACH SPACE

=inf {il x+m Ii: m E M}

+inf {II y+m' II: m' E MJ
=/1 x+M II + II y+M /.
Hence the result (iii).
{iv) To prove liCIt (x+M) II = I or. I • /I x+M \I.
II IX (x+ M) II = inf { II or. (Xi m) : m E M}
=inf { I IX I . II x+m II : m E M}
[For IIClxu=lat/.l/xll
= I III I . inf {II x+m II: m E M}
= I CltI·ll x-r M II·
From what has been done it follQws that N/M is a norm~d
linear space.
Further suppose that N is complete (or Banach space). To
prove that N/M is complete (or Banach space). Here we make
use of the fact that a Cau:hy sequence is convergent iff it has a
cOll·vergent sub~equence
Let <s,,+M> be aCauchy:\cquence in NIM. Takee=l,
by def of Cauchy sequence. given E= i > 0 3 ni e N s.t.
I, m ~ "1 =- II (s,+MI- (s.n+M) II < E=i
=> iI (s +M)-(s",+M) II < •.
"1
Write S
nl
=x1• Then !I (Xl +M)-(sm+M) II <i.
Similarly for 1=(1)2, 3 a positive integer n2 > ni
S.t. II (s+ M)-(sm+M) II < Ul for m > n 2 •
II.,
Set sn 2 ;"'''2' then 11 (x 2 +M)-(s... +M) II < (nl .
Gene"aling this concept, we have positive integers n.> ... >
> na > 12 > n l s.t. "
II (x:. t-M) (s",+ M) II < (f)"
where x,.=s
• 11k
and 111 > nk •
Finally we have obtained a subsequence <x.. +M> of
5"quence <s•• + M> in N{M 5 t
Hence II (XHl +M) -(Xk+M) II < (J)k. . .. (7)
Remains to show that this subsequence convt:rges to a point
in N!M
+
We chooo;e a vector YI E x I M. [Link] choosing Yl' we select
Y2 E x 2 +M s.t. ,
\I )'1-Y2 II < J. We select Y8 E xa+M 5.t. II )'2- Y3 H < (i)I.
Continuing in this manner, we get a sequence <Y.. > in N S.t.
\I y,,- Y"+t II < n)" for n= I, 2, 3,... ...(8)
This assumption is possible due to (7).
bANACH SPACE 315

We claim < l'n> is a Cauchy s~quence.

Let ~ > 0 and choose a po~itive integer 11~ E N so large that
I
2,,0--1 < e S.t. n, m ~ no

nYm-Y .. /1=11 (Y"'--Y"'+I)-!-()'m+l-Ym+2)+-··+Cv .. l-Y") II

~ II y".- Ym+l iI ... II Y"'+l- )'mH !i+ ... + 1Y"-l-)' .. 1I
< 0)'" H!)"Hl+··.+(i)"-l, by (8),
< (t)"'·t(Hm+1+ .. ·+0)n-L t- ......
_ (V'" _ I I
-1-1~2m-=t < iilO=l< e
< y .. > is a Cauchy sequence. But N is complete.
lim lim
:. 3 yEN S.t.
11_<'..0
y,.= y. Consequently
ll--'?OO
x,,= y.
Hence x ..+M ·y+M E NIM.
Thus < sn+M > is a Cauchy sequence in NIM [Link] has a
CO'lvergent subsequence converging to a point y+M E NIM.
Consequently the Cauch sequence <5 n +lvl> converges to a
point in Normed linear space NIM, meaning thereby NIM is
complete
Consequently NIM is complete.
Problem 2. Let C(X) dCllote the linear .~pace of all bounded
continuolls scalar l'alued functions defined 0" a topological space X.
Show that C(X) is a Banach spact! under the nurm
i~fll =sup {If(x) I : x EX}, whuefE (X).
[hombay 1976; Meerut 79]
Solution. Let c(x)={f: f is l. continuous and bounded map
from Xto X}
where X is a topological space.
Let.t~ gEe (X) are arbitrary.
Define ilfli =Sup { If\x/ I : x EX}.
It is easy to verify that C(X) is a linear space over opera-
tions defined by
(f+g) (x)=f(x)+g (x) (vector addition)
«(1f) (x)=~f (x) (scalar multiplication)
To show that OX) is a normed linear space.
(i) Ilfll ~ O. For I fix) I ;;a 0 >,f x
(ii) IIfll =0 => 1=0 (zero function)
For II III =0 ~ sup { I f,x) I : x E X}=O <:> II(x) I ~- 0 V x.
<:> f(x)=O V x <:> f= .
3I 6 BANACH SPA(E

(iii) II/+gl1 ~ \1/11 + IIgll

For II/+g II =sup { I (!+g) (x) I : x E X}
=sup { ! lex) +g(x) I : x E X}
..;; sup { I f(x) I + I g(x) I : x E X}
For I a+b I ~ I a I + I b I
=sup { If(x) I: x E X}+sup {I g(x) I :x E X}
= II f II + II g ". Hence the result (iii).
(iv) II "fll = IIX I .lIfll·
For II afil =sup { I (IX!) (x) I : x E X}
=sup { I a.f(x) I : x E X}
=,sup { I Cl I . I fix: I : x E X}
== I Cl I .sup i I f{x) I : x E X}
= IIX I .lIfll·
Hence C(X) is a normed linear space.
Also C(X) is a metric space w.r.t. induced metric d s.t.
d (f, g)= IIf-g" =sup { I /(x)-g(x) I : x EX}.
Remains to prove that C(X) is complete as a metric space.
Let <f" > be a Cauchy sequence in C(X).
Then each fn : X -x is a continuous map.
By def. of Cauchy sl'quence.
given < > 0; 3 nil E N ~.t. tn, n ~ "0
~ d (f, .. f,,,) < •
.:> IIf.. -f", II < E
~ sup { If,,(x)-fn, (X) I: x E X} < •
~ Ifn(xJ-fm(x) I ~ sup I !fn(.A)-f",(x) I: x E X}<_
-:> I f,.(x)- fm(x) I < E ¥ X E X.
This proves that <f,.(X) > is a uniformly convergent sequ-
ence of continuous functions.
Hence <[II> must COl verge to a continuous Illap f on X.
Accordingly CLX) is a complete normed linear space and hence a
Banach space,
Problem 2a Pro!'e that th~ conjugate spact! 01 a normed linear
space is a lJanach spare. I Kanpur [Link]. F. 1981]
Solution. Let N be a normed linear space over that field K
of scalars. Here K =C or C. Also Rand C both are Banach
spaces. By def. N*, the conjugate space of N is the set of linear
and continuous functions from N into K, i.e.,
N*={f:f N-K is a linear continuous map}.
Replacing C(X) by N* in the above Problem 2, we get the
proof of present problem.
BANACH SPACE 317

Problem 2b. If Nand N' are normed linear spaces, then the
set B (N, N') of al/ continuous (or bounded) /inepr transformations
of Minto N' is itself a normed linear space w.r.t. pointwise linear
operations and the norm is defined by
IITII=sup{IIT(x)II:lIxll~ I}.
FurTher if N' is a Banach space, then B (N, N'),is also a Banach
space. [Jiwaji 1983; Kanpur 86, 91; Meerut 88, 87]
Solution. To prove that B (N, N) is a linear space. Let S
be the set of linear transformations from a normed linear space
N into a normed linear space N'. Let T 1 , T2 E S and a, J3 E K
and x E N. Then S is a linear space w.r.t. the linear operations
(T1 +T2 ) (x)=Tt (x)+T2 (x)
and (GtT1) (x) =«T] (x)
Here K is a field of scalars.
Let T], T2 E B (N, N'), then Tl and 1'2 are continnous linea.r
m'tps from N into N'.
Then 3 kl' k2 > 0 s.t.
il T](x) II ~ kl II x II , II T2(x: II ~ k2 II x II
[For continuity do boundcdness]
Tl and T2 are linear mars => (1.T1 +.81'2 is a linear map.
" (a.T1 +.BT,) (x) " = II a.T1 (x)+fiTs (x) II
<; II «Tl (x) II + II () T2 (x) II
= I Gt I . II T1(x) I! + I .B I . " T 2(x) II
~ I Gt I .kL " x Ii + I Gt I .k? \I x II
or "(GtT1 +()T2 ) (x)" ~ [ I a. I kl+ I ~ I .k~] II x II
This => crT1 +()T2 is bounded
=> GtTI +~ T2 is continuous map
=> CltT1 +fiT2 E B (N, N'J.
Thu~ T 1 , 1~ E B (N, N') => «T1 +.8T2 E B (N, N').
This proves that B (N. N') is a subspace of a linear space S.
Hence B (N, N') is a linear space.
II. B (N, N) is a normed linear space.
Let T 1 • T2 E B (N. N') and x E N S.t. II x II < I.
(i) :I T) " ~ O. For \I TI II =sup { 1/ T1(x)!t: II x II ~ J} and
" T1(:c) /I ~ 0, by def of norm.
318 BANACH SPACE

(ii) II Tl II =0 ~ Tt=O
II Tl II -0 <'> SUp { II T1(x) II : II X /I ~ I}=O
<'> II T1(x) 1/ =0 ¥ X <'> T1(x)=O ¥ X
<::> T1=0
(iii) II Tl + T2 II ~ II Tl II + " T2 "
For "T1 +T2 11 =sup {" T t (x)+T2 (x) II: II X II ~.1}
~ sup { 0 Tlx) II +
1/ T 2(x) II : II x jI ~ I}
[For I a+b I ~ I a i Ib I] +
=sup { !i T1(x) " : " x II I} <
+sup { 1\ T 2(x) n : II x It ~ I}
="
Tlli + II T211
(iv) "rxTt /I = I at I . II Tl "
For "rxTt " =sup { II (ceTl) (x) 11 : II x II ~ 1} .
= , at I .sup { II TJ(x) II : II x II <;; l}= I II I . \I Tl n
From what has been done, it follows that B (N, N ) is a
normed linear space.
fII. Remains to prove that B (N, N') is a complete space,
where lV'· is a Banach space and so it is complete as a metric
space.
Let < 7~ > b~ a Cauchy sequence in B (N, N,).
Then Tn E B (N, N') *
Tn N-+N' is a continuous linear
may "f n.
By def. of Cauchy sequence,
II T,. - Tm ,,-+0 as n, m-+oo ... (1)
For every x E N, T..(x) E N' ¥ n
Also" T..(x)-Tm(x) Ii =
"(T.. --T,,,)(x) " ~" T.. -Tm II. !! x II
-+0 as m, n-+ oo , by (1)
:. < T,,(x) > is a Cauchy sequence in N' which is complete.
Hence:l T(x) E ,v' s.t. lim T..(x) = lex).
n-+oo ...
(2)

This defines a map T: N-+N'.

Now we shall show that T is continuous and linear. If
lim
x, YEN, then T(lIx+Py)= T.. (otX+~y)
IJ-+OO
=atT.(x)+~T..(y)as Tn is iinear
=zT(x)+,9T(y), by (2).
:. T is linear.
BANACH SPACE 319

SinceI a-b I ~ I a I- Ib I
Hence I (II Tn II - 1\ Tm 1\) I ~ II T .. -··T,,, II - 0 as m, n-+ oo ,
by (I)
or I II - II Tm II I -0 as m, n -+00.
1\ Tn
This ~ < II Tn II > i .. a Cauchy sequence of real numbers
=> < II Tn II > is convergent sequence
~ < II Tn !I > is a bounded sequence
=> 3 K > 0 S.t. Ii Tn II ~ K y. n. .. (3)
Consequently, II T(x) II = 111im T,,(x) :1 =lim II T"(x) II
~ lim II Tn II . II X 1\ ~ k II X 1\ by (3)
or II T(x) II <K. 1\ X II
This => T is bounded => T is continuou~.
As < T,,(x) > is a convergent sequence with limit T(x) and
so by def.
II Tn(x)- T(x) 11 < f l.,f n ~ no ... (4)
II T-T" II =sup. (II (T-T,,) (X\ II: II x II ~ I} < 6, by (4)
Thus II T - Tn II < 6 "I n ~ no
This => lim T.. =T
n-+-oo ... (5)
T is linear and continuous ~ TE B (N, N')
lim ,
=> T.. EB(N, N). by (5)
n-+oo
Thus B(N. N') is complete and ~o that it is a Banach space.
Problem 3. Let N be a IIOIl-zero lIormed linear space and
M={x : XE Nand" x II = I}. Then prm'e that
N is a Banach space ~ M is complete.
[Kanpur 1991,89, 83 ; Meerut 79)
Solution. Let N be a Banach space.
To prove thar M is complete.
N is a Banach space => N is complete as a metric space.
Let < x" _ be a Cauchy sequence in M, then
II x .. II =1 'rll ..• (1)
< x" > is a Cauchy iicquence in MeN
=> < x .. > is a Cauchy sequence in N.
Also N is complete
=> 3 x E N S.t. X" "x. But norm is a continuous function.
:. II x" II -+ II x II
lim
or II oX II = n ... oo II x" II = I , by (1)
320 BANACH SPACE

or II x!l =1, XEN

By def. of M, thIs ~ XE IH.
This Cauchy sequence < x .. > in M converges to x E M.
Hence M is complete.
Conversely. Suppose that M is complete. To prove that
N is a Banach ~pace. Since N is a normed linear space and so it
is enough to prove that N is complete. Let < x .. > be a Cauchy
sequence in N. Then, by def.
lim
m, n-oo II X,,-X m II =0 ... (1)
But I II x .. II - II Xm II I ~ II X,,-X m II _ O~ by (1).
lim
m, n-+oo III x" II - II Xm 111=0.
This:? < :I x" iI > is a Cauchy sequence.
Thus < II X .. II > is a Cauchy sequence of real numbers.
Also R is complete. Hence 3 « e R s.t.
lim
n_oo II x" II =ot ... (2)
x"
. y,,=--
W fIte "f n (3\
II x.1I ... )

II Y.. - Ym II = 1\ II;: II - \I ~: II Ii
=
x" [Link]"')
II (1ix..T!- (Xm Xm) III
.. 1I + IiX:f - II Xm \I I
~ nx .. -Xm II + nx~JUL~~ II -/I Xm 1111
"'" II x.. II D x" II • II XmII
~ II X"-X,,, II + 11 X,,-X m 11 by theorem 1
"'" II x .. II II x,,11 '
or II II ~ 211 X"-X m II •
-
Ym '<:::
.V.. II x" It
Making n, m-+oo and noting (I) and (2), ~ve get
11 Y .. -Ym II ~ 0 as m, n~ao.
But II Y.. -Ym n ~ O. Hence.
lim
m, n_oo II Y.. -Ym II =0.
This:? < y .. > is a Cauchy sequence in N

II )'.. ~lln ~: II 11= ::~: : = 1 or II y .. II = 1. '" (4)

BANACH SPACE 321

This =-
y .. E M.
Finally, < Yn > is a Cauchy sequence in M which is given
lim
to be complete. Henee 3 y E M s.t. y,.=y. Then
n-oo
x,.
y,.~ y =- \I x,. II - Y ~ x,. -+ IX}" by (2)

=- nlim
.... oo
x,.=acy
••• (S)
Y E M, IX is a scalar, M is linear space ~ aty E MCN=-[Link].
Now (5) proves that the Cauchy sequence < x. > in N con-
verges to N at a point exy E N, so that N is complete and so N is
a Banach space.
Theorem 6. Let Nand N' be normed line'lr spaces and T a
linear transformation of N into N ' . Then the following conditions
are equivalent to one another:
(i) T is continuous
(ii) T is continuous at the origin
[i.e., xn-+O .. T(x,,)-+O]
(iii) T is bounded
[i.e., 3 a real number k ~ 0 s.I.
nT(x) II ~ k II x II \f x E N]
(iv) If S={x : II x II -I} is closed sphere in N, then its image
T(S) is bounded set in N'. [Meerut 1988; Kanpur89, 85; Jiwaji 83)
Proof I. (i) ~ (ii).
Firstly we prove (i) ~ (ii).
Let T: N-+N' be a continuous map and < XII > be a
lim
sequence in N s.t. xn=O.
n-+oo
By continuity of T,
xn-+O ~ T(x ft )-+ T(O). But T(O)=O for any linear map
:? T(xlI)-+O.
This proves that T is continuous at origin.
Now we claim (ii) ~ (i).
Let T be continuous at the origin. To prove (i).
. N S.t. lim X,,=X E N •
L et < X,,:> b e a sequence 10
n_oo
Then x,.~x => x.-x-+O:? T(x,.-x)-+T(O)=O.
[For T is continuous at origin]
322 BANACH SPACE;

=> T(x,,) T(x)--+O => T(x,.) .... T(x)

Thus X,,-+x ~ T(x .. ) , T(x).
By def. this proves that T is continuous.
II. To prove that (ii) ~ (iii).
To prove (ii) => (iii). Let T be continuous at x=O.
To prove (iii). Suppose not. Then T is not bounded.
Then for each positive integer n, we can find vector x_EN
S.t. II T(x.) II > n II x" II

... (1)

·
T a k log Yn= IIx" ' we get UT(y,,) II > 1. ...(2)
n x" II '

II x..
Also II y .. II = n II x,.III'I = nII \Ixx....II 1\ =,;'1
•
•• lim II y" II =0
n-+oo .. (3)

as ~=O,
CX>
lim But
As norm is continuous function and so
n-+oo
lim
n-+oo
II T(y,.) II > I, by (2).

Finally, y ..-+O and T(y .. ) does not tend to zero as n-HLJ.

A contradiction. For T is continuous at x=O.
Hence (iii) is true.
To prove (iii) => (ii).
Let l' b! bound\!d so that :I a real number k ;;;;t 0 S.t.
II T(x} II ~ k II x 1\ V x E N. ...(4)
Lim
Let < x .. > be a. sequen<::e in N S.t. n ...oo x.. =O E N.
This => II x .. II -+ 0 as n "00. ...... (5)
For norm is a continuJus map.
By (4), \I T(x,,) II ~ k \I x,./I.
Using (5), we find that \I T(x .. ) \I _ 0 as n -+00.
This ~ T(x .. )_O as n -+00.
This shows that T is continuous at the origin.
BANACH SPACE 323

III. (iii) <:> (iv).

To prove (iii) => (iv).
Let (iii) be true, i.e., 3 k ~ 0 S.t.
II T(x) 1/ , k II x II V x € N ... (6)
~ k II x II 'VX ESC N.
(6) => II T(x) II . .. (7)
But XES => T(x) E T(S).
Hence (7) proves that T(S) is [Link].
To prove (iv) => (iii). Let (iv) be true, then 3 k ~ 0 s.t.
1\ T(x) II ~ k II x II 'V x € S. . .. (8)
If x=O, then T(x)=O so that (8) is true for x=o. . .. (9)
x
If x*O, then put Y=1fXj so that

II x 1/
I/y II ==fiTi=I.
Consequently yES and so by (8), 1\ T(y) 11 E:;; k II y n
or
" T ( II ~ II ) 11, k j II/ ~ II 11=k : ~ :: =k
or T(x) 1/ ~ k 1/ x II 'V x~O.
1\ . .. (10)
By (9) and (10), II T(x) n , k II x II ¥ X E N.
This => T is bounded. Hence (iii).
This completes the proof.
<x, Problem 3a. Let X and Y be normed linear spaces. Show that
a linear trans/ormation T on X into Y is bounded iff T is continuous
at a point Xo E X. [Kanpur 1979]
Solution I. Let T: X -+ Y be a linear map s.t. T is bounded.
To prove that T is continuous. By assumption, 3 k>O s.t.
II T(x) II E:;;k II x II ¥ X E X. . .•(1)
Let < x .. > be a sequence in X s.t.
lim
Xn=X EN.
n--? 00
As norm is continuous function and so
Lim
n-oo II x" 11=11 xii.
Lim
This => II x .. -x II =0. ••• (2)
n~oo

By (1), II T(x. -x) II ~ k nx.-x II.

Making n--? 00 and using (2), we get
324 BANACH SPACE

II T(xn-x) II ~ 0 as n-..oo.
But II T(x.. -x) II ~ 0 y. n.
Lim
Finally, II T(x .. -x) 1=0
n"'OO
Lim
or T(xn)=T(x).
n-400
Lim
Thus, Xn-4X => T(xn)= T(x).
1J~00

By def. this proves that T is a continuou~ map.

[Link] T be a continuous linear map.
To prove that T is bounded. Suppose not.
Then T is not bounded so that for every positive integer n,
we can find a vector x .. E X s.t.
II T(x .. ) II > n II Xn II
I
or n \I Xn II II T(x n ) II > I. .. (3)

Write )'n= -~~--, then (3) ~ !I T(Yn) II > I y. n ... (4)

n II XII h
.Also
II )'n I= II n IIx..x .. II !\ I= n IIII x"x" IIII =n1
Lim I
n .. oo
11)'.. 1/=··---=0.
00 . .. (5)
.
A s norm IS . Lim Yn= 0.
contmuous,
n~oo

By (4), Lim T(y .. )

n-4 co
> I, according to (4).
Finally, Y1l-'0 => T(y .. } does not tend to zero as n-4OO.
A contradiction. For Tis continuou'l at x =0.
Consequently T is bounded.
Problem 4. Suppose that a Banach space B is the direct sum
of the linear suhspaces M and N, so that B=MffiN and let z=x+y
be the unique expression of a vector z in 8 as the SUm of vectors x
and y in M and N. Shf)w that a new norm can be defined on the
linear space B by II z II' = II x II + II y II and verify that it is actually
a norm. If 8' symbolizes the linear space equipped with this norm,
prOl'e that B' is a Banach -space if M and N are closed in R
[Meerut 1988]
Solution. Note that Band B are same sets. (B, /I II) is
a Banach space. Let B= M(fJ V and let M and N be closed as a
BANACH SPACE 325

metric space in D. Then z=x+y with x e M, yeN is unique

representation of z. Let
II Z 1'= II x II II Y n + ... (1)
To prove tltat (B', II 11') is a Banach space, where D' =B.
Let Z, Zl' Z2 E D', then
z=x+Y, ZI==XI+Ylt z.=X2+Y2 where
x, Xl' X 2 E M and y, Yl> Ya E N.
Then XI +X2 EM, YI+Ya eN
'is M and N are subspaces. Let CIt be a scalar.
(i) II Z 11 ';;:' O. For II X n .11 Y H ~O ~ II Z II' ~ 0, by (1)
(ii) II Z 11'=0 ~ z=O.
For 11 zl\'=O <:> II X II + II Y II =0 ~ II x II =0= II Y II
~ x=O=y <=> z=x+y=O+O=O.

(iii) II ZI+Z2 11'=11 (XI+ yJ+(x2+y.) 1I'=:I(xl+xJ+(YI+YJII'

= II Xl +X 2 II + II Yt + Y211 , by (I)
~ II XIII + II XI II + II Yl II + II Yz \I
= ( II X~ II + II YI II ) +( II XI II + /I Y2 II )
= II Zl II' + II Z. II'
or IIzI+z211'~lIzllI'+lIz2il'
(iv) \I rl.Z 11'= I at 1·11 Z II .
For CXZ=I1(X+Y)=llx+atY
:. II atZ 1\ ' = II atX II + II atY II = I at I . II X II + I at I . II Y II
= I 11 I .( II X II + II Y (1)= I II I . II Z \I '
(v) From what has been done it follows that (D', II II') is a
normed linear space and is also metric !.pace w.r t. the induced
metric d defined by d(Zl' zz)= II Zl-ZZ II '.
To prove that D' is complete.
M and N are closed subspaces of complete space ~ M and N
are complete,
Let <z,,> be a Cauchy sequence in 8'.
Then z.. =X.. +Yn with x" E M. Yn E N
Zn Zm=(Xn +Yn)-(x":+ y",) =(xn-x m )+ (y"-y,,,). ...(2)
By def. of Cauchy sequence, :I no E N s.t.
m, n ~ "0 ~ II Z"-z,,, It ' < I.

This 0 II X"-X,,, II +
II y,,-y ... 11 < E, by (1) and (2)
::> II XII-X", 1\ <" ,/2 and II Yn - Ym II <6/2 .
¥ n, m ;:;, no
Hence < XII > and < y. > are Cauchy sequences in M and
326 BANACH SPACe

N respectively which are given to be complete. Hence 3 p E M

and q E N s.t.
X,,-+p E M, y,.-+q E N
This=> x,,+y,,-+p+q => z-+p+q E B'.
Thus every Cauchy sequence < z,. > in B' converges to
p+q e B'.
Hence B' is complete.
Consequently it is a Banach space.
Problem 5. If M is a closed linear subspace of a normed linear
Ipace Nand T : N- NI M is a natural map S.t.
T(x)=x+M.
Show that T is continuous and II T II ~ I.
[K'anpur 1994488, 87; Meerut 80; Amritsar 82]
Solation. Let M be a closed subspace of normed linear
space N, then NIM is a normed linear space with the norm
defined by
II x+M II =inf { II x+m /I : m E M} ... (1)
NIM is a linear space w.r.t. the operations
(x+M)+(y+M)a::(x+y)+M ... (2)
a(x+M)=ax+M ... (3)
wnere x, yEN and a. b ~re scalars.
T(ax+by)=(ax+bvY+M, by def. of T
-=(ax+ .M')+(by+ M) by (2)
==a(x+M)+b(y+M) by (3)
-aT (xH bT (y).
This => T is linear.
n T(x) II = IIx+MII cooinf{ IIx+m II: m € M} <;; II x+mll
Thus II T(x) II ~ II x+m II ¥ m E M. ...(4)
In particular for m=O, we get
x"
II T(x) II ~ II ... (5)
or II T(x) II ~ I. II x II "I x E N
By def. this proves that T is bounded and so T is linear, by
problem (30.).
Second part. " T" = sup { II T(x) II : x E N, II x II ~ I}
~ sup { 1\ x II: n x II < I}, by (5)
=1.
or IITII~ I ..
Problem 6. Let N(K) and N' (K) be normed linear spaces.
Suppose that T: N-+N' is linear map. Theil ker (T), null space of
T. is linear manifold and ker (T) is closed if T is conti"uous.
 BANACH SPACE 327

Soultion. By def. ker (T)==N(T)={x eN: T(x)-O} ... (1)

Observe that 0 E Nand T(O)=O
:. 0 E ker (T). . .. (2)
Let x, y E ker (T) and a, b E K. then
T (x)=fl= T(y)
T (ax+by)=aT(x)+bT(y) as T is Iienar
=aO+bO=O.
:. a~+by E ker (T).
Finally, x. y € ker (T) => ax+ by € ker (T). ...(3)
By (2) and \3). it follows that ker (T) is linear manifold.
Sceond Part. Let T be continuous map. By def. of inverse of
an element, T-l[{O}]=ker (1 )
N' is a nonned linear space and is a metric space w.r.t. induced
metric d s.t. d(x', y')= II x'-y'li 'of x', y' E N'
In every metric space !>ingleton set is a closed set and so {OJ is a
closed set in N'. Also inverse image of a' closed set under a conti-
nuous map is a closed set. Hence T-l [{O}] is closed and therefore
ker (T) is closed.
Porblem 7. If T Is cotinuous linea, transformation on a
normed linear space N Into normed lmear shace N' and if M is its
null space. show that T induces a natural linear map
T': NIM_N' and II T' II-II Til
[Meerut 198Z]
Solution. Suppose T: N ... N' is a continuous linear map.
where Nand N' are normed linear spaces ovt:r the same field K.
Let ker (T)=M then M is a closed linear subspace of N.
, (Refer Problem 6). Hence NIM is a normed linear subspace w.r.t.
the norm defined by
II x+MII =inf{" x+m II: me M}. ...(1)
We define a map T' : NIM~N' st.
T' (x+M)=T(x) ... (2)
Let x+M, y+M E NIM and a, be K.
T' [a(x+MJ+b (y+M)]=T'[(ax+M)+(by+M)1
= T'[(ax +by)+M]= T (ax+by)
=aT( x) +bT(y) as T is linear
=aT'(x+M)+bT' (y+M)
by def. of T'.
This proves that T' is linear.
By (1).11 x+M II -inf { II x+m II : In E M} ~ II x+m II •
 328 BANACH SPACE

In pArticular for
m=O E M. II x+M II ~ " x II V x E N (3)
" T' II =sup { II T' (x+M) II : II x+M II ~ I ¥ x EN}.
Using def. of T', we get
II T' II =sup { II T(x) II : II x+ Mil <; 1 V x E N}
or 1/ T' II =sup t " T(x) II: II x+m II ~ 1, x E N, m E M} ... (4)
x E N, m EM=> x, mEN as MeN ~ x+m E N
II T 1/ =sup { II T(x) Ij : 11 x II .;; I; x E N}
=sup{IIT(x+m) u:nx+mll~ 1,xE N,mE M}.
But T(x+m)=T(x)+T(m)=T(x)+O=T(x
as m EM=> T(m)=O
/I Til =sup { II T(x) II: II x+m II .;; 1, m EM}. . .. (5)
By (4) and (5), II T' II = II TO.
Theorem 7. Let Nand N' be normed linear spaces and let
T: N-N' be any linear map. If N is finite dimensional then Tis
~ontinuous map (or bounded). [Kanpur 1989; Meerut 76, 73]
Solution. Let dim N=n and {elf e2 , •• , en} be a basis for N.
Then and x E N is uniqu¢lY expressible as
..
X= 1: at e" where aI' a2,00., a. are scalars.
1-1

It can be proved that II x II 0= max I at I is a norm of N. This

norm is called the zeroth norm.
,. II

II T(x) II = II E a, T(e,) II
i_l
~ E
i-l
I a, I . II Te, II

,.
~ 1: II x II o· II Te, II
1_1

..
Since the basis is fixed and so 1: U Te, II is a positive constant .
• _1

II
Let k==l: 1/ Tel II . Then /I T(x) II ~ k II x H 0
i-I

This => T is bounded

=> T is continuous. (Refer Problem 3a).
Theorem 8. (Hahn-Banach Theorem). Let M be a linear sub-
8pace 0/ a normed linear space N. LeI f be a linear functional on
 BANACH SPACE

M. Thenf can be extended to a function F defined on the whole

spare N s.t. II FII = ilfll.
[Meerut 1988,82; Kanpur 82, 78, 79; Indore 79J
Proof. We first prove the following lemma.
Lemma. Let M be a linear subspace of normed linear space
Nand Xo E Nand Xo $. M. Letf be a functional on M. If
Mo=(MU{xo})={x-t-ClXo : x E M, 1& is real}
is the linear subspace generated by M and Xo. thenf can be exten-
ded to a functionalfo defined on Mo such that
lifo II = IIfll . [Kanpur [Link]. F 19°41
Proof of the lemma. I. Let N be a real normed linear space.
Define a map fo : Mo-R. . .. (1)
s.t. flu)=fo(x+lltXu)=/(x)+cxro ... (2)
where " is an arbitrary but fixed real number, R is the set of real
numbers. u, v E Mo are expressible as
u=x+cxxo, v=y+~xo
.vhere x, y e
M. Let a, b E R=set of reals.
au+bv=a (x+Clxo)+b (yt f3xo)=(ax+by)+xo (ocx+b[:i)
x, y ~ M and a, b e R => ax+hy E M [For M is subspace]
10 (au+bl')=/o[(ax+by)+xo (acx+bf3)]
=/(ax+by)+(acx+h~) '0' according to (2).
=aj(x)+bj(y)+(acx+ hfJ) '0' asf is linear.
=0 [/(x)+cx'o]+b [/(Y)+~'ol
=afo (x+lltXo)+b/o (y+~xo), by (2)
=0/0 (uH blo (v).
This proves that 10 is linear. Now (I) proves that 10 is a
linear functional on Mo and
fo(x)=/(x) ¥ x E M, by putting cx=O in (2). Hence
10 is an eXlension off.
II. Remains to prove that II 10 II = UI \I.
11/011 =sup {[Link](x) I: x EMu, II x II ~ I}
~sup ( I/o(x) I : x E M, II x II ~ I} as Mo ::> M
=sup { If(x) I : x E M, II x II <. IJ as 10=1 on M
= Ilf II·
This:::> 11/0 II ;a 11//1· ... (3)
III. Now our problem is to show that If 10 " .;;;; II f II under
suitable value of '0.
330 BANACH SPACE

Let Xl' X 2 E M, then

/(Xl)-/~X2)=/(Xl-.'C·J ~ 1/(Xl -X2) I ~ lI/n·1I xl -x2 11
s;;: II/II • II (Xl +XO)-(X2+XO) II

II . Ii [ II Xl +Xo II + II X 2 + Xo II ]
< II /
II/II . II XI+Xo II + II/II. II X2+Xo II
=
or -/(x~) - 11/11.11 X 2 +Xo II ~-/(XI)+ IIlll.11 Xl+X o /I.
lt is true Y Xl' XI e M.
This proves that
sup { /(y)-lIl 11,11 y+xo II} ~ inf {-/(y)+11 / 11·11 y+xo II}·
yEM yeM
Choosing a real number s.t. '0
sup {-/<y)-II/ 11':1 y+xo II} ~'o:s;;inf {-/(y)+II/II·1I y+xo II}
yeM yeM ... (4)
For any u E M o, we can write u=x+axo with x e M.
Putting y=~ in (4) and noting that
at

/(y)=/
I x' 1/(x). we, get
,;)-=«
-!f(x)--
~
If 11.1 l~+x.
1
Ot'
II ~ro ~-!." /\x)+I1/ /.·\1:,11~ +Xu II .••
(5)

If 11. > 0, then (5) gives

I
- [--f(x)- 11/11.11 x+:xxo II]
lit
I
~ ro ~ -ot [-/(x)+II/1i .11 x+ClXo II J.
This ~-/(.'()- II/II. II u /I ~ IJro <;. /\x)+ 11/11.11 u II
=> -r.f(x)+IJ'o· EO. II/II. II u II
and /(x)+Qru ~ u [Link]"."
Taking the sec~nd inequality,
/(x)+aro ~ IIfil . II u II
or /o(x +orXo) ~ II/II ." u II
or /o(u) ~ II/II . II u II. ...(6)
Similarly if c,( < 0, then by taking first inequality in (5), we
get (6).
Replacing u by - 11 in (6).
-/o(u) E;;; O/[Link]-ull-II/[Link] ull
-/.(u) .;;; II/II. II u II. ... (7)
By (6) and (7),
BANACH SPACE 331

Ilo{ u I ~ II I II . II U II ... (8)

nfo II=sup ( I/o(u) I : U E M o, II u II < I}.
Using (8), 11/0 II ~ 11/11. .. (9)
By (3) and (9), we get II fo II = II I II .
The same result can be proved "hen N is a complex norm(d
linear space.
Proolol the main theorem. Let P denote the set of all ordered
pairs (/~, M 1 ) where f), is an extension of I to the subspace
MI ::J M and II IA II =111 .
We define a partial order ~ 0 .. P by setting
(II, M I ) ~ (/~, M~) iff M" c
M,.
and/'A=I. on M A• Since ( f, M) E P and so P =F- 0. Let
Q={(j" Mt}}
be a chain in P, then Q has an upper bound (g, U M,) where
g(X)=Ji(x) ¥ x E M,.
Here U M, is a subspace of N. By Zorn's lemma. 3 a maximal
element (F, H) in P. It can be shown that H=N.
Here we have used Zorn's lemma which states that, "Every
non-empty partially ordered set in Which each chain has an upper
bound has a maximal element". By the lemma, \I ,. II = 11/11.
Theorem 9. Generalized Hahn-Banach Theorem. Let X be a
011 X .J t.
real vector space, p a real valued function
p(x+y) .-;. b(x)+p(y)
p(ax)=ap(x) for a ;,il 0
and Y a subspace of X. {f f is linear Oil Y alld
f(y) ~ p(y) V Y ~ Y,
thtll there exists a linear function F on X s.l. F(J')=j(y) 011 Y and
F(x) e;;;; p(x) on X.
lKanpur 79, 78; Meerut 89, Indor 79)
Proof. T he main problem is to extend f to g defined 011
Y1= Y U {xo} where Xo ~ Y but XI) E X so that
Ily)=g(y) ¥ Y E Y.
This can be easily done.
The existence of F will be proved by the help of Zorn's
lemma.
Let E be the family of all extensions g of f defined a~ ahow.
Define partial order relation ~ on E as follo",s :
gl EO;; g~ if g2 is an extension of gl
 332 BANACH SPACE

i.e., if g\(X)=g2(X) and doma n of gl C domain of g2

where gl' g2 E E.
Then (E. ~) is a partially ordelcd set. Let C be a chain
(Totally ordered subset) in E so that C C E. Define a linear
functional T S.t.
(i) Domain of T=union of domains of g [Link] g E C.
(ii) If x E domain of T, then we define
T(x)=g(x) .•. (I) where gEe and x E domain of g.
From the constructions of E, C, T it is clear T is a linear
extension off and
T(x)=g(x)=flX) ~ p(x)
T(x) ~ p(x) on its domain ... (2)
so that TEE and it is an upper bound of C. By Zorn's lemma,
E has a maximal element F. Consequently F is an extension of
f and its domain is X and
F(x)=f(x) on Y, by (1)
and F(x) ~ p(x) on X, by (2).
Theroem 10. If N is a normed linear space and Xo is non-zero
vector in N, then 3 fUllctional fo in N* s.t.
fo (xo)= II XU II and II fo II = I. [Meeurt 79, 76]
Proof. Supposc N is a normed linear space over the field K
of scalars. Let Xu E N s.t. xo~O.
Wirte M={axo: a E K}.
Then M is a linear subspace of N and IS generated by Xo'
Define a map
f: M-+ K S.t. f( x o)= II XU II·
(X (I)
LeT II, v E M, then u=a;xo• I'=~x•• \\here Of, fJ E K. Let
P. q E K.
pu-+-q~'=p /Xxo+q 8xo=(px+qP) X o_ Pr1.+qf3 E J(.
By (1),f(pu-1 qv)=(~+q~) II Xo II=pcr; IIXo ,I +q{J /I Xo II
=pf (7X..)+ qf(f3xo)= pf(u>+qftv).
This proves that f is linear
I feu) 1= I Of III 111= IIX I . 11 Xo I = II ( Xo \1= II u 1\ < \I 2u II
XO
or I < 2 I u II.
If(u) ..(2)
By def., this provcs that f is bounded a~d ht'nce continuous
by Problem 3'(a).
II! II=sup { I feu) I : U E M, /I U II ~ I}
=sup { 1/ U II : U E M, II t.I II <" I}= I
or IIfil=1. Also!(xo)=HxoilontakingOt=l. #'
BAN '.CH SPACE 333

Finally, r: M~K s t. f «(1X o)=(1. II Xo II

is continuous linear tran~formation,
/ Hence.! is a functional on A! with the property that
.!(Xo)= II Xo /I, IIfll =1 .. ,(3)
By Hahn-Banach theorem, the function f can b;! extended to
a functio:lal fo on N S.t. "/0 II = II f II ... ( 4)
By (3) and (4) lifo II =1. ... (~)
fo is an extension off ~ /O=fon A!
=> fu(xo)=/(x o) as Xo E M.
Using (3), /o(xo)= II Xo II ... (6)
Finally, 3 fo E N-It< S.t. /o(x) = II XO II and II fo II = I. This
completes the theorem.
Remark. x,t:y => x-y#O => II x y II =0
=> If(x-y) I = II x-y II ~O, by (2)
=> f(x--y).:;t:O => f(x)*fIY)·
This proves the following result:
if x.:;t:y- and x, YEN, then 3 fo E N if, s.t. fu(x)-~J~(Y).
Theorem 11. If M is a c/oJed sub.)pace of a normed linear space
Nand Xo fI. M, then 3 functional/o ill N :' s.t.
foe M)= 0 and fo(.'(o)=r-0.
[Kanpur 1980; Meerut 80]
Proof I. Let M be a closed linear subspace of a normed
linear space N, then N/A! is a normed linear space w.r.t. the
norm detlned by
II x+M!1 =inf { II x+m II: x E N}
Xo fI. A! => xo+Af-:/'M "'" xo+M is not a zero vector of N/M.
Hence,3 a functional f E (NI M)* S.t.
f(xo+M)= II xo+M II ... (\)
and IIfll =1 ... (2)
(Refer Theorem 10)
II. Define a map T: N-.N/M s.t.
T(x)=x+M ••• (3)
It is easy to see that T IS linear.
For T (ax+by)=(ax+by)+M'=(ax+M)+(by+M)
=a (x+M)+b (y+M)=a T(x)+b T(y)
II T(x) /I = II x+M II =inf { II x+m II: mE M} E;; II x+m II
or II T(x) " ~ " x+m " '<I m E M
In particuhr for m=O E M, we have
II T(x) II ~ " x II or II T(x) II ~ 1 II x II
334 BANACH SPACE

By def., this proves that T is bounded.

Hence T is continuous, by Problem 3 (a).
III. Since T: N+N/M./: N/M-+-K.
This ~ f T : N ~K.
Now we define
fo(x)=( fT) (x) "" x EN. ... (4)
Composite map of two linear maps is a linear map. Also
composite map of two continuous maps is a continuous map.
Consequently fo is continuous linear functional on N. Hence
fo E N* .;.(5)
Also fim)=(fT) (m)=f(T(m)]=f(m+M)= II m+M II =0.
[For M=O' is a zero vartor of N/M and
mE M => m+M=M=O'].
:. fo(m)=O "f m EM .. (6)
This ~ fo(M)=O ... (7)
!o(xo) =( fT) (xo)-l fT(xo)] =f (xo+M)= II xo+M 11*0, by (1)
[For xo+M=I=M i.e., xo+m.,t:zero vector of N/M]
or fo(x o)=1=0 ... (8)
From (5), (7) and (8), the required result follows.
Problem 8. Let M be a closed linear subspace of a normed
linear sp:zce N, and let x" fI. M. If d =d (Xo, M), then 3 functional
fo E N* S.t. fo(M)=O, fo(xo) = I, 11.1;11 =-l/d
[Meerut 1988, 87,86; Kanpur 80; Delhi 81]
Solution. Suppose M is a closed linear subspace of a normed
linear space Nand Xo fI. M.
Write Mo=(M u {x,})
={x /-cxxo : « is any real number}.
Then Mo is a linear subspace generated by M and {xo}.
Xo fI. M ~ any y E Mo has a unique representation
y=x+cxxo ... (1)
fOl' some real number Cl.
Let d=distance of x. from M. Then
d=inf { "x-xo II: x E M} ~ II x-xo II V x E M ... (2)
Xo ft. M ~ x=l=xo y. x EM=> II x -- Xo II > V x e M °
;::.d>O
d ~ II x - Xu II => d.!... ;;;. 1
II x -Xo II
BANACH SPACE 335

/ Ii x -xu II ~
} ¥ x E M. . .. (l)
Define a map I: Mo-..K s.t.
jiy)= ( ... (4)
where y is given by (I).
Since representation of y is unique and so map I is well defined
l(xo)=(O-l-lx~}=1 or I(xo) = I
mE M => Itm)=/(m +[Link]) =0 ~ I(M)=O.
Thus we hav' proved that
j(M) = O,f(xo) = 1 ... (5)
I/(y) 1= I x 1= ac I II Y II I ct I . II y II I at I . II y II
II y" II x+atXo II l7ITxIOl+xo \I
II y II with _::.. to M
IIxo-(-x/ex)ll Ct

~ i! ~J! , according to (3)

or I fiy) I EO; II y IL ... (6)
d
Il/il=sup{ Ily>1 :yE M [Link]<: I}
~ sup {~ : y E Mo. II y II E;; 1 }
This saggests that II I I= I jd.
Finally, we have shown that f is a linear functional on Mo S.t.
!.M)=O,f(xo)=I, 11/11 =I/d.
Hence, by Hahn-Banach theorem, the function I can be
extended to a functional/o on the whole space N S.t. II 10 II = II I II
lifo iI = 11/11 and fill II =Ijd ~ 11/0 II =Ijd.
10 is exten~ion of I=> fo =/on Mo
=> fo~=1 on Mo :J M
=> fo(M)=/(M). But j(M)=O
=> lo(M)=O
Also J.,(xo)=.!i xo)' But I,x o)= 1.
Hence .f~(xo)= I.
This completes the proof.
Definition. Let Nand N' be normed linear spaces with the
same scalars. Then a one-one linear map T of N into N' S.t.
II T(x) II = II x II ¥ X E N,
is called an i'Jomlltric isomorphism of N into N', If such an
isometric isomorphism cxi,ts, we say that N is isometrically
isomorphic to N.
 336 BANACH SPACE

Theorem 12. Natural imbe«!ding of N into N-. Let N be a

normed linear space. Thm each vector x in N induces a lunctional
Frr; on N*' d lined by
FflJ ( I )=i(x) V lEN*' s.t. 1\ FflJ II = II x If •
Further the map J: N-+N'I<* s.t. J(x)=F. V x E N delines
an isometric isomorphism 01 N into N*''*'.
[Kanpur 199.0; 87; Meerut 86]
Proof. Let x E N and IE N*, then I: N_K is a linear
continuous map.
Let F.,: N*'-Kbe a map s.t. FflJ(/)=/(x). . .. (1)
I. To prove that FflJ is functional we have to show that FflJ is
linear and continuous.
LetJ, g E Nit< and a, b E K be arbitrary, where K is a field
of scalars.
FflJ (ol+bg)=(al+bg) (x), by def. of FflJ
=al(x)+bg(x)=a Fz (f) b Fz(g)+
Thl:' => F" is a linear map
! FaI(/) 1= I/(x) , E;;; 11/11.11 x II
Taking K= II x II =constant for the map F." we get
'Frr; (I)' ~ KII/II
By def. this proves that F", is bounded and hence continuous.
Thus F", is linear and continuous and so F:r; is a functional on Nil<.
Consequently F:r; E (N'*')*=N'*'*.
II. II F:r; II =sup UFaI(f) , : 11/11 ~ I}
. =sup { 'I(x) I : 11/11 ~ I}, by def. of F",
<sup {II/II . II x II: IlIIi ~ I}
as '/lx) , ~ IIltlll x II
~IIXII
or II F", II ~ II x II ¥ x E N
III. Let x be any non-Zl!ro vector. Then:l a functional
g E N'*' s.t. g(x)= II x II ... (3)
and If g II = 1 (Refer Theorem 10)
Also g(X)=F",(g), by (1). Now using (3), F",(g)= If x II ==g(x).
This proves that
1\ x /I =g(x) ~ sup { , g(x) I : gENII<, II g II ~ I}
= sup { I F",(g) I : g EN"", II g II ~ I}
=IIF",/I
or /lxll';;;IIF",1I ..(4)
 BANACH SPACE
337
By (2) and (4y "x
II Fill I! = II ••• (5)
IV. To show that the map J : N + N ••
s.t. J(x)=F. [Link] x E N ..• (6)
is an isomorphism.
Fc.+&fI) (f)-/(ax+by)=af(x)+bf(y)
=a FfIJ(f}+b F,I (f)=(a F.·+b F,) (f).
Using (5), J(ax+bYJ=a J(x)+b J(y}.
This .. J is linear. . .. (7)
By (5) and (6),
II J(x} " = " F. II = II x Ii or II J(x) II = II x II ... (8)
This proves that J preserves the norm.
(8) =- II J (x-y) n= "x-y" ~ /I J(x)-J(y}II=lIx-y 1/ ... (9)
J(x)=J(y) .. J(xJ"-J(y)=O => J{x-y) .... O
.. II x--y II =0, by (9)
.. x=y.
:. J is one-one map. ...{l0)
From (7), (8) and (10). it follows that J is an isometric
isomorphism of N into Nit ••
This completes the proof.
Deduction. Since J is an isometric isomorphism of N into
NIt:* and therefore we may regard N as a part of N·. without
changing any of its structure a~ a normed linear space. We
write N C N**. Hence the map J is called natural imbedding
(or canonical map) of N into N·*. [Kanpur 1978]
Not~. If J is onto as well, then Nand Nit'!< are isometric.
At this moment N=N*·, the sign of equality in the same sense
of isomorphism under the map J.- Then the map J is called
reflexive map and also N is called Norm reflexive or simply
reflexive.
Lemma 13. Let Band B' be Banach spaces and T a continuous
linear map of B onto B'. Then the image of each open sphere
centered on the origin in B contains an open sphert' centered on the
origill in B'. [Kanpur 1978; Meerut 77]
Pro..r I. Let Sr and S,' denote respectively the open spheres
centered at the origin and radius r in Band B' respectively.
Then Sr={xeB: II x" < r}. Sr'={xeB : II x 1/ < r}
Observe that
Sr={ XE B: til- < I}. Takip.g x=ry.
Sr={rYEB: \I y 1/ < l}=r {yeB : " y II < I}
338 BANACH SPACE

or S,=r SJ
This ~ T (S,)=nr Sl)=r T(Sl)
or T(Sr)=r T(Sl) ... (1)
as T is a linear map.
Aim. T(S,) = image of Sr contains some S,'
II. First we try to show that T(Sd contain'! Sr'. For each
positive integer n, consider open sphere S.. in B. Then

As T :,B -+B' is onto and linear and so

B'=T(B)=T [ ~l S.. J= "~1 T(S.. )

00
or B'= u T(S.. ). ... (2)
.. =1

Being Banach space, B' is complete as a metric space and so

it is of second category, by Baire's category theorem [Refer
Definition 18'21]. In this event (2) proves that the set T(Sm) for
some value m of n is not non·dense set in B'.
int [T(')'",)]~I/I

T(Sm) has an interior point, say, Yo s.t. Yo E T(S.,.).

It is ea,y to see that the map
f: B'-B' s.t·[.y)=y-yo
is a homomorphism.
Then [T(Sm)) Yo has the origin as an interior point ... (3)
Now we assert that
T(Sm)-Yo C T(S2fR).
Let y E T(Sm)-Yo be arbitrary. then i x E Sm s.t.
y=T(x)-yo ... (4)
Again Yo E T(S... ) => 3 Xo E Sm S.t. Yo = T(xo)
Now by (4), y=T(x)-T(xo)=T(x-xo) ... (5)
where x, Xo E Sm.
This => " x -xo II < 2m=diameter of open sphere Sm
=> jc-x~ E S2fR ~ T(x -xo) E T([Link]")
=>:1, E T(S2"')' by (5).
Thus i E T(Sm)-Yo ~ y E T(S_)
 BA~ACH SPACE 339
:. T(Sm)-Yu C T(S2fIl)=2m T(St), by (1)
or T(S.. )-y. C 2m T(Sl)
This => [T(S... ) - Y.] C [2mT(Sl)]=2m T(S1)
~ [T(S.. )]-yo C 2m [f(SI)] .•. (6)
From (3) and (6), it follows that the origin is also an interior

point of T(Sl)' Hence:l E > 0 s.t. c T(Sl)' S: ... (7)

We complete the proof by showing that
So' C T(S.). . .. (8)
Let yes: be arbitrary, then I~ Y II < E, by (7).
Y E T(Sl), this ~ Y is an adherent point of T(Sl)
~ 3 Yl E nSl) s.t.1l Y-Yt II < E/2.
By (7), Yt E r(St) ~ 3 Xl E Sl s.t. Yl==T(x 1 ), 1/ ~ a< 1/2'.
Again we observe from (7) that

S'(012) C [T(SU2)j
Also II Y-Yl II < ./],.
This ~ l'-Yl e S"/2 c (T(St/.)].
Proceeding a above, 3 Y2 E T(Sl/2) s.t.
II (y - YJ) -Y2 II < (E/2!) where Y.= T{X2)
and II XI II < i·
Continuing in this way, we obtain a sequence < x. > in B
I
s.t. II x,. II < 2"-1 ... (9)
E
and 11 Y--(Yl+Y2+···+Y.) n < 2W ..• (10)
. where X,,= T(x,,).
..
Write s,,= X Xl.
t-l

For n > m, we have II s"-s,,, n=11 t_l' 1 Xi II

i: x,- '-E
or II s,,-s... H ~, Xt II ~ II Xm+t 1\ + II x.+.11 + ... + II x. n
=11 i_m+l
I
< ~ + 2-+1 1+"'+2,,-1
1 (h' h '. . G P )
W IC IS In "

= i,; (l--~)::
1
_1__ .-!-=0
2-- 2,,-m-l 1
1-2"
as m, n~oo.
340 BANACH SPACE

Consequently <S,,> is a Cauchy sequence in a complete

Lim
space B and so 3 x E B s.t. s,,=X.
n~oo

II s" /I = II i~l x,1I ~ II Xl II + ... + II x" II

1 I I-U)"
< I +i+ P +"'+ 2,i="1 =-1-.--

=2 [I-~l < 2
or II s" II <:: 2
II x II = II lim sro!l ==;lim II Sn II <2<3.
This ~ xESa
" n
x=lim Sn ~ x=lim 1: Xi -= E Xi
4_1 i=l

00 (1)

~ T(x)= E T(Xj)= E y, as T is continuous

i=1 '=1

-> T(x =Iim (Y1+Y2+ .. +y,,)=y, by (10).

Thus, y=T(x), 1\ X:I < 3 and so YET(Sd)'
Fin!lliy, we [Link] shown that
YES.' ~ yEO T(Sa)
Of So' C T Sa)'
This completes the proof.
Theorem 14. The open mapping theorem.
If 8 and 0' are Banach spa::es and f T is a continuous linear
map of B ihto B', t"en T is an npen map.
[Meerut 1987; Delhi 81 ; Buodelkhand 82; Kanpur 79]
Proof. Let Band B' be Banach spaces over the same field K
of scalars. Let
T: B~B' be a continuou,> linear map.
Aim. T is a'1 open map.
Let G be an open set in 8. Then we have to show that 1'(G)
is an open set. Let y ~ T(G) , then 3 xEG s.t Y= T(r).
G is open, XEG ~ 3 an op~n sphere sex. r) in B s.t.
r) C G sex, ... (1)
S(x, r) stands for open sphere centered at x and radius r, S,
~ANACH SPACE 341
/
,tands for open sphere centered at the origin and radius r. Then
Sex, ,)=x+S,.
Now (I) becomes x+S, c G. ...(2)
By the Lemma 13, 3 open sphere S' in B' S.t.
S' C T(S,)
'1
or y+S'
'1 C y+T(S,) ... (3)
'1
(\) -> T(x+ S,) C T(G) Q 7'(x) + T(Sr) C T(G)
or y+ T(Sr) C T G)
or Also, by (3), y+ S' c y+ T\S,) C T(G)
'1
or y+S' C T(G)
'1
or S'( ) C T(G).
y, '1
Thus we have shown that
e T(G) ~ 3 open sphere S (
any y
y, '1) in B' S.t. S'(,).-, '1) C T(G).
This ~ T(G) is open set.
Theorem 15. A one to olle t:ontinuous linea, map of Banach
.fpace B onto a Banoch space B' is a homeomorphism. In parti-
cular, if a one-one linear map T of a Banach space onto Itself is
continuous, then its inverse T-J is automatically continuous.
[Meerat 1976J
Proof. Let Band B' be Banach spaces, Let T: B-B' be
continuous linear map s.t, T is one-one onto ma.p.
To prove that T is a homeomorphism, we have to prove
that
(i) T i., one-one onto
(ii) T is continuous
(iii) T-t is continuous.
Our allsumption :> (i) and (ii).
Remains to prove (iii).
To prove (iii), it is enough to prove that T is an open map.
Prove it as in Theorem 14.
Second part follows atonce from first part if we put B' =B.
Remark. This theorem is sometimes known as Banach
theorem.
 342 BANACH SPACE

Theorem 16. Suppose P Is a projection on a Banach .space B

lind M and N are its range and nulL spaces respectively. Then M
and N are closed linear sub3pace3 of B st. B=M(J)N.
Proof. Suppose P is a projection on a Banach space B, then
PI=P and P :.B-+B is a continuous map.
Let M and N be range and null space of P, then
M={x e B: P(x)=x}={x e B: P(x)=/lx)}
or M={x E B : (P-/) (x)=O} ... (1)
N={x E B: Px=O} ... (2)
Evidently N=p-l {O}, M=(P-l)-l (0)
B is Banach space ~ B is metric space
~ every singleton set is closed
:> {OJ is closed set
=> p-l {OJ is closed as P is continuous
=> N is closed.
Beip.g identity map, I is continuous map and so p-/ is con·
tinuous. Consequently M is closed
any x E M n N ~ x E M, x EN=> Px=x, !*x=O
·.'x=O
:. M n N={O} ... (3)
any z E 11 => z=P(z)+t-P(z)=p(z)+(I-P) (z)
Taking m=P(z). n=(/-P) (2'), we get
z=m+n ... (4)
P(m)=PP(z)=PI(i)= f(z)=nt
P(n)=P(I-P)(z)=(PI-PI)(z)=(P-P)(z)=O (z)=O
Thus P(m)=m, P(n) =0. this => m Eo: M, n e N.
Now (4) => B=M+N.
By (3) and (5), B=M(fJN. . .. (5)
Theorem 17. Let B be a Banach space and supp03e that M and
N are two closed linear subspaces oj B 3.1. B=M$N. Ijz=x+y
is the unique representation oj a vector in B as a SUm of vectors in
M and N, then the map P defined P(z)=x is a projection on B whose
range space and null space are M and N. [Meerut 1970j
Proof. Our3Ssumption. P(z)=P(x+y)=x where x E M,
y E Nand
M={x e B: Px=x}:=arange space
N={x E B: Px=O}=nulI space
BANACH SPACE 343

P2(Z) = P(PZ) = PX=X, by def. of M.

=P(Z). by def. of P
:. P2=P. It can be shown that P is a linear map.
Remains to prove that P is continuous. Let D' denote the
linear space B equipped with new norm" /I' defined· by
1/ Z 11'==11 x 11+/1 y /I.
Then B' is a Banach space (Refer Problem 2).
Also II P(z) 11=11 x II ~ II x 11+11 y 1,=11 z I/'
or \I P(z) " <. II z '"
This => P is bounded and is continuous. Remains to prove
that Band B' have the !lame topology. For this let T: D-+.D' be
an identity map S.t. T(z)=z.
Then II T(z) 1/=11 z 11=11 x+y \I ~ II x 1/ + lIy" = II zl/'.
or II Ttz) II ~ II z ;1'
This => T is bounded => T is continuous.
Also T is one-one onto map.
Finally Tis. one-one Onto and continuous map.
This => T is a homeomorphism (Refer Th. 15).
Hence Band B' have the same topology.
The proof is now complete.
Theorem 18. The closed graph theorem Let Band B' be
Banach spaces. Let T: B- B' be a linear map. Then, T is continuous
iff its graph is closed. [Kanpnr 1988. 84; Meerut 82, 88]
Proof. Let Band B' be Banach spa(es. Let T: B-+B' be a
linear map.
Then the graph of T is
TG={(x, T(x» : x E B} C B x B'.
Recall that a set A is closed if A=A where A=closure of A.
I\,lso A C A is true V- A.
I. Let T be continuou<;. To prove that TG is closed. ·Let
(x, y) E fa be arbitrary.
(x, Y) E TG= Tr::r U D(TG)
=- (x, y) E TG or (x, y) E D(TG )
=> (x, y) is a limit point of TG
,.. :I se'luence < x .. , T(x .. ) > E TG s.t.
lim
n-+oo (x", T(x,,» == (x, y) ... (1)
But x,,-x => T(x.. ) - T(x) ... (2) as T is continuous.
Abo T(x,.) -+ y •.• (3), by (I).
344 BANACH SPACE

Combining (2) and (3), we get T(x)=y.

:. (x, y)=(x, T(x» E To
Thus (x, y) E To *
(x, y) E TG
:. T C To. But To C T G is always true. Hence T c= To,
this ~ To is closed.
II. Let Tu be closed. To prove that T is continuous. Band
B' are Banach spaces => Band B' are complete spaces
=> BxB' is complt!te.
Also closed subspace of a complete space is a complete space.
Hence To is a complete space. Define a map
I: To-.B s.t.f[(x, T(x))]=x
Evidently I is one-one onto and linear map. Also
II/«x, T(x»] 11=// x II ~ II x 1/+/1 T(x) 11=1/ (x, Tlx») II
or 1I/[(x, T(x))] II ~ I " (x, T(x)) "
This => lis bounded => I is continuous map;
Then I is homeomorphism. (Refer Theorem 15)
Consequently 1- : B-.1'o is continuous and so it is bounded.
1

(Refer Problem 3a).

II T(x) \I C;; " T(x) 1/+/1 x 1/=1\ (x, T(x» 1\ =i//-1 (x) II
<; ".f- " • " x II
1

or "T(x) II .; "f- 1 II • " x II

Also f- 1 is bounded and so T i~ bounded and therefore Tis
continuous.
Theorem 9. The uniform bounded ness theorem (Banach
Steinhaus Principle). Let B be u Banach space and N Q normed
linear space. If {T,} Is u non-empty set of continuous linear
transformation B_N s.t. {T.(x)} is bcunded subset 01 N ¥ x ~ B,
then { " T, "} Is a bounded set 01 numbers, i.e. {Ti} is a bounded
.fubset ofD (B, N). [Kanpur 1988,82]
Proof. T,: B .. N is a linear map and so
T. (x) E N ¥ x E B.
For every positive integer, n, we define
F,.-{x : x E B, " T, (x) " ~ n "" i} ... ( I)
S, (x)=open sphere with centre x and radius r
S" (xl-closed sphere with centre x and radius r
x E F,. '*
II T, (x) " ~ n, T. (x) E N
9ANACH SPACE 345

~ Tt (x) E Sn [0] ~ x E T,-l {S" [OJ} ¥ I

~ x E n Tt-I {S" [O]}

...
i

Fn= n T,-l {Sn [OJ} .. (2)

t

We know that
S. [0] is closed set in Nand Ti : n. . . N is confinuous
=> T.-l {Sft (0]1 is closed.
Being an arbitrary intersection of clo~t:d sets, F,. is closed
according to (2) ..
...(3)

Also B= U Fn- ".(4) according to (2).

n is Banach,. B is complete
.. B is of second category, by Baire's Theorem.
[Refer Def. 18'21 (ix)]
~ B is not of first category.
=> one of the F.. 's, say Fm is dcnlle set.
=> int (F",):;t:{21
[But Fm i<; clo~ed => F",=Fr..J
.. int (F.,,)::I=0 => (F",t:r=0
=- 3 Xu E (F",t => Xu i<; an interior point of £;"
=> 3 closed sphere Sr [xu] s.t.
s, [xo] C F,,, where S, [xo] s.t.
If /I y II fiib I, and ;=y, then II ~ 1'1-11 y Ii ~ I
or " ~ r, this 0 Z € Sr [0]
z II
~ z+x" E Sf' Ixo] C Fm
.. z+xo E F", ~ II T, (z+xo) II .;;; m ... (4)
xoE: (Fmt. Also(Fm)OC Fm ¥ Fn>
... XO E F. ... this => II Tt (x o) II ~ m ... (5)
Also II T, (y) /1=1/ T, (z/r) 1/= Ifr II T, (z) 1\
=1/r II T, [(z+x o) xo] II
or II T, (y) II=I/r [II T, (z+xo)-T, (xu) II
<.1/r [ U T, (z+xo) 11+11 T, (xu) II J
<'1/r [m+m]=2m/r, by (4) and (5).
or II T, (y) U<; 2m/r,
346 BANACH SPACE

Thus II T. II=sup { II T. (y) II : II y II ~ I} ~ 2m/r

or II T, II (: 2m/r ¥ r.
Thus the set { II Ti II } is a bounded set of numbers.
Remark. This theorem is also known as Banach-Steinhaus
theorem.
Theorem 20. A non-empty subset X of a normed linear space
N is bounded <;!>f(X) Is bounded set o/nun/bersfor eachfin N*.
[Kanpur 1981, 80; Meerut 88, 80]
Proof. Let X be a non· empty subst:t of a normed linear
space N. Letf E ,..* be arbitrary, then/: N~K is a continuous
linear may, K is a field of s~alars.
f is continuouli do / is bounded. (Refer Problem 3a)
.¢> a> 0 S.t. IIf(x) II ~ a II x II ... (1) ¥ x E N
I. Let X be bounded, then 3 b :> 0 s.t.
<
II x II b V x E X. . .. (2)
To prove that/tX) is bounded.
By (1) and (2). I f(x) I , ab y x € X
This ~ /(X) is bounded set.
II. Letf(X) be bounded
To prove that X is bounded
For convenience, we write X = {Xi}
Then flX) is bounded Q {/(x,)} is bounded. ..(2)
Detine a map Fx, : N*~K S.t.
Fx. (/)=/(Xt)
1 hen Fx. is a continuous linear map (Refer theorem 12).
This ~ Ex, E (N*)*=N**
Now. by (2). {Fx,(f)}. i.e. {f(xd} is bounded ¥ f e N*
Also N* is compl!;'t. (Refer Problem 2a)
Applying uOllform boundedness theorem 19, we find that
{ II Fx, II }
is bounded subset of N-it<*. But Ii Fx, Ii = II Xi II
(See cquatlOn~ of Theorem (2)
{ II x, } is bounded set. But {x.}=X.
This => X is bounded ~et
Definition. 18'24.
Conjugate of an operator.
lKanpor 1983, 81J
Let N be a Dormed linear space ov'er the field K of scalars.
Let T: N~Nbe a continuous linear map. LetfE N*.
Then T: N~N,f: N~K
This.. / T : N~K ~ /T e N*
BANACH SPACE 347

Define a map T* : N* ... N* s.t.

T*' (/}=IT V lEN·
This ~ [T~ (/)1 (X)=( IT) (x)
T~' is defined as conjugate or adjoint of T
T*islinear. ForifO!,8 E Kand/,gE N*,then
[T* (<<I+l3g)] (x)=(O!/+ Ilg) T(x)='IT(x)+~gT(x)
=.<Gt (IT) (X) +~(gT) (x)

=~ [T· (I)] (x)+p [T· (g)] (x)

=[zT* (f)+~P (g)] (X)
or P «(11+ ~g)=-a.T* (I )+f3P (g).
Claim. /I T 11=ll T* Ii
For each non-zero vector x E N,3 functionalf E l\ '" f:.t.
11/11= 1 and IllT(x)] \==11 T(x) iI. .. (I)
(Refer Th. 10)
II T\I=sup { II T(x) II : II x II , I}
=sup {.II r~x~!1 :X:;CO}
-sup { 1f~1~~)U : IIf!l=l, x~o} ... (2) by (I)
But If(T(x)] 1=1 (T·(f») (x) I=li pel) Ii . II x II
Using this in (2),
1\ T lI=sup {II T· <I) II : 11/11= I}=II P II

or II T II = II T" 1/.
This equation is expressible as :
the map T: ~ T*' is a norm [Link] mapping of B(N) into
BUV· ). [K aopur 1983]
Problem 10. Let 1 be an operator on a Banach Space B. Show
,hal
T has inv!'rse T-' <=> conjugate r* of T has all inverse (T·r 1
and th~t (T,i'<)-l=(T-l)lt,.
Solution. T has an inverse T-l
<=> TT-l=T-1T=1
~ (Tr-l)"'=(T-1T)*=/~

<=> (T-l)* T*=T* (T-l)*_1

<:> inverse of T-it is (T-l)t,
For AB=BA=I ~ A=B-l ('If B-A-·I
<:> (T·)-l:a=(T-·l).,
348 BANACH SPACE

Problem. If {x ..}. {y,,} are sequences in N and {a,,}. {,sIll are

scular sequences such that x,,~x, y .. ~y
and Cl,,~ ot, ~ ....... ~, then
"nx.+ P"r...... u+J:y. [Kanpur J989, 86]
Solution.
II o"x" + p..y,,-(u+py) II
-II Of"X.-tlt"x+""x-u+ P")'"-fJ,,y+fJ,,y-~y II
=11 Cln (X,,-X) +X (ot"-Il)+P,, (Yn-Y)+y (~n-~) II
~ I a .. I • Ii x .. - X II +lorn - ex I • " X II

-10.1. 0 +0.1 X
+ I ~n I· II y" - y
11+/ P 1.0+11 y 1/.0
,,+ /I Y II • I ~,,- # I
=0 as n_oo
... Lt
n-+oo
(arnx.. + !J"Yn)=a.x+fJy
Exercises
1. la) Define a normed linear space. Prove that I. 1he linear
space of all sequence X= < Xl' XI' ... > of scalars s.t.
-I
J::
.. -1
Xn I" < 00 with norm defined by

r 1 1/ p

II X I!,= ~ r:
L....1
I x" I" ~
j
,(I ~P< 00)

is a complete Hormed iinear space.

(b) If M is a ciol>t:d linear subspace of normed linear space
N, and if
onto
T: N --+ N/M defined by
7(x)=x+M. show that T is continuous linear transforma
1ion. for which !I T Ii E:: 1. lKanpur 1988]
2. (a) State and prove the Banach-Steinhaus PrlOciple of
'Uniform Boundedness".
(b) Prove that a normed linear space N is separable if Its
conjugate space N* is separab!c. [Kanpur 19l18]
3. (a) Define weak convergence and strong convergence in a
"'ormed linear space. Prove that strong convergence implies
weak convergence, but the converse is Iflot tlue in generaJ.
(b) State and prove the closed graph theorem.
[Kanpur 1988]
4. (a) Define a Normed linear space (NLS). Prove the follow·
mJ in a NLS (1'1, II 11):
BANACH SPACE 349

(i) ," x /l-IiYi!I"lIx-Yll.

(iiI If < x" > be a sequence in N s.t. x" ....... x. then
i: x;. I: - !I x "
(iii) If <x.>, <y,.> are sequences in Nand <7,,>. <fJ,,>
are scalar <;equences s t.
..
x;.~x. Yn->Y. at" -+Of, ~" ... {3, then ot"x .• +~ y ..-Gfx+fJy.
(b)· Prove that the linear sp'lce of all bounded sequences
x=·<x•• Xl' .. > of scalars with norm degncd by
II X II =sup ! x I
is a Banach spllce. [Kaopur 1989, 87]
5. (a) Explain tl!e natural imbedding of a NLS N in its second
conjugate N** and show that it is an isometric isomorphio;m.
(b) If M ic; a closed linear sub-space of a NLS N. and If T is
the natural mapping of N onto NI M defined by T(x) =x +M.
show that T is a continuous linear [Link] for which
II Til <= I. [Kanpu... 1987J
6. (a) Define weak convergence and strong convergence in a
NLS. Prove that in a finite dimensional space the notation
of weak and strong convergenc.! are equivalent.
(b) Let N be a non-zero NLS. Prove that N is a Banach
space? {x : II x II = I} is complete. lKanpur 1987J
7. Prove that the linear space C [0, 1J of all continuous real
functions defined on the closed unit interval [0. I] is a
normed linear space (NLS) w r.t. the norm as an elementi
defined by

"(II=C liU)ldr
but not a Banach space. ·lKanp.... 199.0,86]
R. Let Nand N' be normed linear spac:!s with the same scalars.
and Bf N, N') the set of all bounded linear operators of N
into N. Prove that B(N, N') is it~elfa NIS w.r.t. the point-
wise linea .. operations and the norm defined by
II Til =sup f " Tx II: II x n ~ I}.
Further, if N' h a Banach space. then B (N, N) is all>o a
Banach space. (Kanpur 1986]
9. (a> Define a normed linear [Link] (NLS). Prove /" the linear
space of all sequences
x={xIo x 2 .···.x., ... }
00
of sca1a-rs such that 1: I x" Ip < 00
R:=1
350 BANACH SPACE

I" 1'"
1
with norm defined by II x 11.= rEI
1 11=1
x"
f
is a complete Normed linear space.
(b) Let Nand N' be normed linear spaces and T a linear
transformations of N into N . Prove that the following conditions
on T are all equivalent :
(i) T is continuous
(ii> T is continuous at the origin
(iii) There exists a real number k ~ 0
s.t. " Tx /I <;k " x II ¥ x E N.
(iv) lf S is a closed unit sphere in N, then its image T(S) is
bounded in N'. [Kanpur 1985]
10. (a) Define a Banach space. Let M be a linear subspace
of a normed linear space N and let F be a functional defined on
M. If Xo is a vector not in M and if
Mo=M+{xo}
is a linear subspace spanned by M and Xu then prove that F can
be extended to a functional Fo defined on Mo such that
IIF.,II= II F II.
(b) Describe the natural imbedding of N into N** where
N is a normed linear space and show that it is an isometric
isomorphism. [Kanpur 1984]
11. State and prove the closed graph theorem.
[Kanpur 1984, 83]
12. (a> Define a Banach space. Let N be a non-zero normed
linear space. Prove that N is a Banach space
<:> {x : II x /I = I} is complete.
(b) Define the conjugate space of a normed linear space. If
N is a normed linear S1'lace, then prove that the closed unit sphere
S* in N* is a compact Hausdorff space in the weak* topology.
[Kanpur 1983]
13. (a) Define a normed linear space. Prove that in a nor-
m.:d linear space
1 IIxll-lIyll I <;;lIx-yli
and use it to show that norm is a continuous function.
(b) State and Prove Hahn-Banach theorem.
[Kanpur [Link]. F. 1982]
14. <a> Let N be a normed linear space. Describe the natural
imbedding of N into N** and show that is an isometric isomor-
phism.
BANACH SPACE 35(

(b) State and prove the uniform boundedness theorem.

[Kanpur [Link]. F. 1982]
15. (a) Prove that if M is a closed linear subspace of a nor-
med linear space X, and the norm of a coset X + M in the quotient
space XI M is defined by
II X+MI! =inf{ II x+m 1/: m EM}
then XIM is a normed linear space.
(b) Prove that the conjugate space of a normed linear space
is a Banach space. [Kanpur [Link]. F. 1981]
16. Define the conjugate of a bounded linear operator Ton
a normed linear space X into itself.
Prove in detail that a non-empty subset A of a normed linear
spac:! X is bounded if and only if J(A) is bound~d for every JE X*.
[Kaopur [Link]. F. 1981]
t 7. (a) Prove that if M is a closed linear subspace of a nor~
med linear space Nand Xo is a vector not in M, then there exists
an.!;' in N* such that
!o(M) =0 and !1I(XO).t:O.
(b) Prove that a normed linear space N is separable if its
conjugate space N* is. [Kanpur [Link]. F. 1980]
18. (a) Prove that the closed unit ball S* in the conjugate N*
of a normed linear space N is a compact Hausdorff space in the
weak* topology.
(b) Explain the concept of the graph of an operator. State
and prove the closed Graph Theorem. [Kaopur M.S.c. 1980]
19. (a) State and prove the uniform boundec ness theorem.
(b) Prove that a non-empty subset X of a normed linear
space N i5 bounded if and only if f(X) is a bounded set of num-
bers for each f in N*. [Kapour M.S.c. F. 1980]
 19
Hilbert Space

19'1. Inner product space. Let V be a vector space (linear

space) over a field K (real or complex). Let a, b E K and u, v.
WE V be arbitrary. Then < ,> is called an inner product on V if
the following are satisfied:

(11) < u, , > E K

(12) < au+bv, IV >=a < u, IV >+b < v, IV >
(/a) < u, v > = <v, u>, the complex conjugate of <u, v>
(/4) <u, u > ;;t 0 and < u, u >=0 iff u=O.
The vector space V with an inner product is called an illner
product space. A real inner product space is called a Euclidean
space, ard a complex inner product space is called Ii unitary
space.
19'2. Hibert space [Kanpur [Link]. F. 1984, 83]
A Hilbert space is a Banach space in which norm is defined
from a scalar valued function called an inner product.

A veetor space together with inner product is called inner

product space. An inner product space is a normed vector space.
If this normed vector space is complete under the norm obtained
from the inner product, then it is called Hilbert Space. That is to
say, a complex Banach space H(K) together with complex func-
tion < x, y >, called inner product is called Hilbert Space if the
following conditions are satisfied:
(i) < x,> E K
y
(ii) < x, y> = < y, x >
(iii) < aX+J1y, Z > =ot < x, z > +fJ < y, z >
(iv) < x, x >= II x W V- x, y, Z E H and ex, fJ E K,
 HILBERT SPACE 353

Example. Consider the Banach space I." consisting of all

n-tuples of complex numbers with the norm of a vector
x=(xl> xI> "" x,,) defined by

II x II =( ._1.E I Xi /2)1/2 and

< X, y > = E" X, }i,

i=1

where Y=< )'1' Y2, ...• y" >, Z=< ZIt Z2' ... , ZIt > ; Xi, Yi, at, fl
ar~complex numbers. Evidently
(i) < x, Y > E C

=(Xl Vl)+(X 2 y~)+ ... +(x" )i,,)

n
=X1Yl+XIlY2+"'+X" y,,= 0=1
1: y, x,=<y, x>

(iii) OIx+flY=1I (Xl> ... , x,,)+~ (Yl' ... , y,,)

=(oeX1 +/lYh "', u,,+/ly,,)

< ax+fly, Z >= .. ""

E (oex,+fly,) Zt=1X E Xi z,+fl E Y. Zf
1=1 .=1 i-I

=a: < x, z >+fl < y, z >

f\ '"
(iv) < x, x > = 1=1
E Xi X.= E / Xi
t::cl
/1= II X III
Hence 1'1" is a Hilbert space.
19'3. Adjoint operator. [Kanpur 1983]
Let T be a linear operator on a Hilbert space, then its
adjoint denoted by T*, is a unique operator on H satisfying the
condition
< Tu, v >= <U, T* v > "f u, v E H.
To prove that T* is linear.
Let u, v, w E H and a, b be scalars.
< u T,'" (av+bw) >=< Tu, av+bw >, by def. of T*
=< Tu, av >+< Tu, bw >
=0 < Tu, v >+b < Tu, w >
 354 HILBERT SPACE

==a (u, T* v)+/j (u, T*' w)

=(u, aTf:. v)+(u, bT"" w)
_=(u, aT"" v+bT* w)
This is true ¥ u E H. Hence the last gives
T* (av+bw)=aT* v+bT"" w
This proves that T* is linear.
The set of all linear operators on the Hilbert space H is
denoted by B (H) It can be shown that B (H) is a Banach space..
19-... Definitions of different types of [Link].
Let T be a linear operator on a Hilbert space H.
Then (0 Tis DOrmal if TT-»=T""r
(ii) T is unitary or orthogonal if T*T=TT""=/..
(iii) T is self adjoi nt or Hermitian if r = T* •
(iv) Tis positive if (Tu, u) ~ 0 ~ u E Hand T=T*'.
1"5. Convex set
Let H(K).be a real or complex Hiibert space. A non-empi)
subset S of H is said to be convex if
x, YES ~ (I-Gl) X +ay E S where IJI is rea) and 0 ~ Cl ~ 1.
If we take 2=i, tben x. yES .. xt y E S.

Tbeorem. A closed [Link] of a complete space is complete.

19'6. Conjugate space H *.
Let H be Hilbert space~
A continuous linear map f: H~C is called linear fuuctional
or more precisely functional on H. The set L (H, C) of all func-
tionals on H is denoted by H* and is caned cODjugate space of H.
19·7. Continuity. .
An operator T on a Hilbert space H is said to be continuous
at Xo e H if given any sequence (xIt ) e H
s.t. X"-+X o E H ~ TXIt -+ Txo in H.
19'8. Ortbogonality.
Let u, v be arbitrary elements of a Hilbert space H. Also let
W, U be subspaces of H.
(i) Definition. The vectors u, v are said to be ortbogonal if
(u, v)=O.
Result 1. The relation of orthogonality is symmetric.
For u..Lv ~ (u, v)=O => (v, u)=(u, v)=O=O. [Link]
u.L v means "u is orthogonal to v".
Result 2. The zero vector is orthogonal to every vector.
For (0, u)=(O U, u)=O (u, u)=O. .
Result 3. PytbagoreaD tlaeorem. Iff is orthogonal to g,
then II/+gI1 2 =lIfIP' +ilgll l •
Proof. lis orthogonal to g => (r. g'==O, (g,/)=O ... (1)
HILBERT SPACE 355

n/+g 111=< I+g.l+g) = (.f. I+g) +(g.l+ g)

=(/.1)+<1. g)+(g./)+<g, g)
=111112+11 g 11 1 +<1. g)+(g.l)
=11 1111+11 g Ill. by (i).
(ii) Definition. Any vector u is said to be orthogonal to W.
i.e. u .1 W jf (u, w)=O y w e w.
(iii) Definition. The spaces U ~nd Ware said to be ortho-
gonal i.e. U.l W if (u, w)=O ¥ u e U and w e w.
(iv) Definition. The orthogonal complement of U, denoted
by UJ., is defined as the set of vectors v e H s.t. v is orthogonal
to every u E U.
Symbolically. UJ.={x E H: (x. u)=O yo u E U}.
199. Orthogonal set.
Let {u • ...• u.. 1 be a subset of a Hilbert space H s.t. each
u;;cO. The set {u,} is said to be orthogonal set if the vectors are
pairwise orthogonal, i.e. if
(Ui, u/)=O for i:;ej
The set {ud is said to ,be orthonormal set
'If ( Ut, )"
Ul =Oi}=
{I0 i=j
h=j
19'10. Complete orthonormal set.
An orthonormal set is said to be complete if it is not contai-
ned in any orthonormal set.
19'11. Definition.
Let T be a linear operator on a Hilbert space H. then
(i I T preserves norm'i if II T(a) II = II ex H "t at E H
In this case Tis also called isometry.
(ii) T preserves inner product if
(Tat, 713)=(11, ~> yo at, ~ E H
'F' ; .. called isomorphism if T is one-one onto and T pre-

serves inner prouu,",,~,

19'11. Perpendicular projection.
Let T be a linear operator on H. Then T is said to be projec-
tion 011 U along W if
TfV)= T(u+w)=u with u+w=v E H=U~W, u E U. we W.
A linear operator Ton H is said to be perpendicular proJec-
tion on U along UJ. if
H=UffiUJ. and Tv=u where ,=u+w, vE H, uE U. weUJ..
Remark. (i) Since UJ. is uniquely determined by U and
hence we only say that T is a perpendicular projection on (j
instead of sa"ing 'T' is a perpendicular projection on U along UJ.
(ii) Range of T= U. Null space T UJ..
356 HILBBRT SPACE

19'13. Parallelogram law.

If ac, fJ are two vectors, then
"1I+fJ 112 +11 «-fJ 111=2 [II «IP'+II fJ In
Theorem 1. (Some properties of a Hilbert space). In a Hilbert
space H, prove that
(i) (a.x-fJy, Z)=II (x, z) -~ (y, z)
(ii) (x, fJy+yz)=13 (x, y)+.y (x, z)
(iii) (x, 0)=0 and (0, x)=O
¥ x, y, Z E H a"d II, {J are scalars.
Proof. (i) ( ..x-{Jy, z)=(<<x+( -~) y, z)
=11 (x, z)+( -~) (y, z)
=/l (x, z)-{J (y, z).
(ii) (x, {Jy+)'z}=(~y+)'z, x)={J (y, x)+" (z, x)
=13 (y, x)+y (.x) r
=13 (x, y)+y (x, z).
Put y=O and so
(x, ~y)=j:1 (x, y)
where as (~x, y)={J (x, y).

(iii) (x, O)=(x, 00)=0' (x, 0)=0 (x, 0)=0

(0, x)=(OO, x)=O (0, x)=O.
Theorem 2. (Schwarz inequality). If x, yare two vectors in
~ Hilbert space H, then
I (x, y) I ~ II x" II y II·
[Kanpur 19l9~9'O,188, 86, Meeruti 69)
Proof. Evidently inequality is valid when x=o or y=O.
Let us suppose that x,cO and
(y. x) x
z=y- 11 x W • ...(1)
/ (y,x)x '\.
Then (z, x)="y - II x II_ ,x/
(y,x)
=(y, x)- IfX1f (x, x)=O.
For (x, x)=11 X 112
:. (z, x)=O' so that (x, z)=O. ... (2j
Now " z II ~ 0 or (z, z) ;;;. 0
·;>r /y_ (y, x)~ z';;;. 0
" IIxW' /
or (y, z) - ~r~ ~I? (x, z) ~0
HILBERT SPACE 357
Using the result (2), we get
/ (y, x> x",,-·
(y, z) ~ 0 or ""- y, y- II X 112 / ~ 0

(y, x) 0
or (y, Y)-"if"Xli2 (y, x) ~

or II Y \12 \I x 111- I (y, x) III ~ 0

Gr II y II 2 II X III ~ I <x, y) I I
or I (x, y) I ~ II x II • 1\ y II •
Theorem 3. In a Hilbert space the inner product is jointly
continuous, i.e.,
Xn~X, yn~y => <xn, yn)~<x, y). [Kanpur 1988, 90J
Proof. I (xn, Yn) -(x, y) I
= I (x .. , Y.. )-(xn, y)+(x.. , y) -(x,y) I
= I (Xn, y .. -y)+(x.. -x, y) I
~ I (x .. , Yn-y)I+l<x .. -x, y) I
as I a,+tJ I ~ I a, 1+1 fJ I
~[Link],.-yll+llx.. -[Link] ... (1)
by Schwan. inequality.
Now Xn~X, Yn-+y when n-+oo.
Hence II xn-x II - 0, \I Yn-y II ~ 0 as n-+oo.
Now (l) gives
I <Xn,yn)~(X,y) I ~ 0 for n~oo.
But I (x n , Y .. ) -(x, y) I ~ 0 is always ture.
Lim Lim
Hence
n~oo
I (xn, Yn) - (x, y) \=0 or
n-+oo
(Xn' y .. )=(x, y)
Problem lao (Law of parallelogram). If x, yare two vectors
of a Hilbert space H, then
I x+y 11 2 + \I x-y 112=2 [II x 112+ II yin
and 4(x, y)=11 x+y :I2-1! x- y Il ll +i II x+iy II L i II x- iy III.
[KanIJur 19118]
Solution. Ii x+y 112=(x+y, x +y)
=(x, x+Y)+(Y. x+ y)
=(x, x)+(x, y)+(y, x)+(y, y)
0, II x+y 112=1: x 112+11 y iI 2 +(x, y)+(y, x). . .. {l)
Also II x-y 11 2 =(x-y, x -y)=<x, x-y) -<y, x-y)
=.(x, x) --(x, y)-[(y, x)-(y, y)]
or \I x - Y i1 2 = \I x :13..t- II y IL2 -<x, y) - (y, x). ...(2)
Adding (I) and (2), we get the result.
358 HILBERT SPACE

1/ X+y 1/2+ II x--y 1!1=2 [ 1/ X 112+11 y 11

Second Part. (1) - (2) gives
II X+y 112_ II x-y 112=2 [(x, y)+(y, x)] ... (3)
Replacing y by iy, we get
II x+i y \12_ II x-I y 111=2 (x, iy) -(iy, x)]
=2 [T(x, y)+i (y, x)]
or, "x +iy 11:1- II x- iy 1/2=-2; [(x, y) - (y, x)]
as ;=-;
Multiplying both sides by I.
i" x+iy 112-i II x iy 11"=2 [<x, y)-(y, x)] . .. (4)
(3) + (4) gives
II x+y 111- II x-y 112+i 1/ x+iy 112-1 II x-iy 112
=4 (x, y).
Problem lb. If B is a Complex Banach space whose norm
obeys the Parallelogram law, and ifinner product is defined on
Bby
4 <x, y)= II x+y II 2_ II x-y 111+; II x+iy III-I" x-iy 1111
then B is a Hilbert space. [Kanpur 1986; Meerut 77J
Solution. Let x, y. z € B be arbitrary and ex scalar. Also
let
II x+y 112+ II x-y 111=2 [ II x 112+ II y 112] (1) . • .

4(x, y)= II x+y II 2_ II x-y 112+i II x+ iy 1\2-i 1\ x-iy 1\11

... (2)
To Prove B is a Hilbert space, we have to prove that

(i) (x, x) = !I x II 11

(ii) <x, y)=(y, x)

(iii) (x+y, z)=(x, z)+(y, z)
(iv) (ex x, y)=1& (x, y).
Here note that the Properties (iii) and (iv) are equivalent to
the single property
(l&x+[3y, z)=ot (x, z)+11 (y, z)
(0 Put y=X in (2),
4 (x, x)= 1/ 2x IPI- II 0 II 2+; II x (1 +i) II 2-i II x (I-i) II :I

Using the fact that II a u II I a I II u II

= and
11+1 1=\1'(1)>+12) =\I'2= l l-i I
we get
4 <x, x)=4 II x I/I+i 2 II x /!I-i 2 II x /12=4
 HILBERT SPACE 359

or (x, x)= 1\ xIII. Hence the result (i).

(ii) " Til = II 1 II . Since norm of any number is a real number
and hence II x+iy 11= II x+ly II .
In view of this, we take the conjugate of both sides in (2);
4 (x, y)=:\ x+y 1/ 1 -- 1\ x'-y 111-1 1\ x+iy 1\2+1 II x-Iy III
... (3)
Interchanging x and y in ,(2),
4 (y, x)= 1\ y+x 112_ II y-x 111+1 II y+lx 112_1 II y-Ix "I
using the fact that II y+lx /1 2 = II I (x-iy) /1 2 = 1 i 1111 x-Iy III
=lIx-iylll
and II y-ix 11== II -I (x+iy) 1/1= I -I Pili x+iylll= II x+ly 1/ 1
II y-x 111= II (-I) (x-y) nl = I -111/1 x-y 111= II x-y 1/ 2
we get
4 (y, x)= II x+y 111- II x- y 1I11-t ill x-iy 112-1 /I x+iy II·
... (4)
Comparing (3) and (4), we find tllat
4 (x, y)=4 (y, x) or (x, y)=(y, x)
. Hence the result (ii).
(iii) Aim (u+v, w)=(u, w)+(v, w)
By (2),
4<u, w)= II u+w 111- II u-will+i II u+iw Ir'-i II u-iw /I'
... (5)
and 4<v, w)= II v+w /1 1 - II v-w 111 +/11 v+iw 1/1-1 II v -Iw III
... (6)
Adding (5) and (6), we get
4[(u, w)+(v, w)]={ II u+ w 11 1 + II v+w III} -{ II u-wll l + II v -w liZ}
+1 { II u+lw /11+ II v+iw III}
-I { II u~iw 112+ II v-iw III} ... (7)
By (2),
4(u+v, w):;: II (u+v)+w "1- II (u+v)-w 11 1+/11 u+v+iw III
-I II u+v-iw /1 1 ... (8)
By (I), II (uf v)+w II '= II (u+w)+v 112
=2 [II u+w 11 1 + II v II I J- I' u tw-v" t
=2 [ II u+w 11 1 + II v 11 2J- II u+(w~ v) ~ I
again by (I),
=2 [ II u+w II 1+ II v II'J
-{2I1u 11'+211 w-v jl2 - 1/ u-(W_V)!!2}
=2 [lIu+wli l + I v 112 -II u III-II w-v 112]
+lIu+v-wll Z
360 H-II:BERT SPACE

or "u+V+W 111- II U+v-w 112=2 [ II U+W 11 2+ II v 112_ 1\ U 112

- II W-V II 2 ••• (9)
Interchanging u and v, we get
" U+V+w \1 2 _ II u+v-w 111=2 [II v+w It+ II u 112_ II V 112
- II w-u 121 ... (10)
Adding (9) and (10) and then dividing the whole equation by 2,
n u+v+w 112_ II u+v-w 111= II U+w 11 2 + II v+w 112
- II w-v II 2_ II w- U II 2 ... (11)
Replacing w by iw and the mUltiplying by i on both sides,
ill u+v+lw 111-/ II u+v-iw 112
=i [ I' u+iw 112+ II v+iw 112_ II iw-v 112_ II iw- U112]
But II iw-v II = II v-iw II
:. ill u+v+iw III-i II u+v-iw 112
=i [II u+iw 111+ II v+iw 112_ II v-iw 112
- II u-iw 112] ... (12)
Adding (11) and (12) and then using (8)
4(u+v, w)= n u+w I: 1+ II v+w 12_ II v-w" 2_- I u--w 112
+i [ II u+iw 1 2+ II v+iw 112 __ II v-iw 112_ II u-iw n 2J
Comparing this with (7), we find that
4(u+v, w)=4[(u, w)+(v,w)]
or (u+v, w)=(u, w)+(v, w).
or (x+y, z)=(x, z)+(y, z) .. _(13)
Hence the result (iii).
(iv) We establish this result in several steps.
(a) When lit is a positive integer.
Put x=y in (13), (2x, z)=(x, z}+(x, z)
or (2x, z)=2(x, z)
or (2x, y)=2(x, y)
Thus the result (iv) is true for :%=2.
Let it be true for a.=n (Positive integer), then
(nx, y)=n(x, y)
«n+l) x, y)=(nx+x, y)=(nx, y>+(x, y), by (iii)
=n(x, y) +(x, y)=(n+ I) (x, y)
:. (iv) is true for a.=n+ I if it was true for lIt=n. Hence (iv)
is true for all positive integer a.. ~
(b) When Gt is a negative integer.
Replacing x by -x in'(2),
4( -x, y)= II -x+y 112_ II -x- y 11 2 +; U -x+ly 112
- i II -x-iy 112
= II x- Y 1\2_ II x+y 1\2+; II x-iy II 2_; II x+iy II 2
HILBERT SPACE 361

Comparing this with (2), 4( -x, y)= -4(x, y)

or (-x, y)=-(x, y) ... (14)
Let «= - {J, where fJ is positive integer.
«xx, y)=(-~x, y)=-(fJx, y), by (14)
= -fJ(x, y), by (a)=«(x, y)
(c) When« is a rational number.
Let «=~ where p and q are integers and q:;t.O.
q
/p
«lX, Y)='i. x, y
"-
/=,/ P (X)
q ,Y"/=P ""q'
/x y /, ,
by (a) and (b)
/x "-
or (aX, y)=P" q' y / ... (15)
X
-=z, then x=qz
q
(qz, y)=q(z, y) by (a) and (b).
or (x, y)=q(z, y) or 1 (x, y)=""ii'
q /x y "-
/.

Putting this in (15), (ax, y)=~ (x, y)=«(x, y).

q
Similarly it can be proved that
(<<x, y)=cx(x, y) when « is real.
(d) When« is a complex number.
Replacing x by Ix in (2),
4(ix, y)= II ix+ y iI2-- II ix- y !12+1 II Ix+ly \l 2-i II i'x-iy 1,1
= I i i2 11 x-iy 112_ I i 12 lI.x+iy 112+i I i 1
2 " x+y hi

-; I i III \I x - Y li2
= II x-iy 1:2_ II x+iy 111 +1 II x+y 112 -i II x --y 112
Comparing this with (2),
4(ix, y)=4i(x, y) or (ix, y)=i(x, y) .:.(16)
Let «=CX1+;«2 where Cltl and 01 2 are real.
<OIX, y)=«cxl+/~2) x, y)=(CX1X+icx2 x, y)
=( CXI X, y)+(;cxzx, y)~ by (iii)
=Clt1(x, y)+«2(ix, y), by (c)
=cx1(x, y)+iCX9(X, y), by (16)
=(«1+ 1<<2) (x, y)=«(x, y)
Hence «(lX, y)=«(x, y) for every scalar a.
Hence B is a Hilbert space.
360 HILBERT SPACE

or /I U+V+w 111- II u+v-w 111=2 [ " U+W 11 2+ II v 112_ II U 112

- II W-V II 2 ••• (9)
Interchanging u and v, we get
II u+V+w 112_ II u+v-w 11.=2 [II v+w /1+ II u 112_ II V 112
- II w- u / 2] ... (10)
Adding (9) and (10) and then dividing the whole equation by 2,
nu+v+w 112_ II u+v-w II Z= II u+w 111+ II v+w III
- II w-v II 2_ II w- u II 2 ..• (11)
Replacing w by iw and the multiplying by i on both sides,
i" u+v+iw" I-I II u+v-iw II 2
=i [II u+iw" 2+ II v+iw 111- "iw-v 112_ II iw-u III]
But II iw - v II = II v- iw II
:. i" u+v+iw III-i II u+v-iw II 2
=i [ II u+iw 111+ II v+iw 112- II v-iw 112
- II u-iw 112] ... (12)
Adding (11) and (12) and then using (8)
4(u+v, w)= II u+w I: 1+ II v+w 11- II v-w 112~ I u-·w 112
+i [ II u+iw 1 1+ II v+iw II 2__ II v-iw 112_ II u-iw n 2]
Comparing this with (7), we find that
4(u+v, w)=4[(u, w)+(v,w)]
or (u+v, w)=(u, w)+(v, w).
or (x+y, z)=(x, z)+(y, z) ... (13)
Hence the result (iii).
(iv) We establish this result in several steps.
(a) When at is a positive integer.
Put x=y in (13), (2x, z)=(x, z)+(x, z)
or (2x, z)=2(x, z)
or (2x, y)=2(x, y)
Thus the result (iv) is true for :1=2.
Let it be true for rx.=n (Positive integer), then
(nx, y)=n(x, y)
«n+l) x, y)=(nx+x, y)=(nx, y)+(x, y), by (iii)
=n(x,y)+(x, y)=(n+l) (x, y)
:. (iv) is true for rx.=n+ I if it was true for at=n. Hence (iv)
is true for all positive integer at. .
(b) When at is a negative integer.
Replacing x by -x in ·(2),
4( -x, y)= II -x+y 112_ II -x- Y II I +i U-x+iy 112
-ill-x-iYIl2
= II x-y 1\2_ II x+y 11 2 +i II x-iy II I-i II x+iy 112
HILBERT SPACE 361

Comparing this with (2), 4( -x, y)=-4(x, y)

or (-x, y)= -(x, y) .. ,(14)
Let ot= - p, wherefJ is positive integer.
(cxx, y)=<-~x, y)=-(fJx, y), by (14)
= -fJ(x, y), by (a)=cx(x, y)
(c) When ot is a rational number.
Let IX=~ where p and q are integers and q:;t:.O.
q
/p
(lXX, y)=,q. x, y
, (X) y"/=P "i'
, / Pi'
/=" /x y /, '
by (a) and (b)
or /x
(aX, y)=P" (i' y "/ ,., (15)
x
-=z, then x=qz
q
(qz, y)=q(z, y) by (a) and (b).
or (x, y)=q(z, y) or
1
q /x"
(x, Y)='ii' y / '

Putting this in (15), (ax, y)=~ (x, Y)=IX(X, y).

q
Similarly it can be proved that
(CltX, y)=IX(X, y) when CIt is real.
(d) When IX is a complex number.
Replacing x by Ix in (2),
4(ix, y)= II ix+y iI2-- II ix-y 112+i II ix+iy 112-i II ix-iy 1,1
= / j j211 x-iy 1/ 2 _ / i 1211.x+iy 112+i / ; /2 nx+y h2
-; / i \2 II x-y li2
= II x-iy 1: 2- II x+iy 111+1 II x+y \12_; II x-·y 1/ 2
Comparing this with (2),
4(ix, y)=4i(x, y) or (ix, y)=i(x, y) ,:.(16)
Let IX=a 1+;(1(2 where Cil:1 and (X2 are real.
«(Xx, y)=«CX1+;:J2) x, y)=(CZ1X+icx2 x, y)
=(CXIX, y)+(icx 2x, y)~ by (iii)
=otl(X, y)+Clt2(ix, y), by (c)
=a1(x, y)+iot9(X, y), by (16)
=(Clt1+iClt2) (x, y)=cx(x, y)
Hence (<<x, Y)=IX(X, y) for every scalar CIt.
Hence B is a Hilbert space.
 362 HILBBR'r ~PACE

Theorem 4. An inner product space is a "ormed vector space.

Proof. Let V(K) be inner product space. Let u, v E V and
a E K be arbitrary.
We define II u II=v(u, u).
But (u, u) is defined on V.
:. II u II is defined on V.
(i) (u, u) ~ 0 and (u, u)=O iff u=o.
Consequently, II u II ~ 0 and II u II =0 ifu=O.
(ii) II au ii=/ a! I! u i/.
For I! ou !1 2 =(ou, ou)=a (u, ou)=aii (u, u)
= I a 12 I! u II·.
Taking square root, I! au I! = I a III u II.
(iii) I! u+ v II ~ II u il + I! v II
.: I! u+v I/I=(U+V, u+v)=(u, u+v)+(v, u+v)
=(u, u)+(u, v)+(v, u)+(v, v)
= I! U ~/2+
I! v 112+(u, 1')+(u, v)
= II u //2+
I! v 111+2 real part (u, v)
~ II u //2+
II v //2+2 I (u, v) I
~ I! u 112+
II v 112+211 u /III v I! ,
by Schwarz's inequality
=( I! u I! + II v I! ).
...II u+v II ~ II u II + I! v :1, .: norm of any vector ~ O.
Thus the norm defined on an inner product space satisfies all
the conditions of a normed vector space and hence it is a normed
vector space.
Theorem 5. A closed convex subset C of a Hilbert space H
contains a unique vector of smallest norm.
[Kanpur [Link]. 88, 84, 83; Meerut 70]
Proof. Write a= inf {II x II : x E C}. Then there exists
Lim
sequence (x .. ) E C s.t. n .. oo II x" II =a so that
II x" II -. 0 as II -+ C1J •
Let x"' x'" belong to the sequence (x,,)
. bs
IS a convex su et 0
fH ~ X-..2 + X..
- .E C

Hence, by def. of a, a <; I, x",t Xm II,

this => II X .. +Xm 1/ 2 ~ 4a 2 or - II X .. +Xm UI ~ -4a2•
 HILBERT SPAC~ 363

By parallelogram law,
II x .. +x'" III+I! x"-x,,, 112=2 [11 x" Il2+11 ~'" tll)
or II X"-X,,, li 2 = -Ii X,,+X'" 112+2 [/I x" 112+11 [Link] l ]
~ -4a2 +2 [II Xn 112+11 x'" 112].
Making m, n-«>,
x'" Ij2 ~ - 40 2 + 2 [a 2 +a2 ] = 0 as n 00.
Ii x" -
Lim '
Hence U X"-x,,, 11=0.
m, n-?oo
This => (x,,) is a Cauchy sequence.
Since H is complete and C is closed subset of H. Hence C is
complete so that every Cauchy sequence in C converges at some
point in C so that X,,-?Xo E C, say.
Now II Xo Ii=Illim x" I =lim II Xn I\=a.
Hence Xo is a vector in C with the smallest norm a.
Uniqueness of xO' If possible let xo. Yo E C s.t.
II Xo lI=a, "Yo I =a.
To show that xo=Yo'
C is conveX ::. X,,;Y o E C => a I
~ 1 ~oiYo\ i
=> 4al <II Xo+Yo 112.
By parallelogram law,
II Xo-)'o 1/ 2+ II xo+)'o 1'2= 2 [II Xo 1,8+ II Yo \1 2]=2 [all f al]
or II Xo- Yo 112=4a 2 -11 Xo+ ,1'0 112 ~ 4all -4a2=0
or II xo-Yo 112 ~ O. But I xo-Yo 1\ ~ O.
Hence II Xo- Yo 1\ =0, this => Xo= Yo·
Tbeorem 6. Let M be a closed linear subspace of a Hilbert
space H. Let x 11. M but x E H. Let d be the distance from x to
M. Then there exists unique Yo e M S.t. I x-Yo ll=d.
[Meerut 19 8; Kanpur [Link]. F. 79; Banaras 71]
Proof, By def. of distance
d=d (x, M)=inf{1I x Y I: Y E M}
Then:l sequence (y,,) in M s.t. 1\ x-y" II-?d.
Consider tv. 0 vectors y", y", of sequence (y,,).
. a l'mear su bspace -..... y,,+Y
M IS - 2 -... E M .

By def. of d, d ~ .\Ix- (y"+y,,,)

2 I\
364 HILBERT SPACE

or 4d 2 ~ II 2x-(Yn+ Ym) liS

or 1/ (x- Yn)-l--(X-Ym) 1\2 ~ 4d 2 ... (1)
By parallelogram law,
II (X-Yn)+(X-Ym) 1\2+11 (X-Yn)--(X-Y... ) WI
=2 [1/ X-Yn ./1 2 +11 X-Ym 11 2J
=2 [d2+d21~_tl.s m, n_oo
or II Yn-Y'" 1\2=4d2-112x-(Yn+x",) 1\ as m, [Link]
~ 4d2 -4d2 =0 by (1)
But \I Yn - Ym 112 ~ 0 ¥ n, m.
Hence \I Yn-y ... l1= 0 as m, n-+oo
This => (Yn> is a Cauchy sequence in M.
Being a closed subset of a complete space, M ill complete.
Hence (Yn) converges to some point, say)'o e M
:. lim Yn=Yo
x-Yo 11=11 x-lim Yn 11=\1 \im (x-Yn) \I = lim
\I \I X-Yn /:=d
Hence Yo is a vector in M S.t. II x- Yo lI=d.
Uniqueness of Yo. If possible, let 3 Yo, Yl e M s.t.
II x--Ylll=d, \I X-Yo \I=d. ...(2)
To show that Yl=YO'
[Link], d~llx-YltYOil
or 112x-(Yl+YO) 112 ~ 4d2 • .._(3)
By parallelogram law,
\I (x- Yo)+(x- Yl) 1/2+ II {x-Yo)-(x-yJ III
=2 [Ii X-YtlII2+11 (x-yJ III]
=2 [d 2 +d1 j, by (2)
or \I Yx -Yo 112=4d~-112x-{Yx+Yo) Ipi
~ 4d2 -- 4d 2 =0, by (3)
or II (Yx - Yo) 112 ~ O. -But II )'c Yo II ;;;a O.
Hence II YI-JO 11=0, this => 31=Yo'
Theorem 7. If M is a proper closed subspace of a Hilbert
space H, then 3 zo;t 0 E H S.t. Zo 1. M.
fKanpur [Link]. F. 19.3, 89 ; Meerut 88J
Proof. M is a proper subspace of H
=> :I x E H but x ft M.
Let d be the distance from x to M, then
d=,d{x, M)=inf {II x-Y II: Y E M}.
HILBERT SPACE 365

Since M is closed and so 3 unique vector Yo E M s.t.

II x-Yo lI=d (Refer theorem 6).
Write zo=x-Yo, then x, Yll E H ::;. X-Yo=Zo E H.
Also II Zo lI=d.
Cliam. Zo .1.. M.
Let Y E M and ac be a real number.
Yo. Y EM!:> Yo+acy E Mas M is subspace
=> dl <; " x--(Yo+cxy) Ill, by def. of d
=> dB <; II Zo-cxy 111 as ZII=X-Yo
or (Zo-«Y, Zo-«Y) ~ d2
or (Zo. Zo-GlY) -II (Y. Zo-IXY) ;;;:;: d2
or (Zo. ::0)-1£ (Zo' Y)-IX [(y, Zo)-ct (y, y)] ;;. dB
[ac is real => GC=ii!]
or II z(' IIIl-cx (zo, y)-cx (y, zu)+Cl 2 11 Y hll ~ dB
or n zo 112- ac [(zo, Y)+(Zo, y)]+«211 y il2 ~ dB
or d2 - 01.2 Re (zo, y)+acll II y 112 ;at dB [Re=real part]
or -2« Re (zo, y)+ml! II y 1/ 2 ;at 0 for every at.
Hence this proves that Re (zo. y)=O
Replacing y by iy, 1m (zo. y)=O
Accordingly, (zo, y)=O "rI y E M
:. zo.L M.
Theorem 8. Let W be a non-empty sllbset of a Hilbert space
H. Then Wi is!l closed linear subspace of H. [Meerut 1970]
Proof. By def. Wi={X Ell: (x, y)=O y YEW}.
Let x, y E Wi be arbitrary and two scalars GC, fl.
Also let u E W be arbitrary.
Then (x, u)=O=(y, u)
Then (a.x+{:Jy, u)=rx. (x, u)+~ (y, u)=rx..O+{:J.O=O
:. IXx+{:Jy E Wi, showing thereby Wi is a subspace of H.
Remains to prove that Wi is a closed subset of H. For this
we must show that
Xu is any limiting point of Wi => Xo E Wi
Now Xo is any limiting point of Wi
=> 3 sequence (x.. ) in Wi S.t. lim x,,=xo
YEW, XII E WL => (x,,, y) =0 for every y
366 HILBERT SPACE

lim (x n , y)=O => (lim x n , y)=O

n-+oo
[For inner. product is continuous map]
::> (xu, y)=O. Also yEW
o Xo E Wol => Wol is closed.
Theorem 9. If M is a linear subspace of a Hilbert space H,
show that
M is closed <:> M =MJ.J.. [Kanpur 1987]
Proof. MJ.J.=fx e H: (x, y)=O Y y E Mol)
I. Let M=Molol. To prove M is closed.
By theorem 8, Mol is closed subspace of H so that
(Mol)J·=Molol is again closed subspace
But M=Molol. Hence M is closed.
II. Let Mbe closed. To prove M=Molol.
Mol={x E H: (x, y)=O Y y E M}
and x EM=> (x, y)=O ¥ y E Mol
:::> x E Ml.l.
:. Me Moll.. ...(t'
Now suppose the relation (1) is proper so that M;f:.MolJ..
Now M is a closed proper subspace of a Hilbert space MJ.J..
Hence, 3 a non-zero vector Za E [Link] s.t. zo..L.M.
(Refer Theorem) zo..L.M ~ z" E Mol.
Thus Zo E Mol and Zo E [Link] so that Zo E MJ. n Molol ={O}
This ... zo=O. A contradiction. For zo;f:.O.
Hence M=MJ.J..
Theorem 10. Let S, SI' S2 be sub$eta of a Hilbert space H.
Th~" (i) {O}J.=H. (ii) HJ.=={O}, (iii) sns.1.={O},
(Iv) S. C S. :::> S20l C SI J..
Proof. (i) {O}ol={x e
H: (0, x)=O}=H
For (0, x)=O ~ x e H
(ii) Hl={X E H: (x, y)=O ¥ y E H}
= set of all' tbose vectors of H which are pel pendi-
cular to all vectors of H inclUJ<ling the~selves.
={O}, for 0 is the only vector whIch is self orthogonal.
(iii) Any x E Snsol => XES, x E Sol
.. (x, x)=O .. x=O
(iv) Let SI C S.
Any x E S~ol ::> x ..L. S2 => (x, y)=O ¥ y E S2
HILBERT SPACE 367

• in particular (x, y)=O

~ x J. S1 .. x e S1.l
:. any x e StJ. .. x e S1J.
This S1. C S1.l.
Theorem 11. If M and N are closed linear subspaees of a
Hilbert space H st. M LN, then M +N is also closed.
(Meerut 1987,81; Kaupar [Link]:. F'. 80]
Proof. Let z be a limit point of M + N. If we show that
zeM+N, the result will be proved.
z is a limit point of M +N ~ 3 sequence (Zn) in M +N s.t.
lim Zn=Z
MiN=> MnN={O}. ForSnS.1={O}
• M+N=M(BN .
• a unique elements x"e M, Yn => )v s.t. Z. = x,. +Yn.
z",-zn=(x..+y.,.)-(Xn+Yn)=(x.-X")+Cv,,, y,,)
Also x", --x.. E M, YIlt -Yne N, M J. N
This => (Xm -x..)L(Y.-YII)'
Hence, by Pythagorean theorem,
II (x", -x,,)+(Ym-Yn) 1I1:.~11 x.-x.. I1 I +/l y .. -y.. /II
or "zm--z.. I1I~/! X,,,-x .. HII+l! y .. -y.. /III
(zoo) is a convergent sequence => <z.. ) is a Cauchy sequence
=> 1/ z.,.-z.I/_ 0 as m~oo • • ~oo.
:. Nx", -XII 111+// y.-y.. III _ 0
or II x," -XII 1/=0=// Y",-Yn" as.m, n-+oo.
This ~ (X,.)€ M, (y,,) e N are Cauchy sequences .
• 3 XE M, ye N s.t. lim XII=X, lim YII= Y
=> z=lim z.. =lim (x,,+y,,)=xt-ye M+N
• z€ M+N..
For M and N are closed subspaces of a Hilbert space H, fe.,
complete space. Hence M and N are also complete Hence every
Cauchy sequence in M er N is convergent to some point in
M or N. Also
x€ M, YEN. x+yeM+N.
Theorem 11. If M ;s a closed linear subspace of a Hilbert
space H, then
H-M(f)MJ.. .
[Kaupar [Link]. F. 1984 ; Meerut 82, 87]
368 HILBERT SPACE

Proof. Let M be a closed subspace of a Hilbert space H.

Then ML is also a called subspace of H (Refer Theorem 8). Also
we know that [Link]={O}. [Refer Theorem 10 (iii)].
Hence if we show that H=M+M.L, the result will be proved.
For H=M(f)M.L ~ H=M +M.L, MnMl.={O}.
Evidently M +[Link] H (I). Remains to prove HCM+M1.
suppose not. Then M+M1. is a proper closed subspace of H.
Hence 3 zo,cOe Hs.t. [Link]+M1.. (Refer Theorem 7) so that
zo.f.M and zo.LM1..
This ~ ZoE M1. and zoEMl..L <> ZoE M1.nM1.1.={O}
=> zo=O. A contradiction. For zo#O.
Hence HCM+M1. ... (2). Combining (1) and (2), we get
H=M+M.L.
Problem 2. If S is a non-empty subset of a Hilbert space, then
S1.=S1.1.1.. [Kanpur 1978]
Solution. (i) By def.,
any xe S => x.LS1. => xeS.L.L
or S1.CS1.1. ... (1)
(ii) Replacing S by S1. in (I), we get
S.L C(S.L ).L.L
or S.L CS.L.L.L ... (2)
But ACB => [Link].L, by Theorem 10.
In view of this (I) gives
(Sl..i.).LCS.L
or S.L.L.L C S.L ... (3)
Combining (2) and (3), S1.=S.L.L.L.
Problem 3. If S is a non-empty subset of a Hi/bert space H,
SIIOW that S.i.L is the closure of the set of all linear combinations 0/
the vectors in S, i.e.,
[Meerut 1976]
Solution. We know that SCS.L.L ... (1) (Refer Problem 2).
It means that S.L.L is a closed subspace of a Hilbert space H S.t.
SCS.l.L.
By def. of closure, [S] is the smallest closed subspace of H
containting S. .
HILBERT SPACE 369

It follows that .[S] C SJ..1. ... (1)

We have S C [S1.
But ACB ~ BJ. C A.1. ... (2), by Theorem 10.
:. [S]J. C SJ.. Again applying (2),

SHe [S1 J.J. ... (3)

[S] is a closed [Link] ~ (S1=[S)J..1. ... (4)

(Refer Theorem 9)
By (3) and (4), SHe [SJ ... (5)

Combining (I) and (5), we get [S]=S.1..L.

Theorem 13. (Bessel inequality). Let {Xl> x 2 , ... Xn} be ajinite
orthonormal set in a Hilhert space H. If xE H, then

.E
• ~l
I(x, x,) 12~ /I x II!.
[Kanpur [Link]. F. 1984, 87; Meerut 88]
Proof. By definition of orthogonal set,
t JI if i=j
(x,. xi)=8 '=10 if i,ej
..
V/rite y=x- E (x, Xi) X,. Evidently YEH.
i-I

fI
/I Y 1I 1=(y, Y)=(X- E (X, Xi) Xi, y)
'_1
fI
=(x, y)- E (X, x,) (XI, y)
1=1

=" /
't~l
X, X - ~
fI "
(x, X,) X, / - E.. (X, Xt)
~_1
(XI' y)

=(x, x) - ]: (x,
.. Xi) (X, XI)
(_1

i: (x, x,) XI
- E(X, XI) [ (" XI, X- 1-1
~I
>J
fFor (x, IXv)=a (x, y)]
370 HILBERT SPACE

+ E" (X,
~_1
X,) (X, X,)

"
=11 X 111- ,-I
l: I (X, X.> r'
Since II y Ipi ;;;;t 0 and so \I x iiI - l:" I (x,
~_1
XI) II ;iii 0

or " I (X,
II X 112 ;;;;t l: X,) :- Proved.
. 1-1

Theorem 14~ An orthogonal set of non-zero veclf!rs is linearly

independent. [Kanpur [Link]. P. 1977,75,70, [Link]. 'F: 12J
Proof. Let S={Xl> x z, .... x,,} be an orthogonal set of noo-
zero Vf'ctors in a Hilbert space H.
To prove S is linearly independent.

Syppose l:" a,x, =0 ... (1) where a, are scalars.

I-I

From which, ( i a~. ~,. xr/'=<o, Xr)

'-'-I '

l:" a, <X,. x,)=o

,-1
By a'i~umption, (Xt, x,)=O ¥ i except i =r.
 HILBERT SPACE 371
Hence ar (xr, xr)=O, or a,,=O for r= 1,2, ... , II.
Now (l) declares that S is linearly independent.
Theorem 15. A" orthonormal set of vectors is linearly in-
dependent.
Proof. Let S={x1 • x 2 • •• , x .. } be an orthonormal set in a
a
Hilbert space H. To prove that S is linearly independent .
..
Let E a,x.=O ... (1) where a, are scalars.
:I-I

Then .( ~ a~ x" Xr/" =(0, x,)

,,=1
.
~ 0, (x,. x,)=O
.=1

or ~ a, 8',=0. But ~',=O "I i except i=r

.=1
or 0,8',=0 or ,ar.l:...,O or a,=O.
This => a;=O for i-'I, 2, ... , n.
Now (I) proves that S is linearly independent.
Theorem 16. (Gram Schmidt orthogonali.&ation process). Let
S={U., u." .. , u~} be a linearly independent set (or basis) in a
Hilbert space H. Then there exists orthonormal set of vectors
Sl={Wl> W2 , ... , W,.} S.t. L(S)=L(SI)' [Meerat 19701
Proof. Let S={u1 , ... , u..} be a L.I. set,in a Hilbert space H.
To construct an orthonormal set
SI={Wl> .. " w.. } S.t. L(S)=L(Sl)'
L.I. means 'Linearly independent'.
L.D. means 'Linearly dependent ..
Now we shall construct an orthogonormal set Sl-{K!t, 'wa, ... ,w,.}
in H by the help of elements of S. The main idea behind this
construction is that each w, is a linear combination of vectors
belonging to S.
S is basis .. Sis L.l.
~ u,:;cO for 1 .so;; i .so;; n.
(u 2 , VI) lit·
Set V1=Ul> Va=U.- II Villa •

vlI=O .. U2 is a scalar multiple of VI =Ul

=> {UIl, u1} i6 L.D.
972 HfLBERT SPACE

A contradi ct i on. For {u�. u,} is L.I.

Since every subset of L.I. set is L.I.
Consequently v,:;co.
( Vi' VI >/ (Ul• VI) VI "-
=" U2- IIVI V
lit • , /
"
_ / (UII• UI) III
-,,-u.--U�'
<
> (u2, UI)
U2, Ul - II (Ul• UI)
=
UI lit
=0. For (u .. uI)= 1/ Ul )1'.
Hence V2 is orthogonal to VI and therefore {VI' V2} is ortho·
gonal set.

� (ua• v.) Vc
Set va=ua - '_1
'" -II -
VI IIii
(ua. VI) (ua. V,) V9
Ua
=
- rrv;rra - -I
V3:;i:O. For va=O implies that {UtI U�, U3) is L.D. which is impossi­
ble. For {ul• Ul, ua} is L.I.
/ ' (ua• VI) VI (ua• V2) Vz V "­
Now <Va. VI) =" Ua - I/
II Vt II' - Tv;fI"
< a'VI) - <Ua,Vl)<I'I,vl)_<UI,V2)'(�'2,Vl)
=
U --rv;n2"--- --!I v� li2
= ( Ua• Vt) -(Us. VI) -0. F,or (v2, v,)=O.
=0.
Similarly (va, V2)=0.
Thus we have shown that

(v" vs)=O s.t. i:/:j and I, j= I, 2, 3.

Thus we have constructed an orthogonal set {VI' v� • l
. . .• Vio

of k (where k < n) vectors s.t.

Then (v" VI)-O s.t. i;t;j and i, j= 1,2, . . . , k.

Ie (UH1' Vi) VI .
Set Vk+l=UHl- �
- '-1 II Vi III�
Then Vt+I:;i:O. For Vk+l=O implies that {ul• U2, .. ·, Ilk+1}
L.D. A con+radiction. For {ul .. . . U1:+1} is L.I.
HILBERT SPACE 373
If 1 <j E; k. then
:. <U~+l; v,) VI
(VH1> VJ
) =,/ , Uk+!- 1-1
~
"
II V, 1/1 • VI /
I '

=<U/c+1' V,)- E <UHI' II.) _(VI. VI) •

I - I I / v , III

( ) (UHIt v,) (VI, VI)

= UII+1. V, 1/ VI " • •

[For (v,. ",)=0 s.t. l#j and l,j=l. 2.... , k]

=(Uk+1' VI)-(UII+1> 11/)=0 .

:. (v'+l v,)=O.
Consequently (VI. v/)=O s.t. l::j:j and 1<;1. j~k+ I.
Our aim is now completed by induction. Thus we have cons..
tructed an orthogonal set {VI' vz•..• , v.. } of n vectors.
Put

Then {WI. w2 •• •• , w..} is an orthogonal set in which every

vector is unit vector. meaning thereby Sl:=::twi. • w.. } is orthoe
normal S&;t of vectors in V\f'). Since an orthonormal set is L.I.
Hence Sl is L.I.
By the above. expressio~.
L[{u l .••• u}]=L[{w l • W ••.••• wr }].
Hence by induction. the required construction is obtained.
Problems related to operators.
Theorem 17. Every non-zero Hilbert space contains a complete
orthonormal set.
Proof. Let H be a non-zerQ Hilbert space. Then
3 x E: H s.t. *~O.
Normalize the vector x by wtitiDg Xl . II ~". Then {Xl} is
clearly an orthonormal set in H.
Thus every non-zero Hilbert space contains orthonormal
set. Consider the collection or all possib!e orthonormal sets in
H. By Zorn's lemma. this collection hilS a maximal number M.
We assert that M is complete. Let us suppose y.:;;e.O s.t. y4..M. Set
y
Yl=lfYil'
 374 HILBERl' SPACE

Then MU {Yt} is also an orthonormal set s.t.

M U {yil ::l M;
contradicting the maximaJity of M as M is complete
Finally Y..LM ~ y=o
:. M is complete orthonormal set.
Theorem 18. Let H be a Hilbert space. and leI {eel be an or-
thonormal set in H. Then the following conditions are all equivalent
to one another
(1) {e,} is complete
(2) x..L{e,} ~ x=o
(3) if x is, an arbitrary v(!ctor in H, then
x=E (x, e,) e,
I
(4) if x is an arbitrary vector in H, then
,
n x III=E I (x, e,) 12.(Kanpur 1990,8!; Meerut 82, 79]
'

Solution. (1) ~ (1). It is given that the orthonormal set {e,1

is compl~te. Suppose that xJ. {e,}. To prove that x=O, Suppose
not. Then' ¢O. Write e=" ~ II' Then e is a unit vector s.t.
e..L{e,} so that {ec, e} is ~mhonormal set which properly contains
{e,l. It means that {e,l is not complete orthonormal set. A con-
tradiction. Hence x=O
Finally x..L{e.} ~ x=O.
(2). (3). It is given that
x..L,e} => x=O.
Let x E H be arbitrary • To prove that
x::.£E <x, e,).
I

Take Y7x-E (x, e,) e,. Then

I '

(y, el)=(x-E (x, e,) e" e,)

I

-(x, el)-E (x, e,) (~" el)

I
=(x, e/)-(x" e,> (eh el>. For
I, i=j
(e" el)= { 0, i=t=j
-(x, e,)-(x, el)=O
HILBERT SPACE 375
This .. y .i el .. Y .1 {e,l ... y=O, by assumption.
01> x-:E (x, e,) e,=O
I

o x=:E (x. e.) e,.

I

(3) 0 (4). It is given that

x=:E (x, e,) e,.
I
To prove that
II x II I=:E I (x. e,) I •
By as.)umption,
II x II 2=<X. x)=(:E (x, e,) elt x)
I

=E (x, l',) (eh x)=E (x, e.) (x. l',)

I I

=E I (x, e.) Ii as z'z ..; I i 1'1.

(4) => (1). It is given that
II x II '=E I (x, l',) II
I
II x e: H.
To prove that {e,} is complete.
Suppose not. Then {e,}. is not complete so that it is a proper
subset of ali or'tho~ormal set {e, e}. Consequently eJ. {e.} 80 that
(e. e,)=O II i.
By assumption.
II en ·=E I (e. e,) I ·=0
I
or or
nell =0. 1=0.
This contradicts the fact that e is a [Link] vector. Hence {e,} is
complete.
Problem. Let E={e,} be an orthogonal set in a Hilbert space
H. Show that B is compete iff
E I (x. e,) I 1= II x UI [Meerct 1987]
lEI .
Solution I. Here write (4)·... (I) of the above Theorem 18.
II. Given {e,} is complete.
Aim. II x II ·=E I (x. e,) II
I

Let y=x-E (x, e,) e, ... (1)

I
376 HIL~ERT SPACE

Tllen prove as in above Theorem 18 that (y, e1)=O.

This .. y J.e1
But {e.} is complete.
Consequently y=O. Now (1) gives
x=X (x, e,) e,
... II x 1I 1
•
=(x, x)=(}; (x, e,) ej, x)
i
I

,
==X (x, e,) (e,. x) .

=X (x, e,) (x, e.)=E I (x, 1',) II.

I I

• Problem 4. Prove that Hilbert space is finite dimensional

* every complete orthonormal set is basis.
Solution. Let H be a Hilbert space over a field K of scalars.
Let S={e,} be a complete, orthonormal set in H.
Then S is linearly independent . .. (1)
(Refer 'theorem 15)
For any x € H, we have
x=X (l(, ec) e" , ... (2)

by Theorem 18.
Case I. Let dim H=n=finite
To prove that S is basis
(2) .. S generates H. . ... (3)
By (1) and (3), it follows that S is a basis, dim H=n also
proves that number of vectors in S must be n.
Case 11. Let S be a'basis.
To prove that dim J/=finite.
We claim that S is a finite set.
Suppose S is infinite. Then 3 A caSu. A ill denumerable.
Take A-{e1, ea,··., eft .... }.
QO e GO
Consider that series X
It-I
-i-
n
1: Uft.
_-1
Then
1
i IItl I == nl as ! eft I = 1 ¥ n.
All<> X ~ is a convergent series (with p-2 > I).
HILBERT SPACE 317

Hence, by Weierstrass M-test, the series X e~ is uniformly

n
convergent and so converges to some 'vector x E H so that

... (4)

AI~ S is basis for H.

Hence we can write
x=aAeA+ ... +a~e~. ... (5)
Letj be a positive integer having value different from the
values of indices A, ... , 1-'.
By (5), (x, e,)=a). (eA> e,) +...
+a,. (e,., e,)
=a~.O+ ... +a,..O, by orthogonal properlY
or (x, eJ)=O .. (6)

By (4), (x. e,)=( E e:, eJ '/

, .. =1 n

t
or (x , eJ)=-
P . ... (7)

By (6) and (7), -:z1 =-=0' or ~=o which is not possible.

J J
Hence S is finite. Also S is given to be basis. This .. dim
H=finite.
Remark. The above problem proves that: an infinite ortho-
normal set in a Hilbert space cannot be a basis.
Problem 5. Prove thaf any two complete orthonc rmal sets in a
Hilbert spare H halle the same cardinal number.
~olution. Let S] and S2 be 1\\ 0 complete orthonormal seta
in a Hilbert space H.
Case I. Let S}={el' e2 • ... , e .. } be finite.
S} is complete orthonormal set
.. S} is basis of H. (Refer Problem 4)
• di:n H=n
S2 is complete orthonormal se't in l-J
~ S2 is basis of H
378 HILBERT SPACE

But dim H=11 (already proved)

=> S, has n vectors.
Thus SI and S2 both have 11 vectors and so their cardinal
number is the same.
Case n. When SI is infinite
Let x E S2 be arbitrary and
S2(X)=(Y : y E S2 and (y, x)=;eO} ..• (1)
Then S2(X) C S2 and S.(x) is countable.
Let z E S9 be arbitrary, then by'Parseval's identity, we have
II z :1'= E I (z. x) II
XESt
as SI is complete orthonormal set.
But z E S2 • z is unit vector => II z 11=1
:. )= E I (z, x) II
xeSI
This => iI x E SI S t. (z, x)=;eO
=> z E S2(X) for some x E S10 by (I)
Finally.
This suggest that
S2= U St(X) ... (2)
xeSI
Let card (SI) = 11 1 , card (S2) =n 2•
(2) => card S2=card [ U S2(X)] <; card SI
XESI
=> n 2 ~ n l, ... (3)
interchanging the roles of SI and S2' we get
"1
E;; n2• . .,(4)
By (3) and (4), we fet "1="2 .
ot ca'rd Sls=cud S2'
°
Problem 6. Jf ani I be :!ero and identity operator3 on a
Hilbert'space H. Ihen 0*-'0, J*=I.
Hence show that if T is non-singular ope;'alor on H, then T* is
also non-sil1gular and in this case (T*)-l=(T-l)*.
Solution. Let x, y E H. By def., O(X)FO ¥ x and
{(x)=x "<I x.
I. (O(X), y)=(O, y)=O ... (1 )
(O*(x), y)=(x, O(y»=(x, 0)=0 ••. (2)
BY' (l) and (2),
(O(x), y)=(O*(x), y)
HILBERT SPACE 379

Therefore from the uniqueness of adjoint operator, we have

O(x)=O*(x) or 0=0*.
II. (/(x), y)=(x, y) . .. (3)
(f*(x), y)=(x, I(y» = (x, y) . .. (4)
By (3) and (4),
(/(x), y)=(l*(x), y).
This => I(x)=l~(x) => 1=1*.
Uf. tet T be non-singular so that T is invertible,. then
TT-l=T-IT=f.
This =- (,rr- 1 )*=(T-IT)*=:I*
=> (T-l)* T*=T*(T-l)*=I. " (5)
For (AB).::B*A*
=> (T*)-I=(T-l)*
For B*A*'=A'li:B~=1 => (A*r 1 =B;'t.
(5) proves that T* is invertible and hence T* is non-singular.
Theorem 19. The sc(f adjoint operators in B(H) form a closed
real linear subspace of B(H) and therefore a rcal Banach space
which contailts the identity trallsformation.
[Kaopur 1985, 80 ; Meerut 76 j
Proof. Let B(H) denote the set of all operators on H, then
B(B) is a Banach space. Let S be the set of self adjoint opera-
tors on H. Then
S={T E B(H) : T*=T}. . .. (1)
Let ex and f3 be real numbers. then
ex=iX, S=~. . .. (2)
Let At. A~E S be arbitrary, then
Al*=Al' A!*=A2' ... (3)
(exAI +f3A2)*= iiAl *+ "fJ,42 *
=«A 1 +fJA 2 • by (2) and (3)
This => 02A t + (3A z E S, by (1)
Thus At. A2 e: S and ex, fJ are real numbers
=> exAt +IJA2 e: S.
This proves that S is real linear subspace of B(B).
Remains to prove that S is closed.
Let (A .. ) be sequence of operators in S s.t.
A..-A, then A,,*=A...
380 . HILBERT SPACE

To show that A is self adjoint

II A -A* n = II A-A .. +A.. -A ..*+A ..*-A* II
<II A.--A.. II + II A .. -A,,* II + II A,,*--A* II
Using the fact that A,,=A,,* and II Til == II -T II we get
II A-A* II E; !I A-An II+"
:::::211 A A.. II.
(A-A .. )* II
For 11 T* II = II Til ¥ T.
Making n~ 00 and notin!! that lim A~=A, we get
II A -A'* II =0 or A=A* or A E S.
Thus A,,--+A E S.
This '0 S is closed.
Thus S is a closed and real linear subspace of a Banaeh
space B(H).
Consequently S is itself a real Banach space. Also 1*=1,
this ~ IE S.
:. S contains identity transformation.
Problem 7. If Al and A2 are self adjoint operators on H, then
their product Al 4~ is self adjoint <¢' A1A2=A2.41
* A1 and A2 commute.
[Bundelkhand 1981]
Solution. Let A1 and A~ be self adjoint operators on a Hil-
bert spac~ H. then Al"'=Att A2*=A~.
A!l\O (Al.4~)*=A/ A1*=A2A1=AtA2
iff A1.42=.42.41
(A 1 4 2 )*=A1 A 2 iff A1 A 2 ==.4 2 A 1 •
This proves that
A1A2 is self adjoint <: AI and .42 commute.
Theorem 20. The real Banach space of all self adjoint opera-
tors on a Hilbert space H is a partiolly ordered set whose linear
structure and order structure ore related by the following proper-
ties:
(0) ff A) ~ A2 Ihe'l A 1+.4 <; .4 2 +.4 ¥ A
(b) A] ~ Ai and IX ~ O. then tlAl ~ IIA 2 • [Meerut 1976]
Proof. Suppose that B is real Banach space of all self adjoint
operators on a Hilbert space H. Let TJ> T2 .1 T3 E B and x E H
be arbitrary. We define '
T. ~ T2 <=> (T}x, x) E; <T~x, x) ... (1)
Then relation ~ on B is
(i) Reflexive. Tl ~ T.
HILBERT SPACE 381

For (TIX, x)=(T1x, x) $ T 1= Tt

~ TJ ~ T 1•
(ii) Symmetric. Tl <;; TI , T'I. ~ T 1 .,. 1"1 = Ts
For Tl <;; T2 • T9 ~ Tl ~ (TJx, x) <;; (T2X. x)
and (TaX. x) EO;. (TIX, x)
.,. (TIX, x)=(T2x, x)
=> T 1 =T2•
(iii) Transitive. Tl C;; T~. T2 ~ Ta => Tl ~ Ta
For Tl <; T2 , T2 < T. ~ (Tlx, x) ~ (T2X, x)
and (TaX. x) < (TaX, x)
$ (Tlx, x) <;; (T2X, x) <;; (TaX, x)
~ (T1x, x) <;; (T"dx, x)
=> Tl ~ T3
The above facts prove that (B. ~) is a partially ordered set.
Second part. (i) Tl < T2 => 1"1 + T ~ Ta +T
For T1 ~ T2 ::> (T1X, x) ~ (TaX, x)
~ (T1X, x")+(Tx, x) .;;; (T2x, x)+(Tx, x)
~ «T1 + T) (x) x)_ .;;; «T 2 + T) (x), x)
~ Tl + T ~ T 2 + T, by (I)
(ji) Tl ~ T2 and at ~O ~ otTl < otT2
For Tl ~ T2 ~ (TIX, x> ~ (T\!x, x>
~ 0: (Tlx, x) ~ ot (TaX, x)
o «a 1'1) (X), X) ~ ««T2) (.x), oX) Y. a ~ 0
=> «01: TI ) (X), X) ~ «OtT2 ) (X). X)
=> OtTl ~ acT2 \I IX ~ 0, by (1).
This completes the proof.
Theorem 21. If A is positive operator on H, then I+A is non-
singular. In particular, I+T* T and 1+ TT* are non-singular for
arbi:rary operator T on H. [Meerut 1987]
Proor. Let A be positive operator on a Hilbert space, then
A=A* ... (1)
and (Ax, x) ~ 0 ... (2)
Let x,y€ H.
(i) To show that
(/ +A) (x)=O => X=O ¥ X € H
(I+A) (x)=O ::> 1{;"(;)+Ax=O ~ x+Ax=O
• Ax=-x ~ (Ax, x)=< -x, x)= - II x /II
=> (AX, x)= -/I x /II.,. -II X 112 ~ 0, by (2)
.. IIxdl~O
 382 HILBERT SPACE

But II X 112 ;;;;;e 0 is always true

:. "x /12=0 or II x II =0 or x=O
(ii) To show that I+A is one-one.
(I +A) (x)=(I+A) (y) => (I+A) (x- y)=O
=> x-y=O, by (i)
=> x=y
:. I+A is one-one.
(iiI) To show that I+A is onto.
Let M be the range space of 1+ A. Then
M={(l+A) (x) : x e HI ... (3)
Now we claim that if is closed.
" (I+A) (x) 11:r= III (x)+A (x) 112= II x+.4 (x) II!
=(x+Ax, x+Ax)
=(x, x+Ax)+(Ax, x+Ax)
=[(x, x)+(x, Ax)]+[(Ax, x)+(Ax. Ax)]
= " x1l2 + II A¥ W'+(Ax, x)+<x, Ax)
= II x 112 + II Ax II 2+2 (Ax, x)
[For (x, Ax)=(A'*x, x)=(Ax, x)
as A*=A]
= II x 112+ II Ax 11 2 +2 (Ax, x) ;;;;;e II x II a
[For (Ax, x) ;;;;;e 0, by (2)]
or II (l+A) (x) /1 2
I! x " 2<
or " x II ~ " (1+A) (x) II ... (4)
Let «(1+ T) (x .. » be a Cauchy sequence in M. Let m, n be
any two positive integers
II Xm -x.. " < II (1+14) (xm--X,.) "by (4)
= /I (1+A) (x m ) -(l+A) (x.. ) II _I) as m, n .. ex>
This follows from the definition of Cauchy sequence.
This => (x .. ) is a Cauchy sequence in H.
But H is complete and so 3 x E H S.t.
lim x .. =x and so (I+A) (X) E M
lim (I+A) (x .. )=(i+A) (lim x .. )=(1+A) (x) E M, by (3)
Thi~ => (1+A) (xn)-(l+A) (X) E M.
Thus the Cauchy sequence «/+A) (x n » in M converges to
(J+A) (x) e M. Consequently M is complete.
But every complete subspace of a complete space is closed.
This means that M is closed. Next we claim that M=H. Suppose
not. Then M#-H.
HILBERT SPACE 383
It means that M is a proper closed suhspace of H. Then EI
non·z:!ro vector x .. E H S.t. Xo.l.M. (Refer Theorem 7)
Xo E H ~(I+A) (xo) E M, by (3)
Also [Link].
These two statements prove that
Xo 1 (l +A) (x o), this =>
(U+A) (xo), xo>=O
or (xo+Axo, xo)=O
or (x6 • xo>+(Axo, xo)=O
or II x 1j2+(Axo• xo)=O .. (5)
Also II Xo liz ~ 0 and (Axo. xo> ~ ()', by (2).
Now (5) => II Xo iI =O=<Axo x o)
=> xo=O, in particular.
A contradiction.
M=H.
ronsequently T+A is onto map.
From what has been done it follows that 1+.4 is one-one
and onto map. [Meerut 1987J
This ~ l+A is isomorphism
=> I+A is,non singql!J.r..
Second Part. Let T be an arbitrary operator. on H. Then
TP and T*T both are positive operators. :. It follows from
the first part that 1+ TT* and I+- T* T both are non-singular maps.
Theorem 22. An operator Ton H is unitary ¢> it is an iso-
metric isomorphism of I{onto itself. [Meerut 1974]
Pj\)~f. Let T be an operator on a Hilbert space H. Let
x,yE H.
I. Let T be unitary.
Aim. T is an isometric isomorphism.
T is unitary => TT*=T*T=l
=> T is one-one onto S.t. f*= T-l
=> T is isomorphism.
(Tx, Ty)=(x, (f* F) (y»=(x, I(y»)=(x, y)
Taking x=y, (Tx, Tx)=(x, x)
or I Tx 112=11 Xii"
or 1\' !x 11 =:1 xU
This proves that T is isc~etric.
II. Let T be isometric and iSOl~osphism.
384 HILBERT SPACE

To prove that T is unitary.

T is isomorphism => T is one-one and ORtO.
=:. 1'-1 exists.
Tis isometric =- 1I Tx '1=1\ x /I ~ II Tx III=I! X 112
=> (Tx, Tx)=(x, XI
=> «T-*TI (x), x)=(l(x), x)
0:> «T:tT-I) (x), x)=O
.:> T*'T-/=O => T*T=1
=> T is unitary.
Remark. The following problem can also be proved with the
help of the above theorem.
If T is an operator on a Hilbert space H, then the following
are equivalent to one another.
(i) T*T=I
(ii) (Tx, Ty)=(x, y)
iii) 1\ Tx II = II x II.
Problems 8. If H is a finite dimensio'lal Hilbert space. show
that every isometric isomorphism of H into itself is unitary.
Solution. H is finite dimensioal space => T is onto map.
N ow T: H - onto • h'
--+ H is isometric Isomorp Ism.
Then T is one-one onto so that 11 exists. This follows from
the fact that T is isomorphism. .
T is isometric ~ II Tx 1l=11 x II ~ II Tx 111=/1 X 11 2
00> (Tx, Tx)=(x, x)

=> «T*T) (x), x)=(/(x), x)

::. «T*T-I) (x), x)=O
=> T*T-l=O ~ T*T=I
Similarly we can prove that T*'T=I.
Thus T*T=TT*=I, showing thereby T is unitary.
Problem 9. Show that an operator T on a Hi/bert shace H is
unitary
.;:> (T(ecH is a complete orthonormal set whenever {e,l is.

[Meerut 1986]
Solution. Let lei} be a complete orthonormal set in a Hilbert
space Hand T be an operator on H. Then
, {I
<et, e'>=~H= 0 for i::j:j
for i=j

I. Let The unitary, then TT*==T*T=I.

 HILBI!RT SPACE 385

Aim. {Tef} is complete orthonormal set

(Tx, Ty)=<x, (T*T) (y»=(x, I (y»=(x, y)
This proves that
(Te" Te;)=(e" e,)=Il"
This ~ {T(e,)} is orthonormal set in H.
Remains to prove that {Te.} is complete
xl {Tei} 0> (x, Te.)=O ~ (T* x, ef)=O
~ T*x 1. ret} and {e,l is complete.
~ T*x=O (Refer Theorem 18)
~ TT*x=TO=O ~ J(x)=O ~ x=O.
Now x 1. {Ted ~ x=O.
This proves that {Tei} is complete orthonormal set. (Refer
Theorem 18).
II. Let {Tee} be complete orthonormal set in a Hilbert space
H.
To prove that T is unitary.
First we claim that T is isometry i.e .•
II Tx /i=/i x I! '" x E H.
But II T(O) 11=/1 0 1/
Hence II Tx 11=11 x I) for x=o. . .. (1 )

Let x~O, then { M; II f is an orthonormal set in H.

By assumption,
{eel is orthonormal ~ {T(e,)} is orthonormal.

Hence { T ( 11; II )} is orthonormal set

or J T(X)}
lliXiI .
IS orthonormal set.

This ~ \ l~~~i \!= 1 .. II T(x) iI=1I xII

~ 11 T(x) I!= 1/ x If for x:;t:O ... (2)
By (1) and (2), 11 T(x) 11=11 x II Itf x ... (3)
.'. T preserves norms or, T is isometric. Let M be the
range of the map T : H-+H, then T(H)= M, then M is a subspace
of H. We claim M is closed.
Let y be the limit point of M, then 3 a sequence T(x.. in M < »
386 HILBERT SPACE

s.t. T(xn)-+y so that (T(x,,» is convergent. Consequently (T(xn»

is a Cauchy sequence. . .. (4)
By virtue of (3),
II Xn -x", 112= II T(xn -XIII) 112
=/1 T(x,,)-T(x ...) W! as Tis a linear.
-0 as m, n-. 00, by virtue of (4)
II X,,-X m II -+0 as m, n-+oo.
This =- (x,,) in II is a Cauchy sequence.
But H is complete. Hence (x,,) converges to a point x E H.
Hence T(x) E M.
Now y==' lim T(x,,)= T (Um X,.) all T is continuous
n~co

=T(x) EM
or y EM
y is a limit point of M. y E M.
:.
This proves that M is a closed subspace of a complete space
H. As a result of which M is complete. Now we shall show that
T(H)=M=H.
Suppose not. Then M #-H so that M is a proper closed sub-
space of H.
By theorem 7. 3 non-zero vector Yo e H S.t. Yo 1. M.
Now Yo -L M, M is complete. J'0=0 (Refer Th. 18)
A contradiction.
:. M-H.
This. -T: H-+ H is onto ~p.
Also T is isometric & T is one-one. by virtue of (3).
:. T is isometric and isomorphism of II onto H. Hence T
is unitary. (Refer Theorem 22).
Theorem 23. Riesz representltiea tlaeerem for c,ontinuous
linear junctionais on a Hilbert space. Let H be a Hilbert Space
and let / E H* be arbitrary. Then there exists a unique ve:-tor
J' E H s.I./=/". i.e./(x)=(x, y) Y x E H.
[Kanpur [Link]. [Link]', 13; .....y Measure Theory 66;
Meerut 88, "'J
Pr&of. Firstly we shall show that if such y exists, i.e.,
f(x)-(x, y) ... (1)
then :t is unique. If possible let y be not unique and so 3 y' with
the property that f(x) = (x, y'). Also ftx)==(.~, y). Then
HILBERT SPACE 387

(x, y)=(x, y') or (x, y-y')=O V- x E H.

Taking x=y-y', then (y-y', y-y')=O.
This => y-y'=O => y=y', showing thereby y is unique.
Step II. To prove the existence of y occurring in the equation
I(x)=(x, y).
Ifl=0 (zero functional), then choosing y=O, theorem imme-
diately follows. So we suppose
1#0. Write M={x E H :/(x)=O}.
Then M is a null space of/so that M is a closed subspace of H.
Consequently 3 yo,t:O E H s.t. Yo .1 M (Refer Theorem 7). Then
we shall show that y=rxy.. will serve our purpose for suitable value
of scalar ex.
Yo 1. M => (x, Yo)=O ¥ x E M ... (2)
Now if x E M, then
(x, y)=(x, cY.)=« (x, Yo>
=i (i}-I, , ,", (l~. ~
Also I(x) =0, by def. of M
:. l(x)=O=(x, y> or f(x)~(x, y)
so that (I) is satisfied ¥ x E M and ¥ II.
To find the condition for (I) for X=Yo,
l(yo)=(Yo, ~Yo)=« (Yo, yo)=li II Yo 1/ 1
- f(yo) f(yo)
01
«=il1o lit or IX=II Yo ,11 ... (3}
Thus the vector
/(Yo)
y=CltYo= 11-'-,I'Yu ... (4}
Yo !,
will satisfy (I)
¥ X E Mand X=y•.
Call this fact by the name (*).
Step III. To prove/(x}=(x, y) "f x, Y E H.
Any x E H can be expressed in the form x=m+lly, for
SOme scalar 11 and vector m E M.
:. /(x)=/(m+~yo)-/(m}+Il/(Y.)
as / is linear
=(m. y) +~ (Yo, y), by (*).
=(m+.Byu. y)=(x, y)
when y is given by (4).
388 HILBERT SPACE

, _f(Yof
FinaHy )0-11 Yo 11~'Yo
is a unique vector in H satisfying the condition
flx)=(x y)
Hence the theorem.
Theorem 24. Let H be a Hi/bert space. Let T* be the adjoint
of operator Ton Hand QI a sell/or. Also let S he an operator on H.
Then
(i) (T+S)"*=T*+S*, (it) {rtT)*=dT*,
(iii) (TS)*~-=S Til: (iv) (T*)*=1'.
[Meerut 1988; Kaopur M Sc. F. LiBear Algebra 76]
Proof. Let u, v E H be arbitrary.
(i) «T+S) (U), v)=(T(u)+S(u). v)
=(T(u), v>+ (SlU), v)",--(u, T*(v»)+(u, S*(v)/
=(u, (T*) (V)+S*(v»=(u. (T*+S*) (v».
But «T+S) (u), (v»=(u, (T+ S)iI':(v»
:. (u, (T+S)*(v» ~ (u, (TJi<+S'At) (v».
Uniqueness of adjoint implies
(T+S)*=T·+S· lGuru Nanall. 82; Meerut 72]
(ti) «aT) (u), v)=(aT(u). v)=a (T(u), v)
=a (u, T-~(v»=<u, aT*(v»
=(u. (aT*) (v».
But «aT) (u), (v»=(u, (aT)* (v»
• (u, (aT·) (v»=(u, (aT)* tv»
Uniqueness of adjoint ~
aT·=(aT)* [Meerut 70, 71]
(iii) «TS) (u), v)=(T(S(u», v)=(S(u), T*(v»
=(u. S*T* (v»=(u, (S*T*) (v»
Also «TS) (u), v)=(u, (TS)* (l'»
(u, (TS)* (v» =(u, (S*T·) (v»
This => (TS)*=S*T*. [Kaopur 1984; Meerut 72]
(iv) ': (T*(u), v)=(u, T(v»
Also (T*(u), v) =(u, (T·)* (v»
.: (u, T(v) ;(u, (T·)* (v».
This ~ T=(T*)*. [Meerut 1970]
Theorem 25. Let T be a linear operator on a Hilbert space H.
Then (I) IIT!/-= II T* 1\
(il) II T 112=11 Tr* II.
[Kanpor 84; Gurullaoak 82: Meerut 88, 87J
 HILBERT 5:PACE 389
Proof. Let x, y € /I be 3rbitrary.
(i) II 111= <T*y, T*y> = <TT*y, y> as T**=T
T*y
II TT*y II· I y II, by Schwarz's lemma.
~
~ II Til·!i T*r II· ily II
Dividing by !I T*y Ii, we get
/I T*y 1\ ~ II T \1-/1 yll ~ \I T II if II y !I < 1
or :1 T*y 1/ <; II T I ¥ Y s.t. II y /I ~ 1
sup {IIT*y II: II y II ~ I} ~ sup II TII=II Til
or UT*lIoE;IITIi ... (1)
Replacing T by T*,
II T \I ~ II r*'II· ... (2)
Combining (1) and (2), II T 11=11 T* II· ...(3)
(ii) Since 11 ~T II ~ !I S iI·1/ T /I.
Hence II IT* II ~ II Til II T* 1/= I T Ipl, by (3)
or II IT* II ~ II T III •.• (4)
II Tx IPI=<Tx, 'Tx > = <T*Tx, x>
..; Ii T*Tx 11·11 x \I, by Schwarz's lemma.
" II T*T II . II x II . II x II
= II T*T II lI:x II 2
J ~ 1/ T*T II if II x II <; 1
or II Tx II J CO;;; II T--T II if II x II ~ I.
Taking supremum of both sides and noting that
sup \I T*TII = nT*TII ,we get
sup { II Tx II 2: II x II Et: I} ~ II T*TII
or [sup {II Tx II: II x II Et; 1}]2 '" II T*TI~
or II T II 2 ~ II T*T II ... (5)
Combining (4) and (5), we get
II T 111= II T*T H•
Similar Problem. Prove that the mal' T-+ T* is a norm pre-
serving mapping oj 8(N} into 8(N·) [Xaopar 1983]
Solution. Here prove that II Til=- \I T* II •
Theorem 26. 1/ NI> N:. are normal operators on a Hilbert space
H with the property that either commutes with the adjoint 0/ the
other, then NI+N. line! NI NI are normaloperatf!rs.
[Meerut 1979; BUndelkband 83;
(Kanpur 199,o,Kaopur M Sc. F. Linear Algebra 77]
Proof. By assumption NI N 1*=NI *N). N2* N2=Na NI* ... (1)
(N} N2*=N2* Ni> N I • N2=N2 N I "') ... (2),
(i) To prove that N I +\N2 is normal
390 HILBERT SPACE

(N1+Nz) (Nt +N2 )*=(Nt +N2 ) (Nt*+N2*')

=NtNI*+NtNz*+N2NI*+NzNz* ... (3)
Also (N1+N2 )* (N1 +N2 )=(NI*+Nz*) (NI+NJ
=~*~+~*~+~*~+~*~
=NI Nt*+NI * N 2 +N2* Nt+Nz N z*, by (I)
=N1 Nt*+N. Nt*+NI N z*+N2 N2*' by (2)
-(N1 +N2) (NI +N2 )*, by (3)
This. 1I1 +N2 is normal.
(ii) To prove Nt Nz is normal.
(NI N,) (Nt Nz)*=(NI N z) (Nz* NI*)=Nt (Nz N'J.*) NI*
-NI (N2* N z) N t *, by (!)
=(NI N 2 *) (N2 Nl*)= (N2* Nt) (Nt * N z), by (2)
=N2* (Nt N I · ) N2=N2* (NI* N I ) N 2 • by (I)
=(N2* N I *) (NI N.)=(Nt N~)· (NI N~)
:. (NI N2 ) (NI Nz)*=(Nl N,,)· (Nl N2 ) •
. This => Nt Nz is normal.
Problem 27. Let T. be a linear operator on complex Hilbert
Ipace V. .
Then T=O ~
<T(at), «>-0 Y II € V.
[BundelkhluMll983; Kanpur [Link]. Final SO, 9",86]
Proof. Let Tbe a In1ear operator on a complex Hilbert space
V(F) and let II e: V be ,arbitrary.
(i) Suppose T=O.
To prove <T(Cl),II>=O
<T(Cl), «>=<0"(<<), «>=<0, «>=0.
(Ii) Suppose <T(<<), «>=0 Y gr.. ...(1)
. To prove T~O.
(1) => (T(II+~), rt+~=O. {3 E v.
=> (T(cx)+T({3), «+IJ)=O
=> (T(<<), «+f3>+<T(~), «+f3)=O
Q (T(<<), «)+(T(o.), f3>+(T([3), Cl)+(T(~}, f3>-0
=> O+(T(<<), ~>+(T(fSkll)+O=O. by (I)
~ (T(<<). f3)+ (T(P), Cl)=O. . ... (2)
This is true V lit, f3 E V.
Replacing ~ by i{3 in (2)
(T(ll), if3)-t <T(i~), 11)=0
HILBERT SPAC~
191

or 7 (T(Il), f3)+i (T(8), at)=O

or -i (T(ac), /3)+i (T(P), at)=O
or -(T(<<), ~)+(T(IJ), P)-O For I~O
or (T(~), P) -·(T(~), at)=O. : .. (3)
(2)+(3) gives (T(a), p)=O ¥ II, IJ E V.
Taking ~= T(at), we get
(T(Il), T('1)=O ¥ at e V.
This ~ T(CIt)=O y. a. e V [For (u, u) = 0:0- u=O]

=:> T=fl.
Remark In the above theorem, we have used complex num-
bers to prove the result [Link] hence this theorem may fail to be
true for real inner product space.
Theorem 28. Let T be a symmetric (self adjoint) linear trans-
formation on Hilbert V(F). Show that
(T(<<) ,«)=0 ~ T=O. [lanpur [Link] F.1980J
Proof. Let T be a symmetric linear transformation on Hilbert
space V(F) so that T=T*. Let at, P e V be arbitrary.
(i) Let T=f).
To prove (T(Il), Gr.)=0.
Now (T(at), Gr.)=(O(at), 11)=(0, at)==O.
(li) Let (T(at), 4) -0. ... ( I)
To prove T=O
(T( a.+ P), «+/I)-(T(4) + T(/I), at+/I)
=(T(at), «+/I>+(T(/I), «+~)
+
=(T(at), at)+(T(at), /I> (T(I1) , II)
+(T(~), P),
By virtue of (I). this becomes
O=O+<T(at), /I> + <T(I1), at>+O
or <T(<<), 11> +<T(~), rx>=O
or < T(at) , (J>+<~, T*(ct»=O
or < Tt«), /3>+</J, T(ct) > =0 ... (2) For T*=T
Ca.,e I. When V(F) is a real inner product space so that
F=R.
In this case (at, fj)=(fJ. at)=(fJ, at) for (fj, at) E R.
i.e. <II, /I>=<~, at>.
 392 HILBERT SPACE

Using this in (2),

<T(<<), p>+<T(<<), {J>=O
or 2<T(<<), fJ> =0 or <T(IX), ~>-O
or <T(<<), T(IX) > =0, on taking {J=T(rt.).
This => T(IX)=O ¥ IX E V
A

o T=O.
Case II. When V(F) is a complex Hilbert space then F=C.
Replacing fJ by ;{J in (2), we ¥et
<T(<<), i~>+<ifJ, T(m) > =0.
This => 7<T(<<),.(:J>+i <~, T(Ot) > =0
=> i[-<T(re), {J>+<P, T(<<»]=O
=> -<T(<<), !J>+<{J, T(<<) > =0
=> <T(<<), fJ>-<<<, Tlex»=O. . .. (3)
Adding (2) and (3),
<T(IX), fJ> =0 V (j E V.
=> <T(<<), T(<<) > =0, on taking {J=T{fI.)
This => T(IX)=O ¥ CIt E V=> T=O".
In either case 1'= O. .
Theorem 29. Let T be a linear trans/ormation on Hilbert
space V(F). Show that
T-O <:> <T(2o), P> =0 V IX, P e V.
Proof. Suppose T is a linear operator on a Hilbert space
V(F). Suppose at, fJ E V.
(i) Suppose T=O.
To prove that (T(at), (j>=0.
(T(ex), 11>=<0(2), {J)=(O, r;>=o.
(it) Let <T(ex), fJ)=O V ex, (J e V. . .. (1)

To prove T=O.
In particular (I) => (T(2o), T(at»=O, where ~=T(Gt)
.. T(Ot)=O V at e Y
• T=41
Theorem 30. Suppose T is self-adjoint operator on Hilbert
space V(K).
HILBERT SPAC8 393

Then S*TS is self adjoint for el'ery linear operator Son V.

Further if S is invertible and S*TS is self-adjoint, then T is also
selfadjoint. [Kanpur [Link]. F. Linear Algebra 1976J
Proof. Let T and S be arbitrary linear operators on a Hilbert
space V(K).
Step (i). Let l' be self-adjoint so that T = T*.
To prove that S*TS is self-adjoint
(S· TS)* = S*T* S, *= S*P"S= S*TS
(S'1',S)*=S*TS.
From this the required result follows.
Step (ii). Let·S be invertible so that S-1 and (S*)-1 exist.
S iTS is self-adjoint. => (S* n)* = s* TS.
~ S*T'*'S';'~=S*TS ~-S*'T*S=S*TS
S*T*(SS-I)=S*T(5S-})
::>
(S*)-1 S*T*I=(S*)-1 S*TI
::>
~ IT*I=ITI ;> T*=" T::> l'
is self-adjoint.
Theorem 31. Every linear operaTor T 011 a complex Hilbert
space is uniquely expressible as 1'= T} + iT2 , where Tl alld T2 are
Hermitian (self adjoint) operators alld i= v( -1) fKanpur 1991;86]
Proof. Let T be a linear operator on a complex Hilbert
space.
Evidently T=i (T+T*)+; Gi (T -1'*) 1
Taking T)=1 (1'+T*), T~=~i (1'-1'*), we get
T~, T 1 +iT2 •
Remains to prove that Tl and T2 are Hermitian operators
on V.
T/I<=1 (T+T-.\O)*=' (T*+T)~' (1'+1'*)=1'1'
I 1 I
T 2*=-i{ (T--T*)*=-2i (T*-T)=2; (T-P)=1'z•
:. Tl·~Tl' T'1,*=T2-
This => T) and T2 are Hermitian.
Uniqueness. If possible, let l' =P+ iQ be anol her represent-
ation where P*=P, Q*=Q.
Then r*=(p+iQ)*=P"'+'iQ*=P-iQ
394 HlLBEIlT ~PACE

N T=P+iQ l:)Ti-l'*=p T T*' =Q

ow T"*'=P-iQf 2 '2;
~ T 1 =P, T2 =Q·
Hence the representation T= T1 + iT2 is unique.
Theorem 32. A necessary and sufficient. condition that Q linear
operator T on a (omplex Hilbert Space V (unitary space) be self-
adjoint (or Hermitian) is that <T(a), at > be real for every «.
[Delhi 1981; Kanpur [Link]. F. 79]
Proof. Let T be a linear operator on complex Hilbert space
V(F).
(i) Let T be self-adjoint so that T=T*.
To prove that <T(cx), «> is real for every. E V.
<'T(<<), cr)=(<<, T·; «(7.»=(<<, Tt.»=(T~CIt), «)
or (T{CIt), (I.)=<T(at), ex).
ThIs ~<T(IX), IX> is real ¥ at E V.
(ii) Let < T(CI), at> be real ¥ cx E V so that
<T(IX), at> E R ¥ at E V. . .. (1)
To prove that T=T*.

(I) .. (T(<<), a)=(T«(I.), «)=(<<, T(a»

=(T*(at), «)
=> (T(<<), Q[)=<T~(cx), (I.>
~ (T(CIt), a)-<T*(Gt), «)=0
.. (T(IX)-T-!«(a), cr)=O
~ «(T~T*) (a), «)=0
=> T-T.==O T=T*.
Theorem 33. Let U be a linear operator on a Hilbert space
V( K). Then the following are equh·alent.
(1) U*=U-1,
(2) UU*=U*V==I,
(3) <Vert), V(v) > =(u, ).) ¥ lI, Y E Y, i.e., V preserves
inner products
(4) IIV(u)!I='/UI!¥UE v
l Kanpur M Sc. Final Linear Algebra 1976]
Proof. Let U be a linear operator on a Hilbert space V(K).
Let lI, )' E V be arbitrary.
We shall show that
(I) 0:> (2) ~ (3) ~ (4) • ~l).
(I) ~ (1).
U·=U-1". VU=*UU- 1, U.U=U-1U
HILBERT SPACE:. 395

=> UU*=I. U*U=!

=> UU*=U*U=I.
(2) => (3).
UU*=", U·V=I => < (U*V) (u), v>= <u, v> .: I(u) = "
~
<V(Il), V**(v»=<u, v>
=> <V(II), U(V) > = <II. v>.
(3) => (4) .. " l'
<V(II), U(v)=(u, v) => (V(u), U(u»=(u, u)
=> /I U(U) 112=:! u /1 2
=> II U(U) !/=/I u II.
(4) =- (1)
II V(u) Ii=il u!! ~ ii U(U) 1:2=11 u !12
=> <U(U), V(U»=<U, u>
::0- «U*U) (U), U>=<U, U>=</(U), u":;
=> «U*V-/) (U), U> =0
=> U*U-I=O By Th. 21.
~ U*U=I+O=1
~ U*UU-l=IU-1
=> U*/=U-1 .. U*=U-1.
Corollary. U is unitary
¢o II U(u -VI II = II u-v II

..;:. distance between the vectors u and V

=distance between U(/I) and U(v)
<* U is an isometry.
Ex. 10. Suppose T is unitary (orthogonal operator on a Hilbert
shace Vand W is a T-im·ariant. Then _.how that W.l. Is also
T·invariant.
Solution. Suppose T is a unitary operator on a Hilbert space
V(F) so that TT*=T*T=I.
Let W be T-invariallt so that We v
any I\' E W => T~ 1\') E W.
To prove that Wi jc; T-invariant. it is enouguh to show that
any WE Wi=> T(lr) e Wi.
Let It' E Wand u E Wi be arbitrary; then
<II, w>=O and T(W)=Wl E W.
Tis u ,itrary,
< T,u), 1\'\ >~ = < T~u), T(II'» = <II, 11'> =0
 :'196 HILBERT SPACE

Thus, < T~u). wl > = 0 ¥ wl E W.

This * T(u) 1. W => T(u) E W J..
:. any u E WJ. => T(u) e WJ. Proved.
Theorem 34. (a) let T be a linear operator on a complex
lIilbert space V. Theil
T is normal - 11 T*(u) 11= II T(u) II vuE V.
[Deihl 1980; Bundelkhand 81; Meeurt 80J
Proved. Suppose 7 is a linear operator on a complex Hilbert
space V.
(i) T is normal => TT*= T*T. «rr*) (u), u)=«1 *T)(u), u)
.. (T(T*(u», u)=(1fr(T(u)J, u)
=> (T*(u), T*(u»==(T(u), T**(u»
=> " T*(u) lill=(T(u), T(u»=I: T.(u) 1111
=> II T*(u) 11=11 T(u) II·
(U) Again 11"'(u) [Link] T(II) II. II T*(u) 112=11 T(u) III
=> (T"(u), T*(u»=(T(u), T(u»
=> «TT*) (u), II)=«T*T) (u), u)
=> «TT*~ T"'T) (u), u>=O.
Also V is wtnplex Hilbert space
=> 1T*- T*T=O (Refer Theorem ':J.7)
=> TT*=T*T
T is normal
=> Proved,
Theort:m 34. (b). If T is a normal orerafor on Hi/bert space
H, thm II TlII= 1/ T 11 2 •
Proof Let U E /l be arbitrary. Then T is a normal operator
=> I: Tu I/=I! T*u /I .. (1), by Theorem 34a.
II 12 11== SIJP iii 1'u !I : 1 Ii I ~ J}
=sup (i T(Tu) 1/ : I II I! ~ I}. using (J)
"..:sup {II T*(TII) Ii : iI u ii ~ I}
=t;UP {I! (T*7) (u' Ii : i! Ii I! ~ I}
= :! T*T 1i=I!T III, by Theorem 25.
Theorem 35. Suppose T is a linear operator on a Hilbert space.
Then T Is normal iff its real and imaginary parts commute.
[Delhi 1981; Kanpur [Link]. F. Linear Algebra 67, 77J
Proof. We know that every linear opetator T on a Hilbert
space is expressible as T=A+iB, where A=i (T+T*),
B=~ (T--T*) are self-adjoint operators so that A*=-·~.·f B=B*.
HILBERT SPACE 397
T*=(A+iB}*=A*+TB*~=A iB.
TT*,,,-(A+iB) (A-iB)~"'A2+ B2 +j (BA·-AB),
T*T=(A.-iB) (A+iB)c--=A2+B2+i (AB-Bo4).
T i" normal ~ TT*~-~ T*T
--. A2 +-B2+i (BA -AB)'==A2+B~+i (AB-BA)
e> BA - AB:-·=AB-BA ¢ 2AB=2BA
<=> AB=BA <=> A and B commute
¢¢> real and im:lginary part') of I commute.

Ex 11. Show that the set of all unitary operators on a Hilbert

space V(K) is group under the composition of operators.
[Kanpur [Link]. F. 1985, 82; Meerut 69, 70]
Solution. Let G denote the set of all unitary operators on a
Hilbert space V(K). Let T, S, FE G be arbitrary so that T, S, F
are unitary operators on V(K). Consequently
, Tr*=T*T=I, SS*=$"*S=I.
Firstly we shall show that TS is a unitary operator.
(TS) (TS)*=(TS) (S*TI(:)=T(SS*) T*'=TIT*
=TT*-=/
(TS)· (TS)=(S*T*) (TS)==S* (T*T) S=S*}S
=S*S=/
(TS)* (TS)=(TS) (TS)*=I.
This ~ 1"S is a unitary operator.
=> TS E G.
(i) T. S E G -> TS E G.
(if) 3 identity element I (identity operator) -= G S.t.
Tl=IT=T.
For 1/*= 1=1*/
(iii) Every element T E G has its inverse j1 E G.
For TT*=T*T=I ~ T*=T-1.
Also T-I (T-1)*=T* (T*r 1
=1.
Similarly (T-1)* (T-l)=i.
Thus (T-1) (T-1)*=I=(T-1)*, (T-1)
This::;. T-1 is a unitary operator
::;. T-l E G.
(iv) (TS) F= T(SF).
For [(TS) FJ (u)=(TS) [FlU)), uE V
= T lS(F(u)}]
398 HILBERT SPACE

= T [(SF)(/I)]
=[T(SF)](u)
[(TS)F] (u)=[T(SF)](u) ¥ u e V.
This => (TS) F=T(SF).
The results (i), (ii), (iii) and (iv) taken together imply that G
is a group w.r.t. operation of product of operators.
E~. 12. 1fT is an arbitrary linear operator on afinite dimen-
sional complex Hilbert space and if a and b are complex numbers
such that, a 1=1 b 1=1, then aT+bT* is normal.
{Meerut 1971; Kanpur 88]
Solution. Let I a 1= I b I= I so that
laj2=/bI 2 =1 or aa= bjj = 1
(aT+bT")*={aT* +bT*)*=aT""'+ bT
(aT+bT*)*{a r +bT*)=(aT""'+bT) (aT+bT*)
=aaT*T+ab (T*)2+baT2+bbTT*
=T*T+ab (T*)2+abT2+TT* ... (1)
(aT+hT*) (aT+bT*)*=(aT+bT*) (aT*+bT)
=aaTI*+abT2+ba (T*)2+bbT*T
=TT*+abT2+ab {T*)2+T*T
=T·q+ab (T"",)2+al)TI + TT.. ...(2)
Equating (1) to (2), we get
(aT +bT*) (aT+hP)*=(aT +bT*)* (aT+bT*).
This ~ aT +hT'iI< is a normal operator.
Ex:. 13. Let T he a normal operator on a Hilbert space and
fa polynomial with complex coefficients. Then the operator f(T)
is normal. [Kanpur 1987J
Solution Letf=ao+a1x+a2x 3 + ... +a",x"', then
/(T)=ool+a 1T+a 2 TJ+ ... +o",T"', ... (1)
(T*)A=T*T* ... T* (k factors)
=(TT... T k factors)*
=(Tk)*.
Hence (T*)k=(Tk)* ... (2)
[f(T)]*=(aol+a1 T+ ... + a", T"')*
=dol+d1T*+ ... +d", (1"")*
or [J(T)]*=dll+o1T*+ ... +a... (T*)",.
Tis norm:!1 ~ TT*= T*T .. (3)
=> T(T*f=T[T*(T*,)t-l]=(TT*) (T*YH
=(T*'T) (T*) ..-1
HILBERT SPACE 399

=T*[T(T*)·-IJ= ...
= TX<[Tiit1 (T*)k-S]
~ T(T'*)"=(Ti'/<)2 [T(T;\~)a-2]

=(T:Jt:)t T
~ T(T*)"=(T*" T. .. (4)
Similarly we can show that T"T*= T:Jt:7".
By virtue of (3) and (4). it is easy to show that
[f(T)) [f(T»)*=[f(T)]* [f(1')]·
This =-f( T) is a normal operator.
DeC. Projection If P is a project:on with range space M
and null space N, then
M={x E H; Px=x}, N={x E H; Px=O}.
Here we also say that P is a projeJtion on M along N. Also
H=M~N.
A projection P on a Hilbert space H is called perpendicular
projection if M and N are orthogonal. In this case N .L M and
so N=Ml or M==;Nl..
Thus a perpendicular projection P on M is a projl!ction on AI
along Ml. rn the theory of Hilbert space, we shall consider only
tho~e projections which are perpendicular. Hence we omit the
repeated use of the word perpendicular. Thus by a projection P
on n, we shall always mean a perpfndicular projection on H.
An operator P on a Hilbert space H is a projection iff P2=P,
p=p*
Also Px=x ¥ x E M.
Important result. (i) Px =X \rI x E M
(ii) II Px II ~ Ii x \I
(iii) pa=p*=p
(iv) (Px, x)= II Px III.
Theorem 36. P is a projection on a closed linear subspace M
on H ~ I-P is a projecTion on Mi.
Proof. Let P be projection on a closed linear subspace M
of a Hilhert space H.
Aim. I-P is a projection on Mi.
We have pZ=P*=P, Px=x >.,f x E M.
(1_ p)I=/S+P2_IP-P/=J+P_P_1'
=I-P
(/-P)*=I*-P*=l-P
Let N be range space of-I :-P.
 400 HILBIlRT SPACE

any x EN.¢> (1-- P) (x)=x ¢> J(x) -P(x)=x

<=> x-P(x)=x ~ --P(x)=O
~ P(x)=O ~ x is in the null space of P.
~ x E MJ.
.•. N=MJ. •
Finally, we have proved that
(l-P)2=I-P, (/-P)*=l-P,
(I_·P) (x)=x ¥ x E Ml..
These facts prove that 1- P is a projection on MJ..
Conversely. Suppose that J-P is projection on MJ., where
M is a closed linear subspace of H so that MJ.J.=M.
It follows from the first part that l-(l-P) is a projection
on (M.l) 1 =MJ.J. i e., P is a projection on M.
Problem 14. If P is a projection on a closed linear subspace
Mof H, then
x E M ~ P(x)=x 4> " P(x) /1=/1 x II
Solution. P is a projeetion on M ~ p2=P*=P
I. Let Px=x where x E H.
x E H ;... Px E M as M is range space
=> x E M as Px=x.
Thus Px=x ~ x E M
Conversely suppose that x E M: To prove Px=x.
Let Px=y, then PPx=Py or Py=P 2x=Px
or Px=Py or Px-Py=O or P(x-y)=O
This ~
x-y is in null space of P
x-y E MJ. => 3 z E Ml. S.t. x--y=z
~
~ x=y+z.
Px EM:$> Y E M as Px=y
Now x=y+z with y E M, z E Ml.
where H=MffiMJ..
Also x=x+O where x E M (given) and 0 E MJ..
Thus x+O=x=y+z
Also x has unique representation. Comparing both sides,
z=O, x=y. But y E M and so x E M.
Thus x EM=> Px=x
Finally x E M <:> Px=x.
II. Aim. Px=x <:> " Px 1/=11 x"
Obviously Px=x => /I Px !/=Il X" ...(1)
 HILBERT SPACE 401

Conversely suppose that II Px li=1/ x 1/. ••• (2)

To prove that Px=x.
Evidently II Px+(I-P) x 11=11 x II. ...(3)
But Px E M and (I-P) (x) E Ml..
This ~ Px .l (I-P) (x).
By Pyttlagoras theorem.
I Px+(l-P) (x) 111=11 Px 111+11 (I-P) (x) III
using (2) and (3),
II x 112=11 x 111+ ,,(/-P) (x) liZ
or II (I-P) (x) !I =0 or "x-P(x) li=O
or x-P(x}=O or Px=x. . .. (4)
Finally x=Px <:> II Px 11= II x II, by (I) and (4).
Theorem 37. (Projection Theorem). q P is a projection on a
Hi/bert space H with range space M and null space N, then
M .L N <:> P is self adjoint
and in this case N=Ml.. [Meerut 1973; Kanpur 84, 81]
Proof. Let P be a projection on a Hilbert space H, i.e., P is
a projection on M along N so that
H=MffiN, and z E H
is uniquely expressible as
z=x+y with x E M, YEN.
Also Pz= P(x+y) =x.
I Let M I. N, then
(x, y)=O=(y, x).
Now (Pz, 'z)=(x, z)=(x, x+y)=(x, x)+<x, y)
=(x, x)+O=I! x Ill. . .. (1)
Also (P*z, z)=(z, Pz)=(z, x)==(x+y, x)
=(x, x)+(y, x)=(x, x)+O=il x II·. ... (2)
Equating (I) to (2),
(Pz, z)=(P*z, z)
or «P -P*) z, z)=O yo z e H.
This 0 P-P*=O
~ p=p* 0 P is self adjoint.
U. Let P be self adjoint. To prove M .l N.
x E M~ Px=x,
By def. of M.
y E No Py=O,
by def. of N.
(x, y)=(Px, y)=(x, p·~y)=<x, Py)
402 HILBERT SPACE

=(x, 0)=0
:. (X, y)=O
Also x E M, YEN.
This => M .L N.
[Link] M.l N. To prove N= Ml. ... (1)
any x E N ~ x..LM ~ x E M.l.
/. Ne Mi. . .. (2)
Suppos e (I) is not true, then N~M.1.. . .. (3)
By (2) and (3), N is a proper subset of Mi.
~ N is a closed subspac~ of Hilbert space M.l.
{For N is null space ~ N is closed]
~ a
zo-:cO E Ml. S.t. zo...LN.
Now Zo E MJ.. => zoA..M. Also [Link].
Hence we can conclu de
[Link] as H=M( JjN.
This is possibl e only if zo=O. A contrad ication .
Hence N=M.1..
Problem 15. If T is an arbitra ry linear operator 011 afinite
dimensional comple x Hilbert space and if a al1d b are complex num-
bers such that I a 1=1 b I, /hen aT+b. * is normal. (Meeru t 1971]'
Solution. Let I a / :-~ I b / so that
I a 12~1 b /2 or ad=bli .
(a r +bT-iF)* =dT*+ b(T*)* =dT*+ bT
(aT+b T*)* (aT+b T* =(dr* +iiT) (aT+b T*)
=daT* T+dh (T*)2 +b'1r2 +bbTT * ... (1)
~aT+bT*) (aT+b T*)*= (aT+b T*) (dT*+ liT)
=aiiTr *+abT 2+bd (T*)2+ bbT*T .
Using ad = bfj, in first and last term,
(aT+b T*) (aT+bT *)*=bb TT-'F+ afjT2+ bd(T*) 2
+adr* T. . .. (2)
Equati ng (I) to (2), we get
(aT+b T*) (aT+b P')*=( aT-t-b T*)* (aT+bT~).
Thi'l ~ aT+br *' is a normal opera.t or.
Proble m 16. A closed linear subspace M of a Hi/bert space H
is invariant under an operator T <:> MJ. is invariant under T".
Solutio n I. Let M be invaria nt under T. To prove that M.l.
is invaria nt under T'*'.
Let XE MJ.., yE M, then (x, Y) =0.
 HIL8ERT SPACE 403

By def. of adjoint,
(T*'x, y)=(x, Ty)=(x, Yl)=O
where Yl=TyEM.

Thus (Tx:'F, y)=O.

Also yEM.
Hence this declares that T*'xeMJ..
Thu,; xEMl. ~ T*xEMJ..
:. MJ. is T*-invariant.
II. Let M be a closed linear subspace of H so that MJ.J.=M.
Let Ml. be' invariant under T*.
Aim. M is invariant under T.
It follows from the first part that (Ml.)J. is invariant under
(T*)*
i.e. Ml.l. is invariant under T**
i.e., M is invariant under T.
Theorem 38. If P is a p:rojection on closed linear subspace M
of a HUbert space Hand T is a linear operator on H. then
M reduces T <:> TP=PT. , (Meerut 1987. 86)
Proof. M is a closed subspace of H
=> H=M(f)Ml..
Suppose P is a projection on M along Ml.,
then Pz=x, Px=x, Py=O ... (1)
where any zE H is uniquely expressible as
z=x+y with xEM, yEMl..
I. Let M reduce T so that M and Ml. both are invariant
under T.
To prove TP=PT.
By assumption, Tx E M, TyE MJ.. . .• (2)
(PT) (z)=(PT) {x+y)=P [Tx+Ty].
Take TX=X1EM. Ty=y,EMl..
(PT) (z)=P(x1+Yt)=:=PX1+PYl=X1+O=X1 ••• (3), by (1)
(TP) (z) = T(fz) = Txd.x 1 .. (4) by (1).
Equating (3) to (4), (PT) (z)=( TP) (z).
This => PT= TP.
II. Conversely let TP=PT.
To prove M reduces T, i.e., to prove M and Ml. both are
invariant under T.
 404 HILBERT SPACE

P (Tx)=(PT) (x)=(TP) (x)=T(Px)=Tx, by (I)

P (Tx)=Tx, this::;. Tx E M
P (Ty)=(PT) (y)=(TP) (y)=T(Py)=T(O)=O, by (I).
P (Ty)=O, this TYEMJ.
•• xEM => TxEM
YEMJ. => TyEMJ..
Hence M reduces T.
Theorem 39. Let El and E2 be perpendicular projections on U1
a71d U2 respectively, where U 1 and Ua are closed subspaces of a
Hi/bert space V(K).· Then sholl' that
El and £2 are orthogonal <=> U1 and U2 are orthogonal.
E 1E 2 =0 or E2 E1 =0 <=> U}.1U2•
[Kanpur [Link]. 81, 78}
Proof. Suppose E} and E2 are perpendicular projections on
U1 and U2 respectively which are subspaces of a Hilbert space V
so that
E}~=El=El* }
E22=E2=E~* ' .. (I)
E1(U1) =U1 "IE VI L
<=>
1:.~(U2) ="2 ~ "2 E U2 r ... (2)
Step (i) Let E1 and Ea be orthogonal so that
E1E~=O, E2 E1=0. ...(3)
Let "IE VI' U2 E U 2 be arbitrary.
To prove that VI and V,. are orthogonal, it is enough to show
that <"1'
".,)=0.
ul E VI ~ E1(U1)=U1 }, by (2).
u.E V 2 => E 2 ("-4)=U 2
(u1o u2 )=(E1(U 1 ), £2(U2 )=«E2 * E 1 ) (u 1 ), u2 )
=«E2E1) (U1), u2), by (I).
=(0 (u 1 ), u~), by !3)
=(0, u2 )=0.
Step (ii) Let VI and V 2 be orthogonal so that
vd.U2 => U1 =U.J. ... (4)
To prove that E1 and E2 are orthogonal, it is enough to show
that
H1LBER T SPACE 40S
Let U 1 E VI be arbitrary, then UtE U2 .L, by (4)
"lEUI :::> £1(U1)=U1 :::> E1(u1)eUt =U2.L
=> £1(u 1)EU2.L => E1(U1)¢U2 For U2 nU2J.={0}.
=> £2(£I(U1»=0 For u¢U2 => E2(u)=0
=> E 2E 1(U1 )=0 ¥ ul E U1

=> E2E 1 =0
=> (£2£.)*=(0)*=0
:::> El*E2*=0,. E1 £2=O, by (1). Pr,oved.
Theorem 40. If P is a projection on a closed linear subspace
M of H, then M is invariant under an operator T
~ TP=PTP.
Proof. Let P be the projection on a closed subspace M of H
so that
I'X=X ¥ xEM ... (1)
and ze H =>' PZE M. ... (2)
I. Let M be invariant under an operator T, then
xEM => TxeM ... (3)
Aim. TP=PTP
ZEH:::> PzEM => T(PZ)EM, by (3)
=> r(pz)=pz ... (4)
Further (PTP)(z)=aP [r(pz)J=P(Pz), by (4)
_p2Z =PZ as PI=P
-T(Pz), again by (4).
=(TP) ,z)
:. PTP=TP.
n. Let PTP=TP .. ,(5)
To prove that M is invariant under T.
Let xeM, then Px=x. Now (5) gives
(PTP) (x)=(TP) (x) or (PT) (Px)=T(Px)
or (PT) (x)= Tx or p(Tx)e::. Tx
This :> TXEM, according to (I).
Thus xe M => TXE M
This proves that Mis T-invariant.
Theorem 41. If Pl' P 2• .. , P.,. are projections on closed linear
sub,\p~lces M 1 , M 2 ,··., Mn of H, thell
 406 HILBERT SPACE

II
p= ~ P, is projection ~ P,'s are pairwise orthogonal
,-,
and in this case P is a projection on M = M J + M2 + ... + M ".
lKanpur 1981, 78J
Proof. Let P, be projection on closed linear subspaces of
M, of H, then
P,I=P,=P,· ... (1)
and Pi (X,)=x, ¥ x, E Mt

Also let P=PI +P2 + ... +P,,= "~ P,

i_I

P*=(
'£ ~ P,=P
_1 p,)* = '=£1 Pt*= .-1
or P*=P ... (2)
I. Let P,'s be pairwise orthogonal so that
p,p,=O for i::pj ... (3)
Ai.... P is a projection.

Then P'==(
,i-l
£ P, )1 =~ i
P.2+2
i::pj
~ PiP,

=~ P.+O=P, by (1) and (3)

Finally, J'2=P, p=p*

This pr~es that P is projection.
11. Let P be projection.
To prove that PIs are pairwise orthogonal.
If T is any projection and zE H, then
(Tz, z)=(Tlz, z)=(TTz, z)=(Tz, T*z)
=(Tz, Tz)=1i Tz ii-
Thus T is projection ~ <Tz, z)= I Tz 112 ... (4)
Let xER(Pl)=range space of P, so that Pi (x)=x

~"
nx 112=\1 P,(x) IP' ~ 1_1 II P,(x) 11 2 = E" (PI x, x), by (4)
'_I
==(P1 x, x)+ ... +<P" x, x)
-«PI+ ... +p,,) x, x)=(Px, x)== II Pxll t , by (4)
or \I x III ~ II Px liZ. But II Px III .;;; II X 11 2•
(Refer problem 18)
 HILBf:RT SPACE 407

From this we conclude that the sign of equality must hold

throu~hout the above calculation. Hence
to
II P,x
.
W= 1-1
~ II P,(x) 1/2

This ~II PJ(x) ill=O ¥ j s.t. j=l=i

= Pj{X)=O if j::p.i
=> x E Null space of P1 => x € M,L ifj¢:i
=> x J..MJ ¥ j s.t. j#i where x € M,
This proves that M,.1. MJ for i:j:i
This ~ p,P,=O for i#}.
Ill. Finally we have to prove that
to
P is projection on M = ~ M,
J-1

For this we have to prove that R(P)=M.

any x E R(P) => Px=x ~ ( ~ P, ) (x)=x

ox=P1X+P 2 X+ .. +p"x
But PIX e Ml • PzX e Ma etc.
Therefore x E M1+M.+ ... +M,,=M
Thus x E R(P) => x € M
.'!. R(P) C M ... {S)
We have proved that II py 11=11 x II ¥ X € M, and each M,
is contained in the range of P and so M is also contained in the
range of P.
:. Me R(P) ... (6)
By (5) and (6). R(P)=M
This completes the proof.
Problem 17. If P and Q are prujections on closed linear sub-
SpUCf:S M wnd N of H, 'pnJl'e that
PQ is projection ... PQ=QP
n N.
In this case, show that PQ is projection on M
[Kanpur 1981, 78; Meerut 88.87]
Solution. Suppose P and Q are projections on closed linear
subspaces M and N respectively. Then
P2=P=P*, Q2=Q=Q*. ...(1)
 408 HILBERT SPAce

Px=x ¥ x E M Ilnd Qx=x ¥ x E N ... (2)

Step I. Let PQ be a projection.
Aim. PQ=QP.
By assumption, (PQ)*=PQ or Q~p·=PQ
Using (I), QP=PQ.
Step II. Let PQ=QP ... (3)
To prove that PQ is projection.
(PQ)2:a(PQ) (PQ) = (PQ) (QP), by (3)
=p(Q2) P=PQP, by (I)
=P(QP)=P(PQ), by (3) =P"Q=-PQ, by (I)
or (PQ)I=PQ ... (4)
(PQ)*=Q*P*=QP, by (I) =PQ, by (3)
or (PQ)*=PQ .. (5)
By (4) and (5), PQ is projection.
Step III. To show that PQ is projection on Mn-N
For this we must show that R(PQ)=MnN
and x EM n N ~ x € M, x € N ~ Px=-x, Qx=x
=> (PQ) (x)==P(Qx)=Px=x
~ (PQ) (x)=x ~ x € R(PQ)
:. M nNe R(PQ) ... (6)
• J
Agam any y € R(I-Q) =-> (PQ) (y)=y ... (7)
=> P((PQ) (y)]==Py ~ (P2Q) (y)=Py
=-> (PQ) (y)=Py, by (I)
Usjn~ (7), we get Py=y, this => Y E M
Using PQ=QP in (7), (QP) (y)=y ... (8)
This ~ Q[(QP)y]=Qy ~ (Q2P) (y)=Qy
~ (QP) (y)=Qy ~ Qy=y, by (8) ~ Y € N.
Thus y E R(PQ) =-> y E M, YEN ~ y E M n N
:. R(PQ) C M n N ... (9)
By (6) and (9), R(PQ)=M n N
This proves the required result.
Problem 18. .If P is projection on Q Hilbert apace H, then
(i) P is positive operator, i.e., P ~ 0
(ii) O~P~I
(iii) II Px II E; II x II ¥ X e H.
(iv) II P II ~ 1.
Solution. P is proj~ction on a subspa~e M of a Hilbert space
H.
 HILBERT SPACfl 409

=> PX=p "". X E M ... (1)

and p2=P*=P ... (2)
To prove (i) Let x E H
<Px. X>=<P2X, x>=<PPx, x>=<Px, p·x>
=<Px, Px> .... i/ Px !I~ ~ 0
Now <Px, x> ~ 0 and PI=P
By def. this ~ P is positive operator 01> P;;' 0 ... (3)
To pro"e (ii) P is projection on H
~ /- P is projection on H (Refer Theorem 36)
~ I-P ~ 0, by (i)
=> I ~ P .. (4)
By (3) and (4), 0 E;.; P <; I
To prOle (illl Let x E H.
If R(P)=M, then R(I-P)=Ml., by Theorem (36).
Evidently Px E M and (I-P) (XfE M~
Thi~ => PXJ...(I-P) (x)
By Pythagoras theorem,
" Px+(I-P) (x) 1!2= II Px 1i1+1i (I· x) (x) 1111
or I x 1i 2 =1t Px 112+1i x--Px !l'! ~ " Px j:~
or II x 112 ~ II Px liZ
or II Px II ~11 x II
To prO\'e (i\') By def.
" P ii-sup { II Px II: /I x II <;; I}
But II Px II ~ !! x /I , by (iii) part
II P II ~ su p { II x II : II x \I ..; I} ~ I
or II P II ~ 1.
Problem 19. If Pond Q are projcctiultJ on closed linear sub-
spaces M and N of a Hilbert space H, thell 5how that tile folluwing
are equivalent:
(I) P ~ Q.
(Ii) II P,'{ II ~ II Qx II 'V x e H
(iii) MeN.
(ill) QP=P.
(v) fQ=P. [Meerut 1980J
Solution. In order to prove the Problem, we shall prove
that
(i) => (ii), (ii) => (iii), (iii) => (iv) , (iv)~ (V), (v) => 0)
<Px,x>=<P'x, x>=<PPx, x>=<Px, p*x>-:.'«Px, Px>
or <Px, x>=lil'x ii~ ... (1) v- x E II.
 410 HILBERT SPACE

(i) ~(ii)
P ~ Q ~ Px ~ Qx ~ < Px, x > <; < Qx, x >
=> \I Px 112< II QX 11 2, by (I)
=> II Px II ~ /I QX "
(ii) '"" (iii)
'>.f x EM==- Px=x ~ II Px 112=/1 X 112
-::> II X 1/=/1 Px II EO;; II QX H=> II X II< II QX II ... (2)
But II QX" E; " X II ... (3) by Problem 18.
Bv (2) and (3), II QX 1:=1/ X II
This => Qx=x, by Problem 14.
=>xEN
1 hus any ~ EM=> X E N : . MeN.
(iii) => (iv). It is given that MeN.
Px E M=range space of P. But MeN.
:. Px E N and so Q(Px) E N.
This preves that Q(Px)=Px. For yEN => Qy=y
or (QP) (x)=Px ¥ X E H
:. QP=P.
(iv) ~ (v). It is given that QP=P.
This ::;> (QP)*=P* => P'~Q*=P*

=> PQ= P. For P and Q are projections.

(v) => (i). It is given that PQ=P.
Let x E H. Then by (I),
<Px, x>=11 Px 112=il (PQ)X 112 as PQ=P.
=11 P(Qx) 1P' <" II Qx 1/2= <Qx, x>, by (I).
[For II Py II < II y II ¥ yand '>.f P]
<Px, x> < <Qx. x>.
or
This => P" Q.
Problem 20. ~f P and Q are projections QII closed linear sub-
spaces M a11d N of a Hilbert ~pace H, then prove that
Q -P is a projection => P ~ Q. [Meerut 1974]
Also prove that Q--P is Projection on NnMJ..
Solution. Let Q-P be a projection
To prove that P ~ Q.
Since every proje(;tJOn on a Hilbert space is a positive
operator and so Q-P ~ 0, by problem 18
or Q ~ P, .or P ~ Q.
Conversely. Suppose that P E;; Q.
Aim. Q-P is a projection.
P is projection => pI = p* =-' P
In view of this, we have
HILBERT SPACE 411

(Q -P)*=Q*-P*=Q-P
(Q_P)2=(Q _ P) (Q_P)=Q2+p2_PQ_QP
=Q+P-PQ-QP=Q+P-P-P=Q-~
For P ~ Q => PQ=P, QP=P, by problem 19.
Finally, (Q_P)2=Q_P, (Q-P)*=Q--P.
This proves that Q-P is a projection.
Second Part. To show that Q- P is a projection on NnM.1..
For this we must prove that
R (Q-p)=NnM.1.
and x E R (Q-P) _.. (Q-P) (x)=x ... (1)
=> Q (Q--P) (X)= Qx
"'" (QI_QP) (x)=Qx.
But Qll=Q; QP=PQ=P
=> (Q-P) (x)=Qx
=> x=Qx, by (1) => xERtQ) => xEN
Again by (I), P(Q-P) (x)=Px
=> (PQ_P2) (x)=Px => (P-P) (x)=Px
:;> O=Px "'" x E null space of P => x C M.1..
Finally, x E R (Q-P) .. x E N, x E M.1. => x E NnM.1..
:. R (Q-P) C NnM.1.. . .. (2)
Again any y E NnMJ.. "'" YEN, Y E M.1.
.. Qy=y, Py=O
.. (Q-P) (y)=y-O=y
=> y e R (Q-P).
:. NnM.1. C R (Q-P). ...(3)
By (2) and (3), NnM.1.=R(Q-P).
This completes the proof.
Theorem 42. If N is a normal operator on a Hilbert space H,
then II N 1/ 2 = II N2 II. lMeerut 19801
Proof. Let x E H be arbitrary.
N is normal':> II Nx 1/=11 N*x II ... (1), by TIl 26.
Replacing x by Nx, we get II NNx 1/=/1 N*Nx II
or II N2x II = II N* Nx II .. (2)
. By def. II NI II =sup t II N2x II : II x II ~ I}
=sup { \I N*Nx II: II x II .;;; I}, by (2)
= IIN*NII
or II NI II = II N* N n
using the fact that II N 111= II N*N II is true for any operator N,
we get
II NI \I = II Nlil. Pro"ed:.
412 HILB£;(;. T SPACE

Exercises
1. (a) Define a Hilbert space. Prove the following in a
Hilbert space H :
(i) I <x, y> I .;;. II x II • H y II
Oi) XII- X. Yn-+)' => (X n , YII) ~ (x, y)
where <XII>, <Yn> are sequences in H and x, Y E H.
(b) Prove that a closed convex subset C of a Hilbert space
H contains a unique vector of smallest norm. IKanpur 1988]
2. Define normal and unitary operators on a Hilbert
space H. If T is an arbitrary operator on H and if IX and ~ are
scalars S.t. i IX I = I fJ :, show that «.T+fJT-~ is normal.
[Kanpur 1988]
3. (a) Define a Hilbert space. Prove the following in a
Hilbert space H :
(i) Jf M is a linear subspace of the Hilbert space H, show
that M is closed <* M=Ml..i.
(ii) If {P,} is an ortho normal set in H,
then ~ I < x, et > il
i
< II X liz, x E H

(b) Define normal and unitary operators on a Hilbert space

H. Let T be a normal operator on a Hilbert !>pace il and p(x) a
polvnomial with complex coefficients. Prove that p[TJ is a nor-
mal operator. [Kanpur 1987]
4. Define a positive operator r on a Hilbert space H.
Show that it hHs on unIque positive sequare root. lKan, ur 19871
S (a) Define a normed linear spaces (NLS). Prove the
following5 in NLS (N, II II).
(i) / [I x II - II Y ii/ ~ II x-J II
(ii) lf {Xn} be a sequence in N, such tllat
"',,-X, then II Xn II ... II x
{Iii) If {'''n}, {Y .. } are sequences in Nand
{IX n }. {(3.. } are scal~r sequences such that
x n- x, J,,'-~Y and Cl" .... CIt. Bn-B. then
ex"x,,+(;")'n-:- x+r~.r. [Kanpur 1989)
(b) Prove that the linear spa:e C lO, 11 ot all contilluoUS real
functions defined 011 the dosed unit interval [0, 1] is a normed
linear space (NLS) w.r.t. the norm oran element/defined by

. 11/11 = f: 11(t) I dt,

but not a Banach space. [Kanpur 19861
HILBERT SPACE 413

6. (a) Define a Hilbert space. If B is a Banach space whose

norms obeys the parallelogram law and if an inner product is
delin d on B by
4(x, y)=iJ x+y l/a_11 x-y :il+i" x+;y 111-; II x-Iy II!
then prove that B is a Hilbert space.
(b) If x and yare any two elements of a Hilbert space H,
prove that
I (x, y) I ~ II x II . II y "
Explain when does equality hold. [Kanpur 1986]
1. (a> Show that an operator T on a Hilbert space H can
uniquely he expressed as T= T1+iTjj where Tl and T2 are self
adjoint operators H.
(b) If T is an operator on a Hilbert space H for which
(Tx. x)=O for all x, then show that T=O.
(c) Prove that every Hilbert space i<; reflexive.
[Kaopur 1986J
8. (a) Define a Hilbert space. Prove the following in Hilbert
space H:
(i) I (x, y) I < "x II • II y II
(ii> Xn-X, Yn-y ~ (x .. , Yn)-(X, y)
where !xnl, {Y .. } are sequences in H and x, Y E H.
(b) Let H be a Hilbert space and {e,,} be an orthonormal set
in H. Prove that the following conditions are all equivalent to
one another:
(i) fed is complete
=-
(ii) x..L {ei} x=O
(iii) x E H => x=l; (x, e,:) e,
(iv) x E H => II x i I=l; I (x, e,) 12 LAaopur 1985]
9. (a> Define self adjoint operator on a Hilbert space. Prove
that the self adjoint oper!ltors in 8{H) from a closed real linear
subsp!lce of B(H) which c.:mtains the identity transformation.
(b) Define normal and unitary operators on a Hilbett space
H. Prov~ that unitary [Link] on H form a group.
(Kanpur 1985]
10. Define a Hilbert space H and prove that a closed
convex subset C of H contains a unique vector of smallest norm.
[Kaopur 1984)
11. (a) If M is a closed linear subspace of a Hilbert space
H, then prove that H=M(£)Ml., stating all the theorems that are
used in the proof.
(b) State and prove Bessel's inequality. [Kaaput 1984]
U. (a> Define the adjoint operator r-~ r* on B (H) and
prove that
414 HILBERT SPACE

(i) (Tt T2 ):'«=T2*' Tt '1:

(ii) =
II T* II 17 Til
(iii) II T* Til = II Till.
(b) If P is a projection on a Hilbert space H with range M
and null space N then prove that
M J. N <=> P is self adjoint. [Kanpor 19K4]
13. (a) S:ate and prove the closed graph theorem.
(b) Define the conjugate T* of an operator T on a normed
linear space N and prove that the mapping T-+T*' is a norm pre-
serving of B (.TIl) into B (N*'). [Kanpor 1983]
14. (a) Define a Hilbert space and prove that a closed
convex subset C of a Hilbert space H contains a unique vector of
smallest norm.
(b) If M is a proper closed linear subspace of a Hilbert
space H. then prove that there exists a non-zero vector Zo in H
s.t. zoLM lKaDpar 1983]
IS. (a) Let H be a Hilbert space and/be an arbitrary func-
tional in H*. Prove that there exists a unique vector yin H such
/(X)=(x. y) Y x E H.
(b) Define the adjoint operation T-T~ on B (H) and prove
that it is one-one onto as a mapping of B (H) into itself.
[Kanpar 1983]
16. (a) If {ej} is an orthonormal set in a Hilbert space B,
then show that
1:: I {x. e,} 12 ~ II X 1!2
for every vector x in H.
(b) Prove that a Hilbert space H is separable <:> every
orthonormal set in H is countable. [Kanpor [Link]. F. 198Z]
17. (a) Lei: B be an ordered b:lsis of a Hilbert space H.
Describle how using this fixed basis B a matrix [T] can be assig-
ned to each operator Ton H. Such that:
(i) the mapping T-+[T] is a one-one mapping of the set of
all operators on H onto the set of all matrices.
(ii) algebraic operations can be defined on the set of all
matrices in such a manner that the mapping T-+[T] preserves
the algebraic structure of B(B).
(b) Show that the dimension of B( H} is fll, where n=dim H.
[Kanpor [Link]. F. 1981]
18. Prove that a closed convex subset lof a Hilbert space
contains a unique element of smallest norm.
[Kanpur [Link]. F. 1981]
 HILBERT SPACE 415

J9. (a) Prove that P lo P20 ••"0 P n arc projections on closed

linear sub;paces M i , M 2 , .••. , M .. of a Hilbert space H, then
P=PI+Pa+ ... + 'n isa projection ifand only if P,'S are pairwise
orthogonal. When this is the ca'>' , show further that P IS the
projection on M = MI + M2+'" + M ...
(b) If P and Q are projections on closed linear subspaces
M and N of a Hilbert space H, prove that PQ is a projection if
and onlv if PQ=QP. In this case, show that PQ is a projection
on M n N. [Kanpur [Link]. F. 1981, 78]
20. (a) If P is projection on H with range M and null space
N, then show that
MJ.. N <> P is self adjoint
and in this case N=ML.
(b) Show that the unitary operators on H form a group.
[Kanpur [Link]. F. 1982]
21. Prove that a Hilbert space H is separable if and only if
every orthonormal set in H is countable. [Kanpor [Link]. F. 19R11
(a) If M is a proper closed linear subspace of a Hilbert
space H, prove that there exists a non-zero vector Zo in H, such
that zoJ.M.
(b) Prove that if M and N are closed linear subspaces of a
Hilbert space H such that M .lN, then M +N is also closed.
[Kaupur [Link]. 1980, 78]
22. (a) Let H be a Hilbert space. Prove that the self adjoint
operators in B(H) form a closed linear subspace of 8(H).
(b) If an operator T on a Hilbert space /J is such that
(Tx, x)=O for all x in H, show that T=O.
[Kanpur [Link]: F. 1980, 79J
 20
Finite Dimensional
Spectral Theory
20'1. Der.
Let T be a linear operator on Hilbert space H(K). Any
scalar I. E K is called an eign ~alue of T if 3 non-zero vector
v E H s.t. TV=AV. Such a vector v is called eign vector of T
corresponding to the eigen value ". The following have the
same meaning: proper vector, eigen vector, characteristic vector,
latent vector.
. Similarly the following have the same meaning: latent value,
elgn value, proper value, characteristic root, spectral value.
The set of eign values of an operator T is called the spect-
rum of T and is denoted by aCT).
Let M denote the set of all veetors of H.
s.t. Tv='\v
i.e. M={v E H: Tv="v}
Then M is called eign space of~. [Kanpur 1988, 85]
20'2. Spectral Solution.
Let T be a linear operator on a Hilbert space H. If:l distinct
complex numbers "10 "I' .. ..
~ and non-zero pairwise orthogonal
projections PI' Pz, "', P .. S.t .
... (1 ) and
;=1
Then the expression (I) is called spectral solution for T.
20'3. Invariant.
Let T be a linear operator T on a Hilbert space H. A closed
subset M of H is said to be T-invarlant or invariant j under T if
any x EM=> Tx E M.
20'4. Reducibility.
Let T be linear operator on a Hilbert ~pace H. If M is a
closed subspace of H, then T is said to be reduced by M if M
and MJ. both are invariant under T.
If M is invariant under T, then we also say sometimes that
T is reduced by M or M reduces T.
Theorem I. Let A be an eigen value 0/ an operator T on a
FINdE DIMeNSIONAL SPECTRAL THEORY 417
Hilbert Space H. LeI M be the eigen space of A, then M is a closed
subspace of H.
Proof. We have M={x E H: Tx=Ax}.
Let a, b be scalars.
x, y E M ~ Tx=Ax, Ty=Ay
=> T (ax+by)=aTx+bTy
as T is linear
==aAx+bAy=A (ax+by)
=> T (ax+by)=A (ax+by). ax+by E H
~ ax+by EM.
This proves that M is a subspace
M={x E H: Tx=Ax=Mx}={x E H: (T-M)(x)=O}
=Null space of linear map T-M
=c1osed space. For null space of linear map is
always closed. Hence the result.
Theorem 2. The eigen space of a linear operator T on a Hilbert
space H is invariant under T.
Proof. Let M be an eigen space of T corresponding to eigen
value A. Then
M ={x E H : Tx=Ax}
;\ E K, x EM=> Ax E M. by def. of subspace
~ Tx EM, by def. of M
.. x E M~ Tx E M.
Hence M is invariant under T.
Theorem 3. The eigen values of a self adjoint operator are a/l
real. , I [Kanpur [Link]. 1974, 71]
Proof. Let S be a self adjoint operator on a Hilbert space
H (K) Let Tx=;\x, then A is an eigen value of T relative to eigen
vector x.
To prove A is real.
;\ <x, x)=<Ax, x)=(Tx, x)=(x, T*x)
=(x, Tx)=<x, ;\x) as T=T*
=i<x, x)
or (A-i.') (x, x)=O.
This => ;\-):" =0 => ,\=~
=:. ,\ is real.
Theorem 4. If'x, yare eigen vectors of a self adjoint opera-
tor T on a Hilbert space H relative to different eigen values of T,
then x .L' y. [Kaapur [Link]. P. 1971, [Link]. F. 90]
Proof. Let T be self adjoint operator on a Hilbert space
H (K). Let
Tx=Ax, Ty=IJ.Y where Ai-II-'
418 FINITE DIMEN5lONAL SPECTRAL THEORY

The prove
T is self adjoint

(Refer Theorem 3)
,\ (x, Y)=(AX, y)=( Tx, y) =(x, T*y) =(x, Ty)
=(x, I-'y) =p (x, Y)=/L (x, y)
or (A-I') (x, y) =0
But '\:F 1-'.
Hence (x, y)=O or x.1y.
Tbeorem 5. LeI T be a normal operator on a Hilhert space H.
Then
Tu=,\u ~ T*U=AU, u E H
[Kanpur [Link]. F. Linear Algebra 1978, 76]
Or Every vector of a normal operator T is also characteristic vector
of T*. [Meerut 1988J
Proof. Let T be a normal operator on a Hilbert space H,
then TT* = T*T. The result will be proved in two steps.
(i) Let x E H and a scalar
II TtIC 1I1=(Tx, Tx) =(K, T* (Tx,)=('(, (T*T) (x»
= (x, (Tf*) (x» (T*x, T*x)=11 T*x 112
This ~ \I Tx 11=11 T·x II. •.. (1)
(ji) To prove T -aI is normal
(T-al)*= T*-(a/)* =T*-aI·=T*-iiJ
(T-al)(T-al)*={T-al) (T*-tii)
=TT*.:aTI-aIT*+auI2
= IT* -a r
aT* + aul .. (2)
(T-:lI)*(T-all=(T*-aJ) (T-a/)
=T*T-aT*I-uIT+aaI2
=T*T-aT*-ar+aii, ... (3)
Equating (2) to (3) and noting Tr* -=T*T, we get
(T-al) (T-a/)*=(T-aI,* (T-aI).
This ~ T -aI is normal.
(iii) Now we can replace T by T-AI in (I).
Hence II (T -AI) (u) II = II (T -Al)* (u) II
or II Tu -AU II = IIlT*-M I (u) II
or II Tu-,\u II = II T*u-fu II
FINITE DIMENSIONAL SPECTRAL THEORY 419

This declares that

II Tu-Au \I =0 <=> • T*II-AU H =0
er Tu =A.u ~ T*u=Xu.
Theorem 6 If T is a normal operator on a Hilbert space H.
then eigen veclors of T belonging to distinct eigen values are
orthogonal. [Kanpur 1988; Meerut 88]
Proof. Suppose T is a normal operator on a Hilbert space
H(K). Let
Tx=oU, Ty=p.y where p.-:FA.
To prove xl.y
Tx=).x, Ty=p.y => T*x=Ax, T*y=j;).,
by Theorem 5.
). (~, y)=(Ax, y)=(Tx, y)=(x, T*y)=('IC, p-y)
=p. (x, y)
or (.\- p.) (x, y) =0
giving <x t y)=O,
this => xl.y.
Remark. The above problem can also be expressed as
follows:
If Ml and M. are eigen spaces of a normal operator T relative
to two ei!!en values ).10 ).2, then Ml and M. are orthogonal.
Theorem 7. Supp',se T is a normal operator on a Hilbert space
H(K). Let M be an eigen space of .\, where)' is an eigen value of
T. Then M and Ml. both are invariant under T.
[Kanpur [Link]. F. Topology 1977, Linear Algebra 78]
Proof. We have
M={x E H : Tx=).x}
). E K, x EM=> Tx=).x.'\x E M
as M is a subspace
=> Tx E M
:. M is invariant under T.
T is normal => Tr*-T*T
T [T* x] =(17'*) (x) -(T*T)(x)-T*(Tx) = T*(.\)C)
=).T*x
.or TlT*x]=A (T*x).
420 FIN1TE DI\fENSIONAl SPECTRAL THEORY

By def. of, M, this => T*x E M.

Thus ¥ x EM=> , *x E M.
This proves M is invariant under T*.
Now M is invariant under T*
=> M 1. is invariant under (T*)* = T
(Refer remark written just below this theorem)
~ M1. is invariant under T.
Hence the result.
Remark. LeI M be invariant under any linear operator T so
that
x EM=> Tx EM.
¥
To prove M1. is invariant undt-r T*.
Let x E M1., Y E M. then (x. y)=O.
By def. of adjoin{.
(T*x. ?)=(x. Ty)=(x. YJ)=O
where Yl=Ty E M
[For x E M1., YI ~ M => (x. YJ)=O]
Thus (T*(x), y)=O
Also Y EM.
Hence tbis declares that T* x E M L.
Thus X EM1. => T*x E M1..
• M.l. ts T*-invariant.
Tbeorem 8. (Spectral Theorem for normal operator). Let T
be an operator on a finite dimensional Hilbert space H. Let
"1>"a..... "" be distinct eigen values of T, let MiJ M 2 • .... Mil be
their eigen spaces; and let Pl. Pt , .... Pn be the projections 011 these
eigen spaces. Then the following statemen!s are equivalent to one
another:
(/) M/s are pairwise orthogonal and .'pan H.
(ii) P;'s are pairwise orthogonal,
n n
I=E. PI, T=E "IPI,
I-I 1_1

(iii) Tis numal. [Meerut 1988 ; Kanpur [Link]. F. 83. 79. 77!
Proof. (i) => (ii). i e .• suppose (i) is true.
To prove (ii).
By assumption. H=M 1(£; M 2 (fJ ... (fJMn •
FINITE DIMENSIONAL SPECTRAL THEORY 421

Hence any x E H is uniquely expressible as

n
X= E XI with XI E M, ¥ I.
n-l

P,'S are projections on M,'s and Mt's are pairwise orthogonal,

~ P;'s are pairwi!>e orthogonal ... (1)

=> P;Pj=O=PjPI for l::I=j.

P; is projection on M; => PI(X)=XI ... (2)

n
This => 1= ~ PI ... (3)
I-I

Tx=T ( i
I-I
XI )' = i
'-1
TXI as T is linear

n
= ~ A;xl as XI E MI => TXI=}.lx,

.. (4)

From (I), l3) and (4), the required result follows.

To prove (ii) => (iii), i.e., suppose (ii) is true.
To prove (iii).

Tt=[ f AIP,r = i: ~I PI*= 1: -"I PI

as PI is projection => P,=P,*=P;I, ... (5)

Tr*=( 7' A;P) ( fx,P/)-f f AlxjPIP,.

(ii) => Pt's are pairwise orthogonal => P,P,=O for ;::I=j.
:. TT*=E '\/5:; P;P,=E 1'\1 \2 P,2=E I '\; II Pi by (5).
; ; I

Thus TT*=E I Ai \2 Pi'

t
422 FINITE DIMENSIONAL SPE'TltAL tHEORY

Similarly T*T=E I A, \' PI'

,
Hence TT*=T*T, showing thereby by T is normal.
To prove (iii) => (i), i e., suppose T is normal
To prove (i).
T is a normal operator on Hilbert space Hand M/s are eigen
spaces of T.
=> M/s are pairwise orthogonal, (by Th. 6). ...(6)
• P,'S are pairwise orthogonal projections on M/s,
(Refer Theorem 39 of Chapter 19)
From (6), the first part of the r~sult is established.

Write
\

Then M is a closed subspace of H associated with projection

n "
P= 2' P, s.t. P,P,=P;P,=O for i¥=j.
'-1 .
T is a normal operator on H, M, is eigen space of T
=> M, reduces T 101 i, by Theorem 7.
Also P, is a projection on M,
• TP,= PIT of i, by Theorem 9.
TP=T ~ P,=E TP,=~ P,T=(~ P,) T=PT
I , , I

or TP=PT, also T IS operator on M.

Hence M reduces T, by Theorem 38, Chapter 19.
Hence M and MJ. both are invariant under T.
Let T' be the restriction of T to MJ.. Then T' : ML-+Ml. S.t.
T'(x)=T(x) ¥ x E MJ.. ... (1)
Let x be an eigen vector of T' relative to eigen value A, then
T' (X)=Ax. ...(2)
But T' is defined on MJ.,
(2) => x E M_L
=> T(x)=T' (x), by (I)
using (2), we get T (X)=Ax
=> x is an eigen vector of T.
:. Every eigen vector of T' is also an eigen vector of T. But
T has no eigeo vector in MJ. as all the eigeo vectors of T are in M.
Also M n MJ.={Ol.
Consequently T' : ¥J.~M.J. is an operator with no eigen
vectors and so no eigeo values. :. M~={O}.
FINITE [Link] AL SPECtRAL lHEORY 423

This Q H=M as M(£;Ml.:::aH

=> M,'s span H.
"
Also H=M(£;Ml.=M= EM,.
/-1

This Qo M,'s span H.

Problem J. Let T be a normal operator on a finite dimen-
sional Hilbert space and let p be a polynomial. Show that if A is
an eigen value ot T, then p~A) is an eigen vahle of p(n.
[Meerut J987]
Solution. Let T be a no, mal operator on a finite, dimensional
Hilbert space H.
Let Tu=Au, i.e., A is an eigen value of T.
I. To prove that PlT) is a normal operator. Prove this as in
Ex. ) 3, page 398.
II. To prove that peA) is an eigen value of p(T).
p(x) =ao +a,x +' a.x l + ... +amxm
PlT)=Qol+a1T+al T1 + ... + a,,,Tm .. (5)
Tu=Au => T'u= T. Tu=T.(Au)=A(Tu)=[Link]
=> T 2u=A2 u
Generalising tbis, ~e get T"u=A"u.
In view of this (5) gives
plT)u=aolu +a1Tu+ollT2u+ ... +arnT"'u
= 0ou+a1Au+ a.;\'u+ +arnAmu
=(oll+alA+a.;\2+ ... +amAm)u
_p(A)u
or, p( T)u=p(A)u.
This => p(;\) is an eigen value of p(T).
Problem 2. Let T be a normal operator on a Hilbert space
H.
Let {A1' A2 • .... An} be the set of eigen values of T. Suppose
M1-={x E H: ,Tx=A,x}. Prove that M,'s span H.
[Meerut 1986]
Hint. Here prove (iii) => (i) of Theorem 8.
Problem 3. Let T be 0 linear operator on a [Link]/
complex Hilbert .~pace V(K). Then T is normal iff the adjoint T*
is a polynomial in T. [Meerut 1982, 79)
Solution. Step (i). Suppose T is normal.
To prove tbat 1* is polynomial in T. .
424 FINIfE DIMENSIONAL &PECTRAL THfOHY

Since T is normal and hence T is expressible as the spectral

n
solution namely T= 1: c/Ej where c/ E K.
i-I

and E/ is perpendicular pairwise orthogonal projections so that

E j *=Ej =E/2.
T is normal => TT*=T*T

T*==(
/-1
f; CjE/ )* = f; CiEi*=
i-I
i: CiEI
i-I

n
or T* = }; CiEi'
/~1

We know that each E/ is a polynomial in T. Consequently T*

is a polynomial in T.
Step (ii). Let T* be a polynomial in T so that T*=/(T}.
To prove that T is normal.
TT*=T/(T}, T*T=!<T). T=T/(T)=TT*
T*T=TT* so that T is normal,
Theorem 9. The spectral resolution 0/ a normal operator on a
fi nite dimensional non-zero Hilbert space is unique. [Kanpur 1991,
'90]
Proof. Let T be a normal operator on a finite dimensional
non-zero Hilbert space H.

Let ... (1)

be a spectral resolution of T, where P;'s are pairwise non-zere

orthogonal projections S.t.

..,(2)

and "/s are distinct complex numbers

Step t. W ~ shall show that '\;'s are eigen values of T.
There exists a non-zero vector x in the range of Pi S.t. PiX =X.

TX=( ,f;
J-I
AiP,) X= f
1-1
>'JPjPI (x)

using the fact that PiP 1= 0 if i ¥.j,

we get TX=>.;P;2 (X}=-.\i PiX as Pil=p/
=.\/x or Tx=.\x
 FINITE DIMENSIONAL SP CfRAL THEORY 425

This proves that A/ 'S an eigen \'alue of T ¥ i.

Step II. Write A ={Ah A2 , •••• A,,).

Aim. All the eigen values of T are precisely members of T.
If possible, let A be an arbitrary eigen value cf T s.t. Tx=Ax.

Then Tx=A lex) or ( £ AiPI )

i-I
X=A ( i
i-I
P; ) x

" II
or E AiPI(X)=A E PI(x)
i-I I-I

or };" A; p)p/ (X)=A E" PjP; (x)

'-1 1-1

or >.) Pl (X)=A P/ (x)

or A,P, X=AP, (x) or AjX='\X
or (A}-A) x=O. But x~O.

•• >',=A. l-ience every eigen value of T is a member of A.

Thus we have proved that in the spectral resolution (I) of T,
the scalars J..;'s are dIstinct eigen values of T.
Step III. Remains to prove the uniqueness of the solution (i).
til
Let T= E be another spectral resolution of T, where
.-1 I)(.;QI ••• (3)

ext's are distinct eigen values of T and Qt's are pairwise orthogonal
projections. We can also write (3) as

= E AI 2P,"= 1: A,2P;
f ,

or T2= 1:" A,2P,' ... (5)

'-I
n
This:;> TI= 1: A/l Pi ... (6)
1_1
426 FINn E DIMENSIONAL SPiCTRAL THEORY

Let get) by any polynomial \\ ith complex coffident in the

complex variable I. Taking Irnear combination, we get
n
g(T)= E g ('\i) PI
/=1

p; is a polynomial s.t. Pi ('\,)= { ~ if i=j

if i =1) .. (7)
Taking p; in place of g,

Thus p,(T)=Pj •
If we take
po (t)= -<1-'\1) (I - '\2) ..• (I+'\n)
I (,\/-,\d (,\;-,\2) ••• (,\/_oA n )
then p; sat isfies (7).
If we apply this discussion for Q/s then we shall get Q/ ""' p,(T)
for each i.
Therefore p/= Q; for each i.
Hence the two spectral resolutions of T are the same.
PlOblem 4. If T is an operator on a finite dimensional Hilbert
.Ipace If, then prove the following statements.
(0) V is singular ¢> 0 E (J (T), lKanpur 1984,81, 7S]
i.e. TOis s ngular (/fO is the spectrum ofT.
(b) If T is non-~ingular, then
A E (J (T) ::;.. ,\-1 E (J E (T-l) [Kanpur 1984]
(c) If A is [Link], then
(] (ATA-1)= (J(T)
(d) If A E (] (Tl, and if p is any polynomial, then
p (A) E a [p I T»). [Meerut 1982, 80]
Hence (J (T) denote~ tire spectrum of T, i.e. the set of all eigen
values of T.
Solution la)o T is a sipgular operator
~ 3 x E H S.t. Tx=O where x.;iO
¢> Tx=Ox ¢> 0 is eigen value of T

¢> 0E (J (T).
(b) T is non-singular, " E (J (T)
:::. ,\ is eigen value of non-singular operator T
~ A=FO eigen value of non·~ingular T
FIN I 'E DIMf:!I"SIONAL SPECTR \L THEORY 427
~ 3 x E If S.t x#O and Tx=Ax
~ X=T-l (Ax)=AT-l (x) :;.. A-I X=T-l (x)
=> A-I E a (T-l).
(c) Let S=ATA-], Then
S-Al= ATA-I-A1AA-I=ATA-I_A(AlA·-I)
=A [T--AT] A-I
det(S-.\l)=det A det (T-Al).det A--I
=det A. det A-I. det (T - AI)
=det [Link] (T -AI) as det A det A-l=det AA-l
= I det (T- AI)
or det (S~,u)=det (T-A1)
:. det (S-,u)==-O ,'", det (T-Al)=O .. (1)
But det (S-M) -=- 0 ," A E a (S).
:. By (1), A E a (S) <=> A E a (T)
or a (8)=a (T)
or a (ATA-l)=a (T).
(d) Let ,\ E a (T), then 3 non-zero vector x S.t.
Tx=Ax .•. (2)
(1) T 2x= T (Ax)=ATx=A AX=,\2x
:;;10

~ 1'2x =A'x ~ T (Px)=T (AlIX)=.VTx

=> T3 X =A 2 . Ax ~ T3 X =A 3X.
Finally, TX=AX, T2 X=A 2 )(., T8X=AIX.
Generalising, we get TnX=A ..X.
Let p (1)=aO+alt-~ a2(2+ ... + 0 ,,1" be a polynomial in t of
degree n \\-here an¥O. Then
p tTl x=(ool +a1 T+aaP+ ... +anP) x
==aOx+a 1 Ax+a2 A2x+ ... +a.A n x
= (au+olA+02A2+ ... +anA") x=p (A) x
or p (T) xc=p (A),X with r,,=AX.
This proves that
,\ E (J (T) => P (A) E (J [p (T»).
Problem 5. Let T be any operator 011 a fillite dimem;onal
Hilbert space Hand N 0 normal operator on H. Shmv that if T
commutes with N. then T alw commutes with N*.
Solution Let T commute with normal operator N, then
NN*=N*N ... (J) and TN",-NT .. (2)
To prove tbat TN*=N*T
I. We claim TN'''=NmT ... (3)
428 FINITB DIMENSIONAL SPECTRAL THlORY

The result is true for k = 1, by (2). Let (J) be true for k =m,
then TN'" = Nn'T. ...(4)
Now TNtn+l=(TN"') N=(N"'T) N, by (4)
=Nm (TN, =Nm(NT) = Nnl-tlT
or TNm+l=Nm+l T which proves that the result is true for
n=m+ 1 if it is true for n=m. Hence tbe result (3) is true by
Induction.
Let p (t)=a O+a l t+a.t 2+ ... +ant" be aby polynomial with
complex coefficients.
II. Tp (N)=T [ao/+alN+azNI+ ... +anNn]
=a"T + alTN +a~TN2+ ... +anTNn
=aolT +alNT+a2 N2T+ ... +anNnT, by (3)
=(oo/+a l N+a 2N2+ ... + a"N") T
or Tp (N)=p (N) T. ...(5)

III. Let N = 1:" Ai Pi be the I>pectral resolutIOn of the normal

;-J
operator N, where PI is projection and so that
p/2=PI*=p/.
Then N* = (I' A/PI)· = 1: )./p/* =1: ~/P/
I I

or N*=}; ~jPI' ... (6)

i

But each operator Pi is a polynomial in N.

Hence, by virtue of (5), N* commutes with T.
Problem 6 Let T be a normal operator on a finite dimens;onal
Hilbert space H with lipearllnl AI. A2, ... , Am. Then prove the/ollowing
statements
(a) Tis Sl:'lj adjoint ~. each '\; is real,
(b) T is positive "c- each "I
~ 0,
(c) T is unitary «> I A; I= I ¥ i.
c[Kanpur 1989, 91; Meerut 82J
Ii

Solution. Let T= I' "iP, ... (1)

1=1

be the spectral resolution of T, where AI. "', An are eigen values of.
l' and PI, P 2 , ,Pn are non-zero pairwise orthogonal projections.
This,* P?=P;*=P; ¥ i
and P;P,=O for i~j.

Here we also ha\e I'" P;=T. ...(2)

1-1
FINITE DIMENSIONAL SPECTRAL THEORY 4
T*=(E A;P;}*=E ).IP'" -.c=E )./p;
/

or T"'=E ~Pi ... (

(a> let T be self adjoint, then T = T"'. ... (
Putting the values from (I) and (3) in (4),
we get E (i;P;) = E AlP"
i

or E ().;-A/) P;=O.
i

Multiplying both sides by P j and nothing that

p,p) ='1 ¥ i except i=j, we get
C~/-Ai) P/'=O
or (~J-'\~) P/=O
or ("i,-A;) Pj=O or '3:,-A j =O
or ,\j= 1./.
This ==- Aj is real ¥ j.
Conversely Suppose that A; is real [Link] i, then A/= )./. . •. (5
Aim. T is self a~joint.
By (3) and (5), T'" =E AiP/,
By (I) and (6), T"'=T, showing thereby T is self adjoint.
(b) Let x E H be arbitrary.
<Txt x>=<Tx, Ix>=<E Ai Pi x, E Pi x>
J
=2-' E .\/ < p/ x, Pj X >
i i

= E A/ < Pi'" p/ x, x> ... (7.

/' j

• {Pil for i =- j •
But P; Pj=PjP/= 0 &' " (Pj =PJ ).
lor ':1=J
Usin~ this in (7), we get
< Tx. x >=E A/ < P,2 X, X >
I

But <Pit x, x>=<p/ x, Pi'" x>=<Pi x, PiX>

=\1 PI x 112.
Using this in (8), <. Tx, x>=E A; \I Pi x \1 2 • , . .. (9)

Let T be positive, then < TJI, x> ;;:: o. .•• (10)

Aim. AI ~ 0 ¥i.
By (9) and (10), E A. 1 PIX il 2 ~ O.
, ••• (11)

Let us suppose that x is in the range of Ph then

p.x=x, Plx=O ¥ i except i= 1.
4JO FINITE DlMLN310NAL SPECTRAL THEORY

Then (II) gives Al ~ O.

Similarly we can prove A2 ~ O.
Gelleralising, we get Ai ~ 0 ¥ i. ... (12)
Conversely suppose that Ai ;;;:. 0 ¥ i.
Aim. T is positive.
By (9) and (12), <Tx, x> ~ O.
Th is ::> T is positive.
(c) TT· = (.r Ai P;) (E)./ PJ)
i

= E AI ~'J PI Pj=E Ai ii P I2
I' } I
or TT*=E I AI 12 PI. ... (13)
I
T is unitary -¢> TT*=I
-¢> E I '\1 \2 P I = E Pi> by (2) aDd (13) ..
I

~ E (IA/12-1) PI =,0.
I

Multiplyi"g both sides by PI and noting that PiP} =0 ¥ i

except i =--j. we get
(I ",\2_]) PJ=O
-¢> I A} \2-1=0 {:> IAj 1=] vj.
Exercises
1. (a) Define eigen value. eigen vector, eigen space and spectral
resolution of an operator T on a Hilbert space H. Prove that. if
A be an eigen value of an operator T, on a finite dimensional
vector space H. then I A I ~ II Til. [Kanpur 1988J
2. (a I Define a Banach Algebra. Show that the set G of all
regular elements in a Blnach Algebra is an open set. Further
show that the mapping x-*x- 1 is continuous on G. [Kanpur 1989J
(b) For an element x in a Banach Algebra A, prove that
a (x) is non-empty. [Kanpur 1988]
3. Define spectrum and spectral resolution of an operator T
on a non-zero Hilbert space. Prove that eigen values of an
operator T constitutes a non-empty finite subset of the complex
plane and the number of points in thi~ set does not exceed the
dimension of the space. [Kanpur 1987]
4. (a) Define a Banach Algebra A. Also define regular and
singular points in A. Prove that every elem:nt x for which
II x . 1 II < ] is regular and its inverse is given by the formula
00
x- 1 =1+E (I-x)". [Kanpur 1989]
....1
FI"ITE DIMEN~JON"L SPECTRAL THEORY 431

(b) Define spectral radius rex) of an element x of a Banach

Algebra. Prove that
lim
r (.x) =
n-+oo \I x' I lin . [Kaopur 1987]
5. (a) Define eigen vector, eigen value, eigen space, spectrum
and spectral resolution of an operator T on a non-zero Hilbert
splce H. Prove that jf A be an eigen. value of a operator T on a
finite dimensional Hilbert space H, then I A I ~ II Til.
(b) Let T be a normal operator 00 a Hilbert space H with
spectrum {At. A2 , ••• , Am}. Use the spectral resoJution of T to prove
th ~ following statements.
(i) T is ~elf adjoint -= each Al is real.
(ii) T is positive ¢> Ai ;?- 0 for each i,
(iii) T is unitary -= I A/ I= I for each i. lKaopur 1986]
6. (a) Define eigen value, eigen vector, eigen space and
spectral resolution of an operator T on a Hilbert space H.
(b) Prove that the spectral resolution of a normal operator
on a finite dimensional non-zero Hilbert space is unique.
[Kanpur 1985]
7. (a) Prove that two matrices in A. are silT ilar';:> they are
the matrics of a single operator H relative to different bases.
(b) D~fine the spectrum (} (T) of an operator T on Hand
prove that
0) T is singular.;:> 0 E a (T).
(ii) If T is non-singular, then A E a (T) ¢> >,-1 E a (TI).
{Kaopur 198']
8. (a) Define regular and singular elements of a Banach
Algebra A Prove that every element x for which II X -I II < I is
reg'Jlar and the inverse of such an element is giver) by the formula
""
X-I = 1 + E (I-x)n.
n-l

(b\ For an element x in the Banach Algebra A, prove that

a (xl is non·empty. [Kanpur 1984]
9. State and prove the spectral Theorem. [Kanpur 198J]
10. (a) If S denotes the set of singular element of a Banach
Algebra and Z denotes the set of all topologkal divisors of zero
then prove that
(i) Z C S.
(ii) the boundary of S is a subset of Z.
(b) Prove the following:
(i) If r is an element of R then 1 -r is regular.
432 FINITE DIMENSIONAL SPECTRAL THEORY

(ii) If t -xr is regular then 1 -rx is also regular.

. [Kaopur 1983]
11. Let G and S denote the set of regular elements and the
set of singular elements respectively in a Banach Algebra A. Prove
the following:
(a) G is an open set.
(b) The mapping X~X-I of G into Gis homcomorpliism of
G onto itself.
(c) The boundary of S is a subset of the set of all topologi-
cal divisiors of zero. [Kanpur [Link]. F. 19821
12. <a) Prove that a linear operator T on a finite dimensional
Hilbert space H into itself is singular if and only if 0 is in the
spectrum. of T.
(b) Desctibe the group algebra of the additive Z of all in-
tegers. Prove that it is a Banach algebra. lKaopor [Link]. F. 19811
13. (a) If S is the set of all singular elements of a Banach
algebra, prove tQat every boundary element S is a topological
divisor of zero.
(b) Derive the formula for the spectral radius of an element
of a Banach Algebra. [Kaopor [Link]. F. 1981]
14. <a) Define the set S of singular elements and the set Z
of topological divisors of zero in a Banach Algebra.
Prove that Z is a subset of S and that the boundary of S is a
subset of Z.
(b) Define the spectrum of an element of a Banach Algebra
and show that it is ~ever empty. [Kaopor [Link]. F. 1980j