An Introduction to

LINEAR ALGEBRA

V. KRISHNAMURTHY
Professor of Mathematics

\
V. P. MAINRA
Assistant Professor of Mathematics
and

J. L. ARORA
.Assistant Professor of Mathematics
All at the Birla Institute of Technology and Science,
Pilani

lm
AFFILIATED EAST-Wl!ST PRESS PVT LTD
NEW DELHI · MADRAS
 Priae Rs 22.50

No reproduction In any form of thi• book. in whole or in part

(except for brief quolations in critical articles or reviews).
may be qlade without written permission of tbe publlshen.

Printed in lnclia at l l . _ Pri1110n•• New Delhi-110020

..,_bUsbed t,,y Afftlia_. Eat-West Preis Prlftte Unaitccl
104 Ninnal TOWeT. 26 'Barakbam'ba Road. New Delhi-110001
To C>u.r Pa.rents
Contents

PUFACI ii
1 SETS AND FUNCTIONS 1
1.1 Sets 1
1.2 Operations on Sets 5
1.3 Relations 8
1.4 Functions 10
1.S Binary Operations 11
1.6 Algebraic Structures 20
1.7 Operatio~s on Functions 24
2 VECTORS 33
2.1 Space Coordinates 33
2.2 Vectors-Addition and Scalar Multiplication 31
2.3 Dot Product of Vectors 48
2.4 Applications to Geometry 55
3 VECTOR SPACES 63
3.1 Vector Spaces 63
3.2 Subspaces 69
3.3 Span of a Set 14
3.4 More about Subspaces 11
3.5 Linear Dependence, Independence BS
3.6 Dimension and Basis 95
4 LINEAR TRANSFORMATIONS 107
4.1 Definition and Examples 101
4.2 Range and Kernel of a Linear Map 114
4.3 Rank and Nullity 118
4.4 Inverse of a Linear Transformation 122
4.5 Consequences of Rank-Nullity Theorem 111
4.6 The Space LC,U, Y) 129
4.7 Composition of Linear Maps 133
4.8 Operator Equations 1J8
4.9 Applications to the Theory of Ordinary U1111r .
Ditrcrential Equations 141
 Viii / CONTl!NTS
5 MATRICES 141
5.1 Matrix As4iociat~d "ith a Linear Map 148
5.2 Linear Map Associated with a Matrix 154
S.3 Linear Operations in M,,., ,. 158
5.4 Matrix Multiplication 16.3
s.s Rank and Nullity of a Matrix 174
S.6 Transpo.,e of a Matrix and Spec ·al Type4i of Matrices 180
S.1 Elementary Row Operations 185
S.8 Sy~tem~ of Linear Equation, 195
S.9 Matrix lmer~ion :00
6 DETERMINANT5 203
6.1 Definition 203
6.2 Fundamental Properties of Oelerminanb 207
6.3 Proof, of Theorems 212
6.4 Cofactors 2 /6
6.5 Determinant Minor., and Rank of a Matrix 214
6.6 Product of Determinanb 227
6.7 Applications to Linear Equations 230
6.8 Eigenvalues, Eigenvector& 233
6.9 Wrom,kians 238
6.10 Cro<i, Product in VJ 241
7 MORE MATRIX THEORY 250
7.1 Similarity of Matrice!-. 250
7.2 Inner Product Space& 258
7.3 Orthogonal and Unitary Matrices 264
7.4 Application to Reduction of Quadrics 266
APPBNDJX Ordinary Linear Differential Equations 276
ANSWERS TO PROBLEMS 297
INDEX 325
 Preface

In Indi.in universities the emphasis at the undergraduate level has

been much more on calculus than on linear algebra and matrix theory.
Today, however, the need for linear algebra and matrix theory as an
essential part of undergraduate mathematics is recogni5ed. The Bi•
national Conference in Mathematics Education and Research (June 1973)
recommended Elementary Linear Algebra as a compulsory course for aJJ
students at the undergraduate level. It has since been generally agreed
that before a studept begins to specialise in the discipline of his choice-
whether it be mathematics, science, engineering, social science, or
management-he must be exposed at least once to both calculus and
linear algebra; such an exposure will familiari!e him with the concepts
and techniques of continuous mathematics (calculus), and the concepts,
method&, and logic of modern discrete mathematics (linear algebra).
This book is the outcome of a planned effort to teach linear algebra as
a second course in the mathematics curriculum introduced at the under-
graduate level several years ago at Birla lmtitute of Technology and
Science (BITS), Pdani The students who take this cour!e have had a
semester of elementary calculus and analytical geometry. Ho\\ever, a
knowledge of tlie fundamental properties of continuous and differentiable
functions in terms of their addition, scalar multiplication, and multipli-
cation is sufficient for an understanding of this volume.
The treatment throughout is rigorous yet lucid. The fact that the
majority of students who would use this text may not ultimately become
mathematicians or physicists has ·not inhibited our developJrent of the
subject. We strongly believe that present-day users of mathematics,
instead of being content with a hybrid of mathematical tools and
gymnastics which merely graze the subject, should delve deep .bY. training
in concrete matter-of•fact arpments. The conceptual framework of linear
al9ebra and matrix theory provides the most efficient means for this
trainina; for, in one sense, matrices and linear equations form a concrete
foundation, and \ICCtor spaces and linear transformations give the flavour
of the abstract arandeur of modern mathematics. At the same time, as the
freshmen we are addressiq may not have had any grounding in abstract
X / PRl!f'ACB

mathematics, we have made a special effort to smoolhtn their fir~t

encounter with methods of proof.
Theorems are proved in full (the end of proof is indicated by I>, except
in rare cases \\here they are beyond the scope of the book. In these
instances the student is suitably instructed. Where certain consequences
of earlier results are stated as FACTS, the ~tudent will fir.d he has been
sufficiently equipped in advance to prove them himself. The large number
of worked.out examples which have been wo,c'l info the text help the
student to move back and forth from the concrete to the ab!itract. The
sets of problems-numerical, objective, anJ theoretical -interpolated at the
end of almost every article are a drill on the text. Anf>wers to the
numerical problems appear at the end of the book; the obje~tive questions
which are of the "true-false" type are intended to help the student in a
self•assessment of his conceptual understanding of the subject under
study.
Chapter 1 deals with sets and functions and gradually introduces the
language of modern mathematics. A tead1cr may adju~t his puce in this
chapter to suit the standard of his class. Algcbr.1ic blructures, such as
groups, rings, and fields, have been discussed only to the extent needed.
Chapter 2 provides the concrete geometric structure of 2· and 3·dimen-
sional vector spaces. It leads the student to the problems of geometry
through vectors and prepares the ground for Chapter 3 which gets into
the essence of the 1,ubject. Here the theory of vector spaces, and the con-
cepts of linear dependence and linear independence, <limem,ion and bac,1b
are treated elaborately. Though infinite-dimensional vector spaces are
also considered, the emphasis throughout is on finite-dimen~ional vector
spaces.
Chapter 4 aims to familiarise the 1,tudent with the fundamental pro•
perties of linear transformation~. The rank•nullity theorem and its
consequences are presented in deta·I. The theory developed so far is
applied to operator equations and, in particular, to differential equations.
This application discloses that the solution space of the n•th order normal
homogeneous linear differential equation is an n-dimensional subspace of
the space of n-times continuously differentiable functions. The further
application to the theory of ordinary linear differential equations is detail•
ed in the Appendix. However, we have not attempted to make the
treatment of differential equations self-contained.
The elaborate build•up on vector spaces and linear transformations
begins to pay dividends in Chapter S which starts by establishing the link
between linear transformations and matrices. In our experience. the
welter of mathematical detail on matrices in this chapter is easily assimi-
lated by the student because of the knowledge of linear tunsformations
be has pined in Chapter 4. He is thrilled to see that the elementary
 PREfACI! / xi
(apparently trivial) row operations on matrices finalfy result mthe solution
of linear equations in all their ramifications. Naturally, the chapter ends
with matrix inversion.
Now the student is ready for determinants, presented in Chapter 6.
When he comes to determinant minors and the rank of a matrix he realises
the importance of the emphasis in Chapter 3 on the concept of linear
dependence and linear indcpendem:e. The theorem giving the connection
between the rank of a matrix (already defined in Chapter S by means of
independence concepts) and the order of its nonzero minors is the crux of
the content here. The ea~e with which it is proved justifies the eJforts
taken in the development of the subject in earlier chapter1,. Applications
to linear equations, and a brief account of eigenvalues and eigenvectors of
matrices, Wronskian~, and the cross-produi:t in V8 give an idea of what
dcterminanb can do.
Chapter 7 give, a glimp"e of the theory of orthogonal and unitary
n;atm.es, t,1milanty tra111,formation1, and their application to the geometry
of quadrics. When the student reaches this chapter, he easily recogni&es
the connection be.tween linear algebra and geometry.
The student o;hnuld guard against conceptual errors of three types :
'finite dimc:n,ion' vcr,;uc; 'infinite dimen1tion'; 'real scalar' versus 'complex
scalar'; ,md ·non-empty set' ,e,~u1, 'empty set'. When in doubt regarding
hypothe,e!> he ,hou\d invoke what may be culled an 'emergency axiom' :
The r.uitahle ahern.iti,e in each relc,ant pair(~) is mcluded as an additional
hypothesis.
The text can be adapted to suit different curricula : as a one-year
coun.e runmng three hours a \\eel; as a one-1teme1,ter courr,e runnin~ five
or six houri; a week (as is the practice at BITS); or, by a judicious selec-
tion of topic,;, as a onesemcster courr,e running three hours a week. It can
be m,ed during .my year at tl1e undergraduate le,·el or as part of a first
course in linear algebra at the postgraduate le,el, if no matrix theory has
been done till then. However, a,; the topics have been arranged sequen-
tially, any student who wishcc; to change the order of topics will need
guidance.
We wish to thank Dr. C.R. Mitra (Direc!or, BITS) and others at BITS
who encouraged our effort~ in writing this book under the Course
Development Scheme and provided all the necessary ar,sistance in dupli•
cating and cla,!>•tec;ting its earlier ver~ions over a period of three years. To
the BITS students-about 1200-of these three years, we are al'it- indebted
for their li':ely rer,ponse to our experiments in pedagogy.
We are grateful to Messrs. 0. R. Verma and B. L.,Soni who patiently
typed the several drafts; Mr. A. P. Ra~togi who drew the illuc;trationo;; and
the National Book Trust, India, for i.ubsidising the p11blication of the
xii/ PRIPACB
book. To Mrs. J. Fernandes of East•West Press, our appreciation and
thanks are due for a thorough editing of the manuseript.
Our last words of atrectionate gratitude are reserved for our wives and
family members who continually cheered our years of effort at the project.
We shall welcome all suggestions for improvement of the book.

V. Kri&hnamurthy
Pilani V. P. Mainra
September 1938 J. L. Arora
Chapter 1

Sets and Functions

We shall begin with a discussion on •sets'. Immediately, we get into

the first requirement of mathematics, viz. proper definitions for
technical terms. A major concern of mathematics is precision not only
in the calculations of quantitative information, but also in the communi•
cation of thought. This is why importance is given to definitions in
mathematics. In this book we shall come across a large number of
definitions. Each definition will introduce a new technical term or a new
concept in terms \Of preceding definitions. The very comprehension
of a definition may often depend on the logical development of the subject
up to that point.
However, the first technical term, namely, 'set' will be introduced
without a precise definition. The reason is obvious : The moment we
attempt to define 'set', we get into words such as •collection' or 'aggregate',
which themselves need to be defined. We have to draw the line some-
where!

I.I SETS
The meaning of 'set' is intuitively understood as a well-determined
collection of objects, called its 'members' or 'elements'. The objects
(members or elements) are said to 'belong to' the set or to be 1in' the set.
Here all the words in quotation marks are taken to be undefined terms.
To illustrate the meaning of •set', let us consider some examples.
Example 1.1 The collection of the lhree boys : Arun, Mohan, and Ram.
Example 1.2 The collection of the three symbols : 6., □ and a.
Example 1.3 The collection N of all natural numben.
E,ca,nple 1.4 The collection Z of all integen.
1?%t1111Ple 1.S The collection Q of all rational numben.
&am,le J.6 The collection R of alt nal numbers.
Eample 1.1 The collection of all the past presidents of Dldia.
B:u,mple 1.8 The collection of all the tint year 1tudent1 of Birla Institute
of Teohnoloo and Science (BITS).
 l I sm AND FUNCTIONS
Example 1.9 The aggregate of the livina men in the world whose height
exceeds 2 metres.
Example 1.10 The aggregate of the roots of the equation x'' - 1 - 0.
Example 1.11 The aggregate of the integers that leave a remainder 2
when divided by S.
Example 1.12 The group of cricketers who were out for 99 runs in a
test match.
Example l.JJ The coJJection of all positive prime~.
Example 1./4 The collection of derivatives of all ord,rs of the function
et-.
Example 1.15 The collection C of all complex numbers.
All these are examples of sets. We can construct i-everal such exam-
ples. Let us now consider two collecuons which are not sets :
. (i) The collection of some natural numbers.
(it) The collection of the politicians of India.
In (1) it is not clear which numbers are included in the collection.
Whether the number 2 is in the collection or not cannot be answered
without first explaining the word 'some'. Again, in (1i) the question
whether a specific person is a politician or not would get d1trerent responses
from different persons. Thus, collections (i) and (ii) arc not 'weJl deter-
mined'.
In contrast to these two examples it is worthwhile to analyse Examples
1.12 and 1.13. In Example 1.12 we note that every cricketer is either in
the group or he is not in the group. We do not have to check any records
to say this. In Example 1.13, again, either a number is a prime or it is not
a prime. Here it matters little whether it is known that a particular number
ia a prime or not. In other words, it is immaterial whether we can
answer the question : Is this particular object in the given collection or
not ? What matters for specifying a set is to know unambiguously that
only one of the two answers is t,ossiblc : The object' in question belongs
to the aiven collection or it does not bclona to it.
Thus, we can elaborately describe a set as a collection of objects
which is well determined in the sense that, for every object, there should
be only two possibilities available unambiguously: either it belonas or it
docs not belona to the coJlection.
If .4. is a set and" is an element of .4., then we write x E .4. (E is read
as 'belongs to' or 'ia in' or 'ia an eleme11t of' or 'ia a member of'). The
noption of tbia ia denoted by x tie .4. ( te i1 read aa 'docs not belong to' or
'i• not in' or 'is not an element or or 'ia not a member of').
J.n•tead orauch a detailed description of aeta, two kinda of symbolic
descriptiOJla are very oftn used. OU is by listina, if possible, all the
 1.1 sm / 3
-
elements of the set within braces, e.g. sets of Ex11mples 1.1 and J.2 aro
respectively written as
{Arun, Mohan, Ram}
and {.t.i.. □. n} .
The other is by using a dummy element, say x, and uiting the character-
istic properties of x which precisely make it an element of the set. Thus,
the set of elements characterised by the prorerties, say P, (!, ... , is written

{A I x satisfies P, Q, ... }
or {.'t : x satisfies P, Q, ... } .
(' I • and ·:'are reeJ as 'such that'.) Thus, the sets in Examples I.II and
I. 12 may respectively be "ritten a~
{x I xis an integer and x -= Sk + 2 for some integer k}
and { X .• x• is a cricketerh who was out for just 99 runs} •
1n a test mate
Note that in \the first method the order in which the elements arc listed
is immaterial. Thus, {Aron, Mohan, Ram} and {Mohan, Aron, Ram} are
the same sets. We shall make this precise by defining the equality of
sets.
1.1.t Deftnltion Two sets A and Bare said to be equal if every member of
A is a member of B, and every member of Bis a member of A. In
such a case we write A = B. ·
For example, {O, 1, 2, 3} = {2, 1, 0, 3}. Also Z = {x Ix is an in-
teger} = {... , -3, -2, -1, 0, I, 2, 3, ... }.
1.1.2 Convention All our definitions are 'if and only if' (ijf) statements.
For example, if a definition reads 'A triangle is said to be equilateral
if all its sides are equal', we actually mean that 'I\ triangle is said to
be equilateral (ff all its sides are equal'. . .
Jo, view of_ this convention, Definition I. l. I means that two sets A and
B are said to be equal iff every member of .4 is a member of B, and every
member of Bis a member of A ..

SUBSETS
Let A and B be two sets such that every member of A is alao a member
o( B. Then .4. is said to be a subset of B. In symbols we write A C B
('C' is read a■ 'is a subset of' or 'is contained in'). Whenever .4. is. a
subset of B, we also say that Bis a superset of .4.. In symbols we write
B ::::> ..4. (1 :J' is read as 'is a superset of' or 'contain■'). •
Obviously, every set ia a subset (superset) of itself.
 4 / lffl AND FUNCTIONS

Exercise Prove that A .,. B 1/1 ..4 C Band B C ..4.

Thia exercise is immediate and the reader can prove it himself. In
practice, whenever \\e want to prove that two sets A and Bare equal, we
prove both the inclusions, i.e. A C B and B C ..4.

EMPTY SET
Consider the set of women presidents of India elected before
31st December 1974. There was no such woman president. So this set
bu no members in it. Such a set, i.i,. a set containing no elements in it,
is called an empty set or a null ,et. It is denoted by +.
It may be noted here that there is only one empty set. As a clarifica•
tion of this, note that the set {x : x is a real number satisfyins x• + 1 = O}
is also empty and we can write.
The set of all women presidents of India elected before
31st December 1974 = ♦
== {x : x is a real number satisfying x 2 + 1 - O}.
+
The set {x : x is a real number satisfying x1 x - O} consists of only
one member, namely, zero. So this set can be written as {O}. Note that this
is no, the empty set.
Nonempty sets that consist of only a single member are called singleton,.
The set in Example 1.14 is a singleton as itconsists of only one element e".
It would be interesting (at least for cricket fans) to find out whether the
set in Example 1.12 is empty or a singleton or bas more· members than
one.
+
Finally, we note that C A for all sets A. Thus, given a set A, it has
two extteme subsets. One is A itself and the other is +. Any subset
+
of A other than A and is called a proper mbset of A.

Problem Set 1.1

I. Let ..4, B, and C be three sets such that A C Band B C C. Then prove
that.AC C.
2. Let 81, 81, ... , S,. be n sets such that 91 C S1 C ... C S,. and S,. C 81,
Then prove that S1 - S1 = ••• - S,..
3. Prove the exercise pven on pap 3.
4. Determine the set of real numben x 1atisfyin1 the following :
(a) r<1 (b~ z{X + 1) <; 0 (c) ~ +~ ~ 0
(d) r - 4 >0 >
(e) (xi + 2) (.x' - 1) 0 (0 x 1 JC.+ <
5. Determine tbo aet of complex numben s utilfying the following :

(a) I • I <. 4 (b) 1!..±..U - i Cc) I z + ii < l

,-,-:::TT I• - ' I
 1.2 OPBRATIONS ON SETS / 5

(d) z i + 2z + 2z -
S == O
6. Determine all the subsets of the set :
(a) {O, 1, 2} (b) {«, ~. y, 8}.
7. Let X be a set containing n elements. What is the total number of
distinct subsets of X ?
8. Determine whether each of the following statements is true or false
(N, Z, Q, and Rare defined in Examples 1.3, 1.4, 1 5, and 1.6, res•
pectively):
(a) NC Q. (b) {N, Q, Z} C R. (c) NC {N}.
(d) N E {N, Q, Zt (e) R C {N, Q, R}.
9. Given .4 = =
{O, 1, 2, 3, 4}, B -= {x E NI x ~ IO}, C {x E R Ix~ 1O},
determine whether each of the following stateu,ents is true or false :
(a) A C B. (b) BC C. (c) CC .4.
(d) 1 ~ B. (e) -1 E B. (f) -1 EC.
(g) 3 e: B. (h) C C B. (i) 8 = C.
,1.2 OPERATIONS ON SETS
The standard operations on sets, which yield new sets from old ones,
are (i) union, (ii) mtersection, and (iii) complementation.
Given two sets A and B, the union of A and B, written as A U B, is
defined as
A U B = {x I x e: A. or x E B} .
Here •x E A or x E B' means xis in A or in B or in both. Throughout
this book the word 'or' will be used in the inclusive sense as here, ex\!ept
when otherwise stated.
The intersection of two sets A. and B, written as A n B, is defined as
A. n B == {x I x E A and x E B} .
The complement of B in A, written as A '-. B, is defined as
A '-. B =- {x I x E A and x ~ B} .
A.'-. Bis also denoted by c,.(B).
Example 1.16 Let A - {I, 2, 3, 4, 5, 6} and B = {O, -1, 2, -3, 4, -5, 6}.
Then
A U B = {-5, -3, -1, 0, 1, 2, 3, 4, 5, 6}
An~=- {2, 4. 6}
A'\. B = {1. 31 S} and B'\.A. - {O, -1, -:3, -5}.
The geometrical iUustrations in Figure I.I, called Venn diaarams, 'are
helpful in undemanding theso operations.
6 / SETS AND FUNCTIONS

The shaded part represents The shaded part represents

AOB AUB

The shaded part represents The shaded part represents

A'-B X "- A ,f Ac::: X
flOURE 1.1

CARTESIAN PRODUCT
Let A and B be two sets. Consider the set of all ordereJ pairs (x, y),
where :x E .4 and y E B. This set is called the ,:artesian p1wJuct of sets
A and B, written as A x B. In symbols we write
A x B - {(x, y) I x E A, y E B} .
Example 1.17 Let A = {l, 2} and B = {x, y, z}. Then
A x B = {(1, x), (2, x), (l, y), (2, y), (1, z), (2, z)}.
Example 1.18 Let R be the set of real numbers. Then R x R (also written
as R') is the set or all ordered pairs (x, y}, where x and y are real num:
bers, that is,
R x R == {(x, y) Ix, y are real numbers}.
Geometrically, R represents the set of points or a straipt line (called
the real line), and R x R represents the set or points of a plane.

Problem Set 1.2

1. Let A, B, and C be lhe sets a. defined io Problem 9 (Problem
Set J.1). Determine
 1 2 OPERATIOr--8 ON S1:1S / 7

(a) A u B An B
(b) (c) AnC
(d) BL1 C (el A" B (f) B" C
(g) C'., A (h) A < B.
2. let .1 =-- (O, 1, 2, 3, 4}, B -= l~ E N I x -~- 20},
C - t.\ E: NI~ 3}, D = t~ E N ! x 1" d1\IS1ble hy 7f.
Determmc
(a) A I J B (h) AuD (c) Br IC
(d) 8(1 D (<') A I IC (fl A n D
(g) A'-. B (h) A.'\./> lll B'.,C
(J) B '"\. D (k C ' . D t I) AX B.
3. If the c;eti, A, B, C, and Dare a, defined in Problem 2. then deter-
mme ,,hethcr each of the following citatcments 1s true or fal!ie:
(a) B I C r _ D. \b) A .... B. (c) A C C.
(d) A I 1 /J - +. (e) A { ( r: B (f) C''\. A C D.
4. Let A - [u, ~. ·r}, B = f~. ,.,, O}, C ~", ~-. e:}, Determine
(a) A '< R (b) A y C' (c) B ;< C.
5. Dc~cr1be }he follo\\mg ,;ubsets of R
(ct) { ~ I , > 7 J l J [A I x < OJ (b) fY I x .. 1} U {x I x < I}
tc) 'x Ix - l} J t~ I\ l} (d) (r Ix >-1: n {x Ix< 1}
M {x : x < - 11 r) {x !x > I} (f) {x I x > 11 n tx I x ), O}.
6. If A, /J, C, and D are any seh, prove that
(a) (A X B) ("I (C X D) == (A. (' C) X (B ft D)
(b) (A ,,, B) .J (C X D) 1...: (A U C) x (BU D).
7. Give an e>.ample in which
(A X B) U (C X D) :;c (A U C) X (B U D)
8. lf A, B, and C are an) sets, prove that
(a) A U (B n C) = (A U B) n (A U C)
(b) A n (B l.J C) == (A n B) \) (A n C)
(c) AU (BU C) = (AUB) UC
(d) A n (B n C) =-= (A n B) n C.
9. Prove DeMorgan's Theorem:

(a)
(b)
s'
If A and B are both subsets of a set S, then
(A u B) = (S '- A)
S'- (A n B) = (S '\. A) U (S ' . B).
n (S' B)
1O. True or false ?
(a) If .4 and B are two sets, A n B is the largest subset of A U B1
which is contained in both A and B. '
(b) If A and B are two sets, f U B is the smallest superset or .4 n B.
'Which contaias both .4 and B.
8 / SBTS AND FUNCTIONS

(c) If ..t is nonempty, -4 '\. B can never be empty.

(d) The union of two intervals in R is an interval.
(e) ..t x B is not equal to B x ,4.
(0 This text has not yet given the definition of an infinite set.
(1) This text has not yet considered an infinite set.
(h) R x R is the set C of complex numbers.

1.3 RELATIONS
Let ,4 be a nonempty set. A subset )If of Ax A is called a relation
on .4. If (x, y) E JI, we say x 1& related to y by the relation )If and
symbolically we write x )l y. For example, JI = {(I, 2), (l, 3), (2, 3)}
defines a relation on the set ..t = {1, 2, 3}. Here I JI 2, 1 :7f 3, 2 :rt 3.
Obviou1ly, it is the usual relation'<', because 1 < 2, l < 3, 2 < 3. On the
same set A, the relation•~• is described by the set {(l, l), (l, 2), (1, 3),
(2, 2), (2, 3), (3, 3)}.
We note that'<', '=', '>',•~•.etc., are relations on R, N, Z, and Q.
'Is the mother of', 'is the brother of', 'is married to' are relations on the set
of all human beings.
1.3.t Definition Let :7f b~ a relation on a set ..t.
(a) If x Jlf x (1 e. (x, x) E Jlf ), for every x E A, )If is said to be a
r,jlexi,e relation.
(b) If, whenever x )I y, it ,s also true that y JI x (i.e. whenever
(x, y) E JI, (y, x) also belongs to JI), 1l is said to be a symmetric
relation.
(c) If, whenever x Jf y and y :JI z, it is also true that x 9f z (i.e. if
(x, y) E 9'l and (y, z) E 911, then (x, z) E 911), 911 is said to be a
transiti,e relation.
(d) A relation 911 on A that is reflexive, symmetric, and transitive is
called an epl,alence relation on A..
We shall consider several examples of relations.
Example 1.19 Let ..t ... Z, the set of all integers Consider the subset 91
of Z X Z defined by
911 = {(x, y) : x - y is divisible by 3} •
Here x 91 y llf (x-y) is divisible by 3.
(a) DI is reflexive, becauso x-y is divisible by 3 for every integer x.
(b) x-y is divisible by 3 clearly means y-x is also divisible by 3.
Hence, :,c II y implies y 91 x. So JI is symmetric.
(c) If x-y is divisible by 3 and y-z is divisible by 3, it is certainly
+
true that x - s - (~ - y) (y - s) is aleo divisible by 3. Thus, x 91 y
and y I! • Imply x 91 •· Hence, 91 la transitive.
 1.3 RILAnONS / 9

Therefore, 9' is an equivalence relation.

Example 1.20 Let .4 == R, the set of real numben. It is obvious that•-•
, is an equivalence relation.
>Example 1.21 Define the relation Jf on Z '- {1} by saying x ,ry ;Jf x and
y h11_ve a common factor other than 1. This relation is reflexive and
syftintetric, but not transitiye. ·
Example 1.22 On Z, definex 9' y to mean x > y. This relation is neither
reflexive nor symmetric, but it is transitive.
Example 1.23 On Z, define x 9' y to mean x ~ y. This is reflexive and
transitive, but not symmetric.
Example 1.24 On Z, define x ,r y to mean x ~ y + 1. This relation is
reflexive, but neither symmetric nor transitive.
Example 1.25 On Z, define x 9' y to mean x == - y. This relation is neither
reflexive nor transitive, but it is symmetric.
Example 1.26 On the set F of all fractions of the form alb with a, b :p 0,
define a/b !7f c/d iff b = c. This relation is neither refl.exive nor symmet-
ric nor transitive., •
Example 1.27 Let A be the set {l, 2, 4, 6, ... }. Define the relation ,r by
saying x !1f y ;Jf x and y have a common factor other than 1. This rela•
tion is symmetric and transitive, but it is not reflexive, because 1 !7f 1 is not
true.
Example 1.28 On C, the set of complex numbers, define z !1f w to mean
Re (z) ~ Re (w) and Im (z) ~ Im (w). It is reflexive and transitive, but
not &ymmetric.

Problem Set 1.3

1. Determine whether each of the relations defined below on tho set
Sis reflexive, symmetric or/and transitive Pick out the equivalence
relations.
(a) X JI y ~ X ~ y + 1 : S == R
(b) x Jlf y ~ x - 2y : S == Q
(c) x Jf y ~ x - y is divisible by 2 : S - Z
(di.) x JI y ~ x is the brother of y : Sis the set of all human beinp
(e) x )t y * x C y : S is the set of all subsets of R
(0 x Jt y * x is married to y : S is the set of all human beinp
(g) X 1lY * "- I y': s - R '
(b) "Jt y * X - -y : s == Q
(i) " JI y *sin" - sin J' : S = R
(j) z 11 w .- I z I - I w 1 : S - C
(k) x JI y *x and y are studenb of BITS : S is the set of all
students of India.
 JO / srrrs AND FUNCflONS

f.4 FUNCTIONS
The reader D'lusl have already been exposed, in his earlier training, to
a proper definition of functions and operations with functions. Hoy.ever,
as the concept of a function is very fundamental to mathematics and its
applications. we ,hall now give the basic ideas about functions relevant
to our subject.
1.4.1 De&nitioo Let A and B be two nonempty ,ets. A function (map)
f from A to B 1s a rule whlch, to each element of A, associates a
unique element of B. Symbolically, Y.e \\rite
f:A ➔ B.

For each x E A, the unique element of B associateti with x if) denoted

by /(x) and is called the image of x by (under) for the /:image of x or the
w,1,,e off ot x. We al&o say that x is mapped into /(x) by/. x itself is
said to be a pre-image of /(x) under f. A is called the domai11 off v.ritten
as D(f). It is the set on wb1cb / is defined. B is cc1.1led the target set of/.
lt is the set to which the images belong.
For clarity, we often draw a diagram as in Figure l.~.
A ......---r---, 8

FIGURI! 1.2
We shall set up a number of con,entions on usage, by means of a
simple example of a function : Con&ider the function /: R ➔ R di=fined by
the rule, which, to each x e: R, associates the value )."~. So /(x) = xJ for
all x E R. By abuse of language, Y.e sometimes say that the function is x 1 ,
and we even use the symbol /(x) for the function. Very often, we use the
more precise language that the· function is given by /(x) -= x 2 , x E R.
Alternatively, we denote this function as the set {(x, x 1) I x E ~ of or-
dered pairs. Thia set is also called the graph of the function; because, if
we plot the points (x, x2 ) in a plane, we will get the geometrical curve
nspresentina the function.
We shall now illustrate the foregoing ideas by listing si=ven ways of
npressina this x1-function. The first four methods arc technically perfect,
aad. methods (v) to (~ii) are sanctified by custom and Ulage.
(i) /: R-=+- R defined by x t-+ x•.
(ii) •/: x ..,... x• for all x E R.
 u FUNCffONS I 1J
(iii) /: R-1 ➔ R defined by /(x) a= x•.
(iv) / == {(x, x 1) I x E R}.
(v) The function x= defined on R.
(\'i) The graph of the function is {(x, x 2 ) I _x E R}.
(vii) The graph of the function is as in Figure 1.3.
y

fJGURB 1.3
Note that the symbol ·t--►' used in (i) and {ii) is different from the
symbol •-• for a function. The symbol '-+' is used when we want to
specify the domain and target set off, \\ hereas 'f-+' is used when we want
to indicate the image of a single element of the domain. Thus, x t-➔ x•
means that the image of x is x~. On the other hand, when we write
/: A -+ B, we are not specifying the image of a single element.
Though we shall allow oursel\'es the liberty of using any of these
methods to sp1:cify a function, in the sequel, more often than not, we shall
adhere to methods (i), (ii), and (iii). In case there is doubt while using the
other methods, the reader should go back to the first three methods for
clarification.
It may be noted that, if it is possible to list, without ambiguity, all the
values of /(x) as x varies in the domain, we do so, and we also- ri.ay that
this list is the function.
Now we emphasise the word 'unique' occurring in Jhe definition of a
function. for each x E A, f(x) should be unique. ln other words, f
cannot h.ave t\\o or more values at a given x. For example, if A - {1 1 21
3, 4, S} and B - (-1, 2, -3, 4, -5, 6, O}, then the association 1 ...... -1;
12 / SliTS AND 'FUNCTIONS

2 t-+ 2 and O; 3 ~ -3; 4 ~ o. 5 ~ -5 is not a function, since. the

imaae of 2 is not unique.
Example 1.29 A.-= {l, 2, 3, 4, 5, 6} and B .. {-1, 2, -3. 4, -5, 6, 0}.
Define/: A. ➔ B by saying /(1) == -1./(2) = 2. /(3) = -3, /t4) - 4,
/(S) .... -5, and /(6) - 6. This function is nothing but {(I. -:-1), (2, 2),
(3, -3), (4, 4), (S, -5), (6, 6)}.
Example 1.30 Given .A and B as itt Example 1.29, define/: A. ➔ Bas
follows: /ti).., -1,/(2) = 0,/(3) = -3, /{4} == 0,/(5) ~ -S, /(6) = 0.
Thia function is nothing but {(l, -1), (2, 0), (3, -3), (4, 0), ( 5, -5),
(6, O)}.
Example 1.31 Define/: R ➔ R by Jtx) = Ix I . This is called the absolute
,alue function.
Example 1.32 Define/: Rl ➔ R by /tx1 , x1) = x1 + x1 • To each point
(x1 , x 1 ) in JP, this function associates the real number x1 +
x 2• In other
words, to every pair (x, y) of real numbers, this function associates their
aum x + y. Therefore, this function is called addition in_R.
Example 1.33 In the same manner as in Example 1.32, multiplication in
R oan be written as/: Rl -+ R defined by /(x1, x 1) = x1x1 •
Example J.34 Fix a real number A. Then/: x f-+ Ax for aJI x E R is a
function, called (scalar) multiplication by A.
Example 1.35 Let IX be a fixed real number. Then/: x ~ x + IX for all
x E R is a function, called tranllation by IX.
Example 1.36 f: x~ 0 for all x E R is a function, called the zero func-
tion or the zero polynomial on R. It is denoted by O. Note that /(x) = O
for all x E Rand its graph is just the x-axis.
Exatnple 1.31 Let A. and B be two nonempty sets and b0 E B be a fixed
element. Then the function/: A ➔ B defined by f(x) - b0 for all x E A.
ia called a constant function. Note that Example 1.36 is a special case of
this function.
Example 1.38 The function/: A ➔ .A defined by /(x) == x for all x E .d
ia called the identity fi,ln~tlon on A. and ia usually denoted by 111..
Example 1.39 Let n be a nonneptive integer and a 0, a 1 ... , On be fixed
elements in R. Then the function
p :x ~ a0 + a1x + ... + a_X" , x E R,
a c.:alled a real polynomial junction on R or simply a real polynomial on
R. Note that it ia a map from R to R. By abuse of lanauage, a. -+, aix
+ ... + ~ itaelf ia called a polynomial. To denote the face that
o., Oi. •••, a. are real, it ia alao called a po'1nomlal with real coej/iciellt1.
If a. '6= 0, then n ia uid to be the ,Jqree of the palynomial a. + "1"
+ ... + ~- The aet of all real polynomials ia denoted by fl, and the
 J.4 nJNC'IMa / II

• of all naJ polynomials of delfw not areater than II i1 dlnotecl l,y g.,.
Note tbat tbe zero polynomial, 0, belonp to botb !I and g.. Tbe delftl8
of tbe ~ polynomial'i1, by con,eatioo, uaipod to be -ai,. Similarly,
1: C .... C aiven by JI(~) - a.+ 11sz +...
+~.where a.. ai, .•.• a. ue
complex numbers. ii ca.Jled a complex polynomial on C.
&.nple 1.40 Define/: U.➔ Jt by /{J,) -p(O) for all p e fl. For inl,.
taace, ir the polynomial is p: x ...., 2r + I, then ft.I,) - I. Note tbal
tbia is not a coastant function. (Why ?)
Bzam,le 1.41 Let N bo the set of natural· numben. Then any function
/: N ➔ R can be written u the set of ordered pain {(1,/(1)). (2. /(2)).
(3,/(3)), ... }. Such a functioa i1 called a Ufl/MIU of real numben. '11111
11quence ii alBO written u •
.f(l), J{2), /(3), ...
Por instance, 2. 4, 6, 81 ••• i1 a aequenoe. It i1 just the function .x-t-+ 1x
onN.
In tbe dilCUUion of any function /, there are four fundamental quea-
tioaa: \
{i) What i1 tbe domain of/ 7
(ii) What i1 th1t raaae of/ 7
(iii) 11 the function •onto' or not ?
(iv) 11 the funetion •oae•one' or not ?
We have already introduced the domain of/. We 1ball now introduce
the conceptl related to other queatiom.
1.4.2 Delaltloa Let /: .4 ➔ B be a function and , e B. The pre-tmap
of y i1 tbe aet {.x E A. I Jtx) - y}.
t.4.3 Rellluk An accepted notation for the pre-imaae of y under/ ia
/·1(1). However, we shall not 111e tbil notation in thi1 IIDII.
Bza,,,ple 1.42 Define /: R ➔ Jt by tbe rule /(x) - x•.
The pre-imap
of 4 ia {2, -2}, and the pre-imqo of O i1 {O}. The pie-imap of-1 i1
the empty set +.
1.4.4 Delllltlea Let/: A. ➔ B be .a funetion. The aet { /(,c) : ,c e .4} ii
called the ,_,. of/ and ii denoted by RCn.
It ii in fact tho let of all/-imapa.
l.4.5 Delaltla A map / : A.. ➔ B ia said to be 01llo (1~ive) if
Jt(/) - B.
Clearly, tbe followina statement ii true.
1.4.6 Fact A •Pl: A ➔ B ii onto f6 either one of the followina proper-
tia bolda: •
(a) For every 1EB, tbae 1ai111 at leut one ~EA IIICh tbu
/(,c)-1. (I)
14 / 51TS AND FUNCTIONS

(b) Por every y E B, the pre.image of y is a nonempty sublet

or .c. c2>
The proof of this fact is left to the reader.
We now consider Examples 1.29 to 1.41 with reference to the foresoing
definitions. The function/: x -+ x• from R to R is not onto. because the
negative real numbers have empty pre-images.
In Example 1.29 R(/) - B ' . {O}. So/ is•not onto as also the function
in Example 1.30, because •
R (/) = {-1, -3, -s, O}.
In BxaJDple 1.3 l R (/) - {x I x :> O}. So/ is not onto. But in Example
1.32 the ranse iaR. since every number in R can be expressed as x 1 +.:r.
for suitable real numbers x1 and x1 in ar least one way. Hence, the map
boo~ •
The function of Example 1.31 is onto because the range is R- The
reason is similar to the one gived for Example 1.32.
In Example 1.34, if A-::/;: 0, the range is R, because every number in R
can be expressed u u for a suitable x in R. The map is onto in this case.
If A = 0, the range is {O} and the map is not onto.
In Example 1.35 the map is onto, since the ranie is R. (The reader
abould reason this out.r But the map of Example 1.36 is not onto, as the
range is {O}. This is an extreme case of a function not being onto.
The map in Example 1.37 is, in general, not onto as the range is {b0}.
It becomes an onto map if Bis a singleton. In Example 1.38, since the
range is .C, the map is onto.
In Example 1.39, unless p is specified, the range cannot be calculated.
For example, the polynomial p : x 1-+ x + 1 is an onto function, whereas
tin, polynomial 9 : x 1--+-r + 1 is not. (Why ?)
In Example 1.40 the range is R (why?), so the map is onto, whereas in
Example 1.41 the map is not onto, because the ranae is {/(1), /(2), ... }.
It is a proper subset of R. The proof of this is beyond the scope of our
treatment. •
1,4.7 De8altloa A map/ : .C ➔ B is said to be one-one (injective) if
distinct elements of .C are mapped into distinct elements of B.
The function x I-+ x', x E R is not one.one. becaUle we can ftnd several
pain of distinct elements x1 , x1 that have the same image, e.g. both 2 and
-2 map into 4.
Clearly, tho following statement is true. The proof is left to the reader.
1.4.1 fact A map/: .4 ➔ Bis one-one If/ any one of three propertio1
hold1:
Ea) ;,r ¢ Y • /(;,r) ¢ /C,). (3)
(b) /f.x) - .ft,) • x • y. (4)
 1.4 PUNCTIONI / 15

(c) The pre-imap of each element in the raqe of / ii a single-

, ~ w
Note that whenever we want to prove that a certain function ia one-one,
we bave to prove (3) or (4) or (S).
t.4.t Remark The name 'one-one' comes as follows : Already in the
definition of a function, given x, Jtx) is unique. The requirement
of Definition 1.4.7 further asllitts that, given /(x), x is unique.
Thus, the correspondence x .., Jtx) is unique both ways. Such
a correspondence is called a one-one correspondence. Hence tho
name one-one function.
Let us now check which of the examples from 1.29 to 1.41 are one-
one.
In Example 1.29 the function is one-one. In Example 1.30 it is not,
because the pre-image of O is {2, 4, 6}.
In Example 1.31 /is not one-one, because for any ,c E R, x and -x
both map into the same image.
In Example 1.32 / is not one-one, because a number can be expressed
as the sum of two real numbers in more than one way. For instance,
7 .... 3 + 4 = S + 2. So/ (3, 4) = 7 = / (S, 2). For a similar reason, the
multiplication function of Example 1.33 is not one-one.
In Example 1.34, if .A :I= 0, (scalar) multiplication by .A is one-one, because
.Ax1 == .Axa • X1 == x1 , which proves (4) in Fact 1.4.8. If .A = 0, the map
is clearly not one-one.
In Example 1.35 translation by 11' is a one-one map, because
1

X1 + Cl = X1 + 11•X1 == X1,
The maps of Examples 1.36 and 1.37 are clearly not one-one, because
all members of A. are mapped into one member of B. In fact, this is tho
extreme case of not being one-one. On the other hand, the map of
Example 1.38 is one-one.
In Example 1.39 whether or not the function is one-one depends upon
the polynomial, because the polynomial p : :,c I-+ x + l is one-one, where.
u the polynomial q : x I-+ x• + 1 is not one-one. (Why ?)
The map of Example 1.40 is not one-one, because (x+l) and {:ic1 +1)
have the same imaae,
In Example 1.41 also the one-oneness depends upon tho sequence.
.... tho sequence 1, 1/2, 1/3, 1/4, ... is one-one, whereas the, equence
1, l, 1, ... is not one-one.
L4.18 Delaldoa Two aetl A. and B are said to be in ope-on, corr,qond-
.,,_ if there nilts a . . . , : A. ➔ B that ia one-one and onto.
IA.II D1l11dia If a Ill .4. ii iD one-one COfflllpOJldenc with the let
n. 2. .....11l. for 10me DOlitive inteaer 11 • ..4 ii said to be a Jlntt, 111.
16 I sm AND l'UNCTICJNS

Sets that are not finite are called i,ifinite set,.

For eumplo, the set of all real roots of the equation x' - 2.:r' - x + 2
- 0 and the act of vertices of the triangle A.BC are in one-one corres-
pondence. Both are finite acts, since they are in one-one correspondence
with the set {l, 2, 3}.
N, Q, R, Z, and C are all infinite sets. (Why ?)

Problem Set 1.4

1. Determine which or the following subsets or R x R are functions :
(a) {(l, 1). (1, 2), (1. 3). (1. 4), (I, S)}
(b) {(l, I), (2, 1), (3, 1), (4, 1), (S, 1)}
(c) {(l, w), (n, 1), (0. e), (e, 1,}
(d) {(n, e), (TC, 1), (1, e), (e, 1)}.
2. For the aiven sets A and B, define, if possible, a function /: ,4 -+ B
such that (i) / is onto but not one-one, (ii) / is neither one-one nor
onto, (iii) / is one-one but not onto, (iv) / is one-one and onto :
(a} A = {a, b, c, d}, B """ {O, 1, 2, 3}
(b) ,4 = {I, 2, 3}, B == {O, I, 2, 3}
(c) A = {O, 1, 2, 3}, B = {I, 2, 3}
(d) A =- R, B = R.
3. Determine the largest subset .4 of R for the function / : ,4 ➔ R
defined as rollows :
(a) /(x) = v'(x' - 1) (b) /(x) = v'(4- r)
(c) /(x) = x/(1 + x) (d) /(x) = r - 2x + 7
(e) fl..x) =- x/{x1 + 1) (f) /{x) = v'(x' - x 1).
4. Determine the largest subset ,4 of C for the function / :
dofined as rollows :
(a) /(z} = I z I (b) l{z) = z / I z I
(c) /(z) .... Re (z) (d) /(z) - exp(i Im(z))
(el /(:t) -= exp(lle (z)) (f) f(z) == -;.
5. Determine the range of each of the functions in Problems 4 and s.
6. Determine which of the functions in Problem 4 are one-one and
whicbonto.
7. Truo or fabe 'l
(a) There ia a one-one oorretpondence between ..4 x B and B x ..4.
(1,) Tbere ii a one-one oornapondence between C and R x R,
(c) If ..4 aacl B ue tlnite lets, tbm miry one-oae •P /: ..4 ➔ B ii
allo oaio.
 t.S BINAI.Y OPIII.ATIONS / 17
(d) If A. and B are finite sets, tben every onto map /: 4 -+ Bis
also one-one.
(e) A nonempty subset of a finite set is also finite.
(0 Every nonempty subset of an infinite set is infinite.
8. Determine which of the following graphs represent functions:
V V

•
X X
0 0
Ca) Cb)
V V

~
X
0 0
Cc) Cd)
V y

-ioi---------x
0
m
1.5 BINARY OPERATIONS
We shall now introduce a very important class of functions which in•
eludes the familiar operations of arithmetic, namely, addition and multi-
plication. If we look back carefully at the opeiation of addition or
numbers, we find it is essentially a process of combining two number■ and
obtaining another number in an unambiguoua way. Thf: aame ia true of
multiplication. A pneralisation of this idea is the followina definition.
1.5.1 Delllldaa Given a nonempty Mt A., any function from A x A. to A.
is called t. b"""1 opm,tlon OD A..
18 / IBTI AND PUNCTIONS

In other words, to each ordeNd pair (x, y) of elements :c and y of A,

the binary operation • on ..4 associates a unicfue element z of A. We
write z - x • y,
The functions defined in Examples 1.32 and 1.33 are binary operations
on R. But the (scalar) multiplication by a fixed Ain Example 1.34 is not
a binary operation on R, because it is not a function on R x R. It does
not associate a number with every pair of numbers in R.
Note that though addition is a binary operation on R or Nor Q or
the set E of even integers, it is not a binary operation on the set of odd
iDtegers, because tho sum of two odd integers is not an odd integer. We
formalise this situation by the following definition.
1.5.2 Definition Let B C A. Let • be a binary operation on ..4. If for
each pair of elements (x, y) in B, x • y belongs to B, we say that
B Is closed under •••. If there uist x, y in B sucli that x • y ~ B, we
say that B Is not closed under '• '.
Note that the situation 'B is (not) closed under ••• ' is sometimes
0

referred to as • '•' is (not) a closed operation in B'.

To illustrate Definition l .S.2, let us consider a couple of examples.
Example 1.43 The set of odd integers is not closed under the usual addi-
tion of numbers, but it is cJosed under multiplication.
Example 1.44 The set N is not closed under division, but it is closed
under addition.
The reader should note that unless B is closed under '•', it cannot be
a binary operation on B. We now give some examples of binary opera-
tions.
Example J .45 J : (x, y) ~x - y is a binary operation on R. It is also
a binary operation on Q, Z, and C, but it is not a binary operation on N.
Example 1.46 f: (x, y) 1-+ x/y is a binary operation on I'\. {O}. We
cannot have the whole set R here, because x/y is not defiired when Y = 0.
It is also a binary operation on C '\. {O}, Q '\. {O}, but not on Z '\. {0}
or N. (Why?)
Example 1.41 f: {x, y) t-+ x +
y - xy is a binary operation on C, R,
Q, aod Z, but not on N. (Why ?)
Example 1.48 f: (x, :,) J--+ x" is a binary operation on N, but not on Z
or Q or R. (Why?)
Example 1.49 f : (x, y) ~ I x - y I is a binary operation on Q, Z, R,
andC,
t.5.3 DelaldoD . A binary operation • on a set A is said to be
(a) eomm111at1N if a • b - b • a for all o, b e A;
 1.5 BINAR~ OPERATIONS/ 19

(b) a11ociatlN if a·• (b * c) - (a • b) • c for all a, b, c E .(.

Whenever '•' is associative, we can ~rite a .,b • c in place of a • (b • c)
or (a • b) • c.
The usual addition and multiplication of numbers are both commuta•
tive and associative. The operations defined in Examples 1.45, I 46, and
1.48 are neither commutative nor associative. But in Example 1.47 the
operation is both commutative and associative. In Example 1.49 the opera•
tion is commutative, but not associative. The reader should check these
statements carefully.
We shall not discuss nonassociative binary operations in this book,
but we shall deal with some important noncommutatl~e binary opera·
tions.

Problem Set 1.5

1. Define addition •+' and mult1phcation 'x • on the set S = {«, ~. y}
with the help of Tables 1.1 and 1.2.
TABLE 1.2

+ Gt ~ y X Gt ~ y

« « ~ y GC « « GC

~ ~ y Gt ~ Gt ~ y
y y ~ y y

'
Gt Gt

Prove that the binary operations •+• and • x • as defined by these

tables are commutative and associative.
2. Define the binary operation •• • on R, the set of real numbers, as
follows:
X • Y =- xy + X + y.
Is '•' commutative ? Is ••• associative ?
3. Give an example of a binary operation, which is
(a) associative but not commutative
(b) commutative but not associative
(c) neither commutative nor associative.
4. Ddne addition •+• and multiplication• x • in R x Ras foUowa:
(x1, x,) + CY1, yJ - (x1 + Y1, x, + Y,J
(Xu X.) X (11, Y1) - (x, Y1, X■ ya).
Prove tliat tbele are commutative and aaociative binary operations,
20 / sm AND FUNcr1on

given that (x" x1) .. C,1, y1) meam x1 - Y1 and x1 == y 1 •

5. Define the binary operations•+• and' x' on R x Ras follows :
(x1, X1) +<11, Ya) = (X1 + Y1, X1 + Ya)
(xi, x,) X (yJ, )'1) -= (Xi.)'1 - XaJ,'1, Xl,)11 + X.Y1),
Prove that both theae operations are commutative as well as associa-
tive. (Equality is defined as in Problem 4.)
6. Let Z, be the set of nonzero real numbers. Define addition•+• and
multiplication ' x • on Z x Z as follows :
(x1, x 1 ) + ()'i, yJ - (xi11 + ,'t.Y1, XaY1)
(Xi, X1) X (Y1, Ya) == (X1Y1, x._v.).
Prove that these are commutative and associative binary operations,
aiven that (x1, x1) == CY1, Y1) means x1y1 = X■>'l'

1.6 ALGEBRAIC STRUCTURES

FIELDS
The standard properties of real numbers with respect to addition and
multiplication arc so fundamental 1n mathematics that whenever any set
X with two binary operations•+• and•.• (called addition and multipli-
cation) satisfies these properties, we give a special name to such a system.
It -is called a field. To understand these properties, which arc listed in
the following formal definition, the reader is advised to think of X as the
s~t R of real numbers and the operations'+' and'·' as the usual addi-
tion and multiplication in R.
-1.'91 Detlnltloa Let X be a nonempty set on which there are two binary
operations •+' and '• • , called addition and multiplication, respec-
tively. Then the set X together with these operations is said to be
a field, if the following axioms are satisfied :
Fl Addition is associative.
F2 There exists an element OE Xwith the property that O+ a = a
+ 0 == a for all a E X.
F3 For each a E X, there exist, an x E X such that a x = 0 +
-x+a. ·
F4 Addition is commutative.
FS Multiplication is associative.
F6 There exists an element I E X such that 1 •a "='" a == a• l for all
aex.
F7 For each a (¢ O) in X, there exists an element y E X such
tbat tl•J - I • Y•a.
 J.6 ALGUR.\IC STRtlCTURU / 21
F8 Multiplication is commutative.
F9 Multiplication is distributive o,er addition, that is, for all
a, b, c EX,
a • (b + c) = a •b + a •c
and (a + b) • c ..,. a • c + b•c •
In the case of X - R the element x mentioned m F3 is just the fami-
liar -a, the negative of a, and the element y mentioned m F7 is just
the familar 1/a, the reciprocal of a.
Axioms Fl to F9, which are vahd in R, are fundamental. Startin& from
them, all the formulae of elementary algebra can be derived.
The classical examples of fields are, besides R,
(i) the set Q of rational numbers under the usual addition and multi-
plication and
(ii) the set C of complex numbers under the usual addition ana multi-
plication.
In this book we shall be interested in only these three fields, namely,
R, Q, and C. '
Z is not a field under add1t1on anJ multiphcation, because F7 is not
satisfied. So also N 1s not a field. (Why ?)

GROUPS
Now we take up another structure in algebra, which is simpler than a
field. Suppose we restrict our attention to only one binary operation on
a nonempty set X and denote it by•+•. If this satisfies axioms Fl, F2,
and F3, we say that the system is a group. If, in addition, F4 is satis-
fied, we call it a commutative group. We shall now give the formal defi-
nition.
1.6.2 Oe&nition Let G be a nonempty set on which a binary operation •••
is defined. Then G is said to be a group, under the operation '•', if
the followiog axioms are satio;fied :
G 1 ••• is associative.
G2 There exists an element e E G such that e • a = a = a • e for
all a E G. e is called an idtntity for •••.
G3 For each a E G, there exists an element x E G sudl that
a• x = e ... x • a. x is said to be an inverse of a for the optta-
tion •••.
If, in addition,
04 The operation ••• is commutative,
then the aroup ii said to be commutatlH or ahllan.
22 / SETS AND FUNCTIONS

Let us consider four examples :

(i) R, Q, and Z are groups under the usual addition operation.
(ii) N, under addition, is not a poup, because 02 and 03 are not
satisfied.
(iii) Q'\,{O} and R'\,{0} are both groups under the operation of usual
multiplication. The number 1 plays the role of 'e' in the definition,
(iv) The set Z'\,{0}, under multiplication, is not a group, because 03
is not satisfied.
In all these four examples of groups the operation is commutative. •So
they are all commutative groups. However, noncommutative groups are
of areat importance in mathematics and its applications. Watch for some
examples of noncommutative groups as we go along.
Before we proceed, we shall establish the uniqueness of the 'identity'
referred to in 02 and tho "inverse' refemd to in 03.
1.6.3 ntorem (a) In a group G there ts only one identity, t e. the Identity
f• unique.
\b) For a given elem€nt a in a group G there II only one inver,e, I.e.
1he inverse of an element a E G ts unique.
Proof: (a) Let e and e• be two elements in G, both having the
property stated in G2, namely,
e • a == a == a • e for all a E G (1)
and
e' • a == a == a • e' for all a E G. (2)
Then taking e' for a in Equation (1), we get

• e • e' = e' • e - e',

and taking e for a in Equation (2), we get
e' • e = e • e' - e .
Hence, e == e', i.e. the identity is unique.
(b) Given a E G, let x and y bo two elements in G, both havin1 the
property stated in 03, namely,
(3)
and
"*Y == Y•a =- e. (4)
Therefore, y = y•e (by 02)
=- y •(a• x) (by (3)).
But the operation ••• is associative in G. Hence, we get
Y = (y •a)•"
(by (4))
(by 02).
 U ALG■BI.AJIIC STRUC.'TURII / 23

Thia prove, that the inverse of each oJement is unique. I

Thus, hereafter, we shall speak of the identity in a sroup G and also of
the ,,,.,,,., of an element a E G.
Whenever the operation is denoted by •+•, the identity is denoted by
•o•, and the inverse or •a• is denoted by •-a'. It is customary to write
a-bfora+(-b).

ltINGS
Another important structure is that or a ring. Let us once apin
consider the nine field axioms. The first four axioms-Fl, F2, P3, and
F4-are pertinent to addition, whereas the next four axioms-FS, F6, F7,
and F8-are pertinent to multiplication. The ninth axiom, F9, pertains to
both addition and multiplication.
If there is one binary operation on a set X and the four axioms Fl to
F4 are satisfied, then the system is a commutative group. If a set X bas
two binary operations and' all the nine axioms are satisfied, we say tlle
system is a field. Id between these two situations, we llave a ring
structure.
Suppose a set X has two binary operations. Let, for the first operation,
the system 'be a commutative group. The minimum that we require of the
second operation, in order to get a meaningful system with two operations,
is the associativity FS and the distributivity F9.
So whenever· there are two operations, say • +' and •• •, on a set X, and
Fl to F4, and FS and F9 are satisfied, we call such a system a ring.
We shall not give a formal treatment of rings in this book. It suffices
to note that
(i) A set X with two operations •+• and •.' such that (a) it is a
commutative group for •+• and (b) FS and F9 are satisfied is called a
ring.
(ii) A rin1 in which F6 bolds is called a rl11g with Identity (unity).
(iii) A ring in which F8 holds is calletl a commutative ring.
For example, C, R, .Q, and Z are rings under the usual operations of
addition and multiplication. All these are commutative rings and have
unity. The set E of even integers is a commutative rin1 under the opera-
tions of addition and multiplication. But it bas no unity.
Watch for further examples of rings as we go along.

Problem Set 1.6

1. In each of the Problems 1, 5, and 6 (Problem Set l.S) prove that
•
 24 / SBTS AND FUNCTIONS
the set under consideration, alon1 with the operations of addition
and multiplication defined therein, is a field.
2. Let G be an abelian group with addition •+ •as the binary operation.
Prove the following identities in G :
(a) -(-o - b) - a + b (b) a - 0 .. a
(c) -(a - b) - b - a ~ (d) (c - bl - (a - b) - c - a.
3. Let A be a rina with '+' and •.• as binary operation,. Prove that,
for a, b, c, d E .A.
(a) (-a)• (-b) == a• b
(b) (a + b) • (c + d) = a • c + b • d + a · d + b • c.
4. Prove that a ring contains at most one unity.
S. Prove that a • 0 -= 0 for all a in a ring .A.
6. Give three reasons why N is not a field under the usual addition and
multiplication.

1.7 OPERATIONS ON FUNCTIONS

The 11tandard operations on functions, which yield new functions from
the given ones, are (i) composition, (ii) addition, (iii) scalar multiplication,
and (iv) multiplication.
Before we begin a study of these operations, we have to make precise
what we mean when we say •two functions are equal or are the same'.
1.7.1 Ddaltton Two functions/: .A ➔ Band g:A ➔ B are said to be
eq11al, written/== g, if /(x) == g(x) for all x E A.
For example, the function x I-+ x' defined on (-1, 1) and the function
x I-+ x 1 defined on [O, I] are not equal, because their domains are
different.
1.7.2 Deflnltion Let /: .A ➔ B and g : B ➔ C be two functions. Define
the function h : .A ➔ C by
h(x) - g(/(x)) for all x E A •
Then h is called the composition of g and f and is denoted by g of.
which is read as 'g composite p. · Thus. we have (g o /) (x)
- g ( / {x)) for all z E .A. Symbolically, the composition can be
represented as
I g gof
.4 ➔ B ➔ C-A--+C.
EJuim,le I.SO Let/: R ➔ Rbe debcd by x t-t- ,c + 1 andg: R ➔ R
be defined by z I-+ xi. Then 10/ ii defined, and (lo/)(x) - g(Jlx))
--1<x+l) - (x + 1)1. Henr:e,10/:x ...... (.x ..._ l)'iaamapfrom.ltoR.
 1.7 OPlliTIONt ON fUNCl'IONI / 25
Example l,$1 Let f: R ➔ R be defined by x f,-+ x + 1 and g : Jt+- ➔ R
be defined by x f-+ ..;x. Here .R+ is tbe set of all positive nal numberL
In this case g ofis not defined, but fog is defined, since
g I fog
R+-➔ R-R .. R+__..R.
and (fog) (x) = f(g(x))
= /(yx) == vx + 1.
Thus, /o g: x ..➔ y'x + I is a map from Rf" to R.
Example J.$2 In some cases both fog and gof may be defined. In
Example 1.50 / o g is also defined and (/o g) (x) f(g (x)) /(.r) =
= x• t 1.
Thus, in this case / o g : R ➔ R is defined by x f-+ x1 1, but +
go/:R-Risdefinedbyx t-+(x + 1)1.
These examples show that when / o g is defined, go f
may or may not
be defined. Even when both fog and go/are defined, they need not be
equal. In other words, composition of functions is not commutative.
\
Not&tions The set of all functions / from A to B is denoted by :1(..4, B).
If b - R, the set of real numbers, then the function f: A ➔ R is said to
be a real-valued function. The set of all real-valued functions on .4 is
denoted by :1(.4R). The notation :11t(A) is also used to denote this set
of real-valued functions. If R is understood from the context, the symbol
:1'(.4) can be used. If A is an interval, ::f'{A) is written also as :1[a, b]
or ::F(a, b).
t.7.3 Definition Let f: A ➔ R and g: A ➔ R be two real-valued func-
tions. Then the sum f + g is defined as a real-valued function as
follows: f + g: A ➔ R such that x 1-➔/(x) + g(x). Io other words,
(f + g) (x) =/(x) + g (x) for all x EA.
f + g, as defined, is called the pointwl.se sum of f and g. Accordins
to this definition, f + g is also a member of :f(A).
Example 1.53 let f: R ➔ R be defined by /(x) = x• and I: R ➔ R
y

v•ainx

FIGUU 1.4
26 / SITS AND FUNCflONS

be defined by a(x) - x + 1. Then/+ g is the function x

for all x ER. .
Example l .S4 Let us add the sine function and the cosine function. These
functions are aiven by the graphs in Figures 1.4 and 1.5. Therefore, the
y

'I a cos x

PIGUltB l .S

function sine + cosine is siven by the rule (using Definition 1.7.3)

(sine + cosine) (x) == sin x + cos x .
The graph of the function sine + cosine is as in Figure 1.6.
y
'I • sin x + cos x

J-4-~+~--¥-l-.f:-----1Hn:--,..-J--r--.x

= sinx

FIGURE 1.6

t.7.• De8nltioa Let A be a real number and/: A ➔ R be a real-valued

function. Then 'A.f. called the scalar multiple of / by A., is defined
as a real-valued function
A./ : A ➔ R such that x I-+ 1/(x) .
In other words, (A/) {x) == AJtx) for an x e A. 'A/ is called tbe poinhllae
A-multiple of J. Accordin1 to this definition, 1/ is also a member of :f(A).
Example J.SS Let/: R ➔ R be given by x 1-+ x'. Then A/ is the func-
tion from R to R auch that x 1-+ Ar.
Example 1.56 The ·araph of the function 1/2 sine, i.e. A-multiple of aiae,
with A .;. l/2 ia aiven in Fi,ure 1.7.
 1.7 Ol'DATIOlfl ON ,UNC'IIONI / 27

.-"".A'l~:--1~-:;;IJ.:"-+---:::E--+-:~--+--7'-::_~x

' ,...~ y
slnx
FIGURE 1.7
A special case of scalar multiplication when 'A = -1 gives ( -1) / as
the function
((-1)/)(x) - (-1) (/(x)) - ---/(x) for all x ER.
( -1 )/ is also denoted by - /.
1.7.5 Definition Let/: ..4 ➔ R and g: .A.-+ R be two real-valued func-
tions. Then /g, called the pointwise product of/ and g, is defined
as a real-valued function
Jg : .A. ➔ R such that x ....,.. /(x) g(x) •
In other words, (/g)(x) = / (x) g(x) for all x E ..4.
Example 1 57 Let/: R ➔ R be defined by x· f-t> x• and g: R ➔ R be
defined by x t-+ x + 1. Then the function Jg : R -+ R is defined by
x I-➔ x• (x + 1).
Example I .58 The araph of the product of sine and cosine is aiven
in Fiaure 1.8,

.v
y • 11nx cosx

y ■ cosx -· L,. slnx

PIGURB 1.8

1., R•ark
1. We have constructed functions / + g, Af. and /g.' The
construction of / + g from / and g is nothing but an operation
which. to each pair of functions/and g of :1(.A.). aaociatet • func-
tion / + r in f(..4). Tb111, thi• operation is a binary operation on
.:f(A). It ia called addition. Tile acoond operation, that of Iormina
A/ rrom Aand J. called acalar multiplication, ii not a binary opera•
 28 / Slff'I AND FUNC'T~S

tion on :f(A). The reason is that we do not take two functions in

:F(A) to produce a function in :F(A). However, the multiplication
of functions / and g is a binary operation on :F(A).
We shall refer to Definitions 1.7.3 1 1.7,4, and 1.7.S aspointwise addition.
pointwise scalar multiplication, and pointwise multiplication of functions,
respectively. Throughout the book, without further explanation or m••
tion, addition, multiplication by a scalar, or multiplication of functions is
to be understood pointwise as already,eAplained.
The following properties of addition in :f(Jt) arc easy to check. The)'
should not be taken for granted. The reader should carefully check
them.
(i) Addition in !1(A) is associative.
(ii) G2 is satisfied for additibn in ,-(A), because the zero function,
0, acts as 'e' of 02.
(iii) 03 is satisfied for addition in ::F(A), because - / acts as the in-
verse of/ for addition.
(iv) Addition in ::F(A) is commutative.
Thus, :1'U), for pointwlse addition, is a commutative group.
Analogously, we have, for multiplication, the following properties.
The reader should carefully check them.
(i) Multiplicuion in :ICA.) is associative.
(ii,) 02 is satisfied for multiplication in :l(A), because the constant
function C : A - R such that x t---1- 1 acts as the identity in :J'(A). For
we have
(C/) (x) = = I /Cx) == /(x) for all x EA.
C(x)/(x)
Hence, by Definition 1.7.1, Cf = /for.all/ in ,-(A).
(iii) Multiplication in ,-(A) is cC1mmutative.
(iv) Multiplication is distributive over addition in ,-(A).
But :1 (A) for pointwise·multiplication is not a group, because 03 is
not satisfied as we shall now show.
Suppose f: A.-+- Risa function such that /(x0 ) = 0 for some x0 e A.
In order to get a function g : A -+ R such that Jg = C, we must have
C/1) (x) == C(x) for all x E A
or '{.¥) g(x) -= C(x) -= 1 for all x E A.
But at x .. x., thia means 0,Cxo) - 1. No real number g(x0) can
•lilfy tbil requirement. Therefore, no , exist& with the property /g - c.
Tlril .... tbere are functioaa / in :1(..4) which have no inverse for multi•
pllcalioa.
 l.7 OPDATIONS ON FUNCl'lONS / 29
If we Jook at all the propercies of additioa and multiplication which
we have Ji1ted, we find that :1(.A) u a coffllnllta#ve ring with unity for
thae two operatlou.
The question whether :1(A) is a group for scalar multiplication does
not arise, because scalar multiplication is not a binary operation. But
scalar multiplication does satisfy certain properties, which we now list.
, Given /. g € :F(A), and A, p. e: R, we have
(i) A(/+ &) = 1/ + Ag,
(11) (,\ + #')/ == }if+,.,.,.
(iii) ,\ (#'/) - (,\ p.)/ = p.(1/).
(iv) l (/g) = ('A/) g = /{Ag),
(v) 1/ - f.
The reader should check these carefully. Just for guidance we check (ii),
namely,

Since this 1s an equality between functions, we have to prove

((,\ + tJ,)/) (x) == (.\/ + p./) (x) for all x € A.
The left-band side == ((,\ + p.)/) (x)
== (,\ + p.) (/(x)) (by Definition 1.7.4)
== 'A/(x) + p.f (x) (by F9 since R is a
field)
== (,\/) (x) + (!'/) (x) (by Definition l. 7.4)
== (,\/ + p./) (x) (by Definition 1.7.3).
1.7.7 Remark By now it must be clear to the reader that addition. multi·
plication, and scalar multiplication of functions are not the same
as add1t1on, etc., of ordinary numbers, though they may appear
to be so. He should be careful to understand, at any stage of the
working, whether the symbol that 1s being used is just a number or
a function.
All the results in thie article are also true for ,.c(A), 1.e. the set of all
complex valued functions on ...t. The only precaution we have to take
is that 'A, instead of being just a real number, can be any complex number.
We shall end this article by introducing an important concept, called
lnver,e funrtion.
1,7,1 Delaltloll I.et f: A ➔ B be a one-one and onto.function. Then
the in,er,e of/, written as/ ·1, is the function
1-1 : B ➔ A
defined by
 30 / SITS AND FUNCTIONS

/·l(y) - that unique :x E A. for which/tx) - y.

Note that Definition J.7.8 is valid, because by Remark 1.4.9 it
follows that, since/ is one-one and onto, there is a one-one correspondence
between .4 and B.
Example 1.59 Let /: R - R be defined by .«c 1-i> ·" + I. Tben/·1 : R - R
r
ii' given by 1(y) - y•l for all y E R.
If the function/: A. - Bis not one-one or not outo, the inverse does
not exist as a function from B.
If the function/: A. ➔ B is one-one
, but not onto, the in\"erse does not
exist as a function from B. But the sttuation can be retrie,ed by modifying
the function/, Write/,,: A ➔ R(f) by defimng.fi.(x) - /(x) for all :x E A.
Essentially, Ii is the same as /. The only difference between/ and Ii is
that the target set B ·of/ has been restricted to R(f) C B without any
damage.
Obviously,/1 is one-one and onto; so / 1 has an inverse. This inverse
is defined on R(/). Therefore, some authors say (legitimately!) that
though the inverse of/ does not exist as a function from B, it exists as a
function from R(fl, a subset of B.
We shall now give two important properties. If / : A ➔ B is one-one
and onto, then
(i) / o 1·1 = la and
(ii) 1-1 0 / = 1.4.
The proof of (i) is left to the reader. We shall prove (1i) :
/: A ➔ Bis a one-one and an onto map. Therefore, /· 1 : B ➔ A exists.
The composition /·1 o/is a map from A to A such that
( / ·1 o/)(:x) -
/·1 (/(x)) (by Definition 1.7.2)
... J- (y)
1 (by writing Y. = I (x))
== x (by Definition 1.7.8)
for all x E .A. Hence, /· 1 of == /,4.
As mentioned in Remark 1.4.3, the symbol /· 1 is also used in certain
contexts in mathematics for denoting the pre-image under /. But we aball
not use /· 1 in that context. We shall uae /· 1 only when it is the invene
of/.
Prob,.m Set 1.7
1. Prove that the composition of f unctioaa is associative, i.e. for any
three functions 1,: A ➔ B, 6: B ➔ C, and/: C ➔ D, we baw
(/01)01, • /o(go/,).

...
1..

2. Let /(%) - sin:t,-•<:1e<•

,C.x) - xt. -co < " < co•
 1.7 OPDAflONI ON PUNC'ftONI / 31

Describe gof. Is fog defined? Justify your answer. Cbanae the

domain of g so that / o g can be defined. Then describe / o ,.
3. Repeat Problem 2 for /(x) - 1 + x 1, -1 < x < 1, and g(x) - ,Ix,
x>O.
4. Prove that the set of all· real-(compJex) valued continuous functions
defined on [O, 1] is a group under pointwise addition of functions •
.S. Prove that the set of alJ real-valued differentiable functions on (a, b)
is a group under pointwise addition.
6. If / and g are defined as in Problem 2, describe / + g, / - g, and
fgfor -n ~ x ~ n.
7. If/ and g are defined as in Problem 3, describe/+ g, / - g and
/g foJ O :E;;X ~ l.

8. In HZ de:fine equality, addition, and scalar multiplication u follows :

Equality (x1, x 1) = O,i, Ya) if X1 = Y1, X 1 == Y1 ;
A..dditicJn (x11 Xt) + (Yu Ya) = (X1 + Yi, X1 + Y1) ;
Sealar multipli~a,ion 11(x1, x1) = (ax1, ax1), where a is any real
number.
A function /: R3 ➔ R'1 is said to be linear if it satisfies the folJowing
properties :
(1) /((Xi, x.) + (Y1o }'9)) = /(X1, x.) + /(yi, Ya)
(ii) /{a(xi, .\'1)) = a./(x1 , x1 ),
for all re.ii a.
Prove that each function .f: JP ➔ JP defined below is linear.
{a) /(x1 , x 1 ) = (x1 , x1 )
(b) /(x1, x 1) - {x11 -x1)
(c) /{x1, x 1) = (x1 + x 1, x 1 - x 1).

9. Prove that for the linear function /: R1 -+ 1(1 defined in Problom a,

we have
/(0, 0) = (0, 0).
10. Consider :l(G, H), the set of all functions from a arc,up G to a non-
abelian group H with a binary operation•.•. For/, g E :1(0, H),
define /1 u (/r) {x) - ~) • ,<x). Is /1 - 1/ 'l '
11. True or falle ?
(a) The function /: C ➔ R defined by ft.a + lb) - o is aa onto
function.
(b) TIit l'lulalion. /: C ➔ R defined by /(a + lb) - b ii a one-
one function.
 32 / SITS AND FUNCflONs
(c) The function / : C ➔ R x R define~ by /(a + lb) • (a, b)
is a one-one onto function.
(d) It is possible to define multiplication of functions, which is
noncommutative.
We conclude this article by drawing the reader's attention to the two
ideas in this chapter that will form the central theme of this book. They
are "addition' and •scalar multiplication' of functioi,s. The study of these
two concepts in depth and in more general situations is the study of linear
algebra.
 Chapter 2

Vectors

In this chapter we shall study vectors in a plane and in space. These

concepts originally arose in mechanics. A vector in mechanics stands for
a line segment directed from a point P (called the initial point) to a point
➔
Q (called the terminal point). It as denoted by the symbol PQ, where the
arrow indicates that it is a vector. The addition of vectors is done by the
➔
familiar law of parallelogram of forces. The negative of P Q is denoted
by QP. From this the rule of subtraction follows. The study of vecton
starting from this approach is called the geometric approach.
The cdternative approach, called the algebraic approach, is more rele-
vant to the subject matter of this book. Whereas the geometric aPJ)roach
leans heavily on the fundamental geometric concepts of 'length' and •angle',
the algebraic approach uses only the properties of R, the set of real
numbers.
Today we find that the use of vectors transcends its oriainal geometri-
cal background of physical space and finds its application in various fields,
including social sciences. For this reason, our study of vectors will be
mainly· algebraic. The geometric aspect will, however, be presented side
by side, in order to SUPPiement the understanding. In fact, keeping track
of the geometrical penpective is very useful in that it provides a deeper
insiaht into the study of linear algebra. At the end of this chapter some
elementary applications of vectors to geometry are also presented.
We are familiar with &be cartesian coordinate system in a plane. Before
studying vectors we shall introduce tho cartesian coordinate system in
space.

2.1 SPACE COORDINATES

We identify a point in a plane with the help or two axe,, called the
~-axis and tbe 1-ui1. To locate a point in space, we need three mutually
perp111dicular lines X'OX, 1'0 Y, aod Z'OZ tbroqb a point O (Fisun 2.1).
34 / Vl!C'l'OU

This point O is~called the oriain and the lines are called x-axis, y-axis, and
z-a1ii. respectively. These three mutually perpendicular lines, called the

FIGURE 2.1

coordinate axes, determine three mutually perpendicular coordinate planes

XOY. YOZ, and ZOX, called the xy-plane. the yz-plane, and the xz·plane,
respectively. These three planes di,ide space into eight parts, called
octants. Note that
x-axis is the intersection of the xy-plane and the zx-plane;
y-a:x.lS is the intersection of the .¥)1•plane and the yz-plane; and
z-axis is the intersection of the yz-plane and the zx-plane,
We choose the positive directions OX, DY, and OZ of these axes in
such a way that, if a right-banded screw placed at O 1s rotated from OX
to DY through 90°, it moves in the direction of OZ. The axes so chOHD
are aaid to form a right-handed system. Figure 2.2 shows a ript-haoded
system and Figure 2.3 shows a left-handed aystnn.
Let P be a point in space (sle Figure 2.1). Then the distance of P from
the xy-plane is called the z-coordinate of point P. Let Q be the Coot of
the perpendicular from P on the xy-plane. Then the distance of Q from
the y•axil is called tbe x-coordiaate of point P and tbat from the x-axis
Ille ,-coordioate of point P. Obvioutly, the x-coordinate or point Pi• it•
 2.1 SPACE COORDINATES / 35
dastance from tbe yz•plane and the y•coord1nate is its distance from the
:x•plane.
z l

f-1<,URf 22 FIGURE 2.3

If ,, v. and l c1re respecuve)y the T-coordmate. y-coordmate. and
-coordinate ot pomt P, we ll!IC the ordered tnple (x, y, z) to denote P. It
z

FIOURI! 2.4
36 / VECTORS

is customary to say that the coordinates of P relative to the origin O are

(x, y, z). Thus. corresponding to each point in space, there exists a unique
ordered triple of real numbers which gives the distance of P (x, y. z) from
the three coordinate planes. Similarly, corresponding to each ordered
triple of real numbers, there exists a unique point in space, which is situated
at a distance of x units from the yz-plane, y units from the zx-p)ane, and z
units from the xy-plane. Obviously, the coordinates' of the origin are
(0, 0, 0).
Distance between the points Pifx1 , y 1 , z1) and P~(x1 , Y2, z1)
To find the distance between the points P 1 and P,., draw P,.Q perpendi·
cular from P 9 to the plane P 1 RQ parallel to the xy-plane and passing
through P 1 , as shown in Figure 2.4. Further, draw QR and P 1 R parallel
to the x-axis and y-axis, respectively. Then
P1P1 = v (P1QY + (PsQ}
1 2

= v ((RQ)• + (P R) + (P Q)
1 2 2 2

= v (x2-x1>9 + (y,-y1)2 + (z,-z1)2 . (1)

Example 2.1 The distance between the points P(l, -1, 3) and Q(2, 1, - 7)
is v
(-10,:: + (22 ) + 0)2 = v 10s.

Problem Set 2.1

l. Locate the points:
(a) (1, 1, 1) (b) {I, 1, 2) (c) (-1,3,-2)
(d) (-1, -2, -3).
2. Describe the loci :
(a) y = constant (b) z = constant
(c) X == 1, y ,... 2z (d) y == -S, z• - 4x
(e) y 1z~ = l, x -= 2 (f) X ~ 0
(g) X ;;> y (h) y ;;> 2z, x == 3.
3. Find the length of the segment AB, witljl end points:
(a) ..4(0, O. 0), B(l, I, l)
(b) ..4(1, 2, -1), B(O, -1, 1)
(c) ..4(3, 2, -1), .8(3, 1, 2)
(d) ..4(-1, 3, -2), B(I, 2, 0).
4. Look at the rectangular parallelepiped in Figure 2.S. If its length
..4B - 3.S units, breadth BC - 2 units, and boipt BD - l ,S unit&,
and the coordinates of ..4 are ( I., 2, 3), find die coordinates of all its
vertices and the length of it• diaaonala.
 2.2 VECTORS-ADDITION AND ICAf..AR MULTIPLICATION/ 37

FIGURE 2.5

2.2 VECTORS-ADDITION AND SCALAR

MULTIPLICATION
We start with the geometrical concept of a vector. A vector is a direct-
-+
ed lane segment PQ with an initial point P and a terminal point Q. In
--,. -+
Fiawe 2.6 PQ and RS are two vectors. The length or magnitude of Tcctor

FIGURE 2.6

-
PQ is the distance between P and Q and is denoted by I PQ I . A vector
of unit lenatb is called a u11it ,ector.
Two vccton o; and o•J• are said to be parallel if the line segments
OP and O 'P' are parallel. Two parallel vectors may have the same' or
opposite direotions. In aeometry two parallel vectors are considered to
be equivalent if they have the same len9th and the same direction. In
Pi1ure 2.7 o1. O'I, and i1 are pan.Del wc:ton. and o1 and i'J aro
equivalent vecton, HoWftll', in mechanics the need may arile to eonsider
38/ VECTORS

P'
flOURil 2.7
-+ ~
them as different. For example, OP and RS may represent two forces
acting at two different points O and R. In our study of vectors we shall
ignore this distinction between initial points. We shall assume that there
-+
is always a coordinate system with an origin 0. Given any vector AB,
-+
unless otherwise stated. we shalJ always think of a parallel vector OP
-+ ~
with an initial point O and a terminal point P such that OP and AB have
--+ -+
the same directions and I OP I ==- I AB I , as in Figure 2.8.
y z
B

lal X (bl
-----...v
FIGURE 2.8
-+ -+
Note that, given AB, we arrive at only one OP in this way. Thus, all
our vectors shall start from a fixed point, namely, the origin, in the chosen
coordinate system. Such vectors arc also called free vectors.
y y

o'-----------x
lal lb)

FIGUBI 2.9
 2.2 Vt!C~--ADDITION AND SCALAR MULTIPLICATION/ 39
~
Consider the vector OP in a plane, where P is the point with coordi•
nates (a1 , aJ relative to the chosen coordinate system referred to O as
origin (Figure 2.9).
The sum of two vectors -+ -+ is given by OR,
OP and OQ - the diagonal of
the parallelogram whose adjacent sides are OP and OQ (figure 2.10). If
P and Q have coordinates (,1 1, a 2) and (h1 , b1 ), respectively, tbeo elementary
geometry tells us that the coordinates of R are (a1 + bu a1 + b1). (Wby ?)
Suppressing the directed 1ine segments and writing only the coordinates
of tbe vertices, we have Figure 2.10 (b) ,

V V
R "
R<a,•b1,a2+b2>
I
I
•P<•1••2>
I
I
I
---1
I
I
I
-
Q(b1,D2)
I I

X0 l(
<a> lb)

flGUllE 2.10

--+
Geometrically, AOP, with A> O. is defined as the vector whose length
-+-
is ~-times the length of OP, i.e. ~ I OP I , and whose direction is the same
as that of ----- --+ denotes a vector of length _-,... I ~
OP. If ,\ < 0, then .\OP OP l in
-+
a direction opposite to that of OP. As before we get the pairs of diagrams
as shown in Figures 2.1 I (a) and (b), and 2.1·2 (a) .ind (b}. (In these dia-
grams I A I > I.)
An analogous situation arises in space, where the point P bas the
coordinates (ai, a,, a 1 ).

A>O. OP'=AOP V
V
p•

I
I
----1II

I
I

0 )( 0 ' X
Cb)
fIGVRI 2.11
 40 / VECTOllB
7'<0, 6P'=A~
V
Vt

p • P(a1'"2>
,------- X 0 X
: 0
I
I
I
p' •Pha,,A•z>
<a> (bJ

FIGURE 2.12

If we carefully analyse the transition from (a) to (b) in Figures 2.10,

2. t 1, and 2.12, we observe that the coordinates a 1 and a 1 in the plane (and
the coordinates a1, a1 , and a 3 in space) of the point P perhaps contain every-
-+
thing that we want of the vector OP. It is this observation that leads to
the following algebraic definition of a vector and to the succeeding defini-
tions in this chapter.
2.2.1 Definition {a) A plane i·ector is an ordered pair (a1 , a 11 ) of real
numbers.
(b) A spaee ,·ector 1s an ordered triple (a1o a1 , a 3) of real numbers.
We shall not make any distinction between the plane vector (a1, a.) and
-+
the directed line segment OP, where O is the origin and P is the point
whose cartesian coordinates are (a1, a1). In fact, we shall very often
write
" .=- (au a.) -
= OP.
In this case the vector (a1, a 1 ) is also called the poaition vector or P. The
vector (0, O) is called the zero vector in a plane. Similarly, in the case of
-+
space vectors, we write • - la1, a1 , 0a) - OP. The vector (0, 0, 0) is
called the zero ,ector in space.
-+
The lenath or magnitude of tho vector • - (01, a.) =- OP is tbe lenath
of the sepncnt OP. It is v(al + al), In the case or the space vector
-+
• ... (a,, a., a.> == OP, we have I • I - v(ol + a: + al).
&ample 2.2 The Ien1th of the vector (3, 4) i1 v'(9 + 16) - 5.
Ba.ts'III 1.J TIie l•ath of the vector (1, 3,- 1) is v(l + 9 + 1) - vll.
Bzlun/lle 2.4 Tbe vector (v2/10, -7\/2/10) is a unit vector. 1iace its
lenatb ii 1. .
._,,,,le 2.S
iuleD&lbisl.
Tbo we&or (1/vl. 1/v'3, -lf'\/3) it a unit vector, lioc:e
 2.2 VICTORI-ADDinON AND SCALAR MULTIPLICATION/ 41

2.2.2 Deftaltloa (a) The set of all plane vectors (i.e. the set of all ordered
pairs of real numbers) is denoted by J"1 •
(b) The set of all space vectors (i.e. the set of all ordered triples of
real nun1bers) is denoted by J"1 •
Note that J"1 is the cartesian product R x R. For this reason J"1 may
also be denoted by Ra. Similarly, Y8 - R X R X R - ~.
2,2,3 Definition (Equality) (a) Two plane vectors (01, a 1) and (bi, b1)
are said to be equal if 0 1 = b 1 and a 1 - b1, i.e. (ai, a.) = (bi, b.)
if a, - b,, I -= 1, 2.
(b) Two space vectors (ah a., aa) and (bi, b1 , b3) are said to be
= =
equal if a1 == "1, a, b1, and aa ba, i.e. (ai, a1, a3) = (bi, b1, ba)
if a. = b,, i = l, 2, 3.
The direction of a nonzero vector in V1 is the radian measure 8,
O ~8 < 2 ,c, of the angle from the positive direction of the .x-axis to
-+ -+
the vector OP measured counter-clockwise. If • == (a1, a.) = OP, then
the direction e of • ls given by

sin O - va, 2 a.
~ a.1 ' cos 8 = v0 1.., + a.•
01

In V3 the direction of a vector is given by its direction cosines, which

we shal1 study in § 2.4. Note that the direction of the zero vector is
undetermmed.
Example 2.6 The direction 8 of the vector (1/vl, 1/v'2) is given by
sin 8 = l/...J2 - cos 8. So 8 = ,c/4.
Exampl~ 2.7 The direction of the vector (3, 4) is given by sin 8 - 4/S,
cos 8 == 3/S.
Now we shall define the addition of two vecton in a plane as well as
in space. These definitions correspond to the geometrical definition of
addition.
2.2.4 Deflnltion (Addition) (a) Addition of vecton in J"1 is defined as
(au a1) + (b1, b1) · - (a1 + b11 a, + b1)
for all vectors {a1, a,), {blt b1) in V1•
(b) Addition of vccton ia J"1 is defined as
{I.Ii, Ga, a,) + {b1, b11 b1) :i=s (Gi +
b1 , O. + b1, Ga + bi) ,
ror all vectors (q11 a., Ga), (b., b1, ba) in Y,.
Example 2.8 Tbe sum of the vectors (3, -1) and (-1, 1) is the vector
(3, -1) + (-1. 4) -= (2, 3).
Example 2.9 Tho sum of the vecton (-1, 0, 7) and (3, 2, l) is the
+
vector ( - l, O, 7) (3. 2, 1) - (2, 2, 8).
42 / VEC'l'OIS
Obviously. these operations of addition •+ •are binary operations in Y1
and Y1, respectively. We shall now prove that Y, and Ya are commutative
groups for these operations.
2.2,5 Theorem V,.(V8), the set of all plane (1pace) ,ectors, under the ope-
ration of addition, as in Definition 2.2.4, is a commutall,•e group.
Proo/: The proof for Y3 is left to the reader. To prove that 'V1 is
a commutative group, we check the following :
GI Vector addition is associatl'le
let • - (ai, a,), • - (bh b2 ), and w - (c1, c1) be three vectors in
Y1 • Then
(II + •) + " = ;:. h 1)
((a1 +
e 1, (a1 + b8) + c1 )
and • + (• + w) =
(a1 + (b1 e1J, a 1 + (b. +
<"J)) • +
But we know that for real number& a,, b1, and e, (cf§ 1.6)
(a, + b,) + e, = a, + tit, + c,), i = 1, 2 .
Therefore, (• + •> + "' = • + (• + w) for all u, •• "' E J' 1•

02 Existence of identity
For any • = (a,. a 1 ) in V1 , we have
(0, 0) + (a1, a,.) = (a11 a1 ) -= (ai, a 1 ) + {0, 0) •
So the zero vector (0, O) plays the role of identity. The zero vector
usually denoted by O.
03 Existenc.e of inverse
Given a vector 11 - (a1, a,), the vector z - (-ai, -a.) satisfies
z+•== 0 == ■ +.¥.
This vcctur " is called the additive in"lerse of • or the negati,e of • and
ii denoted by-•·
04 Vector addition is commutative
Let u - (a1, a.) and • =
(61 , b1) be two vectors in Y1 • Then
• + , ==
(01 +
61, a1 + b1)
aad •+• -
(61 +
a1, b1 +
a.) .
But we know that for real numbers a 1 , a 1 , 61, b8 (cf§ 1.6)
+
a, + b, === b1 a1, I - 1, _2.
Therefore, • +, - , + • for all •• , E Y1•
This completes the proof that v,.
under the operation •+\ ii a
commutative group. I
It may be noted that the additi'vo identity of the group Y• ii &ho vector
(0, 0, 0). We shall use the same symbol O for tbil vector al10. la "• if
• • (Oa, ._, a.), tben -• - -(Oa, a., Ga) - (-11i,-0p -Ila).
1M D tJlfha (Scalar ""'1tt,Ntot1on) Mvltipliation or vecton ia Y1 by
 2.2 VICl'ORS-ADDITION AND SCALAR WLTIPLICATION J 43
a real number is defined u
i\(a1, a1) = (Aa1, ,\a.)
for every (alt a.) E Y1 and every real number A- Similarly, in Y11
A (a1, a.. a 1) -. (A. a1, .\ Os, .\ a,)
for every (a1, a 1 , a 8 ) E J/8 . and every real number A.
Example 2.10 If the vector • = (3, -]). then the vector 2u - (6, -2)
and Ju = (1, -1/3).
Note that scalar multiplication is not a binary operation. (Why?)
However, it does satisfy certain natural properties which we shall list and
prow.
2.2.7 Properties of Scalu Multiplication
For all 1•e,·tors 11. " in V2 ( or in J/3) and real nw11bers at, (i, we have
(a) a(u + v) =-= ar1 + a,.
(h) (.x + ~ II = all + ~u.
(c) a(~u\ = (11~)u = ~(all).
(d) lu = u.
(e) Ou == 0.
Proof: We shall prove (a) as an illustration of the method of proving
these properties. Further, we shall pro,e this m Va only. The rest are
left to the reader.
Let • = (a1 , a 1 , a 3) and , --= (b 1, bi, b.) be two vectors in Y3 , and a.
be a real number. Then
• +, ::::a (a1+ bi, a2 + b1 , aa + ba)
and a(u +- •) = +
(a(a1 +
b1), a(a1 +- b1). a(a1 b1))
= +
(aa1 + ab1• aa1 ab1, ixa3 + ab,) (1)
by properties of real numbers (cf§ 1.6). Further, we have
a■ + a,
=-' a(a1 , "2, 0 0) + a(bi, b,, ba)
== (aa1, «tis, aa:i) +
(arb1, a.b11t abs)
- (C&Di + 0W1, fl"t + @1, Ma + f&ba)
= + «<• •> (by (1)). I
The last two properties. namely, td) and (c), may look trivial to the
reader. In fact, they just say
l(a.. a.. aJ - (lai, la., laa} (Definition 2.2.6)
- (Oi,O.. a.)
and
0(4a, ... -.) - (0.a. Ga.., Goa) (De&nitioa 2.2.6)
- (0, 0. 0).
44 /VEctOU
Let the reader reserve his opinion until be comes to the discussion or
an abstract vector space in Chapter 3. Then he will see the sipiftcanc~
or properties (d) and (e) and also the delicate difference between them.
2.2.8 Deflllitlon The difference of two vectors• and,, written as• - ,, is
defined by• - , - • + {-,).
Exampl, 2.11 u,t • = (3, 4), , = (-1, '3). ~hen• - , = • + (-,)
- (3, 4) + (1, -3) - (4. 1).
-+
We sbaU now find the vector • - , geometrically. Let • = OP
-+
- (ai, a.) and , =
OQ == (b1 , b2). The negative of ,, i.e. -,, ia
-+
(-b,, -61). Geometrically, it is the vector QO which is the same as
-+
OQ' (see Figure 2.13), where Q' is the point whose cartesian coordinates
v·

fIGUlll! 2.13
alO ( -bi, -b1). The vector • - , is the diagonal of the parallelogram
-+
formed by • and ,. It is the vector OR = -QP -= lD1 - b1, a, - b1).
Analogo111ly1 we can find • - , in V1 •
For every nonzero vector ■, the vector r¾i- • is a unit vector in the
direction of•· If • == (a1, a.) :I= 0, the unit vector in the direction of• is
( a1 a1 )
va,• +a.•. vo,• + a.• .
If• ("1, a.. a1) :¢= 0, the unit vector in the direction of• is
( a, a. a1
'"...; "11 + a.1 + 0»1 • v'iil" + a.• + a.• • 'Va,•+ a.•+ a,i).
Example 2.12 If• - (3, 4), the unit vector in the direction of • it:
(3/S, 4/S).
lb«lln/lle 2.13 If • - (1, 1, 1), the unit vector in the direction of • la
(1/4/3. 1/4/3. 1/4/3).
 2 2 verroas-ADDfflON AND SCALAll MULTIPLICAnON / 45

In a plane the umt vector in the direction of an angle 8 is (cos e. sin 8),
as shown in Fi1ure 2.14. It is also calJed the unit vector at an angle 8 with
y
P <cos e. s,n e>

IOP1=1

f-lGURE 2.14

the positwe direction of the x-axis. The umt vectors 10 the positive
directions of the x-axis and the y-axii are (). 0) and (O. 1), respectively.
They are denoted by i andj, respectively, 1.e i = (l, 0) and J - (0, I),
as in Figure 2 I S.
V
(0,1>
j

<1,0)
X X
0 0

z z
(0,0,1)
k

X
•
FIGURE 2.15
In apace the unit vectors along tbe positive directions of the x-axis,
the 1-axi1 1 and the r-axis are respectively (1, 0, 0), (0, l, 0), and (0, 0, 1).
They aro denoted b>- i,, J, and i, respectively, i.e. i - (1. d. O),
J - (0, I, 0), and i ""' (0, O, 1). Though we use the same letters I and
j hero aa in the case of plane vectors, thero cannot be any confusion,
becau• tbe context will always ■bow whether we are talkioa of plane
vectors or apaoe vccton. Tbo importance of unit veeton is brouabt out
by the tollowina theorem.
~ 46 /VICTOU

2.Z.9 'J'lleore• (a) ENry pl0111 NCtor Is of th, form a11 + aJ and nay
r,ctor of this form Is a plan, Hctor.
(b} EHry space rector is of the form aif + a. J + a,> and ner:, Hctor
of this form 18 a space ,ector.
Proof: We shall prove (b). Tho proof or (a) is analogous. Let
(ai, "-• aa) be a vector in Y1 • Then
(Os, a., a,) - (a1, 0. 0) + (0, "a, 0) + (0, 0, a 1)
== 01(1, 0, 0) + a.CO, I , O) + a1(0, 0, I)
== 41i + aJ + a,.k .
Thus, every vector in Y1 is of the form a11 + aJ + a1k. On the other
hand, a vector of the form a 1i + aJ + ask is the sum of vectors a11,
aJ, and aJi of Y1 • Since wctor addition is an associative binary opera-
tion, a11 + aJ + a1t is a vector in Y1 , i.e. a space vector, I
Geometrically, Theorem 2.2.9 means that every vector can be represent-
ed as a sum of scalar multiples of unit vectors along the coonimate axes,
as shown in Figure 2.16.
V Z

<•>
flGURB 2.16

The numbers ai, a 1 are called the component, of the plane vector
• - (a1, a.) == a'J.I + a.J. a1 is called the I-component (or x-component)
and a. is called the }-component (or 7-component). We also say that
a.
ai and are the coordinates of • with respect to I and J.
Similarly, a1, a.,
and a. are called tbo co,npoMnts of tho space vector
• - (Os, a., + +
a.> - "1,I aJ ¥• a1 is the I-component, a. is the
}-component, and a. the I-component. We tllo aay that "1, "9, aa4 "a aro
coordinates of• with respect to l,J, and I. •

Problem Set 2.2

1. Pind the mqnitudo aad diNction of tu followq plane W1eton :
(a) (:Z. 3) (b) (3, -1) (o) (3, 6) (d) (-1, 2)
(e) I + I (f) 21 - U {I) -I + 4J (II) -41 -',/.
 2,2 VBC'IOU-ADDITION AND IICALil. MULTIPLICATION / 47

2. Find the magnitude of the followin1 apace '¥'cccort :

(a) (2, - 1, 3) (b) (3, 0, 4)
(c) 3i + 2} - k (d) - I - 2J + 4k.
3. Simplify
(a) 3(2, - 3) (b) 4(1, - 1, 3)
(c) 2(1, - 1) - 7(5, 1) (d) 3(2, - 1, 4) + 4( - 1, S, 0)
Ce) 7(31 + '1J - k) - 4(21 - J + k).
4. Express the following vectors in terms of the unit vectors i, }, and l :
(a) (2, 3) (b) {-1, 4) (c) (-3, -5)
(d} (S, 3. -2) (e) (2 ,0, 1) (0 (-1, 2, 0).
5. Express the following space vectors as sums of scalar multiples of
• -= (1, 1, 1), • == (1, 2, 3), and " == (2, 3, 5) :
(a) (I, 3, 5) (b) (7, - 1, 3} (c) (2, -2, 1)
(d) i + j - k (e) - 2i + J -Sk (0 3i - 7} - 2.t.
(Hint : Assume ,each vector to be equal to (« ■ + , • + r 11), where
«, J. and y are real numbers.)
6. Find a unit vector ■ in the direction of each of tile vectors in
Problems 1 and 2.
7. Find the unit plane vector in the direction :
(a) 8 == -3s/4 (b) 8 - s/6 (c) 8 -
7n/4.
-+
8. Given a point A, determine a point B such that the vector .A.B 11
equal to the vector , in the following :
(a) • =- (1, -2. 3), .A. (0, 1, -1)
(b) , == (3, 1, -2)1 A (-1, 21 3)
(c) , == (3, -7, 1), A (1, 5, 0)
(d) • - (1, -1. 1), .A. (2, 3, 5)
(e) , = (1, -2), A (1, -2)
(f) • = (-1, 5), A (1, -3). ·
9. Given the midpoint of AB. determine tbe points A and B such that
-+
the vector A.B is equal to the ~tor , in the followin1 :
(a) , - (1. 3), midpoint of Aa ii (1, 2)
(b) , - (-1, 2), midpoint of/.A.B is (0, 0)
' .
(c) , - (3, -1, 2), midpoint of Alls (3. -1, 2)
(d) , • (-2, 0, 1), midpoint of A•
ia (2. 3, -2). ·
, 10. Prove tbat Y1 topther with die opllalion of ftCtOr addition is a
comm11tatift poup.
48 / VECTOllS

-+ -+
J1. Let • == OA· and • - OB be two vectors. Then prove that
-+
•- • = B.A.
12. Prove the properties of scalar multiplication (b) and (c) stated
in§ 2.2.7.
13. Prove that a•
= 0 iff « = 0 or • = O.
14. Let • be a nonzero vector in Y1• Then prove that the set
S = {au I « E R} is a group under vector addition.
15. Let u and , be two nonzero vectors in V1 • Then prove that the set
S == {rxu + ~• \ «. ~ E.: R} is a group under vector addition.
16. True or false?
(a) i(2, 0, 4) = (1, 0, 4).
(b) (a + b, b + + + (c, a, b) =
c, e a) (a -+ b + c) (1, 1, 1).
(c) 21 + 3j = 3j + 2i.
(d) ; + J + k is a unit vector.
(e) 0 - (a, b)== (-a, b).
-+ -+ -+
(f) OP - OQ = PQ.
(g) If u and • are vectors, then u bas not been defined so far.
(h) «b = O.
(i) Ou = o.
(j) The zeros on both sides of (e} in § 2.2.7 are different.
-+ -+ -+
(k} Let .ABC be a triangle. Then AB + BC + CA == O.

2.3 DOT PRODUCT OF VECTORS

We start this article with the definition of the angle betV1een two
vectors.
-+ -+
2,3.1 Deftnltlon Let " =
OP and , = OQ be two nonzero vectors.
Then the angle between • and ,, u ::/=, ;\ ,, is defined as the angle
-+ -+
of positive measure 8 between OP and OQ interior to the triangle
POQ (Figure 2.17). If u = A,, then the angle 8 is defined as 8 = O,
if A > 0, and 8 = w, if A < O.
It follows from Definition 2.3.1 that O ~ 6 ~ ir. We shall now prove
the following theorem, which sives the ansle between two vectors,
-+ -+
3,3,2 Tlaeor• Let • -= (aa, a1 , aJ - OP and • =
(bi, b1, b.) - OQ
be two nonzero Yectors In Y.. Let 8 be the angle between • and ,.
11,en
-+ -+
+
I OP I I OQ I cos 8 - a,b1 aJ,1 a-61 • + (1)
Proof : Ir O ~ 8 < 1r, consider tho trianaJe POQ (ace Fiaure 2,17).
 2.3 DOT PllODUCT Of VECTOIU / 49

)(

FIGURE 2.17

or 21
--+
We have QP

--+
OP
- - = -
OP -
- OQ. From geometry, we have
.....
I OP 12 + I OQ lz - 21 OP I I OQ I cos 6 = I QP 11
--+
I I OQ I cos 8 = I -OP I 2 + I OQ
- --+
12 - I QP j2
- -
I (ai, a2, a 3) I2 + (bi, b2, b3) 12
- I (a1 - bi, aa- b2, a 3 - b 3) 1 1
= (a? + al' + a + (b/ + bl + bf)
3 2)

-(a1 - b1)] - (a2 - b2)2 - (a3 - b:)

Therefore, - -
I OP I I OQ I cos 8
= 2a1b1
== a 1b1
+ 2a b + 2a b
+ab +ab
Now let 6 = 0 or O = n. In these cases we have OP = ~OQ,
2 2
2 2

3 3,
3 3•

--
where ~ > 0, if~ = 0, and ~ < 0, if 6 = 1t, i.e. ~ = I ~ I cos 6. Thus,
(a 1, a2 , a 3) = ~(b 1, b 2, b3), i.e. a, = ~b,, i = 1, 2, 3.

Hence I~
OP- -
I I OQ I cos 6 == --+ 2
I ~ I I OQ I cos 8
= I ~ I (cos 8)(b/ + b/ + b32 )
= ~(bi'+ bl + bs2)
= a 1b1 + a2b1 + a 3b3 ,
because a, = ~b;, i = l, 2, 3.
A similar result can be obtained for plane vectors. We have only to

and , == (b1, b1) -

suppress the third coordinates a8 and b3 • In this case, if r,
= OQ, we have
-+- I I OQ
I OP == a1b1 + aab,,
- I cos 8
= (a1 , 1a)
--+
~ OP

(2)
where 6 is the angle between the vectors • and ,.
so I VICJ'ORS
Example 2.14 let • = (3, 2) and , = (2, S). Then the angle 8 between
the vectors• and , is given by
I (3, 2) I I (2, 5) I cos 8 == 6 + 10 .
So 8 16 16
cos = v'l3 v'29 """ v'377
Example 2.15 Let • == (1, -1, 3) and , = (3, I, I). Then the angle
8 between the vectors u and , is given by I (1, ~I, 3) I I (3, 1, I) I
cos 8 = 3 - 1 + 3. So cos 8 = S/11.
Now we shall define the dot product of two vectors. Actually, there
are two different ways of multiplying two vectors. In one the product is
a real number, whereas in the other the product is a vector. In this article
we shall consider the first kind of product. The second kind will be dealt
with in Chapter 6.
2.3.3 Definition Let u = -OP and .., = OQ -+ be two nonzero vectors and
6 be the angle between them. Then the dot product (also called scalar
product or inner product), written as " • ,,, is the real number defined
by
-+- -+-
,, · • = I OP I I OQ I cos 8 = I r, I I ,, I cos 8 . (3)
If either" or Y is the zero vector, then r, • • = 0.
This definition is based on the geometric concept of the angle between
two vectors. However, Theorem 2.3.2 gives an algebraic expression for
the dot product of" and ,, in terms of their components. Hence, Definition
2.3.3 may be equivalently written as follows.
2.3.4 Definition Let " = (a1, a 2) and v = (h1, h1) be two vectors of Y1 •
Then the dot product 11 • Y is defined by
11 • • = a1b1 + a,.b 8 • (4)
If u = (ai, a8 , a 3 ) and , = (b1 , b1 , b3) are vectors of V3 , then 11 • ,
is defined by •
(S)
Note that the formation of the dot product is not a binary operation.
(Why?) However, it satisfies the following properties.
2.3.5 Properties of the Dot Product
Let "• •• w be vectors in Y3 ( or In Y1 ) and«, ~ be real numbers. Then
(a) u · • = I• I 2; hence 11 • • ;> 0. (6)
(b) u · 11 =- 0 i/J• = 0. (7)
(c) • • , = , • 11. (8)
(d) • • (, +
w) =r • • , + ." · w. (9)
(e) (a) • , • cz <• •,) - • . («,). (10)
 2.3 DOT PllODUCT OF VECTORS / 51

Proof: We sbaJI prove these properties for vectors in V8• Analogously,

by suppressing the third coordinate. we can prove tbese properties for
vectors in V1 • Let • = (ai, 0 1 , a8 ), • =
(b1, b1, b8), w == (c1 , c1, c1), and«
be a real number.
(a) • · • == a12 + a.8 + a;1 = I• I•.
(b} If • = (0, 0 1 0), then " • • - 0. Conversely, 1f • • • = at'
+ a.8 + aa" == 0, then a 1 .... a1 = a8 = 0, because a11, al'.
and as" are all nonnegative. Hence, 11 == (0, 0, 0) .... O.
(c) • • , - a 1b1 + a1b1 + a8hJ == b1a 1 + b1a, + b3a3 = • • • •
(d) • • (• + w) = a 1(b1 + c1) + a1,(b1 + <'1) + a3 (b3 + c3 )
=- (a1b1 + a-61 + aaba) + (a1C1 + a.e, + Oa<'sl
=a•,+ a• w.
(e) (u) · , == (aca1)b1 + (a:a1)b8 + (a:u 3 )b3
= a:(a1b1) + ot(a1 b1 ) + a.(a3b.&)
== a:(a 1b1 + a1b1 + a 3b3) = a:(u • •).
Similarly, • • («•) == a:(u · ,). I
Another important property sati&fied by the dot product is the follow-
ing.

2,3.6 Schwarz laequality

For any two vectors II and, of V3 (or V) 1, we have
1•·•1~1•11,1.
i.e.
(11)
Proof: If either • or • is the zero vector, there is nothing to prove.

cose ==
..
So we assume that r, and • are nonzero vectors. Let 8 be the an&le
between• and ,. Then, by Definition 2.3.3, \\e have

l•I l•l
,
But I cos 8 I ~ I, Therefore,
l•·•l~l•I 1•1,
i.e. <• · •>"
~ I • I " I , I 1 = <• · •> (t
· I •>·
Exampk 2.16 The dot product of the vectors (2, 1) and (-1, 6J ia
(2. 1) • (-1, 6) == -2 + 6 == 4.
Example 2.11 The dot product of the vectors (I, -1, 3) and (0, l, -3)
is O + (--1) + (-9) - -10.
Example 2.18 For any real numbers "1, a..
a1 , b1, b1, and b1, wo have
(a,,b1 + a,J,1 + a,J,1)'- <;. ("1.1 + a.• +
a.■) (bi' b11 + b11). +
(Hint : Apply Schwarz inequality to the vectors • - "lai, "-• a.) aud.
, - (bi, ht, ba).)
 52 / VECTORS

Applying the definition of dot product to the vectors i, J, and I, we

have
i • i = j • } =k•k== I {by 2.3.S (i))
and i·J-=j•lc=lc•i=O.
We know that O. • = 0 = • • 0 for any vector •· The converse is
not true, i.e. • • , -= 0 need not imply that either of the vectors • or , is a
zero vector. For, i • j = 0 but ; and j are both nonzero vect.lrs.
. In general, if • and , are nonzero vectors and jf • • , = 0, then in view
of Definition 2.3.3, cos 8 = 0, where 8 is the angle between • and ,. In
other words, • and , are perpendicular to each other. Thus,••• - 0
implies that either • = 0 or , = 0 or • and , are perpendicular to each
other.
2.3.7 Definition Two vectors• and, of Vt (or Y3) are said to be orthogo-
nal if • · , == O.
Note that O is orthogonal to every vector • and i, J, k are pairwi•e
orthogonal.
Example 2.19 Let• = (I, 2, 3) and , -= (2, - 7, 4). Then • • , =2-
14 + 12 = 0. So • and, are orthogonal.
Example 2.20 Let• = (1, 2, 3) and• = (2, - 5, 4). Then 11 • , = 2
- 10 + 12 -=I:- 0. So II and, are not orthogonal.
Every vector • - (a1 , a1 ) in a plane can be expressed as the sum of
two orthogonal vectors a1i and aJ. a1i and aJ are called the resolved
parts of • respectively along and perpendicular to the x-axis. In general, let
• and , be two vectors. If we express • as a sum of two vectors w1 and "•
such that w1 is parallel to •• and "'• is orthogonal to ,, then w1 and "'• are
called the resolved parts of• respectively along and perpendicular to ,.
In such a case, we have
II == WJ + W1,
where w1 = ~, for some real number~ and"• • , == O.
Hence,
■ •• = (W1 + •1) • • == W1 • , + W1 • P
= (~,) . , -= A<• • ,) •
Thus,

Therefore,
•·•
"1- ,-:-, , and "• - • - "1 ... • -
.., ,
- ., , (Fi,ure 2.18).
 2.3 DOT PRODUcr OF VECTORS / 53
p

We have thus proved the following theorem.

2.3.8 Tlaeorem Let , be a gi,en nonzero.,ector. 71,en any nonzero ,ector
• can be expressed as the sum of a ,ector parallel to , and a ,ector
orthogonal to ,.
In the foregoing discussion the vector WJ == ,•~
., , is called the ,ector
projection of• along,.
'
The scalar projection of • on , is I • I cos
between• and,. Obviously, the scalar projection of• on, is
..
8, where 8 is the angle
T,T
, and

.. ..
the vector projection of • along , is
" " "
;:-;•-~r,r
== (scalar projection of • on ,) times tho unit vector
in the direction of ,.
Thus, the magnitude of the vector projection of • along " ;s the absolute
value of the scalar projection of • on " and the direction is the same as
that of ,, if 8 < 1r/2, and opposite to that of"• if 8 > 1r/2. If 8 .... TC/2,
the vector projection of r, along , is O.
Example 2.21 Let ■ == (I, 2, 3) - i + 2} + 3k and , == (-2, 3, 0)
= -21 + 3} + Ok. Then

scalar projection of • on • is
■ ·,
T,T" = -2 +6
v' 13 == v'
4
13
. • .• 4
vector projection of• along , 1s ,.--:-, • - 13 (-2, 3, 0)
8 • , 12 •
= - 13 I ;- 13J ...,.. ua.
I ftL

Abo
8 12 .
u = (1, 2, 3) == i + 2} + 3i =- (- 13 1+ 13 J+Oi)
+ ( f~ I + ~: J + 3k)"
54 / VICTOllS
This is the resolution of • along and perpendicular to ,. The reader
8 12 . ( 21 I
can check that ( - U; + 7"f J + Ok ) 11 along , and. 13 +

:: } + 3.t) is perpendicular to,.

Problem Set 2.3
Work out Problems 1 through 4 for each pair oi vectors • and , given
as follow1,
(a) • = (l, 1),, - (-1. 3)
(b) • = (-2. S). , ..., (1, -6)
(c) • - f + 'JJ, , - 2i - 3J
+
(d) ■ = -3i SJ. , =-= 61 4J +
(e) ■ == (1, 3, S), , = <-3, 1, I}
Cf) • -= (2, -1, 4), , = Cl, 2, -1)
+
(g) ■ == -i + j k, , = 31 - 2} + k
(h) • = +
21- 3} k, , .., I J + k. +
1 Find ■· ,.
2. Find the cosin~ of the radian measure of the angle between • and ,.
3. Find the scalar projection of • on ,.
4. Find the vector projection of , along •·
S. Find the real number II such that the vectors • and , given as follow
are orthogonal.
(a) ■ = (2, G, 1), , """ (4, -2, -2)
(b) u = (1, -2, 1), , == (2, I, 11)
(c) u == (O. 3, -2J, , -= (a., 4, 6)
(d) • = (flt -3, 1), , = (11, 11., 2).
6. For any vector u, prove that
• = (• · i)I + t• · J)j + (■ · k}k.
7. Let• and , he two vectors of V8 (or V1}. Prove that
(a} I• + •I ~ I• I + I • I
(b) I • · , I = I • I I , I 1/f • is parallel to ,.
8. Let • be orthogonal to both , and w. Then prove that • is orthogo-
nal to each vector of the set {11, + ~., I a, ~ any real numbers}.
!>. True or false ?
{a) ■ • , = Oimplies either • == 0 or , - O.
(b) If ■ • , • Ofor atl, e Y., then• - o.
(c) • • (, • ") is meanlngleu.
 2.4 APPLICATIONS TO GEOMETR.Y / 55
(d) (11 • ,)w is a vector parallel to "'·
(e) In space if• •} - 0 and" • k == 0, then• =- i.
(f) O, -1) · ( - I. 1, O) = 0.
(g) I " I - o I.ff• = o.
2.4 APPLICAY.IONS TO GEOMETRY
The main applications that \\C consider in this article are equations to
lines and planes and certain associated problems. First, we shall work out
a few examples to illustrate the po\\er of vector methods.
Example 2.22 In trigonometry, the formula
cos (8 - +) .... +
cos 8 cos + sin 8 sin cj,
is fundamental. We shall prove this formula by using vectors.
Proof: The unit vectors -OQ and --+
OP (see Figure 2.19) at angles 6 and +
V

fIGURB 2.19
with the positive direction of the x-axis are respectively
cos 8i + sin 6} and cos cj,i + sm ♦J.
The angle 8 - +between the vectors OQ
-+ and OP
- 1s given by
-+ ~
COS (8 - ♦) = ~
OP• OQ
- = COS 8 COS cj, + SID. OSin. ♦ ,
IOPI I OQI
Example 2.23 Prove, by using vector algebra, that the three altitudes of
a triangle are concurrent.
Proof : Let AD and BE be the two altitudes of a triangle ABC, meoting
at the point O (see Figure 2.20). It 1s sufficient to prove that CO i's
perpendicular to AB.
Clearly.
-+ -+
AO• BC - 0,
and
_,. ....
BO, CA= O. (1)
56 / VBCTOllS
A

B D C
FIGURB 2.20

Note that we are working with vectors that do not have the same
initial point. Now
-+ -+ --+ -+ -.
CO • AB = CO • (AC + CB)
= -+
CO
-+ - -+
• AC + CO • CB
-+ --+ -+ -+ -+ -+
- (CB+ BO)• AC+ (CA+ AO)· CB
-+
-
-+
CB · AC + -+
BO
--+
· AC + -CA · CB
--+
+
-+ -+
AO • CB
-+ --+ - _.
= CB• AC+ CA • CB (by (I))
-+ -+ -+--+ -+ -+
= CB • AC - AC • CB (because CA = -AC)
-+-+
= CB • AC - -
CB ·-AC = 0
{by 2.3.S {c)).
We shall now take up the standard applications of vectors to equations
of lines and planes in analytic geometry. First, we note that if P (x1 , Y1)
-+
and Q(x1, Y1 ) are two points in a plane, then the vector PQ is given by
--+ -+ --+
PQ == OQ - OP - (Xa - x1~y1 -Yi) = (X1 - X1)i +(Ya -Y1U •
Similarly, if P(x1, Yu z1) and Q(x1, y1, z1) are two points in space, then
-+
the vector PQ is (x1 - x1 )i + (y1 - y1)J + (z1 - z1)k.
2.4. t The Equation of a Straight Line
Consider a straight line L parallel to a vector ■ and passing through
a point P (Figure 2.21).

FlGURB 2.21
 2.4 APPLICATIONS TO GIDMl'l'&Y J 57

·-+
Let Q be a point OD £. The vector r - 0Q (0 ii the oripn) i■ called
the position vector of point Q. !:et • be the po,ition vector ot point l'.
Then
~ -+ -+
r-OQ-OP+PQ
... • + a vector parallel to •
-,+r■, (2)
where, is a real number. This equation is satisfied by all point■ Q on the
line and by no points off the line. (Why?) It is called the vector equa-
tion of a straight line through a pven point P and parallel to a given
vector ■• fbe followina special cues are worth notins-
2.4.2 Line I■ a Pine
Let P be the point (x1, 1 1) and • be tho vector (a, b). Then , = -+
OP
=- (xu 11) and r = OQ - (x, 1). Then the equation of the line takes
the form
(x, y) = {x1, y 1) + l(a, b) = (x1 + ta, Yi + tb) .
This equation can be written as a pair of equations :
x = Xi+ ta, 1 = y1 + tb,
i.e. (x - x1) == ta, (y - y 1) = tb.
These are called the parametric equations of the straight line. Eliminating
t, we get the cartesian equation
a(y - y 1} = b(x - X1) • (3)
If a = 0 or b == 0, vector • becomes a vector along one of the coordi-
nate axes and the line bas the equation x == x1 or y = Yi•
If a ¢ 0 and b #- 0, Equation (3) may be written in the symmetric
form as

(4)

If • is a unit vector at an angle 8 (:;t: ,c/2), then• - (a, b) - {cos 8,

sin 8). Hence, Equation (3) may be written as
Y - Yi == tan 8 (x - X1)
or
1 - Y1 = m(x - xJ • (S)
whero m is the slope of the straight line L. Thus (S} is the equation of
the straiaht line through PCXi, y1) and having slope m.
Thus, given a vector, we have found the equation to a atraipt line
parallel to the vector. The convene 1ituation is contained in the follow-
ing example.
 SB/ VICJ'ORB
+
Example 2.24 Let Ix + my n - 0 be the equation of the ,traiaht
+
line L in a plane. Then the vector -ml /j is paraUel to L and the
vector II + mj is perpendicular to L.
Case l n# 0.
If either / == O or m = 0, then the stiaight hne L JS parallel to one-
of the coordinate axes and the result JS true.
Now suppose that I =I= O and m r 0. Io this case the two pointa
-+ n. n
P(,-n/1, 0) and Q(O, -nlm) be on L. The vector QP 1s - 7• + ,;-J,
Im -+ -+
The vector - QP 1s parallel to the vector QP and hence to the line
n
L. This vector is -mi + /j.
The vector Ii + mj 1s perpendicular to the vector -+
QP and hence to the
line L, because

(Ii + mj) (- ~ i -t -;, J) = - n+n - 0.

Crue 2 n -= o.

Shift the hoe Ix + my == 0 parallel to itself and obtain Ix + my

+ n• - 0, n' #- 0. Using Case 1, we get the desired result.
2.4.3 Line in Space
Let P be the pomt (x1, y1 , z1) and II be the vector (a, /1, c) Then
~
, = OP :::!: (X1, Yi, Z1)
and = (x, y, z) = (x1, Y1o z1) + t(a, b, c) •
r - (x, J 1s the position vector of any point Q on the hne L. This
vector equation may be written in the parametric form
x = Xi -+ ta; Y = ;y1 + tb; z == z1 + tc. (6)
Eliminating t, we get the symmetric form

(7)

provided a, b, and c are nonzero. Thus, we get another form of the equation
of a straight line through P(x19 Yi, z1) and parallel to the vector
•=-al+ b} + ck.
If any two of the numbers a, b, c are zero, say a == b = O, then the
vector• == (0, 0, c) becomes a vector along the z-axis. The equation of
the straight line, from Equation (6), is
X - Yi•
Xi, y - (8)
Note that these two together represent the straight line.
 2,4 APPLICATIONS TO OIIOIIIITRY / 59

If just one of the numbers a, b, c is zero, say a - O, then ■ is the vector

b} + ck. The equation of the straight line, from Equation {6), is
J1-J11 Z-Z1
(9)
X = "1' b - = C 0

These two together represent the straig~t line.

2.4.4 Direction Cosines
. ~

Let «, ~. y be the inchnations of the vector• == OA. = (a, b, c) to the

positive directions of the coordinate axes OX, OY, OZ, respectively, as in
Figure 2. 22. Then cos «, cos ~. cos y are called the direction cosines or

y
•
FIGURE 2.22

the vector • == ai + bJ + ck. These are also called the direction cosines
of a line parallel to • and in the same direction as •· The ordered set
{cos«. cos~. cosy} is called the set of direction cosines of the line re-
presented by vector•· Clearly,
" .i a
I. I I I I -
cos « = (10)
va + b' + c3
1

" .J b
I• I Ii I = va1 .+b"+c1
cos~= (11)

cosy= •·•
I• I Iii
==
va'-+h'+c•
C
(12)

It follows that cor « + cos1 (i + cos• y == 1. For any real numt,er

k, the ordered set {k cos«, k cos (i, k cos r} is called the set of db·ection
ratios of a line parallel to -. Clearly, a, b, c are the direction ratios of tho
line parallel to the vector ai + b} + ck and hence of the line •
X- X1 == JI - Y1 Z- Z1 ,
a b == c
 60 / VBCTOU

2,4.5 TIie Equation of a Pine

Consider a plane pass1n1 throul,h a point Q(xi, Jiu z1) and perpendi-
--+
cular to a vector • - O.A - (a, b, c) (see Fisure 2.23), Let P(x, :,, z)

FIGURE 2.23
--.
be a point in the plane. Then the vector QP =
(x - x1)i +(
y - y1)j
+ (z - z1)k lies in tho plane and is perpendicular to the vector ■ = ai
+ bj + ck. Therefore,

which gives
a(x - X1) T b(y - Y1) , e(z - Z1) = 0. (13)
This rs satisfied by all pomts P on the plane and by no pomts oft the
plane. Equallon (13) 1s called the equation of the plane through Q(xu Yi, z1)
and perpendicular to the vector• = (a, b, c).
Equation (13) of the plane can further be simplified to the form
ax + by + cz + d = 0•
where d = -ax1 -by1 -cz1• Thus, the equation of a plane 10 space 1s
a lmear equation in x, y, and z. Conversely, consider a hnear equation
A.x+ By+ Cz+ D = 0, (14)
where A, B, and C are not ,dl zero. Let A. =I: 0. Then we can write
Equation (14) as
D
A(x+ y) + B(y-0) + C(z-0)-= 0,
which 1S the equation of a plane through the point (-D/..4., 0, 0) and per-
pendicular to the vector A.i + BJ +Ck.
Two planes, if not parallel, intersect 1n a straight line. Thus, a straight
line may also bo considered as the intersection of two planes, and hence
 2.4 APPUCATIONS 'ro OIONBTI.Y / 61

its equatiOJl may also be given as a pair of equat10-.s of planes, e.g.

01X + bJ)I + C1Z + di = 0
c1nd a2x + ba)' f- c1z + da = 0
represent a Jine. From these equations, we can get the equation of th•
line in the symmetric form
X - Xi y - Yt z - Zi
a = b - c
We illustrate this by the following example.
Example 2.25 Find the symmetric form of the equation of the line of
intersection of planes
X -1- 3y - Z + S =0 (1S)
and Sx - 2y + 4z - 8 = 0 . (16)
Eliminating z from the two equations, 1'Ye get
9x +lOy + 12 = 0
or 9x = - lOy - 12 .
Ehmmating x from Equations (1 S) and (16), we get
9z•
y--11 17 0 + 33
= •
or -lOy - 12 = - ~
17
+ 126
17
So, we get
90z - 126
9x=-10y-12= - 17
x-0 y 61S +
=z-17/90
-- -1/S-
or ti9 ==
-1/10
x-0 y + 6/S z - 1/S
or ==
-=Io == 9 17

Problem Set 2.4

1. Prove by vectors that the diagonals of a rhombus meet at right
angles.
2. Prove by vectors that the line segment joining the midpoints of two
sides of a triangle is parallel to the third side and its length is one-half
the lenath of the third side.
3. Find the direction cosines of the line pasains through the po~nt■
All, 2, 1) and .8(3, -1, -1).
4. Find the vector equation of the line passing tbrouah two 1iven
points. ·
--+ -
5. Prove that the points A, B, Care collinear ii/ OC - &OA + •,OB,
--+

where 11 +
J - 1.
 62 / VECl'OJrS
6. Prove that the perpendicuJar distance from the point P to the line
-+
r - , + t• is f,
+ (OP-,)·"• -
••II
DP/·
7. Find the angles between the Jines
x- a y - b z -c x - a' y - b' z - c'
I = m 11 ' /' = ~ -= n'
8. Find the angles that the vector • = 2i - 2} + k makes with the
coordinate axes.
9. Find the equation of the plane perpendicular to the vector 11 - 2i
+ 3} + 6k and passing through the terminal point of the vector
i + Sj + 3k.
10. Determine a vector perpendicular to the plane containing the vecb>rs
r, = 2i - 6} - 3k and , = 4i +
3J -- k.
11. Find the angle between the planes
(a) 2x - y + 2z = 1 and x - y = 2
(b) 3x + 4y - Sz = 9 and 2x + 6y + 6z = 1.
12. Prove that tile distance of the point P 0(x0, Yo, : 0) f1om the plane
ax + by + ez + d = O is
I ax0 + by0 + ez0 + d I .
va'+ll-+c•
13. Prove that the equation of the line of intersection of the plants
4x + 4y - Sz = 12 and 8x + 12y - 13z =- 32 can be written in
the form
x-1 y-2 z
-r-= 3 =4·
14. Prove that the equation of the sphere of radius r with centre
P(x11 Yu Z1) is (x - x1)1 + (y - ,Y1)z + (z - z1)2 = r 2•
Chapter 3

Vector Spaces

In Chapter 2 we saw that the set of all plane (space) vectors forms a
commutative group relative to addition and, further, relative to scalar
multipJication it satisfies certain properties (Properties 2.2.7). All these
properties of vectors are so fundamental in mathematics that whenever
any system satisfies them we give it a special name, namely, vector space.
The precise definition of a vector space follows.

3.1 VECTOR SPACES

3.1.1 Definition A nonempty set Vis called a real vector space (or a real
linear spare or a real linear vertor space) if the fo]lowing axioms are
satisfied:
VSl There is a binary operation•+• defined on V, called 'addition',
VS2 There is a scalar multiplication defined on V. (This means, to
every real n1.1mber a. and every element u of V, we tan associate a
unique element of V and denote it by acu )
V~3 Addition and scalar multiplicat1on satisfy the foil owing :
(a) Vis a commutative group for addition (i.e. GI, 02, 03, and 04
hold for addition m Y).
(b) a(u + ,) = a.u + a.v and (« + ~)u = a.u + ~u for all real
numbers«,~ and all u, , E V.
(c) a.(~u) = (a.~)u == ~(a.u) for all real numbers «, ~ and all
uE V.
(d) lu = u for all u E V.
3.1.2 Remuk A complex vector space is defined analogously by using
complex numbers instead of real numbers in Definition 3.1.1. '
3 1.3 R•uk The real or complex numbers used for scalar multiplication
in the definition of a vector space are called scalars., Throughout
we shall deal with only real vector spaces and complex vector
spaces.
 64 / vacroa SPA.CBS
In Chapter 2 we saw that Y1 and Y1 have addition and scalar muJtipJi•
cation defined on them and VS3(a) holds for addition in both (Theorem
2.2.5); and that (b), (c), and (d) of VSJ hold for scalar multiplication in
Y1 and Y3 (Properties 2.2. 7). Since these properties of plane vectors and
space vectors have been abstracted to provide the general definition of a
vector space, we shall hereafter use f!le word 'vector' to mean 'an element
of a vector space'. In other words, elements of a vector .,pace shall be
called vectors. Plane vectors and space vectors are only specific cases.
3.1.4 Remark Note that here we have not used 11, ,, w for denoting
vectors as we did in Chapter 2. We shall con,1stently use the letters
11, v, w, ... for vectors, and the Greek letters ti, {1, y, ... for scalars.
However, we use the boldface O to denote the zero vector of the
space to distinguish it from the scalar 'zero'.
Clearly, V1 and Y8 are ,ector spaces. There are many other examples
of vector spaces. We shall note a few of them here. In fact, the reader
should be able to visualise several vector spaces before we go deeper into
the subject. Visualisation of a vector space involves five steps :
(i) Consider a nonempty set Y.
(ii) Define a binary operation on Y and call it 'addition'.
~iii) Define scalar multiplication on V as described in VS2.
(iv) Define equality in V.
(v) Check whether (a) to (d) of VS3 are satisfied relative to the
addition and scalir multiplication thus defined.
Example 3.1 Let v. be the set of all ordered n-tupks of real numbers.
Thus, an element of Yn would be of the form (x1, x 1, ... , Xn), where the
x,'s are real numbers. Define addition, scalar multiplication, and equality
in Y11 as follows : If u = (x1, .x1, ... , Xn) and v = (Yi, )'1, ... , Y11) are two
elements of V11, then :
Addition
+V =
U (X1 + >'1, X1 + )'z, ..• , X11 + )'11), (1)
(Note that u + v E Y11, because the right-hand side of Equation (1) i!i an
ordered n-tuple of real numbers. This is called coordinatewise addition.)
Scalar multiplication For a real· scalar 4 1 au is defined as the ordered
n-tuple {1XX1, «x.. .., «Xn), i.e.
(2)
which is again in V,.. (This is called coordinatewise scalar multiplication.)
Equality 11 = (x1, x 1, ... , x.) and , == (Y1, Y1, ... , y,.) are defined as
equal if x, = y,, I == 1, 2, ... , n
Since addition, scalar multiplication, and equality in Y11 have been
 3 I VICTOR. SPACIS / 65
defined, what remains is to check whether (a) to (d) of VS3 are satisfied.
We shaJJ leave most of this checking to the reader. However. in order to
give ham an idea of what such checking involves, we shall now go throuah
some of the detc1iJs.
Commutativity of addition
Let u == (xi, x1 , ... , X11) and " = (11 , y1, ••• , y.) be two members of Y,..
Then u + v = (x1 + y 1, x 1 + y 1 , .. , x. y,.) +
and v + =
u (Y1 + X1, Y1 + X1, ... , Yn x.) • +
The real numbers Xt + y,, and Yt + Xt arc equal, since the commut.1tive
law of addition holds for real numbers. As this argument is true for
every k, it follows from the definition of equality that u + v == v + u.
The alert reader would have now noted that, fundamentally, 1t is the
commutative law of add1t1on among real numbers that gives the commuta-
tive law of addition in Y.. If this fact is remembered, we can check that
u + v = v + u even mentally.
Existence of additive identity
We have to detemune an element 0, which has the property
0 + u = u. In this case we find that O - (0, 0, .. .. 0), because
(O. 0, ... , O) t (x1, x 1 , .. , x,.) == (x11 x 1 .... , x,.) for all (x1, x1 , ... , x.)
EV,..
Existence of additive inverses
Here we have to determine an element -u for the vector u €'" Y. such
that -11 + u ..., 0 = (0, 0, .... 0). Ir u = (x1, x,, , .. , x 11 ), then we can
easily see that -u = (-x1 , -x1 , ... , -xn),
We assume that the reader can check the remaming properties. Once
(a) to (d) of VS3 are checked, 1t follows that v., for the addition and scalar
multiplication defined an (I) and (2), is a vector space.
It may be noted that, by Theorem 1.6.3. 0 = (0, 0, ... , 0) is the unique
zero of the space and -u == (-x1, -xi, ... , -x.) is the unique negative
of u - (x11 X1, ... , x.).
Thus, V,. is a real vector space, since we have used only real scalan.
Cao we use complex scalars in this case? No. ·for, suppose • is a
complex number. Then "" == (ca11 ca1, .. , ca,.) is not in v.,
because
the numben u 11 are complex and v.
contains only n-tuples of real numben;
10 scalar multiplication is not defined. See, however, Problem 2.
Tbe 1pecial cue, n - 2 and n - 3 of :Example 3.1 live u1 the vector
IJ'ICII Y1 and Y1• The special caeo n - 1 gives the apace. Y11 which ii
aotbiaa but the apac:o or real numben, where addition ii the ordinary
or
acldidoa real n•ben and ICllar multiplication i1 the ordinary multi-
plication of real n-amben.
66 / VECTOR SPACBS

Example 3.2 Let :1(1) be the set of all real-valued functions defined on
the interval l With pointwise addition and scalar multiplication (cf§ J. 7),
:1(1) becomes a real vector space. The zero of this space is the function
0 given by O(x) =
0 for all " E /.
Example 3.3 If, instead of the real-valued functions in Example 3.2, we
use complex valued functions defined on T and pointwise addition and
scalar multiplication, then we get a complex vector space {using complex
scalars). We denote this complex vector space by ~ c(/).
Example 3.4 Let 9(1) denote the set of all polynomials p with real
coefficients defined on the interval I. Recall (cf Example 1.39) p is a
function whose value at x is
p(x) = ,; + a1x + .. . +11.X" for all " E /,
where «,'s are real numbers and n is a nonnegative integer. Using point-
wise addition and scalar multiplication as for functions, we find that 9(1)
is a real vector space. If we take complex coefficients for the polynomials
and use complex scalars, then we get the complex vector space 9c(l). In
both cai.es the vector O of the space is the zero polynomial given by
O(x) == 0 for all x E I.

NOTATIONS

~ [a, h] == the set of all real-valued functions defined and cont1•

nuous on the closed mterval [a, b].
~ <1l(a, b] = the set of all real-valued functions defined on La, b] and
whose first derivatives are continuous on [a, b].
Y!/ <•l[a, b] = the set of all real-valued functions defined on [a, b),
differentiable n-t1mes and whose n-th derivatives are
continuous on [a, b]. These functions are called n-times
continuously diff'erentiable functions.
Example 3,5 g='[a, b], ~ <11[a, b], ~ <•>[a, b] are real vector spaces under
pomtwise addition and scalar multiplication. We have to use the fact from
calculus that the sum ot two continuous (differentiable) functions is conti-
nuous (differentiable) and any ~alar multiple of a continuous (differenti-
able) function 1s continuous (differentiable).
By changing the domain of definitions of continuity and differentiability
to the open interval (a. b), we get, similarly, the real vector spaces '6' (a, b)
and '6'<•>(a, b) for each positive integer n.
3.1.5 Remark By changina real-valued functions to complex-valued func-
tions and using complex scalars. we get the complex vector apace■
~ c(a, b] and <G' c(a. b).
3.1.6 Remark In each of tbele caaea the reader should convince himself
that the sets concerned are •vector spacca' under tbe definition■
 S.1-VECTOR SPAClll / 67

mentioned. He should check the axioms in each case until ho reaches

the stage where he can quickly say \\hether the axioms are satis-
fied or not. In any case no axiom should be taken for granted. In
all cases the zero element of the space and the negative of any given
element of the space should be clear m the reader's mind.
Example 3.6 Let ~<oo >[a, b) stand for th~ set of all functions defined
on [a. b] and having derivatives of all orders on [a, b]. This is a real vector
space for the usual operations. It 1s called the space of infinitely differenti-
able functions on [a, b].
We conclude this article with a theorem giving certain immediate conse-
quences of the definition of a vector space. Recall that, in the definition
of a vector space, we have property (d) of VS3, namely, tu - u for all
u E V. It should be noted that this apparently trivial axiom is crucial"to
the development of the theory of vector spaces.
3.1.7 neorem ln any vector space V,
(a) «O •-= 0 fqr every scalar «.
(b) Ou = 0 for every u E V.
(c) (-J)u - -uforevery u EV.
Proof: (a) «O = «(O + 0) (by 02)
- «O + «O (by VS3(b)).
Adding -(«O) to both sides, we get
O = -{atO) + («O + «O)
= (-(atO) + (110)) + «O (by 01)
= o+«o (by 03)
= «O (by 02).
(b) Ou = (0 + O)u = Ou+ Ou (by VS3(b)).
Adding -(Ou) to both sides, we get
0 =-~ -(Ou) + (OU+ Ou) =- ( --(Ou) + (Ou)) + Ou (by 01)
= 0 + Ou (by 03)
= Ou (by 02).
(c) (-l)u + u = (-l)u + lu (by VS3(d))
== (-1 + l)u {by VS3(c))
= Ou = 0 (by (b)).
So by 03 and the uniqueness of the negative. {-l)u is the ,neptive of 11,
i.e. (-l)u .... -u. I
It is convenient to write 11 - , for u + (-,).
68 / V!!CTO.R SPACES

Problem Set 3 1
1. Let u1= (1, 3, 2, 7), Us - (0, 2, -3, 5), u3 =- (-1, 3, 1, --4), and
"• = (- 2, 16, -1, S) be vectors of v.. Then evaluate
(a) u1 + Ui (b) 2ui + 3u8 - 1u• (c) u1 + 2u8 + 3u8 - u,
(d) 3u1 + (u,-u1) (e) (au 1-bu2 ) + au3 ,
2. Consider the set VS of all ordered n-tuples of complex numbers. By
defining equality, addition, and 'iCalar mult1phcat1on (with complex
scalars) as in Jlf;, prove that V;
is a complex vector space. Is it a real
vector space ?
3. Let R+ be the set of all positive real numbers. Define the operations
of addition and scalar mult1plicat1on as follows :
u + v =- u•"for all u, v E R+
a.u = 1P for all u E R+ and real scalar a..
Prove that R"' 1s a real vector space.
4. Which of the follo\\ing subsets of V, are \ector spaces for coordinate-
wise addition and scalar multiplication ?
The set of all vectors (Xi, x8, x3, .x,) E V, &uch that
(a) x, = 0 (b) x1 == 1 (c) x 2 > 0
(d) xi :>
0 (e) .xf < 0 (f) 2x1 + 3x2 =- 0
(g) X 1 + fx 8 - 3x3 + x, = 1.
S. In any vector space prove that rxu = 0 iff either rx = 0 or u = O.
6. Which of the following sub,ets of f/l are vector spaces ?
The set of all polynomials p such that
(a) degree of p ~ " (b) degree of p = 3
(c) degree of p :> 4 (d) p(l) = 0
(e) p(2) = I (f) p'(l) - 0
(g) p has integral coefficients.
7. Which of the following subsets of'(; (0, 1] are vector spaces?
The set of all functions / E W [0, l] such that
(a) /(1/2) =- 0 (b) /(3/4) 1 =
(c) /'(x) =- xf(x) (d) /(0) - /(1)
(e) /(x) =- 0 at a finite number of points in [O, l]
(f) / has a local maxima at x = 1/2
(a) / bu a local extrema at x - 1/2.
8. Let Y be a real vector apace and X an arbitrary set. Let y.r be the set
of all functiona /: X-+ Y. Prove that y.r ia a real vector apace for
pointwile addition and scalar multiplication. where definitiona are
analogous to tboae for real-valued fu6ctioo1.
 3.2 SUBSPAC!S / 69
9. True or false?
(a) In a vector space V, (-1) <-u) - u for all u E V.
(b) In a vector space V, --u - v ::a -v -.u for all u, v e: Y.
(c} In a vector space V, -u - v == -(u ..1.. v) for all u, v E V.
(d) In a vector space V, -(-u) = u for all u E. V.
(e) R x R 1s a vector space.
(f) If the scalars are complex. numbers, then ~(/) 1s a complex vector
space.
(g) In C, the set of complex numbers considered as a complex vector
+ ~•
space, 1f ot • 1 = 0, then « == O == ~-
3.2 SUBSPACES
3.2.l Deftnitlon Let S be a nonempty subi.et ot a vector space V. S 1s
said to be a subspace of V ,r S 1s a vector space under the same
operations of add1t1on .ind i.c.c1l,tr mult1phcauon as m V.
In order to unq_erstand this defimuon as well as the future concepts 10
the subJect, we shall repeatedly consider the concrete cases of V1 and V3 •
Y1 1s the Buchdean plane. Take r1n) ,;tra1ght hne S through the origin
O. Any point P on this straight bne can be considered a& a ,ector OP of

-
~ -,.
V1 in S. The sum of two such vectors OP and OQ, \\here P and Q both
he in S, is agam a vector OR, where R hes m S. S1mllarly, a scalar multi·
pie of any vector in S 1s agam a vector m S. All other axioms of a vector
space are automatically satisfied 1n S. So S 1s a vector space under the
same operations as an Y1 • Thus, S is a subspace of V1 • In other words, C\er1
line through the origin is a subspace of V,.
In the same manner, in V3 we find that any plane S through the origin
1s a subs1'ace of V3 • Also, every straight hne L through the OJt81n 1s a sub-
space of V1 • The followmg example~ further illustrate the concept of a
subspace.
Example 3.7 Let L be the set of all vectors of the form (x, 2x, -3x, x)
in Y,. Then L is a subspace of v,.
Proof: If u == (x, 2x, -3x, x) and v = (y, 2y. -3y, y) are two elements
of L, then clearly
u +, =- (x+ y, 2(x + y), -3(x + y), x + y)
ucl «u == (cxx, 2(ocx), -3(cu), atx)
allo beloq to L. The zero element (0, o. 0, 0) 11 also of this form and
_ heace belonp to L. The negative of u is -u - (-x•.-2x, 3x, -x),
which ia apin of the same form and bcooe belon,p to L. The other law,
of auociatMty and commutativity for additioa, clitvibutivo laws, and the
ftnal uiotn lu • u are all true in L, becauae elem.Inti of£ are elemcnta
 ot V,. and in V, a\\ these \aws are true. Thus, Lis a subspace or Y1 , Tn
fact, L is the 'line' through the point (1, 2, -3, 1) and the or1gm
(0, o, o, 0) in r•.
... .
Example J 8 Generalising Example 3. 7, we can say that all scalar multi-
ples of a givea element u0 of a vector space V form a subspace of Y.
Proof: Let the set of all scalar multiples of u0 be denoted by [uo]. If u and
v are two vectors in {UoJ, then u = cxu0 for some ex and v = ~U0 for
some~-
u +, = cxu0 + ~u0 = (cx + ~)u0 , i.e. u + , is a scalar multiple of
Uo- Hence, u + v E [uJ. Again, Au = A(cxu0 ) = ().cx)U0 , i.e. Au is also
a scalar multiple of u0 • Hence, Au E [u.,] for all scalars;. and u E [u0].
0 E [u0], because O = 011o by Theorem 3.1.7. If u E [u0 ], then
-u = (-l)u by Theorem 3.1.7, and so -u E [uJ. The other axioms,
which are only interrelations, are true for all elements of Y and so are
true for all elements of [u0]. Hence, [u0] is a subspace of V.
Example 3. 7 is a special case of Example 3.8. The subspace of V,
considered in Ex.ample 3.7 is just [(1, 2, -3, 1)).
In these examples we note that to prove that S is a subspace of V we
explicitly checked only the following :
(i) The sum of any two vectors in S is again in S, i.e. addition is
closed in S.
(ii) The scalar multiple of any vector in S is again in S, i.e. scalar
multiplication is closed in S.
(iii) The existence of O in Sand the existerice of a negative for each
element in S.
The other axioms were not explicitly checked, because this, as the
following theorem shows, was not necessary. In fact, the theorem says
that even (iiy need not have been checked.
3.2.2 Theorem A nonempty subset S of a vector space V is a subspace of
V iff thefollowi11g conditions are satisfied:
(a) If u, v E S, then u + v E S.
(b) /Ju E Sand ar. a scalar, then cxu E S.
In other words, a subset S of a vector space V is a subspace of V IJf it
is closed under addition and scalar multiplication defined in V.
Proof: Let S be a subspace of V. Then S is a vector space under
the same operations as those of V. Hence, S satisfies (a) and (b). Con-
versely, if (a) and (b) are satisfied, then we have to prove that S satisfies
all the axioms of a vector space. VSl and VS2 are satisfied for S,
because this i1 exactly what (a) and (b) 1ay. We shall now show that
VS3 for S follows from VSl and VS2 for S.
 3 .2 SlJBSPACIS / 71
Ou -= 0 for any " E: J' and therefore for any u0 E S. Takin1 any
Uo E: S, we see that, from (b), 0 - Ou0 E S. Similarly, (-l)u == -u
for any u E J' and therefore also for a siven "u E S. Hence, -r,0 e S
for every Uo E S. Thus, 02 and G3 hold in S.
Axioms GI, G4, and (b), (c), and (d) of VS3 are automaticaUy satisfied
for elements of S, since these laws already bold for elements of Y. Note
that this needs no checking, because these axioms are interrelations of the
elements. In the cases of G2 and 03 the checking was necessary, because
we had to show the existence of certam elements in S. I
3.2.3 Remark In view of Theorem 3.2.2, whenever we want to prove
that a set S for certain operations 1s a vector space, we try to look
at S as a subset of Y, where Y is a vector space and the operations of
addition and scalar multiplication in S are the same as the opera-
tions in V. Once we have done this, it is enough to prove that the
operation& are closed in S.
Example 3.9 Take <'6...>[a, b] of Example 3.5. <lJ [a, b] is a subset of
:J'[a, bJ, which is a vector space by Example 3.2. So to prove that
W[a, b] is a vector space (which we have done in Example 3.5), it is
enough to prove that 6:7 [a, b] is a subspace of :J'[a, b]. Therefore, by
Theorem 3.2.2, we only need to prove that 'G[a, b] 1s closed for addition
and scalar multiphcation.
Since the sum of two continuous functions is continuous and any
scalar multiple of a continuous function is again continuous, we find that
addition and scalar multiplication are closed in Cf! [a, b]. This observa-
tion not only provt:s that W [a, b] is a vector space, but also that it is a
subspace of :1 [a, b].
3.2.4 Remark The spaces y,/[a, b], ~< 1 >[a, b], '2:'(n>[a, b], and B>[a, b]
are subspaces of :f[a, bJ.
Further, note that
(a) tl[a, b) is a subspace of W[a, b].
(b) <Ct1>[a, b] is a subspace of 1G'La, b].
(c) W<n>[a, b] is a subspace of g:::'[a, b] for every positive integer n.
(d) W<n>[a, b] is a subspace of 'ii;-'<m>[a, b] for every m < n.
(e) 9[a, b) is a subspace of W <n>[a, b) for every positive integer n.
(f) Similar results are true for functions defined on (a, b).
Example 3.10 The set S of all polynomials p E 9>, which vanishes at a
fixed point .¥0, is a subspace of 9>.
We have S - {p E 9> I p(Xo) - O} •
If p and q are two members of S, then p(xo) - 0 and q(x.) - o. So
(p + q)(Xo} - ,Cx.,) + 'l(x0) =- 0, which means that the polynomial
p + 'l also vanishes at .¥t and ao p + tJ E S. So addition is cl01ed in S.
12 J VECTOR SPACES

Similarly. if" ia a scalar and p E S, then (a.p) (x0) - a.(p(x0)) - 11 • 0 = 0.

So q ES. This completes the proof that Sis a subspace of !P.
Before concluding this article, we shall identify what are called trivial
subspaces of a vector space V. The set containing just the zero element
of V and nothing else is clearly a subspace, since it satisfies conditions (a)
and (bl of Theorem 3.2.2. Similarly, V, itself considered as a subset of Y,
satisfies the conditions of Theorem 3.2.2 and so is a subspace of itself.
These two subspaces {O} and V of J, are called trivial subspaces of V, since
they trivially satisfy conditions (a) :.nd (b). All other subspaces of V are
called nontrivial subspace, of V. The trivial subspace {O} of V is denoted
by J/0 and is also called the zero subspace of Y.
Example 3.11 Consider the equation
(1)
where at/s are real constants and xa's are rral unknowns. A solution of
this equation can be represented as an n-tuple (x1, XJ, ••• , x11), which is a
vector of V,.. Now let S be the set of all vectors (X1t x1, ... , x11) E Y,.,
which satisfy Equation (1). We can pro,e that Sis a subspace of V11 • The
proof is left as an exercise for the reader.
The special cases of Example 3.1, when n = 2 or n = 3, 11re interest-
ing. Take V1 • The set S of vectors (x, y) E J/1 , which satisfy the equa-
tion u + ~)' a= 0, is clearly a straight line through the origin in J/1 •

It i1 therefore a subspace of V1 , as pointed out in tho discussion following

Definition 3.2. t. This is now corroborated by the general case considered
in Example 3.11.
Again, in V8 the set S of all vectors (x, y, z) E J/8, which satisfy the
equation u + ~Y + yz == 0, 1s a plane through the origin and hence a
subspace of V8 (also explained in the discussion following Definitfon 3.2.1).
This fact is now corroborated by the result in the general case.
Finally, suppose in J/8 we consider the set S of all vectors (x, y, z), which
satisfy the equation ax + ~Y + yz - 1. This, of course, is a plane, but
since it does not contain the vector (0, 0, 0), the zero of the space. it is not
a subspace.
From the foregoins argument we can see that to prove that a siven set
is not a subspace (when it is not!) is really easy. For, from among the
several axioms that it has to satisfy we choose one tbat is not satisfied by
it and we are done. For example, in the previoua paragraph, it is obvious
that (O. O, 0) don not belona to the set ud this observation alone is
enoup for ua to coaclude that the set is not a subspaae.

Problem Set 3.2

I. Pl'Ofl that conditions (a) and (b) of TllearOID 3.2.2 caa be ~
by tbe ahape condition
 3.2 SUBSPACES / 73

«u + ~v E
S for an u, v E S and all scalars ar:, ~-
2. Let W - {(xi, x., ... , x,.) E Y,. I x 1 == O}. Prove that Wis a s'Ub-
space of Y,..
3. Prove that
w= {(Xi, X2, .. ' Xn) E v~ I Gt1X1 + «1XJ + ... + lly,,Xn = 0,
«,•s are given constants}
is a subspace of~-
4. Which of the following sets are subspaces of Y3 ?
(a) {(x1, x 1 , x 3 )I x 1x 1 = OJ
(b) {(Xi. X2, X3) I Xz = v2}
-'"1
(c) {(X1, Xi, Xs) I v2X1 y'3X2} =
(d) {fa1, x2, X3} I x 3 1s an integer}
(e) {(Xi, x., x,f) I xi + Xi + X1 <: n
(f) {(xi, x 2 , x3) l X 1 + X 2 + xa '.> O}
(g) {(xi, x 1 , x 3) I X1 =- y2l'1 and xt = 3x2}
{h) {(xi, X 1 , X 3) I X1 = 2x: or X3 = 3x2}

(1) {(xi, x 2 • x,) I x 1 - 2x2 = x3 - 3~J } •

5 Which nf the following !>ets are !>Ub~p,il.C!> of :.J 1

(a) {p E 9' I degree of p = 4}
{b) {p E £f I degree of p ::;; 3}
(c) {p E 9' I degree of p S} >
(d) {p E £1' I degrre of p '-:::: 4 and p'(O} - O}
(e) {p E if I p(l) = O}.
6. Which of the followin~ sets are subspaces of-{/ (a, b)?
(a) {/ E 75-? (a, b) I /(x0) = 0, X 0 E (a, b)}
(b) {/ E 2;' (a, b) I /'(x) = O for all ·" E (a, b)}

(c) {/ E <t (a, b) I /( a ~ b) - I}

(d) {/ E '~ (a, b) I f'(x) == xi f(x)}

(e) {/ E '6-'(a, b) I 2/"'(x) + 3xf"(x) - /'(x) + x'/(x) = O}
(f) {f E <c'(a. b) I f!/(x)dx - O}.
7. Prove the statements in Remark 3.2.4.
8. (a) If U and Ware subspaces of a ,ector space V such that U C· JY.
thtn prove that U is a subspace of W.
(b) If Wis a subspace of a vector space V and U is a su1?5pace of W,
then prove that U is a subspace of Y.
9. True or false ?
(a) In Y1 every infinite subset is a subspace.
 14 / VICTOR SPACES

(b) V1 is a subspace of J-'1 •

(c) V1 X V1 - V1 •
(d) Every subspace of a vector space is a vector space.
(e) V0 is the only real vector space containing a finite number ot ele-
ments.
(f) The derintives of e"' form a vector space.
(g) The set {~ sin x + ~ cos x I «. ~ are real numbers} is a vector
space.

3.3 SPAN Of A SET

In Example 3.8 \\e saw that an element u0 of a vector space V gave rise
to a subspace [u0] of V. Looking at 1t another way, we can say that here
is a subspace of V, namely, [u0], that is fully identified once we know the
single element u0• Jn this article we shall see subspaces that are fully
identified by a subset of V, \\bich is smaller than the subspace. To be more
preciM:, we need the following definitions.
3.3.1 Definition Let u1, u1, . . . u. be n vectors of a vector space V and
let «1, «1, ... , «. be n scalars. Then
« 1u1 + ot1 U1 + ... + «.u,.
is called a linear combination of 111, u3 , ... , "•· It is also called a linear
combination of the set S = {u1, 111, •••• Un}• This being a linear combi-
nation of a finite set is also called a.finite linear combination
In the case of an infinite set S a linear combination of a finite subset
A of Sis referred to as a finite linear combination of S.
3.3.2 Definition The span of a subset S of a "ector space V is the set of
all finite linear combinations of S.
In other words, if S is a subset of V, the span of S is the set
{
111U1 + Ila"• ... +
+ I
«i, «., ... , «,. any scalars, n any posi-}
11,. UII t1ve
• •
integer, s .
and Uu u1, ... , Un E
The span of S is denoted by [SJ. If S contains only a finite number of
elements, say 111, "■• ••• , U., then [SJ is also written as (u11 "■• ••• , u,J.
As illustration, take V = V1 and S == {(l, 0, 0), (0, 1. O)}. Any linear
combination of a finite number of elements of S is of the form 11( 1, o, O)
+ ~o. 1, O) =- (11, ~. 0). The set of all such linear combinations is [SJ.
Actually, [SJ -=:{(11, ~. 0) I 11, ~ any scalars}. In this case [SJ can be written
also as ((1, 0, 0), (0, 1. O)]. We can see that it 1s a subspace of Y1 • In fact,
this is true in all cases. We shall state and prove this assertion in the
form of a theorem.
3.3.! Tlleor• Let S be a noMmpty 1ub,et of a Hctor ,pace V. Then (SJ,
the qan of S, 11 a 1ublpace of Y.
 U SPAN OP A SIT / 75
Proof: By Theorem 3.2.2, we lllve only to prove that [SJ is closed
for addition and scalar multiplication. Let u and , be any two vecton in
[SJ. Then
u - at1u1 + a 1u. + ... + &,au. for some scalars 11,1 some
u,'s E S, and a positive integer n,
and v - ~1v1 + ~v1 + ... + ~-'• for some scalars~,, '«>me
vectors r,'s E S, and a positive integer m.
Hence, u + v = «1u1 + ... + «.u. + , 1, 1 + ... +- ~..,... This is again
a finite line~r combination of S and so u + v € [SJ. Similarly,
«u = (a.«1)U1 + ... + (u11)u,. is again a finite linear combination of
S and so it is in [SJ. Hence, [SJ is a subspace of Y. I
3.3.4 Remark A non\rivial subspace always contains an infinite number
of elements. So [SJ (:¢: V0) always contains an infinite number of ele-
ments. But S itself may be a smaller set, even a finite set. By con-
vention, we take [ +] -= V0 •
3.3.5 Theorem If S is a nonempty subset of a rector space V, then [SJ u
the smallest Jubspace of V containing S.
Proof: Clearly,' [SJ is a subspace by Theorem 3.3.3. It contains S,
because each element u0 of S can be written as l flo, i.e. a finite linear
combination of S. To prove that [SJ is the smallest subspace containing
S, we shall show that if thcrr· exists another subspace T containing S, then
T contains [SJ also.
So Jet a subspace T contain S. We have to prove that T contains [SJ.
Take any element of [SJ. It is of the form «1U1 + + ... +
«1u1 «11u11, where
«.'s are scalars, u1, u1 , ••• , u11 are m S, and n is a positive integer. Since
S C T. each u, also belongs to T. Since Tis a subspace, «1 u1 + a.1 u1 + ...
+ «.u,. should also belong to T. This means that each element of [SJ is
in T, I
Example 3.12 In V11 show that (3, 7) belongs to [(l, 2), (0, 1)], but doe•
not belong to [(l, 2), (2, 4)].
Clearly, (3, 7) belongs to [(I, 2), (0, 1)), if 1t is a linear combination of
(I. 2) and (0, 1), i.e. if
(3 1 7) = «(l, 2) + ~(0; 1) = («, 2«, +~)
for some suitable a. and ~- This is possible only if
« = 3, 2at + ~ = 7. (1)
Solving Equation (1), we get a. = 3, ~ = 1. Thus, (3, 7) = 3(1, 2) +
1(0, 1). Hence, (3, 7) E ((1, 2), (0, J)].
Further, it (3, 7) E ((1, 2), (2, 4)], then
(3, 7) == A(l, 2) + ~(2, 4) = (a + 2~. 2« + 4~) ,
for some suitable II and ,. This gives
fl + 2, - 3, 211 + 4' -= 7. (2)
76 / VECTOR SPACES

But these equations cannot hold"at the same time. So (3., 7) ~ ((I, 2).
(2, 4)J.
Example 3.13 In the complex vector space Vf show that (1 t, 1 - I) +
belonp to ((1 + I, 1), (l, 1 - i)].
[(1 + i, 1), (1, 1 - I)] -= {«(I + ;, ]) + ~(l, I - i) I rx, ~ complex scalars}
== {(« +
~ + a.i, « + ~ - ~i) I «,~complex scalars}.
(1 + f, J - I) belongs to [(I + t, 1), (1, 1 - I)] if
(1 + i, 1 - I) = (at + ~ + «f, IX + ~ - ~i)9
for aome scalars «, ~- Thus,
1 + i ... « , ~ + «i = at(l + t) + ~
1 - i = Gt ~ + -
~i = a + J(l - i) .
Solving for a, ~. we get cc = 1 + i, ~ = 1 - i, show mg that (1 + i,
1 - i) belongs to [(1 + #, 1), (1, 1 - I)].

Problem Set 3.3

1. Let S = {(l, 2, 1), (1, 1, -1), (4, 5, -2)}. Determine which of the
following vecton are in [SJ :
(a) (O, 0, 0) (b) (1, 1, 0) (c) (2, -1, -8)
(d) (-i, -¼, !) (e) (1, 0, 1) (f) {1, -3, S)
2. Let S ={x', x2 + 2x, x' + 2, 1 - x}. Determine which of the follow-
ing polynomials are in [S'J :
(a) 2x' + 3x' + 3l. , 7 (b) l.' t 7x + 2

(c) 3r +" + S (d) x1 _ !.x9

2
+ ~2
(e) 3x + 2 (f) x8 + x1 + 2x + 3.
3. If S is a nonempty subset of a vector space Y, prove that
(a) [SJ - S ljf S is a subspace of Y
(b) ([SJ] = [SJ.
4. Let , 1 , , 1, ••• , ' • be n elemtnts of a vector space Y. Then prove that
(a) (,1, '•• ... , ,J -= (111, 1, ._,,, ••• , 1111,J, «. :I= 0
(b) [Pi, ,.] - c,. - '•· + V1 ,J
(cl If'• E [,1 , '•• •••• ,,._1), then
[v1, '•• ... , '.1:-1, l'.t, '.1:+1, ... , ,.] = [vi, 'u, ... , •1:-1, V.1:+10 ... , , ..).
5. Let S be a nonempty subset of a vector space Y and u, , E Y. If
u E (SU {,}] but u - [SJ, then prove that, E [SU {u}].
6. What is the span of
(a) x-axis and y..uis in Y, ?
(b) x-axis and xy-plane in Y1 ?
 U MORf ABOUT SUBSPACIS / 77

(c) xy-plane and yz-plane in Ya?

(d} x-axis and the plane x + y = O in Y• ?
7. True or false?
(a) Span'of x + y = 0 and x - y = O in Y8 is Y,.
(b) Y7 bas a subspace consisting of 7 elements.
(c) The span of the set {x, x1 , r, x•, ... }1s 9'.
(d) In V8 , [y-axis U z-axis] = V3 •
(e) In V8, [k] == [i, k, i t- k] n [J, k,j + k].
(f) In Yf, (-1, 2, 0) E [(I, -i, 1), (0, -i, -1)).

3.4 MORE ABOUT SUBSPACES

Let U and W be two subspaces of a vector space V. Their intersection
U n W cannot be empty, because each contams the zero vector of V.
Now, if u and v are two vectors of U n W, then u + v E U and
u +vE W, because U and W are subspaces of V and u, v both belons to
U as well as W. Hen1e, u + v E Un W. Similarly, if« is any scalar,
then au is in both U and W. Hence, «u E U n W. This shows that If
U and Ware subspaces of V, then U n Wis also a subspace of V.
This result can be generalised to any number of subspaces. More
precisely, if Ult U1, ... , Un are n subspaces of V, then their intersection
U1 n U1 n ... n Un is also a subspace of Y (see Problem 1). We shall
use this fact 10 an interesting way in the following example.
Example 3.14 Let W be the set of all vectors (x1, X1, ... , Xn) of Vn satisfy-
ing the three equations
«1X1 + «1X1 + .. . + «nXn == 0 (1)
~1X1 + ~X1 + ... + ,nXn = 0 (2)
Y1X1 + Y1X1 + ... + "fnXn = 0, (J)
Then W = W1 n W1 n W8 , where W1 11 the solution set of Equation
(]), W1 is the solution set of Equation (2), and Wa is the solution ,et of
Equation (3). Each W1 is a subspace by Example 3.11. So, by the foregoing
argument, W is a subspace of Yn•
Thus, we have proved that the intersection of two subspaces is always a
subspace. On the other band, the union of two subspaces need not be a
subspace. For example, take U = x-axis and W = y-axis in V1• U and
Ware subspaces of V1 • Here (1, 0) E U and (0. 1) E W. So both (I. 0)
and (0, 1) belons to U U W. But (1, 0) + (0, I) - (1, 1} - U U W.
This shows that U U W is not a subspace of V1•
Thus, U U Wis not in general a subspace. However, we know. by
Theorem 3.3.S, that (U U W] is the smallest subspace of V containing
 18 / VECTOR SPAC~

U U W. Let 111 now analyse the subspace CU U WJ. Any elemnt of

dJis subspace is a linear combination of a finite subset of U U W. In other
word,, a vector , of [U U WJ is of the form
P = + ... +
«11"1 +
«,.U,. ,1W1 + ... + ,,.,.,_:, (4)
wbere 11/1 E U, w, 's E 11: n and m are nonnegative intesers, and ar:'s, P's
are scalars. But 11i111 + ... + «,.u,. is a vector in U and ,S1wJ + ... + ,S..w111
is a vector in W. Therefore, ,. ca11 be expressed as u + w, where II E U
and w e W. So we can say that [ U U WJ conrists of elements of the form
u + w, r, E U, and w E W.
3,4.l De8altlon (Addition of sets) Let A and B be two subsets of a vector
space Y. Then the sum of A and B, written as A + B, is the set of
aU vectors of the form u + v, u E A, and " E B, that is,
A + B = {u + " I u E A., 11 E B} •
Example 3.15 In Y2 let .4. =- {(l, 2), (0, l)) and B = {(l, I), (-1, 2),
(2, 5)}. Then
A+ B = {(I, 2) + (1, 1), (I, 2) + (-1, 2), (1, 2) + (2, 5), (0, 1)
+ (1, I), (0, 1) _,_ ( -1, 2), (0, I) + (2. S)}
= {(2, 3), (0, 4), (3, 7), (l, 2), (-1; 3), (2, 6)}.
Example 3.16 In Y1 let A - {(2, 3)}, B = {t(3, 1) I ta r,catari. Then
A. + B = {(2, 3) + t(3, 1) I ta scalar}
= {(2 + 3t, 3 + t) I , a scalar} .
Geometrically, Bis a line through the origin and A. is a set containing
one vector (Figure 3.1). A + Bis n line parallel to Band passing through
the point (2, 3).

X
FIGUllB 3.1
Example 3.11 In Ya Jet .4 -= {c(l, 2, 0) l « a scalar}, B - {P(O, 1, 2) I fl
a scalar}. Then
.4 + B -= {«(1, 2, 0) +11(0, 1, 2) I cx, Pscalars}
- {(«, 2«, +~, 2fl) I «. fl scalars} •
Geometrically• .« avd B ate lines throu&b the oriain in Ya and.« B +
is a plane containing these lillOI and puaina through the origin (Fipre,3.2).
 3,4 MORI! ABOUT SUBSPACU / 79
In this case ..4 ..J.. B = {(I, 2, 0), (0, 1, 2)).

FIGURI! 3.2

3 4.2 Theorem If U and Ware two subspaces of a vector space V, then

U t- Wis a subspace of V and U + W -= [ U U W].
Proof: Obviously, U + WC (U U W], because each vector of U + W
is a finite hoear combination of U U W. We shall now prove that [U U W]
C U + W. Let v E [ U U W]. Then the argument preceding Defini-
tion 3.4.1 shows that v 1s of the form u + w, for some u E U and w E W.
Hence, v E U + W. Thus, U W + =[
U U WJ. This automatically
makes U + W a subspace of V. I
3.4.3 Remark From Theorem 3.4.2 it follows that U + Wis the !lmallest
subspace of V contammg U U W, 1.e. both U and W.
Example 3.18 If V - Y,, U = x-axis, and W = y-axis, then U Wis +
the set of all those vectors of V3 that arc of the form «(I, 0, 0) + ~(0, 1, 0).
Therefore, U + W = {(at,~. O) I «, ~ are scalars}. On the other hand,
[U U W] consists of vectors of the form cxu + ~w, where cit, ~ are scalars,
u E U, w E W. This gives '
[ U U W] == {a.(1, 0, 0) + fl(O, 1, 0) I at, fl are scalars} ,
which is the same as U +
W. This illustrates Theorem 3.4.2. Note the
interesting relation arising from this example :
[x-axis U y-axis] =
x-axis + y-axis xy-plane.= (5)
80 / VBCIOR. SPAC!S

DIRECT SUM
We have just seen that if U and W are subspaces of a vector space Y,
then the sum U + W is also a subspace of Y. If, in addition, U fl W
= Y0 = {O}, the sum U + W is called a direct sum. The direct sum of
U and W is written as U EB W.
Example !J.19 We can check the following additions in Y8 :
xy-plane + yz-plane = Ya (6)
xy-plane+ z-axis = V3 • (7)
The sum in Equation (7) is a direct sum, because (xy-plane) n z-axis
= {O}. So Equation (7) can be rewritten as
(xy-plane) EB (z-axis) = Ya • (8)
The sum in Equation (6) is not a direct sum, because
(xy-plane) n (yz-plaoe) =- (y-axis) =#: {0} •
Any vector (a, b, c) E V8 can be written as
{a, b, c) = (a, b, 0) + (0, 0, c) , (9)
where (a, b, c) E xy-plane, (0, 0, c) E z-axis. Thus, (a. b, c) is the
sum of two vectors, one in the xy-plane and the other in the z-axis. The
advantage of the direct sum lies in the fact that the representation (9) is
unique. That is, we cannot find two other vectors, one in the xy-plane and
the other in z-axis, whose sum is (a, b, c). The reader can check this for
himself.
On the other hand, in Equation (6) any vector {a, b, c) can be
written as the sum of two vectors, one in the xy-plane and the other in
the yz-plaoe, in more than one way, for example,
(a, b, c) = (a, b, 0) + (0, 0, c) (10)
and
(a, b, c) - (a, 0, 0) + (0, b, c) • (11)
The fact asserted in this example is generalised in the following
theorem.
3.4.4 neorem Let U and W be two subspace, of a vector space Y and
Z - U + W. Then Z = U Ea W iJf the following condition Is
satilfied:
.4ny vector z E Z can be expressed uniquely a the sum
z == u + w, u E U, w E W. (12)
Proof: Let Z - U EB W. Since Z = U + W, any vector z E Z
can be written aa
z=u+w, (13)
for some II e U and w E W.
Suppose it is possible to have another representation
z-u•+w•, (14)
 S.4 MOU ABOUT SUBSPACES / 81

for some ,I E U and w' e W. Then

u+w=u'+w' (14)
or
u-u'=w'-w. (IS)
But u - u' E U and w' - w E W. From Equation (IS). we find that
u - ,I = w' - w E U n W = {O}. since Z is a direct sum of U and W.
Hence. u - u' = 0 = w' - w. which implies that u - u' and w - w'. So
no such second representation for Equation (13) is possible. In other
words, condition (12) is satisfied.
Conversely, let condition (12) hold. We shall now prove that Z is
a direct sum of U and W. Since Z = U + W, we have only to prove that
un W= {O}.
Let, if possible, Un W contain a nonzero vector v. Then " E U, ,, E
W, and , ="
+ 0 E U + W with , E U and O E W. Also " = O +, E
+
U W with O€ Uand v € W. These two ways of expressing a vector , in
U + W contradict the hypothesis (12). Hence, V n W = {O} and Z =
u EB W. I \
3.4.5 Deftoitioo If U is a subspace of a vector space V and , a vector of
+
Ji", then {v} U, also written as v + U, is called a translate of U
(by Y) or a parallel of U (through v) or a linear variety.
U is called the base space of the linear variety and v a leader. Note
that{•}+ U is not a subspace unless v E U. (Why 'l)
As illustration. take the line y = JC through the origin in Ji"1 • Call it U.
Conside1 the point, = (1, OJ. The translate,+ U of U by vis the line
y = JC - 1 through the point (1, 0) (Figure 3.3). It can also be obtained
by add mg (J, 0) to the vectors in y = JC.

y y

/
JC 0 v•ff.0) X 0
Ca) Cb>

FJouu3.3
Aa a second illustration, take the plane y = 0 in Y1 and call it U. Con-
sider the point, - (1, 1, 1) e: Y1 • (1, 1, 1) + U is the set or all pointl of
82 / VEC10R SPAC!S

+
V1 given by (I, 1, )) u, where II E U. Geometrically, it is the plane
.1,1arallel to y = 0 through the point (1, I, 1) (Figure 3.4).
z

(a> Cb) Cc>

FIGURE 3.4

The following theorem gives the fundamental properties of a linear

variety.
3.4.6 Theorem Let U be a &Ub$pace of a vector space Y and P = , + U
be 1l1e parallel of U through v. Then
(a) for any win P, w + U = P. In other worda, any. vectcJr of P can
be taken as a leader of P;
(b)_ twJ vectors Vi, ,, € V are in the same parallel of U ,ff v1 - v1 E U.
+
Proof: (a) Let w be a vector of P. Since P = v U, w can be express·
+
ed as v + 111 for some u1 E U. Sow= v u1 or, equivalently, v = w - u1•
Let z E P. Then z has the expression z = 11 + Us for a suitable u2 U. e:
;)ince v = w - uh we have z = (w - u1) +
u1 - w + (u1 - u1). Here
111 - 111 E U. Thus, every vector z of P has the form w +(some vector
in U). So P C w + U. Conversely, let y E w +
U. Then y is of the
form y = w + u for a suitable u E U. The vector y = w 11 = " + +
(111 + u) is of the form v + (a vector of CJ). Soy v e: +
U = P. This
shc.,ws that w +UC P. Hence, w + U = P.
(b) Let 111, v1 be in the same parallel of U, namely, 11 + U. So
• 1 = v + u1 for some u1 E U, and "• = 11 + Ila for some "• e:
U. Then
, 1 - v1 = (v + u1) - (v + u.) = r,1 - u1 , which belongs to Cl. Con•
versely, if 111 - ,~ E U, then 111 - 111 =
u for some u E U. So
=
v1 v1 + u. Therefore, v1 E 111 + U. Also , 1 = "• + o.
Therefore,
+
"• E "'• U, since OE U. So , 1 and , 1 belong to the same parallel
"• + u. I
As illustration, take Y = =
Ya and U yz-plane. Let•= (1, 1, 1).
Then P =,+ U is the parallel given by the set
{(1, 1, l) + (0, 11, y) I ~. y arbitrary scalan}
+ +
,.. (0, 1 P, 1 y) I p, y arbitrary scalan).
 3.4 MORE ABO\JT SUB.~PACF.S / 13

Part (a) of Theorem 3.4.6 says that to describe this set we could take
instead of (l, I, I), any other vector from P. Let us take vector (1, O, OJ,
which is also in P. The theorem says that everv vector
(1, 1, I)+ (0, ~. y)
can also be written in the form
(J, O, 0) + (0, ~•. y')
for suitable~• arid y'. Clearly, ~• = I t ~ and y' ~ 1 -t y.
To continue the illustration, both (1, I, 1) and (1, 0, 0) are an P. The
difference (0, 1, 1) is obviously m U. Part (b) of the theorem says that
whenever the difference of two vectors belong11 to U, then they both belong
to the same parallel and conversely.
3.4. 7 Remark Starting from a subi.pace U, we can obtain many parallels
of U. This i11 the same as starring from one straight line throuah
the origin and drawing many parallel straight lines through different
points in the plane. Given a vector v E JI, we get a parallel v + U.
If we take another vector v' E V, the parallel v' + U obtain.:d from
+
,• will be d1fferenl from v U iff v' 1s not in V U. +
Problem Set 3.4
1. Let U 1, U1, '"i Un be n subspaces of a vector space Y. Then prove
n
that U1 n U1 n ... n U,. = n U, is al~o a i.ubspdce of V.
i=I
2. If U and Ware subspaces of a vector space V, prove that
(a) U n
W is a subspace of W
(b) U U W is a subspace of V iff U C W or W C U.
3. If S is a nonempty subset of V, prove that" [SJ is the intersection of all
subspaces of Y containing S.
4. Prove that the set of vectors (x1, x1 , ... , x,.) E Y,. satisfying the follow-
ing m conditions is a subspace of 'Y• :
«11X1 + «11X1 + •.. + «1n-Xn = 0
«11X1 + C11.ztX1
... + ...
+ «1nXn = 0
... ...
«1a1X1 + «m1'\"1 + ... + «111nXn = 0'
«,,'s are fixed scalars.
5. Find the intenection of the given sets U and W and determine
whether it is a subspace.
(a) U ,_ {(xi, x.) E v. >
t x, O}, W ==- {(X1, X1) E t x, <; O} y•
(b) U = {/E W(-2, 2} IJt-1) - O},
W ... {/E ~(-2, 2) 1./U) - O}
 84 / VECTOR SPACES

(c) U == {/E. ~(-2, 2) I Jim Az) == OJ,

z+l
W = (/ E g:7(-2, 2) I lim l(x) == 1}
,x+2
(d) U == fl, W == (IE <?(-00, ~) 1/tx) ==Jt-x)}.
6. Describe A + B for the given subsets A and B of Y1 and determine in
each case whether it is a subspace or a parallel or just a subset of Y1 •
(a) A = {(J, 2), (0, 1)}, B = {(1, 0), (3, -1)}
(b) A = {(I, -2), (5, I)}, B == {(3, 5), (l, 3), ( v'2, 1t)}
(c) A == {(l, I)}, B = segment joining (1, 1) and l2, 3)
(d) A = {(2, 3)}, B = {t(3, 4) I 1 <: t < 2}
(e) A= {(3, 7)}, B = {(t{-1, 2) IO< t < I}
=
(f) A = {(l, 2)}, B {t(3, 0) I t >
O}
(g) A = {(2, I)}, B = {(x, y) I 2x + 3y = l}
(h) A= {(I, 5)}, B = {(x,y) Ir+,~= I}
(i) A = {t(3, 4) I 0 <: t < 1}, B = {t(2, 5) I 1 < t < 2}
(j) A = {t(l, 0) I 0 < t < l}, B = {t(0, 1) I 2 <: t < 4}
(k) A = {r(l, 0) I ta scalar}, B = [(l, 2))
(I) A = line x = 3t, y = 4t, B = line 2x -t- Sy = 0.
7. Describe A+ B for the given subsets .4 and B of Ya- Determine in
each case whether A + B is a subspace or a parallel or just a subset of
Ya.
(a) A = {(I, 2, 1)}, B = {t(l, 2, 0) I t a scalar}
(b) A = {(3, 1, -1)}, B == {(x, _v, z) I x + y + z = 0}
(c) A = {(I, -3, 4)}, B = [(I, 2, 3), {0, 0, I)]
(d) A = [(I, 2, 3)], B = ((3, 1, 0)]
(e) A = ((1/2, 2/3, 1)), B = plane 2x + 3y + z = 0
(f) A = [(5, 0, 2)], B = [{1, 2, 3), (0, 1, 2)1
(g) A = [(1, 0, -1 )], B = [l2, .5, 8), (2, 3, 4)).
8. If U and W are two subspaces of a vector space Y, prove that
U + W = U ijfW CU.
9. Let A and B be two nonempty finite subsets of a vector space Y. Then
prove that
(a) [A n BJ C [A] n [BJ
(b) [A U BJ = [A) +
[BJ.
10. 1f Ui, U1 , and U1 are three subspaces of a vector apace Y, prove that
(U1 n
U,) + cu. n U.J C (U1 + U,) n u,.
11. (a) Prove that two parallel straight lines in Y1 are parallels of tho
same nontrivial subspace of Y1•
{b) Prove that the sum of two distinct in~necting linea in V1 is V..
II it a direct sum 'l
 j,~ LINEAR. DEhNDl!NCE, INDBPBND!NCII / 85
J2, (a) Show that the set of all solutions of the equation 3x1 + 2%1 -x1
+ x, = 5 can be expressed as (J, -2, 2, 8) + W,[wher(W is
the space of all solutions of the equation 3x1 + 2x1 - x1 +
x, = 0•
. (b) Let W be the set of all solutions of the equation ix1x1 + 11ax1 +
... + ixnx11 = ~- Then show that Wis a parallel in J/11, Find
its base space and a leader.
13. Show ;that the gh·en set Pis a parallel in ('( o. 2). In each case
find the base space and a leader.
(a) P = {/ I /'(x) = 3x'}
(b) P = {/ I f'(x) = xe•2}
(c) P = {f I/"Ix) = :;in x}
(d) P = {/1/'(x) = x' + 2x + 3}
<e> P == {/I/Ct> = 21.
14. Let U = {/ E <c,"l-a, a) I /is odd}, and
W = {/ E rt \-a, a) I/ is even}.
Then describe (a) U n W and (b) U + W, and determine whether they
are subspaces of 'C(-a, a). Is '6,,,(-a, a) CU+ W? Is U + W
a direct sum ? Is W r-a, a) = U EB W?
15. Prove that a linear ~ariety is a subspace iff its leader belongs to the
base space.
16. True or false ?
(a) For any subset A o( a vector space Y,
[A] =-= [A + A].
DO
(b) If u,, i = 1, 2, ... are subspaces of Y, then n U, is a subspace
i=l
ofY• .
(c) The intersection of two linear varieties is a linear variety, pro-
vided it is nonempty.-
Cd) The sum of two linear varieties may not be a linear variety.
(e) The union of two linear varieties need not be a linear variety.
(f) The set of all solutions of the equation :~ = g, where g
is a constant, is a linear variety in 'if (0, 00 ).
(&) In Eiample 3.14 Y. cannot be replaced by Y~•

3.S LINEAR DEPENDENCE, INDEPENDENCE

We atart this article with tho definition of 'triviar and •nontrivial'
linear combinations.
 86 / VBCTOR SPACES

3.5 J DelJaltJon If Utt "2, ... , u,. arc n vectors of a vector space Y, then
the Hnear combination
«1111 + «alls + ... + «1111,. (I)
is called a tri,ia/ /i11ear c~ mbina1ion if all the scalars «i, «,.... , «11
1

are zero. Otherwise the linear combinationI

is said to be nontrMal.
In other words, a nontrivial linear combination is of the form (I),
where at least one of the «'s is not zero. As illvstration,
Ou1 + Ou1 + ... + Ou,. is a trivial linear c.:-mbination;
+
0111 Cuj t- ... + Ou.-1 +Ju. is a nontrivial linear combination;
and
1,.,1 + 2uA + 3ud t- ... + nun is al11>0 a nontrivial linen combination.
Note that the trivial linear combination of any set of vectors is always
the zero vector, for
Ou 1 + OuJ + ... + Ou,. = 0 + 0 + ... + 0 = O •
The queiition arises whether a nontrivi41 linear combination of a set of
vectors can give the zero vector. The answer is in the affirmative. Consi•
der the following examples.
Example 3.20 Let (1, 0, 0), (2, 0, 0), and (0, 0, 1) be lhrec vectors in Y1 •
Ther, we have
Ill, 0, 0) + (-l) (2, 0, 0) + 0(0, 0, 1) = (0, 0, 0) = 0.
Thus, a nontrivial linear combination may give the zero vector.
Example 3.21 Let (l, O, 0), {O, 1, 0), and (0, 0, 1) be three vectors of V1 •
Take any linear combination of these vectors, say
cx(I, 0, 0) + ~O, 1, 0) + y(O, 0, 1) ,
which is the same as («, ~. yJ. If this were to be the zero vector, then
(ex, ~. y) = (0, 0, 0). Therefore, a. =-= 0 =-= ~ = y.
In this cai,e the only linear combination that gives the zero vector is
the trivial linear combination. The same is true in the following example.
0

Example 3.22 Let (I, 1, 1), (1, I, 0), and (1, 0, 1) be three vectors. As
befort, if
a.(l, 1, 1) + ~(l, 1, 0) + y(l, 0, 1) = (0, 0, 0),
then {« + ~ + 'Y, « + ~. « + y) = (0, 0, 0) .
Therefore, ex + ~ + y = 0, « + ~ = 0, and « + ·r = 0. These give 11 =
0 ='= 'Y·
It is clear that these three examples fall into two categories. Examples
3.21 and 3.22 are of one kind, whereas Example 3.20 is of another. In
Example 3.20, (1, 0, 0) is a linear combination of (2, 0, 0) and (0, 0, 1):
(1, o, 0) == 1(2. o, 0) + 0(0, 0, 1) •
In other words, (1 1 0, 0) dopends on (2, 0, 0) and (0, 0, 1). We say that
these three vecton are linearly dependent. On the other hand, in Examples
 3.5 UNl!AR DBPEND£NCI!, IHD.EPBNDINCJI / 81

3.21 and 3.22 there is no such dependence of one vector on the others. We
say that they are Uoearly independent.
We formalise these concepts of linear dependence and linear indepen-
dence in the following definitions.
3.5.2 De8Ditlon A set {ui, Us, ... , u.} of vectors is said to be linearly
dependent (LD) if there exists a nontrivial linear L"ombin1ition of
Ui, Us, ... , u. that equals the zero vector.
Example 3.23 Prove that the vtctors (1, 0, l), (1, l, 0), and (-1, 0, --1)
are J..D.
To prove linear dependence we must try to find scalars et, (i, y such that
«(l, 0, 1) + ~(1, 1, 0) +
y(--1, 0, -1) = 0 -= (0, 0, 0) .
Therefore, IX + ~ - y = 0, ~ = 0, and IX - y = 0. This can happen iJf
« = y. Any nonzero value for IX, say 1, will do. Thus
+
1(1, 0, 1) 0(1, 1, 0) + 1(-1, 0, -1) = 0.
So there exists a nontrivial linear combination of the given vectors, which
equals the zero vector. Hence, the vectors are LD.
Once we find a nontrivial linear combination equal to the zero vector,
linear dependence is proved. Very often it is possible to guess such a
linear combination.
3.5,3 Definition A set {u1 , Ui, ••• , un} of vectors i1 said to be linearly
independent (LI) if no nontrivial linear combination of ui, "~· ... , u,.
equals the zero vector.
'No nontrivial linear combination equals the zero vector' means the
following:
If at all there is a linear combination that equals the zero vector, then
it must be the trivial linear combination. Now recall that the trivial
linear combination i& always zero. So the statement withm quotation
marks means 'The only linear combination that equals the zero vrctor is
the trivial linear combination'. In view of thi1, Definirion 3.S.3 can
be reworded as follows.
3.5.4 Detialtioa (Reworded) A set {ui, Us, ... , 1111} of vectors is said to
be linearly independent (LI) if tho only linear combination of
u1, U., ... , "• that equals the zero vector is the trivial linear
combination.
By convention, the empty set is considered to be LI. Note that linear
dependence and linear independence are opposite concepts. ·
Example 3.24 Prove that the vecton (J, 0, 1), (1, I, 0), and (l, 1, -1)
are LI,
If possible, let
«(l, 0, l) + 11(1, 1, 0) + y(I, 1, -l) .. 0 == (0, 0, 0) . (2)
This gives • + + I) +
y - 0, It y =- 0, and 11 - y - o. Thercfore, IX = 0
 88 / VJfC'J'OJl SPACB.9

=·fJ == y. This means that the only values of «. ,. r that satisfy

Equation (2) are«= O = ~ = y, i.e. only the trivial linear combination
equals the zero vector. Hence, the given vectors are u.

HOW TO CHECK LINEAR DEPENDENCE OR INDEPENDENCE

The foregoing discussion tells us that when we are given a set
{uu u1, ••• , u11 } of vectors and we want to check whether it is LD or LI,
the following procedure will be the natural one.
Fint, we assume that some linear combination orui, u1, ••• , u,. is equal
to the zero vector, that is,
«1U1 + 1X■U1 + ,., + Gt,aU11 = 0 , (3)
Then two cases arise. One is when it is possible to.find scalars « 1, «a, ... «11
with at least one of them not zero such that Equation (3) is satisfied. In
· this case we conclude that the vectors u11 u.• ... , u11 are LD.
The other case is when we can prove that our assumption automatically
implies that «1 = 0 = «a = ... = «11 • In other words, the only values of «.'s
that satisfy Equation (3) are«,= 0 for each i = I, 2, ... , 11. In this case
we conclude that Uu fla, •.. , Un arc LI.
In the first case, even guesswork in finding the suitable scalars is
enough, whereas in the second case a proof is necessary.
Example 3.25 Check whether the following set of vectors is LD or LI :
{(I, 0, 1), (I, I, 0), (1, -1, 1), (I, 2, -3)}.
Suppose
«(), 0, I)+ ~(1, I, 0) + y(l, -1, 1) +
8(1, 2, -3) = 0-= (0, 0, 0)
or (« + ~ + y + 8, ~ - y + 28, IX + y - 38) = (0, 0, 0)
or « + ~ + y + 8 = 0, ~ - y + 28 = 0, IX + y - 38 = 0 .
This system has a solution « = 58, ~ = -48, y = -28 for each choice
of 8. Take 8 = 1. This gives
5(1, 0, 1) - 4(1, I, 0) - 2(1, -1, 1) + 1(1, 2, -3) = (0, 0, 0).
Hence, the.given set is LD.
Example 3.26 Check the linear dependence or linear independence of the
set{,., r•} in ~••>(-OC>,00),
Suppose that
ru" + ~elat = 0 (x) =
0 for all x E (-OC>. OC>).
Thia aives, on differentiation,
«e~ + 2~eta- =
0 for all x E (-OC>, co).
Solvina these two equations for at and ~. we get
~el• = 0 for all x E (-co. oo) ;
but ,t• ii never zero and hence ~ =
0, It then automatically follows that
• ii also zero. Thus, the only linear combination that equals the r.ero
 3.5 LINEAR. DEPENDENCE, INDEl'ENDENC8 / 89
vector is the trivial one. Hence, the given set is LI in ~-'c• '(-00, CO).
We also say that the set is u over (-co. co).
Example 3.27 Check the linear dependence or linear independence of the
Ix I} in
set {x, ~·c-1.
I).
Suppose «x + ~ I x I = 0.
Since the function I x I is not differentiable at uro, we cannot use
the method in Example 3.26.
atX +
~ I x I = 0 holds for all x in ( -1, 1). So choosing two di.fl'e•
rent values of x. say x = 1/2 and x = -1/2, we get

2
«
+ 213 = 0 and -
« 13
2 + -2 = 0.
Solving these two equations, we get at = 0 = 13. Hence, the set is u
over (--1, I).
We shall now take up the geometrical me,10ing of linear dependence.
J'or this we need the following definitions in a vector space V.
3.5.5 Definition Given' a vector v -/= 0, the set of all scalar multiples of
v i~ called tire line thl'ough v.
Geometrically, in the cases of V1, V1 , and V3 it is nothing but the
Hraight line through the origin and ,.
3.5.6 Definition 1 wo \'ectors v1 and v1 are collinear if one of them lies in
the 'line' through the other.
Clearly, 0 is collinear with any nonzero vector v.
3.5.7 Definition Gh-en two vectors v1 and V1, which are not collinear,
their span, namely, [v1, , 1] 1 is called the plane through r1 and v1•
Geometrically, in the cases of V1 and J/8 it is nothing but the plane
passing through the origin, , 1 and v1 •
3.5.B Definition Three vectors vi, v1 , and v8 are coplanar if one of them
lies in the 'plane' through the other two.
Clearly, 0 is coplanar with every pair of noncollinear wctors.
Example 3.28 The vectors , and 2, of ·a vector space V are collinear,
because 2v lies in the 'line' through v, i.e. 2v is a scalar multiple of v. In
particular, sin x and 2 sin x are collinear in :1(1).
Example 3 29 The functions sin x and cos JC in :1(1) are nC't collinear,
because neither of the two lies in the 'line' through the other, i.e. one is
not a scalar multiple of the other. Their span, namely,
[sin x, cos x] =- {« sin x+ ~ cos JC I «, ~ any scalars}
is the plane through the vectors sin JC and cos JC,
Examp/1 3.30 The functions sin x, cos x, tan x in :1(/) are obviously not
coplanar, because none of them tica in the 'plane' through the other two.
90 / VBCl'OR SPACES

Example 3.31 The functions sin9x, cos:x, cos hare coplanar. cos 2x lies
in the plane through sin1x and cos•x, since cos 2x is a linear combination
of cos1 x and sin1x.
3.S.!> Theorem Let V be any ,ector 1pace. Then
(a) The 1el {11} i8 LO i.ff I' = o.
(b) The set {111, , 1} 1, LD i.ff 111 and'• are collinear, I.e. one of them;, a
scalar multiple of the other.
(c) The set {,1 , '•• v8} is co iff Iii, , 1 , and'• are coplanar, i.e. one of
them is a linear combination of the oth,r two.
Proof: (a) {v} is LD iff there exists a nonzero scalar cx such that
cz,.• = O. This is possible (Problem 5, Problem Set 3.1) ,j/ v = O.
(b) Suppose {v11 v1} is LD. Then there exist scalars«,~ with at least
one of them, say « -1- O, such that «•·1 + ~"• = O. Therefore, v1
= (-~/cx, 11,, which means v1 is a scalar multiple of , 1, i.e. , 1 lies in the Jine
through"•· Hence, 'i, "•
are collinear.
Conversely, if v1 and v1 are collinear, then, by definition, one of them,
say"" hes in the line through v1 • Therefore, v1 is a scalar multiple of"•·
So ,[ = °'"••
i.e. I • v1 - «v l -=- O. Hence, 1·1 and "• are LD.
(c) Suppoi.e {v1 , Va, v3} is LD, Then there exist scalars ex, ~. and y
with at least one of them, say ex al= O, such that
GtV1 + ~"• + Y"■ = 0
or "1 = (-(i,'«) "•+ (-y/«) Va •
This means v1 E [v1 , v3]. Hence, 111 lies in the plane tb1ough "• and v3 • So
v1 , "•• and v3 are coplanar.
Conversely, if , 1, "•• and "• are coplanar, then one of them, say
, 1 E [v2, v:J, i.e. 111 = «1• 1 + ex1v1 for suitable scalars ex1 and ex 8 • Therefore,
1v1 - ex8111 - ex3v8 = O. Hence, , 11 111, and 118 are LD.
As illustration, consider the three vectors (1, I, ]), (1, -1, I), and
(3, -1, 3). They are LD, for
1(1, 1, 1) + 2(1, -1, 1) - 1(3, -1, 3) = 0.
Hence, by Theorem 3.5.9, the plane through (1, 1, 1) and (3, -1, 3) con-
tains the point (1, -1, 1). Let us verify this.
The plane through (1, 1, 1) and (3, - 1, 3) is
((1, J, 1), (3, -1, 3)) == {(« + 3~, « - ~. • + 3~) I•,~ any scalars}.
This set contains (I, -1, 1), for we have only to choose cx and I! such that
ex + 3~ = 1, Gt - ~ = -1, 11 + 315 == 1.
This gives et = -1 /2 and ~ == 1/2.
Before we proceed, we shall record some simple facts about linear
dependence.
3.9.10 fact In a vector space Y any set of vecton containiq the zero
vector is LD.
 3.5 LINEAR DEPENDENCE, INDa,BNDINCB / 9J
For, if {•u "•· ... , .,.,) is a set and,, = 01
Lhen
Ov1 + Ov1 + ... + Ov,_1 + 1,,
+ Ovf+1 + ... + Ov.,
is a nontrivial linear combination resulting in the zero vector.
3.5.11 Fact In a vector space JI', if " is a linear combination of
,., 111, ••• , 11., i.e. v E [,1, 111, ••• , vJ, then {v, v10 ... , ,·.} is LD.
For, " E [vi, "l• ... , v11] means
"= «1V1 + Cls"a + •·· + «.,11. ,
i.e. 1v - « 1v1 - ... - «,.11 11 = o.
3.5.12 Fact In a vector space JI', if the set {vi, 111 , ... , 11,.} is LI and
v ~ [v11 v1 , ••• , v,.], then the set {v, 111, 111 , ... , v,.} is LI.
The proof is lefL to the reader.
In the case where the set consists of just two or three vectors, we can
look at these fctcts from the 'geometric' point of view of Theorem 3.5.9.
For the purpose of understanding, imagine that JI' = V8 (though this is not
necessary for the argument). Since O lies in the lme through v, {O, 11} is
always LD by Theorem 3.5.9 (b). Similarly, 0 always hes in the plane
through any two vectors 1111 v1 • Hence, {O, vi, v11} is always LD by Theore10
3.5.9 (c).
The idea contained in Theorem 3.5.9 does not stop with just three
vector&. We <.an in fctct work out a general result on the same lines for
a linearly dependent set of n vectors. Before we do that, let 01 look at the
theorem once again. Actually, in the proof, we have indicated the
following:
(1) {v1 , v1} is LD iff one of the vectors is a scalar multiple of the
other and (ii) {v1, v11 , v1 } is LD i.ff one of them is a linear combination of
the others.
In general, we shall now prove in Theorem 3.5.14 that if {v11 v1 , ... , 11,.}
is LD, then one of the vectors is a linear combination of the others, the
converse of which is Fact 3.S.11. Before that, we state the followina&
simple theorem and leave its proof to the reader.
3.5.13 Tbeorea (a) // a set 11 LI, then any subset of it is also LI, and
(b) If a set is LD, then any 1uperset of it i• also LD.
3.5.14 Theorem In a ,ector space V 1uppo1e {, 1, '•• ... , 11,.} 11 an ordered set
of vectors with , 1 :/=, 0. The set is LD I.ff one of the 'Vettori
1111 , 8, ... , ,,., ,ay 111:, belongs to the span 0/111, ... , '►1o i.e. 111c E (1111
, 1, .. , Vt_ 1] for some k = 2, 3, ... , n.

Proof: Spppose v., E (,1, v., ... , "•-1). Then 111: is a linear combi-
nation of 111• 111, ... , 111-•• Thus, the set {v1, ,., ... , v,r-1o v.} is LD, Hence,
by Theorem 3.S. 13 (b). {v1, v1 , ... , v,.} is LD
 92 / VIICT<m SPACES
Conversely, suppose that {,1, ,,, ... , , 11} is LD. Consider the sets
Si= {vi}
S1 = (v.a, , 1 }

s" = {,1, '•· ...• '"} •

By Theorem 3.5.9 (a), S1 is LI and by assunaption S,. is LD. So we go
down the list and cholise the first lineiirly dependent set. Let S,, be that set.
Then S,, is LD and s.,_1 is LI. Here 2 < k < n.
Since S,, is LD, we have
111 "1 + 111111 + ... + «1:'111: = 0 •

with at least one of the CA.·s not zero. Surely «1. ::fo 0, for if «,., = 0, then
S,r,-1 would become a linearly dependent set contradicting our statement
that S,., is the first linearly dependent set. Therefore, we can write

V1: ~) V1 + (-..!!,_)v1
= (- 11,t «1:
+ ,,. + (- Clt-1) l'.t-1 •
«.
which means v11 E [111, v2, .... v,1: _J.
3.5,15 Corollary .4.,finite Jubset S = {vi, v1 , ... , vn} of a vector space JI
containing a nonzero vector has a linearly Independent Jubset A such
that [A] = [SJ.
Proof: We assume v1 ::fo O. If S is u, then there is nothing to
prove, as we have A = S. If not, then, by Theorem 3.5.14, we have a
vector 11: such that v,, E [111, •••• V,1:_ 1]. Discard ,,,. The remaining
set S1 = {v1, ... , Vt-1, V1:+1, .... vn} has the same span as that of S (see
Problem 4 (c), Problem Set 3.3). If S 1 i11 u. then we are done. If not, then
repeat the foregoing process. This must finally lead to a linearly indepen•
dent subset A such that [A]= [SJ. (Why?)
Example 3.32 Show that the ordered set {(l, 1, 0), (0, 1, 1), (1, 0, -1),
(1, 1, 1)} is LD and locate one of the vectors that belongs to the span of
the previous ones.
Consider the sets
S1 =
{(1, 1, 0)}
S1 = {(1, 1, 0), (0, 1, I)}
S1 = {(I, I, 0), (0, 1, 1), (1, 0, -1)}
S• = {(1, 1, 0), (0, I, 1), (I, 0, -1), (1, 1, I)}.
Obviously, S 1 is LI. Sa is also LI, because neither of the two vecton in S1
is a scalar multiple of the other. S 1 is LD, because
1(0, 1, 1) - 1(1, 0, -1) - (0, 0).
1(1, 1, 0), - o.
Hence. (1, o, -1) E [(I, I, 0), (0, I, I)), s.
is LD, because of Theorem
3.S.13 (b).
 J,5 UNBAR DIIPIINDBNCE, INDIPIINDINCI / 93

Exampl~ 3.33 In Example 3.32 find the largest linearly independent

subset whose span is [SJ.
As proved in Example 3.32, (I, O, -1) e [(I, I, O), (0, I, l)J and
hence (1,:0, -]) E [(I, I, 0), (0, I, J), (I, J, 1)). Therefore, we discard
(I, 0, -1). The span of the remaining set A= {(I, J, 0), (0, I, I), (I, I, 1)}
is the same as [SJ. Now ~e check for the linear independence of ..4.
Suppose
«(I, J, 0) + ~co, I, J) + y(I, J, I)== (0, 0, 0)
or (11 + r, 11 + ~ + y, ~ + r) = (0, o, O) •

This means 11 + r = O, 11 + ~ + y = 0, ~ +
y = 0. Solving these equa-
tions, we get only one solution, i.e. 11 = 0 = ~ = y. Hence, the set is LI.
Therefore, A = {( 1, 1, 0), (0, I, 1), (I, J, 1)} is the largest linearly inde-
pendent subset and [A) = [SJ.
Finally, let us extend the concept of linear dependence and linear
independence to intmite sets.
3.5.16 Deflaitloa An i,ifinite subset S of a vector space Y is said to be
linearly inden_endent (LI) if every finite subset of S is LI.
Sis said to be linearly dependent (LD, if it is not LI.
Example 3.34 The subset
S = {I, x, x•, x•, ... }
of fl' is LI.
For, suppose a 1x"1 + o,xk• + ...
+ anit'" = 0 with k 1, k1 , ••• , k. being
distinct nonnegative integers. Note that the equality is an algebraic
identity, since the right-hand side is the zero polynomial. So, either
by giving various values to x or by repeated differentiation of both
sides of the identity, we get
a1 = 0 = a1 = ... = a,,. •

Problem Set 3.5

I. Which of the following subsets S of Y3 are LI ?
(a) S = {(I, 2, I), (-1, I, 0), (S. -1, 2)}
(b) S = {(I, o. 0), (I, J, 1), (I, 2, 3)}
(c) S = {(I, I, 2), (-3, 1,0), (1, -1, J), (I, 2, -3)}
(d) S = {(I, S, 2), (0, 0, 1), (I, l, O)}
(e) S == {(1/2, 1/3, ll, (0, 0, 0), (2, 3/4, -1/3)}.
2. Which of the following subsets S or
Y, are LD ?
(a) S == {(I, 0, 0, 0), (1, 1, 0, 0), (I, I, 1, I), (0, 0, I, I)}
(b) S = {(1, -1, 2, 0), (I, I, 2, 0), (3, 0, 0, 1), (2,•I, -1, O)}
(c) S ""' {(1, I, I, 0), (3, 2, 2, I), (I, 1, 3, -2), (I, 2, 6, - ,),
(1, 1, 2, l)}
94 / VECTOR SPACES

(d) S = {() /2, 2/3, 3/4, 4), (0, 01 2, 0) 1 (1. 2 1 O, l), (1/2, 2/3, 3/41 4)}
(el S = {(1 1 21 3, 0), (-1. 7, 3, 3), (I, -1, I, -1)}.
3. Which of lhe following subsets S of fl are LI ?
(a) S = {r - 1, X + 1, X ·- I}
(b) S={l,x+r,x-x',3x}
(c) S = {x, x8 - x, x' + x'l, x + x• + x' + 1/2}
(d) S = {x1 - 4, X + 2, X - 2, .r/3}.
4. Which of the following subsets S of ('7'(0, ooJ are 1.1?
(a) S = {x, sin x, cos x}
(b) S = {sin1 x, cos 2x, I}
(c) S = {sin x, cos x, sin (x + I)}
(d) S = {In x, In r, In r}
(e) s = {re2 , xeZ, (x' + .'"C - l)eZ}.
S. Prove that the vectors (a, b), (c. d) are LD iff ad= be.
6. Prove that
(a) 1 and I arc LD in the set C of complex numbers considered as
a complex vector space
(b) 1 and i are LI in the set C of complex numbers considered as a
real vector space.
7. Show that the set S = {sin x, sin 2x, ... , sin nx} is a linearly indepen·
<lent subset of 'i; [-n:, n:] for every positive integer n.
8. Prove Theorem 3.5.13.
9. In the proof of Corollary 3.5.15 answer the question 'why'.
10. If u, v, and ware three linearly independent vectors of a vector space
V, then prove that u + v. v + w, and w +
u are also LI.
1I. Find a linearly independent subset .4 of S in Problem I such that [.4)
= [SJ.
12. Proceed as in Problem 11 for the sets S in Problem 2.
13. Proceed as in Problem 1 I for the sets S in Problem 3.
Jt Proceed as in Problem 11 for the sets Sin Problem 4.
15. In Problems 1, 2, 3, and 4 1 whenever a set is LD, locate one of the
YC'Ctors that is in tbe span of the others.
J6. True or false ?
(a) {sin x, cos x, 1} is LD in 7,'(0, 1).
(b) {i +J. i-J. 2i + 3j} is LD in v•.
(c)· Jf u1• u1, u8 are LD, then
«1U1 + a,u1 + «1111 = 0 ~ one of the «·s is not zero.
Cd) Every set of three ~ectors in J/1 is LD.
(e) A set of two vectors in J/1 is always LI.
(f) If a vector 11 is a linear combination of u10 u., ...• u11• then
{uh "■• ... , u.} is LI.
 3.6 DIMENSION AND BASIS / 95

(s) A subset of a linearly dependent set can never be LD.

(b) Every superset of a linearly independent set is u.
(i) x and I JC I are u in WCI, 4).

3.6 DIMENSION AND BASIS

We shall begin this article with the definition of a 'basis'.
3 6.1 Deftnltlon A subset B cf a vector space JI is said to be a basis for
Vif
(a) B is linearly independent, and
(b) [BJ = JI, i.e. B generates V.
Another definition of a basis, which is used by some authors, will be
introduced as an equivalent property in Theorem 3.6.8.
Example J.J5 Take JI = V3• Consider the set B = {i, j, k}, where
I= (I, 0, 0 ,j = (0, I, 0), k = (0, 0, 1).
The set B is u, b~ause
+
Iii+ ~j yk = «(I, 0, 0) + ~(O, 1.0) + y(O. 0, I)= 0 = (0, 0, 0)
implies that («, (3, y) = (0, 0, 0), i.e. « = 0 = ~ =-= y. Also B spans Jl1 ,
because any vector (x1o x,, x1 ) of V1 can be written as a linear combination
+ +
of i, j, and k, namely, (xi, Xz, X3) = ·x 1i xjj x 3k. Hence, B is a basi11
for V1 •
It may be seen that the set B1 = {(I, I, 0), (I, 0, I), (0, I, I)} is also a
ro5.is for J/3 , because any vector (x 1, x1, x 3 ) E J/3 can be written as
(Xi, X1, X3) = Xi + ~- - Xa (I, ), 0) + Xi + ~3 - x. ll, 0, I)
• + Xz + ~• - X1 (0, I, I)

and «(I, I, 0) t ~(I, 0, I) t y(O, I, I) = (0, 0, O)

implies Gt= 0 =~ = y, i.e. Bis u.
This example shows that a basis for a vector space V need not be
unique.
The two properties of a ba.,is B, namely. (i) Bis LI and (ii) B gene-
rates V are not entirely unrelated especially when B is finite. For, if a set
B of n elements generates V, then we can prove, with some effort, that no
linearly independent set can have more than n vectors. This resultJs the
content of the following important theorem.
3.6.2 Theorm In a ,ector apac, JI if {vi, '•• •.. , v,.} generate, V and fl
{wi, w1 , ... , w.} t, u, then m <: n.
In ocher words, we cannot have more linearly independent vecton than
the number of elements in a set of pneraton.
 96 / VECTOR SPACES

Proof: Let us adopt the foJJowing notations. If Sis a finite linearly

dependent ordered set of vectors u1, u1, ••• u.,, with U1 =I= 0, then
(i) by Theorem 3.5.14, there is a vector u., 2 < i < p, which is a hn~ar
combination of its predecessors. Let S' denote the ordered set \\htch
remains after the deletion of such a u, and
(ii) for every vector w, let wS denote the ordered set {w, u11 u,, ... , u.,,}.
Now construct the set
S1 = {Wm, V1, v1, ... , v,.} .
S1 bas the following properties :
(i) [S1] =- V, because {vi, ... , n,.} spans V and Wm E JI,
(ii) S 1 is LD, since Wm E V = [vi, v1, ... , v,.J, and
(iii) Wm =/- 0.
So, applying Theorem 3.S.14, we see that S~ can be formed by deleting
vector v,, which i,; a linear combination of w,.., Vi, v,, .. , v,_ 1 •
Now consider the set S,. = Wm_1 S~. Since s;
spans V, S2 also does 50,
Further, S1 is LD, since Wm-1 E V = [Sa and Wm_1 =I 0. Therefore, by
another application of Theorem 3.5.14, we form S~. Then v.e construct
.s;
the set S3 = Wm~ 1 and continue the process of constructing new sets S
and S'. Smee the set of w's is LI, every time the di,;carded element must
be av.
If all the w's are used up in this process, then m < n. Otherwise the
set {Wm-11, Wm-n+i, ... , Wm-h Wm} would be LD, which contradicts the hnear
independence of the w's, I
3.6.3 Corollary 1/V has a basis nf n elements, then e1•ery set of p vecto1s
witlr p > 11, is LD.
Proof:Suppose B = {vi, v2 , ... , v,.} is a basi-; for V. Let A
•
= {ui, u2, u.,,} be a set of p vectors, p > n. If A is LI, then p .--; n by
•• ,

Theotem 3.6 2. Hence, A is LD. I

3.6.4 Corollary I/ V has a basis of n elements, thrm every other ba,is for V
also has n elemellts.
Proof: Suppose B1 = {vi, v2, •• , vn} and B 2 = {wi, Wa, ... Wm} are
two bases for V. Then B1 and Ba are LI and [B1] = V, [B1] = V.
Since [B1] = V and B1 is LI we have, by Theorem 3.6.2, m < n. Since
[B1] = Vand B 1 is LI we get, hy the same theorem, n < m. Thus, m = n, I
The number of elements in a basis is therefore unique and hence the
following definitions.
3 6,5 Deflnldon If a vector space V has a basis consisting of a finite
number of elements, the space is said to be finite-dimensional; the
number of elements in a basis is ca1led the dimension of the space and
is written as dim J'. If dim V == n, Vis said to be n-dimtnsional. If
V is not finite-dimensional, it is called lnfinite-dimtnslonal.
 3.6 Dl.ft11!1'(SJON AND BASIS / 97
If Y = J-'0 == {O}, its dimension is taken to be zero.
3.6.6 Remark It folJows from Corollary 3.6.4 that, if a vector space Y
is n-dimensional, (a) there exist n linearly independent vecton ip Y
and (b) every set of n + l vectors in Y is LD,
Example 3.36 V1 is 2-dimensional, because (I, 0) and (0, 1) form a basis
for V1 • Ya is 3-dimcnsional, because Va has a basis of 3 elements, namely,
i,j, and k.
Y11 is n-dimensional, for, consider the elements
el= (1, 0, 0, ... , 0)
e1 = (0, 1, 0, ... , 0)

t'n = (0, 0, 0, ... , 1)

with e, as the vector, all of whose ccordinates are zero except the i -th,
which is I. It is em,y to see that eh e2, , en are LI and every n-tuple is
a linear combination of e., e2, ... , e".
Thus, the !let {eu e2 , ... , e"} is a basis for V". The basis {e1, e1, ... , ,.}
is called the standard basis for Y,. and will be used in the sequel without
any further mer,tion. In particular, note that {i,j, k} is the standard basis
for V3 • Jn other worJs,
e, = (I, 0, 0) = i
e8 =-= (U, I, 0) = j
e, = (0, 0, 1) = k
Example J.37 Find the dimension of the space 9'""
Every polynomial in £/111 is a linear combination of the functions
1, x, x2, ... , xn. rurther, this set is u. (Why ?) Hence dim f},. = n + 1.
3.6.7 Theorem In an n-dimensional vector .1pare Y, any set of n linearly
independe11t vectors is a ba.,ir.
Proof: Suppose B = {1•1,
v1 , .... v,.} is a set of n linearly indepen-
dent vectors. To prove that B is a basis we have only to show that B
spans V.
Take VE Y. The set {vh v1, ... , v.. , v} is LD, because V is n-dimen-
sional (cf Remark 3.6.6). Hence, by' Theorem 3.S.14, one of the vectors
v2 , v, • ...• v.. , say v, is in the span of all its predecessors. Obviously, this
one cannot be·any one of v1, v8, ... , v,., for in that case we shalt be contra-
dicting the linear independence of {v1, v1, ... , v,.}. Hence, v E {v1, ,., ••• , v,.).
Therefore, B spans Y. I
Example 3.38 Prove that the set l(l, 1, 1), (1, -1, 1), (0, 1, 1)} is a basis
fur~. .
To prove this we have to prove only that the set is LI, because Y1 is 3-
dimensional. Suppose
 98 / VECTOR SPACES
«(I, J, 0 + f3(J, -J, I)+ y(O, J, J) - (0, 0, 0)
or (at + 13. « - f3 + Y, 11 + 13 + y) - (0, 0, 0).
So ix + f3 = u, ix - 13 + y = o, 11 + f3 + y = O. These equations give
or: = 0 = r, = y. Hence, the set is LI and is therefore a basis for Y,.
We shaJJ now take up the theorem referred to immediately after Defi·
nition J.6.J. This theorem gives rise to the definition of a basis used by
some authors (see Remark 3.6.9).
3.6.8 neorem · Jn a vector space Y let B == {vi, v1, ... , Vn} span V. Then
the following two conditions are equil'alent :
(a) {vi, V:a, ... , Pn} is a linearly independent set.
(b) 1/ v E JI', then the expression v = «1v1 + «,v1 + ... + «nVn is
unique.
Proof: .Assume (a). We shall prove that any expression for v in
terms of vi, v2 , ... , 1111 is unique. For, if
I' = «1V1 + «1V1 + ... + «nVn (1)
and also
(2)
then «1V1 + CXzVa + ... + «nV11 = f31V1 + f'ava -j- ••. + ~nVn
or («1 - ~ 1)v1 + («1 - f31}v1 + .. + (ex. - 13.)v. = 0 .
But the v.'s are LI, so «. - ~. = 0 for all i. This gives «, = ~' for all i and
hence expression (I) is unique.
Conversely, assume (b). Suppose now that the v."s are not LI, i.e they

~ 1v1 + f3tv1 + ... + ~.v. ,

are LD. Then there exists a nontrivial linear combination, say

which equals the zero vector. But 0, on the other hand, is already equal
(3)

to the trivial linear combination

Ov1 + 0111 + •·• + Ovn • (4)
Thus, we get two different expressions, (3) and (4), for O. This contradicts
(b). Hence, {vi, v1 , ••• , v,.} is LI. I
3.6.9 Remark Note that the foregoing proof can be adopted for infinite-
dimensional cases also.
Further, note that a basis B has two properties:
(i) B is LI and
(ii) . [B] = Y.
Hence, the following conclusion is obtained : 'A set B is a buia ror a
vector space Y #ff [B] == Y and the expression for any , E J' in terms
or
of elements B is unique:
We now deftne the term coordinate vector.
 3.6 DIMENSION AND BASIS/ 99
3.6.10 Deftaitl• Let B = {111, 111, ••• , 11,J be an ordered basis for Y. Then
e:
a vector II JI can be written as v = «1v1 + «1v1 + ... + «11v11• The
vector (ixi, «1, ... , ix,.) is called the coordinate rector ofv relati,e to the
ordered basis 8. It is denoted by Ma. «1, ixa, ... , ix,. are called
the coordinates of vector v relative to the ordered basis B.
The coordinates of a vector relative to the standard basis are simply
called the coordinates of the vector.
It should be noted that Ma is unambiguously fixed in view of Remark
3.6.9.
Example 3.39 Find the coordinate vector of the vector (2, 3, 4, -1) o(
V, relative to the standard basis for V,.
The standard basis for JI, is {ei, e1 , e3, e,}, where e1 =
(I, 0, 0, 9),
e,. = (O, 1, 0, 0 J, e3 = (0, 0, I, 0), and e, == (0, 0, 0, I). Obviously,
(2, 3, 4, -1) = 2e1 + 3e2 + 4e1 - I e,. So the coordinate vector of
(2, 3, 4, -1) relative to the standard basis is (2, 3, 4, - I). Therefore,
2, 3, 4, and -1 are the coordinates of the vector (2, 3, 4, -1).
If the reader finds that this example is not very illuminating, he should
study the following example.
Example 3.40 Find the coordinates of (2, 3, 4, -1) relative to the
ordered ba5is D ::: {ll, I,o, 0), (O, I, I, O), (O, o, 1, I), (I, o, o, O)} for v•.
(The reader can verify for himself t'1at B is a basis for V,.) To get the
coordinates of (2, 3, 4, -1) relative to B, we write
(2, 3, 4, -I)= «(), I, 0, 0) + ~(O, J, I, 0) + y(O, 0, I, 1) + S(I, 0, 0, 0)
or (2. 3, 4, -1) = (« + 6, « + ~. ~ + y, y) .
This gives ix + 6 = 2, « + ~ = 3, ~ + y = 4, y == -1. Solving these
equations, we get « = -2, ~ = S, y = - J, and 6 = 4. Hence, the co-
ordinates of (2, 3, 4, -1) relative to the ordered basis B are -2, S, -1,
and 4. (-2, S, -1, 4) = [(2, 3, 4, - l))s,
We now prove a theorem on the extension of a linearly independent set
to a basis for a vector space Y.
3.6.11 Theorem Let the set {vJ.! v1 , ... , v1} be a linearly Independent subset
of an n-dimen.slonal ,ector space V. Then we can find vector,
•1+1• ... , v11 in Y ,wh that the .set {111, .. , v1, V1:+1t ••• , •J is a ba1i1
for Y.
Proof: By Theorem 3.6.2, k < n. If k == n, then, by Theorem
3.6.7, {11, 11, ••• , 1 11,.11,} ia a basis for Y. If k < n, then {111, "•• ••• ,-vt} is
not a basis for Y (Corollary 3.6.4). But the set {111, v._ ... , ,.} is Lt.
Therefore,(,., , 1, ••• 1 •al#:, Y. Hence, [v1, 1,, ... , ,.;i is a proper subset of
Y. Tbus, there exists any nonzero vector V..+1 in Y such' that ,.,., •
[,,, '• ... , ,.i.
 JOO/ VECl"OR SPACES
Hence, the set {111, v1, ... , 111 , 111 ~1} is u; otherwise, by Theorem 3.5. 1_4,
one of the vectors v1, 118, ••• , 11.t, say 11L+1t is in the span of all its
predecessors.
Now, if k + I = n, we are done. If not, we repeat the foregoing
process until we get n linearly independent ,ector11 {111, '•• ... , V1,
Vt-t-1t •• , 11
11 } . This forms a basis for V by Theorem 3.6. 7. I
3.6.12 Remark We can produce any number of bases, because we can
start from any nonzero vector c1nd extend it.
Example 1.41 Given two linearly independent vectors (1, {I, 1, 0) and
(0, -1, 1, 0) of v,, find a basis for Y, that includes these tYio ,ectors.
((1, 0, 1, 0), (0, ---1, I, 0)) ={(at,-~. at+~. 0) I at,~ any scalars}.
As in Theorem 3.6. ll, so too now we choose a vector outside I his span
and get an enlarged linearly independent set.
Since the fourth coordinate is always zero for vectors in this span,
certainly (0, 0, 0, I) is not in this span. Thus, we get an enlarged lmearly
independent r.et {(I, 0, 1, 0), (0, -1, 1, 0), (0, 0, 0, 1)}, whose !.pan is
((1, 0, J, 0), (0, -- 1, 1, 0), (0, 0, 0, l)] = {(at, - ~. at + ~. y) I at,{\, y are
scalars}. Now we have to identify one element outside this span. Given
«, ~. y, the third coordinate in the elements of this span is always « + ~-
So we shall look for a vector for which this is not true. Clearly,
(1, -2, 0, 0) is not in the span of the earlier set. So we have a set
B = {(1, o, I, 0), (0, -], J, 0), (0, 0, 0, 1), (1, --2, 0, O)},
which is u and, by Theorem 3.6.7, it is a basis for V,. The reader is
advised to verify that this is indeed a basis.
Example 3.42 Let {(I, 1, 1, I), (I, 2, 1, 2)} be a linearly independent subbet
of the vector space V,. Extend it to a basis for Y,.
We have
[(I, 1.:1, J ,), (1, 2, 1, 2)) = {(at+~. at+ 2~. at -t- ~.
a. + 2~) I at, ~ are bcalars} .
Since the first and third coordinates are equal for all vectors in the span,
we find that (0, 3, 2, 3) is not in the span. Thus, we have an enlarged
linearly independent set {(I. I, 1, I), (I, 2, 1, 2), (0, 3, 2, 3)}, whose span is
[(I, 1, I, 1), (1,2, 1, 2),(0,3,2, 3)) = {(« + ~. « + 2~ + 3y, « + ~ + 2y,
« + 2~ + 3y I ac, ~. y are scalars} •
Obviously, the vector (2, 6, 4, S) is not in this span. Hence, the set
{(I, I, 1, )), (1, 2, 1, 2), (0, 3, 2, 3), (2, 6, 4, S)} is u. Thus, by Theorem
3.6.7, it is a basis for Y,:
We shall now prove two theorems on dbnensiom of subspaces.
3.,.13 'l'lleor• Let U 6, a 111bl,occ of a /inltNlimnulonal ,·,ctor ,poet Y.
TINn dim U < dlm Y. Bfuallty hold, only t1·htn U - Y.
 S6 DIMENSION AND BASIS/ 101
=
Proof: Let B {vi, v1, ••• , v,.} be a basis for Y. This generates Y
and has n elements. Any set of linearly independent vectors in Y and
therefore any set of linearly independent vecton in U cannot have more
than n vectors.. Therefore, dim U , dim Y.
When dim U = dim V, B1 , a basis for U, is a set of n linearly indepen-
dent vectors of V, whose dimen.;ion is also n. So, by Theorem 3.6.5, it
follows that B1 i, a basis for Y. This means V =[BJ= U. I
3.6.14 Theorem 1/ U and W are two subspaces of a finite-dimensional
vector space Y, then
dim (U = dim U + dim W -- dim (Un W).
+ W)
Proof: Let dim U =- m, dim W = p, dim (U n W) = r, and dim
V = n. By Theorem 3 6.13, m < n, p <: n, r < n. Let {v1, v2 , .. , 1•,} be
a basis for U n
W. This is a linearly independent set in U W and n
therefore in U as well a'> ID W So 1t can be extended to a basis for U, say
{v19 Vi, .. , r,, u,Ttt ... , u,.l \5)
and to a basis for W, say
(6J
We shall now prove that the set
A = {vi, v2, ... , v,, Ur+u ... , Um, Wr+t, ... , w11} (7)
is a basis for U + W. In fact, this will complete the proof of the theoren1,
because in that case, the dimension of U + W will be the number of
elements in A, i.e. r + (m - r) + (p - r) = m + p -- r. This is what the
theorem asserts.
To prove that A is a basis for U + W, we have to prove that (,1.) .4 is
u in U + Wand (b) [A] = U + W. To prove (a) let us assume lhat
r m
I «,v, + I ~.u, + (8)
1-1 i•r+l
This gives
r m A (
I «,v, +l=r+l
I "'•"• = - ~
l•r+l
y,w, (9)
== ,, say.
Tho vector , is in U, because the lef't-hand side of Equation (9) is in U. v
is also in W, since the right-hand side of Equation (9) is in W. Thus,
, e U n W. Therefore, , can be expressed uniquel) in term, of

(10)

for auitable 8's. Hence,

r II
:E ~v,
1-1
+1-,+1
E y,w, - 0. (ti)
102 / VECTOR SPACES
But {v1, , 1 , ••• , ,n Wr+i, ••. , w.,.} is u. So each of the 6.'s and r.'s is zero.
Substituting Yr+i = Yr+i = ... = y.,. = 0 in Equation (9), we find that
r m
I «1111 + E ~.u, = 0 . (12)
i=-1 i=r+l
Again, {111, ••• , llr, ur♦ 1 , ••• , Um} is LI. So each of the rx.·s and ~.'s is zero.
Thus, Equation (8) implies that each scalar involved is zero. Hence, A :s
LI, which proves (a).
To prove (b), let z e: U + W. Then z = u 1 - w, where u e: U and
w € W. This gives
r m r P ,
z= E «1 111 + I ~.u. + E at;v, + E ~,w1 (13)
i=l i=-r 1-1 i=I i=r+l
for suitable scalars at,'s, ~.'s, at~'s, ~~•~. Simplifying expres&ion (13), we see
that z € [A). Hence, U l- WC (A] The reverse in equality is obvious,
because A C U + W. I
3.6.15 Corollary If U and W are subspaces of a fimte-d,mensional vector
space V such that U f"\ W - {O}, the11
dim (U E9 W) = dim U + dim W .
The proof is obvious and is left to the reader.
As ilJustration, take U =- the xy-plane and W-= the J'z-plane in Y1 .
Clearly, U and W are subspaces of Ya and dim U =-= 2, dim W = 2.
U n W = y-axis, whose dimension is I. By Theorem 3.6.14, we have
dim (U + W) = dim U dim W - dim (U n W) • +
But U + W, in this case, is Ya. So
dim Ya = (xy-plane) + dim ()'z-plane) - dim (y-axis)
or 3=2+2-1.
This verifies the result of Theorem 3.6.14.
On the other hand, if we take U = xy-plane and W = z-axis, then
U n W = {O} and U + W = Ya. Also,
dim Y1 = dim (U (B W) = dim U dim W , +
M 3=2+1. .
This verifies the result of Corollary 3.6.1 s.
Now we give a comprehensive example which illustrates different ideas
studied in this chapter.
E;tample 3.43 Take Y == fl 1, the space of all real polynomials of degree at
moat 3. Let U be the subspace of fl 1 consisting of those polynomiaJs of {}1
that vanish at x == 1. Let W be the subspace of fl1 consistiq of tboso
polynomials of fl, whose first derivatives vanish at x == 1. Study the sub•
apac:ea u. W, U n W, U + Win terms of dimension, basis, and the exten•
1ion of each of these bases to a basis fm J' - 111•
 S.6 DIMENSION AND BUIS/ 10!
Let p stand for an arbitrary polynomial. Recall that R stands for the
set of real numbers. Then
V = {p I degree of p < 3}
+ + +
= {« px yx1 Sx' I «, Ii, Y, S E R} .
We know that dim Y = 4. Also
U = {p e: Y I p(l) = O}
+ +
= {at Px yx2 + Sx' I « + Ii + y + 6 = O;
«, Ii, Y, 6 € R}
+ +
= {( -Ii - y - S) + !ix rx 2 Sx8 I ~. y, 6 e: R}
+
= {P(x - 1) + y(r - 1) S(r - 1) I~. r, SER}.
This bhows that the set {x - 1, x 2 - 1, x 8 - 1} spans U. Since it is
easily seen that this set is LI, it also forms a basis for U. So dim U = 3.
This basis can be extended to cl basi!1 for V = fl'a by taking another
element that is not in the span U of these polynomials. Such a polynomial
is a constant polynomial, say I. Thus, by Theorem 3.6.7, {], x - 1,
,c1 - I, x• - I} is a basis for Y = f} 8 • Again,
e:
w = {p y I p'(l) = 0}
-= {at + li.i + yx• + sx• I ~ + 2r + 36 = O; at,~. y, 6 e: R}
--= {at + y(.tJ - 2x) + 6(.~ - 3.x) I «, y, S E R} •
This shows that Wis spanned by {l, x2 - 2x, x' - 3.x}. Since this can be
shown to be Lr, it follows that {1, .JC1 - 2:x, x8 - 3:x} is a basis for Wand
dim W = 3.
The extension of this basis to a basis for 9'1 is got by looking for just
one vector that is not in W. Such a vector is a polynomial x. Thus,
{1, x, x8 - 2x, x' - 3x} is a basis for 9'8 • Again,
U n W = {p E 9'8 I p(l) = 0, p'(l) = O}
+ + + + +
= {« + ~x yr Sr I at ~ r S = 0,
Ii + 2y + 36 = O; at, ~. y, S E R}
= {y(l - 2x + x8) + 6(2 :- 3,c + x') I y, S E R} .
Clearly, 1 - lx + r and 2 - 3x + x8 are u. So dim U n W = 2 and
{1 - 2x + x•, 2 - 3x + x8} is a basis for Un W. Now we look for
elements that arc not in U n W. One such element is ,cl. Now consider
the span of the linearly independent set {1 - 2x + x', 2 - 3x + x8, x8}.
The span of this set is still not Jf, because dim Y = 4. So look for one
more element outside this span. One such element is x'. (Why? Can
you find others ?) Hence,
{l - 2x + +
,cl, 2 - 3x xi, r, x8}
is a basis for Y. Finally,
U +W- +
{p f IP E U, f E W}
+
- {a&i(¥ - I)+ t&a(Jt' - 1) t&a(x' - I)+ (11
+ +
'8fx' - 2¥) '8(x8 - 3x) I ••• -.. -., '•• '8, ,. e: .I} •
104 J VBCl'OR SPACES
So B - {(x - l), (x' - 1), (x' - 1), 1, (or - 2.t), (x' - 3x)} spans the
+
space U W. But
dim (U + W) = dim U + dim W - dim (U n W)
==3+3-2=4.
Therefore, B is LD. Applying Theorem 3.5.9 to the ordered set C
= {I, x - t, x 1 - I, x• - 2x, x• - 1, x 3 - 3x}, which is LD, we find
that x" - 2x E (1, x - 1, x 1 - 1], for
~• - 2x = ](xi - I) -- 2(x - 1) - 1(1) .
Hence, r- 2~ can be discarded. Wr: are thus left with the set
C' = {l, x -- 1, ."tJ -- l, x3 -- 1, x• - Jx} ,
which has S elements. This 1s still LD (fiince dim (U + W) = 4). Again,
by Theorem 3.S.9, there exists a vector that is a linear combination of all
its predecessors. It cannot be l or x -- I or x2 -- 1 or x3 - 1, since these
four are u among them!iclves. So let us check \\ hcther x3 - 3x is a
linear combination of its predeces!iors. We have
+
x' - 3x = l(.'t:i - 1) 0(x~ - 1) - 3(x - ]) - 2(1) .
Hence, A. 3 - Jx e: (), x - I, x~ - J, x 3 - 11 So x 1 -- 3x ran he dis-
carded. We arc thus left with the set
D = {I, x - l, A.1 - I, x3 - 1},
whi\lh is obviously a basii. for U + W. Since dim (U + W) = dim V = 4,
we get U + W --= V, by Theorem 3.6.13.

Problem Set 3.6

I. Which of the following subsets S form a basis for V3 ? In case S is
not a basis for Vi,, determine a basis for [SJ.
(a) S = {(I, 2, 3), (3, 1, 0), (--2, 1, 3)}
(b) S = {(l, 1, I), (1, 2, 3), (-1, 0, l)}
(c) S = {(0, 0, I), (1, 0, 1)_, (1, - 1, 1), (3, 0, l)}
(d) S = {(I, 2/S, -1),(0, I, 2), (3/4, -1, I)}
(e) S = {(-1, 3/2, 2), (3, 2/3, 3)}.
2. Which of the following subsets S form a basis for the given vector
space V?
(a) S = {(I, -1, 0, I), (0, 0, 0, 1), (2, -1, 0, )), (3, 2, I, 0)},
Y== V,
(b) S == {(O, 1, 2, 1), fl, 2, -1, 1), (2, -3, 1, OJ, (4, -2, -7,-5)},
Y= Y,
(c) S = {x -- I, ,cl + x -I, x• - x + l}, V == fP1
(dJ S = {l, x, (x -l)x, x(x - 1) (x - 2)}, Y = fP1
(e) S == {1, x, (3r - 1)/2, (Sx'-3x)/2, (35.¥' - 30x' + 3)/8},
y = fl.
(f) S == {1, X - 2, (X - 2)1, (X - 2)'}, V == f/ 1
 3.6 DIMBNSION AND BASIS/ 10S

(g) S = {l, sin x, sin2 x, cos• x}, y = ~[-ff, ff)

+
(h) S = {(I, I, 1 I), (1, I, 1 - i), (11 -I, 1)}, Y = Yi.
3. Determine the dimension of the subspace [SJ of Y8 for each S in
Problem 1.
4 Determine the dimension of the subspace [S] of Y for each S in
Problem 2.
S. Extend the set {(3, --1, 2)} to two different bases for Y1 •
6. Given S 1 = {(I, 2, 3), (0, 1, 2), (3, 2, I)} and S1 = {(I, -2, 3),
(-1, 1, -2), (I, -3, 4)}, determine the dimension and a basis for
(a) [S1] n [S2] (b) [S1] + [S1].
7. Given Sas a finite subset of a vector space Y, prove that
(a) If S is LI and every prorer superset of S in Y is LD, then S is a
basis for V.
(b) 1f S spans Y and no proper subset of S spans Y, then S is a
basis for Y.
8. Find a basis for a subspace U of Yin the following cases :
(a) U = fp E 9' 1 Ip• = O}, Y = 9',.
(b) U = {p E 9' 1 I p(x0 ) = p'(x0 ) = O}, Y = f/' 5
(c) U = {(xi, x •• ... , x,.) € Vn I «1X1+ ... -1- «nXa = O; «i, .•. , «.
are any n scalars}, Y = Y,.
X:1 + Xa =
(d) U = {(Xi, X2, Xa, ..-,, x,) E YI• 3xXi+
1 - x, + 7x6 =
0}
0 '
y = Ya
(e) U = {p E 9\ I p(x0) = O}, V = H',.
9. Let U and W be two distinct (n - ))-dimensional subspaces of an
n-dimensional vector space Y. Then prove that dim (U n W)
== n - 2.
10. (a) Prove that every I-dimensional subspace of Y1 is a straight line
through the origin.
(b) Prove that every 2-dimensional subspace of V1 is a plane
through the origin.
(c) Deduce that the intersection of two distinct planes through the
origin is a straight line through the origin.
11. Let ...., = at,e.t - at,,t,, I = 1, 2, ... , n, and for a fixed k( <: n), be n
vectors of v. with ra. 11 ¢ 0. Then prove that Y1, '•• ... , '.t-1• Y1+1, •··•
v,, is a basis for U of Problem 8(c).
12. Find the coordinates of the following vectors of Y1 relative to the'
ordered basis B == {(2, 1, 0), (2, I, I), (2, 2, I)} : •
(a) (I. 2, I) (b) (-1, 3, 1) (c) (x1, x1, x1)
(d) (-v'2, w, e) (e) (-1/2, 11/3, S) (f) (2, 0, -1).
106 / vscroa Sl'ACES
13. Find the coordinates of the following polynomials relative to the
ordered basis {I - x, 1 + x, l - x'} of fl 1 :
(a) 3 + 1x + 2x8 (b) x - 3x3 (CJ x8 + 2x - I.
14. Find an ordered basis for v, relative to which tho vector
(-1, 3, 2, I) has the coordinates 4, 1, -2, and 7.
15. Let B1 = {v1, v1, ... , v11} and B1 = {u1, Us, ... , u11} be two ordered
bases for an n-dimensional vector space V and let
U1= «11V1 + «111'1 + ... + «nil'"
= «111'1 + 0Ca1V1 + •" + «111Y11
.. ...
U9

Un = «1nV1+ «a,1V1 + ... + «1111V1J,

Then find the coordinates of a vector v e: Y relative to the basis B1 ,

if its coordinates relative to the ba11is B1 are xi' x1 , ... , x•.
16. Find the dimension of vr.
17. Construct two subspaces A and B of V, such that dim A = 2,
dim B = 3, and dim A n B = I.
18. Let B1 = {u1, u1, ... , Un} and B1 = {vi, v1, ... , Vn} be ordered bases
for an n·dimensional vector space V i.uch that {u1 - v" u1 - V.a, •• ,
Un - v11} is Lo. Then prove that there exists a nonzero vector u E V
such that [u]a1= [u]a1,
J9. True or false?
(a) Every vector space has a finite basis.
(b) A basis can never include the zero vector.
(c) If two bases of V have one common vector, then the two bases
are the same.
(d) If every set of p vectors of V with p > n is LD, then Y has a
set of n generators.
(e) A basis for Va can be extended to a basis for Y,.
(0 A basis for Va is {i +J + le, i +i, i}.
(g) A basis for fl 1 is {I, 'lx, (x - 1)1}.
(h) A set of generators of fla is
{(x - l)a, (x - 1)1, (x - ll. r- 1, I}.
(i} M• is independent of B.
(j) {(l, 0, O), (0, 1, 0), (0, O, l)} is a basis for the.complex vector
space Yi.
(k) In Y1, if (2, 3) is the coordinate vector or (2, 3) relative to the
ordered basis B, then B is the standard basis.
Chapter 4

Linear Transformations

The significance of vector spaces arises from the fc1ct that we can pass
from one vector space to another by means of functions that possess
a certain property called linearity. Thtse functions are called linear
transformations.

4.1 DEFINITION AND EXAMPLES

4.1.t Definition Suppo11e U and V are vector spaces either both real or
·both complex. Then the map T: U ➔ JI is said to be a linear map
(tramfo,mation, operator), if
+
T(u 1 + 118) -= T(u1) T(u2) for all Ui, u1 E U 0)
and T(ocu) =-- ocT(u) for all u E U, and all scalars «. (2J
A linear map T: U - U is also called a linear map on U. Whenever
we say T : U -~ V is a line.,r map, then U and V !>hall be taken as vector
spaces over the same field of scalars.
4.1.2 Remark In Definition 4.1.1 the 'plus' in u1 + u2 denoles the addi-
tion in space U, and the 'plus' in T(u 1J + T(u1 ) denotes the addition
in space V. We shall not elaborate these delicate points in the sequel.
A similar remark is true for the two scalar multiplications implied
in Equation (2).
Example 4.1 Define T: Va ➔ Y1 by the rule
T(x1r x1, x1) = (xi, x1, 0) .
It is called the projection of J/1 on the x1x.-plane. To prove that it is a
linear map, we have to show that
T(x + =
y) T(x) + T(_v) and T(«x) =
«T(x)
=
for all x, y E V1 and all scalars «. Let x (x1, x8, xa) and)' = <Y1, Y1, .Ya),
Then ·
+ + +
x y == (x1 Y11 x, Y,, Xa Ya) +
and «x - («xi, OIX1, «Xa) •
108 / LINBAR TaANSPOlOtATJONI

Now, by definition of T, we have

T(x + y) == T(xi +Yi, X1 + Ya, x, + Y,)
= (X1 + Y1, x. + , •. O) •
On the other hand,
T(x) + T(y) = T(xi, Xa, Xa) + T(yi, .Y1, Ya)
= (xi, x1 , 0) + b'1t y1 , 0) (definition of T)
== (x1 + )'1, x 1 + )'1, 0) (addition in J',).
Thus, T(x + y) = T(x) + T(Y) for all x, y E J',.
This verifies one requirement for the linearity of T. Again,
T(cxx) = T(«Xi, CXX1, «Xa)
= (cxx1 , «x1 , 0) (definition of T)
= «(x1, x 1, 0) (scalar multiplication in V1)
and
cxT(x) = cx7'lx1, x 1 , xJ} = «(xi, x 1 , 0) (definition of T).
Thus, T(xx) = «T(x) for all x E Y3 and all scalars «.
Hence, both the conditions for linearity of Tare verified. So Tis linear.
The reader should note that we have gone through the verification
rather leisurely. We shall not be able to afford such a leisurely pace in
the sequel. But every time we say that a map is linear the student would
do well not to take it for granted but •~ther to go through the verification
in as much detail as his understanding demands.
Example 4.2 Define T: J'8 -➔ Y1 by the rule
T(Xi, X2, Xa) = (X1 - Xa, X1 + X8) ,
This is a linear map as we shall now show.
If x = =
(x1 , x1 , X 3) and Y (Yi, Y1, Ya), then
T(x + y) = T(Xi + Yu Xa + Ya, Xa + Ya)
= (X1 + Yi - X1 - Ya, Xi+ Ya+ X3 '+ Ya),
whereas
T(x) + T(y) = (Xi - X1, Xi + x,) +
(yi - .Ya, Yi + Ya)
==- (x1 - X1 + .Yi - Y1, X1 + Xa + Y1 + Ya)
== (xi + Yi - X1 - Ya, X1 + Y1 + X3 + y,J .
Hence, T(x + y) == T(x) + T(y) for all x, y e:
Y,. The remaining part,
namely, T(«x) = «T(x) can be completed by the reader. Then it will
follow that Tis linear.
Example 4.3 Define T: Y8 -+ Y1 by the rule
T(xi, x•• xJ = xr
+ xi + xi ,
This is not a linear map, because for x - y =
(l, 0, O)
T(x + =
y) (x, + yJI + (x1 + y 1) 1 + (x1 + y.)1 - 4 ,
whereas
T(x) + T(y) =- xf + xi + xi + yf + yl + yl -= 2 •
Note that even if one or the two requirements of the definition of linearity
fails, Tis not linear.
 4.1 DEFINITION AND DAMPLII / 109
Example 4.4 Define T : U - JI (U and Y being vector spaces) by the
rule
T(u) = Ov
for all u E U. This is a lineair map. It is called the zero map, because it
maps every vector to the zero vector of Y.
Example 4 5 Let U be a vector sp&ce. The identity mop lu : U ➔ U
defined by the rule
Iv(u) "'= u
is also linear.
Example 4.6 Define T: V1 - V1 by the rule
T(x10 x 1 ) = (x1 , -xJ) .
This is a linear map and is called the teflection m the x 1-axis (Figure 4.1).
><2

r•····
0 I
I
I
! •
--
,
"I

>- cx1,-x,>
FIGURE 41
Example 4.7 Define D: ~ CU(a, b) - > t (a, b) by the rule
D(f) =I'•
where/' is the derivative off. The facts that D(f + g) = (( + g)' = f' + ~•
and that D(a.f) =- (a./)' = r,f' are elementary but important rei.ults in
calculus. In view of these results, we see that D is linear. This trans-
formation D is called the Jijferelllial operator.
Example 4 8 Define J : r (o, b) - R by the rule
b
c'l(f) = J/(x) dx.
a
Again, by the properties of integral, we find that J is a linear map.
Example 4.9 Suppose Ne -f O is a fixed vector of a \'ector space U.
Define T : U -➔ U by the rule
T(x) = x + u0
for all x E U. This map Tis not linear, because
T(x + y) = (x + y) + u0
=I= (x + 110) + (y + u0) = T(x) + T(y) •
This map is called traMlatlon by the vector u0• •

4.1.3 Remark The function/: R ➔ R defined by fC.x) - x + a ('a' fixed)

ii customatlly called a linear function, because its sraph in the xy-
l 10 / LINEAR TllANSfORMATIONS

plane is a straiaht Hile. But it is not a linear map from the vector
space V1 to itself, in the sense of Definition 4.1.l.
4.1.4 neorem Let T : U -> Y be a linear map. Then
(a) T(Ou) = Oy, (bl T(-u) =
-T(u), and
(c) 1T«1U1 + ... + «11u,.) = «17'(.u1) + ... + «.T(uJ.
In other words, a linear map T transforms the zero of U into the zero
of Y and the ne,ative of every u E U into the negative of T(u) in V.
The proofs of (a) and (b) are left to the reader. The proof of (c) is
obtained by induction on 'n', starting from the fact that T(ixu) == «T(u) and
using the property
T(«u 1 + ~u2 ) --= T(«u1) +
T(~u1 ) = ixT(u1) + ~T(u1) •
In viev. of (c), we get a standard technique of defining a linear trans-
for.nation Ton a finite-dimensional vector space. Suprose B == {u11 u1 , ... ,
u,.} is a basis for U Then any vector u e: U can be expres!led uniquely in
the form
u = cz 1 u1 + «.uJ + ... -1- ix,.u,. .
So, if T : U - V is a linear map, then
T(u) = T(«1 u1 + «1 u11 + ... + «.u,.)
= «1 T(111) + «1 T(u1 ) + ... + a:,. T(u,.) •
Thus, T(u) is known as soon as T(u1); T(112), ... , T(u,.) are known. This
is formalised in the following theorem.
4.1.5 Theorem A linear transformation T is compe/tely determined by its
values on the elements of a basis. Precisely, if B = {ui, u11 , ... , Un} is
"n
a basis for U and v11 v8 , .. , be n vectors (not n;,cessarily distinct) in
Y, then there exists a unique linear transformation
T: U-~ V
such that
T(u,) = v, for i = I, 2, .. , n. (3)
Proof: let u e: U. Then u can be expressed uniquely in the form
= ac 1u1 + «1u1 + ... + a:,.u,. .
u
We define
T(u) = x1v1 + at1 v8 + ... + at,.11,. • (4)
We now claim that this tran!lformation T is the required transformation.
To prove our claim, we have to show that (i) T is linear, (ii) T satisfies
Equation (3), and (iii) Tis unique.
(ii) is obvious, since u, = Ou1 + ... + Ou,- 1 + Ju, + ... + Ou., and so
T(u,) = h, = v, for all i.
(iii) follows, because if there were another such linear map S with
S(uJ ==,,,then
S(u) = S(at1u1 + at1U1 + ...
+ at,.u.)
= «1S(u1) + izsS(ua) + ... + «.S(u,.) (S is linear)
- 111'1 + Gt~"• + ... + «-'•
- T(u).
 4,1 DEPINITION AND.EXAMPLES/ 111
This is true fo, every u E. U So S = T.
It only remains to prove (IJ, whicb is just a verification of the two
relatfons
T(u + v) = T(u) + T(v) and T(atu) == cxT(u)
for arbitrary u, v E U and all scalars ex. Let u, v be two vectors of U.
Then
U =- IJ.1U1 + ... + «,.U,u V = ~1U1 -t ... -1- ~,.u,.
and we have
U +V = («1 -j- ~1)U1 + ,,, -t- («11 -J- ~,.)U11 •
Hence, by the definition of T, we have
T(u + v) = («1 + ~1)V1 + ... + (1J.11 + ~11)1'11 •
Also, T(u) -1- T(v) = (« 1v1 + ... + o:nv,.) + (~1l'1 + ... + ~-,l'11)
== («1 -j- ~1)111 + • • I- («,. + ~ ..),•11 •
Therefore, T(u + 11) = T(u) + T(v). Again,
T(«u) = «« 111 1 + .. . 1· «1,.11,.
= «(«11'1 + •· -1-- 11 P,.) 11

= «T(u).
Thus, T is linear aad the theorem is proved. I
4.1.6 Remark Theorem 4.1.5 will be used in the following way : To de-
fine a linear map, very often we shall be content with listing
T(u1), Ttu1 ), •• , T(u,). This m1;ans the value of Ton a general u is
to be obtained by the prr ce&s shown in the theorem, namely, if

then
T(u) = 1J.1T(u1) + tJ.2T(u9) t- ... + tJ.aT(u,.) •
This proce-,s of defining Ton the basis and extending it to the re-
maining elements is called linear t."(tension </ T. So we shall simply
say •Define T on a basis and extend it linearly' or, equivalently,
'Define the linear map T by specifying its values on a basis'.
Example 4.10 Suppose we want to define a linear map T : V1 - Y,.
Take a basis for Y1 , say {(I, 1), (1, -]}}. We have only to fix T(l, l}
and T( I, --1) in v,. In fact, every ordered pair of vectors in Y, will give
us one linear map T. We shall cite_ a few in Table 4.1.
TABLE 4.1
------- - - - - --- ---- ------
Linear map-
- Ti Ta Ta T,
Value at
...
(0, o. o. 0) (0, 1, o. 0) (I, I, l, 1)
(l, 1)
. (1 1 l, 0, O)

(1, -1) (0, 0, 0, 0) (I, 0, 0, 0) (-1, -1, (0, o, 0, 0)

-1, -1)
112 / UNBAR T'tNIPOJUQlJONS
Let us now linearly extend each of these maps to any u E YI The
actual values of T(u) will depend on the coordinates of u relative to the
basis {(I, 1), (1, -1)}.
Let u = (x, y) E V1 . Then
x+y x--y
(x,y)-== 2 - (I, I) 1-- 2 -(1, -1).
Thus, we get
T1(x,y) = -X+J' X -y
2 T1 (I, 0 +-2- T1 (I, -l)
x+y x-y
= 2- - (0, 0, 0, OJ + 2 - (0, 0, 0, 0) --= (0, 0, 0, 0)

=O.
Therefore, T1 is the zero map.
X - y
T1(x,y> = X----~2 }'
T2(1, I)+ -- 2-- T1(1, -1)
x+y x-y
= -2- (0, ], 0, O) + --T (I, 0, 0, 0)
x-y x+y
=-=<--2. -2--,0,0).
't' -I- )' X -- )'
T8 (x, y)-=- - 2 - T3 (1, lJ -t - 2 - T3 (1, -1)

= x+J'
--f - (I,
X-}'
I, 1, I) +(-2-) (-1, -1, -1, -1)
=- (y, y, J', y} .

T,(x, y) = -X -f-y
2 -Til,
X - y
I) -t - 2 - T.(1, --1)

= --x+y
2 (I, 1, 0, 0) + x-y
- 2 (0, 0, 0. O)
xi..y x+v
- ( 2 • -2-• 0, O) .

Problem Set 4.1

1. Let U and V be vector spaces over the same field of scalars, and Ta
map from U 10 V. Then prove that T is linear if/ T{«u1 t u1)
= «T(r,1) + T(u2) for all ui, u1 E U and scalar «.
2. Which of the following maps are linear ?
(a) T : V1 ➔ Ya defined by T(x) = (x, 2x, 3x)
(b) T: V1 ➔ Ya defined by T(x) = (x, x•, :r)
(c) T: Vi-+- Yi defined by T(x, y) = (x +
«, y, 0), « ,¢: 0
(d) T: Y1 ➔ Y1 defined by T(x, y) = (2x + 3y, 3x - 4y)
(e) T: Y8 ➔ Y1 defined by T(x, y, z) == (x1 xy, xy, yz) +
(f) T: V1 ➔ Y1 defined by T(x, y, z) = (x, y)
 4.1 DEPINITIO~ AND EXAMPLES/ 113

(a) T Vi -+ Jli defined by T(x, y) = (y, x)

(h) T Ya ➔ Va defined by T(x, y, z) = (x + y + z, 0)
"(i) T fJ ➔ fJ defined by T(p) = p• + p
(j) T fJ ➔ 9' defined by T(p)(x) = xp(x) p(l) +
(k) T : 11,"' [O, 1] -+ V2 defined by T(f) = ( /(0), /(1))
(I) T 9' -➔ [f' definP.d by T(p) = p(O)
(m) T fl' - fl' defined by T(p)(x) =- 2 + 3x + 7x3 p(x)
, x•
{n) T: g, -► g, defined by T(p)(x) = p(O) t xp'(O) + - 21 - p•(o)
{o) T: fl' _,.. 9' defined by T(p) = p'
(p) T: '(( 1" 1(a, b) -> 7:-'(o, b) defined by T(/) = a11 J + ai.f' + ...
+ a.,f'"I, a;'s are fixed scalars
(qJ T: '%-' 121 (a, h) - ► t./ (a. b) defined by Ttf) = (3x 2 -i- 4)/'
+ (7x + 3)/' -I (3x -I- 5)/.
3. Determine whether there exists a linear map in the following cases,
and where it doe!> exist give the general formula.
(a) T: V2 - ~ Vi such that T( I, 2) -. (3, l1) :ind T(2, J) = (1, 2)
(b) 1': V2 ,,. V2 such that T(2, I) - (2, J) and 71 I, 1) =-= (4, 2)
(c) T: V3 :- V~ such that F(O, I) -= (.\ 4), T(3, I)= (2, 2), and
T(J. 2) = (5, 71
(d) T: VJ > V., c;uch th.tt 7'(0, 1. 2) -__, (3, I, 2) and T(,, I, J)
~=- (2, :', 1)
(e) T: H' 3 • ... ,'J\ ~uch that T( I ·! x) · 1 -: x, T(2 t x)
-- x -l 3x2, and T(x~) = 0
(i) T: 9, - ➔ fl. 1 such that T(I I x) -= I, T(x) =s 3, and T(x 2) • 4
cg) T: Vf ,. Vi such that T(i, i) ~ (I +;,I).
4. Determine a nonzero linear transformation T: V8 ► V2 , which maps
all the vectors on the line x ::.:: y onto the origin.
5. Determine a linear tranformation 7': V~ • i·2 • which niaps all the
vrctors on the I ine x 1 y = 0 onh> thcmsch'cs l T ,: /).
6. Let T: V~ - ► V~ be defined by T(-x1 i· i~i, «! + i~11 ) :-:.: <«u «2 ).
Then prove or disprove that Tis linear.
7. Prove that a Ji near transform.it ion on a 1-dimem.ional vector space is
nothing but multiplkation by a fixed :.calar.
8. Prove Theorem 4.1.4.
9. True or false ?
(a) There exists a linear transformation T: V1 ➔ V, sue.ti that
T(O, 0) = (1, 0, 0, O).
(b)" Scalar multiplication is the only linear transfonqation from Y1
to Y1•
(c) T : fJ -,. fJ defined as T(p(x)) == xp(x} is not a linear transfor-
matio:i.
\ \4 I LlNEAll TllANS'fOI.NATlONS
(d) T: V1 -J-'1 defined by T(J, J) = (I, 0, OJ, T(2, JJ -= (0, I, 0),
T(O. I) =
(0, 0, I) is not linear.
fe) Rotation cf coordinates in V1 defined as (x, y) i---+ (x', y'),
wher~ x' = .-c cos O + y sin 8, y' = -x sin 8 + y cos 0, is a
linear transformation.
(f) Let R+ be the vector space proved in Problem 2 of Problem Set
3.1. l(t T: V3 - R... be a linear map. Then T(0, 0, 0) = I.
(g) Linear transformations cannot be defined from the real vector
space C to th!= complex vector space C.
(h) Let T: U ~ Y be a map such that T(Ou) ::/=- Ov. Then Tis not
linear.
•
4.2 RANGE AND KERNEL OF A LINEAR MAP
With reference to a linear map T : U ➔ V, two sets are important. One
is the range of T, denoted by R(T) and already defined (cf Definition
1.4.4) for any function T as the ::iet of all T-images. The other is N(T),
the kernel of T, defined :l!, follows.
4.2.1 Definition Let T: U ,. V he a linear map. The kernel (null space)
of Tis the set
N(T) = {u E U l T(u) =
Or}.
lt' is. also denoted as kerT.
In other words, N(T) is the ,;et of all those elements in U that are
mapped by T mto the zero of V. Note that this i'l nothing but the
T-pre-image of Ov (cf Defimtion I .4.2).
As illustration, let us find R(T) and N(T) for each of the linear trans-
formations T defined in Examples 4.1, 4.2, and 4.4-4.8.
Example 4.11 In Example 4. I we have T: Va -• V3 defined by T(xi, x 2 , x3 )
= (x1, X:, 0).
Here R(T) is the set of all elements of the form (x" x 1 , 0), which is
nothing but the x,x1-plane in V3 • This also says that T is not onto
(cf Definition 1.4.5).
To determine N(T), the kernel of T, we want all those vectors
(x11 x., x 3) for which T(xi, x2 , x3 ) = 0. This means (Xi, x 1 , 0) = ((J, (\ 0).
So x1 • .-c1 = 0. In fi.ct, any element of the form (0, 0, x3 ) would be
mapped by T int.> (0, 0, 0). No other element would be so mapped.
Therefore, N(T) is the set of all elements of the form (0, 0, x3), which is
nothing but the Xa•axis in J/3 •
Example 4 12 In Example 4.2 T: V1 - V1 is defined by T(x1, x1, x1 ) ==
(X1 - X1, Xi +
x.,).
. In this case R(T) consists of vecton of the form (x1 - x1 , x 1 + x1). We
waat to determine the vectors of Jo', that are of Chis form. For this, take
 4.2 RA.NOi! AND KERNEL Of A LINFAR MAP / 115
a vector (a, h) in Y1 and solve the equation
<xi - X1, X1 + X 3) = (Q, b) .
This means x 1 - x 8 = a and x 1 + x8 = b. Solving these. we get
x 1 = x1 - a, x3 = b - x1 •
Hence, T(x 1, x 1 - a, b - x 1) = (a, b). This shows that every vector
(a, b) of Y1 is in R(T). In other words, R(T) = V~. So this is an onto
map.
To determine the kernel, we solve the equation T(xi, x2 , x 8 ) = (x1 -
Xz, X1 + Xs) = (0, 0). Thi" gives x 1 === Xz -= ·- x 3, i.e. all vectors of the
form (xi, X1t -x1) will be mapped into zero. So
N(T) = {-A 1 (1, I, -1) I x 1 any scalar}= [(I, I, -1)].
This is the subspace of V3 generated hy (I, I, - J).
Example 4./3 For the zero map (set' Ex.1mplc 4.4) the range is fOy} and
the kernel is U. This is clearly not nn omo map.
Example 4.14 For the identity map Iu (,;ee Example 4.5) the kernel is
{Ou}. The range is U. so the map is onto.
Example 4.15 In Example 4.6 T: V2 - ~ V2 • is defined by T(x 1 , x~)
= (xi, -"\'2). Here R(T) :.:a V2 and N(T) = t(O, 01}. Thi!. is an onto map.
Example 4.16 In Example 4. 7 n: '(7 <1>fo, h) 'r' (a• b) j,; defined by
'.>,

D(/) ~= /'. In thi,; case R(D} -.:: '( (a,/,), ,;ince every continuous function
K on (a, b) pos!tesses an antideri,ative and hence D is an onto map.
N(D) is the set of all cunstant functions m 'f, lll(a, h).
Example 4.17 In Example 4.8 /J : f, (n' b) - , R ii defined by J(/)
= J'a' f(x)dx. Here the range is the whole of R, since every real number can
be obtained as the algebraic area under some curve )' =- f(x) from a to b.
Therefore, it is an onto map. The kernel is the set of all those functic.ns
'/' for which the area under the curve}' = f(x) from a to b is zero. Jt is
difficult to say anything more than this about the kernel.
We shall now check whether the linear tram,formations discussed in
Examples 4.11 to 4.17 are one-one lcf I .4. 7 for the definition of one-one).
In Example 4.1 l N(n is the x3 ~axis. So all points on the Xa·axis go
into (I\ 0, 0). So this map is not one-one.
In Example 4.12 N<n =[(I, I, -1)]. So, many poiats go into
(0, 0, 0). This again means r is not one-one.
[n Example 4.13 N(T) = U. So the zero map is not one-one. kcause
all elements go into the zero of V.
In -Example 4.14 the identity map is one-one, because, if x =I= y. then
certaialy /(x) ef:. l(y). It may be noted that Nm i"' the zero subspace of
u,
 116 / UNbAlt 1KANSFOlUIATlONS

In Example 4. J5 the linear map T is one-one, because if (xi, x,J

Observe that in
# U'.1•J'a), then (.-c1, -x1 ) is also not equal to (Yu -y1 ).
this case also N(TJ is the zero subspace of Vs,
In Example 4.16, since different functions (say those that differ by
a constant) have the same derivative, the map Dis not one one. Observe
that N(T) is a nontrivial subspace in this case. In Example 4. 17 the linear
map T i~ not one-one. (Why ?)
Summari~ing these observation'i, 1t appeh"S that T is one-one when
N(T) is the zero subspace and conversely. This is, in fact, true as a gene-
ral statement, as is borne out by the following theorem which gives, in ad·
d1tion, more information about R(T) and N(n
4.2.2 'l'heorem Let T: U -+ V be a linear map. Then
(a) R(T) is a sz,bspace of V
(b) N(T) b a subspace of U.
(c) T if 011e-011e iff N(n is the zero sub.space, {Ou}, 1,J U.
(d) If (111, u2 , ... , u,.J = U, then R(TJ =- [T(u1), T(u:), .. , T(un)].
(e) If U is ftmte-di'!'ensional, then dmi R(T) < dim U.
Proof: (a) Let v1 , v1 € R(T). Then there exist vectors u 11 u1. in U
such that T{u1) = l'1 and T(u 8 ) = v2 • So
1•1 + vl = T(u1) 1· T{u1,) --= T(u1 -t u1,) ,
since T is linear. But u1 1 U2 E U, i,,incc U is a ,ector ,;race. Hence,
v1 + v1 is the image of an element of U. Su v1 + v1 € R(TJ. In the same
way, .:zv1 = ac.T(111) =- T('Xll1), 'iince Ti, linear. But ixu1 € U, because r; is
a vector space. Hence, :1v1 E R(T). Thus, R<n is a sub'ipace of V.
(b) Let u1• uJ € NlT), Then T(u1) = Ov and T(u~) =- 0,-. bec.1u~e thiii
is precisely the meaning of their being in N(T). Now
T(u 1 + U.1) = T(u1) t- T(ua), since Tis linear
=-= Ov -I Ov =
Ov ,
wbi(:h 1,how,; that u1 ~- 111 E N(n. Similarly, for all 'icalar,; rx, we ha,·e
T(ixu1) =
«T(u1), since Tis linear
= «Ov = Oy (Theorem 3.1. 7)
which shows that 1111J E N(TJ. Thus, N(T) is a subspace of U.
(c) Suppose T 1s one·one. Then T(u) = T(11) implies u v. If =
e: = =
u N(T), then T(u) = Ov T(Ou). Therefore, u Ou. This means no
nonzero veaor u of U can belong to N(T). Since Ou in any case belongs to
N(T) (why ?), it follows that N(T) contains onJy Ou and nothins else.
Hence, N(T) is the zero subspace of U. •
Conversely, suppose N(T) - {Ou}. Then. to prove that Tis one-one,
we have to prove that T(u) - 7'(,) implies u - ,. Suppose T(u) - T(vJ.
Then
T(u - ,) • 7\U} - T(P) - o., .
 4.2 RANOB AND ICBllNBL OF A UNIAII MAP/ 117

Sou - , e:N(T) = {Ov}. Sou - , = Ou. i.e. u == ,. This proves that T

is onN>ne.
(dJ Let [Ui, u1, ... , 11,.) = U. Then each vector u can be expressed as
a linear combination of vectors Ui, Us, •.. , u,.. The vectors T(u1),
T(u1), ... , T(un) are in R(T). So, obviously, [T(u1 ), T(11a), ... , T(.u.)1 C R(T).
Let, e: R(T). Then there exis.ts a vector u e: U :,ud..1 that T(u) -= v. Since
u e: U = [u., "1, ... , u,J, we have
u = at1u1 -1- at1 u1 + ... + «,.Un •
Therefore, v = T(u) = T(«1U1 + GL11U1 + ... + «nun)
= «1 T(u1) + «1T(u1) + ... + «nT(un) •
So v E (T(u1), ••• , T(un)J. This proves that
R(n = [T(u1), T(111), •••• T(un)] •
(e) The proof is left to the reader. I
We shall conclude this article with a definition.
4.2.3 Definition Let T : U ➔ Y be a linear map. Then
{a) If R(T) is finite-dimensional, the dimension of R(T) is called the
rank of T and i6 denoted by r(T).
(b) lf N(T) is finite-dimensional, the dimension of N(T) is called the
nullity of T and is denoted by n(T).
We shall study these concepts at length in § 4.3.

Problem Set 4.2

1. Determine the range of the following linear transformations. Also
find the rank of T, where it exists.
(a) T: Y1 -+- Ya defined by T(x 1, x 1) = (x1 + x 2, Xi)
(b) T: Y1 ➔ Y3 defined by T(x10 x 1) == (xi, X 1 + x 1 , x 1 )
(c) T: Y1 ➔ Ya defined by T(x1, x1 , x3) = (lx1 x1 x1 , + +
x1 -lx1, x1)
(d) T: Y1 ➔ Y1 defined by T(x1 , '"•• x 1 ) = (x1 , Xa, X1)
(e) T : V, -+- Y1 defined by T(x11 x1 , x1 , x,) = (x1 - x,, x1 +x 3,
x1 -xJ
(f) T : Y1 - Y, defined by T(x1, x 1 , x 1) = (xi, x 1 + x1 , x 1 + x,
+ Xao x1)
(g) T : Y, ➔ Y, defined by T(xi, x1 , x 1, x.) = (3x1 + 2x1, x 1 - x.,
ix1 - x,. x1)
(h) T : 9' -+ fJ defined by T(pXx) = xp(x)
(i) T: fJ -+ 9' defined by T(p)(x) = xp'(x)
(j) T: 9' ➔ 9' defined by T(p)(x) = p•(x) - 2p(x)
(k) T : ~ l 0, 1) ➔(jj' (0, 1J defined by T(/)lx) = /(~ sin x
(I) T: ~IU(0, 1) ➔ ct (0, I) dotined by 7'(/)(x) - r(x}e".
2. Determine the kernel of the linear transformations of Problem 1•
(a){I). Also find the nullity or T, where it exists.
J J8 / UNBAR TKANSFC>JIMATJONS

3. Let T : Y - W be a linear map and U a subspace of Y. Define

nu) = {w E W I w = T(u) for some u E U}. Then prove that
T(U) is a subspace of W.
4. Let T: V - W be a linear map and W1 a subspace of W. Then
prove that the set {11 E VI T(v) E W1} is a subspace of Y.
5. Let at11, at11, ••• , at11,. Gtzi, ••• , ata,11 ••• , at111, ••• , atJI" be any pn fixed
scalars. Let T: V,. - Y11 be a linear map defined by
T(e,) = (at1., Gtsu ••• , cx 11,), i = I, 2, ... , n.
Then prove that
(a) T is not one-one if p < n
(b) T 1s onto when p = n and (ctu, ••• , cx 111), («m ••• , «111), ••• , (cx111,
••• , atJIJI) are u.
6. Find a linear transformation T: V8 - ► Y1 such that the set of all
vectors ('.t11 x 1 , Xa) satisfying the equation 4x1 - 3x2 + Xa = 0 is the
kernel of T.
7. Find a linear transformation T: V3 ➔ V8 such that the set of all
vectors (x11 xi, x8) satisfying the equation 4.~1 - 3x2 + x 8 = 0 is
the range of T.
8 Pick out the maps in Problem 1 that are
(a) one-one (b) onto (c) one-one and ,;,nto.
9. True or false ?
(a) Every constant map from one vector sp&.ce to another is both
one-one and onto.
(b) A hnear transformation T: Y1 - V, defined as T(l, I)
= (I, 0, 0, 0) and T( I, 2) = (2, 0, 0, 0) is one-one.
(c) Every linear transformation from V:i to itself is onto.
(d) No linear transformation from Y2 to Va is onto.
(e) There exist one-one linear transformations from Va to Yz.
(0 If T : U - V is a linear transformation and 110 E Y, then the
T-pre-image of 110 is a subspace of U.

4.3 RANK AND NULLITY

A careful scrutiny of Table 4.1 shows that the image of a linearly
independent set by a linear map need not be LI. The following theorem
throws light on this situation. Essentially, it says that a one-one linear
map will preserve linear independence, whereas under any linear map the
set of pre-images of a linearly independent set of vectors is LI.
4.3.1 neon• Let T : U ➔ Y be a linear map. Then
(a) 1/ T Is one-one and u1, "•• ••• , u,. are linearly independent vectors
of U, tl1en T(uJ, T(u.), •..• T(u,.) are LI.
(bJ If 111, ,,, ••• , •• are linearly lndependtnt Hctors of R(T) and
 4.3 RANK AND NUW'n / 119

u1, "•• ... , u. are vectors of U sue/, that T(u1) = v1 , T(u1) = "•• ... ,
T(u.) = "•• then U.1, u., ... , 11. are LI. •
Proof: (a) Let T be one-one and llu 111, ••• , u,. be linearly indepen-
dent vectors in U. To prove that T(uJ, T(u1), ••• , T(u11) are LI, we
assume
«1T(u1) + «1 T(u1) + ... + «.T(u..) = 0
or T(«1u1 + «1u1 + ... + «.u,.) = 0,
since Tis linear. So «1111 + «.au. + ... + «..11. = 0, since T is one-one.
But u1 , "•• ••• , u.. are u. So «1 = 0 = «1 = ... = «11•
Thus, «1Tlu1) + «1 T\u1) + ... + «..T(u.) = 0 implies that each oi:, is
zero. Hence, T(uJ, T(11.), .•• , T(u.) are LI.
(b) Let u1, 111, ••. ,"•and v1 , ••• ... , v.. be as stated in Theorem 4.3.1.
To prove that 1110 111, ••• , "• are LI, suppose
«1111 + «1U1 + ... + otnU11 = 0 .
Since T is linear, we have
= +
Ov = T(Ou) T(«1u1 «1111 + ... + «,.u.)
or «1 T(111 ) + oi:1 T(u1) + ... + x.T(r,,.) = Ov
or «1v1 + «1v1 + ... + cz11v,. = Ov .
But ".1• ,·1 , ••• , v11 are LI. Therefore, at1 = 0 = at1 = ... = at,..
Thus, «1111 + r;U. + ... + «.u,. = 0 implies that «1 = 0 = a, = ...
= «,.. Hence, 111, 111, ••• , "" are LI. I
Example 4.18 Prove that the linear map T: Y1 - Y1 defined by T(e1)
= e1 - e1 , T(e1) = 2e1 + e1 , T{e1 ) = e1 + e1 + e1 is neither one-one nor
onto.
Since [eh e1, eJ = V1 , by Theorem 4.2.2 (d),
R(n = [T(e1), T(e1 ), T(e8)] = [e1 - e2 , 2e1 + e3 , e, + e1 + e8]
= (e1 - e1 , 2e2 + e1] ,
because e 1 + e1 + e1 is a linear combination of e1 - e1 and 2e, +
e1 •
Now we see that e1 -- e1 and 2e1 + e1 are LI. So dim R\T) = 2.
Therefore, R(T) is a proper subset of Y1 • Hence, T is not onto.
To prove that T is not one-one, we check N(T). N(7) consists of
those vectors (x1, Xu x1 ) in Y1 for which
T(xi, Xe, x8) = 0
or T(x1e1 + x1e1 + x1e1 ) = 0
or XiT(ei) + x1 T(e1) + ..c1 T(e1 ) = 0,
because Tis linear. Thus,
(X1 + Xa, -Xi + 2x1 + X1, X1 + X1) = (0, 0, 0) ,
i.e. x1 + x1 = O, x1 + x1 = 0, and -x1 + 2x1 + x1 == 0. · Solving these,
We get X1 = Xa == -Xa, nerefore,
N(T) = {(x 1, Xi, -xJ I Xi an arbitrary scalar} = [(I, I, -1)) .
Hence, by Theorem 4.2.2, T it not one-one.
 120 / UNBAR TRANSFORMATIONS

In ~pie 4.18 the linearly independent vectors ei, '•• and e, span the
domain space Ya, but lheir images e1 - e1 , 2e1 + ea. and e1 e, + +
e, are
LD and dim R(T) = 2.
Thus, we find that the effect of the linear map T on U = V, is to
shrink V1 to a 2-dimensional subspace R(T) of V = V1 • What happens
to the remaining 'one' dimension ?
Observe, in this case, that the kernel of T is [(I, I, -1)) and so
dim N(T) = I. Thus, it appears that in this example
dim R(T) + dim N(T) = dim V8 •
This is not an accident. A general result of the same kind is true for
all linear maps whose domain space is finite-dimensi.onal. We shall now
record this as a major theorem.
4.3.2 Theorem (Rank-Nullity Theorem) Let T: U - V be a linear map
and U a.finite-dimensional vector space. Then
dim R(T) +
dim N(T) == dim U. (IJ
In other words,
r(T) + n{T) = dim U (2)
or rank + nullity == dimension of the domain space.
Proof: N(T) is a subspace of a finite-dimensional vector space U.
Therefore, N(T) is itself a finite-dimensional vector space. Let dim
=
N(T) n(n n and dim U p (p ~ n). Let B {u1, u. •... , u.} be a
= = =
basis for N(T). Since 11, E N{T), T(11,) = 0 for each i == l, 2, ... , n. B
is u in N(T) and therefore in U. Extend this linearly independent set of
U to a basis for U. Let B1 == {111, 111 , ••• , 11., ll11+u ••• , 1111} be a basis for
u.
Consider the set
A = {T{Un+J), T(11n+t), ••• , T(u11)}.
We shall now prove that A. is a basis fol' R(T). Observe that, if this is
proved. the proof of the theorem is over; for. this means
dim R(T) = p - n = dim U - dim N(T) ,
which is the samo as Equation (1).
It is therefore enough to prove
(i) (A.) - R(T), and
(ii) A. is LI.
To prove (i) we proceed as follows: Since [B1] == U, it follows from
Theorem 4.2.2 (d) that R(T) = [7\uJ, T(u1), ••• , T(1111) 1 T(u..+1), ••• , T(u.)].
But T(u,) = Ofor I - 1, 2, ... , n. Hence,
R(T) = (7'(11,a+1), T(u..+a), ••. , T(1111)] •
To prove (ii), consider
«n+1TCu.+1> + .. . +«.71:u.,) = o • (3)
Using the fact that Tis linear, we 1et
TCc.+-1"-+ + .... + c.,u.) == o,
 4.3 aANE AND NULLITY / l 21
which means that «11+11111+1 + ... + «,,u,, E N(T). Therefore. «n+1un+1
+ ... + «,.M,, is a unique linear combination of the basis B for N(T).
Thus,
«n+1Un+1 + •, · + «,,U,, = ~1U1 + ••• + ~nlln ,
i e. ~1111 + ••• + ~nUn - «n+1U +1 - ••• - «,,Ur, = 0 .
11

B1 being a basis for U is LI. Therefore.

~1 = ~. = ... = ~n= «n+1 == •·• = GtJI = 0.
Thus. Equation (3) implies «n+i = ... = «,, = 0. Hence, A is LI. I
Example 4.19 Let T: v, -~ V be a linear map defined by
3 T(e1 )
= (1, 1, I), T(e1) = (1, -1, I), T(e1 ) = (1, 0, 0), T(e,) = (1, 0, 1). Then
verify that r(T) + n(T) = dim U ( = V,) = 4.
We know that R(T) = [(I, 1, 1), (I, --1, I), (I, 0, 0), (1, 0, I)).
(I, 1, I), (1, -1, I), (I, o. 0), and (I, 0, I) are LD, because a set of four
vectors of V3 (dim V3 = 3) is always LO. We find that
(I, 0, 1) = l(I, 1, I) t 1(1, -1, I)+ 0(1, 0, 0).
Hence, we can discard the vector (1, 0, I), &o that
R(T) = [(J, I, 1), (I, -J, 1), (J, C\, 0)).
To check whether (It. I, I), (1, -1, 1), and (1, 0, 0) are Lr, we suppose
at 1(1, I, I) + «1 (1, -1, l) + oc3(1, 0, 0) = 0--=- (0, 0, 0)
or (« 1 + «11 + «3 , «1 -- «~, oc1 + oc1) = (O. r, O) .
Solving this, \\e get at1 = 0 = «1 = at 3• Hence, (1, 1, I), (I, -1, 1), and
(I, O, 0) are LI and dim R(T) = r(T) = 3,
Now to find N(T). we suppose that T(u) = 0 = (0, 0, 0). If
u = (x1 , x1 , x3, xJ = x1e1 + x11e1 + x3e3 + x,e, ,
then T(x1, x 1, X 3 , x,) = T(x1e1 + x.,e,. + x 3e8 + x,e,) = (0, 0, 0)
or (x1 + x~ + Xa + x,, Xi - x 8 , Xi + x 1 + x,) = (0, 0, 0) .
Solving this. we get Xi = x 1 == -x.f2. x 3 = 0. So N(T) contains the
vectors of the form (xJ, x 1, 0, -2x1), i.e. N(T) = [(I, I, O, -2)). So
n(T) = dim N(T) == 1. Hence. r(T) + n(T) = 3 + I = 4, and the
theorem is verified.

Problem Set 4.3

1. Let U be·a vector space of dimension n and T: U ➔ V be a linear
and onto map. Then prove that T is one-one iff dim V == n.
2. If T : U ➔ V is a linear map, where U is finite-dimensional, prove
that
(a) n(T) < dim U
(b) r(T) < min (dim U, dim V).
3. Let Z be a subspace of a finite-dimensional vector space U, and V a
finite-dimensional vector space. Then prove that Z will be the
kemel of a linear map T: U -+ JI iff dim Z >
dim U - dim V.
 122 / UNBAR TRANSl'OIHATIONS

4. Prove Theorem 4.3.2 by the following method : Assume a basis

(vi, "•• ... , ,,} for R(T) and Iii, u., ... , llr in U such that T(u,) = .,,,
I= J, 2, ... , ,. Assume a basis {w1 , w,, ... , w,.} for N(T). 1hen
prove that {u1, ... , Ur, Wu ... , w,.} is a basis for U.
S. True or false?
(a) No linear transformation from v, to V1 cari be one-one.
(b) If T: V, ➔ V1 is linear and one-one, then it is onto.
(c) Let T: V1 -,. Y, be a linear map. Then R(T) can be a 4-dimen-
sional subspace of V,.
(d) Let T: U -+ V be a linear map (U and V are finite-dimen-
sional vector spaces). If Tis one-one, then dim U < dim V.
(e) Polynomial functions p of degree less than or equal to 3 such
that (. p(:A.)dx = 0 form a 3-dimensional subspace of 9'1 •

4.4 INVERSE OF A LINEAR TRANSFORMATION

Linear transformations that are both one-one and onto play an

important role. We give them a special name in the following definition.
4.,U DeBnltloa A linear map T : U ➔ V is said to be nonsingular if it is
one-one and onto. Such a map is also called an isomorphism.
We know that any function has an inverse i.ff it is one-one and onto
(cf Definition 1.7.8). Hence, we have the following fact.
4.4.2 Fact_ A linear transformation is nonsingular i.ffit has an inverse.
Let us illustrate this by the following cumples.
Example 4.20 Consider the map T: 9'1 ➔ V8 defined by T(«o + «1x
+ «ax') = («0, «1, «1). Clearly, T is a linear map (check I). This is
onto, because given a vector (P1, '•• ~.) in V, we can get a polynomial
~ 1 + ~ + ~• of which ~ 1, ~•• ~) is the T-image.
Further, T is one-one, because if P1 = 0 = ~ = Pa, then the poly-
nomial (i1 + P.x + Par also reduces to the zero polynomial of 9'1• Thus,
Tis an isomorphism. It is easily seen that r-1 : V1 ➔ r,, is defined as
7'"l(ati, CIC■,«.) = «1 + y + atar •
Example 4.21 In Example 4.IS we bave T: V1 ➔ V1 defined by T(x10 X1)
""' (x10 -xJ. We have already seen that Tis one-one and onto.
To calculate 7'""1(y10 y1), we have to find the element that maps into
(y1, y 1) by T. The answer is (11, -11), because
T(yi, -yJ = <Y1, - (-,Y■)) = <Yu YJ ·
Thus, T-1: V1 ➔ V1 is defined by 7'""1 CY10 YJ == (y1, -yJ.
~ 4.22 We have already seen that the identity map Iu : U - U
delned by /u(u) .. u tor all II e: U is one-one and onto. So the invene
 4.4 INVLRSE OP A LINEAR ffANSPOllMATION / 123

of J, i.e. Ji•, exists as a map from U to U. Obviously, 1;;1 (u) = u for

all u E u. Thus, Ju 1 = Ju.
In Examples 4.20-4.22 we can check that r-1 is linear. In faet, this is
in general true as proved in the following theorem.
4.4.3 Theorem Let T: U - V be a nonsingular linear map. Then
r-1 : V - U is a linear, one-one, and onto- map.
Proof: To prove the linearity of r-1, let v., v1 E V. Let
r-1(v1) = u1 and T"" 1(v1) = u1 • Since T is one-one and onto, u1 and "•
e:mt uniquely. Thus, v1 = T(111) and v1 = T(u1). So
v1 +v =
1 T(U1) + T(u.) = T(u + u 1 1) ,

since Tis linear. Therefore,

r- 1(v. + V1) = U1 + "• = T-1(1•1) + T-1(v.) ,
Again, ar.v1 = ar.T(u1) =T(ar.u1). So
7-•c«v.) = «u. = «T-1(v1) •
Hence, r- 1 1s linear.
r- 1 is onto because, if u e: U, then T(u) = v belongs to V, and
r-1cv> = u.
The rest of the proof is left to the reader. I
The linear map T: V3 - V3 , defined in Example 4.18, does not have
an inverse, because it is neither one-one nor onto.
A linear transformation T: U - V has an inverse, if the following two
properties hold :
(i) T is one-one.
(ii) Tis onto.
Jr or.e or both of the properties fail to exi~t, then r-1 does not exist.
In order to check whether Tis one-one, we have to find N(T). If N(T)
is the zero subspace, then Tis one-one, otherwise it is not. The second
property, namely, T is onto, holds ijf R(T) = V. Tllis involves the
determination of R(T), the range space of T.
4.4.4 Remark Instead of calculating the kernel, if we can somehow find
(perhaps by guesswork) more ,ban one vector in u. which maps into
the same vector in V by T, then this would be enough to prove that
r-1 does not exist. In the case of Example 4J8 T(l, I, 0)
= (1, 1, 1) and T(2, 2, -1) = (1, 1, 1).
Juat from this we could have concluded that T does not have an
inverse. However, the method tJaat involves finding the kernel is re-
commended to the reader, since it. is applicable in most ases.
We shall now consider two more examples in which r-.1 does not mt,
because in one case T is one-one but not onto, and in the other case T is
onto but not one-one.
124 / UNIWl TRANSFORMATIONS

Example 4.23 Let Ube the set of all infinite sequences {x1, x1, ... , Xn, ... }
of real numbers. Define the operations of addition and scalar multipli-
cation coordinatewise as in Y., i.e. if
X == {Xi, X1, ••. , X,., ... }
and Y == {.Yi, Y1, ... , Yn, ... } ,
then X + ,Y == {x1 +
,Yi, x. + Y1, .. , Xn + .Yn, ••• }
and
The set U with these two operation, becomes a real vector space (check!).
Note that the sequence {0, 0, ...,o, ... } is the zero of U and {-Xi,
-x1 , ... , -x,u ... } is the negative of x.
Let T : U -+ U be defined by
T(x) = T({xi, x~, ... , x., ... })
= {Xa, X3, ... , Xn, ... } ,
It is easy to check that T is linear. Here R(T) = U. For, take the
sequence {.Yi, y 2 , ••• , y,., ... }. Its pre-image by T can be any sequence of
the form {z, ,Yi, .Y2, ... , .Yn, ... }, where z ran be a real number. In parti-
cular, z can be zero. Hence, T 1s onto.
But T is not one-one, because all the &equences {z, .Yi, y 1, ... , y.., ... }
map into {y1 , y 1 , ... , y,., ... }. Further, N(T) is the set of all sequences of
the form {z, 0, 0, ... , 0, ... }.
Thus, though T 1s onto it does not have an inverse, since T is not
one-one.
Example 4. 24 Let U be the vector space of Example 4.23. And let
T : U - U be defined by
T({Xi, x1, ... , X,., .. }) = {0, Xi, Xa, ... , Xn, ... } •

Obviously, Tis linear (check!). Now Tis one-one, for N(T) is the set of
all sequences of the form {O, 0, 0, ... , 0, ... }. There is only one such
sequence, namely, the zero element of the space U. So N(T) = {Ou}.
But Tis not onto, because the element {J, 1, 1, ... , 1, ... } has no pre-
r
image in U. So R(T) '=I: U. Hence, does not have an inverse.
In Examples 4.23 and 4.24 we prod11ced situations where just one of the
two conditions, namely, (i) T is one-one, (ii) T is onto, holds and the other
does not hold. Further, note that in these examples the space U involved
is not finite-dimensional. (Why?) When U is finite-dimensional, we may
not be able to produce such an example, because of the following
theorem, which essentially says that if T : U ➔ Y is a linear map and
dim U == dim Y, then the two conditions (i) T is one-one and (ii) T is onto
are implications of each other.
4.4.5 Theorem // U and Y are finl/e-dimen,lonal vector space, of the same
dimension, then a linear map T : U ➔ Y is one-one iff It l1 onto.
 4.4 INVDSB OF A LINEAR TRANSPOlllfATION / 125

Proof: T is one-one • N(T) = {O} (Theorem 4.2.2)

•n(T) = 0 (Definition 4.2.3)
•r(T) =
dim U =dim Y (Theorem 4.3.2)
•R(T) = y (Theorem 3.6.13)
co Tis onto. I
Now we shall take up an example where the inverse exists and we
proceed to calculate it. Recall that a linear map is completely determined
as soon as its values on the elements of a basis are specified. Therefore,
in order to determine the linear map r-1, it is enough to determine the
values of r-1 on the elements of a basis.
Example 4.25 Prove that the linear map T : Y:, -+ Y8 defined by T(ea)
== e1 + e1 , T(e1) = e1 + e8 , T(e8 ) = e1 + e1 + e3 is nonsingular, and
find its inverse.
First, let us find the value of T at a general element u = (x11 x1, x1 ) :
T(xi, x1 , x1) = T(x1e1 +
x1e1 + x1e3)

If T(x1, x1 , x3) =,O, then

X1 + X3 = 0, X1 + X-, + X3 = 0, X1 + Xa = 0.

Solving these, we get x 1 = 0 = x1 = x3 • So N(T) = {0y1 } and hence T

is one-one. It follows from Theorem 4.4.5 that T is also onto. Hence, T
is nonsingular and T"1 exists.
Now we shall give·two methods to find the inverse, r-1, which is also a
linear, one•one, and onto map from V3 to V3,
Method I We have T(e1l = e1 + e2 , T(e8 ) = e2 + e3 , T(e8 ) = e1 + e.
+ e1• Therefore,
e1 = T,"1(e1 + e8) = T""1(e,) + T""1(e1)
t'9 = r-1(e. -f- ea) = r 1(e.) + r-1(eaJ
ea = r-1(e1 + ea + ei) ·
= r-l(e1) + r-1(e1) + 1l(ea) •
because r-1 is linear, one-one, and onto. Solving these three equations
for r-1(,.). r-•ce.), and r 1(e1), we get
T"1(e1) = ea - t1 =(0, -1, I)
:,-1(t1) = e1 + e1 - ea== (1, I, -1)
'.i"1(eJ
e1 - = e1 == (--], 0, I).
Now we extend r-1 linearly and obtain
r
T""1(Xi, X1, Xa) = 1(x1e1 + .X1e1 + .x_eJ
== x1 r-1(e1) + x,r-1(e1) + x 1T-l(e.J
- (x_ - .X., .X1 - 'X1, X1 - X. !f- x.) •
Method 2 Let T-1(.xi, x11 x,J - C,1, Y■, y1). Then
7lY1, 11, Ya) - (xi, Xa, x.)
 126 / LINEAR TRANSFOPMATIONS

or 7fy1e1 + y,111 + y1e1) = (x,, x,, x,)

or y1T(e1) + y1.1{e1) + y.,.1{,1 ) = (xi, x,, .t,)
or (y1+ y 3 1e1 I- (y1 + y 3 + y1 lea + <Ya + Ya)ea = (Xi, Xi, X3 )
« fY1 + Ya, Yi + Yt -I- Ya, Ya + Ya) -= (Xr, X:, Xa) •
This gives y 1 t- .V.1 -= x., Yi + YJ + Ya = X1, and Ya + Ya = x,. Solving
these, we get y 1 = xl - x,, Ye = x 5 - X11 and Ya """' Xi - x1 + x3 • So
r 1(x., X1~ X,1) = (X1 - X3, X 1 - Xi, Xi - X1 -f- Xa) •
Problem Set 4.4
I. Let R, S, and T be three linear maps from V3 to Vi. defined by the
values in Table 4.2. Determine which of them a.re nonsingular and
in each such case find· the inverse.
TABLE 4.2
- - - -------
Value at -1o

Linear maps
J.
R e, I eJ e1 - "• + f';i :e. ;- 4e.l

s e1 -- e_ e2 t'1 ,_ el - 7e3

T e1 - e,. + ea 3e1 - 5e1 3f'1 - 2e1

------ ----- - - -----
2. Show that each of the following maps is nonsingular and find its
inverse:
(a) T: Vi1 .,. V1 defined by T(x 1, xJ) = («1x 1, «1 x 1), where «1 and ~,
are both nonzero.
(b) T: Y3 -► Va define~ by T(xi, x!, x3) ~- (X1 + x 2 t X:i, x3 -I :c.,
X3).
(c) T: ~'1 - ► H'1 defit,cd by T(o 0 t 111x I- cxi.'t') -= (ao t «1) -t
(cx1 -I 2cx1)x + (, 0 t ct 1 1- 1«l)x'.
3. Let Ube the subc,;et {p E g, I p(O) =-· O} of V', Then rro,e that the
derivative D is a nonsingular hne11r map from U to 9', and the
integral (c!J(p))(x) fo p(x)dx is its inverse.
4. Let T : U ► V be a nonsingular linear tram,formation. Then prove
that (1- 1,- 1 --- T.
S. True or fab,e ? •
(a) Fvery linear map from V3 to V1 has an inverse.
(b) The inverse of a nonsingular linear map is nonsingular.
(c) A nonsingular linear map transforms linearly independent sels
into linearly independent sets,
 4,S CONSEQUENCES OF RANIMilULLITY THEOREM/ 127
td) A nonsingular linear m1p transforms linearly dependent sets
into linearly depertdent sets.
(e) Every translation of Y1 to Y1 has an inverst..
(f) Every translation of Y2 to Y1 is an isomorphism.
(g) Given tw_o vectors ui, u. e: U, there exists an isomorphism
T: U ➔ £1 such that T(u1) =- "i·

4.5 CONSEQUENCES OF RANK-NULLITY THEOR6t1

The results discu!\&cd in § 4.4 and certain allied facts are important for
further disc1ni;ions. We collect them tClgether in the followir,g theorem.
4.5.1 Theorem let T: U - ► V be a linear mop and dim U =-· dim V = p.
The? the following statemenu qr'! cqui,r.lent :
(a) Tis nonsingular (an irom,rpl,ism).
(h) Tis u11e-one.
Cc) T transforms /,nearly indepc·ndent substtJ of i· into linearly inde-
pendent subsets of V.
(d) T transforms e1·ery basis for U·inttJ a basis fur V.
(e) Ti., onto. •
<O ,en=- p.
(g) n(T) •- 0.
(h) • T I exMs.
Proof: (a) :- ( I,) hy Ddinition 1'.4.1.
(h) =-> (c) by Theorem 4.3.l(a).
(c) :- (d). Let T tran'lform hnearh independent subsets of U into.
line..irly independent i.ubst:ts of R(T). Now, let {u1, llz, .... u,,} be a b&sis
for U. Then 7'(u1 ), T(u:), ... , T(u,,I arc u by hypothesis. But dim Y
• f'. So {T(u1), T(ut), .. :. T(u,)} l!t a basii, for V by Theorem 3.6.7.
(d) -> (e). let {u1 , u2 • ... , u,,} be a basis for U. Then, by hypothesis,
{T(u,), T(u.), ... , T(up)} is a bai.is for V. This means, by Theorem 4.2.2 (d),
R(T) --- V. Hence, Tis onto.
(1') ~ (fJ. T ii, onhl --> R\T) :=-- V ~ r(T) -- p.
(I) , (g) by the Rank-NuJlily theorem.
(g) -> (h). n(T) --- 0 means NtT) - {Ou}, i.e. Tis one-one. So Tis
onto by Theorem 4 4. "· Hence, r- 1 exists.
(h) ~ (a) by Fact 4.4.2. I , •
We shall c~clude this article by exhibiting an important isomorphism
of linear algebra.
4.5.'1 De&aitlon • Two ,•ector spaces U and Y are said to t>e isomorphic if
u
there exists • an isomorphism from U to Y.. · Cl and 'JI are ise-
naorphic. then we write U m Y.
 /.28/.UN&Uf :nuNSPOIINATIONS

4.5.3 Tlleoreaa E,ery real (complex) ,,ctor IJJ«t of dimension p Is lso,

morphic to JI.,(~. ,
Proof: Let U be a real vector space or dimension p, B == {u1,
u1 , u.,} an orde~ basis for U, u an arbitrary element or U, and
... ,
(«1, «1, .,., «11) the coordinate vector of u relative to B. Consider the map
7 : U -+ Y., defined by
T(u) = («1, «,, ... • «.,) •
This map is ·easily seen to be linear; for, if
U = ll1U1 +
0,"9 + ,.. + 11,Pp
and v = ~1u1 +
~~u• + ... + ll11u11,
then + + + +
u v = («1 ~1)u, (111 fl1)U. + ... + («11 +
f!.,)1111 •
Hence. the coordinate vector of u + v relative to B is («1 +~ 1,
a, + ~••... , « 11 + ~.,).So
T(U +i•J = («1 +
~b «1 + ~I• •••• 1Xp+ ~p) '
But T(u) +T(v) = («11 «1 , ••• , «11) + (~i, ~..... , ~.) .
Therefore, T(u + v) = 7tu) + T(v) .
Similarly, T(«u) = 11T(u).
Further, T is• one-one, because T(u) = («1, «1, .... «,.) = Ov means
«1 = 0 = « = .. . = «
1 11, which implies u = Ou1 + OU. + .. . + OU 11
= Ou, i.e. N(n = {Ou}. Thus, by Theorem 4.5.1, Tis an isomorphism.
Hence, there exists an isomorphism from U to Y II and consequently
U =
Y.,. The proof of the other part is left to the reader. I
Example 4.26 In Example 4.20 we h1ve seen that T: f}1 ➔ Ya defined by
T(ix 0 +
ix 1x +
« 1r) = (« 0, «i, «1) is an isomorphism. Thus, f/1 Y1 • =
Theorem 4.5.3 says that not only f/ 1 but also any real vector space of
dimension 3 is isomorph:c to Ya,

Problem Set 4.5

1. Prove that ' = ' is an equivalence relation.
2. Let .A be the subspace of Y, defined by
A. = {(X1 , X1 , Xa, x,) I X1 = O}.
Prove, by exhibiting an isomorphism, that .A =
Y1 •
3. let B be the subspace of 9', defined by
B =- {p I p'(I) == O, p•(I) = O}.
Prove, by exhibiting an isomorphism, that B Ya.=
4. Prove, by exhibiting an isomorphism, that A: =
B, where ..4 and B
are subspaces of Problems 2 and 3.
S. True or falle ?
(a) The diff'ercntial operator D : g. -+ g _ 1 bu nullity zero.
(b) There exist two isomorphisms from £11 to Y1•
 46 THE SPACB L(U, Ji')/ 129

(c) In V1 all nontrivial subspaces are isomorphic.

(d) Rotation in V1 is an isomorphism.
(e) Let T: V ➔ V (Vis a finite-dimensionaJ vector space) be a linear
map. If R(T) n N(T) = {Ov}, then V = R(T) EB N(T).

4.6 THE SPACE L (U, V)

SUM OF TWO LINEAR MAPS
Let T : l/ -+- V and S : U ➔ V be two linear transformations. Con-
sider the map M : U ➔ V defined by
M(u) = S(u) + T(u) for all u E U.
We shall prove that Mis linear, that is,
M(u1 + u1) = M(i.1 1) + M(u2)
and M(«u 1) = «M(u1 ) ,
for all ui, u2 E U and all scc1lars « .
We have
M(u1 I- U2) = S(u1 + +
u2) T(u 1 u2)+ (definition of MJ
+ +
= (S(u1) S(u11)) + (T(u1) T(14)) ,
because S and Tare linear maps. On the other hand,
M(u1) + M(u2) = (S{u1) + T(u + {S{ui) + T(u
1)) 2))
(definition of M)
= (S(u1) + S(u2)) + (T(u1 ) + T(u1))
by associativity and commutativity of addition in V. Thus,
M(u1 + U:) = M(Ui) + M(u1) for all u10 u2 E U.
Again, M(or:U1) = St«u1) + T(or:u1) (definition of M)
+
= cr.(S(u1)) «(Tlu1)) (S, Tare linear)
= cr.(S(u1) + T(u1))
by properties of scalar multiplication in V. Thus, M(«u1) =--:: cr.M(u1) for
aJI u1 E U and all scalars «. This proves that M 1s a linear map.
The map M defined above is called the sum of S and T and ii. denoted
by S + T. Thus,
(S + T)(u) = M(u) = S(u) +
T(u) for all u € U. (I)
We have thus shown that the sum of two linear maps is linear.

SCALAR MULTIPLE OF A LINEAR MAP

Let S: U ➔ V be a linear map and« a given scalar. Note that U and
V are vector spaces over the same field of scalars, and « also belongs to
the same field. Consider the map P : U ➔ V defined by •
P(u) = «(S(u)) for all u E: U.
We sJaall prove thai Pis linear,
130 / LINEAR TRANSFORM.\TIONS

Let ui, uJ be two elements of U and }. be a scalar. Then

'
P(u1 -+ u1) = cx(S(u1 + u1 )) (definition of P)
= ac(S(u1) + Slu1)) {S is linear)
+
= 11(S(r11)) ac(S(u2}) (Vis a vector space)
= P(u1) + P(u1) (definition of P).
Again,
Pp u1) = ac(S(}.u1)) (definition of P) •
= ac{A(S(u1)J) (S is linear)
= }.(11t(S(u1))) • ( V is a vector space)
= A(P(u1)) for each sca~ar i (defimtion of I'),
This prov~ that P is a linear map.
The map P defined above is called the s<"alar multiple of S by « and is
denoted by 11S. Thus,
(«S)(u) = P(u) = ac(S(u)) for all u E U. (2)
Thus, the scalar multiple of a linear map is linear.
Example 4.27 Let T: V8 ➔ v1 and S : Vd ➔ V1 be two linear maps
defined by
T(Xi, Xa, x 3) = (x1 - X1 , X 1 + x 3)
and S(x1, x 1 • x 3) = (2xi, x 1 - Xa) •
Then (S + TJ : Va ➔ V1 is given by
(S + T)(x1 ,.x1 , x.) = S(x1t x 1 , x 3) +
T(x 1, X2 , x 8)
= (x1 - x1 , x 1 + x 8 ) + (2xi, x 1 - x3)
= (3x1 - x 1 , 2x1) ;
and «S : V3 ➔ V1 is given by
(«S)(xi, x 2 , Xa) = «(S(xi, x 2 , x 3))
= CJ(X1 ·- X 1 , Xa + X3)
(11x1 - «x1, acx1 + «xa) •
=
Example 4.28 Let T : V3 ➔ V 8 and S : V1 ➔ Va be two linear maps
defined by
T(e1) = e1 + e1, T(e1) = e8, T(e8) = e1 - e8 ;
and S(e1 ) = e8 , S(e1 ) = 2e1 - e8 , S(e8 ) = 0.
Then S + T : V8 -+ V8 is given by
(S + T)(e1) = S(e1) + T(eJ = e1 + e2 + e8 ,
(S + TJ(e1 ) = S(e1) + T(e1) = le1 ,
(S + T)(e1 ) = S(e1 ) +
T(e8) = e 1 - e8 ;
and 2T : V1 ➔ V8 i,i given by .
(2T)(e1) == 2(T(eJ) 2e1 =
2e1 , +
(27')(e1 ) =
2(7{e.)) = le1 ,
(2T)(e.) = 2(T(e8)) = 2e1 - lea •
The set of all linear transformations from V to Y is denoted by L(U, Y),
Wo now have three ditreroat objccta belore ua (aee Fipre 4,2), namely,
 4 6 THI? SPACE L(U, V) / 131

(1) the vecror space U,

(ii) the vector space V, and
(iii) the set L(U, V), whose elements are linear maps T: U _.., V.
The sum of two hnear maps and the scalar multiple of a linear map of
the foregoing discu11,sion give us the operations of addition and scalar
multiplication in L( U, V). Our claim ,s that L( U, V) is a vector space for
this addition and scalar multiplication.
LCU,V)

IC

• II
II
S(u)+T(u)

FIGURE 4.2

Since the sum of two linear transformations is a linear transformation

and the scalar multiple of a linear tramformation is linear, we have only
to check axiom (VS3). First note that L(U, V) is a commutative group
for addition. The zero map O : U ➔ V, defined by O(u) = Ot for all
u E U, 1s in Ll U, J.') (see Example 4 4) and plays the role of identity for
addition in L(U, V). The map (-S): U ➔ V is defined by (-S)(u;
= - (S(u)) for all u e: U. It is easily seen that ( - S) E L(U, V) and
-S = (-l)S.
further, if Sand Tare two members of L(U, V) and «, ~ two scalars,
then
(i) «(S + T) = «S + «T,
(ii) (« + ti)S = «S + ~s.
(iii) •(~S) = («~)S = ~(«S),
(iv) lS == S.
We shall leave it to the reader to verify these axioms. Once they are
verified, we will have proved the following theorem,
.
4,Ci.1 Tbeorem The set L(U, Y) of all lintar trancformations from U lo Y
1og11hn with th, op,rat/0"8 of addition anti ,calar multiplication
deJi,,ed In Stat1mnt1 (i) and (ii) Is a ,ector 1pac1. •
4,6.2 Reauk It should be noted that L(U, Y) is a real vector space if
both U ad Y are real vootor spaces, and it is a complex ~cctor.apace
132 / UNEAll TRAN~R>llllATIONS

if both U and V are complex vector spaces.

We shall prove in Chapter S that if U and V are finite-dimensional
veqor spaces, then dim L(U, V) = dim U x dim V.

Problem Set 4.6

l. Let the linear maps T : V1 ➔ V1 and S : V1 ➔ V1 be defined by
T(xi, x 1) = (x1 + x 1, 0),
S(x11 x1) = (2x1, 3x1 + 4x1).
Determine the linear maps
(a) 2S + 3T (b) 3S - 71'.
2. Let the linear maps T : Y3 ➔ Ya and S : Ya ➔ Y3 be defined by
T(xi, x 1 , x 8) = (2x1 - 3x8, 4x1 + 6x2, x8)
S(e1) = e1 - e3 , S(e1 ) = ei, S(e3 ) = e1 + e1 + e3 .
Determine the linear maps
(a) S + T (b) 3S - 2T (c) «S
and find their values at (x1, Xa, x 3).
3. Let the linear maps R, S, and T be defined as in Problem I, § 4.4.
Then find the linear maps
(a) R + 2S (b) 2R + ST (c) S - T
(d) R + S + 2T (e) a.R + ~s + yT.
4. Prove that the set of all linea1 maps from Y2 to V8 which map the
vectors on the line x + y = I) onto the origin is a subspace of
L(V8 , V2).
S. Let Ube a subspace of a vector space V. Then prove that the set of
all lij)ear transformations from V to V that vanish on U is a subspace
of L(V, V).
6. Let T: V1 ➔ V1 be a nonzero linear transformation. Then prove
that L(V11 V1) = [T].
7. Determine two linear transformations T and S of rank 4 from V, to
V, such that
(a) r(T + S) = 3 (b) r(T - S) = 2
(c) r(T + 2S) -= I (d) r(T - S) = 0.
8. Prove Theorem 4.6.1.
9. If Sand T belong to L(U, U), then prove that So TE L(U, U).
10. True or false ?
(a) To prove «(S + T) = a.S + a.T, where S and T are linear maps
from U to V, it is enough to prove that S + Tis linear.
(b) L(U, R) is a proper subset of~JI..U}.
(c.1) If S, T : {I ➔ Y ar~ linear, then S - T is linear.
 4.7 COMPOSMON OP LINEAR MAPS/ 133
(d) For a linear map s. -S(u) = S(-u).
(e) Tu E L( U, Y).
(f) Rank of (S + T) = rank of S + rank of T.
(g) Rank of («S) =«rank of S.

4.7 COMPOSITION OF LINEAR MAPS

Let T : U ➔ V and S : V ➔ W be two linear maps. We know that the
composition S o T : U - W is defined by
So T(u) = S(T{u)) for all u E U (Definition l.7.2).
So T T S
Symbolically, U --➔ W = U ➔ V ➔ W.
This map is linear, because with the usual terminology
So T(u 1 + u2) = ~(T(u1 + U.1)) (definition of composition)
= S(T(u1) T(u2))+ (T is linear)
= S(T(u1') + S(T{uz)) (S 1s hnear)
= (S o T)(u1 ) + (So T)(u2) {definition of
composit10n)
for all u1o u2 E U.
Again, (S o T)(«u1) = S(T(«u1)) (definition of composition)
= S(«(T(u1))) (Tis hnear)
= ii(S(T(u1))) (Sis linear)
= «(So T)(u1) (definition of composition)
for all u1 E U and all scalars "·
Thus, the composition of two linear maps is a linear map.
Example 4.29 Let a hnear map T: Va ➔ V, be defined by
T(e1) = (1, 1, 0, 0), T(e1) = (~ -1, 1, 0), T(e8) = (0, -1, 1, 1),
where {e1, e1 , ea} is the standard basis for Va, and let a bnear map
S : V, ➔ V1 be defined by
S(/1) = (1, 0), S(f.) = (1, 1), S(/a) = (1, -1), S(/,) = (0, 1),
where {/1. /1 , /8,/,} is the standard basis for V,. Then the lineat map
S o T : V8 ➔ Ys is obtained as follows (see Figure 4.3) :
(So T)(et) = S(T(eJ) = S(l, 1, 0, 0) = S(/i + / 1)
== S(/1) + S(f.) = (1, 0) + (1, 1) == (2, 1),
(So T)(e1) = S(T(e1)) == S(l, -1, 1, 0) == S(fi - f., + /a)
== S(/J - S(f.) + 1(/i) '
= (1, 0) - (1, 1) + (I, -1) == (1, -2),
(So T)(e1) = S(T(e8)) =- S(O, -1, 1, 1) = S(-r/1 +Ji+ /J
== -S(/1) + S(/a) + S(/,)
.. -(1, 1) + (1, -1) + (0, 1) - (0, -1).
134 / LINEAR TRANllFOllMATIUNS

We shall hereafter ui,e ST for S o T and call it the product of S ond T,

rather than the compo11ition of S and T.
We know that if ST is defined. then TS need not be defined. Even if
both are defined, they need not be equal. Thus, the commutative law of
the product is not in general satisfied. The other laws of multiplication
are easily seen to hold.

FIGURB 4.3
4.7.1 neorem Let T., T1 be linear maps from U to V. Let S 11 S1 be
linear maps from V to W. Let P be a linear map from W to Z, wl,ere
z
U, V, W, and are vector spaces over the same field of scalars. Then
(a) S1(T1 + T1) = S1T1 + S1T1 .
(b) (S1 + s.)T1 = S1T1 + s.T1 •
(c) P(S1T1) = (PS1)T1 •
(d) («S1)T1 = «(S1T1) = Si(«T1), where « is a scalar.
(e) lvT1 = T1 and T 1 lv = T1 •
Proof: (a) T1 + T1 : U - V and S1 : V ➔ W are linear. The pro-
ducts S1(T1 + T1) and S1 T1, S1 T1 are defined. So both sides of (a) make
sense. Now, if u is a vector of U, then
(S1(T1 + T,))(u) == S1((T1 +
T1)(u)) (definition of product)
. == S1(T1(u) + T1(u)) (definition of addition
in L(U, J'))
== S1(T1(u)) + S1(7i(u)) (S1 is linear)
-= (S1T1)(u) + (SJTJ(u) (definition of product)
== (S1T1 + S,T,)(11) or
(deftnition additi.on
in L(U, JI)).
Hence,
s.c1i. + r,> - s,Ti + s,r, .
The proofs of tile remaining parts are Jef't to tbe reader. I
Note •~Y the diapam for part (e) of Theorem 4.7.1 :
 4 7 COMPOSlTION ClP UNPAR MAPS/ 135

1i. J., Ti
U--+ V-- ► V= l'-- ► V
Iv T1 Ti
U - .. C,-~V=-U-+V.
Let T: U _.,. V be a non~mgular hnear m.ip. 1 e T I\ linear, one-one.
an,;1 onto. Then we know that T-1 : V -• U exists and is lmear (Theorem
4.4.3). Further. TI 1 = /r and T- 1T -- Ju (ct§ 1.7). In fact, th,s
cbaracteru,e1,, nonsingular1ty a1,, shown in the followmg theorem
4,7.2 Theorem A. lmear map T: U - V ,s nonsingu/ur if/ there exl\ts a
linear map S : V ~ U .,uch tl,at TS = Iv and ST = lu,
In such a case S = T-1 anti T = s-1 •
Proof: Let T be nomingular. The existence of S, namely, T-1,
follows from the observation• made immediately before the theorem.
Conversely, let S and T exist with the properties stated, i e. TS= Iv
and ST= Ju.
Let u e: N(T). Then T(u) ~ 0 and consequently S(T(u)) =- O But
S(T(u)) = lu(u) = u. Therefore, u = O. So N('T) 1s the ztro subspace
of U. Hence, T is one-one.
Now let v E V. Then T(S(v}) -= (TS)(v) = /r,(v) = ,, Therefore,
there exists an element u = S(v) in U such that 7l.u) = v. Hence, T11
onto. Thus, T H nonsingular. Therefore, T-1 exists by Fact 44.2.
Further, using Theorem 4.7.1, we get
T-1 = T-1(/v) = T-1(TS) = (T- 1T)S = luS = S .
Similarly, T = S-1. I
Finally, we prove the following interesting theorem.
4.7.31beorem Let T: U-+ V and S: V-+ W be two linear mapr Then
(a) If S and Tare nonringular, then ST Is nonsingular lllld
(ST)-1 = T-1S-l.
(b) If ST Is one-one, then T Is one-OM.
(c) If ST is onto, then S 18 onto.
(d) If ST is nonsingular, then T Is one-one and S is onto.
(e) q U, Y, War, of the same finite dl~nslon and ST Is nonsingular,
then both S and Tare nonsln,ular.
Proof: (a) Since Sis nonsingular, S-1 is defined and SS-1 = Iw
and S-1S - Iv. '
Since Tia nonsingular, T-1 ia defined and TI'-'= Iv and T-1T = Iu.
By virtue of(c) and(e) of Theorem 4.7.1, we have
(STX.T-18-1) - S(7tJ'-IS-l)) .. S((TT-1)S-')
- S(I..s-1) • Ss-1 - 'fr· , /
136 / LINEAR 1RANSFOllMATIONS

Similarly,
(T-1s 1>csn = T-1cs-1csn> = r-1«s-1s,n
= T- 1(1vT) = T- 1T == Ju.
=
Hence, by Theorem 4.7.2, ST is nonsingular and (SD-1 T- 1s-1.
(b) Let u € N(T). Then T(u) = Ov. So S(T(u)) == Ow, i.e. (ST)(u)
== Ow. But ST is one-one. Therefore, u = Ou. i.e. N(T) = {Ou}. Thus,
T is one-one.
(c) Let w E W. Since ST is onto, there exists a vector u E U such
that (ST)(u) = w. Therefore, S(T(u)) = w. Hence, there exists a vector
v = T(u) € Y such that S(v) = w. Thus, S is onto.
The proofs of (d) and (e) are left lo the reader. I
Notations (a) The set of all linear operators on U is denoted by L(U).
(b) If Tis a linear transformation on U, then the composition 1T is also
denoted by TJ. Similarly, TL = T"- 1T for any positive integer k. By
convention, ro = I.

Problem Set 4. 7
1. Let T: Y3 - Ya be defined by T(x1, x 1, x 3) = (x1 -t Xa + x 3 , x 1)
l'nd S: V1 ➔ ¥:a be defined by S(x1, X2) ~ (x11, x 1). Then determine
ST.
2. Let S and T be as in Problem I,§ 4.6. Determine
(a) ST (b) TS (c) st (d) -rs.
3. Let S and T be as in Problem 2, § 4.6. Determine
(a) ST (b) TS (c) STS (d) TST.
4. Let R, S, T be as in Problem I, § 4.4. Determine
(a) ST (b) RT (c) RST (d) R(S + T) (e) r
(0 'PST. Also verify that R(S + T) = RS + RT.
S. Determine two linear transformations S and T on V1 such that ST
== 0 L(f'a) and TS =I= 0£(Va).
6. Let S and T be two linear maps on U such that ST == TS. Then
prove that
(a) (S + 1')1 = S2 + 2(ST) + T8
{b) (S +TY,= S" + nc1sn-1T + ... + nc.T-.
(Hint : Use induction.)
7. Let Y be a 1-dimensional vector space and S, T two linear maps on
Y. Then prove that SI' = TS.
8. Let T be a linear map on a I-dimensional vector space Y. Then
prove that ra
== •T for some fixed scalar •·
9. Find the ranae, kernel, rank, and nullity of the followins linear
 4.7 COMPOSlTlON Of Ll'NEU MA.PS / 137
maps, where R, S, Tare as in Problem 1 of Problem Set 4.4 :
(a) RS (b) RT (c) RST.
10. Let T be a linear map on a finite-dimensional vector space V. Then
prove that
(a) R(T) n N(T) = {O} iff T 1x = 0 ~ Tx ::, 0
(b) If r(TJ) = r(T), then R(T) n
N(T) = {O}.
1J. If a linear transformntion Ton V satisfies the condition T'- + I= T,
then prove that T-1 exi&ts.
12. Let T be a linear map on V:1 defined by T{e 1) = e1 , T(e1 ) = e11
T(ea) = e2• Then rrove that TJ =- T 1 •
13. Let T : U ·➔ V and S : V -► W be two linear maps. Then prove that
(a) If Tis onto, then r(ST) :;;: r(SJ
(b) If Sis one-one, then r(ST) = r(T).
J4. A linear transformation T on a vector space V 1& said to be idempo-
tent if T 2 = T. For example, the zero transformation and the
identity tran&formation are idempotent.
(a) Let S and T be two linear maps on V3 defined as
T(x1 , x 2 , xJ) = (0, x2 , Xs)
S(x1, Xz, x 3) - {x1, 0, O) .
Then prove that both S and T are idempotent.
{b) If S1 and s~ are idempotent on a vector space V, then find the
conditions under which S/;1 and S 1 + Si are idempotent.
(c) If S 1s idempotent on a vector space V, then I - S is alro
idempotent on V.
{d) Determine two idempotent transformations Sand Ton a vector
space V such that ST = TS = 0, but neither S =-- 0 nor T -= O.
(e) If T 1s an idempotent transformation on V, then pro\-e that
N(T) = R(I - T) and R(T) = N(I - T) .
I 5. A linear transformation T on a vector space Y is said to be nilpotent
on Y if T" = O for aome integer n > l, and the smallest such integer
•n• is called the degree of n;/potence of T.
(a) Prove that the differential operator D is nilpotent on £1' •· What
is the degree of nilpotence of D 'l
(b) Is D 1 + D nilpotent on £1',? If yes, find its degree or nilpo•
tence.
(c) Let T : Y1 -► V, be defined by
7lx1, X:, X1 , Xe) = (0, X1, X1, Xa) •
Is T nilpotent ?
138 / LINEAR TRANSPORMATIONS

(d) Let T : Ya ➔ V1 be defined by

T(e1) = 0, 7{e1) = e1 , T(eJ = 2e1 3e1 • +
Is T nilpotent ?
(e) If S and Tare nilpotent transformations on a vector space Y
and ST ,.._ TS, then prove that ST is nilpotent on Y. What is
the degree of nilpotence of ST?
What can you say about the nilpotence of S + T ?
16. Let T : Yt ~ Y, be define~ by
+
T(x1, x1 , x1 , x,) = (0, 2x1, 3x1 2x1 , x1 + 4x8) •
Theo prove that
(a) T is nilpotent of degree 4
(b) For a nonzero scalar A, I+ A7' is nonsingular and(/ -f AT)- 1
, ; I - >.T + A"'T' - >.•r. In particular, / + T and / - Tare
nonsingular.
17. Let T: 9'1 - ► fl d be defined by
T(ix0 -t « 1X + « 1 x1 + ix 3x1 ) =-= (x0 -t 2ix 1).x 1 ·t- (a 1 + 0tl)x 8 •
Then prove that
(a) Tis nilpotent tj degree 3
(b) For a nonzero scalar 'A, I+ 'AT is nonsingular. Fmd its inverse.
18. True or false ? ,
(a) If S : U - ► Y, T : V ➔ Ware linear, then
(To S)(«u) = ((ixT) o («S')Xu).
(b) T(S(«u)) = (-&T}(S(u)).
(c) If S is one-one, then TS i1t one-one.
(d) Every idempotent operator is nonsingular.
(e) A nilpotent operator can never be nonsingular.
(f) T:, == T ~ T-1 =
(T-1) 1 , if T·1 e1:ists.
(g) If Tis idempotent, then T.,, = T for all positive integen k.
(h) If S : U ➔ ,. and T : Y -+ U are linear and TS Ju, then =
s = r-•. •
(i) If R, S. T are three linear transfonnations such that RSI' is
defined and is one-one, then Tis one-one.

4.8 OPERATOR EQUATIONS

Let T : U -+ Y be a linear map from .the vector space U to the vector
space Y. In this article we shall discuu the solutions of the equation
7{11) - '• , (1)
where "• is a fixed vector in Y. Equation (l) is called an operator equation.
 4.8 OPlRATOR EQUATIONS/ J39
It js said to be homogeneoru if v0 = Ov, The set of solutions of the
equation
TC~=• ro
is simply the kernel of T, i.e. N(T). If the equation is not homogeneous,
i.e. Vo .:/= o., in Equation (I), then (I) is called the nonlwmo,:eneotts {NH)
equation and (2) the homogeneous (H) equation as&0ciated with f I) In
this connection we have the following theorem.
4.8,1 Tbecrem Let T: U - V be o linear map Given ,,0 -;I Ot in V, 1/,e
nonhomogeneous equation
(NH) T(u) = v0
and the associated homogeneous equation
(H) T(u) = Ov
have the following properties :
(a) If v0 • R(T), then (NH) hos no solution for u.
(b) // v0 E R(T) and (H) has the r~ivial solution. namely, u - Ou, as
its only solution, then (NH) has a unique sol11tio11.
(c) If v0 E R(T) and (H) has a nontrMul .volution, 11umdy, a volutinn
u --t- Ou, then (NH) has an infinite 1111mber of solutionr, In this catt'
if u0 ;s a solution of (NH), then the ftt of all solutions of lNH) is
+
the I wear ,,ariety u0 K, »-hl're K =-- .'\·( f) is the stt of oil
solutions nf (H). ·
Proof: (a) is obvious. Recall the definition of R{ f).
(b) If l'o E R(T), then T(u) =-= l'o has a !tolut1on. It 7(11)-=- Ov
has only one solution, i.e. u = Ou, then NtT) --= {Ou}, i c. Tis one-one.
This means T{u) = ,,0 cannot have more than one ~,,Jution. i.e. the
solution of (NH) is unique.
(c) If T(u) =-= o,.. has a nonLero !tolutiun, then N(Tl t {Ou}. Let
u0 E U be a solution of (NH). It exi!tt!i because ,,0 E R(T). Then T!11i,)
= )•0 • Now if u,. E N(7), then
T(u0 + u,.) =- T(u0 ) I T(uA)
= v0 + Oy = l'o •
Therefore, u0 t- u1: is a solution of {NH). This is true for every
u,. € N(n, and since the latter has an infinite number of elements in it,
(NH) also has an infinite number of solutions.
From this discussion it is obvious that u0 +
K, where K =:: N(T), is
contained in the solution set of (NH). Conversely, if w be an;t other
solution of (NH), then
T(w) = v0 = T(u0)
or T(w - u0) Ov • =
This means w - lie e:N(T) ""' K. So w and &lo belong to the same
 140 / UNBAR T.RANs.R>lUIA TIONS

parallel of K, namely, u. + K. Thus, the solution set of (NH) is precisely

11e+x.1
Note tbat Uo + K is the T-pre-image of v0•
Example -I.JO Let D: iz1 1°co, 21r) ➔ 16',,(0, 21r) be the linear d.iff"erential
operator. Consider the operator equation
D([Xx) = sin x •
To solve this, we Jook at the associated homogeneous equation
(H) Df= 0.
The solution set of this equation is the set of all constant functions. So
K = {/ I /(x) = b for all x E (0, 2n:) and b a constant} .
One solution of D(f")(_x) = sin x is the function fo, where fo(x)
== - cos x. So the solution set is
fo + K.
In other words, the set of all functions g, where g(x) = - cos x + (a
constant) is the solution set of D(/)(x) = sin x.
To solve a nonhomogeneous operator equation
(NH) T(u) = 110 ,
where Tis a linear operator, we go through three steps :
Step 1 Form the associated homogeneous equation (H).
Step 2 Oet all solutions of (H). It is the kernel of T, i.e. N(T).
Step 3 Get one particular solution u0 of (NH).
Now the complete solution of (NH) is u., + N(T).
Example 4.Jl Let T : V6 - Va be a linear map defined by
T(e1) = ½/1, T(e2) = i/1, T(e.) = J.,
T(e,) = / 1 , T(eJ == 0,
where {e1 , e1, e3 , e,, e1 } is the standard basis for V1 and {/j,/1, /3} is the
standard basis for V3• Then solve the equation
. T(u) = (I, I, 0).
We first calculate the value of T(u), i.e. T(xi, x1, x1, x., xj :
+
7lxi, x1, x1, x,, xJ = xiT(e1) + x 1T(e,J x1T(e1) x.T(eJ +
+ x,T(e11\ = ( -X1 +
2 - , x1 + x,, 01 •
X1 ,\

The associated homogeneous equation leads to lhe equations

Solving tbeie, we get x1 = -xi, x1 == -x,.

Thus, the kernel of 7'is the
set of all vectors of the form (xi, -xi, x1, -x1, x1), i.e. x1(1, -1, 0, 0, 0)
+ x,C0, 0, I, -1, 0) + x.(O, 0, 0, 0, 1). Hence,
· N(T) - [fl, -1, 0, O, O), (0, O, I, -1, 0), (0, 0, O, 0, 1)) ,
 4.9 APPLICATIONS TO DIFFl:kENTIAL EQUATIONS/ 141

One particular solution of T(u) = (1, 1, O) is u = (2 o 1 o O)

which is obtained by putting x1 =- - 2 x2 =- O x =- 10 x -~ o' x' ·• o'
•
So the complete solution of the equation
• ' 3 ' ,1- ' .= .
T(u) :a- (1, 1, 0)
is the linear variety (2, 0, J, 0, 0) f- N(7), i.e. the &et
(2, 0, J, 0, 0)+ {(a, -a, b, -b, c) I a, b, c arc reul numben.} ,
which i& the same as
{(a+ 2, - a, b -1- J, -b, c) I a, h, care real numbers}.
In other words, the T-pre-imagc of (I, I, 0) is this linear variety.

Problem Set 4.8

I. Determine the range, kernel, ,md the prc-im.1ge of (3, 1, :) for the
linear tr,111sformations R, S, T of Prob 1cm I, Problem Set 4.4.
2. Find the T-pre-image of (I, 2, 3) under the lmear tran,;formatiom, T
defined in (b) through (e) of Pruhlem J Prnblem Set 4.2.
3. Find the T-prc-image of the followinp; two VCltor,; under the linear
tr.1n.,formatios defined in (h) through (I) of Problem 1, Problem
Set 4.2:
(a) x (b) .r.

4.9 APPLICATIONS TO THE THEORY OF ORDINARY

LINEAR DIFFERENTIAL EQUATIONS
In this article we propJse to appl) the theory of operator equations to
the important orcr 1tor equation in mathem.tt1cs, namely, the linear ordi-
nary d1fferentu,I eqm:.t1on. For this we &hall develop the theory of
ordinary linear differential equation, using the necessary concepts from
linear algebra.
The simplest linear ordinary differential equation of the first order is of
the form
a0(x) di_
,,.,; + a1(x)y · g(x) , ())
where a 0(x), a 1(x), and g(x) arc continuous on an interval 1. If a 0(x) is
nowhere zero in/, then Equation()) is called a 11ormal linear differential
equation of fint order. If Equation (I) is normal, we can conveniently
write it in the form
dy +
a 1(x) g(x)
(2)
dx 0o(i) y Do(X) •=
This is usually written as
(3)
 142 / UNEAR. TRANSFOAMATIONS

where P and Q arc continuous functions of x alone. We can rewrite jt in

operator form as
(D + P)y = (J
or
Ly=~, (4)
where L stands for the operator D + P which is a linear operator from
r,;ycu(I) to <if([). Since Lis a linear operator, we can Jook at Equation (4)
as an operator equation. Therefore, we can apply the results of§ 4.8. In
particular, we know that the general solution of Equation (4) consists of
two parts. One part is the solution of the homopneous equation
(H) Ly- 0
and the other is a particular solution of
(NH) Ly= Q.
The general solution of (H), i.e. of
: +Py=O (5)

is called the complementary function and is denoted by )'e. A particular

solution of (NH), i.e. of Equation (3) is called a particular Integral and is
denoted by YP·
Therefore, the complete solution of Equation (3) is of the form
Y=h+h, 00
where Ye is actually the kernel of L and )'p is some particular solution of
Equation (3).
To solve Equation (3), we first find Ye• We write Equation (5) in the
differential form
Pydx + dy = 0 (7)
or Pdx + _I)' dy = 0.
Integrating Equation (7), we get
+
fPdx In y = In C , (8)
where C is an arbitrary constant. Equation (8) can also be written as
y == ce-1Pu, (9)
which is the required complementary function Ye• Writing Equation (9) as
yef Pd~= C, we find that it is the solution of
~ (yJPdz) == O
dx '
i.e. pyef Pb + elPdx ~x= 0 •
This gives the clue to •the solution of Equation (3). We multiply both
sides of Equation (3) by efl'dz and get
ix Cye'Pa, - QJNx.
 4.t APPUCATIONS TO DIFFEIENRAL EQUATIONS / 143

which on intearation gives

yelNz = IQefPdxdx •
We are not interested in any arbitrary constant now because we are
looking for only one y,. So •
YP = e-lPdx JQelPdx.dx •
Thus, the complete solution or Equation (3) is
y =Ye+ YP :!:: ce-lPdx + e-JPdx IQef Pdx dx. (10)
where C is an arbitrary constant.
Example 4.32 dy - -=L = x
dx x1 - l ·
X
Here P =- xA _ 1, Q = x, and JPdx= -- J x 2 X_ 1 dx
= - ! In I x=I- - 1 I .
So ef Pdx = e-i In I A•-11 = l/vx•--=-1 and e-fPdx = vxi""=l.
Thus,
Yo= Cvr - 1,
Jp = v'~ - -•I_;-=-
vx~-1
d:c =- (.'t2 - I)•
and the complete solution is y = Ye -1- YP•
E:'l(amp/e 4.33
t +Py=Qy", n:;i-1. (11)
This equation is the well-known Bernoulli's equation. It is not linear, but
it can be reduced to the linear form by the substitution
dY dy
y = y•-", dx = {I - n)y-" dx •
Equation (11) then becomes
I dY
l--n
--d X
+PY=Q.
This is a linear differential equation and can be solved by the method of
Example 4.32.
Equation (I 0) represents a family of solutions (actually it is a linear
variety). For each value of C, this is a curve in the xy-plane. If, in
addition, we require a solution· that satisfies the condition y(x0) = Ya, then
the solution will be represented by a unique member of the family passing
through {x0, yo). The problem of finding a solution y = y(x) for Equation
(3), which also satisfies the initial condition y(xo) = Ya, is called u. initial
val11e problem for the normal first order linear differential equation.
The foregoing discussion is summarised in the followinJ theorem.
4.9.1 Theorem Every Initial ,ahle problem involving a normal first order
ord nary I near differential fflllalion has one and only one solution.
 144 / LINEAR TRANSFORMATIONS

This is called an exist,nce and uniqueness theorem-existence. because

the theorem as&erts that every initial value problem has a solution; unique•
ness, because 1t further says that the solution is unique.
Existence and uniqueness theorems are important in the theory of
differential equations. Theorem 4 9.1 was easily provable. The corres-
ponding theorem for the n-th order linear differential equation is beyond
the scope of thii; book. However, we shall now state it (without proof),
since we intend to use it in the theory that follows.
4.9.2 Theorem let
d"l' dn Ir
aufx) dx•• -1- a 1(x) dxn 1 t- ... -\ a,.(x)y -= g(x), (12)
a0(x) -/ 0, .Y E / be a normal 11-th order fir.ear d(ffi.'rmtial equation
defined. 011 a11 illlen·c.l /. I.et x 11 be a Ji n·d point in I. let at0 , 0t1, ••• ,
at,._ 1 be n arbitrur; re al 11umhc:rs. 111,·11 th,•re exists one and only one
solution of Equation ( 12) satis{J ing the inittal <·omfitums J (x0} :.- at0 ,
y'(xu) -- Ot1, .. ' y<11-ll(xu) - at, t•

The problem of finding a solution of Equation (12) i:.at1sfying the initial

conditions, listed in Theorem 4.9 2, is called an initial value p1oh/em for
the equation.
This theorem is at the basis of all fur1her theory of linear differential
equations. It says that every mitial value problem involving a normal
n-th order linear differential cquaticn has precisely one solution. We are
now ready to state the main the<)rcm of this article.
4.9.3 Theorem J he solution space of any normal 11-th order homogeneous
li11ear ordinary di.ffi'rentiul equation
d 11y d" 1y
au(x) cb.." + a1(X) axn-l
+ ... I un(x)y -= 0. (13)
a0(x) /- 0, x E /, defined on an int. n•a/ I is an n-dimcn.,ional sub-
space of 'I '" 1(/).
To prove this the(.lrem, we need the following lemmas.
4.9.4 Lemma The solution space of Equation (13) is a subspa ·e of (, 1,.1(1).
P,oof: We need only to note that the differential operator
L -=- a0(x)D,. -I- a 1(x)D,. -1 + ... -I a,,(x) is a linear operator from </,'Cn>(J)
to '/' (/), and so the solution space of Equation (J 3) is the kernel of L.
Therefore, it is a subspace of 'C <n>(J).
4.9.5 Lemma Let Y1, Ya, ... , Yn be n functions in <t,, 1" 1(/) and x0 a fixed
point In I. If the vectors v,(x0) = (y,{x0), y;(.Y0), ... , y, 1"· 11(x0)),
i = J, 2, ... , n of Vn are u, then Yi, y 2, ... , Ya ore LI over I.
Proof: To prove that Yu Ya, ... , Yn are LI over/, assume
C1Y1(x) + c.,y1(x) + ... + C,.Yn(x) 0 =
for all x e: /. (14)
 4.9 APPLICATIONS TO DIFI-ERJ!lNTIAL EQUATIONS/ 145

The assertion is proved if we can show that Equation (14) holds only
when c1 = c, = ... = c,. 0. =
Differentiating Equation (14) successively (n - 1) times, which is
justified because Y1(x), y 1(x), .•. , Yn(x) belong to 1( / 1•>(/), we get the
following equations :
C1Y1(x) +
CsYs(X) I- ... + Cn)',.(X)
c1y;(x)
. +
c8y~(x) + ... + c,.y~(x) .. ..
~1Y1 1"-11(x) + C2Y2'"-11 (x) + .. + C,J',. 1"-l>(x) = 0,
for all x E /. Thus, for x 0 E /, we obtain
C1Y1(x0) + CaY1(x0) + ... + c,.y,.(x0) =0,
CJ.Y~((x0) + CzY~(x0 ) + ... + c,.v~(x0 ) = 0,
C1Y1'"-U(xo) + C2Y2'"-ll(xo) + .. + c,.y,.'" ll(xo) = 0,
which 1s equivalent to
C1(Yt(Xo), y'1(Xo), ... , Y1 1"- 11 (x0)) t ciy1(x0), J'~(x0), ... , Y21"-0 (x.,)) +
.., t- c,.(y,.(x0 ), y~(x0 ), ... , y,. 1"- 11 (x0)) = 0.
In other words, c1,·1(x0) ~ c,1v 2(xo) + ... + c,.v,.(x0) = 0. Since Y,(x0),
i - 1, 2, , n are LI in V,., all the scalars Ci, c2 , , c,. are zero. Hence,
Y1t Ya, .. , Yn are LI o-ver J. I
Proof of Theorem 4.9.3 By Lemma 4.9.4, the solution space- of
Equation (13), denoted by K, i,; a subspace of l 1"1(/). We shall now prove
that dim K -= n.
Let x 0 be a fixed point in I. Then, by the existence and uniqueness
theorem, there exists a unique solution Yt satisfying the initial conditions
y;
Y1(Xo) = 1, Y~ (x0 ) = 0, (Xo) = 0, , , , Y1 <n-ll(x0) = O.
Similarly, there exist unique solutions y 2 , y 3 , •• , y,. ,;atisfymg respectively
the initial conditions
y 8(X0 ) = O,y~(x0) = 1, y;(x0 ) = 0, ... ,y21"-11(x0) = O;
y 3(X0) =- 0, y;(x0) = 0, y;(x0) = 1, •·, y 31"-ll(x0) = 0;

Yn(Xo) = 0, Y~ (X0 ) = 0, y;(xo) = 0, ... , y,. 1" 0 (x11) = 1 .

From this it is obvious that (y,(x0), y;(.A 0), ... , y. 1"-ll(x0)), i = 1, 2,
... , n, are the vectors e1 , e8 , .. , e" of V". Therefore, the vectors (y,(x0 ),
y;(x0), .... y, 1"-11 (x0)), i = 1, 2, ... , n, are LI in V,.. Hence, by Lemma
4.9.S, Yi, Y1, ... , y,. are LI over /.
We shall complete the proof of the theorem by showing that the
solution_s Yi, y1.... , y,. generate the space K.
Let y be an element of K, satisfying the initial conditions
Y(Xo) = «1, y'(x0 ) = 111 , ... , y<•-11 (x0) = at,..
Considei- the fun~ion «i11 t l&sYi + ... + GtJ,.. This is clearly a solutiQQ
146 / UNBAR J'KANSFOlcMATIONS

of Equation ()3) satisfying the said initial conditions. Hence, by the

uniqueness part of Theorem 4.9.2, y = «1Y1 + ... + «,J,'.. So y E C,,1,
Ya, ... ,yJ.
Thus, the set {yi, y 1, .•• , y,J is LI and spans K, which means that the
set {yi, y1, ••. , Ynl is a basis for K. Hence, dim K = n.
From Theorem 4.9.3, the following corollary is obvious.
4.9.6 Corollary If Yi, y 1 , ••• , y,. are n linearly independent solutions of
Equation (H), then any soluti,m of Equation (JJ) can be written as
CJY1 + Ca11 + ...
+ C,,y,..
In Theorem 4.9.3 and Corollary 4.9.6 we have discussed in detail
the solution of the homogeneous normal linear differential equation
Ly = 0, where L = a0 (x)D" + a1(x).D"-1 + ... + a.-1(x)D + a,.(x). We
now consider the nonhomogeneous equation
Ly= g, (15)
Since L is a linear operator from ~~''"'(/) to ~ (I), rquation (15) can be
considered as an operator equation. Hence, by Theorem 4.8.1, the
general solution of Equation (15) is the L-pre-image of g. It is the linear
variety YP + Ye, whereyp•is a particular solution of Equation (15) and Ye
is the kernel of L. We summarise this result in the form of a theorem.
4.9.7 Theorem Let
d"y d•-ly --
ao(X) d~ + a1(x) dx•-i + ... + a.(x)y - g(x) (16)

be a nonhomogeneous normal linear differential equation on an interval

I. Then the set S of all solutions of Equation (16) is a linear variety
S = YP + Ye in '6' 1" 1(1), where the leader YP is any one solution of the
equation and Ye is the set of all solutions of the associated homogeneous
equation.

Problem Set 4.9

1. Solve the following differential equations :
~ ~+x ~
(a) dx - y = xe (b) X di + y = r + x• cos X
(e) (4-3t) : - Sy= x'/(3x- 4)1 11 (d) -t + cxy == ~
(e) (X + 2y8) * == y (0 : =- Q + Jy + 8
(g) (tan-1y - x) f == 1 + y8 (b) secy dx - (x+cos y)dy == O
(i)<,8- l)dx + (2x-r + l)dy- 0
(j) coa y b + (X sin 1 - I)'>' - 0
 4.9 APPLICATIONS TO DIFFEllENTIAL EQUATIONS/ 147

(k) (r + 1) : + 2xy = sin x + xe- + r

dy 1
(I) (r + 1) -d--
x
+ xy = x(l----.
+x 2)

2. Solve the following differential equations :

(a) (1+ r) : = xy - y3 (b) ~: = x3y3 - xy
dy .. -x•
(c) -- + re = xy (d) e1 dx = x(2xy + e )dy
11
dx
(e) cos y dx + x(x - sin y)dy = 0.
3. In each of the following differential equations, verify that the given
functions are solutions. Determine whether they are LI and whether
they form a basis for the solution space.
(a) y" + S)I' -1- 6y = O; e•la:, ,.-3- on (-oo, co).
(b) y"' - 4y' = O; eb, e-2• on (-ex:,, oo).
(c) x'y# - hy' t 2y"""' 0; 2.x' -x, x2 t 2.t, r
on (- co, co).
(d) y'" - y• - , 2y' = 0; c-111 , sinh x - ~/2, 2e2"', 1 on ( - ex,, oo).

(e) y" - 2xy'/(l - r) =- O; 1 - tanh- 1x, J + In /: ~ ~,. 2 on

(-1, i,.
(f) (D4 - 4D.1 -l 7Dl - 6D T 2)y = 0; l'e111, e4, ea:( 1 - sin x),
em cos x, e'(sin x - x) on (- oo, uo ).
(g) (x2JJ2 +3xD -1- I))' = 0; J /x, In x/x on (0, oo).
(h\ (xD3 - D1)y =-" O; I. x, x 3 on (0, co).
Chapter 5

Matrices

Matrices form an important tool in the study of finite-dimensional

vector spaces. Determinants form an important tool in the study of
matrices. We shall study matrices in this chapter ar.d determinants in the
next.
Hereafter, though for convenience we deal with real vectors and the
,cal vector space Y,11 all our arguments and dcfio1tions, unlrss otherwise
restricted, will apply also to complex vectors and the complex vector &pace
vc••
5.1 MATRIX ASSOCIATED WITH A LINEAR MAP
For the theory of matrices, it is sometimes necessary to viiiualise
vectors of V,, (any p-dimensional real vector space 1s isomorphic to Vp)
as column vectors, i.e. a vector («1t «1 , ••• , «21 ) of VP may be written in the

r«.
form
«1 7

I
L «,, .J
where coordinates of the vector are written in a vertical column read from
top to bottom. It may be noted that so far we have been writing this as
a row vector, i.e. in a horizontal row from left to right. To save space,
column vectors are also written as («1, «2, ... , «,,)r, where the letter T
denotes that the vector is a column vector. (The appropriateness of the
letter Twill become clear later.)
For example, the vectors e1, e1 , and e1 of V1 may be written as
 S.t MATRIX ASSOCIATED WllH A LlNEA'll MAP/ 149
r17 ro7
107
l:JLJ··l:J
or as (I, 0, O)T, (0, 1, O)T, and (0, 0, J)T. If 11 = 2e1 - 3e1 + e3 , then
the coordinate vector (2, - 3, I) of II relative to the basi~ {ei, ei, e3} can be
written as
r 2 -1
I -3 or (2, -3, J)T.
L l _J
Before we discuss the general definition of a m~trix, let us take a simple
example : Let B 1 = {ui, u8 , u3} and B 3 -- {111, 111, 113 , 11,} be ordered bases
of Ya and v., respectively. Let T: V3 - v. be a linear map defined by
T(u1 ) =- 111 - 2111 +
113 - 11•
T(u1 ) -= 111 111 t 211,
, T(u3 ) = 2112 +
3118 - 11, •
Then the coordinate vectors (relative to B 1 ) of T(u1), T(u2 ), ar.d T(u.),
written as column vectors, are respectively
r

L-l_J
17
-2
1
r

L
-1
0
17

2_j
, and I 07
2
3
L -1 _J
Everything about the linear map T is contained in these 3 x 4 = 12
numbers, written as above. The arrangement of these 12 numbers in the
form
r I
I -2 -1

L-1
is called the matrix of T relative to B1 and B1 •
We shall now give the general definition of a matrix.
5.1.1 Deflaltlon Let U and V be vector spaces of dimensions n and m,
respectively. Let B1 = {u1, ... , u,.} and B1 -== {111, ••• , 11,,.} be ordered
bases of U and V, respectively. Let T : U - V be •a linear map
defined by T(u,) = «,,111 + «11v1 + ... + ac.,11,,., j = 1, 2, •· , n so
that the coordinate vector of T(u,) written as a column vector is
ISO I MATlUCliS

I «11
«11

L «...,
7

_J
1·

Write the coordinate vectors of T(u1), T{u1 ), ... , T(u,), .. , T(u.)

successively as column vecto'l'S in the form of a rectangular array as

I
follows :
«u «11 ... «u

«in «aa ••• «s,

. .:
: I
I
L«m1 «ma « ..., at...._!
This rectangular array is called the matrix of T relatiw to the
ordered bases B1 and 8 1 , and is denoted by (T: Bi, Bll).
It may be noted that in this definition «11 is the i th coordinate of T(u1)
relative to the basis {v1, v1 , •• , , ...}.
5.1.2 Remark The numbers that constitute a matrix are called its entries.
Each horizontal hoe of entries is called a ~ow. Each vertical line of
entries is called a cofomn. The above matrix has m rows and n
columns. The matrix is therefore called an •m x n' (read as m by n)
matrix. The order of the matrix is said to be 'm x n'. Thej-th
column of the matrix is a column vector («1,, «J,, ... , «...,)T, and may
be considered an element of v.... The i-th row of the matrix is a row
vector {«11, cx,11 , •• , ex.,,), ,md may be considered an element of v.
The matrix is denoted, in i,hort, by the symbol (« 11) 1 ... 1, •. , ,,. or
J"" I, .. , n
simply («11 ) 111x 11, «, 1 is called the ij-th entry, and it is at the inter-
section of the i-th row and the j-th column.
Example S.J Let a linear transformation T : V1 ➔ V8 be defined by
T(x11 x1 ) =- (x1 + x., 2x1 - x1 , 7x1 ).
If B1 = {ei, e1} and B1 == {/11 1,.,/a} are the standard bases of V1 and
Y1, respectively, then
T(e1) = /1 + 2ft + 0/a
and T(e1 ) == Ii - / 1 + 7/a .
The coordinate vectors of T(e1) and T(e,) are respectively

I:l
Lo _J
and [ - :
7
J7.
 5.t MA1RIX ASSOCJATEI> WITH A UNIIAR MAP / 151

Therefore, the matrix of T relative to B1 and B1 is

L0
r: _:1-
7_j
Exampie 5.2 Let a linear map T : Va ➔ Ya be defined by T(x11 x1, Xa)
= (X1 - "• + Xa, 2x1 + 3xa - ix,, X1 + x 1 - 2x3).
If B1 = {e1, e2 , e8}, the standc1rd basis, and B 1 = {(I, I, 0), (1, 2, 3),
(-1, 0, I)}, then T(e1 ) = (], 2, 1) = 2(1, 1, 0) -l- 0(1, 2, 3) + 1(-1, 0, I).
So the coordinate vector of T(ei) relative to B1 is (2, 0, I). Similarly, the
coordinate vectors of T(e1 ) and T(e3 ) relative to B,. are (6, -j/2, 11/2) and
(0, -1/4, -5/4), respectively. Writing these successively as column
vectors, we get the matrix of T relative to B1 and B~ as

Ir: _: _:7J·
LI ¥ -t
Example 5.3 Let a:1inear map T: V3 ➔ V1 be C:cfined b} T(e1) = 2/1
+
- f., T(e1 ) == / 1 + 2/a, T(e3) == 0fi 01,., where {e1 , e1 , eJ} and {.'1, fa}
are standard bases in V3 and V1 , respectively. Then the matrix of T
relative to these bases is
r 2 1 07
I I•
L-1 2 0_j
Let us find the matrix of the same linear map T relative to some other
bases, say B1 == {(I, I, 0), (I, 0, 1), (0, 1, 1)} and B1 = {(I, 1\, (l, -lJ}.
We have
T(x1, x 1, x 3 ) == x1T(e1) + x 1 T(,1 ) + xJT(e3)
== (2x1 + x,) Ii + (2xa - X1) h
= (2X1 + l'1, 2X1 - X1) ,
Therefore, T(l, 1, 0) = (3, I)= 2(1, 1) + 1(1, -1)
T(l, 0, I) = (2, -1) = i(l, 1) + f(I, -1)
T(O, 1, 1) = (1, 2) == 1(1, 1) - i(l, -1).
Hence, the matrix of T relative to B1 and Ba is
r2 i 17
I I.
Ll f -l_j
Noto that the matrix of T changes when we change the bases. Wo
shall pursue this matter in Chapter 7.
Example 5.4 Let a linear map T: Ela ➔ !J1 be defined by TC«o + «ix +
V + «1 x') == oc1 + («. + «3)x + (oc0 + «1).x1•Let us calculate the
152 / MA11UCl!S

matrix of T relative to the bases B1 = {1, (x - I), (x - 1)1, (x - 1)1} and

B1 = {l, x, x1 } :
T(l) =r = 0· I + Ox + I r
T((x - I)) =
0= 0 • I Ox + +
Ox1
T((x - J) J = x - .\ = O • I
1 1 Jx - 1x1 +
T((x - 1) ) -=- l - 2x
1 2 +
2x = I • 1 - 2x + 2x1 •

r: ~ :_:1
Hence, the matrix of T relative to B1 and B1 is

Ll O -l 2_J
Example 5.5 let D: 9', - 9'1 be the differential map D(p) = p'. let
us calculate the matrix of D relative to the &tandard bases {I, x, x•, x 3} and
{l, x, x1}:
D( I) --= 0 = 0 , l -t Ox Ox 1 +
D(x) = 1 ~ 1 • 1 + Ox t Ox1
D(x11) = 2x = 0 • l l- 2x + Ox8
D(x3 ) = 3x2 = 0 • I +Ox+ 3x1 •
Hence, the matrix of D relative to the standard bases is
ro 1 o 07
I o o 2 o
LO O O 3_j
Example 5.6 We can easily check that the matrix of the identity map
T: U ➔ U (dim U = n) relative to a basis B(B1 = B,. = B) is the n x n
matrix
0 0 07
1 0 0
(I)

0 0 l_J
nis matrix is called the identity matrix and is denoted by 1,. (or simply
1. if n is understood).
The ij-th entry of this matrix is usualJy denoted by 611 , where 611, called
tho Kronecker delta, is defined by
6,, -_{I,0 , if i = J
jf i #: J • (2)
So 1. is the matrix (8, 1).x.,
Example $.1 Usina tho same procedure as in Example 5.6, we can check
 5 1 MATRIX AS!OCIATE~ WITH A UNEAR MAP / I 53
that, whatever the bases used, the zero map O Y, Y"' will have the
matrix
ro 0 07
0 0 0
(3)

I
LO O •• O_J
Each entry 10 this matrix 1s zero. Th11i, matrix 1s called the zero matr,x
or order m x n and 1s denoted by O,,.,..n or simply O, 1r the order 1s under-
stood.

Problem Set 5 1
In Problems I through 7 dete1mme the matrix (T. B" BA) for the given
hnear transfo1 mallon T c1nd the bases B 1 and Bl
1. T: v~
- ► v 1 , T(Y, >) = (x, -y)
(J) B1 -= {e1 , el}, B11 -= {(I, I), (1, -1)}
(b) B 1 - {(I, I), (I, 0)}, B2 = {(2, 3), (4, 5)}
2 T V3 ► v~. T(x, y, z) -=- (x .J J, y-+ z).
(t) B1 =--{(I, I, 1),(1,0,0),(1, 1,0)},~= {ei,e2 }
(b) B1 = {(I, I, t), ( ~I, 2, - l), (2, 3, I)}. BA-= {( '3), (I, l)}.
3. T: V, - V5, T('- 11 X2, x3, x,) = (2.\ 1 + x2, '\1 -x3, x3 + x,, Yi,
, 1 + X 11 , 3.\3 -t- x,).

B1 = {(I, 2, 3, I), (I, 0, 0, I), \I, I, 0, 0), (0, I, ', 1)}

B~ = {ei, eA, ea, e,, e6}
4. D: fl,,, --,. fl,,,, D(p) -= p'
Bi = B. = {1, Y, x .. • , A."}.
S T. fl, ➔ CJ,, T(p)(¥) = 1; p'(t) dt.
(a) B1 = B1 =- {I, x, r, r, t'}
(b) B1 = {I, A., r, x 3, x'}, B2 = {x - I, x + I, x8 - x', x• + x',
x• + x}.
6. T : fl 1 ➔ Ela, T(p) = xp(x).
B1 = {1, 1 + x, 1 - x + x2}, B1 = {I, I + x, x2, 2x - r}
(a)
(b) B1 = {l, x, r}, B1 = {1 + x, (I + x)9 , (1 + x) 1 , 1 -.or}.
1. T: Vf -+ Vi, T(z1, z1) = (z1 + 1z1, Z1 - iz,.). B1 and B1 arc stan-
dard bases.
8. Determine, relathe to the standard bases, the matri~ of each linear
transformation T an Problems 1(a) through 1 (g) of Problem Set 4.2.
9 True or false ?
J54 / MATJllCIS

(a) Let T: V, ➔ Y8 be a linear map. Then the matrix of T relative

to any pa,r of bases is of order 4 x 6.
(b) When referred to the standard basis, the matrix of the linear
map T; V1 ➔ V1 defined as (x, y) I-+ (x', y'), where x ==
x' cos 0 - y' sin 8, y = x' sin 8 + y' cos 8, is
1 cos 0 -sin 07
I !.
L:i.m G cos o_:
(c) The matrix of the identity map lu : U -► U relative to any pair
of bases is the identity matrix.
ro o o 7
1 O O
II '" the matrix of the linear map T : Y3 ➔ Y3
LO O O .J
defiraed by T(e1) = 0, T(e11 ) = l'i, T(e3 ) = O relative to the
standard b.ises.
ro o 17

(e) I1 0 0 Iis the matrix of the linear map T : V ➔ V defined 3 3

LO l O J
by e1 = T(e3), e1 = T(e 1 ), e8 = T(e1 ) relative to the standard
bases.
(f) To every linear transformatior, there corresponds a unique
matrix.

5.2 LINEAR MAP ASSOCIATED WITH A MATRIX

In § 5.1 we started with a linear map and then defined its matrix.
However, in order to define a matrix, it i!> not necessary to start from a
linear map. Matrices can be considered as entities in their own rights as
the following definition will show.
5.2.1 De8nitlon Any rectangular array of numbers such as
1~11 ~12 ~JI ~ltl 7
.
u,

~l ~ lit, ... ... ~

:
~fl ~,. ~-, ~, .
.
L~.1
~-- ~-, ~--.
 5,2 LINIIAR MAP ASSOCIATIID WITH A MATRIX/ 155
is called an III x n matrix. If m = n, the matrix is called a square
matrix.
Two matrices A. = <«.,).,)(,. and B = (~ 11 ).,x,. are said to be eq11al,
written as A. = B, if «, 1 = ~., for every ; and j.
All the statements in Rei;nark S.1.2 also apply to a general matrix as
defined in Definition S.2.1.
A matrix whose entries are all real (complex) numbers is called a real
( omplex) matrix.
As already mentioned, our discussions m this chapter deal with real
vector sraces, i e. Vn and V"'. Also, the matrices used will be real.
However, all the d1scumons hold for complex matrices if the real vector
spaces V" and V111 are repl.tced by the complex vector spaces Vf and V~.
Let us start with a matrix B = (~11) 11u<n and ask whether we can find a
lmear transformation T: U-+ V, where U and V are suitable vector
spaces with ordered ba~es, say B1 and BJ, so that B = (T: Bi, B8).
The answer to this quei.tion is yes, because the process of obtaining the
matrix from a hne~r transformation (see Definition S.1.1) 1s reversible.
Let us g1> through thl." steps of this reverse process.
Let the rn,ilnx B = (~, 1)mxn be given. Ui,ing lhts matrix B, we shall
define a linear map S : U -.. V, where U and V are vector spaces of
d1menMoni, n and m, respectively.
Let B1 -= {111, "i, .. , un} and B1 =- {v1 , v1 , ... , v.,} be ordered bases for
U and V, rei.pecuvely Then we define a linear map S : U -+ V by pres-
cribing \alue!!i of Son the vectors of B1 as follows:
f (u1) = + ~11Va + ,.. + ~m1Vm ,
f11V1

~(u,) = ~11V1 + ~,111 + ... + ~,.,v., ,

.. . ..
S(u,.) = ~111111 + ~znV1 + ... -1- ~..,.v,. ,
Extend S hnearly to the whole of U. The linear transformation S is
unique (Theorem 4.1 .S). By this very construction, it 1s clear that
B = (S : B1, B1). For this purpose, we have only to note that the coordi-
nate vector of S(u1) relative to B1 is thej-th column of B. S: U-+ Vthus
defined is called the linear map associated with the matrix B relati,e to
B1 and B1 •
Note that an m x n matrix gives rise to a linear map from 11.n ,._
dimensional vector space to an m-dimensional vector space. Further, note
that, if we choose bases other than B1 and B1, the hnear map associated
with the matrix B will also be different.
Example 5.8 Consider a 4 x 3 matrix
 156 I MATRICES

r 2 -3
I 0

L I
-2
2 -2_J
I -:1-
This matrix will give rise to a Jinear map S : V3 ➔ v,. Suppose
B1 = {ui, Ila, ua} and B1 -= {vi, v1, v3 , v,} are ordered bases for V1 and v,,
respectively. Write
S(u1) = 2v1 + v1 - 2v3 + v,
S(ua) = -3v1 + v3 + 21•,
S(u3 ) = 4v1 - 1·1 - 2v,.
The linear extension of this definition of S on the basis elements is the
required map S. If BJ and 8 1 are the standard bases {e 1, e1 , e3 } and
{/11/1./1,/,} for Va and V,, respectively, then we get a linear transfor-
mation T : V1 ➔ V, given by
T(e1 ) = (2, 1, -2, I), T(e1 ) = (-3, 0, I, 2), T(e3 ) = (4, -1, 0, 2),
i.e. T(Xi, Xz, Xa) = (2X1 - 3x. + 4xa, Xi - Xa, -2X1 + X1,
X1 + 2xa + 2x3) •
Let M.,,.. ,. denote the set of all m x n real matrices. Let U and V be
real vector spaces of dimensions n and m, respectively. Fix ordered bases
B1 for U and Ba for Y. Then the process of determining the matrix. of a
linear map and the linear map corresponding to a matrix shows that the
map
't': L(U, V) ➔ M..,.,,. (1)
defined by -r(T) = (T : Bf, B 2) is one-one and onto.
This result merely says that, if bases B1 and Ba are fixed, then to each
linear map T : U - V there exists a unique m x n matrix and to each
m x n matrix there exists a unique linear map from U to Y. This enables
us to pass from linear transformation to matrht and vice versa.
When U = v. and Y = V.,. and the bases B1 and B1 are the standard
bases in the spaces, then the matrix associated with T: Y. ➔ Y.. is called
its natural matrix.
Note (cf Example 5. 7) that -r(O) = O.xa• Further, note (cf Example
5.6) that, if
't' : L(U) - t.\,., " (2)
then -r(I) = I,., where I is the identity transformation on V, and r. is the
n x n identity matrix.

Problem Set 5.2 .

ID Problems 1 through 4. for each given m x n matrix .4 and bases B1
and B1• determine a linear transformation T: v. ➔ Y.. such that
.4 = (T : B1, B1 ).
 5.2 LINEAR MAP ASSOCIATED WITH A MATRIX/ 157

I. A= r: ::
Ll. 2 O O__I
(a) B1 and Ba are standard bases for V, and V8 , respectively
(b) B1 = {(l, 1, 1, 2~, (1, 1, 0, 0), (0, 0, I, 1), (0, I, 0, O)},
B1 -c.. {(1, 2, 3), (I, -J, I), (2, I, I)}.
r1 o o 1
2. A= 0 1 0 I·
LO O 1 _J
(a) B1 = BA - {e1, e2, e3}
(b) B1 - {(I, l. 1), (I, 0, Ol, (0, l, O)},
H2 =-- {(), 2, 3), (l, 1, 1), (2, 1, I)}
(c) B 1 -- {(l, 2, 3), (), -1, I), (2, I, J)},
B 2 =- lO, I, I), (), 0, 0), (0, 1, O)}.
r 1 1 2-1
3. A:- I I.
I 1 1 0 I
(a) B1 and B2 are the btandard bases for V3 and V2, respectively
(b) B1 :- {(I, I, I), (I, 2, 3) (I, 0, O)},
Bi · {(1, 1), (I, - l)}
(c) B1 -- {(I, 1, 1), (I, 2, 0), (0, - I, O)},
B1 -- {(J, 0), (2, I)}.

4. A= r~ : :7
L- I 3__1
(a) B1 and B2 are standard bases for Va and VJ, respectively
(b) B1 - {(l, I), (-1, l)},
B, -- {(1, J, 1), (1, -1, 1), (0, 0, 1)}
(c) B1 = {(1, 2), ( - 2, 1)},
B2 - {(I,-· l, -1), (I, 2, 3), (- 1,0, 2)}.
I
5• If Lsincos 8 - cos 87 1s the matrix o f a Jmear map T : V.11 ➔ V.1
sin 0_J
8
relative to the btandard bases, then find the matrix of T-1 relath e to
the standard bases. '
6. True or false ?
(a) The matrix of the linear map T : U ➔ V is square if/ dim U =
dim V.
(b) An identity matrix is a square matrix.
(c) A zero matrix is a square matrix.
158 / MATRICES

r1 0 07
{d)

LO
0 0 1
1 O..J
Iis the matrix of the identity transformation

I: Va - Va relative to the bases B 1 = {ei, e1 , e3} and B1 = {ei,

e3 , e1}.

(e) I; :1
LI 1..J
determines a linear transformation from Y, to Y,

defined as T(e1) = ft, T(e1 ) = Is, T(e3 ) = Ii + I,. relative to the

standard bases (e1 , e8 , e3 } and {.'1, J.}.

5.3 LINEAR OPERATIONS IN».,.,.

Throughout this article U and V are real vector spaces of dimensions
n and m, respectively; and B1 = {u1 , u1 , ... , u,.} and B1 = {v1 , v8 , ... , v,.}
are ordered bases for U and V, respectively.
Let T and S be two linear maps from U to V. Then T + S is also a
hnear map. If the matrices of T and S relative to B1 and B1 are A. =
(«u)111x11 and B = (~ol..x,u respectively, then
m
T(u1) = l: cr:.,v, U = I, 2, ... , n) {I)
f=l
and
m
S(u1 ) = I
l=J
~.,11, U = 1, 2, ... , n). (2)
So (T + S) (u1) = T(u,) + S(u1)
m
= l: (er:,, + ~, 1)v, (j = 1, 2, ... , 11) • (3)
l=J
Thcrefore, the j-th column of the matrix of T + S relative to B 1 and B 1 is
(«1' + +
~11• 11fl ~II• ... , 11 11 + ~"' ,.. , «mi + ~m1)T •
Hence, the matrix of T + S is
1111 + ~11 «11 + 1311 «u + f3u +
«111 f3111 7
«11 + 1311 «.1 + f311 «.1 + ll.J ... + f3111
.. .:
(4)
°''1 + fi11 «,. + f3,. ... + 13,, 13,.
"''' "''" +

Lac..,+ Pt.., -...+P.a., ... ..., + Ila, ... °'-- +,..

 S.3 LINBA.R OPBllA.tlONS IN M,_, 11 / 159
This motivates the following definition of addition of matrices.
5.3.1 Dethdtlon Let ..4 = (or.,,) and B = (~,,) be two m x n matrices. Then
the .rum ..4 + B is defined as the m x II matrix (or.1, + ~.,).
In other words, the sum of two matrices (of the same order) 1s obtained
by adding the corresponding entries.
Example 5.9 Let
rt 3 27 r-1 2 17
A= I B= I I.
LO -1 1_JI ' L 3 1 2 __J
1-1 - 1 3 +2 2 + 1-1 ro s 37
Then A+B=I I -= I I.
LO+ -1+1 1 + 2_J
J L3 0 J_J
5.3.2 Remark The argument preceding Deftmhon S 3.1 says only that the
map T : L(U, V) _,. tA.,.,,. of§ S 2 preserves add1t1on, 1 e. T(T + S) =-=
T(T) + T(S) for all T, S € L(U, V).
Let T: U - V be a hnear map and or. a scalar. 1hen cxT: U -► V 1s
also a linear map. If the matrix of T relc1.tJve to B1 and B1 1s (i,. 1) ..x,.,
then
m
Ttu,) = I «,,..,, (j = 1, 2, ... , n) • (5)
' 1
m
So («T) (u,) =- cxT(u,) = ~ (C1CX.,)i•, (J = 1, 2, ... , n) • (6)
' t
I herefore, the j-th column vector of the matrix of «T reld.tive to B1 and
B, •~
(««u, O'IX 20 , .. , cxix., .... , ixix,.,)T •
Hence, the matrix of ar.T 1s
CXIXi,

(7)

L ixa:,.1 171Xm1 otot,., otcx,...

This motivates the following definition of scalar multiplication for
matrices.
5.3.3 Deflnltloa Let A = (ac,,) be an m x n matrix and « a scalar. Then
the ,c•alar multiplication of ..4 by ac denoted by acA is. defined as the
m x n matrix («ac,1).
In other words, multiplication Of a matrix by a SQalaf a is dono bf
inllltiplyins each ntry by -.
160 / MATRICES

Example 5.10 Let

r1 3 27
A= I I,
L.0 1 l_J
r-l -f -17
Then <-i).4 =I 1.
L o -~ -LJ
5.3.4 Remark The argument preceding Definition ~-~.2 says only that
the map -r : L(U, V) ➔ :tA,.,,. of§ 5.2·preserves scalar multiplication,
i.e. -r(«T) = «T(T) for all T E L(U, V).
From the definition of addition and scalar multiplication it immedi-
ately follows that :tA,,., ti is a real vector space. (The reader is advised to
go through the various steps of verification.) Note that the 2'ero clement
of this vector space is the zero matrix 0 111,cn of Example 5. 7.
Further, Remarks 5.3.2 and 5.3.4 show that the map T : L(U, V)
➔ M,,. t1 is linear. Since we have already proved in § S.2 that -r is one-
one and onto, it follows that T is an isomorphism. We collect all these
resu'its in the form of a theorem.
S 3.5 Theorem (a) ""'',. is a ,ea/ veetor ,pare fur ,/,e Jore~oing
definitions nf addition and :,calar multiplication.
(b) -r : L(U, V) ➔ :t.\,.., n defined by -r(T) = (T: B1, B 2 ) is an iso-
morphism.
The implication of this theorem is that, for all practical purposes of
linear algebra, linear transformations between finite-dimensional vector
spaces are just matrices and vice versa.
We shall now determine the dimension of L(U, V). By Theorem S.3.S,
it is the same as the dimension of ,A,..,,,, because isomorphic vector !.paces
have the same dimension. (Why ?) So we have only to prove the
following theorem.
5.3.6 Theorem The dimens·on of the vector ~pace :tAnn II if mn
Proof: Oiven l and j, define the matrix £,, as the m x n matrix
with 1 in the ij-th entry and zero in all the other entries. We claim that
the set B of matricei;
B =- {E11 , .... £ 1•• £ 11 , .... E 111 .... , £,,11 , ... , £,..}
= {Eu E :tA,.," Ii= 1, ... , Pfl;j = 1, ... , 11 }
is a basis for :-A,,., 11•
To prove that B is 11 in )A,., ,., we assume «uE11 + ... + «1,.£ 111

+ «11E11 + ••• + ata11E111 + •" + 111111E..1 t ... + «..,.E..,. = ot11X11• This

mean• that
 S 3 LINEAR OPERATIONS IN tA ..

0 ..•
0 ...
...
0
0
..
l 11 / 161

L IX1111 1Xma O •• 0 _j
Hence, Gt,, = O for all ; = I , ... , m and j -== J, ... , n. Thus, B is u.
Again, to pro,e that [BJ =- M. 111 , ' " let us take a general matrix A ==
(Gt 11) m M. 111 , 11 • Then Jt can be eas1ly verified that
+ IX1,.E1t, + Gt~1E11 + ••• + «.nE111 + Gt1111Em1
+ ... t Gt11111E11111 •
Hence, B J& a bas,-.. Hence the theorem. I
The bac;1s {E, 1 I , I, ... , m; j = 1, ... , n} is called the .,tandard basis
for M.111 • ,..
Comhmmg Theorems 5.3.S and S 3.6, \\e have the foUowing corollary.
S.3 7 Corollary 1/ U and V are fin,te-dimensiJnal vector spaces, then
dim L(U, V) dim U x dim V.

Problem Set 5.3

In Problems 1 through 4 determme «A + ~B for the given scalars«
and~ and the matnces A and B.
r- 1 2 ., 1 r1 o
I. A=I l,B=I
L-1 1 t _J L2 1
(a) Gt = 2, ~ -= 7 (b) Gt = 3, ~ = -2.
rt 27 rt 07
2. A = I I,B = I I·
L3 l_j L2 -l_j

3. A.=
(a) Gt

r
I
=

-1
3, ~

1 2 3 47
1 1 I
= S

I·
(b)

B =
Gt

I -= 2,

3 -1
1
~

s ,
= -3.
2 07
3 I·
L 3 1 2 0.J L2 1 C 1.J
(a) II= 2,f3 = -6 (b) ar. == 3, f3 == S
(c) ar. - -7, f3 - 3.

4. A=
r1
j2
1-3
-1
-3 -1
2
17

OJ
I, B
r1

==\ 0 l
L3 1
0

-1
:1
162 / MA TRICES

(a) « = 7, ~ = -5 (b) « = 1/2, 13 = 2/3

(c) « 1/3, f3 = 4/5.
=
5. Let Ta, T 1 : U - V be two linear maps. Let B1, B, be ordered
bases for U :ind V, respectively. Then prove that
(«1T1 t- « 1 T1 : Bu B8 ) = «1(T1 : B1 , B,) + «1 (T1 : B1, Ba),
6. Let S, T : V3 - ► V, be defined as
x,, x 8) = (x1 + x 2, x 1 -
S(x1 , 2x1 + x 8 , x 1 + 3x3 , X 1 + x,)
T(xl' x., x,) = (X1 + 2x,, Xi - x •• 3x. + Xa, Xi + x. + X3),
Determine the matrix of 3S - 4T relative to the standard bases by
two diff'"rent methods.
7. Let S, T: 9'3 ·-► ,'I', be defined as
(S(p))(x) = (x' - l)p'(x)
(T(p))(x) = (3x + 2)p(x) - 1; p'(t)dt.
Dctermi11c the matrix of 4S + 2T relative to the ordered bases
B1 = {I, x, x", x8} and B2 {(I - x), (1 = + x), (1 - 't)1 , (I - x)1,
x'/2} by two d_ifferent methods.
8. Let B1 = {uu u1 , ... , 11"} and B1 = {,1 , 1·1 , ••• , v,.} be ordered bases
for the vector spaces U and V, respectively. Define T11 : U--► V,
I < i < m, 1 < J < n such that
{ O if k i= I
T.,(ua,) = ,, ·r
1 k -- •
- ,.
Then prove that
(a) (T,, : B1 , B8) = E, 1 (cf Theorem 5.3.6)
(b) {T.,} is a basis for L(U, V).
9. Define T: tA:, 1 -► t.\ 1 , 8 such that

[
«11 «11] [«11 + «11 o «11 + «s1] .
0
T «11 «:11 = «11 «11 + «11
Prove that T is linear and determine its matrix relative to the
standard bases for t.\1 , 1 and t.\1, a•
10. Repeat Problem 9 for T: M,, 8

T[
«11 «11 «117

«.1 «.1 «....J

I-
--► t.\ 1 , 1

- [«i1
«.1 + «.a
defined as
+ «11 + «11
1111
1111 J .
11. Let V be the subspace of ~c•>(-00,00) spanned by the functions
sin x, cos x, sin x cos x, sin1 x, cos' :c. Determine lhe dimension of
V, and prove that the differential operator D• maps Y into itself for
every positive integer n < m.
Determine the matrix of (a) 2D 3, (b) 3D' - D +
4 relative +
to the basis of Y obtained from the aiven 1pannin1 set or Y.
 5.4 MA1RIX MULTIPLICATION/ 163
12. True or false 'l
12 3 17 14 67 16 9 17
(al I I+ I 1=1 I,
L4 5 7_j LS -1 _J L9 4 7.J
r « 3«7 ,1 37
(b) I I = IX~ I I.
L2~ 4~_J L2 4_j
(c) OA = Om.<n for any m X n matrix A
(d) The set of all square matrices 1s a vector space.
(e) Every matrix can be written as the sum of two matrices, one of
which 1s two times the other.
(t') The set of all 3 x 3 matncei. of the form
ro
I
IX

Iy 0

L6 p
1s a group under the operation of add1t1on.

5.4 MATRIX MULTIPLICATION

Hereafter, we shall use only standard bases in the vector spaces Yn,
unless otherwise stated.
In this article let S : VP --+ Yn and T : V" -➔ V,. be hnear maps. Let
B1 = {ei, e., . , en}, B1 = {/ 1, I,., .. , f m}, and Ba = {gi, g1 , .• , g,} be the
standard bases for Vn, V,., and V,, respectively.
Let the natural matnx for T be A = («.,)mxn, and that of S be
B = (~ii )nxp•
We know that TS : Y, -► V"' is also a hnear map. Let the natural
matrix of TS be C = (y.,}mxp• We shall now calculate Y,; in terms of ar.'ir
and ~•s. We have, for each k = I, 2, .. , n,
m
T(ek) = I «.d.
1=1
= IX1kf1 + 1X1·1Ja + ... + IX,d, + ... + «,,.,J,,. (J)
and, for each J = I, 2, .. , p,
n
S(g1)= I ~t1e1
k=l
= ~11e1 + ~,ea + ... + ~1:,ek + ... + ~11,e.. . ' (2)
Al10, for each J = I, 2, ... , p,
Ill
(TS) (g,) =
,-1r,Jc
E

• ru/1 t r11ft + ... + r,,/, + ... + y,,.,/.,,, • (3)

164 / M.\TIUCES

Now, for each j = J, 2, .. ,p, we also have

(TS)(g,) =- T(S(g 1))
= T(~ue1 + ~1,e1 + ... + ~11,e1: + ... + ~..,,,.) (by (2))
= ~11 T(e1) + ~1,T(e1) + ... t ~.1:,T(e,1:)
+ ••• + ~ ,T(e 11 11) , (4)
bec1i use T is lir.ear. Each of the terms on the right-hand side of Equation
(4) is, by Equation (I}, a linear combination of the vectors fi.,f,_, ,/.,..
Collecting the 1erms involving/., we sec that the:; coefficient off. on the
right-hand ~ide of Equation (4) is
~1,«o + ~2,«,2 + ~.1:,«,.1: + ... + ~..,«... '
t ...
-= «,1~1, + «••~., -t ... + «,1:~t + ••• + «...~..}
J
n
=--= I «,1~11:, . (S)
k- I
Comparing this with the cceffi.cient of /. on the right-hand side of
Equafr1n (3), we get
n
Yu =- :& «,rr~u (6)
k-1
for every i -= I , 2, .. , m and j = I, , , . , p.
Using this as motivation, we define the product of two matrices as
follows.
5.4.l Definition Let A = (ac,,) be an m x n matrix tnd B =- (~1,) an
n X p matrix. Then the product AB is defined a11 the m x p
matrix C = (y,,). where
n
y., E = 11,.,~1,,
i = l, 2, ... , m; j = l, 2, ••. , p. (7)
k-1
In orc'er to undentand and use this <!efinition, we shall define the
•inner product' of two vecturs in V11 •
5.4.2 Peftnitlon Let u = (xi, x 1 , ••• , x,.) and , = (J,i, y 1, ••• , y 11) be two
vectors in v... Then the inner product of it and , denoted by u • , is
the scalar
u• "= Xi)'1 + x.y + ... + x,.y••
1

For example, if u = (I, 2, -3) and"= (0, -1, 2), then u · , = l x0

+2 X (-1) + (-3) X 2 = -8. ~.
By this definition, the exprusion for y" in Equation (7) becomes
r" == c.1P11 + 11,.~., + ... + 11, .., . ,
c:::i. (GC11, «c■, '"Is,.) • (~lit Ju, •••• ,_,)
= (t·th row vector of A) • (}-th colUDln vector of B), (8)
J:,oth being considtred vcctrrs in Y••
So Ddinition 5 4.J can be reworded as follnws,
 5.4 MATRIX MULTIPLICATION / 165

5.4.3 Defl■ltloll (Reworded) Let .A = (11,,) be an m x n matrix and

B= (~0 ) an n X p matrix. Then the product AB is the m x p
matrix C = ("r;,), where y" is the inner product of the i-th row
vector of .A. and thej-th column vector of B, for i == l, 2, ... , m and
J = 1, 2, ...• p, b:>th being treated like vectors in V,..
Example 5. I I Let
,-3 27

A=
0

1
-1

0 -2
1
L-2
I
,B
J_j
=I
r 1

0 -:1.
-2
L2 O_j

Herc .A is a 4 x 3 matrix and B is a 3 x 2 matrix. So AB is the 4 x 2

matrix
Yu
Yu

Y81
:::1
'Y111
•

LYn Y,z_
where
y11 = (1st row of .A) • (1st column of B)
= 3 X 1 + l X O + 2 X (-2) == - I
y11 = (1st row of A.) • (2nd column af B)
= 3 X (-1) + 1 X 1 + 2 X 3 = 4
y11 = (2nd row of A.) • (ht column of B)
= 0 X l + (-)) X O + 1 X (-2) = -2,
and so on. Completing the calculations for all the y's, we find that

AB-r=: _:7J·
L 2 -4
Bzo,nple S.12 Let
rt+, _, 27 .
r o l -17
A=-1 landB== 1 2
. L-4 + I 1 - I O...J
LI - I 1 _j
Here ~ is a 2 x 3 matrix and B i• a 3 x 2 matrix. So AB is the 2 x 2
matrix
166 / MATRICES

r«u «i17
I I .,
L«11 G&u.J
where «11 = (1 + i)O - i(l) + 2(1 - i) .,. 2 - 31
at11 = (1 + i)(l - i) - 1(2) + 21 == 2

ac21 = (4 + i)O -t (1 - i)l + 0(1 - i) = I - I

1111 = (4 + ,,o - t) + (1 - t)2 + o; = 1 - Si.
,-2 - 3i 2 -1
So AB= I I-
LI - i 1 - Si.J
5.4.4 Relllark The discussion preceding Definition 5.4.1 proves that if
T: L(V,.) ~ Mon, 11 1s defined by T(T) = the natural matrix of T,
then T(TS) = -r(T)-r(S} for all T, Se: L(V11 } and -r-1(.4B) = T-1(A.)
-r· 1(B} for all A, B E >A,., 11•
In other words, the comr,osition of transformations corresponds to
matrix multiplication.
Let us further analyse the proce:,s of matrix multiplication, namely,
r ... ~s, ... -1
••• ~11 •••

r= = = = 7 7

l ~. ~.. · · ~-.. ~- I
Li : : i -1
••• ~ltl ••• Yu ...
: .J
1·

L ·•· ~Ill ••• .J

Here attention is f 1.,cussed on the f.. ct that the ij-th element of AB is (I-th
row of A} • (j-th column of B). But recall that we can form the inner
product of two vectors only if both have the same number of coordinates.
In other words, m order that matrix multiplicJtion be possible. we
should have the number of entiies in the i·lh row of A equal to the
number of entries in thej-th column of B, i.e. the number of columns of
A should be the same as the number of rows of B. Thus, an m x n
matrix A can be multiplied on the right by a q x p matrix B tJI 11 IJ· =
If n =I=, tJ, then the product AB is not defined.
A 1 x n matrix A is simply a row vector cif the form (•1, «., ... , ••>·
An n x I matrix B ia simply a column vector of the form (P1, Ji., ... , p.)'.
The product AB of these two matrices is a 1 x 1 matrix [«1Pt + «.Ji.
+ .•. + •.JJ, whose only entry is nothins b11t a 1calar, which is the
inner product of vectors ..t and B.
In seneral, matrix multiplication is not commutative. For, if
 5.4 MA11llX MULTJIUCAflON / 167

r1 07 rJ 07 [ ' -27 rt - 1],

A=l l,B=I 1.C== l,D-1 0
L2 l_J LI -l_J 2 -l_J Ll
then
07 r3 07
I and B..t == I I.
-l_J LI -1-1
r-1 -17
Henct, AB #a B.4.. But we find that CD== I I= DC.
L I -2_j
On the other hand, we have the 6>1lowing theorem.
5.4.5 Tlaeorem (a) Matrix multiplication is asSflciative, that is, (AB)C
== A(BC), whenev,•r , ither side is defined.
(b) Matrix multip/icatiun is distributive over additio,,, that i.r,
A(B -f C) = AB -1- AC, whenever either side is defined. Al.ro,
(B +
C)D = BD +
CD, whenever either side i.r defined.
The proof is left to the reader. Note the followinJ :
(i) 0111xnAnl'j = 0111x11 and
AtlXpOp,<111 = Onxm•
(ii)/,.A ..x• = A,.x.. and
AnX.,J• == ..t.x..
5 4.6 Definition If A and B are square matrices of the same order such
that A.B == I = BA, then B i'i called an Inverse oj A.
Example S.13 Let
r I 2 37 r l
A= 0 1-21,B=I 1
L-1 1 1-1 L-l

Then
0
1 0
01 = B.A..
0 l_J
Therefore, B is an inverse of .A. and .A. is an inverse of B.
5.4.7 Tlaeorem If an Inverse of a matrl-c .A. exiat.r, then it i.r unique.
Proof: Suppose that B and C are two inverse~ of .A.. Then ,U1 == I
= BA. and A.C == I = C.A.. Thus, we have
C = CI = C(.A.B)
= (C.A.)B (Theorem 5.4.S(a))
== IB =- B.
Hence, B == c.1
168 / MATRICES

In view of this theorem, \\e speak of the (unique) inverse of A and

denote it by ,4- 1• Whenever A 1 exists, A i~ said to be Invertible.
By Theorem S 4.1, it foUo,\s that if ,4- 1 is the inverse of A, then A is
r
the inverse or .4-1, i e. (A- 1 1 =- A Both the matrices are therefore
invertible.
S.4.8 Theorem An n x n matrix A is invertible if/ the corresponding linear
transformation T (via tlze standard bases) is nonsingular.
Proof: Suppose A i<i inverllble. Then there exists an n x n
matrix B such that AB - =- I,. -= BA. Let the lmear transformation
corresponding to B be S: Vn - ► V,. Then, by Remark S.4.4,
't'-l(AB) = 't'-1(/,.) = 't'-l(BA)
or .. l(A}'t-l(B) -=- ,.-1un> = 't'-l(R)-r-l(A)
or TS -= I ==- SJ' .
So, by Theorem 4.7.2, Ti-, nonsingular.
Conversely, if T is nom,ingular, then there exists d lmear transfor-
mation S: V,. _.,, V,. such thdt TS"' I-= ST. Therefore, by Remark
5.4.4, -r(TS) -=-- -r(/) - -r(ST). If B is the naturc1l matrix of S, we have
AB = I,.= BA. Hence, Ah mvertible-1
For this reason, a ~quare matrix that is invertible is called a non-
singular matrix. A square matrix that is not invertible is called a
singular matrix.
The foregoing ideas can be used in solving cert,1in linear sy,tems of
equations. To do this, we first write the system in matrix form.
Suppose that

«11 «11 ... «an

A= and x =

L «.1 «,... ··· «,..n _J L Xn _J

Then Ax is the m x 1 matrix, i.e. a column vector with m entries :

+ «11X1 +, ... + «11,Xn 7
J
r«11X1

AX==
«11X1 + «11X1 + ••• + Gt■11X11

L~1X1 + ~x• + ... + ~.X11

Further, suppose we have m linear equations inn unknowns :
 5 4 MAIR.IX MUL11Pl ICATION / 169

«11X1 -i «111X1 + ... + «,"x n - : Yi

«11X1 + «11X11 + •·• + «:,.n•'°n = YJ

(9)

«1111X1 -f- «,,.sXJ -I -j- «,.,. ,,. - )'m •

The system t9) can be written in the matrix form
«11 ••• «111 1 ,- '\"1 -1 r·
I
1«11 Y1
17 n 'faa ••· «,111 XJ. I Ya
(I,)

L«m1 «ma ... «,,.n .J L Xn _

Il IL }',. _!
or Ax - - y , (II)
where x 1s an unknown column vector and y the g1,en column vector
(y1 • y 2, •• , y,,.) 1• If the S:)stcm has " cquat,on'i m 11 unknown~. then A is
a square matrix, and x and y are n x I mdtrtces.
Further, if A 1.i in,e1tible, then A-• exii-b. Multiplying both sides Clf
Equation ( 11) on the left by ,4- 1• we ~ct
A" 1(Ax)-=- A- 1y
or
(A- 1A)x =-= A- 1y (Theorem 5.4.5)
or J,.x -- A · y
1

or X - A 1y.

Problem Set 5.4

I. Determine the mner product u a• m the followmg ca.,es :
(a) U = (1, -1), V - (2, 3)
(b) u ::.: (1, 2, 3), v - (3, 0, 2)
(c) u:... ( -1, I, 2, 4), v -=- (I, 2, -1, ]).
2. fa•aluate the following products :
r 37
2 37 r 17 rJ 1 2 -17

l:
r1

I I,
-1
,.> 0

0
1
O_j L -1.J
2
(b)

LO 2 1
3 1
0
4.J
I 1
L 2_j
r 17
I
r 17 '

2 2
(c) [2 -1 I 3) (d) (3 1· 2)
-1 -1
l I _J
L O_j L
\10 f MATlUCl!S
rt 17 r 2 7
(e) I I I I·
Li -l_J L I ...J
3. For the given matrices A. and B, determine the product AB or BA,
whichever exists.
ro 17
(a) A=
rt
I
L2
3
:J, B = 13 21

I :_:7
LI 5_1

, I 3 07
(b) A= I I, B =
1_2 1 3_j -S 1

5 77
-1 0 •B
L

=
r·I3

O • 2
2 37
4_1

I
2 l_j L3 7 O_I
s7 r1 2 3 4 57
l,B= I I
l_j LO 1 3 0 l_J
r o 2 17
r1 2 3 0 17
3 S 7
(e) A == I O 1 0 1 0
-1 1 0
l,B=
L3 O 1 l S_j
L O 0 -2_j
r1 17 r-1 17
CO A=I l,B=I I-
LI -LJ L-i - l_j
4. Determine A.· 1 for the given matrix A. in each of the following :
ro 17 r1 27 r« ~7
(a) A =I I (b) A =I I (t') A =I I , ~ -:/- 0
LI O_I LO l_l L~ OJ
r1 o -27
r« 07
(d) A == I
L~ y_l
I , aty :/= 0 (e) A ==
~1 0
2 I
0
2..J
I·
S. Compute A1 , .41, and A' for each given square matrix A.
 5.4 MAntx MULTIPLICATION f 171

(a)
r1
l 0
L3 -1 0.....1
2 37
2 I (b)
r1
I
L.3 1-J
27
I (c> IL
3 2
-1
1
0

s
17

0
7_j
I·
6. Determine all 2 x 2 matrices that commute with
r 1 27 ro 17 r1 07
(a) I I (b) I I (c) I I
L-1 0.....1 Ll -LJ LO 2_j
r 0 17
(d) I I.
L-1 0.....1
ro 17 r3 17 r1 07
7. If A= I I ,B = I 1.c = I I.
L2 l_j L2 -l_J L2 l_j

r 1 27
D=
r1
I
2
37I ,E = l 1 0 I•solve the rollowing matrix
LO -I 1J I
L-1 l_J
equation41:
(a) 3X- AB= C (b) (BXD + AXD)E =- C 1

(c) EXD = I.
rA1 0 07 r1L1 0 0 7
I

8. .A.=
0 A
- 0
,B =
0 I-ta 0

LO 0 A11_J LO 0 !-t11.....I
Compute (a) AB (b) BA. (c) A.- 1, if it exists.
9. An n x n matrix ..4. is said to be nilpotent if A11 = O for some positive
integer n. , The smallest positive integer n, for which .A.11 0, is =
called the degree o/nl/potence of A.
Check whether the following matrices are nilpotent. In the case
or nilpotent matrices find the degree of nilpotence.
ro 0 0 l 2 -17

(a)
•
~ ~ 0
Ly, Y, Y1 0...J
0 0
:1 ~, r: - LO 0 0
0 I
0 0
2
1
0.....1
172 / MATRICES

5 -27 r 1 -3 _4-7

I
r1
(c) I 1 2
I -1 I (dj -1 3 --41,
L3 6 -3.J L. t -3 -4_1
10. If A. and Bare square matrices of the same order, then p,ove that
(a) A.:1 - /J2 -=-= (A -- B)(A .L B) i/J AB = BA.
(b) A.1 ±- 2.AB t Bl -= (A .±. B)2 ijf AB = BA.
f - IJtu IXu !Xia. 1
11. Denote the matrix A = I oc 21 11 11 «za I
L_«,u 'lt3z 11aa - I
r-An Au7
as A =-=- I I , where A 11 --= the matrix [«iJ, A11 =- [«11 «18],
L_Azi A12 I
r-«u -i r«iii «u7
Au= I I, A 21 =I I . Similarly, define B, Bu, •·· by
L«u J Lot.12 «J3 •. i
replacing «" by (i,,. Then pro"e that
r An+ Bu A12 -1- Bu7
A+B==I I
LA11 -f Bu A11 + B22_I
r-A.uBu + AuBu A B12 + Ai:1B117
11
and AB= I 1-
I_AinBu + AaB21 A11 Bu + A21B22_J
12. If«, ~ are any scalars, then prove that A2 - (« + ~).A + a.~/
= (A. - rxl)(A - ~I). where A is any square matrix of order n and
I= I,..
13. If «, ~ are scalars such that A = a.B + ~I. then prove that
AB= BA.
14. A square matrix A is said to be involutory if A.1 =- /. Prove that the
r1 «7 1 07 r
matrices I I and I I are involutory for all scalan
LO -l_J L« -l_J
GC,

Determine all 2 x 2 involutory matrices,

IS. Let T: L(Y,.) ➔ tA,.,,. be defined as in expression (I),§ 5.2. Then
prove that T "1(.4B) = ·T-1(A.)T-1(B) for all A., B € tA,.., ,...
16. Let p(x) = «0 + «1:,c + ... + «,.X" be a polynomial of degree n,
and
A be a square matrix of order m Then the matrix polynomial p(..4)
is defined as p(A) == .•,,I + «1.4 + ... +
v.,.A.11, where/== 1•.
If ft..x) = 7r - 3x + 5, g(x) -·3r - 2x1 + Sx -1,
 5 4 MATlllX MULTIPUCATION / 173
r 1 27 r-1 17
A = I I , and B = I I , evaluate
L-1 J_J L I -2_j-
(a) /(A) (b) g(A) (c) /(B) (d) g(B) (e) /(2A + 3B)
(f) g(JA - 7B).
I 7. Prove Theorem S.4.S.
J8. Using matrix methods, solve the follow mg systems of linear
equations:
(a) x + 2y -= 3 (b) CXX + ~y = a (c) x - 2z = I
y-=l ~x =b 2~ t y =2
X -j- 2z = 3.
19. If A and B are two nonsmgular matrices of the same order, then
prove that AB 1s nonsingular and (AB)- 1 - B- 1A- 1 •
20. Prove that the set of the following eight matrices forms a group for
matrix mult1phcat1on :
rt 07 r -1 07
I -=- 'I I, Di= I I,
LO J_J L_ 0 - )_J

D:a= I
r -i 07
I, Da
,i
= I
07
I,
L 0 j_J L _O -i_J
f 0 17 r o -17
D, I I ,D,.Jt.-- I I'
L-1 0 -
I
I LI O_J
I- 0 -r1 ro ;7
D, = I I, D7 = I I.
[_-i O_J L_i O_j
Write the mult1rhc 1110n tab 1e of the group.
21. True orf albe?
ra b7 ,- -a b7 r-a• +bd ac + bd7
(a) I I I I= I I-
Le d_j L c d_J L - ac + bd c9 + d8 _J
(b) If AB = I, then .A. 1s invertible.
(c) If A and B are square matrices of the same order, then
AB= BA.
(d) lhe system of t\\o equations 2x + Jy == land 4x + 6y, - 2
cannot be solved by the method discussed in this article.
(e) Let A be 2 x 3 and B be 3 x 2 matrices. Further, let C bo
2 ,1(, 2 and D be 3 x 3 matrices. Then
(i) AB is defined but not BA.
(ii) AB and B~ arc both dctlnc4.
 J14 / JIATRICBJ

(iii) .BC is defined but not CB.

(iv) DA. is defined but not AD.
(f) For every integer n > I, there exists a nonzero n x n matrix
which is nilpotent.
(g) For every integer n > 0, there exists a nonzero n x n matrix
which is nilpotent.

5.5 RANK AND NULLITY o, A MATRIX

We have seen that every m x n matrix A corresponds to a unique
linear transformation T: v. - ➔ Vm• where we have agreed to use the
standard bases in both spaces. We shall now show that the value of T at
an arbitrary vector x is just the effect of matrix multiplication of A. by x
in that order.
Let A. == («o)mx•· Let the standard bases in Y11 and V,.. be {e1,
e1 , ••• , e,.} and {/1 ,/,., ••• ,/,,.}, respectively. Jf -~ = (x1 , x 1 •••• , x 11) is an
arbitrary element of v., then x = x 1e1 + x 1e1 + ... + x11e11• Let
T: V11 --> V,,. be the linear transformation associated with A. Then, for
eachj = 1, 2, .. , n,
T(e,) = «i;/1 + «21/1 + ... + «m,/,. = («u, «:u, , .. , «...,) , (I)
because {/1,/:, ... ,/,.} is the standard basis for V,,.. Writing these T(e~)•s
as column vectors, we get
T(x) == T(x1t1 + "•"• + ... + x.e,.) = x 1 T(e1) + x 1 T(,1) + ... + x 11T(,,.)

(2)

r«uX1 -t- «11Xs -1- ... -1- «1.Xn 7

= I «11X1 + «u,Ta + ... + Gtz,.x.
I:
L«...1X1 + «.,.,x, + ... + «,,.,.X11-i
r«u «11 «i11 7 rx, 7

(l)

L-.1 -.. ••11...J Lx• ...J

Tli\1!, tho v~tor T(x) iJ obtaiiiod by niultipl)'iDJ tho matrix .4 b7 tbt
 5,S RANK. AND NULLITY OP A MATaJX / 175
column vector x =-= (xi, x1: ... , Xn)r in that order. In this sense there-
fore a matrix can be identified with the linear transft\rmation it represents.
In other words, if A is an m x n matrix, then the corresponding linear
transformation T : V,. -+ V• represents the operation of matrix multipli•
cation of A by vectors of the space V,. (assuming that we use standard
bases throughout).
Let A be an m x n matrix («.,). Instead of talking of the linear
transformation T associated with A, we talk cf A itself as the lii:ear
transformation, the understanding being that the value of the linear trans-
formation A at x = (x 1 • x 1 , .•• , x,,)T is just the matrix product Ax.
5,5.t De8nition The range, kPrnel, rank, and nullity of a matrix A are
defined as the rang¢, kernel, rank, ard nullity of the linear
transf,.mnation A.
Range of A The range of A : V,. - ,. V,.. is the i.et of all vectors of the
form Ax € V,,., x E V.,. For A ==- («.,). it is the set of all column
vectors of the form

r:::
I =
+ «1n.:\n 7
f Gt1"X,. I
I•
L«,.11X1 -I «.,.zXa + •·• 1 cx,,.,.Xn J
r«1/7 r«u -, 1«111 -1
I
I O'n
I e. + x., : + ... -l x,,

L«.,u_J L«m1 J I_«,,.,,

which are jui.t linear combinations of the column vectors of A.
Obviously, column vectors of A are vecton of V,11 • 1 bus, we have the
foil owing result :
The range of an m x n matrfa is the .•pan of its column vectors.
Rank of .A The rank of a matrix .A is the dimension of the range of A,
i.e. the span of column vectors of A. This shows that the rank of .A is the
maximum number of linearly independent column ve, tors of A.
Example J.14 Let
3 1 27
2 -1
A=
1
0 ::J
 l 16 / MA TRICES

The range of A is [(3, 2, l, O), (1, - J, J. I), (2, 0, l, 2)). The rank of A
is 3, since the three vectors listed are u.
Example 5.15 Let
r3 I 4 07
A= 0 2 2 0
l_l - I 0 0 _J
We ha\e R(A) - ((3, 0, I), (I, 1, -1), (4, 2, 0), (0, 0, 0)) = ((3, 0, I),
(1, 2, -1), (4, 2, O)J = ((3, 0, I), (I, 2, - l)J, since (4, 2, O) = (3, O, I)
+ (I, 2, - I). The rank of A is 2, because (3, 0, I) and (1, 2, -1) are LI.
Kernel of A The kernel of A is the set of all vectors x E V,. such that
Ax= 0,.,,<1.
/1.'ullity of A The nullity of A is the dimension of the kernel of A.
Example 5.16 Let
r 2 17
A == I -I 2 I.I
L 3 O_l
This is a linear transformation from V2 to V8 • We have
[ - 2 17 I X1 7 I 2x1 + X2 7 r- 0 7

I -I 2 I I
l_ 3 O_J Lx2
J"-' -xi
L 3x1
+ 2x,
_j
= I
L O
O
_j
I·
Thus, Ax = 03 x 1 implies
2x1 + x 8 = (l, -x1 + 2x1 = 0, 3x1 = O. .
This gives x 1 = 0 = x 2 • Therefore, the kernel of A is {O} and the nullity
of A is 0.
Example 5 .17 Consider the matrix of Example 5.15. We have
rx17 13x1 + Xa + 4xa7
I 4 07

2 2 o I Xa \ =:
LI -1 0 O_l x, I I
Lx,-1L X1 - X1 _J
Thus, Ax.= 0 implies 3x1 + x 1 + 4x8 == 0, 1x1 + 2x3 = .0, x 1 - x 1 == 0.
This gives x1 == x1 == -x8• So the kernel of A is the set of all vectors
of the form (xi, Xu -x1 , x,) = x 1{1, J • -1, 0) + x.(0, 0, 0, 1). It is
therefore the span [(I, 1, -J, O), (0, 0, 0, 1)). So the nullity of A is 2.
In Examples 5.14-5. l 7 rank + nullity = dimension o: V11 • This
theorem, which is true for linear transformations between finite-dimen-
sional vector spaces, is also true for matrfoes, because of the identification
9f matri~ as linear transformations in the above disc:ussion.
 .5•.5 llANK. AND NULLllY OP A MATUX / 177
Recall that a linear transformation from v. to v. is one-one lffit is
onto. So an n x n matrix A will be a one-one hnear transformation lff
its range is the entire space, 1.e. iff its rank is n. This means that the
maximum number of linearly independent column vectors 1s n. il other
words, the column vectors of A are LI. Since the foregoing argument is
reversible, we have the following theorem.
5.5.2 neorem A square matri'lc is nonsingular iff its column vector,
are LI,
Example 5.18 Let
l 0

L -1 1
The range of A 1s [(I, I, - I), (0, 2, 1), (I, 3, O)]. For the kernel of A,
consider

This gives x1 + xJ =•0, x + 1x,. + 3x

1 1 = 0, -Xi-! Xs == 0. Solving
these equations, we get x 1 =-· --.x,i, \ 1 = x 8 So the kernel of A. is
[(l, I, -1)) Therefore, the nullity of A 1s I. So the rank of A is
3 - I -=-· 2.
The s.ime result 1s obtained by mspectmg the range of A. The range
of A is [(I, I, - J), (0, 2, 1)], because (1, 3, 0) = (I, I, -1) (0, 2, 1). +
Hence, the matrix A 11, singular.
Example 5 19 Let

A-[: _: ~:]
The ranse of A is [(2, I, 1), (I, 2, -1), (--1, 0, I)]. The rank of A is 3,
because we can easily check that these three vectors are u. Hence, A is
nonsingular. This can also be arrived at by showing that the kernel of A
is {O}.
r1 27
Example 5 20 Prove that A == I I is nonsingular and find its
LO 1.J '
inverse.
The column vectors (I. O)T and (2, J)r arc u. Tberdore, tho matrii
 178 / MATRICES
rat f37_
Then
is nonsingular. Assume that the inverse of A. is I
LY BJ
,,., r1 27 rGt f37 r1 07
I I I I= I I,
LO l_l Ly 3_j LO l_j
r ix + 2y ~ + 2s 7 r1 07
i.e. I I= I I-
L y 3_J LO I_!
So at I- 2y = I, f3 + 28 == 0, y == 0, 3 = I, which gives « = I,
f3 = - 2, y =- 0, 8 = I. It is easily seen that
r 1 -- i- 1 r 1 27 r 1 2 7 r 1 -27 r1 07
I I I l=I 11 l=I l-
LO l_J LO l_J LO l_j LO 1_1 LO 1.J
Hence,
rt
A.-1 =I
LO
Example 5 21 Find the inverse of the matrix
I
I 2 3-1

A =1 0 I 2
L-1 1 1.J
The reader can check that the column vecton, of A. arc
.
and so A. 1s LI
.
nonsingular. To find the inverse, we soh,e the matrix equation
r 1 2 37 r«1 «z «a-1 r-1 0 07

IO I 21 IP, ll, P,J--/0°10.

L-1 I I_! LY1 Ya Ya.J LO O l_J
We get n ne equations in nine unknown"· Solving, we get
I i -l -17
A.-1 = I
L-i
1 -2
I -i
1 j.
Alternate methods (which are more powerful) of finding the iuverse of
a nonsingular matrix will be described in§ S.9.

Problem Set 5.5

Find the range, kernel, rank. and nullity of the following matrices :
1.
r 1 3 27 12 o 17
(a) I -1 7 2j (b) 1L3 1 -l 271(c)
-2 5.J
7 1 2 I
1 0 1 L3 -1 l.J
 S S RANK AND NULLln OF A MATRIX/ 179
2 3 1 2 07 r1 -1 0-1

(d)
0
) -3
3 ·- 1 2 I
2 4 3
II
I
(e)
2
l
3 -1
5 2
1 l
0 \
I
L2 3 0 3 OJ L_O 0 1 I _I
r-1 1 17
I
(f) 3 1 -1 I
IL 2 2 1J
2. Prove thc1t the following mdtrtce-. are nonsingular and find their
mverses:
( 0 0 ,-1 -1 17
r I 27 I
(a) I
L-1 O.J
I (b) I2
I
I
0
I
(c)
I3 0 1
I
L.3 4 ">
.. ..J
I
LO 1 1
r-1 2 37 2 2 I

(d) I0 2 1
(e)
0 ~ 1

II 3 -I 0 0 l
LI -I O_J LO () 0
3 Fmd the values of oi: and ~ for wtuch the followmg matrix 1s inverti-
ble. Fmd the inverse when 1t ex1st'i
101: ~ 07

0 «_J
0
L~
°' ~ I·
4. Prove the following :
(d) If two rows of a matrix are interrhanged, then the rank doe,
not change
(b) If a row of a matrix 1s mlllt1phed by a nonzero scdlar, then the
rank does not change.
S. True or false ?
(a) The range and kernel of a square matrix are of the same
dimension.
(b) There exists a 7 x 12 matrix whose rank is 10.
(c) If two columns of a matrix are interchanged, then the rank does
not change. ·
(d) If a column of a matrix is multiplied by a ponzero scalar, then
tbe rank does not change,
J 80 / MATRICl:S

(e) The rank of an m x n matrix having r rows of only zeros is

less than or equal to m - r.
r1 2 3 47 r1 o 07
A-loo o oi,s=J o o I· 2
LOO O OJ L4 0 OJ
(f) The kernel of A is 2-dimensional.
(g) The kernel of B is 2-dimensional.
(h) The ranks of A and B are the same

5.6 TRANSPOSE OF A MATRIX AND SPECIAL

TYPES OF MATRICES
In this article the matrices involved may be com,idered rFal or complex.
Let A be an m x n matrix (oc.,). Writing the rows of A as columns
(and therefore the columns of A a& rows). we get a new matrix called the
transpose of A, which is denoted by AT. Actually, the lj-th element of
AT is «,,, as can be &een frc,m the details of the following entries ·
I IXn Otu •·• tXu •·• «1,. 7 I 1111 «111 .•. «,1 ••. «,.1 7
«11 I «u «12 • . • «,1 «,.1 I
A=
:x,. • Ar =-

I
L«,..1 «,..1 ... ar:,.., a.,.n_l Lo-111 «.,. ... «,,. ... «,..,.J
Clearly, AT is an n x m matrix whose ij-th element 1s ar:,.. the ji-th
element of A. It follows immediately from the definition of transpose that
P. = I,. and O~x. = O,.x,... •
5..6.t Theorem If A and B are two m X II matrkes, then
(a) (..4 + B)T = AT + BT,
(b) (aiA)T = aiAT, and
(c) (.A.T)T = .A..
The proof is straightforward and left to the reader.
5.6.2 Theorem If A is an m x n marrix and B Is an n x p matrix, then
(.AB)T == BTA.T.
Pre of : First, note that AB is defined and is an m X p matrix. So
(.A.B) 1 is a p x n matrix'. Again, Ar is an n x m matrix and BT is a
p x n matrix. Therefore, we have p x m matrices on both sides of the
octuality in the theorem and so the equality is mcaninsful.
 S.6 TRANSPOSE OF A MATID: / 111

To prove that the equality ltolds1 it suffices to prove that the ij-th
element or (AB)T is equal to the ij-th element or BTA.T. Let A.= («.,>..x.
and B := (~,,)11Xt1• Then AT = c«:1)11)(111 and BT = (~;,).,x11, where «~
= "" and = ~~, = ~,,. The ij-th element of BTAT i:1
n n
t ~~Glflll :::. .& Gl1t~k1 ,
k=I k ... l
which is the ji-th element of AB and hence the ij-th element of (AB)T.
Thus, (.AB)T =- BT.AT. I
S.6.3 Theorem If A is a nonsingular square matrix, then AT is also no,r-
slngular and (A')-1 == (A.-1)T. ·
Proof: Since A is nonsingular, there exists a matrix B such that
AB =I= BA. Therefore,
(AB)T = JT = (B.A)T .
This gives BTAT = I--= A BT. Therefore, .A has the mverse, namely,
BT, and (.A.1)· 1 = BT ,... (.A-1)T, because B =- ,4-1 . I
5.6.4 Corollary The· columns of a SCJUare matrix A. are LI if/ its rows
are LI.
Proof· Columns of A are LI .A is nonsingular (Theorem S.5.2)
A is nonsingular (Theorem 5.6.3) -e:> columns of AT are LI (Theorem
-e:>
5.5.2) -e:> rows of .A are LI (definition of AT). I
We shaU prove in Theorem 5.7.5 that in any matrix, not necessarily
square, the maximum number of linearly independent columns is the same
as the maximum number of linearly independent rows.
Before we conclude this article let us familiarize ounclves with certain
special types of matrices.
A square matrix of the form
rA1 0 0 ... 0 0 7
0 A.0 ... 00

0 0 A. .:. 0 0

o o o ... o~.
is called a diagonal matrix. In other words, A = (1111) 11x• is a diagonal
matrix if «u = 0 whenever i :;: J, i.e. «u == .,.._Bu,
A square matrix of the form
182 / IIATRICBS

r >. 0 ... 07
0 ·A ... 0
:': : =='Al

LO O ... A_J
is called a scalar matrix. In other words, A = (at")11x11 is a scalar matrix
if 11.11 == >.811 for a fixed scalar 'A.
A square matrix of the form
r«u at11 at11 ... Gt111 7

0 IX111 «113 ••• «111

0 0 «aa ... «an

LO 0 0 ... «1111_J

is called an upper triangular matrix. In other words, A = (:x,,)nx11 is an

upper triangular matrix if «u = Owhenever i > j.
A square matrix of the form
IX11 0 0 ... 0 7
at11 Glas O .. · 0

Latn1 at•• «111 ... «... _j

is called a lower triangular matrix. In other words, A. = (11 1,)11x11 is a lower
trianauiar matrix if at 0 = 0 whenever I < j.
A square matrix A. is said to be Idempotent if A.1 = A..
A square matrix A is said to be symmetric if A. = AT, i.e. «u == «11 for
all l,J.
A square matrix ..4. is said to be skew-symtMtric if A = -.4. , i.e.
""' == -ex,, for all l,J.
IC A = («u).x11, the matrix B = («u)111x11• where the bar denotes
complex conjugation,· is caned the conjugate of A. It is denoted by A.
The matrix (A)T is called the transposed conjugate of ..4. It is denoted
by A•. Obviously, (A)'= (.41).
A matrix A is said lo be Hermitian if A == .4*, i.e. if «11 = i 11• It is
11id to be llcew-Hermltian if A= -A•, i.e. «11 == -«11 • Por example,
 .U DANSPOSII Of A MATIUX / J83
r o r
I
L-i O_J
17
I and I
1-i7
LI+ i 2
1
J
are Hermitian matrices

ro 17
and I I is a skew-Hermitian matrix.
LI O_j
From these t!efinitions of special matrices, we get the following results :
(1) The product of two diagonal matrices of the same order n is again
a diagonal matrix of order n and
r"lo O •. • 0
>-. ..• o
711-'ol O •• • 0
1,1,a ••• o
7I 0
I=
. • . . I
• : : : • I
I
LO O >..,._J LO O 11,11 _J

(ii) Any two diagonal matrices of the same order commute with each
other.
(iii) A diagonal matrix is nonsmgular iff none of tl:e diagonal entries 1s
zero.
(iv) The in~erse of the diagonal matrix
0 ... O 7 r-1r>-1 0 .. 0 7

[ >-a

0
.

LO
0

0
A,._J
A, /:- 0, i = 1, 2, ... , n.
l'i

1/>n_J
0 l/'>-1. ..

..
0

(v) The transpose of an upper (lo~er) triangular matrix 1s lower

(upper) triangular.
(vi) In a skew-symmetric matrix all the diagonal entries are zero.
(vii) In a Hermitian matrix all the diagonal elements are real.
(vili) Multiplication of a matrix ..4. by the scalar matrix >./ is equivalent to
multiplication of ..4. by the scalar >..

Problem Set 5.6

1. Prove Theorem 5.6.l.
r l 2 37 r 1 17.
2. ..4. = 1-1 1
L3
I
2 ,B= Ir1
0
-1
-17
2J•C=
·1
-• I·
L ~ I -2_J L-2 3_J
J 84 / IIA 'IJUCES

D = r: -:7J,E= rl-: : :7J

L3 2 1 J 3
For these matrices, evaluate the foil owing :
(a) AT (b) BT (c) ABT (d) (CB)T (e) (A.,C) 1 (f, BA 1
(I) CTBT (h) B + DT (i) (C + D)T.
3. For the matrices in Problem 2, verify the followmg :
(a) (A+ E)T =- AT+ ET (b) {D'(C + B1)}T = (C 1 + B)D
(c) (3C + SD)T = :CT + SDT (d) (AC)T = erAT
(e) (CD')' = ocr.
4. For m x n matrices A, B, and C, prove that
(a) (AB')T = BAT (b) (A.TB)T = BTA.
(c) ((A + B)C1)T = C(AT + Bf) (d) ((A :t- B)TC)T = CT(A -t- B)
5. Determine A.• for the matrix A in the following :
r 1 +, 3 - ;7 r1 - , 2 - 3; J i ·7 +
(a) I I (b) I I
L - t -- I 2 - i_J L2 + i 1 + 4i 3 - 2i_J

(c)
r2
I •-
+ 3; 2- ; 7
2i 2 - 2i (d)
r- 1
I 2 - i
1- ;
; 2i=/l
L3 + 4i 2-ti_J LI +i 2 +; 3 + 2i_J
1a h g-7
6 .. If .A = Ih b f I and X -=--a (x, y, z), evaluate XAXT.

Lg f c_J
7. If A is an m x n matrix with complex entries such that AA• = 0,
prove that A = A = o.
8. If A and B are square matri~s of the same order and A. is symmetric,
prove that BTAB is also symmetric.
9. (a) Prove that a triangular matrix whose leading diagonal entries are
all zeros is nilpotent.
(b) Determine a matrix that is both upper and lower triangular.
10. Let A be a square matrix. Then prove that
(a) A + A• is Hermitian.
(b) .4 - ...4.• is skew-Hermitian.
11. (a) Prove that every square matria: can be expressed as the sum of
a Hermitian and a skew-Hermitian matrix.
 5.7 ELEMENTARY llOW OPERATIONS/ 185

(b) Prove that every square matrix ca11 be expressed as the sum of
a symmetric and a skew-symmetric matrix.
12. P,ove that the set of all n X n diagonal matrices is a subspace of
tA. •• •·
13. Prove that the set of all n x n invertible diagonal matricei. is a group
under multiplication.
14. True or fab,e ?
(a) The set of all n x n diagonal matrices 1s a group under multi-
plicat1on.
(b) If A 1s mvolutory, then its transpose is also involutory. (A is
mvolutory means A 2 = /.)
rO I 27
(c) The matrix - 1 0 3 1s skew-symmetric.

'- -2 -3 l_J
(d) Every !tyhlmelnc matrix is Hermitian.
(e) A nonLero idempotent matrix 1s not nilpotent.
(f) An idempotent matnx A 1& singular unless A = I.

5.7 ELEMENTARY ROW OPERATIONS

In this article we shall study certain operations on a matrix that help us
to determme ll'I rank. First, we start with the familiar method of solving
a system of s1mult:meous lmear equations in three unknowns.
Two systems of (simultaneous) linear equahons are said to be e1JUi•al-
11nt (an symbols, e:t ) 1f they have the same set of solutions.
Let us solve the system
2x- 3y + z = -1
3x + Oy + z -= 6 . (A)
X + 2y - 2z == -1
Interchanging the first and third equations, we get the system
x+ 2)' - 2z = - l (El)
3x + Oy + z == • 6 (Bl)
2x - 3y + z == -1 (B3)
*(Bl): x + 2y- 2z - -1 (B4)
(El) x (-3) + (E2): Or - 6y + 1z - 9 (BS)
(Bl) x (-2) + (B3) : Ox - 1y + Sz - 1 (E6)
* (E4): x + 2y - 2z = -1 (B7)
(ES) x (-l): Ox+ y-fz - -I (ES)
(E6) : Ox - 1y + 5• - l (E9)
186 / MA.TRICBS

0
(E7): X + 2y - 2z == -1 (BIO)
Ox+ y- ¼z = -t
.
(ES): (Ell)
(El) X (7) + (E9) Ox + Oy - 1/z == -Y (El2)
(EIO): x+ 2y- 2z == -1 (B13)
(Ell) Ox+ y .... h=-t (B14)
(El2) X (- i\) Ox+Oy+ Z= 3 (EJS)
0
(1:\13) : Ix+ 2y- 2z = -1 (EJ6)
(EIS) x U> + (El4): Ox+ ly + Oz= 2 (E17)
(EIS): Ox -f- Oy + lz= 3 (El8)
0
(El7) X (-2) + (El6) Ix+ Oy - 2z = -S (EJ9)
(E17): Ox -t ly + Oz= 2 (E2O)
{EIS) Ox+ Oy + lz = 3 (E21)
0
(Ell) x (2) + (J::19) 'Jx + Oy + Oz= I
(E2O) Ox + ly + Oz = 2 (B)
(E21) Ox + Ov + lz = 3
which gives x = J • y = 2. and z = 3.
It can be seen that throughout we have worked with only the coeffi-
cients in the equations and the numbers on the right-hand side. The pre-
sence of the symbols x, y, and z and the sign of equality does not at all
atrect the working. Now, deletmg the symbols x, y, and z and the sign of
equality from the pattern, which otherwise remains unchanged, we shall
write only the numbers involved in the form of a matrix A. Repeating
the foregoing sequence of steps, we get the following sequence of matrices,
where the symbol - is an analogue of the symbol o, and the letters r1• r1 ,
and r3 stand respectively for row 1, row 2. and row 3. Further. r 1 + kr1
means 'add k times the j-th row vector to the I-th row vector' • We have
r2 -3 1 -17 11 2 -2 -17
matrix A = 1' 3 0 1 6 Iinter;ange
r1 and , 1
3 0 1 6 I

r
LI 2 -2 -l_J L2 -3 1 -l_J
l
,,+£ o
-2 3i,, -6
2
7
-17 [l
91 •• x(-t) o
2-2 I I -i -I
-17

'• + (-l)ri LO -7 S l_J O -7 ! l_J

r1 2 -2 -1 7 r1 2 -2 -17

,, + Io
7, 1 1 -i· -t j ,, x C-1") lo 1 -f -t I
LO O -¥ -¥ _J O O 1 3..J
 ::, 1 ELEMENTARY ROW OPERATIONS/ 187
r1 2 -2 -17 r1 0 -2 57

r8 ,. ., lra I 0
+ 1 0 2 ! ,...,
rt+ (-2)r2 :
Io 1 0
2 I
LO 0 3..J LO 0 1 3..J
rt 0 0 17

'1 .-+. . 2ra I 0 1 0 2 I- matrix B.

I 0 0 1 3 I
1 he systems (A) and {B) have the same solution, hecam,e one has been
obtained f1om the other by certain operations that do not affect the
solution set. The same operations when rerforrned on the rows of matrix
A produce matrix B. 1 hese Clperations are Clllled elementary row
operations. 1hey are of three typei. :
Type I Interchanging two rows.
Type II Multiplying a row by a nonzero scalar.
Type III Add'mg to a row a scalar-times another row.
S 7.1 Definition When a mJtrix A 1s subjected to a finite number of
elementary row operations, the re,;ultmg matrix B is said to be
row-equivalent to A. We write this as B,...., A.
A moment':. thought will show that ,_, is an equivalence relation.
5.7.2 Definition 1 he row rank of a matrix A is the maximum number of
linearly mdependent row vectors of A.
5.7.3 Remark In this sense the rank of a matrix A, defined in § 5.5, n1ust
be called the column rank of A. But ultimately (see Theorem 5.7.5)

0 , 2 0 2 0 2 0 0 0 0

0 0 0 1 -1 0 3 a a o -2

0 0 0 0 0 1 0 0 0 0 2

0 0 0 0 0 0 a 0 0 0

a 0 0 a 0 0 0 1 0 ~

0 0 0 0 0 0 0 0 0 .._
1 __ 2

0 0 0 0 0 0 0 0 0 0 0
188 / MATIUCl!S
we shall prove that for any matrix A. the column rank is equal to the
row rank; so we can afford to call both of them simply 'rank'-. Till
then,' we shall have to distinguish between the row rank defined in
5.7.2 and the rank(= column rank) defined in§ S.S.
The aforesaid process of obtaining matrix B from matr;x A is called
the process of row reduction. In practice \\ e usually aim at getting B in a
standard form called row-reduced echelon form.
Consider the following 7 X 11 matrix and the 'stairs' or 'steps' marked
therein:
Such a matrix is said to be in row-reduced echelon form. We shall
now give its precise definition.
5.7.4 Definition A matrix is said to be in roH"-reduced echelon form if it
satisfies the following criteria :
(a) The fir!i.t nonzero entry in each nonzero row is 1.
(b) If a column contains the first nonzero entry of any row, then
every other entry in that column is zero.
(c) The zero rows (i.e. rows containing only zeros) occur below all
the nonzero rows.
(d) Let there be r nonzero rows. If the first nonzero entry of the
i-th row occurs in column k, (1 =-= 1, 2, ... , ,·), then k 1 < k,.
< ... < k,. .
Draw horizontal and vertical partition lines such that below and left
of these lines there are only zeros and such that at the point where the
vertical line 1s followed by a horizontal line there are 1's, namely, the first
nonzero entries of nonzero rows. These turnmg po111ts are called steps.
See the foregoing 7 x 11 matrix, where there are s•x steps.
We now state the main theorem of this article.
5.7.5 'fteorem (al The ,ow rank and column rank of a matrix A are the
~e,me. In other words, the m:iximum number of /im,arly independent
row vectors is equal to the maximum number of linearly independent
column vectors and is equal to the rank of the matrix.
(b} The rank of a matrix A. la the number of nonzero row, In its row-
reduced echelon form.
 5.7 ELEMENTARY BOW OPERATIONS/ 169
We shaU prove this theorem by establishing a 1.uccession of lemmas.
The sequence of proof can be easily understood from p:gure 5.1 :

Row rank of A Column rank of A

CD
to,:
II)
-...
r,..
tri

II I• ..J

.:
Row rank of 8 Column rank of B
• 4-
by Lcll'm!l 5 710
I\~
<.,~~
Q
d'.,.
.> '0~
.~
~C)(,:,

-..;

~umber of Number of
nonzero rows steps
of B ,n B

A. Given matrix

,.,, ,.,
B: Row reduced echelon form of A <Lemma 5. 7 &,

FIGURE 5.1
5.7.6 J.emma Ei·ery matrix A i$ roM•equivalent to a row-~educed echelon
matrix.
In other words, every matrix can be reduced to the row-reduced
echelon form by a finite sequence of elementary row operations.
We shall omit the proof of this lemma, as careful scrutiny of the
process of row reduction will convince the reacer that this lemma is true.
5.7.7 Lemma // a matrix is in tl,e row-reduced echelon form, its row rank
is the 11umber of nonzero rows in it.
The proof is left to the reader.
5.7.8 Lemma 7he row rank of a matrix A is equal to the row rank of the
row-reduced echelon matrix B, obtained from .4.
Proof: B has been obtained from A by a finite sequence of
elementary row operat:ons. We .shall show that these row operations do
not affect the row rank of A. It is clear that the row operations of type
I and type Udo not alter the row rank (see•Problem 4, Problem Set 5.5).
It is therefore enough to prove that an elementary row operation of type
III does not change the max:mum number of linearly independent row
vectors. Supposo we Jdd «-times a row vector •i to rnot)l~r row vector
 I 90 /MATRICES
v1 • Let the other row vectors be , 9, ,,, •· , v,... Examine the two sets of
vectors
P = {111 , 112 , ••• , v,,,} and Q = {v., v8+ix1'i, V3 , ... , v,..} •
It can be easily checked that (1) ,f P is u, I hen Q 1s also LI and (ii) if P is
LD, then Q js LD. This shows that a type Ill oper,uion does not affect the
maximum number of UnearJy independent row vectors, and so the row
rank is unaffected by such an operation (sec Problem 9). I
5. 7.9 Lemma // a malrbc is in the row-reduced echelon form, the,, its
.
(column) rank is the number of ·steps' in it .
rroof: Let the number of steps hep. Every column that occurs
hefore the first step is a zero vector, and every row that occurs after the
last step is also a zero vector. So the nonzero column vectors can be
con,;idered vectors in V,. Therefore, the column rank is less tha1i or equal
to p. On the other hand each step contributes one nonzero column
vector. The set of these column vectors is LI. This means the column
rank is greater than or equal to p. Thus, the column rank is p. I
5.7.10 Lemma If a matrix is In the row-reduced t!chelon form, then its
(column) ro,rk is equal to its ro,v rank.
Proof: Note that the number of steps in such a matrix is equal to
the number of nonzero rows. Using lemmas 5.7.7 and 5.7.9, we find that
its (column) rank is the number of step~ m it, i.e. the number of nonzero
rows, i.e. itc; row rank. I
5.7.11 Lemma Tlie (column) rank of a matrix A i, equal to the (column)
.
rank of the row-reduced <che/012 matrix B, obtained from A .
Proof: To facilitate understanding cf this lemma, wc shall prove
it for a matrix lf 01der 3 x 4. By using proper notations, the same
proof can be extenc!ed to any m X n matrix.
We have only to prove that the th1ee types of elementary row
oper.ations do not affect the column I ank. The fact that types I and II
Ol}lralions do not affect the column rank is easy to prove and is left to the
reader. To prove that a type Ill operation does not affect the column
rank, com,ider the matrix

: :: : :~l
L.a.31 11.az «aa «u.J
Let us pe1form a type III operation on P1 namely, let us add ix-times row
3 to row 2. The resulting matrix Q 9,ifl be
r «n

I
IXJS

a= + CI.IX31
ixa1 IXaa + ««11 «.a + ix«aa
1.- «.1
 5.7 ELEMENTARY ROW OPERATIONS / 191

Suppose the column 1ank of Pis 3 (any other value of the rank can be
handled analogously). Then in P there exist three linearly independent
column vectors and every four column vectors are LD. We shall prove
that the same phenomenon occurs in Q also.
Let the three linearly independent column vectors of P be the first three
(1f they are any other three, they can be brought to the first three positions
without affecting tl•e argument). Let us call them C1, C1 , and C3 •
If Fi, F2 , and F 3 are the first three column vectors of Q, then the vector
equation ~1 F1 + ~2F2 +~ 3 F3 = 0 gives
~1 1111 + ~21Xu + ~3IXJ3 = 0
81(IX21 + IXIX31) + f32(IX22 +. 0'0'32) + ~3(0'23 ,- IXIX33) = 0

~10tJ1 + ~,ocaz + ~.11X33 = 0 ·

Adding ( -1X)-times the third equation to the second equation, we get
the equivalent system
~10'11 f- ~3IX12 + ~JIXt3 = 0
~1IX21 + ~21X22 + ~a1Xaa = 0
~1IXa1 -j- ~21X32 -j ~,10t33 = 0 •
This system is nothJng but the vector equation ~1C1 ~aCa +.~3Ca = +
O. Since Ci, C2 , and C3 are LI, we get ~1 = 0 = ~" = ~3• Thus, Fi, Fa,
and F3 are u.
Agam, ~e prove that when Ci, C1 , C 3, and c, arc LD, then a.1.1tomati•
catly F1 , F,, F 3, and' F4 are also LD.
The vector equation ~1 F1 -+ ~"F2 -f ~.1Fa f ~,F, = 0 gives
~ 1(111 + ~.&ixa -I ~Joc 13 + ~41Xu = 0
• ~L(ixu + IXIX 31) + ~b:n t- uu) -t ~a(0t2a 1 IXIXaa) + ~&(ocz. + 0to.3J = 0
~LIXJl T ~2'132 _J_ ~3IX31 t ~40t34, = Q•
Adding (-1.1t)-times the third equation to the second equation, we get
an equivalent system which is nothing but the vector equation ~1C1 +
+ ~3Ca + ~.c, = O.
~.c.
Since C 1, C 2 , Ca, and C, are LD, at least one~ is not zero. Hence, F 1,
F.,_, F 3 , and F4 are 1.0.
Thus, we see that the type III operation does not affect the column
rank of matrix P. I
Proof of Tl1eorem 5.7.5 : (a) let B be the row-reduced echelon
form of A. Then ·
row rank of A = row rank of B (Lemma 5.7.8)
= column rank of B (Lemma 5.7.10)
= column rank of A (Lemma 5.7.11).
(b) Rank of A = column rank of A {Remark 5.7.3)
= row rank of A (part (a) of this theorem)
= row rank of B (Lemma 5. 7.8)
;:::: nu1Dber of nonzero r9ws in B (Lemma 5.7.7). I
 J92 / MATRICES

5.1. IZ Corollary An n x n matrix A is nonsingular iJI its row-, educed

,cl,,/on form is /,..
Proof: Let the row-reduced echelon form be 1.. Clearly, its
column vectors are u. So the rank of / 11 is n. Hence, by Lemma 5.7.ll,
the column ranlc' of A. is n, i.e. column vectors of A. are u. Therefore, by
Theorem S.S.2, ..4 is nonsingular.
Conversely, let A be nonsingular. So the rank of A is n. Therefore,
the row-reduced echelon form B of .4 will heve n linearly independent
column vectors (Lemma 5.7.IIJ. Since the number of columns inB is n,
it follows that B = 1,.. I
Example 5 12 Determine the rank of the matrix
r 1 2 -1 o7 I

A=
-1
2
3
I
0
3 -2
-41
I
L 1 J I - I.J
by reducing it to row-reduced echelon form

A.= l. _: : -: _:7 . +... ·:

2 1 3 -2 2' 10
'• -- Ti 0
2

.
S -I
-3
-1

S -2
07
-4

r,
1 1 I -I_J LO -1 2 -I_J
r1 2 -t 2 -1 07
,...,
inter•
changing
, 1 and r,
0 -1
0 -3
: -~
7
j,,x(-1) 1
:
1 -2
-3 5 -2
I

LO S -I -4 LO S -:• -4.J
2 -1 07 r1 2 -I 07
,..., 0 1 -2 1 ,._, 0 -2 I
r1 + Jr1, r, t 9r3
r , - fr1 0 0 -1 1 0 0 r-1 1
LC 0 9 -9.J LO 0 0 0.J
r1 2 -1 07 r1 2 0 -1,
0 -2 0 0 -1
r1 X (-1) 0 0 1 -1 0 0 l -1

LO O O U,.J 0 0 0 O
 5,7 ELIIMENTAllY llO\\' OPBllATIONS / 193
1 0 0 17
....., 0 l 0 -1 I
'1 - 2,. 0 0 1 -1 I
0 0 0 O_J
Hence, the rank of A is 3.
Recall the explanation at the beginning of this article. The row reduc-
tion process is just an abstract imitation of the elimination method we
adopt in solving simultaneous linear equations, except that we now do it in
an organised way. The row reduction worked out in Example 5.22 is there-
fore also the method of solvmg the system
X -j 2y - - Z = 0
-x + 3y = -4
2x I- y + 3z - -2
X Z = -1, + )I+
giving the solution x =- I, y -- -1, z -=- -1.
This takes us to the subject of the solution of a syi.tem of linear equa-
tions, which we st-all deal with in § 5.8. '

Problem Set 5. 7

,~
1. Reduce the following matrices to the row-reduced echelon form :
-1 17 rJ 2 3 17
(a)

L3
-1
1
2
1 _J
I (b)
I4
L2
3
1
5
I
2
O_J
I
r1 3 2 07
r1 2 3 4 -17
2 0 1 I
{c) I 4
1
0
s
2
--6
-2
101
6_J
(d)
I
2 3
0 I
L2
L3 -1 0 l_J
r-1 1 1 07
r1 2 0 07
I I I
l:
1 -1
(e)
0 I 2 -1 II
I
(f) 1

2
-1

1
_;J
L 1 l 0 - 2_J
'r3
I
1 -17 r o 6 6 17

I 2 3 -8 .7 2
(8) (h) •

b.1
4 0
s
I
3.J b. 1
-3 2
l -1
l
:J
 JH / IIATRICD
ri o 1 97
5 2 2 JO
(i)
3 -2 3 11
L2 -1 3 8_J
2. Determine the column rank of the matrices in Problem I by two diffe-
rent mMhods.
3. Determine the row rank of the matrices in Problem 1 by two different
methods.
4. Determine which or the square matrices in Problem 1 are nonsingular
and in each case find the inverse.
S. Solve the following systems of linear equations by using the row•
reduction method :
(a) 'J X - Jy - l (b) y- 2z = 3
2x- Y+ Z= 2 3x + z=4
3x+ y-2z= 1 x+y+ z-=-1
(c) x- y+ Z= 0 (d) x - y + 3z = 1
2x+ y- 3z= 1 2x+y- Z= 2
-x+ y + 2z = -1 3x - y + 2z -== 2
(e) x+ y-2z= 3
3x+ y- Z= 8
2x- y+ Z= 0.
6. Prove that •,...,• is an equivalence relation.
7. In the 7 x 11 matrix on p. 187 find the numbers k1, k1, •• , k, or
Definition S.7.4.
8. Prove Lemma S.7.7.
9. Write down the details of the proof that the row operation of type Ill
does not affect the row rank of the matrix.
10. True or false 1
(a) For any matrix .tC, the ranks of ..4. and .A.1 are the same.
(b) The row-reduced echelon form of a diagonal matrix ..4. is .4 itself.
(c) Every upper triangular matrix is row equivalent to a diagonal
matrix.
(d) There exists a 12 x 7 matrix of rank 10.
(e) ne row-reduced echelon form of a symmetric matrix is also
symmetric.
•
(f) The type DI operation performed on the columns of a matriJ
~ves its rank unchan,ed.
 S.8 SYSTEMS Of LINEAR EQUATIONS / 1'5
5.8 SYSTEMS OF LINEAR EQUATIONS
In this article we shall study the general system of m linear equations in
n unknowns.
«11X1 + at11X1 + ,. • + «111X11 = b1 }

-~~:~~-~-~~-- ~~--~.: :..~..:~~~--~=..~~--

«1111X1 + «maX1 + ... + «mnXn bm
. (NH)

This is-called a nonhomogeneous system (NH) if at least one b 1 is not zero.

The associated homogeneous system (ff) is
«uX1 + «uXa + •·· + «1,.Xn = 01
«uX1 + 1Xz~X2 -1 •.. + IX2nX11 = 0 ~ •
(H)
IX:~~~•+••~~:~~•;•:::.+-~~~;~•~••() j
These systems can be written in matrix notation as
Ax= b (bi= O)
and Ax-= 0,
r IXu «12 «tn 7

where A=

L«m1 «m2 Otm,._1

X = x,.)T E Vn, and b = (bi, ba, ... , bm)T € Vm,
(Xi, X1, ... ,
This operator equation is to be solved for x. The matrix A is called
the coefficient matrbc of the sy!>tem. The matrix obtained by adjoining the
column vector b, at the end, to the matrix A, is called the augmented matrix
of the system (NH) and is denoted by (A, b). As usual, we consider A a
linear transformation from V,. to V.,.. By Theorem 4.8.1, we have
(i) Ax = b has a solution iff b E range of A. This is equivalent to
saying (cf§ 5.5) that b E span of the column vectors of A. In other words,
A and the augme::ited matrix (A, b) have the same rank.
(ii) If Ax= b has a solution, then the solution is unique I.ff Ax 0 =
has the trivial solution x -= 0 as its only solution. This happens ijf the
kcmel of A is {O}, i.e. ifJ nullity of A is zero. This is equivalent to saying
that the rank of A is n.
Thus, we have proved the following theorem.
5.8.1 Theorelll (a) (Existence) TM system (NH) ha, .a solution' IJf the
matrix A. and the augmented matrix (A, b) ha,e the ,ame rank.
(b) (Uniqueness) If tlw ,y1tem (NH) has a aolutlon, th.n the so/11-
IIOn f6 .,,,.,. IJf th, rmrk o/ A f6 equal to 11,
 196 / llfATRICES

This theorem and Theorem 4.8. I, properly interpreted for our system,.
of equations (NH) and (HJ. give the following comprehensive theorem.
5.8.2 Theorem Consider the systems of equations (NH) and (H). Let the
rank of A be r. Then
ta) (NH) has a solution iff (A, b) has rank r.
(b) If r - m, then (NH) always has a solution, whatever may be b
E ~- •
{c) If r - m =-= n, then (NH) has a unique solution, whatever may
be b E V,,.; and further (H) has a unique .•olution, namely, the trivial
solution.
(d) If r •- m < n, whatever may be b E V.,., (NH) as well as (H)
ha,•e an infinite number of solution:r. Tn fact, r of tht: unknowns can
be determined in terms of the remaining (11 - r) unknowns, whose
,,afues can be arbitrarily chosen.
(e) In the cases (i) r < m =--= n, (1i) r. < m < n, and (hi) r < n <
m, if tNH) ha, a solution, then there is an infinite number of solutions.
In fact, r of the unknowns can be determined in terms of the remain-
ing (n - r) unknowns, whose values con be arbitrarily chosen. Further,
(H) hos 011 infi,ute number of solutions.
(f) In the case r -- n < m, (H) has a unique solution, namely, the
trMal s lution, and if (NH) has a sol11tion, then that solution is unique.
(g) If m :..... n, (H) hos a nontrivial solution iff A is singular.
Proof: Parts (a) and (f} are only restatements of Theorem 5.8. I.
They arc included here for completeness. Part (b) follows, since r = m
implies the range of A is V,,.. Part (c) follows from part (b) and Theorem
5.8. I (b). Now we ha\e to prove only parts (d), (e), and (g).
To prove part (d), first note that, since r = m, the range of A is V,..
and so every b E V,.. has an A pre-image in V,.. The kernel of A has the
dimension n -- r > 0. So the kernel K, being a subspace, has an infinite
number of \ectors in it. By Theorem 4. 8.1, it follows that the solution set
of (NH) is a trans1atc of K. So it also has an infinite number of vectors.
The fact that r unknowns can be determined in terms of the remaining
(n - r) unknowns can be seen from the row reduction process of A (see
Example S.23).
To pro\e part (e), note that the proof is the same as that of part (d),
except that, first, we should know whether there exii.ts a solution. Once
the solution exists, the rest of tte argument is tJ-e same.
Finally, part (g) follows from parts (e(i)) ar.d (c) once we recall that an
n x n square matrix is singular iffits rank is less than n. I
The result of this theorem cln be presented as in Table S. I.
Let the rank of matrix A be r and that of the augmented matrix (.4, b)
be , 1• Further, Jct the rank of the kernel be k = n - r. Obviously, r <:
r1, r <: m , r <; n. (Numbers within brackets refer to the pa'11 of
Theorem S.8.2.)
 $ 8 SYSTEMS OP UNBAR EQUATIONS/ 197
TABLB 5.1

----------The system has no solution

-- --------
(a)

The system has a solution and the following hap-

pen&.
Homogeneous system
(H)
I (a)
Nonhomogencous sys-
tem (NH)
- --.--- ---- -~- !x_ = _O___ Ax-b,b+O
r = n, k -- 0 has only the trivial I bas the unique solu•
&olutJon ('-) I t1on (c)
_ _ _ _ _ _ !_ _ _ _ _ _ _ __
---- I
has mfimteQunany sol• I ha• mfi.nJtely many
ut1ons Solution space solutions ~olut1on
11 of d1men&1on I
set 11, a hncar VdrJety
m = n r < n, 1 wbo,;e base spctce JS ot
k n r ;;> 11 k = n - r (e(t)) l d1mens1011 J., (e(1))
-
1 We can find r unknowns m terms of the remam•
I mg A. ( 11 - r) unknowns, chosen .arb1trartly (e)

r =- n, k -= 0 has only the trlVlal sol• has the unique solution

ut1on (t) (f)
------·------ - -
has mfimtely many sol- ha• mfinitely many sol-
utions Solution ,;p.a"e ut1on!> Solution set 1s
1s of d1mens1on .i hnear variety ...,hose
m >n r . ___ n, bai.e <,pd.(..C I!> of d1men-
k = n - r;;>J k (""- n - r) (e(111)) s1on k (e(m))
- -
We can ftnd r unknowns m term!> of the rema•'I\•
mg k ( = n - r) unknowns, chosen ,ub1trarzly (e)
:--- ------1-----
has mfimtely many sol- bets mfimtely many sol•
utions Solution space uhon'l Solution set is a
1s of d1mens1on hnear variety who&e
k (= n m) (d) base spa"e 1s of d1men-
r = m, s1on "- (d)
k=n-m
We can find r ( ~ m) unknowns in terms of the
remammg k ( - n - m) unknowns, chosen arb1t-
rc1rily (d)
m<n
has infinitely m.any sol- has mfimtely many sol·
utions Solution space ut1ons Solullon set 1s
JS of d1mens1on a bnear variety whose
base space 1s of d1men•
k (= n - r) (e(d)) SIOD k (e(n))

We can find , unknowns in terms of the remain•

iq le(==. n - r) unknowns, chosen arb1tranly (e)
I 98 / MATRICES

Example J.23 Consider the system

2x1 + Xa - x, + x, = 2
Xi + Xa - x, + x,= 1
12x1 + 2x1 + 8x8 + 2x = 12.
6

The &uJDlented matrix (A, b)

r2 0 1 -1 1 27 ro 0 -1 1 -1 07

=I 1 0 l -1 1 1
l ,. ., l
'1 - 2,••
I
1 0 1 -1 1 1
'a - 2ra
L12 2 8 0 2 l2_J LO 2 -4 12 -10 oJ
,...,
r1 0 1 -1 1 17
interchange
r 1•and , 1 ,
ra Xi
I 0
LO
0
1 -2
-1 1 -1
6 -S
ol
O_J
I
,...,
r1 0 0 0 0 17

'ra1 +- '•·
2ra I0 0 -] 1 -1
0 I
LO I 0 4 -3 O_J
11 0 0 0 0 17
,...,
interchange
r 1 and r3 I0 1 0 4 -3
0 I
LO 0 -1 -J • O_J

,...,
rt 0 0 0 0 17

ra X (-1)
I0
LO
I

0
0
I -1
4 -3
1
0
O_J
I·
This shows
Xi = I
Xa + 4x, - 3x = 0 6
Xa - x, + X = 0. 1

So Xi= I
x11= -4x, + 3x1
X8 = x, - x1 •
Rank of the coefficient matrix A = 3 = rank of the
augmented matrix.
S -3 - 2 unknowns can be arbitrarily chosen. They are x, and x1 • Three
unknowns x1, x 1, and x 1 are determined in terms of x, and x1 • There is an
infinite number of solutions. This is case (d) of Theorem S.8.2.
The set of solutions can be written as
{{I, -4x, + 3x1, x, -x" x,, x.> I x,, x 6 arbitrary scalars} are
-= {(I, 0, 0, 0, 0) + x,(O, -4, 1, 1, 0) + x1(0, 3, -1, O, 1) Ix,, x1
are arbitrary scalan}
 5.8 SYSTl:.MS OP LINEAB. EQUATIONS / 199

= (1, 0, 0, 0, O) + ((0, -4, 1, 1, 0), (0, 3, -1, 0, l)].

It is a linear variety. Here (J, O, 0, 0, 0) is one particular solution of the
system and [(O, -4, 1, 1, 0), (0, 3, -1, 0, 1)] 1s the kernel of the coefficient
matrix .4 of \Jle r.ystem.
Example S 24 Consider the system
x1 + 2 t"1 + 4x8 + x, =4
2x1 x8 - 3x, =4
X1 - 2l.1 - Xa -=- 0
3x1 + XJ - x8 - Sx, = S•
The augmented matrix (A, b)

r~ _: =: -: :7 rt 47
2 4 1
,...,
r11 - 2r1, 0 -4 -9 -S -4
= 'a - ri,
r, - 3r1 0 -4 -S -1 -4

L3 1 -1 -S 5.J LO -S -13 -8 -7_j

r1 2 4 1 47
,..,,
'• X (-l)
0
0 -4 -S -1
1 t
t _:j •
LO -S -13 -8 -7

,..,, r1 o -I -I 27
r1 - 2,1, 0 1 t t 1
r, + 4r 1,
r, + 5r1 0 0 4 4 0
LO O -1 -1 -2.J
r1 o -l -f 27
,..,, 0 1 t t
r8 X l,
r, X (-f)

LO
0 0
0
1
1 1
1
:J
-,,
,.'1 +- ,,....
r1
0
0
1
0 ;-1
0 -1
27
1

,,- '• 0 0 1 1 0

"LO O f..J
O O
The last row shows that Ox1 + Ox1 + Ox. + Ox, ,.. 8/7, which is
-200 / MATRICES

absurd. Otherwise also it sliows that the rank of A is 3 and the rank of
(A, b) is 4. (Why?) So "the system cannot have a solution. In other
words, the equations are inconsistent.

Problem Set 5.8

1. Determine whether the following systems of linear equations are con-
sistent. Discuss completely the solution in the case of consistent
systems.
'.a} x 1 - ~-2 + 2x3 + 3x4 = (b) x1 + 2x + 4x3 + x,=
2 4
2xi+2x2 + 2x, = 2x1 x3 +3x,= 4
4xd- x2 -- x 8 - x, = l X1 - 2x2 - X3 - 0
xi+2x2 + 3x3 l 3X1 -f-- X2 - -5X,=
X3 7
(c) 2xi+ x 3 - x, ..i x 6 a-.: 2 (d) x 1 -t 2x2 - x 3 --2x,= 0
xi-+ x 3 -- x, + x 6 = 1 2x1 -f- 4x3 + 2x8 +4x,= 4
12xi+ 2x 2 -j 8x3 -t 2x6 -- 12 3x1 -t 6x2 -l 3x3 +6x4 = 6
(e) X1 - ,X3 1 (f) x1 + 2x3 = 1
2x1 -+ x 2 + x8 -- 2 2x1 + x2 + 2x 3 -- I
Xa - Xa - 3 x2 - 2x3 = l
xi-f-x2 + x 3 4 X1 + 1
Xz ·-
2x3 - x3 0 Xi - 1
Xa -1 4xd
(g) 2x1 !-j- x 2 + :\'a -t .\"• 2 (h) X1 -j 3Xz - 3X3 -j 2x4= 1
3x1 - Xa-t X3 - X4 = 2 4x1 + X 2 - 2x3 + X 4 = 1
Xi +2x2 -x3 x, + =- 1 6xi t 5x2 + 10x3 +3x,= 15
6x1 +2x2 +x3 + x, 5 x 1 -t 2x2 + 3x3 + x,= 6
(i) X1-2Xa-X 3 =- -} (j) 3x1 + 6x2 + 3x3 +6x,= 5
2xi -- x 3 -3x, -- I x1 + 2x x 3 - 3 -2x,=-1
3x1 +X 2 - x 3 -Sx,, - I 3x1 + 6x2 + x3 +2x,= 3
2x1 +3x3 + x, -- 0 x + 2x + 2x
1 2 3 +4x,= 3
(k) X1 -x3 - 2 (I) 2x + x + 2x
1 2 3 = 1
x1 + x1 +2x3 - 4 X1 -j- Xa = 0
x1 + x1 -2x8 4 x1 - 2x + 6x 2 3 = 3
x1 + x1 +x3 = 4 x 1 2x - 3 - 1
x1 + 3x8 -x3 -- 8 x 1 - x + 4x 8 3 = 2
(m) x 1 + x1 - x 8 -6x,+h,= - 19 (n) x 1 - 3x2 + x3 x,= 7
Xi + 1x,-1x6= 28 2x1 + 4x1 - x3 +6x,= -6
2x1 -3x8 +18x,-4x 6 = 24 2x1 + x1 + x,= 0

5.9 MATRIX INVERSION

In this article we apply the method of row·reduction to find the inverse
of a nonsingular matrix. Let A be an n x n matrix, which is nonsingular.
Then. A-1 exists. Look at the equation Ax = b, where 6 ¥' 0. To solve
this we can proceed as follows :
 .S 9 MATRIX INVDSION / 201

Ax= b.
Therefore, ..4· 1(Ax) = .4.- 1b
or (..4-1 A).l. =-- ..4-1b
or l.x = .i4-1b
or x-= ,4- 1b.
So, solving the equation Ax = bis equivalent to finding .i4- 1b. But the
solution of equations can be done by the process of row reduction. There-
fore, the method of row reduction wlll also be adaptable to find A-1b and
consequently A-1 • We can write the equation ..4x = bas
A1e = 1,.b . (1)
Therefore A-1Ax = .4-•1"b,
i.e. l"x -=-- .4-1b . (2)
In the process of reduction we always work from A and arrive at I,.
(see Coro lary 5.7.12). Equations (1) and (2) md1cate that, ,r the same
row reduction 1s applied to I,., the 1dentaty matnx, then we end up with the
matrix .i4-t on the right-hand side of (1). Thus, we have the followmg
method of mvertmg A..
Perform a sequence of elementary row operations on A so that it
reduces to I,. Perform the same sequence of elementary row operations
on the matrix 1,.. The resulting matnx 1s ,4- 1 •
In prart,ce, we perform the sequence of elementary row operations on
A and In simultaneously, keeping them side by side as illustrated JD the
followmg example.
Example 5.25 Invert the matnx
1 r
1 07

A= j 1 -1
LI -1
1
2_j
I·
Note Whether the matrix is nonsingular or not w,11 also be clear by the
process of row reduction. If the final rnatnx has all its rows nonzero, then
the original matrix should be nonsmgular. (Why ?)
We write the 3 x 3 matrix A and the matrix la side by side and
obtain a 3 x 6 matrix with a veruc:al line separating the entries of the
matrices A and / 1 as shown below. Then we perform the elementary row
operations on this 3 x 6 matrix in r,uch a way that the entries on the left-
hand side of the vertical line ultimately form la, The entries on the right-
hand side of the vertical bne then give the required inverse of A. The 3 X 6
matrix used here 1s called the enlarged matrix of A. '
r1 1 o 1 o
ILI
l - 1
-1
1
2 0
0 l
0 1..J
202 / MATRICES

r1
I
1 0 1 0 07

lLO 0
-2
--2 2
1 -1

-1
1 0

0 l_J

ra
'a
,...,,
X
X
(-½),
(-½)
I~
LO
1
1 --¼
1 -1
0 1

l
0

o -l_J
07

0 I
,...,, r1 1 0 l O 07

I 0
1 --l i -l OJ
LO o -I o l -l
11 l 0 0 0-7
,...,,
r3 X (-2)
I

Io
LO 0 1
l -l
0 -1
0
l_J
I
1 I 1 0 1 0 07
,..._,

I
I

I 0
1 0 -1
l
LO 0 1 0 -1 l_J

.....,
1 l 0 0 t 1 -½7
Io 1 0 I -1 ½ I.
I
LO 1 0 -1 l_J
1 -½7
I
Hence,

The reader is advised to check that

LO -1
.AA-1
:J
is indeed I, and so also is A-1A.

Problem Set 5.9

1. Using the row reduction method, determine which of the square
matrices of Problem 1, Problem Set 5.5, are nonsingular and in each
of these cases find the inverse.
2. Repeat Problem 1 for the matrices of Problem 2, Problem Set 5.5.
3. Repeat Problem 1 for the square ma1rices of Problem I, Problem Set
S.1.
Chapter 6

Determinants

6.1 DEFINITION
We start with some preliminaries on permutations of the set {I, 2, ... ,
n}.
6.1.1 De&nltion An ordered pair (p, q) of distinct poc;itive integers p
and q is said to be an im·ersion if p > q.
For eilample, (5, 3) 1s an mvers1on but (3, 5) 1s not.
6.1.2 Definition Given a permutation P =- U1, ii, .. , in) of the set
{1, 2, ... , n}, we define the bet
cilp-= {(j1,j2), (ji,j3), ... , (j1,jn),
Uz,ia), •··• Ul,jn),

(jn-1,inlJ·
The number of inversions in <l>p ib called the number of inversions in
the permutatio11 P.
Example 6.1 Let P be the permutation (3, 4, 1, S, 2). Then <l>p =
{(3, 4), (3, 1), (3, 5), (3, 2), (4, 1), (4, 5), (4, 2), (1, 5), (J, 2), (S, 2)}. Count·
ing the inversions, we see that the number of inversions in Pis 5.
6.t.3 De8nition A permutation (ji, j 2, ••• , in) of the set {l, 2, .. ,, n} is
said to be an even (odd; permutation if the number of inversions in it
is even (odd).
Examp!~ 6.2 The permutation (3, 4, 1, S, 2) in Exampre 6.1 is an odd per•
mutation, whereas (3, 1, 4, 5, 2) is an even permutation.
Now let us take up the definition of a determinant. · To each n x n
matrix A. of numbers we associate a number called the determinant of A. in
the following manner. Let
204 / DETERMINANTS

Form all products of the type

« Iii «2;, «nj,. ,
in that order (the ascending order of row indices). Here (ji,j2, •• :, j,,,) is a
permutation of the set {1, 2, •..• n} of column indices. Note that each such
product contains one (and only one) term from each row and each column
of the matrix. To each such product assign a 'plus' sign or a 'minus' sign
according as the permutation (ji,j2 , ••• ,j,.) is even or odd. The algebraic
sum of all these signed products is called the determinant of A. It is de-
noted by det A or I A I or
«11

This last symbol is itself called a determinant of order n. The precise defi-
nition is as follows. ·
6.1.4 Deftaltion If
r«u «11 «1..7

«11 «.1 «.11

A= (1)

Lot,.1 « ... a.,.,._J

then
det A = I E(J 1.
It 1, • •• 1 "n1 ) ix1,·l ix-,,1 ... otn' ,
.,: !lft.
(2)
where the summation is taken over all possible products of the form
«11. 11211
... fn,J,.• in that order, and

if (Ji, .f1 , ... , J,.) is

an even permutation
if <J1, 11, .. ,, J,.) is
an odd permutation.
The ript-hand side of Equation (2) is,called the expan,ion of the deter-
minant of A.
 6.1 DEFINITION/ 205
Note that there are nl terms in the expansion of det A, because there
are nl permutations of the set {I, 2, ..• n} of column indices, and to each
such permutation U1tl1., ... , in) we get a product« 111 «211 ... «nJ,.· In fact,
the permutation Ci1, i,, ... j,.) tells us the order in which the column mdi-
ces are chosen to form the product.
6.1.S Remark Det A. 1s defined only when A 111 a square matrix. If A is
not a square matrix, det A. is not defined.
Example 6.3 Let

A= !123 1.
l_«,i e,dJ 17 J1

Here the six possible products are « 1111 11 or,u, cxua i.1:>'u, «u11u,- 31 , «11:x11«88,
«11«11«a■, «1a«11«a1• The &ign, to be attached to these products are
determined ac; follows : for example, t.tke «u«u!X11 The order of the
columns chosen here I!, (2, 1, 3). This i!, an c,dd permutation. So the
sign, to be .:1.ttached to th1c; product 1'1 'mmus' Domg this for every pro-
duct, we finally get \
I A I = «11«1211,, 1n2i1'a2 -t txu:t2,u,1 - O'u0121«aa
+ rL13!X910CJ2 - 1i,OC211IX31 • (3)
As a numencal examplt, we ma) take
2 -1 3
=2 X 4 X 2 - 2 X OX I (-1) XO X +
1 4 0 (-]) - (- 1) X 1 X 1 + 3 X 1 X 1
3 X 4 X ( -1) = 32 .
-1 1 2.
A careful arrangement of the six term1, m Equc1t1on (3) gl\,es

= rx11(at11«aa - 1X1,10taz) + «11(«1aata1 - «11tzaa)·

-f- «1,1(«11«.t11 - «11«,J •

«n
Clearly,

~1 ~.
if A. is a 1 X 1 matrix [«11, then det .A. = «1,
When it comes to considering determinants of the fourth and ,higher
orders, the calculation invoJved in writing out all the terms becomes prohi-
bitive. So we study further properties of determinants and arrive at
shorter methods of calculating det .4. However, certain special detcrmi•
nants ~ be eval •.&ated directly from the definition, for example,
206 / DETERMINANTS
A1 0 0 0

0 >-1 0 0

: .. = >-1~ •.. >-••

0 0 0 >-.
In particular, det I. = 1 and det O.x,. = 0.

Problem Set 6.1

I. Determine whether each of the following permutations is even or
odd:
(a) (I, 3, S, 61 41 2) (b) (3, 4, S, 2, I, 6)
•
(c) (2, 6, 4, 3, 5, I) (d) (1 1 4~ 3, 2, 5),
2. Write down all possible permutations of the set A and separate the
odd and even ones :
(a) A = {l, 21 3} \b) A -- {l, 2, 3, 4} .

r: : :]
3. Find det A for the given matrix A :
•

(a) A = (b) A = [: : :] .
Li o o 1 1 o
4. Evaluate the following determinants :
I x x2 I X yz a b
1• cl
I:
(a) 1 )' y• (b) y zx (C) • C a b,.
I
1 z z'I. z X)' I b C a1
s. Prove that the solution of the system
«1X1 + «1X1 = 81
~1X1 + ~zX1 = 8~
I81
8a
"a
fia
I I«1
~1
81
s.
I
is = = ,
"~•- I
X1 X2

I I
"1
~1
provided that «1~ -I= «1~1-
"1

~1
•

:1
6. True or false ?
(a) If (a, b, c) is an even permutation, then (b, a, c) is odd.
(b) The sign attached to the product or:11a.u«,1 of a third order de-
terminant is 'plus'.
 6.2 FUNDAMPNTAL PROPU.nas o, DBTDMINAND / 207
0 1 31
(c:) 2 0 -4 1 =0
-5 5
ol .
0 1 1
(d) 2 0 0 = 0.
-5 s s
12 - ).
(e) The equation 3
1 - ).
j -= 0 has no real root.
4 I
(f) The determinant of a t1iangular matri,c is the product of its
diagonal entries.
r1 21
(g) I I -= -2.
L3 4_j

(h) I 1
-I -1
; I= 0.

6.2 FUNDAMENTAL PROPERTIES

OF DETERMINANTS
The fundamental properties of determinants, which foil ow immedia•
tely from the definition, are given in this arllcle in the form of theorems.
Of these, the proofs of Theorems 6.2.1 and 6.2.8 (see§ 6.3) are rather com-
plicated, and a beginner may skip them on his first reading. In this article
A. = (at, 1) stands for an n x n matrix.
6.2. l neorem // the matrix B is (!btained from A. by an elementary row
vyv, 1i1••v•• VJ •.Tr• ,.. , ..... ••••--• ,.,,. •.,.,.6•••5 .,,.,_ • ., ... .,,, .......... _.,. - -
For the proof see § 6.3.
Example64 1 2 3 4 1 2 3 4
-1 0 1 2 1 -1 2 3
=-
0 1 4 0 0 1 4 0
l -1 2 3 -1 0 1 2
because the determinant on the right-hand side is just the determµiant on
the left-hand side with the second and fourth rows interchaqed.
,.2.2 Tlleorelll I/ A.11118 two identical row,, then det A. = 0.
~oof: A. can be considered as being obtained from itseJf by inter-
cbanams the two identical rows. So, by Theorem 6.2.1 9 det A. - -dot 4'.
lfeDce. det A. =- o. I
208 / DETERMINANTS

Examp/e6.J 2 3 -l 4
4 0 -I
- 0
1 2 -3
4 0 -- I
without any further calculation.
6.2.3 Theorem If A has a row of zeros, then det A - 0.
Proof: det A = >; E • • ) (1!I/ «2 .... « .. If the k-th row is the
lh,12, , ,,. 1 11 n1,.
zero vector, then ,xk. is zero for every choice of it• Therefore, each
"JI:
product is zero and thi'I gives det A - 0. I
Example 6 6 2 1 3,

without any further calculation.

6.2.4 neorem If the k-th row vector rA. of A is the sum of two vectors bir
and C1,: of Vn, then I A I I B I I I C I , where B is the matrix
<1btained from A when r,. is replaced by b1r., and C ir the matrix
obtained from A when r1: is replaced by c,,.
Before we take up the proof, we shall mustrate the result by a simple
example.
Example 6.7 or.1 «a oc 3 «,

b1 b8 b3 b, c1 C'z c3 c,
II 31 3a 33 3, 31 31 83 3,
Proof of Theorem 6.2.4 Let b11 = (b11u b111, ... , b11J and ca: = (cu,
c111, ••• , c11J, Then r1.kJ1, = blrJ,, + ck/11 for every k. This shows that each
product "JJi «211 ••• «k/11 ••• «n1,. can be written as (r1.1/a «211 ••• bk/1: ••• r1.ttJ)
+ (atl/1 "1.J•... Olk/• ... Gt"'·>·
 6.2 FUNDAMENTAL HOPl!P.TJES OP DlTBIUONANTS / 209

The alaebraic sum of the first type of products with the appropriate
signs gives det B, and the algebraic sum of the second type of products
with the appropriate signs gives det C. Hence the result. I
6.2.5 Theorem If the matrix B is obtained from A. by an elementary
operation of type II (i e. multiplying any row ,·ector by a nonzero
1calar c). then 'det B = c det A.
Example 6.8 a1 · h1 c1 I n1 b1 C1 1

ca1 cb8 cca - c az b 1 c1 1

.

a3 b3 c8 a9 b3 c3 I
Proof of Theorem 6.2.5 Let the k-th row of A. be multiplied by c.
Then the k-th row of B is (c«u, c2., 2 •••• , cxtn). So, in the expansion of
det Beach i,roduct is of the form :it 11 «1'1 .•. (c«k.) ... « .• This shows that
I I Jlt 11.Jn
every term of the expan,ion of det B i'i a multiple of the corresponding
term by a factor c, in the expansion of det A. Hence. det B -== c detA. I
Note that Theorem 6.2.5 holds even if c -:.- 0.
6.2.6 Theorem // tf,e matrix B is obtained from A. by an elementary row
operation of tJpe Ill (i.e. adding c times the s-th row to the k-th row),
then det B --= der A.
Examplc·6.9

<"1 C1. Cs

Example 6 10 2 3 1 2 3 I 2 3

4 4 4 4 4 4 - 4 4 4 =0,

S 6 7 5-1 6-27-3 4 4 4
by Theorem 6.2.2.
Proof of Theorem 6.2.6 Let the Jc-th row of B be ,,. + er., where r.,
and ,. are the k-th and .r-th rows of A. So, by Theorem 6.2.4,
at11 at11 at1,. ~11 «ui «1..
.:
at11 «11 «1.. c«,1 COl,z c«.,. -k-th row.
I
detB = ·+
Giel 111•• 111...
I «.1 «,. "•• ➔ 1-th row.

. I!
11.., «., ... ix,.• • «111
"•• at,."
210 / DETERMINANTS

Applyina Theorem 6.2.S, we obtain

/ «n «11 IX1n

I :
I •

I a.1
I

det B = det A ~ c I:
I
I
1 «•1

I
.
:

I r, nl 1Xr,J 1Xnn

=-=- det A (by Theorem 6.2.2). I ,

6.2.7 Remark Theorems 6.2.1, 6.2.5, and 6.2.6 tell us that the effect of a
finite number of elementary row operations (of any or all of the three
types) on det A i,; to multiply it by a nonzero scalar.
6.2.8 Theorem det AT ·-= dct A.
For the proof 11ee § 6.3.
Example 6.11 2 3
1
4 2 s
I 5 1 R 3 7 - l,.
I3 2 4--1 21
6.2.9 Theoren1 Theorems li.l.J to 6.16 ore true if e,•erywhert· tl.e word
•,ow(s)' 1.s <hanged to 'c11/rm111(s)".
Thi,; is a consequence of Theorem 6.2.8.
6.2.10 Remark As a consequence ('f Theorem 6.2.9 the elementary
operations can be performed a& \\Cll on the column!. of a determi-
nant, with the same effect ac; that on rows.
Example 6.12 Evaluate
2 3
I
IAI== 2 -)
: 1·
--1 0
2 3 2 3
I
IA I = I 2 -1 0 == 0 -5 -6
I
1--1 0 I O 2 4
by r1 - 2r1 and r8 +r 1• Using Theorem 6.2.8, we get
1 0 0

l .4 I= 2 -s 2 =1X (-5) X 4- I X 2 X (-6) = -8

3 -6 4
~s all other products are zero.
 6 2 FUNDAMENTAL PROPERllES OF DETEkMINANTS / 211

Problem Set 6.2

1. If A is a square matrix of order n and ix a scalar, then prove that
det (a.A) = 01." det A.
2. Let A be a skew-symmetric matrix of odd order. Then prove that
det A =-= 0.
3. For a triangular matrix A, prove that det A i-. the: product of its dia-
gonal entries.
4. Prove that
I a. ix2 10
I
1 ~ ~l -= (..t - ~) I l ~

l 6 62
0 1 01. +~I
-= (et - ~) (fi - 6) (6 - r,.) 0 0 1 I
1 6 52 I
s. Without cxpandinf, prove that
l 1 3 b +C l a
(a) 2 9 1 =0 (b) l -1- a 1 b =0
4 11 7 b ta 1 cl
a b C y b q X y z'I
(c) X y z - X a p = p q rl
I
I
p q r z C r I a b cl

I:,I :;i y
z I 1 1
x2 ,-2 z2 1
1 I
(d) ya
lyz
= xa ya :za'
:zx
a 1-- b b+c e +a a b C

(e) p-j-q q+r r -t p =2 p q r

x+y y+z z+x X y z

x+3 x+6 4 7 10

(0
X

x+l x+4 x+1 =0 (g)

I10 13 16' =0
I
x+2 x+s x+8 I 20 23 26
212 / DETDMINANTS

1 a a• -be x-y y-z z-x

(b) 1 b b'- - ca =0 (i) y-z z-x x-y =0.
1 C c1- - ab z-x :x -y y-z
6. True or false ?
(a) If every row of a determinant of order n ( I 1) is multiplied by
k and every column by k', then the ,alue of the determinant is
multiplied by kk'.
(b) If the diagonal entries of a determinant are multiplied by k,
then the value of the determinant is multiplied by k.
a b c ab atb1I
(c) 2a lb 1.c = 6abc. (d) C d c+d =0.

3a 3b 3c e f e + r
a,+ a1 P1 + Pa 11 -t 1. 01 P1 11 I Oa Pa I,
(e) b1 -1- b1 ql + ql m1 t- ma - b1 ql ml + I
b1 q, m1
I

C1 + Ca r, + '2 t- nz
n1 C1 T1 n1 I c1
'• n1
a a' a' a a' a'
(0 b b' b' = - b b' b'
, ,
'c c' C C c' C

a 0 01 Id 0 0

(g) b 0
0 I=I leIo I
0 0
0 d e I
a b

I01
a,. Oa i a1 - b1 a1 -b1 a:s -ba

(b)
(1
C1
ba ba•
Ca Ca j
= h1 -

I C1 -
t'1

01
b1 - Ca ba -
c1 - a 1 Ca - aa
Ca

6.3 PROOFS OF THEOREMS

To prove Theorem 6.2.1 we need the following two lemmas.
6.3.1 Lllllma Let p be the number of lnverslon.r In a permutation
P = (J1,J1, ... ,J.) of the integer, I, 2, 3, ... , n, Let Q be thep,nnu-
tatlon obtained from P by changing two adjacent entrte, of P. Then
tM number of lnvtr.rlon, Ill Q is p ± 1.
 6.3 PROOFS OP TBBOIU!MS / 213

Proof: Let l• and i1t+1 be interchanged in P to produce Q

= (/1, ia, •··• ii.-1,, i"Jc+i, l1c, it+1, ... , /,.). We have
(l)p == {U1, /a), (}1, ia), ···• (ji, /11-1), Ui,it,), (ji, /rtt-1), •··, U1t JJ,
Cia, /a), •··, (j.,j1c-1), Ua, J1c), Ua,JttJ, ... , U1,JJ,
... ... ..... ....................................
U1t.-11 i1t.), U1t.-1, i"Jc+1), .. , Uri, J.),
U11, /t;+1), · •• , U1c:JJ,
..........................
(J,._1, JJ}.
The set (l)Q is obtained from '1Jp by interchanging it and /u1 • Careful
scrutiny shows that Cl>Q 1s different from Cl>p in only two ways, namely,
(i) cl)Q does not contain (j1c,Jm), whereas Cl>p contains it, and
(u) () P does not contain Ut+i, it), whereas (l)Q contains it.
Since either Ure, Au) or Um, J,,) is an inversion, it follows that when
we change from P to Q the number of mversions is either increased by
one or decreased b)' one. Hence the result. I
6.3.2 Lemma If the matrix Bis obtained from the matrix A -= (et11),.x,. by
interchanging two adjacent rows, then det B = -det A.
Proof: Suppose the k·th and (k + 1)-th rows of A. are inter•
changed to produce B. The typical term in the upansion of det A. is,
keeping the row indices in the natural order,
etlj1 et2j1 ... 0tk/1c etk-t l,h:+1 etk+2,Jt-tt ... etnj,. •
Denotmg the entries of B by ~1,, we have the corresponding typical pro-
duct m the expansion of det B :
~JJi ~2/1 ... ~lci1c ~k+l,it+1 ... ~'!In= "'Iii r,.2J, ••• etktl,h, 0tkJ1c+, ••• etn/n
(because the k-th row of B is the (k + 1)-th
row of A. and vice versa)
= etlii et2J, ... 0tk/1t+1 etk+ l,i1 ... a.nJn "
By Lemma 6.3.1, we have
f.(/1,/1, ... ,ik,J1tt1, ... ,/,.)= -f.(/1,}1, ... ,/11+1,/1, ... ,/,.) 0

Hence the result. I

Proof of Theorem 6.2.1 The theorem has just been proved in
Lemma 6.3.2 for interchanges of two adjacent rows. Now, suppose the
k-th and .r-th rows are interchanged to get B. This transition fro111 .4 to
B can be made by carrying out the following operations successively ;
Let , > k + 1 and let there be t rows between the k-th row and the
,-th row. Let r,, stand for the p-th row, p = l, 2, ... , n. ,, can be
brought to a position immediately above ,,,, by t + 1 successive inter·
changes of adjacent rows. After this, r 1 can be brought to the position
214 / DETEJUIINAN'CS

originally occupied by,,. by making t successive interchanges of adjacent

rows. Thus; the interchange of r1 and r, is equivalellt to making (t IJ +
+ t = 2t + l successive interchanges of adjacent rows. By Lemma
6.3.2, the value of the determinant undergoes a change in sign (21 + 1)
times and therefore changes sign ultimate)). Hence, det B = -det A. I
Proof of Theorem 6.2.8 A and AT are both square matrices of the
same ~rder. So the expansion cf det A and det A 1 have the same number
of terms. Each term is a product obtained by taking one (and on1y one)
entry from each row and each column in the matrix, and attaching a
certai11 sign to it. In view of this, we see that, ignoring these signs
temporarily, every product that occurs in the expansion of det AT also
occurs in that of det A and vice versa. In order to complete the proof we
have therefore to show only that the signs attached to the products also
coincide, irrespecthe of whether they are obtained in the expansion of
det AT or in that ofdet A. We shall prove this in the following manner:
Let A = (a,,),.xn• Then AT ~ (a;,)n-<n, where ()';, = rx.,,. A typical
term in det A T is
E( , • • ) (I. '1 ot •. .. • ex • (I)
h,/1, .. ,J,. 1/1 2Ja nJ,.,
Taking product (I), without the sign, we have
I , I

rx.li, r,.2j1 ... rt.nj,.

= IX/ I
1
!X 1
~•
z .. . ex •
J,.n
• (2)
Here (ji, j 21 ••• , j,.) is a permutation of the set { I , 2, ... , 11}. Hence,
h, 1 = l for some k1 E {I, 2, ••• , n}
ika == 2 for some kJ € {I, 2, ... , n}, where k 2 / k 1 ,
... . ..... ... .................................
jkn = n for some kn E {1, 2, .. , n}, where k,. / k,.- 1 , .. , kl, k 1 •
So product (2), without the sign factor, is
r,.ik/'1 ,.,'ik,"• ... «Jk/• •
i.e.
«u.1 rt.2k1 ··· rt.nkn" (3)
Product (3) is the form of the product as it occurs in det A. The sign
attached to product (3) in det A is E(ki, k,, ... , kn). The sign attached
to product (2J in det AT is, from product (1), E(l"lt 1·It •••• 1·n)·
If we can prove that these two signs are the same, we would be
through. These signs depend on the number of inversions in the permu-
tations P = (j11 j 1 , ••. ,j.) and Q-== (ki, k 2, ••• , k,.). So we shall now
analyse the process by which we obtained the rearrangement (3) from (2).
Both (2) and (3) represent the same product. The entries in product
(2) have the column indices in the natural order I, 2, ... , n, where&s the
entries in product (3) have the row indices in the natural order. The
order of rows in product (2) is P and the order of columns in product (3)
 6,J P.ROOFS OF THEOREMS / 21'
is Q. The passage_ from product (2) to product (3) 1s nothing but the
rearrangement of P m the natural order. It is presented schematically m
Table 6.1.
TABLE 6.1
-----------
Order nf l"ows Order of co/umr,s

p = U1oil, .. ,,in) (1, 2, .. , n)

= U1,iz, ... , iki' ,in) (I, 2, ... , k1, ... , n)

t t mterchanges t
(jk1' i1, ... , i,._1-I, ik1+l, (k1, 1, 2, ., k1 -- ], k1 + I, ... , n)
... ,i..)
( ik 1' j1t .. , 11.,; ·• ,j.. ) (ki, 1, 2, ... , k 2 , • , n)

s interchanges
\

( ik,' jk,' j1,j,, ... ,j,.)

t t
t t
(jk1' j/..1' .. ' jk.) (k1o k~. .., kn)
= (l, 2, ... , n) =Q

The tin,t line ID T.ible 6.1 md1cdtes the order of rows (--PJ and the
order of columns (ndtural order) in product (2). The last line md1cates
the order of rows (natural order) and the order of columns (=Q) in
product (3).
The passage from product (2) to product (3) presented in Table 6.1 can
be realised by successive interchanges of adjjcer.t entries in the respective
permutations. Since P = (j1 ,j2 , ••• ,i,.), look forJi<1 = 1 and brmg it to
the first position by interchanging 1t successively with all lhe entries
preceding it. Thu, needs a certain number of interchanges, say t. The
effect of these interchanges is to reduce the number of inversions in P by t.
If we make the sami kind and number of mterchanges in the order of
columns, namely, the natural order 1, 2, 3, ... , n, we get (k1 , 1, 2, ... ,
k1 - 1, k 1 + I, ... , n), the effect of which is to increase the nut'nber of
inversions in (I, 2, ... , n) by t.
Now we look for ik = 2 and bring 1t to the se~ond position by
making successive interchanges. Suppose this requires s interchuges.
The same s interchanges carri~ out on the order of columns will bring k 1
216 / DBTBRMINANTI

to the second position. The effect of this is to r,duc, the number of

inveniona in the order of rows by , and Increase the number of invenions
in the order of columns by ,.
Carrying on this process, we arrive at CJ1c1, }"•' .... J1c,.), i.e. the natu-
ral order on the left and at Cl == (k11 k1, ... , k,.) on the right. Thus, the
number of interchanges required to bring P to the natural order is the
,ame as the number of interchanges required to bring (1, 2, ... 1 n) to
(ki, k1 ,... , k,.) == Cl. Since the number of inversions in ( I , 2, ... , n) is
zero, it follows that the number of invasions in P is the same as the num-
ber of inversions in Q. Hence, I( 1·1, JI•
1 1· \ == e(k11 k 1, • • • 1 k R) and the
•• • I fl/
proof is over. I

6.4 COFACTORS
The expansion of an n-th order determinant is the sum of all the n!
terms, each of which is a product obtained by taking one (and only one)
element from each row and each column. Let A. = (ac,1) 11 x11 , Suppose we
collect all the terms in the expansion of det A., in which the fixed element
«u appears as a factor, and write down their sum in the factored form
•uA". Here A." denotes the factor remaining when the element ac11 is fac-
tored out. We call A." the cofactor of «u in det A..
Example 6.13 «u «11 «11 == «11(«..«aa - ataa«aa)

«11 «.1 «.a

«a1 1111 «.a + «11(«11«11 - Gtt:1at~,i) •
So Au = cofactor of «u = «11«aa - «11«11
A 11 = cofactor of «11 = «11«11 - «11«11
A. 11 == cofactor of «11 = «11«11 - «.1«11.
If we want any other cofactor, say ...4 18 , we have to rearrange the above
expansion and collect the terms containing 1118 • Thus, we see that Au
.. •u«11 - 11111111•

An explicit formula for the cofactor of «u is given by Theorem 6.4.2.

Por this we need the following definition.
.-.4.1 De8nltloa The determinant obtained by deleting the I-th row and
the }-th column in det A is denoted by A,,.
Au Is also called the minor of the element ,,,,, in det A, but we shall
not use this terminology.

&,.,,,le 6.14 t,·or the determinant «.1 -.1 -.1 ,

«u «.1 ._
 6.4 COFA.CTOllS / 217

A11 = I «11
ataa
at1•1
«1:a
and so on.
Example 6.15 If .4. = (a.11).x,., then

«a, n-1

Gtn-1, 1 «n-i, 1

6.4.2 neorem Let A = (o:,,),.xn• Then for each i = 1, 2, ... , n a11d

=
J 1, 2, ... , n,
(l)
Proof: First, we shall prove a special case of the theorem, namely,
that A,.,. =-=- &,.". To prove this we collect all the terms involving «,...
They are of the form
c(Ji, J;,,,
• ... , Jn-lt
• ) « . « . ... exn- 1, 1,..
n 1It 21• 1
«,.,. •
where (ii,j1, ••• ,/,. 1 ) is a permutation of the set {I, 2, ... , (n - I)}. The
sum of all these terms is r:, ,.,.Ann• But
s(J" jz, ... , j._i, n) == cUu ja, ... , J._J ,
since the integer n occurring last does not increase or decrease the number
of inversions in ( /1, j 1, ... , 'j __ 1). Therefore,
Ann =I . J1, , .. ,J,.-1
c{ h, . ) IX lJ1 «211 .. • IX
n-1,/n-1

CXu «11 or;l, n-1

«11 «u ... «a, 11-1

= .. (by definition of a
determinant)
cx,._i, 1 «n-lt I «n-1, 11-1

= A,.,..
Now we take up the general case and reduce it to the special case. We
have
p Q R
.......
detA. = ·······
s . a.u { T
...... : t

·······
u V . w
where P, Q, R, S, T, U, V, and Ware mere abbreviations-for the remainiq
entries. U, Y, and W have (n -- I) rows; R; T, and W have (n - J)
2J8 / DETERMINANTS

columns; S and T have only one rcw, and Q, V have only one column.
Clearly,

(2)

Shifting the Hh row by successive interchanges \\lilh the (n - i) rows of

U, V, and W, we get
p Q RI
det A = (- I),._, U V W (by Theorem 6.2.1).

S ex11 T
Again, by another series of interchanges of the j-th column with (n - j)
columns of R, W, and T and by apply mg 1 heorems 6.2.1 and 6.2.9,
we get
p R Q

det A = (- l)n-• ( - 1),. ' u w V

~ T ex,,
So the cofactor of ex., in det A
'P R Q
= (-- l)•+J X cofactor of «,, in u w V

= (- l)it' X I'Pu wRI •

by the preceding part of the proof. Hence, A,,= (- J)t+1 fl,, by
Equation (2). I I

Example 6.16 Consider

ah 8/
h b / 1 •

a I cl
The cofactor of h occurring in the second row, first column, is

(-1)1+1 \; : I= -(he - gf) =fg - ch.

The cofactor of/ occurring in the third row. second column, is

(- J)l+I I: ; I= gh - qf.
 64 COFACTORS/ 219
Since this determinant is symmetric, the cofactors of elements,
which are symmetrically placed with respect to the diagonal, are the aame.
So, for example, we need not distinguish between ac11 -= h and «ti == h.
Both h1Ve the same cofactor. (Check!) Denoting the cofactors by
capital letter-i, we have
A -= be - /2, B = ca - g1 , C = ab -- h1 ,
F = gh - af, G = hf - bg, H = fg - ch.
The n! terms th.it occur m the expansion of llet A can be separated
into n groups ai, follows :
•
(I) all tenns contaimng «11
(2) all terms containing ac 11

(n) all terms containmg «1 n •

Edch of the,;e groups contains (n - I)! terms, since this is the
number of terms m any Am which 1s a determinant of order (n -1).
The,;e group, arc mutually exclushe, because when a term contains 11,,.
11 cannot conta•n any other element from the first row. Thus, there a1e
n mutually exclusive g1oups, each containing ( n - I)! terms. Therefore,
the total number 'of terms m aJI these groups s n X (n- I)! = n!.
Hence, 1t follows that all the terms in the expansion of det A are exhaust-
ed by this groupmg. _Thus, we have
det A = «uAu + «11.A.11 + ... + «1nA1n , (3)
If, m,;tead of the first row, we focus our attention on the elements of
the i-th row and collect terms mvolving cx 11 , 11,1 , .. , etc, \\e ,et, by the
same argument I s that which resulted an Equation (3) :
det A = ac, 1A, 1 + cx, 1 A, 1 + ... +
cx, 11 A,n • (4)
Samilarl), we can work with the elements of any column, say theJ-th
column, to get
det A= au.Au + cx11 A2, + ... + cx,. 1 An,. (S)
Thus, \\e have proved the following theorem.
6.4.3 n~re• If A =. (ac,1) ts a squa,r matri~ of order n and A 11 denote,
the cofactor of a,, in det A, then,jbr each , = 1, 2, ... , n
ft

det A = I cx,1:A,1: (6)

k=l
ar1d,for each)= 1, 2, ... , ,;,
n
det A == l cx11A11, • (7)
k•l
Equation (6) is called the expansion of det A In t,rm.r of the I-th
row and Equation (7) is called the expansion of det A In •~rm, of the J-th
column.
220 / DBmtMINA.Nff

Example 6.17 Evaluate

a h g
A= h b f
g IC
Expansion tn terms of the first row gives
A = aA. + hH + gG .
Expansion in terms of the second row gives
• A = hll + 1-,B + JF,
and expansion in terms of the third column ghes
A=gG+JF+cC.
Taking any one of these, we get
A= aA + hH+ gG
= a(bc - /2) + h(fg - ch) + g(fh - bg)
= abc + 2/gh - tJ/2 - bg2 - ch1 •
In order to evaluate a determmant, we usually use the theorems of
§ 6.2 to introduce as many zeros as possible in any row (column) and then
expand in term,; of that row (column).
Example 6.18 Evaluate
2 1 -3 4
1 0 -1 2
3 -2 1 0
-1 4 2
Multiplyjng the first row by 2 and adding it to the third row, we get
2 l -3 4 2 1 -3 4

1 0 -1 2 1 0 -1 2
fl = =
7 0 -s 8 r, - 4r1 7 0 -s 8
-1 4 2 1 -9 0 14 -ts
Expanding fl in terms of the second column, we get
- 1 -1 2 1 0 0

fl== 7 -S 8 == 7 2 -6 •
c,_+ C1o
-9 14 -IS c1 - 2e1 -9 S 3
where c, denotes the i-th column. Now expansion in terms of the fint

row &ives4 =- I: -: I= - (6+ 30) = -36 ,

 CUCOPAcrGU / 221
6.•.• 'l'lleonll
r
Let the row rtt:tor, of tl,e 1q,1are matrix A .. («u).x. 6,
Let the ,ector R, denote tM Ytctor of cofactor,, In
1, , 1, ... , ,..
det A, of the efement of,,. In other word11
r, == (at,i, ate■, .. , «11, ••• , «,.)
and R, = (..4.,,. A 11, ••• , A.,, .•. , ..4.,,.).
Then r, • R, == det A for oil I= I, 2, . , n,
and r, • R1 = 0 ifl rfo j.
In 1hort, r, · R1 = 811 det ..4. for all i, J == 1, 2, .. , n.
Proof: The fir!lt result is only a restatement of Equation (6). To
prove the second result, let us construct a matrix B by replacing the )-th
row of A by the i-th row of A. Note that the Hh row of A remains as it
is in this construction. So

cz,n --+- i-th row

det B =
cz,1 cz,1 •• cz,. -► j-th row

IZnt IZn1 1Znn

This determinant is zero, because two rows are 1dent1cal. On the other
hand, we can expand det B m terms of its j-th row. Smee the deletion of
~heJ-th row and the k-th column of det B gives the same 111-,, as th"
deletion of the J-th row and the k•th column of det ..4., we get
0 = det B = cz, 1A,1 + cz,1A,1 + ... + «111 A, 11
= r, · R1• I
6.4.5 Remark A similar result as in Theorem 6.4.4 is true for columns.
6 .•.6 Corollary Let A be the matrix (cz.,)11x11 and B the matrix (~11).x.,
wMre ~., = A1,. Then
fl 0 0 07

AB= BA=
0 t,. 0
U=M.,
0 0 0
where t,. == det A.. Consequently, det (AB) = t,,.• == det (B..4.).
'lbe proof is obvious when we carry out the multiplication of .4 and
B and repeatedly use Theorem 6.4.4.
 222 / DETERMINANTS

6.4. 7 Delaltloa The matrix B cf Corollary 6.4.6 is called the adjoint of

A and is written as adj A.
Example 6.19 Find the adjoint of
r1 -I 07

A=l0 l 1/.
LI 2 l_j
The cofactors of the different entries in det A are
A11 = -1, A11 = I, A13 = -1,
..411 = 1, A11 = 1, Asa =- -3,
A11 = -1, A 31 = -1, Aaa = I.
Therefore,
r-1 1 -17
adj A= I
I
-I I·
L-1 -3 ]_J
The adjoint is of sigo.ificance in finding the inverse of a nonsingular
matrix, as will b;: shown at the end of§ 6.S.

Problem Set 6.4

l. In Cor,>llary 6.4.6 we repeatedly use Theorem 6 4.4. How many
times do we use it ?
2. Prove Remark 6 4.5.
3. Evalu:He the following determinants :

(a)
2 I 0
0 3
II 2 4
(b) 12 3
3

(c) 6 I; I
0 3 2 13 4 2 10 20

I
3 2
15 29
41

2 141
I 3 8 7 6
I
7 4 10 2
I 1
I 2 3 4
1 I
s
(d> I (e) (fl I 3 6 10 15
I
I
16 19 3 17 6 8 s 8
I 4 10 20 3S
: 33 3~ 8 38 19 s 3 9
I SIS 35 69
4. Without expanding, prove th,tt
I yz y + z 1 I I

1 ZX Z +X - X Y Z

I X)' X +y
 6 4 COFACTORS / 22J
S. Prove that
1 X x1
(a) 1 y y• = (x - y) (y - z) (z - x)

z z1
X JZ

(h) 1 y zx = (x - v)(y - z) (z - t')

z xy
XTO b C d
a x+b C ' d
(cl = "t3(x -t a + b + c + ti).
a b ~+c d
a b C x+d
6 If c.>1, c.> 2, and c.> 3 are the three cube roots of unity, then prove that

Ix, t'3
x. "•
X1 l.:a -
3
JI (X1 t Xie.>, -f X461!)
I
I I

I XA Xa Y1
This determm,int 1s called a c,rcula111 ot the third order Write
dow,1 a circulant of order n Write al!.o its value
7. (a) Prove that the equation of a ui cle throurh three pomts (xi, y 1),
(x1, l 1), and (Y3 , h) 1s g1~en by
x~ + y' x y l I
xf t- yf X1 Yi I
0
l.: I+ y: 1 11 __

XA >'•
YI + J'i 'ta Ya l I
(b) Determine the equation of the sphere passmg through the points
(Y1,) t• z1), (l.,. Yz, Za), (xa, Ya, Z1), and (x., Y,, z,) 1n the deter•
mmant form
8 Solve the equation
X t- a b C

C x+b a == 0.

a b x+c
 221 / DBrBR-.1 NANTS

9. Without expanding. prove that

I+a 1 t I
I I+ b I l
= abcd(l + -aI + -bI + -cI + -)
I
d '
I I+ c 1
I I+ d
10. Calculate the adjoint of matrices of Problem 2, Problem Set S.S.
JI • True or false ?
r cos 8 -sin 8 7
(a) In the matrix I I the cofactor ..411 is -sin 8.
L sin O ,cos 8 .J
(b) There exist determinants none of whose entries is zero, but f..>r
which every cofactor is zero.
(c) If a matrix is nonzero, then its adjoint is also nonzero.
(d) adj (adj A) = A.
(e) adj («/,,,) = «-11,,,.
(f) If adj A is a diagonal matrix, then .A. is also diagonal.

6.5 DETERMINANT MINORS AND RANK

Of A MATRIX
Given a matrix A, we can get many smaller matrices from it by simply
deleting a certain number of its rows and/or columns. Matrices thus ob-
tained are called submatrices of A.
6.5.1 Deftnltlon The determinant of a square submatrix B of A is called
a determinant minor of A or simply a minor of A.
Example 6.20 let
r1 2 3

A== ls 9
6
10 11
1
12.J

ThenB=
r•
I
2 37
I is a submatrix of A and I119 .1131 is a minor of
L9 10 lt_j
A and also of B.
6.5.2 Lemma (a) Let A be an m x n matrix. Apply som? elementary row
operations 10 obtain a ne,v matrix Ai, Delete some columns of Ai to
get a submatrix B1 of Ai, The same submatrlx Bi could also luwe
been obtained if we l1ad first deleted the same column., from A Itself
to obtal11 a aubmatrlx B r,nd then applied the same elementary row ope•
ration, to B,
 6.5 DETERMINANT MINORS AND RANE OF A MATRIX / 225
(b) I/ Bi la a square matrix (and therefore B also), then det B ta tf/Ulll
to some nonzero scalar multiple of det B1 :
The proo! is obvious, because of the way elementary row operations
work. Deletion of a column or columns will not affect the result of the
elementary row operations on the remaining submatrix.
Part (b) is qnly a restatement of Remark 6.2.7.
6.5 3 Theorem If A is an m X n matrix, then the rows of A are £1 tJf A has
a nonzero minor of order m.
Proof: Let A1 be the row-reduced echelon form of A. Suppose the
rows of A are LI. Then .A.1 would contain m nonzero rows. The m
columns containing the 1's at the steps will aiso be nonzero. Delete the
remaining columns of .A.1 and obtam an m x m submatrix B1 of A 1 • Ob-
viously, B1 = I,. and det B1 = det T.,. = 1. By Lemma 6.5.2 (b), the cor-
responding square submatrix B of the matrix A will have the property that
det B is a nonzero scalar multiple of det Bi, which is 1. So det B =I= o.
Thus, A has a nonzero minor of order m.
Conversely, suppose A has a nonzero minor of order m. Let B be the
corresponding submatrix. We shall now prove that A1 contains m nonzero
rows. Suppose, if possible, A1 has a zero row. The same operations
that reduced .A. to A1 will reduce the submatrix B to B1 • Since B1 is of
order m x m and at least one of the m rows of .A.1 is zero, it follows that one
of the r.)WS of B1 is zero. So det B1 = 0. But, by Lemma 6.5.2 (b), det B
is equal to some nonzero scalar multiple of det Bi, which is zero. This is
a contradiction. Thus, all the rows of A1 are nonzero. In other words,
the rows of A are LI. I
6.5.4 neorem The rank r of an m x n matrix A is the order of the largest
nonzero minor of A.
In other words, the matrix A is of rank r ijJ there exists a minor of
order r which is nonzero, and every minor of order s > r is zero.
Proof: Let p be the integer such that there exists , minor of order
p, which is nonzero, and every larger minor is zero. We shall lk>w prove
that r = p.
Since .4 has rank r, r rows of A are u. Choose such r rows. Let B be
the submatrix of .A obtained by deleting the remaining rows. B is an r x n
matrix, all of whose rows are LI. Therefore, by Theorem 6.5.3, B contains
a nonzero minor of order,. Hence, r < p. ,
On the other hand, look at the nonzero minor of order p. This gives
rise to a p x p submatrix B. This ht turn gives rise to a p x n submatrix
C bf~- C contains a p x p nonzero minor. Hence, by Theorem 6.5.3, the
rows of C are u. This meaos that .4 co,tains p LI rows and so p < r.
Thus,,== p. I
226 / DE'IERMJNANTS

,.s.s ·corollary A square matrix A of order n is nonsingular iff det A ;/::- O.

The proof i'i left to the reader.
rJ -1 27
Example 6.21 Fmd the rank of A = I 0 I -31 by examining the
L6 -1 l_J
determinaht minors.
3 -1 2 ,3 -1 2

I -3 0 1 -3 =- 0.
6 -1 1 0 1 -3

So the rank of A is less than 3. The minor I -3,

-1
1
1 .:/- 0.
So the rank of A is 2.
INVERSE OF A, MATRIX
We are now ready to give another method of finding the inverse of a
nonsingular matrix :
I.et A - (lll.,),.... n be a nonsingular matnx. By CoroJlary 6.4.6, we
bave
AB= BA= Al,
where B = adj A and fl = det A. Since fl "::f::. O, we get

A(~B)~ t!B)A=l.
~
T her.,ore, A-1 = 1 B
'ii: = fll a d'J A

= det
1
A ~~ 7J·
LA1,. A1,. A,.,.
Example 6 22 Find the inverse of the matrix of Example 6.19.

r
We have
1 -1 07 --1
A==
[
0 1lj and adj.if== 1
l 2 1 L-1
lf~re det ..4 = -2. Therefore,
 6 6 PRODUCT OF DETBDIINANTS / 227

r 112 -112 1/27

A-1 = I -1/2 -1/2 1,2 rI.
L 1/2 '3/2 -1/2_J

Problem Set 6.5

I. Give explicitly the proofs of the •1f' part and the •only if' part of
Theorem 6.5.4.
2. Use determinant minors to calculate the rank of the matrices of Pro-
blem 1, Problem Set S.S.
3. Using determinants, prove that the matrices of Problem 2, Problem
Set S.S, are nonsingular. •
4. Use Theorem 6.5.3 to solve Problem& I and 2, Problem Set 3.5.
5. Calculdte the inverse by adjomt method for nonsingular matrices of
Problem 2, Problem Set S.S.
6. True or false ?
(a) If k-th order minors of a matrix A are all zero, then the rank
of A is k - 1.
(bJ If the rank of A 1s k, then every k-th order minor is nonzero.
11 2 37
(c) In the matrix A =I I 2 4 Ithe minor I: : Ii• zero.
LO O O_J
Thec-efore, the rank of A is 1.
(d) If a row and a column of an n ..< n matrix A are zero, then the
rank of .A is less than or equal to n - 2.

6.6 PRODUCT OF DETERMINANTS

In this article we shall study the determinant of the product of two
square matrices In this connection we have the following theorem.
6.6.t Deorem If .A and Bare two square matrices of the same order, then
det (AB) == (det A) (det B) •
The proof of this theorem is rather complic1ted. So we shall be satis-
fied with verifying the theorem for 2 x 2 matrices.
rc,.1•117 r,11 1117
LetA-=I
L•11 «.s...J
I and B-=l
Lf!.i IS.a
j·
228 / DITIIIMINANTS

Then det .4 = «11«11 - «11«11

and det B ~u~n - = ~11f3.1•
We shall now calculate det (.4B). Since
r «11~11 + «11~1 + «1113.17 «11~1
.4B == I I.
L«.1~11 + «.113.1 1111~11 + «■ 1f3.1_J
we have det (AB)= {1111~11 + «1113.1)(«.~11 + ~
- (,xll~ll + IX11~)(«.1~11 + IX11~J •
A merciless simplifk,ation will show that det (..4B) == (det ..4) (det B).
Though we have not proved Theorem 6 6.1. we shall be using it in the
sequel A complete proof can be found in advanced texts on the subject.
6.6.2 CorolllrJ 1/ ..4 is a non,ingula_r matrix, then
1
det (A-1) = (det A)-1 det A .

Proof: We have ..4..4-1 == /.

So det (..4..4-1) == det I== 1
or (det A) (det ..4-1) == 1 •
Therefore, det A -:I= 0 and the result follows. I
Theorem 6.6.1 can be rewritten in the following form.
6.6.3 Theorem (Reworded) If ..4 == (11 11) and B == (~11) are square matr,-
ce, of order n, then (det A) (det B) = det C, where C = (c11).x11 and
n
Cu = I «o:~lld•
k•l
Example 6.23 Let

-l~ n o -11 r-1

I
2 07
A
1 2
3 4
l_J
and B==

L
3 6
1 s
1
3_j
I·
0 -1 I O -1

1bmddA =:; 3
2
4
l
=
r1 - 2r1,
0 3
0 2
6
2
'• - '1

-1: :1- -1 2 0
-6.

and det B- 3 6 I -=
I ,,'• +
+ 3r1,
,,
' 3
 6·6 nooucr OP DBl'llltlalWm / 229

Also A.B ""' l

r-2 -3
11
6 19
42

det (AB) =
-2 -3 -3
11 42 IS =
-2 -3\ --3
11 42 15 =
-2 -3 -3\
1 27 0
6 19 5
'•+Sri
0 10 -4 0 10 -4

l
0 51 -3

== 1 27 o == - 51 - 3 1__ (-204 + 30) _ 174

'1 + 2,. 10 -4
0 10 -4
== (det A) (det B) .

Problem Set 6.6

1. In the proof of Corollary 6.6.2 justify the step 'Therefo1e
det .4 o·. +
2. If A. and Bare square matrices of order n, then prove that
det (ATB) == det (AB') == det (A.TBT) == det (AB).
3. If A is a square matrix, then prove that det (A•) ... (det A.)• for all
positive integers n. (Hint : Use induction.)
4. Prove that the determinant of an idempotent matrix is either 0 or 1.
5. Evaluate det A.1 if A is a nilpotent matrix.
6. Prove that
1 a a• • 3
(a) l b b1 == a + b + c a• +·b' + cl al + b1 + c'
1 C c1 r+IJl+c• a•+b'+cl a'+b'+c'
0 z y 1 y•+r' xy ;,cz

(b) Z O X == xy yz
y X 0
2}'1' - :JI:' zl
zx
,. yz "'+1•
X y I 1

(c)
,.
z' 2zx - y'
:JI:' 2xy - z'
,:I = Y
Z X y
Z X •
230 / DETERMINANTS

7. Prove that
«11 «u oc1a i"·~
I

«n «22 «2a ' = Au A22 A23 ,

I
«a1 «aa 1733 ' Aai Aas Asa
where A;1 is the cofactor of a.;,.
8. Trueor false ?
(a)If A is involutory, then det A = I.
(b) If (det A) (det B) = J. then AB = I.
(c)det (A + B) = dct A + det B.
(d) (det (cA)) (det (B)) = c(det A) (det B), where c is a scalar.
(e)det (AB) = det ( BA).
(0 det E ~ R (?<A,., ,.).
(g) If A and Bare lW<' square n1.1trices such that A = C-1 BC for
some invertible matrix C, then det A = det B.

6.7 APPLICATIONS TO LINEAR EQUATIONS

Recall Theorem 5.8.2. The systi:m of" equ:itions inn unknowns
rl11X1 +«12X2 + ... + «1nXn = ~1
..
«i1X1 -j «22X2
..
+ ,·· + Otz,.X,. = ~a. (I)
ct,11X1 + ctn2X:a + ·•• + «nnXn = ~ ..
h. s a un:que i,olution ifJ the m, trix A = (:x.,),.x,. is nonsingular, i.e.
det A =r- 0 (cf Coiollary 6.5.5). We shall now give a method, using
determinants, of finding the solution when det A =r- 0.
Let the cofactor of «;1 in det A be A,,. Multiply the first equation of
system (I) by A11 , the ~econd by A11 , and so on, the last equation by A,.i,
and add. We get
X1(«11Au + «21A11 + ... + «,.1A..1)
+ Xa(«iaAu + «aaAai + ... + «,.zA,.1)
+ ...............................................
+ x,.(«1,.Au + «a,.Au + ... + «,.nA111)
= ~1A11 + ~2Au + ... + ~.. A,.1,
In view of Theorem 6.4.4 and Remark 6.4.S;this equation becomes
X1 dct A == ~1A11 +
~aA11 + ... + ~,.A,.1
~l «11 «111

-
 6 7 APPLICATIONS TO LINEAR EQUATIONS/ 231

The last equality can be seen by exp,.inding this new determinant in terms
of its first column :

} ~J IX29 IXu
So Xi= det A

(3,. a:,., OlriJ OCnn

In the same manner, if v.e multiply the fir!>t equ 1t1on by Au, the
second by A2, and so on, the last by A 111 , and add, we get
IXu 0'12 ••• IXi, r l 131 IXi, ,tt °'111

I Cl21 ot22 ' ' ' ot2, i -1 132 Ol~, 1 H CX.2n

x, = det A I' (2)

IXnt ot,.2 ... IXn, ,-t ~ . or.,., ,+I ... 0Cnn

I

for allj = I, 2, ... , n.

This method of solving a system of n equations in n unknowns is
known as Cramer's rule. This can also be derived as follo\\s :
System (I) can be written in matrix notation as
Ax= b, (3)
where A is the square matrix (ixo)nxn, x = (x10 X2, ... , x,.)T,
b = (~1, ~1 , ... , ~n)T, Since A is nonsingular, we get
A-1(Ax) = A-1b
or
X = A-1b. (4)
But A-1 = de! A adj A (cf§ 6.4).
Hence,
x = de: A (adj A)b (5)

rx1 7 Au .Au A"1 7

x. Au A11 A"• rf31 7
1 f31
or = det A
x, A11 A11 Am

: ..• L..f3• .J
x• .J t..A1• A,. ... A.. .J
Therefore,
1
"' .... det A (~1A11 + ~A11 + ... + ~.A"')
232 / DBTBRMINANTS

at11 ll11 ... ~1 ... «1.

1 «11 etas ~ ... «••

= det A ..
IXn1
"•• ~-
t
j-th column
«••

Example 6.24 Solve, by Cramer's rule,

2x - 3y + z = l
x+ y-z= 0
X ·- 2y + % = -1 .
r2 --3 17
Here A. = J 1 1 -I
LI -2 I_J
and det A. == 1. Therefore,
1 -3 1 2

0 1 -1 = 1. JI= !1 0 -1 = -1,
1 1 -1 1
2 -3 I
and z =--r1 I 1 0 = -4.
-1 -1 I
Note that if det A = 0, then A. has a rank less than n and so Ax == 0
bas nontrivial solutions. In t.bis case Ax = b may or may not have a solu-
tion. Even if it has a solution the solution will not be unique. Converse-
ly, if Ax == 0 has nontrivial solutions, then rank A is less than n. So
det A. == 0. Thus, we have proved the following theorem.
6. 7.1 Theorem q A. == (a:u)11X•• then A.x =0 has nontrMal solution, i/1
det A =- o.
6.7.2 R•art When det A = 0, Cramer's rule cannot be used. (Why?)

Problem Set 6. 7
I. Use Cramer's rule (if applicable) to find the solutions of the following
systems of linear equations :
(a) 3x + y = 1 (b) 2x - 3y = 1 (c) X + 2y + 31' - 3
5x+2y==3 x+4y-l 2x - s-4
4x+2y+2z=5
 6.8 DGBNVALUl81 mGl!NVBCTOU / 233
(d) x - y + 2z = 1 (e) X + y + 2z = 3
2x +2z=3 2x+2y+2z==1
3x + y + 3z = 7 3x + 4y + 3z = 2 •
2. Repeat Problem l for the systems "r linear equations of Problem 1,
Problem Set 5.8.
3. Find the inverse of the coefficient matrix for the systems of linear
equations of Problem I by any method, and use Equation (4) to solve
these systems.
4. If t'he coordinate axes in a plane are rotated through an angle «, then
the old coordinates (x, y) are expressed in terms of the new coordi-
nates (x', y') as
x=x'cos« -y'sin «
y = x' sin « + J/ cos « •
Use Cramer's rule to express (x'. y') in terms of (x, y).

6.8 EIGENVALUES, EIGENVECTORS

An 11 x n real matrix ha& been considered as a linear tran~formation
from VII to V11• The simplest type of linear transformation is the one that
merely multiplies all vectors x in V11 by a fixed scalar, that is: Ax =- ).x for
some fixed A Nuw we shall discuss those situations where the effect of a
linear transformation is simply scalar multiplication, at least in part of the
domain space. In the process we shall see that determinants come in very
handy. In this anicle we shall deal with complex \lector spaces and com-
plex scalars. Consequently, the matrices may have complex entties also.
First, we shall take up a 3 x 3 matrix :
r «11 Gtllll «1a 7

A = I «11 «a1 «ta I·

L «11 «a■ «11 _j
Our aim is to find out whether for some nonzero vectors x == (x1t x1, x 3) 1 ,
it is possible to have
Ax= AX (I)
for some suitable scalar >.. From Equation (1) we have
Ax-AX==O,
or
Ax - >,.(Ix) == (A - )J)x - 0 '
(2)
r«u - ).. rx17

lI
«11 «11

i.e.
I
Gt11 «n - ).. Gt13 X1 1-•o.
L «11 «.1 «u-l...l Lxa...l
234 / DETl!RMINANTS

Such an equation has nontrivial solutions, by Theorem 6.7.1, iff

I .Xu - A
I
I
=0. (3)

IX31 IXaz IX33 - ).

Expanding this determinant, we find that Equation (3)-reduces to an

algebraic equation of the third degree in )., Since our scalars are complex
numbers, we are assured of three roots for this equi..tion. Let us call these
roots ).1 , ).2 , and ).3 • f"or each of these ).'s, it follows that Equation (I) has
a nontrivial solution. Denote the set of these nontrivial solutions of Equa-
tion (I) for a given ). by E(>..).
The values >..1 , >.. 2 , and ).3 of). are called eigenvalues of the matt ix A., and
each vector belonging to E'().) is called an eigenvec·tor ·corresponding to the
eigenvalue >••
Corresponding to eJch ).,, i = I, 2, 3, we have a set of eigenvectors,
denoted by £().1), i = I, 2, 3.
The precise definition is us follows.
6 8.1 Definition If A is a squnre matrix of order n, then the values of).
for which the equation
Ax--= b· (4)
has nontrivial solutions are called tl~e eigenvalues of A. If ). is an
eigenvalue, then the nonze10 vectors x, for which Equation (4) holds,
are called eigenvector., corresponding to the eigenvalue>..
Eigenvalues are also r.alled 'characteristic values' or 'proper values'.
Similarly. the terms 'characteristic vectors' and 'Proper vectors' are also
used.
By Theorem 6. 7.1, it follows that eigenvalues ). are simply the roots of
the algebraic equation
det (A. -- >.IJ == O • (S)
Equation (5) is called the characteristic equation of the matrix A.
det (A - )./) is a polynomial in >. and is called the characteristic polynomial
of the matrix A. Note that this polynomial has degree n. Hence, there
cJn be at most n distinct eigenvalues.
Example 6.25 Let
r1 -1 27

A== I 0
Ll ::J
To find the eigenvalues of .4, we solve
 6 8 EIGENVALUES, EIGENVECTORS / 235
I - 'A -1 2
0 1 -A
1 2
This gives
(1 - 'A)((l - >.) 2 -I (--2)) = 0
or (I - >.)(>-2 - 2>. - - 1) =-- 0 , •
i.e. ). = 1, 1 + v2, 1 - y2.
So the eigenvalues of A are
11 == 1, >-2 =-= I -I v2, >-a =- I - v'2.
To find E(A 1) = E(l}, we write
r-1 - l 27 r-x 1 7 1 -1 fl 0 r-x
I O O II Xi !~ I I I:~ l I O
x~ 1
LI 2 I _J L_x3 J L xJ _l L._ 0 0
r-o -1 r1 1x1-1 107
i.e. I o, 0 0 j I Xi : - ! 0
I_ I 2 O_J I_ ,J _J I _0
or - .\J + 2x 3 = 0, ., 1 -I 2xl =- 0.
This gives x 1 =- 2.:\ 2 nnd AJ = 2xJ, The eigenvectors corresponding to
the eigenvalue I a1c Lherefore of the form ( 4xJ, :ZxJ, x 3), 'whe1e x 8 =t= 0.
Therefo·e,
l,(I) = [(--4, 2, 1)1\{0}.
To find .E(i.2) - ~ E( I ~ \/2), we wtite
r- 1 -1 2 -1 •'"1 7 r- r-.,.1 7
IO 1 0 I ! ·'"2 I = (I !- y'2) X2 I,
I

LI 2 l_J LXa_J LxJ _J

This gives
1 -v'2 -1 2 7 f-X1 -I 107
I 0 -y'2 0 Yi = 0
IL 1 2 -,v2_J
I Lxa_J
I
LO _J
I
I

Solving this, we get

E(l + v2) = [(v2, 0, l))\{O}.
Similarly, E(l-y'2) = [(-v2, 0, l))\{O}.
6.8.2 Theorem /f). is an eigenvalue of an n x n matrix A, then the set
K'A = E(A) U {O} is a subspace of the vector space ~ of n-tuples of
complex numbers.
236 / DITBRMINANTS

Ki is called the elgenspace corresponding to l., The proof is left to the

reader. Note that K'A is the set of all solutions of the equation Ax .. >.x.
It is clear from Theorem 6.8.2 that if II belongs to E().}, then every non-
zero scalar multiple of , is also a member of E(l.). Therefore. if
111 E E('A1) and'• e: E(>.1), then {111, 111} is LI, because neither is '111 a non-
zero scalar miltiple of 111 nor is 111 a nonzero scalar multiple of 111 • Thia
idea does not stop here. It can be further extended. More precisely, we
have the following result.
ti.8.3 Lemma Let A be a square matrix of order n having k diltlnct
eigenvalues >-1t >.1, , >.1 • Let v, be an eigenvector corresponding to
the eigenvalue>.,, i = I, 2, ... , k. Then the set {111 , 118, ... , '111:} i8 u.
Proof: We shall prove this lemma by induction. The lemma is true
for k - I, i e. {111} is LI, because v1 -=f; O. Now suppose that it is true for
k ..,, r, i.e. {v1, 118 , •• , vr} is LI. We shall now prove that it holds also for
k == r + 1, i.e. {vJ, , 1 , ••• , ' " 'r+i} is LI. Let
«1111 + «1V1 + ... + Ctr1'r + Ct,+1Vrf-1 == 0 , (6)
Th~refore, A.(et11'1 + et1V1 + ... + Gtr1'r + Gtrf-i'rf-i) = AO = 0 ,
i.e.
et1A1V1 + «1As'1 + ••• + etr>-,.1'r + lltrf-1~1"r+-1 == 0~, (7)
because A is linear and .A.,, = i,v,. Multiplying Equation (6) by >.,.+.1 and
subtracting from Equation (7), we get
«1Cl.1 - >..-+-1)1'1 + «1('A1 - >.,.+.Jv1 + •.. + ar:,.('Ar - >.rf-1)Vr == 0 .
Since {v1, '•• .. , Vr} is LI and 'A, =f::. >.r+-11 I == 1, 2, 3, ... , r, we get ar:1 == 0
= ar:. == .•• == «r• Substituting these values of «,'s in l:!quation (6), we get
«r+i'r+1 == O. Therefore, "r+1 = 0, which completes the proof. I
We shall now prove a fundamental property of eigenvalues and eigen-
vectors, which will be useful in the further development of the subject.
6.8.4 Tbeonm Let A be a square matrix of order n having n distiMt
eigenvalues >.1, }a, ... , 1,.. Let v1 be an eigenvector corresponding to
the elgen,olue 'A,. i = 1, 2, ... , n. Then the set {,u 111 , .. , v.} i8 a
basis/or the domain space of A. The matrix of the linear transfor-
mation A with respect to the basis {v1 , , 1, ... , v.} Is
>-1 0 0 0
0 }aO 0

0 0 >.a 0

LO O O >..
P100/: The set {1111 v1, ••• , v.} is u by Lemma 6.8.3. Since the
domain space is n-dimensional, it follows. by Theorem 3.6.7, that the set
 6 8 EIGBNVALUU, EIOENVBCl'ORS / 237
{vi, , 1 , ... , v"} is a basis for the domain space of A. For the second part
of the tbe'lrem, note that Av, = >.,v, = Ov1 +
Ov1 + ... + 0v,_1 + >.,v, +
Ov,+1 + ... + Ov". Thus, the i-th column of the required matrix is
(0, 0, ... , ).., .. 1 O)T •
t
,-th place
Hence the theorem. I
l11 such a case the matrix is said to have been diagonalised.

Problem Set 6.8

1. Determine the eigenvalues and the correi.pondmg eigenspacec; for the
following matrices :

(a)
r3
I
17
I (b) I
1-0 -1-1
I (c)
I 0
I I
3 1
(d)
13 l2
2
o2
47
L6 2_J L_l O_f L2 - I _J
L4 2 3_J
r 3 2 l 0"'1
:1 0 07 r-o ; I ]

I
I
0 l 0
(e}
·l2
1
0
(f)
Ii 0 l
I
(g)
0 2 -l 0
L3 2 O_J Li i O_j
LO 0 0 i.J
2. Diagonalise the matrices of (b), (cJ, and (g) m Problem I.
3. Find the characteristic polynomial of the matrix
r l 0 0 07
-1 i 0 0

2 ½ -I 0
'
L t -i '11: -lJ
Diagonalise this matrix, if possible.
4. If). is an eigenvalue of the matrix A, prove that
(a) ).1 is an eigenvalue of A 1
(b) )." is an eigenvalue of A"
(c) «>. is an eigenvalue of «A, whe1 e « is a scalar
(d) gQ.) is an eigenvalue of g(A), where g is a polynomial.
S. If x is an eigenvector of A corresponding to the eigenvall•e ).1 prove
that
(a) x is an eigenvector of A" corresponding to the eigenvalue >.•
(b) ~ i$ an eigenvector of g(A) corresponding to the eigenvalue I(>.)
238 / DBTEllMJNANTS

6. Prove that>,. = 0 is an eigenvalue of the matrix A iff A is singular.

7. If>.. is an eigenvalue of the matrix A, prove that
(a) >.. is also an eigenvalue of A 1
(b) I/>.. is an eigenvalue of A-1, if A js nonsingular.
8. If x is an eigenvector of A corresponding to the eigenvalue >., prove
that
(a) x need not be an eigenvector of A 7 corresponding to the eigen-
value i.
(b) xis an eigenvector of A - 1 ·(if A is nonsingular) corresponding
to the eigenvalue If">..
9. Prove that the eigenvalues of a triangular matrix are its diagonal
clements.

6.9 WRONSKIANS
In this article we shall use determinants to discuss dcpendeuce anb
independence of functions. These ideas help in an understar.ding ot the
theory of ordinary linear differential equations. First, we shall c!efine the
Wronskian of functions.
6.9.t Definition If Yi, Yh • , y,, are n functions in •(/ '"- 11 (/), tl--en the
determinant
Y1(x) Y2(X) y,.(x)

y;(x) y~(x) y,.(x)

W(x) = ..
y/"-ll(x) Ya'"- ll(x) y,.ln-ll(x)
is called the Wronskian of Yi, y 2 , •• , Yn at x, and is denoted by
W[y1(x), Ya(x) ... , y,.(x)J.
Note that the Wronskian is a function of x.
Example 6.26 The Wronskian of the functions x ard cos x is
i
W[x, cos x] = I1
X COS X

-sinx 1
= -x sin " -- cos x .

The linear dependence or independence of the functions Yi, y1 , ... , y,.

plays an important role in the solution of an ordinary linear differential
equation. The following theorem tells how the dependence or indepen-
dence offunctions Yi, y,, .. , y,. is related to their Wronskian.
6.!>.2 Tlleorem Let )'11 Y,, ... , Yn E <?I' 1" 1(/), and let W(x) be thtir
Wronslcian,
 6.9 WllONSKIANS / 239

(a) 1/ there exists x0 E I such that W(x0) =I= o. then Y1t >'•• ... , Yn are
LI over I. •
(b) _J/ there exi,ts X 0 E I S&U'h that W(x0) = 0 and if ,Yi, y1 , •• , Yn
are solutions of the dijferentlal equation
a0ylnl + ai11t•-11 + ... + Dn-i y' + DJ = 0 , ( 1)
where "o• Di, ••·• a,. e: '"(( (1), then Yu y 1 , .... y,. are LD over I.
(c) If Yu Ya, .... y,,. are solutions of Equation (1) and are LI over 1,
then W(x} neh r fJanishes on I.
(d) // Y1t Y1, ••• , Yn are solutions of Equation (1), then either W(x)
= 0 over I or W(x) has no zeros in I.
Proof: (a) Since W(x0 ) /- 0, W(x0) considered as a square matrix
is nonsingular by Corollary b.5.5. Hence, by Theorem 5.5.2, the column
vectors
,.(xo) = (y,(Xo), ,;cxu), ...• y, Cn- 11 (xu)). i = I, 2, .. , n (2)
arc LI in V,,.. Therefore, y1, y., ... , y,. are LI over/, by Lemma 4.9.5.
(b) If W(x0) = 0, the columns or W(x0 ), namely, ,,(x0) =--= (y,(x0),
y;(xJ, ... , y 1«n-1>(x0)), i = I, ... , n, are LD in V,., by Theorem S.S.2. There-
fore, there exist seal us C1, C1, ••• , C,,, not all .zero, such that
• C1v1(x0) + C1 v11(x0) t- ••. + C,.1 ,.(.T0 ) = 0.
1

This means
CiY1'1:1(xo) + c.y.li:l(xo) + ... ~ C,.y,." 1(Xo) =0, k - 0, I, 2, ... , (n - I),
(3)
for at least one scalar C,. not zero. Consider the linear combination y ==
C1y 1 + C1y 1 + ... + C,.y,.. Since y 1 , y~, •· , y,. are solutions of Equation
(I), y is also a solution of (I). But by Equation (3), we have
y(x0) = y'(x0) = ... = y«n 11 (x0) = 0. (4)
Thus, y is a solution of Equ.ition ( I) and satisfies the initial conditions (4).
But the zero function also satisfies Equation (I) and the initial conditions
(4). Hence, by Theore111 4.9.2 (existence and uniqueness), y(x) = 0 for all
X in/.
Thus, C1y 1(x) + C1 y 1 (x) + ... + C"yti(x) = 0 for all x E / and at
least one C., =p 0. Therefore, Yi, )"1, .. , .Vn are LD over /.
(c) This is the contrapositive statement of (b).
(d) It follows from (b) and (c). I
6 9.3 '11leorem If y 1 , are solutions of
)'1
"9Y + 01)' + D1Y = 0,
11 1
(5)
where a., a1, and a1 belong to fjj' (1), and a0 is noM here zero on I, "then
_ J a1{xl dx.
W(y1(x), y 1(»))'- Ce t1t(x) (6)
Proo/: ,, and 1, •re soh1tione of (S), Therefore,
240 J DBTERMINANTS
• Qa)'~ + 01Y; + Os.Yi = 0
and ao1; + Ol)'~ + a1y8 = 0.
Also WC,,i, Y1J = I1; y;
Y1 Ya 1 ,
I -= Y1Y1 - Y1Y1 •
I
,

dW ,· • • N , , • •
So dx = Yi>'• + YiYa - Y1Yz - Y1Ya = Y1Y1 - >'i.Y•
= Yi ( -- a1
Ya, -- -a, Ya ) - Ya { -
Oo Oo

01 (
= -- ,
Y1Y1
')
- Y1J'~ = -- -01 W•
Oo Oo
Hence, on integration, we obtain
. - l Oi(x) Jx
W[y1(x), y 2{x)] --= Ce a,(x) ,
where C is a:1 arbitrary constant. I
6.9.4 Remark The foregoing result can be extended to a differential
equation of the n-th order. Here "'e state the general result. If
y., Yx, .. , Yn are n solution& of the n-th order normal hnear differen-
tial equation ao1'"1 -t a 1y 1" ·u + ... + an1· -= 0, on a'h interval I,
- I a1~xl Jx
then W(y1(x), J2(,t), , Jn(x)] - Ce a.(x) (7)
Equations (6) and (7) are known as Abel's/ormulae.

Problem Set 6.9

1. Use the Wronskian to prove that the following sets of functions are
u over J:
(a) es", eb1 ; a =I= b ; I any interval
( b) m,x m.,r m.x
e ' e ' ·• · ' e . ;. ; I any intervtlt
m1 , m., ... , mn all distinct
(c) 1, x, x• ; I any interval
(d) 1, x, r, ... , X"- ;1 I any interval
(e) e-, sin 2x; I any interval
GIX CU
(f) e , xe I any interval
IXX • «X
(g) e sm ~x. e cos ~x ; I any interval
(h) 1n x, x In x ; I = (0, co).
2. Use Abel's formula to find the Wronskian of the linearly independent
solutions of the following normal linear differential equations :
(a) ,• - 4y' + 4y = O
(b) y" +
2xy- 0
 6.10 CROSS PRODUCT IN Y1 / 241
(c) (1 - x 1)y• - 2xy' + 2y = O; y 1(0) = y;(O) = 2
Ya(O) = - y;(O) = 1 •
(d) x 1• - 3xy' + y
1 = O;Yt(-1) = Yi(-1) = 2,y1(-1) =O
y;(-1) = -1. •
3. What is the value of C in Equation (7) if J'i, y 2, ••• , y,. are LD?

6.10 CROSS PRODUCT IN Va

In this article we c;hall dii.cus.s the product of vectors m V3 , mentioned
in§ 2.3.
6.10.1 Definition If u = (a1 , 0 2 , aJ) and v 0 - (bi, b2 , bJ) are two vectors
in Va, their cross product, written as u x v, i!, defined as the vector
u X 11 ~ ( \ 02
I b2 bJ
°a : • \ "a
I b3 b1
0
1 I•I I) . 0
1
b1 b2
OJ

Since the cross product is a vector, it is aho called vector procluct.

In simplified form Definition 6. l O. I say!>
(ai, a3 , 0 8) X (bi, 'b=, hJ) -- (a~bJ aJb~, a3b1 - a 1h3 , a1b8 - a1b 1) •
If we u<;e the unit vectors i, j, k, we obtain
(a1i + ad +
a3k) x (bii t-- b 2 j -t bJk)

i
b3 1

which is easily remembered in the determinant form

I j k
(1)

• b1 b,. bJ
To obtain the cross product we expand (I} in terms of the first row as I/
it were a determinant. Note that (I) is not a 'pukka' determinant.
From Definition 6.10. l it follows immediately that
i X j -=-= k, j X k = i. k X i --= i.
j X i = --k, k X j = -i, i ,< k-= -j,
; x i = 0 = J x j = k x k.
.
and
Note that the cros~ prodcct is a binary operation on Va.
Example 6.27 (1, 2, -1) x (0, -1, 3) .

- (I -1
2 -1
3
I· 1-l I· 11
3 0·
1
0 -1
21)
== (S, -3, -1).
242 / DBTIIMlNAN'II
6.10.2 Theorem (Properties of cross product) ut u, ,, w be any three
•ectors In Y1 • Then
(a) u x , = -(, x u).
(b) u X (• + w) = u X v w.+U X
(C) For (lt E R, GtU X ' = «(u X v) = " X GtY,
(d) u • (u x v) = 0, and v • (u X v) = 0, i.e. u X 11 11 orthogonal
to both II and 11.
(e) I II x v 1== I u I 2 I II I 11 - I 11 • 11 I 11-
Since the proofs are straightforward, we shall prove only (e).
Proof o/(e): Let 11 = (alt a 1 , a3 ) and v = (b1t b1 , ha). Then ,
lul 1 l11l 1 -l11•vl 1
=- (~ + a: + a!)(b! + b: + b:l - (a1b1 + aaba + aaba)1
= E a~(b: + b:) - 2 I a 1b1asb1
= I (~b= + a:b!) - 2 I a1b1asb1
= E (a1b1 - a-61)1
== lux,1•. I
6.10.3 Corollary 1/8 is the angle between two nonzero vectors u and II in
Y1, then
I u x II I = I u I I , I sin 8 •
Proof: We have
u • v = Iu I Iv I cos 0 (cf Definition 2.3.3).
Hcnc:c, from (e) of Theorem 6.10.2 it follows that
I u x II I 11 = I u I I I v I 1 - I 11 11 I II I I cos11 8
= I II I I I v I a sin1 8 .
So I u X II 8 <: 1t. I
I = I II I I II I sin 8, since O <
Note that I u I I v I sin 8 is the area of the parallelogram formed by
vectors u and ,. Hence, I u x v I represents the area of the parallelo-
aram formed by u and v.
DIRECTION OF u X v
Given two vectors u and v, we have just proved that
(i) I u X. v I = I u I I v I sin 8, where 8 is the angle between u and v,
and
(ii) u X • is perpendicular to both II and v and therefore to the plane
containing u and,. ·
(i) specifies the magnitude of II x v, but (ii) leaves us to decide which
of the two directions perpendicular to the plane of u and , is the direction
of u X •· We shall now settle this question. •
We start from the ordered triple of vectors (i.J, i). Any ordered
triple of linearly independent vectors (u, .,, w) is said to be a poritiN triple
if, b7 f8dually,changing the directions of u, .,, w b11t without making them
 6.10 CllOII HGDIJCI' DI f'1 / 243
linearly dependent (i.e. without making any of the veetora cross the plane
of the other two), we can bring them into coincidence with the directions
of I,}, and le. Otherwise it is said to be a negatlie triple.
Examples of positive triples arc
(i, }, k), (}1 k, i), (k, i, }).
Examples of negative triples arc
(}, i, k), (k,J, i), (i, lc,J).
6.10.4 Lemma Three linearly ind,pendent vectors u = (ai, a1, a1),
• = (bi, b1, ha), w = (ci, c11 c3) form a positive triple iJf
a1 a2 o1 l

b1 h1 b81 I> 0.
Ci Ca Ca •

Proof: A positive triple, which actually coincidet. with the direc-

!ions of l, }, k, 111 of the form
u - oi, v = bj, w = ck,
where a, b, and c are J>Oi.ihve. So the determmant of the lemma is
1
a o o
b 0 = abc,
1:
0 C
which is positive Now any other positive triple is obtained by gradually
changing the directions of u, v, and w, but without ever making them w.
This proce'!; continuously changes the value of the aforesaid determinant,
but without ever making the determinant take the value zero (since such a
value will correspond to making u, v, and w LD). Hence, every positive
triple has a po&itive \'alue for th1,; determinant and vice versa. I
Now we can see that u x v, as defined, has that direction perpendi-
cular to the plane of u, v that makes u, v, u x v a i,ositive triple; for the
determinant of Lemma 6 10.4 now becomes
UllV

a1b1 - a3b1 a 8b1 - a1b1 a 1b1 - a1b1

whose va luc is I (a1b1 - a.,b2) 2, \\ hich is positive.
Thus, u X v is given by Figure 6. 1.
In other words, if the angle between u and v is 81
0 < 8 < "R, then the direction of u x • is the o~ in
which a right-handed screw would move when rotated F1ouu6.l
through the angle 8 from II to ,.
244 / DBTBRMINANTS

6,J0.5 Delllltfoa If u, ,. and w are three vectors in Y,. u • (v x w) is

called the Nalfll' triple product of u, v, and w.
Note that, even if we remove the parentheses in Definition 6.10.S and
writer, • v x win place of u · (v x w), there cannot arise eny ambiauity,
since (11 • v) x w is meaningless. (Why ?)
6.10.6 Tbeonm (Properties of scalar triple product} Let 11 = (a11 a1, a1),
, == (b,. b1• b3 ), and w = (C1t c1 , c3) be three ,ectors in Va, Then
, a 1 a 1 a1

, c1 Ca Ca

(b) II • 1' X w= " w 1' • w X u

X 1' • = = 1' X w. II
=-W•UXl'=WXU•V,
{c) I u · v X w I represents the volume of the parall,lepiped formed
by the vectors u, ,, and w.
_ Proof: (a) Using the definitions of dot and cross product, we
have 11 • , x w

"'"'01
'ha b1
I C1 Ca
II+ a. Iba
t'1
b1

Ci
I + a, I
I b1

Ci
b1
c.
I
a1 a. aa I
I
= bi b1 ba ! (by expansion of the determinant in
terms of the first row).
C1 C1 c. I
ai a1 Di,

(b) U•l'XW= bi b. b,
£'1 C1 Ca

bi ba ba
= c1 c1 c8 ( by two interchanges of rows)
a1 a 1 a 8
=1'·WXU.
Similarly, 1' • W X 11 =
u X 1'■
W •
Thus, 11 • , x w = v • w x 11 = w • u x v •
Since the dot product is commutative, these are also equal to , x w • u
== W )( U • V = r, X 1' • w.
(c) If one of the \!eCtors is zero or , x w ... O, the proof is obvious.
We theref'orc usume that none of these vectors is zero and so also
,xw~o.
 6.10 CkOSS PllQpUCT IN Ya/ 245
Hence, the vector v x w is perpendicular to the plane detemuned by
the vectors v and w (cf Theorem 6. I0.2 (d)). So either " x w or -(v x w)

..
makes an angle ; < n/2 with u, as shown in Figure 6.2.
,,

,, ,,
,,,,,,,.. __

f
------·,,,
,,.,,,
,, ,,, I
.,, ' .,, I
.,,.,, f ,,," I
,,, I ,,,, I
_,, I / I
- - - - - -,,- - - - - - - - 7 I

I
f ' f
I ' f
f '
f I I
I I I
I ,' f
------ -t- -----~
I ,,
I , .,,
I _,,
I
I .,,
I /
,.,,,,,
"I
FIGURE 6.2
So lu•vxwl= lul lvxwlcos,f,
= ( I u I cos ,f,) I V X w I
= (height of the parallelepiped formed by u, v, and
w) x (area of the parallelogram formed by v
and w)
= volume of the parallelepiped formed by u, v,
and w. I
6.10.7 Corollary u • v X w = 0 if! u, v, and ware LD.
Proof: From the proof of rart (c) of Theorem 6.10.6, it follows
that
Iu . V X wI = I u I I V X w I cos ,f,
= I u I I v I I w I sin 8 cos ,f, •
Hence, u • v x w = 0 means one of the three vectors is O or sin 8 = 0
or cos ,f, = O. If one of the vectors is 0, then u, v and w are LD, If
sin 8 = 0 or cos ,f, = 0, then the three vectors are coplanar and hence LD.
Sou • v x w = 0 implies that the three vectors are LD. The converse is
obvious. I
We shall now consider two important applications of the cross product
to geometry.
(i) Vector equation of a plane through three glvtn points P1 , P1 , Pa in
space : Let Q be a point in the plane P 1 P1P1 (see Figure 6.3). The vectors
 2# / 1>61DIII.NANTS

--+- --+
P1P1 and P,P1 lie in the plane. The problem now reduces to finding the

FIGURE 6.3
equation of a plane through P 1 and perpendicular to the vector

- -- -
~ - -➔
N = P1P1 x P1P3 • Hence, the required equation is P1Q · N = 0,
i.e. P1Q · (P1P1 X P.,P1) = 0 • (2)
This equation is satisfied by all points Q on the required plane and by no
points oft' the plane.
If Pu Pa, and P 3 are (x1 , Yu z1), (x1 , Ya, z1 ), and (xJ, Ya, z3) respectively,
then Equation (2) becomes
X - X1 y - Y1 Z - Z1

Ya - Ya Zs - Z1
Xa - Xa

Example 6.28 The equation of the plane through P 1(0, 0, O), P 1 (1, -1, 1),
and P 1(2, 1, 2) is
X y Z
1 -1 1 = 0,
1 2 1
i.e. x - z = o.
(ii) Shortest distance between two llne.r : If we have two lines in a
plane, then either they are parallel or they intenect. On the other band,
two lines in space can behave in any one of three ways : (a) they are
 6.10 CROSS f'RODUCI' IN Ya / 247

parallel or (ltl they inter1iect or (c) neither are they parallel nor do they
intersect. In case (c) they are called &kew lines. Let L1 and La be two
skew lines (see Figure 6.4).

FICiURE 6.4
Let Pi, Q1 be two points on L,,, and P 1, Q1 two points on La- 'I'he
-- -+
vgctor P 1Q1 x P1 Q1 is perpendicular to both L 1 and La- The scalar pro-
--+
jection of P 1P1 qn this perpend1culur is the required shortest distance
between L1 and £.. It is given by
J -+
P1P1 -
• P1Q1 X -P.Q. I
- --➔
I P1Q1 X P.Q. I
Example 6.29 Let the two lines be
x-1 y-2 z-1 x-3 y-2 z+4
L1 3-
: -- = - 2 = S- and La: -,- = - 2- = -r.
Since L 1 is parallel to the vector (3, 2, S), and La is parallel tc the
vector (1, 2, 2), the vector N = (3, 2, S) x (J, 2, 2) is perpendicular to
both L1 and La- We can take P1 as (1, 2, I) and Pa as (3, 2, -4).
Hence, the shortest distance between L,, and La is
2, 0, -5) • (3, 2, 5) X (1, 2, 2)
I (3, 2, SJ X (I, 2, 2)
2 0 -51
= 3 2 S I+ \/'36 + 1 + 16
I 2 2 j
= I -12 - 20 I + v'53
32
=- --=- •
\/'53
,.18.I Remark The method in Eu.mple 6.29 fails in ~ the lines L 1 and
La are parallel. (Why?) In this case let P 1 and (21 be two Points
 248 / Dll'IDJIINAN18

on Li, and P1 a point on La (FJ,ure 6.5). Then the scalar projection

-+
of P1P1 on L,, gJVes
-➔

--
--
d = J P1P1 P1Q1 I
1 P1Q1 I

-----.,,---....-•------L1
.,,,, ,,
01

.,,,,
,,,,,,

FIGURE 6.5
Hence, the (shortest) distance p between £ 1 and La is given by

P = JI Pi~ I 2 --~ 2

I =....- -
= JI I --+
P1P,. 2 -
I P1P1 · __
__ --:...::.+.
P,Q1_I 11
I PlQl I 1

--
---+ -~
I P1P1 X P,Q, I
= (cf Theorem 6.10.2 (e))
1 P1Q1 I
Problem Set 6.10
1. Evaluate u x v for the vectors u and " 10 the following :
(a) u = i + +
J k, v = 2i - 3} k +
(b) u = i - 2k, v = 2j + 1k
(c) u = (1, 3, O), 11 = (2, -1, -1)
(d) u == (-1, 2, 3), v = (I, I, 2).
2. Prove that O x " == 0 for every vector v in V1 •
3. Prove the distributive law u x (11 + w) == (u x 11) + (u > w) and
verify it for 1he vectors u, v, and win the following:
(a) u == i - 2} + k, v == 3i - j, w = 'li + } - 4k
(b) u = (1, 2, 3), v = (3, 0, 1), w = (2, 4, 5)
(c) u = (-1, 1, 2), ,.. = {I, -1, -1), w = (41 2, -3).
4. In the discussion of direction of the vectOl' u x 11 we have taken
O < 8 < ff, What happens when 8 == 0 or 8 == ff ?
5. Determine whether the vectors u x II and w lie on the same side of
 6 JO CROSS l'lt.ODUCT IN f,'1 / 249
~he plane containing the vectors u and v for the vectors u, v, and w
m Problem 3.
6. Fmd the equation of the plane passing through the three pomts
(a) A(l, 2, -3J, B(J, -1, I), and C(2, I, I)
(b) A(l, 2, 3), B{J, 0, 1), and C{2, 4, 5)
(c) A(l/2, 1, -1), B(l/2, -1, -1), and Cl4, 1/3, -3).
7. Prove that for vectors u, v, and w of V,1
(a) (u x v) x w = (u w)v - (v • w}u
(b) u X (v X w) - (u • w)v - (u • v)w
8 Evaluate the product
(a X b) X (c X d)
of the four vectors a, b, c, and d of V8 • P1 ove that this 1s a vector in
the plane of a and bas well as of c and d.
9 Find the shortest distance b,!tween the skew Imes
X-2 L-t-1
3 -= ., 2 =- I - z and x = y - z .

10. Fmd the equation of the plane contammg the hne

X -- 2 = y_±__! -=- I - z
3 2
and rarallel to the lme x = y = z.
11 . Show that the ltne
L . X ..±1 = y =l - z t- 2
. 2 3 4
1s pa1allel to the pl,me
P : 2x - 4y -I •2z = 3 ,
and find the equation of a hne lying m the plane P and perpendicular
to the lme L.
I~. Find a plane through P(l, 2, 3) and perpendicular 1o the line of
mtersection of the planes 2x + 3y - z = 1 and 3x - Y - 7 = 4.
13. Find a plane through the pomts P(l, --1, 2) and Q(2, 0, 1) and
• perpendicular to the plane 3x - 4y + z ~ 0.
 Chapter 7

More Matrix Theory

In this chapter we shall discuss some miscellaneous topics to give the

reader a glimpse of what lies beyond. The intention is not to give an
exhaustive coverage but to further our knowledge by giving a brief account
of the following topics :
(i) Similarity of matrices;
(ii) Inner product spaces;
(in) Orthogonal and unitary matrices; and
(iv) Application to reduction of quadrics.

7.1 SIMILARITY OF MATRICES

The central idea here is to find out the effect of working with different
bases in a finite-dimensional vector space Y. Let two ordered bases
F = {u1, "•• •··, U11} and G ={•1, ••• •.. , •11}
be given in V. We pose the following questions.
Question 1 Given a vector x E V, what is the relationship between its
coordinate vectors [x)F and [x]c. ? (Recall Definition 3.6.10.)
Question 2 Given a linear operator T : Y - Y, it has different matrices
when referred to the different bases F and G. What is the relation bet-
ween these matrices, namely, lT: F, F) and (T: G, G)? •
Question 3 Given a coordinate vector («1, «., ... , «.), we can construct
two different vectors
" n
x = ! «,u, and y - l: ac,,,
l=l l-1
in Y. What is the relation between x and y ?
Question 4 Given an n x n matrix icu) of scalan, we can define two
linear operators R and S by
 7.1 SIMILARlTY OF MAllUCES / 2Sl
II II
R(u1) =I c11u, and S(v1 ) = :£ c.,v,, J = J, 2, ... , n.
i=l l=l
What is the relation between R and S ?
Essentially, these four questions can be unified by asking for the rela-
tion between
(iJ two coordinate vectors arising from the same element x of r·;
(ii) two matrices arising from the same linear operator on V;
(iii) two vectors in V arfr,ing from the same coordinate vector; and
(iv) two linear operators on V arising from the same matrix.
In order to answer these questions we construct the important linear
operator A on V, which transforms the basis Finto the basis G. In fact,
A : V ➔ V is defmed by A(u,) = v., i = I, 2, •· , n. Denoting the matrix
(A : F, F) by (ot.,), we can write the definition of A as
n
v1 -- A(u,' = I ix.,u., ; =- I, 2, . , n. (I)
1-1
n
Note th:.t A is one--one, since, if x -= I «,u., then Ax = 0 gives
i- 1
n n n
O- A(I ot,u,) =-- I ot,A(u,) -=- I «,v1
i=I 1=1 1=1
or ,x, =- 0 for all i = I, 2, .. , n, i.e x = 0. Thus, A is a nonsingular
hnear transformation and so (et.,) 1s a nonsingular matrix. Thi~ matrix is
called the matrix of change of basis from the F-basis to the G-basis.
These four questions can now be respecti~ely answered by the follow-
ing four theorems.
7.1.1 Theorem Let x € V. Let [xjp and [x]G be the coordinate vectors,
written as column vectors, of x, relativt: to the bases F and G, respec-
tivtly. Then
tx],- = («,,)lx}G , (2)
Mhere the matrix («,1) Is defined by Equation (I).
Proof: Let
[x], = («i, ms, ••• , cx,.)r and [x]G = (~1, ~..... , ~..)r •
Then I cx,U. = x = I ~1v1 = I ~1A(u1)
I J J
== I ~, I cx,,u, = I (l: ~,0t.,)u, •
i , , J
Since the u.-s are LI, it follo\\s that
ex, == I «11~1, i == I, 2, ... , n,
J
 212 / ll()U MATIIX fHBORl'

i.e. f~11-
L«n-1
(•u) I;~l
L~,. J
' This pro,es the theorem. I
7.J.2 Theorem Let T he a /i11ear ope1ator on V. Let (T: F, F) he the
matrix B _:,. (~1, ) and (T: G, GJ he the maim: C ~ (y.,). Then, with
(«, 1 ) defined as i'n Equation (I),
B(oc,1) = («u)C • (3)
In other words,
B = (oc,1)C(«.,)-1 • (4)
Proof; We have
T(u,) = I ~.,u., and T(v,) = I r.sv., j = 1, 2, ... , n.
i '
Now, for eachj = 1, 2, .•. , n, \.\e have
T(v1) = T(A(u,)) , (\\here A is defined by (I))
= TlUtkJU1:) • (again by (1))
" =
= Iocr..,T(u.,) I«A:,(I~,11:u,)
k k l
= I(I~ 1:ot1c,)u, .
1 (S)
I k
On the other hand, for each j = 1, 2, •··, n,
T(v,) = Iy,.,i•,. = Iy1: 1A(u1:)
k "k
= Iy.,,Iot,11:u, (by (1))
k I
= I(Iot,.1:Yt1)U, • (6)
I k
Equating (S) and (6) and observing that th u,s are u, we get, for each
=
J I, 2, ... , n,
I,
I~,1:ot1:, = Iot..tYld •
k k
This gives Equation (3) and proves the theorem. I
7 .1.3 Theorem If («1 , «a, ... , oc,.) is a given coordinate vector and » =
Iot,u., y = I«,v., then the two vectors x and y are connected by the
I I
relation
.Ax= Y, (7)
where A Is defined by Equation (1).
Proof: Since A(u,) = v. for all l = I. 2•.•.• n. it r~uows that
 7.1 SIMIUltlTY OP MATRICIIS / 253
Ax = A(I«,u,) = I«,A(u,) = I«,v, = y. I
I I I
7.1.,t neorem Let (c11) be a given n X n matrix. Define linear operators
Rand S by •
R(u,) = Ic,1u, and S(v1) = Ic, 1v., J = 1, 2, ... , n.
I I
Then
SA. =AR. (8)
In other words,
S = ARA-1 , (9)
where A. Is defined by ()).
Proof: We have, for eachj = I, 2, :.. , n,
(SA)(u1) = S(A(u 1)) = S(11.1)
= Ic, 1v, = !c,,A(u,) = A(u.,u,)
I ' I
=
A(R(u1)) =- (AR)(u1) •
Hence SA =-= AR. I
Equations (3), (4), (8), and (9) provide the motivation for the f, llow-
ing definition.
7.1.5 Definition (a) Two n X n matrices Band Care said to be similar
if there exists a nonsingular matrix P such that BP = PC or
B= PCP- 1•
(b) Two linear operators Sand Ron V are said to be similar if then,
exists a nonsingular linear operator A on V such that SA = AR or
S = ARA.-1•
In the light of this definition the answers to Questions 2 and 4 can
now be briefly stated as : The matrices or the linear transformations in
question should be similar •
let U\ now look at a comprehensive numerical example, which illus-
trates Theorems 7. 1.1 to 7.1.4.
Example 7.1 let V = V3 • Consider the ordered bases
F = {(0, I, I), (1, 0, 1), (1, I, 0)}
and G = {(-1, -2, I), (l, 3, -1), (0, -1, 2)}.
(a) Given x = (1, 1, I), verify I esttlt (2).
(b) Given a linear operator T: Ya - V8 defined by T(x1, x,, x 1) =
(x1 + Xa, x 1 + 2x1 + Ka, x1 - x 1), verify result (3).
(c) O,ven the coordinate vector (4, S, -1), verify result (7).
(d) Given the matrix
r
1 -1 07
(c:,> _ I o 1 2 1·.
L-1 -2 -l_J
254 / MOU MATILD[ TBBORY

verify result (8).

Fint, we have to construct the matrix («11) defined in Equation (1),
correspondins to the linear transformation A. : Y1 ➔ Y1 given by
.4.(0, 1,1) = (-1, -2, I), .4(1, 0, 1) = (I, 3 1 -1), .4(1, I, 0)
= co. -1. 2).
So we have to find «,,'s such that
«i 1(0, 1, 1) + «11(1, 0, I)+ «.1(1, I, O) = (-1, -2, 1)
«.1(0, I, 1J + exu(l, 0, I)+ «11(1, 1, O) = (I, 3, -1)
«,,1(0, l 1 1) + «.1(1, 0, 1) + «11(1, I, O) = (0, -1, 2) .
In other words, we have to.solve the following three systems, each of
three equations in three unknowns :
O« + + y = -I
~ 1 0
11 + ~ + y = -2 3 -I
er.+ P+Or= 1 -1 2.

ro
The matrix for these three systems is
1 1 -1 I 07
II
Lt 1 0
O I -2
-1
3 -1
2_J
I·
Applying the row reduction process to this matrix, we get
rt o o o l 17
I o 1 o
LO O 1·
1 _,
-2 I -t.J
, I·
, o l 17
So (cr.u) = 1 -I f I,
I

L-2 t -f_j
(a) Let [x), be (s11 &a, 11a)r. Then
(1, 1, 1) == s 1(0, 1, 1) + «.(I, 0, 1) + s 1(1, 1, 0).
From this we get s 1 == 1/2 == &a .. &a·
Let [x]a == CP1, ~ ~r. Then
(I, 1, 1) == ~1(-1, -2, 1) + ~(I, 3, -1) + '-CO, -1, 2),
which 1ives ~ - -1, 13. - 0, P, == I. Thus!
. [x), = Cl, l, ur
and (x]a == (-1, 0, l)T.
We have now only to check
•
 7.1 SIIIILAlll'lY OP IIATIUCl!S / 255

i7 r-17

which is easily verified.

-f
·J I L 1-J
O I•
(b) We have to compute the matrix B = (T: G, G) = (13,,). Fint,
observe that, from the hypothesis,
T(O, 1, 1) = (1, 3, 1), T(J, 0, I)=-- (2, 2, -]), and T(], 1, 0) = (I, 3, O).
The scalars ,.,'s then satisfy
(I, 3, 1) = f311(0, 1, I)+ 1321 (1, 0, 1) + 1331(1, 1, 0)
(2, 2, -1) = '311(0, 1, I)+ 1312(1, 0, 1) 1381(1, I, 0) +
(I, 3, 0) = f313(0 I, 1) + f3aa0, 0, 1) 1333(1, 1, 0). +
To find the f3 1,'s we have to solve three systems, each of three equations in
three unknowns. Setting up the matrix for the row reduction process, we
get
ro I I 2 17
,I • o • 3 2 31
LI 1 0 1 -1 O_J
which, on row .-eduction, finally leads to
rt o o I -l 17
I O 1 0
LO O I
-l -l -1
f I 2
J.
r t -i 17
Hence, B = (1311) = I-i -l -l I·
2_J L I t
Again, to compute the matrix C =
(T: G, G) = fo,), observe that
T(-1, -2, 1) = (0, -4, -1), T(I, 3, -1)-= (0, 6, 2), and T(0, -1,2)
= (2, 0, -1).
lbe equations to be solved here a.-e
- at+ = f3 0 0 2

-2ct + 313 - r= -4 6 0

at - f3 + 2y = -1 2 -1.
The matrix of this system is
r-1 1 0 0 0 27
I-2 3 -1
L 1 -1 2
-4 6
-1 i -1
oJ'
 256 / MOlll! MATRIX TRl!OllY

which, on row reduction, finally leads to

r1 o o -I 7
I O I O -f 7
LO O l -i
· r-t 1
Hence, C = (Yu) = -t
L-i I
I 7

It is now easy to verify that B(«u) = (11.u)C.

(c) x = I11.,u, = 4(0, I, 1) t 5(1, 0, I) - (I, I, 0)
I •
= (4, J, 9)
y = I«,v, = 4(-1, -2, I)+ S(l, 3,
'
Now Ax= .4(4, 3, 9)
-1) -

= .4(4(0, 1. l) + 5(1, 0,
(0, --1, 2) = (I

I) - 1(1, t, O))
..
8 -3)
.
= 4.4(0, I, I)+ SA(l, 0, I) - A(I, I, 0)
= 4(-1, -1, I)+ 5(1, 3, -1) - (0, -1, 2)
== (I, 8, -3) == y.
(d) We have (SA)(O, I. l) = S(A(O. 1, I))
= S(-1. -2, 1) = Ic,111,
i
= 1(-l, -2, 1) - 1(0, -1, 2}
= (-1, -1, -1)
and f.4R)(0, l, 1) = A(R(0, 1, I)) == .4.(Ic0 u,)
I
= A(l(O, 1, 1) - 1(1, 1, 0))
= .4.(0, 1, I) - .4.(1, I, 0)
= (-1, -2, I) - (0, -1, 2) = (-1, -1, -1).
Slmilarly. we could ,check that S.4. and AR coincide on the other basis
elements Ila and Ila also. Hence, S.4. = .4. R.

Problem Set 7.1

1. R.epeat Example 7.1 fdr the following data:
Y= J/1
F == {(1, -1), (I, l)}
G = {(2. -3), (3, 3)} .
(a) X - (l, 0)
(b) T : Y1 ➔ V1, T(xu x1) = (x1 + x1, x1 - x.)
(c) The given coordinate vector is ( - I, - I)
 7,1 SIMILAlllTY OF IL\11UCIIS / 257
r-1 -27
(d) The given matrix is I I.
L 2 -l_J
2. Repeat Example 7.1 for the following data :
y == Y,
F = {(1, 1, 0, 0), (1, 0, I, I), (0, 0, I, 1), (1, 0, -1, 0)}
G == standard basis.
(a) x == ( I , I , I , 1)
(b) T: V, ➔ Y,, T(x1, Xa, x8 , x,) = (x1 + x1 + x1 + x,, 0, 0, 0)
{c) The given coordmate vector is (I, -2, 3, 0)
(d) The given matrix is
r 1 o o 17
-1 1 0
0 0 0 -1
L 1 -2 3 4_J
110 27 111 07
3. Prove that I I is s mdar to I I via the nonsinrular
L2 7_l L 0 6_J
matrix
r2 17
v'S v'S
1 2
Lv'S 'V"S.J
0-1
r r3 0
I
1 2 -27

4. Prove that I 2 1 2 is similar to I 0 3 0 I via the non-

L-2 2 l_J LO 0 -3_j

2 1 7
0 - \/'6 y3

1 1 1
singular matrix v'l - ~ - \/'3 .
1 1 1
Lv'l ~6 v'3...J

5....... tllatj _:
L-10
-1 -107

10
7 10
-2_J
I· olmilar to\
16

LO
o. -12
0
:1
0 18..J
258 / MORE_MATalX THEORY

1 1 17
v12 v'-6 v3
1 1
via the nonsingul1r matrix
v'2 y6 - v3
2 1
LO v6 - y3_J
6. Prove that o;imilar matrices have the same eigenvalues.

7.2 INNER PRODUCT SPACES

In§ 2.3 we talked about the orthogonality or vectors in Y1 and V8 • In
this article we shall gene"Jlise this concept to al) finite-dirr..ensicnal vector
spaces and to certain more 8t'neral spaces called inner product spr..ces.
7.2.t Definition Let V be a (real or complex) vector sp ce. An inner
product on V assigns to each ordered pair of vectors u, v in V a
scalar (complex ir Vis a complex vector space and re .. t if Vis a 1eal
vector sp·:ce), written as u · v, satisfying the following properties :
(IPI) u • (v + w) = u • l' + u • w for all u, l', w •n V
(IP2) («u) • v = «(u v) for all u, v E V and all
scalars t1
(IP3) u . v =-= v • u
(iP4) u · u ;> 0 and u II= 0 ijf u =-= 0.
7.2.2 Definition A vector space V together with an inner product 1s called
an inner product space.
A finite-dimens10nal real inner product space is called an Euclidean
space and a finite-dimensional complex inner product space is called a
unitary space.
We shall now give some elementary consequences of Definitions 7 2.1
and 7.2.2 in the form of a theorem. Note that we state the results for
complex inner product space•. The corresponding results for real inner
product spaces are obtained by merely omitting the complex conjugations,
7.2.3 Tbeorem Let Y be an inner product space, u, v, and II be any thrte
-vectors in V, and t1 a scalar. Then
(a) (u + Y) • w = u • w + , • w.
(b) u • (t1v). = ii(u • Y).
(c) 0 • u == 0 == u • O.
Proof: (a) (u + -v) • W = w~-(u+-,) = w • u + w • v
;mw=-ii +~=II• W + V • Wo
 7,2 INN& PaODUC'I' IPACllll / 259
(b) u · («I) -= («•) • u = «(• • u)
= i(.--:,;) == «(u . •) .
(c) 0 .:...!f. =- (O,} • u == O(, • u) == O •
The second part of (c) can be proved similarly. I
Example 1.2 In ~ we can define an inner product as follows:
If u = (xu x 1, ... , x.) and • = (y1, y1, •• , y.), then
U • 1' = X1J1 + XtJ1 + ,.. + X"'• • (I)
For a real vector space Y., the analogous inner product is
U • II = X1Y1 +
Xl)'1 + .. . +
XJ• • (2)
Note that this has already been defined (cf Definition S.4.2).
It 1s routine to verify that Equations (l) ,nd (2) are inner prod~• on
Y~ and Y., respectively. Clearly, the fam.iliar dot product on Y1 and Y1
is a special case of the inner product thus defined on Y.. Therefore, Y1
and Y1 , endowed with the usual dot product, are inner product s ~ .
This also explams why we use the notation u · , for the ioiaer product in
general inner product spaces.
7.2 4 Remark If u and" are considered column vectors, thee u • , caa
also be wr1ben as the matrix product .,r-,. On the other hand, jf
they are coni.1dered ruw vectors, then u • v == uvr.
7.2.S Remark The inner product defined in Example 7.2 is called the
natural inner product on Y~ (V.). Problem S, of Problem Set 7.2,
gives another inner product on v•.
ExamJ le 1.3 Consider the vector space '6'[0, I]. Define
1
f, g =-- J0 f(t)g(t)dt,
where/, g € W[O, 1]. It can be verified that this is an inner product on
the space.
Analogously, on ~ (0, 1) we can define an inner product
C

I· g = 10l /(t)g(t)dt.
-

Let Y be an inner product space. Then for each vector u E Y, u • II

is a nonnegative real number irrespective of whether Y ;s a ..,1 or a CQID-
plex vector space. We write
Hull== v~ (3)
and call it the nol'l&i¥J', In Eumples 7.2 ao.d 7.3 we have
H(x1, X1, ••• , xJ II - ( I X1 I 1 +
I X1 I 1 + ... +
I x. 11)111 (4)

and I/ d == ( JO I /(t) I I dt)111

1
(5)
260 / MORI MATRIX Tlll!OB.Y
Equation (4) suggests tb&t the norm of an element u may be considered
the •distance of u from the zero of the space' or the 'length of the vector
u', We shall, without further ado, call II u II the length of II even in the
general case. The properties satisfied by II u U in an inner product apace
are siven in the following theorem.
7.2.6 Theorem Let Y be an inner product space. Then for arbitrary
,ectors u and v in Y and scalars 11,
(Nl) II l&U II = I (& I II u u
(N2) 11 u II >
0 and 1111 11 == 0 iff u = O
(N3) I u • v I < 11 " 11 11 , 11
(N4) II u + v 11 < 11 u II + 11 v II •
(N3) is called Schwarz', inequality and (N4) is called the triangle
inequality. The triangle inequality simply generalises the familiar
pometrical fact that in a triangle any side is lefll than the sum of the
other two sides.
Proof: (NI) II 1111 II 1 = (1111) • (1111) == C«ti)(u • u)
=l«i ■ uu11 1 ,
which gives (NI).
(N2) This is a restatement of (IP.J),
(N3) This is clearly true if u = O. Suppose u ::/, o. Consider
M' == 11 - «II ,

where 11 =. ., ".111" • Then

11
0 < 11 w 11 1 = w • w = (v - 1111) • (, - 1111)
== 11 • v - , • («u) -(cxu) • 11 + («u) • (1111)
11(11 • 11) + («i)(u • u)
= 11 • v - -c..,. • u) -
•>
= II v 11 1 - "i(ll • u) - «(II • + I ex I • Uu 11 1 •
Substituting the value of «, " e get
• 21••111' 1••111 1 •
O<Uvll - Hull* + 11u11' Hun
I jv•r,fl
==ll•II - 01111 1 '
which gives (N3).
(N4) Uu + • U1 = (u + •> • (u + v)
== II • II + II • 'P + 'P • ·r, + 'P • 'P
=u • u+u • ,+.;-:-.+, • •
- UII H1 + 2 Re (u · ,) + U• U1
<Uu11 1 +21u · •I +u,a•
<RuD 1 +211uft ll•B+H•H 1 (by (N3))
==(UuU+H•U)I,
"biob aivea (N4). I
 7.2 INNER PRODUC? SPACIS / 261

Now we are ready to take up the concept of orthogonality of vectors

in a general inner product space. Note that these ideas have already
been developed for Y1 and Y1 in § 2.3, where we used the dot product
without mentioning that it is an inner product on the space concerned.
7.2.7 Delaltloa (a) If u and , are two vectors in an inner product space
Y, they are said to be orthogonal if u • , = 0.
(b) A set of vectors is said to be orthogonal if each pair of distinct
vecton of the set is orthogonal.
7 2 8 Theor• Any orthogonal set of no zero vectors in an inner product
space is u.
Proof: Let A be an orthogonal set and B == {u11 u1, •, •, u11} be a
finite subset of A. Consider
«1111 + «1"8 + •• + «11U11 = 0 ,
Then, for every i = 1, 2, ... , n,
0 =0 • "·
= («1111 + «aU1 + .. + «11u,.) • 11,
= «1("1 • u,) + «1(u1 • u,) + .. • + «11(1111 • 11,)·
= «,(u u,), since u. • u = 0 for I el= J.
1 • 1
Since 11, el: 0, we get «, = 0. Hence, every finite subset of A is u and
therefore A is LI. I
The converse of Theorem 7 .2.8 is not true. In other words, not every
linearly independent set is orthogonal. This can be seen from simple
examples, such a&
{2i, 21 3}} +
in Y1 • But, given a linearly independent set, we can arrive at an ortbo•
gonal set in a standard way, which is very important. This constructinn
is called the Gram-Schmidt orthogonalisation proce11. We buald this in
the proof of Theorem 7 .2.10. But first we shall introduce the concept of
vector projection, analogous to the situation in Y1 and Y, (cf Theorem
2.3.8).
1.2., Detlaltlo■ If u and , belong to an inner product space Y and
, ::/:: 0, then the vector
u. '
T,jjl'
is called the ,ector projection ofu along,.
7.l,10 Tlleona E,ery finite-dlmeuional Inna product 8/JtlCI luu on
orthogonal basil.
Proof: Let {"1, u., .... u.} be a basis of the inner product space
Y. We shall construct an orthogonal set {111 ••• ••• , •11} !)f vectors in Y,
which is a basis for Y. Write , 1 =- 111• To construct'•• i.ubtract from Ila
its vector projection along , 1• So
262 / MORE MATlllX THEORY
U1 • Vt
\I.a = ... - II V1 11 1 Vi •

Thia 1ives v1 • '1'1 = 0 and so {vu 111} is an orthogonal set.

Now to construct va, subtract from u1 its vector projections along '1'1
and 111 • Thus,
Ua ' V1 Ua • 11,
V3 = 113 - llv,li I \11 - iv.ii7 Va.
This gives 111 • 111 = 0 = 113 • 111• {Check.) ftence, {111, "•• 'l'a} is an
orthogonal set. Observe 1hat we have so far used only three of the
vectors of the given basis. Proceeding thus and using all the vectors
'41, flt, ... , u11, we construct the orlhogonal set B = {v11 v1 , ••• , 1111}, where
l ; l U, • v, . 2 3
V1 = U1 - ~ -11--1I l',, I = ' , .. ,, n,
i=- I "1 I
None of the vectors of B is zero, because if v., were zero, then u., would
be a linear combmation of v1 , ,·1, ... , ''k- 1 and therefore of u1, Ua, .•. , U1-1,
By Theorem 7.2.8, Bis a linearly independent set and hence a basis for
the.n-dimensional space V.
7.2,11 Deftnltlon An orthogonal set of nonzero ,ectors {ui, u1 , .. , u11 } is
said to be orthonormal if II u, II =-1, i = 1, 2, ... , n.
Clearly, {i,j, k} is an orthonormal set (in fact, basis) of V3• From
Theorem 7.2. IO, it follows that every finite-dimensional inner product
0

space has an orthonormal basis.

7.2.12 Remark In the Gr, m-Schmidt construction, described in the proof
ofThrorem 7.2.10,we started with n vectors, which formed a basis
for V. Even if we had started with fewer than n elements, the
construction would have worked, provided those elements were LI.
The resulting set would have been merely an orthogonal set and
n~t an orthogonal basis.
In Example 7.4 we shall apply the Gram-Schmidt construction to
produ~e an orthonormal set from a given linearly independent set (which
ia not a basis).
Example 7.4 Orthonormalise the set of linearly independent vectors
{(I, O, I, I), (-1, 0, -1, 1), (0, -J, I, J)} of v•.
Let ,, == (I, 0, I, I). Then
v1 == (-1, 0, -1, I) - (- I, O, -!,_!l ..:J.~~. I, _l) (I, 0, I, I)

= (-f. o. -f, t).

,1 - (0, -I, I, I) - (O, -J, J, l)3 ' (l, O, J, I) (l, 0, I, I)
 7.2 INNEll •aooucr SPACD / 263
_ (0, :-1, J, 1) :._ (-2/3, o, -2/3, 4/3) ( 2/3 0 2/
24/9 - , , - 3, 4/3)
= (-1/2, -1, -1/2,0).
The resultint'orthogonal set is
{(I, 0, l, J), (-2/3, 0, -2/3, 4/3), (-1/2, -1, -J/2, O)}.
The corresponding orthonormal set is
{(l/v'3, 0, l/v'3, 1/v'3), (-1/v'6, 0, -1/./6, 2/v'6),
c-1/vJ, --2/v'3, -t/v'3,o)}.
Example 7.5 Find an orthonormal ba!,is of 9' 3 [-1, l] starting from the
basis {I, x, x•, r}.
u.,e the mner product defined by
f ·g = J~ 1 /V)g(t)dt.
We take vJ = J.
X • 1
Theo v. = X - -
2
l

=-- x - (l J~ 1 tdt) =--= .\.

x2,1 xz,x
Va = x• - -2- l - -2/3 X

= x2 - i J~ 1 t 2dt - ~x ( 1 t8dt = x 2 - i.
v, = "3 _ ~~ • _! 1_~~•_xx_ x• • (x~ - IJ3\.r .:_ i)
2 2/3 2/5
= X 3 - fx,
Thus, the orthogonal basi:,, 1s
{I, x, x 2 - l, X 3 - fx} .
To get the corresponding orthonormal basis, we divide these by the res-
pective norms and get
I
{ v2' v'3 3v'S 2 Sv'7 3 I }
,/2 x, 2,/2 (x - l), 2y'2 (x - x) '

Problem Set 7.2

1. Use the Gram-Schmidt process to ortbonormalise the linearly inde-
pendent sets of vectors given in Problem I, Problem Set 3.S.
2. Repeat Problem 1 for Problem 2, Problem Set 3.S.
3. Repeat Problem 1 for Problem 3, Problem Set 3.S.
(Take the inner product in the space to be J~ 1 f(t)g(t)dt.)
4. fn an inner product space Y show that
(a) If 11 • u = 0 for all u € Y, then 11 = 0
=
. (b) If 11 • 11 w • 11 for all II E Y, then 11 = w.
2'4 / KOU Ill.DIX THIORY
5. Show that an inner product can be defined on Ya by
_,.
(%1t ""II ,... \ (x1 -!!...XY1 - y.) + (x1 + x.)(y, + y.)
111 =
• Vl1 4 4 , •
In thi1 inner product space calculate·
(a) • 1 • •• (b) (1 1 -1) • (1, 1).
6. (This is a generalisation of Problem S.) Let B be a basis of an
n-dimensional vector space V. Define. for arbitrary x, y E V1
· x • y = [xJa · (yJ.,
where the inner product of coordinate vectors on the right-hand side
is the natural inner product of~- Prove that this definition gives
an inner product on V. In this inner product space prove that the
basis B is orthonormal.
7. Prove that all the eigenvalues of a symmetric matrix are real.
8. Let the set {•i, 111 , ••• , ,,.} be LD. What happens when the Gram-
Schmidt process of orthogonalisation is applied to it ?

7.3 ORTHOGONAL AND UNITARY MATRICES

We shall discuss two special types of matrices in this article, and
assume the natural inner product on V,. and ~.
7,3.l Delaltloa (a) A real square matrix H is called an orthogonal
matrix if
HTH = I == HHT •
In other words, HT H-1 • =
(b) A complex square matrix U is called a unitary matrix if
u•u = 1 == uu• .
In other words, u• == u-1•
For enmple,
r 1
r
I
cos&
•
sin 87
I is orthosonal and
v'2 J271
1 is unitary.
L-11n8 coa 8.J i
L,v'2 ,v'2_J
Orthogonal and unitary matrices are important because they correa•
pond to certain linear operaton that have sipificant properties. But.
1ince we ahall not be discussing such operaton in tbe general settina of a
vector apace. we aba11 be content with deriving certain fundamental
properties of ortho10aal and unitary matrices.
7.3.2 'l'lllonla 4' real (co,nplu) 1fUt1H matrix 11 orthogonal (unltory)
the row, of tM matrl% form an orthonormal ••• of wctor, or f6
co.,.,.. of,,,. matra form an ortlrononnal i.t of ,.ctor1.
*f6
 7.3 ORTHOGONAL AND UNffAllY MATRICIIS / 265
Proof: We shall prove the theorem for the complex case. By
removing all the complex conjugations in the proof, we get the proof for
the real case.
Let U1, "•• .... u,. be the column vectors of a complex square matrix
U. Then
r uf 7
-T
U,a

L u! _J
where u, - (uu, u8,, u,.,)T and u,1 = (,, 11, aia., ••• , u,.,).
r iiI 7

So

L urII _J

ufu,.7
li{u,.

Lu!u1 il!Ua uTu

ft II
_J
This is the identity matrix I iff
iifu, = 0 for i =I= j
and u~u, = I for all i = 1, 2, ••. , n.
This means that the set of column vector& {ui, u1, ... , u,.} is orthonormal.
To prove that this requirement is equivalent to saying that the rows
are orthonormal, we have to note only that u•u = I implies uu• = I
(and vice versa), and this latter is true iff the columns of u• are ortho-
normal. I
7.3.3 Theorem (a) The orthogona, n x n matrice, form a groflP for
multiplication.
(b) The unitary n x n mDtrlcesform a group for multiplication.
The proof is left to the reader. The groups mentioned in the theorem
are called the orthogonal group of order n and the unitary group of order
n, respectively.
We shall end this article by giving a definition.
7.3.4 DelaltlOD (a) Two square matrices A and B of the same order
266 / MOUi MATRIX THBOI.Y

are said to be orthogonally similar if A. = HBH- 1 for some ortho•

gonal matrix H,
(b) Two square matrices A and B of the same order are said to be
amitarily similar if A --= U su-1 for some unitary matrix U.

Problem Set 7.3

l. If H is orthogonal, prove that det H = ± 1.
2. If U is unitary, prove that I det U I = 1.
3. If U is unitary, show that U. u7 , and u• (k a positive integer) are
also unitary.
4. If H is orthogondl, show that HT and H 11 (k a positive integer) are
also orthoi;onal.
5. Show that
ro 17 ro -i7 r1 07
I I •I I,I I
LI O_j Li O_j LO -I_J
are unitary, involutory, as well as Hermitian .
. 6. Prove that if U is unitary and U*AU and U'BU are both diagonal
matrices, then AB= BA. Is this reilult true if U is replaced by a
real orthogonal matrix H ?
7. Let T be the linear operator on V3 defined by T(x, y, z) = (x', y', z'),
where x' = x cos ,f, - y sin ,f,, y' = x sin ,f, + y cos ,f,, z' = z, with
respect to a cartesian coordinate system. Prove that Tis given by
an ort!1ogonal matrix.
8. True or false ?
(a) Every real symmetric matrix is orthogonal.
(b) Every Hermitian matrix is unitary.
(c) If for a complex matrix A, AA* = A*A, then A is unitary.
(d) The eigenvalues of a unitary matrix have absolute value 1,
(e) If a matrix is unitary, it is also Hermitian.
(0 The columns of
r 1 i 7
·~
I_,
Lv2
v'2

-~_J
1

form an orthonormal set of vectors.

7.4 APPLICATION TO REDUCTION OF QUADRlCS

In this article we shall use the concept of an orthogonal matrix to
reduce a real symmetric matrix to the diagonal form by a similarity
 7.4 APPLICATION TO REDUCTION OF Ql!AI>lllCS / 267

transformation. An analogous reduction of 81:rmitian matrices to

diagonal form by a similarity transformation using unitary matrices is
possible; such a reduction and related topics are covered in advanced
textbooks. Since the problem of reduction of a real symmetric matrix to
diagonal form is the same as the geometrical problem of reduction of a
quadric to its principal axes, we start this article with a discussion of the
geometrical aspect.
In 2-dimensional analytic ~eometry the equation of a central conic
referred to its c.ntre as origin is of the form
ax2 + 2hxy + by3 ----= k .

[x
,a
This can be written in matrix form as

yJ I
h7 rx7
I I I= k .
Lh b_J LY J
In 3-dimensional analytic geometry a similar equation represents what
is called a central quadric. It is of the form

[x Y z] 1: : ;1 1:1
L_g I c_J Lz J
d ·

ln detail, this equation is

ax 2 + by2 + cz + 2fyz + 2gzx
2 I- 21,xy =- k•
In general, even in n dimensions, we can write a similar equation.
The surface represented by it is called a quadric in n dimen!lions or simply
a quadric. Its equation is of the form
u Au---= k,
where

U=

Lx,._J
and A. is a real symmetric matrix of order n.
In discu\sing a quadric in V,. we shall use geometrical ideas like
tangent and normal. A tangent to a quadric is a line that meets the
quadric at two or more coincident points. If a straight line is perpendi-
cular to the set of all tangents at a point, we say it is a normal at the
point. We assume the natural inner product on Vn and begin with a
fundamental lemma.
7.4.t Lemma /fu1 fa a poi11t on the quadric 11TA.11 = k, then the normal
at "• is in the direction of A11e.
268 / MOllll llAl"RIJ( THEORY
Proof: Figure 7.1 represents
the situation in the 2-dimens.ional
case. Let L = {u0 + tv I t E R}
be an arbitrary tangent, in the
direction of "• at 11o on the surface.
£ clearly passes throuBh the point
u0• Any other point, say u0 + tv, 0
where this line L meets the quadric, ~
will be given by those values of t
that are roots of (u0 + t,)TA
(Uo + tv) = k. This is a scalar 'V
equation, though written in matrix FIGURE 7.1
form. It simplifies to
utAUo + 111T AUg + u{A(tv) + t•vrAv = k •
But u{Au0 = k, since 110 lies on the quadric. So the equation becomes
t(vT .4Uo + u{Av) + t1 11TA11 = 0.
One root of this is t = 0, which corresponds to the point 110 itself. In
order that L may be a tangent to the surface, we require the other root
also to be zero. This gives
vTAUo + u{AY = 0.
But v1 Au1 and u[Av represent the same real scalar. Hence,
vTAu0 = 0 = ufAv.
The first equation means AUg is perpendicular to v, i.e. to 110 + tv, an
arbitrary tangent at u0 • Thus, Au0 is perpendicular to all tangents at llo·
In other words, A.u0 gives the direction of the normal at u0• I
Let u0 be a point on a quadric. Then the vector u0 is said to deter-
mine the direction of a principal axis if it is normal to the quadric at the
point u0• By Lemma 7,4.1, it follows that the principal axes are in the
directions 110 , where A11o = Nl0 for some real scalars >.. Thus, we have
established the following theorem.
7.4.21heorc• In the IJUQdlic uTAu = k the principal axes of the flllldric
are In the direction of the eigenvector, of A.
Example 1.6 Find the directions of the principal axes of the conic
10.x' + 4xy + 7y1 == 100 •
In matrix form this is written as
r10 27 r.x7
[x 11 I I I I == 100 •
L 2 7.J LY.J
r10 27
Jbe eiaenvalues of A = I I are aiven by
L 2 7.J
 7.4 APPUCATION TO llEDUCllON OP QUADIU<:11 / 269

110 2- ). 7 -2 ). 1-- O,
which gives >. = 11, 6.
Corresponding to the eigenvalue ). = 11, the eigenvectors are given by
r-1 27 rx7 ro7
I I I I= I 1.
L 2 -4_J Ly_J LO_J
r27
This gives an eigenvector I I.
Ll_J
Again, corresponding to the eigenvc1lue ). = 6, the eigenvectors are
given by
r4 27 rx7
I 11 I= I
L2 l_J Ly_J
r 17
This gives an eigenvector I
I . Therefore, the axes of the conic are
L-2_J
+
along the dtrections (2i j) and (i - 2j).
In Example 7.6 we have determined two eigenvectors corresponding to
two distinct eigenvalues. Naturally they are orthogonal. Let us normal-
ise these vectors and construct the matrix
r-2 17
vS vs = H (say).
] 2
Lv'S y'S_J
Observe that this 1s an orthogonal matrix (since its columns are ortho-
normal). Use this as a similarity transformation on the given symmetric
matrix
r10 27
A= I I-
L 2 7_J
Calculate HTA.H. We obtain the matrix
ru 07
D:; I I,
LO 6_J
Note that this matrix is a diagonal matrix whose diagonal elements are the
r10 27
eigenvalues of A.. Thus, A =I I is orthogonally similar to the
L 2 7_J
r11 07
diagonal matrix D =I I• The passage from ..4 to D is called
LO 6_j
27,0 / MORI MATIIX TIIIIORY

d/conall8atlon. The 1eometrical implication of this dia1onalisation can

be understood by going back to· the equation
10x1 + 4xy + 7,-1 == 100 .
1bia equation contains tne xy term; so we do not kl\OW the exact size and
hMatioJl of Jhe conic except that it is an eJlipse, since its discriminant
/,i - ab - --66 < 0. To find the length of the axes we transform the
rx7 r~7
coordinates I I to the aew coordinates I t by the formulae of trans-
LY-1 Ly'_J
formation
x = x' cos O y' sin 6 +
y == - x' sin 8 +
y' cos 8 ,
where 8 is the angle of rotation of the axes. We know from analytic
geometry that if 8 is suitably chosen, the xy term will vanish in the new
coordinates, and the lengths of the axes can be found by inspection of the
transformed equation. The theorem on tbe reduction of a real symmetric
matrix, which we shall now prove, will produce this value of 6 without the
be4p of geometry. This theorem of linear algebra says that the required
r-x7 rx'7
transformation matrix from I I to I I is nothing but the orthogonal
LY.J Ly'.J
matri& ot,taiacd from the normalised eigenvectors of .4.. In Example 7.6
it is just
r 2

"= II vs I
Lv5
Usu,g this, we get

rx7
I I=
Ly.J
r
lvS 2

1
17
v'S
2
Irx'7
:
Ly'.J
I.
v3 v"1-1
Write this as u == Hv. Then UTAU = k beoomes vTHTAHv = k. Thia
simplifies to
rn 07 rx'7
[x' y'l I I I I == 100.
LO 6.J L.>"-1
i.,. 1b'1 + 6y'1 = I 00
x'' ,••
er c101v11>1 + c10/v6)1 == 1 •
nia ia Ill ellipse with Jcnaths of semiaxes lO/v 11 and 10/../6, Thia
81181,.a shews that
 7.4 APPLICATION T9 I.EDUCTION 01' QUADRICS / 271

(i) the eigenvectors of A give the directions of the principal axes of the
conic;
(ii) the eigenvalues >., and ).1 of A specify the lengths of the semiaxes
vk/">.1 and v'kf>.2;
(iii) the orthogonal matrix l/ constructed out cf the normaHsed eigen-
vectors of A is the matrix that reduces A to the diagonal form; and
(iv) the matrix A is orthogonally similar to a d agonal matrix, whose
diagonal elements are the eigcnvatues of A.
We shall now prove that this situauon is true in the n-dimensional
case also, at least for the ca!>e where the matrix has distinct eigenvalues.
The result of the theorem i\ true even for general real 1,ymmctric matrices.
But we lea,e the 1heoretical clevelopmcnt of this case to more advanced
textbooks.
7.4.3 Theorem (Diagonalisation of .l real i,ymmetric matrix) Let A be a
real symmetric rnatrh: with distinct dgenm/ues. Let the normaliud
eigenvectors of A be ll'ritten as column ve1·tors of a matrix H. Then
(a) H is orth<?gonal.
(b) HTAH(~-D) is a diagonal matrix, whose diagonal entries are
the eigenvalues of A.
(c) A is tJrthogonally similar to the d1a!;o11a/ matrix D.
Proof: (a) First, we prove that the eigenvectors of .A corresponding
to distinct eigenvalues are orthogond. Let u1 and 112 be e genvectors
corresponding to distinct eigenvalues A1 and )2, respectively Note that
both are real (see Problem 7, Problem Set 7 2). Then
(A1 - A2)(u1 • U2) = ),1(U1 • Ue) - Az(U1 · U2)
=- (A1U1 U2) - (u1 • AzU2)
rlz) - (u1 · Au2)
=-= (Au 1 •
= (Au1)Tua - uf(Aua)
= uf A 1ua - ufAu2
=0 (beca~se A 1 = A).
Since >.1 -=f:: >.a, it follows that u1 • u1 = 0. lhis coupled ·, ith Theorem
7.3.2 shows that His an orthogonal. matrix.
(b) H = £u1 Ua •.. Un) ,
where "' are the eigenvectors of A written as column vectors.

So H'A - r:~ r:;~ J

7
A=
7

,.
LuT ...J ,.
L~TA
212 / MOU MATRIX THBORY

r<Aui)T -7

= I (Au2) (where the >../s are the corres-

ponding eigenvalues)
I : I
L(Au,.)T .J L(>-,.u,.)T _J
r;..1 o 0 7 ruf7
0 Al
0(
LO O >-11 .J Lu~.J
-~ DHT ==- DH l •
J'.ence, H 1 AH-= Das required.
(c) 1 his is only a restatement of (b). B
Example 7.7 Reduce the quadric
r 1 --1 -10- l rx7
[x y z] I
L-10 10
-1
-2.J Lz.J
7 10 I l I=y 36

to its principal axes. Note that when this equation is written in full it
takes the form •
7x + 7y - 2z + 20yz - 2<'zx - 2xy = 36 .
2 2 2

-r _; -: -::1
According to the preceding discussion we need to reduce the matrix

L-10 10 --2.J
to the diagonal form. The eigenvalues of A are 6, -12, and 18. The
corresponding eigenvectors are (1, 1,0), (1, -1,2), ard (1, -1, -1).
Hence, the normalised eigenvectors are
(l/y'2, l/v'2, 0), (1/y'6, -1/y'6, 2/\/6), and (l/y3, -l/v'3, -1/y'J).
These three vectors give the following orthogonal matrix
I r1 I 7
v'l y-6 v'3
I l I
H= v1 - '1ti -- v3

Lo
 7.4 APPLICATION TO l\BDUCTION OF QUADI.IC:S / 273

Hence, HT AH
r l 1 0 7 r 1 l 17
~i. J2 v'6 yJ
r 7 -1 -107 1 V' 2
==

I
1
\16
l
- vo
l
2
v'6
l
IL-10 -l 7
10 -2
to J'l
I
..;1
I
l
-,J6
1
---
v3
vJ --
2 1
Lv'3 y3_J LO y6 - v'l.J
0

-12
LO
is the required diagonal matrix.
S.o the equation of the quadric referred to its axes is
6X1 - l2yl + 18~ = 36 I
x3 y• z•
i.e. T - T -t T = 1 .
This is a hyperboloi,J of one sheet.
We shall conclude this article with another example where the matrix
does not pos&e<1ci distinct eigenvalues. Though we have not developed the-
necessary theory for this, the following example will show lhat a reduction
to diagonal form is possible even m ciuch a ca!le,
Example 7 8 Reduce the matrix
r 1 2 -27

A= I 2 2 I
L- 2 2 l_J
to the diagonal form, and hence reduce the quadric
x• + y1 + r 1- 4yz - - 4zx 4xy '"" 21 + (2)
to its principal axes.
The eigenvalues of the matrix A are 3, 3, and - 3. Note that in this
case 3 is a repeated eigenvalue. So Theorem 7.4.3, which depends on
the fact that the matrix has distinct eiaenvalues, cannot be applied. How-
ever, an orthogonal matrix can be obtained by looking at the eigenvectors
corres?()nding to 3 and -3.

r-2 2 -27 rxl

The eigenvectors corresponding to the eigenvalue 3 are given by
ro7
I
L-~
2 - 2 2
i -2_J Lz_J
ii y == I l·.O
t...0.J
 274 MOiii MATRIX THEORY

This gives only one equation :

x-y+z=OI
The solution set of this is the subspace {(y - z, y, z) I y, z E R} of Y,.
We can choose two linearly independent vectors in this 2-dimensional
space by giving suitable values toy and z. Taking y = 1 and z = 1, we
aet (0, 1, I); and taking y =- l and z = 2, we get (- I, t, 2). Since
these two are LI, we have only to orthogonalis, them. Using the Gram-
Schmidt process, we get two orthogonal vectors
(0, I, I) and (-1, -1/2, 1/2).
Normalising them, we get
(0, 1/v'2, l/v2) and (-v2/v'3, -1/v'6, l/v'6).
The third eigenvector 1s the one that corresponds to the eigenvalue -3
and is (I, -t, 1).
Normalising this vector, we get (1 / y'3, -1 / v'3, 1/ y'3). Without
checking, we can say that this will be orthogonal to the two eigenvectors
corresponding to 3, because eigenvectors corresponding to distinct eigen•
values are orthogonal. Thus, we have the orthogonal matrix
ro v'2
-v'3
17
v'3 i

H..,,,,
I
1
v'2 -
1
y'6 -,/3 1
.
I _1 1 ] I

Lv'2 y6 v'3.J
Hence, HTA.H
0
1
y2 v2
17 ro v2
-v3
17
v'3
r 1 2 -27

- --
v'2
v'3
l
- v'6 v'6
l
IL-2
2 1 2 I 1
v'2
--1
"6
--v'31
2 J.J
L
1
vJ -- I
v'3 0-1
1
Lv'2
1 l
v6
1
:J)_j
rJ 0 07

- I0
LO 0 -3.J
0
3
I·
Thus, the quadric (2) reduces to
3x1 + 3y' - 3z' == 27 ,
~ ,.. z'
i.o. -+----1.
9 9 9
 7.4 APPLICATION TO I.EDUCTION Of QUADRICS ( 275
It may be noted here that Equation (2) can be written as
1 1 2 -27 rx7

(x y z] I 2
L- 2 2
2 \ \ y \ - 21 .
1_J Lz _I

Problem Set 7.4

t. Reduce the following matrices to diagonal form :
rs 1 17 r 6 4 -27 r4 3 37
(a) 11 5 - I I I(b) 4 12 -4 I(c) I 3 0 -I i.
LI -1 S_j L--2 -4 l3_j L3 -1 3.J
2. Reduce the following conics to their principal axes :
(a) 7x1 + S2xy - 32y2 = 180
(b) 17x1 + 312xy + 108y1 = 900
(c) 145x1 + 120xy + 180y1 = 900.
Appendix

Ordinary Linear
Differential Equations

In§ 4.9 we saw that the general solution of the n-th order normal
linear differential equation
Jny ti" ly
Oo(X) iJxn + 01(X) dX"- 1 + ••• t a,.(x)y = g(x) (J)
is of the form y = J'o + yp, where Ye is the solution of the associated
homogeneous equation
dny d•-ly *
a..(x) dA" + a1(x) dx" 1 -t •.. + a,.(x)y = 0, (2)
and }'p is one particular solution <'f (1). y,. is called the complementary
function and is the kernel of the linear differential operator
d• d-1
L = Oo(X) d£• + 01(X) dx"-1 + ... -I On(X) • (3)
Thus. the method of solving Equation (I) involves two steps. namely.
finding Ye and YP• Further, it may be noted that the solution space of
Equation (2) is n-dimensional (cf Theorem 4.9.3).
The solution of the first order normal linear differential equation has
thus been completely discussed in§ 4.9. In this appendix we shall develop
methods of finding Yo and YP for differential equations of arbitrary order.

Al HOMOGENEOUS LINEAR DIFFERENTIAL

EQUATIONS WITH CONSTANT COEFFICIENTS
In this article we shall discuss the method of finding Yo for linear
differential equations with constant coefficients. i.e. the method of solving
• 'lbe zero on the riaht-band side of Equation (2) should be 0. But throuahout the
appllldia we ■hall u1e O as ii the practice in writi111 • ditrerential equation. 111,
contest wlll make it clear whether it la Oor O.
 , Al HOMOOBNEOUS DIFFDBNTIAL EQUAtlONS / 271
.
homogeneous linear differential equations with constant coefficients. To
start with, let us consider the second order differential equation
ao1• + ai1' + OiY == 0 , (4)
where a0, ai, and 01 are constants, and a 0 =I- o:
lhis equation may also be
written as
(a 0 .DI + a1D + a2)y = 0 (S)
or
Ly= 0, (6)
where L = ae.DI + a1D + a 1
is • linear operator from <if 111 (/) to '67(1).
It may be noted that Lis a hnear operator also from i..,1 11(1) to ~'(I).
C C
In § 4.9 we saw that the equation
dy
dx t P(x)y c::a Q(x)
has the complementary function y 0 = CeS P(x)tlx • When P(x) is a constant
p, this reduce, to Ye = Ce,,., Thus, the first order homogeneous linear
differential equation
!9'_
dx
+ ay = 0
has the sol1,1tion Ce"'. Using this as an analogy, we try e- (where m is a
constant) as a solution of the second order homogeneous linear differential
equation

This gives
(a,,m2 + a 1m + a2)e"''" =-- 0 .
But e""'' is never zero. So
a0 m' + a1m + a2 = 0 . (7)
Equation (7) ts called the indicial equation or the auxiliary equation. Tlh:
values of m for which Equation (7) holds will lead to solutions e"'" of the
differential equation (5). Let the roots of Equation (7) be m1 and m_.
Then e""-x and e~ are two solutions. Their Wronskian is
emlx em.x
W[em,x, em.x ] = = (mJ - m1)e(m1 + 111a1.r •
m1emtx m~em.x
The different roots of the auxiliary equation (7) give rise to tbe follow-
ing three cases.
Case 1 m1 =I= m. In this case the Wronskian is never zero (cf Theorem
4.9.2). So em,.r and e"'-.r are two linearly independent solution~ of
Equation (S). The solution space has dimension 2. Therefore, ,m,.r and
e"'-" form a basis for the solution space. Hence, the gener~ solution is
Yo - C1e'"l" + c,e"'-" ' •
 ffl / UIINDII
01# 1 • 1 - .,,,,. la du, ~ it foJlowl from Equdma (7) daat
2-i - -tlJ11e and one ,aJutlon of Equation t,) II ,i - .-.-,
111J - -11J2t,,. Let:,, be tJJe otber solution auch that ~1 and Ya are u. TJlea
I••
W[e•11t , y.J - Ce- •• tbt .., ci1111" ,
Dy Abel"s formula (cf Remark 6.9.4). This, on simplification, pvea
e"'11ty; - mi,at"'"' .. Ce2111,." •
Heaoe, y; -
mi11 . . Ce-.;& •
Solvina th11 fint order linear difl'enntial equation, we obtain a puci-
cular solution Ya - ex,•"' . This aives us Y■ -= '""" as a aecond
10lutlon (taking C == 1) of Equation (5). Obviously, y1 - ~ and
71 - "'""" are LI. (Why ?) Hence, the general solution ii
Ye == C1 ,•.;t
+ C,,c ,l"Jl' •
Thu, we have obtained Yo•
&,,,,,,1, A..l Obtain Yo, the complementary function for the differential
equation
(D' - 2D - l)y - sin x •
Tbe auxiliary equation of the associated homogeneous equation ii
m•-2m-1-o.
Its roots are m1 • 1 +./2, "'• • 1 - ./2. So ,u +vZ).1 and ,<1-v~
an two linearly independent solutiona of the auociated homopaeou1
equation. Honce,
Yo ... c11U+v'2 z + cil-v2)r .
._,,,,,, A..2 Consider the differential equation
(4D' + 12D + 9)y ,... ~ , . .
The auiliary equation of the associated homopneou equation ii
4m' + 12m + 9 - 0.
Its IOOtl are m, =- -3/2 - ,,,_, Al the roots are equal, the two lilleady
indlpmdent solutiom are e-■-'8 and. x.-..,., 1hu, the complementary
funotion it
Y• - Ci,-,. 11 + C1,crlafl •
.,_,,,,. A..J Consider the diff'erential equation
(D' - 4D + 13)y ... ~ + cos 2x •
The auiliarY equation of the auociated homopneous cciuation ii
,,.a - 4ffl + 13 - 0 •
Its fOl)tl are ffla - 2 + 31, m. - 2 - 31. Sm ms rt/: -.. two linearly
lndepmlclnt solutiom are ,,....,,. and r._..,.. nu., tbe ccaplemenlary
fUDOliollla
 A I BOIIIDGINIOUI DlnllDmAL tlQuA'IIGNI / 2'19
Tbll ii a OQllpla aolutlon. We •hall now di1e11■ tbe method of lltdDa
1111 IDlutlolll from complex tolutiou.
Al.t LteM I/ u(.x) + 1¥(.x) II • compl,,c '°"'tlon of a homo,...,
,,,,_, tllff,rntlal ,.,.,Ion Ly - 0 wld& nol co,Jficien11, then llotlt
1(.¥) 11114 ,(.x) 11,a,11t1ly 1111#1/y the quatlon Ly ... O.
Proof : Since u(z) + l,(z) is a solution or Ly - O, we have
L(u(z)-+ l,(x)) - 0.
£ it a liaeu operator. Therefore,
L(l(x)) +
IL(,(x)) - O •
Equadna the nal and imapnary parts, we get
L(r,{x)) = 0 and L(,(x)) = 0 •
HIDCI tho lemma. I
Ca.r, J Auxlllrlry lfllllllon with complex roots Jr « + ,, is a root of the
auxiliary equation (with real coefficients), then 11 - , , is another root.
lince m, ::p "'-• two linearly independent solutions are .C•+IJ)1t ud
,<•-•> 11• These are complex solutions.
\
As
,<•+IIS)1t = •" cos Jx + l,u un ll.x
ii a solution of.the d.ift"erential equation (5), it follows from Lemma Al.I
tbat ,■" ms ~% and ,- liD ~ are two real solutiom of Equation (5).
Further. they are linearly indepeadent, because
W[e" cos~. e" sin jlx] - ,,"

1• - cos ll,¥ + C.eu sin ,x.

ii never z•o- Heace, the complementary function can also be written u
c,,.,,
If C1 and C1 are take:i to be real, then we get all real solutions of
Equation (5). For instance, in Example A.3 the real complementary
fuaetion it
Ye - ~cl cos 3x + c. sin 3.x) •
We now state (without prooO the extension oftbe aforesaid method of
6Ddin1 1• for dift'crcntial equations or arbitrary order. Conaider the n-th
order liaear dift"erential equation
(,,..0- + ,,._,,._. + ... + o.-1D + oJy - 0 , (8)
when -.. a,, .•. , •• are real constants. In thla cue also we try y - . -
u a sohltlon or Equation (8) ancl obtain the auxiliary equation
11-,,,. + a,.wl'-1 + ... + 0.-,.111 + "• ... 0 • . (9)
Tbil bu n roota (which may be complex). Some or the roots may be
...,_114_
Pot each real root • with multiplicity r (i.e. repeated ·, times) or the
 280 / .APPINDIX
auxiU,ry equation, the wrrespondins part of the complementary function is
e-(A, + A1X + ... + Ar_1r1) ,
where .A., '41, .•. , A..-1 are arbitrary constants.
For each pair of complex roots at ± ;~ with multiplicity r, the corres-
pondiDs part of the complementary function is
eu(A1 + .41x + ... + Ar-1:xr-1J cos~¥
+ eti•(B0 + B,_x + ... + B,-1.xr-1) sin ~x.
Let us illustrate these rules throush a few examples.
Example A.4 Find the complementary function of the dift'etential
equation
y'" + y' = 2.r + 4 sin x •
Tbe auxiliary equation is m1 + m = 0. Its roots are m = 0, i, -i. So
the complementary function is
Ye = C1 + C1 cos x + C1 sin x .
Example .4.J Consider the differential equation
(D' - 10D1 + 430' - IOOD1 + 131D1 - 90D + 2S)y = x8.
The auxiliary equation is
m• - 10m1 + 43m• - 100m8 + 13lm1 - 90m + 25 = 0
or (m - It1(m1 - 4m + 5)1 = 0.
lta roots are m1 - m1 = l; ma = me= 2 + i; m1 = m1 = 2 - I.
Corresponding to the repeated root 'I', the complementary function is
e"(C1 + C1x) ,
and corresponding to the repeated complex roots, the complementary
function is
e1•{(Ca +
c.x) cos X + (C. + C,x) sin x}.
Hence, the complementary function is
I
Ye = ,..-(C1 + C1x) + e'-{(C1 + c.x) cos x + (C1 + C_x) sin x} ,

Problem Set A 1
l. Solve the following homogeneous linear dift'erential equations :
(a) y'' + 4y' + 4y = 0 (b) y" + 4y == 0
(c) y"' + y' == 0 (d) 3y'' + y' -2y == 0
(e) y'" - y == 0 (f) y" - 2y' - 3y = 0
(I) y" + 2y' - y 0 = (b) y'" - y" - y' + y - 0
(i) (D' + 4D1 -- SD!)y = 0 (j) (D1 - D' - 6D)y = 0
(t) (D' + 4D + 11)1y O = (I) (D' - l)'y == 0
(m) (D' - 3D + 2)1y - 0 (n) {D' + 3D + l)y - 0
(o) (2D' + SD' - 3D)y =- 0 (p) (D' - 4D' + 4D)y - O.
 A2 VARIATION OF PAI.AMl1ERS / 281

Al METHOD OF VARIATION OF PARAMETERS

In this and the next two articles we shall give various methods of
finding y, for a nonhomogeneous ordinary linear differential equation.
One method univenally applicable to all linear equations (even to some
nonlinear equations) is the method of ,arlation of parameters.
We shall illustrate the method by taking a second order equation
(a0 D1 + 0i.D + a 1)y == g(xJ, a 0 ':/:- 0. (I)
Let the complementary function Ye be
J'o == Ci.Y1(x) + C.,y1(x) ,
where C1 and c. are arbitrary constants, ard Y1(x) and y1(x) are two
linearly independent solutions of the associated homegeneous equation of
Equation (1). In this method we assume the particular solution YP to be
of the form
(2)
where u1(x) and u1(x), yet to be chosen, are twice differential functions of
x. Our assumption will be justified if we can show that u1 (x) and u.(x)
can be chosen such that (2) is a solution of Equation (I). Since we have
to determine two functions u1(x) and u1(xJ, we will have to specify two
conditions. Ob-.,iously, one of these is that y.p satisfies Equation (I).
The other is to be chosen so as to facilitate the calculations. Dift'erenti-
atin1 Equation (2) with respect to x, we get
yp(X) == u1(x}J•1(x) + u.(x)y;(x) + 11;(x)y1(x) + ~(x)y1(x) .
In order to simplify the calculations, we shall impose the condition
u1(x)y1(x) + 11~(.x)y.(x) == 0 , (3)
This implies
yi,(x) == u1(x)y;(x) + 111(.x)y~(x) •
So y;<x) == u1(x)y;(x) + u.(x)y"(x) + u;(x)y;(x) + ~(x)y~(x) •
Substituting yp(x), yP(x), and y;(x) in Equation (I), we get
a,(ui.Y; + u.,; + 11;,1 + "';, + 01(IIJY~ + ...,,;) + O.(U1Y1 + 11.,y,)
=- ,C.x). (4)
SiJK:C y 1 and y 1 arc solutions of the associated homoaeneous equation of
{I), we get
a.,; ai.Y;
+ + a.Yi = 0 and a.,; aJ.Y; a.,y
+ + 1- 0.
So Equation ( 4) becomes
a,MY; + ~.) - ,Cx)
or
u~; + .,;y~ = ,Cx)/a0 • (5)
Equationa (3) and (5) give two conditions 1overnin1 t1 e
choice of 111 and
,,., If we can choose u1 and u. so as to satisfy these two conditiona, then
2'2 / A.IIBNDIX
we an throu1b. Wriaiq t.bese to1et.ber, we hav•
u;(x})l1(x) + u;(x)y1(x) - 0
u;(x)y;(x) + u;(x)y;(x) .. 6(x)/a, •
Solving for u1(x) and u~(x) by Cramer's rule, we bave

Ua(X) ==
I O y,,(x)
,<x)fa, y~(x) =_ y,(x),C.x) (6)
IY1\X) Y,(x) a0 W1Y1lx), Y1(x))
Jy;(x) y;(x)
and

(7)

Carrying out the integrations, we get

,« g~t)y1(l)t/t
"i(x) =- J a 0 Wty1(1), y1(t))

lla(x) = ix g(tlY1rdt ) •
a,W'11lt , y,(t)
Thus, a particular solution of Equation (I) is
_ _ ( ) I" g(t)y,(t)dt ( ) I" g(t)}'1(t)dt
.)'p{X) - 11 X a.W[y,(t), Y,V)l + >'• X a.W[11V). Y1ll))
=- I" Y■lXlY1(t) - Y1(x)y.(t) g(t)dt • (8)
a,W(y,(tJ, Y,V»-
Hcnce, the complete solution of Equation ( l) is
Yi = Ci1,(x) + ca1.(.x) + 1" Y,(.x)y,(t) - y,(x)y.(t) g(t)dl •
a,WIY1(t),y,(t))
Al.I Remark The same· method is applicable to cquationa of arbitrar,
order. The only modification is that in the case of the n-th order
equation we shall need n arbitrary functions 11i(x)1 u,(x), ... , u.(~) in
Jp, since there are n arbitrary constants C1, C., ... , c. in the comple-
mentary function. We shall impose on it (n - I) conditions, one at
each sua:essive differentiation, in addition to the flnt condition that
YP """ "1.)'1 + ""• + ... + "-"• satisfies the nonhomogeneous dift'er-
ential equation.
Instead of carryioy out these details in the aeneraJ case, we ahall
illustrate the method by wortiDa out some suitable eumples.
Eza,npl, .4.6 (D1 - 2D - l)y - sin x.
Tlae complementary function ia (cf Example A.I)
Y• - c,,U+\1'2l¥ + c_.U-\l'Z)ll.
 A2 VAllATION OF itAliMBTliS / 21.1
Sowoauumo
1, - 11i(x)e(l+v'l).r + lla(x)e(l-vi)x.
We ~ now _obtain u1(x) and u.~x) by going through the procesa already
described. Since we have done this for the general equation of the second
order, we shall merely plus in the functions in Equation (8). We have
WCY1(x), Y1(x)J = W[e(H v'l)x, e'1-v'l)x) = -2y2e2x .
Hence
yp(x) -= r ----e(I-.J2)x /1+v'2)1
-
(l-v'2)t (I+v'2)x
-----,,,...--=-e=-_ _e____ sin t dt
-2y2ell

Sx [il+v'2)x e-(v'2+1)t e(l-v'2)x (v'2-l)t] . ~

- e s1ntul

== - { - (sin x - co& x) .

A2.2 Remark The reader is advised not to use result (8) for findina ,,.
He should carry out all the steps needed to amvc at Equation (8) as
illustrated again in Example A.7.

Example A 7 ~; + ~f == 2x1 + 4 SID x.

The complementary function is {cf Example A 4)
Ye = C1 + C1 cos x + C1 sin x .
So we assume
yp(x) = u1(x) + 11s(x) cos x u.(x) sin x . +
Then
)';(x) = u1(x) + U:(x) COS X + u;(x) sin X - Ms(X) sin X + fls(X) COS X •
When we impose the cond1t1on
u;cx) + ~{x) cos x + u,(¥) sin x = 0 , (9)
we set
y;{x) = -u.(x) sin X + u,(x) COS X •
Differentiatins again, we get
y;(x) == -u.(x) cos x -111(x) sin x - u.(x) sin x + u1(x) cos x .
We nnw impose another condition
-u;(x) sin x + u~(x) cos ~ == 0 • (IO)
So yp(x) - -u.(x) cos x - u1 (x) sin x
and y;'(x) == -~(x) cos x - u (x) sin x + u1(x) sin x - u1(x) cos x .
Tbo third condition on u1(x), u1(x), u.(x) is that )'p should satisfy the
equation
y"' + )'1 - 2x1 + 4 sin X •
Substituting y,, we obtain
-r,;(xl eos x - u;(x) sin x = 2x1 + 4 sin x . (11)
284 / APPENDIX
Writing Equations (9), (10), and (I I) together, we get
u;(x) +
u1 (x) cos x -t u~(x) sin x = 0
- U1(x) sin x t u~(x) cos x = 0
- u;(x) cos x - u~(x) sin x = 2x1 + 4 sin x •
Solying by Cramer's rule, we obtain
u;(x) = 2.l" 2 -I 4 •JD x, u1(x) =- -(.2x1 -t 4 sin x) cos x
and u1 (x) --(2 ~i + 4 sin x) <,in x .
This, on integration, gives
2x,
U1(X) = J - 4 COS X (12)

u2(x) -- -2 &in2 x - 2xl sm x -- 4x cos x + 4 sin x ('13)

and
u 3(x) = 2x2 cos x - 4x s n x - 4 cos x - 2x + sin 2x. (14)
2x8
Thus yp(x) - 3 - 4 cos x ~ uh.) cos x + u1(..c) sin x
2x3
-= 3 - 4 cos x - 4x -- lx sin x.
Hence, the complete solution 1s
y = C1 + C., cos x + C3 sin x -+ 2; 1 - 4x - 4 cos x -- 2x sin x

= C1 + D1 cos x + C1 sin x + + x1
Note that in this exJmple we needed three conditions, because the
- 4x - 2x sin x .

differential equation involved is of order 3.

Problem Set A2
1. Using the method of variation of parameters, find the general solution
for the following differential equations:
(a) (D1 + 4)y = x sin 2x (b) (D1 - 4D + 4)y = xe-•
(c) (D' + 4)y = sec 2x, (d) (308 + D - 2)y == 3,-
0 < < n/4
X
(e) (4D' + l)y = sec9 x/2, (f) (D1 + l)y == sin x
O<x<n
(s) (D1 + 4D + 4)y = xe-•• (h) (6D' + D - 2)y = xr"
(i) (J>I + D)y == x• (j) (D' - D)y == x cos x
(k) y"' - 3y" - y' + 3y == eh (I) 2y"' + y" - y' = e• 11•
2. Let y1(x), y1(x), ... , y ..(x) be n linearly independent solut1\lns of the
associated homogeneous equation of
(a.»" + a 1.l)a-1 + ... + a,.)y B(x) • =
 A3 METHOD OF UNDBTEllMiNED COEFFICIENTS / 285
Use the method of variation of parameters to prove that
yp(x) = J.x K(x, t)_g(!l._ dt, where
Oo
J'1(t)

y;(o
.
Ya(t)
'
y,.(t)

Yz(t) y~(t)
K(x, t) = 1
W(Yi(t), Yz(t), ,. , y,.(t)] :
Y1 <n-2>(t) Ya'n-2l(t) y,. In -ll(t)

Y1(X) Yz(X) y,.(x)

A3 METHOD OF UNDETERMINED COEFFICIENTS

The method of undetermined coefficie-nts is applicable whenever the
right-hand member g(x) of the nonhomogeneous n-th order linear d1ft'er-
ential equation with constant coefficients
(a 0D" +
a1D" -1 + ... +
a,.)y =- g(x) (1)
is a finite linear combination of
(i) powers of x;
(ii) sin ixx, cos ocx;
(iii) e«:11; and
(iv) finite products of any of the function,; in (i), (ii), and (iii).
In order to outline this method, we need the following definition.
A3.1 Definition Given a function/(x), a !'>Ct of thm,e linearly independent
functions whose finite linear combinations give the function/ and Sill
its derh,atives is called a derivative familv of /(x). It is denoted by
D[fJ.
Example A .8 Here we list the derivative families of four functions :
{a) D[x'"] = {x'", xm-1, .. , x 2, x, 1}
(b) D[e~] = {elu}
(c) D[sin 0tX] = {sin 0tx, cos «x}
(d) D[cos ~x] = {cos ~x, sin ~Y}.
We shall be interested in only those functions / for which D{/) is a
finite set; for example, the functions listed in Example A.8 and their
finite products, namely, x'", elJt:lll, sin OtX, cos 0tx, x"' sin 0tx, x'" cos 0tX,
x-e«•, ti'-- sin ~x, e«" cos ~x. x"'f!Z- sin ~x, and x'"ect:1 cos ~x.
For each of these functions, a derivative family can be obtained as
illustrated in Example A. 9.
example ,4.9 D(eU cos ~x] = {/(x)g(x) I /(x) E D[e«•J, g(x) E:: D [cos ~x)}
= {e«" cos ~x. ,«ta sin ~x} .
286 / APPINDIX
D[ra sin u) == {/(x>,(x) I/(x) E .D[.X-J, g(x) E D[sin u)}
=- {.X- sin u, X-- 1 sin acx, .•., x sin acx, ain u,
,X- cos «x, .x--1 COi O&X, •••• X COi u, COi acx} •
D[.X-,U sin (b] == {/(x>,(x)h(x) I /(x) e: D[.x""],
g(x) e: D[e"-], h(x) E D[ain jb]} •
.U.2 R•ark It can be easily verified that for two functions p(x) and
q(x)
D[p(x) + f(x)] == D[p(x)J U D[q(x)J •
Now we are ready to enunciate the method of undetermined coeffi-
cients for finding the particular integral YP of
(o.,,- + 111D--1 + ... + a.)1 == g(x) •
litep 1 Write
,Cx) - l1(x) + l1(X) + ... + g.(x) ,
where each 1,(x) is a function of the type mentioned ilDlllediately after
Eumple A.8.
Stq 2 Find the derivative family DC,,) for each I == I, 2, ... , k.
Step J Find the complementary function ')'c of tho differential equation
in question. Let
Ya == Ct11(x) + Ct11(x) + ... +
CJ.(~ .
For each I == I, .... k, check whether any member of D(g,J is already
one of the y,'a. i.e. a solution of the associated homogeneous equation.
Ir 10, then multiply each member of D[g,J, for that I, by the least power of
x such that the new set thus obtained does not contain any y 1• Call this
new set the modified D[g,J.
If no member of D[r,l is a y 1, t'hen leave that D[gJ unchanged.
Step 4 Let SC,] be the union of all those D[gJ's that have not been
modified and all the modified D[g,)'s.
Step S Assume
YP == l: A.,h,(.x) ,
wbere the A.,'s are arbitrary constants, and h,(x) varies through all the
members of SC,]. In other words, ')'p is a linear combination of all mem-
bers or S(gJ.
Step 6 Substitute YP in tbe diff'erential equation and determine the coeffi.
cients A/a by equating coefficients of identical terms on both sides.
Wo illustrate this method through the following eumplcs.
Example A.10 (IJI - 2D - l)y - sin x.
'lbe 00111plementary function is (cf Example A.I)
, .... c,e(l+v2),s + cae<•-v'l),& •
 A3 IIBTHOD OP UNl>lmll'MINED COl!fflCllllnl / 287
To aet YP, we proceed u follows :
Step 1 It is not necessary here.
St,p 2 D[sin x] = {sin x, cos x}.
Stq J No modification of D[sin x) ii needed, because neither sin x nor
cos x occ:un in the complementary function.
Step 4 S[sin x] = {sin x, cos x}.
Step $ Assume y.,(x) = .4 sin x + B cos x.
St,p 6 y~(x) """ .4 cos x - B sin x,
y;(x)-= -.4 sin X - B cos x.
Substitution in tM equation gives
(2B - 2.4) sin x - (2A + 2B) cos x - sin x .
Therefore, 2.4 + 2B == 0 and 2B - 2A == I. This gives ..4 - -B
- - 1/4. Hence, 1,(x) = (cos x - sin x)/4. Thus, the complete solution
is
y = c1,C1+v'l).r + C,eCl-v'l).r - (sin x - cos x)/4.
B,unnple A..11 (4D' + 12D + 9)y == xe-,.1•. The complementary func-
tion Is (cf Example A.2) 0

Ye = C1r'-l1 + C.xr ..11 •

Step 1 It is not necessary here.
Stq 2 D(xe-'-11] _ {xr,.I•, e-b/'}.
St,11 J Both the runctions in the set D[xr,.11] appear in the comple-
mentary f11nction. So we multiply them by x 1 . (Note that multiplication
by x will leave one fuoction in the set to coincide with one in the comple-
mentary function.) So the modified derivative family is {x'e-,./1, x'r'-/1},
Step 4 S[xr'-/1] = {x'e-..,., x'r.. '1}.
Step$ Assume Y, == .4x'r.. ,.+ Bx•,.....,,,
, 3..4 3B
Step6 y == r• 1• ( - - x • + 3Ax' - - x ' + 2Bx)
P 2 2
• 9..4 9A
y
P
.. ,-■.ii( -
4
x' - - - x•
2
+ -9B
4
9.4
x' - 3Bx - - 2 x•
+ 6.4x - 3Bx + 2B) •
Substitution in the equation gives
,-../1(24Ax +BB)= xr-1•.
Hence. B =- O and .4 == 1/24. Thus, yp(x) = (x'/24).-,.11 and the
complete solution is
y = (C1 + C1x + .x'/24)r•t1 •
._,,.,. ..4.12 (,DI - 4D + 13)y - .x' + cos 2.x. The complementary
ftlDc:lioa (real) ii (cf Exampl8 A.3)
288 / APPBNDJX

Yo - e1S(C1 cos 3.x + C, sin 3¥).

Step 1 g(x) = r + cos 2x = l1(X) + f1(X),
where f1(X) = r, r,(x) = cos 2x.
Step 2 D[gJ = {r, x, I}, D(g1] == {cos 2x, sin 2x}.
Step 3 No modification of the derivative families is needed.
Step 4 S[g] == {.x1, x, 1, sin 2x, cos 2.x}.
Step S Assume yp(.x) == Ax1 t Bx + C + D sin 2x + E cos 2x.
Step 6 Substitution of Jp in the equation gives (2A -- 4B + I 3C)
+ (-8A + 13B).x + 13.fx' + (13£ - 8D - 4£) cos 2x + (13D + 8£
- 4D) sin 2x = r
+ cos 2x.
This gives A = 1/13, B = 8/169, C = 6/2197, D =- -8/145, E = 9/145.
Hence,
_ £_ + ~ + -~ _ + 9 cos 2.x - 8 sin 2x
Jp - 13 169 2197 145 •
~
and the complete solution is
y = r"(C1 cos 3.x + c. sin 3x) -I YP.
Example A.13 (D' + D)y = r -t 2et- sin x.
The complementary function is
Ye = Co + C1 cos x + C1 sin x (cf. Example A.4).
Step I g(x) = g 1(x) + g 1(x), where g 1(.x) ""' x". g 1(x) = 2e sin x.
Step 2 D[g1] = {.r, x, l}, D{g1] == {e1' sin x, e" cos xl.
Step 3 Since 'l' is included in the complementary function, D(.r] is
moddied as {x1 , x2, x}.
Step 4 S(g] = {r, .r, x, r sin x, r cos x}.
Step S Assume YP -= Ax1 +- Br +
Cx + De:• sin x -f Er cos x.
Step 6 Substituting yp in the equation, we get
6A + 3Ax1 + 2Bx + C - (D + 3E,e- sin X + (3D - E)r cos X
= x" + 2e" SID :C •
Thus, A = 1/3, B-= 0, C = -2, D = -1/5, E = -3/5.
Hence,
x• 1 3
Yp= --2.x- ~sinx- -e"cos.x
3 S :) '
and the general solution is
1 = C0 + c, cos x + c, sin x + ~• - 2x - e' •;n x - 3e" 5°• x .
 A4 OPERATIONAL MITHODS / 289
ProblematA3
1. Using the method of undetermined coefficients, find the seneral
solution for the following differential equations :
(a) y" - 2y' - 3y == ~ (b) y" + 4y = sin 2x
(cJ y" + 2y' + 2y = xe" t x (d) (D1 + 2D - l)y-= 2 cosh x
1

(e) (.D3 - D1 - D +
l)y = sin x + 1 +
xe •
(0 (D + 9D)y = x
1

(g) (D' + D' - D - l)y = x sinh x + x' + 3

(h) (.D3 +D)y = 2 cos x
(i) (D' + 4.D3 -- 5D2)v = x + x sin x
(j) y"' - y" - 6y' = xe-18 + sin x +
x -t e-a..
(k) (D + 4D + 11 )y = xe-1• + e- sin 3x.
1 18

A4 OPERATIONAL METHODS
The two method<1 discussed in § A2 and A3, namely, the method of
variation of parameters and the method of undetermined coefficients,
suffice to take care of all problems of finding a particular solution. How•
ever, in this article we add to the reader's armour another tool. which
often comes 1n very useful: Consider the equation
f(D)y = g(x) , (1)
where /(D) = 00 D11 f- a 1/)fl- 1 + .. . + a._1D + a,,, and Do, ai, .•. , a,. are
constants. To obtain one particular solution of this equation. we are
tempted to write formally
1
y = 7(D'f g(x) . (2)
This is only a formal step, which would be useful only if it enables us to
calculate y. First, we have to give a meaning to Equation (2). Second,
we have to relate Equation (2) to Equation (1). We achieve both purposes
simultaneously by defining Equation (2) as follows.
A4.t Definition If /(D) == aJ)" + a1D•-1 + ... + a,., where Do, a1 ... , a,.
are constants, then
1
/{D)g(x)
is defined as a function y for whichf(D)y = g(x).
This definition gives meaning to Equation (2) and also relates it to
Equation (lJ. •
We shill now show that /tb) g(x) can indeed be cal~lated easily in
certain favourable situations.
290 / APPENDIX

The simplest case occ1Jrs ,, hen /(D) = D. Now ~ g(x) is that func•
tion y for which Dy - g(x). This means
dy
dx = g(x).
So y -= J g(x)dx. Smee we are mterested :n only one particular solution,
we shall ignore the arbitrary constant. Thus, we have
1
D g(x) =I g(x)dx . (3)

Extending this we see that

1
D;- .rr(x) = I (I g(x)dx)dx , (4)
and so on
We now take the case /(D) =D - at. Then D ~ a. g(x) is evaluated

as follows : Let
I
y =D -=-«g(x) .

Then (D - 0t)y = g(x), which is a linear differential equation. So

y = e«z I e-~g(x)dx (cf § 4 9).
Thus, we have
t
D _ 0t g(x) = e«z I e-«"'g(x)dx . (5)

E:x:imple A.14 Calculate D -1 - e1z sin ~x. Using Equation (5), we get
- at

- 1 - .."" si.; ~x = e«z I e-«z,«z sin ~x dx

D - a.
e'" cos ~x
= - --~---
To handle more complicated expressions for /(D), we fir,t note the
following result.
44 2 Theorem The set of all operator polynomials /(D) = a0 D" + a1D"-1
+ .. + a,.1, where a 0 , a1 , ... , a,. are in general complex constants, is
a vector space (real or complex, according to the scalars used) Here
I is the identity operator.
The proof is left to the reader. Note that each sucb/(D) is itseJfa
linear operator on~(•).
If /(D) and g(D) are operator polynomials, then /(D) o ,CD) or
simply Jt.D)g(D) is defined, just as we defined S o T in Chapter 1, i.e,
(Jt.D)g(D))y = Jt.D'J<.g(DJy). This multipti~ation is
 A4 OPERATIONAL llffllODI / 291
(a) tWOClathe, i.e
}tD)((,CD)h(D)} = (/tD)g(D))h(D) ,
(b} "8trilnldff o••r addition, i.e.
/(D)(,CD) + h(D)) == /(D)g(D) + ft.D)h(D)
and (/(D) + g(D))h(D) == /(D)h(D) + r(D)h(D) I

(c} commutatt.e, i.e.

/{D)g(D) == g(D)/{D) •
A.4.3 R,mark It may be noted that the multiplication of operator poly-
nomials is not commutative if the coefficients are functions or"·
For example,
(xD + l)D = xD• + D but D{xD + /) == xD1 + 2D •
Because of the aforesaid properties, we may use the results of ordinary
algebra pertaining to factorisation of polynomials. Note that D - I can
be written as D - 1 without any damage to the working. In general,
(D) = ap + a1D"- 1 + a..I + ...
can be written as
/(D) = a.,D" + a1DR- 1 -¼ ••• + a• •
as we have been doing all along. In particular, we can factorise
D' - SD + 6 as (D - 2)(D - 3).
With this backaround, let us now handle/(~ g(x) • We illustrate
the method through the following example.
1 .
Example A.15 Eva1uate u--=-Tsm x.
We write
,DI-
1. 1 .>
1 SIRJC={D-1) D +1 SIDJC
c•
=D~ 1 (r• Je"' sin x dx) (by (5))

== 1 1 (.
2 D _ 1 •un x - cos x
)

= ! ti' I r" (sin x - cos x)d,c (by (5))

= - 21 .
s1nx.
To solve such problems we can also use partial fractions u follows :
1 . 1 .
"JI _ l SID X == (l> _ l)(D + l) SID JC
1 ' j • •
- -(-- - -)SID~
2 D-1 D+l
292 / APPBNDIX
1 I . 1 I .
= -r D- l SJDX- 2 ff+ 1 SIDX
= i; e9 I r' sin x dx - f e _. I el' sin x dx •
This, on simplification, gives
1 . 1 .
g_ 1 smx=- 2 s1nx.
Actually, there should be two arbitrary constants, but we omit them since
we are interested in only a particular integ1al.
In such operator methC1ds the follo~ ing theorem is useful.
A4,4 Theorem If f(D) is the operator polynomial a,,D" + a1D"- 1 + ... + a.
with constant cotfficients, then
(a) f(D)e&- ==- f(a.)e«-.
(b) /(D)(eOg(x)) = ecr/(D + a.)g(x).
(c) /(D') sin a.x =-= f( -a.1) sin a.x.
(d) /(D') cos a.x = /(-a.1) cos a.x.
The proof of this theorem is left to the reader. Using Theorem A4.4
and Definition A4.1, the next theorem immediately follows.
A4.5 Theorem Let f(D) = a0D" + a1 D"-1 + ... + a,. be an operator
polynomial with constant cotfficients. Then
I eP
(a) /(D) eP = /(a.) , /(«) :I 0.

(b) /(~) ~"g(x) =-= e&- f(D ~ «) g(x).

') I . I • /( 1 )-'-0
\C ft.D') SID «x = /( -«') SID a.x, -Gt r •
1 1
(d) f'(D') cos a.x = /(-«8} cos a.x, /(-«1) '=I= O.
1 1 t
(e) /(D) (P(x) + Q(x)) = J'(JJ) P(x) + /(D) Q(x).
t I 1
(f) /(D)g(D) P(x) = /(D)~D/(x)).
With this theorem, let us find particular integrals yp for the differential
equations in the following example, some of which we have already
worked out by earlier methods.
Etample A.16 Consider the differential equations
(a) (D' - 2D - l)y == sin x
(b) (4.D' + 12D + 9)y = xr.. ,.
(c) (D' - 4D + 13)y - x' + cos 2x
 A4 OPlllATIONAL MBTHODS / 293
(d) (D' + D)y = 2x' + 4sinx.
(a) YP = D'-=----21-»--=1 SlD. X
IP+ 2D - 1 .
== (DZ - -1)1 - 4D1 SID X

-= (D1 + 2D -1) ( (Da = 1 ~~ .::-4jji sin x)

= -l (D? + 2D - I) sin x (by Theorem A4.S)

= {- (cos x - ~in x) .
It may be noted that, as a working rule, we can use also the following
method:
I .
YP = 1)1 _ 2D _ 1 sm x

= _1_ 1 _
20 1 sin ~ (by writing -1 for 1)1)
1 1 .
= -T»+l smx

= _.1
I
D
DI - 1
~ sin x = _.1(D - 1) ( -1 - sin x)
I D2 -- 1
= ¼(D - 1) sin x (by Theorem A4.5}
=
-¼(sin x - cos x).
Note that in this method we replaced Dz by -1. In general, to
evaluate ~(~) sin ax or /(~) cos ax, we replace D2 by -a2 ia1 /(D) pro
vided (D1 +a 1) is not a factor of /(D).
1 --1:1 ,s 1 k /ii
(b) Y, = 4D1 + J2D + 9 xe = (2D + 3)11 xe-
1
= e--la/1 -40- X (by Theorem A.4.5)
x'e-k/1
=74•
1
(c) YP == oz - 4D + 13 (r + cos 2x)
l 1
== D' - 4D + 13 xi+ D' - 4D + 13 cos 2x.
Taking the second part, we have
1 l
D1 - 4D + 13 cos 2x = 9-.= 4D cos 2x
2'4 / APPIINDJX
9 +4D
== 81 - 16.0- cos 2.x

= (9 + 4D)(8 I _ I 16.ij! cos 2x)

_ (9 + 4D) "- _ 9 cos 2.x - 8 sin 2x
- 145 cos~ - 145 •
To tackle the first part, we have no operational method so far. But
the following •formal' method, for which we cannot give any justification
within the •cope of this appendix, is successful :

D1
1
4D+ 3 .c'= ~• 4D x2
- l 13(1 + -=--)
13
1 D1 - 4D (D' - 4D)1
== 1T (l- 13 + 169 + ...)r
(by a formal long division)
= _1_ (x11 _ !-=- 8x + .E_)
JJ 13 169
1 Sx 6
= 13 Cr + 7T + 16!Y •
Hence,
Y,=E. + 8x + _!_+9cos2x-8sin2x
13 169 2197 145

A4.6 Remark The working of n•· -1D + jj x', though not justifiable
at this Mage, can be very powerful. The reader is Wied to use it
with the full knowledge that be is using only a formal method, the
validity of which is beyond the scope of this appendix.
In any case the reader will have already noted that wherever the
operational method can be used it is really powerful.
(d) YP = jji ~ D (4 sin x + 2r)
-= D ~ D ( 4 sin x) +
1 .& ~D (2r)

-YPI +Y.,.·

Here '"' == 4 JJtD'i + J)

.
sin x = 4 (.& i+ l) c•D .
sm x
>
1
== 4D,. + 1(-cos x)
J
== -4D1 + l COi X •
 M otDATIONA.L IIB'IIIODI / 295

We cannot apply Theorem A4.5. In this case we cont4nue u follows :

I
Yp1 == -4 R.e ( D• + 1 e'8)
1
== -4 Re (eca (D + I)• + l 1)
1
- -4 Re (el- (D + 2i)D- I)
I
= -4 Re (f'"' D + 2;-X·) •
D
ef•
== -4 Re (2-. (I - -:-
2 +-4D'. 1 - ... )x)
' ' ' .
== - 4 Re (f',.(- •; I -! ))
1 x sin x
= -4( 4° COS X + l ) = - 2x SIQ X - COS X •

Now Yp. == D' ~ D(2x1) = D' -}-1 ( 1( 2x1 ))

_ 23 DJ +
-, 11 ~
1
= 2f (l - DI + D' -- .. )x1
=- 3
i
(r - 6x) = 3 -
2r 4x .
Thus,
.YP = -cos x - 2x sin x --1- 2; - 4x.
It may be noted that the term (-cos x) can be omitted in view of the
fact that cos x is a part of Ye•

Problem Set A4
1. Factorise the operator in each of the following cases and hence find
a particular integral :
(a) (D9 - l)y == e9
(b) (J)I - JD + 2).)' == x + el-
(c) (J)I + 4D9 - SD)y == x + sin x
(d) (SD' + lW - 2D - 3)y = x + xe •.
2. Usina operational methods, determine a particular integral for the
rollowiq differential equations :
(a) (Ji - 3D' + 1D + S)y == ,_..
(b) (D' + 9D)y - cos 3x
(c) (.DI+ 3D +
l)y == r' sin 2x
296 / APPENDIX

(d) (IJI +
4D)y = s:n 2x
(e) (IJI + 3D - 4)y = e"'(3x2 + Sx - 1)
(f) (D - 2)2y = e-u cos x
(g) (2.D8 + SD3 - 3D)y = e-az
(h) (D1 - 4D + 4)2y = x3e2z.
3. Prove Theorem A4.2.
4. Prove Theorem A4.4 .
•
S. Prove Theorem A4.5.
Answers to Problems

Problem Set 1.1

4. (a) {x E R I -1 < x < I} (b) {x E R I -1 < x < O}
(c) {x E R I - I < ."IC < 2} (d) {x E R I ;)x 2 or x < -2}
(e) {xERlx:>lorx<:-1} (f} f-•
5. (a) The set of all complex numbers represented by the points inside
a circle of radius 4 centred at the origin
(b) The set of all complex numbers represented by the points on a
cirde of radius 4/3 centred at ( - 5/3, O)
(c) {z E C I Im z < O}
(d) The &et of all complex numbers represented by the points on a
circle ofr'adius 3 centred at (-2, 0).
6. (a) ,f,, {O}, {I}, {2}, {O, 1}, {O, 2}, {I, 2}, {O, I, 2}
(b) rf,, {«}. {~}, {rJ, {&}, {«, P}, {rx, r}, {tx, 3}, W, r}, {~, 8}, {r, BJ,
{ex, /3, y}, {ex, 13, 3}, {ex, y, 3}, {~, y, 3}, {ex, 13, y, 3}.

Problem Set 1.2

I. (a) {O, I, 2, 3, 4, S, 6, 7, 8, 9, 10} (b) {1, 2, 3, 4}
(c){O, l, 2, 3, 4} (d) C (e) {O} (f) ,f,
(g) {x e: R I x < 10, x c#: 0, 1, 2, 3, 4}
(h) {(O, 1), (0, 2), .•. , (0, 10), (J, 1), (1, 2), ... , (1, JO),
(2, I), (2, 2), ... , (2, 10), (3, 1), (3, 2), ... , (3, I 0),
(4, )), (4, 2), .... (4, 10)}.
2. (a} {O, 1, 2, ... , 20} (b) {O, 1, 2, 3, 4, 7, 14, 21, 28, ... }
(c) {x e: NI 3 < JC < 20} (d} {7, 14} (e) {4} (f) ,/, (g) {O}
(h) A (i) {1,2,3} (j} {JCENIJC<:20,x*7,14}
(k) {JC E N I x > 3, x not divisible by 7}
(I) {(O, 1), (0, 2), ... , (0, 20), (I, 1), (I, 2), ... , (I, 20),
(2, 1), (2, 2), .•• , (2, 20), (3, 1). (3, 2), .... (3, 20),
(4, 1), (4, 2), ... , (4, 20)}.
4. (a) {(ex, ~). (a, 3), (11, 8), ~. ,), (~. 3), (~. 8), (y, ")• (y, 3), (y, 8)}
00~~~~~~~~~~~~~~~~~~
(c) {(f3, ex), (I', y), (/3, c), (3, a), (3, y), (3, c ), (8, ex), (8, y), (8, c)},
5. (a) {x I x > 7 or x < O} (b) {x I x * I} (c) R
298 / ANSWIRS ro PROBLEMS
(d) (x I -1 < x < J} (e) ~ (f) (x I .r > J}.
1. .4 == {I}, 8 = {1}, C = {2}, D - (3}.

Problem Set 1.3

l. (a) Reflexive (b) None
(c) Reflexive, symmetric, transitive, and hence an equivalence
relation (d) Transitive (e) Reflexive and transitive
(0 Symmetric (s) Transitive (h) Symmetric
(i) Reflexive, symmetric, transitive, and hence an equivalence
relation
0) Reflexive, symmetric, transitive, and hence an equivalence
relation
(k) Symmetric and transitive.

Problem Set 1.4

1. (b) and (c) are functions.
2. (a) (i) Not possible (ii) /(a)= /(b) = /(c) -=a/(d) = I
(iii) Not possible (iv) /(a) =- 0,/(b) = 1,/(.c) = 2,/(d) = 3
(b) (i) Not possible (ii) /(1) = 0,/t~) = /(3) = 1
(iii) /(1) = 0,/(2) = 2,/(3) = 3 (iv) Not possible
(c) (i) /(0) = /(I) = 1,/(2) = 2,/(3) 3 =
(ii) i(o) = /(I) = 2. /(2) = /(3) == J
(iii) Not possible (iv) Not possible
(d) (i) /(x) = tan x (ii) /(x) = sin x
(iii) /(x) = e" (iv) /(x) = 3.x -+ 4.
3. (a) (-oo, -1) U (). a)) (b) (-2, 2) (c) {x e: RI x el- -1}
(d) R (e) R (0 {O} U (), co).
4. (a) C (b) {z e: C I z el- O} (c) C (d) C (e) C (f) C.
5. 4(a) [O, co) 4(b) (0, 2) 4(c) {x E RI x :¢. 1} 4(d) (6, 00)
4(e) (-1/2, l /2] 4(f) (0, oo)
,Ca) {z e: C I Im z = O, Re z >
O} S(b) {z e: c I I z I = I}
5(c) {z E C I Im z = O} S(d) {z E C I I z I = l}
S(e) {z E C I Im z = 0, Re z > O} 5(0 C.
6. (c) is one-one. None is onto.
8. (a), (c), and (0 represent functions.

Problem Set 1.5

3. (a) In R define a • b =a
(b) la R X R define (x1• x1) • <Y1o Y1) - C-1 i 11•
(c) la R define a • b - " - b + ab.
 ANSWEU TO PR.OILIIIS / 299
Problem Set 1.6
6. Additive identity, additive inverses, and multiplicative inverses are
not in N.

Problem Set 1. 7
2. go/: x ~ sin• x, -1t < x < 1t, f o g is not defined since R(g)
- D(/). If D(g) = [- vn,
\/1t], then/ o g is defined and
Jog: x ..._.. sin xS, -v'n < x < y'n.
3. g of: x t-+ v'(l + x 1), -1 < x < I. /o g is not defined since
R(g) ~ D(f). If D(g} = (0, 1), then / o g is defined and
+
IO g : X ~ l x, 0 < X < I.
6. (/ + g)(x) = sin x + x 1 ; ( / - g)(x) = sin x - r;
{/g)(x) = x• sin x.
+
1. (/ g)(x) = 1 + x• + y'x; ( f - g)(x) = 1 x1 - y'x; +
(/g)(x) = {l + .r)\/'x.
10. No.

Problem Set 2.1

2. (a) A plane parallel to the xz-plane
(b) A plane parallel to the xy-plane
(c) A line passing through {l, 0, 0) and pcirallel to the line y == 2z
in the yz-plane
(d) A parabola in the plane y == -5; the axis of the parabola is a
line parallel to the x-axis and passing through (0, - S, 0)
(e) The projection of the rectangular hyperbolas yz = 1 anJ
yz = - 1 (lying in the yz-plane) on the plane through (2, 0, 0)
parallel to the yz-plane
(f) The half space on that side of the yz-plane in which x < 0, and
the yz-plane
(g) x = y is the plane perpendicular to the xy-plane and containing
the line y = x in the xy-plane; the required set of points is the
union of the plane x = y and the half space on that side of this
plane that contains the positive x-axis
(h) The projection of the line y = 2z {lymg in the yz-planel on the
plane through (3, 0, 0) parallel to the yz-plane along with the
half plane determined by this projection and containing the
point {3, 1, O!:_ '
3. {a) v'J (b) v'14 {c) vfo {d) 3.
4. (I, 2, 3), (-1, 2, 3), (-1, 11/2, 3), (1, 11/2, 3), (1, 2, 9/2),
<-•. 2, 9/2>, c-1, 1112. 912>, (1, 1112, 9/2>: vJ11v12.
300 / ANSWERS TO PROBLl!MS

Problem Set 2.2

I. (a) \113; tan 8 = 3/2, 0 < 8 < n/2
(b) \110; tan 8 = -1/3, 3n/2 <. 8 < 2n
(c) 3\15; tan 6 = 2, 0 < 6 < n/2
(d) \IS; tan 6 =- -2, n/2 < 6 < n
(e) \12; 0 = n/4 (f) \113; tan 8 = -3/2, 3n/2 < 6 < 21t
(g) y'l7; tan O =-= - 4, 1t/2 < 8 < n
(h) 2v'S; tan 0 - 1/2, r. < 0 < 31t/2.
2. (a) y'l4 (b) 5 (c) y'l4 (d) y'21.
3. (a) (6, -9) (b) (4,-4,.12) (c) (-33, -9) (d) (2, 17, 12)
(e) l3i + 18j - l lk.
4. (a) 2i t- 3j (b) - i t· 4j (c) -Ji - Sj (d) Si 3j -- 2k +
(e) 2i 1- k (f) - i -1- 2j.
5. (a) -u + 2, (b) 3u - 20, I- 12w (c) -u - llv + 7w
(d) 311 + 2v - 2w (c) 4u + 12Y - Q,v (f) --2u - 25, + 15w.
6. l(a) (2/y'l3, 3/v'U) l(b) (3/ \110, -1/y'ioT l(c) (I/ \15, 2/v'sf
l(d) (-lf-\15,2/y'S) l(e) (i-1-j)/v'2 l(f) (2i--3j)/y'l3
l(g) (-i+4j)/v'17 l(h) (--2i- j)/v'S
2(a) (2/y'14~ -1/yl4, 3/y'f4) 2(b) (3/5, 0, 4/5)
2(c) (Ji+ 2j - k)/v'l4 2(d) (-i -- 2j + 4k)/y'21.
7. (a) (-1/\/'2, -I/y2) (b) (yJ/2, 1/2) (c) (l/y2, - l/y2),
8. (a) (1, --1, 2) (b) (2, 3, l) (c) (4, -2, l) (d) (3, 2, 6)
(e) (2, -4) (f} (6, 2).
9. (a) A(l/2, 1/2), B(3/2, 7/2) (b) A(l/2, -1), B(-1/2, l)
(c) A(J/2, -1/2, 1), B(9/2, -3/2, 3)
(d) A(3, 3, -5/2), B(I, 3, -3/2).

Problem Set 2.3

1. (a) 2 (b) -32 (c) - 4 (d) 2 (e) 5 (f) -4
(g) -4 (h) 0.
2. (a) 1/y'S (b) -32/v'lO73 (c) -4/y'65 (d) -l/y'44i
(e) 5/y'385 (0 -4/3vi4 (g) -4/\/42 (h) 0.
3. (a) 2/yiO (b) -32/v'37 (c) -4/v'l3 (d) 1/v'Ll
(e) 5/\111 (f) - 4/ \16 (g) -4/y.14 (h) 0.
4. (a) (l, I) (b) (64/29, -160/29) (c) (-4i - 3j)/S
(d) (-3i + Sj)/17 (e) (1/7, 3/7, 5/7)
(0 (-8/21, 4/21, -16/21) (g) (4i - 4J - 4k)/3 (h) 0.
5. (a) 3 (b) 0 (c) any real number (d) 2, 1.
 ANSWl!RS TO Pll08LIIMS / 301
Problem Set 2.4

3. cos«= -2/'\/17, cos~= 3/y17,cos y = 2/vl"f:

4. r = • + t(, - u), where u and , are the position vectors or the
given points.
7 8 _ II' + mm' + nn'
. cos - y(l'-+ ma + n') vo·~ + .,, 2 t 11'2) •
8. cos-1 (2/3), cos 1 (-2/3), cos-1 (1/3).
9. 2x + 3y /- 6z = 35.
10. 3i - 2j + 6k.
11. (a) n/4 (b) n/2.

Problem Set 3.1

1. (a) (I, 5, -1, 12) (b) (16, - 100, 2, - 6) (c) (0, 0, 0, O)
(d) (-3, 19, -12, 13) (e) (0, 6a -- 2b, 3a + 3b, 3a - Sb).
2. Yes. 4. (a), (d), and (f) are vector spaces.
6. (a), (d), and (f) are vector spaces.
7. (a), (c), (d), and (f) are vector spaces.

Problem Set 3.2

4. (c), (g), and (i) are subspaces of V8 •
~. (b), (d), and (e) an, subspaces of fl'.
6. (a), (b), (d), (e), and (f) are subspaces or re (a, b).
Problem Set 3.3
1. (a), (c), and (d) are in [S]. 2. (c) is in [S].
6. (a) xy-plane (b) xy-plane (c) V1 (d) V,.

Problem Set 3.4

5. (a) {(xi, x 3) E V1 I x1 = O}; subspace of V1
(b) {/E ¥1(-2,2) 1/(1) =/(--1) = O}; subspace of'G'(-2,2)
(c} {/ E 'G' (-2, 2) I !i!\f(x) = 0 and ;~2 /(x) = 1}; not a
subspace
(d) {p E f} I p(x) = p(-x)}; subspace of fl.
6. (a) {(2, 2), (4, 1), (1, 1), (3, O)}; subset
(b) {(4, 3), (8, 6), (3/2, I), (11/2, 4), (1 + y2, 1t - 2), (S +
y2,
n + I)}; subset
(c) {(S/2 - t, 11/3 - 2t) I O < t < l}; subset
302 / AHIWIU 10 ftO.IUlfl
(It is the Jin-. segment joining the points C3/2, ,/3) and (S/2,
JJ/3).)
(d) {(2 + 31, 3 + 41) I l < / < 2)}; subset
(ft is the line segment joining the points (S, 7) and (8, 11).)
(e) {(3 - /, 7 + 2t) I O < t < I}; subset
(It is the line segment joining the points (3, 7) and (2, 9).)
(0 {(x, 2) I x ;> I/2}; subset
(g) {(x, y) I 2x + 3y = 17}; parallel
(h) {(x, y) I (x - 1)1 +(y - 5) 1 = I}; subset
(It is the unit circle centred at (I. S).)
(i) {{311 4- 2tt, 411 + St,) I O -:: 11 < 1, 1 < '• < 2}; subset
(It 1s th interior of the parallelogram PQRS, where the points
P, Q, R, and S arc (2, 5), (5, 9), (7, 14), and (4, JO), respec-
tively, including the edges PS and QR.)
(j} {(ti, ta) IO < 11 <. I, 2 < t, < 4}; subset
(It is the rectangle with vertices (0, 2), (0, 4), (I, 4), (I, 2), and
its interior.)
(k) V1 ; subspace (1) V1 ; subspace.
7. (a) {(I + t, 2 + 2t, I) I t a scalar}; parallel
(It is the hne through the point (I, 2, 1) and parallel to the
vector (1, 2, 0).)
(b) {(x, y, z) I x t-- y + z = 3}; parallel
(It is the plane through the point (3, I, -1) and parallel to the
plane x + y + z = 0.)
(c) {(I + cc, -3 + 2cc, 4 + 3cc + ~) I er., ~ scalan}; parallel
(It is the plane through the point (I, -3, 4) and parallel to the
plane 2x - y = 0.)
(d) {(cc + 3~. 2« + ~. 32) I ai:, ~ scalars}; subspace
(e) V1 ; subspace (0 V.1 ; subspace (g) B ; subspace.
n
12. (b) Base space: {(x., x., ... , x.) I =
,-1
E ai:,x, 0}

Leader : ~ ei, ex, -=I- O.

at,
13. (a) Base space : set of an constant functions Leader : x•
(b) Base space : set of all constant functions Leader: ,1t1f2
(c) Base space : {«x +
~ I ex, ~ scalars} Leader : -sin x
(d) Base space : {cx.x + ~ I ai:, ~ scc1lars} Leader: %4/12 t r/3
+ 3x•12
(e) Base space : {/ e: ~ (0, 2) I /(1) = O} Leader : x + I.
14. (a) Un W == {O} (b) U + W == {/1 + f.1/i,J. e: 'Z'(-a, a),
/ 1 is odd and/. is even}; U + W = '61(-a, a) U (9 w. =
 ANSWDI to ROIL&II f '303

Pro~lem Set 3.5

1. (a), (b), and (d) are u.
2. (a), (c), and (e) are LD.
3. (a) and (c) are u.
4. (a) and (e) are LI
11. (a) S (b) S (c) {(1, 1, 2), (-3, 1, 0), (1, -1, t)} (d) S
(e) {(1/2, J/3, 1), (2, 3/4, -1/3)}.
12. {a) {(1, 0, 0, O), (1, 1, 0, 0), (1, 1, 1, 1)} (b) S
(c) {(1, I, 1, 0), (3, 2, 2, 1), (l, 1, 3, -2), (1, I, 2, l)}
(d) S (e) {(I, 2, 3, 0), (-1, ~. 3, 3)}.
13. (a) S (b) {I, x + x', x - x'}
(c) S (d) {r - 4, x + 2, x - 2}.
14. (a) S (b) {sin' x, cos 2x}
(c) {sin x, cos x} (d) {In x} (e) S.
15. l(c) (I, 2, -3) l(e) (0, 0, 0) 2(a) (0, 0, I, 1)
2(c) (I, 2, 6, -5) 2(e) (1, -1, 1, -1) 3(b) 3x J(d) x1 /3
4(b) l 4(c) kin (x + 1) 4(d) In x•.
Problem Set 3.6
I. (a) Not a basis ; {(I, 2, 3), (3, 1, O)}
(b) Not a basis; {(I, l, 1), (I, 2, 3)}
{c) Not a basis; {(O, 0, I), (I, 0, 1), (I, -1, I)}
(d) Basis (e) Not a basis; S.
2. (a) Basis lb) Basis (c) Not a bJsis (d) Basis
(e) Basis (f) Basis (g) Not a basis (h) Basis.
~ 00 2 00 2 00 3 00 3 00 ~
4. (a) 4 (b) 4 (c) 2 (d) 4 (e) S (f) 4 (g) 3 (h) 3.
5. {(3, -I, 2), (1, 0, 0), (I, I, 0)}, {(3, -1, 2), (0, I, 1), (2, l, 2)}.
6. (a) 1 ; {(3, 2, 1)} (b) 3; {(l, 2, 3), (0, 1, 2), (1, -2, 3)}.
8. (a) {I, x} (b) {(.r - x0) 1, (x - x0) 1 , (x - x0)', (x - Xo)1}
(c) See Problem 11
(d) {{I, 0, -1, 3, O), (0, I, -1, 0, 0), (0, 0, 0, 7, J)}
(e) {(x - x.), (x - x0) 1, (x - x.}1 , (x - x.)'}.
12. (a) -1/2, -1/2, 3/2 (b) -3/2, -S/2, 7/2
(c) lx1 - x,, Ix, - x1 + x1, -ix, + x1
l l 1
(d) - v'2 - e, - ;ti - ff + e, vl + ff
(e) -21/4, 1312, 47/12 (f) 2, 0, -1.
I).· (a) --1, 6, -2 (b) -2, -1, 3 (c) -1, l, -1,
304 / ANSWBU TO ftOILIIIS
14. {(I, 0, 0, O), (-S, J 4, -6), (0, 0, 1, 0), (0, 0, 0, 1)}.
• • n
15. % atuX" E at1,x, .... , E ot.,x,.
1... 1 l=l i I
16. 2.
17. A== [(I, 0, 0, 0), (0, 1, 0, O)], B = [I, I, 5, 2), (1, 2, 3, 0),
(1, 1, I, 1)).

Problem Set 4.1

2. (11'. (d), (f), (g), (h). (j), (k), (I), (n), (o), (p), and (q) are linear.
3. (a) Yes i T(x, y) = (-x/3 + Sy/3, 4x!3 - 2y/3)
(b) Yes; T(x, y) = (2y, y) (c) No
(d) Yes; T(x, y, ?) = (y -t z, x/2 + 2y - z/2, 2x - 2y + 2z)
(e) Yes; T(at0 + ot1X + «.~2 + «iX') = (--«0 + 2«,) + at1X
+ 3(ot0 - at1)x1
+
(0 Yes ; 7{«o + ot1X «1X -t otar + ot,x•) = - 2ato + 3ot1 +
2 4ata
(1) Yes ; T(x, y) == (.x - iy, x - y - iy).
4. T(x, y) = (0, y - x).

S. T(x, y) = (3x-2+Y
--, y).

Problem Set 4.2

1. (a) Y1 ; 2 (b) [(I, 1, 0), (0, I, 1)) ; 2 (c) J/1 ; 3 (d) J/1 i 3
(e) Y1 ; 3 (0 ((1, 1, l, 0), (0, I, I, 0), (0, 0, 1, I)] ; 3
(g) Y, ; 4 (h) {p € g I p(O) == O} (i) {p € g I p(O) - O}
(j) fl (k) '6'(0, l) (1) if (0, 1).
2. (a) Y0 ; 0 (b) Y0 ; 0 (c) Y0 ; 0 (d) J/0 ; 0
(e) [(I, -1, 1, ))] ; l (() J/0 ; 0 (s) J/0 ; 0 (h) J/1 ; 0
(i) Set or
all constant polynomials ; 1 (j) J/0 i 0
(k) J/0 ; 0 (1) Set or all constant functions ; I.
+
6. 7lxi, .x1 , .x1) == (4x1 - 3.x1 x1 , 8x1 - 6.x1 lx1 , 0). +
7. 7l.x1 , x1 , x1) = (x1 + x11, x1 +2x1 , -4x1 + 3x1 + 2xa).
8. (a) : (a), (b), (c), (d), (f), (g), (h), (j), and (k) are one.one
(b) : (a), (c), (d), (e), (g), (j), (k), and (I) are onto
· (c) : (a), (c), (d), (I), (j), and (k) are one-one and onto.

Problem Set 4.4

1. All are nonsingular;
R-1(.xi, x1, x.) =- t(4x1 + x1 - 3x1 , 4x1 - 4.x1 - 3x1,
-x1 + x1 + 2x1)
 ANSWERS TO PM>BL!Mll / 305
s-1(x1, x,, x,) = ½(7x1 +
X8 , 7x1 + 7x1 2x1 , -xa) +
T- 1(x1o x,, x 3) =
l(-9x1 , - 2x1 - Sxa - 3x3 , Sx1 + 8x1 + 3x,).
l. (a) T->(xi, x1) = (xJ«1 , x1 /a.a)
(b) T-1(x., x 1 , x3) = (x1 - X 2 , X 2 - x3 , x3)
(c) T· 1(<7 0 + «iX +
«1X1 ) = (rx 0 /3 - 0( 1 + 2-:t. 2 /3) + (2a.0 13
+ «1 - 2«1/J)x (-«J3 + «J3)x2.+
Problem Set 4.6
1. (a) (2S + 3T) : (xi, x (7l"1
2) . . - +
3x2 , 6x1 8x2) +
(b) (3S - 11') : (x1, x 1) 1---+ (-x 1 - 71(2 , 9x1 + 12.'t1 ).
2. (a) (S + T) : (X1, Xz, X3) . . - (2x1 - 2x2 +
x 3, S\"i +
6x1 I X3,
-x1 + 2x3)
(b) (3S - 2T) : (Xi, ¥ 2 , x 3) ~ (- 4x 1 + Qxl + 3x3 , - Sx1 -- I 2x1
+ 3x1 , -3x1 t- x 3)
(c) (a.S) · (xi, x 1 , x 3) ~ (O(x11 I u,, ax 1 +
«x3 , - «x1 + «x.).
3.
- - - - - - - - -- ---

(a) R f- 2S

(b) 2R + ST

(c) S - T -e3

(d) R + S + 2T 4e1 -- 2e1 + 2e3 10e1+ e 7e 1 - 1

(e) «R + ~s (3« + ~ + 3y)e1

+rT + ~e, + (4«- 7J
- 2y)e,
------------------- -
7. (a) T(x1, X2, x3 , x,) = (x1 + 2x2, - X1 + Xa - x,, -x, + .¥a,
-x + x,), 8

S(xi, X2, Xs, X4) = (X1 - X2 + X3, X1 + x, -f- Xa + x,,

Xa + x,, 2x1)

(b) T(X1, Tz, Xa, x,) = (x2 r x,, Xi -f- X3 + Xo -x. + Xa, Xa),
S(x1, x1, x8, x,) -= (xi - 2x1 - 4x3, 2x3 + 2x,, Xi+ Xa - x,,
x~ + ~xJ
(c) Tlxi, x 1 , x 8, x,) = (x2, x 2 + x3 - x,, 2x1 + Xs, Xa + 6x,),
S(x1> x1 , xJ, x,) = (Xi + xd/2 + 3x,/2, -x,/2 - Xa/2 + xJ2,
x1/2 + 3xJ2, 2x1 +· X 1 + xJ2)
(d) T = S.
306 / ANIWDI TO PJtOBUllfl

Problem S•t 4.7

1. (ST)(x1 , x1 , x.) - (x1 , x1 + x 1 + x.).
2. (a) (ST)(xi, x.) =- (lx1 + 2x1 , 3.x1 + 3.xa)
(b) (TSJ(x1, x.) == (5.r1 + 4x1 , O)
(c) S'(x1, x.) = (4xi, 18x1 + 16x.)
(d) (T1S)(xi, x.) == (5x1 + 4x1 , 0).
3. (a) (ST)(xi, x1 , x1) -= (4x1 ➔ b 1 + x 1, lx, - 3.Y, + x1 ,
-2x1 + 3x1 + x.)
(b) (TS)(Xi, x1 , x 1) - (-3x1 + 2x1 - X1 , 6x1 + 4x1 + 10x1 ,
-Xi+ X.)
(c:) (STS)(.x1, x1, x.) == (5x1 + 4x1 +
l lx1 , -4.x1 2.x1 , +
2x1 - 2x1 + 2.x1)
(d) (TST)(xi, ~, x 1) = (2x1 + 2lx1 - x1 , 28x1 6x1 + 10x1 ,+
-2x1 + 3x1 + xJ.
4.

11
'• '•
- - ----
{a) ST le, - '• - 1,. -2e1 - 8e1 + 35e1 e1 - Sr, + 14e1
(b) RT 3e1 + 2r1 + 3r1 -J2e, + 3,. -le1 + le1 - &.1
- 20e.
(c) RST -20e1 + le1 95e1 + 6e1 38e1 + 6e1 + 5le1
- 29ea + l32e,
(d) R(S + T) 3e1 + 4,1 + 2,1 -1Je1 + 2e1 -2211 + lea
- 19e1 - 35,,
(e) 7" e, - '• + 4e1 -12,, - 3,. -3e1 - 3e1 + 7e1
+ 13e1
co rsT 35e1 + 22e1 -lle1 - 79e1 19e1 - 28e1
- 54,. + 133,, + 37,.
5. S(.x1, ~) = (x1 - x 1, x1 - x 1), T(x10 x1) == (x1, x.).
9. (a), (b), (c:) Range : Y1 ; kernel,: J/1 ; rank: 3 ; nullity : 0.
14. (b) S1S1 is idempotent if S1S1 -= S 1S1 ; S1 + S 1 is idempotent if
s,s. + s.s, == •
(4) 7{x1, x1 , .x.> = (x1, x1 , 0), S(x1 , x1 , xJ = (0, 0, x1 ).
15. {b) Yes; 5 (c) Yea (d) No
 A.NSWIU TO ftOILl!MI / 30'1

(e) Lesa than or equal to the minimum of the degree■ of n.ilpotence

or S and T ; S + T is nilpotent.
17. (b) (I + >.7')-1(«1 + ll1X + «1x' + ac1x') = «0 + ac1X + (st - ACe
- 2~a1)x' + («1 - >.ac 1 - >.at1 + >.1ato + 2)....)zl.

Problem Set 4.B

J.
------
Rang, Krrn,I Pre-image o/(3, -1, 2)
------ -- - -
R Y, Y, (I, 2, 0)

s v. Yo (23/7, 18/7, -2/7)

T Ya Y, (I, -7/9, 13/t)
- - - -------- - - - - - - - - - - - - - - - - -
2. (b) I (c) (8/7, - 18/7, 3) (d) (1, 3, 2)
(e) (I, -1, 3, 0) + [(I, -1, 1, I)].
3. (a) : (h) i (i) x + [I] (j) -x/2 (k) .:¥ coscc x
(I) -(I + x)r" + [I)
(b) : {h) x (i) (r/2) +
(f] (j) -(1 -I x')/2
(k) r
cosec x (1) -(2 + 2;,c + x'),-. + [I].

Problem Set 4.9

1. (a) 2y - (<' + eX1)e"
(b) y - ~X + .r3_ + 2 COS JC + (JC - _!)
%
sin %, X "F O , Y - 0
when JC== 0
(c) 3y - C(3x - 4)--IIJ - .x'(3x - 4)-111, .:¥ =I 4/3
(d) Y = c,-u: + tFJ(ac + P), cz +~-:I: 0; y == (C + x>--u
if ' == -at (e) X == Cy + 'Y'
,Br
-r· ~ -:/- o; ,
co , = c,- - T - at+~ atX
== atx'
2 + ax + c
if' == 0 {I) -= C,-tan-l y + tan l y - 1
X

(h) x- c,• 7 + e• 11,-un:, cos• y dy

(i) (y - l)x - (y + l)(C - y + 2 In (y + 1)), y > l
{j) x == C COi y +
sin y (k) y(I + ~ .,. C - cos ,t
+ (X - 1)1" + x'/3
. (I) 1v'(l + .r) - C + i In (v'(I
v(I + - I
+ ~) +
,
1),
2. (a) ,CC + ainh-1 .r) .... v'(I + r)
308 / ANSWERS TO PROBLIMB

(b) y'(Cr1 + 1 + xi) = I

(c) r(C + 2x) = r 9

(d) x(C - y') = e' {e) x(C cosy+ sin y) = I.

3. (a) LI ;yes (b) LI ; no (c) no (d)
LD ; LD ; no
(e) LD ; no (0 LD ; DO (g) u ; yes (b) LI ; yes.

Problem Set 6.1

I. (a) 1Li1 -•7l_j (b)

r-t -17
I
L t f_j
I.
r2 1 27 r-i -1 -1-1 7
2. {a) I I (b) I I.
L2 0 I_J L \'- 4 23_j
4 2 3 17
-1 0 I 0
3. 4 1 0 2

1 I I 0

13 2 2 5..J
ro 1 o o 7
0 0 2 0
4.
0 0 0 n
LO O O O _J
ro -1 -1 -1 -17 0 1 0 1 07
0 0 0 0 0 0 -1 0 -1
5. (a) 0 O 1 O O (b) 0 O O 1 -1
0 0 0 1 0 0 0 0 1 0
LO 0 0 0 l_J
-1 -37
1 3
1 -1

o -l..J
 , ANswns TO PllOBLIIMS 1 309
rt l7
7. I I.
LI -I_J
r1 07 r1 I 17
r1
8. (a) I
LI
17
oJ (b)
ILO l_JI
1
1 (c) I 1. - , o I
LO O l...J
r1
fi ~l
0 0 0 -17
I

(d) 0 (e)
I0 1 1 o II
LO 1 O_J LO 0 l -1.J
0 07 r3 2 0 07

(f)
r:
LO 0
1 0
1 l
l_j
(g)
l 0
LO l
0 -I
0
0
-1
0

0.J

Problem Set 5.2

I. (a) T: V, -+- V1 ,
T(xi, x., X 8 , x,) = (x1 + x1 + 2x3 + 3x,, X 1 + X:, - x,,
X1 + 2x1)
(b) T: V, - V3 ,
T(x1 , x1 , x 3 , x,) = (7x1 + 2x2 -I llx3 - Sx,, lh1 + 7x1 + 22x 1
- 19x,, 13x1 + 8x1 + 30x, - 23x,).
2. (a) T: V1 ➔ V1 , T(x1 , r 1 , x1 ) = (x1 , X1., x1 )
(b) T: V1 ➔ V1 , T(x 1 , x1, x3 ) = (x1 + 2x1 -- 2x3, - x1 + x1 + 2x1 ,
X1 -t Xa -f- Xa)
(c) T: V1 ➔ V8 , T(x1 , xJ, x3) = 3(-x1 - 4x, + 6xa, 3x1 + 3x1,
- 2x1 + x1 t 3x1 ).
3. (a) T: V1 ➔ V1 , T(x1, x1 , x1 ) = ("1 - X1 + 2xa, 3x1 + x.)
(b) T: V1 ➔ V1 , T(x11 x1 , x 1 ) = (2x1 + 8x1 - 6x1 , 2x1 - 8x1
+ 4x.)
(c) T: V1 ➔ V,, 1(x1 , x1 , xJ = (Sx1 - 2x1 , -xi - 2x1).
4. (a) T: V, ➔ V1 , T(xi, x1) = (x1 + 2x1 , x1 , -xi+ 3x.)
(b) T: V1 ➔ V8 , T(x1, x.) = (2x1 - x 1, x1 , 3x1 - 3xJ
(c) T: V1 ➔ 1'1 , T(xi, x1) = 1(2x1 + 4x1 , -Xi - 2x,, -17x1 + x.).
r cos 8 sin 87
s. I I.
L -sin e COi O_J
JlO / ANSWIU 1V PaOBUIMI
.
Problem Set 5.3

I. (a)
r
I
9 4
Lil 9
137
I
-S..J
(b)
r
I
L--7 I 5.J
I 6 77
,.
rs 67 r-1 47
2. (a) I I (b) I 1.
Ll9 -2..J L O 5.J
r-16 10 -6 87 rt8 1 19 127
3. (a)
I
L -6
-8 -28 -40 -161 (b)
-4 4
2
ILl9 28 38 18 I
-6..J 8 6 5.J
2 -17 -15 -287

(c)
I
L-15
r 2
10 8

-4 -14
-7 -37
14 2

3.J
r:
I

I.
I
-l 1.;-7
4. (a)
114
L6
-26
9 -•:J (b) I · -t i I
Ll l --1..J

(c) rt: -J
-i
ff7
l-. .
L\' u -t.J
-S 07

1-•-1 -2
6.

L-1 -4
0

-1 -17 -33
-9
_:J 147
s s S -4
7. 0 10 44 -6 .
0 0 -14 -2
L O 0 0 36
 ANSWERS TO Pll08LEMS / 311
1 1 0 07
0 0 0 0
0 1 0 1
9.
0 1 0 0
0 0 1 1
ILO 0 0 0 _J
r1 1 l 0 0 07
I

10.
Io
I
0 0 l 0 0
.
I
I
0 0 0 0 0
I
LO 0 0 0 l_J
11. dim V .,_- 5. The matrices with respect to the ordered bai,is {sin x,
co:. :x, sin x ~os x, sin1 x, cos• x} are
rJ -2 0 0 07
I
I 2 3 0 0 0

(a) 0 0 3 4 -4
0 0 -2 3 0
I
LO 0 2 0 3 _J
- I I 0 0 07
-1 l 0 0 0

(b) 0 0 -8 -2 2

I 0 0 1 -2 6
I -2
L 0 0 -1 6

Problem Set 5.4

1. (a) -1 (b) 9 (c) 3.
2. (a) (2, 3, 3)T (b) (8 1 19 1 7)T (c) The matrix [ -1)
1 27
2 4 r17
(d)
1-:
L 3
-1
1
-2
2-1
(e) I
Li -1
I.
312 / ANIWIU TO PI.OBLBMS
r 2 l
3. (a)· AB = Ir9 17
I , BA ==
/
1
L6 19.J I
l!.11

(b) BA=
r_:
I _; 9
-3
67

-8 21
L 11

I
r22 56
(c) AB= 3 5
L3 9
r3 11
(d) AB - I
L7 IS 24 28
r 3 2 4

24 11 16 19
(c) BA -
-1 -1 -3 1
L -6 O -2 -4
ro 07 r-2
(f) AB == I
I ' BA. = I
LO O_J L-21
ro 17 r1 -27 r o 11~ 1J
4 • (a) L1 oJ (b) Lo 1J (c) I I

r-: :
L J/~ -«/~• _J
17
(d) j l/« O 7 (c) -1 I.
r:
L-~/«y J/y_J ,. I
L-¼ O ,_j

5. (a) A.• - rt: : :7J .

L 3 4 8
A' =
L27
21
10 12 , ,
6
337

13_j
r124 59 967

"'-I 4S 26
L 66 53 87_j
371
 ANSWIRS TO PllOBLIIMS / 313

(b) .4• = j7
L6 7-J
47 , ..t• = j 19
L27
187 ,A'= 173 567
l9_j L84 73_j
r 10 5 107 r 35 60 107

(c) A' =I -5 4 -I I• A'~ I -20 3 -12 I•

L 5 4S SO_J L 20 340 3S5_!
17 125 520 5957

A' ==. I -75 -54 -104 1•

~ 75 2455 2505j ,•
rlX -2~ 7 JIX ~ 7 r« 07
6. (a) I I (b) I I (c) I I
L~ IX+ ~_J L~ X - [i_J LO ~_j
r IX ~7
(d) I I , where «and ~ are arbitrary scalars.
L-~ IX_J

7. (a)
r 1 -i7
I I (b)
r
I
. \· -l7
I (c) Solution does not exist.
L1/- f _j L-io LJ
rA11L1 0 0 7 r!A-1>-1 0 0 7
0 >-21Lt 0 0 (,l,d>-2 0
8. (a) (b)

LO 0 >.,.(L,._J LO 0 1-'n>.,._J
r-11>-, 0 0 7
I

(c) O i,i0.,-ft0,i=l,2,.,.,n.

I
L O O 1/>.,._j
9. (a) Nilpotent; 4 (b) Nilpotent; 4
(c) Nilpotent; 3 (d) Not nilpotent.
r 1 - IX1 7 IX P7 r r1 07
14. IL~ IX p-
-IX _J
I. I ·-=-~
L
at• -
at
I· ± I
_J LO
I,
1_J
r-1 07 .
±I I , where IX and ~ are arbitrary scalars with P'I: 0.
L O l_J
r -5 so7 r-21 607 r 22· -247
16. (a) I I (b) I I (c) I I
L-" ~_J L-~ ~_J L-~ %_j
314 / ANSWt.u 10 PllOBLl!MS

r-2S 3S7 r 64 -707

(d) I I Ce) I I
L 35 -60..J L-10 S4..J
r 4119 -25467
(f) I I.
L-25460 37217..J
18. (a) x--=1,y=l (b) x = bf~. y = (a, - bx)/~1 , ~ -I= 0
((.) X -- -y -= 4z = 2.
20.
-----
D,I Di Da D, D• D, D.,
----------- - -- - -
I I Di D1 Da D, D,. D, D.,
Di D1 I Da Da D• D, D, D,
D, D1 D1 Di I D, D, D\ D,
Da Da n. I D1 D., D, D, D,,
D, D, D,, D, D, Di I D1. Da
Da D, D, D, D, I »1 Dd ».
D, D, D, D, D,, Da D1. Di I

D, D, D, Dr, D, ». Da I »1

Problem Set 5.5

I. (a) Range: [(l, -1, 1), (3, 7, 0), (2, 2, 1)) = V1 ; kernel : V0 ;
rank : 3 ; nullity : 0
(b) Range: [(I, 3), (-1, -2)) = 1'1 ; kernel: ((-1, 1, 1)]; rank:
2 ; nullity : 1
(c) Range: [(2, 7, 3), (0, 1, -1), (1, 21 l)] = V1 ; kernel. J'1 ;
rank : 3 ; nulhty : 0
(d) Range: ((2, o, 1, 2), (1, -1, 2. 0), (2, 2, 4, 3), I, 3, 0)J ~o.
== Y, ; kernel : [(- l5, 1, 9, 9, -12)) ; rank : 4 ; nullity : 1
(e) Range: ((1, 2, 1, 0), (-1, 3, ~. 0), (1, -1, 2, I), (0, 1, 0, 1)]
= V,; kernel : Y1 ; rank: 4; nullity: 0
(0 Range: [(-1, 3, 2), (1, I, 2), (1, -1, 1)) =
Y1 ; kernel: Y0 ;
rank : 3 ; nulhtf : 0.
 ANIWIRS TO PROBLEMS / 315

r 2 0 07 •
ro
2. (a) i I
LI
-27
I {b)
l.J
•I L
-4
5
2
-4
0
l_l
I
r-1 I --17
I

(c) i ! 3 -1 s ·I
I

r: - :
L 3 --1 3.J
r 7 -13 8 0 07
-3 9 0 :~J. Cc) l
-1 o I
4 -4 8 -12 0 0 o 2 I
I_ 2 10 -8 6 LO 0 2 -2.J
r- ~17
I
ex 11 --ex~

3• Ma+ r.ta f (I. __ !__

- I:' , Cl3 + ~I
I
ex~ --«~
a,1 _j
I·
L--Cl~ ,

Problem Set 5.6

r1 -I 37 r 1 37 r-2 77
I I

2. (a)
L3
I
I
I
2 -2J
1 2 (b) 1
'
L-1
0

2.J
(c) I -3 0
L s 4.J
I
-• I I
I
r 4 -2 77
I r-3 -4 s-1
(d) -1i I -3 (c) I
L 8 4 - 4_J
I
L I -3 8.J
r-2 --3 S7 r 3 --27 r 2 2 27
(f) I I (g) I I (h) I I
L 7 0 4.J L-2 1,0.J L2 --1 4.J
r2 3 17
(i) I I.
LO -I S.J
+;
r1 2-i7
s. (a)
r t - I -1+17
I
L3 + I 2 + i.J
I (b) ·12 + 31 I-4ij
Ll-i 3 + 2ii
r2 - 3; 1 + 2; 3 - 417
(c) I I
L2 +; 2 + 21 2 - '.J
316 / ANSWllts TO fl.Olt.md
r 1 2+1 1-17

(d) I 1 +I -i 2 - i I·
L-1 - 21 I 3 - 2/__J
6. ax• + by + 1 er
+ 2hxy + 2/yz 2gzx. +
9. (b) A diagonal matrix.

Problem Set 5. 7

1. (a)
r1 o
Io
I
LO O
1 ; -: -:1
0 0 O_j
1 0 0 07
0 1 0
0 1 0 0
1 1 0
(c) [ : - : ] (d) 0 0 1 0
0 0 1
0 0 0 l_j
r1 o o o 1 r1 o 07
0 0 -27
Io
(e)
0 1 0 0
0 0 1 0
0
1 0

I
1j
-3
(s)
I

I
1
O O
o
t,
LO O O l_j LO O O__J

(b)
rt
0 1 0
0 0
-1.l
1\·
(i)
r1
0 1 0
0 0 27
-I
0 0 1 -ii. 0 0 1 1
LOO O 0 _J LO o ·o O_J
2. (a) 2 (b) 2 (c) 3 (d) 4 (e) 4 (0 3 <s> 3
(h) 3 (i) 3.
3. Same as the answer to Problem 2.
r 1 -s 1 57

4. (d) Nonsingular; t
4
-3
1 -3
1 4 -1 -IJ
L 1 16 -6 -9
 ANSWIIU TO PROBLEMS / 317

r 3 3 -3 37
7 3 -S 4
(e) Nonsingular; l
-1 0 2 -1
L
5 3 -4 2-1
5. (a) X = 14/19, y = 3/19, Z = )3/19
(b) x = 1/4, y = 1/2, z = -S/4
(c) x = 1/9,y = -2/9,z = -1/3
(d) x = 5/7, y = 1, z = 3/7
(e) X = 8/5, y = 5, Z = 9/5.
7. k1 = 2, k1 == 4, k3 == 6, k, = 8, k, = 9, k, = 10.

Problem Set 5.8

1. (a) Consistent; x 1 = 5/16, x1 = 1/16, x 1 == 3/16, x, = 1/8
(b) Consistent; x1 = 2, x1 = I, x 1 = 0, x, = 0
(c) Consistent; (1, 0, 0, 0, O) + [(O, -4, 1, 1, 0), (0, 3, -1, 0, I)]
(d) Consistent; (1, 0, 1, 0) + ((-2, 1, 0, 0), (0, 0, -2, 1))
(e) Not consistent (0 Not consistent
(g) Consistent; 1(2, 1, 1, O) +
[(2, -8, -S, 9)]
(b) Consistent; x1 = 7/6, x1 = 41/6, Xa = 1/3, x, = -59/6
(i) Not consistent (j) Not consistent
(k) Consistent; x 1 == x1 == 2, x 3 = 0
(I) Consistent; x 1 = -x1 == J, x1 = 0
(m) Consistent; (7, 6, 14, 3, 0) + [(O, 14, 14, 1, 1), (7, -1, 12, 3, 4))
(n) Consistent; (1, -2, 0, O) + [(-4, 7, 26, 1)] •

Problem Set 5.9

I. (a) Nonsingular; {3 7 -3 -87

-7
-1
3
-4J
10

l
r-3 I
(c) Nonsingular;
¼ I
1 _:1
10 -2 -2.J
318 / ANS'WBU TO PI.OBLltNS

r 16 7 -1 -77
-6 I 4 -1
(e) Nonsingular; i,
1 -6 5 6

L-7 6 -5 23.J

(f) Nonsingular; 11-: -:J

L-4 -4
-1
3
27

2. Same as the answer to Problem 2 of Problem Set 5.5.

3. Same as the answer 10 Problem 4 of Problem Set S.7.

Problem Set 6.1

1. (a) Eftll (b) Odd (c) Odd (d) Odd.
2. (a) Even : (I, 2, 3), (2, 3, 1), (3, I, 2)
Odd: (I, 3, 2), (2, I, 3), (3, 2, 1)
Even: (1, 2, 3, 4), (I, 3, 4, 2), (1, 4, 2, 3), (2, 1, 4, 3),
(b)
(2, 3, l, 4), (2, 4, 3, 1), (3, 2, 4, 1), (3, l, 2, 4),
(3, 4, I, 2), (4, 2, 1, 3), (4, 3, 2, I), (4, 1, 3, 2)
Odd: (1, 2, 4, 3), (1, 3, 2, 4), (1, 4, 3, 2), (2, I, 3, 4),
(2, 3, 4, 1), (2, 4, 1, 3), (3, 2, 1, 4), (3, 1, 4, 2),
(3, 4, 2. 1). (4, 2, 3, 1), (4, 3, l, 2), (4, l, 2, 3).
3. (a) 6 (b) 10.
4. (a) (x - y)(.y - z)(z - x) (b) (x - y)(y - z)(z - x)
(c) a' +
b' + c' - Jobe.
Problem Set 6.4
1. n".
3. (a) -20 (b) -4 (c) 6 (d) 6 (e) 500 (0 o.
~l x. x. x.
x. Xi x. x.-1
n
6. X•-1 x. Xi x.-. == ;11... 1 (x 1 + Xs,.,, + X 1Ci>! + .. . + x.,.,:,-1) 1

X1 Xa Xt Xi
where ,.,1t fa>a, ... , c.,• arc n n-th roots of unity.
 ANSWBI.S TO PI.OILIMS I 3\9

1x'-+ Y1
I
+ z2 X y z
1xf + y~ + zf X1 Y1 Z1 :1
7. (b) Ix!+ Y! t- z1 Xa y, z, 1 = 0.
Ix:+ y: -1- z: Xa Ya Za 1
I
Ix! 1- v: + z, x, Y, z, 1
8. x = 0, -a - b - c.

10. (a}
r-o
I
Ll
-27
I
l_J
(b)
r
I
I
L
-4
2

s
0 07

-4
2
:-1
(c) r-: :J
L 3
-l

-1
1 -17

r-1 13 - 8 --97 -2 2 0 07
I
3 -9 0 -3 0 -2 2 0
(d) (e)
-4 4 -8 12 0 0 0 -4
L--2 -10 8 -6_J L 0 0 -4 4_j

Problem Set 6.5

2. Same JS the am,wer to Problem I of Problem Set 5.5.
4. Sctme as the answers to Problems I and 2 of Problem Set 3.5.
5. Same a,; the answer to Problem 2 of Problem Set S.S.

Problem Set 6. 6
5. 0.

Problem Set 6. 7
1. (a) x-= -1,y = 4 (b) x = 31/11,Y-= -5/11
(c) x - - 2, y ==-- 29/2, z = -8 (d) x = -1/2, y = 5/2, z == 2
(e) x-= 2S/2, y = -17/2, z = -1/2.
2. (a) X1 = 5/16, X2 = 1/16, Xa = 3/16, x, = 1/8
(b} x1 = 2x2 = 2, X,1 = x, = 0
(h) X1 =- 7/6, x 2 = 41/6, x3 = 1/3, x, = -S9/6.
3. Inverser. of the coefficient matrices :
r,-1 -1 17
r 2 -17 r 4 37
(a) I
L-5
I (b) ·l'r I
3.J
I
L-1 2.J
(c)
I 4

L-2 -3
s -1
2~
I
320 / ANSWDI TO P&OBLIMI

r-1 t -17 ,--1

I
I -17
(d}
I
L 1 -2
0
LJ
-I
l -i
I

The solutions are the same as those of Problem 1.

I I
(e}
L
0 -I
:J
4. x' == x cos 11 + y sin •• y' = -x sin at + y cos«.
Problem Set 6.8
1. (a} 0, S; K9 -= [(I, -3)), K1 = ((1, 2))
(b) I, -i; K, =
[(I, l)), K_, = [(- i, I)]
(c} -3, 2 ; K-1 = [(11, - l)J, K1 = [(3, 2))
(d) -1, -1, 8 ; K_1 = [(I, -2, 0), (0, -2, 1)), K1 = [(2, I, 2)1
(e} 1, 1, 0; K1 = ((0, 1, 2)), K 9 = [(O, 0, 1))
(f) -i, -i, 21; K_, = ((1, 0, -1), (0. I, -1)], K11 = [(I, 1, 1))
(g) -1, i, I, 3; K_1 = ((1,0, -4, O)],K1 , 1 = ((-8, 6, 8, -3)),
K1 = [(-3, 2, 2, O)], K1 = [(I, 0, 0, O)].
r, 07 r-3 07
2. l(b) I I l(c) I I
LO -I_J L O l_J
r-1 0 0 07
0 i 0
1(1)

L
0 0
O O 0
r1 0
I
:J 0 07
Io ' 0 0

.
3. >,.• - 1;
lo 0
_, 0
I
LO 0 0 -LJ

Problem Set 6.9

2. (a) Ce'- (b) c (c) -4 / I 1 - r I (d) -2 I x I 1,
where C is an arbitrary constant.
3. 0.

Problem Set 6. 10

1. (a) 41 + J - 51 {b) 41- 7J + 21 (c) (-3, 1, -7)

 ANSWBRS TO PROBLIIIS / 321
(d) (1, 5, -3).
4. U X V = 0,
S. (a) No (b) Yes (c) Yes.
6. (a) 8x + 4y - z == 19 (b) y - z + I 0 (c) Px - 1z • 11. =
8. (a • b X d)c - (a · b x c)d or (a · c x d)b - (b • c x d)a.
9. 11/v'26.
10. 3x - 4y + z = 11.
11. x/11 = y/2 = (3 - 2z)/14.
12. 4x + y + llz = 39.
13. 3x + 4y + 7 z - 13.

Problem Set 7.2

1 2 1 7 4 1
1. l(a) {(\ib' ,J6' y6), ( - v'66 I - ,16°6 ' - v'66 ),
1 1 3
<- vir' - v11 · v11 n
1 t 1 1
l(b) {(1, 0, O), (0, v'l' \1 2), (0, - v'2' ..jJ.)}
1 S 2 1 5 13
l(d)
Hvlo · ;_t30-· vlo->· <- vf9s: - v19s' -;;195>•
S I
( ,126' - \126 'O)}.

2. 2(b) {( 1
v6' - v6' v6'
_1 2 o>, c-L. _ ~-•
v3o • v3o v3o
~-.o>.
( 12 0 _ 6 s > <-2=. 0• __ ! --, _ --!-n
v12os • • J20s' -v26s • v41 • v4°f v4I
1 2 1
2(d) {(0, 0, 1, 0), (v'6' v6' 0, v'6)'
17 46 O 109 )}
<- vt4286' - vI4286' •v142s6 ·
- - S ...s + x - l-
3. 3(a) { \/'15 _. _ v' 151 _
4 ""- 4 4 ""- 4.
S I 1
- 2y'2 x' + v'2 " + 2v72}
3(c) { '\1'3 5'\1'7 x' _ 3'\1'7 X 3'\1'35 (xi + x•),
y2 X, 2 v'2 2 v'2 ' 2 v'94
v'S (47 - 42(.xi + .x'))}.
\111562
322 / ANSWERS TO PROBLEMS

5. (a) 0 (b) 0.
8. If v, is the first vector in the sp.1n of the earlier ones, then the
corresponding vector in the proce<;s of orthogonalisation will be zero.

Problem Set 7.3

6. Yes.

Problem Set 7.4

r6 o 07 r1s o 07 r1
1. (a) : 0 6 0 I (b)
I
o 9 o I cc) Io
I
LOO 3J L O O 4.J LO
2. (a) 4:c2 - = 36 or
9i· 2 .,... 4y =- 36-Qx-2 2

(b) 9~• - 4y1 -- 36 or - 4xi \ 9v'II :- 36

{c} 9.r + 4y 1 = 36 or 4x 2 t 9yi --= 36.

Problem Set A 1
1. (a) )' c,e-M + C3xe 2 ' (b) y =- C1 cm, 2.'- -t- C1 sin 2x
(l') J' - c, + cl co, .\ + Ca 'iin A {d) J' = • C1e2-' 11 -j- c,.e-..
(e) Y "- l\<'-' -t- e "' 1 (C2 coi, (\, 3x,2) l C, sin (y3t/2))
(0 y - C1e3r -t CJe ., (g) y C,e- (l+v2 1ll' -1- C1e- (I- v2)x
(h) -. C1e
y C2 x~
1 + CJe., +
(i) C1 -t ('4x -l C 3t.. I- C4e
l' -- 61

(j) Y = C, t C~e 9 r ; C,e !;ii

(k) J' -= ((C, + cl 't) cos (" 7x) -+ (Ca t- c,x) sin ( v7x))e- 1"
(I) )' (C1 I CzY)e-" I- (C, t C,x)e-"'
(m) J' = (C1 ·I Clx)e2rr f- (C3 -t C,x)e..
(n) y --:: C,e-<HvS 1Y/2 +
C2e-(3-'\,1 S1A/2
(o) y = C 1e? 12 + C8e a.r +C 3 (p) y = C1 + (C, + C3x)e'"'.
Problem Set A2
1. (a) y = C1 cos 2x + C1 sin 2x - (r/8) cos 2x + (x/16) sin 2x
(b) y -=- C,e•• +
C1 xe'-v: (2x +
l)e-111/32 +
(c) y =-= C1 cos 2x C1 sin 2x + +
(l/4) cos 2.x In cos 2x
+ (x/2) sin 2x (d) y == C1e""JI c,r- + 3e"/2 +
(e) y = C1 cos (x/2) C1 sin (x/2) + +
sin (x/2) In (sec (x/2)
+ tan (x/2)) - l (f) y == C1 cos x C1 sin x - (x/2) cos~ +
 ANSWERS TO PR.OBLBMS / 323
(g) Y -= C1 r 1"' + CJxe-2:IJ + x3e 2'"/6
(h) y C,e'"t 2 I- C2 e 2 ·03 -t (Jx -1-- l l)e-.-/9
(1) Y --,, G\ -1- C2 CO'i ~ -1- C3 sin x - 2x + x9/3
(j) Y -- C1 I- C2~ + C 3 e "' - cos x - (x/2) 1>1n x
(k) J' - C1e3z 1-- C,e· -1-- C3e-z + xe'z/8
(I) y -= C, t- Cae"'' I- C,e ~ t ht '• 1 /3.

Problem Set A3
l. (a} y -= C 1 e3,. + Cle .. - e2J /3
(b) y =- C1 coi, 2x 1 cl '!tin 2:c (.t/4) cos 2t
X 4
{c) y ---= (C, co,; x -1-- CJ ~in .:\)e' -t - er - - ~
~ l~
l 2 I
+-y..\-\+:?
(d) y - C1e 1 - 1 I , '21 '" -I c~/ -I \ 2~ t- r.mhx
(e) y =-- C 1e'" I- Cixe' I- CJe-" l- (2 cor. x -t 2 -;in " l- he-..
+ ..\ie ' I- 8)/8
(f) y = C1 t- cl CO'J 3t -f CJ l>in 3x +- "t 2/l8
(g) y -.., (( 1 c1>s (y'3x/.?) l- C, .,m (\'h/2))e--· 4 !- C 1e"' + C,e-
x~ +h - 5 t- ftHh t· 2x2 ) coi,h ~ - (6x + xl)
':.inh x}
(h) y -- C 1 I- Ca cos X + C3 !.Ill X - X cos X

+ CzX + CJez + C,e .'.,, 25 x ~

(I.) y -=-= C I 2 x1 107
- .iO I 338 cos x

27. t-1 3 .
- 169 smx 13 xcos 2
x t 26 X SIO X

G) y = C1 t C,e.a + C e- 3 2"' - x 2/12 + x/36 t ;,{ cos x

1 . I 7 l
+ 50 sm x - fg e-llz + I00 xe-2'" + 20 x2e-2z

{k) y = {C1 cos {v7x) +C 1 sin ( y7x))e 2111 + 4-xe-z.

- 1e-1.11 sin 3x.

Problem Set A4

1. (a) -¼- (2x - l)e- (b) T (2x + 3 + 2e'•)

3'-' / ANSWIIU TO P&OBLBMI
(c) -x'/10 - 4:x/25 - : 3 sin x + :6 cos x
I
(d) 9 (3x + 2)r" - 91 (3x - 2).

2. (a) -e--'•/29 (b) 1~8 (sin 3x - 6x cos 3x)

{c) 1~1 (sin 2x - 10 cos 2x)em

{d) - 2~ (2 cos 2x + sin 2x)

t
(c) --(250.xl
1250
+ 475x1 - 1940x + 388)e•
1
(f) -e1• COS X (g) - - (7x
147
+ 2)e-l■ (h) 1 'e'-
840 X •
 Index

Abel's formula, 240 coplanar vectors, 89

addition function, 12 Cramer's rule, 231
additive inverse, 42 cross product, 241
associative operation, J9, 26, 134, direction of, 242
167, 291 properties of, 242
auxiliary equation, 277
DeMorgan's theorem, 7
Base space, 81 derivative family, 285
basis, 9S modified, 286
orthonormal, 262 determinant, 204
standard, 97, 161 expansion of, 204, 219
Bernoulli"s equation, 143 minor, 224
binary operation, 17 product of, 227, 228
diagonalisation, 237, 270, 271
Cartesian product, 6 diffe1ential equation, ordinary li-
central quadric, 267 near, 141, 144, 276
characteristic homogeneous,276
equation, 234 nonhomogeneous, 280
polynomial, 234 differential operator, 109
value, 234 dimension, 96, 100, 101
vector, 234 direct sum, 80
circulant, 223 directed line segment, 37
coefficient matrix, 195 direction cosines, 59
cofactor, 216 distance between points, 36
collinear vectors, 89 distributive operation, 21, 134, 167,
column, JSO 291
rank, 187 domain, 10
vector, 148 dot product, 50
commutative operation, 18, 20, 291 properties of, 50
complement, 5
co111plementary function, 142, 276 Echelon form, 188
composition, 24, 133 eigenspace, 236
'°ordinate vcetor, 99 eiaeqvalues. 234, 236
326 / lNDBX

eigenvectors, 234, 236 natural, 2.S9

elementary row operation, J87 space, 258
equality in Yn, 64 intersection, S
equivalent system, I 8S of subspaces, 77
Euclidean space, 258 inverse
existence and uniqueness theorem, of clement, 21
144, 239 of function, 29
extension of a basis, 99 of linear transformation. 122,
I 23, 125
Field, 20 of matrix, 167,177,200,226
finite-dimensional, 96 inversion, 203
function(s), 10 matrix, 200
absolute value, 12 invertible, 168
addition of, 25 isomorphic, 127
constant, 12 isomorphism, 122, t 27, 160
equal, 24
linear, 3t Kernel of a linear transformation,
one-one, 14 114, I 16, 139
onto, 13 matrix, 176
pointwise sum of, 25 Kronecker delta, 152
polynomial, J2
real-valued, 2S Leader, 81
scalar multiplication of, 26 left-handed ~ystem, 34
zero, 12 line
equation of, 56
Oram-Schmidt orthogonalisation, in vector space, 89
261, 2n linear
group, 21 combination, 74
abelian, 21 finite, 74
commutative, 21 nontrivial, 86
orthogonal, 265 trivial, 86
unitary, 265 equation, system of, l 95
extension, ·111
Homogeneous equation, 139 operator, 107
hyperboloid of one sheet, 273 transformation, l 07
idempotent, 137
Identity, 21 natural matrix of, 156
function, 12 nilpotent, t 37
map, 109, 115, 122 nonsingular, 122, 127, 13S
image, 10 nullity of, I 17
indicial equation, 277 product of, 134
infinite-dimensional, 96 range of, I 14, 116
initial value problem, 143 rank of, 117
inner product, 50, J64, 258 1calar multiple of, 129
 similar, 253 singular, 168
sum of, 12.9 skew-Hermitian, 182
variety, 81, 139 skew-symmetric, 182
linearly square, 155
dependent, 87, 88, 93 sub-, 224
independent, 87, 88, 93 sum of, 159
symmetric, 182
Map, 10 transpose of, 180
linear, 107, I 55 transposed oonjugate of, 182
zero, 109, 115 unitary, 264
matrix, I 49, I 55 upper triangular, 182
addition of, 1 j9 zero, 153
adjoint, 222 minor, 224
augmented, t 95 multiplication fonction, 12
characte1istic equation of, 234
characteristic polynomial of, 234 Natural matrix, 156
chat acterist1c value of, 234 negative triple, 243
characteristic vector of, 234 mlpotence, degree of, 137, 171
conjugate, 182 nonhomogeneous equation, 281, J39
diagonal, I 81 nontrivial solution, 139
diagonalisation of, 237, 270, 271 norm, 259
eigenvalue of, 234, 236 normal, 141
eigenvector of, 234, 236 n-tuples. 64
entries of, 150 addition of, 64
equal, 155 null t.t•t, 4
Hermitian, 182 null space, J 14
idempotent, 182, 229 nullity, 117, 176
identity, 152
involutory, 172 1 230 One-one corre!>pondence, 15
lower triangular, 182 onto. 13
multiplication of, 164 operational methods, 289
nilpotent, 171,229 operator equation, 138
nonsingular, 168 polynomial, 290, 292
nullity of, 176 order, 150
orthogonal, 264 _orthogonal !let, 261
polynomial, 172 orthogonalisation, Gram-Schmidt,
product of, 164, l 6 5 261
prop"r value of, 234 orthogonally similar, 266
proper vector of, 234
range of, 175 Parallel, 81
rank of, 175, 188 particular integral, 140. 280
scalar, 18.2 solution, 140, 276
scalar multiple of, I 59 permutation
similar, 253 even. 203
328 / JNDIIX

odd,. 203 Scalar, 63

perpendicular distance, 62 multipl~. 26, 39, 42, 64, 129, 159
plane multiplication, ru~tion, 12
equation of, 60, 245 multiplication, properties of, 29,
in a vector space, 89 43
vectors, 40 product, SO
addition of, 41 projection, 53
pointwise product, 27 triple product, 244
scalar multiplication, 26 properties of, 244
sum, 25 Schwarz inequality, 51, 260
polynomial, degree of, 12 sequence, 13
position vector, 40 set(s), 1
positive triple, 242 addition of, 78
pre-image, 13, 141 empty, 4
proper equal, 3
values, 234 finite, 15
vectors, 234 infinite, 6
null,4
Quadric, 267 of complex valued functions, 66
tangent to, 267 of continuous functions, 66
of infinite sequenceq, 124
R.an!le, 13, 114, 116, 175 of lin~ar transformations, 130
rank, 117, 175, 188 of m x n matrices, 156
column, 187 of n-times differentiable func-
row, 187 tions, 66
rank-nullity theorem, 120 of n-tuples of complex num-
reflection, 109 bers, 68
relation, 8 of n-tuples of real numbers, 64
equivalence, 8, 128, 194 of plane vectors, 41
reflexive, 8 of polynomials with complex co-
symmetric, 8 efficients, 66
transitive, 8 of polynomials with real coeffi-
right-handed system, 34 cients, 1,2, 66
ring. 23 of real-valued functions, 25, 66
commutative, 23 of space vectors, 41
with identity, 23 shortest distance, 246
row, ISO 1r,ing)eton, 4
equivalent, 187 skew lines, 247
operations, 187, 207, 209 space
rank, 187 coordinates, 33
-reduced echelon form, 188 vectors, 40
reduction, 188 addition of, 41
vector, 148 span, 89
 INDBX / 329
stain, 188 column, 148
steps, 188 components of, 46
straight line coordinates of, 99
in a plane, 57 difference of, 44
in space, 58 direction of, 41
vector equation of, 57 equal, 41
submatrix, 224 free, 38
subset, 3 initial point of, 33, 37
proper, 4 length of, 37, 40
span of, 74 magnitude of, 37, 40
subspace, 69 norm of, 259
nontrivial, 72 orthogonal, 52, 261
trivial, 72 parallel, 37
superset, 3 plane, 40
symmetric form, 57, 58, 61 product, 241
projection, 53, 261
Target set, 10 resolution of, 52
translate, 81 row, 148
translation function, 12, 109 scalar multiple of, 39, 42
triangle inequality, 260 sum of, 39
trivial solution, 139, 196 terminal point of, 33, 37
unit, 37, 44
Undetermined coefficient, me- vector space, 62
thod of, 285 complex, 62
union, 5 finite-dimensional, 96
uniqueness, existence and, I 44, real, 62
19.S
unitarily similar, 266 Wronskian, 238
unitary space, 258
Zero vector
Variation of parameters, 280 in plane, 40
vector(a), 37, 64 in space, 40
anale between, 48 of vector space, 64
Errata

f.t. = from top

f.b. = from bottom
Page Line h place of Read as

4 6 fb. page 3 page4

5 I f.t. zz I 2z t- 2: 5 =0 ii + 2z + 2z - S = 0
8 6 fb. ••. be~clUSe X • J' ... • becaui,,e x - x ...
11 1 f b. B - ( .. } ... B- { ... }
12 6 f.b. s called ... is called .. .
14 s f.t. •• / : l' - x• ... . .. f:xt-,X' ...
16 8 f.b. . Problems 4 and 5. •.. Problems 3 and 4.
16 7, 6 f.b. .. Problem 4 are one-one . .. Problem 3 are one-one
and which onto. and which are onto.
18 13 f.t. •.• elements (x, y) ... ... elements x, y ...
20 7 f.t. . .. real numbers. ... integers.
20 8 f.t. . .. Z X Z ... ...ZXZ0 •••

22 9 f.t. In all these four examples In all the examples of

of groups the operation groups listed above the
1s commutative . operation is commutative.
25 9 f.t. •.. (/ o g)(x) /(g(x)) ... (/ o g)(x) = f(g(x))
==f(r) =f(r>
25 19 f.t. . .. by :1(.4R). •.. by :J'(A., R).
26 I f.t. ••. function x x• + x + 1 ... function x f-+ X'
+x+l
26 2 f.b. ... function 1/2 sine ••• ... function (1/2) sino ...
Pase Line In place of Read as

6 f.t. ... /-l(y) = y-1 .. .../-l(y) =y - l

44 l 3 f.t. ... by • and ,. ... by u and -,.
48 1 f.b. •· O<8<n ... ... O<0<n, ..
49 4 f.b. . .. (a1, ~u) ... ... {a., a11) .. .

51 20 f.t. ... (or V)11 ••• ... (or 1'1) .. .

54 19 f.t. ... follow .. follows

83 3 f.b. ... E VJ ••• ... e: v, ...
85 8 f.t ... in ('/;' 0, 2). .. in ( '(0, 2).

92 3 f.h 1( I , 1, 0), -- I (0, I , 1) 1l I, 1, O) - 1(0, 1, 1) ...

96 10 f.t. ... hecause {1'J, .. , nn} • • .. because {vi, ... , v,.} ...
97 10 f.b. Take VE V. T.ike v E V.
97 8 f.b. l'l, l',, ... , l'r,Say l', ...
99 '1 f.b. .. . rxii.ts any non1ero ... . .. exi&ts a oonzerc> ...
100 1 f.t. .. • ici u; otherwise ... is 11; bcc-ausc
otherwise, ••
102 11 f.t. The reverse in equality ... The revcrt.c inclu11,ion , .•
102 22 ft. dim Vd - (xy-planel dim Va=-,, Jim
t ... (.~y-plane) + ...
120 J f.b. 7lci,.1-1u... t- ... 1 7(7n;-1U,..1 -t ,., + «..,u..,)
+ «,.up) =- 0 == 0
126 3 f.t. (Y1 + Ya e1 --1 ·••
126 2nd row c>1 --
-e
in Table
4.2
133 7, 8, 11, So T (So 7)
28 f.t.
151 last entry -t -t
of 2nd
matrix
rJ 07 r
.3 07
167 3 f.t. BA. - I I BA=- I I
LI -LJ L-1 -1.J
Page Line In place of Read as
- - - - - - - - - - ---
174 22 f.t. ... +xi ... ... + x • ...
181 4 f.t. . .. and -= ~;, = ~" ... . . . and ~~, = ~,, ...
182 3 f.b. ... (A)T ~ (A 1 ).

187, 188 The 7 x 11 matrix on page 187 should be read after

line 9 f.t. of page 188.
190 I f.b.

251 13 f.t. = A(u1 = = A(u1) =

267 12 f.b. u Au= k, u 1.Au = k,