MATRICES Teoria y Problemas

SCHAUM'S
OUTLINE
THEORY aiMl>ROBLEMS
SERIES
of
MATRICES
by
FRANK AYRES,
JR.
including
Completely Solved
SCHAUM PUBLISHING
NEW YORK
CO.
in
Detail
SCHAVM'S OUTLINE OF
THEORY
AI\[D
PROBLEMi;
OF
MATRICES
BY
FRANK AYRES,
JR., Ph.D.
Formerly Professor and Head,
Department of Mathematics
Dickinson College
^CÔ
-^
SCHAIJM'S OUTLINE SERIES

McGRAW-HILL BOOK COMPANY
New
York, St. Louis, San Francisco, Toronto, Sydney
('opyright
United
1962 by McGraw-Hill, Inc.
States
of America.
No part
AH
Rights Reserved.
Printed in the
may
be reproduced,
of this publication
stored in a retrieval system, or transmitted, in any form or by any means,

electronic, mechanical, photocopying, recording, or otherwise,
written permission of the publisher.
02656
78910 SHSH 7543210
without the prior
Preface
Elementary matrix algebra has now become an integral part of the mathematical background
necessary for such diverse fields as electrical engineering and education, chemistry and
sociology,
as well as for statistics and pure mathematics. This book, in presenting the
more
essential matedesigned primarily to serve as a useful supplement to current texts and as a handy reference book for those working in the several fields which require some knowledge
of matrix theory.
Moreover, the statements of theory and principle are sufficiently complete that the book
could
be used as a text by itself.
rial,
is
The material has been divided into twenty-six chapters, since the logical arrangement is
thereby not disturbed while the usefulness as a reference book is increased. This
also permits
a separation of the treatment of real matrices, with which the majority of readers
will be concerned, from that of matrices with complex elements. Each chapter contains
a statement of pertinent definitions, principles, and theorems, fully illustrated by examples. These, in
turn, are
followed by a carefully selected set of solved problems and a considerable number
of supplementary exercises.
The beginning student in matrix algebra soon finds that the solutions of numerical exercises
are disarmingly simple. Difficulties are likely to arise from the
constant round of definition, theorem, proof. The trouble here is essentially a matter of lack of mathematical maturity,'
and
normally to be expected, since usually the student's previous work in mathematics has
been
concerned with the solution of numerical problems while precise statements of principles and
proofs of theorems have in large part been deferred for later courses. The aim of the
present
book
is
lems
in
to enable the reader,
if he persists through the introductory paragraphs and

solved probany chapter, to develop a reasonable degree of self-assurance about the material.
The solved problems, in addition to giving more variety to the examples illustrating the
theorems, contain most of the proofs of any considerable length together with
representative
shorter proofs. The supplementary problems call both for the solution of numerical
exercises
and
for proofs.
Some
of the latter require only proper modifications of proofs given earlier;
more important, however, are the many theorems whose proofs require but a few lines. Some are
of the type frequently misnamed "obvious" while others will be found to call for
considerable
None should be treated lightly, however, for it is due precisely to the abundance of
such theorems that elementary matrix algebra becomes a natural first course for those seeking
to attain a degree of mathematical maturity. While the large number of these problems
in any
chapter makes it impractical to solve all of them before moving to the next, special attention
ingenuity.
is
directed to the supplementary problems of the first two chapters.

mastery of these will do
to give the reader confidence to stand on his own feet thereafter.
much
The author wishes

Publishing
Company
to take this opportunity to express his gratitude to the staff of the
for their splendid cooperation.
Frank Ayres,
CarHsle, Pa.
October, 1962
Jr.
Schaum
CONTENTS
Page
Chapter
MATRICES
Equal matrices.
Products by partitioning.
Matrices.
Chapter
Sums
of matrices.
Products of matrices.
SOME TYPES OF MATRICES
10
Scalar matrices. Diagonal matrices. The identity

matrix. Inverse of a matrix.
Transpose of a matrix. Symmetric
matrices. Skew-symmetric matrices. Conjugate of a matrix. Hermitian
matrices.
Skew-Hermitian matrices. Direct sums.
Triangular matrices.
Chapter
DETERMINANT OF A SQUARE MATRIX.
20
Determinants of orders 2 and 3. Properties of determinants,

and cofactors. Algebraic complements.
Chapter
Minors
EVALUATION OF DETERMINANTS
32
Expansion along a row or column. The Laplace expansion. Expansion

along the first row and column. Determinant of a product. Derivative
of a determinant.
Chapter
EQUIVALENCE
39
Non-singular and singular matrices. Elementary

transformations. Inverse of an elementary transformation. Equivalent
matrices. Row canonical form. Normal form. Elementary matrices.
Canonical sets under equivalence. Rank of a product.
Rank
Chapter
THE ADJOINT OF A SQUARE MATRIX

The
Chapter
of a matrix.
adjoint.
The adjoint
of a product.
Minor of an
49
adjoint.
THE INVERSE OF A MATRIX.
55
Inverse of a diagonal matrix. Inverse from the adjoint. Inverse from

elementary matrices. Inverse by partitioning. Inverse of symmetric
X w matrices.
matrices. Right and left inverses of
Chapter
FIELDS
Number
64
fields.
General
fields.
Sub-fields.
Matrices over a
field.
CONTENTS
Page
Chapter
LINEAR DEPENDENCE OF VECTORS AND FORMS
67
Vectors. Linear dependence of vectors, linear forms, polynomials, and

matrices.
Chapter
10
LINEAR EQUATIONS
75
System of non-homogeneous equations. Solution using matrices. Cramer's

Systems of homogeneous equations.
rule.
Chapter
11
VECTOR SPACES
Vector spaces.
section
space.
85
Sub-spaces. Basis and dimension. Sum space. InterNull space of a matrix. Sylvester's laws of nullity.
Bases and coordinates.
Chapter
12
LINEAR TRANSFORMATIONS
Singular and non-singular transformations.
space. Permutation matrix.
Chapter
13
94
Change of
basis.
Invariant
VECTORS OVER THE REAL FIELD
100
Inner product.
Length.
Triangle inequality.
Schwarz inequality.
Orthogonal vectors and spaces. Orthonormal basis.
Gram-Schmidt
orthogonalization process. The Gramian. Orthogonal matrices. Orthogonal transformations. Vector product.
Chapter
14
VECTORS OVER THE COMPLEX FIELD
110
Complex numbers.
Inner product. Length. Schwarz inequality. Triangle inequality. Orthogonal vectors and spaces. Orthonormal basis.
Gram-Schmidt orthogonalization process. The Gramian. Unitary matrices.
Unitary transformations.
Chapter
15
CONGRUENCE
115
Congruent symmetric matrices. Canonical forms

of real symmetric, skew-symmetric, Hermitian, skew-Hermitian matrices
under congruence.
Congruent matrices.
Chapter
16
BILINEAR FORMS
125
Matrix form.
formations.
Chapter
17
Transformations. Canonical forms. Cogredient transContragredient transformations. Factorable forms.
QUADRATIC FORMS
Matrix form. Transformations. Canonical forms. Lagrange reduction.
Sylvester's law of inertia. Definite and semi-definite forms. Principal
minors. Regular form. Kronecker's reduction. Factorable forms.
131
CONTENTS
Page
Chapter
lo
HERMITIAN FORMS
Matrix form.
146
Transformations.
Canonical forms.
Definite
and semi-
definite forms.
Chapter lif
THE CHARACTERISTIC EQUATION OF A MATRIX

Characteristic equation and roots.
Chapter
20
21
Invariant vectors and spaces.
SIMILARITY
156
Similar matrices.
Chapter
149
Reduction to triangular form.
Diagonable matrices.
SIMILARITY TO A DIAGONAL MATRIX
163
Real symmetric matrices. Orthogonal similarity. Pairs of real quadratic

forms.
Hermitian matrices. Unitary similarity. Normal matrices.
Spectral decomposition.
Chapter
22
Field of values.
POLYNOMIALS OVER A FIELD

Sum, product, quotient
common
divisor.
Least
of polynomials.
multiple.
common
172
Remainder theorem. Greatest

Relatively prime polynomials.
Unique factorization.
Chapter
2o
LAMBDA MATRICES
179
The X-matrix or matrix polynomial. Sums, products, and quotients.

Remainder theorem. Cayley-Hamilton theorem. Derivative of a matrix.
Chapter
24
SMITH NORMAL FORM

Smith normal form.
Chapter
25
Invariant factors.
invariants.
derogatory matrices.
26
Elementary
divisors.
THE MINIMUM POLYNOMIAL OF A MATRIX

Similarity
Chapter
188
Minimum
polynomial.
Companion matrix.
Derogatory
196
and
non-
CANONICAL FORMS UNDER SIMILARITY
203
Rational canonical form.

second canonical form. Hypercompanion
matrix. Jacobson canonical form. Classical canonical form. A reduction
to rational canonical form.
INDEX
215
INDEX OF SYMBOLS
219
'
chapter
Matrices
A RECTANGULAR ARRAY OF NUMBERS

"2
enclosed by a pair of brackets, such as
and
-1
l'
(a)
(b)
and subject to certain rules of operations given below
is
called a matrix.
The matrix
(a) could
be
(2x + 3y + 7z =
linear equations
\ X- y + 5z =
considered as the coefficient matrix of the system of homogeneous
i2x + 3y = 7
-
or
as the augmented matrix of the system of non-homogeneous linear equations
we
Later,
shall see
how
the matrix
may be used
dinates of the points (1,3,
1),
(2, 1,4),
to obtain solutions of these
we might consider
The ma-
rows as simply the coor-
and (4,7, 6) in ordinary space.
The matrix will be used

same plane with the
or
such questions as whether or not the three points

on the same line through the origin.
later to settle
origin or
systems.
y = 5
its
could be given a similar interpretation
trix (b)
\x-
lie in the
In the matrix
^11 ^12
'ÎS
'
(1.1)
*mi "m2 ""m3
the numbers or functions a^- are called its elements. In the double subscript notation, the first
subscript indicates the row and the second subscript indicates the column in which the element
stands.
Thus,
all
elements in the second row have
the fifth column have 5 as second subscript.
"m by
order
(In
used.
ra"
or
as
first
subscript and all the elements in
rows and n columns
said to be of
),
and double bars,
||
||,
are
sometimes
shall use the double bracket notation throughout.)
At times the matrix (1.1) will be called "the mxra matrix

[a^-]".
is
mxra.
indicating a matrix pairs of parentheses,
We
matrix of
When
we
the order has been established,
SQUARE MATRICES. When m
n, (1.1) is
[a^
"
or
"the
shall write simply "the matrix
mxn
matrix A =
4".
square and will be called a square matrix of order n or an
re-square matrix.
In a square matrix, the
The sum
elements
of the diagonal
a^,
022.
" are called
elements of a square matrix A
its
diagonal elements.
is called the trace of A.
MATRICES
EQUAL MATRICES. Two
matrices A = [a^] and B =
[CHAP.
[bij]
are said to be equal (A =
B)
if
and only
if
they have the same order and each element of one is equal to the corresponding element of the
other, that is,
if
and only
if
Thus, two matrices are equal
ZERO MATRIX. A
'V
^^J
and only
if
= 1,2,
if
one
which
matrix, every element of
ro;
1,
n)
duplicate of the other.
is a
is zero, is
When ^
called a zero matrix.
we
matrix and there can be no confusion as to its order,
is a zero
shall write A = Q instead of the
mxn
array of zero elements.
SUMS OF MATRICES.
is
If
defined as the
4 = [aÂ and S =
mxn
corresponding elements of
1.
31
It
[fe.^-]
[cÂ
mxn
are two
matrices, their
is the
sum
(difference),
sum
A B,
(difference) of the
'\
Example
C =
where each element of C

A and B. Thus, AB
[Oy bij]
matrix
14
and =
2
I
then
-12
+B
0'
Lo+(-
1)
5
2
+3
+2
+ 5j
and
Two
1-2
2-3
3-0
O-(-l)
1-2
4-5
matrices of the same order are said to be conformable for addition
matrices of different orders cannot be added or subtracted.

(b) above are non-conformable for addition and subtraction.
The sum
of k matrices
is
a matrix of the same order as
times the corresponding element of A.
We
define:
If
2.
If
[:
'11
A+A + A
subtraction.
A and each
(a)
Two
and
of its elements is k
k is any scalar (we call k a scalar to dis-
tinguish it from [k] which is a 1x1 matrix) then by kA = Ak

A by multiplying each of its elements by k.
Example
or
For example, the matrices
is
meant the matrix obtained from
then
-2
3A
A-3
and
-5A
r-5(l)
-5(-2)"|
r-5
10-1
L-5(2)
-5(3) J
L-10
-15j
by -A, called the negative of /4, is meant the matrix obtained from A by mulby -1 or by simply changing the sign of all of its elements. For
every A, we have A +(-A) = 0, where
indicates the zero matrix of the same order as A.
In particular,
tiplying each of its elements
Assuming
(a)
(b)
(c)
(d)
that the matrices
A,B,C are conformable
for addition,
we
state:
A + B = B + A
(commutative law)
A + (B+C) = (A + B)+C
(associative law)
k(A + B) = kA + kB = (A + B)k, A- a scalar
There exists a matrix D such that A + D = B.
These laws are a result of the laws of elementary algebra governing the addition of numbers
and polynomials. They show, moreover,
1.
Conformable matrices obey the same laws
of addition
as the elements of these matrices.
CHAP.
MATRICES
1]
MULTIPLICATION. By
the product
AB
in that order of the
Ixm
matrix A = [aî a^g ais
and
a^m]
fell
the
^31
mxl
matrix
meant the 1x1 matrix C = [on fen + 012
is
fi
fesi
+ aimfemi]
fell
feî
That
[an
is,
Oi
'îmj
Lîi
fell
"*"
^12
feji
imi]
-|^^2
a^^fe^^J
femi
Note that the operation is row by column; each element of the row is multiplied into the corresponding element of the column and then the products are summed.
1'
Example
(a)
3.
[2
[2(l) + 3(-l) + 4(2)]
4]
[7]
2.
-2
(b)
[3
-1
[-6 - 6+ 12]
4]
3J
By
is
the product
meant the
mxn
AB
in that order of the
matrix
C =
\c;;'\
<-
iiîj + "t2 ^2; +
Hj
-'
mxp
matrix A = [a^-] and the p
xn matrix B =
[bij]
where
p
^^^"ikbkj
"-ip i>,pj
(f
1, 2,
,m-;
1,
re).
Think of A as consisting of m rows and B as consisting of n columns. In forming C = AB

each row of A is multiplied once and only once into each column of B.
The element c^ of C is then
the product of the ith row of A and the /th column of B.
Example
4.
^11
*^ 1
."31
'011611 + 012621
O32
[611
612I
621
622J
021611+022^21
."31611+032621
11 ^12 + "12 622

0^21
^12 +
'^22
^22
"31612+002600
The product ^S is defined or A is conformable to B for multiplication only when the number
columns of A is equal to the number of rows of S. If ^ is conformable to B for multiplication
{AB is defined), B is not necessarily conformable to A for multiplication (BA may or may not
be
of
^^îned)-
See Problems
Assuming
that
A,B,C
are conformable for the indicated
(e)
A(B + C) = AB
(/")
(A +
AC
(g)
A(BC) = (AB)C
B)C
AC
BC
3-4.
sums and products, we have
(first distributive law)

(second distributive law)
(associative law)
However,
(A)
(i)
(/)
AB
AB
AB
i=
=
=
BA, generally,
does not necessarily imply i =
or S =
does not necessarily imply B = C.
0,
AC
See Problems
3-8.
MATRICES
PRODUCTS BY PARTITIONING.
A=
Let
[CHAP.
be of order
[a^J]
mxp
8=
and
forming the product AB, the matrix A is in effect partitioned into
Other partitions may be used.
into n matrices of order pxl.
[b^j]
be of order pxn.
In
matrices of order Ixp and B
For example,
A and 6 be
let
parti-
tioned into matrices of indicated orders by drawing in the dotted lines as

(piXTii)
(mixpi)
(m]^xp2)
(P2X%)
Jmgxpi)
(m,2Xp2)
(p2Xra2)
(m2Xp3)_
(psxni)
Am
(pixnj)
(m-Lxps)
A^2
(p3Xre2)
Aig
"21
O22
A,
Aî
I
"31
In
any such partitioning,
in exactly the
necessary that the columns of A and the rows of 6 be partitioned

m^, mg, re^, rig may be any non-negative (including 0) integers
and rei+ ^2 = n. Then
it
is
same way; however
such that mi+ m2 =
^llOll + '4i2021
^8
"32
""
'4l3"31
A-^-^B-12
+ A-^qBqq + '^13832
Cii
A-2_iiB^^
A^^B^^ + ^22^22 + ^423632
_C21
C12
c
A21B11 + 422^21 +
Examples.
Compute
/4S, given
/I
1^22.
1110
and B
1
110
12
Partitioning so that
^12
11
A21 A22
AB
we have
'0
10
11
[A^-iB-iî
111
^12
fill
and
S22
B21
+ A-12B21
A-ij^B'12 + A-i,
=
1
^21^12 *
AqiB^^ + ^22^21
1
i4'
'
+ [l][2
1]
-^2^
B,
[2]
[3
3]
[1
0]
[?
1"
[1
0]
_[l
3-1
[4
1]
To
+ [l][2]
[2
o]
1]
To]
Tol
[0]+[2]
5JL0J
[3
2] [2]
See also Problem
Let A, B.C... be re-square matrices.
9.
Let A be partitioned into matrices of the indicated
orders
(pixpi)
(piX P2)
(P2X pi)
'
""
^"
(P2X Ps)^
(piX Ps)"
All
A-^2
(P2XPS)
A21
A22
'
j
.(PsXpi)
and
let
8, C,
...
(psXp2)
...
I
(PsXPs)
/loo
be partitioned in exactly the same manner. Then sums, differences, and products
the matrices Aî, A^2< : Sii. 5i2. - ^n, C12
may be formed using
CHAP.
MATRICES
1]
SOLVED PROBLEMS
2-10
(a)
1.
2I
3
2_
2-2 3-1
'l
-1
O'
3-412
2-5
12
2.
"l
-1
o"
-2
-5
-3
Then D
= 4,
[4
[J
4
(c)
[1
3]
-6
-7
8
5
[
{2
.
Let
<4
-1
'2
-2
2
-3 -2
such that A + B -
2-p
-9
-1-s
"0
T)
0.
0"
-2-p
9-u_
4-r
and p = -2,
.0
A +
9.
[17]
2(5)
2(6)-
3(5)
3(6)
-1(4)
-1(5)
10
12
12
15
18
.-4 -5 -6
-1(6).
6"
-11 -8
1
(-6) + 2 (-7) + 3(8)
r2(l) + 3(2)+4(3)1
Ll(l) + 5(2) + 6(3)J
'
2J
1(9) + 2(10) + 3(-ll)
1 (6)
+ 2(7) + 3(-8)]
poT
L29J
ri(3) + 2(l) + l(-2)
l(-4) + 2(5) + 1(2)]
[4(3) + 0(1) + 2 (-2)
4(-4) + 0(5) + 2(2)J
8]
[s -I2J
1"
.1
A""
3-5
-2
1-3
2+1
4-r
9-
3(4)
10
"gl
_2
2J
0-2
-4 -4]
C
l]
(e)
2 + (-l)
+3
0"
-1
1(4) + 2(0) + 3 (5)
[19 4
(rf)
2(4)
=
e]
[4
(b)
D =
[4(2) +5(3) + 6(-l)]
6]
6,
3. (a)
find
-2
r
2-2- ?1
4-5-s
6+3- u.
3+1-r
_5+4-
and
4-202
5 2 4
5
4-741
0+2
1+3
2+0
-1 -2
1
-4
-2 -1
-2
5 -1 -2
and 6
A-\-B-D
-1 +
2+4 -1-1
2-0
0-5
-5+2
1-3
1-3
4-1
2-2
-15
+2
12
1-3-p
If
0+5
-5 + (-2)
6-3
-5
-5
2+(-4)
4+1
2-2 3-1
ri + 3
=
-1
If
15
2.
-4
-5
(d)
(b)
(c)
Then
1.
-1
The reader
\2
-1
1
will
r
2
1-
show
that
"5
-3
_3
-1
r
4
and
2.
A.a'^ and ^^./l^ = /4^.^^.
-3 f
-1
-1
1
.1
1'
2
1.
11
=
-8
8-18
.8-43
MATRICES
5.
Show
2
2:
k=i
a,-,-,
j=i 1=1
22
a--
1=17=1
(c)
that:
2
(b)
[CHAP.
aife(
2 (2
b^hCj^j)
^
h=t
h=i
k=i
aif,bkh)chj
^
2
(")
J,
"il^î +
ife(*fei +'^fej)
+ '^t2(*2i+'^2i)
'^lj)
2
(6)
(ilî+''i2*2i)
(il<^17+'î2'^2i)
2 2
a--
t=i j=i
(a^j^+ a-^ + a -p)
(an + 0^2 +
(Oil + O21)
+ (a^t + 092 +
a^g)
<223)
i=i
''
a-
"
i=l
+ (12 + "22)
2
+ (ai3 + "23)
=22
2
a +
a.
= l ^2
i = i 13
o--
j = ii = i
This is simply the statement that in summing all of the elements of a matrix, one may sum
elements of each row or the elements of each column.
2
(c)
^_j
222
^-^
(oil
=
a-,(bi, c
tfe^
fei
ij
6,
fee
&,
fe3
2JI
,^
Prove: It A = [a^-]
= AB + AC.
The elements
62J+
of the
j'th
fenj
+ 012(621^17 + b2QC2J + 623C3J)
(2
a{fefefe2)c2f
-^
fe=i
fe=i
ai<^b^^)c3j
**
-^
/.l/li^^^'fe'^^'^'^r
is of order
c'2j
oifefefei)cii
O
37'
+ ai2*2i)cij + (0^1612 + aiQh.22)cQj + (aiifeis + ai2b23)cgj
fell
k=l
feijj+cij,
the
a., (
b,,c,-)
tfe^^^j^ kh hy
0-il(^ll'^lj + ftl2C2J+ ÎsCsj)
6.
first
row
mxn
of
and
if
are a^^
+%j- Then
S =
[fe^-]
^
a^^,
...
and C =
[c--] are of order
nxp, then
%i)
C)
+C
are
a^^ and the elements of the /th column of S
the element standing in the ith row and /th column of
oil(feij + Cij) + afe(fe2j + C2j) +... +o.i(feni +
/}(B +
'
^^^"ikHj+^kj)
AB and ^C.
S^îifefefej +^2âifeC;^j,
A(B + C)
is
sum
Of
the
the elements standing in the ith row and /th column of
7.
Prove: If i = [aij] is of order

then ^(6C) = (iS)C.
The elements
p
mxn,
S =
[6^-] is of order rexp,
and
if
C =
[c.-A is of order
throw
and the elements ofthe/th column
of /4 areai^.a^g, ...,a-
BC
of
h=
b^j^c^j-,
hence the element standing

P
P
n
^^^^j_"ik''kh)<^hj
This
is the
...
and then
(/),
Using
(/")
and then
(e),
*n/i<;/ij
j'th
(A+B)(C+D)
(A+B)(C+D)
+ (Xâif^bf^2)<:QJ
+ (^^"ikbkp)<^pj
row and /th column of (AB)C; hence, A(BC)
in
+ (A
+B)D
(A
A(C +D) + B(C +D) = AC + AD ^ BC + BD

= AC +BC +A.D +BD.
(AB)C.
AC + AD
AC +BC +AD +BD.
two ways that (A +B)(C +D) =
+B)C
P
n
^^"ik<-j^,^hh<^hj^
(2^^0ifefefei)cij
Assuming A, B ,C,D conformable, show

(e)
J^
n
=
element standing in the
Using
+ tn
6ih c^
i
row and /th column of A (BC) is
in the ith
ait^^b^chj +ai2^^b2h%j
=
are
P
b^f^Cf^-
pxo
-^
P
of the
8.
if
BC
BD.
'iJ'
CHAP.
MATRICES
1]
o'
"l
l'
9. (a)
10
0'
10
'1
10
1
o"
"4
2"
[0]
2
:]
:o]
[0
j2
10
10
2]
10 3 10
[3
10
oil
10
o'
'l
2_
'
(b)
1
1
_3
'1
10
"l
[0]
[0:
[0]
3]
[0
"][o
2
3
12
10
18
0'
"1
0^0
-4- i
3|4
6_
4 15
8.
_0
4'
(c)
1
+i
-^-
Hiyi+
%i
Let
%2
[l]-[6 5 4]
'7
11'
"13"
10
13
16
19
"35
24
37
26
39"
20
33
22
13
13
.13
13
[6
10.
[l]-[8 7]
31
7"
[1]-[1]
11
13
13
16
19
41"
31
28
30
20
33
22
10
35
24
37
26
39 41
28 30
13
.13.
13
13
13
13
13
13
4]
[l]
"1272
11^1
= aîy-i + 0^272 be three linear forms in y^ and yj ^-nd let
%3
os
yi +
% 2 72
new coordinates
linear transformation of the coordinates (yi, j^) into
0]^2^2
be a
.J1
(z^, zg).
O21 Zi +
The
h-2
result of applying
the transformation to the given forms is the set of forms
(on ii2 +
1) ^1
"*"
(^2
("31^11 + ^32^21) Zl
("3
1^12 + "32 ^22)
%i
(o-Li
%<2
((^2 1 C^l 1
X3
6^j +
"*"
aj^2
621) Zi
^2 2 ^2
^12
*i
Using matrix notation, we have the three forms
"*"
'^12^22) ^2
^2 2 ^22) ^2
%1
012
"21
"22
r-
Z2
-.
Vl
and the transformation
Vr,
^3
611
612
O21
022
The
'^3
result of applying the transformation is the set of three forms
"11
'xi
%2
"2
"^2
Os
^^3
pri ii2
^2
Thus, when a set of m linear forms

formation of the variables with matrix B
in n variables
,
622
-1
/,
^',
with matrix A is subjected to a linear transm linear forms with matrix C = AB
there results a set of
'
MATRICES
[CHAP.
SUPPLEMENTARY PROBLEMS
-3'
"l
Given A =
-1
.1
1_
-1
-2-4
Compute:
12.
Verify:
(d)
Find the matrix
-1
-1
-2
0-S
(A+B)-C.
D such
-3
Given A =
-51
t-2j
-4
-2
-10
A + {B~C)
(c)
3_
5-3
4j
-2
J -2
A-C
2I
-3
and C =
3_
-2 -4
(6)
-f
-1
_3
2I
_2
Compute: A + B =
(a)
A+D^B.
that
and B
AB
compute
=B-A
Verify that
= Q
-(A-B).
and
BA
-1
-2
-1
-11
-22
12
-11
that
AB
Hence, ABi^BA
generally.
\ -3
13.
Given A =
14
-3
-3 -1
-1
15.
Using the matrices of Problem
Given
Show
[
20.
BA
use the results of
Ans.
19.
AB
show
(b)
A, -I,
I,
AC
= 0,
(a) to
2
i
show
+ B^
AC
AC
or
more
A(BC).
and
(A+B)C
AC +BC.
-2 -3
= C.
ACB
as n
-2 -4
-13
(AB)C
that
- B^ ^ (A-B)(A+B).
and A^
and
show
CS^,
A'^
B"^ =
(A
-B)(A +
B),
= -1, derive a formula for the positive integral
any two
AC. Thus, AB
ij
AB
CA
that
A(B + C)
2AB
-3 -5
= A,
show
-A according
that the product of
that
-13
where
show
3 -4"!
0-2
[2
-13
-3 -4
^["12
and C
(ABf ^ A^
(a)
that
11,
-3 -5
-14
Given A =
18.
-1
16. Explain why, in general,
17.
1-1-2
-2 -1 -1
2-5-1
2
3
-1
Given A =
and C
does not necessarily imply B
14.
10
111
1-212
=4p,
4p-l-l,
ip +
2,
4p-^3, where
(A
Bf
powers of A
/=
[0
ij-
-'"'[;;][-:;] [rM'i
= A'^ + b'
-"]
[0 -]- ["0 3-
is a matrix of the set.
[? 0]
Given the matrices A of order mx, B of order nxp, and C of order rx^, under what conditions
on p, 9,
and r would the matrices be conformable for finding the products and what is the order of each" (a)
ABC
(b)ACB, (c)A(B + C)?
Ans.
(a)
p =
r;
mx-q
(b) r = n = g;
x p
(c)
= n, p = q;
x g
CHAP.
21.
MATRICES
1]
Compute AB, given:
10
A
(a)
10
11
0^
and
10
Ans.
and
Ans.
1
--1
12
1
10
and
I
'2
'
(a) trace (A
10
Ans
1
22. Prove:
10
(c)
o"
(b)
"
+ B) = trace A + trace B,
(6) trace (kA) = k trace
A.
h
=
V,
2n+r.-3y.
r -zi +
Y^l
y,\
^;^^^l^
[2
-3
[2
-2
1
-1
-3
722"]
[-221 - 622J'
24. If
-4
and B = [fe^J are of order
= [aiji
25. Let
= [a^-] and
/4
[fej,-;,]
where
(i
x n
1,
and
if
m;
= [c^j] is of order nx. p,
1,
p;
A;
show
1, 2, ...
,n).
that
(/4+B)C
Denote by
/S-
=^C
the
+ BC.
sum
of
Pi
182
the elements of the /th row of B, that is, let fij =
j)^.
Show
that the element in the ith row of
Pi
sum
is the
of the
Problems 12 and
26.
(II)
check the products formed
to
Determinative
Either a is in the relation to 6 or a is not in the relation to
Reflexive
a is in the relation to a, for all
(iii)
Symmetric
If
Transitive
If a is in
is called an
Show
relations.
in
congruency) between mathematical entities possessing the following properties:
(iv)
Show
Use this procedure
13.
A relation (such as parallelism,

(i
27.
elements lying in the sth row of AB.
b.
a.
a is In the relation to b then b is in the relation to a.
the relation to b and b is in the relation to c then a is in the relation to
c.
equivalence relation.
that the parallelism of lines,
Show
similarity of triangles,
and equality of matrices are equivalence
that perpendicularity of lines is not an equivalence relation.
that conform ability for addition of matrices is an equivalence relation while con form ability for multi-
plication is not.
28. Prove:
If .4, B,
are matrices such that
AC
CA and BC
CB,
then
(AB
BA)C
C(AB
BA).
chapter 2
Some Types
THE IDENTITY MATRIX. A

lar;
of Matrices
square matrix A whose elements a^: =
a square matrix A whose elements

"ii
"12
"13
tip o
tZo o
a^=
for
upper triangu-
Thus
upper triangular and
is
lower triangular.
is
both upper and lower triangular,
is call-
a,
...
It
is called
lower triangular.
is
which
Ogj
ed a diagonal matrix.
i>j
for
is called
"2n
03S
The matrix D
i<j
will frequently be written as
diag(0]î, 022, 033,
'
^nn )
See Problem
If
in the diagonal matrix
D above,
Oi^ = Ogs =
k,
and
is
ci
in addition, k = l, the matrix is called the identity matrix
is
called a scalar matrix;
denoted by
/.
1.
if.
For example
'1
t"]
When
/+/ +
...
and
1
1
the order is evident or immaterial, an identity matrix will be denoted by /. Clearly,

to p terms = p / = diag (p,p,p
= I-I ... to p factors = /. Identity map) and
[1.
A.L
hAh
A, as the reader
may
readily show.
10
3I
(5
then l2-A =
CHAP.
2]
SPECIAL SQUARE MATRICES.
A and B
If
are square matrices such that
called commutative or are said to commute.

matrix,
it
commutes with
11
is a
It
AB
= BA, then A and B are
simple matter to show that
if
is
any ra-square
itself and also with
See Problem
A and B are such
If
matrix
AB
that
-BA,
i =
matrix
1,
A and 6
are said to anti-commute.
,fe+i
for
which A
where
A,
A:
is a positive integer, is called periodic.
If
k is
See Problems
3-4.
k-t-i
the least positive integer

e for which
If
th^ matrices
2.
then A is said to be of period
A,
k.
so that A^ = A, then A is callled idempotent
for
aP
4=0,
which
where p
least positive integer for which A'^ =
0,
a positive integer, is called nilpotent.
is
ther
is said to
If
p is the
be nilpotent of index
See Problems
5-6.
THE INVERSE OF A MATRIX.
If A and B are square matrices such that AB = BA = I, then B is called the inverse of A and we write B = A'^ (B equals A inverse). The matrix B also has A
as its
inverse and we may write A = B~^
Example
1.
Since
-1
12
-1
-2 -3
o'
/,
each matrix in the product is the inverse of
( 1
the other.
We
shall find later (Chapter?) that not every square matrix has an inverse.
A has an inverse then that inverse is unique.
We can show
here, however, that if
See Problem
7.
If A and B are square matrices of the same order with inverses A^^ and B~^ respectively,
then (AB)''^ = B'^-A~^, that is,
1. The inverse of the product of
verse order of these inverses.
matrices, having inverses, is the product in re-
tijvo
See Problem
A
utory.
matrix
An
A such
that A^ = I is called involutory.
An
identity matrix, for example, isinvol-
own inve r$e.
involutory matrix is its
See Problem
THE TRANSPOSE OF A MATRIX. The

columns
of
an mxn matrix A
is
matrix of order
and /th column of A stands

If
In
^[-e]
in the /th
IS
i' =
nxm obtained by interchanging the rows and

A and is denoted by A' (A transpose). For
Problems 10 and
II.
11,
The transpose
(A'Y =
we
4'
6.
Note that the element
aIr
in the ith
A and
and
B,
a scalar, we have immediately
and
if
(b)
(kAy = kA'
A:
is
prove:
of the
row
row and ith column of A\
A' and ZTare transposes respectively qf

(a)
9.
called the trahspose of

1
example, the transpose of A
8.
sum
of two matrices is the
(A +
BY
A'+ S'
sum
of their transposes, i.e.
12
[CHAP. 2
and
in.
The transpose
transposes,
of the product of
two matrices
product in reverse order of their
is the
i.e.,
(AB)'
B'-A'
See Problems 10-12.
SYMMETRIC MATRICES. A
A =
matrix
is
a^j]
/.
Thus, a square
For example,
4-5
3-5 6
Problem
IV.
symmetric and so also is kA
is
In
square matrix A such that A'= A is called symmetric.

symmetric provided o^-- = a,-,^ for all values of i and
k.
we prove
13,
If
any scalar
for
is
an ra-square matrix, then A + A'
symmetric.
is
A square matrix A such that A'= -A is called skew-symmetric. Thus, a square matrix A is
skew-symmetric provided a^.- = -Uj^ for all values of i and /. Clearly, the diagonal elements are
"
-2
For example,
zeroes.
3"
skew-symmetric and so also
is
is
kA
for
any scalar
k.
-3 -4
With only minor changes in Problem 13, we can prove
V.
If
is
any re-square matrix, then A -
A' is
skew-symmetric.
From Theorems IV and V follows

VI. Every square matrix A can be written as the sum of a symmetric matrix B = ^(A+A')
and a skew-symmetric matrix C = ^(A-A').
See Problems 14-15.
THE CONJUGATE OF A MATRIX.
Let a and b be real numbers and let i = V-1 then, z = a+ hi is

The complex numbers a+bi and a-bi are called conjugates, each
the other. If z = a+ bi, its conjugate is denoted by z = a+ hi.
;
called a complex number.
being the conjugate of

z^
If
= a + bi
and
Z2
z^
a~
Z2 =
then
bi,
z-^
a-
bi
= a+
bi,
that is, the conjugate of
the conjugate of a complex number z is z itself.
z.i=a+bi and
If
(i)
z-L
Z2
= c+
+ Zg = (a+c) + (b+d)i
sum
that is, the conjugate of the
(ii)
z^- Zg
then
di,
and
of
= (ac-bd) + (ad+bc)i
z^
= (a+c) - (b+d)i =
z^
two complex numbers
z^^
and
is the
sum
(a-bi) + (c-di)
of their conjugates.
= (ac-bd)- (ad+bc)i = (a-bi)(c-di) = F^-i^,
that is, the conjugate of the product of two complex
numbers
is the product of their conjugates.
When i is a matrix having complex numbers as elements, the matrix obtained from A by replacing each element by its conjugate is called the conjugate of /4 and is denoted by A (A conjugate).
l
2i
then
3
If
A and B
(i)
-21
and
(ii)
(.4)
above,
2-3i
are the conjugates of the matrices

(c)
Using
Example 2. When A
= A
we may prove
and
3i
A and B and itk

(c?)
is
any scalar, we have readily
(X4) = ~k-A
CHAP.
2]
13
_VII. The conjugate of the sum of two matrices is the sum of their conjugates,
(A + B) = A + B.
Vin. The conjugat e of the product of two matrices

(AB) = A-B.
is the product, in the
same
i.e.,
order, of
their conjugates, i.e.,
The transpose
We have
IX.
A, i.e.,
Example
3.
of
is
denoted by A^(A conjugate transpose).
Tê transpose
(Ay = (A').
Prom Example
1
(AY
of the conjugate of
- 2i
3
2+
square matrix
Hermitian provided a^j = uj^ for
Hermitian matrix are real numbers.
1
square matrix
^ =
2i
and
[0^^]
[l-2i
3
2
such that
i and /.
A*.
= A
1
=
-i
A'
(A')
Zi
values of
all
1-i
+j
(AY
+ 3iJ
is called Herraitian.
Thus,
/I
Clearly, the diagonal elements of an
is Hermitian.
-I
if
+
i
is
kA Hermitian
sometimes written as
equal to the conjugate of the transpose of
A'
3i
HERMITIAN MATRICES. A
Hermitian provided
while
-i
Is
is
is
Example 4. The matrix A
It
k is any real number? any complex number?
^ =
such that
[0^^]
a^^ = -o^--
for all
1= -A
values of
is
called skew-Hermitian.
and
Thus, ^
is
skew-
Clearly, the diagonal elements of a

skew-Hermitian matrix are either zeroes or pure imaginaries.
Examples. The matrix A
l-i
-1-t
3i
-2
is
/.
skew-Hermitian.
Is
kA skew-Hermitian
it
k is any real
number? any complex number? any pure imaginary?
By making
X.
If
minor changes in Problem 13,
is
we may
an ra-square matrix then 4 +
Z is
prove
Hermitian and
A-T is
skew-Hermitian.
From Theorem X follows

XI. Every square matrix A with complex elements can be
written as the sum of an
Hermitian matrix B={(A + A') and a skew-Hermitian matrix
C=i(A-A').
DIRECT SUM.
Let A^, A^
As be square matrices of respective orders m^,
ization
^1
...
A^
...
diag(î,^2
...
of the diagonal matrix is called the direct
A,
sum
of the Ai
As)
,ms.
The general-
14
Examples. Let
/4i=[2], ^5
B:}
and An
-1
-2
[CHAP. 2
12
The
sum
direct
Problem 9(6), Chapter
A^,A^,Aq
of
diag(^3^,
is
/42, /4g)
12-1
2
1-2
illustrates
1,
XII. If A = Aia,g {A^,A^,

same order for (s = 1, 2
the
...,Ag)
s),
S = diag(Si, S2
and
AB
then
^s).
= diag(îSj^, A^B^
where Ai and Behave
^s^s)-
SOLVED PROBLEMS
Or,
1.
Since
ii
"12
2i
"22
''ml
"7)12
'm
2n
"tn
11^1
OooOoo
22 21
''WB "Ml
"^22 ''2n
the product
AB
of
^mm"nn
''wm"m2
an m-square diagonal matrix A = diag(oii, a^^

an) and any mxn matrix B is obtained by multiplying the first row of B by %i, the second row of B by a^^- and so on.
2.
Show
This follows from
Show
-13
that
4.
Show
that
"^l
-2 -4'
-2 -3_
'
P^+*'^
-2
-4
'"^
-2 -4"
-13
.
-2 -3_
AB
A and BA =
B,
(AB)A
A-A
ABA
that if
.45.4 =
show
commute
for all
values of
'^l
["
o, 6, c, J.
*1
is idempotent.
-13
^1
-2 -3_
\"
and
-2 -4
2
3.
that the matrices
is idempotent.
A^ and
then i and
= .4(34) =
-2
-3
AB
are idempotent.
=
then 4^ =
.4
and 4
is idempotent.
Use BAB
to
CHAP.
5.
2]
Show
that
is
Let
8.
A,B,C
Prove:
By
-1
-3
3"
-1 -3
A^ =
0,
a''
2,
...
^"
that
AOAf=
= 0.
Then A{IAf
(AB)(ABf
)AB
Supposed
is involutory if
11.
is involutory; then
(A+B)' =
A' +
A{lnA)
(AB)' =
B^-I-B
(AB)^
A-A
A(B-B''^)A~^
and only
A^ =
= [6^].
-2 -1 -3
AnA^
= A.
iCA)B =
CUB)
so that B
B-'.B
=
B-'.A-'
(IA)(I+ A) =
if
and
/I
0.
is involutory.
and (I-A)(I+A) =
We need
S' and (A+B)' are respectively a^,
Prove:
any positive integer.
for n
=
I-A^
I-I
= 0.
B'.
A = [ay] and B
Let
Now
/.
(AB)'^ is unique; hence,
7,
matrix
Suppose (I-A)(l+A) = I-A^ = 0; then A^ =
A'.
b\a''-A)B
AB(B~'^-A~'-)
Prove:
B~^-A'^
{ABf{AB)
definition
By Problem
10.
-1 -1 -3
be square^ matrices such that AB = 1 and CA = 1. Then

^ is
the unique inverse of ^. (What is S~^?)
(AB)^ =
5 = C= ^
and
Prove:
A^.A
-1 -3
-1
show
and
-2 -1 -3
(B' -A
9.
is nilpotent of order 3.
nilpotent of index
Thus,
C.
Since /? =
7.
5
-2
6. If
15
bjj_.
only check that the element in the ith row and /th column of
and aj^+
bj^.
S'i'.
Let A = [ay] be of order mxn,
[6y] be
element standing in the ith row and /th column of
of order
AB
is
cy
nxp
=
then
AB
[cy]
is of order
mxp.
The
Jâife. b^j and this is also the element stand-
ing in the /th row and ith column of (AB)'.
The elements of the
/th row of S'are
Hn- Then the element
"i2
in the /th
iy, b^j
Thus,
12.
Prove:
(AB)' =
bnj
and the elements of the ith column of ^'are a^^,
row and ith column of
B'/l'is
B'A'.
(ABC)' = CB'A'.
mite
ABC
(AB)C.
Then by Problem
,
1 1
(ABC)' = \(AB)C\' = C'(AB)' =
C'B'A'.
SOME TYPES OP MATRICES
2g
13.
Show
A =
that if
[o^j]
B =
is n-square, then
[bij] =
[CHAP.
+ A' is symmetric.
First Proof.
The element
oj^j]
hence,
in the ith
row and /th column of
The element
bij = aij + aî.

bji
in the /th
a^^ + aij.
Thus,
and the corresponding element of
is aij
.4
row and ith column of A is
and B
bij = bji
is
aî
/I'
hence,
is aji;
and the corresponding element of
.4'
is
symmetric.
Second Proof.
By Problem
14. Prove:
If
10,
A and B
(A+Ay
A' +
(A')'
A'+A
A + A' and
are re-square symmetric matrices then
(^ +.4') is symmetric.
AB
symmetric
is
and only
if
if
A and B
commute.
Suppose A and B commute so that
Suppose
is
Then (AB)' =
BA.
Now
symmetric so that (AB)' = AB.
B'A' =
BA
(AB)' = B'A' =
AB and AB
BA
AB
hence,
the m-square matrix A is symmetric (skew-symmetric) and

symmetric (skew-symmetric).
15. Prove:
If
is
If
.4
is
symmetric then (see Problem 12) B'
If
,4
is
skew-symmetric then B' = (P'AP)'
-P'AP
If
A and B are
16. Prove:
commute
for
Akl
if
P'AP
and B
BA and
the ma-
AB
=
= B.4
A and 5 commute
if
and only
and
AB - k(A+B)
+ kÎ
BA - k(A+B)
+ kÎ
(B-kr)(A-kl)
(B
and B kl commute.
(A-kI)(B-kI)
BA. and A and B commute.
is
mxn then B =
symmetric.
and B is skew-symmetric.
Suppose 4 -A;/ and B -kl commute; then
is of order
k.
{A-kI)(B-kI)
Thus,
(P'AP)' = P'A'(P')' = P'A'P
re-square matrices then
every scalar
Suppose A and B commute; then
AB
symmetric.
is
A and B commute.
trices
P'AP
AB
AB
AB - k(A+B)
+ kÎ
BA - k(A+B)
+ kÎ
-kr\(A-kI)
if
A~
kl and
B-
kl
CHAP.
21
17
17.
Show
that the product of
two upper (lower) triangular matrices
BA
18. Derive a rule for forming the product
19.
See Problem
Show
that the scalar matrix with diagonal element
where the order

20.
If .4 is
show
that
23. (a)
If
1.
-1
Show
that
that
a).
Wand that kA
klA =Aia.g(k,k
are idempotent.
that
(a)
o"
Show
that
29.
Show
that each of
2/1
0.
-4^
but
0.
9/ /
0.
-1
/.
-1 -1
is periodic, of period 2.
is nilpotent.
and
-12
-1 -6
2
commute.
_-l -1 -4_
-1/3'
2/3
and
-3/5
2/5
1/5
7/15 -1/5
1/15
commute.
anti-commute and (A+
Bf
The only matrices which commute with every n-square matrix
all matrices
(a) diag(a,fe, c)
A^ +
B'^
anti-commutes with the others.
31. (a) Find all matrices which
Ans.
= 0.
0-3
28.
Find
BA
"112'
30. Prove:
5! =
- 94
-1 -1
-1 ~1 -1
(b)
- 2A
-4
'12
that
4.A
-3 -4
Show
.4
AB
idempotent and that
-3 -4
-13
that
is
-2 -6
-3
Show
I-A
(b)
diag(aii,a22
and q are positive integers.
_3 _5
[-1
Show
that
show
that
27.
-4J
show
-1 -1 -1"
26.
-13
B
and
-3
show
is idempotent,
(b) If
25.
Using A and B, show that the converse of Problem 4 does not hold.
(b)
24.
can be written as
A:
A^ where
a'^ a'^ = a'^
-14
If
S and ^
matrix
row order of A
2-3-5
A
that
upper (lower) triangular.
1.
of / is the
re-square,
Show
21. (a)
22.
mxn
of an
Hint.
is
commute with diag(l, 2, 3).

which commute with diag(aii,a22
where a,6,
are arbitrary.
a^^).
are the n-square scalar matrices.
k) A,
32.
[CHAP. 2
18
Show
(a)
that
"3-2-1
-4
is the inverse of
-2 -4 -5
10
2 10
4 2 10
(b)
"
34.
1.
-2100
is the inverse of
_ 8
-1 -1
to find the inverse of
[3:][:Ê3
1.
^"'-
4]-
[3
that the inverse of a diagonal matrix A, all of
Show
10
0-2
-2311
33. Set
-1
-1/2]
[3/2
whose diagonal elements
diagonal matrix whose diagonal elements are the inverses of those of
A and
are different from zero, is a
in the
same
order.
Thus, the
inverse of / is /
-1
35.
Show
that
4-3
3-3
3"|
-1 0-1
-4 -44-3J
and
4
4
are involutory.
10
36. Let
10
I2
38. Prove:
(ABCy
39. Prove:
(a) (A-^)'^ =
/s
(c) (^^)' = (/l')^ for p
(b) (kA)" = kA',
C'^B-Â-'^.
A,
(a)(A)=A,
41. Prove:
a positive integer.
ABC=iAB)C.
Hint. Write
(c) (A'^Y^ =
(b) (kA)-^ = -j-A'^,
(a)
2- 3
-1 +
(A'^y forp a positive
integer.
iB is Hermitian,
that
{AB) = A B.
is
Hermitian,
is
skew-Hermitian,
-1
(<f )
+
2i
(c)
show
(d)
-i
1
i
-2-3J
is n-square,
+ 3
(c)(kA)=kA,
is Hermitian
A + B
(b)
B)
(b) (A +
42. Show:
that every real symmetric matrix is Hermitian.
Show
If
that
0-1
(a)(A')'=A,
43.
Show
'2
/I21
37. Prove:
40.
/j.
by partitioning.
-1
and B is skew-Hermitian.
(a)
AA' and A' A
are symmetric,
(6) A-irA',
Hermitian and A is any conformable matrix then (A)'
44. Prove:
If
45. Prove:
Every Hermitian matrix A can be written as B + iC where B
is
HA
AA', and A'A are Hermitian.
is Hermitian.
is real
and symmetric and C is real and
skew-symmetric.
46. Prove:
and C
Show
Every skew-Hermitian matrix A can be written as A = B + iC where B is real and skew-symmetric

(6) A' A is real if and only if B and C anti-commute.
is real and symmetric.
47. Prove:
48.
(a)
If
A and B commute so also do ^"^ and B'
that for
m and
n positive integers, ^4
A' and
B\
and S"^ commute
if
and A' and B"
A and 5 commute.
CHAP.
2]
19
A
A
49.
Show
raA
(6)
If
51. Prove: If
is
symmetric
is
nX
A"
skew-symmetric then AA'= A'A
or
2n('-l)A
n-i
,ra
(a)
50. Prove:
nA
symmetric so also is
+6/1^
a-4
+.-.+/
and
where
are symmetric.
a, 6
are scalars and p is a positive
integer.
52. Prove:
Every square matrix A can
53. Prove:
If
54.
is real
if
is
Hermitian and C is skew-He rmitian.
complex and skew-Hermitian then iA are Hermitian.
is
that the theorem of Problem 52 can be stated:

Every square matrix A can be written as A =B+iC where B and C are Hermitian.
If^
are such that AB = A and

^ = B = / if ^ has an inverse.
A and B
If
idempotent,
(c)
show
is involutory.
Y=
(A
(a)
that k(.I+A) and
(A')
(A)^ = A-\
(b)
BA
kO-A)
n-square matrix A has an inverse A
57. If the
58.
and skew-symmetric or
B +C where B
/I
Show
55. Prove:
56.
written as
bfe
matrices which commute with (a) diag(i.
all
I.
(a) B'A'= A'
and
j(I+A)
A'B"= B\
^(I-A) =
(b) A"
and B' sue
0.
show:
Find
Ans. (a) dia.g(A. b.c).
then
(r)-^=(^-V
(c)
Prom the transpose
AA~^
are idempotent and
Hint, (a)
of
obtain
as the inverse of
(A'''-)
(6)
i, 2, 3),
diag(l.
/!'
1, 2, 2).
where A and B are 2-square matrices with arbitrary elements and
(b) dia.g(A.B)
b. c
are scalars.
A^.A^
59. If
with
A^
are scalar matrices of respective orders
diag(î, ^2
Ans. dia.g(B^.B^
AB
60. If
= 0,
mj^.m^
m^.
find all matrices
which commute
'^s)
B^) where Si, S2
are of respective orders m-^.m^,
-85
m^ with arbitrary elements.
where A and B are non-zero n-square matrices, then A and B are called divisors of zero.
A and B of Problem 21 are divisors of zero.
Show
that the matrices
61. If
...,
s),
diae(Ai.A2
show
A^)
(a)
A+B
(b)
AB
diag(îBi. /I2S2
(c)
trace
AB
and
di&giB^.B^
B^)
where ^^ and B^ are of the same order,
diag(î+Si,^2 + S2
=
62. Prove: If
^ and B
63. Prove: If
-^s
trace /liB^ + trace /I2S2 +
+ trace A^B^.
...
are n-square skew-symmetric matrices then
is n-square
and B = rA+sI, where
and
AB
is
symmetric
are scalars, then
if
and only
if
A and S commute.
A and B commute.
65.
Show
66. If
1, 2,
+ Bs)
Let A and B he n-square matrices and let ri, rg, si, S2 be scalars such that
ri4+siB, C2 = rÂ+SQB commute if and only if A and B commute.
/I
A^B^)
64.
(6)
(J
that
risj
7^
rssi.
Prove that
Ci =
that the n-square matrix A will not have an inverse when (a) A has a row (column) of zero elements or
has two identical rows (columns) or (c) ^ has arow(column)whichis the sum of two other rows(columns).
A and B
are n-square matrices and
A has an inverse, show that
(A+B)A~'^(A-B)
(A-B)A''^(A+B)
chapter 3
Determinant of a Square Matrix
PERMUTATIONS. Consider
together
the 3! = 6 permutations of the integers 1,2,3 taken
123
(3.1)
and eight of the
4!
132
312
231
213
321
= 24 permutations of the integers 1,2,3,4 taken together
4123
4213
3124
3214
2134
2314
1234
1324
(3.2)
we say that there is an

a given permutation a larger integer precedes a smaller one,
the permutation is
(odd),
even
is
inversion. If in a given permutation the number of inversions
there is no inversince
even
called even (odd). For example, in (3.1) the permutation 123 is
permutation
the
312 is even since in
sion, the permutation 132 is odd since in it 3 precedes 2,
If in
it
2,
and 3 precedes 2. In (3.2) the permutation 4213

4 precedes 1, 4 precedes 3, and 2 precedes 1.
precedes
is
even since in
it
4 precedes
Consider the re-square matrix
THE DETERMINANT OF A SQUARE MATRIX.
'11
"12 "IS
Uo 1 Oo p ^03
(3.3)
-'ni
"ne "n3
and a product
(3.4)
^j, "'^L
~^3
1 ~-'2 3Jo
'^"
"^n
comes from any row and one

of n of its elements, selected so that one and only one element
of convenience, the factors
matter
as
a
column.
In
from
any
(3.4),
comes
one
element
and only
re;
the
have been arranged so that the sequence of first subscripts is the natural order 1,2
of
the
intepermutations
re!
of
the
one
some
then
is
subscripts
second
sequence j^, j^
/ of
(Facility will be gained if the reader will parallel the work of this section bere.
gers 1,2
ginning with a product arranged so that the sequence of second subscripts is in natural order.)
For a given permutation /i,/2, ...,^ of the second subscripts, define ^j^j^....j^ = +1 or -1
according as the permutation is even or odd and form the signed product
(3.5)
-W-2-
By
k'^k ^^h
'^Jn
signed prodthe determinant of A, denoted by U|, is meant the sum of all the different
Ul, which can be formed from the elements of A; thus.
ucts of the form (3.5), called terms of

(3.6)
le Jlj2-
where the summation extends over
p=n\
The determinant
'^h
permutations hk---Jn of the integers 1,2,,
of a square matrix of order
20
re
is called
a determinant of order
re.
CHAP.
DETERMINANT OP A SQUARE MATRIX
3]
DETERMINANTS OF ORDER TWO AND THREE. Prom

-"ll
"12
^21
"22
(3.7)
12
^11^22
(3.6)
we have
21 '^12^^21
"'"
21
for n
11^22
and n = 3,
CE-j^2^21
and
"13
"12
'll
(3.8)
21
"31
"?
P*^
"32
"S3
^123
ail0223S
+
231
flu 022 %3
^132 Oll02s'^32
%2 02331
312 Ol3'%lC'32
^11023032
^213 Oi221<%3
^321
%3 ^^2 %!
012^1033
%l(22 033 - 02S32) " "12(021 "ss " ^S^Sl) + ÎS (a2lOS2 - 022^31)
022
^3
"^32
^3
O21
053
Example
"31
21
<^2
'^Sl
^^32
Oi
<^33
1.
21
-11
(a)
1-4
2-0
(fc)
2-3
(-1)3
-2
+ 3
3
2
5
10
(O
+
2
11
2(0-0
2
(d)
11
01
11
-1-1) - 3(1-0 -1-2) + 5(1-1-0-2)
2(-l)
- 3(-2) +
5(1)
-3 -4
0-2
2{0(-6)-(-2)(-5)! - (-3){l(-6)- (-2)0! + (-4) {l(-5) -
O-o!
-5 -6
-20
18
20
-18
See Problem
PROPERTIES OF DETERMINANTS.
nant
Ul
is
Throughout this section, A
is the
1.
square matrix whose determi-
given by (3.6).
Suppose that every element of the sth row (every element of the/th column) is zero.
Since
every term of (3.6) contains one element from this row (column), every
term in the sum is zero
and we have
I.
If
every element of a row (column) of a square matrix A
is zero, then U| =0.

Consider the transpose A' of A. It can be seen readily that every term of
(3.6) can be obtained from A' by choosing properly the factors in order from the
first, second, ... columns. Thus,
II. If 4 is a square matrix then
U'l = \A\; that is, for every theorem concerning the rows
of a determinant there is a corresponding theorem concerning
the columns and vice versa.
Denote by B the matrix obtained by multiplying each of the elements of

the ith row of A by
Since each term in the expansion of |5| contains one and only one
element from its
row, that is, one and only one element having A; as a factor,
a scalar k.
fth
\B\
1
P
Thus,
\: ,
,
a. , n^n
\
*%'2---i'iii2J2----anj^!
.
Jc\A\
DETERMINANT OF A SQUARE MATRIX
22
[CHAP.
every element of a row (column) of a determinant U| is multiplied by a scalar k, the

determinant is multiplied by k; if every element of a row (column) of a determinant \A\ has k as
If
III.
may be factored from
a factor then k
a7']^]^
ltd AQ
O\
OrQ
/l'
t*
CI
\A\
For example,
QO
f (X'
Let S denote the matrix obtained from A by interchanging its ith and (i+l)st rows. Each
\A\ is a product of |s|, and vice versa; hence, except possibly for signs,
product in (3.6) of
expansion of
(3.6) is the
term of
\b\,
\b\.
sign changed is a term of
its
IV.
If
In
counting the inversions in subscripts of any term of (3.6) as a
before i+1 in the row subscripts is an inversion; thus, each product of (3.6) with
|s|
and
= -
\e\
\A\.
Hence,
obtained from A by interchanging any two adjacent rows (columns), then \b\ =
is
- \a\.
As a consequence
V.
VI.
|s|
If
is
If
of
Theorem
we have
is
obtained from A by carrying
its ith
its
rows (columns), then
= -\A\.
\b\
row (column) over p rows (columns), then
= (-i)ÛI.
VII.
If
two rows (columns) of A are identical, then
Suppose that each element

(j
IV,
obtained from A by interchanging any two of
= 1,2,
...,n).
row of A
of the first
is
\A\
expressed as a binomial
%,
^li
b-^:
c-^j,
Then
in
^Jd2
^^îi ^
"î^ "2 J23 J5
p^kk^ Jn ^^k'^k'^k-h,.
ii2
^13
^m
21
O22
"23
02n
""in
Os,
Os-,-
"njn^
"Jn
'In
Oni
n2
"na
'^nn
-^n2
'-"ns
In general,
VIII.
If
every element of the ith row (column) of A is the sum of p terms, then
\A\
can
be expressed as the sum of p determinants. The elements in the ith rows (columns) of these
p determinants are respectively the first, second, ..., pth terms of the sums and all other rows
(columns) are those of A.
The most useful theorem

IX.
If
is obtained from
is
A by adding
to the
elements of
its ith
tiple of the corresponding elements of another row (column), then
Cfrii
T"/CCEog
^22
Clnr
Clf,',
{-
Karin.
Clf=n
Clr^r,
Cfg ^
row (column), a scalar mul= \a\. For example.
\b\
H'Qri^
^Q'2
"^
i^Clr
Zgg +
ka^^
See Problems 2-7.
FIRST MINORS AND COFACTORS. Let A

by
(3.6).
When from A
be the re-square matrix (3.3) whose determinant \A\ is given
the elements of its ith row and /th column are removed, the determinant
of the remaining (re- l)-square matrix is called a first minor of
or of \a\
and denoted by
\M^j\.
CHAP.
3]
More frequently,
cofactor of
Example
2.
a^-
If
The signed
called the minor of Oy.

denoted by a;
it is
and
is
23
minor,
is
(-if'*'^ \Mij\
called the
"V
11
12
''13
"21
'^SS
"23
"'31
^^32
''SS
"23
22
M 11
"21 "23
ki2
21
22
"31
"32
kic
"32 "33
"31
"33
and
1+1
1 + 21
kll
(-1)
1^11
(-1)
ai2
1^12!
14-31
I
ki3
(-1)
Then
\m.
(3.8) is
Ml
In
Problem
X.
The value
9,
aiilA/iil
"11*11
O12IM12I
ai20ii2
aislMigl
"130^13
we prove
of the determinant U|, where A is the matrix of (3.3), is the sum of the products obtained by multiplying each element of a row (column) of
U| by its cofactor, i.e..
n
(3.9)
Ul
aiiflii
(3.10)
Ml
"îfîf +
a^g a^g
în * in
a^jd^j
^ aik'kk
n
Using Theorem
XI.
ra-square
Example
VII,
021*2/ +
(ij, = 1,2
fe?i"fei*j
n)
we can prove
The sum of the products formed by multiplying the elements of a row (column)
matrix A by the corresponding cofactors of another row (column) of A is
zero.
3.
If
is the matrix of
Example
and
2,
of
an
we have
"310^31
"320^32
"330^33
"12*12
"22*22
"32*32
"31*21
"32*22
"SS*23
"12*13
"22*23
"32*33
Ul
'4
I
while
and
See Problems 10-11.
MINORS AND ALGEBRAIC COMPLEMENTS. Consider the matrix (3.3). Let

i^
arranged in
h, i^
order of magnitude, be m, (1 < m < ra), of the row indices
1,2
n and let j^, j^
/^ arranged in order of magnitude, be m of the column indices. Let the
remaining row and cofumn indi,
ces, arranged in order of magnitude, be respectively
i^,^,,
i^
and /+;+2
a separation of the row and column indices determines uniquely

two matrices
a.-
.Jl.j2.-
(3.11)
im
'l'J'2
'î.
i2.ii
''2.
i2 :h
"^m Ji
%J2
72
Jm
^n
^n
Such
24
[CHAP.
and
''m+i'J'm+i
Jn
+ i'Jm+2'
.7ra
''m
+ S'Jm +
+ i'Jffl+s
n.7n
^m+2>7m+2
^ra+2>Jn
^n'j7ii+2
^ri'Jn
'-m
(3.12)
^^Jm-i-i
called sub-matrices of ^.
A and the
of each of these sub-matrices is called a minor of
The determinant
Jm
J1.J2.
J'm-'-i'Jn^-Q
Jn
^m+l'
''r
pair of minors
and
are called
complementary minors
Examples. For the
I
i
1,3
'^2,5
each being the complement of the other.
of A,
5 -square matrix
[iij.
'^21
'^23
f^Sl
63
and
'^l, 3.4-
'
I
I
"I'Z
%4
'^15
"32
^^34
'^SS
"42
"44
^^45
2,4,5
'-m-^2
are a pair of complementary minors.
Let
U +
(3.13)
In
+ in +
and
(3.14)
i-n
Ji, is.
The signed minor
fm+2 +
+1 +
(-1)
A,-
+ *n +
7(4-1.
and
/I.
(-l)'^
7n
is called
''m,+
,2 4 5
A.{3\i
of
Jn
^n
'to+2
the algebraic complement of

J1.72
Jm
H>h
^n
13
2+-5 + 14-31
complement
l>^m+2> ''-
Exainple4. For the minors of Examples.

Of
+ /m+2 + " + In
is called the algebraic
Jm-'-i'Jn-i-2
J+i'Jm,+2
J,
^ra
and
(-1)
l(-3 + 4-^2l-446
(-1)
.2,4,5
A-^ 34
Ap' g
=
I
-I
,1.3
/i^
I
2 4 5
îjg',4
is
the algebraic complement
is
the algebraic complement of
1.3
^îs
Note that the sign given to the two complementary minors
is
the same.
Is this
always true?
Ji
.Ji
When m =
l,
(3.11)
becomes
J2.J3.
[%ii]
and
"HJi
an element of A.
The
Jn
complementary minor
is
Ml
in the notation of the section
above, and the
algebraic complement is the cofactor
OL-Lj.
minor of A, whose diagonal elements are also diagonal elements of A,
is
called a principal
The complement of a principal minor of A is also a principal minor

braic complement of a principal minor is its complement.
minor of A.
of A; the alge-
CHAP.
DETERMINANT OP A SQUARE MATRIX
3]
Examples. For the 5-square matrix A
25
H'
"?2
"11
"15
31
S3
and
^1.3
are a pair of complementary principal minors of
A,
,t
"24
(^25
0^2
044
045
52
"54
055
A What is the algebraic complement of each ?

The terms minor, complementary minor, algebraic complement, and principal minor as defined above for a square matrix A will also be used without change in connection
with U
.
See Problems 12-13.
SOLVED PROBLEMS
1- (")
!l
4|
1-1
2-4
- 3(-l)
4
(b)
15
21
5l
(1)
(l)(4-7
71
5-6)
-2-4
+ 2(3-6
4-5)
-6
(c)
11
1(4-21
-18
11
id)
2.
5-1)
first
column the corresponding elements
-4
-4
-4
-4
-4
-4
-4
-4
-4
of the other
columns,
I.
Adding the second column

using Theorem
4.
1(3-3
Adding to the elements of the
by Theorem
3.
to the third,
Vn
removing the common factor from this third column and
a+b+ c
+a
a+h
a+b+c
a+b+c
(a + b + c)
Adding to the third row the first and second rows, then removing the common factor
2; subtracting
the second row from the third; subtracting the third row from the first;
subtracting 'the first row
from the second; finally carrying the third row over the other rows
26
Oj^+fej
a^+b^
ag+63
62+ ^2
^3+^3
i^+Ci
62 + ^2
&3+^3
Cj+a^
C2 + a2
C3+a3
a5^+5]^+Cj^
a2+fe2+"^2
"s+^s+'^s
6^ +
Without expanding, show that
02
Ul
Subtract the second row from the
l-2
'^3
Oi
02
^2
Og
Og
-(Oj^
Oi + a2
02
ar,
(a^
02
flg
Cl
C2
Cg
bi
^2
63
a-,
Oo 03
C]^
C2
Cg
'^s) ('^s
%)
a2
02)
O2
ag
by Theorem
III
Og
Og
oj^-oj is a factor of U|.
and
a^^
then
first;
^2
fl-j^
^3
- 02) (02 ~
b^+Cr
62
^3+^3
62+^2
C-L
b^+c^
b-i
63
62
a^
5.
bi + c^
6^+C-L
bi
[CHAP.
Og
Similarly,
a^-a^ and ag-a^
are factors.
Now M|
is of order three in the
letters; hence.
\a\
(i)
A:
k(a^^-a^)(a^-as)(as-a^)
The product of the diagonal elements. 0^02. is a term of \a\ and, from (i), the term is -kaâ^. Thus,
-l and \a\ = -{a^-a^){a^-a2){as-a^). Note that U vanishes if and only if two of the a^, og. os are
|
equal.
6.
Ul
7.
is
Since A
is
Prove:
If
skew-symmetric and of odd order 2pskew-symmetric, -4'= -A; then

Ut = -U! and Ul = 0.
\-A
- Ul- But, by Theorem
(-if^-^-UI
Ul; hence,
is
Hermitian, then \A\ is a real number.
Since A
is
Hermitian. A = A'
Prove:
If
then
Now Ul
8.
then Ml =
1,
Ul
and
(-ir'\l
12
,2+1
(-ly
3+
(-ly
II.
%;
Ul
|eii;2---Jnîii"2i?--^"in
12
2,
U|
îij2...j"yi%
is a real
ai2 =
a22 = (-1)
age
But
if
'^
+ *'
"
"
^'
number.
(-1) 1+2
U| by Theorem
U'| =
?1
\a\ =
Ul
requires 6 = 0; hence.
For the matrix A
a,,
|i
2+2I
(-1) 3+2I
2I
-2,
-1,
Mi3 =
1+3
(-ly
a,23
(-1)
a 33
(-1)
2+3
11
3+3I
2I
31
II,
CHAP.
3]
Note that the signs given
to the minors of the
elements
in
forming the cofactors follow the pattern
where each sign occupies the same position

cupies in ^,
9.
27
in the display as the element,

Write the display of signs for a 5-square matrix.
whose cofactor
is required, oc-
Prove: The value of the determinant U| of an re-square matrix A is the sum of the
products obtained
by multiplying each element of a row (column) of A by its cofactor.
We
Now
e.
shall prove this for a row.
kh---Jn~
^jbjb- -in
The terms
î"''^ i"
of (3.6) having a^^ as a factor are
^ P"'^tin
lJi..fe.
-in.
the
Then
is in natural order.
(a)
may
be written as
(6)
where the summation extends over the cr= (n-i)! permutations of the integers
2,3
022
2S
n,
and hence, as
<2n
"an
(c)
"ii
"11
'twill
"ns
^n2
Consider the matrix B obtained from A by moving its sth column over the first s-l columns.
By Theorem
\B\ = (-1)
U|. Moreover, the element standing in the first row and first column of B is
a^s and the
minor of a^^ in B is precisely the minor \M^\ of
a^s in A. By the argument leading to (c), the terms of
ais mis\ are all the terms of \b\ having a^s as a factor and, thus, all the
terms of (-1)^"^
having a.^
VI.
a factor.
Then
U|
the terms of
ais!(-if
are all the terms of \a\ having a^s as a factor
M^/isli
as
Thus
(3.15)
+
"11*11
since
(-1)
,s+i
= (-1)
m.r,JWlTr\T
B be the
ments. Let
umn over
the first
We have
+ ai5!(-lf||/,3|! +
+ ai2i2
(3.9) with
shall call (3.15) the expansion of
U| along
its first
"' ''^ '^ ^^^^^ '^' ^^-^^ '' ' = '^ '^ '''î"^'J by f^P^^ting the above arguJ'*' îT^
matrix
obtained from A by moving its rth row over the first r-1
rows and then its .th col^
s-l columns. Then
T~l
(-1)
The element standing in the

^ Thus,
the minor of a.rs in
are all the terms of
first
row and the
s-l
(-!)
first
for
having a^^ as a factor.
r.
\a
column of 5
(-1)
is
a and
the terms of
u
the minor of a^^ in
^^
Then
,l,-rkU-l)
and we have (3.9)
+ ai !(-!)'*" M,
+ ain*in
We
...
r+fel
M.rk\
2
k=i
a^jâ.
rk^rk
is yreciseiy
i
precisely
10.
[CHAP.
28
When
A =
is the cofactor of aij in the ra-square matrix
oLij
'll
(i)
fC^GC^-j
"T
fCiÔL^j
"T
[a^j]
show
that
%,j4-i
^'i.n
(h,j-i
02
*2n
"W,j-i
<%,
%,f-i
"12
"'1
i +
itj^dj^j
'î
^2
'^n
In making these
with k^
0^7 with Atj,
This relation follows from (3.10) by replacing a^j with k^, 027
^2J
appearing is affected since none contains an element
(t^j
replacements none of the cofactors OLij.QL^j.
,
from the
th
column of A
By Theorem
when A^= a^^,
VII, the determinant in {i) is
and VII, the determinant in
(i) is
when
+ fea^g,
Ay
\7
Write the eauality similar to (0 obtained from (3.9)
by
= 1,2
(r
and
= 1,2
and
when the elements
By Theorems Vin,
j).
/).
of the ith row of
are replaced
/ci,/c2.
11. Evaluate:
(a)
04
-5
3 4
-2
-4
(c)
28
(e)
25
38
42 38
65
56 47
83
3-4
5-6 3
2-3
4
4 8
-2 1 5
(b)
((f)
-3 2 4
(a)
(r
Expanding along the second column (see Theorem X)

1
-5
Cl-^^fX^Q
022^22 ^ ^32 32
"f"
ai2 +
a22 +
(-5)o;g2
1
1 2
34.2I
-5(-l)-'
-10
5(4-6)
3 4l
(b) Subtracting
twice the second column from the third (see Theorem IX)
4 8
-2 15
-3 2 4
8-2-4
5-2-1
4-2-2
3 4
-2 5
-4
-3
second row from the
3(-l)''
1 3
-2 +
2(1)
-4
5-11
-2 -4
5-3(3)
=
+ 2(2)
-2 -3
-4+
27
8
-11 -41
-2 - 11
to the third
-2 -4
9
2(3)
-32
in (c)
-4 + 4(1)
2-2(1)
1
5-2(-ll) -11
3 + 4(-ll)
4-2(-2)
-2 -3 + 4(-2)
column from the second and then proceeding as
3-4
5-6 3
2-3
4
-3
and adding twice the second row
first
3-3(1) 4-3(2)
(-4 + 36)
(d) Subtracting the first
-42
-3(14)
(c) Subtracting three times the
-2
27
-41
-31
-11
CHAP.
(e)
3]
Factoring 14 from the
28
first
column, then using TheoremIX to reduce the elements in the remaining columns
38
25
42 38 65
14
47 83
56
-2
25
38
38
65
47 83
Show
-12
6
P + q
Now
13.
25-12(2) 38-20(2)
38-12(3) 65-20(3)
47-12(4) 83-20(4)
-1
-14
-1
-14(-l-54)
770
that p and q, given by (3.13) and (3.14), are either both even or both
odd.
Since each row (column) index
(-1)
14
14
4-13
12.
29
is
found in either p or 9 but never in both,
+ 2+-+7J) +
(1
(l
2+-"+n)
2-kn{n + l)
n(n +1)
even (either n or n + 1 is even); hence, p and q are either both even

and only one need be computed.
p+q
= (-1)^
is
For the matrix
[..]
12
11
12
13
14
10
15
16
17
22
18
19
23
24
.21
'4^'4^'5|
complement
of
Ao's
Thus,
is
20
25
13
,^2+3+2+4| .1,3,51
,
(-1)
Ml,4,5l
and the algebraic complement of
the algebraic
or both odd.
16
18
20
21
23
25
(see Problem 12)
is
I12 141
14.
Show
that the permutation 12534 of the integers
1, 2, 3. 4,
is even,
24135
is odd,
41532
is even,
odd, and 52314 is even.
15.
List the complete set of permutations of
16.
Let the elements of the diagonal of a 5-square matrix A be a.b.cd.e.

diagonal, upper triangular, or lower triangular then \a\ = abcde.
17.
Given
-4
[j J]
each product is 4.
18.
and B = [^
(a)
-1
2 4
-1
^^^^
AB^BA^ A'b
27
4 AB' ^
a'b'
Show, using
^ B'a'
(b)
2-2
12
2 3
2 3
=
(c)
-2
-3 -4
(3.6), that
when ^
is
but that the determinant of
1,
is
2,3,4, taken together; show that half are even and half are odd.
6J
Evaluate, as in Problem
2
^^^^
1,
53142
30
19.
(a)
Evaluate
\A
1 2
10
2 3
[CHAP.
4 5 11
(b)
Denote by B
Evaluate \B
to verify
(c)
Denote by
Show
that
2 7
by multiplying the elements of
Obtain from \A
2 3 4
4 5 8
4 5 3
.4
by interchanging
|o| =
the determinant
2 3
and third columns.
its first
by subtracting three times the elements of the
Evaluate
to verify
Theorem
/I
I
\A
I
multiply the first column by three and from
has been tripled.
an n-square matrix and 4
20.
If -4 is
21.
Prove: (a)
If
(b) If
22.
Compare with
then
= k.
\a\ = k =
(a)
Count the number of interchanges

and thus prove the theorem.
(b)
Same,
for
Theorem
it
subtract the third column.
not confuse (e) and
use (3.6) to show that
is a scalar,
is skew-Hermitian, then
Do
(e).
Evaluate
Evaluate to show that
(g).
U^
/t^M
|.
\a'\.
|
-4
is either real or is a pure imaginary
of adjacent
number.
rows (columns) necessary to obtain 6 from A
in
Theorem V
VI.
23.
Prove Theorem VII. Hint: Interchange the identical rows and use Theorem V.
24.
Prove:
If
first
IX.
subtract twice the first row from the second and four times the first row from the third.
the resulting determinant.
(g) In
Evaluate
5-1
column from the corresponding elements of the third column.

(/) In
5.
thus verifying Theorem VIII.
second column by
its
2 3
2 3 5
(e)
-4
III.
the determinant obtained from
(d)
Theorem
Theorem V.
to verify
the determinant obtained from
any two rows (columns) of a square matrix A are proportional, then
25.
Use Theorems
26.
Evaluate the determinants of Problem 18 as
VIII, III,
,4
o.
and VII to prove Theorem IX.
in
Problem
11.
a b
27.
Use
(3.6) to evaluate
\A\
d
e
".P/.
then check that \a\ =
Thus,
if
= diag(A-^. A^).
where
g h
A^,
A^ are 2-square matrices,
U4i|
2/3
-2/3 -2/3
1/3 -2/3
2/3
-2/3
-1/3
28.
Show
that the cofactor of each element of
1/3
-4
29.
Show
that the cofactor of an element of any row of
is that element.
-3 -3
is the
corresponding element of the same
numbered column.
30.
If
is
symmetric then
(b) If
is
n-square and skew-symmetric then
Prove: (a)
<Xij=
0^j^
when
aij = (-1)
"P when
Ot--^
CHAP.
31.
3]
For the matrix A of Problem

(a)
show
(b)
form the matrix
that
32.
12
(^^22
32
^23
Otgg
the result in (6) is
show
that
Without evaluating show that
known as soon as
a
a^
6^
ca
52
c^
c2
ab
be
ab
ca
ab
ca
be
ca
be
ab
a^
bed
6^
aed
e^
d^ d
a-^
a^ a
abd
c^
c'^
abc
d d^ d
(a
Show
that the ra-square determinant
a\
n-2
ra-i
n-2
a^
1 ... 1
0...1
...0
remove the common factor from each of
1...1
("-!)
re-i
known.
(a) is
respectively by a,b.c
34.
AC = [.
and show that
be
Multiply the columns of
the rows to
33.
why
_'^13
(c) explain
31
b)(a
c)(a
...
...
...
d)(b
...
...
d)(e
- d).
n-1
(-1)
c)(i
(n~l).
1
1
ar, 1
35.
Prove:
n-i
2)(i
as).
(ai-a^H(a2- 03X02-04)...
na^+b-^
na^+b^
na^+bg
nb-j^+e^
nb^ + c^
nb^+Cg
nc-i^+a.^
nCQ+a^
ncQ + a^
X
37. Without expanding,
show
that the equation
38.
a+ b
a
b
+6
(n
x+c
{na + b).
+ l)(rf--n +
1)
O]^
02
O3
61
62
is
Ci
Cg
Cg
xb
x-c
,n-i
Prove
x-i-a
x+b
-a^j
Without expanding, show that
+6
(02
ra-2
36.
S(i
has
as a root.
7i-i
- "ni
chapter 4
Evaluation of Determinants
determinants of orders two and three are found in Chapters. In

Problem 11 of that chapter, two uses of Theorem IX were illustrated: (a) to obtain an element 1
or 1 if the given determinant contains no such element, (b) to replace an element of a given
PROCEDURES FOR EVALUATING
determinant with
0.
For determinants of higher orders, the general procedure is to replace, by repeated use of
Theorem IX, Chapters, the given determinant U! by another |b| -= \bij\ having the property
that all elements, except one, in some row (column) are zero. If &>, is this non-zero element
'Pq
and ^p^
is its cofactor.
Then
bpq
is treated in similar
the minor of bp
i-ir
dp^
minor of
^Pq
b.
'pq
fashion and the process is continued until a determi-
nant of order two or three is obtained.
Example
1.
3-2 4
3-2 12
2 + 2(3)
-2
3 4
3-3(3)
2-3(-2)
-2
-2405
+ 2(-2)
8-1
=
(-1)"
+ 2(1)
-2
4 + 2(2)
2
+ 8(-l) -1
-6
-2
-6 +
8(8)
-2
-2 +
8(4)
'^
-(-l)- = (-l)P'
30
0-2
-2405
3-2
-6
4-3(2)
3-3(1)
8-10
0-10
+ 8(-l)
-2 + 8(8)
58
62
+ 8(4)
30
37
-286
37
See Problems 1-3
For determinants having elements of the type in Example 2 below, the following variation
divide the first row by one of its non-zero elements and proceed to obtain zero
may be used:
elements in a row
Example
0.921
or
column.
2.
0.185
0.476
0.614
0.517
0.667
0.201
0.201
0.517
0.667
0.123 -0.384
0.782
0.157
0.527
0.138
0.782
0.157
0.527
0.138
0.872
0.484
0.637
0.799
0.872
0.484
0.637
0.799
0.309
0.196
0.217
0.555
0.841
0.448
0.312
0.555
0.841
0.448
0.492
0.680
0.240
0.921
0.921
0.312
-0.320
0.123 -0.384
0.921
0.309
0.196
0.217
0.921(-0.384) 0.309
0.196
0.217
0.492
0.680
0.240
0.492
0.680
0.240
0.921(-0.384)
0.309
0.265
0.217
0.492
0.757
0.240
0.921(-0.384)(0.104)
32
-0.037
0.921(-0.384)
0.309
0.265
0.492
0.757
CHAP.
4]
THE LAPLACE EXPANSION.
The expansion
of a determinant
33
of order n along a row (column) Is
a special case of the Laplace expansion.
bered
I'l,
i^
J2
Instead of selecting one row of \a\, let

when arranged in order of magnitude, be selected. Prom these
n(n-
m + l)
m
l)...(ra
1-2
....
minors
can be formed by making all possible selections of m columns from the n columns.
minors and their algebraic complements, we have the Laplace expansion
>
|(-1)^
(4.1)
i.
Jrr.
Jn+i' jm+2
i.
Example
Using these
Jn
^m4-i'% + 2
where s = i^ + i^+... + i^ +A + /2 +
column indices taken m at a time.
m rows numm rows
^n
and the summation extends over the/? selections of the
+h
3.
Evaluate
3-2 4
2
3-2 12
using minors of the
two rows.
first
3 4
-2405
Prom
(4.1),
1+2+1+21
(-1)
1 + 2+1 + SI
U1.2
_l + 2 + l + 4|
-2
1 + 2 + 2 + 31
+
,,1+2 + 2+4. .2,4
.13
,
(-1)
Mi'jl-Us'.
-2
(-1)
-2
-2
_2
(-13)(15)
.2,31
^1,:
1 + 2+3+4
(-ly
2,41
,1,31
"Mi',|-U
1,2!
3,4!
1,4
(-1)
(-1)"
U-5aI
^3,4
_2
-2
- (8)(-6) + (-8)(-12) + (-1)(23) -
.34
*l,2r 1^3, 4I
-2
(14)(6)
-2 4
+ (-8)(16)
-286
See Problems 4-6
DETERMINANT OF A PRODUCT.
If
A and B
Us
(4.2)
are n-square matrices, then
Ul-lsl
See Problem 7
EXPANSION ALONG THE FIRST ROW AND COLUMN.
If
^ = [aid
n
(4.3)
Where
^11 ^1
i=2
a^
is the
cofactor of
o^ and
a^^-is the
n-square, then
is
ii
IJ
IJ
j,-=2
algebraic complement of the minor
"iiîil
of^.
ii Otj
I
DERIVATIVE OF A DETERMINANT. Let

tiable functions of a variable x.
the ra-square matrix
Then
A =
[a^,-]
^tj.
have as elements differen-
34
The
[CHAP. 4
\A\, of \A\ with respect to x is the sum of n determinants obtained

dx
by replacing in all possible ways the elements of one row (column) of
by their derivatives
with respect to x.
I.
derivative,
Example
4.
x^\
x^
1
dx
2x-\
2x-\
-2
x2 x +
2x
=
x""
x^
4x
-2
x+
3^^
-2
2x-i
x-"
6x
12x^
See Problem 8
SOLVED PROBLEMS
2
-2
-2 4
-3 10
-2
4-2(3) -3-2(-2) 10-2(4)
2(2)
3-2
-2
-2
1.
(See Example 1)
-286
3
There are, of course, many other ways of obtaining an element +1 or -1; for example, subtract the
column from the second, the fourth column from the second, the first row from the second, etc.
-2
2+2
2-2(1)
-2-2(2)
+2
1-2(2)
-3
+ 3 -3-2(3)
18-9
-1 +
4-3
1-
1+
2 +
Multiply the second row by
: i
and the third row by
(-1+3j)|-^|
2J;
-4 + 7i
1+i
25 -
+j
5j
-6
1+2;
5-J
-4 + 7j -10 +
2j
I
14i
then
5-J
,4
1+2j
+i
2
5
and
4-3
-11 -4
2j
(l+J)(l+2j)Ml
-9-3(-3)
8-3(4)
-4
2-3i
2i
-5
-72
-11 -4
\A\
-6-2(-3)
4-2(4)
1-3(4)
-4
4-6
4-3
4-3
3-2(4)
=
-9
-5
Evaluate
-6
3.
10
-1
first
18
+J
82J
14j
2J
25 -
5j
-4 + 7J -10 + 2i
CHAP.
4.
4]
Derive the Laplace expansion of
\A\
Now
into the first
row
(I'l- 1)
numbered
which the row and column indices are arranged
of 1^1 in
%\
in
interchanges of adjacent rows of |^

the row numbered ii can be brought
row. by i^- 2 interchanges of adjacent rows the tow numbered is can be brought into the second
-m interchanges of adjacent rows the row numbered % can be brought into the mth row. Thus.
i;
by
after
by
m<n
Jrn
A;
H-''^
order of magnitude.
of order n, using minors of order
\aij\
ji'h
Consider the m-square minor
35
+ (ig-
2)
i^
I'l,
+ (ijj -m)
11 + ^2+
occupy the position of the
+ 'm-2'"('"+l) interchanges of adjacent rows the rows

m rows. Similarly, after /i + j^ + + /^ - \m(m+l)
first
interchanges of adjacent columns, the columns numbered /i,/2

occupy the position of the first m col/jj
umns. As a result of the interchanges of adjacent rows and adjacent columns, the minor selected above occupies the upper left corner and its complement occupies the lower right comer of the determinant; moreover.
I
A has changed sign

[1
+ ^2+ +
in
= j- + i +
+
+ /i +
+ ii + + in changes. Thus
cr
Ji.J2'
/i
Jm
Ji~h-
'-m
'm-H'''m +2-
Jn
Jn-H' Jm+2'
-m)! terms of (-1) \a\
m!(ra
is
equivalent to
yields
w!(n-
or
' Jn
n+2'
"TOJ-l'
times which
+ /-"('" +1)
yields
(-ir
(a)
Jn
Jm+i'Jn-i-2
AH'''^
/2
m)! terms of \a\.
n(n~l)...(n-m + l)
different m- square
l'2....m
m\{n m)\
minors may be selected. Each of these minors when multiplied by its algebraic complement yields m!(/j-m)'
terms of U|. Since, by their formation, there are no duplicate terms of
U| among these products.
Let
I'l,
in be held fixed.
12
Prom these rows
S(-i/
I
,
5.
where
/i, 72
in of the column indices.
i^
+ i^+
12 3 4
12 1
11
3 4 12
(-1)^
Jn
different selections
using minors of the first two columns.
21
|1
ll
(-2)(1)
(-ir
21
ll
41
ll
(-If
11
13
41
ll
(5)(-l)
A,B, and C are re-square matrices, prove
C B
Prom the
7.
and the summation extends over the
+ in
(-3)(1)
6. li
Jm-1' Jvn
I'm
Evaluate
+ i^ +j^ + j^ +
Jn
rows of |P| only one non-zero n-square minor, U|, can be formed.
Hence, by the Laplace expansion, |p| = |.4|.|b|.
first n
plement
is
Prove
\AB\ =
|s|.
Suppose
Problem 6
= [a^j]
Its
algebraic com-
S
and
= [bij]
are n -square.
Let
= [c^j] =
AB
so that
^Hkhj-
i^rom
36
"m
2n
"nn
ail
012
021
022
"ni
n2
-1
-1
[CHAP. 4
6ii
612
..
hi
621
^22
..
?,2
-1
To
the (n+l)st column of \P\ add fen times the first column, 621 times the second column
the nth column;
On
012
"m
Cii
"21
"22
2ra
C21
"ni
"n2
"nn
Cm
-1
Next, to the
(n
+ 2)nd column of |P| add
times the nth column.
fei2
times the
Let A =
h,
622
..
62
column, 622 times the second column,
Hence,
\p\
in
Cii
C12
2n
C21
C22
"nn
Cm
Cn2
Its
(-i)Vlf" ^"^"lc|
^13
^m
*23
b^n
\P\
zero n-square minor, 1-/| = (-l)'^ can be formed.
8.
first
612
We have
Continuing this process, we obtain finally
= (-lf'2"^^'|c|.
-1
Oil
6i times
we have
Prom
the last n rows of
only one non-
algebraic complement is (_i)i+24--"+ra+(n+ii+-"+2n|(-|

=
and
\c\
\c\
\ab\ = U1.|b|.
%2 %S
021 '%2 'â
where
Oif
aij(x),
(i, j
= 1,2,3),
are differentiable functions of x.
031 032 033
%l"22*33 +
and, denoting -ra;:
dx
^^
by
a,',-,
^J
122Sa31
%s"S2"21
"
(îx"^"^ "
0:l202lOs3
"
(hs^Ô'Sl
Then
CHAP.
4]
'^ll"22''S3
dx
+ 22ll"33 + ''33"ll"22
+ %3032021
+ O32OISO2I
aÔQiOg^
Oil
0^11
o-L20î2
OisOi^g + a^-^dj^x
"*"
oil 012 Ol3

022
031
032 033
by Problem 10, Chapter
023
021
Oj^jCggOgg
11
O12 Ol3
21
O22
O23
31
O32
O33
Ogââ^^
022(^22
>i23'^n''S2
0210-12033
+ O31O12O23
+ "l2"23"31 + 023012131
021013032
37
airâ^â^.^
+ 023(^23
"^
Og-LCXgi
OgjOj^jOgg
02205^303^
Ogjô^gOjg
+ 032(^32 + 033(^33
"13
'11
"12
i2i
"22 "23
'31
"32 "33
3.
9.
Evaluate:
(o)
11
-2
-1
3
1
-4
-2
-1
156
-2000
113 4
1116
16
12 9
ic)
41
(d)
7
3
11.
5
3
-2 -2
show
Is n-square,
that \AÂ
is real
aJ"'
Let
"^
and B =
[^-02 ojj
4B|
Use |^B|
(6)
Let
''
1^-62
|.4|"|B|
i.4|-|B| to
show
+ t03
a2+iaA
-Og + ja^
Oi-iogJ
tOj^
and non-negative.
inii.
\ts\
Use \AB
to
first
two rows; also using minors from the
'^
0000
bjj
that
and
118
-4 -3 -2 -5
-2
2
2 -2
Evaluate the determinant of Problem 9(o) using minors from the

first two columns.
12. (o)
-304
{b)
10. If
-4
4 -3
-4 -5
(<
+ 02)(6i+62)
(oi
61 + 163
(0161-0262)
+ (0261+0163)
62 + 164
[-62 + 864
6i-j6gJ
222222,2,2
express (01 + 02+03 + 04X61+62+63+64) as a sum of four squares.
13.
Evaluate
using minors from the
first three
rows.
Ans. -720
38
112 12
14.
12
^1,^2
15. If
111
110
112
Evaluate
using minors from the
first
two columns.
.4ns. 2
11
square matrices, use the Laplace expansion to prove
^"^^
^^s
[CHAP. 4
|diag(4i,.42
As)\
Uil-U,!
....
a^ a^ Og a^
16.
17.
*1
*2 ^3 *4
a-^
a^
flg
a^
*1
^2
^3
*4
a^
a2
62
Expand
using minors of the
two rows and show that
60
Use the Laplace expansion

A
first
to
show
62
6..
A
B C
that the n-square determinant
\A\
umn
to
is fe-square, is
zero when
aiiOiii + ai2ai2 + OisOîs
expand each of the cofactors a^g.
+ ai4i4.
.-^
.^
ffi,-
is the
algebraic complement of the minor
"11
"12
"21
22
"ni n2
Use
92
ire
Pi
?i
92
"2ra
P2
Pi "11
'^12
"
"nra Pre
"712
Pn "ni
|.4|
If
"m
Pilj
.^
'J
t=i j=i
Cti
^V
"nn
Zx +
I
2*:
2a:
-1
a:
(c)
;<:
x-l
a;
a:
2a:
+5
3-2
Prove
?n
find the derivative.
x^
(b)
(a)
that the bordered determinant
(4 .3).
()
-4ns.
show
[a^,-],
In
For each of the determinants
2x
"ij
denotes the cofactor of a^j in the n-square matrix A =
li
Hint.
first col-
11 "IJ l^lj
îl^lj^lj
Of
"ii
19. If a^j
along its
ti
1=2 J=2
where
OLâ. tti*
show
^11^11
21.
where
> 2"-
18. In
20.
63
+ 9a:^-
8a;=^
A and B are
(6)
6a:
x^+l
+ 21a:^ +
x+l
12a;^
real n-square matrices with
15a:*,
(c) 6a:^
A non-singular and
5*"^
if
ff
x^
- 28x^ +
=
4 + iS
9a:^
20a;
is Hermitian, then
chapter 5
Equivalence
THE RANK OF A MATRIX. A
non-zero matrix A is said to have rank
minors is different from zero while every (r+l)-square minor,

said to have rank
Example
1.
if
r if
at least one of its r-square
any, is zero.
zero matrix is
0.
The rank
of
'l
3"
is
r=
-1^0
since
while
= 0.
See Problem
An
A
A is called non-singular if its rank r=n,

The matrix of Example 1 is singular.
re-square matrix
called singular.
is
Prom
I.
AB\
that is,
if
A ^
0.
1.
Otherwise,
A\-\B\ follows
The product
of
two
more non-singular re-square matrices
or
is non-singular;
the prod-
uct of two or more re-square matrices is singular if at least one of the matrices is singular.
ELEMENTARY TRANSFORMATIONS.
on a matrix do not change either
(1)
(2)
order or its rank:
The interchange
of the ith and /th rows, denoted
The interchange
of the ith
The multiplication
The
(3)
The following operations, called elementary transformations,

its
by Hij;
and /th columns, denoted by K^j
of every element of the ith row
by a non-zero scalar
k,
denoted by H^(k);
multiplication of every element of the ith column by a non-zero scalar k, denoted by Ki(k).
The addition
elements of the sth row of

denoted by Hij(k)
to the
of the /th row,
k,
a scalar, times the corresponding elements
The addition
to the elements of the ith column of

ments of the /th column, denoted by K^j(k)
k,
a scalar, times the corresponding ele-
The transformations H are called elementary row transfonnations;
the transformations
are
called elementary column transformations.
The elementary transformations, being precisely those performed on

determinant, need no elaboration.
order of a matrix.
In
Problem
It
2, it is
the rows (columns) of a

an elementary transformation cannot alter the
shown that an elementary transformation does not alter its
is clear that
rank.
THE INVERSE OF AN ELEMENTARY TRANSFORMATION. The
inverse of an elementary transformawhich undoes the effect of the elementary transformation; that is, after A
has been subjected to one of the elementary transformations and then the resulting matrix has
been subjected to the inverse of that elementary transformation, the final result is the matrix A.
tion is an operation
39
EQUIVALENCE
40
2.
Let A
Example
[CHAP.
4 5 6
7 8 9
2
"l
The
effect of the elementary row transformation
H2i(-2)
is to
produce B
3"
10
.7
8 9
The
effect of the elementary row transformation ff2i(+ 2) on B is to produce A again

Thus, ff2i(-2) and H2x(+2) are inverse elementary row transformations.
The inverse elementary transformations

-1
"ij
are
îj
(2')
Hi\k) = H^(i/k)
(3')
H--(k)
= HÂ-k)
K^jik)
= Kij(-k)
We have
II.
same
The inverse
of an elementary transformation is an elementary transformation of the
type.
EQUIVALENT MATRICES. Two
matrices A and B are called equivalent, A'^B,

from the other by a sequence of elementary transformations.
one can be obtained
if
Equivalent matrices have the same order and the same rank.
Examples. Applying
in turn the elementary transformations
W2i(-2), ^siCD. Ws2(-1).
-1
12-14
-1
5-3
-1
-2
-7
-7
-3
-3
-1 -2
Since all 3-square minors of B are zero while
I
I
t^
0,
12-1
5-3
'^
the rank of
is 2
hence,
^ is 2. This procedure of obtaining from A an eauivalent matrix B from which the

rank is evident by inspection is to be compared with that of computing the various minors of -4.
the rank of
See Problem
ROW EQUIVALENCE.
lone,
If
is said to
3.
a matrix A is reduced to B by the use of elementary row transformations abe row equivalent to A and conversely. The matrices A and B of Example 3
are row equivalent.
Any non-zero
(a)
matrix
of rank
r is
row equivalent to a canonical matrix C in which
one or more elements of each of the
first r
rows are non-zero while
all other
rows have
only zero elements.

(b) in the ith row, (i
this
(C) ]\
<
=1,2, ...,r), the first non-zero element

element stands be numbered
't
is 1;
let the
column
in
which
;'-
/2
<
(d) the only
<
j^.
non-zero element in the column numbered
the ith row.
j^, (i
=1,2
r),
is the
element
of
CHAP.
EQUIVALENCE
5]
To reduce A
(ii)
If
aij
171
a; J
to C,
7^
suppose
use
0,
//i(l/oi,iji
o^
^^7
but
/i is
(is)
If
(ii)
Use row transformations
Vi
0,
the
number of the
reduce
to
use
41
it
to
non-zero column of A.
first
when necessary.
1,
and proceed as in
ffij,
(i^).
of type (3) with appropriate multiples of the first row to obtain
zeroes elsewhere in the /^st column.

If
non-zero elements of the resulting matrix B occur only in the
wise, suppose
use ^2(1/^2^2) as in
in
(ii);
if
column
first
but
&2J2=
bqj^
B =
first row,
Other-
C.
which this does not occur. If &2j ^ 0,

use H^^ and proceed as in (ii). Then, as
in
0,
clear the /gnd column of all other non-zero elements.
(il),
If
number of the
72 is the
non-zero elements of the resulting matrix occur only in the
Otherwise, the procedure is repeated until
Example
4.
The sequence of row transformations

to A of Example 3 yields
-1
-2
-7
having the properties
first
two rows, we have C.
is reached.
ff2i(-2), ffgiCD
-1
^\j
-3
-3
2(l/5)
-1
'\^
17/5
1
'%^
-3/5
-3
applied
//i2(l). //ssC-S)
-3/5
(a)-(rf).
See Problem
THE NORMAL FORM OF A MATRIX. By means

r
>
can be reduced
to
of elementary transformations any matrix A of rank
one of the forms
(5.1)
/.
called its normal form.
\l%
zero matrix is its
"'"'
own normal
form.
Since both row and column transformations may be used here, the element
obtained in the section above can be moved into the
first
column can be cleared of other non-zero elements.

row can be brought into the second column, and so on.
first
For example, the sequence

/^32(-l), ^42(3)
ff2i(-2),
^31(1).
column.
of the first row
Then both the
Similarly, the element
Ksi(l),
^2i(-2),
applied to 4 of Example 3 yields
first
of the
row and
second
K23, K^{\/%),
X4i(-4).
the normal form.
See Problem
ELEMENTARY MATRICES.
4.
5.
The matrix which results when an elementary row (column) transforma-
tion is applied to the identity matrix /^ is called an elementary
row (column) matrix. Here, an

elementary matrix will be denoted by the symbol introduced to denote the elementary transformation which produces the matrix.
0'
1
Example
5.
Examples
of elementary matrices obtained from
/g
1
1_
"0
1
= K,
_0
'1
0'
1_
Ha(k)
0"
k_
0'
'l
K-sik).
H^g(k)
k
1_
K^^yk)
EQUIVALENCE
42
Every elementary matrix
The
To
A by an elementary
To
mxn
matrix
A can be produced by
matrix.
effect a given elementary row transformation on
to form the corresponding elementary matrix
Ijn
(Why?)
is non-singular.
effect of applying an elementary transformation to an
multiplying
to
[CHAP.
effect a given elementary
of order
mxn, apply the transformation

left by H.
and multiply A on the
column transformation on A, apply the transformation to

K and multiply A on the right by K.
/ to
form the corresponding elementary matrix

~1
Example
6.
When A
3~
4 5 6
_7
H^^-A
4/^13(2) =
3'
"l
_2
"7
9"
interchanges the
4 5 6
7 8 9_
_1
first
and third
3_
"723'
o"
10
8 9
3~
4 5 6
0_
4 5 6
'l
1
1
9_
1
rows of A
"O
_25
1_
adds to
t(
the
16 5 6
first
column of A two times
39J
the third column.
LET
AND B BE EQUIVALENT MATRICES. Let the elementary row and column matrices corresponding to the elementary row and column transformations which reduce /I to 5 be designated
as //i./Zg
where //^ is the first row transformation, //g is the second, ...;
^s< J^\,T^Q.
K^ is the first column transformation, K^ is the second
Then
'^t
(5.2)
IU-H,.A K-i-K^
//.
PAQ
K,
where
H^-H^
Ih
(5.3)
and
We have
III.
P and Q
Two
matrices A and B are equivalent
defined in (5.3) such that

"1
Example
7.
When
2
_1
-1
["100"! r
1
and only
if
there exist non-singular matrices
^3i(-l) //2i(-2) -^ ^2i(-2) -Ksid) .K4i(-2) -K^sd) .Ks(i)
2J
1-200
o"j
0-2
ij
[jî
if
= B.
2~|
5-23,
2
PAQ
~1
10
10
0~
2"
10
10
"1000"
10
10 1
10
"1000"
10
5
iJ
1
_0
1_
_0
1_
_0
1_
1-25-4
10
[:;=}
o"l
1
PAQ
[1
10
1
oj
Since any matrix
is
equivalent to
its
normal form, we have
IV. If ^ is an re-square non-singular matrix, there exist non -singular matrices

as defined in (5.3) such that PAQ = 1^
P and Q
See Problem
6.
CHAP.
EQUIVALENCE
5]
43
INVERSE OF A PRODUCT OF ELEMENTARY MATRICES.

P
as in (5.3). Since each
H^...H^-H^
and
Let
and
K-^-K^-.-Kt
has an inverse and since the inverse of a product is the product
in reverse order of the inverses of the factors
P~^
(5.4)
H;\hI\..H^^
Q''
and
Let A be an re-square non-singular matrix and

=
let
Kt...Kt-Kt-
P and Q defined above be such
that
PAQ
Then
A
(5.5)
P'\PAQ)Q^
P'^-k-Q^
P'^-Q'^
We have proved
V.
Every non-singular matrix can be expressed as a product
of
elementary matrices.
See Problem
From
7.
this follow
VI.
If
VII.
If
is non-singular, the
P and Q
rank of
Two mxn
(also of
In
Problem
8,
BA)
PAQ
are non-singular, the rank of
CANONICAL SETS UNDER EQUIVALENCE.

VIII.
AB
is that of B.
is that of A.
we prove
if
and only
if
they have the same rank.
A set of my.n matrices is called a canonical set under equivalence if every mx-n matrix is
equivalent to one and only one matrix of the set. Such a canonical set
is given by (5.1) as r
ranges over the values 1,2
m or 1,2. ...,re whichever is the smaller.
See Problem
RANK OF A PRODUCT.
matrices
Let A be an mxp matrix of rank
P and Q such
NQ
Let S be a
AB
pxre matrix
(5.6)
By Theorem
By Theorem
III there exist
non-singular
that
PAQ
Then 4 = P
r.
9.
P''
and consider the rank of

=
P~'NQ'^B
AB is that of NQ'^B. Now the rows of NQ~'b consist of the firstr

m-r rows of zeroes. Hence, the rank of AB cannot exceed r
the rank of A
Similarly, the rank of AB cannot exceed that of S. We
have proved
rows of
VI, the rank of
Q B and
IX.
The rank
of the product of two matrices cannot exceed

the rank of either factor.
suppose iS =
then from (5.6). NQ-'b = 0. This requires that the
first r rows of Q'^B
be zeroes while the remaining rows may
be arbitrary. Thus, the rank of Q-'b and, hence
the
rank of B cannot exceed p-r. We have proved
:
X. If the mxp matrix A

rank of B cannot exceed p-r.
is
of rank r
and
if
the pxn matrix
is
such that
AB
the
EQUIVALENCE
44
[CHAP.
SOLVED PROBLEMS
1.
(a)
The rank
-4
The rank
12
of
is 2
since
and there are no minors of order three.
-4
2 3
(b)
2 3]
sj
of
since
is 2
and
2 4 8
"0
(c)
The rank
Show
since
is 1
4 6
each of the nine 2-square minors
0,
is 0, but nov
6 9_
_0
every element
3"
of
^0.
is
that the elementary transformations do not alter the rank of a matrix.
We
shall consider only row transformations here and leave consideration of the column transformations
Let the rank of the mxn matrix ,4 be r so that every (r+l)-square minor of A, it any, is zero.
Let B be the matrix obtained from .4 by a row transformation. Denote by \R\ any (r+l)-square minor of A and
by Is] the (r+l)-squaie minor of B having the same position as \R\
as an exercise.
Let the row transformation be H^j

two of
in the
Its effect on |/?| is either (i) to leave

interchange one of its rows with a row not of \R\
its rows, or (lii) to
case
(ii),
= -\r\ =
\S\
case
in the
In the
case
(ii) to
(i),
\S\
interchange
= \r\
=0;
except possibly for sign, another (r+l)-square minor
\s\ is,
(iii),
unchanged,
it
of l^l and, hence, is 0.
Let the row transformation be Hi(k). Its effect on \R\ is either (1) to leave
A:.
Then, respectively, |S| = |/?| = o or |S| = ;i:|/?| = o.
unchanged
it
or (ii) to multi-
ply one of its rows by
Let the row transformation be Hij(k). Its effect on |/?| is either (i) to leave it unchanged, (ii) to increase
one of its rows by k times another of its rows, or (iii) to increase one of its rows by k times a row not of S|.
In the cases (i) and (ii), |S|=|ft| = 0; in the case (iii),
\s\ =
=
/?| + A: (another (r+i)-square minor of /I)
|
k-0
0.
Thus, an elementary row transformation cannot raise the rank of a matrix.

lower the rank
tor,
did, the inverse transformation
if it
would have to raise
On
it.
the other hand, it cannot

Hence, an elementary row
transformation does not alter the rank of a matrix.
For each of the matrices A obtain an equivalent matrix B and from
it,
by inspection, determine the
rank of A.
"1
(a)
3"
1
'-^-/
"1
-3 -3
_3
1_
-4 -8
_0
2_
The transformations used were

"1
(b)
2 3
0"
2 4 3 2
3
_6
i
+
2j
0"
-3
-8
-44 -11
8 7 5_
Note.
1+i
(c)
1
''V.
1_
_0
//2i(-2). ffsi(-3);
''\j
-4
-8
-3
H^(-l/3), HQ(-l/i); Hgî-l).
0'
0"
-8
-3
5_
-3
p -4 -11
5_
_1
3"
-""J
"1
'~^
r^
-4
The rank
is 3.
-8 3
S.
-3
The rank
is 3.
'~^
2j
l+i_
'1
3~
"^
2j
2J_
'-Vy
2J
B.
The rank
is 2.
The equivalent matrices 5 obtained here are not unique. In particular, since in (o) and (i)
only
row transformations were used, the reader may obtain others by using only column
transformations.
When the elements are rational numbers, there generally is no gain in mixing row and column
transformations.
CHAP.
4.
EQUIVALENCE
5]
(a)
(b)
A =
13
12 6
113
2 3 9
2 3 9
113
13
2 4
1
.1
5.
C row equivalent
Obtain the canonical matrix
-2
-2
-3
-1
Reduce each of
(a)
12
6_
GOOD
-2_
3"
-1
'^
10
4
13-2
'->-'
-2 3
l"
of the given matrices A.
13
-1
p -1
each
13-2
13-2
-2
6 4
113
--v^
to
5_
45
'10
10
10
^\y
2
1
-1
--O
pool
4_
10
1
10 0-1
10 2
7~
2_
the following to normal form.
-1
1
^\y
-1
-2
7 2
"l
fl
o"|
0-2
7 2 sj
o"
1-2
p-
Lo 2
o'
'l
10
1-2
11
7 3_
o'
'l
10
'V.
-7_
0_
Us o]
The elementary transformations
are:
^2i(-3). 3i{2); K2i(-2), K4i(l); Kgg; Hg^-^y. ^32(2). /f42(~5); /fsd/ll), ^43(7)
(6)
[0234"! P
23 5 4^-0
4 8 13 I2J
|_4
354
2
"1000"
'-^
8 13 12
2 8 13
13
13
'Xy
2 3 4
"1000'
^\j
10
p
4.
"1000"
<^
10
0_
0_
are:
^12: Ki(2); H3i(-2); KsiC-S), X3i(-5), K4i(-4); KsCi); /fs2(-3), K42(-4); ftjgC- 1)
6.
1
2 3-2
2-2 1 3
3
04 1
Reduce A
Since A is 3x4,
we
to
normal form
A'
and compute the matrices
shall work with the array
seven elements and each column transformation
is
and
(^^
Each row transformation
such that
is
P-i_AQ^
-3
-2 -3
1
1
12
2-2
10
-6
-5
-6
-5
3-2
1/3 -3
-1/6
1-57-210
0-1-11
-5
7/6
10
10
1
or
0-210
0-1-1
-2
0-1-1
1/3 -4/3 -1/3
10
1
-6 -5 7 -2
-6 -5 7 -3
0-1/6 -5/6
1
1
Pi
A^.
performed on a row of
performed on a column of seven elements.
10
P^
EQUIVALENCE
46
[CHAP.
1/3 -4/3 -1/3
1
1
Thus,
Pi
-1/6 -5/6
-2
-1
-1
7/6
10
and
PiAQ-i =
10
10
N.
7.
Express A =
.1
4_
as a product of elementary matrices.
H^-H^-A-K^-K^
//2i(-l).
=
^siC-D; ^2i(-3). 'f3i(-3) reduce A
Hl-H'^-K^-Kt
(5.5),
Prove:
If
other.
Two mxn
9.
that is,
[see (5.2)]
ij
[l
if
3"|
010
ij
[p
and only
if
ij
fl 3
O'
010
1_
_0
same rank, both are equivalent to the same matrix (5.1) and are equivalent
^ and B are equivalent, there exist non-singular matrices P and Q such that B
A and B have the same rank.
Conversely,
VII,
fl
0"j
010
A and B have
By Theorem
fl
0"1
110
_0
8.
ff3i(-l)-^2i(-l)-'4-X2i(-3)-?f3:i.(-3)
fl
Prom
to
the
if
to
each
PAQ.
canonical set for non-zero matrices of order 3 is
[!:][::]
[i:l-[:i:]
A
canonical set
tor
non-zero
Vh o]
10. If from a
3x4
[:=:]
square matrix
nm
matrices is
of order n and rank
is selected, the rank r^ of
is
r^,
a submatrix
equal to or greater than
The normal form of A has n-rj^ rows whose elements

whose elements are zeroes. Clearly
'A
from which follows
>
- n as required.
r^
&:]
B consisting
+ s -
of s rows (columns) of
n.
are zeroes and the normal form of
6 has s-rg rows
CHAP.
EQUIVALENCE
5]
47
2
4 5
11.
Find the rank of
(a)
2
(c)
(c) 4.
(b) 3,
Show by considering minors
13.
Show
14.
Find the canonical matrix row equivalent

Tl
A,
that
2-3"]^ri 0-7]
[o
-4j
[2 5
A. and
A'.
-2
10 11 12 13 14
4-16
15 16 17 18 19
have the same rank.
(b)
3 4
3 4
12
11/9
-1
-1
(b).(c)
[/g
/a
12
12
^V/
(e)
-2
o]
P^
(d)
(e)
j]
From
Prom
10
1
10-1
[I, 0],
(a)
Let A
1/9
(c)
-3
-3
10
10
12
10 0-12
10
normal form of each of the matrices of Problem
15. Write the
Ans.
10
2j
-1
1/9"
1
'X^
14.
P'
jl
form Z/^^. /Ô). //i3(-4) and check that each HA effects the corresponding row transformation.
form K^^. Ks(-l). K^Ô) and show that each AK effects the corresponding column transformation.
(c) Write the inverses H^l,
H^
{?,),
H^lc-i) of the elementary matrices of
(d) Write the inverses K^l. ifg^C-l).
K^lo)
Compute B
^12
ft,(3) -//isC-
4)
(a).
of the elementary matricesof (6)
"0
(e)
(d)
each of the following:
to
(6)
5-46
.4'
12
_4
(a)
that the canonical matrix C, row equivalent to a given matrix A, is uniquely determined by A.
(a)
16.
(d) 2
12.
(d)
-1 -3
7 4 6
(a) 2,
(b)
4 3 4
Ans.
12-23
Check
Check
0'
-4
that for each
"0
and C
H^^(-4:)-H^(3)-Hi
1/3
(/)
17. (a)
(b)
Show
that
BC
Show
Show
that
/?',-=
that if
/?
CB
H-.
H,H-H~^^l
KK~^ = I
that for each K.

1
4"|
0.
ij
K-(k) = H^(k).
is a product of
and K^,(/t)
= H^-(k)
elementary column matrices. R'is the product in reverse order of the same
elementary row matrices.
18.
Prove:
(a)
(b)
19.
UP
AB
AB
and
and
BA
BA
are non-singular if
are singular
if at
.4
and B are non-singularra -square matrices.
least one of the n-square matrices
A and B
is singular.
and Q are non-singular, show that A,PA,AQ, and PAQ have the same rank.
P and Q as products of elementary matrices.
Hint. Express
20.
Reduce B
13 6-1
14 5 1
15 4 3
to
normal form
/V
and compute the matrices
P^
and
Qr,
such that P^BQ^ = N
EQUIVALENCE
48
21. (a)
(6)
Show
Show
m+l
22.
that the
number of matrices
number
and n+1.
that the
12
13
Given A
2 5 6
10
in a canonical set of n-square matrices under equivalence is
of matrices in a canonical set of
of rank 2.
23.
The matrix A
24. If the
25.
mxn
of Problem 6 and the matrix
AB
matrices A and B are of rank
show
Let A and B be n-square matrices,
Show
rj
and
rg
respectively,
show
Find P and Q such that B - PAQ.
that the rank of
that
\AB\ = |^|
(a) If at least
A+B cannot exceed
By considering each
of the
|fi|
one is singular show that

\AB\ = \a\-\B\
\AB\ =
|/4|-|s|
(6) If
both are
that equivalence of matrices is an equivalence relation.
29. Prove:
Hint.
= 0.
A_
Let ^ be an arbitrary n-square matrix and S be an n-square elementary matrix.
28. Prove:
30.
of Problem 20 are equivalent.
non-singular, use (5.5) and Problem 25 to show that
27.
such that
7^
are arbitrary.
six different types of matrix S,
26.
n+l.
matrices under equivalence is the smaller of
abed
_e
and take
Q-'b
where a.b
mxn
Find a 4-square matrix S
Follow the proof of Theorem
Hint.
[CHAP.
The row equivalent canonical form
of a non-singular matrix
Not every matrix A can be reduced
to
is I
and conversely.
normal form by row transformations alone.
Exhibit a matrix which cannot be so reduced.
Show how
to effect
types (2) and
31. Prove: If
.4
on any matrix A the transformation H^: by using a succession of row transformations of
(3).
is an
mxn
matrix, (m ^n), of rank
theorem when the rank of A is < m.
m,
then AA' is a non-singular symmetric matrix.
State the
chapter 6
The Adjoint of a Square Matrix
THE ADJOINT.
Let A =
[a^ ]
be an n-square matrix and
adjoint
(6.1)
adj
oij-
be the cofactor of a-; then by definition
12
^22
'îre
'^sn
Note carefully that the cofactors of the elements of the ith row (column) of A are the elements
of the jth column (row) of adj A.
Example
i.
For the matrix A
11=
2
3
ai2 = -2.
6,
Cl-lS
-3.
fflsi
= 1,
0122 =
and
adj
a^g
-5,
6
ffai
= -5,
Otgg
4,
agg = -1
1-5
1
-2 -5
-3
= 3,
-1
See Problems
Using Theorems
and XI of Chapter
"In
^^2
0271
'^
diag(M),
Examples. For the matrix ^
/I
By
of
Example
U|.
ni
1,
adj
CE 7
A-L
1^1)
6
1 -5I
-2 -55 4
-3 33 -ij
/I
|,
There follow
I.
(6.4)
If
is re-square
and non-singular, then

adj
I
-7 and
U|
(adj A)
f-T
4
I
49
.
I
-7/
-7
[_
adj
0'
0-7
we have
taking determinants in (6.2),
(6.3)
1^1
(adj /I)
find
%1 %2
i(adj A)
(6.2)
we
3,
1-2.
50
If
II.
is of
If
A and B are
AB
adj
IV.
In
Problem
îi.i2
^l'''-2
Let
7m + i' 7m+2'
let
we
is of rank ra-1, then
Jn
%.
^2
adj
adj
-square minor of the
be
Jn
its
complement
re-square matrix
in A,
A =
the m-square minor of adj A

dei
denote
in adj
A as those
Jm
occupy
of
j\
If in (6.6)
S2
+ +
in A.
,?i.
M
When m =
2, (6.7)
Jn
Jm+i' Jm + 2
Jn
(-D^'i^l'
%+
/i
is non-singular,
(6.7)
+ +
72
Jt,
then
' Jm
J2<
(-1)
H, %,
uaH-i
I
i-
^M+1' 'm+2. . ^n
becomes
Jg>J4-
Jn
(_l)V*2+Jl-'j2m
(6.8)
H'J2
^2'J2
.J1.J2
\A\
When m = n-l,
(6.7)
Jl'j2
When m =
n, (6.7)
algebraic complement of
becomes
Jre-1
M,
(6.9)
tjj
whose elements occupy the
Jm + 2
Jvi + i'
Jm
Ji' J2
M:
s
[o,-,-].
and
Then
where
3-
prove
Ji' J2
(6.6)
See Problem
is of rank 1.
Ji'J2
same position
adj
we prove
6,
be
^m+i' ^ra+2'
let
4,
If
re-square matrices,
(6.5)
MINOR OF AN ADJOINT.
0.
Problem
In
(ad}A)A
rank < ra-l, then adj A =
THE ADJOINT OF A PRODUCT.

III.
and singular, then
is re-square
A(a,diA)
If
[CHAP.
becomes
(-1)
(6.4).
Ul
a.
CHAP.
6]
51
SOLVED PROBLEMS
1.
The
adjoint of
i4
a h
'^ll
IS
c
0^21
d -b
3 4
4 3
4 3l
3 4
-7
2.
The
A =
adjoint of
[;]
IS
-2
3.
Prove:
If
is of order n
we
First
-1
-1
and rank n-\, then adj A
rank
is of
1.
note that, since A is of rank n-\. there is at least one

non-zero cofactor and the rank of adj A
By Theorem X, Chapter 5, the rank of adj^ is at most n-{n-l) =
i.
Hence, the rank is
IS at least one.
exactly one.
4.
Prove:
adj
^S
= adjS
adj
^S-adjS-adj^
and
(adjB-adj^)4fi
we conclude
5.
Show
that
By
(6.2)
^Badj^S
^y(6.2)
Since
y4
.4
(S adj
S ) adj
/I
adjS {(adj^)^lfi =
that
adj
adj(adji) = U|
and
\ab\-I
-A,
if
^S
\a\
{a.aiAB)AB
^(l^l-Oadj/i
jslcâdj/^) =
adjS-U|./.B
U|{(adjB)Si
adj
adj
|b|.U|./
=
Ab\
\AB\-I
/I
0.
?^
(6.4),
adj
adj (adj
^)
diag(| adj ^|,
diagd^l
adj/l|
|adj^|)
Ul
1^-1
Ul
Then
adj (adj ^)
and
6.
Prove:
Ji'
Let
A;
adj (adj
^)
\aT' -A
adj (adj
^)
UT'^-^
^^ ^" m-square minor of the re-square matrix

I
Jm+i. Jw4-2
in
^m+i. ^m+2
^n
let
nH.i2
UT'^-^
in
;
'"n
^^^
in\
A =
[a:-]
tj-
be its complement in A, and
^*ê the m-square minor of adj

J1.J2
position in adjv4 as those of
/I
whose elements occupy the same
is
^H,i^, ...,i
occupy in A.
Then
52
u
where
s =
Ji
+ f2+ +
Ji.
J/2'
H'
^2-
+/1 + 72+
Jra + i' Jii
Jm
>
'
(-ifur
^m
[CHAP.
î
'
Jn
7m
Prom
a,-
a,-
a,-
a,,
"im.im
"^mJ's
Wl'A
''m+iJs
''m+iJm
'
"W2
"ht.ji
in.i
"^m^n
'^m+iJm+i
'
"
"^raJm+i
"traJn
tti,
*i
a,-
a,-
a,-
a,-
a,-
>
'00
,/2
ai
a,-
a,-
''m+i-Jn
"V.im4.i
Ul
^^2
h-j9
"^ra.ii
1-
h.Jmî.
\a\
Ul
''i2Jn
" "^-Jn
"^M'Jnî
"V+i'in
"%-i-i'Jn-î
^ri'^m+i
"in.in
by taking determinants of both sides, we have

Ji. Js'
(-ifUI- M %
where
7.
s is
Prove:
If
as defined in the theorem.
is a
H\
Jn
Jm+1' Jm-t2
4.
^2
Prom
this, the required form follows
skew-symmetric of order
^m+l' ''m+2
''w
2ra,
immediately.
then \A\ is the square of a polynomial in the elements
of i.
By
its definition.
above this polynomial
The theorem
2.
By
we
a\ is a polynomial in its elements;
is true for n = l
Assume now
order2A: +
are to
show
that under the conditions given
is a perfect square.
since,
when A = r
\
that the theorem is true
partitioning, write
oi
,1
\A\ = a
\,
2
.
\-a Oj
when
P^l
\
n = k
and consider the skew-symmetric matrix A = [a^j] of
E==
where
'
r
\
-zriî.-zii^-zy
'^2k^..2n^\
Then S
is
skew-sym-
"
CHAP.
THE ADJOINT OP A SQUARE MATRIX
6]
metric of order 2k and, by assumption,

If
\b\ =
a^j denotes the cofactor of a^j in
Moreover,
/I,
*2fe+l, 2fe+l
0^2^+2. 2fe+l
f''2fe+l,
0'2fe4-2,
Otsfe+s.sfe+i =
2fe+2
-fflgfe+i^sfe+s
1^1/
where /is a polynomial
we have by Problem
elements of B.
in the
Chapter
6,
53
and (6.8)
3,
0^2^+2, 2fe+l
^2^+1, 2/!e+2
2fe+2
hence,
Ll2/fe+l.
and
2fe+2
a perfect square.
8.
Compute the adjoint

3~
"l
(o)
~1
(b)
0_
_0
of:
(i)
_3
10
4
-2
110
(rf)
1_
1_
_0
9.
-2
0-2
(a)
(c)
1_
Ans.
2*
'l
_0
2"
S'
(c)
-2
-5
-2
(rf)
-16
10 3-5
-2
10
Verify:
(a)
(b)
(c)
The
The
The
adjoint of a scalar matrix is a scalar matrix.

adjoint of a diagonal matrix is a diagonal
matrix.
adjoint of a triangular matrix is a triangular
matrix.
10. Write a matrix
11. If
is a
/I
7^
of order 3 such that
adj^ =0.
2-square matrix, show that adj(adj/4) = A.
-1 -2 -2
12.
-4
-2
Show
that the adjoint of
13. Prove:
If
-2
2-2
an n-square matrix A is of rank
14. Prove: If
is
15. Prove: If
is Hermitian,
16. Prove: If
is
-4 -3 -3
is 3^' and the adjoint of
<n-\.
then
adj^
is
itself.
= 0.
symmetric, so also is adj-4.
so also
is adj/1.
skew-symmetric of order
, then
adj^
is
symmetric or skew-symmetric according as . is

odd
54
theorem similar to that of Problem 16
17. Is there a
18.
[CHAP. 6
skew-Hermitian matrices?
for
For the elementary matrices, show that

-1
(a) adj Hij =
-Hij
-1
H^ (k)
diag(l/A:. l/k
(c) adj Hij

Hjk(k)
Hij(k),
HJ
(6) adj
19. Prove: If
1,
l/k
stands in the ith row
with similar results for the K's.
-1
Use
n-l
and
H^...H^
if
-H-^
-A K-^-K,2---^t
where \
is
then
adj
20.
where the element
l/k),
an n-squaie matrix of rank n or
is
or
1/k,
adj
Xi
-1
adj
-1
K^
K^
adj
-1
adj
adj H^
-1
adj
H^
-1
adj H^
the method of Problem 19 to compute the adjoint of
1110
(o)
of
12
Problem 7 Chapter 5
(b)
2-2
-14
7
Ans. (a)
-3 -3
-1
-1
14
= [a^--]
23.
is n-square then
Let A^ =\.%j\
example,
1-1
and B = [^k-aÂ be 3-square matrices.

S(adj-4)
22. Prove: If
-2
(b)
-7
21. Let
-2
and
S(adjB)
adj (adj
^)
If
S{C) = sum of elements of matrix C, show that
\b\
UI
k S(adj,4)
Ui
be the lower triangular matrix whose triangle is the Pascal triangle; for
('/ = 1. 2, ...,n)
10
110
12 10
13
3
Define
bij
"V
= (-1)^ âi,
Hj
and verify
for n
adJ-4
(i)
24.
Let B be obtained from A by deleting
where
=2,3,4
o!^-
its ith
îj
^pj
iq
pq
is the cofactor of a^j in \A\
that
[6.^,-]
tjj
and pth rows and /th and gth columns.
Show
that
chapter 7
The Inverse of a Matrix
IF A
AND B
A
is
are n-square matrices such that
called the inverse of B, (A =

In
Problem
I.
An
II.
THE INVERSE
If
ra-square matrix
BA
A has an Inverse
if it is
of a non-singular n-square matrix is unique.

is non-singular,
then
AB
AC
of a non-singular diagonal matrix
:î' -^2
is called the inverse of A,
and only
if
implies
B =
diag (i k,
diag(i/k^,l/kr,.
^^
=1,
(B=A~^) and
we prove
l,
The inverse
AB
B''^).
non-singular.
(See Problem?, Chapter2.)

C.
A)
diagonal matrix
is the
1A)
...,
are non-singular matrices, then the inverse

of the direct
diag(Al^.
A^
sum diag(/}
A~s)
Procedures for computing the inverse of a general

non-singular matrix are given below.
INVERSE FROM THE ADJOINT. From
adj
i = U|
/.
if
is non-singular
n/M|
a^^/\A\
...
ai/U|
i2/U|
a22/!.4|
...
a2/UI
am/Ml
a^n/\A\
...
ann/\A\
adj^
(7.1)
Example
(6.2)
1.
From Problem
2,
Chapters, the adjoint of A
"7/2
Since
Ml
= -2,
A~^ =
4 3
-1
-1
-2
-3
^'^^ -^
-k
2
-7
is
^J
See Problem
55
2.
THE INVERSE OF A MATRIX
56
INVERSE FROM ELEMENTARY MATRICES.
[CHAP.
Let the non-singular n-square matrix A be reduced to
by elementary transformations so that
H^.-.H^.H^. A-K^-K2...Kt
Then A ^ P
Q^
Example
2.
From Problem
Q-P
(P'^-Q'Y
K^-K^... KfH^... H^-H^
K^K2H2'i
o"
-1
In Chapter 5
-3 -3
1
-10
that a non-singular matrix can be reduced

Then, from (7.2) with Q =1, we have
normal form by row
to
H,...H^.H,
(7.3)
That
-1
was shown
it
transformations alone.
-1
1
1
1__
"106" '100"
-3
_0
-3"
'l
-3
1_
-3
"l
-1
-1
Chapters,
7,
HqH^ AK^Kq
Then
(S~V^ = S,
by (5.5) and, since
A'^
(7.2)
PAQ
is,
III.
If
is
reduced to
by a sequence of row transformations alone, then A~
equal
is
to the product in reverse order of the corresponding elementary matrices.
Examples. Find the inverse
of
.4
4 3
13
Write the matrix [A
Ir,
3 3
4 3
3 4
of
Example
using only row transformations to reduce
/I
to/.
and perform the sequence of row transformations which carry A into
on the rows of six elements.
[AI^]
We have
1
1
"X^
1
1
-1
-1
-3
-"o
'\j
-1
-li
-3 -3
1
[/3^"']
7
by
(7,3).
Thus, as A
is
reduced to
/g, /g is
carried into
-1
-1
-3 -3
1
1
See Problem
INVERSE BY PARTITIONING.
Let the matrix A = [a^j] of order n and
its
inverse
B =
titioned into submatrices of indicated orders:
41
A12
^11
"12
(pxp)
(pxq)
(pxp)
(px?)
^21
A22
(qxq)
"21
S22
(qxp)
(qxq)
and
(qxp)
where p + q = n
[6.^,-]
3.
be par-
CHAP.
7]
Since
AB
BA
(i)
57
we have
I^,
/4iiSii + /lisBgi
4iiSi2 +
Ip
(iii)
Bgi'îi + B^z'^si
(iv)
Bjiiia + Bss^âs
(7.4)
(ii)
Then, provided
/4ij^
!Bxi
where
.422
A-^^B^-z
Iq
is non-singular,
'4ij^
(A.^^^
"-i-zK
^21
(^^si '^n)
~C
(^21 'û)
- AîÂli A^^).
See Problem
To
taken of order ra-l.
In practice, A.^^ is usually
used.
4.
obtain A^^, the following procedure is
Let
[Oil
%2l
2i
^âj
"11
Osi
'
31
%2
13
022
O23
32
G^ =
11
%2 %3 %4
021
022
023
Ogj
Ogj
033
6(34
41
%2
'''43
''44
33_
'^24
After computing C^^, partition G3 so that /422 = [o33] and use (7.5) to obtain C^^. Repeat the proc
ess on
G4. after
partitioning
so that ^22 = [044], and so on.
it
Example 4. Find the inverse of A
Take ^11
11
^=
[!'].
^22
4 3
3 4
^12
'42i(^;' ^12)
using partitioning.
^421 =
[13],
and ^22= [4]-
[4]-[l3]M
and
[1],
Now
f^
[l]
Then
Sii
4\
S12
-('4i\î2)r^ "
('4i\/ii2)rwîi)
S21
-^
(^2i'4la)
522
r'
[j
i]^[o]fii-'^^^
oj
[-1,0]
[1]
"
fill
fil2
and
-1
-3 -3
1
S21 B22
_
-1
See Problems
5-6.
58
THE INVERSE OF A SYMMETRIC MATRIX. When A

tors
need be computed instead of the usual
[CHAP. 7
symmetric, aij =
is
ciji
and only ^(n+1) cofac-
n^ in obtaining A~^ from adj A.
If there is to be any gain in computing A~^ as the product of elementary

matrices, the elementary transformations must be performed so that the property of being symmetric is preserved.
This requires that the transformations occur in pairs, a row transformation followed immediately
by the same column transformation. For example,

'0
...
b
.
Ht
...
...
^l
b
c
î(- ^)
^î(- |)
However, when the element a in the dia^gonal

and K^{l/\Ja).
H^(l/\Ja)
is not
In general,
replaced by
is
1,
the pair of transformations are
^Ja is either irrational or imaginary; hence, this
procedure
recommended.
The maximum gain occurs when
the method of partitioning is used since then (7.5) reduces to
(7.6)
where
A^^ - A.^^(Af-Â^^).
See Problem
When A
is
is
not symmetric, the above procedure
symmetric, and then the inverse of A
to find the inverse of A'A,
which
found by
is
A'^
(7.7)
may be used
7.
(A'AiÂ-
SOLVED PROBLEMS
1.
Prove:
An
re-square matrix
Suppose A
that
is
non-singular.
By Theorem
PAQ=I. Then A=P~^-Q'^ and

-1
Supposed
A
A has an inverse
A''^
if
IV,
Q-P
and only
Chapter
5,
if it is
non- singular.
there exist non-singular matrices
-1
exists.
is non-singular.
The A- A
=1
P and Q such
exists.
is of rank n.
If
/4
were singular,
_i
/l/l
would be of rank <
n; hence,
CHAP.
2.
(a)
7]
When A
2 3
(b)
3.
When A =
!]
[?
r4-3l,andi"=r4/5-3/5-]
=5, adji
[-1
2 3
-5
-5
then U! = 18, adj A
2 4
3 6
2 5
-3
_4 5
14
4 =
2 4
4 5
14
1 1
2/5J
-5
and A
-5
14_
0'
10
[_-l/5
2j
Find the inverse of
[AIJ
/I
I
59
3/2
3 6
2 5
-3
14
14
1
"K^
1/2
1
1
^\j
3/2
1/2
'
1/2 -1
-3/2
-1 -5
-1
10
-2
1
'
- 3 10
10
-3
3/2
-1 -5
0'
1/2
1
]
'Xy
_4
-1
-3
10
10
18
1
'X'
-3/2
7/2
11
5/2
1-1-5
-1
1-2
5-5
-3
-2
"1
'V^
1/2 -1
h
-7 -2
13
-7| -4
-2 -3
0'
10
18
-7
-2
-7-2
-4
-3
'
--10
10
11
1_
-23
29
10
26/5
7/5
-12
-2
6/5
2/5
-2
3/5
1/5
100]
'^-/
'-\J
13
0'
-2
-5
-2 3/5 1/5
-64/5 -18/5
Ih
The inverse
(i)
4.
Solve
is
iiiSn
(..,
/fiiS,2
(11)
Set
A-']
S22 = f"^
^ti - A'lÂ^^B^^ =
29
10
26/5
7/5
-12
-2
6/5
2/5
-2
3/5
1/5
+ iisSsi
^
+ A^^B^^
Prom
4\
-64/5 -18/5
-23
(ii),
(iii)
Sîîi + B^Â^^ =
(iV)
^21^12 + 52,42
for
B^^ = -(-4^1-412)^"^;
from
(iii),
Bii,Si2,S2i, and S
= /
B21 =
- ^~\A^Ât^y,
and, from
+ (^lVi2)f~\'42iA\).
Finally, substituting in (iv),
-f
(-42i'4ii)î2 + ^'^422
*.
and
'^'2
('^21'*lll)'^12
(i),
B-^^
THE INVERSE OP A MATRIX
60
5.
12
13
Find the inverse of A
[CHAP.
by partitioning.
1111
(a)
Take Gg
and partition so that
[24],
-1
Now
r 3 -2l
jj.
^_^
11
,f
Then
^22 -
Sii = 4l\ +
,-1
,
^11-41. -
'42i('4iiî2)
[3
-2
|^_^
(.4lVi2)rV2iî\)
3L-3
4]r
[j
Hr-i] [2
"^l +
-'^
f (/I2A1)
and
D01
A^^
3 3
4 3
-3
Oo
-421
=[
-6
0],
A~'^
0-1
and
1 ]
/I22 =
[ I ]
AîA-^-^
l](i-)
and
&
3'
^1
3 [3]-|[2
B12 =
-3 2]
-1
-1
"
"X"L2
3 2J,
-6
C'^=[3]
3'
-1
[-2
S21 =
B22
-2].
4
=
1_
1-210
[Sii
S21
S12
1-2
2-3
1-11
B22J
-2
-2
1-2
-2
-1
6-9
-2
0'
-3
_
and
4-fl
0-1
-1
[o oJ
[l]-[l
.f
B,
ij
Then
=[-l/3]
A^Q
3-6
4i
[2 0],
Now
-2
-2] _ [2 ol
-1
3-6
that
r 3
f
3
0]
i[2
-3
A so
[24]
and
[-3],
[3]
3J
B12 = ('4ii'4i2)C
(6) Partition
'421-4 11
[3] - [2
and
3-2
[3]
CHAP.
6.
THE INVERSE OP A MATRIX
7]
Find the inverse of A =
We cannot
take
4 3
-4,
by partitioning.
since this is singular.
By Example 3,
61
3 3
4 3
3 4
7 -
-3
the inverse of
//
-3 -3'
B^Ho
"1
0'
1
1
-1
-1
7
is
-3
-3'
1
Then
Thus, if the (ra-i)-square minor A^.^_ of the n-square non-singular matrix A is singular, we first bring a
non-singular (n-l)-square matrix into the upper left corner to obtain S, find the inverse of B, and by the proper transformation on B~^ obtain A''^
1-12
2
7.
Compute
the inverse of the symmetric matrix
13
-1
2-3
1-1
2-3-1
1
-1
-1
Consider
first the
submatrix G^
partitioned so that
-421
^22=
[-1 2],
[1]
Now
^22 - ^2i(Aiî2) =
Then
[l]-[-l
2]rM
[-2]
and
f^=
[-i]
B,
B12
["jj.
B21
[-k
10
-5
-1
-5
-5
and
B22 = [-i]
k].
Consider now the matrix A partitioned so that
'^11
A<
_1_
10
-1
-1
-5
3-1
Now
-5
-5
[2 -3 -1],
-1/5
2/5
-2
<f= [18/5
],
[4]
[5/I8].
62
5-7
Then
5-15
JS^
[CHAP.
-7
^[1
521
-2 10],
B22 =
[5/I8]
11
5-71
5-1
and
-7
5-2
11
10
-2 10
8.
Find the adjoint and inverse of each of the following:

"
2-1"
"2
4"
_1
3"
"1
3
(a)
-1
-1
(b)
2
1_
(O
(d)
n
w
n
U
n
^
1
J.
3-15
Inverses
(a)
-10
^ -153-1
5
-5
9.
10.
Find the inverse of the matrix of Problem
8((?)
-3
-3
-1
-2/3
1/3
(d)
1/2 -1/6
2-1
-J
(c)
-4 -14
15
^'î
1/3
as a direct sum.
Obtain the inverses of the matrices of Problems using the method of Problems.
13
3-1
13
11
3
1
11.
Same,
for the
matrices
(a)
-4 -5
(a)
16
-6
-144
- 30
-1
48
-10 -44
4 -13
30
-2
-1
(c)
11
-1 -7 -3
48
-26
16
-1
-1 -1
Example 4
(d)
36
60
12
21
-5
-20 -12
-4 -12 -13
48
12 -12
3
30 -20 -15
25
-18
21
30 -11
(^)i
-5
-8
-30
-15
-9
12
6-9
15
-7
-6 -1
-1
12
12.
Use the result
13.
Obtain by partitioning the inverses of the matrices of Problems 8(a), 8(6), 11(a) - 11(c).
of
11-1
1-2-12 2
41
2
(c)
(b)
-1
(b)
22
18
-4
-5
to obtain the inverse of the matrix of
Problem 11(d) by partitioning.
CHAP.
14.
7]
63
12-12
12
2-11
112
Obtain by partitioning the inverses of the symmetric matrices (a)
-1
-1 -1 -1
-1 -1
1
-1
-1 -5 -1
Ans. (a)
-1
15. Prove: If
16.
Show
(a)
17.
andS,
Show
that
if
AÂ
Hint:
18.
Show
A und B
(b)
20.
(AA"^)' =
AC
is said to
and only
Find a
if
The rank
is the
A and B commute, so also do

B
Hint, (a) A (AB)A
(A~^)'A
is
symmetric, so also is
/I
(BA)A
^.
A and B commute, then
(a) A~'^B.
(b) AB"'^.
and
(c) A~^B~^
have a right inverse B if AB = I and a left inverse C it CA = I. Show that A

has
A is of rank m and has a left inverse if and only if the rank of A is n
13 2 3
14 13
13 5 4
if
one exists.
of A is 3 and the submatrix S
4x3 matrix
= C.
a.nd
17
if
right inverse of
{A'^ B)' = {BA^'^)' = (A'^)'B' = A~^B.
Hint: (a)
a right inverse
implies
that if the non-singular symmetric matrices
An mxn matrix A
Hint.
(c)
the non-singular matrix

:r
are symmetric.
19.
AB
then
that if the non-singular matrices
1-12
3-3 2
3-4 4-2
-3
4-5 3
2-2 3-2
(b)
is non-singular,
.4
-1
-3
-1 -1
(b)
-1 -1
2
is
non-singular with inverse S
right inverse of
-9 -5
-4
-1
L J
Show
that the submatrix
3 3
21,
4 3
3 4
of A of Problem 20 is non-singular and obtain
-1
-1
-1 -1
-3
23.
Show
|/+
l|
that
24. Prove: If
25. If
-310b
13
14
2
|.4ii|
0,
0,
then
;^
26. Prove:(i) of
as a
inverse of
left
has neither a right nor a
/111
A21
+
where a,b. and
then
.4)-^
and
(I
Problem 23, Chapter
6.
(/
^12
Aq2
|.4ii|
- A) commute.
left inverse.
-1^22 - ^^siîi/ligl
as another
right inverse of A.
22. Obtain
-3 -3
c are arbitrary.
chapter 8
Fields
NUMBER
collection or set S of real or complex numbers, consisting of more than the ele-
FIELDS. A
ment 0, is called a number field provided the operations of addition, subtraction, multiplication,
and division (except by 0) on any two of the numbers yield a number of S.
Examples
number fields
of
(a) the set of all rational
(b) the set of all real
are:
numbers,
numbers,
(c) the set of all
numbers of the form a+b\f3
(d) the set of all
complex numbers a+
bi,
where a and b are rational numbers,

where a and b are real numbers.
,
set of all integers and the set of all numbers of the form
The
bvs
where
i is
a rational
number, are not number fields.
GENERAL FIELDS. A
collection or set S of two or more elements, together with two operations called
addition (+) and multiplication
is called a field
().
provided that
a, b, c,
being elements of
...
F, i.e. scalars,
a + b is a unique element of
a+b
= b+a
a + (b + c)
= (a +
b) + c
For every element a in
For each element a in
As
ab =
a-b
is a
F
F
there exists an element
in
F such
there exists a unique element
unique element of
-a
in
that
a+
F such
+a =
that
a.
a + (-a) =
0.
ab = ba
(ab)c = a(bc)
For every element a in
For each element a ^

a~
Di
in
?^
such that
there exists a unique element a'^ in
1-a =
F such
that
o.
a-
a'''^
-0=1.
a(b +
there exists an element
c)
= ab+ ac
(a+b)c = ac + bc
D^'-
In addition to the number fields listed above, other examples of fields are:
(e) the set of all quotients
(/) the set of all
2x2
^^^
Q{x)
of polynomials in
matrices of the form
ti
x with real coefficients.
where a and
b are real
numbers.
0+0=0.
This field, called of characteristic 2, will be excluded hereafter.

customary proof that a determinant having two rows identical
not valid. By interchanging the two identical rows, we are led to D = -D or 2D =
(g) the set in
which
In this field, for example, the

is
but
is
is not
necessarily
0.
64
CHAP.
FIELDS
8]
SUBFIELDS.
S and T are two sets and
If
65
every member of S is also a member of T, then S is called
if
a subset of T.
S and T are fields and
If
if
S is a subset of T, then S is called a subfield of T.
For exam-
the field of all real numbers is a subfield of the field of all complex numbers; the field of
ple,
all rational
numbers
is a subfield of the field of all real
numbers and the
field of all
complex
numbers.
MATRICES OVER A FIELD, When

'Mis over F". For example.
1
A =
1/4
1/2
2/3
elements of a matrix A are in a field F, we say that
all of the
and
over the rational field
is
11
- 3i
is over the
complex
field
Here, A is also over the real field while B is not; also A is over the complex field.
Let
A.B, C,
is,
let
F be
the smallest field
the elements are rational numbers, the field
if all
rational field and not the real or complex field.

fined
F and
be matrices over the same field
...
contains all the elements; that
An examination
which
is the
of the various operations de-
on these matrices, individually or collectively, in the previous chapters shows that no

F are ever required. For example:
elements other than those in
The sum,
difference, and product are matrices over F.
If
If
A ^l then
If
is non-singular, its inverse is
P and Q over F such
there exist matrices
is over the rational field
over F.
and
rank
is of
r,
its
that
PAQ
and
is over F.
rank is unchanged when considered over
the real or the complex field.
Hereafter when A is said to be over

taining all of its elements.
it
will be
assumed
that
is the
smallest field con-
In later chapters it will at times be necessary to restrict the field, say, to the
real field.
At other times, the field of the elements will be extended, say, from the rational field to the real
field.
Otherwise, the statement "A over F" implies no restriction on the field, except for the
excluded field of characteristic two.
SOLVED PROBLEM
1.
Verify that the set of all complex numbers constitutes a field.
To do
this
we simply check
the unit element (M^) is
product
(A/ 1)
is
1.
(a+bi){c +
If
the properties
a + bi
di) = (ac
and
di
A^-A^, Mi-Mg, and D^-D^. The zero element

are
-bd) + (ad + bc)i
a + bi
a^
bi
+62
two elements, the negative (A^) of a +

;
the inverse (M5) of a +
a^+
b^
hi
a^ + b^
Verification of the remaining properties is left as an exercise for the reader.
biô
is
bi
(/I4) is
is
-a-bi.
and
the
66
FIELDS
[CHAP.
2.
Verify (a) the set of all real numbers of the form a + b\r5
(6) the set of all quotients
numbers and
Verify (a) the set of all rational numbers,
numbers a + bvS, where a and b are rational numbers, and

numbers a+bi, where a and b are rational numbers are subfields of the complex
(b) the set of all
(c) the set of all
4.
b are ra:ional
of polynomials in x with real coefficients constitute fields.
Q(x)
3.
where a and
Verify that the set of all
Show
where a and
that this is a subfield of the field of all 2x2 matrices of the form
5.
Why does not
6.
set
a ring.
ring.
2x2 matrices of the form
2x2
the set of all
b are rational
numbers, forms a
w]\eK a and
h are real
field.
field.
numbers.
matrices with real elements form a field?
of elements a,b.c.... satisfying the conditions {Ai, A^. A^. A^, A^; Mi.
D^, D^) of
M.y,
Page 64
is called
To emphasize the fact that multiplication is not commutative, R may be called a non- commutative
When a ring R satisfies Mg, it is called commutative. When a ring R satisfies M^. it is spoken of as
a ring with unit element.

Verify:
even integers 0,2,4,
(a) the set of
(b) the set of all integers

(c) the set of all
(d) the
...
0,+l,2,+3,
is an
n-square matrices over
set of all
2x2
example of a commutative ring without unit element.
is an example of a commutative ring with unit element.
...
is an
example of a non-commutative
matrices of the form
where a and
ring with unit element.
are real numbers, is an example of a
commutative ring with unit element.

7.
Can the set

ments 1
8.
Problem 6 be turned into a commutative ring with unit element by simply adjoining the ele-
(a) of
to the set?
By Problem
4,
the set (d) of Problem 6 is a field.
Is every field a ring?
element a field?
9.
Describe the ring of
10.
=
.
show
all
that
Let C be the field of

u, V are real
Is every
commutative ring with unit
numbers.
2x2
LA
all
To
matrices
= A.
Call
L a
ol
^
where a and
left unit element.
complex numbers p + qi and
Take the complex number a +
bi
b are in
F.
If
is
A.
any matrix of the ring and
Is there a right unit element?
be the field of
all
2x2
-b^
and the matrix
matrices
where
p, q,
^
as corresponding elements of
the two sets and call each the image of the other.
[2
-3l
To
-4l
2j
L4
Oj
(b)
Show
that the image of the
(c)
Show
that the
(d)
What
is the
(e)
What is the image of the inverse of
3+ ^^2i,
sum (product) of two elements of
image of the identity element
image
5.
of
of the conjugate of a + bi?
"!]
[:
This
is
is the identity
an example of an isomorphism between two sets.
is the
sum (product)
element of C.
of their
images
in C.
chapter 9
Dependence
Linear
and Forms
of Vectors
pair
of real numbers (%, x^) is used to denote a point Z in a plane. The same
2-vector
or
vector
two-dimensional
the
to
denote
used
here
will
be
as
x^],
written
[x^,
of numbers,
THE ORDERED PAIR
OX
(see Fig. 9-1).

X2
''^3(*11
^21..
^12 +
^2!
X.2(X^1. X22)
X(Xi. Xq)
Fig. 9-1
If
.Yi
= [%i,xi2]
sum (see Fig.
their
X^=
and
[x^^.x^q]
are distinct 2-vectors, the parallelogram law for
9-2) yields
A3
A^ + A2
^11
Treating X^ and Xg as 1x2 matrices, we see that this

en in Chapter 1. Moreover, if k is any scalar,
kX-i
L rC%'\
is the familiar multiplication of a vector
VECTORS. By
-1
"^
-^12
"^
-'^221
merely the rule
is
-^21
fv
X-y
for
adding matrices giv-
by a real number of physics.
an n -dimensional vector or re-vector
over
is
meant an ordered set
of n
elements x^
of F, as
(9.1)
The elements
x^,X2
Later we shall find
%
it
and
[a;i,x2, ...,%]
are called respectively the first, second,
...,
reth
components of X.
more convenient to write the components of a vector in a column, as
(91)
Now
I X-i
X2,
Xi
same vector; however, we shall speak of (9.1) as a row vector

We may, then, consider the pxq matrix A as defining p row vectors
(9.1) and (9.1') denote the

(9.1')
as a column vector.
(the elements of a row being the
components of a
67
17-vector) or
as defining g column vectors.
68
The
whose components are
vector, all of
The sum and difference

tor are
of
zero, is called the zero vector and is denoted by 0.
two row (column) vectors and the product of a scalar and a vec-
1.
Consider the 3-vectors
(a)
2X-^-iX^
(6)
IX^+X^.
(c)
2X^
{d)
2X^ -
2[3,l,-4] - 5[2,2,-3]
2[2,2,-3]+ [-4,-4.6]
^3= [0,-4,1],
-^2= [2.2,-3],
A:i= [3,1,-4],
-3X2-^3
X^- Xq+
X^
are said to be linearly dependent over
[-4,-8,7]
that if
primed to denote col-
is
^mJ
|.%]^,%2'
each bracket
provided there exist
elements
h-^^.k^,
,k^
of F, not
such that
+
k.^X-1
(9.3)
Otherwise, the
k^X^^
k^X-^
vectors are said to be linearly independent.
Example 2. Consider the
Example
four vectors of
so also are X^, X^, and X^ by
The vectors X^ and

fel^l
Then
3fti + 2k^ =
and then ^2 = 0.
n-vector
+
0,
Xq^.
(c)
[Zk^ + 2k^,
+ 2^2 =
ft^
By
(6) the vectors
^2 ^"d
Xj^_
are linearly dependent;
{d).
however, are linearly independent. For. assume the contrary so that

=
;c2A'2
1.
and the entire set by
and
0,
k^+
-ik^ -
2k^, -'iJc.i_~ Zk^l

lik^
= 0.
Prom
[o,
0,0]
the first two relations A^ =
are linearly dependent.
and the n-zero vector
vector ^,B+i is said to be expressible as a linear combination of the vectors Ai, X^
there exist elements
k^, k^, ...,k^
Xfn+i
BASIC THEOREMS.
If
in (9.3),
î
(9.4)
-I5]
[o,0.o] =
re-vectors over
Ai
Any
10,
LINEAR DEPENDENCE OF VECTORS. The m
all zero,
-S] - [lO,
[6. 2,
^4= [-4,-4,6]
and
The vectors used here are row vectors. Note

umn vectors, the results remain correct.
formed by the rules governing matrices.
Example
if
[CHAP.
=
Xi
k^
?^
0,
F such
of
%A]^ +
we may solve
- T"!^!'^! +
=
SiJVi
X^
that
:2A2
k^Xf^
for
+ î-i^t-i + -î+iî+i +
+ s^_iZi_i + s^j-iî+i +
+ ^m^m!
or
+ s^J^
Thus,
I. If m vectors are linearly dependent, some one of them
a linear combination of the others.
may always be expressed as
CHAP.
9]
II.
If
vectors X^, X^
ing another vector

bination of Xi, X^,
is
X^j,-^
, X^
Examples. Prom Example
X^ are linearly independent while the set obtained by addlinearly dependent, then Jt^+i can be expressed as a linear com-
If
X^.X^.^niXg
Zg=2Aî-3^2-
the vectors X^, and X^ are linearly independent while
2,
Xg=
linearly dependent, satisfying the relations 2X.i^-2X^-
III.
69
among the m vectors
X^ there
X^, X^
is
0.
Clearly,
a subset of
r<m
are
vectors which are
linearly dependent, the vectors of the entire set are linearly dependent.
Example
4.
By
(b) of
Example
the vectors
1,
X^ and X^ are linearly dependent; by
(d),
the set of four
See Problem
vectors is linearly dependent.
1.
IV. If the rank of the matrix
%2
Xi 1
m<n
(9.5)
associated with the
^n2
*TOl
vectors (9.2) is r<m, there are exactly
vectors of the set which
are linearly independent while each of the remaining m-r vectors can be expressed as a
linear combination of these r vectors.
See Problems 2-3.
V.
is
necessary and sufficient condition that the vectors (9.2) be linearly dependent
that the matrix (9.5) of the vectors be of rank
r<m.
If
the rank is m, the vectors are
linearly independent.
The
If
set of vectors (9.2) is necessarily linearly dependent
if
m>n.
the set of vectors (9.2) is linearly independent so also is every subset of them.
A LINEAR FORM
over
in
ra
variables
n
(9-6)
x^, x^
OiXi
is a
a^x^
polynomial of the type

+
a^x^
aÂ:^
i=i
where the coefficients are

Consider a system of
in
F
linear forms in n variables
/l
CIu% + %2^2 +
(9.7)
1^1
/m
0~foXo
'^2n*re
%n^n
and the associated matrix
-'ml
If
there exist elements k^.k^
Kk
k^
+
^^^2/2
"m2
not all zero, in

+
...
A;4
F such
=
that
70
[CHAP. 9
the forms (9.7) are said to be linearly dependent;
independent.
to the linear
Example
5.
otherwise the forms are said to be linearly

independence of the forms of (9.7) is equivalent
independence of the row vectors of A.
Thus, the linear dependence
dependence
The forms
/i =
or
2xi
- 2 +
3*g,
or
2%
x.^+
/2 =
ix^
/g =
4^=3.
Tx^ + Xg
are linearly depend-
2-13
A =
ent since
The system
-7
is of rank 2.
(9.7) is necessarily
dependent
Here,
if
3/^
"if^
fs =
m>n. Why?
SOLVED PROBLEMS
1.
Prove:
If
among
the
vectors X^,X^, ...,X^ there is a subset, say, X^,X^
linearly dependent, so also are the

Since, by hypothesis,
k^X-^ +
k^X^
k^X^ +
+ k^X^ +
X^,
r<m,
which is
vectors.
+ k^X^ =
with not all of the k's equal to zero, then
+ k^X^ +
0-.Y^+i +
0-.Yot
with not all of the k'a equal to zero and the entire set of vectors is linearly dependent.
2.
Prove:
If
the rank of the matrix associated with a set of
ra-vectors is
r<m, there are exactly
vectors which are linearly independent while each of the remaining m-r vectors can be expressed
as a linear combination of these
vectors.
Let (9.5) be the matrix and suppose first that m<n

If the r-rowed minor in the upper left hand comer
equal to zero, we interchange rows and columns as are necessary to bring a non-vanishing r-rowed minor
.
is
into this position and then renumber all rows and columns in natural order.
11
11?
21
25-'
Thus, we have
Consider now an (r+l)-rowed minor
*11
12
%-r Xiq
%1 %2
*rl X^2
xp^
where the elements xp; and

...,A;^+i =
by (3.10)
xj^q
Xpr,
Xqt x^q
^rr Xrq
.
xpf xpq
are respectively from any row and any column not included in A.
be the respective cofactors of the elements x^g. x^q
x^q. xpq,
Let h^,k^,
of the last column of V.
Then,
CHAP.
9]
Now
+ krXri + ^r+i*î
k^x^q + k^xQq +
and by hypothesis
u.
k2X2i +
fciîi
= 1,2
(i
+ krx^q + krA-ixpq
71
r)
V be replaced by another of the remaining columns, say the column numbered

cofactors of the elements of this column are precisely the k's obtained above
column of
let the last
The
not appearing in A.
so that
k^x^n + ^2*2W +
+ ^rXru + ^r-n^pu
Thus,
and,
summing over
all
k^2t +
k^x^j;
values of
"
/i:,^+i
was any one

binatlon of
0,
ji^
Xp
is
a.
X^
k^X^ +
linear combination of the
of the m-r vectors
^j^,
(t
f'r-n'^pt
= 1,2
n)
t,
k^X^ + k^X^ +
Since
f'r'^rt
Î
^r+2
-V^+i,
k^^-^Xp
linearly independent vectors
hence, each of these
X-^^.
X^
X^.
But
^^j
may be expressed as a linearcom-
X^.
m>n.
consider the matrix when to each of the given m vectors m-n additional zero compoThis matrix is [^ o]. Clearly the linear dependence or independence of the vectors and
also the rank of A have not been changed.
For the case
nents are added.
Thus, in either case, the vectors

tors
3.
X-^.X^.
...,
X^
as
was
X^
Xr+^
are linear combinations of the linearly Independent vec-
to be proved.
Show, using a matrix, that each triple of vectors

X^ = [1,2,-3.4]
X-L
^2 = [3,-1,2,1]
(a)
and
^2=
(b)
In
1,-2]
[2, 3,
^3= [4,6,2,-3]
^^3= [1,-5,8,-7]
is linearly dependent.
= [2,3,1,-1]
each determine a maximum subset of linearly independent vectors and
express the others as linear combinations of these.
(a)
Here,
12-34
3-121
1-5
is of
rank
2;
there are two linearly independent vectors, say
-1
j^
Consider then the minor
Here
1-1
1-2
2-3
;
The minor
-3
3-12
1-5
respectively -14,
(b)
and X^
8-7
1
X.^
7,
and -7.
Then -1^X^ +
is of rank 2;
we interchange
The cofactors
of the elements of the third column are
1X2-1X2=
and Xg
-2X^ + X^.
there are two linearly independent vectors, say
the 2nd and 4th columns to obtain
The cofactors
of the elements of the last
column of
-1
-2
are
2-113
2-213
4-326
and
Xg
for
Xi +
X,
Now
-1
-2
which
2,2,-2 respectively. Then
4-32.
2X^ + 2X2 - 2Xs =
X^ and Xg.
the
5^0.
72
4.
Let Pi(l, 1, 1), PsCl, 2, 3), PsiZ, 1, 2), and P^(2, 3, 4) be points in ordinary space.
and the origin of coordinates determine a plane tt of equation
X y z
12
The
points
Pi,
P^
1111
(i)
[CHAP. 9
2y +
1
1
Substituting the coordinates of P^ into the left

2
member
1111
1110
12
12
We have
The
tt.
Any
verified:
we have
4
significant fact here is that
[P^.,
Px.Pq
]'
is of rank 2.
three points of ordinary space lie in a plane through the origin provided the matrix
of their coordinates is of rank
Show
(i).
Thus, P4 lies in
of
2.
that Pg does not lie in v.
5.
Prove:
If
vectors
-^m+i is linearly
6.
Show
X,^.
X^
are linearly independent while the set obtained by adding another vector
X-^
dependent, then ^m+i can be expressed as a linear combination of X^.X^
that the representation of /^^+i in
m
Hint: Suppose
^7^+1
Prove:
kiXi
1,
and consider
siXi
(A:^.
- s^ ) X^
i=i
i=i
necessary and sufficient condition that the vectors (9.2) be linearly dependent
(9.5) of the vectors be of rank
Hint: Suppose the
vectors are linearly dependent so that (9.4) holds.
converse, see Problem
is that the matrix
r<m.
product of the first row by
8.
is unique.
i=i
7.
Problems
X^^.
s^,
the product of the
second row by
In (9.3) subtract
S2.
from the ith row the
^s Indicated in (9.4).
For the
2.
Examine each of the following sets of vectors over the real field for linear dependence or independence. In
each dependent set select a maximum linearly independent subset and express each of the remaining vectors
as a linear combination of these.
^1
Xj_
[2,-1,3,2]
X^
[1,3,4.2]
X2 = [2,1.4]
(a)
(a)
Xq
2X^ - X^
[2,1, 3,2, -1]
^2 =
[4,2 1.-2 .3]
[0,0 5.6, -5]
(c)
(6)
Xa
[4,5,6]
Xs
^4
[1.8.-3]
X^
[6,3 -1,--6,7]
Xs
x^
2^1--X^
2X2--^1
X3 = [3,-5,2,2]
Ans.
^1
[1,2,1]
A3 = 2a^ +
(b)
A. A
A.Q
5a 1 2a o
(c)
CHAP.
9.
10.
9]
Why can
Show
73
there be no more than n linearly independent -vectors over F'>
that if in (9.2) either Xi = Xj
X^ = aXj, a
or
in F. the set of vectors is linearly
dependent.
Is the
converse true?
11.
Showthatanyn-vectorZandthen-zero vector are linearly dependent; hence,

k^-O =
where fc^ = o and ftj ^ 0.
A"
ando are considered
proportional.
Hint: Consider k^^X +
12. (a)
Show
that
[l.l+i,i], X^ = [j,_j,i_i] and X^ = [l-i-2i,l-i, 2-j
X._ =
are linearly dependent over
the rational field and, hence, over the complex field.

(b)
Show
Zi
that
[l.l+i.i], X^ = [i.-i.l-i], and Xq = [o.l-2i.2-i]

complex field.
are linearly independent over
the real field but are linearly dependent over the
13. Investigate the linear
()
dependence or independence of the linear forms:
Xg + 2Xg + x^
/i
fz
2::i
+ 3x2 - Xg+ 2x^
/a
5x^
Arts, (a)
14.
3% -
3/i
95C2
2/2
+ 8xq -
2Xx
f^ =
3%-L
+ 2^2 -
fg =
5Xj^
- X2+ 2Xq +
fx
(b)
x^.
3Xg +
4A!g
- 2*4
2x3 + 5*4
X4.
/g
Consider the linear dependence
or
Hox
independence of a system of polynomials
+ aij^x
+ ai_^x
+ a;
(i =
1,2
m)
and show that the system is linearly dependent or independent according as the row vectors of the coefficient matrix
"10
"11
20
21
"^0
"ni
nn
are linearly dependent or independent, that is, according as the rank
of
is less than or
equal to
polynomials of either system are linearly dependent, find a linear combination which is identically
15. If the
zero.
Pi =
(a)
~ 3x2 + 4^ _
- 6 +
P2 =
Ps =
Ans.
16.
2x2
2*2 +
2Pi + Pg
(a)
Pj_
(b)
2x* +
P2 =
P^ +
(6)
P^-
2Pg
-4x^+5x
x +
X* +
Pg =
2P3 =
3:c^
2x-
+ 3
2x2- Zx +
x^ +
X + 2
Consider the linear dependence or independence of a set of 2x2 matrices M.
[::]-Ci]-[::]
over F.
Show
that
feîl/i
A:2^^2
+ ^3^3 =
when
not all the k's (in F) are zero, requires that the rank of the
abed
matrix
g h
be <
3.
(Note that the matrices
components,)
Extend the result
to a set of
mxn matrices
M-^.Mz.Mq
are considered as defining vectors of four
74
17.
Show
that
18.
Show
and
are linearly dependent.
3
-
any 2x2 matrix can be written as a linear combination
that
[CHAP.
of the matrices
[o
oj'
[o oj
and
[i
oj'
Generalize to nxn matrices.
X^
X^^.X^
19. If the ra-vectors
are linearly independent,
show
that the vectors
Y-^.Yr,
1^
where
7^
20. If
aijXj.
is of rank
where
21.
C^,
C^
(6)
(c)
Show
Show
Does
Show
if
and only
if
\_aij'\
is non-singular.
show how to construct a non-singular matrix B such that AB

C^ are a given set of linearly independent columns of A.
r,
Given the points Pi(l,

(a)
22.
are linearly independent
that the rank of [Pi, P3]' is

that [P^, P^. P3,
PaY
is
C2
C7-,
o]
Ps(2, 2, 2, 2), and /VO, 4. 5, 6) of four-dimensional space,

so that the points lie on a line through the origin.
of rank 2 so that these points lie in a plane through the origin,
Pjd,
1, 1, 1),
= [C^,
2, 3. 4),
1
P5(2, 3. 2. 5) lie in the plane of (6)?
that every n-square matrix
A^ +
A over F satisfies an equation of the form

k-Â^
+ kr,A^
+ kp^Â + kpl
...
where the k^ are scalars of F

Hint: Consider
23.
/,
4,^^,/!^
in the light of
Problem
16.
Find the equation of minimum degree (see Problem 22) which

(a)
Ans.
24. In
(a)
4^-24=0,
Problem 23(b) and
(b)
(c),
thus verify: If A over

cients are scalars of F.
L J,
[:;]
4^-24
is satisfied
A
'-[:-:]
(b)
+ 27 =
0,
(c)
\_
A^ - 2A +1
multiply each equation by 4"^ to obtain

is non-singular, then
A'
(c)
\.
by
[;:]
=
(b)
A'''^
I-Â.
(c)
A~'^=2l-A,
and
can be expressed as a polynomial in A whose coeffi-
chapter 10
Linear Equations
DEFINITIONS. Consider a system
of
linear equations in the n
022X2 +
OqiX-i +
(10.1)
"Wl*!"'" Ob2*^2 +
+ c!2n*-n
^2
Aa
OIj,
unknowns
xi.a;.
>
*r?
x^
in
in
which the coefficients
By
a solution in
(o's)
and the constant terms (A's) are in
of the system is meant any set of values of %,%2.
isfy simultaneously the
equations.
When
the system has a solution,
otherwise, the system is said to be inconsistent.

tion or infinitely
many
it
is
F which
sat-
said to be consistent;
consistent system has either just one solu-
solutions.
Two systems
of linear equations over F in the same number of unknowns are called equivevery solution of either system is a solution of the other. A system of equations equivalent to (10.1) may be obtained from it by applying one or more of the transformations: (o) interchanging any two of the equations, (b) multiplying any equation by any non-zero constant in
alent
if
F, or (c) adding to any equation a constant multiple of another equation. Solving a system of
consistent equations consists in replacing the given system by an equivalent system of prescribed form.
SOLUTION USING A MATRIX.
In matrix notation the
system of linear equations (10.1) may be written
as
^11
02
(10.2)
\'
1
02 2
*2
^2n
hm
r".
or,
more compactly, as
AX
(10.3)
?here
A =
[o^^-] is
Consider now
(10.4)
the coefficient matrix,
for the
xj\ and H
[x^.Xr,
[h^h
A^]'
system (10.1) the augmented matrix
Oil
ai2
"i?i
021
02 2
^271^2
ml
0^2
<^nn
[A
H]
(Each row of (10.4) is simply an abbreviation of a corresponding equation of (10.1); to read the
equation from the row, we simply supply the unknowns and the + and = signs properly.)
75
LINEAR EQUATIONS
76
[CHAP. 10
the system (10.1) by means of (10.4), we proceed by elementary row transformations

A by the row equivalent canonical matrix of Chapter 5. In doing this, we operate on
rows of (10.4).
To solve
to replace
the entire
Example
3xi +
Solve the system
1.
x^
2X2
^.X-^
Xg
2x-j^
+ ^Xq + 2xq
^Xq^
The augmented matrix VA H\
-2
-3
-1
-5 -5
-5 -5
-]
2'
"1
-1
1
1
1
Thus, the solution
is the
equivalent system of equations:
pressed in vector form, we have
= [l,
0, l]
1'
1
1
x-i
.0
OJ
.0
-11 -5 -5
0.
'1
0.
=1, 2
= 0,
xq =
1.
Ex-
FUNDAMENTAL THEOREMS.
When the coefficient matrix A of the system (10.1) is reduced to the

A:]'.
row equivalent canonical form C, suppose {A H] is reduced to [C K], where K= ^1,^5
If A is of rank r, the first r rows of C contain one or more non-zero elements. The first non-zero
element in each of these rows is 1 and the column in which that 1 sta,nds has zeroes elsewhere.
The remaining rows consist

the variables x:
bleS X:
Jr+1
If
X:
Jr+2
X,-
On
a solution.
0,
...
...
Prom
of zeroes.
r rows of [C K], we may obtain each of

Chapter 5) in terms of the remaining varia-
the first
(the notation is that of
,xj
Jr
X; and one of the i^, Uq
Jn
k^
= k
If
values for
kj; 7^
x:
then (10.1) is consistent and an arbitrarily selected set of
0,
together with the resulting values of
X.
ac,-
k^.
the other hand,
if at
least one of
V+i'
"r+s
'
%Ji
,,
x
,
...
J2
is different
X- constitute
Jr
from zero, say
the corresponding equation reads

Qx.^
and (10.1)
0%
+ 0*71
7^
is inconsistent.
A and
In the consistent case,
have different ranks.

I.
[A
have the same rank; in the inconsistent case, they
H]
Thus
A system AX
of
unknowns is consistent if and only

system have the same rank.
linear equations in n
the coefficient matrix and the augmented matrix
if
of the
r<n, re-r of the unknowns may be chosen

When these n-r
r unknowns is of rank r.
are
uniquely
determined.
unknowns
other
r
values,
the
whatever
assigned
any
unknowns are
II.
In a consistent
system (10.1)
of rank
so that the coefficient matrix of the remaining
3*^3
+ 3^2 +
Xi +
Example
2.
For the system
aci
2^1 +
2*2
5a:2
~ 4^4
xg
2x4.
2%3
5x^1
10
CHAP.
LINEAR EQUATIONS
10]
12-3-4
[A
12-3-4
13 1-2
2 5-2-5
H]
77
">-/
10_
-11
"l
-V
-11 -8 10
2-2
3-2
'Xy
2-2
10
_0
lO"
0-2
[C K]
10
^0
Since A and [A H] are each of rank r = 3, the given system is consistent; moreover,
the general solution contains n-r = 4-3 = 1 arbitrary constant. Prom the last row
of [C K], x^= 0. Let xs = a, where a is arbitrary; then aj-l = 10 + 11a and xq = -2-4o.
The solution
= [lO
of the
system
-2-4a,
+ llo,
a consistent system of equations over
If
is over
F.
If
by x^ = 10 + lla, xj =
is given
-2-4o,
xg = a,
x^,
or
a, o]'
the system has infinitely
has a unique solution (Example
many solutions (Example
2)
it
1) that
solution
has infinitely many solu-
F when the arbitrary values to be assigned are over F. However, the system has
many solutions over any field 2F of which F is a subfield. For example, the system
of Example 2 has infinitely many solutions over F (the rational field) if o is restricted
to rational
numbers, it has infinitely many real solutions if a is restricted to real numbers, it has infinitely
many complex solutions if a is any whatever complex number.
tions
over
infinitely
See Problems
NON-HOMOGENEOUS EQUATIONS. A
a-^
is called
non-homogeneous
equations provided
1-2.
linear equation
Xi + 0^X2 +
+ n*n
A 7^ 0. A system AX = H is called a system of non-homogeneous

zero vector. The systems of Examples 1 and 2 are non-homogeneous
if
is not a
systems.
In
Problem 3 we prove
ni.
A system
of
re
non-homogeneous equations
provided the rank of its coefficient matrix
in n
is n, that is.
unknowns has a unique solution

provided \a\ ^
0.
In addition to the
method above, two additional procedures for solving a consistent

system
non-homogeneous equations in as many unknowns AX = H are given
below. The first of
these is the familiar solution by determinants.
of n
(a) Solution
by Cramer's Rule. Denote by 4, (i = 1,2

n) the matrix obtained from A by replacing Its Jth column with the column of constants (the
h's). Then, if \A\ y^ 0, the system
AX = H has the unique solution
<1"'5)
777
X2
'
"'
See Problem
X2 + Sxg +
2xi
Example
3.
X2
Solve the system
We
*4
- 3x3 - 4x4
6x2 - 2x3 +
X4.
Xl +
3x:i
2%-^
+ 2*2 + 2x3
using Cramer's Rule.
- 3x4
find
1
-3 -4
-2 1
2-3
-120.
-3
-4
-2
-3
-240
4.
LINEAR EQUATIONS
78
-1 -3 -4
8-21
-24,
2-3
2
and
Then
-240
x-i
[CHAP. 10
15
1-1-4
2-3
-3 -1
-2 8
2
-96
2_
A^
-24
Ml
-120
= 0,
-120
and
-120
-96
^4
(b) Solution using
^.
-120
If
^ #
0,
AX
A~^ exists and the solution of the system
is
given
by
A-^-AX
(10.6)
A-^H
or
A-^H
2xi + 3X2 +
Example
The
4.
coefficient matrix of the system
From Problem
2(6), Chapter 7,
x^ +
2%
3^1 +
a;2
A
18
+ 3xg
+
2a:3
-5
-5
-5
A-^H
AX
18
-5
is
J_
Xg
-5
-5
Then
"35'
["9"
29
18
L8_
5_
solution of the system is x^ = 35/18, x^ = 29/18, x^ - 5/18.
The
See Problem
HOMOGENEOUS EQUATIONS. A
is called
homogeneous.
A system
"2*2 +
unknowns
+ ''n'^n
of linear equations
AX
(10.8)
in n
linear equation
'^1*1 +
(10.7)
5.
system of homogeneous equations. For the system (10.8) the rank

and the augmented matrix [A 0] are the same; thus, the system is
=
is always a solution; it is
Note that X = 0, that is, %i = xs = =
is called a
of the coefficient matrix A
always consistent.
called the trivial solution.

If
the rank ot
unique solution
is
xj^
r<n. Theorem
=
II
is n,
X2=
then n of the equations of (10.8) can be solved by Cramer's rule for the
= x^=
and the system has only the trivial solution. If the rank of
...
V.
Thus,
assures the existence of non-trivial solutions.
A necessary and sufficient condition

solution is that the rank of A be r < n.
IV.
trivial
for (10.8) to
have a solution other than the
necessary and sufficient condition that a system of n homogeneous equations in
n unknowns has a solution other than the trivial solution is
|/4
=
|
0.
VI. If the rank of (10.8) is r < n, the system has exactly n-r linearly independent solutions such that every solution is a linear combination of these n-r and every such linear
See Problem 6.
combination is a solution.
CHAP.
LINEAR EQUATIONS
10]
79
two distinct solutions of AX = H. Then AX^ = H, AX^ = H, and A (Xx- X^) =

Y = X^~ X^ is a non-trivial solution of AX = 0.
LET Iiand X^he

Thus,
then
Zy, +
'P
AY
0.
Conversely, if Z is any non-trivial solution of AX =

and if X^ is any solution of AX = H,
X = Xl+ Z is also a solution of AX = H. As Z ranges over the complete solution of AX = 0,
Z ranges over the complete solution of AX = H. Thus,
VII.
the system of non-homogeneous equations
If
AX
solution of the system is given by the complete solution of
AX
lution of
Example
5.
E
AX
is
consistent, a complete
plus any particular so-
= H.
i
system
In the
solution is
A:^,
2x2
Xi
Xj +
= [o,
+ 3x3
^2 + 2 Xg
set x^ = 0; then xg = 2 and x^ =
The complete
1, 2]'.
solution of < *^
\^Xi
where a
~
+
*^
Xq + 2*3 =
[-7a,a,3o]' +
[-7a,
[O, 1,2]'
IS
-^
Then the complete solution of the given system
is arbitrary.
1.
particular
[-7a,a,3oJ
is
+a, 2+3a]'
Note. The above procedure may be extended to larger systems. However, it is first
necessary to show that the system is consistent. This is a long step in solving the
system by the augmented matrix method given earlier.
SOLVED PROBLEMS
X2 ~ 2xs +
xi +
X4 + 3 K5
+ 2x4 + 6*5
+ 2 X2 - 4 Xg - 3 X4 - 9 xg
2i -
ail
^2 +
2%
tion
The augmented matrix
1-2
2-1 2
'l
[A H]
-4 -3 -9
"1
l'
2
3_
-3
-1
-v-
-6
8
3
-2
-2
0-1
1
'\J
0_
1-2
-18
-2
-18
-6
1
1
1
-2000
13
Then
1, x^- 2xs = 0, and

may be given as xx=
x^ -
solution
x^ +
Solve
^4
+ 33=5 =
x^â,
X4
s - 2x4
+ 3xg
X2 + 2Xg +
5%
Take
2a,
2%i + 3x2 -
4xi +
0.
x^^
1.
xg = a and x^ = b. where a and b are arbitrary; the complete

x^ = -3b. x^ = b or as AT = [l, 2a, a, -3 6, 6]'.
Snliitinn*
'11
[A H]
The
last
3-1-22
row reads 0-xi +
'112
5"'
7
O-^tj +
"V-
1-5-4
1-5-4
-8
-13
1
1
13
-5 -4 -8
-5
O-^g + 0-4 = -5; thus the given system
is inconsistent
and has no solution.
LINEAR EQUATIONS
80
3.
A system AX
Prove:
provided \A\
If
7^
non-homogeneous equations
of n
is non-singular,
[/
K].
it
is
equivalent to
Then X
When A
/.
reduced by row transformations only to
is
is a solution of the
X = L is a second solution
non-singular, K = L, and the solution is
Suppose next that
4.
unknowns has a unique solution
in n
0.
[A H] is reduced to
Since A is
[CHAP. 10
/,
suppose
system.
AK
of the system; then
and
AL
H, and
AK
AL.
unique.
Derive Cramer's Rule.

Let the system of non-homogeneous equations be
(1)
il*l + a^jAiQ +
+ ain*n
h^
uqiXi + 022*2 +
+ 2n*n
^z
"ni*i + an2*2 +
+ "nn^n
^n
Denote by A the coefficient matrix
[a.
let a^,-be the cofactor of
and
the last equation
tion of (J) by ttn, the second equation by Ksi
S
X
which by Theorems
ai^di-^x-^ +
ai20iii% +
and XI, and Problem

hx ai2
22
^^2
^n
""no.
10,
Chapter
.S
1=1
i=i
i=l
3,
''is
Oqi
'^23
/12
\A\.x^
and add.
n
Multiply the first equa-
We have
ainîlXn
1
=1
reduces to
"m
'^2n
so that
x-i
^11j-
^1
"n
n2 and sum to obtain
Next, multiply the equations of (i) respectively by aî, îi

"11 ^1
in
ttni.
by
"m
"2n
so that
Continuing in this manner, we finally multiply the equations of
(i)
*<,
respectively by a^n. 2n
(^,71
and sum to obtain

Oil
"l,n-i "1
"21
"2,n-i ^2
so that
x^.
'ni
2Xi +
5.
Solve the system
X2 + 5 *3 +
*^4
X2 ~ 3x3 - 4*4
Xi +
:4
3 Xi + 6 Xj - 2 aig +
+
+
3x4
2a;3
2x1
2x:2
-1
using the inverse of the coefficient matrix.
Solution:
The inverse
of
15
1-3-4
3
2
6-2 1
2-3
2
120
-120'
120
-69 -73 17
120 -15 -35 -5
.24
80
40
-40.
Then
CHAP.
LINEAR EQUATIONS
10]
"120
120
.24
(See Example
Solve
40
1/5
-40.
"
-1
80
4/5
2.
3.)
s +
*4
i + 3*2 + 2xq +
4a;4
Xj^
Xl +
6.
5"
-120"
120
-69 -73 17
-15 -35 -5
81
^2 +
Xg -
2Xi
Solution:
\A
U\
-1
o"
1
'X/
1,
6 =
x^ =
What can be said
7.
1
''\J
i 1
The complete solution of the system is

we may obtain exactly n-r = 4-2
taking a =
0.
and then a =
X2 = -2, *3 =
3,
1,
11110
13
= 2
6 =
X4 =
-i
3
2
-â + 16, ^^
x^ =
is 2,
first
-2 -1 -3
11110
1
o"
-! - |6, xs=a, x^ = h. Since the rank of

One such pair, obtained by
linearly independent solutions.

is
and
of the pair of solutions obtained
x^ = -1, x^ = -3, ^3 =
by taking a =
b =
and a =
3,
:x:4
b =
37
Prove: In a square matrix A of order n and rank n-1, the cofactors of the
elements of any two rows
(columns) are proportional.

Since
AX
1^1
(A'X =
=0,
the cofactors of the elements of any row (column) of
are a solution X-^ of the
system
0).
Now the system has but one linearly independent solution since A (A^) is of rank n-l.
Hence, for the
cofactors of another row (column) of A (another solution X^ of the
system), we have X^ = kX^.
8.
Prove:
If
/"i,
f^
f^ are
m<n
linearly independent linear forms over
in n variables, then the
linear forms
^j
are linearly dependent if and only if the
The g's
are linearly dependent
if
mxp
and only
0-=1.2
^^'ijfi-
p)
matrix [5^] is of rank r<p.
if
there exist scalars a^,a^
a.
in
not all zero, such
that
lgl + "2g2 +
+apgp
ai
^^Hxfi
.2
1=1
si^fi
.|
2
J=l
a;:S--)f.
J
^J
...
ap
.2
sj^fi
LINEAR EQUATIONS
82
[CHAP. 10
Since the /'s are linearly independent, this requires
j?/i^v
Now, by Theorem IV, the system of m homogeneous equations
trivial solution
9.
Suppose A =
[a-
apV
= [a^.a^,
] of order n
AB
order n such that
and
if
only if [^^j] is of rank
Show
is singular.
sider any one of these,
that there
in p unknovras
say AB^
1,
m)
always exists a matrix R =
Find
of
AB^
Then, by hypothesis,
B.
^65
..
AB^
ii&it + "iQ^'st +
"m ^nt
ni*it + n2^2t+
"nn^nt
Similarly,
AB^
= 0,
AB^
= 0,
+ *3
2^:5
...
(c)
3*'4
i>nt
Xj^
X^ +
/4ns.
(a)
! =
(b)
xi =
(d) Xi
11.
Find
\2x^
+ 5
1 + 2a
-7a/3 +
% = 1,
Xq +
Xg
2 Xg
3C2
:;3
x-i
1 x^
+ %2 +
^1 + %2 +
ajj
x:2
X\
-^
Xq
X-i
Xq +
~ 2%3
7^3
->=3
+ %4
-^S
x^ +
:>:4
-4
+ x^
Xr,
= o, X3 = b, x^ =
4a/3 =
5/3, Xg
2x^
*1 + 2*2
+ 3x3
(c)
<
12*1
2*1 +
+ 5*2 + 6*3
3*i
2x-i
(.h)
(6) *i =
(d)
2*1
*2
5
*i
*3
J.
*3
- 4*2 + 5*3
*i = -3o, X2 =
3,
"*'
+ 2*2 +
0,
2*1
*g = a
= a
*o =
*4
3*3
X2 + 3*3
4*1
X2 + ^x^
3*1 + 22 +
*i
(o)
of
Con-
has solutions other
all non-trivial solutions of:
(a)
/4ns.
-^
(d)
+ 3c,
17/3,
x^ +
2%i + 5%2
x-^
(b)
= 0.
have solutions, each being a column of S.
x-^
Xj_
?^
all solutions of:
{a)
[h^j]
= 0, or
10.
has a non-
s^j xj -
r<p.
Since the coefficient matrix ^ is singular, the system in the unknowns h^t,^'2t
than the trivial solution.
0.
B be the column vectors
Let Bi.Bg
(i
+ "p'iP
î^-ii
*3
+ 2*3 +
*3
^ *47*2 4*3 5*4
3
*2
11*5
1*3
5*4
CHAP.
12.
LINEAR EQUATIONS
10]
Reconcile the solution of 10(d) with another
A =
13. Given
6_
the solutions of
14.
15.
AX
AX
= c, x^ = d, x^ =
x-i
of rank 2
AB
such that
^c -
= 0.
Hint.
x^ =
c +d.
o
Select the columns of
B from
= 0.
Show that a square matrix

Let
find a matrix
83
and only
is singular if
if its
rows (columns) are linearly dependent.
be a system of n homogeneous equations in n unknowns and suppose A of rank r = n-1.

[aî, a^j
OLinV of a row of ^ is a solution of AX = 0.
Show
that any non-zero vector of cofactors
16.
Use Problem
ixg
3%
\2x-^ + 5%2 + 6*3

Xg
^1
Hint.
third
To
row of
Ans.
17.
15 to solve:
- 2%2 +
(b)
the equations of (o) adjoin

"l
-2
+ 2x^
y'ix^
4%2
+ 2%3
Xg
Oxi + Ox^ + 0%3
(c)
%2 + 6 Xg
+ 3*2 + 4
2.^1
1 2xi
and find the cofactors of the elements of the
a""
-27a. X2
(a) xi =
/ 2xj_
9a
= 0, X3 =
[3a,
or
-a]',
0,
(6) [2a,
-7a, -17a]',
[lla, -2a, -4a]'
(c)
Let the coefficient and the augmented matrix of the system of 3 non-homogeneous
equations in 5 unknowns
be of rank 2 and assume the canonical form of the augmented
matrix to be
AX =H
1
1
613
614
fcj^g
ci
623
624
625
C2
_0
with not both of
AX
tion of
ci, C2
equal to
Then choose
= H.
First choose
0.
tain other solutions X^,Xg, and X^.
18. Consider
that
for (O,
19.
20.
IS a
AX
if
Hint. Follow
If ^ is an m xp
Use Theorem VI.
Prove:
x^ = xg =
1,
AX
0,
s,X^ + s^X^ + s^X^ + s^X^
and only
complete solution of
Prove: Theorem VI.
Hint.
21.
i^
the linear combination
is a solution of
X3 = x^ = x^ =
and obtain X^ = [c^, c^, 0, 0, o]' as a solualso X3 = x^ = 0, x^ = 1 and X3 = x^ = 0, Xg == 1 to obshow that these 5-2 + 1=4 solutions are linearly independent.
X3 =
if
s, +.2 +-3 +^4 =
(i)
r^
of the solutions
Thus, with s
of
Problem 17
s,. .3,
Show
.4 arbitrary except
= H.
Problem 17 with c^ = c^
matrix of rank
1-
and S
is a p
= 0.
matrix of rank
r^
such that
AB
= 0, then
r,
^-<
r^
n
f
Using the 4x5 matrix
.1 = [a^^-]
of rank 2, verify: In an x matrix A of
rank .. the r-square determinants formed from the columns of a submatrix consisting
of any r rows of A are proportional to the
r-square
determinants formed from any other submatrix consisting
r
Hint.
(7 =
of
rows of A
two rows are linearly independent so that a^j = p^.a^j
+Psâ,.. a^.-pâ^,f^
ij +p42a2,ha2 2j.
j
Evaluate the 2-square determinants
j
Suppose the
1.2
5).
first
"17
ai5
\'^3q
agg
a^q
025
and
"4(7
"AS
22. Write a proof of the theorem of Problem 21.
'
^'"
'''''"''
''
f?ctors''hcJ?'"
(a)
where (h,i,j,h =
1,
n).
*'' ""''"'''' '""'"'
a^^-a^^
a^J^(x^,
'" ' '^"' ""'

(t)
*^'" "'' '''^'"^ ^^1^""^ ^"-""S
a^âJJ
a^jUji
"^
LINEAR EQUATIONS
84
11114
123-4
21126
10
24.
Show
that
system
m>n
-3
-,,.
row equivalent to
ij
01000
00100
00010
00001
From B =[A H]
infer that the
[_0
unknowns has 5 linearly independent equations. Show that a system of

unknowns can have at most re +1 linearly independent equations. Show that when
of 6 linear equations in 4
linear equations in n
there are
25. If
IS
32-1-13
122-2
[CHAP. 10
ra
AX =H
the system is inconsistent.
is consistent
and of rank
r,
for
what set of
variables can one solve?
equations in n unknowns with coeffi26. Generalize the results of Problems 17 and 18 to m non-homogeneous
and the augmented matrix of the
coefficient
If
the
r
to
prove:
cient and augmented matrix of the same rank
Xn-r-n are
rank
r and if X-^^.X^
have
unknowns
=
in
n
equations
non-homogeneous
AX H of m
system
linearly independent solutions of the system, then
siî + 51^2 +
^n-r+i^n-r+i
n-r+i
where
1,
a complete solution.
is
i=i
i and
27. In a four-pole electrical network, the imput quantities
2 and
Iq
by
E^
Show
6/2
"11
cE^ + dlQ
_/lJ
oEq +
'eI
h.
'b
I^,
and
I^,
and E^
AX
29. Solve the set of linear forms
Now
write
down the solution
of A'
'l
-1
X2
_s_
1"
0,
71
matrix with
72
AX
= K.
Show that the
for the x^ as linear forms in the y's.
p.
= E^,
(i
1,
^n^%
n).
where îs the
Identify the matrix [S^, Sg
and let S^ be a solution of AX - E^, {i -1,2,.

and whose other components are 0. If K = \k^. k^.
k^Si + A:2S2
is a solution of
have a unique solution.
m<n
m-vector whose ith component is
>-
"n"
'xi
A be n-square and non-singular, and let S^ be the solution of AX

n-vector whose ith component is 1 and whose other components are 0.
Let 4 be an m X
'e2
30. Let
31.
Let the system of n linear equations in n unknowns 4X = H, H ^

system AX = K has a unique solution for any n-vector K ^ 0.
and
that
Solve also for 2 and
28.
are given in terms of the output quantities
m).
,
S^].
where E^ is the
k^' show
,
that
Chapter
11
Vector Spaces
UNLESS STATED OTHERWISE,

played,
we
all
now be column vectors. When components are disThe transpose mark (') indicates that the elements are
vectors will
shall write [xj^.x^
a^]'.
to be written in a column.
A
them
if
F is said to be closed under addition if the sum of any two of

Similarly, the set is said to be closed under scalar multiplication
set of such 7z-vectors over
is a vector of the set.
every scalar multiple of a vector of the set is a vector of the set.
Example
1. (a)
The set of
all vectors [x^. x^. x^Y of ordinary space havinr equal

components (x-^ = x^^ x^)
closed under both addition and scalar multiplication. For, the sum
of any two of the
vectors and k times any vector (k real) are again vectors having equal
components.
is
(6)
The set
of all vectors
[x^.x^.x^Y of ordinary space
is
closed under addition and scalar
multiplication.
VECTOR SPACES. Any
set of re-vectors over
plication is called a vector space.

linear combinations
<"!)
Kî
is a vector
F which
Thus,
if
is
X^, X^
KX^
closed under both addition and scalar multiX^ are n-vectors over F, the set of all
KX^
(kiinF)
space over F.
For example, both of the sets of vectors (a) and (b) of Example
1 are
Clearly, every vector space (ll.l) contains the zero
re-vector while the zero
re-vector alone is a vector space. (The space
(11.1) is also called a linear vector space.)
vector spaces.
The
totality VîF) of all re-vectors over
is called the re-dimensional vector
space over
F.
SUBSPACES. A
set V of the vectors of V^(F) is called a subspace

of V^(F) provided V is closed under addition and scalar multiplication. Thus,
the zero re-vector is a subspace of F(F)- so
also
IS V^(F) Itself. The set (a) of Example
1 is a subspace (a line) of ordinary
space. In general
If X^, X^, ..., Z^
belong to V^(F), the space of all linear combinations
(ll.l) is a subspace of
vector space V
vided (a) the Xi

vectors X^, X^
lie in
X^_
is
said to be spanned or generated by the re-vectors

(b) every vector of F is a linear
combination
V and
X^,
X^
(11.1).
X^ proNote that the
are not restricted to be linearly independent.
Examples. Let F be the field R of real numbers

so that the 3-vectors
X^
^3= [1.3,2]' and X^= [3,2,1]' lie in ordinary space S =
[i.i.lY X^
Any vector
V^R).
s can be expressed as
85
2 i]'
\i
cY
\a
b
,
ot
VECTOR SPACES
86
[CHAP. 11
3y4.
+ 2X2 + Sys +
274.
y2 +
yi +
+ 72^2 + ys'^s +
Ti ^1
yi
y4-^4
3y2 +
yi +
ys
2y3
Ja
since the resulting system of equations
yi +
y2
yi +
(i)
yi
Thus, the vectors
is consistent.
ys
2y2 +
3y2
Xj^,
3x4
3ys + 2y4
2y3
Xj. Xg,
y4-
X4 spanS.
and ^2 are linearly independent. They span a subspace (the plane

real numbers.
S which contains every vector hX^ + kX.^. where /i and k are
The vectors
of
tt)
X-y
The vector X^ spans
hX^. where
a subspace (the line L) of S which contains every vector
A is a real number.
See Problem
BASIS
1.
maximum number of linthe dimension of a vector space V is meant the

number of linearly inminimum
same thing, the
early independent vectors in F or, what is the
is considered as
space
ordinary
geometry,
dependent vectors required to span V. In elementary
it as a
considering
been
have
we
Here
(a, 6, c).
a 3-space (space of dimension three) of points
is of
line
L
the
and
is of dimension 2
3-space of vectors ia,h,c \. The plane n of Example 2
AND DIMENSION. By
dimension
A
we
1.
vector space of dimension
consisting of 7i-vectors will be denoted by F(F).
When
n,
shall agree to write p;(F) for %\F).
basis of the space. Each veclinearly independent vectors of V^(F) is called a

of this basis. All bases of
vectors
the
combination of
tor of the space is then a unique linear
independent vectors of the
linearly
r
V^(F) have exactly the same number of vectors but any
set of
space will serve as a basis.

Example
The vectors
3.
X^.
Xr,.
X^
of
X-y
71-^1
Example
span S since any vector [a,
+ j^Xq + ygâ
yi
ys
ys
yi
2y2
3ys
3X2
2ys
yi +
ys =
,^ + 3X3 =
yi + 2y2
3X2 + 2xs =
!yi
xi
The
resulting system of equations
Xi+
y2 +
b. c ]'
of S can be
unlike the system
expressed
i),
has a u-
are not a
The vectors X^.X^.X^ are a basis of S. The vectors X^.X^.X^
X^. X^
set
the
basis
is
whose
tt
Example
2,
of
(Show this.) They span the subspace
nique solution.
basis of S
Theorems I-V
of
Chapter 9 apply here, of course.
In
particular,
Theorem IV may be
re-
stated as:
I
If
X X^
..
X^
the rank of the raa:ro matrix

Unearly independent vectors may be selected. These
are a set of re-vectors over
F and
if r is
of their components,' then from the set r

the remaining m-T vectors lie.
T vectors span a VîF) in which
See Problems
2-3.
CHAP.
VECTOR SPACES
11]
Of considerable importance are

II.
If
^m4.2. .
JYi,
^n
^2,
...,
Zm
m<n
are
linearly independent n-vectors of
are any n-m vectors of VîF)
independent set, then the set
87
X^
X^,
which together with
X.^,
X^
VîF) and
J^+j^,
if
X^ form a linearly
is a basis of VîF).
See Problem
III.
If
Z^.Zg,...,! are m<7i linearly independent -vectors over F, then the

p vectors
m
î
are linearly dependent
IV. If
4.
Zi,
.2, ...,
p>m
if
or,
i^'ijî
when p<m,
if
(/=l-2
P)
[s^,] is of rank r<p.
are linearly independent re-vectors over F, then the vectors
n
Yi
=
7
are linearly independent if and only
IDENTICAL SUBSPACES.
only
is
If
X(F)
,V^(F) and
each vector of X(F)

a subspace of the other.
if
if
a^jXj
"
(i
= 1,2
re)
[a^-] is nonsingular.
are two
subspaces of F(F), they are identical if and

and conversely, that is, if and only if each
is a vector of ^V;[(F)
See Problem
SUM AND INTERSECTION OF TWO SPACES.

sum
is
meant the totality of vectors
5.
Let V\F) and f3^) be two vector spaces. By their

where X is in V^(F) and Y is in VîF). Clearly, this
X+Y
is a vector space; we call it the sum space

V^îF).
The dimension s of the sum space of two
vector spaces does not exceed the sum of their dimensions.
By
the intersection of the two vector spaces is meant the

totality of vectors common to the
Now if Z is a vector common to the two spaces, so also is aX;
likewise if X and
y are common to the two spaces so also is aX^bY. Thus, the intersection of two spaces
is a
vector space; we call it the intersection space V\F).
The dimension of the intersection space
of two vector spaces cannot exceed the smaller of
the dimensions of the two spaces.
two spaces.
V. If two vector spaces FV) and V^(F) have

V^\F) as
section space, then h + k = s + t.
sum space and V^(F) as
inter-
Example 4. Consider the subspace 7f, spanned by X^ and X^ of

Example 2 and the subspace tt^ spanned
by Xs and X^. Since rr^ and tt^ are not identical (prove this)
and since the four vectors span
S, the sum space of tt^ and tt^ is S.
Now 4X1 - X2 = X^: thus, X^ lies in both tt^ and 7t^. The subspace (line
L) spanned
by X^ IS then the intersection space of 77^ and 77^
Note that 77^ and 77^ are each of dimension
2, S IS of dimension 3, and L is of dimension 1.
This agrees with Theorem V.
.
See Problems 6-8.
NUIXITY OF A MATRIX.
For a system of homogeneous equations AX = 0, the

solution vectors X
constitute a vector space called the null space of A.
The dimension of this space, denoted by
^
A'^
IS called the nullity of A.
,
Restating Theorem VI, Chapter

VI.
If
A has
nullity N^
10,
then
we have
AX
has N^ linearly independent solutions X^. X^
VECTOR SPACES
88
Xti
A
AX
such that every solution of
[CHAP. 11
is a linear
combination of them and every such
linear combination is a solution.
basis for the null space of A is any set of N^ linearly independent solutions of
AX
0.
See Problem
Vn. For an
mxre matrix
of rank rj
(11.2)
rA
SYLVESTER'S LAWS OF NULLITY.
AB
>
of order
'^
Nab > Na
(11.3)
Nab
El = [1,0,0
N/^,
and respective ranks q and
ra
rg
the
satisfy the inequalities

''AB
BASES AND COORDINATES. The
Nji
A and B are
If
rank and nullity of their product
and nullity
9.
<
- "
'B
Nab > Nb
Na + Nb
See Problem
10.
ra-vectors
1]'
En = [0,0,0
0]',
E^ = [0,1,0
0]',
are called elementaiy or unit vectors over F. The elementary vector Ej, whose /th component
The elementary vectors E^, E^
constitute an
is 1, is called the /th elementary vector.
important basis for f^^Cf).
Every vector
uniquely as the
^nl' of 1^(F) can be expressed
[%,%
sum
1=1
Of the elementary vectors.
X
X
XxE-^
xiEi
The components %,
2-^2
x^ oi
x^
are
now called the coordinates of

we shall assume that a vector
relative to the E-basis. Hereafter, unless otherwise specified,

is
'^nEr,
given relative to this basis.
Zn be another basis of 1^(F).
Let Zi, Zg
in F such that
exist unique scalars
Then there
%,
03
a^
X
i
These scalars
JY^
= [oi, og
is the matrix
If
02
Zg +
o are called the coordinates of

a^Y, we have
Examples.
01,05
(11.4)
where Z
a.^Z.^
aiZi
Zi =
Z^ = [l,
2,
-1
]',
Z3
= [l, -1,
dj^Zj^
relative to the Z-basis.
-1
Zi,
]'
Writing
Z-Xz
Zn]Xz
[Zi.Zg
whose columns are the basis vectors
[2, -1, 3]',
=i
is
Z.
Zg
a basis of
Fg
(F) and
Xz
[1.2.3]'
is a vector of Vg(F) relative to that basis, then
[Zi, Zg, Zg]A:^
-1
-1
relative to the -basis.
-1 -1
"1
[7,0,-2]'
-2
See Problem 11.
CHAP.
VECTOR SPACES
11]
Let
W^,
be yet another basis of f^(F).
(11.5)
From
(11.4) and (11.5),
=
=
X^
(11.6)
where P =
JV^
IT
Suppose
WX^
\_\,\
89
Zj^
-r
K\' so
= [61,^2
that
Xy
and
^'^.Z-X^
Xig
PX,
IF"^Z.
Thus,
VIII.
If
a vector of f^(F) has coordinates Z^ and X^ respectively relative to two bases
of P^(F), then there exists a non-singular matrix
given by (11.6) such that
Xf/
P determined
,
solely by the two bases and
= PX^
See Problem
12.
SOLVED PROBLEMS
1.
The
set of all vectors
A'
= [%,
x^, Xg,
x^Y,
where x^ + x^ + x^ + x^ = Q
is
a subspace V of V^(F)
since the sum of any two vectors of the set and any scalar multiple of a vector of the set have
components whose sum
2.
is zero, that is, are
vectors of the set.
Since
is of rank 2, the vectors
X.^
= [1,2,2,
1 ]',
X^ = [3,4,4,3]', and X^ = [1,0,0,
are linearly dependent and span a vector space
(F).
Now any two of these vectors are linearly independent; hence, we may take
and Xg as a basis of the V2{F).
14
13 12
12
2
3.
Since
1 ]'
X-^
and X^, X^ and
A:g.
or
X^
is of rank 2, the vectors
^"1
= [1,1,1,0]',
A's
= [4,3,2,-1
]',
A'g
= [2,1,0,-1]',
-1 -1 -2
and
^"4
= [4,2,0,-2]' are linearly dependent and span a V^(F).
For a basis, we may take any two of the vectors except the pair Xg, X^.
4.
The vectors
X^, X^, Xg of
Problem
2 lie in V^(F).
Find a basis.
For a basis of this space we may take X^,X^,X^ = [l.O.O.O]', and Xg = [o, 1,0,0]' or X^.X2.Xg =
since the matrices [X^, X2, X^.Xg] and [X^. X2.Xg. Xy] are of rank
[1.2,3.4]'. and X; = [1,3,6,8]'
4.
5.
[CHAP. 11
VECTOR SPACES
90
Let Zi = [l,2,l]', Z2 = [l,2,3]', A:3 = [3,6,5 ]', Y^ = [0,Q,lY, i'2=[l.2,5]' be vectors of

Show that the space spanned by X-^, X^, Xq and the space spanned by Y.^, Y^ are identical.
we
First,
note that X^ and X^ are linearly independent while
of dimension two. say
2Zj^
X^..
Thus, the X^ span a space
Also, the Yi being linearly independent span a space of dimension two. say
iF|(''').
Thus, any vector

iX.^, Y^ = 2^2- ^li î = î^ - 4^1. X^ = Y^- 271.
(50 + 26)^2 - (2a + b)X.^ of iPg^CF) and any vector cXj^ + dX^ of iff(f)
+ d)Y2 - (4c + 2<i)Ki of QVi(F). Hence, the two spaces are identical.
Fi =
Next,
2lg^(F)
(c
Xq
6. (a) If
kXz -
is
a vector
= [%, 2.%]' lies in the Vg(F) spanned by
-2% +
Vs(F).
X^ =
aY-^
fcK,
1,-1,1]' and X^ = [3,4,-2]',
of
is a vector
then
5X2 + IXg
(b) If
= [xi,x^,X3,x^Y lies in the Pi(F) spanned by X^ = [1,1,2,3]' and X^ =
3Ci
-2
%4.
xi
4.
%
1
Since
2.
= - 2*1 + 4% -
this requires
0,
then
1,0,-2,1]',
and
%+
2*2 -
s;*
These problems
linearly independent
7.
rank
is of
verify: Every ^(^ ) may be defined as the totality of solutions over F of a system
homogeneous linear equations over F in n unknowns.
of n-k
two vector spaces Vn(F) and I^(F) have Vn(F) as sum space and VîF) as intersection
space, then h + k = s + t.
Prove:
If
Suppose
s
= h; then Vr!XF) is a
h+k. The reader
will
show
subspace of (^^F) and their sum space

that the
same
Suppose next that t<h. t<k and let X^^.X^

Yh so that X1.X2
^t.i't-n
A:i,A:2
^t.Zt+1
2fe span I^'^F).
^t^-i, yt-+2
Now suppose
is true if
is 1^' itself.
X^ span
=h
and
Then by Theorem H there exist vectors

Z^ so that
Zj^^. Z^+j
V^(F).
span l^(F) and vectors
Yh
X
1=1
"iXi
.S
a^Yi
t=t+i
ft
i=i
i=t+i
biZi
i=t+i
i=t+i
The vector on the

V^(F). But X^.X^
left
belongs to P^(F),and from the right member, belongs also to ^(F); thus
X-t
span Vn(Fy, hence,
from (11.4),
2
i=i
a^+i = at+2 =
"f,
it
belongs
= 0.
Now
there exist scalars a's and 6's such that
(11-4)
to
Thus, s=k.
= k.
"iXi
2;
b^Z^
i=t+i
But the X's and Z's are linearly independent so that a^ = 05 = = t =

the ^f's.y-s, and Z's are a linearly independent set and span ^(F). Then
''tî =
s
t*^"^,
*t+2 = = ^fe =
as was to be proved.
=h + k-t
CHAP.
8.
VECTOR SPACES
11]
Consider ^FsCF) having X^ = [l,2,2Y and Z2 =
1,1,1
]'
91
as basis and jFgCF) having 71 = [3,1,2]'
113
2 110
1
and
Fj
1.0,1
]'
as basis. Since the matrix of the components
12
space
is V,(F).
Prom
h + k
As a basis, we may take

= s
t.
X-^,
and
X^,
Y^
To
the intersection space is a VsiF).

2
sum
is of rank 3, the
find a basis,
we equate
linear combinations
of the vectors of the bases of iFg(F) and s^'sCF) as
- 3e
take d =
aX;]^
9.
for
convenience, and solve
+ 6^2 = [-1.-2/3.-1/3
Determine a basis
]' is
for the null
^2a
6-
( 3a
- 2e
a basis of the intersection space.
space of A
Consider the system of equations
AX
a = 1/3,
obtaining
113
10
113
The vector [3,2,1]'
basis for the null space of
.4
c =
-2/3.
Then
is also a basis.
x^ =
which reduces
to
\x2+
A
6 = -4/3,
Xs + 2x^ =
independent solutions [1.2,0,-1]' and [2,l,-l,o]'
is the pair of linearly
of these equations.
r >
10. Prove:
AB
Suppose
r
A
- n.
first that
A has the form
the remaining rows are zeros.
PAQ
By Problem
r,
rows of
Chapters, the rank of
10,
the special
Solution
(a).
Find
AB
AB
is
are the first
^45
>
+ %
'k
r.
rows of B while
"
aZ^ + 6Z2 +
Thus
coordinates relative to a new basis Zi =
its
cZr^.
that is.
+ h
c = 2.
[l 1 0]'
Write
1
case when
Let X=[l,2.lY relative to the S-basis.

Z2 = [1,0,1]', and Zg = [1,1. l]'.
(i)
first
Suppose next that A is not of the above form. Then there exist nonsingular matrices P and
Q such that
has that form while the rank of PAQB is exactly that of AB (why?).
The reader may consider
11.
Then the
relative to the Z-basis,
we have X^
!a
[0,-1,2]'
Then
+ i + c =
a
b
and a =
0,
6 =
1,
VECTOR SPACES
92
Rewriting
Solution (6).
(i)
as
-1
X^
Z X
-1
Let X^ and
be the coordinates of a vector

Z3= [1,1,1]' and f4 = [l,l,2]',
X-^
Z2=[l,0,l]',
trix ? such that X^ = PX^
Then
Zg]
\_Z^, Z^,
W'^Z
1
1
-1
_ _
with respect to the two bases
ff^
= [2,2,1
]',
Ifg
= [1,2,2
]'.
7,^
= [1,1,0]'
Determine the ma-
Here
[0,-1,2^
2
1
.
12.
11
we have
{Z-^.Z^.Z^^X^ = ZXg,
[CHAP.
'l
l'
2-3 2
0-1
2
and
-3
by (11.6).
0-3
13.
Let
[x-i^.
x^. x^.
X4,y
be an arbitrary vector of Vi(R). where
R denotes
the field of real numbers.
Which of the
following sets are subspaces of K^(R)'?

(a)
All vectors with Xj_= X2 = X3 = x^.
(6) All vectors

(c) All
with x^ = x^.
vectors with %4 =
Ans. All except {d) and
x^=2x^.
(d) All vectors with
x-^
(e) All vectors with
x^.Xr,.Xs.xj^
and [2,3.3,2]' are a basis of the
fi^(F) of
Problem
Determine the dimension of the vector space spanned by each set of vectors.
]'
integral.
2.
15.
1.1.1.1
0.
Show
(e).
14.
that
Select a basis for each.
[1.1,1.1]'
[l.l.O.-l]'
[1,2,3.4.5]'
(b)
[5.4,3,2,1]',
(a)
16. (a)
(6)
[3,4,5,6]'
^'^
[2.3.3,3]'
[1.1,1.1,1]'
Ans.
[1,2,3,4]'
(a), (b), {c).
[1.2,3,4]'
[1,0,-1,-2]'
r= 2
Show that the vectors X-^ = [l,-l,l]' and X^ = [3,4,-2]' span the same space as Y^ = [9,5,-1 ]' and
72= [-17,-11.3]'.
Show that the vectors X^ = [ 1,-1,1 ]' and A'2 = [3,4,-2 ]' do not span the same space as Ti = [-2,2,-2]'
and K, = [4,3,1]'.
n. Show
that if the set
X^.X^
Xfe is a basis lor Vn(F). then
sented uniquely as a linear combination of
X-^,
Hint.
Assume
51
aiXi
biXi-
X^
X^
any other vector Y of the space can be repre-
CHAP.
18.
VECTOR SPACES
11]
4x4
Consider the
93
matrix whose columns are the vectors of a basis of the Vi(R) of Problem 2 and a basis of
3. Show that the rank of this matrix is 4; hence. V^R) is the sum space and l^(R), the
the \i(R) of Problem
zero space, is the intersection space of the two given spaces.
19.
Follow the proof given
20.
Show
Problem
in
that the space spanned by [l,0,0,0,o]', [0,0,0,0,1 ]', [l.O,l,0,0]', [0,0,1,0,0]' [l,0,0,l,l]' and the
space spanned by [l,0.0.0,l]', [0,1,0,1,0 ]', [o,l,-2,l,o]', [l,0,-l,0,l ]', [o,l,l,l,o]' are of dimensions
4 and 3, respectively. Show that [l,0,l,0,l]' and [l,0,2,0,l]' are a basis for the intersection space.
Z^=
21. Find, relative to the basis

(a)
[l.l.o]',
Ans.
[1,0, l]',
(b)
(a) [2,-1,0]',
Ans.
Let X^ and
trix
Zi=
PX^
[l/3, 1/3,
]'
(c) [0,0,l]'.
[-6,7,-2]',
(6)
that Xj^ =
(c)
]',
Zi=[o,l,o]', Z2=[i,l,l]', Z3=[3,2,l]' the coordinates of the vectors

(c)
[-1/2, 3/2, -1/2
]'
with respect to the given pair of bases.
[1,0,0]',
Z2=[i,o,l]',
Z3=
[1,1, il-
= [0,1,0]',
[^2= [1,2,3]',
If'3=
[1,-1,1]'
Zi = [0,1,0]',
^^^
1^1
(a)
2j
is a solution of
AX
= [l,1.0]'.
2^
23
= [1.1,0]',
1^2= [1,1.1]',
= [1.2.3]'
1^3= [1,2.1]'
-2"1
-1
!fi
ij
n
=
Ej
(j = 1,2
n).
then
hjPj
is a
solution of
AX
= H.
where H =
KV-
[''1.^2
Hint.
(6)
24. Prove: If Pj
41
Ans.
Determine the
ea)
25.
-1
be the coordinates of a vector
X^^
P such
[l.l.l]'.
(6) [4/3, 1/3,
(b) [1,-3,5]',
(a) [-2,-1,1]',
[l,1.2]', Zg = [2.2,l]', Zg = [l,2,2]' the coordinates of the vectors
(c)
[-1/3,2/3,0]',
(a)
22. Find, relative to the basis
23.
Chapter 10, to prove Theorem HI.
8,
hÊj^ +
hÊ^+
+hnE^.
The vector space defined by all linear combinations of the columns of a matrix A
is called the column space
of A. The vector space defined by all linear combinations of the
rows of A is called the row space of ^.
Show that the columns of AB are in the column space of A and the rows of AB are in
the row space of
26.
Show
that
AX = H a system of m non-homogeneous
H belongs to the column space of A
equations in n unknowns, is consistent
the vector
27. Determine a basis for the null
space of
Ans.
28. Prove:
Hint:
[1,-1,-1]',
(a)
(a)
r^g
5.
r^g
Consider n~r^g
29. Derive a procedure for
Problem
1,1,-1, -i
N^,>N^. N^^>N^
(a) /V^g
(b)
(6)
<
and only
if
the
1111
1-1,
(a)
if
fi.
(6)
12 12
3
[l, 2,-1, -2]'
]',
(b)N^,<N^^N^
r^
and
rg
using the theorem of Problem 10.
Problem 16 using only column transformations on A = [X^. X^, y^ Y^].
Then resolve
chapter 12
Linear Transformations
DEFINITION. Let
%]' and Y =
[x,., x^, ....
JnY ^^ ^^ vectors
lyx. y^.
ordinates being relative to the same basis of the space.
of l^(F), their co-
Suppose that the coordinates of
are related by
yi
'^ll'^l
"^
^12 "^2
"^
^^Yi J^
df^Yj.^ri
(12.1)
AX
or, briefly,
where A =
[a^.-]
Then (12.1) is a transformation T which carries any vector X

Y of the same space, called its image.
is over F.
of
V^(F) into (usually) another vector
(a)
(fe)
If
(12.1) carries X^ into F^ and X^ into Y^, then
it
carries kX^ into ^y^, for every scalar k, and
it
carries aX^ + bX^ into aY^ +
every pair of scalars a and
feFg. for
b.
For this reason, the
transformation is called linear.
Example
Consider the linear transformation Y =
1.
(a)
(6)
The image
of
A"
= [2,0,5]' is
The vector X whose image
112
Since
BASIC THEOREMS.
if
then
I.
If in (12.1),
Y =
12
13
[a^j.a^j
5
3
is
= [\,Q
"nf]'-
'l
13
in ordinary
A"
space Vq(R).
'12"
27
[12,27.17]'.
17
= [2,0.5]' is obtained
10
10
AX
by solving
'l
'x{
*2
.1
3_
3.
'2'
=
5
13/5
11/5
[13/5,11/5,-7/5]'.
-7/5
0]'='i then
Hence,
Y =
an,
linear transformation (12.1) is uniquely determined
Ogi,
...,
a^J'
and, in general,
when the images (Y's)
of the
basis vectors are known, the respective columns of A being the coordinates of the images
See Problem
of these vectors.
94
l.
CHAP.
12]
95
if the images of distinct vectors Xi

Otherwise the transformation is called singular.
linear transformation (12-1) is called non-singular
are distinct vectors Y^.
II.
linear transformation (12.1) is non-singular if and only if A, the matrix of the
transformation, is non-singular.
in.
See Problem
2.
non-singular linear transformation carries linearly independent (dependent) vec-
tors into linearly independent (dependent) vectors.
See Problem
3.
Prom Theorem HI follows

k
IV. Under a non-singular transformation (12.1) the image of a vector space Vn(F) is a
vector space VjJ,F), that is, the dimension of the vector space is preserved. In particular,
the transformation is a mapping of ^(F) onto itself.
When A
is non-singular, the inverse of (12.1)
A'^y
carries the set of vectors Y^, Y^, ...,\ whose components are the columns of A into the basis
vectors of the space. It is also a linear transformation.
V. The elementary vectors ^ of \{F) may be transformed into any set of n linearly
independent n-vectors by a non-singular linear transformation and conversely.
VI. If
CZ
IT
into
Y = AX carries a vector X into a vector F, if Z = BY carries Y into Z, and if

Z into W, then Z = BY = {BA)X carries X into Z and IF = (CBA\X carries
carries
IF.
VII. When any two sets of re linearly independent re-vectors are given,
there exists a
non-singular linear transformation which carries the vectors of one set into the
vectors of
the other.
CHANGE OF
BASIS. Relative to a Z-basis, let 7^ = AX^, be a linear transformation of ^(F). Suppose

changed and let X^ and Y^ be the coordinates of X^, and Y^ respectively relative to the new basis. By Theorem VIH, Chapter 11, there exists
a non-singular matrix P such
that X-^ = ?X^ and Yy, = PY^ or, setting ?~^ = Q, such that
that the basis is
X^
Then
Y^
Q-%
QX^
=
and
Q-^X,
Y^
QY^
Q-ÂQX^
BX^
where
(12.2)
Two
Q-ÂQ
matrices A and B such that there exists a non-singular matrix

We have proved
for
which B = Q'ÂQ
are called similar.
vm. If Y^ = AX^ is a linear transformation of V^(F) relative to a given basis (Z-basis)

and Yjf = BX^ is the same linear transformation relative to another basis (IF-basis),
then
A and B are similar.
Note. Since
will be
for
might have been written as B = PAP-^. A study of similar matrices

There we shall agree to write B = R'ÂR instead of S = SAS'^ but
P"^, (12.2)
made
later.
no compelling reason.
96
Example
Let
2.
AX
[l.2,l]', W^ = [1,-1.2]',
be a linear transformation relative to the -basis and
= BXjf
IFg
(a)
to the W-basis.
(b)
corresponding to
AX.
(c)
let
W^
Given the vector X = [3,0,2]',

Find the linear transformation
Use the result of (b) to find the image ijf of Xy =
[1,-1,-1]' be a new basis,
image relative
find the coordinates of its

Yjr
[CHAP. 12
[1.3,3]'.
1
Write
[W^.W^.W^] =
(a)
Relative to the
The image
of
If -basis,
is
-1 -1
-1
1
9
-1
the vector
AX
then
-2
-1 -3
= [3,0,2]' has coordinates
[9,5,7]'
Xff =
W X
which, relative to the IF-basis is
Yf^
= [l,l,l]'.
=
W Y
[14/3,20/9.19/9]'.
(b)
w\
(c) Yj.
WÂX
36
21
-15
21
10
-11
-3 23
-1
(W~ÂW)Xjf = BXj^
36
21
-15
21
10
-11
-3 23
-1
6
=
[6,2,7]'
L.
See Problem
5,
SOLVED PROBLEMS
1. (a)
{h)
(c)
(d)
(a)
(6)
Set up the linear transformation

and 3 into Fg = [2,1,3]'.
carries E^ into
Y^
= [1,2,3]',
E^ into [3,1,2]',
Find the images of li= [1,1,1]', I2 = [3,-1,4 ]', and ^3 = [4,0,5]'.

Show that X^ and Zg ^-^^ linearly independent as also are their images.
Show that Xi, X^, and Z3 are linearly dependent as also are their images.
By TheoremI, A
The image
of
image of Xg
(c)
Y = AX which
The rank
is
= [y^, Fg, K3]
X^=
K3
[l,l.l]' is
the equation of the linear transformation is
Y-^
13
[6.4,8]'.
The image
of
=^
AX
-Yg is
Ys = [8,9,19]' and the
=[ 14.13,27]'.
of [A'^.Xg] =
-1
_1
is 2 as also is that of [î, Kg]
19
Thus, X^ and X^ are linearly
independent as also are their images

(rf)
We may compare
the ranks of \_X^. X^. X^] and {Y^.Y^.Yq\; however,
both sets are linearly dependent.
X^
X^ + X^ and Fg = Fi+Zg so that
CHAP.
2.
12]
Prove:
linear transformation (12.1) is non-singular
97
and only
if
if
is non-singular.
Suppose A is non-singular and the transforms of X^ ^ X^ are Y = AX^ = AX^. Then A{X-i^-Xi)
homogeneous linear equations AX = Q has the non-trivial solution X = X-^-X^. This
sible if and only if .4| = o, a contradiction of the hypothesis that A is non-singular.
the system of
and
is pos-
3.
Prove:
non-singular linear transformation carries linearly independent vectors into

linearly in-
dependent vectors.
Assume the contrary, that is, suppose that the images Yi = AXi. (i = 1,2
ent vectors X-^.Xr,
Xp are linearly dependent. Then there exist scalars s-^.s^
of the linearly independ-
p)
sp
not all zero, such that
^lYi +
H'î
s^Y2+
1=1
s^Yfy
^^
f^
P
'
.|^ ^(-4^1)
Since A is non -singular,

linearly independent.
4.
s^X^ + s^X^ + -. + spXp =
Hence, the
Y^ are linearly
certain linear transformation
[3.0,-2]', and I3
iZ
F =
But this
o.
carries Z^ =
=[
a + b +
aX.i^
+ bXr,+
cXg= E^;
-6
then
a + b
and
its
imageis
Y2 = 1-1.3,1
Y,
-^2, 3,-1
and the image of Eg
Yy
= AXf, =
ter 11, find the
X^
]'
1,0,1
into
[2,3,-1]',
Find the images of
and a
-^,
6 = 1,
H-2.7.-1 ]' =
The equation
]'.
7^ =
BX^
fj^,
^^s
Chapter
11,
Xf^r
PX^
-1
P''X,.,
c =
-1
-2
i.
Thus,
]'
E^= -hX^ + X^
1"
^z-
Then
0_
-1"
-1
^!i
Q^w
-2
PY
=[ 1.-1.1
into
and write the equa-
Q-ÂX,
Problem
relative to the f'-basis of that problem.
--3
and
,!
^Xg
[l,2,-2]'. Similarly, the image of 5 is

of the transformation is
[Y^.Y^.Yg]X
12,
hypothesis that the Xi are
is a linear transformation relative to the Z-basis

of
same transformation
From Problem
is contrary to the
c =
is K, = [l,l,l
+ 2c =
]'+ [3,0,-2]' +
5. If
+ spXp)
independent.
1.2,-1]' into [-2,7,-1]'.

tion of the transformation.
Let
A(Sj^X^+ S2X^+
Q^^Q^r.,
\^
14 -6
14
-9
X,
12,
Chap-
98
[CHAP. 12
Problem
6. In
show:
(a) the transformation is non-singular,
(b)
Af
= [1,1,2]',
A Y
carries the column vectors of
into
the elementary vectors.
7.
Using the transformation of Probleml, find (a) the image of

[-2.-5.-5]'.
-4ns.
(a) [8,5.11]', (b) [-3.-1. 2]'
8.
Study the effect of the transformation Y = IX. also Y = klX.
9.
Set up the linear transformation which carries E^ into [l,2,3]', 5 into [3.1. 2]', and 3 into [2,-1,-1]'.
Show that the transformation is singular and carries the linearly independent vectors [ 1,1,1 ]' and [2,0,2]'
into the same image vector.
10.
11.
Suppose (12.1) is non-singular and show that

ages Y^.Y^
Y^.
Use Theorem
changed.
to
III
show
....
X^
X whose
image
is
are linearly dependent so also are their im-
that under a non-singular transformation the dimension of a vector space is un-
Hint. Consider the images of a basis of P^ (F).

1
12.
X-^.X^.
if
(h) the vector
Given the linear transformation Y
-2
dependent vectors
î=[l,l,l]',
JVg =
X.
show
]',
and A:3=[i.2,3]' are linearly dependent,
(a) it is singular,
(6) the
images of the linearly
in-
3 5
[2.I.2
(c) the
image
of V^{R) is a Vs(R).
13.
Given the linear transformation Y
2 4 X.
show
(a) it is singular,
(6) the
image of every vector of the
113
2
V,
14.
(R) spanned by
]'
Yi,
= 1,2
]'
lies in the K,(fl)
spanned by [5,7.5]'.
Prove Theorem Vn.

Let Xi and
Hint.
Y
15. Prove:
16.
and [3.2.0
1.1,1
Let
BZ
(j
n) be the given sets of vectors. Let
carry the E^ into
AX
^2
12
_1
1_
= [1.0.1]',
Z3
(6)
X. when referred to the
= [14,10,6]' is the
= [1.1.1]'
image of
new
A^
carry the set Xi into E{ and
be Chosen.
Yy =
PX
and Y^ = PY.
Q ÂQX.
AT
=P
basis, say
= [1,2,3]' relative to the E-basis.
basis, has coordinates X^, = [-2,-1.4]' and
where P
where Q
Let
new
Show
Z, =
that
under the transformation.
Y has coordinates
Y^ = [8,4,2]'
0-1"
(d)
AX
be a linear transformation relative to the -basis and let a
A^
(a)
X^
Similar matrices have equal determinants.
[1,1,0]',
(c)
Kj.
-1
-1
iZ^,
Zf2,
Zî
^.
0"
1
17.
Given the linear transformation
7^
1
1
1
1
Xjf
relative to the IF-basis:
W^= [o,-l,2]',
IK,= [4,1,0]'
CHAP.
IK5
121
[-2.0,-4]'- Find the representation relative to the Z-basis: Z^ = [i,-l,l]', Z2 = [l,0,-l]', Z3=[l,2,l]'.
-10
Am
18.
If.
99
2-5
-10 2
2
in the linear transformation
whose vectors transforms
AX. A
is singular,
then the null space of A is the vector space each of
Determine the null space of the transformation of
into the zero vector.
123"
(a)
Problem
-4ns.
19.
12.
(6)
Problem
13.
(c)
X.
(R) spanned by [l.-l.l]'
(a)
(6)
I^\R) spanned by [2,l,-l]'
(c)
I/^()
spanned by [2,-l,o]' and [3,0,-1
]'
If y = AX carries every vector of a vector space I^ into a vector of that same space, v^ is called an Invariant space of the transformation. Show that in the real space V^{R) under the linear transformation
-f
(a)
12
X. the
\l^
spanned by [l.-l.o]', the V^ spanned by [2,-1.-2]', and the V^ spanned by
[1.-1,-2]' are invariant vector spaces.

2
(6)
13
spaces.
X. the Vq spanned by [l.l.l]' and the
^ spanned
by [l,0,-l]' and [2,-l,0]' are invariant
(Note that every vector of the V^ is carried into itself.)

1
(c)
10
X, the li spanned by [l,l,l,l]' is an invariant vector space.
-14-6
20.
Consider the linear transformation

tion of
1,
n.
PX
yi
(i
^h
1.2
n)
(a)
Describe the permutation matrix P
(b)
Prove: There are n! permutation matrices of order n.
(c)
Prove:
If
Pj and fg are permutation matrices so also are P3 = P-lP2
(d)
Prove:
If
(e)
Show that each permutation matrix P can be expressed as a product

umn matrices K^2, ^28
^n-T-nWrite
P = [^^, E^^.
^^] where ii, ij
% is a permutation
(/)
is a permutation matrix so also are P' and
= [s, 1, 4,
/^ is
/1./2
a permuta-
and
P^^P^Pt.
of a
number of the elementary
/.
1,2
of
and ^
col-
are the ele-
Find a rule (other than P~ = P') for writing P~

For example, when n = 4 and
2], then P'^ = [2. '4. 1. 3]; when P = [E^ E^. 1, 3], then P~^ = [g, 2, 4, 1].
mentary n-vectors.
PP'
which
in
chapter 13
Vectors Over the Real Field
INNER PRODUCT. In this chapter

If Z = [%,%, ..., x^y and y
all
vectors are real and l^(R) is the space of
= [ji,
all real re-vectors.
are two vectors of l^(R), their inner product is
YnY
72, ,
defined to be the scalar
X-Y
(13.1)
Example
1.
x^y^
^2=
For the vectors X^=\\.\,\\',

(a)
X^-X^
1-2
(6)
X-^-X^
1-1
+ l(-2)
(c)
X.yX^
1-1
(rf)
^1-2^:2
1-1
1-4
1- 1
1-
[2.1,2]',
+1-1
1-4
XnJn
^3 = [l.-2.l]':
=
=
1-1
1-2
x^j^
10
2(^1-^2)
Note. The inner product is frequently defined as
X.Y
(13.1')
X'Y
Y'X
The use of X'Y and Y'X is helpful; however, X'Y and Y'X are 1x1 matrices while
X-Y is the element of the matrix. With this understanding, (13.1') will be used
In vector analysis, the inner product is callhere. Some authors write Z|y for X-Y
.
ed the dot product.
The following
(13.2)
rules for inner products are immediate
(a)
X^-X^ = X^-X^,
ib)
X^-(X^ + Xs)
(c)
(X^+ X^)
0.
(X^+Xs)-X^
(Xs+ X^)
ORTHOGONAL VECTORS. Two

product is
X^-hX^ = HX^-X^)
X and Y
vectors
The vectors Z^ and Xg
THE LENGTH OF A VECTOR X of

thus.
ner product of X and X
X^-X^
î(R),
of
X^-X^ + X^-X^
=
+
X^-X^ + X^-Xs + X^-X^
of VîR) are said to be orthogonal if their inner
Example
denoted by
\\
are orthogonal.
X\\
is
defined as the square root of the in-
(13.3)
Examples. Prom Example
II
1(c),
^11
\\
X^\\ =
\/
V3
X-X
\/
xl +
xl+
---
X.
See Problems
100
1-2.
CHAP.
13]
Using (13.1) and
(13.3),
x.Y
(13.4)
A
are
may be shown
it
vector
X whose
length is
that
iil|z+y||'
||z||
101
WxW
- ||y|h
is called a unit vector.
The elementary vectors E^
examples of unit vectors.
THE SCHWARZ INEQUALITY.
If
and Y are vectors of ^(/?), then
\X-Y\
(13.5)
<
\\x\\.\\y\\
that is, the numerical value of the inner product of two real vectors is at most the product of
their lengths.
See Problem
THE TRIANGLE INEQUALITY.
If
and Y are vectors of
(13-6)
ORTHOGONAL VECTORS AND SPACES.

n-vectors and
0.
If
X^,
CiZi + ^2^2+ ...+ c^^ =

Since this requires 0^ =
for i = 1,2
if
Any
I.
set of
m<
vector
0,
IUII + ||y||
X^
X^
then for
then
are
= 1,2
m<n
mutually orthogonal non-zero

+ c^X^) Xi =
m. (c.lX^+ 0^X2+
we have
n mutually orthogonal non-zero re-vectors is a linearly independent
set and spans a vector space
<
l!^+yll
)/(/?),
3.
is said to be
Ijf(/?).
orthogonal to a vector space Vn(R)
orthogonal to every
if it is
vector of the space.

II.
If
a vector
to the space
is
orthogonal to each of the re-vectors X^, X^
X^,
it is
orthogonal
spanned by them.
See Problem
HI. If Vn(R) is a subspace of I^(/?),
which
is orthogonal to
k>h. there exists
at
least one vector
4.
of V^CR)
V^\R).
See Problem
5.
Since mutually orthogonal vectors are linearly independent, a vector space V'^(R), m>0,
can contain no more than m mutually orthogonal vectors. Suppose we have found r<m mutually
orthogonal vectors of a V^(R). They span a VîR), a subspace of V*(R), and by Theorem HI,
there exists at least one vector of V^(R) which is orthogonal to the I^(/?). We now have
r+l
mutually orthogonal vectors of l^(R) and by repeating the argument, we show
IV.
Every vector space V^(R), m>0, contains m but not more than
mutually orthog-
onal vectors.
Two vector spaces are said to be orthogonal if every vector of one is orthogonal to every
vector of the other space. For example, the space spanned by X^ = [1,0,0,1]' and X^ =
[0,1,1,0]' is orthogonal to the space spanned by X^ = [ 1,0,0,-1]' and X^ = [0,1,-1,0 ]'
since (aX^ + bXr,) (0X3+ dX^) =
for all a,b,c,d.
102
The set
V.
tor
[CHAP. 13
of all vectors orthogonal to every vector of a given Vn,(R) is a unique vec-
space
Vn'^(R).
^
n
J
ggg Problem
6.
We may associate with any vector ^ 7^ o a unique unit vector U obtained by dividing the
components of X by \\X\\ This operation is called normalization. Thus, to normalize the vector
X = [2,4,4]', divide each component by ||^|| = V4 + 16 + 16 = 6 and obtain the unit vector
.
[1/3,2/3.2/3]'.
basis of Vn(R) which consists of mutually orthogonal vectors is called an orthogonal baif the mutually orthogonal vectors are also unit vectors, the basis is called a
sis of the space;
The elementary vectors
normal orthogonal or orthononnal basis.
are an orthonormal basis of ^(R).
See Problem
THE GRAM-SCHMIDT ORTHOGONALIZATION PROCESS.

V^(R).
Suppose
X^,
X^
are
7.
a basis of
Define
y,
X,
^S
Y
^3
Y V
-'m
'^
^g'^3
-'w-l
Then the
unit vectors
Gj =
(i
^1-^3
%-l
Xi
= l,2,...,m)
V V
ll*^!
"
y'
"'l
are mutually orthogonal and are an orthonormal
1^11
basis of
Example
F(/?).
3.
Construct, using the Gram-Schmidt process, an orthogonal basis of V2(R). given a basis
A'i= [1,1,1]', a:2= [1,-2,1]',

(i)
Y^
X^ = [1.1.1]'
(ii)
Y^
X^- ~^Yi
(ill)
^3 =
^3 -
[1,-2,1]'
^Y, - ^Y,
The vectors
G2 =
Xs=[i.2.zY.
-^,
Note that Fg =
-^2
^1-1 =
[l/\/l. l/\/3, l/^/3l
[l/Ve, -2/V6, l/Ve]'
here because
Each vector
X.^
and
[1,-2.1]'
- -^y, - ^[1,1.1]'
[1,2,3]'
G^ = -jpj =
are an orthonormal basis of ^(fl).

0.
and
Gg
[-1,0,1]'
]',
-ii-
[-I/V2,
0,
1/\A2]'
G^ is a unit vector and each product G^
A^2 a^re
Gj =
orthogonal vectors.
See Problems 8-9.
CHAP.
13]
Let
^m be a basis
^2
Zi,
of a f^(/?) and suppose that
103
X^,
Xg,(l< s<
X^
m),
are
Then, by the Gram-Schmidt process, we may obtain an orthogonal basis

of the space of which, it is easy to show, Yj^ = X^, (i = 1,2
Thus,
s).
mutually orthogonal.
y^, Yg
}^
VI. If
X-i^,
X2,
exist unit vectors
, Xs,(l< s<m),
X^^^,
^m
are mutually orthogonal unit vectors of a Vn(R), there

i" the
X^^.^-
space such that the set
X^, X^, ...,X^
is an
orthonormal basis.
THE GRAMIAN.
Let X^, X^
be a set of real n-vectors and define the Gramian matrix
Z>,
A^ A^
A^ A2
A2 Aj X2 A2
'
'
Xp- X^
Xp- X2
...
'
(13.8)
A>)
A]^
A2
'
Ajj
A2 A^ A2 A2
X-^ A,
X^Xp
Xp- Xp
XpX-i
Clearly, the vectors are mutually orthogonal if and only if

In
Problem
if
the vectors are linearly dependent.
14,
Chapter
17,
we
ORTHOGONAL MATRICES. A
XpX^
...
XpXp
is diagonal.
shall prove
For a set of real re-vectors
VII.
A^ A^ A2
A-L
Z^, X^,
....
Xp,
G >0. The
|
square matrix A is called orthogonal
AA'
(13.9)
A'A
equality holds
if
and only
if
that is, if
(13.9')
A'
Prom
(13.9)
it
is clear that the
A'
column vectors (row vectors) of an orthogonal matrix A are
mutually orthogonal unit vectors.
Example
4.
By Examples, A
l/\/3
l/x/6
l/\[Z
-2/\/6
l/\/3
\/\[&
-1/^2
is orthogonal.
l/x/2
There follow readily

VIII.
If
the real re-square matrix
is orthogonal, its
column vectors (row vectors) are
an orthonormal basis of V^(R), and conversely.

IX.
X.
XI.
The inverse and

The product
the transpose of an orthogonal matrix are orthogonal.
of two or more orthogonal matrices is orthogonal.
The determinant
of an orthogonal matrix is 1.
ORTHOGONAL TRANSFORMATIONS.
(13.10)
Let
AX
104
[CHAP. 13
be a linear transformation in Xi(R) and let the images of the n-vectors I^ and
X^ be denoted by
Yi and Y^ respectively. Prom (13.4) we have
x^-x,
u\\x,^x,f -
\\X,f
Y,-Y,
k\\\Y,^Y,f -
II
\\X,f]
and
Comparing
and
right
left
members, we see that
products, and conversely.
XII.
A
lem
if
y,
Y,f]
(13.10) preserves lengths
it
preserves inner
Thus,
linear transformation preserves lengths
linear transformation
Y=AX
is called
and only
if
orthogonal
if its
if it
matrix
preserves inner product s.
is orthogonal.
we prove
10,
XIII.
Examples. The linear transformation Y
AX
if
l/\/2
l/v/6
l/\/3
-2/\/6
I/V3
l/\/6
and only
if its
In Prob-
matrix is orthogonal.
-I/V2
ii3
orthogonal.
The image of
l/v^
[a,i,c]'is
"
_^
26
_f
_1_
c "]/
[
and both vectors are of length yja^ + b^ + c^
XIV. If (13.10) is a transformation of coordinates from the -basis to another, the Zbasis, then the Z-basis is orthonormal if and only if A is orthogonal.
SOLVED PROBLEMS
1.
Given the vectors
= [1,2,3]' and X^ = [2,-3,4]',
Zj.
(o) their inner product,
(b) the
find:
length of each.
2
(a)
X^-X^
XiX^
[1.2.3] -3
1(2)
2(-3) +
3(4)
(6)
\\X^f
X^.X^
lA-jf
2(2)
X[X^
[1,2,3]
+ (-3)(-3) + 4(4)
29
14
and
and
\\X^\\
\\Xji
V29
vTi
CHAP.
13]
105
Show
(b)
that 1 = [1/3, -2/3, -2/3 ]' and Y ^ [2/3.-1/3, 2/3]' are orthogonal.
Find a vector Z orthogonal to both X and Y.
(a)
X-Y
2. (a)
2/3
Xy
[1/3,-2/3,-2/3] -1/3
and the vectors are orthogonal.
2/3
1/3
(6) Write
2/3
-2/3
column of zeros.
3.
Then by
and compute the cofactors -2/3. -2/3, 1/3
2/3
Clearly, the theorem is true
of the elements of the
0_
(3.11)
Prove the Schwarz Inequality:
If
O"
-2/3 -1/3
[A:,y,o]
= [-2/3, -2/3, 1/3]'
If
if A"
is
and Y are vectors of

F
or
is the zero vector.
orthogonal to both
Vn(R), then
Assume then
A:
\X-Y\ <
that
and K.
||A'||.||y||
and Y are non-zero vectors.
a is any real number,
llaA^
yf
(aX + Y)-(aX + Y)
[ax^ + y^.ax^+y^
(a x^
a^xf
+ 2ax^y^ +
axn+yn]-[ax^+y^. ax^ + y^
+ (a^x^ + Zax^y^ + y^)
yj^)
+ 2aX.Y
axn+jnY
+ (o^^ + 2a%j^ + y^
\\Yf >
Now a quadratic polynomial in a is greater than or equal to zero for all real values
of o
discriminant is less than or equal to zero. Thus,
i(X.Yf - i\\xf- \\Yf
and
4.
\x-y\
Prove:
If
a vector
is orthogonal to
each of the n-vectors
Since
Xi-Y
= 0, (i = 1,2
orthogonal to the space.

orthogonal to that space.
5.
Prove:
which
If
a^X^)-Y
X^
X^,
Thus, Y
m).
In particular, if
a^X^-Y + a^X^-Y +
is
is
X^.
it
a^X^-Y
is orthogonal to the
Then
orthogonal to every vector of the space and by definition is

orthogonal to every vector of a basis of a vector space, it is
a ^() is a subspace of a V^(R), k>h. then there exists at least

one vector
is orthogonal to the f^(fi).
^^*
-Î'-^s
^h be a basis
of the
FV).
let X^,^^
be a vector
in the
vM)
A"
of
v\R)
"
but not in the P^(R)
consider the vector
<*)
The condition
that
if its
m-\\Y\
<
of the space spanned by the X's can be written as a^X^+a^X^-^

^-c^Xtji
{aiXi + a^X^+... +
and only
<
space spanned by these vectors.
Any vector
if
a^X^ + a^X^ +
be orthogonal to each of X^.X^
...
Xf,
aj^X,^
+ a^^^Xf,^^
consists of h homogeneous linear equations
and
106
[CHAP. 13
a^X^-X^ + a^X^-X^ +
...
+ af,Xh-X-, + a,,^^Xf^^^. X^
a^X^.X^ + a^X^.X^ +
...
+ a^Xh-X^ + af^^^Xf,^^- X^
ft + l
unknowns a^.a^
ay,^^. By Theorem IV, Chapter 10, a
non-trivial solution exists. When these
values are substituted in (i), we have a non-zero (why?) vector
X orthogonal to the basis vectors of the kA)
and hence to that space.
In the
^ ^^^ ^^''^'''^ orthogonal to every vector of a given V^{R) is a unique

vector space
""^^
Vn-k'J'^^
^^* -î-^s
be a basis of the V^{R).
A-fe
system of homogeneous equations
The
..-vectors
X^.X=o.X^.X=Q
^'>
Since the
Jt,-
n-k linearly independent solutions (vectors) which span

a K"-\/?).
system
(i) is
of rank k
hence, there are
(See Theorem VI, Chapter 10.)
Uniqueness follows from the fact that the intersection space of the
VîR) and the V^'^(R)
space so that the sum space is Xi{R).
7.
Find an orthonormal basis of
V^(R),
given
is the zero-
= [1/^6, 2/x/6, l/x/6]'.
Note that A- is a unit vector. Choose Y = [l/Vl,

Then, as in Problem 2(a), obtain Z = [l/^Jl. -l/VI,
0,
-IA/2 ]'
1A/3]'
8.
satisfy the
Xk.X=0
are linearly independent, the coefficient matrix of the
X^^
orthogonal to each of the
to
another unit vector such that
complete the
-Y =
set.
Derive the Gram-Schmidt equations (13,7).

^^'
^^' -^2
be a given basis of V^(R) and denote by
>^n
Y^.
Yr,
the set of mutually orthogonal
Y^
vectors to be found.
(a)
Take
Y-^^
(b)
Take
Y^ = Xr, + aY-^
X^.
Since
Y-^
Y^.Y^
and
(c)
Take
o =
- -i-2
Y^ =
and Y^ are to be mutually orthogonal.
X3 +
aYr,
Thus.
+ bY^
K, =
Y^-X^ + Y^-aY^
Y X
^Sill
X^ ~ ^1^
Xo-
Since
Y^. K,,
Y^
Y^-X^ + aY^-Y^
Y^ are to be mutually orthogonal,
yi-Yj,
Yi-Xg + aY^.Y^ + bY^-Y^
Y^-
X^ + bY^-Y^
Y^.Y^
Y^-Xs + aY^-Y^ + bY^.Y^
Y^-X^ + aY^-Y^
and
Then
(d)
Continue the process until K,
- ^ll^
is obtained.
and
7, =
Z3 -
^ki^ K _
2W^ v
CHAP.
9.
13]
Construct an orthonormal basis of
Take
Y^ =
X^
= [2.I.3Y.
K,
X^ -
^î-i
"2
[1.1.1]'
Fg,
107
given the basis Z^ = [2,1,3]', X^ = [1,2,3]', and Xg = [1,1,1]'.
Then
=
^3 V
- ii[2,1.3]'
[1,2.3]'
[-6/7.15/14,3/14]'
^1 '"^3
- ^
Normalizing the y-s. we have
14
I4J
[2/\/l4. l/i/Ii, 3/v'l4
]',
[_3
3j
[-4/\/42, 5/V42. l/\/42]', [l/Vs. 1/V3.
-1/^3
]'
as the required orthonormal basis.
10. Prove:
Let
y^. Yq
Suppose A
and only
if its
is
orthogonal so that A/l =
Y^-Y^
Theorem XII lengths
YiY^
AX.
Then
/.
(X'Â')(AX^)
X\X^
X^^-X^
are preserved.
Conversely, suppose lengths (also inner products) are preserved.

Y^-Y^
and A
matrix is orthogonal.
be the respective images of X^,X2 under the linear transformation
(1)
and, by
if
X{(A'A)X2
Then
A'A=1
X^X^.
is orthogonal.
SUPPLEIVIENTARY PROBLEMS
H. Given
the vectors A'l = [l,2,l
.Yg =
]'.
[2.I.2]', vYg = [2.1,-4
]'.
find:
product of each pair,

(6) the length of each vector.
(a) the inner
(c)
a vector orthogonal to the vectors X^, X^
Ans.
(a) 6, 0,
-3
Vb,
(6)
12.
Using arbitrary vectors of
13.
Prove (13.4).
14.
Let
a:
= [1.2.3,4]' and
V^
3,
^3(7?).
(c)
X.^,
Xq
[l,0,-l
]',
[3.-2.1]'
verify (13.2).
= [2,1,-1.1]' be a basis of a V^(R) and
= [4,2,3,l]' lie in a V^{R)
containing
and y.
(a)
Show
that
(b) Write
15. (a) Prove:
Z is not in the ^^{R).

aX + bY + cZ and find a
vector of
I^(ft) is
vector
of the V^{R) orthogonal to both
orthogonal to itself
if
and only
if it is
and Y.
the zero vector.
If X-^. X^. Xg are a set of linearly dependent non-zero ^-vectors and

X^ and Xq are linearly dependent.
(6) Prove:
if
Z^ Xg = Xj_-Xq=
0,
then
108
16.
Prove:
vector
is orthogonal to every vector of
P^"(/?) If
and only
[CHAP. 13
is orthogonal to every vector of
if it
a basis of the space.
17.
18.
Prove:
20.
two spaces VîR) and
t^(fl) are orthogonal, their intersection
space is
I^(if).
Prove: The Triangle Inequality.

Hint.
19.
If
Show
Prove:
that
A"
||
!|Z +
1|
<
(||.Y||
+ \y\
|1a:|1
if
Normalize the vectors of Problem

Ans.
[l/Ve, 2/V6. l/\/6]',
21.
Show
that the vectors X.
22.
(o)
Show
using the Schwarz Inequality.
||y||)^,
and only
if
and Y are linearly dependent.
11.
[2/3, 1/3, 2/3]',
Y .Z
X^
that if X^.
Problem
of
X^
2 are
[2/V2i, l/VÎ, -4/V2i
]'
an orthonormal basis of V^{R).
are linearly independent so also are the unit vectors obtained by normalizing
them.
(6)
Show
that
if
the vectors of (a) are mutually orthogonal non-zero vectors, so also are the unit vectors
obtained by normalizing them.
23.
Prove: (a)
is
orthogonal and
.4
is
orthogonal and \a\ = -i, each element
24.
Prove Theorems
25.
Prove;
26.
each element of A
If
(6) If
If
Prove that AA'
.4
1,
equal to
is
of^
its
cofactor in
^4
is equal to the negative of its cofactor in |^|
VIII, IX, X, XI.
and B commute and C
.4
where A
(or A'A),
is orthogonal,
then C'AC and
is n-square. is a
C'BC commute.
diagonal matrix
if
and only
if
the rows (or columns) of A
are orthogonal.
27.
Prove:
UX
28.
Prove:
If
29.
Prove:
If
and Y are n-vectors, then XY'-YX' is symmetric,

and Y are n-vectors and A
is n-square, then
X-(ÂY) = {A'X) -Y
(b)X.X
30.
be an orthonormal basis and
A'
cl+4+...
(a)
X^.
(6)
[3/VI3,
{a)
[1,-1,0]',
(6)
[1.0,1]',
(e)
[2,-1,0]',
VsiR). given
2/\/l3]',
c^X^,
then
(a)
X-X^ =
c^,
(i = 1,2
(a)
X^ =
3/VT7, -2/\/T7, 2/1/17]';
[-4/\^, -3/^34,
[2/VI3.
Vî^R)
0,
(6) [3,0,2]'
3/\/34]'
-3/VT3]', [0,1,0]'
by the Gram-Schmidt process, using the given vectors
in order:
[3.2,1]'
[4,-1,0]',
(a)
[iV^. -iV^,
(b)
[^\/~2, 0,
(c)
[2V5/5, -V5/5,
[4,0,-1]'
[V^/6, V2/6, -2v'V3]', [-2/3,-2/3,-1/3]'
0]',
iV2]',
[0,1,0]',
0]',
Obtain an orthonormal basis of

Hint.
Take
Ans.
[\A3/3, V'3/3, -V^/3]',
Y^ = X^,
[1,-1,-2]'
[2,-1,-2]',
[1,3,1]',
A^
^=1
[0.l/^^2.l/\f2Y,
0,
if
cl
Construct orthonormal bases of
Ans.
32.
Z^
Find an orthonormal basis of

Ans.
31.
A"]^,
[iV^.
-iV2]'
0,
[^/5/5.2^/5/5.oy,
I^(ft),
given
^"1
[o,0,-l]'
=[ 1,1,-1
]'
and
ATg =
[2,1,0
]'.
obtain Y^ by the Gram-Schmidt process, and Y^ by the method of Problem 2(6).

[5\/2. 0, 2\A2]',
[\/6/6, -\/6/3, -\/6/6
]'
n);
CHAP.
13]
33-
Obtain an orthonormal basis of VîR). given
34.
Show
35.
Prove:
36.
Use Problem 35
two ways that the vectors
in
If
skew-symmetric. and
is
[7,_i,_i
10
(b)
-5
oj'
38.
71 =
-5 -10
-10
(b)
5 -I2J'
is an orthogonal matrix
P^ÂPX^
and
it
Prove:
40.
Let
If
and
is orthogonal
= [xj^.x^.xsY
as
or
-11
AP
where P
is
non-singular, then
PB^
is orthogonal.
-basis to an orthonormal Z-basis with matrix P. Y =
7^= BX^ (see Chapter
12).
Show
that if
is
orthogonal so also is
B.
and
ZxF
and
+ .4 is non -singular then
= [yi.yQ.ygY
where
25, 23]'
= (I
- A) (I + A)'"^
is
Jl
21 =
3 ys
,
Z2 =
X^, Y^.
^1
.
the z^ as cofactors of the elements of the third column

mn of
],
Zg =
yi
estciblish:
of
(6)
The vector product
of two linearly independent vectors is orthogonal to each of the two vectors.
(c)
XxY
(d)
(kX)xY
41. If W. X. Y.
(a)
(d)
k(XxY)
XxikY),
k a scalar.
are four vectors of V^{R), establish:
X x{Y+Z)
(b)X-(YxZ)
(c)
is the zero vector.
-(YxX-)
XxY
XxZ
Y-(ZxX)
W-Y
(WxX)-(YxZ) =
X-Y
X-X
(XxY)-(XxY) =
Y-X
=
Z-(XxY)
W-Z
X-Z
X-Y
Y-Y
\XYZ\
of
After identifying
2 yi
The vector product
two linearly dependent vectors
X xY
yi
(a)
be-
skew-symmetric.
be two vectors of VsiR) and define the vector product,

X2
= [21,
^3 ys
AX
and con-
Theorem XIV.
versely, to prove
39.
-51
In a transformation of coordinates from the
comes
is orthogonal.
10
.4
-A)(I + A)~^
given
"5-14
If
= (I
2-3
-12
37. Prove:
12
s"!
(a)
and [5,4,5,6]' are linearly dependent.
+ A is non-singular, then
to obtain the orthogonal matrix
Ans.
]'.
[l. -1.-2, -3]',
[l.2,3,4]',
"
(a)
X-^ =
109
chapter 14
Vectors Over the Complex Field
COMPLEX NUMBERS.
x and j are real numbers and i is defined by the relation j^ = 1, z = xîy

is called a complex number. The real number x is called the real part and the real number y is
If
called the imaginary part of x +
Two complex numbers
fy.
are equal if and only if the real and imaginary parts of one are equal
respectively to the real and imaginary parts of the other.
A complex number
The conjugate
x + iy =
of the
if
complex number
(product) of any complex number and
It
and only
its
if
x = y =
= x+iy
0.
is given
by
= x+iy = xiy.
The absolute value |z| of the complex number z = x+iy is given by

follows immediately that for any complex number z = x + iy,
|z|
(14.1)
VECTORS. Let X
and
> \A
>
\z\
be an ra-vector over the complex field C.

Since ^(R) is a subfield,
the vector space I^(C).
The sum
conjugate is a real number.
\J
z-z =
\fW+y^
\y\
The
is to
it
|z|
such vectors constitutes
totality of
be expected that each theorem con-
cerning vectors of I^(C) will reduce to a theorem of Chapter 13 when only real vectors are considered.
If
is
= [%, x^
XnY and y =
[ji,
y^Y are two vectors of
72
P^(C), their inner
product
defined as
X-Y
(14.2)
XY
%yi
The following laws governing inner products

(a)
(14.3)
I-y =
hold.
2?.{X.Y)
where R(X-Y)
is the real part of
X-Y.
(c)
X-(cY) = c(X-Y)
(d)
X-(Y+Z) = X-Y + X-Z
= 2CiX-Y)
where C(Z-y) is the imaginary part of Z-F.
(e)
(Y+Z)-X =
Y-X+Z-X
See Problem
of a vector
vectors
is
given by
and Y are orthogonal

V^{C), the
X-Y-Y-X
(g)
||Z||
if
\/
X-X =
X-Y = Y-X =
Q.
Triangle Inequality
\\X+Y\\
and the Schwarz Inequality (see Problem

(14.5)
XnJn
(cX)-Y = c(X-Y)
For vectors of
(14.4)
are readily verified;
X-Y+Y.X
(/)
(b)
The length
Two
y^
x^y^
<
||Z||
<
\\X\\-\\Y\\
||i'll
2)
\X-Y\
Moreover, we have (see Theorems I-IV of Chapter 13)
110
\/%%
%^2
+ XnXn-
1.
CHAP.
14]
I.
Any
set of
111
mutually orthogonal non-zero re-vectors over C is linearly independent
and, hence, spans a vector space I^(C).

II. If a vector Y is orthogonal to each of the re-vectors
thogonal to the space spanned by these vectors.
III.
If
X^,
X^
X^,
then
it
is or-
VîC) is a subspace of V^C), k>h, then there exists at least one vector
in
V^(C) which is orthogonal to V^(C).

IV.
Every vector space
J^(C),
m>0, contains m
m mutually
but not more than
orthog-
onal vectors.
basis of I^(C) which consists of mutually orthogonal vectors is

called an orthogonal
If the mutually orthogonal vectors are also unit
vectors, the basis is called a nonnal
or orthonormal basis.
basis.
THE GRAM-SCHMIDT PROCESS.
(14.6)
Let
X,. X^
Y,
X,
In
Xn
Yd
Xn
yn
_
~
X^
^2-^3
Y .Y
l2-l2
be a basis for F^^CC).
Yi-Xs
Y^
Y,.Y,
Y
y
'a-i'^m
y
T;
"Si
'y^
-'m-i
^m-i'-'m-i
The
unit vectors
Gi
7.
(i
= 1.2
m)
Define
are an orthonormal basis for ^^(C).
V. If X^, X^
X^, (l<s<m), are mutually orthogonal unit vectors
of ^(0) there
exist unit vectors (obtained by the Gram-Schmidt
Process) Z,,,, X, ^
;i'
in the snaoe
such that the set Z Z^
Z, is an orthonormal basis.
THE GRAMIAN.
Let
Gramian matrix.
Xp be a
X^
set of
;s-
vectors with complex elements and define the
Xi'Xi Xj^-Xq
X^-Xp
Aj
X-y
X-^Xq
X^ Xp
X^-X^
X^-Xp
Ag X^
XqXq
X2 xp
Xp-Xp
XI X-^ XI x^
x^x.
p^p
X^'X'i
(14.7)
Xp-X^ Xp-X^
Clearly, the vectors are mutually orthogonal if and

only if
Following Problem
14,
Chapter
17,
is diagonal.
we may prove
VI For a set of re-vectors X^. X^

Xp with complex elements,
holds if and only if the vectors are linearly
dependent.
\G\
>
0.
The equality
112
UNITARY MATRICES. An
n-square matrix A is called unitary
The column vectors (row vectors)
fCHAP. 14
(AyA = A(A)'=
if
that is
if
(i)'=
^.
of a unitary matrix are mutually orthogonal unit vectors.
we have
Paralleling the theorems on orthogonal matrices of Chapter 13,
The column vectors (row vectors)
VII.
I,
of an re-square unitary matrix are an orthonormal
basis of l^(C), and conversely.

VIII.
The inverse and
The product
IX.
of two or more unitary matrices is unitary.
The determinant
X.
the transpose of a unitary matrix are unitary.
of a unitary matrix has absolute value
UNITARY TRANSFORMATIONS. The
(14.8)
where A
1.
AX
is unitary, is called a unitary transformation.
XI.
linear transformation preserves lengths (and hence, inner products) if and only
matrix is unitary.
if its
XII.
If
AX
is a transformation of
coordinates from the i'-basis to another the Z-
basis, then the Z-basis is orthonormal if and only
if
is unitary.
SOLVED PROBLEMS
1.
Given X = [l+j, -J, 1]' and Y = [2 + 3J, 1- 2J, i]',

(c) verify X-Y + Y-X = 2R(X-Y)
(a) find X-Y and Y-X
(d) verify X-Y -Y-X = 2C(X-Y)
(b) verify X-Y = Y-X
2+
(a)
X-Y
X'Y
[l-i,f,i;
3/
l-2i
(l-i)(2+3!) + i{l-2i) + 1(0
7 4 3j
Y-X
Y'X
+i
[2-3i.l + 2i.-i]
3i
3i
= X-Y.
2.
Y-X. the conjugate of Y-X. is
(b)
From
(c)
X-Y + Y-X
(7
3i)
+ (7-3i) =
14
2(7)
(d)
X-Y
-Y-X
(7
3j)
-(7-30
6j
2(30 =
(a):
Prove the Schwarz Inequality:
As
in the
\X-Y\
<
2R(X-y)
p||
case of real vectors, the Inequality
2C(X-Y)
11Y||.
is true if
or
= 0.
When X and Y
a'Wxf
+ 2aR{X-Y) +
are non-zero
vectors and a is real, then

\\aX +
Yf
{aX+Y)-{aX+Y)
a X-X + a{X-Y + Y-X) + Y-Y
\\Y\\
>
0.
CHAP.
14]
Since the quadratic function in a is non-negative
\xf\Yf
R{X-Yf If
X-Y
then
0,
|;f
-F
if
and only
<
discriminant is non-positive,
if its
Ra-Y)
and
< ||x|
fl(;5:-y)
||y
X-Y
if
113
<
,i
IkMyll
define
O.
X-Y
c =
\x-y\
Then
R(cX-Y) <
R[ciX-Y)]
3.
Prove:
B =
(B)
4. If
\(A)'AV
i = B + iC
5.
B+iC
is real if
Prove:
If
is
(A'A)
(A)A
while, by (14.3(6)),
R(cX-Y)
and B is Hermitian.
(B + iCUB+iC)
and only
if
BC + CS
= o or
is skew-Hermitian. then
(B + iC)(B+iC)
BC
iA
-CB
if
B and C anti-commute.
The reader
B^ + i(BC + CB) - C^
and only
if
if
B and C anti-commute.
is Hermitian.
(riA)'
thus,
Since A is skew-Hermitian,
(S)'
is Hermitian.
and only
if
(B+iC)' = B + iC; thus.
Hermitian,
Consider B = ~iA.
and S
|UM|y||
forallândy.
Hermitian, show that (I)/l is real
is
(A)A
This
|c||U||-iyl|
U-yl < U||-||y||
Thus,
(A)'A is Hermitian for any square matrix A.
Since
\\cX\\-\\y\\
\X-Y\.
(A)' =
t(Z)'
will consider the case
-A. Then
i(-A)
-iA
= iA.
6.
Given the vectors X^=[i.2i,iY.

find X^-X^ and X^-X^,
A:2 = [l,
1+
/,
o]',
and Xg =
[i, 1
j,
2]'
(a)
(b)
find the length of
(c)
show
(d)
find a vector orthogonal to both
Ans.
(a)
that
each vector Xi
[l-i, -1, i-j]'
2-3i.-i
(b) ^/e
7.
Show
8.
Prove the relations (14.3).
9.
Prove the Triangle Inequality.
that [l + i.i.lV.
orthogonal to both X^ and X^.
is
X^ and
V^ V^
.
[iA-i.oy, and
10.
Prove Theorems
11.
Derive the relations (14.6).
I-IV.
Jf g
(d)
[-1 -5iJ .3 -i]
[l -i.
1. 3j ]'
are both linearly independent and mutually orthogonal.
114
12.
in order, construct
Using the relations (14-6) and the given vectors
[CHAP. 14
an orthonormal basis
for iy.C)
when
the
vectors are
(a)
[0,1,-1]', [l + j,l,l]'. [l-j,l,l]'
(b)
[l + i.i.lY.
Arts,
[2.1-2i.2 + iY, [l-i.O.-iY.
[O.^V2.-iV2]', [1(1 + 0.
(a)
ri
[2(1+0. 2..2
(&)
13. Prove: If
is
/I
T
J.
[3;^.
I. 2
5t
]'.
[-î,
3+
3jy
^d
^(1 + 0.
]'
-6 + 3i
7
-5
-5
3j
-b
1-i
-t
y
L2V30 'aVso- 2\Am^
r
4^. 4^ J-
a matrix over the complex field, then A + A has only real elements and
A- A
has only pure
imaginary elements.
14.
Prove Theorem V.
15. If
(b)
show
is n-square,
(a) A'A is
A'A =
diagonal
if
and only
if
and only
if
Prove:
17.
Prove Theorems VII-X.
18.
Prove:
If
the columns of
and Y are n-vectors and A
16.
If
if
the columns of
is
A are mutually orthogonal vectors.
are mutually orthogonal unit vectors.
is re-square,
skew-Hermitian such that
X-AY
then
A'X-Y
+ A is non-singular, then B = (l-A){I + A)~^ is unitary.

I
19.
Use Problem 18
to form a unitary matrix, given
(6)
"^
i
L-1+.-
_9 + 8i
Ans.
r-l +
|_
-4-2i"|
2j
If
2-4i
4-lOj
^ and B are unitary and
Follow the proof
22. Prove: If
29
-2-jJ'
in
Problem
is unitary
Show that
same
1(1 +
J//3
-k
l/\/3
+t
-1 + j
-10 -4t
-9 + 8j
AB
-16-18i
12i
J-
-2-24J
order, then
13. to prove
and Hermitian. then A

3+
23.
of the
Chapter
10.
-10 -4i
-2-24J
(b)
(a)
20. Prove:
21.
i+j1
and
BA
are unitary.
Theorem XI.
is involutory.
2^
4 + 3i
is unitary.
2y/T5
-^
2Vl5_
-i/y/3
-1
24. Prove: If
is unitary
and
is unitary
and I+A
25. Prove:
if
= .4P
where P is non-singular, then PB
(I
- A)
(,!
is unitary.
+ A)'^
is
skew-Hermitian.
chapter 15
Congruence
CONGRUENT MATRICES. Two
A and B over F
P over F such that
re-square matrices
there exists a non-singular matrix
(15.1)
are called congruent,
over
if
FAP
Clearly, congruence is a special case of equivalence so that congruent matrices have the
same
rank.
When P
is
expressed as a product of elementary column matrices, P' is the product in reis, A and B are congruent provided A can
verse order of the same elementary row matrices; that
be reduced to B by a sequence of pairs of elementary transformations, each pair consisting of
an elementary row transformation followed by the same elementary column transformation.
SYMMETRIC MATRICES.
I.
whose
Example
1.
In
Problem
we prove
1,
Every symmetric matrix A over F of rank

first r
r is congruent over F to a diagonal matrix

diagonal elements are non-zero while all other elements are zero.
Find a non-singular matrix P with rational elements such that
12
10
10
-8
- P'AP
is diagonal, given
reducing A to D, we use [A /] and calculate en route the matrix P' First we use
and XgiC-S), then H^-ii-2) and K^^{-2) to obtain zeroes
in the first row and in the first column.
Considerable time is saved, however, if the three
In
ff2i(-2) and K2i(-2). then //gjC-S)
row transformations are made first and then the three column transformations.
transformed into a symmetric matrix, an error has been made. We have
[AH^
12
10
10
-8
10
1
1
[DP']
115
-2
-1
-1
10
is not then
-2100
-3010
-2001
0-100
4-12
10
0-1-1
0-1-1
c
'V-
If
0-100
-2
10
~\^
-10
-1-110
CONGRUENCE
116
[CHAP. 15
-2 -10 -1
4-1
Then
1
10
The matrix D
to
which A has been reduced
The additional transformations
is not unique.
10
0-100
ffgCi) and Kg(^).
for
D by
example, will replace
the diagonal meitrix
while the
10
10
transformations
H^d) and K^Ci)
replace
0-900
D by
There
however, no pair of
is,
rational or real transformations which will replace
D by
a diagonal matrix having only non-neg-
ative elements in the diagonal.
Let the real symmetric matrix A be reduced by real elementary

transformations to a congruent diagonal matrix D, that is, let P'AP = D. While the non-zero
diagonal elements of D depend both on A and P. it will be shown in Chapter 17 that the number
REAL SYMMETRIC MATRICES.
of positive non-zero diagonal elements
depends solely on
A.
a sequence of row and the same column transformations of type 1 the diagonal elements
may be rearranged so that the positive elements precede the negative elements. Then a
By
of
sequence of real row and the same column transformations of type 2 may be used to reduce the
diagonal matrix to one in which the non-zero diagonal elements are either +1 or 1. We have
II.
real symmetric matrix of rank
r is
congruent over the real field to a canonical
matrix
P
(15.2)
The
'r-p
integer p of (15.2) is called the index of the matrix and
p-(r p)
is called the
signature.
Example
2.
Applying the transformations H23. Kâ and Hîk), Kr,(k) to the result of Example
[A\n
10
0-100
4
-2
-10
-1 -1
and (/AQ = C.
if
Thus, A
is of
rank
C
1
we have
10
-5
0-10
-2
10
IC\(/]
-1-110
= 3,
1,
index p =
2,
and signature
= 1.
III. Two re-square real symmetric matrices are congruent over the real field if and only
they have the same rank and the same index or the same rank and the same signature.
In the real field the set of all re-square matrices of the type (15.2) is a canonical set over
congruence
for real ra-square
symmetric matrices.
CHAP.
IN
CONGRUENCE
15]
THE COMPLEX FIELD, we
117
have
IV. Every ra-square complex symmetric matrix of rank

complex numbers to a canonical matrix
congruent over the field of
is
/^
(15.3)
Examples. Applying the transformations
Hîi)
and
10
-5
0-10
-2
[^1/]
-1 -1
Two
V.
numbers
if
Example
to the result of
10
10
10
2,
-5
-2i
-1 -1
if
SKEW-SYMMETRIC MATRICES.
[D
']
1
See Problems 2-3.
complex symmetric matrices are congruent over the field of complex

ra-square
and only
we have
^&:]
R'AR
and
K^fî)
If
is
skew-symmetric, then
(FAP)'
FAT
r(-A)P
-FAP
Thus,
VI.
Every matrix
B =
FAP
congruent to a skew-symmetric matrix A is also skew-
symmetric.
In
Problem
we prove
4,
Every n-square skew-symmetric matrix A over F
VII.
congruent over
is
to a canoni-
cal matrix
(15.4)
where
D,-
= r
diag(Di, Dj
0^,0,..., 0)
,(f=l,2
t).
The rank
of
/}
is
2t.
See Problems.
There follows
if
Vin. Two ra-square skew-symmetric matrices over

The set
are congruent over
if
and only
of all matrices of the type (15.4) is a canonical set over congruence for re-square
skew-symmetric matrices.
HERMITIAN MATRICES. Two
[^
],
(15.5)
Thus,
or
conjunctive
if
n-square Hermitian matrices A and B are called Hermitely congruent,

FAP
P such
that
CONGRUENCE
118
Two
IX.
[CHAP.
re-square Hermitian matrices are conjunctive
if
and only
if
15
one can be obtain-
ed from the other by a sequence of pairs of elementary transformations, each pair consisting of a column transformation and the corresponding conjugate row transformation.
An Hermitian
X.
matrix
of rank
conjunctive to a canonical matrix
r is
Ip
-Ir-p
(15.6)
The
p-(r-p)
integer p of (15.6) is called the index of A and s =
Two
XI.
is called the signature.
have the same
re-square Hermitian matrices are conjunctive if and only if they
rank and index or the same rank and the same signature.
The reduction
of an Hermitian matrix to the canonical form (15.6) follows the procedures
Problem 1 with attention

troublesome case is covered
to the proper pairs of
of
in
The extreme
elementary transformations.
Problem?.
See Problems 6-7.
SKEW-HERMITIAN MATRICES.
If
skew-Hermitian, then
is
(FAPy
FAP
(PAT)
Thus,
Every matrix B =
XII.
conjunctive to a skew-Hermitian matrix A is also skew-
FAP
Hermitian.
By Problems, Chapter
14,
= -iA is Hermitian
P such
if
is
By Theorem
skew-Hermitian.
that
Ip
FHP
iFHP
Then
= iF{-iA)P =
FAP
= iC
-Ir-p
and
Up
B
(15.7)
FAP
- ilr~p
Thus,
XIII.
which
Every re-square skew-Hermitian matrix A
is the rank of
A and
p is the
is conjunctive to a matrix (15.7) in
index of iA.
XIV. Two re-square skew-Hermitian matrices A and B are conjunctive

they have the same rank while -iA and -iB have the same index.
if
and only
See Problem
if
8.
CHAP.
CONGRUENCE
15]
119
SOLVED PROBLEMS
1.
Prove: Every symmetric matrix over

r non-zero elements in the diagonal.
of rank
can be reduced to a diagonal matrix having exactly
Suppose the symmetric matrix A = [a^,-] is not diagonal. If a^ / 0. a sequence of pairs of elementary
3, each consisting of a row transformation and the same column transformation,
will
transformations of type
reduce A to
"n2
Now
"ns
the continued reduction is routine so long as b^^.c^^.
that along in the reduction,
we have obtained
S+1, s+2
k^
S+2, s+ 1
'^ri
which the diagonal element k^^^^^^^
however, some
k^j
say
s+^ /
0,
Suppose then
hss
in
are different from zero.
the matrix
= o.
s+i
every
If
we move
4y
*s+l, n
"n,
s->-2
= 0,
we have proved
the theorem with s =
into the (s+l,s+i) position by the proper row and
it
r.
If,
column
transformation of type 1 when u = v; otherwise, we add the

(s+u)th row to the (s+t;)th row and after the
corresponding column transformation have a diagonal element
different from zero. (When a^, = o, we proceed
as in the case Ar^+i ^^.^ = o above.)
we
Since
whose
2.
first t
Reduce
In
are led to a sequence of equivalent matrices, A is

ultimately reduced to a diagonal matrix
diagonal elements are non-zero while all other elements
are zero.
the symmetric matrix
'12
.2
to
canonical form (15.2) and to canonical form

(15.3).
each obtain the matrix P which effects the reduction.

1
2
1
u\n
-1
-2
-2
-1
c
1
[0|^']
To
obtain (15.2),
we have
[0
Pi']
-1
-4
-2
1
1
0-1
-2^f2
-2
k\f2
1
i\Pi
\C\P'\
-1
-4
-2
CONGRUENCE
120
[CHAP. 15
-av^ -2
j\/2
and
5a/2
To
obtain (15.3),
we have
[D\Pl]
-1
Find a non-singular matrix ? such that ?'A? is
2J
k\f2
-i
2\/2
2i
-i
2-j
Vc\?'\
iV2
canonical form (15.3), given
in
1
-2\/2
-2V2
and
3.
2-i
10 + 2i
^
1
[^1/]
1+J
2-j
+i
2-i
10 +
2j
'-\J
3-2J
-i
2j
-1-j
10
10
5+12J
2i
2i
-i
7+
13
4i!'
-5+12i 3-2t
13
13
[c\n
-I
7+
4i
13
-5+12i
1
Here.
13
3-2j
13
4.
Prove: Every ?i-square skew-symmetric matrix A over
B
where
It
D^
= Q.
U :]
then S = ^.
(J
If
= 1,2,
-4
0,
diag(Di, Dg
of rank 2t is congruent over

7)^,0
to a matrix
0)
then some
mj
and
/th and second rows; then Interchange the Jth
oy
a..
j
^ by the skew-symmetric matrix
-a,;,
-Oy
g
2
'S*
i
first
-aji ?^ 0. Interchange the sth and first rows and the

columns and the /th and second columns to replace
Next multiply the
first
row and the
first
column by
l/a^^-
CHAP.
CONGRUENCE
15]
121
1
''2
to obtain
-1
and from
by elementary row and column transformations of type
it,
3,
obtain
4
1
Di
-1
F4
If
5.
/v=
0.
the reduction is complete; otherwise, the process is repeated on F4
Find a non-singular matrix P such that P'AP
a^s 4 0,
we need
-2 -1
-4
3
10
10
1-3
and second columns of
10
-4320
0-3
10
10
-4230
-1 -2
to obtain
1
Next, multiply the
first row and first column by 2; then proceed to clear the

two columns of non-zero elements. We have, in turn.
10
0-3
-2230
-2
to obtain
10
Di
1/10 -1/5
0-10
1/2
-1
Thus when
-2i
-1
-1/5
P'
Do
0-2
-1
1/10 -1
-1/5
=
10-2
P'A?
= diag (Di.Dj).
6.
is in
-3 - 2j
two rows and the
-5
Finally, multiply the third row and third column by
1/2
and
first
10
10
-1 -2
10
-2
-2
-2 -1
-1
is obtained.
only interchange the third and second rows followed by the interchange of the
2
third
is in
Using
until
-3 + 2i
first
CONGRUENCE
122
1-j -3 + 2i
1+i
[^1/]
[CHAP. 15
-3 -
-i
2j
10
25
2-3i
5_
13
-13
13
13
2j
00
-l-i
-13
2i
2-3f
13
5V13
5\/l3
\/l3
0-1
2i
Vl3
VT.5
vc\p'^
2+
3-2i
3t
5\/l3
\fl3
13
and
5\/T3
7.
^/I3
\/T3
Find a non-singular matrix P such that P'AP is in canonical form (15.6), given
l
- 2i
2+
1
[41/]
2i
2+
3i
2i
3j
-4 +
13
2i
10
HC
10
13
2f
10
5j
-5i
HC
-2-Zi
-5J
-4 +
5i
-4 - 2s
3f
- 3i
-4 - 2
+ 2i
10
0-1
-4 -
-2-2J
5J
-5/2
10
\/T6
x/Io
x/To
[cin
-4 +
4i
ao
\/T0
10
10
and
vTo
4j
\/To
-2-3!
10
'10
10
10
HC
HC
2j
-1 +
CHAP.
8.
CONGRUENCE
IB]
is in
-l +
-J
1-i
1
The non-singular matrix
2i
-l-2j
-i
is
+ 2t
1
1
Consider the Hermitian matrix
1 +
123
2j
-i
+i
such that P'HP = diag[l,
1,
-l].
10
Then P'AP
= diag[i.
i.
-i
9.
(a)
[-:
(6)
-:]
3-1
(a)
"O
O'
(c)
-2
(b)
.2
"
2"
4
4
(d-)
(c)
is in
2i
ky/2
-k
2i
11.
(a)
1
/I
-1
-2
H- j
+ 4iJ
- 4;
-1 -
(b)
(1
(l-J)/\/2
-1
1
(6)
is in
canonical form (15.4). given
-"
(b)
-3/2"'
(c)
12-2
(d)
A =
-2 -3
-
+ 2i)/l3
(c)
+ 0/2
(-3-2i)/l3
(3
5i
-1
4j
i/\/2
Ans.
-3 -
2i
2(1 -J)
2(-i-2,)]
(a)
(d)
-l-2i
+i

(a)
+j
Ans.
11.
I
'
|_1
i + 2i
1+211
-1 o1
-1
0,
kV2
[::]
10.
"^V2 -^v^ -1
-1 -2_
1
Ans.
is in
0-1
2-3
"0010"
10
10
-2
-2103
2-1-3
'1
(d)
3
2
0-1
-1
-1/3
-2/3 2
1/3
CONGRUENCE
124
12.
Find a non-singular matrix P such that P'AP is
in
+i
tl+
(a)
(6)
'
1-i
Ans.
13.
[1
(a)

2
-i +
sfl
(6)
-i
(c)
in
(c)
A
I
(6)
6i
[:
1
(6)
(-2-J)
-1
1-i
-1-j
-i.
2+j
l-2i
-1
-1-2J
_-2 + i
(c)
-r]
-i
1
^1
(l-J)/\/2
-l'
l/\/2
-1
1,
,0
-2 + 3j
-2 -
(rf)
show
l/x/2
(l-2i)/VT0
i/y/2
(-2-i)/yJT0
-l/\/2'
i/\/2
l/\/l0
(a)
(i)
2i
-1-j (-2 + 51)
-1-i
i
Ans.
(c)
l+i
18
1-j
-l+i
3 + 4t
-1
3 + 2f
i/Vs
L-i+j
2i
3-4i
(2-f)/V5
1-j
-1-i (-5-i)/\/5
Find a non-singular matrix P such that P'AP is
(a)
3i"|
10
3i
[CHAP. 15
that a 2-square matrix
satisfies
C'DC
14
If
15.
Let A be a non-singular n-square real symmetric matrix of index
if
and only
|Cl=l.
if
L-i oj
Show
p.
that
>
if
and only
if
-p
is even.
16- Prove:
Hint.
17.
non-singular symmetric matrix A is congruent to its inverse.
Take P
BB' where B'AB
= I
and show that P'AP = A
Rewrite the discussion of symmetric matrices including the proof of Theorem
to obtain (15.6) for Hermitian
matrices.
18.
19.
Prove:
It
then A is symmetric (skew-symmetric)
Hint.
if
is
symmetric (skew-symmetric).
7")
(S
- T)
Show
is non-
(S+TT^(S-T)
P'SP =[(S-Tf^(S+T)S~'^(S-T)<.S+T)'^r^.
Let S be a non-singular symmetric matrix and let T be such that (S+T)(S - T) is non-singular.
then T is skew-symmetric.
if P'SP = S when P = (S + TTîS- T) and I+P is non-singular,
Hint.
21.
and only
Let S be a non-singular symmetric matrix and T be a skew-symmetric matrix such that (S +

singular. Show that P'SP = S when
20.
if
S(/-P)(/ + P)-^
S(I + PrHl-P).
that congruence of n-square matrices is an equivalence relation.
Show
that
chapter 16
Bilinear Forms
AN EXPRESSION
and
(y^, y^
which
Jri)
is linear
and homogeneous in each of the sets of variables

fonn in these variables. For example,
(x^, x^
m)
is called a bilinear
XxJi
2x172
is a bilinear form in the variables {x^, Xg)
The most general
ISxija
and
4x271
15x272
^s/s
(y.^, y^, y-g).
bilinear form in the variables (x^, Xg
and {y^.y^
x)
Jn)
maybe
written as
or,
more
%i%yi
oi2%y2 +
+ 021^271
022^72 +
+ ami^myi +
Om2%y2 +
X i7?i
%r!
fl2ra%7ra
""mv^n Jn
briefly, as
(16.1)
/(^.y)
t
S ai-x^y
=i
j=i
where
[x^, x^
The matrix A
]',
^ =
[0^^],
'
-^
*2> .
and
i'
%1
012
in
'7i'
Ojl
^22
2n
72
_0'n\
Om2
Omra_
= [yi,72
_Yn_
7]'.
of the coefficients is called the matrix of the bilinear form and the rank of
is called the rank of the form.
See Problem
Example
1.
The bilinear form

1
îXi + îrs + *2yi + ^2X2 + ^373
Ui.
%.%
X'AY
125
Ti
73
y2
1.
BILINEAR FORMS
126
CANONICAL FORMS.
Let the
[CHAP. 16
new variables
x's of (16.1) be replaced by
u's
by means of the
lin-
ear transformation
(16.2)
Xi
(f=l,2
bijUj,
1.
m)
or
BU
and the n y's be replaced by new variables v's by means of the linear transformation
(16.3)
Ti
(f=l,2
c^jVJ,
J=i
or
n)
Y = CV
We have X'AY = (BU)"A(CV) = U'(BAC)V. Now applying the linear transformations U =

V = lY we obtain a new bilinear form in the original variables X\B'AC)Y = X'DY
Two
bilinear forms are called equivalent
and only
if
if
IX,
there exist non-singular transfor-
mations-which carry one form into the other.
Two bilinear forms with mxn matrices A and B over

I.
only
If
if
the rank of (16.1) is
sire
equivalent over
there exist (see Chapters) non-singular matrices
r,
if
and
P and Q such
that
Ir
FAQ
and C =
Taking B = P'
in (16.2)
(16.4)
UXPAQ)V
reduced
in (16.3), the bilinear form is

Ir
U'
U^V^
^2^2
to
U^V^
Thus,
II.
Any
mations over
bilinear form over
of rank
to the canonical form
can be reduced by non-singular linear transfor-
u^^v^
+ u^v^ +
+ uî'r-
Examples. For the matrix
XÂY
of the bilinear form
X'
ofE^xamplel,
10
10
10
10
10 110
110 10
10
1
10
1
1
10
10
1-1-110
-1
10
10
1-1-110
10
0-110
1
p'
-1
Thus,
-1
10
PU
-1
and
QV
-1
1
1
reduce X'AY to
-l'
1
1
1
1
U%V
U-^V-:^
+ U^V^ + Ugfg
CHAP.
BILINEAR FORMS
16]
The equations
127
of the transformations are
X-j^
X2
Xg
Ur,
and
U,
Ti
72
Ts
^1
"3
1^3
V^
^3
See Problem 2
TYPES OF BILINEAR FORMS. A
bilinear form
1 S
X'AY
ij x
'îJj
a,-,-
symmetric
symmetric
alternate
skew-symmetric
according as A is
Hermltlan
Hermltlan
alternate Hermltlan
COGREDIENT TRANSFORMATIONS.
^) and
(i. 2
tion
CU
is
IV.
CV
skew-Hermltlan
Consider a bilinear form X'AY in the two sets of n variables

When the ;'s and y's are subjected to the same transforma-
the variables are said to be transformed cogredlently.
Under cogredient transformations
III.
where A
If
(yi.ys. .3^)-
and Y =
is called
CU
Y = CV,
and
is re-square, is carried into the bilinear form
We have
the bilinear form X'AY.
U\C'AC)V.
symmetric, so also is C'AC; hence,
symmetric bilinear form remains symmetric under cogredient transformations of
the variables.
V.
Two
variables
if
F are equivalent under cogredient transformations of the

matrices are congruent over F.
bilinear forms over
and only
From Theorem
I,
if their
ChapterlS, we have
VI. A symmetric bilinear form of rank

transformations of the variables to
(16.5)
%%yi
From Theorems
II
and IV, Chapter
15,
can be reduced by non-singular cogredient
02*272
OrXryr
follows
VIl. A real symmetric bilinear form of rank r can be reduced by non-singular cogredient
transformations of the variables in the real field to
(16.6)
and
iyi
in the
complex
(16.7)
%y2 +
xpyp
2y2
a;^4.i
7^4.1
x^Jr
field to
%yi +
x^y^
See Problem
CONTRAGREDIENT TRANSFORMATIONS.
Let the bilinear form be that of the section above. When

X = (C~^yU and the y's are subjected to the trans-
the x's are subjected to the transformation
formation
Y = CV,
3.
the variables are said to be transformed contragredlently.
We have
BILINEAR FORMS
128
X'AY, where A
IX.
only
if
The
= (C-K^T
)'U and
X'lY =
bilinear form
x^y^ + x^y^ +
%y
Y = CV,
the bilinear form
U\C AC)V.
is n-square, is carried into the bilinear form
transformed into itself
is
if
and
the two sets of variables are transformed contragrediently.
FACTORABLE BILINEAR FORMS.

X.
Under contragredient transformations
VIII.
[CHAP. 16
Problem 4, we prove
In
non-zero bilinear form is factorable
and only
if
if its
rank is one.
SOLVED PROBLEMS
1.
%yi
+ 2iy2
13%yg
4a^yi + 15x^y^
%. 2
-13]
x^y^
= X'
Ji
-1
4 15
Jx
1
-13l
-4
15
-ij
73
2.
Reduce %yi + 2x^y^ + 3%y3
Ix^y^ + 2%yi
2^y2 + x^y^ + 3x2y4 +
+ 4%y3 + x^y^ to canon-
"ix-^y^
ical form.
2
The
matrix of the form is
-2
-1/6 -5/6
and
-1 -1
By Problem
6.
Chapter
5.
the non-singular matrices
1/3 -4/3 -1/3
10
3-2
2-2
7/6
10
/g
PAQ
are such that
Thus, the linear transfor-
14
mations
P'U
or
X^
:x:2
Ui
2U2
U2
Uq
U3
and
y
6
QV
or
"3
xs =
7a
reduce X'AY to u^v^ +
3.
U2t)2.
Reduce the symmetric bilinear form
mations
X'AY = X'
to (16.5) in the rational field,
12
10
10
-8
(b) (16.6) in
by means of cogredient transfor-
the real field, and (c) (16.7) in the complex
field.
1
(o)
Prom Example
1,
Chapter
15. the linear transformations
-2 -10 -1
4 -1
1
u.
=
1
1
-2 -10 -1
4-1
=
1
10
CHAP.
BILINEAR FORMS
16]
reduce
XAY
to
uj^'i
"2% +
129
ûsfg.
(6)
Prom Example
Chapter
2,
the linear transformations
15,
-5 -2 -1
1
-1
u.
-5 -2 -1
2
1-1
1
reduce X'AY to u^v^ + ug^s

(c)
Prom the
result of
Example
- "3%Chapter
2,
000'
-5
-2
-1
we may obtain
15.
-1
-5202
-2j
-1-110
Thus, the linear transformations
10
10
10
-5 -2J -1
i
-1
2
u.
-5 -2J -1
-1
2
t
reduce X'AY to
U^V^ + U2V2 + UgVg.
4.
Prove:
non-zero bilinear form f(x,y) is factorable
Suppose the form
if
and only
if its
1.
is factorable so that
and, hence, a^j = b^cj.
Clearly, any second order minor of
aij
aj5
a^bj
aib^
fk^'j
fe*s_
= [a^-], as
bib^
Lfei "ks_
vanishes.
rank is
Thus the rank
of
/4
"k
"k
is 1.
Conversely, suppose that the bilinear form is of rank 1. Then by

Theorem I there exist non-singular
linear transformations which reduce the form to U'(BAC)V = u^v^.
Now the inverses of the transformations
=
carry uû^ into
(S rijXj)(l s^jyp
jnr
f(x.y)
and
'^jyj
and so f(x.y) is factorable.
BILINEAR FORMS
130
[CHAP. 16
5.
Obtain linear transformations which reduce each of the following bilinear forms to canonical form (16.4)
(a)
x^j^
- 2^73 +
+ ^272 - 3*273 - ^sTi - ^72
"ix^j^
(6)
-11
-4
7 y.
(c)
-2
6.
-2
-2
X'
5-510
^sTs
10
2-510
(rf)
12
10
Obtain cogredient transformations which reduce
(a)
14
y and
Ans.
(a)
(6)
X'
10
5_
-2 -7
to canonical form (16.6).
'l
(6)
-3 -4V3/3
\/3/3
\/3/3
Ir
7. If Bj^,
B2. Ci
C2 are non-singular n-square matrices such that
B-Â^C^ = BÂ^C^ =
find the transfor-
mation which carries X'A^Y into UA2V.

Arts.
= (B'^BîU,
8. Interpret
Problem
23,
C^C^V
Chapter
5, in
terms of a pair of bilinear forms.

-
9. Write the
transformation contragredient to
.Y
1
1
Prove that an orthogonal transformation
11.
Prove: Theorem IX.
is
/.
.4ns.
-2
i
2
10.
221
~2
contragredient to itself, that is
.Y
PV Y
.
PV.
-1
12. If
X'AY
is a real
non-singular bilinear form then
XA Y
is called its reciprocal bilinear form.
Show
that
when
reciprocal bilinear forms are transformed cogrediently by the same orthogonal transformation, reciprocal bilinear forms result.
13.
4, Chapter 15, to show that there exist cogredient transformations

duce an alternate bilinear form of rank r = 2t to the canonical form
Use Problem
U1V2
14.
W2V-L
+ U3V4.- u^V3 +
+ ugt-i "^t - "sfêt-i
Determine canonical forms for Hermitian and alternate Hermitian bilinear forms.
Hint. See (15.6) and (15.7).
PU. Y
PV
which
re-
chapter 17
Quadratic Forms
A HOMOGENEOUS POLYNOMIAL
of the type
n
(17.1)
whose coefficients
Ît ^2
Example
a^,-
"u
X'AX
are elements of
is called a quadratic form
over
in the variables
^7J'
1.
2
as^^
+ 2*2 -
trix of the form
2
"^x^
The maways according as the cross-product terms -4x1x2
+ 8x^x^ is a quadratic form in the variables xi.x^.xg.
'ix^x^
may be
written in various
and 8x1x3 are separated to form the terms
012X1X2.021^*1
and
013X1X3,031X3X1.
We
shall
agree that the matrix of a quadratic form be symmetric and shall always separate the crossproduct terms so that a
a^ Thus,
The symmetric matrix A =

A
-2
0-7
4Xj^X2
SXj^Xg
is called the matrix of the quadratic form
[o^^- ]
X'
7Xg
"-^s
'
If
the rank is r<n,
and the rank of
the quadratic form is called singular;
otherwise, non-singular.
TRANSFORMATIONS. The
linear transformation over F,
with symmetric matrix A over
(17.2)
(.BYyA(BY)
BY
carries the quadratic form (17.1)
into the quadratic form
YXB'AB)Y
with symmetric matrix B'AB.
Two
quadratic forms in the same variables %, xz
there exists a non-singular linear transformation
one of the forms into the other.

I.
The rank
X=BY
are called equivalent if and only if
which, together with Y =IX, carries
Since B'AB is congruent to A,
we have
of a quadratic form is invariant under a non-singular transformation of the
variables.
II.
Two
quadratic forms over
are equivalent over
if
and only
if their
matrices are
congruent over F.
Prom Problem
1,
Chapter
15, it
follows that a quadratic form of rank
the form
(17.3)
Kî
Kr^
131
+ h^y^,
hi/=o
can be reduced to
QUADRATIC FORMS
132
in
[CHAP. 17
which only terms in the squares of the variables occur, by a non-singular linear transformaX = BY. ViTe recall that the matrix B is the product of elementary column matrices while 6'
tion
is the product in reverse order of the
-2
Example
2.
Reduce
X'
same elementary row matrices.

4
Example
of
to the form (17.3).
-7
We have
[A I]
-2
-2
o'
BY
0-2
-23
-4
0-2
Thus.
0-7
reduces q to
1
1
[D Bl
=
1
q'
y=-2v^+9v=.
See Problems
1-2.
LAGRANGE'S REDUCTION. The
reduction of a quadratic form to the form (17.3) can be carried out

by a procedure, known as Lagrange's Reduction, which consists essentially of repeated com-
pleting of the square.
Example
3.
1
4x^*2
7%;
3
"*"
'
^*1*3
[x^- 'ix^{x^-2x^)\ + 2x1
7^:
{xl-'^x^{x^-2x^) + i{x^-2x/\ + 2x1 - Ixl -
(x^-2x^ + ix/ - 2(xl-8x^x^)
Ux^-2xJ
2Sx
(x^-2x^ + 'ix^f - 2(.xl-Zx^x^-\-\Qxl, + Qxl
(\-2^2+H)'
Thus,
yi
ys
xx
- 2x2 +
xq^
4acs
- 4x3
ot
Xg
73
reduces q to
- 2(A:^-4^3f + Qxl
+
2y2 +
xi
X2
y2 + 4y3
X3
ys
yi
4ys
y^ - 2y^ + 9y^.
See Problem
REAL QUADRATIC FORMS.
3.
q = X'AX be reduced by a real non-singular

one or more of the h^ are negative, there exists anonsingular transformation X= CZ, where C is obtained from 6 by a sequence of row and column
transformations of type 1, which carries q into
Let the
real quadratic form
transformation to the form (17.3).
(17.4)
in
sz^
s,z7
11
s z^
s^z2 2
If
...
Sfyzj,
s*+iZa+i
...
s^;.
which the terms with positive coefficients precede those with negative coefficients.
CHAP.
QUADRATIC FORMS
17]
Now
133
the non- singular transformation
Wi
"'i
VSi Zi
'j
= 1,2
(/
=r+l,r+2
r)
71)
or
'
;-
'''4k-
.1.1
J
carries (17.4) into the canonical form
wl + wl +
(17.5)
w^^
;2^^
wl
Thus, since the product of non-singular transformations is a non-singular transformation,
we have
Every real quadratic form can be reduced by a real non-singular transformation to
the canonical form (17.5) where p, the number of positive terms, is called the index and r
III.
is the rank of the given quadratic form.
Example
4.
In
Example
o
y^
q" =
into
J
reduces
q
"
^12
the quadratic form
2,
a^ + 2x^
8xx
1213
- Ix^ - ixx
3
was reduced
to
The non-singular transformation yi = z^, y^ = 23, ys = 22 carries q'

2y^ + 9yQ
~
+
^^3 ^^^ ^^^ non-singular transformation 21 = w^, 22 = w^/S, zg = wa/v^
922
zf
.
i.
to
"'
2,2
w^+
w^ -
w^.
Combining the transformations, we have that the non-singular linear transformation

Wi + ^W2 + V^M>3
Xl
X2
|mJ2 + 5V^;g
X3
3"^
4/3
v^
4/3
2^
1/3
qr
to
SYLVESTER'S LAW OF INERTIA.
W^ +
In
wl-
The quadratic
q
w%.
Problem
we prove
5,
form is of rank 3 and index
2.
the law of inertia:
IV. If a real quadratic form is reduced by two real non-singular transformations to

canonical forms (17.5), they have the same rank and the same index.
Thus, the index of a real symmetric matrix depends upon the matrix and not upon the
mentary transformations which produce (15.2).
ele-
The difference between the number of positive and negative terms, p - {r-p), in (17.5) is
called the signature of the quadratic form. As a consequence of Theorem IV, we have
V.
Two
and only
if
real quadratic forms each in n variables are equivalent over the real field if
they have the same rank and the same index or the same rank and the same sig-
nature.
COMPLEX QUADRATIC FORMS.
Let the complex quadratic form X'AX be reduced by a non-singular
transformation to the form (17.3).
It is
clear that the non-singular transformation
2i
îVi.
(i
Zj
yj
(7
= r+1, r+2,
1.
r)
...,7i)
QUADRATIC FORMS
134
[CHAP. 17
Ul
1
1, 1,
-'-
''''{/h^'/h;'-
)^
carries (17.3) into
-4--
<
(17.6)
Thus,
Every quadratic form over the complex
VI.
field of rank
can be reduced by a non-
singular transformation over the complex field to the canonical form (17.6).
Two complex
VII.
quadratic forms each in n variables are equivalent over the complex
field if and only if they
have the same rank.
DEFINITE AND SEMI-DEFINITE FORMS. A
if its
rank and index are equal.
positive definite quadratic form can be reduced to
of values of the x's, ^ >
rank and index are equal,
the
q>
a;'s,
be reduced
-yl
i.e.,
y^ + yj
to
=X'AX,
= p<
n.
"^
\a\
=0,
is called positive semi-definite if its
Thus, in the real field a positive semi-definite quad+ y^ r< re,

and for any non-trivial set of values of
,
0.
real non-singular quadratic form
A
i.e.,
-^
q = X'AX is called negative definite if its index p = 0,

= n,p = 0. Thus, in the real field a negative definite form can be reduced to -y^ - y^ - y^ and for any non-trivial set of values of the x's, q < 0.
i.e.,
j\-^ J^-^
0.
real singular quadratic form
ratic form can
q = X'AX, \a\ f 0, in
Thus, in the real field a
y^ and for any non-trivial set
real non-singular quadratic form
n variables is called positive definite
real singular quadratic form q =

<
re,
p =
- y^ Clearly,
X'AX
is called negative semi-definite if its index p = 0,
Thus, in the real field a negative semi-definite form can be reduced

^f ^y non-trivial set of values of the x's, q 5 0.
^"^
y^
if
0.
to
q is negative definite (semi-definite), then -q is positive definite(semi-definite).
For positive definite quadratic forms, we have
=X'AX
VIII. If q
is positive definite, then
U| >0.
PRINCIPAL MINORS. A
minor of a matrix A is called principal if it is obtained by deleting certain

rows and the same numbered columns of A. Thus, the diagonal elements of a principal minor of
A are diagonal elements of A.
In Problem
IX.
6,
we prove
Every symmetric matrix of rank
has
at least
one principal minor of order
r differ-
ent from zero.
DEFINITE AND SEMI-DEFINITE MATRICES. The
matrix i of a real quadratic form q = X'AX is called definite or semi-definite according as the quadratic form is definite or semi-definite. We have
X.
real symmetric matrix
singular matrix
XI.
C such
that
A =
exists a matrix
of rank
is positive definite if and only if there exists a non-
C'C.
A of rank r
A = C'C.
is positive semi-definite if
and only
if
there
such that
See Problem
7.
CHAP.
QUADRATIC FORMS
17]
XII.
If
135
is positive definite, every principal minor of
is positive.
See Problem
XIII. If
is positive semi- definite, every principal minor of
REGULAR QUADRATIC FORMS.
For a symmetric matrix A =
is non-negative.
over F,
[a^,-]
8.
we
define the leading
principal minors as
PO =
(17.7)
In
Problem
Pl=
1,
9,
O12
021
022
P2 =
Oil.
04.2
%3
Ogi
052
O23
Oqi
O32
'^3
%1
Oil
Pa
Pn= Ul
we prove
XIV. Any re-square non-singular symmetric matrix A can be rearranged by interchanges

of certain rows and the interchanges of corresponding columns so that not both p_^ and
Pn-2 are zero.
XV. If A is asymmetric matrix and

have opposite signs.
Example
5.
For the quadratic form
X'AX
Oigs
t^
0;
12
po =
Kg^X
X'
12 14
2
for
which po =
1,
Pi =
1,
P2 =
0.
Ps = -1.
and p
P4=
pi =
1,
1,
p2 =
0,
p3 =
0,
P4 =
m=
1.
yields
'1112
112 2
(i)
then p_2
X,
the transformation
but Pn-i = 0,
Pn-iPn^^
112
112
2
Here
if
1-
Thus,
for
i) not both p2 and ps are zero.
symmetric matrix A of rank r is said to be regularly arranged if no two consecutive p's

sequence po.pi, -.Pr ^^^ zero. When A is regularly arranged the quadratic form X'AX
is said to be regular. In Example 5, the given form is not regular; the quadratic form (i) in the
same example is regular.
in the
Let i be a symmetric matrix of rank
r.
By Theorem
A contains
IX,
least one non-
at
vanishing r-square principal minor M whose elements can be brought into the upper left corner
of A. Then p^ ^^
while p^^.^ = p^^.^ = ... = p = 0. By Theorem XIV, the first r rows and the
first r columns may be rearranged so that at least one of Pr-i and p^_2 is different from zero.
If p^_^ j^
and p^_2 =0, we apply the above procedure to the matrix of Pt--i; if Pt--2^0,
we apply
the procedure to the matrix of P7-_2;
^^^^
so on, until
XVI. Any symmetric matrix (quadratic form) of rank
is regularly arranged.
Thus,
can be regularly arranged.
See Problem
all
XVII. A real quadratic form X'AX is positive definite

leading principal minors are positive.
XVIII.
real quadratic form
each of the principal minors po.Pi
X'AX of rank
Pr
if
and only
if its
rank is
re
10.
and
is positive semi-definite if and only if
is positive.
QUADRATIC FORMS
136
[CHAP. 17
KRONECKER'S METHOD OF REDUCTION. The

into one in
method of Kronecker
which only the squared terms of the variables appear
XIX.
X'AX
q =
is a quadratic form over
for
is
reducing a quadratic form
based on
in n variables of rank
singular linear transformation over F it can be brought to q' =

singular r-rowed minor C oi A occupies the upper left corner of B.
F which reduces
a non-singular linear transformation over

quadratic form in r variables.
X'BX
r,
then by a non-
which a nonMoreover, there exists

in
q" = X'CX, a non-singular
q to
XX. li q = X'AX is a non-singular quadratic form over F in n variables and

a ^ 0, the non-singular transformation
if
Pn-i =
re-1)
1, 2.
^nnyn
or
10
a^n
BY
1
<^n-i,
carries q into
HiJiYj + Pn-xPnYn
which one squared term in the variables
in
has been isolated.

1
-2
-2
4
1
Example
6.
X'AX
= X'
-î
P2 =
-2 ^ 0.
-2
0-7
-2
O^S
non-singular transformation
Xl
xs
ys +
a.13
ys
- 8y3
yi
a23 ys
8y3
2ys
0-2
ss ys
-8
ys -
-8
reduces X'AX to
1
-2
-2
0-7
0-2
Y'
8 -8 -2
in
XXI.
but
which the variable yg appears only
q =
a,
're,?!-! ^
'
0,
-8
in
-8
Y'
-2
-2
36
squared form
X'AX is a non-singular quadratic form over

the non-singular transformation over F
F and
if
o^-i, n-i
"
xi
yi
CLi,n-iyn-i
Xn
^n-i.n yn.
*ra-i
(i
'înyn,
1, 2,
...,n-2)
a n n-i Yn-
,
or
1
1
sy
...
!, ^.
...
a2 n-1
=
1
a.-2,
_i
n-1
...
,0
...
a
'-'-n.-2, n
'-n-i,
a__i
= ^nn
The
CHAP.
QUADRATIC FORMS
17]
n-2
n-2
carries 9 into
(Hjyiyj
1'^n,n-\Pnjn-\yn-
i-i j-\
The
137
further transformation
'
Ji
Jn-x
2
V
1,
n-2)
2n
Zn
n-2
n-2
yields
Zn-i
Zn-i
Jn
(i
Zi,
a^jz^zj +
.2
2a_^_j^p(zn-i
- z^
in
which two squared terms with opp-
J -1
site signs are isolated.
Example
7.
X'AX
^22
- O'ss =
but
reduces X'AX
xi
X2
xq
*i3ys
*23r3
-1
or
0-10
a 32 72
-1
The non-singular transformation
0.
cti272
yi
i^
to
-1
tXg2 =
X'
-1
-]
rSY
yl
+ 2y,y3
The transformation
)'l
72
22
23
-1
73
22 +
23
21
carries
Y'BY
into
1
Z'
-1
o'
1
1
-1
z'
<
2zi
-2
Consider now a quadratic form in n variables of rank r. By Theorem XIX, q can be reduced to 9i = X'AX where A has a non-singular r-square minor in the upper left hand corner
and zeros elsewhere. By Theorem XVI, A may be regularly arranged.
If
pr-i
0,
Theorem
XX
can be used to isolate one squared term
(17.8)
Pr-iVrYr
If pr-i =
but a.r-1, r-i ^0, interchanges of the last two rows and the last two columns
yield a matrix in which the new pr-i = a^_i^^_i f 0. Since pr-2 ^ 0, Theorem XX can be used
twice to isolate two squared terms
(17.9)
Pr-2
r-i,
r-iJ
a r-i., r-iPrYr
which have opposite signs since pr-2 and pr have opposite signs by Theorem XV.
QUADRATIC FORMS
138
If
used
to
and a^_i_r-i = 0,
isolate two squared terms
[CHAP. 17
then (see Problem 9)
pr-L =
a,-,
and Theorem XXI can be
r-i ^
2ar,r~ipr{yr-i-yr)
(17.10)
having opposite signs.
This process may be repeated
until the given quadratic form is
reduced
to
another con-
taining only squared terms of the variables.
sequence pr-i,Pr
presents a permanence or a variation of sign. In (17.9) and (17.10) it is seen that the sequences p^_2, oLr-i. r-v Pr ^"^^ Pr-2' ^r, r-v Pr Present one permanence and one variation
Thus,
of sign regardless of the sign of dr-i, r-i and a.^, r-i
In (17.8) the isolated term will be positive or negative according as the
XXII. It q = X'AX, a regular quadratic form of rank r, is reduced to canonical form

by the method of Kronecker, the number of positive terms is exactly the number of permanences of sign and the number of negative terms is exactly the number of variations of
sign in the sequence
where a zero in the sequence can be counted either
P^,
po, pi, P2
as positive or negative but must be counted.
See Problems 11-13.
FACTORABLE QUADRATIC FORMS.
Let X'AX f
0,
with complex coefficients, be the given quad-
ratic form.
Suppose X'AX factors so that

X'AX
(1)
+ 02*2 +
(oj.%1
+ ;)( iiXi +
b.2_X2
+ + b^x^)
0--
to.;
hi
J
.^
tj
is non-singular.
becomes
Let the
transforms
If
i )
into
y-^y-z
of rank
yx
OiaCi
+ 02*2 +
+ 0
y2
^1*1 + ^2^2 +
+ ^n^n
ys
Xs
Thus,
2.
x^
1) is of rank
2.
the factors are linearly dependent, at least one element a^^f 0.
their coefficients be renumbered so that o^ is ci
y?
transforms (i) into
Conversely,
if
!Y^
a^Xx + 02*2 +
72
X2
of rank
X'AX has rank
Thus, (1) is of rank
1.
or 2
it
Let the variables and

+ (^n^n
Xn
1.
may be reduced respectively by Theorem VI
to y^ or
r^ + y^, each of which may be written in the complex field as the product of two linear factors.
We have proved
XXIII.
linear factors
quadratic form
if
and only
if its
X'AX ^
rank is
with complex coefficients is the product of two
r<2
CHAP.
QUADRATIC FORMS
17]
139
SOLVED PROBLEMS
1.
Reduce
q = X'AX = X'
Prom Example
1,
12
10
Chapter
[^1/]
10-8
15,
12
10
10
-8
PY
10
-2100
0-100
c
^\J
-10
-2
-10 -1
4 -1
iD'.P']
4
-1-110
Thus, the transformation
form (17.3).
to the
reduces q to the required form y^
yl^'^yl-
2.
Reduce
We
q = X'AX = X'
4 8
to the
form (17.3).
find
1
[^1/] C
2
1
-4
PY
-i Y
1
1
3.
[D\P']
reduces q to y^ + 8y^ - 2y^
Lagrange reduction.
(a)
2
,...2
2
2x^ + 5^2 + ,
19*3
2{xf+ 2x^(2x^ + 3x^+2x^)1 + 5x^+ 19x1 " 24*4 + ^S^j^g - 8x^x^ - IGx^x^
2{xl+ 2x^(.2x^+Zx^+2x^) + {2x^ + 3x^ + 2x^f\
- 24x^+
8a:â=2
+ I2x^x^ + Sx^x^ + l%x^x^ - 8x^x^ - IGx^x^
+ bxl + 19*2 - 24*2 ^ jg^^^^

~ 8*2^^
g^^^^ _ ^^
"2*3 _
=
2 (*i
+ 2*2 + 3*g + 2*^)2 - 3
{*^ + 2*2(*g+ 4*4)
2 (*i
+ 2*2 + 3*g + 2*^)2 _
(^ + ^ + 4^^)2
+ 4 (*g
\yi =
Irs
+ *3 - 32*2 -
^q_^
- 2*4)2
*i + 2*2 + 3*3
ryi
_ 2(2*^ + 3v +2*,)2
*2 +
+ 2*4
*s + 4*4
*3-2*t
'Educes,
to
00^
2y2_3y22+4y2.
*4
(i)
For the quadratic form of Problem

9
*2 +
4*^*2 +
2,
we have
4*^*3 + 4*2 +
Since there is no term in *2 or *2 but a
tenii in
16*2*3 + **
*2*3
we use
(*^
+ 2*2 + 2*3)2 + 8*2*3
the non-singular transformation
QUADRATIC FORMS
140
Xl
i )
Zl
A;2
22
[CHAP. 17
Z2 + ^3
Xs -
to obtain
(2i
+ 4z2+2z3f'
+82^+82223
422+2z3)2+8(22+2Z3f - 2z2
(z^ +
Now Y
4 2
1
X
2
Z and
from
i),
-1
'1
Thus, the non-singular transformation
4 2
1
y^
+ ^y^ -
-1
2yi^
-1
ii
1
6"
-4
2
I
X\ hence,
y effects the reduction.
4.
Using the result of Problem
2,
[^1/] C
-4
-2
-2
i
2
and applying the tiansfoimations HgCiV^). K2(i+-\/2) and ^3(5 a/2). ^3(2
[^!/]
0-2
1
1
-2
Prove:
-iVl
-V2
k^f2 -\\f2 y
= (^y
[cic']
0-1
q = X'
reduces
iv^
i\f2
2,2
5.
10
we have
),
10
-|
10
V2
ii\/2
to the canonical form
,2
If
a real quadratic form q is carried by two non-singular transformations into two distinct
reduced forms
(i)
y; +
yg
y^
and
<ii)
y' + yl +
yq
yq,i
yj.2
y'
then p = q
Suppose q>p. Let

mation which produces ( ii
X
).
FY
be the transformation which produces (1) and
Then
F~^X
+ binxn
+ b^nXn
+ bn2X2 +
cii% + ci22 +
C21X1 +
filial
bi2X2 +
621X1
622*^2
bjiixi
+ bnnpcn
and
C222 +
+ C2ren
Cin^n
G'^X
cr!.i;i
cn2*:2
cnnXn
GY
be the transfor-
CHAP.
QUADRATIC FORMS
17]
respectively carry (i) and (ii) back into
(iii)
*i2*2+- +
*in^n)^ +
...
+ {bp^'^^ + bp^x^-\-- + bp^Xyf
(CiiA:^
+ c^2*2+- + c^^f +
+ (cq^x^ +
+ l.l^l'''
('^g
r-q+p <
Consider the
+ 6i2 *2 +
bpiH
By Theorem
+ *;^2*2 +
(*;b+i,i'i
"*"
+1,2*2
equations
^g+i,i*i + '^(7+1,2*2 +
't?+2,l*l
+ 6jbn% =
+ ^^+1,2*2 +
(-^11
i +
%22 +
^ri
+ ^2*2
*i
+ bpî,rf-J^
q<p
6.
will also lead to
'^g
M*re =
+ c^^x^
(Ot^.ttg
0!^^).
When
this
solutionis
(Vll + V22+--- + V?In)
+ c.^n'^nf +
+ '^9+2,2*2 +
+ (<^7ii + ^(?22 +
Clearly, this requires that each of the squared terms be zero.
q<p. A repetition
a contradiction. Hence q=p.
contrary to the hypothesis.
x^f
+ l,/!^^
'^q
+ 6^
x^+- + c
Chapter 10, they have a non-trivial solution, say

iii ), we have
IV,
substituted into
"^g
+ *in*n =
''21*1+^22*2 +
Thus,
q.
(''ii*i
*ii
141
Thus,
Prove: Every symmetric matrix A of rank
has
>an)^
But then neither F nor G is non-singular,
of the above argument under the
at least
assumption that
one principal minor of order
different
from zero.
Since A is of rank
r,
it
has
at least
stands in the rows numbered Ji,i2

matrix and let the columns numbered
Now
one r-square minor which is different from zero. Suppose that it

these rows be moved above to become the first r rows of the
i^ be moved in front to be the first r columns.
I^^t
'rfi,s2
rows are linearly independent while all other rows are linear combinations of them. By
taking proper linear combinations of the first r rows and adding to the other rows, these last n-r rows can
be reduced to zero. Since A is symmetric, the same operations on the columns will reduce the last n-r
columns to zero. Hence, we now have
the first
ili2
H2i2
%i2
"Vii
in
which a non-vanishing minor stands
pal minor of
7.
Prove:
matrix
A
C
in the
"vv
upper left hand corner of the matrix.
Clearly, this is a princi-
of rank
of rank
is positive semi-definite if
and only
if
there exists a
such that A = C'C.
Since A is of rank
r,
its canonical form
'--[t:]
Then there exists a non-singular matrix B
QUADRATIC FORMS
142
such that A
of rank
B%B.
and A
Since
Ai = A^
A'i'=
we have A
[CHAP. 17
= B'NxB = B'N^Ny^B = B'N(-N^B
Set
= A^B; then
is
as required.
= C'C
Conversely, let C be a real n-square matrix of rank
r;
then A = C'C is of rank s<r.
Let
its
canonical
form be
N^
diag(di,
^5,0,0
(^2
0)
where each di is either +1 or -1. Then there exists a non-singular real matrix E such that E'(C'C)E =
CE = B = [bij\. Since B'B = N^, we have
A^g
Set
(i =
1,
ii =
s+l,s+2
s)
and
+
*ii
d;>0 and A
Clearly, each
8.
Prove:
If
is positive
*i2
^jn
0'
semi -definite.
is positive definite, then every principal minor of
n)
is positive.
Let q = XAX. The principal minor of A obtained by deleting its ith tow and column is the matrix A^ of
the quadratic form q^ obtained from q by setting x^ = 0. Now every value of qj_ for non-trivial sets of values of its variables is also a value of g and, hence, is positive. Thus, Aj_ is positive definite.
This argument may be repeated for the principal minors

... rows and the same columns of A.
A^j,
obtained from A by deleting two,
A^^j-^, ...
three,
By Theorem
Ai>
VI,
A^j>
0,
0,
...
thus, every principal minor is positive.
Prove: Any ra-square non-singular matrix A = [ay] can be rearranged by interchanges of certain
rows and the interchanges of corresponding columns so that not both p_^ and p_2 are zero.
Clearly, the theorem is true for A of order
when
a^
p^_^ =
Suppose
(a)
j^
Suppose
0.
some CL^
?^
0.
01^ =
and of order
then either (a) some a^^
2.
Moreover,
it is
true for
of order
n>2
or (b) all a^^ = 0.
?^
After the ith row and the ith column have been
new matrix has p_^ = CC^j^ 0.
moved
to
occupy the position
of the last row and the last column, the
Suppose (6) all aî = 0, Since \A\ J^ 0.

and the ith column into the (n-l)st position.
a
a
n, n-i
^n-i,n-i
and
at
least one aî i^

new matrix
0.
Move
_i,n =
In the
Vi,n
^n-i,n
the ith row into the (n-l)st position
\n-i
^-
-a.n-i,n
^^
(^-^^^ '^ ^^''^
PnsPn
Vi,n
Pn^Ô.
Note that this also proves Theorem XV.
10.
Renumber the variables so
that
q =
X'AX = X'
2
13
is regular.
1111
Here
po
1,
pi =
0,
ps =
0,
Since
ps = -4, p* = -3.
2
of p3
The cofactor B22
?^
pi = P2 =
0,
ps ^
0,
we examine
the matrix
the interchange of the second and third rows and of
CHAP.
QUADRATIC FORMS
17]
the second and third columns of
143
A yields
2
2
11
1111
for
11.
which p^ -
1,
p^ =
p^ = -4, pg = -4, p^ = -3.
0,
Reduce by Kronecker's method
Here, X2 has been renumbered as xq and
12 3
15
as xg
X'
;<:g
X.
Here Pq - 1, p^ - 1, p^
and q is regular. The sequence of p's presents one
3, pg - 20, p^ = -5
permanence and three variations in sign; the reduced form will have one positive and three negative terms.
Since each pj /
repeated use of Theorem
0,
XIX
yields the reduced form
PoPiTi + PiPgrl + PsPsrl + PsP^rl
12.
Here A is
of rank 3 and
fflgg
= X'AX = X'
An interchange
i 0.
y\
12
13
- eo^g -
"^yl
looy^
X.
of the last two rows and the last two columns carries
12 13
2
into
2 4 3 6
13
which
in
5
4 3
3 6 2 9
Now
has been reduced
to
XCX
= X'
12 10
Since S is of rank
0.
3, it
can be reduced to
2 1
4 3
3 5
for
form will contain two positive and one negative term.
which p^ =
Since p =
p^ =
1,
1,
p^ =0, pg = -1.
but
= 4
i^
0,
2 4 3
13
The reduced
the reduced
form is by (16.8)
PoPiri +
Pxy-i-zJi
1-2
13.
-2
2-12
Here p^
tive terms.
- 1, p^ -
1,
p^ -
0,
Pg
Consider the matrix B

L
9,
p^ = 27
-2
ij
yl + 4y| -
4y|
1-1
1112
= X'
y-iiP^yl
X.
the reduced form will have two positive and two nega-
of pg.
Since /^gg =
but /3g2
=-3^0
the reduced for
is by (16.8) and (16.9)
PoPiTi^ + 2/332 Pg (y| - y|) + pgp^y^
yl + 54y|
54^3^
- 243y|
QUADRATIC FORMS
144
For a set of real ra-vectors Xi, X^,
14. Prove:
Ai
[CHAP. 17
Xp,
Ai
A]_
A2
;t2-x
A2 ^2
Ag Ai
'
>
Z^ Zi
Xp- X2
the equality holding
and only
if
Xp- Xp
the set is linearly dependent.
if
P
(a)
Suppose the vectors Xi are linearly independent and
0<Z-Z
and
)( 2 X:xA
X-x.
j=i
1=1
X\X':XAX
t-j'
Since this quadratic form is positive definite,
(6)
let
G >
= [i,2
r(A:^-;f,-)^
t
X^x^
i=i
X'GX
J'
0.
Then there
Suppose the vectors X^ are linearly dependent.
Then Z
xp]' f 0.
exist scalars
kp, not all zero,
k^^^.k^
such that
S
f = i=i
4,,jy.^
X..^
and, hence, such that
+ k^Xj
kiXjj-Xi
Thus the system
X2 +
+ kpXj
Xp
= 1.2
p)
homogeneous equations
of
Xj-X^xx + Xa-Xqxq +
'J
has a non-trivial solution xi=
We have proved
that
k;
G >
1
(i = 1,2
+ Xa-X,
'j
'
p),
P ^P
and \G\
p)
0.
To prove the converse of
0.
= 1,2
we need only
(b).
assume
to
reverse the steps of (6) to obtain
^=
^.
0,
if
= 0,
;?,
= 0, (/ = 1,2
and
=
|
S kjXy^ =
Thus,
f= l^k^Xi.
where
p)
p
i= 1
and the given vectors Xj are linearly dependent.
15. Write the following quadratic forms in matrix notation
(a)
xl + 4*1^2 +
34
(6)
2x1 -
6j;i2
<<^)
^s
^1
" 2^| -
3a:
12
2-3
Ans.
(a)
rP
^JX
-3
X'
(b)
(c)
Ans.
17.
2x1
- 6x^x^ +
Reduce by the method
(a)
X'
Ans.
6-2 X
(a)
of
Problem
+ 2y^ - 48y|
is
(6)
Hint. In (c) and (d) use
y,=
a;i
-3
-5
and by Lagrange's Reduction:
1-13-3
13
012
X
(c)
2^2^
+ 4y^
= zg, X2 = zi, xg = Z2.
-1
(d)
2-10
1-3 1-3
(6)
-3
5<
1111
-2 18
yl
whose matrix
+ 2x^ + 8X2^3
2x.^Xg
xi.x^.xs
-2 -4
-4 -3
16. Write out in full the quadratic form in
+ 4j:i%2 + 6;c^*g - ^x^x^
(c)
y^
72 + 8yg
,,
id)
2
Ji
-2
-2
2,2
-y^z^ya
CHAP.
18.
19.
20.
QUADRATIC FORMS
17]
\x = ^'
145
(a)
Show
that
(b)
Show
that the symmetric matrix of a quadratic form is unique.
Show
X'\
that over the real field
Prove:
\x but the matrices have
different ranks.
xf+xl- ixl+ix^x^ and 9xf+2x^+2xl + 6x^x^-6x^x^-8x^x^
are equivalent.
real symmetric matrix is positive (negative) definite if and only

if it is congruent over the real
field to / (-/).
21.
Show that X'AX

Theorem XIX.
22.
(a)
Show
(6)
Show
23.
of
Problem 12
is
reduced
to
X'CX hy X
RX.
where
Ks^K^^{-5)K^â).
Then prove
two real quadratic forms in the same variables are positive definite, so also is their
sum.
that if q^ is a positive definite fonn in x^.x^
xs and 92 is a positive definite form in xsî. xs+2.
..x^, then g = ?i + 92 is a positive definite form in xx,x2
x^.
that if
Prove: If
is any real non-singular matrix, then C'C is positive definite.

XlX = Y'ClCY
Hint: Consider
24.
Prove: Every positive definite matrix

of
25.
26.
Theorem
X.)
Hint: Consider
A can be
D'AD=I.
written as 4 =
CC. (Problems
23 and 24 complete the proof
Prove: If a real symmetric matrix A is positive definite, so also is A^ for

p any positive integer.
Prove: If ^ is a real positive definite symmetric matrix and
CB is orthogonal.
B and C
if
B'AB =1 and A
are such that
= C'C
then
27.
Prove; Every principal minor of a positive semi-definite matrix A is

equal to or greater than zero.
28.
Show
29.
Verify the stated effect of the transformation in each of Theorems
30.
ax^ - 2bx^x^ + cx\ is positive definite
that
By Kronecker's
if
and only
reduction, after renumbering the variables
a>0 and \a\
if
XX
ac-l^ >0.
and XXI.
when necessary, transform each
of the following
into a canonical form.

1
(a)
-1
2
-1
0-1
-1
X'
12 12
12
r
(c)
1112
2
1
4-4
(b)
X'
-4
-3
10-2
-3 X
(d)
X'
2
2
X'
12
-2
(g)
9
3
(/)
X'
-12
(a)
(b)
(c)
(d)
31.
Show
that
p^=p^=p^=p^=l;
^^j
iyl - 16y| + 16y|
yl
jI
q
^yl
+ 4y= - 3y2
8y,
xf-
Gxl -
6^
- Zxl - x^x^ -
xx
Pq = Pi =
(h)
22 =
-!
P3
Pi=l. 2S=-4. P3
(g)
See
(e).
(A)
4y2 - 16y| + 16y| + 12y^
11
Uxx
3_
X'
17(rf).
Po =
2xx
-2
1111
(/)
+ y| + y2
yf
-2
2 -1"
Hint: In (g), renumber the variables to obtain (e) and also as in Problem
Ans.
X'
2-462
3
(e)
XUx
2
4-
-1
-16;
2
Ti
y^^
2,2
72
^3
+ 128y|- 128y=
^xx^
can be
3
4-
"^
factored
chapter 18
Hermitian Forms
THE FORM DEFINED
by
where
is Hermitian
Hermitian form.
X'HX
(18.1)
v=\
j=i
and the components of
The rank
of
__
h^jXiXj,
= hj^
hij
are in the field of
complex numbers, is called an

If
the rank is Kra, the form is
we
shall find that the theorems
called singular; otherwise, non-singular.

If
and
are real, (18.1) is a real quadratic form;
hence,
here are analogous to those of Chapter 17 and their proofs require only minor
changes from
those of that chapter.

Since H is Hermitian, every hî is real and every h^^^x;^ is real.
cross-products h^j'x^Xj and hj^XjXj^,
h^jx^xj
is real.
hjiXjXi
Moreover,
for
the pair of
+ hijXiXj
hijXiXj
Thus,
1.
The values
of an Hermitian form are real.
The non-singular linear transformation X = BY carries the Hermitian form (18.1)
into an-
other Hermitian form
{mH{BY)
(18.2)
Y(WHB)Y
Hermitian forms in the same variables x^ are called equivalent if and only if there exists
a non-singular linear transformation X = BY which, together with Y = IX, carries one of the
forms into the other. Since B'HB and H are conjunctive, we have
Two
H. The rank of an Hermitian form is invariant under a non-singular transformation of

the variables,
and
HI.
Two
Hermitian forms are equivalent
REDUCTION TO CANONICAL FORM. An
if
and only
if their
Hermitian form (18.1) of rank
matrices are conjunctive.
can be reduced
to
diagonal
form
(18.3)
Kjxn
k^y-iLji
'crfryr'
î^^
and real
by a non-singular linear transformation X = BY. From (18.2) the matrix B is a product of elementary column matrices while B' is the product in reverse order of the conjugate elementary
row matrices.
By a
(18.4)
further linear transformation, (18.3) can be reduced to the canonical form [see (15.6)]
%Zi + i^22 +
+ 'zpzp
146
z^+i2^ + i
z-rV
CHAP.
HERMITIAN FORMS
18]
147
of index p and signature p-(r-p). Here, also,

p depends upon the given form and not upon
the transformation which reduces that form to (18.4).
IV. Two Hermitian forms each in the same n variables are equivalent if and
only
they have the same rank and the same index or the same rank and the same
signature.
if
DEFINITE AND SEMI-DEFINITE FORMS. A
non-singular Hermitian form h = X'HX in a variables

and index are equal to n. Thus, a positive definite Hery^y^ + y^y^ + +Jnyn and for any non-trivial set of values of
is called positive definite if its^rank
mitian form can be reduced to

the x's, A>0.
A
a.:e
X'HX is called positive semi-definite if its rank and index

Thus, a positive semi-definite Hermitian form can be reduced
to
and for any non-trivial set of values of the x's, h>0.
singular Hermitian form h =
equal,
rj= p <
i.e.,
yiTi + y2y2 +
+ yrYr
n.
r<
n,
H of an Hermitian form XliX is called positive definite or positive semidefinite according as the form is positive definite or positive
semi-definite.
The matrix
An Hermitian fom
V.
matrix C such that

VI. If
VII.
If
is positive definite if and only if there exists a

non-singular
= C'C.
is positive definite, every principal minor of
is positive semi- definite,
is positive,
every principal minor of
is
and conversely.
non-negative, and
conversely.
SOLVED PROBLEM
1
1.
Reduce
X'
\-2i
2 + 3i
-4 +
Prom Problem
7,
1-2/
+ 2i
to
canonical form (18.4).
13
2f
15,
2-Zi
-4 -
- 3i
-4-2i X
Chapter
2j
2j
o"
'l
2/\/io
i/^To
j/yTo
(-4-4j)/v^
j/a/TO
1/vTo
'^^
3i
-4 +
2i
13
-1
l_
Thus, the non-singular linear transformation

1
BY
,0
reduces the given Hermitian form
to
0'
HC
2/VlO
(-4 +
4i)/V^
1/VIO
-i/yflQ
-j/VTo
1/v^
Jiyr + Jsys - ysys
HERMITIAN FORMS
148
[CHAP.
18
2.
Reduce each
of the following to canonical form.
+ 2il
l-2i
(c)
+
+
3j
3j
i"
(d)
For
(6), first
Ans.
{a)
(.b)
multiply the second row of
3i
3i
2i
2-J
2
2J
by
+J
and add to the
first row.
(-l-20/\/3l^
=
3t
1+j
3
Hint:
3i
1-i
jx
:]^
x'\
(b)
[1
tl
nyi - y-m
V2
(-1 + 30/3
(c)
-l"
-1
1/3
nn
- 7272
(-l+f)/A/2
'l
(d)
(-l +
Obtain the linear transformation
j_ri
Ans.
X= -^\
that X'
1-t
-1
l/v^
1+i
-1
-3+J
-3-J
5.
Prove Theorems V-VII.
6.
Obtain
7.
for
Prove:
BY
Show
Ti
(-i-2i)/v3i
:^'
v/2"Lo
4.
J<iyi
Jb
which followed by Y = IX carries
is positive definite
and X'
11
xi
X2
^n
X-L
All
hi2
hin
^2
/J21
h22
h^n
Use
(4.3).
K.
K.
^nn
(6).
\y
1+j
l+2i
1-i
l-2i
10
Hermitian forms theorems analogous to Theorems XIX- XXI, Chapter
^n
Problem 2 into
ij'
(a) of
is positive semi-definite.
17, for
quadratic forms.
n
_S Tj^-Xj^Xj
1=1 j=i
Hint:
- 7272 " 7373
2/VlO
.0
3.
30/\/^
(-3-2J)/\/lO
1/^2
where
77-
is the cofaotor of h^: in
= \h^j\
chapter 19
The Characteristic Equation of a Matrix
THE PROBLEM.
Let Y = AX,
where A = [o^],
(i,j
= 1,2
Z=
In general, the transformation carries a vector
X
vectors X
be a lineal transformation over F.
n).
x^Y
[%i,%2
into a vector
Y =
JriY
\.%,J-z
whose only connection with
is
possibility of certain
being carried by the transformation into KX, where A is either

which F is a subfield.
a scalar of
or of
some
We
through the transformation.
shall investigate
here the
field 3F of
Any vector X which by
the transformation is carried into KX, that is, any vector
AX
(19.1)
for
which
XX
is called an invariant vector under the transformation.
THE CHARACTERISTIC EQUATION.
\X-AX
(19.2)
From
(XI-A)X
(19.1),
we
obtain
A di^
O^ 2
^21
^ f^2
~
The system
of
homogeneous equations (19.2) has
\XI-A\
(19.3)
A Oil
a-zi
^271
X a-nn
On2
non-trivial solutions if and only if
Oi2
A 022
c-ni
0-tn
ci^n
A Onn
On2
The expansion of this determinant yields a polynomial 0(A) of degree re in A which is known as
the characteristic polynomial of the transformation or of the matrix A. The equation
<^(X) =
is called the characteristic equation of A and its roots Ai,A2
A are called the characterA = Ai is a characteristic root, then (19.2) has non-trivial solutions which are
the components of invariant or characteristic vectors associated with (corresponding to) that
root.
istic roots of i. If
Characteristic roots are also known as latent roots and eigenvalues; characteristic
vectors
are called latent vectors and eigenvectors.
r2
Example
1.
Determine the characteristic roots and associated invariant vectors, given A
The characteristic equation
characteristic roots are
Ai=
5,
A-2
-2
-1
-1
A-3
-1
-1
-2
A-2
is
A2=
1,
A3=
149
1.
A^'
- tA^ + llA
and the
150
When A
Ai =
(19.2)
5,
-2
-ll Vxi
-2
-1
since
becomes
2-1
-1
^2
or
-1
-1
-2
to
-1
X2
-1
-1
hence, associated with the characteristic root

x-i = x^ = xq = i
one-dimensional vector space spanned by the vector [l,l,l]'. Every vector
solution is given by
A=5
Xl
_X3
row equivalent
is
-1
xs
-2 -1
-1
[CHAP. 19
is the
[k.k.kY of this space is an invariant vector of A.
When A
Two
A2=1,
(19.2)
becomes
-2 -l'
Xl
-2
-1
X2
-2
-1
_X3_
Xl + 2x2
+^3
linearly independent solutions are (2,-1,0) and (1,0,-1).
Thus, associated with
two-dimensional vector space spanned by
Xi = [2,-l,0]'
Every vector hX^+kX^ = [2h+k,-h,-ky is an invariant vector of A.
the characteristic root A =
and X2 = [1,0,-1]'.
is the
See Problems
GENERAL THEOREMS.
In
Problem
3,
we prove a special case
(A:
3)
1-2.
of
Xj^
As
Xk ^f^ distinct characteristic roots of a matrix A and if Xi, X^
the
X's
are
linewith
these
roots,
respectively
vectors
associated
invariant
are non-zero
I.
If
Ai,
arly independent.
we prove a special case
In
Problem
k\
where A is re-square, with respect to A is

II. The Ath derivative of
cfi(X) = \XI A\
times the sum of the principal minors of order n-k of the characteristic matrix when k<n,
is
re!
4,
(ra
3) of
when k = n, and
As a consequence
III.
If
of
when k>n.
is
Theorem II, we have
Ai is an r-fold characteristic root of an re-square matrix A, the rank of Xil A

nr and the dimension of the associated invariant vector space is not
is not less than
greater than
r.
See Problem
5.
In particular
Iir.
is
re
Xi is a simple characteristic root of an re-square matrix A, the rank of Xil

and the dimension of the associated invariant vector space is 1.
If
Example 2. For the matrix A
of Ex.
1,
the characteristic equation is
<^(A) =
(A-5)(A-1) =
0.
The invariant vector [l,l,l]' associated with the characteristic root A=5 and the linearly
independent invariant vectors [2,-l,o]' and [l,0,-l]' associated with the multiple root A=l
are a linearly independent set (see Theorem I).
The
invariant vector space associated with the simple characteristic root
A=5
is of
CHAP.
19]
dimension
1.
multiplicity
2,
The invariant vector space associated with

Theorems III and III').
151
the characteristic root
= l, of
is of dimension 2 (see
See also Problem
6.
Since any principal minor of A" is equal to the corresponding principal minor ot A, we have
by (19.4) of Problem
1,
The characteristic
IV.
roots of
A and
A' are the same.
Since any principal minor of A is the conjugate of the corresponding principal minor of A,
we have
V.
The characteristic roots of A and
of
/I'
are the conjugates of the characteristic
roots of A.
By comparing
VI. If
Xi,
scalar, then
VII.
If
Problem
VIII.
A2
AAi,
kX^
we have
A are the characteristic roots of an ra-square matrix A and

AA are the characteristic roots of kA.
A are the characteristic roots of an n-square matrix A and

\-Lk,\2k
\ik are the characteristic roots of 4 -A/.
Ai, A2, ...,
scalar, then
In
characteristic equations,
7,
If
is a
if
A;
if
& is a
we prove
is
a characteristic root of a non-singular matrix A, then \A\/a is a character-
istic root of adj A.
SOLVED PROBLEMS
1.
If
is re-square,
(19.4)
where s^
We
show
0(A)
(m =
1,
that
|A/-^|
A" + 51 A"-^ + 52 A"-^ +
re-1) is (-1) times the
sum
...
+ s_,A + (-l)"|i|
of all the m-square principal minors of A.
rewrite \\! - A\ in the form
A-aii
0-ai2
0-021
A-a22
0-a2n
O-'^ni
O-Ons
X-
a-in
ar
and, each element being a binomial, suppose that the determinant

has been expressed as the sum of 2" determinants in accordance with Theorem VHI, Chapter 3. One of these
determinants has A as diagonal elements and zeroes elsewhere; its value is A". Another is free of
A; its value is (-1)"|^|
The remaining
determinants have m columns, {m = 1, 2, ....- 1), of -A and n~m
columns each of which contains just one
non-zero element A.
Consider one of these determinants and suppose that

of
-A
its
columns numbered h,
ia
are columns
After an even number of interchanges (count them) of adjacent

rows and of adjacent columns, the de-
terminant becomes
152
%.ii
[CHAP. 19
"iiA
"is.ii
"imA
"inA
"im.-im
(-1)'"
I
%.fe
"in-k
t^.to
im!
is an m-square principal minor of
i\
Aj
i
^'2
(-1)"
as
i^
(ii, is
%''''OT
..
where
^n
'i.'2
..
runs through the
n (n
1)
(n
...
-4
Now
S ^'^
H.^
-m + 1)
n taken
different combinations of 1, 2
at a
time.
-4 -1 -4
2.
Use
(19.4) of Problem
We have
to
1+0-2+6
Sl
-4
-3
1
3.
Let
\XI
-2
that Xi, X2,
Assume
--
(1)
8 + 2
A^'
+ 9A^ -
7A
the contrary, that is,
assume
-1
-4
-2
-1
-1
5-4
+
-2
4-1
2.
that there exist scalars 0%, 02, 03
+ 0^X2 + 03X3
A^^
(11)
aiAiî + osAjA^s + a3A3A^s

(i), (11), (111)
may be
not all zero, such that
aiAiî + 02^2X2 + osAgXg
by A and obtain
(III)
we have
aÂXi + oqAX^ + agAXg
Now
-2
A3, ^3 be distinct characteristic roots and associated invariant vectors of A.

X3 are linearly independent.
by A and recall that AX^ = A^
(II)
Multiply
-4 -4
-4
-1
a-]_Xi
Multiply
A'^- 5
(i)
(iv)
+ 16
-2
+
As, ^s)
Ai, A"!;
Show
- A\
-4 -1
-3
-1
-4
2-5+16-9
-1
SS
-4
-2
-1
Then
-1
-1
-1
S2
Ml
5-4
1-2 3
4-1 6
A =
expand |A/-^|, given
written as
1
aiXi
Ai
Ai
Ao
As
An
A3
02X2
ooXq
x\
\l
A\
"3^3
CHAP.
19]
By Problem
have
[aiX-j,
Chapter
5,
a^X^.asXaY
3,
Ai
As
As
\l
Al
x%
B~
hence,
0;
153
Multiplying (iv) by
exists.
B~
we
But this requires 01 = 02 = 03=0, contrary to the hypothesis.
= 0.
Thus. Xi. X^, Xq are linearly independent.
4.
From
(j!)(X)
\\1
A On
-012
-021
A 022
-ai3
- 123
-<31
-032
A- 033
A\
A-O22
-023
- Osi
- O32
A-O33
A-O22
-O23I
032
A Ossl
the
O12
A -Oil
-021
<^'(A)
we obtain
-031
A -Oil
Os 1
sum of the principal minors

1
<^"(A)
-032
A -022
-02
-Ol3
A
of
-012
-O2I
A 022
of order
-O23
two
OSI
c
A -01 1
A-o22
Ois
023
1
A -Oil
Og 3
XI - A
-ai2
A (322
A-a3g
-O32
A -033
-021
A -Oil
-Ol3
A 033
A -Oil
-oia
1
-0 12
1
2!(A-Oll) + (A-a22) + (A-Ogg)!

2!
times the sum of the principal minors of XI
-A
of order
one
0"'(A)
Also
5.
<j!.^"\A)
= cji^^hX) =
...
Prove: If X^ is an r-fold characteristic root of an re-square matrix A, the rank of X^J -A is not less
than n T and the dimension of the associated invariant vector space is not greater than r.
Since Ai is an r-fold root of <;6(A) = 0, ^SCA^) = ct>(X^) = 4' <.\) = = 4>''^~^\x^) =
and cfy'ÂXi) + 0. Now
(^"^HX^ is r! times the sum of the principal minors of order n-r of A^/ -A; hence, not every principal minor
can vanish and X^l-A is of rank at least n-r. By (11.2), the associated invariant vector space otXj^l -A,
I.e., its
6.
null-space, is of dimension at most
For the matrix of Problem
2,
The characteristic roots
find the characteristic roots and the associated invariant vector spaces.
are
1, 1, 1, 2.
For A =
2:
XI
-2
-A
1
1
sion
1.
r.
2-5
-1
is of rank 3;
-3
-4
The associated invariant vector space
is that
spanned by
A=
1:
XI -A
-1
=
1
1-4
sion
1.
The associated
[2, 3,
-4
-2
-1
For
its null-space is of dimen-
-2, -3]'.
is of rank 3; its null-space is of
1-5
invariant vector space is that spanned by
[3, 6,
-4, -5]'
dimen-
154
7.
Prove:
a is a non-zero characteristic root of the non- singular
If
n-
[CHAP. 19
square matrix A, then \A\/a is
a characteristic root of adj A.
By Problem
1,
a"+ .!"-%
(i)
where
s^.
n - 1) is (-1)
= 1. 2
(i
l/xZ-adj/ll
where Sj
= 1, 2
(/
times the sum of

=
By
and the definitions of
llll-M
A\ = \AY' ^ then
(6.4)
s,-
and
S,-
all
Si/^""^
sum
l'^!
f-square principal minors of A, and
+ S^^^/s +
...
(-if
ladj^l
of the /-square principal minors of adj
Sj = (-1)" |-4|s._
Si= (-l)"s^_.,,
-^
To
Jl 1
and
m"
is (-1)-^ times the
1)
^n-i^ + (-1)
(-if Ml"-^si,
|adj
|;x/-adj A\
(-if !(-if ;x"
+ ._,A^"-' + ._2MlAi"-" +
s,|/irV
...
siMf-V
n-i
and
Mr"|/x/-adj
4|
(-if !i
s,(i^) +
...
\A\
(-if
._i(f^f-'
\A\
(^f Ml
/(/^)
\A\
Now
and by
/(R)
(-if ii
a"/(4^)
(-ifia"+ sia"-'+
+ ,,(1) +
._,(lf-
...
(-if (If Ml!
(i)
...
^_,a
(-i)"Ml
Hence, M|/Ot is a characteristic root of adj A.
8.
AAA
Prove: The characteristic equation of an orthoganal matrix
We have
=
</,(A)
|X/-P|
\\PIP'-P\
|-PA(^/-P')|
is a reciprocal equation.
A"|f/-P|
A"c^(i)
9.
For the following matrices
determine the characteristic roots and a basis of each of the associated invariant
vector spaces.
-1
(a)
(6)
12
2
1
-2
-1
-12
1, [1,
-3
2,
[1,3,1]';
2,
[2,1,0]';
1,
[1,1,-1]';
(rf)
1,
[1,1,1]'
(g)
-3 -9
-12
(0
3,
6-10
(A)
9
6
-6
-4
-3 -3
-5
[1,-1,-2]'
7
n)
-1
-1
_i
1_
-1
-6
-2
-3
-1
-1
-1
3'
-5
'
1+t
(/)
'-1
-1
-1
'2
_0
-5 -4
-3 -2
1, [3, 2, 1]'
(O
1_
-1,-2]';
(b) -1, [1,0, 1]';

(c)
2-1
2, [2,
(/)
112-1
-1,0]';
2-4
2-1
-2 -8 -12
(e)
(rf)
(a)
(A:)
Ans.
(c)
2-i
CHAP.
19]
2,
[2,-1,0]';
0,
[4,-1,0]';
(/) 0,
[3,-1,0]';
1,
[12,-4, -1]'
(e)
(g) 1, [1,0,-1]', [0,1,-1]';

(A) 0,
[1,-1,0]';
-1, [0,1,-1]';
(j)
[l
i,
1+t,
(/) 2, [1,0,1]';
[4,0,-1]'
1,
[1,1,0]'
3,
[0,0,1]';
1,
[1,1,0]'
4,
+ j,l,l]';
[l-i.1,1]'
-J,
2-2j, [1,0,-1]'
1,0]';
[0,
(/c)
1,
[1,0,-1,0]', [1,-1,0,0]';
(/)
1,
[1,2,3,2]';
-1, [-3,0,1,4]'
(m)
0,
[2,1,0, 1]';
1,
[-2,4, 1.2]';
2,
AX
[0,3, 1,2]'
3,
-1, [3,0, 1,2]'
[3,0, 1,4]';
XX
X'AX
10.
Prove:
If A" is
H.
Prove:
The characteristic roots of a diagonal matrix are the elements
a unit vector and
if
155
then
of its diagonal and the associated in-
variant vectors are the elementary vectors E^

12.
13.
Prove Theorems
Prove Theorem
Hint.
16.
17.
=(A-Ai)(A-X2)...(A-A) then
The characteristic
14. Prove:
13.
VI.
VII.
|A/-.4|
If
and
Prove:
If
A and
A'
roots of the direct
! are n-square and
equation.
t:]
Prove:
If the n-stjuare
matrix A is of rank
Prove:
If
A and B
sum
are n-sauare and
r.
\(X
+ k)I
diag(.4i,
-A\
A2
r<n, show
then at least n
= (X + k~ X-O (X + k
(X + k -X^^.
A^) are the characteristic roots of A^ A^
that
-r
-X2)
NA
AN
and
have the same characteristic
of its characteristic roots are zero.
A~'^B and BA~'^ have the same characteristic
roots.
18.
For A and B of Problem
X9.
Let ^ be an n-square matrix.
17,
show
that
Write
B and A'^BA have the same characteristic
\XI -A''^\ = \-XA~'^(Î
-A)\
roots.
and conclude that
iAl IA2
lA
are
the characteristic roots of ^4"^
The characteristic
20. Prove:
Hint.
21.
If
are of absolute value
1.
X'
(PX.Y(PX-)
Prove: If A^ ^ \ is a characteristic root and A^ is the associated invariant vector of an orthogonal matrix
P, then XlXi = 0.
22. Prove:
23.
roots of an orthogonal matrix
A-, X. are a characteristic root and associated invariant vector of P, then
The characteristic
Obtain, using Theorem

=
(-i)"M|
<^ (0)
(-1)
(0)
(-1)
0^">(O)
n!
24. Substitute from
times the sum of the principal minors of order n -
r!
of
times the sum of the principal minors of order n -
of
.4
Problem 23 into
c/3(X)
to obtain (19.4).
1.
II,
0(0)
<P
roots of a unitary matrix are of absolute value
(?i(0)
+ (^'(0)-X +-^(f^'(0)-X^
...
+^0'^'^'(O)-A"
chapter 20
Similarity
TWO
ra-SQUARE MATRICES A and B over F are called similar over F

matrix R over F such that
if
there exists a non-singular
B = R'^-AR
(20.1)
Example
11
2J
R~'ÂR
The matrices A
1.
of
Example
1,
-3 -3]
-110
-1
E7
ij
Chapter
19,
and
li
2
133"
12
13
[1
14
13]
iJ
are similar.
The characteristic equation (A 5)(A
An
invariant vector of
1)^ =
of
B associated with A=
is
is also the characteristic equationof ^.
Y^ = [l,0,0]' and
it
is readily
an invariant vector of A associated with the same

X-L
A = 5. The reader will show that Fg = [7, -2, O]' and Yg = [n, -3, -2]' are
arly independent invariant vectors of B associated with A = 1 while X^ = RY2
are a pair of linearly independent invariant vectors of A associated with the same
= RYi =
that
[1, 1, 1]' is
root
Example
characteristic
a pair of lineand Xg = RYg

root A =
1.
illustrates the following theorems:
Two
I.
shown
same characteristic
similar matrices have the
roots.
For a proof, see Problem

II.
If
A^ of B, then
is an invariant vector of
RY
1.
B = R~ AR corresponding to the characteristic root

A corresponding to the same characteristic
root A^ of A.
For a proof, see Problem
DIAGONAL MATRICES. The
characteristic roots of a diagonal matrix
D =
a^)
diag(ai, 02
2.
are
simply the diagonal elements.
diagonal matrix always has n linearly independent invariant vectors.
vectors E^ are such a set since
DE^ = aÊ^,
(i
As a consequence, we have (see Problems

III.
Any
The elementary
1, 2, ...,n).
and 4
for proofs)
re-square matrix A, similar to a diagonal matrix, has n linearly independent
invariant vectors.
IV.
If
an re-square matrix
A has
re
linearly independent invariant vectors,
it
is similar
to a diagonal matrix.
See Problem
In
Problem
V.
\IA
6,
5.
we prove
Over a
field
an re-square matrix A is similar to a diagonal matrix
factors completely in
F and
if
and only
if
the multiplicity of each A, is equal to the dimension of
the null-space of X.I A.

1/
156
CHAP.
SIMILARITY
20]
157
Not every ra-square matrix is similar to a diagonal matrix. The matrix of Problem 6, Chapexample. There, corresponding to the triple root X = 1, the null-space of A/-i
is of dimension 1.
ter 19, is an
We can
prove, however,
VI.
Every ra-square matrix A is similar
to
a triangular matrix
are the characteristic roots of A.
See Problems
7-8.
As special cases, we have

VII.
If
is
orthogonal matrix
any real
n-
P such
that
square matrix with real characteristic roots, there exists an

P'^ AP = P'AP is triangular and has as diagonal elements
the characteristic roots of A.
See Problems 9-10.

VIII. If A is any re-square matrix with complex elements or a real re-square matrix with
complex characteristic roots, there exists a unitary matrix U such that W'ÂU
UAU is
triangular and has as diagonal elements the characteristic roots of A.
=--
See Problem
The matrices A and P' AP
of
Theorem
The matrices A and U'ÂU
of
Theorem VHI
11.
VII are called orthogonally similar.

are called unitarily similar.
DIAGONABLE MATRICES. A
Theorem IV
is
matrix A which is similar to a diagonal matrix is called diagonable.

basic to the study of certain types of diagonable matrices in the next chapter.
SOLVED PROBLEMS
1.
Prove:
Two
similar matrices have the
Let A and B
R~ÂR
A/-B
(')
same characteristic
roots.
be the similar matrices; then
\I -
R'ÂR
R-'ÛR - R-''aR
R'^(XI-A)R
and
\\I-B\
\R-^.\XI-A\.\R\
\XI-A\
Thus, A and S have the same characteristic equation and the same characteristic roots.
2.
Prove:
RY
If
By hypothesis, BY
X^Y and RB
AX
and
3.
A"
Is an invariant vector of
Prove:
B = R ÂR corresponding to the characteristic root

A corresponding to the same characteristic root \,- of A.
is an invariant vector oi
Any matrix A which
ARY
A,-,
then
AR; then
=
RBY
A corresponding
RX^Y
X-RY
X^X
to the characteristic root X^.
is similar to a diagonal matrix
has
re
linearly independent invariant
vectors.
Let R AR = diag(6i, 62. > b^) = B. Now the elementary vectors fii, Eg
^n ^'^ invariant vectors
Then, by Theorem n, the vectors Xj = RE. are invariant vectors of A. Since R is non-singular, its
column vectors are linearly independent.
of B.
SIMILARITY
158
4.
Prove:
[CHAP. 20
an n-square matrix A has n linearly independent invariant vectors,
If
is similar to a
it
diagonal matrix.
X^ be associated with the respective

L,et R = [X^,X2, ....X^]; then
Let the n linearly independent invariant vectors X^. Xq

teristic roots Ai, Xs. /^n ^ '^^"^
AR
,4A'^ = AÂ'^,
AX^]
[AX^.AX^
0'
[Aiî, ^2^2
Ai
1X1,
R ÂR
Hence,
5.
= diag(Ai,
^n^n^
...
A2
X2
...
'^n-'
A2
[Jf 1,
A)
Z2, ^3!
-1
then
10-1
R'UR
-2
1-2
-3
Example
of
1"
-1
2_
-3
,1
1,
Chapter
19, is
^3 = [1,0,-1]'
Z2 = [2,-1,0]',
Zi = [1,1,1]',
R =
diag(Ai, A2
A).
set of linearly independent invariant vectors of the matrix
Take
charac-
n).
1, 2
and
0"
-1
a diagonal matrix.
6.
Prove:
Over a
field
factors completely in
of
F an n-square matrix A is similar to a diagonal matrix if and only if Xl-A

F and the multiplicity of each A.^ is equal to the dimension of the null-space
XÎ-A.
and that exactly k of these characteristic

B
R-'ÂR
diag(Ai, A2,
^n'
Then XÎ - B has exactly k zeroes in its diagonal and, hence, is of rankn-fe; its
thus, A.^/-^ has the same
null-space is then of dimension n-(n -k)=k. But \l-A = R (XjÎ - B) R~^
\Î
B
rank n -k and nullity k as has
Conversely, let Ai,A2
Ag be the distinct characteristic roots of A with respective multiplicities
suppose that
First,
roots are equal to A^.
r,
'i-''2
Take
where
Xi^.Xi^.
exist scalars a
.,Xir-
as a basis of the invariant vector space Vf..
not all zero,
s).
Suppose that there
XJ'
Now
= 1,2
(i
such that
(aiiAîi + 012-^12 + + "iTiîri)
(i)
the associated invariant vector spaces.
Denote by V^^. V^^,
ri+r2+...+rc
...
each vector
Yi
(osi^si
=
ai^Xu
"^
^s^Xs'z
ai^Xi^ +
"^
+
*
(o2i''''2i
"^
+ 022^^22 +
^'T^ ^STg
a' STg
XciT^))
+ air^Xir^) =
0,
''2'r2^2r2)
(i
1,
s).
for
otherwise,
Theorem I their totality is linearly independent. But this contradicts

X's constitute a basis of K and A is similar to a diagonal matrix by Theorem IV.
invariant vector and by
7.
Prove:
Every ra-square matrix A
is similar to a triangular matrix
(i);
it
is
an
thus, the
are the
characteristic roots of A.
A and let X-l be an invariant vector of A corresponding to

column of a non-singular matrix Qi whose remaining columns
may be any whatever such that \Qi\ ^ 0. The first column of AQi_ is AX^ = X-^X^ and the first column of
O]'. Thus,
QlÂQi is Ql^XiXi^. But this, being the first column of Qi'^XiQi. is [Ai,
Let the characteristic roots of A be Ai, Aj
the characteristic root Ai.
Take
X-i
as the
first
CHAP.
SIMILARITY
20]
QiÂQ^
(i)
where
A-^
is of order n
Since
|A/
hi
Bil
[o
A,j
1.
AQ^\ = (X-Xi)\XI - Ai\,
-ft
159
and Qi^ AQi_ and A have the same characteristic roots, it

A. If n = 2, A-i = [A2] and the theorem is proved
follows that the characteristic roots of Ai are Xg. Ag

with Q = Qi.
Otherwise, let X^ be an invariant vector of Ai corresponding to the characteristic root As- Take X^ as
column of a non-singular matrix Qq whose remaining columns may be any whatever such that IftI /^ 0.
the first
Then
=
Q2Â^Q^
(il)
where A2
is of order n
2.
n =
If
3,
/I2 = [As],
[^2
52I
[0
A^j
and the theorem is proved with Q = Q^[0
Otherwise, we repeat the procedure and, after n -
I2
/i
n-2
(iii)
(?2
such that
8.
(J
/1()
is triangular
Q2J
steps at most, obtain
<?3
Qn-i
and has as diagonal elements the characteristic roots of A.
Find a non-singular matrix Q such that
Q AQ
is triangular, given
9-1
8-9
-5
-1
-4
-4
Here |A/-^| = (A^-l)(A^-4) and the characteristic roots are 1,-1,2,-2. Take [5,5,- 1, 3j', an
invariant vector corresponding to the characteristic root 1 as the first column of a non-singular matrix
Qx
whose remaining columns are elementary vectors, say
^5000'
5
10
10
-10
3
Then
-5
and
8-9
5-1
1
1
-3
Qi AQ^
-15
-12
20
16
Aq
[O
17
characteristic root of Ai is -1 and an associated invariant vector is
-l]'.
[4, 0,
Take ft
then
-20
o"!
ft A1Q2
and
II
J_
20
-15
-48
64
-11
48j
characteristic root of A2 is 2 and an associated invariant vector is
20]
[s, ll]'.
r_i
5I
"d
Take Qs
}
[.:
<?3'
H[-11
8
,!
2/5I
"I
sj
and
ft'-42ft
[2
-2J
then
[CHAP. 20
SIMILARITY
160
Now
32
pi
o"|
[o
Q^j
[h
o1
[o
Qsj
Q''
'
160
-1
11
20
-180 40 -220
1_
"l
-40 40
160_
-7 -9/5"
-1
and
2/5
-2
9.
If
any real re-square matrix with real characteristic roots then there exists an orthogonal
that P''ÂP is triangular and has as diagonal elements the characteristic roots of ^.
is
matrix
P such
X be the characteristic roots of A. Since the roots are real the associated invariant
As in Problem 7, let Qi be formed having an invariant vector corresponding to Aj
as first column. Using the Gram-Schmidt process, obtain from Qt an orthogonal matrix Pi whose first column
is proportional to that of Q-^. Then
Let Ai.Xs
vectors will also be real.
hi
P'lAP^
where Ai
is of order n
fill
and has As, A3, ...Ayjas characteristic roots.
Next, form Qq having as first column an invariant vector of Ai corresponding to the root A2 and, using
the Gram-Schmidt process, obtain an orthogonal matrix P^- Then
TAs
Ss"!
[0
A^j
iji-g
After sufficient repetitions, build the orthogonal matrix
for
10.
which P~ÂP
is triangular
pi
0]
|_o
pj
Pi.
pn-2
"I
P_J
with the characteristic roots of A as diagonal elements.
Find an orthogonal matrix P such that

"2
P-ÂP
is triangular
2_
and has the characteristic roots of A as diagonal elements.
Prom Example
to
P""
1"
Chapter 19, the characteristic roots are
1,
5, 1, 1
and an invariant vector corresponding
is [1,0, -1]'.
"
We take
Qi
10
and, using the Gram-Schmidt process, obtain
-1
1
1
Pi
l/v^
/V2
-l/\/2
an orthogonal matrix whose
We
first
column is proportional to
1/1/2"
[l, 0, -l]'.
find
-I/V2
1/a/2
Pl'^Pi
l/>/2
I/V2"
3^/2"
\0
1/1/2"
l/^/T
-1/V2
1/
V2
21/2"
A^j
CHAP.
SIMILARITY
20]
Now Ax has A
a characteristic root and
[l,
161
-\/2]' as associated invariant vector. From Q2 =
r
we
"2=1 1/V3
obtain by the Gram-Schmidt process the orthogonal matrix
L-2/V6
1/\A2
-1/^^
1/^6
1/a/3
2/^6
-I/V2"
-I/a/S"
1/V6"
P-,
X]
\_
rn
Lv2
r-n
ij
2/v/6
Then
I/V3J
is orthogonal and
P~^ AP =
->/2
11.
Find a unitary matrix
f/
such that
All is triangular and has as diagonal elements the charac-
teristic roots of A, given
1 +t
2 - 2J
-6 -
-4 - 6i
+ 3i
-1 + i
-3 -
5 + 5i
of
is
A(A^
4f
6 + 4f
+(-4-J)A+
2i
i)
and the characteristic roots are

"
1
0,
1-i, 3 +
2i.
For A =
0,
take
[l,
-1, l]' as associated invariant vector and form
Qi
-1
The Gram -Schmidt process produces the
1_
unitary matrix
f/i
i/\fz
i/Ve'
-l/vA3
2/v^
1/V3"
l/\/6"
-i/\/2
1/^2"
Now
-2y/2a-i)
~(26 + 24:i)/y/6
1-i
(2 + 3i)/\A3"
3
so that, for this choice of Qi, the required matrix
12.
Find an orthogonal matrix
P such
that
P ÂP
+2J
= Ui.
is triangular
and has as diagonal elements the char-
acteristic roots of A, given

3
The
[1,1,1]',
[1,-2,1]' respectively.
-1
Now
may be taken as
[l,0, -l]'.
these three vectors are both linearly independent and mutually
Taking
l/V^
P
-1/-/2
find
characteristic roots are 2,3,6 and the associated invariant vectors
orthogonal.
we
-1
-1
P~^ AP
= diag(2,3, 6).
in the next chapter.
l/y/J
l/y/6
l/y/3
-2/^/6
l/y/3
l/yje
This suggests the more thorough study of the real symmetric matrix made
SIMILARITY
162
[CHAP. 20
13.
Find an orthogonal matrix P such that f'ÂP is triangular and has as diagonal elements the characteristic
A for each of the matrices A of Problem 9(a), (6), (c), (d). Chapter 19.
roots of
\/Zs[2
2/3
\/^^pi
2/3
-4/3a/2
1/3
\l\pZ
Ins.
(a)
-l/v/2
I/V2"
(rf)
1/^2"
14.
1/V^ -i/'v^
-i/Ve"'
1/V3"
2/V6"
I/a/2"
-1/\/6"
and (rf) are

Chapter 19 and determine those which are similar to a
diagonal matrix having the characteristic roots as diagonal elements.
of
Problem
9,
For each of the matrices A of Problem 9(j). (/), Chapter 19, find a unitary matrix V such that
angular and has as diagonal elements the characteristic roots of A.
(i)
l/v^
-(l+i)/2
l/>/2
a-i)/2^/2
\/\[2
(1
17.
Make the necessary modification
18.
Let B^ and C^ be similar matrices
is real
S
are similar.
Let
is tri-
(/)
\/^[2
\/y[2
and symmetric and P is orthogonal, then P''^ AP is real and symmetric.
Prove:
If
U'ÂU
-1//!
\l\p2.
i
2
-0/2/2"
16.
of Problem 9 to prove
Suppose C^ =
Hint.
= diag(Bi, Sj)
for
(i
and
-1
i?^
B^ fl^ and form
B
B
to
that
and
VIII.
C)
= diag(Ci, C2
= diagCSj, B^).
Extend the result of Problem 19
Theorem
Show
m).
1, 2
= diag(5i, B2, ...,S)
of Bi and B^ respectively, and define
20.
1/^6"
of Problem 13 are similar to a diagonal matrix while (c)
(6)
Examine the matrices {a)-{m)
Ans.
19.
1/^
I/V3"
1/V2"
Explain why the matrices (a) and
not.
15.
2/>/6
-1/1/3
-1/^/6""
i/Va"
(c)
-l/V'2
(b)
1/-/2
I/V3"
Write
ij
= diag(Bi,
/ = diagC/i./g),
Show
Bg
where the orders of
BTÊK
that
B^).
to prove
/i
and
/g are
B and C are
those
similar.
oJ
= diag(Bi,
Bg
B)
and C any matrix obtained by rearranging the
S^ along the diagonal.

21. If
A and B are n-square, then
Hint.
Let P AQ =N; then
AE
and B/1 have the same characteristic roots.
PABP"^ =NQ''^BP''' and Q~^ BAQ
Q'^BP'^N.
See Problem
15,
Chapter 19
A^ are non-singular and of the same order, show thut A-Â^-.. A^, A2As-.. AÂ^, A3... AÂ-^Â^.
have the same characteristic equation.
22. If A-i.A^
...
23.
Q'ÂQ
Let
of
(a)
B where B
is triangular
and has as diagonal elements the characteristic roots Ai, X2
Show that
S
A Q
is triangular
and has as diagonal elements the kth powers of the characteristic
roots of A.
(b)
.4.
Show
that
X,-
= trace
24.
Show
that similarity is an equivalence relation.
25.
Show
that
'2
_1
2_
-1
2-1
and
-3
-2
have the same characteristic roots but are not similar.
chapter 21
Similarity to
REAL SYMMETRIC MATRICES. The

we
be combined but
study of real symmetric matrices and Hermitian matrices may
shall treat them separately here.
The characteristic
I.
a Diagonal Matrix
For real symmetric matrices, we have:
roots of a real symmetric matrix are all real.
See Problem
The
II.
1.
invariant vectors associated with distinct characteristic roots of a real sym-
metric matrix are mutually orthogonal.
See Problem
When A
is real
and symmetric, each B^ of Problem
9,
Chapter 20, is
0;
hence,
is a real ra-square
symmetric matrix with characteristic roots Ai, \^,

then there exists a real orthogonal matrix P such that P'AP = P'ÂP = diag(Ai, Aj
If
III.
Theorem
IV.
III
If
A^ is a characteristic root of multiplicity
In terms of a real quadratic form,
A,
A).
r^
of a real
Theorem
III
symmetric matrix, then
r^.
becomes
V. Every real quadratic form q = X' AX can be reduced by an orthogonal transformation

to a canonical form
BY
(21.1)
where
the
...,
implies
there is associated with A^ an invariant space of dimension
2.
Aiyi + X^yl +
r
is the rank of
A and
Ai,
Aj
...
A^ are
+ X^y"^
its
non-zero characteristic roots.
Thus, the rank of q is the number of non-zero characteristic roots of A while the index is
number of positive characteristic roots or, by Descartes Rule of signs, the number of varia-
tions of sign in |A/

VI.
^|
0.
real symmetric matrix is positive definite if and only if all of its characteristic
roots are positive.
ORTHOGONAL
SIMILARITY. If P is an orthogonal matrix and 8 = P"^ AP then B is said to be orthogonally similar to A. Since P~^ = P', B is also orthogonally congruent and orthogonally
equivalent to A. Theorem III may be restated as
.
VII. Every real symmetric matrix A is orthogonally similar to a diagonal matrix whose
diagonal elements are the characteristic roots of A.
See Problem
3.
Let the characteristic roots of the real symmetric matrix A be arranged so that Ai ^ A2 =
Then diag(Ai, A2, . A.) is a unique diagonal matrix similar to A.
... Â.
The totality of
such diagonal matrices constitutes a canonical set for real symmetric matrices under orthogonal
similarity. We have
VIII.
the
Two
real symmetric matrices are orthogonally similar if and only if they

roots, that is, if and only if they are similar.
same characteristic
163
have
164
PAIRS OF REAL QUADRATIC FORMS.
AX
IX. If X'
and
X'BX
In
Problem
4,
[CHAP. 21
we prove
are real quadratic forms in
(x-^^, x^
x^) and if
tive definite, there exists a real non- singular linear transformation X = CY
X'AX
X'BX
is posi-
which carries
into
AiXi + A2Y2 +
and X' BX into
22
yi
where A^ are the roots of
+ 72
+ Ay
+ Tn
\\BA\=0.
See also Problems
HERMITIAN MATRICES.
X.
Paralleling the theorems for real symmetric matrices,
4-5.
we have
of an Hermitian matrix are real.
See Problem
XI.
The
7.
invariant vectors associated with distinct characteristic roots of an Hermitian
matrix are mutually orthogonal.

XII.
If
is an re-square Hermitian matrix with characteristic roots Ai, A2.
exists a unitary matrix
U such
is called unitarily similar to

XIII. If
U'HU = U~^HU =
that
U'
HU
A,-
A.) is
r-
an invariant space of dimension
Let the characteristic roots of the Hermitian matrix
Then diag(Ai, A2
A).
...,
A, there
The matrix H
is a characteristic root of multiplicity
there is associated with
diag(Ai, A2,
of the Hermitian matrix H, then

r-
be arranged so that Ai i A2 ^
a unique diagonal matrix similar to H.
The
totality of
i Ansuch diago-
nal matrices constitutes a canonical set for Hermitian matrices under unitary similarity.
There
follows
Two
XIV.
Hermitian matrices are unitarily similar
characteristic roots, that
NORMAL MATRICES. An
diagonal,
and only
if
real
if
and only
they have the same
if
they are similar.
n-square matrix A is called normal
symmetric,
real
is, if
if
AA' = A'A. Normal matrices include
skew-symmetric, orthogonal, Hermitian,
skew-Hermitian, and
unitary matrices.
Let .4 be a normal matrix and f/ be a unitary matrix, and write B = U'AU.

and B'B = U'A'U -U'AU = U'A'AU = U'AJ'U = U'AU -U'A'U = BB'. Thus,
XV.
In
If
Problem
XVI.
mal
8,
is a
normal matrix and
is
a.
unitary matrix, then
U'AU
Then B' = U'A'U
is a normal matrix.
we prove
X^
If
rnatrix A,
is an invariant vector corresponding to the characteristic root A/ of a northen X^ is also an invariant vector of A' corresponding to the characteristic
root A^.
In
Problem
XVII.
9,
we prove
square matrix A is unitarily similar to a diagonal matrix
if
and only
if
is
normal.
As a consequence, we have
XVIII.
If
is normal,
the invariant vectors corresponding to distinct characteristic
roots are orthogonal.
See Problem
10.
CHAP.
211
MX.
If
A- is a characteristic root of multiplicity
ated invariant vector space has dimension
r^
165
of a normal matrix A, the associ-
r^.
XX. Two normal matrices are unitarily similar
if
and only
they have the same char-
if
acteristic roots, that is, if and only if they are similar.
SOLVED PROBLEMS
1.
characteristic roots of an re-square real symmetric matrix
The
Prove:
Suppose that h+ik
which
BX
is real
is a
complex characteristic root of A.
{(h
and singular since
+ ik)I-A\[(h-ik)I-A\
{h
+ ik)I
-A
A are
all real.
Consider
(hl-Af+k'Î
There exists a non-zero real vector
is singular.
such that
and, hence,
X'BX
The vector
X\h!-AfX
(/(/- 4)
+ k^X'X
is real; hence,
A"
{(hi
X'(hl-A)'(hl
-A)X
- A)X\' {(hi -A)X\ ^
+ k^X'X
Also,
0.
X'X>0. Thus,
A:
and
there are no complex roots.
2.
Prove:
distinct characteristic roots of a real symmetric
The invariant vectors associated with
matrix A are mutually orthogonal.

Let
A'l
As of 4.
AX^
and X^ be invariant vectors associated respectively with the distinct characteristic roots Xi and
Then
=
Aiî
AX^
and
= Aj'^s.
A^2'4î =
also
X-iX^X^
and
XjÂX2 = XqXiX^
Taking transposes
^1 A X^ = A 1 A*! ^2
Then
3.
X^X^X^
and
X[X2
and, since Ai ^ A2,
A2îA^2
XqAX-j^ =
X2X2X1
Thus, X^ and X2 are orthogonal.
= 0.
Find an orthogonal matrix P such that P'ÂP is diagonal and has as diagonal elements the characteristic roots of
A given
,
^7-2
-2
10
-2
-2
is
A-7
-1
A -10
-1
A-7
and the characteristic roots are
A - 24A^ + I8OA - 432
6, 6, 12.
- -
For A =
6,
we have
-1
Xl
-4
X2
-1
X3
variant vectors the mutually orthogonal pair X^ = [l,

[1, -2,1]' as associated invariant vector.
or
0,
Xi
- 2x2 + xs
-l]' and
X2
= [l,
1, l]'.
and choose as associated
When A
= 12,
in-
we take X3 =
166
Using the normalized foim of these vectors as columns
1/VT
P
-i/x/T
is left as an exercise to
It
4.
Prove:
X'AX and X'BX
If
show
that
Xiyi + \2y2 +
\\B-A\ =0.
By Theorem
^'B^
^'^'^
^nTn
= diag(l/\/7H, X/sfJT^
for the real quadratic form

it
l/yjl
6, 12).
+ y^, where Xi, Ag, .,
...
GV
Then V =
BGH)W
G'AGH)W
(H
HW
which carries X'BX into
carries
(i)
into
+ wt
there exists an orthogonal transformation
X^y2 + X^y^ +
...
Y (K~ IK)Y
-'i
Since for
values of
all
KY
which
XK'h'g'BGHK - K'h'g'AGHK
=
=
follows that Xi, Xg
Prom Problem
3,
are the roots of |XB-/l|
there exists a real non-singular trans...
and X'BX into
+ Xy^
-'2
'n
diag(X,X
diag(X-Xi,
X)
- diag(Xi,X2
X-X2
X-X)
=0
the linear transformation
\/^pr
+ X^y^
X,
K'h'g'(\B-A)GHK
(GH)W
-l/\f2
l/vA3"
l/sfQ
l/VT
-2/'/6"
IATs
1/v^
\/2\fl
_-l/2>/3
carries
are the roots of
+ Ait;^
X are the characteristic roots ot H G AGH. Thus,

CY = GHKY which carries X'AX into X^y^ + X^y^ +
Y'(K'h'g'BGHK)Y
5.
into
where Xi.Xg
it
) and if X'BX is positive defi= CY which carries Z'/lZinto
transformation
jji^vf + fJ.2V2 +
(H G
Y'(K'h'g'AGHK)Y
formation
l/VT
into yi + y| +
l/i//!^)-
(ii)
carries
-2/sf6
/x are the characteristic roots (all positive) of B.
jjii.jj.^
X/\f&
I/VT
VII there exists an orthogonal transformation
V'(G'BG)V
Let
we have
are real quadratic forms in (xi,x2
(i)
where
1/x/y
= (iiag(6,
there exists a real non-singular linear
nite,
Now
P~ÂP
of
[CHAP. 21
X'BX =
X'
The same transformation
-2
10
-2
-2
1//6"
1/2x^3"
1/3n/2
1/(
1/3-/2
-1/:
1/3\A2
1/f
into
ir'ZIF.
carries
1/3
X'AX
into
1/^r
tr'
0^
X)
CHAP.
21]
Since this is a diagonal matrix, the transformation
Thus, the real linear transformation
and the quadratic form X' AX

into ri + y| +
7l
\kB-A\ = 36(3\-l)(2X-l)^.
6.
CY
KY
167
of Problem 4 is the identity transformation
into ^y^ + iy| + |y|.
A = CP
Every non-singular real matrix A can be written as

symmetric and P is orthogonal.
where C
Since A is non-singular, AA' is positive definite symmetric by Theorem X, Chapter

Q"''
with each
k^) = B
AA' Q = diag(A;i, ij
Q such that
diag(\/^,
VT^
C^
C'ÂA'C'^^ C ^C'^C^
is orthogonal as required.
Prove:
pose
Then
C'Â.
The
QBQ-
QB^Q-^
and
= I
Then there
17.
A:^
>
Thus A
0.
Define
symmetric and
PP'
that
8.
QB^Q-^QB^Q-^
Now C
= QB-^Q''^.
positive definite symmetric and
Define
7.
and
\fk~^)
that
exists an orthogonal matrix

=
show
is left as an excersise to
It
= lY.
X'BX
carries the positive definite quadratic form
{GH)Y
Prove:
Bi
AA'
is orthogonal.
CP
with C
characteristic roots of an Hermitian matrix are real.
Let \j be a characteristic root of the Hermitian matrix H. Then there exists a non-zero vector X^ such
Now X'^HX^ = XjX'j^Xiis real and different from zero and so also is the conjugate transHXj^ = A jX^
.
\^X\X^.
X\HX;j^ =
Thus, X^=X.j andXîsreal.
If X: is an invariant vector corresponding to a characteristic root A,- of a normal matrix A,

then Z^ is an invariant vector of A corresponding to the characteristic root X^.
Prove:
Since A is normal,
(KI-A)(XI-A')
(XI-A)(XI-A)'
-XA+
XXI - XA'
so that XI
-A
is normal.
(SX^)'(SA'i)
By hypothesis, BX;
=
X-B'-BXi
X'iB.B'X^
-A)X^=
= (XÎ
(W X ^)' {W X ^)
XA
XXI - \ a'
a' A
0;
(XI -A)' (XI
Prove:
An
re-square matrix
Suppose A
is normal.
By
AA
-A)
then
and
B'X^
Thus, X: is an invariant vector of A' corresponding to the characteristic root X
9.
is unitarily similar to a diagonal matrix if
(X^/-I')Zi
^.
and only
if
A is normal.
Theorem vni. Chapter 20, there exists a unitary matrix U such that
^1
6i2
bin
As
bin
U AU
....
B'b
By Theorem XV, B is normal so that B' B = BB

is XiXi while the corresponding element of BB
Now
Xfi-i
bn-i. n
the element in the first row and first column of
is
AjAi + 612^12 + bigbis +
...
bmbm
Since these elements are equal and since each b^j b-^j 2 0, we conclude that each b^j = 0. Continuing
we conclude that every b^; of B
with the corresponding elements in the second row and second column,
A). Conversely, let A be diagonal; then A is normal
is zero. Thus, B = diag(Ai,A2
168
10. Prove:
If
[CHAP. 21
is normal, the invariant vectors corresponding to distinct characteristic roots are
orthogonal.
Let Ai,
A*!
and Xj, ^2 be distinct characteristic roots and associated invariant vectors of A. Then
= \^X^ and, by_Problem 8. jI'A'i =Xiî,_I'A'2 = AaATa.
Now ^2/lA'i = Aiâî and,
AXi=\iXi, AX^
X[a' X^
taking the conjugate transpose,

and, since
11.
XiX2=
Xx iXo.,
x^ -
Consider the conic
X[a' X^
But
= X^X'-^X^.
- 4.xi = 40
12xix.2.
X'AX
(i)
referred to rectangular coordinate axes
of
OX^ and OX2.
A-1
(A-5)(A + 8)
For the characteristic roots Ai = 5 and A2 = -8, take
-2/\flf\
/v/T3
'VTs
The conic
form the orthogonal matrix
The transformation
vectors after normalization.
-6~]
40
::]
A is
|A/-^|
Now
Thus, Xi^Â's = Xaî-^s
or
[-:
invariant vectors.
\^X[X2.
as required.
3/x/T3
PY
and
[2, 3]'
tsVTs
'__
2/V/T3I
-2/v/T3
3/vT3j
"
reduces
'^lll
respectively as associated
whose columns
fs
o"!
^-y|
y'\
13 J
are the two
to
(i)
13"!
Vis'
-\y
3/\/l3.
2/\/l3
3/VT3.
[3, -2]'
40
5yl
-Sj
Lo
is an hyperbola.
Aside from the procedure, this
is the familiar rotation of
axes
in plane analytic
geometry to effect the

Note that by Theorem VII the result is
elimination of the cross-product term in the equation of a conic.
known as soon as the characteristic roots are found.
12.
One problem
of solid analytic geometry is that of reducing, by translation and rotation of axes,
the equation of a quadric surface to simplest form.

to
determine the principal directions,
tempting to justify the steps,
The main tasks
are to locate the center and
the directions of the axes after rotation.
i.e.,
we show here
Without
at-
the role of two matrices in such a reduction of the
equation of a central quadric.

Consider the surface
Zx^ + 2xy + 2xz + iyz - 2x - 14y + 2z "
"
and
characteristic equation of
1A/-.11
The characteristic roots and associated

fi =
~.
-;=|
-1
-7
-1
-7
-9
all the terms.
is
A-3
1,
formed respectively from the terms of degree two and from
Ai =
and the symmetric matrices
The
A2
-1
-1
-1
-2
-1
-2
unit invariant vectors are

= 4,
V2
-2,
V3
^N-TfJ
CHAP.
21]
169
Using only the elementary row transformations Hj(k) and H^:(k), where
3
-1
r-
Bi
-1
-7
-1
-7
4,
-4
1
--VJ
-9
ft]
-4
'3x +
Considering B^ as the augmented matrix of the system of equations
X + 2y
\^
from Di the solution x = -1, y =
The rank
of
is 3
0,
z =
C(-l,
or
and the rank of B is
4; the
XtX'^ + X^y'^ + XsZ^ + d
The
x = x'
of translation are
principal directions are
di,
y = y',
1,
we have
D^,
111
4F
z = z'
2Z
find
0, 4).
The required reduced
- 4
4.
Denote by E the inverse of
1-2. i^s-
we
d = -4.
quadrlc has as center C(-l,
equation is
The equations
Prom
0, 4).
0-1=0
+22-7=0
+1=0
y +
\_v\.v.2,v^.
The equations of the
rotation of axes to the principal directions are
\_X
Y Z\- E
[X Y Z]
i/\A3"
-i/i/s"
-1/vT
2/vT
i/vT
i/vT
-i/x/T
i/VT
13.
For each of the following real symmetric matrices A find an orthogonal matrix
nal and has as diagonal elements the characteristic roots of A.
,
-1
2
(a)
(b)
-1
2_
_1
(c)
2_
1/VT
(a)
-4
-4
_ 2
-2
l/\f2
(b)
i/VT
-1/VT
-i/v^T
(d)
14.
i/\/^
Ans.
()
2/3
1/3
2/3
-2/3
2/3
1/3
-1/ 3
--2/3
2/3
21
-2
id)
-1_
\/sf&
l/v/
-2/vT
l/v/s"
1/\A6"
i/a/s"
(e)
4_
(c)
1/vT
P
that /""^.^P
is diago
2"
_2
i/x/T
(e)
"3
P such
"4
-1
-1
_1
-1
r
-1
4_
2/3
1/3
-2/3
2/3
1/3
1/3
2/3
-2/3
2/3
i/vT
2/^/6'
-l/v/T
yV 6
i/VT
-1/\A2
Find a linear transformation which reduces X'BX to

X' AX to Xtyl-^X^y^ + Asy^, where
yl-^y'^-^y'l and
of \XB - A\ =0, given
Xi are the roots
-2
{a)
[7
-2
1
10
-2
l/\/T
/4ns.
(a)
-\/\[2
r2
l\
-2
5=03
7J
\\
1/3VT
1/3
-2/3x^2"
1/3
1/31/2"
1/3
fj
(6)
-4
2J
(6)
-4
-4
-4
-8
_-4
-8
2/3
2/3
l/3x/l0
2/3
1/3
2/3^/10
1/3
-2/3
2/3\/l0
1_
2
4
170
15.
21
Prove Theorem IV.

Hint.
'î
/'"^^P = diag(Ai,
If
-^r + i'
'^1
Xi,
A.1
^1 ~^v)
~'^r+2
'^ ^
Modify the proof in Problem
17.
Prove Theorems XII, XIH, and XIX.
18. Identify
2 to
20x2-
(h)
3x^+ 2%2
24xi:>;2
27%^
+ 3%2
369,
"i'
= diag(0,
0,
r.
(a)
Each characteristic
(6)
+ ^ is non-singular,
(/
root of
\Q?,
(''^
*i + 2^:1x2 + Xg
is either zero or
900,
a pure imaginary.
non-singular.
/ -.4 is
(See Problem 35, Chapter 13.)
+ /iri(/-4) is orthogonal.
normal and non-singular so also is
20.
Prove:
If
Is
21.
Prove:
If
is normal, then
22. Prove:
square matrix A is normal
x\ - 312%ia:>, + 17 %|
(c)
Prove:
Let A be real and skew-symmetric.
B =
P"\Ai/ -/1)P
"^S"
^ri)'
'^r-i-2
each locus:
(o)
(c)
X^ + i-
''^"'^
prove Theorem XI.
16.
19.
[CHAP.
is similar to
and only
if
A~'^
if it
H + iK, where H
can be expressed as
and
commu-
are
tative Hermitian matrices.
23. If
is n-square with characteristic roots
roots of
ty =
for
Prove:
If
X2A2
^n^rv
where
are XiAi,
C/"^<4(7 =
Write
Hint.
24.
AA
J
T*
= [. ],
then
is normal if imd only if the characteristic
Xi, A,2
A.,
is unitary
and T is triangular.
_
C/
Now
tr(rT') =
tr(-4/l')
requires
AA'
is positive definite Hermitian.
Restate the theorem when A
is real
and non-singular.
25.
Prove:
A and B are n-square and normal and
If
if
A and S' commute, then
26. Let the characteristic function of the n-square matrix
and suppose there exists a non-singular matrix
P'^ AP
(/)
Define by
B.-
and Xj,
(j
= 1, 2
s) the
and
BA
are normal.
A be
(X-x^fîX-x^f'^ ...a-'Kfs
<^(X)
^B
P such
that
diag(Ai/.ri' '^2-'r2
n-square matrix diag(0,
-^sV^)
0, ...,
0,/
0) obtained
by replacing A. by
'i
r a
i).
in the right
by
member
T?
E,
Show
DR. ,
PBiP~^.
= 1,2
s)
that
(a)
P'ÂP
(6)
~-
AiBi
A282+
...
(c)
= Ail + A22 + +'^3^5

Every ^ is idempotent.
(d)
EÊj
(e)
1 + 2 + + E3
=0
for
A5B,
^j.
=
rank of ^ is the multiplicity of the characteristic root A^.
(/)
The
(g)
(Xil-A)Ei
(h)
of (/) and define
IfpM
Hint.
= 0,
(J
= 1,2
s)
is any polynomial in x. then
Establish
^^
p(A)
Aii + A22 +
p(Ai)i + p(A2)2 +
xIe
+ P(^s^^s-
Aii + A22 +
+ XÊ^.
CHAP.
21]
Each E^
(i)
(/)
is a polynomial in A.
Define
Hint.
matrix
Hint.
/(A)
-Xi)(A-X2)
commutes with A
fi
B commutes
If
(A
with
if
it
(A -X^)
(k)
If
is normal, then each Ei is Hermitian.
A-^
If
and
/^(A)
/(A)/(A-A^),
(i
= l,2
Then
s).
and only if it commutes with every ^

commutes with every polynomial in A.
(I)
(m)
171
Al%
A-2%
+ A-^fig
...
is positive definite Hermitian, then

,1/2
A^'^
/f
vTii
VA22
...
VAs^s
is positive definite Hermitian.
Equation
(n)
27. (o)
(6) is called the spectral
24
-20
-20
24
10
-10
"
//^
Obtain
Use Problem
Prove:
If
-4/9
4/9
-2/9
2/9
-2/9
1/9
20
29
10
-10
10
44_
38/9
0/9
-10/ 9
26
and
Prove:
The square matrix A
such that
32.
Prove:
33.
Prove:
If .4 is
34.
Let A be normal, B =
4/9
-2/9'
5/9
2/9
-2/9
2/9
8/9
-10/9
23/9_
B commute.
(/).
by H^ = AA'
Hermitian matrix
36. If
5/9
4/9
10/9'
-20/9
-20/9
If
U and a positive
definite Hermitian matrix
f/
= H~'Â.
is
normal
if
and only
if
and U of Problem 29 commute.
is similar to a diagonal matrix if
H~ÂH
and only
if
there exists a positive definite
is normal.
real symmetric (Hermitian) matrix is idempotent if and only if its characteristic roots are O's
(a)
Prove: If
+ 4
-lO"
29
Prove:
Prove:
2/9"
is non-singular then there exists a unitary matrix
Define
Hint.
35.
is unique.
A= HU
such that
31.
4/9
Prove: If A is normal and commutes with B, then A' and

Hint.
30.
49
20
--
29.
it
A-^
Obtain
9_
38/9
28.
that
-4/9
lO"
-10
196
(c)
Show
Obtain the spectral decomposition

"
(b)
decomposition of A.
tf
real
A and
symmetric (Hermitian) and idempotent, then

I
is n-square, the set of
(6)
(6)
Every diagonal element of

of values of A.
(c)
If
(d)
If
is real
is real
= tiA.
is unitary.
~iH)
numbers X' AX where
(a)
I's.
+ A be non-singular, and C = B^^B'.
(S')"^ commute,
is Hermitian, then {I + iHT'^{I
r.
and
of
is unitary.
is
2i
unit vector is called the field of values ot A.
A are in its field of values.

A and every diagonal element
of
U'ÂU, where U
symmetric (Hermitian), every element in its field of values is
symmetric (Hermitian),
values
least and Ais the greatest characteristic root of A.
its field of
is the set of reals
Prove:
is unitary, is in the field
real.
Ai i A < A, where Ai
is the
chapter 22
Polynomials Over a Field

POLYNOMIAL DOMAIN OVER
commutative with
to be
Let X denote an abstract symbol (indeterminate) which is assumed

and with the elements of a field F. The expression
F.
itself
n-i
n
(22.1)
f(X)
where the
a^ are in
cirik
is called
On-^X
a^X +
p
a^X
a polynomial in \ over F.
(22.1) is called the zero polynomial and we write f(\) = 0. If a^ ^

of
degree n and a is called its leading coefficient. The polynomial /(A)
(22.1) is said to be
=
said
to
be of degree zero; the degree of the zero polynomial is not defined.
Oq^
oq ^
is
If
If
every
a =
oj;
polynomial is called raonic.
in (22.1), the
polynomials in A which contain, apart from terms with zero coefficients, the same
terms are said to be equal.
Two
The
totality of polynomials (22.1) is called a polynomial
SUM AND PRODUCT. Regarding
the individual polynomials of
tem, the polynomial domain has most but not

f(X) + g(\)
If
(i)
f(X) is of degree
^(X) + g(A)
is of
and g(X) is of degree
degree
TO
when m.>n,
F[X\ as elements
all of the properties of a field.
of a
number sys-
For example
f(X)-g(X) = g(X)-f(X)
and
= g(X) + f(X)
domain F[\] over F.
n,
of degree at most
m when m=n, and
of degree
re
when m<n.
(ii)
/(A)-g(A)
If
/(A)
If
g(A) ^
QUOTIENTS.
In
I.
?^
is of
while
m + ra.
f(X)-g(X) =
0,
1,
then
A(A) =
A:(A)
we prove
/(A) and g(X) ^
are polynomials in F[A], then there exist unique polynomials
and r(A) in F[A], where r(A)

that of g(A), such that
/((A)
(22.2)
g(X) =
then
and h(X)-g(X) = h(X)-g(X),
Problem
If
degree
/(A)
is either the
zero polynomial or is of degree less than
h{X)-g{X) +
r(A)
Here, r(A) is called the remainder in the division of /(A) by g(A).

to divide /(A) while g(A) and h{X) are called factors of /(A).
172
If
r(A) = 0, g{X) is said
CHAP.
22]
Let
= h(X)-g(X).
/(A)
When
Example
Over the rational field A, -3 is irreducible; over the real field

Over the real field (and hence over the rational field) A^+4
field it is factorable as (\+2i)(\-2i)
1,
THE REMAINDER THEOREM.
Let f(X) be any polynomial and g(X) =
(22.3)
By
is free of A.
n.
When
(22.3),
f(a) =
/(A) is divided by
h(X)-(X-a) +
/(A)
r
when g(A,) = c, a constant,

F is called irreducible over
of degree zero, that is,
non-constant polynomial over
only factorization is trivial.
if its
where
g(A.) is
the factorization is called trivial.
173
r,
is factorable
it
is irreducible;
X-
a.
Then
as (X+\/3)(X-^j3).
over the complex
(22.2)
becomes
and we have
A- a
until a remainder free of
A is obtained, that remainder
is f(a).
m. A
polynomial /(A) has
GREATEST COMMON
DIVISOR.
If
X-
a as a factor
h(X) divides both f(X)
and only
if
and
if
f(a) =
0.
g(A), it is called a
common
divisor of
f(X) and g(A).
polynomial d(X) is called the greatest common divisor of
(i)
d(X) is monic,
(ii)
d{X) is a
(Hi)
common
/"(A)
and g(X)
if
divisor of /(A) and g(X),
every common divisor of /(A) and g(X) is a divisor of d(X).

In
Problem
2,
we prove
IV. If /(A) and g(X) are polynomials in F[X],
a unique greatest
common
not both the zero polynomial, they have
divisor d(X) and there exist polynomials h(X) and ^(A) in F[X]
such that
(22-4)
d(X)
h(X)-f(X) +
k(X)
g(X)
See also Problem
When the only common divisors

is
d(X) =
of
/"(A)
3.
and g(X) are constants, their greatest common divisor
1.
Example?. The greatest common divisor
A +3A+5. and
of /(A)
(A^
+4)(A^+3A+
and g(A) =
5)
(A^
-1)(A^ + 3A+5)
is
(22.4) is
A.=
3A +
i/(A)
- lg(A)
5
We have also
(1-A)-/(A) + (X+4)-g(X)
0.
This illustrates
If the greatest common divisor of /(A) of degree re>0 and

g(X) of degree m>0 is
there exist non-zero polynomials a(X) of degree <m and 6(A) of degree <n such
that
V.
not
1,
a (A) -/(A) +
b(X)- g(X)
and conversely.
RELATIVELY PRIME POLYNOMIALS. Two

common
divisor is
i.
See Problem
polynomials are called relatively prime
if
4.
their greatest
174
VI. If g(A) is irreducible in
F[X] and /(A)
divides /(A) or g(X) is relatively prime to

VII.
If
VIII.
and
UNIQUE FACTORIZATION.
g:(A)
/(A)
h(X), it divides at least
are relatively prime and
if
each divides
one of
/(A)
and
h(X).
soalsodoes /(A)-g(A).
A(A),
Problems, we prove
In
Every non-zero polynomial f(X) of F[A] can be written as
IX.
(22.5)
/(A)
where
any polynomial of F[X], then either g(X)
is
/(A).
g(A) is irreducible but divides
If /"(A)
[CHAP. 22
is a
9i(A)
?2(A)
^^(A)
constant and the qi(X) are monic irreducible polynomials of F[X]-
SOLVED PROBLEMS
1.
Prove:
/(A) and g(X)^0 are polynomials in F[X], there exist unique polynomials h(X) and r(X)
where r(A) is either the zero polynomial or is of degree less than that of g(X), such that
If
in F[X],
(i)
h(X)-g(X) +
/(A)
r(X)
Let
n-i
+ %1-iA.
+
n
/(A)
o^jA
g(A)
h^X
+ aÂ + Oq
and
Clearly, the theorem is true
/(A)
if
7^ A
/(A) =
i^-iA
or if
g(A)
n<m.
6iA +
n>m;
Suppose that
cpX
/i(A)
6o.
^m
?^
then
c^-iA
Co
is either the zero polynomial or is of degree less than that of /(A).

If /i(A)
r(X) = /i(A).
or is of degree less than that of g(A),
we have proved
the theorem with h(X) =
~X
and
Otherwise, we form
/(A)
Â"-$A) - ;Â^-\(A)
/,(A)
f^X) =0 or is of degree less than that of g(A), we have proved the theorem. Otherwise, we repeat
Since in each step, the degree of the remainder (assumed ^ 0) is reduced, we eventually reach
a remainder r(A) = 4(A) which is either the zero polynomial or is of degree less than that of g(A).
Again,
if
the process.
To prove uniqueness, suppose

/(A)
where the degrees of
r(A)
h(X)-g(X)
and
+ r(X)
/(A)
k(X)-g(X)
'c(A)-g(A)
s(A)
Then
and s($ are less than that of g(A).

h{X)-giX) + r(X)
+ s(X)
and
[k(X) -h(X)]g(X)
Now r(A)-s(A)
is of
to or greater than m.
degree less than
Thus, k{X) - h(X)
h(X) and r(A) are unique.
while, unless
=
0,
r(A)
A:(A)
s(A) =
r(X)
s(X)
h(X) = 0.
[k(X)
- A(A)]g(A)
is of degree equal
so that k{X) = h{X) and r(A) = s(A).
Then both
CHAP.
2.
22]
175
Prove:
If f(X) and g(A) are polynomials in F[X], not both zero, they have a unique greatest
divisor d(\) and there exist polynomials A(A) and k(\) in F such that
da)
(a)
/(A) =
say,
If,
A(A) =
then d(X) =
0,
and k(X) = 6
bin,
(*)
is not greater
r^^(X)
=0
r^(A)
If
7^
degree less than that of g(A).
or is of
and k(X)
6^
or is of
degree less than that of
it
or is of
rgCA) =
degree less than that of
'2(A)
it
b'^^giX)
and we have
(a)
with
+ rîX)
/(A)
we have from
= 0,
If r^(X)
ri(A).
g(A)
-92(A) -/(A)
93(A)
'i(A)
and from
then d(X) =
( i
?i(A)-g(A)
we have
(i")
where
= 0,
obtain (a) by dividing by the leading coefficient of ri(A).
r^(X) 4 0,
If
we have
ri(A)
If r^(X)
g2(A)-ri(A)
HiX)
and from
I,
we have
0,
g(A)
/^(A)
(a) with
(ii)
where
By Theorem
than that of /(A).
?i(A)-g(A) +
/(A.)
where
we have
leading coefficient of g(k) and
is the
Suppose next that the degree of g(A)
h(X) =
ka)-g(\)
h{X)-f(X) +
where i^
g(X)
common
r^(X').
If rg(A) = 0,
?2(A)-'-i(A)
TsCA) + rsCA)
g(A)
we have
?2(A)[/(A)
from
i)
and
(ii)
9i(A)-g(A)]
[1 + 9i(A)-92(A)]g(A)
obtain (a) by dividing by the leading coefficient of r2(A),
Continuing the process under the assumption that each new remainder is different from
0,
we have,
in
general,
(v)
'i(A)
9i+2(A)-'-i+i(A)
ri+2(A)
moreover, the process must conclude with

(V)
's-2(A)
9s(A)-'-s-i(A)
&(A)?^0
r^cA),
and
(^^
's-i(A)
By
divides rs-i(A), and by
(vi), rjCA)
's-3(A)
so that rs(A) divides r^-gCA).
both /(A) and g(A).
Prom(l)
If
ri(A)
'2(A)
Prom
(iii),
/3(A)
'3(A)
Continuing,
d(X)
The proof
9s+i(A)-'-s(A)
also divides rs_2(A).
/(A)
91(A)
g(A)
qa(>^)
r^(X)
/!i(A)
[1 + 92(A) 93(A)]/(A)
h3(X)-f(X)
c-\^(X)
rf(A)
-/(A) + /ci(A)
g(A)
[-9i(A)
we have
we conclude
and substituting
93(A)
(vi),
that rs(A) divides
= e~^'s(A).
A2(A)-/(A)
Substituting for r^(X) and r2(A),
(iv).
r5_i(A)
then
c,
[l + 9i(A)-92(A)]g(A)
we obtain
Prom
Thus, by retracing the steps leading to
-52(A). /(A)
=
t^(K)
9s-i(A)-'s-2(A)
the leading coefficient of rs(A) is
in (ii)
+ i2(A)-g(A)
we have
9i(A)
92(A) 93(A)]g(A)
43(A) g(A)
finally,
rs(A)
Then
(v).
hs(X)'f(X)
c-i/5(A)-/(A) + c-iA:s(A)-g(A)
that d(X) is unique is left as an excercise.
ksiX)-g(.X)
=
h(X)-f(X) + k(X)-g(X)
as required.
176
3.
[CHAP. 22
Find the greatest common divisor d(X) of

=
/(A)
and express d(\)
We
and
3A^ + 1)C + llA + 6
Theorem
in the form of
g(\)
)C
2)i-)^-X+2
III.
find
+ 4A^+6A+4)
(i)
/(A)
(3A+l)g(A) +
(li)
g(A)
(A-2)(A^ +
(iil)
A^+4A2+6A + 4
(A-3)(A^ + 7A+10) + (17A + 34)
7A + 10
(Â + p^)(17A+34)
(A'
6A+4) +
4A^ +
+ 7A+10)
(A^
and
(Iv)
A^
The greatest common

Prom
+34
Substituting for A^ + 7A + 10
17A
for
A +
(iii),
17A
and
J-(17A+34)
divisor is
from
+34
+ 4A^ + 6A + 4 from
+ 4A^ + 6A + 4)
(A
(A^
+ 4>f + 6A + 4)
(A
- 3)[g(A) - (A- 2Xi^ +
(A^
- 5A +
(A^
- 5A + 7)/(A) + (-3A +
i(A^ - 5A +
3)(A^ + 7A + 10)
(ii)
7)(A^ + 4>f + 6A + 4) -
(A
4;? +
6A+4)]
- 3)g(A)
+34
17A
(A
14A^
- 17A - 4)g(A)
Then
A +
4.
Prove:
If
common
the greatest
divisor of /(A) of degree
there exist non-zero polynomials o(A) of degree
< m and
+ 14A^ - 17A
4)-g(A)
and g(X) of degree m >

such that
re>
is not 1,
b(X) of degree < n
b(X)-g(X)
a(X)-f(X) +
(a)
i(-3A
7) -/(A)
and conversely.
Let the greatest common divisor of /(A) and g(A) be d(X) ^
/(A)
where
/i(A) is of
degree
<n
and
rf(A)-/i(A)
and gi(A) is of degree

gi(A)-/(A)
then
d(X)-g^(X)
g(A)
<m. Now
=
gi(A)-rf(A)-/i(A)
g(A)-/i(A)
and
gi(A)-/(A)
Thus, taking a(A)
= gi(A)
and 6(A) = -/i(A),
we have
(a).
Conversely, suppose /(A) and g(A) are relatively prime and

polynomials h(X) and ^(A) such that
A(A)-/(A) +
Then, using
[-/i(A)-g(A)]
k(X)-g(X)
(a) holds.
Then by Theorem IV
there exist
(a),
a(X)
and g(A) divides
a(A).
a(X)-h(X)-f(X)
a(A)-A:(A)-g(A)
-b(X)-h(X)-g(X)
a(A)-i(A)-g(A)
But this is impossible; hence,
if (a)
holds, /(A) and g(A) cannot be relatively prime.
CHAP.
5.
22]
177
Prove: Every non-zero polynomial /(A) in F[\] can be written as
/(A)
where
c-q^(X).qâ)
...
qr(\)
constant and the q^(\) are monic irreducible polynomials in F[A].
is a
,^
Write
(i)
/(A)
where a
is the leading coefficient of /(A).
theorem.
Otherwise, there is a factorization
(")
and
are irreducible, then
/!(A)
a-/i(A)
If /^(A)
/(A)
If g(A)
is irreducible,
then (i) satisfies the conditions of the
n-g(A)-A(A)
satisfies the conditions of the theorem.
(ii)
Otherwise, further factor-
ization leads to a set of monic irreducible factors.
To
prove uniqueness, suppose that
"n
are two factorizations with
by a change
Pi(A) which,
Then
Pi(A).
Eventually,
ity is impossible,
?2(A)
r<s.
P2(A)-p3(A)
= s
.
.
a p^(X) PaCA)
and
9r(A)
p^CA)
... ps(A),
it must divide some one of the
Since Pi(A) is monic and irreducible, ?i(A) =
Pi(A)-p2(A)
may be taken as
p^(\).
and, after a repetition of the argument above, 92(A) = P2(A).

r and Pr-î(A)
Pr+2(A) ... p^cA) = 1. Since the latter equaland uniqueness is established.
9i(A) = Pi(A) for
Since 91(A) divides
in numbering,
92(A) divides
we have
9i(A)
...
Ps(A)
= 1,2
6.
Give an example
7.
Prove Theorem
8.
Prove:
9.
Find a necessary and sufficient condition that the two non-zero polynomials
/(A) and g(A)
each other.
10.
If /(A)
in
which the degree
divides g(A) and h{\).
it
()
/(A)
2A^-A=+2A2-6A-4,
(6)
/(A)
A^-
(c)
/(A)
2A^ + 5A^+ 4A=
(d)
/(A)
3A^ -
(a)
A=-2
(6)
A-
A"
4A='
3A=
+ A^
- 11A+
^+1
^"^^
g(A) + h(\).
divides
- ^ (A-
g(A)
g(A)
- A^ - A +
- 5A + 6,
1)/(A) +
A*-A=-A2+2A-2
A=
g(A)
1,
g(A)
in the form of
- 2A" - 2A -
A% 2A%
A= + 2A
-h
in
F[X] divide
Theorem IV
2A +
i (2A=+ l)g(A)
-l_(A+4)/(A) + l-(A^+5A+5)g(A)
to
('^)
degree of either /(A) or g(A).
III.
For each of the following, express the greatest common divisor
Ans.
11.
of /(A) + g(A) is less than the
lo
^(A+4)/(A)-^ ^(-2A=^-9A2_2A + 9)g(A)
jL(5A+2)/(A)
^^{-l5X^+4.A>?~55X+ib)sa)
Prove Theorem VI.

Hint.
Let d{\) be the greatest common divisor of /(A) and g(A)

h{X) is a constant.
then
g(A) = d{X)-h(X)
and either d(X)
or
178
12.
Prove Theorems VII and
13.
Prove:
14.
If
VIII.
15.
A=-
/(A)
A^-
(b)
/(A)
(A-l)(A+lf(A+2).
g.c.d.
A-l;
(b)
g.c.d.
(A+i)(A + 2);
l.c.m,
(A+l)(A+2f(A-3)
g(A)
(a)
Given 4
divides a( A).
Find the greatest common divisor and the least common multiple of
(a)
g(A)
it
g(A) is a monic polynomial which is a multiple of both /(A) and g(A),
The least common multiple of /(A.) and

and is of minimum degree.
Ans.
and divides g(X)-a(X),
/(A) is relatively prime to g(A)
1,
[CHAP. 22
(A^- l)(A^ + A+
l.c.m.
1)
(A- l)(A+ if (A + 2f (A-
3)
show
12 ;2 ij
(o)
(f)(k)
(b)
m(A)
= 0.
A^
SA^
when m(A)
16.
What property of a
17.
The scalar
and only
18.
if
c is
A-
Take
Show
- 4A -
field is not satisfied
(A-c)^g(A).
k>
(a)
of /(A)
if
that if D(A) is the greatest
Prove:
Let
/I.
sA!^
- 9A -
51
by a polynomial domain?
Show
/(c) = 0.
Prove: The scalar e is a root of /(A)
that c is a root of multiplicity
and only
common
if c
if
An n-square matrix A
is normal it
k-l
of /'(A),
(b)
common
divisor d(X). Let
/(A) = s(X)-D(X).
'
Show
that c is a
is a root of both /(A) and /'(A).
be any field containing F.
divisor of /(A) and g(A) considered in K[A], then
d(X) = h(\)- f(\) + k(X)- g(\).
aÂ
in
/(A) and g(A), not both 0, in F[\] with greatest
Hint:
20.
A^
A -
(f)(A)
called a root of the polynomial /(A) if
root of multiplicity
19.
and
c is a factor of /(A).
/(A) =
Suppose
- 9A -
g(A) = t(X)-D(X).
can be expressed as a polynomial
+ a^^Â
...
+ aiA + oqI
D (A)
and D(X)
= d(X).
c{X)-d(X).
chapter 23
Lambda
Matrices
DEFINITIONS. Let F[X] be a polynomial domain consisting

in F.
mxn
non-zero
(23.1)
of all polynomials in A with coefficients
matrix over F[Aj
^(A)
[ij (A)]
oii(A)
ai2(A)
...
ai(A)
asiCA)
a22(A.)
...
CsnCA)
Omi(A)
a^îX.)
...
aj^n(>^)
is called a A-matrix.
Let p be the maximum degree in A of the polynomials a^jiX) of

(23.1).
written as a matrix polynomial of degree p in A,
(23.2)
where the ^^ are

Example
Ap\^
^(A)
mxn
sp-l
A.^X^-^
...
A,\
+ Ao
matrices over F.
A^+A
1.
Then A(X) can be
A* + 2A= + 3A^ +
5]
'^(A)
.
A=
- 4
A^
- 3A=
J
1
A^ +
A +
-J
-4
L
is a A-matrix or matrix polynomial of degree four.
A(X) is ra-square, it is called singular or non-singular

according as \A(X)\ is oris not
Further. A(X) is called proper or improper according
as A^ is non-singular or singular
matrix polynomial of Example 1 is non-singular and
improper.
If
zero.
The
OPERATIONS WITH A-MATRICES.
Consider
two
the
n-square
A-matrices
over F(X)
(23.3)
A^X^
^(A)
.P-'
S^_,
or
matrix polynomials
+ Ai^X + Aq
p -1
and
(23.4)
(j
B(X)
a"^
'
...
SiA + So
The matrices (23.3) and (23.4) are said to be equal, ^(A) = 5(A), provided
= Bp=a andA-VI,'
^ ^
= 0,1,2
p).
The sum A(X)

two A-matrices.
+ S(A) is a A-matrix C(A) obtained by adding corresponding

elements of the
The product /1(A) B(A) is a A-matrix or matrix polynomial of degree at

most p+q. If either
^(A) or B(A) is non-singular, the degree of ^(A) 5(A) and also of
5(A) /1(A) is exactly p + 9.
The equality (23.3) is not disturbed when A is replaced throughout
by any scalar k of F
For example, putting A = A in (23.3) yields
A(k)
Apk^
+ Ap_.^k^~^ +
179
...
A^k ^ Ao
LAMBDA MATRICES
180
However, when \
[CHAP. 23
two results can be obtained due
is replaced by an re-square matrix C,
We
fact that, in general, two ra-square matrices do not commute.
Ag(C)
(23.5)
A^C^
. ^^
CÂ^
.
C^
+ .
A^_,C^
p-i
to the
define
+ A,_C + Ao
...
and
4 (J)
(23.6)
A^
^p
+ ^^î +
ô
called respectively the right and left functional values of A(\).
Example
Let
2.
(X)
A'
Jx^
X
XmI
\X-2
Then
X''
P
[O
+ 2j
Ajf(C)
[o
ij [3
o1.
To
i1.
To
ij
[l
oj
[-2
'
4J
[1
oJ
'
4)
[3
il
and
2j
[-2
_3
4_
[lo
15
14
26
12
[17
27
2J
j,nd
Al(C)
See Problem
DIVISION.
Problem
In
I.
If
1.
we prove
2,
A(\) and B{X) are matrix polynomials (23.3) and (23.4) and
polynomials
then there exist unique matrix
Q2(X), R2(X), where /?i(A) and
Qi_(X), RxiX);
degree less than that of 6(A), such that
/?2(A) are either zero or of
(23.7)
Bq is non-singular,
if
.4(A)
(?a(A)-S(A) + Ri(A)
A(X)
S(A)-(?2(A) + R2(X)
and
(23.8)
S(A) is called a right divisor of i(A);
/?i(A)=0,
If
visor of
/1(A).
Example
3.
jV
A
If
(A)
A"^
+ A -
2A^
A% A% A +
-A
Ta^-i
2A
|_
and
2A''
2A
JL
(A)
A'^+aJ
'1
'
aâJ
ii
A-iirA^+i
6(A) is called a left di-
X%
21
'4(A)
/?2(A) = 0,
if
r 2A
2A+3
j_-5A
-2A_
Q-l(X)-B(X)
then
Ri(X)
and
A(X)
1_
A^ + aJLA-I
B{k)-Q2{X)
ij
Beie, B (A) is a left divisor of A (A)
See Problem
matrix polynomial of the form
is called scalar.
bqX"-!^ + 5,_,A'?"'-/ +
6(A)
(23.9)
trix
3.
scalar matrix polynomial
B(A) =
...
+ b,X
fc(A)-/
+ bol^
polynomial.
If in
(23.7) and (23.8),
A(X)
(23.10)
6(A) = b(X)
4.
Let
then
Q^(X)- B(X) + Rt(X)

rA^ +
Example
I.
2A
B(X)- Q^(X) +
A + ll
and
(A)
LA^^-I
2A+1J
commutes with every
B(A) = (A + 2)/2.
Then
R^X)
i(A)-/
ra-square ma-
CHAP.
LAMBDA MATRICES
23]
rA + 2
If
/?i(A)
If
-ll
To
ll
^(A) = 6(A)- /
in (23.10), then
181
and we have
(?i(A)
matrix polynomial i(A) = [oy (A)] of degree n is divisible by a scalar matrix polynomial S(A) = 6(A) / if and only if every oy (A) is divisible by 6(A).
II.
THE REMAINDER THEOREM.
Let A(X) be the A-matrix of (23.3) and let B =[6^,-] be an re-square

Since Xl - B is non-singular, we may write
matrix over F.
(23.11)
A(\)
(2i(A)-(A/-S) +
A(\)
(A/-fi)-(?2(A) + R2
/?!
and
(23.12)
where R^ and Kg are free of
m.
If
A.
It
can be shown
the matrix polynomial A(X) of (23.3) is divided by \I
re, until
remainders ^1 and R2, free of
-B, where B =[b^-]
is
ra-
''
A, are obtained,
then
Aj^(B)
A^B^
Ap_^B^~'
...
a,
Aj^(B)
BÂp
B^~'a^_^
...
+ BA, +
A,B
+ Ao
and
Examples. Let
^(A)
."^M
LA-2
A='
[a +
\I - B
and
+ 2j
3
IfA-l
[4
A+
a'-JH'
4JL-3
r~^
"^
|_-3
A-4J
-2I
A-4J
"
^
The
flO
isl
[14
26j
[17
27]
'^'
and
'^'^
Prom Example
When ^(A)
2,
a!4J
R^ = Ag(B) and
j?2
(^'-B^Qî^^-^^
= A^(B) in accordance with
Theorem
III.
is a scalar matrix polynomial
'4(A)
/(A)-/
apIX^ + ap.jx''^ +
...
+ ai/A + Oq/
the remainders in (23.11) and (23.12) are identical so that
^1
^2
apB^
ap_^B^~^
...
+ oifi + aol
and we have
IV. If a scalar matrix polynomial /(A)-/ is divided by
free of A, is obtained, then
As
a consequence,
V.
until a remainder/?.
we have
scalar matrix polynomial /(A)-/ is divisible by
CAYLEY-HAMILTON THEOREM.
trix
A/-B
= f(B).
KI^- B
if
and only
if
f(B) =
0.
Consider the re-square matrix A = [o,-,-] having characteristic maijj

Xl - A and characteristic equation 0(A) = \Xl - A\ = 0. By (6.2)
(A/-i)-adj(A/-i)
(f>(X)-I
LAMBDA MATRICES
182
Then ^(A)
6.
(^(A) =
by XI - A and, by Theorem V,
/ is divisible
Vl. Every square matrix A =
Example
[CHAP. 23
[oj,-]
of
Thus,
0.
satisfies its characteristic equation
is
V.i
A -
7>i
32
62
31
31
63
31
31
62
32
+ llX - 5 =
<fi(X)
0.
0.
Now
and
32
62
3l"
31
63
31
31
62
32_
- 7
"2
12
13
12
7.
11
_1
o"
1_
See Problem
4.
SOLVED PROBLEMS
1.
For the matrix
A(X)
[,
A+
^'<'^'
compute
ij
-4
and 4r(C)
^p(C)
when
C=
i o]C
"3
'2
'
o][o
'
2J
c 3
and
2.
there
^(A) and B(A) are the A-matrices (23.3) and (23.4) and if Bq is non-singular, then
are
either
and
RîX)
exist unique polynomial matrices Q-,(X), i(A); QsO^). R2M. where Ri(A)
zero or of degree less than that of 6(A), such that
Prove:
If
(i)
^(A)
QAX)-B(X) +
^(A)
6(A) (?2(A) + RîX)
Ri(X)
and
(ii)
If
p <
q,
then
(I)
and
holds with (3i(A) =
where C(A)
If
is either zero or of
degree at most p -
C(A) is zero or of degree less than

(3i(A)
q,
Suppose that p iq; then
fii(A) = A(X).
4(A) - ApB-^B(X)X^''^
C(A)
1.
we have
ApB'q'-^''^
(i)
with
and
RîX) =
C(A)
CHAP.
LAMBDA MATRICES
23]
C^X +
C(X) =
If
...
where
A(\)
If
s>
q.
form
- ApB-^''B(\)\P-1 -
D(X) is either zero or of degree less than
Qi(\)
we continue
otherwise,
of decreasing degrees,
and we have
To
the process.
we
q,
C^S-1B(A)X^-'?
we have
and
Since this results in a sequence of matrix polynomials C(A), D(A),
(II),
begin with
- B(X)B-^''ApXP~1
This derivation together with the proof of uniqueness will be
rA^ +
and
A + A'
A"
A^ +
find matrices Q-l(X), Ri(X);
A(X)
as an exercise. See Problem
left
1,
Chapter 22.
A'-A-l]
2A=-l
^(A)
|_
L
(a)
A=
A"
(?2(A),
S(A)
L-A^ + 2
A^
- A
R2(X) such that
= Qi(X)-B(X) + Ri(X).
= S(A)-(?2(A) + RîX)
A(X)
(b)
as in Problem
2.
We have
and
Here,
So
-ii
L-1
(a)
and
Tl
-,
S2
ii
[i
ij
2J
We compute
^(A)
C(A)
-^4S;^5(A)A^
- C3S2^S(A)A
1'
-2
A +
1]
-10
-6
and
D(A) - D^B^^BCX)
-6
5"]
(?i(A)
(A^X^ + Cs A
+O2 }B^-^
A +
-1'
f-l
11
C(A)
-1I
11
L
D{X)
-6A-13
-2A-9
A= +
q
+
fA^
P-i
^
A +
5A + 3
3A + 5
fli(A)
--
-l"!
fo
J
3
'1
Then
-9
[_
3J
3]
["-13
^ +
-2
10
A2 +
=
3
(b)
...
(i).
obtain
Given
Ri(X) = 0(A)
ultimately reach a matrix polynomial which is either zero or of degree less than
g
^(A)
3.
D(X)
with
(i)
Aj^B-^XP'^l .C^B'^^X'-^
ApB-'XP'^
+ C^B^^X^-'f
183
4A
2A+
+ 4
4
5]
A^ +
A +
sA +
3A +
^1
e]
We compute
^(A) - B(A)S2^'44A^
(A) - 5(A) B2
E(X)
^sA
and
F(X) - B(X)B2
F^
ftsCA)
LAMBDA MATRICES
184
Then
B2^(A^>? + 3 A +
QziX)
F2)
[A^+4A
4.
Given
and A
also, since
use the fact that A satisfies
2X +
+ 4
6A +
A^ +
[CHAP. 23
2I
3A + 5J
equation to compute A"
its characteristic
compute A
is non-singular, to
A-1
-1
-2
-3
A-1
-1
-2
-3
A-1
A/-^|
and A'
A^-3A^-7A-11
Then
"8
5"
_13
3A +1 A + 11/
^*
3/ +7/
From
11/
UA
-1 A -3-4
+A
"l
2"
8_
42
31
29"
45
39
31
_53
45
-J-7/
1_
5'
13
42_
0"
11
'8
+ 7
"1
_2
1_
42
31
29
45
39
31
53
45
42
193
160
224
177
160
272
224
193
144
we have
'1
"1
0'
34 +/II
'1
2"
1_
_13
8_
11
11
_0
1_
-1
-2
-1
-3
-1
-2
11
i{_7^-i -3I
+ A\
-1"
-3
-1
-2
121
-8
-24
40
-1
"1
- 33
0"
1
1
29'
-24
121
-27
5.
Let Ai, As,
,^n
of degree p in
x:.
40
-8
be the characteristic roots of an re-square matrix A and let h(x) be a polynomial
Prove that
\h(A)\
= h(\)- h(X2)
h(X^).
We have
(i)
\XI-A\
(A-Ai)(A-Ao)...(A-A^)
Let
(ii)
h(x)
c(,si- x)(s2-x) ...(Sp - x)
Then
h(A)
c(sil-A)(S2l-A)...(spI-A)
4
CHAP.
LAMBDA MATRICES
23]
185
and
|A(^)|
c'P\siI-A\-\s2l-A\ ...\sp!-A\
!c(si-Ai)(si-A2)
...
(si-A)i
Jc(S2-Ai)(S2-A.2)
using
(S2-X)!
...
{c(Sp -\t)(,Sp -As)
...
(Sp
jc(Sl-Ai)(S2-Xl) ... (S^- Xl)i

{c(si-A2)(S2-A2) ... (Sp-\2)\
...
ic(si-A)(s2-A)
...
(s^-A)!
...
-X)!
h(\i_)h(\2) ...h(\^)
(ii).
6.
Given
^1
Pr^
[a^+i
A(\) =
_r2\2+2A
(a)
-4(A)
A(\) - B(A)
^(A)-S(A) =
2A -
r2A
-A^l
[a^-a
-ij
2A=+A=+A
L ^ +2A^ (rf)
r2A'' +
B(A)M(A)
1_
7.
Given
Ag(C)
^;?(C)
^^^^1
a
compute:
A^+2A1
+ A + 2
rA* +
(c)
[a +
+ B(A)
[_A^
(6)
B(A)=r^^
sna
A-iJ
2A='
ij
A* + 3A= + 3A=1
-,
A
o
A-'
3A'+A^+A
2A^
+ 3A^ + 3A
A'
2A^J
-a]
2\^
/1(A)
compute:
-2
B;p(C)
-1
-3
[-:
<?i?(C)
3"]
-3I
Bn(C)-4n(C)
i?^
4;f(C)-B^(C)
[l7
î(C)-B^(C)
B^(C)-A^(C)
E
3-
î(C)
[3
-7J
[::;]
:]
--M
where P(A) = 4(A)- B(A) and Q(X) = B(A)M(A).
8.
If /4(A)
and B(A) are proper n-square A-matrices of respective degrees p and q, and
show that the degree of the triple product in any order is at least p+q.
A-matrix,
if
C(A) is any non-zero
LAMBDA MATRICES
186
9.
For each pair of matrices A(X) and B(X), find matrices
|> + 2X
(a)
.4(X)
AiX)
-A.2
'
,,
La%i
f
(b)
"
,,
a^-aJ
+
U.2A-1
.4(A)
A'*
2A - A +
B(X)
(d)
A(X)
(a)
2A
A^
A-2
2A
A+
A"-l
A=-A^+l
A* + A^ +
A^ + A
A+
A^+3
-A+7
A^-1
3A + 5
-3A+2
A-6
A"
A+
A -
81A +
15A -
-7A
3A^+5A
+ 31
-3A -
46
-15A - 30
Verify in Problem 9(b) that
17A -
-21A + 4
-2A+3
loA +
- 2A -
2A^-4A
-A +
llA +
4A - 4
2A +
16A - 16
Ri(A) = .4^(0) and
4A -7
1
+ 3
2A - 2A -
-12A -
+ 8
-2A + 6A - 6
A"
-26A - 30
J,
23
12A -
-A^-A-4
A - 14
7lA +
-85A -
16
ll
figCA) = -4^(0)
A^ + A -
o]
[l
ij
f-
-fjA-3
-A^ +
2A'^
-12A -
A^-2A
fo
ll
-16A + 14
3A'^-7A
+ A
A +
|^_ 1
5A-7
-9A -
-A +
|^_^
A=^
46
4A -
fti(A)
- A -
A''
/?2(A) =
-2A -
fl2(A)
,,(A)=
-3A^-5A-16
''^'''-
f-A-l
-A+1
A^ +
a-i1
-A + 2j'
+2
_J,
3A^ + 6A + 31
<?2(A)
A^
A-2
^=^'^^
.,(A)^0;
10.
2A* + A -
-A-l]
,2
B(X)
ij'
A-l
A^-2A-1
A+1
r 2a
2A-3
Ri(X)
A -
0,(A)=
Qi(X)
A^-A
'^^''-'-
<?2(A)
8A -
A + A^ +
a1
6A+
4A^ +
A""
\^_^^^
(d)
+ 2
+ 4
3AnA"-l
(?i(A)
5A^+2A
A^ + 2A''
f-A
(c)
7A-2
+ A^ +
x4
[o
J'
3A^+ 2A
-ll
FA-I
''''-
3A-1
A^+1
(b)
-aJ
[\+i
Ans.
''''-'I
A'^
3A -
3A''
a]
A^-2A+1
(c)
Fa
-A^ + 2A + 2l
A.
Q^X), Rd^) satisfying (23.7) and (23.8).
(?i(A), /?i(A);
BiX)
[CHAP. 23
where B(A)=A/-C.
..(A).
-5A + 2
A-5
18A-7
CHAP.
11.
LAMBDA MATRICES
23]
3X+
A-2
^2
A^-3A
[\^
Given
compute
(b)
find (3(A) and fi(A) of degree at
most one such that
A^'+SA^+sA
L^- 5A=+llA Given
A 1
A
A+lJ
A +
B(A)- C(A)
Fa^'+SA^-VA-
12.
[A - 3
Â
=1fx
C(X)
''*
+ 2j
(a)
.4(A) =
187
4(A) = Q(X)- B(\) + R(\).
compute as
in
r-9 A +
si
A -
A^
^^1
i7j
|_i2
[I
Prove:
14.
A and B
If
Show
Prove:
If
and B
S^
= diag(Si,
are similar for any positive integer k.
B^) and g(A)
g(B)
15.
!;]
are similar matrices and g(A) is any scalar polynomial, then g(A) and g(fi) are similar.
first that
fi
10J
Problem 4
'1
Hint.
9A +
|_13A-6
ij
10
13.
-A - gl
S(A)
A-6
A='-A^-3A+2j
10
A +
[A + 4
l]
is
any scalar polynomial, then
g(B))
diag(g(Si), g(B2)
Prove Theorem ni.

Verify: A/ - B divides A(\)-Ajf(B).
Hint.
16.
The matrix C is called a root of the scalar matrix polynomial B(A) of (23.9) if B(C)
C is a root of B(A) if and only if the characteristic matrix of C divides B(A).
17.
Prove: If Ai, Aj
A are the characteristic roots of A and
characteristic roots of f(A) are /(Ai), /(A2)
/(A)Hint.
Write
\- f(x)
Nowuse
18.
c(xi- x)(x2-
|*j/-4|
- x)
x) ...(x^
so that
f(A) =
A'^
-2A+3,
given
19.
Obtain the theorem of Problem 5 as a corollary of Problem 17.
20.
Prove:
21.
Let A(t)
If
= [a^,-()]
where the
A of Problem
a^j{t) are real
17, then
polynomials
Prove: The matrix
f(A) is any scalar polynomial in A, then the
\\I
and
(xj-Ai)(::i-A2)...(*^-A)
Find the characteristic roots of
if
= 0.
- f(A)\ = c'^\xj - A\ \x2l - A\ ...\xsl -A\.

c(xi -Xj) (x2-\j) ... (x^ -\j) = \- f(\j).
-1
is an invariant vector of f(A)
in the real variable
Take
t.
A(t)
and differentiate the last member as
if it
nition
22.
Derive formulas
(a)
-r-\A(t)
were a polynomial
Sfnomial with
wit! constant coefficients to suggest the defi,
for:
+ B{t)\;
(b)
-T-\cA(t)\.
dt
where
c is
a constant or c =[ciA;
(c)
^{A(t)-
B(t)\;
dt
(d)
Â
dt
'it).
n
Hint.
use
For (c). write

A{t)-A-\t) = /.
A(t)- B(t) = C(t) = [ci;.(t)]

'
and differentiate
cv,-()
a,-i,(0 6i,,-().
k=i ^^
"^^
For
(d),
chapter 24
Smith Normal Form
BY AN ELEMENTARY TRANSFORMATION
(7)
The interchange
of the ith and /th tow, denoted by
column, denoted by K^j

(2)
on a A.-matrix A(X) over F[\] is meant:
The multiplication
ff|--
the interchange of the ith and /th
of the ith row by a non-zero constant k, denoted by H^(k);
the multiplication of the ith column by a non-zero constant k, denoted by Kîk).

(3)
The addition
row of the product of
to the ith
/(X),
any polynomial of F[\], and the /th row,
denoted by ^^(/(X));
the addition to the ith column of the product of /(X) and the ;th column, denoted by K^j (/(X)).
These
are the elementary transformations of Chapter 5 except that in (3) the word scalar has
been replaced by polynomial. An elementary transformation and the elementary matrix obtained
by performing the elementary transformation on / will again be denoted by the same symbol.
Also, a row transformation on ^(X) is effected by multiplying it on the left by the appropriate
H and a column transformation is effected by multiplying ^(X) on the right by the appropriate K.
Paralleling Chapter
I.
5,
we have
Every elementary matrix in F[\] has an inverse which in turn
is an elementary
ma-
trix in F[\].
II.
III.
Two
If
\A(\)\ = k
The rank
0,
with k in F, A(\) is a product of elementary matrices.
of a X-matrix is invariant under elementary transformations.
re-square X-matrices i(X) and B(X) with elements in F[X] are called equivalent pro-
vided there exist P(\) = H^
ff 2
'
^1 and Q(\) = K^- K^
B(X)
(24.1)
...
K^ such
that
P(\)-A(X)-Qa)
Thus,
IV.
Equivalent
THE CANONICAL SET.
In
mxn
X-matrices have the same rank.
Problems
and
2,
we prove
Let A(\) and B(X) be equivalent matrices of rank r; then the greatest common divisor of all s-square minors of A(\), s r, is also the greatest common divisor of all ssquare minors of B(\).
V.
In
Problem
VI.
3,
we prove
Every X-matrix A(X) of rank
can be reduced by elementary transformations
Smith normal form
188
to the
CHAP.
SMITH NORMAL FORM
24]
189
A (A)
/2(A)
N(\)
(24.2)
...
where each
monic and
/^(A) is
When a A-matrix A(\)
of rank
f^(\) divides f^^^ (A),

r
has been reduced
(i
0_
1, 2, ...
1).
to (24.2), the greatest
common
divisor of
s-square minors of A(\), s ^ r, is the greatest common divisor of all s-square minors of A^(A)
by Theorem V. Since in N(\) each f^(X) divides /^^j(A), the greatest common divisor of all ssquare minors of N(\) and thus of A(\) is
all
g,(A)
(24.3)
Suppose A(\) has been reduced
= l,2,
(s
/i(\)-f2(A)-...-4(A),
.r)
to
/V(A)
diag(/-i(A), /2(A)
/Vi(A)
diag(Ai(A). AsCA)
/^(A), 0,
and to
By
O)
hîX),
(24.3),
gs(A)
Now
we
go(A) =
define
(24.4)
Ai(A)-A2(A)-...-As(A)
g2(A) =/i(A)-/2(A) =Ai(A)-A2(A)
gi(A) =/i(A) =Ai(A),
eral, if
/i(A)-/2(A)-...-/s(A)
1,
so that
U\)
h^W,
in gen-
then
h^(\).
4(^)
s:s(^)/s-i(^)
(s
1, 2
r)
and we have
VII.
The matrix N{X)
of (24.2) is uniquely determined by the given matrix /1(A).
Thus, the Smith normal matrices are a canonical set

A +
Example
1.
Consider
A'+2A^+A
.4(A)
A^ +
It
A +
3A +
2A +
A^ +
gi(A) =
and g3(A)
/i(A)
1,
= i\A(X)\
gi(A)
the greatest
= A^
+X^
1,
equivalence over F[A].

A.
+ 3
2A%3A'^+A
3A^ +
6A
common divisor
is readily found that the greatest
A(X) is
A^+A'^ + A
for
common
+ 3
of the one-row minors (elements) of
divisor of the two-row minors of A(X) is g2(A)
=X,
Then, by (24.4),
/2(A)
g2(A)/gi(A)
/3(A)
A,
g3(A)/g2(A)
A^+A
and the Smith normal form of A(X) is
10
A
A'(A)
A^+A
For another reduction, see Problem
INVARIANT FACTORS. The
polynomials
4.
f^(X),f-2{X)
/V(A)
in the diagonal of the Smith normal
form of A(X) are called invariant factors of ^(A). If 4(A) =

fk(X) = 1 and each is called a trivial invariant factor.
As
a consequence of Theorem VII,
we have
1,
k r,
then
/i(A) = /2(A) =
...
SMITH NORMAL FORM
190
VIII.
Two
[CHAP. 24
re-square X-matrices over F[X] are equivalent over F[X\ if and only if they
have the same invariant factors.
ELEMENTARY
DIVISORS. Let A(X) be an re-square X-matrix over F[X] and
let its invariant factors
be expressed as
(24.5)
{p,(X)\'î^\p^(\)fi^...\PsMi^^^.(i = i.2
f^(\)
r)
where pi(A), p2(X), ., PgCX) are distinct monic, irreducible polynomials of F[X\.
q^j may be zero and the corresponding factor may be suppressed; however, since
r-l;/ = l,2, .s).
/i+i(A), qi+t,j^gij. (f = 1.2
The
factors
!p-(X)S'^*i
which appear
Some
fj
(A)
of the
divides
elementary divisors over F[X]
in (24.5) are called
of A(X).
Example
2.
Suppose a 10-square A-matrlx A(X) over the rational
has the Smith normal form
field
10
10
(A.-1)(A^+1)
'
(A-1)(A^ + 1)Â
(A-l)^(X" + l)Â^(A"-3)
The rank
is 5.
The invariant
=
/i(A)
The elementary divisors
(X-l)^
/2(A)
1,
(A-l)''(A.''+lf A^(A''-3)
/5(A)
(A-1)(A^+1),
/3(A)
1,
(\-l)(X^+lfX.
U(k)
factors are
are
A-1,
A-1.
(A^'+l)'',
(A^+l)^,
Note that the elementary divisors are not necessarily

mentary divisor appears as often as
Example
3.
(a)
Over the
it
A^
(A'^+l),
distinct;
A,
A""
- 3
in the listing
each ele-
appears in the invariant factors.
Example
real field the invariant factors of A(X) of
2 are
unchanged but the
ele-
mentary divisors are

(A-l)"^,
since
(6)
A^-3
A-1,
A-1,
(A^lf,
(A^'
l)''
(A^'
l),
X',
A,
X+x/l
X-\fz,
can be factored.
Over the complex
field the invariant factors remain
unchanged but the elementary divisors
are
A-1,
A-1,
(X-if,
X-i,
(A-1)^,
The
{X + if,
(X + if,
A^,
X-\/l,
A,
X + i.
(X-if.
X+\/l
Invariant factors of a A-matrix determine its rank and its elementary divisors;
conversely, the rank and elementary divisors determine the invariant factors.
Example
4.
Let the elementary divisors of the 6-square A-matrix A(X) of rank

A=,
A"",
A,
(A-l)^
(A-^)^
A-l,
be
(A+l)^
A+
Find the invariant factors and write the Smith canonical form.
To form
/5(A), form the
lowest common multiple of the elementary divisors,

/5(A)
a" (A-1)'' (A
if
i.e.,
CHAP.
SMITH NORMAL FORM
24]
To form
fî\),
form the lowest
191
remove the elementary divisors used

of those remaining, i.e.,
in /sCX) from the original list
and
common multiple
UiX.)
Repeating, /sCX) =
Now
A(X-l).
The Smith canonical form
A^(X-lf(A+l)
the elementary divisors are exhausted; then /sCX) = /i(A) =
1.
is
$XN($
1)
A^(A-lf (A + 1)
A=(A--lf(X+lf
Since the invariant factors of a A-matrix are invariant under elementary transformations, so
Thus,
also are the elementary divisors.
IX. Two re-square X-matrices over F[\] are equivalent over F[X]
have the same rank and the same elementary divisors.
if
and only
if they
SOLVED PROBLEMS
1.
If P(\) is a product of elementary matrices, then the greatest common divisor of all ssquare minors of P(\)- A(X) is also the greatest common divisor of all s-square minors of A(\).
Prove:
It
is
necessary only
to
consider P(A)-.4(A) where P(A) is each of the three types of elementary matrices H.
Let R(\) be an s-square minor of .4(A) and let S(A) be the s-square minor of P(X)- A(\) having the same
position as R(X). Consider P(A) = H^j
its effect on ^(A) is either (i) to leave R(X) unchanged, (u) to interchange two rows of R(X). or (iii) to interchange a row of R(X) with a row not in R(X). In the case of (i),
S(X) = R(X); in the case of (ii), S(A) = -R(X); in the case of (Iii), S(X) is except possibly for sign another
s-square minor of ^(A).
;
Consider P(X)
= H^ (k);
Finally, consider
then either S(A) = R(X)
P(A) = H^j (/(A)).
Its effect
or S(X) = kR(X).
on ^(A) is either
(i) to leave R{X) unchanged,

(U) to increase one of the rows of R(X) by /(A) times another row of R(X). or (iii) to increase one of the rows of R(X)
by /(A) times a row not of R(X). In the case of (i) and (li), S(X) = R(Xy, in the case of (iii),
S(X)
where 7(A) is an s-square minor of
R(X) f(X)-T(X)
.4(A).
Thus, any s-square minor of P(X)' A(X) is a linear combination of s-square minors of /4(A). If g(A) is
the greatest common divisor of all s-square minors of .4(A) and gi(A) is the greatest common divisor of all
s-square minors of P(A)--4(A), then g(A) divides gi(A). Let B(X) = P(XyA(X).
Now
.4(A) =
P"^(A)-
fi(A)
and P~'^(X) is a product of elementary matrices.
Thus, gi(A) divides g(A) and
gi(A) = g(A).
2.
Prove:
If P(X) and Q(X) are products of elementary matrices, then the greatest common divisor of all
s-square minors of P(A) -^(A) -QiX) is also the greatest common divisor of all s-square minors otA(X).
Let B(A) = P(A)M(A) and C(A) = B(X) Q(X). Since C'(A) = Q'(X)- b'(X) and Q'(X) is a product of elementary matrices, the greatest common divisor of all s-square minors of C'(A) is the greatest common divisor
of all s-square minors of s'(A). But the greatest common divisor of all s-square minors of C'(X) is the great
common divisor of all s-square minors of C(A) and the same is true
common divisor of all s-square minors of C(A) = P(X)- A(X)- Q(.X) is
est
square minors of ^(A).
for B'(A)
and B(A). Thus, the greatest
the greatest
common
divisor of all s-
SMITH NORMAL FORM
192
3.
Every A-matrix A(\) =
Prove:
of rank
[a^j(\)]
[CHAP. 24
can be reduced by elementary transformations to
the Smith normal form

fi(X)
/2(A)
N(X)
where each
fi(\) is
L(A)
monic and
f^(X) divides /^^j^(A),
(i
1, 2,
...,r-
1).
The theorem is true for A(\) = 0. Suppose A(X) ^ 0; then there is an element a::(K)
of minimum
J
By means of a transformation of type 2, this element may be made monic and, by the proper interchanges of rows and of columns, can be brought into the (l,l)-position in thematrix to become the new aii(A).
7^
degree.
(a)
Suppose a 11 (A) divides every other element of ^(A).

duced to
Then by transformations
of type
3,
4(A) can be re-
/i(A)
(i)
B(A)
where /i(A) = a 11(A).

(6)
Suppose that an(A) does not divide every element of A(X).

which is not divisible by
an (A).
By Theorem
ay (A)
where
ri^-(A) is
I,
Let
ai--(A)
of degree less than that of aii(A).
Prom the
/th
(l,l)-position as the
new oii(A).
If
row
column subtract the product of q{X) and
now
ri,-(A).
By a transformation
replace this element by one which is monic and, by an interchange of columns bring
2,
first
g(A)an(A) + ry(A)
the first column so that the element in the first row and /th column is
type
be an element in the
Chapter 23, we can write
now an(A) divides every element
of A{X),
we proceed
it
of
into the
to obtain
(i).
Otherwise, after a finite number of repetitions of the above procedure, we obtain a matrix in which every
element
If
in the first
row and the
column is divisible by the element occupying the (l,l)-position.
first
this element divides every element of A(X),
not divisible by
On (A).
Now
row
we proceed
to obtain
(i).
Otherwise, suppose
a,-,-(A)
is
"J
Let
aii(X) - qix(X)- a^îX) and aij(A) = gijXA)- aii(A). From the ith row subtract the product of gii(A) and the first row. This replaces aii(A) by
and aij (A) by aij{X) - qii(X) a^jiX).
add the
jth
to the first.
aij (A)
This leaves oii(A) unchanged but replaces aij(A) by
- qii(X)- aij(X) + aj(\)
Since this is not divisible by aii(A), we divide
îj(X) +
it
gij(A)U- ?ii(A)!aii(A)
by aii(A) and as before obtain a new replacement (the
remainder) for aii(A). This procedure is continued so long as the monic polynomial last selected as aii(A)
does not divide every element of the matrix. After a finite number of steps we must obtain an a 11(A) which
does divide every element and then reach

Next,
we
treat B(A) in the
(i).
same manner and reach

i(A)
/2(A)
C(A)
Ultimately,
we have the Smith normal
form.
Since /i(A) is a divisor of every element of B(A) and /2(A) is the greatest common divisor of the elements
of S(A), /i(A) divides /2(A).
Similarly,
it
is found that
each /^(A) divides fi^îX).
CHAP.
4.
SMITH NORMAL FORM
24]
193
Reduce
X +
^(A)
A='
A^
A+
+ 2A=
+A
3A +
A +
X" + 2
A + 3
+ A
a""
A +
+A
+ 3A^
2A''
+ 6a + 3
a""
Smith normal form.
to its
necessary to follow the procedure of Problem 3 here. The element /i(A) of the Smith normal
common divisor of the elements of A(\); clearly this is 1. We proceed at once to obtain
such an element in the (l,l)-position and then obtain (1) of Problem 3. After subtracting the second column
is not
It
form is the greatest
from the
we obtain
first,
A+
A+
a"
A^
'4(A)
A^ +
A+
"
A% A
2A +
3A%A
2A +
3A^ +
6A
A +
A +
A%A
+ 3
2A^ + 2A.
A + A
Now
the greatest
common
S(A)J
Lo
2A + 2A_
Then
divisor of the elements of S(A) is A.
A+A
2A^ + 2A
2A^ + 2A
A" + A
and this is the required form.
5.
Reduce
A =
-4(A)
A""
+ A
A + 2
A^
A''
- 3A +
A=- 2A
2A
A= + A -
Smith normal form.
to its
We
find
A-
^(A)
A=
A-
A^-3A
+ 2
"
"l
-A -
^\^
A +
A-
A=+2A
A^+A-3_
2"
A +
A+1
1
-V-
-1
-A -1
A+1
A+
using the elementary transformations î2(-l);

l),
A +
"l
A+1
^23(1); Hs2(A +
A +
-A'
^2i(-A), ^3i(-A +
2);
-A
/f2i(-A +
A(A+i)
l),
.Ksi(-A -2);
HqsC-I);
Hî-iy. K32(-A-l), Xs(-l)-
6.
Show
7.
Prove:
that
H.jK^j
H^{k)Kâ/k)
An n-square A-matrix A(\)
constant.
H^j{f{\))
Kj^{-f(,X))
/.
is a product of elementary matrices if
and only
if |.4(X)| is
a non-zero
SMITH NORMAL FORM
194
8.
An n-square A-matrix ^(A) may be reduced
Prove:
to / by
[CHAP. 24
elementary transformations
if
and only if\A(X)\
is a non-zero constant.
9.
Prove:
A-matrix .4(A) over F[AJ has an inverse with elements in F[\]
if
and only
if /1(A) is
a product of
elementary matrices.
10.
Obtain matrices P(\) and Q(k) such that
P(\)- A(\)- Q(X) =
A(Xr^
and then obtain
Q(X)-P(X)
given
A+
=
-4(A)
A+
A+2
A+1
-A-1
A+2
Hint.
See Problem
Chapter
6,
A-1
Ans
5.
-A^ - 3 A - 2
-A
11.
Reduce each of the following
A%A
_2A''
>+
\^-2A
A
A= +
A^
A-1
(b)
+ 2A
>?
_ A^
A^- 2A+
2A=
A^
A+
2A''
- 2A +
3\^
-7A
l_
A=
A-
(c)
a''-
A^
2A
A"
AÂ""
+ 4
2A''
2A''-2A''
A +
A^ + 2
A" + A
A +
A^
+A
A"
A%A^
A=
A^ +
A""
A + 4
A""
- 2A^
a''
\^ + l
A^ +
+ A -
A"
(d)
00
-2A=
a""
- A
A^-
A=
A^ +
A"-
1
1
'-\j
'a=^
A= +
3A
A%
A -
+ 3
A^ +
A= +
4A -
A-2
A-1
A+
3\ +
4A+3
2\ + 2
A-1
A^
A-^
A^
A^ +
2A
A^ +
6A
A%
+ 4
"a^
A-2
-2\
'-VJ
A+
6A -
i_
A ^+ 2A + 3J
Obtain the elementary divisors over the rational

matrices of Problem 11.
3A + 2
\^
a""
A= + 3
(e)
12.
'-Xj
- A -
a""
if)
'Xy
2A-2
A+1
A% A
- A=
A+
A^ + 2
A^
-AÂ
2A
1
'-\>
2A^ - 3A + 2
-A^'-A +
normal form:
A-1
A^-1
X
(a)
to its Smith
2A-1
A^ +
A=( A-
ifOV+l I
field, the real field,
and the complex field
for
each of the
CHAP.
13.
SMITH NORMAL FORM
24]
The following polynomials
195
Find
are non-trivial invariant factors of a matrix.
its
elementary divisors in the
real field.
(a)
X^-X. X=-A^ A^-2A^
(b)
A+1, X^-
14.
(X^-lf. (X^-lf
1,
(c)
X,
X" + X,
A'
(d)
X.
A^ + A,
A''
+ 2A +
Ans.
+ X^
(A
(a)
X^,
(b)
A+1, A+1,
(c)
A,
A,
A^
(d)
A,
A,
A.
A,
A^,
+ 2X'' -
X''
2A^+
A''
X"^
A''+A''+2A'' + 2A +
A,
A^ +
A,
The following polynomials
(A^+lf,
l,
-if,
(A
(A
-if
A-1
(A^'+lf,
l,
+A
A-1
-if, A-1,
(A+lf, (A+lf, A-1,
A^'
A''
(A^
A+1
+ lf.
are the elementary divisors of a matrix
whose rank
What are
is six.
its invariant
factors?
(a)
A,
(6)
A^
.4ns.
15.
A+
A,
A=,
A+2, A +
1,
A,
(A-
1)2,
A+
3,
A-1
-if,
(c)
(A
(d)
A^,
A^
(A
-if,
(A
(A + 2f,
A,
-if, A-1, (A+lf

(A + 2f,
(A+2f
A(A + l)(A + 2)(A+3)(A + 4)
(a)
1,
1,
1,
1,
A,
(b)
1,
1,
1,
A,
A^cA-l),
(c)
1,
1,
A-1, (A-
(d)
1,
1,
1,
AÂ-lf
if, (A- if, (A- if (A+lf
A(A + 2f,
A''(A
+ 2f,
A''(A
2f
Solve the system of ordinary linear differential equations
!Dxx
+ (D + i)x2
(0 + 2)^1
(D + l)x2
where
xj, x^,
unknown
xs are
the polynomials in
computing form similar to

.ffi(-l),
Hs(2). Ks(l/5)
D
D+2
and
4.
D+1
'o'
Xl
-D + 1
D+2
X2
e\
3_
K2i(D +
l),
/I
H2i(-D-2). ff3i(D+l), K^siD). H^si-i). Kîh. K^gi^D
+1).
Hsî-kD),
obtain
k(D +
5D + 6
e*
combine as do the polynomials in A of a A-matrix; hence, beginning with a

that of Problem 6, Chapter 5, and using in order the elementary transformations:
of
-1
PAQ
at
D+1
:i2(-l),
(D+ 2)X2
system is
AX
Now
real functions of a real variable
In matrix notation, the
Hint.
- (D-l)xs
-4
.4
^(5D2+12D + 7)
l)
-jD
-1
-5D2-8D-2 -D 4D + 2
^-^(SD^ + TD)
i(5D2+7D + 2)
\D
A'l
d=+|d + | 5
the Smith normal form of A.
Use the
yi = 0,
y2=-4e*
QY
to carry
AX
into
AQY
and from
PAQY
N-J
PH
get
Finally, use
i
QY
(D2+|D
+ i)y3
6e*-l
and
ys
Kê"^^^ + Kê-^ + \e^ ~ ^

3
4
to obtain the required solution
3Cie-**/5 + it _ |,
^2
12Cie-'^*/5 + Cje"* - i,
Xs
-2Cie-**/s+ ie*+ i
chapter 25
The Minimum Polynomial of a Matrix
THE CHARACTERISTIC MATRIX \I-A
of an -square matrix A over F is a non-singular A-matrix

having invariant factors and elementary divisors. Using (24.4) it is easy to show
If
I.
Problem
In
II.
is a
we prove
1,
Two
diagonal matrix, the elementary divisors of \I- D are its diagonal elements.
A and B over F are similar over F if and only if their charsame invariant factors or the same rank and the same elemen-
re-square matrices
acteristic matrices have the

tary divisors in F[X.].
Prom Theorems
HI.
An
and
II,
we have
re-square matrix
A over F
is similar to a diagonal matrix if and only if
\I- A
has linear elementary divisors in F[\].
SIMILARITY INVARIANTS. The
invariant factors of \I -
are called similarity Invariants of A.
Let P(X) and Q(\) be non-singular matrices such that P(X) (\1-A)-Q(\)
mal form
diag(fi(X), 4(X),
Now
\P(\).(XI-A).Q(\)\
Since <^(X) and
IV.
/^(A) are
...,
monic,
P(A)
(?(A)
Smith nor-
jâ))
\P(X)\.\Qa)\4,a)
is the
A(A)-/2(A)-.../na).
and we have
The characteristic polynomial of an re-square matrix A

XI- A or of the similarity invariants of A.
is the
product of the invar-
iant factors of
THE MINIMUM POLYNOMIAL. By
the Cay ley-Hamilton Theorem (Chapter 23), every re-square matrix

A satisfies its characteristic equation <^(A) =
of degree re. That monic polynomial m(X) of
minimum degree such that m(A) =
is caUed the minimum polynomial of A and m(X) =
is
called the minimum equation of A. (m(X) is also called the minimum function oi A.)
The most elementary procedure for finding the minimum polynomial of i

following routine:
(I)
If
(II)
If
(III)
If
4 =
A ^
oq/,
then m{X) =
al for all a but
A^ ^ aA + hi
X-ao:
i^ = aÂ +
for all a
aj
and b but
're(A)
A''
A^
then
m(A) =
aÂ"^
a^)?
122
1.
Find the minimum polynomial of ^
196
- a^X - uq
+ aÂ + aÎ
- a^X - Oq
and so on.
Example
A^
then
5.^
involves the
CHAP.
25]
Clearly A
Using the
first
a^= i and oq
A^ = iA +
Problem
In
V.
if
If
is
and only
Problem
In
VI.
fn(X) of
is
impossible.
'9
two elements of the
'100"
'122"
=
oo
1
1
row of each matrix, we have
first
+ ao_
9 =
Oj^
8 =
2ai
then
After (and not before) checking for every element of A^, we conclude that
and the required minimum polynomial is X^ - 4X - 5.
= 5.
51
if
any re-square matrix over F and /(A) is any polynomial over F, then f(A) =
the minimum polynomial m{\) of A divides /(A).
we prove
3,
The minimum polynomial m(\) of an

A which has the highest degree.
Since the similarity invariants
f^(X), f^(k)
Vn. The characteristic polynomial

A and certain monic factors of m(X).
of
Set
we prove
2,
oqI
197
(fi(\)
ra-square matrix
is that similarity invariant
fn-iM all divide

A is the product
we have
^(A),
of
of the
minimum polynomial
and
Vin. The characteristic matrix of an ra-square matrix A has distinct linear elementary
if m(A), the minimum polynomial of A, has only distinct linear factors.
divisors if and only
NON-DEROGATORY MATRICES. An
re-square matrix
A whose
polynomial are identical is called non-derogatoiy;

IX.
An
re-square matrix
is
characteristic polynomial and
otherwise, derogatory.
non-derogatory
if
and only
if
minimum
We have
A has
just one non-trivial
similarity invariant.
It
is also
X.
If
easy
to
show
Si and Sg have minimum polynomials
polynomial
and m2(A).
ro(A) of the direct
is the least
This result may be extended to the direct sum of

XI.
let
Let gi(X),g2(X),
Aj be a non-derogatory matrix such that
B = diag(î, ^2
acteristic and
COMP ANION MATRIX.

(25.1)
We
^n)
has
matrices.
^(A)
\XI - Aj\ = \g-(X)\ "i

fii
fgi(A)!
ig2(A)P-ig^(A)!
>
= 1,2,
'^
Let A be non-derogatory with non-trivial similarity invariant
^(A)
C(g)
re
(;
,02
a" +
a-iA"'^
define as the companion matrix of g(X),
and for
multiple of ^^(A)
minimum polynomial.
g(X)
(25-2)
common
be distinct, monic, irreducible polynomials
..gnCA)
and mîX) respectively, the minimum
/rei(A)
sum D = diag (61,55)
= [-a],
if
g(X)
X+a
a^X +
Oq
in
F[A] and
...,m).
Then
as both char-
198
[CHAP. 25
1
1
(25.3)
C(g)
1
1
Oq
Problem
In
4,
-%
-e
-an-2
-'^n-3
-Gj
we prove
The companion matrix C(g) of a polynomial g(\) has g(\) as both

and minimum polynomial.
XII.
istic
(Some authors prefer to define C(g) as the transpose

Both forms will be used here.)
its character-
of the matrix given in (25.3).
See Problem
XIII.
If /4 is
non-derogatory with non-trivial similarity invariant f^(X) = {k-af,

"a
/ =
(25.4)
[o],
if
ra
and
1,
...
if
ra
>
...
f^{\)
then
O'
...
...
has
5.
easy to show
is
It
as its characteristic and minimum polynomial.
SOLVED PROBLEMS
1.
Prove: Two ra-square matrices A and S over F are similar over F if and only if their characteristic
matrices have the same invariant factors or the same elementary divisors in F[X\-
Suppose A and B are similar. From (i) of Problem 1, Chapter 20. it follows that Xl - A and XI -B are
Then by Theorems VHI and IX of Chapter 24, they have the same invariant factors and the same
equivalent.
elementary divisors.
let Xl - A and Xl - B have the same invariant factors or elementary divisors.
Chapter 24 there exist non-singular A-matrices P(X) and Q{X) such that
Conversely,
Theorem
VIII,
Xl - B
P{X)-{XI-A)-Q{X)
or
P{X)-{XI-A)
(i)
(Xl-B)-Q
(X)
Let
(ii)
P{X)
(X/-S).Si(A) + Ri
(iii)
Q(X)
S2(X)-(A/-S) +
/?2
(iv)
Q'\X)
S3(A)-(A/-.4) +
/?3
where R^, Rq, and R3 are free of
A.
(Xl-B)-S^(X)-(Xl-A)
Substituting in
+ Rj^(Xl-A)
i ),
we have
(A/-B)-S3(A)-(A/-4)
(XI-B)Rs
or
(V)
{Xl-B){S^{X)-Sa{X)\{Xl-A)
(A/-B)Rg
R^{XI-A)
Then by
CHAP.
25]
Then
S^(X)
and
Sg(A) =
(Xl-B)Rg
(vi)
since otherwise the
199
left
member
R^(\I-A)
of (v) Is of degree at least two while the right
member
is of degree at
most
one.
Using
(ill), (Iv),
and
(vi)
Q(^)-Q~\\)
Q(\)\Ss(\)-(Xl-A) + Rs\
Q(X.)-S3(X).(X!-A) +
Q(X)
Q(X)-Ss(X)-(Xl-A) + S2(X)-R^-(XI-A) + R^Rg
Ss(X)
(XI
\S2(X)-(XI-B) + R^lRg
- A) + S2(X)-(Xl-B)Rs + R^Rs
or
(vii)
- R^Rs
Now
Q{X)-Sq(X) + S^(X)R-^ =
X while the
right
member
and
is of
/ =
Since A.B.R^, and R^ are free of A.

to be proved.
Prove:
only
If
is
R2R3
Thus, Rg = R^
/?i(A/-^)/?2
F and
any ra-square matrix over
if
the minimum polynomial m(X) of
By
the division algorithm. Chapter 22,
(vii) is of
degree zero
XI - R'ÂR^ and A and B are similar, as was
f(X) is any polynomial in F[X]. then f(A) =
A divides
in
and, from (vi)
ARi2 - RÂR^
R^: then XI - B
/Ji =
RxUXl - A)
since otherwise the left member of
degree at least one.
XI - B
2.
\Q{X) Sg(X) ^ SîX)
if
and
/(A).
f(X)
q(X)-n(X) +
f(A)
q(A)-n(A) + r(A)
r(A)
and then
=
r(A)
Suppose f(A) = 0; then r(A) = 0. Now if r(A) ji 0, its degree is less than that of m(X), contrary
to the
hypothesis that m(A) is the minimum polynomial of ^. Thus, r(A) =
and m(A) divides /(A).
Conversely, suppose /(A) = ?(A)-m(A).
3.
Then f(A)
= q(A)-m(,A) = 0.
Prove: The minimum polynomial m(X) of an ra-square matrix A

which has the highest degree.
Let
gri-i(X)
denote the greatest common divisor of the
\XI-A\
cf,(X)
(n
is that similarity invariant fn(X) of
- l)-square minors
g-i(A)-/(A)
and
adj(A/-/l)
g_i(A)-S(A)
where the greatest common divisor of the elements of B(X)
Now
(A/-^).adj(A/-^)
(f)(X)-I
is 1.
sothat
(X/-A)-gn_^(X).B(X)
gn-i(A)-/(A)-/
(A/-^).S(A)
fn(X)-I
or
(i)
Then XI
-A
is a divisor of f^(X)-I
and by Theorem V, Chapter 23, f^(A) =
0.
of XI
-A. Then
200
By Theorem
V, m(\) divides
Suppose
fn(X-)-
(ii)
fn(\)
Since m(A) =
0,
XI
-A
Then, using (i) and
[CHAP. 25
is a divisor of
m(X)
g(X.)-m(\)
say
/,
m(\)-l
(\I-A).C(\)
fn(X.)-l
9(A)-m(A)-/
5(A)
?(A)
(ii).
(\I-A)-B(\)
9(A)
(A/ - ^). C(A)
and
Now
9(A) divides every element of iS(A); hence ?(A)
=1
C(A)
and, by
(ii),
m(A)
/n(A)
as was to be proved.
4.
Prove: The companion matrix C(g) of a polynomial g(\) has g(A) as both
minimum polynomial.
The characteristic matrix
its
characteristic and
of (25.3) is
-1
-1
^1
"2
an-2
To
the first column add A times the second column,

column to obtain
A + a^_j
A times the
third column,
,n-l
times the last
-1
A
G(A)
Since
|G(A)| = g(A),
a^
Oi
g(A)
-1
a_2
A+a^-l
the characteristic polynomial of C(g) is g(A). Since the minor of the element
g(A)
common divisor of all (n-l)-square minors of G(\) is 1. Thus, C(g) is non -derog-
in G(X) is 1, the greatest
atory and its
5.
minimum polynomial
The companion matrix
of
g(\)
is g(A).
A^ + 2A
A=
+ 6A -
is
10
10
10
1
5-61-20
or, if
preferred.
10
0-6
10
1
10-2
1
CHAP.
25]
201
6.
Write the companion matrix of each of the following polynomials:
- 2\ -
+ \^
- X^
+ 2X
X''
(d)
X""
(b)
(X^-i)(X+2)
(e)
]?()? + I)
(c)
(X-lf
(/)
(A+2)(A^-2A^ + 4A-8)
'o
Ans.
(a)
-1
--3
(d)
(/)
0-10
7.
Prove: Every 2-square matrix
8.
Reduce G(X) of Problem
A=
[a^]
4 to diag (1
16
which (011-022)^ +
for
o'
(e)
-2
-2
o"
(c)
2X!'
o"
(b)
1
1
9.
(a)
g(X)
4012021/0
is non-derogatory.
).
For each of the following matrices ^. (i) find the characteristic and minimum polynomial and
(ii) list
non-trivial invariant factors and the elementary divisors in the rational field.
1
(a)
(b)
'211
if)
Ans.
2
(c)
1"
2-3
(g)
m(X)
(A-l)(A-2)(A-3);
(b) <^( A)
m(,\.)
X^;
if)
if. = e.d
(^(A)
(A-if(A-2);
i.f.
m(X)
(A-1) (A-2)
e.d.
i.f.
(A+ihA-5)
i.f.
A+1, (A+i)(A-5)
(A+l)(A-5)
e.d.
A+i, A+l, A-5
,^(A)
A(A-H)^(A-1)
m(A)
A(A^-l)
(/)(A)
A2(A+1)2
m(A) =
A(A+1)2
(A-2) (A""-
m(X)
A^-A-2
Prove Theorems VII and
A,
i.f.
A+l,
ih)
_1
4-6 3 s"
3-2
2
4-3 4-1 -6
4-2 4 0-4
-1
0-2 1 2
1
e.d. (A-1), (A-2).
(A-3)
2)'';
A^-A
A+l, A+l, A-1
A(A+l)2
e.d. A. A,
A-
A-
e.d. A,
oS(A)
(h)
i.f.
A''-4A
e.d. A, A,
c^(A)
is)
A,
(e)
A-1, A-1, A-2
i.f.
A-1, (A-i)(A-2)
- 4A=
-5
-2
(A-l)(A-2)(A-3);
m(A)
A=
= A=
(/)(A)
m(A) = A^ - 4A
1-3
-1 -6 -3 -6
-3 -3 -4 -3
ie)
id)
1
1
<^(A)
(a)
id)
11.
-6 -2 -3 -2
-3 -1 -1 -2
(c)
10.
-2 -1 -3
(A+D^
i.f.
e.d.
A-2, A'^-A-2, A^-A-2

A-2. A-2, A-2, A+l, A+l
VIII.
Prove Theorem X.
Hint.
m(D)
= diag(m(Bi), m(B^)
requires
m(Si) = miB^) =
thus, mi(A) and
m^X)
divide m(X).
the
202
12.
Prove Theorem XI.
A
Show
n-square and if k is the least positive integer such that A = 0, .4 is culled nilpotent of index
that A is nilpotent of index k if and only if its characteristic roots are all zero.
13. If
14.
is
Prove: (a) The characteristic roots of an n-square idempotent matrix A are either
The rank
(b)
15.
Prove: Let
of
A.B.C.D
same invariant factors

Follow the proof
Hint.
Prove:
number of characteristic roots which are
is the
be ra-square matrices over
P and Q such
matrices
16.
PCQ
that
or the
Problem
in
18.
Use the minimum polynomial
F
it
with C and
and only
if
There exist non-singular
non-singular.
R(X) = \C - A
and
S(A.)
XD-B
have the
1,
noting that similarity is replaced by equivalence.
s,
then A~^ is expressible
slmA.
to find the inverse of the matrix
of
Problem
9(h).
Prove: Every linear factor X-A.J of 0(A) is a factor of m(A).
20.
PDQ
1.
1.
the minimum polynomial m(X) of a non-singular matrix A is of degree
If
The theorem follows from Theorem VII or assume the contrary and
Then {A -\Î)q(A) + rl =
and A -X^l has an inverse.
Hint.
19.
= A.
or
A:.
same elementary divisors.
as a scalar polynomial of degree
17.
[CHAP. 25
^ 0.
Use
-4
Prove:
If
^
to
[o
show
is not the product of the distinct factors of 0(A).
g(\) is any scalar polynomial in A, then g(^) is singular if and only if the greatest
Suppose d{\)
Suppose d(A)
(i)
(ii)
minimum polynomial
i-
common
divi-
oi A, is d(k) ^ 1.
and use Theorem V, Chapter 22.

and use Theorem IV, Chapter 22.
Problem 20 that when
21. Infer from
r,
;]
sor of g(A) and m(A), the

Hint,
minimum polynomial
that the
m(A) = {\-X^)q{X) +
w:rite
g(.4) is
non-singular, then [g(/l)]
is
expressible as a polynomial in A of
degree less than that of m(\).
22.
Prove:
If
the
minimum polynomial m(A) of A over F is irreducible in F\X\ and is of degree

A with coefficients in F of degree < s constitutes a field.
s in A,
then the
set of all scalar polynomials in
23.
Let A and S be square matrices and denote by m(A) and (A) respectively the minimum polynomials oi
and BA. Prove:
(a) m(A) = n(A)
(&) m(A)
Hint.
24.
when not both
and n(A) differ
B.m(^S)./l
Let A be of dimension
at
.A
and B are singular.
most by a factor A when both A and B are singular.
(B,4).m(B.4)
mxn
Let
X.I
and
A-n{BA)-B
m >
n,
and S be of dimension nxm,
the characteristic polynomials of
25.
AB
and BA.
Show 0(A)
(AB)-n(AB)
Xj^ is
26. If the matrices
0.
and denote by 0(A) and 0(A) respectively

(/f(A).
be an invariant vector associated with a simple characteristic root of A.
mute, then
AB
Prove:
If
A and B com-
an invariant vector of B.
A and B commute, state a theorem concerning the invariant vectors
simple characteristic roots.
of
B when A has only
chapter 26
Canonical Forms Under Similarity
THE PROBLEM.
it was shown that the characteristic matrices of two similar n-square

R'ÂR over F have the same invariant factors and the same elementary divisors.
we establish representatives of the set of all matrices R'ÂR which are (i) sim-
Chapter 25
In
matrices A and
In this chapter,
ple in structure and
These matrices,
ical matrix
(ii)
put into view either the invariant factors or the elementary divisors.
four in number, are called canonical forms of A.
'[^
introduced earlier for

'
o]
all
mxn
They correspond
matrices of rank
to the
canon-
under equivalence.
THE RATIONAL CANONICAL FORM.
Let A be an ra-square matrix over F and suppose first that its

characteristic matrix has just one non-trivial invariant factor /(\). The companion matrix
C(f )
of /(A) was shown in Chapter 25 to be similar to A. We define it to be the rational
canonical
form S of all matrices similar to A
Suppose next that the Smith normal form of \I - A

diag
<26-l)
(1, 1
with the non-trivial invariant factor
1,
f^(\.)
f.(X), f.^îX),
...,
of degree s^,
(i
tional canonical form of all matrices similar to
(26-2)
To show
D^ = diag
is
f^(X))
= j.j +
n).
We define as the
di^s {C(fj), C(fJ ^^)
C(/))
A and S have the same similarity invariants we note that C(f-)
that
l,/i(A))
(1, 1
and, thus, S is similar to diag(D^-
l.fja), fj^^(\)
is similar to
By a sequence
D).
.^^
interchanges of two rows and the same two columns, we have S similar
diag (1,1
ra-
of
to
fîX))
We have proved
I.
Every square matrix A is similar to the direct sum (26.2) of the companion matrices
of the non-trivial invariant factors of XI - A.
Example
1.
Let the non-trivial similarity invariants of A over the rational field be

/8(A)
X+1.
fg(X)
A%L
/io(A)
A^ +
2A +
Then
1
1
1
C(fa)
= [-1],
C(f9)
C(fio)
10
-10
-10
203
-2
204
[CHAP. 26
and
-1
1
1
-1
1
diag(C(/g),C(/9),C(/io))
1
1
1
1
-1
Is the required form of
Note.
immaterial.
The
Theorem
-2
I.
order in which the companion matrices are arranged along the diagonal is
Also
-1
-1
1
1
-1
1
1
-2
1
1
using the transpose of each of the companion matrices above is an alternate form.
A SECOND CANONICAL FORM.
Let the characteristic matrix of A have as non-trivial invariant

tors the polynomials f^(X) of (26.1). Suppose that the elementary divisors are powers of t
tinct irreducible polynomials in F[X]: pi(A,), paCX), ..-, PjCA). Let
(26.3)
fi(\)
ipi(X)l^"!p2(A)!^2^..{p,(X)!'îs(pi(A)i'"''iP2(A)i'"''...iPi(
(i
= i,i+'^
dis-
n)
The companion matrix
where not every factor need appear since some of the ^'s may be zero.
C{p?^^) of any factor present has
fac-
\])^(X)\ ^^ as the only non-trivial similarity invariant;
hence,
C{{^) is similar to
diag (C(p^?"
).
Cipl-'i
Cip^^ti))
We have
Every square matrix A over F is similar to the direct sum of the companion matriII.
ces of the elementary divisors over F of XI A.
Example
2.
For the matrix A of Example

(A.+ l)^,
A^-A,+
l,
(X
-A+1)
1,
.
the elementary divisors over the rational field are
Their respective companion matrices are

1
1
[-1],
[-1],
[-:
.;]
[.:
\
1
and the canonical form of Theorem
II is
-3
A+1, X+1,
CHAP.
26]
205
-1
-1
1
-1 -2
1
-1
1
1
1
1
-1
THE JACOBSON CANONICAL FORM.
-3
Let A be the matrix of the section above with the elementary
divisors of its characteristic matrix expressed as powers of irreducible polynomials in F\_\].
Consider an elementary divisor {p(X)!^.
If
q =
use C(p), the companion matrix;
1,
if
q>l,
build
C(p)
C(p)
(26.4)
where
Cq(P)
C(p)
C(p)
same order as C(p) having the element 1 in the lower left hand corner
The matrix C^^Cp) of (26.4), with the understanding that Ci(p) = C(p) is
is a matrix of the
and zeroes elsewhere.
called the hypercompanion matrix of \p(X)]'^.

I's just
..
Note that
in (26.4), there is a
continuous line of
above the diagonal.
When the alternate companion matrix
C'(p) is used, the hypercompanion matrix of jp(A)!''is
C'ip)
C'ip)
C'ip)
Ca(P)
C'ip)
C'ip)
where A' is a matrix of the same order as C'ip) having the element 1 in the upper right hand corner and zeroes elsewhere. In this form there is a continuous line of I's just below the diagonal.
Examples. Let
IpiX.)]'^ = (\^
2\-
Then
1)'^.
C(p) =
M,
M=y
1
1
-2
1
1
-2
CqiP)
1
1
-2
1
1
1
-2
and
206
In
Problem
1, it
is
Thus, Cq(p) is similar
shown
to
[CHAP. 26
that Cq(p) has {pCA)}*^ as its only non-trivial similarity invariant.
C(p^) and
may be substituted
for it in the
canonical form of Theorem
II.
We have
Every square matrix A over F
III.
is similar to the direct
matrices of the elementary divisors over
Example
4.
For the matrix A of Example
A+1 and A -A+1
2,
of
sum of the hypercompanion
\I - A.
the hypercompanion matrices of the elementary divisors A +
are their companion matrices, the hypercompanion matrix of
1,
(A+l)îs
r-1
10
-11
and that of (A^-A + 1)^ is
Thus, the canonical form of Theorem
0-11
nils
-1
-1
-1
-1
1
-1
1
1
-1
1
1
-1
The use
of the term "rational" in connection with the canonical form of
what misleading.
It
was used
Theorem
is
some-
originally to indicate that in obtaining the canonical form only
A are necessary. But this is, of course, true

Theorems n and III. To further add to the consometimes called the rational canonical form.
rational operations in the field of the elements of
also of the canonical forms (introduced later) of

fusion, the canonical form of
Theorem
III
is
THE CLASSICAL CANONICAL FORM.
Let the elementary divisors of the characteristic matrix of A

be powers of linear polynomials. The canonical form of Theorem III is then the direct sum of
hypercompanion matrices of the form
i
(26.5)
..
1
1
C^(P)
..
.
corresponding to the elementary divisor
!p(A)!
1
<^.,
(A-a^)^
This special case of the canonical form of Theorem

canonical form.
III is
For an example, see Problem
known as the Jordan

We have
]
2.
or classical
[Note that C^(p) of (26.5) is of the type / of (25.4).
Let
be the field in which the characteristic polynomial of a matrix A factors

Then A is similar over ^to the direct sum of hypercompanion matrices of the form (26.5), each matrix corresponding to an elementary divisor (X-a^)'^.
IV.
into linear polynomials.
Example
5.
Let the elementary divisors over the complex field of \I
-A
be:
A-i, X + i, (X-i)
(X+i)
CHAP.
26]
The classical canonical form
of
207
is
-1
1
-i
-i
From Theorem IV follows

V.
An
re-square matrix
is similar to a diagonal matrix if
and only
divisors oi XI -
are linear polynomials, that is, if and only if the

is the product of distinct linear polynomials.
if
the elementary
minimum polynomial of
See Problems 2-
4.
A REDUCTION TO RATIONAL CANONICAL FORM.
In concluding this discussion of canonical forms,

that a reduction of any re-square matrix to its rational canonical
form can be
made, at least theoretically, without having prior knowledge of the invariant
factors of \I - A.
A somewhat different treatment of this can be found in Dickson, L. E., Modern Algebraic
it
will be
shown
Benj.H. Sanborn, 1926. Some improvement on purely computational

aspects
Browne, E. T., American Mathematical Monthly, vol. 48 (1940).
ries,
We
in
to
Theo-
is
made
in
shall need the following definitions:
If A is an re-square matrix and X is an re-vector

over F and if g(X) is the monic polynomial
F[\] of minimum degree such that g(A)-X = 0. then with respect
to A the vector X is said
belong to g(X).
If,
vectors
Example
with respect to A
X,AX.A''X
6.
Let A
while
[1,0, -ij',
/4^
-6
-3
-2
= X.
AY
the vector
belongs
to g(X) of
are called a chain having
The vectors
Then
(A
Y;
thus,
and
A"
degree
p,
the linearly independent
as its leader.
AX
= [l, 0, oj' and
-I)X=0
=[-1,0,1]'
= [2,
1, l]'
are linearly independent
belongs to the polynomial A,^-l

For y =
and F belongs to the polynomial A+ 1.
(A+I)Y=0
m(\) is the minimum polynomial of an re-square matrix A,

then m(A)- X =
for every reThus, there can be no chain of length greater than the degree of
m(A). For the matrix
of Example 6, the minimum polynomial is A^ - 1.
If
vector X.
Let S be the rational canonical form of the re-square matrix A

over F.
R over F such that
Then, there exists a
non-singular matrix
(26-6)
R'ÂR
diag(C.,C.^^
C)
where, for convenience, Q/^) in (26.2) has been replaced

by C^.
companion matrix of the invariant factor
/iCA)
Aî
c^ ^_X'i-
c,.
We
A
has the form

-^tl
1
-î2
1
-î3
Ci
.
shall
assume
that
C,-
the
208
Prom
(26.6),
we have
AR
(26.7)
RS
Rdiag(Cj,Cj^^
R-,R-^
Let R be separated into column blocks

have the same number of columns.
AR
A[R..R.^^
[CHAP. 26
From
RJ
R^
C)
so that R^
and C^, (i=/,/ +
n)
(26.7),
IRj.Rj,,
iRjCj.Rj^^C.^^
C)
/?Jdiag(C,,..C^.^^
R,,CJ
and
AR^ = R^Ci.
Denote the
si
(i=j,j +
column vectors of Ri by Ri^, R^s.
n)
Ris- and form the product

t
RiCi
[Rii, Ri2,
, Risi\Ci
st
[Ri2,Ri3, ,Risj,> -
Rikîk]
^__
Since
ARi
A[Ri^,Ri^, -.Risi]
[ARi^.ARi2,. ..ARisil
RiCi
we have
ARn,
Ri2 =
(26.8)
Ris - ARi2 = A Rii,
Risi = A^'-'Ri.
and
-|>i.%
(26.9)
^Ris,
we obtain
Substituting into (26.9) from (26.8),
-i=
A'iRu
c^kA^-'Ri^
l
or
(A^^ + cis.A^'i'^
(26.10)
Prom the
definition of C^ above, (26.10)
(26.11)
...
may be
aÂ
+ cirDRir
written as
fi(A)-Ri^
Let Rit be denoted by X, so that (26.11) becomes fj(A)-X: =0; then, since X-,AXj,
A ^ X^ are linearly independent, the vector X^ belongs to the invariant factor fîX).
Thus, the column vectors of fi^ consist of the vectors of the chain having X^, belonging to /^(X),
Xj^
as leader.
To summarize:
the n linearly independent columns of R, satisfying (26.2), consist of re-/+l
chains
Xi,AXi
whose leaders belong
A^^"'Xi
(i=j,i +
to the respective invariant factors
<
satisfy the condition
s,-
s- ^^
+1
...
/ (A),
n)
fj.AX.)
/^(X)
and whose lengths
^ s.
n
We have
VI.
Por a given re-SQuare matrix A over F:

(i)
let
X^ be
(it)
let
X^_^ be the leader
the leader of a chain
(iii)
let
X^^^he
is linearly
F which
and so on.
Then,
for
^_^
maximum length
for all
ra-
vectors over F\
of
are linearly independent of the vectors of ;
the leader of a chain -n-2 ^ maximium length (any member of which

independent of the preceding members and those of and fi'^.i)
for all n- vectors over
and_i;
of
maximum length (any member of which

preceding members and those of for all n-
of a chain
is linearly independent of the
vectors over
F which
are linearly independent of the vectors of
CHAP.
26]
AR
Example
7.
209
is the rational canonical form of A.
Let
Take ;f=[l,0,o]'; then ^,
4X
= [l,
^;f = [14,25,30]'= 54^X- a:.

m(X) = A - 5A^ + 1 = <jS(A). Taking
Independent while
arly
belongs to
/gCX) =
= Q
and
A'
15
6
we
[3, 5. 6].' are line-
4A',4=A']
[a;.
^Â: =
{A-bA^+I)X
Thus,
1, l]',
find
-3
6-5
-1
AR
[AX.A'^Xâ'^X]
14
25
30
0-1
and
R ÂR
Here A is non-derogatory with minimum polynomial m(A) irreducible over the rational
Every 3-vector over this field belongs to m(X). (see Problem 11), and leads a chain
field.
of length
three.
The matrix R having
Example
8.
Let
the vectors of any chain as column vectors is such that
Tak
A"
then ^ A' =
= [l,-l,o]';
cannot be the minimum polynomial m(A) of A.

11). and could be a similarity invariant of A.
Next, take
F=[l,0,0]',
independent while
A - 5A - 3A +
A^Y
The vectors
= [54,43,46]'
7 = <^(A).
It is,
A-
and
AT
belongs to
A-
1.
S.
NowA-1
however, a divisor of m(A), (see Problem
Y, -4? = [2,
= 5A'^Y +
The polynomial
AT
R'ÂR =
1, 2]',
SAY -77.
^^^ =[ 11,
8, s]'
are linearly
Thus, Y belongs
is not a similarity invariant;
m(A) =
to
in fact,
the first choice of vector belongs to a polynomial which could reasonably

be the
function, it should be considered a false start. The reader may verify
that
unless
minimum
R'ÂR
11
when
[Y,AY,A'^Y]
See Problems
5-6.
SOLVED PROBLEMS
Prove:
The matrix Cîp)
Let C^(p) be of order

so that the greatest
tors of
of (26.4) has !p(A)!'? as its only non-trivial similarity invariant.
s.
common
The minor of the element in the last row and first column of A/ - C
(p) is 1
divisor of all (s-l)-square minors of A/ - C (p) is 1.
Then the invanant fac-
A/ - C^(p) are 1,1

0(A)
l.f^(\).
=
But /^(A) =
|A/-C^(p)|
ip(A)!'?
since
\XI-C(pf
jp(A)!9
210
2.
The canonical form
(a) is that of
X^+4X +6A^ +4X +
divisor being
Theorems I and II, the non-trivial invariant factor and elementary

The canonical form of Theorem III is (b).
1.
-1
(b)
(a)
-1
-4
-1
The canonical form
[CHAP. 26
(a) is that of
-6
-1
-4
Theorem
and the elementary divisors being X +

Theorems II and III is (b).
2,
I,
2,\ -4, A'+3A -4A-12

The canonical form of both
the invariant factors being \ +
X+
2,
X+
2,
X-
2,
X-
X+
2,
3.
-2
-2
4
(b)
(a)
12
4.
The canonical form

X+
2,
X+
2,
(X'^
-3
-3
Theorem III. Over the rational field the elementary divisors

+2X- if and the invariant factors are
(a) is that of
+2X-lf
(X^
(X + 2)(X^ + 2X -
The canonical form
of
Theorem
is (b)
2)(X^+2X-1)^
(X +
1)^
and that of Theorem
II
is (c).
-2
-2
1
-2
1
--2
1
(a)
-2
1
1
-2
1
1
-2
1
1
1
-10
--6
1
(b)
1
ID
1
1
1
11
12
17
-14
--21
-8
are
CHAP.
26]
211
-200000000000
-2
1
1
-1
-2
-4
(c)
1
1
1
1
-6
-2
-1
-6
-2
-1
-1
-1
0_
-1
A =
Let
5.
-2
1
-2
AX
Then
[-2
-9
-6
Take X
-1
pendent while
m(\) = A* - 2X^ +
1, 1, 1. 1, 1]',
A^X
= [1,0,-1,0,0
= [l, 0, -2, 0, 0, -2]' = 2^=A:
-4
A"
and write Xg for X.
-Z;
-ir.
A''
belongs to
=[-3.1.1,1.
A*-2A^ + 1
1,2]'
We
are linearly inde-
tentatively
assume
Y = [o, 0, 0, 1, 0, O]' is linearly independent of the members

of the chain led by
-1, 1,0]' is linearly independent of Y and the members
2,,
= Y
of the chain. Now A^Y
Xg and
The vector
AY
- [-1,0,
Y belongs
Xs
1,
to
A -1.
Since the two polynomials complete the set of non-trivial

invariant factors
When
for Y.
-1
[Xs.AXs.Xg.AXe.A^Xe.A^'Xfi
1
-2
-3
"o
-1
-1
-1
write
-1
R~'ÂR 1
so that
we
2
1
the rational canonical form 'of ^.
Note.
and
AZ
The vector Z =
- [3, 0, -2,
1,
-2,
[o, 1, 0, 0, 0. Oj' is linearly
o]' is
linearly independent of
independent of the members of the chain led by Xg

Z and the members of the chain However A'^Z =
1-1,1,0,0,0,1]'= -AXg + A Xg + Z; then (A'' ~ 1)(Z - AXg) =

and W=Z-AXe = [2.0 -1 -1 ll -i]'
belongs to A - 1. Using this as X^. we may form another R with which
to obtain the rational canonical' form.
-2
6.
A =
Let
Then
r
-1
-1
-1
-2
-1
-1
-4
-4
-1
-2
-2
-2
-2
-2
^.^5
2,
Take X = [1,0,0,0,0]'.
[-2,1,-1,-1,-2]',
X-3X
-2, -2,0J - 2A
polynomial m(A) and label
L-1,
-1
and
as X^.
^^^=[1,1,-1,-1,0]' are linearly independent while a'^X^

3.
We tentatively assume this to be the minimum
belongs to A - 2A^ +
212
[CHAP. 26
When, in A, the fourth column is subtracted from the first, we have [-1,0,0,1,0]'; hence, if y =
Again, when the fourth column of A is subtracted from
-1, o]', AY = -Y and Y belongs to X+ 1.
the third, we have [o, 0, -1, 1, O]'; hence, if Z = [O, 0, 1, -1, O]', AZ = -Z find Z belongs to A+ 1. Since
y, Z, and the members of the chain led by X^ are linearly independent, we label Y as X^ and Z as X^. When
[l, 0, 0,
11-2
-1
11
R
[Xs.X^.X^.AXs,.A%]
0-1-1
-1
0-100
-1
R~ AR
0-3
10
-1
-1
12
0-2
the rational canonical form of A.
7.
For each of the matrices (a)-{h) of Problem
Can any
over the rational field,
9,
Chapter 25, write the canonical matrix of Theorems
of these matrices be
I,
n, in
changed by enlarging the number field?
Partial Ans.
(a)
I,
-11
"o
(b)
II,
m.
I, II,
3)
o"
0_
"o
o'
_0
n, in.
I,
diag(l,2,
_0
(e)
ni,
-1
-1
1
(/)
n, ni.
I.
-1
1
o'
m.
n.
(g)
-1
-1
-2.
-1
-1
'2
(h)
8.
I,
10
10
n. in,
diag(2, 2,2, -1, -1)
Under what conditions will (a) the canonical forms of Theorems I and n be identical? (b) the canonical
forms of Theorems n and in be identical? (c) the canonical form of Theorem H be diagonal?
CHAP.
26]
"o
9.
o"
Identify the canonical form
Check with the answer
to
Let the non-singular matrix A have non-trivial invariant factors

Write the canonical forms of
that of
Theorem
Ans.
(a)
Problem
8(6).
0_
_0
10.
213
(a) X + 1
Theorems I, n,
A,
III
(X + i)^
IV.
-1
1
1
-1
10
1
1
1
1
-10
_0
-2
0_
-1
-1
1
-1
-2
1
n.
-1
1
1
1
1
-
-1
-3
2
6"
-1
-1
-1
-1
1
in,
-1
1
1
--1
-1
-1
1_
o"
-1
-1
-1
a
IV,
where a,^ = ^(li\/3).
a
/S
/3
(b)
X^+
1,
over the rational field and
214
11.
Hint.
12.
Prove: If with respect to an n-square matrix A, the vector

polynomial m(A) of A.
In
Suppose the contrary and consider
Example
show
6,
X,AX. and Y
that
m(A)
A (A)
belongs to g(X) then g(X) divides the minimum
g (A) +
"
[CHAP. 26
r (X).
are linearly independent and then reduce
to its rational canonical
form.
13.
In
Problem
6:
Take 7
(a)
= [o,
linearly independent of the chain led by ^5, and obtain
1, 0, 0, o]',
X4
y_
(3/1
_2/)
A^g
belonging to A+1.
Take
(i>)
= [O, 0,
belonging to A +
Compute R~
(c)
14.
led by X5, and obtain
X^ = Z
-X^
1.
using the vectors X3 and X^ of (b) and
For each of the matrices A of Problem 9(a)-(h), Chapter

ical form of
15.
AR
X^ and the chain
linearly independent of
1, 0, O]',
(a) to build R.
25, find
R such
R' AR
that
is the rational canon-
Solve the system of linear differential equations

dxi
2Xx
X2
Xg
4^1
2x2
3At3
6*1
2a:2
- 3*3 -
2xi^
3ai
X2
2x4.
X4 +
dt
dx2
dt
dxs
dt
dX4
dt
where the x^ are unknown functions of the

TT.
i.
Hint.
Leti
^V
U, . .3. -JT
J
define
(i)
dt
real variable
"^^
dxQ
dxi
|^^,
111
dX_
-6
-2
-3
-2
-3
-1
-1
-2
Since the non-singular linear transformation
dY
RY
Xq
t.
dxA'
dxn
-jf. -jf,
^J
AX
carries
R ÂRY +
and rewrite the system as
(i)
into
R^H
dt
choose R so that R~
AR
is leader of the chain
The elementary 4-vector 1 belonging to A - A

Now
4 yields X2 = E^- Xi + 2AXi belonging to A + 1.
is the rational canonical form of A.
X-i
= Ei, AX^.A'^X-i
while
12-1
with
4-2
8
0-6 4 -12
0-3 2-5
[X.L,AXi,A^Xi,X2]
0'
t
dY^
dt
Y +
yi + 73
y2
-1
-74
Then
2Ci +
C^+it'^
Cge* + Cge-* -
-Ci +
-t_l,2^
J. Cse"Cse^Sr
46"*
and
RY
C2e*+
3(Cs + C4)e"*+
^-2t+l'
2C1 + 2C2e*+ 2(3C3 + 4C4)e-* + ^ -4C1 - 2C2e*- 2(5C3+6C4)e-* - 2^+
-2C1 -
Cae*-
5(C3+C4)e"* -
4 + 2
6t
- 4
2+3{-2_
INDEX
Absolute value of a complex number, 110
Addition
of matrices,
2,
Characteristic roots (cont.)

of Hermitian matrices, 164
of inverse A, 155
of vectors, 67
of real orthogonal matrices, 155

of real skew-symmetric matrices, 170
Adjoint of a square matrix

definition of, 49
of real symmetric matrices, 163
determinant of, 49
inverse from, 55
rank of, 50
Algebraic complement, 24
Anti-commutative matrices, 11
Associative laws for
of unitary matrices, 155

Characteristic vectors
{see Invariant vectors)
Classical canonical form, 206
addition of matrices, 2
fields, 64
multiplication of matrices, 2
Augmented matrix, 75
Cofactor, 23
Cogredient transformation, 127
Closed, 85
Coefficient matrix, 75
Column
space of a matrix, 93
transformation, 39
Commutative law for
addition of matrices, 2
fields, 64
multiplication of matrices, 3
Commutative matrices, 11
Companion matrix, 197
Complementary minors, 24
Complex numbers, 12, 110
Conformable matrices
Basis
change
of,
95
of a vector space, 86
orthonormal, 102, 111

Bilinear form(s)
canonical form of, 126
definition of, 125
equivalent, 126
factorization of, 128
rank of, 125

reduction of, 126
for addition, 2
for multiplication, 3
Congruent matrices, 115
Canonical form
Conjugate
of a complex number, 12
classical (Jordan), 206
Jacobson, 205
of bilinear form, 126
of Hermitian form, 146
of matrix, 41, 42
of quadratic form, 133
rational, 203
row equivalent, 40
Canonical set
under congruence, 116, 117
under equivalence, 43, 189
under similarity, 203
Cayley-Hamilton Theorem, 181
Chain of vectors, 207
Characteristic
equation, 149
polynomial, 149
Characteristic roots
definition of, 149
of adj A, 151
of a diagonal matrix, 155
of a direct sum, 155
of a matrix,
of a product,
12
13
of a sum, 13
Conjugate transpose, 13
Conjunctive matrices, 117
Contragredient transformation, 127
Coordinates of a vector, 88
Cramer's rule, 77
Decomposition of a matrix into
Hermitian and skew-Hermitian parts, 13
symmetric and skew-symmetric parts, 12
Degree
of a matrix polynomial, 179
of a (scalar) polynomial, 172
Dependent
forms, 69
matrices, 73
polynomials, 73
vectors, 68
Derogatory matrix, 197
215
INDEX
216
Hermitian form
canonical form
Determinant
definition of, 20
derivative of, 33
of,
146
147
index of, 147
rank of, 146
definite,
expansion of
along first row and column, 33
along a row (column), 23
by Laplace method, 33
multiplication by scalar, 22
of conjugate of a matrix, 30
of conjugate transpose of a matrix, 30
of elementary transformation matrix, 42
of non-singular matrix, 39
of product of matrices, 33
of singular matrix, 39
of transpose of a matrix, 21
Diagonal
elements of a square matrix,
semi-definite, 147
signature
of,
147
Hermitian forms
equivalence
of,
146
Hermitian matrix, 13, 117, 164

Hypercompanion matrix, 205
Idempotent matrix, 11
Identity matrix, 10
Image
1
matrix, 10, 156
Diagonable matrices, 157

Diagonalization
by orthogonal transformation, 163
by unitary transformation, 164
Dimension of a vector space, 86
Direct sum, 13
Distributive law for
fields, 64
matrices, 3
Divisors of zero, 19
Dot product, 100
Eigenvalue, 149
Eigenvector, 149
Elementary
matrices, 41
n-vectors, 88
transformations, 39
of a vector, 94
of a vector space, 95
Index
of an Hermitian form, 147
of a real quadratic form, 133
Inner product, 100, 110
Intersection space, 87
Invariant vector(s)
definition of, 149
of a diagonal matrix, 156
of an Hermitian matrix, 164
of a normal matrix, 164
of a real symmetric matrix, 163
of similar matrices, 156
Inverse of a (an)
diagonal matrix, 55
direct sum, 55
elementary transformation, 39
matrix, 11, 55
product of matrices, 11
symmetric matrix, 58
Involutory matrix, 11
Equality of
matrices, 2
matrix polynomials, 179
(scalar) polynomials, 172
Equations, linear
equivalent systems
solution of, 75
Kronecker's reduction, 136

of,
75
Lagrange's reduction, 132
system of homogeneous, 78
system of non-homogeneous, 77
Equivalence relation, 9
Equivalent
bilinear forms, 126
Hermitian forms, 146

matrices, 40, 188
quadratic forms, 131, 133, 134
systems of linear equations, 76
Factorization into elementary matrices, 43, 188

Field, 64
Field of values, 171
First minor, 22
common
Lambda
matrix, 179
Laplace's expansion, 33
Latent roots (vectors), 149
Leader of a chain, 207
Leading principal minors, 135
Left divisor, 180
Left inverse, 63
Linear combination of vectors, 68

Linear dependence (independence)
of forms, 70
of matrices, 73
of vectors, 68
Lower triangular matrix, 10

Matrices
congruent, 115
Gramian, 103, 111

Gram-Schmidt process,
Greatest
Jacobson canonical form, 205

Jordan (classical) canonical form, 206
equal, 2
102, 111
divisor, 173
equivalent, 40
over a
field,
65
INDEX
Matrices (cont.)
product of, 3
Null space, 87
Nullity, 87
scalar multiple of, 2

similar, 95, 156
square, 1
sum
of, 2
Matrix
derogatory, 197
diagonable, 157
diagonal, 10
elementary row (column), 41

elementary transformation of, 39
Hermitian, 13, 117, 164
idempotent, 11
inverse of, 11, 55
lambda, 179
nilpotent, 11
nonderogatory, 197
non-singular, 39
normal, 164
of,
w-vector, 85
Order of a matrix,
Orthogonal
congruence, 163
equivalence, 163
matrix, 108
definition of, 1
normal form
217
41
nullity of, 87
similarity, 157, 163

transformation, 103
vectors, 100, 110
Orthonormal
basis,
102,
111
Partitioning of matrices, 4
Periodic matrix, 11
Permutation matrix, 99
Polynomial
domain, 172
matrix, 179
monic, 172
scalar, 172
scalar matrix, 180
of a bilinear form, 125
Positive definite (semi-definite)
of an Hermitian form, 146

of a quadratic form, 131
Hermitian forms, 147

matrices, 134, 147
quadratic forms, 134
Principal minor
order of, 1
orthogonal, 103, 163
periodic, 11
permutation, 99
polynomial, 179
positive definite (semi-definite), 134, 147
rank
of,
39
scalar,
10
singular, 39
skew-Hermitian, 13, 118

skew-symmetric, 12, 117
symmetric, 12, 115, 163
triangular, 10, 157
unitary, 112, 164
Matrix Polynoniial(s)
definition of, 179
degree of, 179
product of, 179
proper (improper), 179
scalar, 180
singular (non-singular), 179
sum of, 179
Minimum polynomial, 196
Multiplication
in partitioned form, 4
of matrices, 3
definition of, 134
leading, 135
Product of matrices
adjoint of, 50
conjugate of, 13
determinant of, 33
inverse of, 11
of, 43
rank
transpose
of, 12
Quadratic form
canonical form of, 133, 134
definition of, 131
factorization of, 138
rank of, 131
reduction of
Kronecker, 136
Lagi-ange, 132
regular, 135
Quadratic form, real
definite, 134
index of, 133
semi-definite, 134
signature of, 133
Quadratic forms
equivalence
Negative
definite form (matrix), 134, 147
of a matrix, 2
semi-definite form (matrix), 134, 147
Nilpotent matrix, 11
of, 131, 133,
Rank
of adjoint, 50
of bilinear form, 125
Non-derogatory matrix, 197

Non-singular matrix, 39
of matrix, 39
of product, 43
Normal form of a matrix, 41

Normal matrix, 164
of quadratic form, 131

of sum, 48
134
INDEX
218
Right divisor, 180
Right inverse, 63
Root
of polynomial, 178
of scalar matrix polynomial, 187
Row
(cont.)
deiinition of, 12
invariant vectors of, 163

System(s) of Equations, 75
Trace, 1
equivalent matrices, 40
space of a matrix, 93
transformation, 39
Scalar
matrix, 10
matrix polynomial, 180
multiple of a matrix, 2
polynomial, 172
product of two vectors (see inner product)
Schwarz Inequality, 101, 110
Secular equation (see
characteristic equation)
Signature
of Hermitian matrix, 118
of real quadratic form, 133
of real symmetric matrix, 116
Similar matrices, 95, 196
Similarity invariants, 196
Singular matrix, 39
Skew-Hermitian matrix, 13, 118
Skew-symmetric matrix, 12, 117
Smith normal form, 188
Span, 85
Spectral decomposition, 170
Spur (see Trace)
Sub-matrix, 24
Sum of
matrices, 2
vector spaces, 87
Sylvester's
Symmetric matrix
law
of inertia, 133
of nullity, 88
Symmetric matrix
characteristic roots of, 163
Transformation
elementary, 39
linear, 94
orthogonal, 103
singular, 95
unitary, 112
Transpose
of a matrix, 11
of a product, 12
of a sum, 11
Triangular inequality, 101, 110

Triangular matrix, 10, 157
Unit vector, 101

Unitary
matrix, 112
similarity, 157
transformation, 112
Upper triangular matrix, 10
Vector(s)
belonging to a polynomial, 207

coordinates of, 88
definition of, 67
inner product of, 100
invariant, 149
length of, 100, 110
normalized, 102
orthogonal, 100
vector product of, 109
Vector space
basis of, 86
definition of, 85
dimension of, 86
over the complex field, 110
over the real field, 100
Index of Symbols
Symbol
Page
Page
88
Hj
E^, (vector)
[î;]
X-Y; X Y
100, 110
\\n
100, 110
103, 111
10
zxy
109
11
115
116
116
131
146
^k
A'
A-'-
A'-
Symbol
A':
11
I;
12
A*- A^
13
\A\;
det
20
22
\îj\
A. A,-
149
23
<^(A)
149
a,.
23
E^, (matrix)
170
39
/(A)
172
h-îj
39
F[A]
172
Hi(k), Kîk)
39
^(A)
H^.{k). K.j(k)
39
-V.
40
îi'i2
ii, is, ....
im
i^
AjfiC),
A^ (C)
179
180
NiX)
189
fi(^)
189
49
m(A)
196
64
C(g)
198
X. X^
67
198
yn(f)
85
203
V^tiF)
86
Cq(p)
205
N
adj
Â
43
87
219

MATRICES Teoria y Problemas

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

MATRICES Teoria y Problemas

Transféré par

Droits d'auteur :

Formats disponibles

SCHAUM'S

Formerly Professor and Head,

SCHAIJM'S OUTLINE SERIES

York, St. Louis, San Francisco, Toronto, Sydney

1962 by McGraw-Hill, Inc.

stored in a retrieval system, or transmitted, in any form or by any means,

written permission of the publisher.

78910 SHSH 7543210

without the prior

to enable the reader,

if he persists through the introductory paragraphs and

of the latter require only proper modifications of proofs given earlier;

directed to the supplementary problems of the first two chapters.

The author wishes

to take this opportunity to express his gratitude to the staff of the

for their splendid cooperation.

SOME TYPES OF MATRICES

Scalar matrices. Diagonal matrices. The identity

DETERMINANT OF A SQUARE MATRIX.

Determinants of orders 2 and 3. Properties of determinants,

Expansion along a row or column. The Laplace expansion. Expansion

Non-singular and singular matrices. Elementary

THE ADJOINT OF A SQUARE MATRIX

THE INVERSE OF A MATRIX.

Inverse of a diagonal matrix. Inverse from the adjoint. Inverse from

LINEAR DEPENDENCE OF VECTORS AND FORMS

Vectors. Linear dependence of vectors, linear forms, polynomials, and

System of non-homogeneous equations. Solution using matrices. Cramer's

Bases and coordinates.

VECTORS OVER THE REAL FIELD

VECTORS OVER THE COMPLEX FIELD

Congruent symmetric matrices. Canonical forms

Transformations. Canonical forms. Cogredient transContragredient transformations. Factorable forms.

THE CHARACTERISTIC EQUATION OF A MATRIX

Invariant vectors and spaces.

Reduction to triangular form.

SIMILARITY TO A DIAGONAL MATRIX

Real symmetric matrices. Orthogonal similarity. Pairs of real quadratic

POLYNOMIALS OVER A FIELD

Remainder theorem. Greatest

The X-matrix or matrix polynomial. Sums, products, and quotients.

SMITH NORMAL FORM

THE MINIMUM POLYNOMIAL OF A MATRIX

CANONICAL FORMS UNDER SIMILARITY

Rational canonical form.

A RECTANGULAR ARRAY OF NUMBERS

enclosed by a pair of brackets, such as

and subject to certain rules of operations given below

considered as the coefficient matrix of the system of homogeneous

as the augmented matrix of the system of non-homogeneous linear equations

dinates of the points (1,3,

to obtain solutions of these

rows as simply the coor-

and (4,7, 6) in ordinary space.

The matrix will be used

such questions as whether or not the three points

could be given a similar interpretation

*mi "m2 ""m3

elements in the second row have

the fifth column have 5 as second subscript.

subscript and all the elements in

rows and n columns

and double bars,