Académique Documents
Professionnel Documents
Culture Documents
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.
Department of Mathematics
Dickinson College
^C^O
-^
('opyright
United
States
of America.
No part
AH
Rights Reserved.
Printed in the
may
be reproduced,
of this publication
02656
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
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
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
much
Company
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.
10
Triangular matrices.
Chapter
20
Chapter
Minors
EVALUATION OF DETERMINANTS
32
Chapter
EQUIVALENCE
39
Rank
Chapter
Chapter
of a matrix.
adjoint.
The adjoint
of a product.
Minor of an
49
adjoint.
55
Chapter
FIELDS
Number
64
fields.
General
fields.
Sub-fields.
Matrices over a
field.
CONTENTS
Page
Chapter
67
Chapter
10
LINEAR EQUATIONS
75
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.
Chapter
12
LINEAR TRANSFORMATIONS
Singular and non-singular transformations.
space. Permutation matrix.
Chapter
13
94
Change of
basis.
Invariant
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
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
Chapter
16
BILINEAR FORMS
125
Matrix form.
formations.
Chapter
17
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
Chapter
20
21
SIMILARITY
156
Similar matrices.
Chapter
149
Diagonable matrices.
163
Chapter
22
Field of values.
common
divisor.
Least
of polynomials.
multiple.
common
172
Unique factorization.
Chapter
2o
LAMBDA MATRICES
179
Chapter
24
Chapter
25
Invariant factors.
invariants.
derogatory matrices.
26
Elementary
divisors.
Chapter
188
Minimum
polynomial.
Companion matrix.
Derogatory
196
and
non-
203
INDEX
215
INDEX OF SYMBOLS
219
'
chapter
Matrices
and
-1
l'
(a)
(b)
is
called a matrix.
The matrix
(a) could
be
(2x + 3y + 7z =
linear equations
\ X- y + 5z =
i2x + 3y = 7
-
or
we
Later,
shall see
how
the matrix
may be used
1),
(2, 1,4),
we might consider
The ma-
or
later to settle
origin or
systems.
y = 5
its
trix (b)
\x-
lie in the
In the matrix
^11 ^12
'^IS
'
(1.1)
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
"m by
order
(In
used.
ra"
or
as
first
said to be of
),
||
||,
are
sometimes
is
mxra.
We
matrix of
When
we
n, (1.1) is
[a^
"
or
"the
mxn
matrix A =
4".
re-square matrix.
In a square matrix, the
The sum
elements
of the diagonal
a^,
022.
its
diagonal elements.
MATRICES
[CHAP.
[bij]
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
ZERO MATRIX. A
'V
^^J
and only
if
= 1,2,
if
one
which
ro;
1,
n)
is a
is zero, is
When ^
we
is a zero
mxn
SUMS OF MATRICES.
is
If
defined as the
4 = [a^A and S =
mxn
corresponding elements of
1.
31
It
[fe.^-]
[c^A
mxn
are two
matrices, their
is the
sum
(difference),
sum
A B,
(difference) of the
'\
Example
C =
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
The sum
of k matrices
is
We
define:
If
2.
If
[:
'11
A+A + A
subtraction.
A and each
(a)
Two
and
of its elements is k
Example
or
is
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,
Assuming
(a)
(b)
(c)
(d)
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.
of addition
CHAP.
MATRICES
1]
MULTIPLICATION. By
the product
AB
Ixm
and
a^m]
fell
the
^31
mxl
matrix
is
fi
fesi
+ aimfemi]
fell
fe^i
That
[an
is,
Oi
'^imj
L^ii
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
4]
[7]
2.
-2
(b)
[3
-1
[-6 - 6+ 12]
4]
3J
By
is
the product
meant the
mxn
AB
matrix
C =
\c;;'\
<-
Hj
-'
mxp
xn matrix B =
[bij]
where
p
^^^"ikbkj
"-ip i>,pj
(f
1, 2,
,m-;
1,
re).
Example
4.
^11
*^ 1
."31
'011611 + 012621
O32
[611
612I
621
622J
021611+022^21
."31611+032621
^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
^^^ined)-
See Problems
Assuming
that
A,B,C
(e)
A(B + C) = AB
(/")
(A +
AC
(g)
A(BC) = (AB)C
B)C
AC
BC
3-4.
(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
[b^j]
be of order pxn.
In
For example,
A and 6 be
let
parti-
(mixpi)
(m]^xp2)
(P2X%)
Jmgxpi)
(m,2Xp2)
(p2Xra2)
(m2Xp3)_
(psxni)
Am
(pixnj)
(m-Lxps)
A^2
(p3Xre2)
Aig
"21
O22
A,
A^i
I
"31
In
in exactly the
is
^llOll + '4i2021
^8
"32
""
'4l3"31
A-^-^B-12
+ A-^qBqq + '^13832
Cii
A-2_iiB^^
_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-i^i
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]
9.
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 A^i, A^2< : Sii. 5i2. - ^n, C12
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
r2(l) + 3(2)+4(3)1
'
2J
1 (6)
+ 2(7) + 3(-8)]
poT
L29J
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
[19 4
(rf)
2(4)
=
e]
[4
(b)
D =
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.
-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^^i +
ife(*fei +'^fej)
+ '^t2(*2i+'^2i)
'^lj)
2
(6)
(il^i+''i2*2i)
(il<^17+'^i2'^2i)
2 2
a--
t=i j=i
(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
(2
a{fefefe2)c2f
-^
fe=i
fe=i
ai<^b^^)c3j
**
-^
/.l/li^^^'fe'^^'^'^r
is of order
c'2j
oifefefei)cii
O
37'
fell
k=l
feijj+cij,
the
a., (
b,,c,-)
tfe^^^j^ kh hy
6.
first
row
mxn
of
and
if
are a^^
+%j- Then
S =
[fe^-]
^
a^^,
...
and C =
nxp, then
%i)
C)
+C
are
/}(B +
'
^^^"ikHj+^kj)
AB and ^C.
S^^iifefefej +^2^aifeC;^j,
A(B + C)
is
sum
Of
the
7.
The elements
p
mxn,
S =
and
if
C =
[c.-A is of order
throw
of /4 areai^.a^g, ...,a-
BC
of
h=
b^j^c^j-,
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)
+ (X^aif^bf^2)<:QJ
+ (^^"ikbkp)<^pj
in
+ (A
+B)D
(A
(AB)C.
AC + AD
AC +BC +AD +BD.
+B)C
P
n
^^"ik<-j^,^hh<^hj^
(2^^0ifefefei)cij
J^
n
=
Using
+ tn
6ih c^
i
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
%3
os
yi +
% 2 72
new coordinates
0]^2^2
be a
.J1
(z^, zg).
O21 Zi +
The
h-2
result of applying
(on ii2 +
1) ^1
"*"
(^2
("31^11 + ^32^21) Zl
("3
%i
(o-Li
%<2
((^2 1 C^l 1
X3
6^j +
"*"
aj^2
621) Zi
^2 2 ^2
^12
*i
"*"
'^12^22) ^2
^2 2 ^22) ^2
%1
012
"21
"22
r-
Z2
-.
Vl
Vr,
^3
611
612
O21
022
The
'^3
"11
'xi
%2
"2
"^2
Os
^^3
pri ii2
^2
in n variables
,
622
-1
/,
^',
'
MATRICES
[CHAP.
SUPPLEMENTARY PROBLEMS
-3'
"l
Given A =
-1
.1
1_
-1
-2-4
Compute:
12.
Verify:
(d)
-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.
Given
Show
[
20.
BA
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),
any two
AC. Thus, AB
ij
AB
CA
that
A(B + C)
2AB
-3 -5
= A,
show
-A according
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
17.
1-1-2
-2 -1 -1
2-5-1
2
3
-1
Given A =
and C
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-
[? 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]
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,
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
= [aiji
25. Let
= [a^-] and
/4
[fej,-;,]
where
(i
x n
1,
and
if
m;
1,
p;
A;
show
1, 2, ...
,n).
that
(/4+B)C
Denote by
/S-
=^C
the
+ BC.
sum
of
Pi
182
j)^.
Show
Pi
sum
is the
of the
Problems 12 and
26.
(II)
to
Determinative
Reflexive
(iii)
Symmetric
If
Transitive
If a is in
is called an
Show
relations.
in
(iv)
Show
13.
27.
b.
a.
c.
equivalence relation.
Show
similarity of triangles,
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,
AC
CA and BC
CB,
then
(AB
BA)C
C(AB
BA).
chapter 2
Some Types
of Matrices
"12
"13
tip o
tZo o
a^=
for
upper triangu-
Thus
is
lower triangular.
is
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
'
^nn )
See Problem
If
D above,
Oi^ = Ogs =
k,
and
is
ci
is
denoted by
/.
1.
if.
For example
'1
t"]
When
/+/ +
...
and
1
1
[1.
A.L
hAh
A, as the reader
may
readily show.
10
3I
(5
then l2-A =
CHAP.
2]
A and B
If
it
commutes with
11
is a
It
AB
if
is
any ra-square
See Problem
If
matrix
AB
that
-BA,
i =
matrix
1,
A and 6
,fe+i
for
which A
where
A,
A:
If
k is
See Problems
3-4.
k-t-i
th^ matrices
2.
A,
k.
for
aP
4=0,
which
where p
0,
is
ther
is said to
If
p is the
be nilpotent of index
See Problems
5-6.
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'
/,
( 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
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.
tijvo
See Problem
A
utory.
matrix
An
A such
An
See Problem
of
an mxn matrix A
is
matrix of order
In
^[-e]
in the /th
IS
i' =
Problems 10 and
II.
11,
The transpose
(A'Y =
we
4'
6.
aIr
in the ith
A and
and
B,
and
if
(b)
(kAy = kA'
A:
is
prove:
of the
row
9.
8.
sum
(A +
BY
A'+ S'
sum
12
[CHAP. 2
and
in.
The transpose
transposes,
of the product of
two matrices
is the
i.e.,
(AB)'
B'-A'
SYMMETRIC MATRICES. A
A =
matrix
is
a^j]
/.
Thus, a square
For example,
4-5
3-5 6
Problem
IV.
is
In
k.
we prove
13,
If
any scalar
for
is
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"
is
is
kA
for
any scalar
k.
-3 -4
With only minor changes in Problem 13, we can prove
V.
If
is
A' is
skew-symmetric.
If
= a + bi
and
Z2
z^
a~
Z2 =
then
bi,
z-^
a-
bi
= a+
bi,
z.i=a+bi and
If
(i)
z-L
Z2
= c+
+ Zg = (a+c) + (b+d)i
sum
(ii)
z^- Zg
then
di,
and
of
= (ac-bd) + (ad+bc)i
z^
= (a+c) - (b+d)i =
z^
z^^
and
is the
sum
(a-bi) + (c-di)
of their conjugates.
numbers
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
Using
Example 2. When A
= A
we may prove
and
3i
is
(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.
same
i.e.,
order, of
The transpose
We have
IX.
A, i.e.,
Example
3.
of
is
T^e 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
is Hermitian.
-I
if
+
i
is
kA Hermitian
sometimes written as
A'
3i
HERMITIAN MATRICES. A
Hermitian provided
while
-i
Is
is
is
It
^ =
such that
[0^^]
a^^ = -o^--
for all
1= -A
values of
is
called skew-Hermitian.
and
Thus, ^
is
skew-
l-i
-1-t
3i
-2
is
/.
skew-Hermitian.
Is
kA skew-Hermitian
it
k is any real
By making
X.
If
is
we may
Z is
prove
Hermitian and
A-T is
skew-Hermitian.
DIRECT SUM.
Let A^, A^
ization
^1
...
A^
...
diag(^i,^2
...
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
A^,A^,Aq
of
diag(^3^,
is
/42, /4g)
12-1
2
1-2
illustrates
1,
the
...,Ag)
s),
S = diag(Si, S2
and
AB
then
^s).
= diag(^iSj^, A^B^
^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
2.
Show
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.
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
Prove:
for n
=
I-A^
I-I
= 0.
B'.
A = [ay] and B
Let
Now
/.
7,
matrix
A'.
b\a''-A)B
AB(B~'^-A~'-)
Prove:
B~^-A'^
{ABf{AB)
definition
By Problem
10.
-1 -1 -3
(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'.
[6y] be
of order
AB
is
cy
nxp
=
then
AB
[cy]
is of order
mxp.
The
"i2
in the /th
iy, b^j
Thus,
12.
Prove:
(AB)' =
bnj
B'/l'is
B'A'.
(ABC)' = CB'A'.
mite
ABC
(AB)C.
Then by Problem
,
1 1
C'B'A'.
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
The element
in the /th
a^^ + aij.
Thus,
is aij
.4
and B
bij = bji
is
a^i
/I'
hence,
is aji;
.4'
is
symmetric.
Second Proof.
By Problem
14. Prove:
If
10,
A and B
(A+Ay
A' +
(A')'
A'+A
A + A' and
(^ +.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
B'A' =
BA
(AB)' = B'A' =
AB and AB
BA
AB
hence,
15. Prove:
If
is
If
.4
is
If
,4
is
-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)
+ k^I
BA - k(A+B)
+ k^I
(B-kr)(A-kl)
(B
and B kl commute.
(A-kI)(B-kI)
is
mxn then B =
symmetric.
and B is skew-symmetric.
is of order
k.
{A-kI)(B-kI)
Thus,
every scalar
AB
symmetric.
is
A and B commute.
trices
P'AP
AB
AB
AB - k(A+B)
+ k^I
BA - k(A+B)
+ k^I
-kr\(A-kI)
if
A~
kl and
B-
kl
CHAP.
21
17
SUPPLEMENTARY PROBLEMS
17.
Show
BA
19.
See Problem
Show
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.
Bf
all matrices
(a) diag(a,fe, c)
A^ +
B'^
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
-3 -4
-13
that
is
-2 -6
-3
Show
I-A
(b)
diag(aii,a22
_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
-14
If
S and ^
matrix
row order of A
2-3-5
A
that
1.
of / is the
re-square,
Show
21. (a)
22.
mxn
of an
Hint.
is
are arbitrary.
a^^).
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
[3:][:^E3
1.
^"'-
4]-
[3
Show
10
0-2
-2311
33. Set
-1
-1/2]
[3/2
A and
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'^B-^A-'^.
A,
(a)(A)=A,
41. Prove:
a positive integer.
ABC=iAB)C.
Hint. Write
(c) (A'^Y^ =
(a)
2- 3
-1 +
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:
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)
are symmetric,
(6) A-irA',
44. Prove:
If
45. Prove:
is
HA
is Hermitian.
is real
skew-symmetric.
46. Prove:
and C
Show
47. Prove:
48.
(a)
If
that for
m and
n positive integers, ^4
A' and
B\
if
A and 5 commute.
CHAP.
2]
19
A
A
49.
Show
raA
(6)
If
51. Prove: If
is
symmetric
is
nX
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
integer.
52. Prove:
53. Prove:
If
54.
is real
if
is
is
If^
A and B
If
idempotent,
(c)
show
is involutory.
Y=
(A
(a)
(A')
(A)^ = A-\
(b)
BA
kO-A)
57. If the
58.
and skew-symmetric or
B +C where B
/I
Show
55. Prove:
56.
written as
bfe
all
I.
and
j(I+A)
A'B"= B\
^(I-A) =
(b) A"
0.
show:
Find
then
(r)-^=(^-V
(c)
AA~^
Hint, (a)
of
obtain
as the inverse of
(A'''-)
(6)
i, 2, 3),
diag(l.
/!'
1, 2, 2).
(b) dia.g(A.B)
b. c
are scalars.
A^.A^
59. If
with
A^
diag(^i, ^2
Ans. dia.g(B^.B^
AB
60. If
= 0,
mj^.m^
m^.
which commute
'^s)
-85
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
61. If
...,
s),
diae(Ai.A2
show
A^)
(a)
A+B
(b)
AB
diag(^iBi. /I2S2
(c)
trace
AB
and
di&giB^.B^
B^)
diag(^i+Si,^2 + S2
=
62. Prove: If
^ and B
63. Prove: If
-^s
+ trace A^B^.
...
is n-square
and
AB
is
symmetric
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 = r^A+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
(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)
4!
132
312
231
213
321
4123
4213
3124
3214
2134
2314
1234
1324
(3.2)
it
2,
precedes
is
even since in
it
4 precedes
'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
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.
le Jlj2-
p=n\
The determinant
'^h
20
re
is called
a determinant of order
re.
CHAP.
3]
"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) + ^IS (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
2(-l)
- 3(-2) +
5(1)
-3 -4
0-2
O-o!
-5 -6
-20
18
20
-18
See Problem
PROPERTIES OF DETERMINANTS.
nant
Ul
is
is the
1.
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
a scalar k.
fth
\B\
1
P
Thus,
\: ,
,
a. , n^n
\
*%'2---i'iii2J2----anj^!
.
Jc\A\
22
[CHAP.
III.
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\.
its
IV.
If
In
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
its ith
its
= -\A\.
\b\
= (-i)^UI.
VII.
If
IV,
= 1,2,
...,n).
row of A
of the first
is
\A\
expressed as a binomial
%,
^li
b-^:
c-^j,
Then
in
^Jd2
^^^ii ^
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
If
is obtained from
is
A by adding
to the
elements of
its ith
Cfrii
T"/CCEog
^22
Clnr
Clf,',
{-
Karin.
Clf=n
Clr^r,
Cfg ^
\b\
H'Qri^
^Q'2
"^
i^Clr
Zgg +
ka^^
(3.6).
When from A
the elements of its ith row and /th column are removed, the determinant
or of \a\
and denoted by
\M^j\.
CHAP.
3]
More frequently,
cofactor of
Example
2.
a^-
If
The signed
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
"^if^if +
a^g a^g
^in * 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
i^,^,,
i^
and /+;+2
.Jl.j2.-
(3.11)
im
'l'J'2
'^i.
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
The determinant
Jm
J1.J2.
J'm-'-i'Jn^-Q
Jn
^m+l'
''r
pair of minors
and
are called
complementary minors
I
i
1,3
'^2,5
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
Let
U +
(3.13)
In
+ in +
and
(3.14)
i-n
Ji, is.
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
Jm
H>h
^n
13
2+-5 + 14-31
complement
l>^m+2> ''-
+ /m+2 + " + In
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
^ijg',4
is
is
1.3
^^is
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
OL-Lj.
is
called a principal
minor of A.
of A; the alge-
CHAP.
3]
25
H'
"?2
"11
"15
31
S3
and
^1.3
A,
,t
"24
(^25
0^2
044
045
52
"54
055
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
-4
-4
-4
-4
-4
-4
-4
-4
-4
of the other
columns,
I.
4.
1(3-3
by Theorem
3.
to the third,
Vn
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^ +
02
Ul
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
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,
are factors.
Now M|
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 -ka^a^. 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
\-A
(-if^-^-UI
Ul; hence,
is
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---Jn^iii"2i?--^"in
12
2,
U|
^iij2...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.
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]
elements
in
9.
27
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.
kh---Jn~
^jbjb- -in
The terms
^i"''^ i"
^ 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
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
U| along
its first
"' ''^ '^ ^^^^^ '^' ^^-^^ '' ' = '^ '^ '''^i"^'J by f^P^^ting the above arguJ'*' ^iT^
matrix
obtained from A by moving its rth row over the first r-1
rows and then its .th col^
T~l
(-1)
first
s-l
(-!)
first
for
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^j^a.
rk^rk
is yreciseiy
i
precisely
10.
[CHAP.
28
When
A =
oLij
'll
(i)
fC^GC^-j
"T
fCi^OL^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
'^i
^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
(i) is
when
+ fea^g,
Ay
\7
by
= 1,2
(r
and
= 1,2
and
By Theorems Vin,
j).
/).
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
-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
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)
3-4
5-6 3
2-3
4
-3
first
3-3(1) 4-3(2)
(-4 + 36)
-42
-3(14)
-2
27
-41
-31
-11
CHAP.
(e)
3]
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.
(-1)
14
14
4-13
12.
29
is
+ 2+-+7J) +
(1
(l
2+-"+n)
2-kn{n + l)
n(n +1)
p+q
= (-1)^
is
[..]
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
the algebraic
or both odd.
16
18
20
21
23
25
is
I12 141
SUPPLEMENTARY PROBLEMS
14.
Show
1, 2, 3. 4,
is even,
24135
is odd,
41532
is even,
15.
16.
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
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
Obtain from \A
2 3 4
4 5 8
4 5 3
.4
by interchanging
|o| =
the determinant
2 3
its first
Evaluate
to verify
Theorem
/I
I
\A
I
20.
If -4 is
21.
Prove: (a)
If
(b) If
22.
Compare with
then
= k.
\a\ = k =
(a)
(b)
Same,
for
Theorem
it
is a scalar,
is skew-Hermitian, then
Do
(e).
Evaluate
(g).
U^
/t^M
|.
\a'\.
|
-4
of adjacent
number.
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
5.
second column by
its
2 3
2 3 5
(e)
-4
III.
(d)
Theorem
Theorem V.
to verify
25.
Use Theorems
26.
VIII, III,
,4
o.
in
Problem
11.
a b
27.
Use
(3.6) to evaluate
\A\
d
e
".P/.
Thus,
if
= diag(A-^. A^).
where
g h
A^,
U4i|
2/3
-2/3 -2/3
1/3 -2/3
2/3
-2/3
-1/3
28.
Show
1/3
-4
29.
Show
is that element.
-3 -3
is the
numbered column.
30.
If
is
symmetric then
(b) If
is
Prove: (a)
<Xij=
0^j^
when
aij = (-1)
"P when
Ot--^
CHAP.
31.
3]
show
(b)
that
32.
12
(^^22
32
^23
Otgg
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
a\
n-2
ra-i
n-2
a^
1 ... 1
0...1
...0
1...1
("-!)
re-i
known.
(a) is
respectively by a,b.c
34.
AC = [.
be
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
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
+6
(02
ra-2
36.
S(i
has
as a root.
7i-i
- "ni
chapter 4
Evaluation of Determinants
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
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
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.
EVALUATION OF DETERMINANTS
4]
The expansion
of a determinant
33
bered
I'l,
i^
J2
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.
^n
+h
3.
Evaluate
3-2 4
2
3-2 12
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
.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)
Ul-lsl
See Problem 7
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
"ii^iil
of^.
ii Otj
I
Then
A =
[a^,-]
^tj.
EVALUATION OF DETERMINANTS
34
The
[CHAP. 4
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
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 +
: i
(-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.
EVALUATION OF DETERMINANTS
4]
\A\
Now
row
(I'l- 1)
numbered
of 1^1 in
%\
in
by
after
by
m<n
Jrn
A;
H-''^
order of magnitude.
\aij\
ji'h
35
+ (ig-
2)
i^
I'l,
+ (ijj -m)
11 + ^2+
occupy the position of the
first
+ ^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!(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
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
S(-i/
I
,
5.
where
/i, 72
i^
+ i^+
12 3 4
12 1
11
3 4 12
(-1)^
Jn
different selections
21
|1
ll
(-2)(1)
(-ir
21
ll
41
ll
(-If
11
13
41
ll
(5)(-l)
C B
Prom the
7.
+ 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
EVALUATION OF DETERMINANTS
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
On
012
"m
Cii
"21
"22
2ra
C21
"ni
"n2
"nn
Cm
-1
Next, to the
(n
fei2
times the
Let A =
h,
622
..
62
Hence,
\p\
in
Cii
C12
2n
C21
C22
"nn
Cm
Cn2
Its
(-i)Vlf" ^"^"lc|
^13
^m
*23
b^n
\P\
8.
first
612
We have
= (-lf'2"^^'|c|.
-1
Oil
6i times
we have
Prom
and
\c\
\c\
\ab\ = U1.|b|.
%2 %S
where
Oif
aij(x),
(i, j
= 1,2,3),
%l"22*33 +
and, denoting -ra;:
dx
^^
by
a,',-,
^J
122Sa31
%s"S2"21
"
(^ix"^"^ "
0:l202lOs3
"
(hs^^O'Sl
Then
CHAP.
EVALUATION OF DETERMINANTS
4]
'^ll"22''S3
dx
+ 22ll"33 + ''33"ll"22
+ %3032021
+ O32OISO2I
a^OQiOg^
Oil
0^11
o-L20^i2
OisOi^g + a^-^dj^x
"*"
031
032 033
023
021
Oj^jCggOgg
11
O12 Ol3
21
O22
O23
31
O32
O33
Og^a^a^^
022(^22
>i23'^n''S2
0210-12033
+ O31O12O23
+ "l2"23"31 + 023012131
021013032
37
air^a^^a^.^
+ 023(^23
"^
Og-LCXgi
OgjOj^jOgg
02205^303^
Ogj^o^gOjg
+ 032(^32 + 033(^33
"13
'11
"12
i2i
"22 "23
'31
"32 "33
3.
SUPPLEMENTARY PROBLEMS
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 \A^A
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
'^
0000
bjj
that
and
118
-4 -3 -2 -5
-2
2
2 -2
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
13.
Evaluate
first three
rows.
Ans. -720
EVALUATION OF DETERMINANTS
38
112 12
14.
12
^1,^2
15. If
111
110
112
Evaluate
first
two columns.
.4ns. 2
11
^"^^
^^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
60
first
to
show
62
6..
A
B C
\A\
umn
to
is fe-square, is
zero when
+ ai4i4.
.-^
.^
ffi,-
is the
"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
x^
(b)
(a)
(4 .3).
()
-4ns.
show
[a^,-],
In
2x
"ij
li
Hint.
first col-
11 "IJ l^lj
^il^lj^lj
Of
"ii
19. If a^j
along its
ti
1=2 J=2
where
OL^a. 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;^
15a:*,
(c) 6a:^
A non-singular and
5*"^
if
ff
x^
- 28x^ +
=
4 + iS
9a:^
20a;
is Hermitian, then
chapter 5
Equivalence
Example
1.
if
r if
any, is zero.
zero matrix is
0.
The rank
of
'l
3"
is
r=
-1^0
since
while
= 0.
See Problem
An
A
re-square matrix
called singular.
is
Prom
I.
AB\
that is,
if
A ^
0.
1.
Otherwise,
A\-\B\ follows
The product
of
two
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)
The interchange
The interchange
of the ith
The multiplication
The
(3)
by Hij;
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
to the
k,
The addition
k,
the transformations
are
In
Problem
It
2, it is
is clear that
rank.
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
H2i(-2)
is to
produce B
3"
10
.7
8 9
The
"ij
are
^ij
(2')
Hi\k) = H^(i/k)
(3')
H--(k)
= H^A-k)
K^jik)
= Kij(-k)
We have
II.
same
The inverse
type.
if
Equivalent matrices have the same order and the same rank.
Examples. Applying
-1
12-14
-1
5-3
-1
-2
-7
-7
-3
-3
-1 -2
I
I
t^
0,
12-1
5-3
'^
the rank of
is 2
hence,
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
Any non-zero
(a)
matrix
of rank
r is
first r
all other
rows have
this
(C) ]\
<
is 1;
let the
column
in
which
;'-
/2
<
<
j^.
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)
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^).
wise, suppose
use ^2(1/^2^2) as in
in
(ii);
if
column
first
but
&2J2=
bqj^
B =
first row,
Other-
C.
in
0,
(il),
If
number of the
72 is the
Example
4.
-2
-7
first
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
>
can be reduced
to
(5.1)
/.
\l%
"'"'
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
ff2i(-2),
^31(1).
column.
Ksi(l),
^2i(-2),
first
of the
row and
second
K23, K^{\/%),
X4i(-4).
See Problem
ELEMENTARY MATRICES.
4.
5.
Example
5.
Examples
/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
The
To
A by an elementary
To
mxn
matrix
A can be produced by
matrix.
Ijn
(Why?)
is non-singular.
multiplying
to
[CHAP.
of order
/ to
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
39J
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
Example
7.
When
2
_1
-1
["100"! r
1
and only
if
2J
1-200
o"j
0-2
ij
[j^i
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
is
equivalent to
its
P and Q
See Problem
6.
CHAP.
EQUIVALENCE
5]
43
H^...H^-H^
and
Let
and
K-^-K^-.-Kt
P~^
(5.4)
H;\hI\..H^^
Q''
and
let
Kt...Kt-Kt-
that
PAQ
Then
A
(5.5)
P'\PAQ)Q^
P'^-k-Q^
P'^-Q'^
We have proved
V.
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
AB
is that of B.
is that of A.
we prove
if
and only
if
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
P and Q such
NQ
Let S be a
AB
pxre matrix
(5.6)
By Theorem
By Theorem
non-singular
that
PAQ
Then 4 = P
r.
9.
P''
P~'NQ'^B
Q B and
IX.
The rank
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
:
is
of rank r
and
if
is
such that
AB
the
EQUIVALENCE
44
[CHAP.
SOLVED PROBLEMS
1.
(a)
The rank
-4
The rank
12
of
is 2
since
-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
0,
is 0, but nov
6 9_
_0
every element
3"
of
^0.
is
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.
case
(ii),
= -\r\ =
\S\
case
in the
In the
case
(ii) to
(i),
\S\
interchange
= \r\
=0;
\s\ is,
(iii),
unchanged,
it
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-
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.
tor,
if it
On
it.
it,
rank of A.
"1
(a)
3"
1
'-^-/
"1
-3 -3
_3
1_
-4 -8
_0
2_
(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
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
-2
-2
-3
-1
Reduce each of
(a)
12
6_
GOOD
-2_
3"
-1
'^
10
4
13-2
'->-'
-2 3
l"
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_
-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]
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'
is
and
(^^
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
10
1
-6 -5 7 -2
-6 -5 7 -3
0-1/6 -5/6
1
1
Pi
A^.
performed on a row of
10
P^
EQUIVALENCE
46
[CHAP.
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_
H^-H^-A-K^-K^
//2i(-l).
=
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.
[!:][::]
[i:l-[:i:]
A
canonical set
tor
non-zero
Vh o]
10. If from a
3x4
[:=:]
square matrix
nm
matrices is
is
r^,
a submatrix
>
- n as required.
r^
&:]
B consisting
+ s -
of s rows (columns) of
n.
CHAP.
EQUIVALENCE
5]
47
SUPPLEMENTARY PROBLEMS
2
4 5
11.
(a)
2
(c)
(c) 4.
(b) 3,
13.
Show
14.
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
(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
Ans.
10
2j
-1
1/9"
1
'X^
14.
P'
jl
form Z/^^. /^O). //i3(-4) and check that each HA effects the corresponding row transformation.
form K^^. Ks(-l). K^^O) and show that each AK effects the corresponding column transformation.
H^
{?,),
K^lo)
Compute B
^12
ft,(3) -//isC-
4)
(a).
"0
(e)
(d)
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
"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
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
18.
Prove:
(a)
(b)
19.
UP
AB
AB
and
and
BA
BA
are non-singular if
are singular
if at
.4
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
P^
and
Qr,
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
of rank 2.
23.
The matrix A
24. If the
25.
mxn
AB
show
Show
rj
and
rg
respectively,
show
that
\AB\ = |^|
(a) If at least
By considering each
of the
|fi|
\AB\ =
|/4|-|s|
(6) If
both are
29. Prove:
Hint.
= 0.
A_
28. Prove:
30.
27.
such that
7^
are arbitrary.
26.
n+l.
abed
_e
and take
Q-'b
where a.b
mxn
Hint.
[CHAP.
of a non-singular matrix
to
is I
and conversely.
Show how
to effect
31. Prove: If
.4
(3).
is an
mxn
m,
State the
chapter 6
The Adjoint of a Square Matrix
THE ADJOINT.
Let A =
[a^ ]
adjoint
(6.1)
adj
oij-
12
^22
'^ire
'^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.
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),
/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
-7 and
U|
(adj A)
f-T
4
I
49
.
I
-7/
-7
[_
adj
0'
0-7
we have
(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
^ii.i2
^l'''-2
Let
7m + i' 7m+2'
let
we
Jn
%.
^2
adj
adj
be
Jn
its
complement
re-square matrix
in A,
A =
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
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
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
rank
is of
1.
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
jslc^adj/^) =
adjS-U|./.B
U|{(adjB)Si
adj
adj
|b|.U|./
=
Ab\
\AB\-I
/I
0.
?^
(6.4),
adj
adj (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'^-^
Jm+i. Jw4-2
in
^m+i. ^m+2
^n
let
nH.i2
UT'^-^
in
;
'"n
^^^
in\
A =
[a:-]
tj-
/I
is
^H,i^, ...,i
occupy in A.
Then
52
u
where
s =
Ji
+ f2+ +
Ji.
J/2'
H'
^2-
+/1 + 72+
Jm
>
'
(-ifur
^m
[CHAP.
^i
'
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.Jm^i.
\a\
Ul
''i2Jn
" "^-Jn
"^M'Jn^i
"V+i'in
"%-i-i'Jn-^i
^ri'^m+i
"in.in
(-ifUI- M %
where
7.
s is
Prove:
If
is a
H\
Jn
Jm+1' Jm-t2
4.
^2
Prom
skew-symmetric of order
^m+l' ''m+2
''w
2ra,
immediately.
of i.
By
its definition.
The theorem
2.
By
we
is true for n = l
Assume now
order2A: +
are to
show
is a perfect square.
since,
when A = r
\
partitioning, write
oi
,1
\A\ = a
\,
2
.
\-a Oj
when
P^l
\
n = k
E==
where
'
r
\
-zri^i.-zii^-zy
'^2k^..2n^\
Then S
is
skew-sym-
"
CHAP.
6]
\b\ =
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/
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.
SUPPLEMENTARY PROBLEMS
8.
"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
11. If
is a
/I
7^
adj^ =0.
-1 -2 -2
12.
-4
-2
Show
13. Prove:
If
-2
2-2
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.
so also
is adj/1.
skew-symmetric of order
, then
adj^
is
54
17. Is there a
18.
[CHAP. 6
skew-Hermitian matrices?
for
-Hij
-1
H^ (k)
diag(l/A:. l/k
Hij(k),
HJ
(6) adj
19. Prove: If
1,
l/k
-1
Use
n-l
and
H^...H^
if
-H-^
-A K-^-K,2---^t
where \
is
then
adj
20.
l/k),
is
or
1/k,
adj
Xi
-1
adj
-1
K^
K^
adj
-1
adj
adj H^
-1
adj
H^
-1
adj H^
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
22. Prove: If
-2
(b)
-7
21. Let
-2
and
S(adjB)
adj (adj
^)
If
\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)^ ^ai,
Hj
and verify
for n
adJ-4
(i)
24.
where
=2,3,4
o!^-
its ith
^ij
^pj
iq
pq
that
[6.^,-]
tjj
Show
that
chapter 7
The Inverse of a Matrix
IF A
AND B
A
is
Problem
I.
An
II.
THE INVERSE
If
ra-square matrix
BA
A has an Inverse
if it is
then
AB
AC
:^i' -^2
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.
A)
diagonal matrix
is the
1A)
...,
diag(Al^.
A^
sum diag(/}
A~s)
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,
"7/2
Since
Ml
= -2,
A~^ =
4 3
-1
-1
-2
-3
^'^^ -^
-k
2
-7
is
^J
See Problem
55
2.
56
[CHAP.
H^.-.H^.H^. A-K^-K2...Kt
Then A ^ P
Q^
Example
2.
From Problem
Q-P
(P'^-Q'Y
K^K2H2'i
o"
-1
In Chapter 5
-3 -3
1
-10
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,
A'^
(7.2)
PAQ
is,
III.
If
is
reduced to
equal
is
of
.4
4 3
13
Write the matrix [A
Ir,
3 3
4 3
3 4
of
Example
/I
to/.
[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.
its
inverse
B =
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'^ii + B^z'^si
(iv)
Bjiiia + Bss^^as
(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 '^u)
- A^i^Ali A^^).
See Problem
To
used.
4.
Let
[Oil
%2l
2i
^^aj
"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
it
Take ^11
11
^=
[!'].
^22
4 3
3 4
^12
'42i(^;' ^12)
using partitioning.
^421 =
[13],
[4]-[l3]M
and
[1],
Now
f^
[l]
Then
Sii
4\
S12
-('4i\^i2)r^ "
('4i\/ii2)rw^ii)
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
[CHAP. 7
symmetric, aij =
is
ciji
...
b
.
Ht
...
...
^l
b
c
^i(- ^)
^^i(- |)
H^(l/\Ja)
is not
In general,
replaced by
is
1,
procedure
recommended.
(7.6)
where
A^^ - A.^^(Af-^A^^).
See Problem
When A
is
is
which
found by
is
A'^
(7.7)
may be used
7.
(A'Ai^A-
SOLVED PROBLEMS
1.
Prove:
An
re-square matrix
Suppose A
that
is
non-singular.
By Theorem
Supposed
A
A has an inverse
A''^
if
IV,
Q-P
and only
Chapter
5,
if it is
non- singular.
-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
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
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
[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'l^A^^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)
S^i^ii + B^^A^^ =
(iV)
^21^12 + 52,42
for
B^^ = -(-4^1-412)^"^;
from
(iii),
Bii,Si2,S2i, and S
= /
B21 =
- ^~\A^^At^y,
and, from
+ (^lVi2)f~\'42iA\).
-f
(-42i'4ii)^i2 + ^'^422
*.
and
'^'2
('^21'*lll)'^12
(i),
B-^^
60
5.
12
13
[CHAP.
by partitioning.
1111
(a)
Take Gg
[24],
-1
Now
r 3 -2l
jj.
^_^
11
,f
Then
^22 -
Sii = 4l\ +
,-1
,
^11-41. -
'42i('4ii^i2)
[3
-2
|^_^
(.4lVi2)rV2i^i\)
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 ]
A^iA-^-^
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.
7]
We cannot
take
4 3
-4,
by partitioning.
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
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(Ai^i2) =
Then
[l]-[-l
2]rM
[-2]
and
f^=
[-i]
B,
B12
["jj.
B21
[-k
10
-5
-1
-5
-5
and
B22 = [-i]
k].
'^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
SUPPLEMENTARY PROBLEMS
8.
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.
8((?)
-3
-3
-1
-2/3
1/3
(d)
1/2 -1/6
2-1
-J
(c)
-4 -14
15
^'^i
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.
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
CHAP.
14.
7]
63
12-12
12
2-11
112
-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
A^A
Hint:
18.
Show
A und B
(b)
20.
(AA"^)' =
AC
is said to
and only
Find a
if
The rank
is the
(A~^)'A
is
symmetric, so also is
/I
(BA)A
^.
(a) A~'^B.
(b) AB"'^.
and
(c) A~^B~^
13 2 3
14 13
13 5 4
if
one exists.
4x3 matrix
= C.
a.nd
17
if
right inverse of
Hint: (a)
a right inverse
implies
An mxn matrix A
Hint.
(c)
are symmetric.
19.
AB
then
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
right inverse of
-9 -5
-4
-1
L J
Show
3 3
21,
4 3
3 4
-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
/111
A21
+
then
.4)-^
and
(I
6.
(/
^12
Aq2
|.4ii|
- A) commute.
left inverse.
-1^22 - ^^si^ii/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
are:
numbers,
numbers,
complex numbers a+
bi,
set of all integers and the set of all numbers of the form
The
bvs
where
i is
a rational
GENERAL FIELDS. A
collection or set S of two or more elements, together with two operations called
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
As
ab =
a-b
is a
F
F
in
F such
unique element of
-a
in
that
a+
F such
+a =
that
a.
a + (-a) =
0.
ab = ba
(ab)c = a(bc)
Di
in
?^
such that
1-a =
F such
that
o.
a-
a'''^
-0=1.
a(b +
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
2x2
^^^
Q{x)
of polynomials in
ti
where a and
b are real
numbers.
0+0=0.
which
but
is
is not
necessarily
0.
64
CHAP.
FIELDS
8]
SUBFIELDS.
If
65
if
a subset of T.
S and T are fields and
If
if
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
field of all
complex
numbers.
A =
1/4
1/2
2/3
all of the
and
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
if all
F and
...
An examination
which
is the
The sum,
If
If
A ^l then
If
over F.
and
rank
is of
r,
its
that
PAQ
and
is over F.
it
will be
assumed
that
is the
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.
To do
this
we simply check
product
(A/ 1)
is
1.
(a+bi){c +
If
the properties
a + bi
di) = (ac
and
di
a + bi
a^
bi
+62
a^+
b^
hi
a^ + b^
bi^o
is
bi
(/I4) is
is
-a-bi.
and
the
66
FIELDS
[CHAP.
SUPPLEMENTARY PROBLEMS
2.
Verify (a) the set of all real numbers of the form a + b\r5
(6) the set of all quotients
numbers and
4.
b are ra:ional
Q(x)
3.
where a and
Show
where a and
that this is a subfield of the field of all 2x2 matrices of the form
5.
6.
set
a ring.
ring.
2x2
b are rational
numbers, forms a
w]\eK a and
h are real
field.
field.
numbers.
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
...
0,+l,2,+3,
is an
set of all
2x2
...
is an
example of a non-commutative
where a and
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,
element a field?
9.
10.
=
.
show
all
that
Is every
numbers.
2x2
LA
all
To
matrices
= A.
Call
L a
ol
^
where a and
bi
b are in
F.
If
is
A.
be the field of
all
2x2
-b^
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
(c)
Show
that the
(d)
What
is the
(e)
3+ ^^2i,
image
5.
of
"!]
[:
This
is
is the identity
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,
OX
''^3(*11
^21..
^12 +
^2!
X.2(X^1. X22)
X(Xi. Xq)
Fig. 9-1
If
.Yi
= [%i,xi2]
their
X^=
and
[x^^.x^q]
9-2) yields
A3
A^ + A2
^11
L rC%'\
VECTORS. By
-1
"^
-^12
"^
-'^221
is
-^21
fv
X-y
for
over
is
of n
elements x^
of F, as
(9.1)
The elements
x^,X2
%
it
and
[a;i,x2, ...,%]
...,
reth
components of X.
(91)
Now
I X-i
X2,
Xi
as a column vector.
components of a
67
17-vector) or
68
The
vector, all of
of
two row (column) vectors and the product of a scalar and a vec-
1.
(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^
[-4,-8,7]
that if
is
^mJ
|.%]^,%2'
each bracket
elements
h-^^.k^,
,k^
of F, not
such that
+
k.^X-1
(9.3)
Otherwise, the
k^X^^
k^X-^
Example
four vectors of
Then
3fti + 2k^ =
and then ^2 = 0.
n-vector
+
0,
Xq^.
(c)
[Zk^ + 2k^,
+ 2^2 =
ft^
By
^2 ^"d
Xj^_
{d).
;c2A'2
1.
and
0,
k^+
-ik^ -
= 0.
Prom
[o,
0,0]
Xfn+i
BASIC THEOREMS.
If
in (9.3),
^i
(9.4)
-I5]
[o,0.o] =
re-vectors over
Ai
Any
10,
all zero,
-S] - [lO,
[6. 2,
^4= [-4,-4,6]
and
Example
if
[CHAP.
=
Xi
k^
?^
0,
F such
of
%A]^ +
we may solve
- T"!^!'^! +
=
SiJVi
X^
that
:2A2
k^Xf^
for
+ ^i-i^t-i + -^i+i^i+i +
+ s^_iZi_i + s^j-i^i+i +
+ ^m^m!
or
+ s^J^
Thus,
I. If m vectors are linearly dependent, some one of them
a linear combination of the others.
CHAP.
9]
II.
If
vectors X^, X^
is
X^j,-^
, X^
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=2A^i-3^2-
2,
Xg=
III.
69
X^ there
X^, X^
is
0.
Clearly,
a subset of
r<m
are
linearly dependent, the vectors of the entire set are linearly dependent.
Example
4.
By
(b) of
Example
the vectors
1,
(d),
See Problem
1.
%2
Xi 1
m<n
(9.5)
^n2
*TOl
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
r<m.
If
linearly independent.
The
If
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^
a^x^
a^A:^
i=i
in
F
linear forms in n variables
/l
CIu% + %2^2 +
(9.7)
1^1
/m
0~foXo
'^2n*re
%n^n
-'ml
If
Kk
k^
+
^^^2/2
"m2
...
A;4
F such
=
that
70
[CHAP. 9
independent.
to the linear
Example
5.
dependence
The forms
/i =
or
2xi
- 2 +
3*g,
or
2%
x.^+
/2 =
ix^
/g =
4^=3.
Tx^ + Xg
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
k^X-^ +
k^X^
k^X^ +
+ k^X^ +
X^,
r<m,
which is
vectors.
+ k^X^ =
+ 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
ra-vectors is
vectors which are linearly independent while each of the remaining m-r vectors can be expressed
vectors.
is
into this position and then renumber all rows and columns in natural order.
11
11?
21
25-'
Thus, we have
*11
12
%-r Xiq
%1 %2
*rl X^2
xp^
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.
x^q. xpq,
Let h^,k^,
Then,
CHAP.
9]
Now
+ krXri + ^r+i*^i
k^x^q + k^xQq +
and by hypothesis
u.
k2X2i +
fci^ii
= 1,2
(i
+ krx^q + krA-ixpq
71
r)
column of
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
0,
ji^
Xp
is
a.
X^
k^X^ +
^j^,
(t
f'r-n'^pt
= 1,2
n)
t,
k^X^ + k^X^ +
Since
f'r'^rt
^I
^r+2
-V^+i,
k^^-^Xp
X-^^.
X^
X^.
But
^^j
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.
3.
X-^.X^.
...,
X^
as
was
X^
Xr+^
to be proved.
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]
(a)
Here,
12-34
3-121
1-5
is of
rank
2;
-1
j^
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
1X2-1X2=
and Xg
-2X^ + X^.
The cofactors
column of
-1
-2
are
2-113
2-213
4-326
and
Xg
for
Xi +
X,
Now
-1
-2
which
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
member
1111
1110
12
12
We have
The
tt.
Any
verified:
we have
4
[P^.,
Px.Pq
]'
is of rank 2.
three points of ordinary space lie in a plane through the origin provided the matrix
Show
(i).
Thus, P4 lies in
of
2.
SUPPLEMENTARY PROBLEMS
5.
Prove:
If
vectors
-^m+i is linearly
6.
Show
X,^.
X^
are linearly independent while the set obtained by adding another vector
X-^
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
r<m.
8.
is unique.
i=i
7.
Problems
X^^.
s^,
second row by
In (9.3) subtract
S2.
^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 =
(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
X^ = aXj, a
or
dependent.
Is the
converse true?
11.
A"
proportional.
12. (a)
Show
that
X._ =
Show
Zi
that
()
Xg + 2Xg + x^
/i
fz
2::i
/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
or
Hox
+ 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
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
[::]-Ci]-[::]
over F.
Show
that
fe^il/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.
components,)
to a set of
mxn matrices
M-^.Mz.Mq
74
17.
Show
that
18.
Show
and
3
-
that
[CHAP.
of the matrices
[o
oj'
[o oj
and
[i
oj'
X^
X^^.X^
show
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.
r,
22.
PaY
is
C2
C7-,
o]
Pjd,
1, 1, 1),
= [C^,
2, 3. 4),
1
A^ +
+ kr,A^
+ kp^^A + kpl
...
23.
/,
4,^^,/!^
in the light of
Problem
16.
Ans.
24. In
(a)
4^-24=0,
(b)
(c),
L J,
[:;]
4^-24
is satisfied
A
'-[:-:]
(b)
+ 27 =
0,
(c)
\_
A^ - 2A +1
A'
(c)
\.
by
[;:]
=
(b)
A'''^
I-^A.
(c)
A~'^=2l-A,
and
chapter 10
Linear Equations
of
022X2 +
OqiX-i +
(10.1)
"Wl*!"'" Ob2*^2 +
+ c!2n*-n
^2
Aa
OIj,
unknowns
xi.a;.
>
*r?
x^
in
in
By
a solution in
(o's)
equations.
When
many
it
is
F which
sat-
said to be consistent;
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.
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)
for the
xj\ and H
[x^.Xr,
[h^h
A^]'
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
To solve
to replace
the entire
Example
3xi +
1.
x^
2X2
^.X-^
Xg
2x-j^
+ ^Xq + 2xq
^Xq^
-2
-3
-1
-5 -5
-5 -5
-]
2'
"1
-1
1
1
1
is the
= [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.
bleS X:
Jr+1
If
X:
Jr+2
X,-
On
a solution.
0,
...
...
Prom
of zeroes.
the first
,xj
Jr
X; and one of the i^, Uq
Jn
k^
= k
If
values for
kj; 7^
x:
0,
X.
ac,-
k^.
if at
least one of
V+i'
"r+s
'
%Ji
,,
x
,
...
J2
is different
X- constitute
Jr
and (10.1)
0%
+ 0*71
7^
is inconsistent.
A and
[A
H]
Thus
A system AX
of
linear equations in n
if
of the
In a consistent
system (10.1)
of rank
3*^3
+ 3^2 +
Xi +
Example
2.
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,
If
is over
F.
If
by x^ = 10 + lla, xj =
is given
-2-4o,
xg = a,
x^,
or
a, o]'
2)
it
1) that
solution
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
if
is not a
systems.
In
Problem 3 we prove
ni.
A system
of
re
non-homogeneous equations
in n
is n, that is.
0.
In addition to the
(a) Solution
<1"'5)
777
X2
'
"'
See Problem
X2 + Sxg +
2xi
Example
3.
X2
We
*4
- 3x3 - 4x4
6x2 - 2x3 +
X4.
Xl +
3x:i
2%-^
+ 2*2 + 2x3
- 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
^.
-120
If
^ #
0,
AX
is
given
by
A-^-AX
(10.6)
A-^H
or
A-^H
2xi + 3X2 +
Example
The
4.
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_
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.
is called a
always consistent.
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,
IV.
trivial
for (10.8) to
|/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
then
Zy, +
'P
AY
0.
If
AX
AX
lution of
Example
5.
E
AX
is
consistent, a complete
= H.
i
system
In the
solution is
A:^,
2x2
Xi
Xj +
= [o,
+ 3x3
^2 + 2 Xg
The complete
1, 2]'.
solution of < *^
\^Xi
where a
~
+
*^
Xq + 2*3 =
[-7a,a,3o]' +
[-7a,
[O, 1,2]'
IS
-^
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
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
x^ -
solution
x^ +
Solve
^4
+ 33=5 =
x^^a,
X4
s - 2x4
+ 3xg
X2 + 2Xg +
5%
Take
2a,
2%i + 3x2 -
4xi +
0.
x^^
1.
Snliitinn*
'11
[A H]
The
last
3-1-22
'112
5"'
7
O-^tj +
"V-
1-5-4
1-5-4
-8
-13
1
1
13
-5 -4 -8
-5
is inconsistent
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
/.
is
is a solution of the
X = L is a second solution
non-singular, K = L, and the solution is
4.
in n
0.
[A H] is reduced to
Since A is
[CHAP. 10
/,
suppose
system.
AK
and
AL
H, and
AK
AL.
unique.
(1)
il*l + a^jAiQ +
+ ain*n
h^
uqiXi + 022*2 +
+ 2n*n
^z
"ni*i + an2*2 +
+ "nn^n
^n
[a.
and
S
X
which by Theorems
ai^di-^x-^ +
ai20iii% +
22
^^2
^n
""no.
10,
Chapter
.S
1=1
i=i
i=l
3,
''is
Oqi
'^23
/12
\A\.x^
and add.
n
We have
ain^ilXn
1
=1
reduces to
"m
'^2n
so that
x-i
^11j-
^1
"n
n2 and sum to obtain
in
ttni.
by
"m
"2n
so that
(i)
*<,
respectively by a^n. 2n
(^,71
"l,n-i "1
"21
"2,n-i ^2
so that
x^.
'ni
2Xi +
5.
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^ =
7.
1
''\J
i 1
0.
and then a =
X2 = -2, *3 =
3,
1,
11110
13
= 2
6 =
X4 =
-i
3
2
-^a + 16, ^^
x^ =
is 2,
first
-2 -1 -3
11110
1
o"
and
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
AX
1^1
(A'X =
=0,
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
linear forms
^j
are linearly dependent if and only if the
The g's
if
mxp
and only
0-=1.2
^^'ijfi-
p)
if
a.
in
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
j?/i^v
trivial solution
9.
Suppose A =
[a-
apV
= [a^.a^,
] of order n
AB
and
if
Show
is singular.
that there
in p unknovras
say AB^
1,
m)
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-
(a)
/4ns.
-^
(d)
+ 3c,
17/3,
x^ +
2%i + 5%2
x-^
(b)
= 0.
x-^
Xj_
?^
{a)
[h^j]
= 0, or
SUPPLEMENTARY PROBLEMS
10.
has a non-
s^j xj -
r<p.
Since the coefficient matrix ^ is singular, the system in the unknowns h^t,^'2t
0.
Let Bi.Bg
(i
+ "p'iP
^i^-ii
*3
+ 2*3 +
*3
^ *47*2 4*3 5*4
3
*2
11*5
1*3
5*4
CHAP.
12.
LINEAR EQUATIONS
10]
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
B from
= 0.
find a matrix
83
and only
is singular if
if its
Show
16.
Use Problem
ixg
3%
^1
Hint.
third
To
row of
Ans.
17.
15 to solve:
- 2%2 +
(b)
-2
+ 2x^
y'ix^
4%2
+ 2%3
Xg
(c)
%2 + 6 Xg
+ 3*2 + 4
2.^1
1 2xi
a""
-27a. X2
(a) xi =
/ 2xj_
9a
= 0, X3 =
[3a,
or
-a]',
0,
(6) [2a,
-7a, -17a]',
(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
AX
tion of
ci, C2
equal to
Then choose
= H.
First choose
0.
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,
and only
complete solution of
Hint.
21.
i^
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
(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
.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
+Ps^a,.. a^.-p^a^,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
'
^'"
'''''"''
''
f?ctors''hcJ?'"
(a)
where (h,i,j,h =
1,
n).
a^^-a^^
a^J^(x^,
a^^aJJ
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]
[_0
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
is consistent
and of rank
r,
for
what set of
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
si^i + 51^2 +
^n-r+i^n-r+i
n-r+i
where
1,
a complete solution.
is
i=i
i and
2 and
Iq
by
E^
Show
6/2
"11
cE^ + dlQ
_/lJ
oEq +
'eI
h.
'b
I^,
and
I^,
and E^
AX
Now
write
of A'
'l
-1
X2
_s_
1"
0,
71
matrix with
72
AX
= K.
p.
= E^,
(i
1,
^n^%
n).
is a solution of
m<n
>-
"n"
'xi
'e2
30. Let
31.
and
that
28.
m).
,
S^].
where E^ is the
k^' show
,
that
Chapter
11
Vector Spaces
we
all
now be column vectors. When components are disThe transpose mark (') indicates that the elements are
vectors will
a^]'.
to be written in a column.
A
them
if
Example
1. (a)
The set of
(6)
The set
of all vectors
is
multiplication.
<"!)
K^i
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
space over
F.
SUBSPACES. A
vector space V
lie in
X^_
is
V and
X^,
X^
(11.1).
[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
yi +
y2
yi +
(i)
yi
is consistent.
ys
2y2 +
3y2
Xj^,
3x4
3ys + 2y4
2y3
Xj. Xg,
y4-
X4 spanS.
The vectors
of
tt)
X-y
hX^. where
a subspace (the line L) of S which contains every vector
A is a real number.
See Problem
BASIS
1.
AND DIMENSION. By
dimension
A
we
1.
When
n,
set of
The vectors
3.
X^.
Xr,.
X^
of
X-y
71-^1
Example
+ j^Xq + yg^a
yi
ys
ys
yi
2y2
3ys
3X2
2ys
yi +
ys =
,^ + 3X3 =
yi + 2y2
3X2 + 2xs =
!yi
xi
The
Xi+
y2 +
b. c ]'
of S can be
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
In
particular,
Theorem IV may be
re-
stated as:
I
If
X X^
..
X^
F and
if r is
See Problems
2-3.
CHAP.
VECTOR SPACES
11]
If
^m4.2. .
JYi,
^n
^2,
...,
Zm
m<n
are
87
X^
X^,
X.^,
X^
V^iF) and
J^+j^,
if
X^ form a linearly
is a basis of V^iF).
See Problem
III.
If
IV. If
4.
Zi,
.2, ...,
p>m
if
or,
i^'ij^i
when p<m,
if
(/=l-2
P)
n
Yi
=
7
IDENTICAL SUBSPACES.
only
is
If
X(F)
,V^(F) and
if
a^jXj
"
(i
= 1,2
re)
[a^-] is nonsingular.
are two
is a vector of ^V;[(F)
See Problem
is
5.
X+Y
By
two spaces.
inter-
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.
.
NUIXITY OF A MATRIX.
If
A has
nullity N^
10,
then
we have
AX
VECTOR SPACES
88
Xti
A
AX
[CHAP. 11
is a linear
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
AB
>
of order
'^
Nab > Na
(11.3)
Nab
El = [1,0,0
N/^,
ra
rg
the
Nji
A and B are
If
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.
Every vector
uniquely as the
^nl' of 1^(F) can be expressed
[%,%
sum
1=1
X
X
XxE-^
xiEi
The components %,
2-^2
x^ oi
x^
are
'^nEr,
Let Zi, Zg
in F such that
Then there
%,
03
a^
X
i
These scalars
JY^
= [oi, og
is the matrix
If
02
Zg +
Examples.
01,05
(11.4)
where Z
a.^Z.^
aiZi
Zi =
Z^ = [l,
2,
-1
]',
Z3
= [l, -1,
dj^Zj^
-1
Zi,
]'
Writing
Z-Xz
Zn]Xz
[Zi.Zg
=i
is
Z.
Zg
a basis of
Fg
(F) and
Xz
[1.2.3]'
-1
-1
-1 -1
"1
[7,0,-2]'
-2
See Problem 11.
CHAP.
VECTOR SPACES
11]
Let
W^,
(11.5)
From
=
=
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
Xf/
P determined
,
= PX^
See Problem
12.
SOLVED PROBLEMS
1.
The
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
2.
Since
X.^
= [1,2,2,
1 ]',
(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-^
A:g.
or
X^
^"1
= [1,1,1,0]',
A's
= [4,3,2,-1
]',
A'g
= [2,1,0,-1]',
-1 -1 -2
and
^"4
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
we
First,
2Zj^
X^..
Also, the Yi being linearly independent span a space of dimension two. say
iF|(''').
Next,
2lg^(F)
(c
Xq
6. (a) If
kXz -
is
a vector
-2% +
Vs(F).
X^ =
aY-^
fcK,
of
is a vector
then
5X2 + IXg
(b) If
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 V^iF) as intersection
space, then h + k = s + t.
Prove:
If
Suppose
s
= h; then Vr!XF) is a
will
show
same
Now suppose
is true if
is 1^' itself.
X^ span
=h
and
V^(F).
Yh
X
1=1
"iXi
.S
a^Yi
t=t+i
ft
i=i
i=t+i
biZi
i=t+i
i=t+i
left
belongs to P^(F),and from the right member, belongs also to ^(F); thus
X-t
from (11.4),
2
i=i
a^+i = at+2 =
"f,
it
belongs
= 0.
Now
(11-4)
to
Thus, s=k.
= k.
"iXi
2;
b^Z^
i=t+i
''t^i =
s
t*^"^,
*t+2 = = ^fe =
as was to be proved.
=h + k-t
CHAP.
8.
VECTOR SPACES
11]
1,1,1
]'
91
113
2 110
1
and
Fj
1.0,1
]'
12
space
is V,(F).
Prom
h + k
t.
X-^,
and
X^,
Y^
To
sum
is of rank 3, the
find a basis,
we equate
linear combinations
- 3e
take d =
aX;]^
9.
for
+ 6^2 = [-1.-2/3.-1/3
Determine a basis
]' is
^2a
6-
( 3a
- 2e
space of A
AX
a = 1/3,
obtaining
113
10
113
.4
c =
-2/3.
Then
is also a basis.
x^ =
which reduces
to
\x2+
A
6 = -4/3,
Xs + 2x^ =
of these equations.
r >
10. Prove:
AB
Suppose
r
A
- n.
first that
PAQ
By Problem
r,
rows of
10,
the special
Solution
(a).
Find
AB
AB
is
^45
>
+ %
'k
r.
rows of B while
"
aZ^ + 6Z2 +
Thus
its
cZr^.
that is.
+ h
c = 2.
[l 1 0]'
Write
1
case when
(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?).
11.
Then the
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
X-^
Z2=[l,0,l]',
trix ? such that X^ = PX^
Then
Zg]
\_Z^, Z^,
W'^Z
1
1
-1
_ _
ff^
= [2,2,1
]',
Ifg
= [1,2,2
]'.
7,^
= [1,1,0]'
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
SUPPLEMENTARY PROBLEMS
13.
Let
[x-i^.
x^. x^.
X4,y
R denotes
Which of the
with x^ = x^.
vectors with %4 =
x^=2x^.
x-^
x^.Xr,.Xs.xj^
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
[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]'
[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
X^.X^
X-^,
Hint.
Assume
51
aiXi
biXi-
X^
X^
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
19.
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^=
[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)
]',
]'
[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
X^^
P such
[l.l.l]'.
(b) [1,-3,5]',
(a) [-2,-1,1]',
(c)
[-1/3,2/3,0]',
(a)
23.
8,
h^Ej^ +
h^E^+
+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
the vector
space of
Ans.
28. Prove:
Hint:
[1,-1,-1]',
(a)
(a)
r^g
5.
r^g
Consider n~r^g
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
]',
(b)N^,<N^^N^
r^
and
rg
Then resolve
chapter 12
Linear Transformations
DEFINITION. Let
%]' and Y =
JnY ^^ ^^ vectors
lyx. y^.
are related by
yi
'^ll'^l
"^
^12 "^2
"^
^^Yi J^
df^Yj.^ri
(12.1)
AX
or, briefly,
where A =
[a^.-]
is over F.
of
(a)
(fe)
If
it
it
feFg. for
b.
Example
1.
(a)
(6)
The image
of
A"
= [2,0,5]' is
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,
Ogi,
...,
a^J'
and, in general,
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.
LINEAR TRANSFORMATIONS
12]
95
II.
transformation, is non-singular.
in.
See Problem
2.
See Problem
3.
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
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
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
X^
Then
Y^
Q-%
QX^
=
and
Q-^X,
Y^
QY^
Q-^AQX^
BX^
where
(12.2)
Two
Q-^AQ
for
which B = Q'^AQ
will be
for
P"^, (12.2)
made
later.
no compelling reason.
LINEAR TRANSFORMATIONS
96
Example
Let
2.
AX
[l.2,l]', W^ = [1,-1.2]',
= BXjf
IFg
(a)
to the W-basis.
(b)
corresponding to
AX.
(c)
let
W^
image relative
[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
[9,5,7]'
Xff =
W X
Yf^
= [l,l,l]'.
=
W Y
[14/3,20/9.19/9]'.
(b)
w\
(c) Yj.
W^AX
36
21
-15
21
10
-11
-3 23
-1
(W~^AW)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)
carries E^ into
Y^
= [1,2,3]',
E^ into [3,1,2]',
By TheoremI, A
The image
of
image of Xg
(c)
Y = AX which
The rank
is
X^=
K3
[l,l.l]' is
Y-^
13
[6.4,8]'.
The image
of
=^
AX
-Yg is
=[ 14.13,27]'.
of [A'^.Xg] =
-1
_1
19
We may compare
X^
CHAP.
2.
LINEAR TRANSFORMATIONS
12]
Prove:
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:
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^
p)
sp
^lYi +
H'^i
s^Y2+
1=1
s^Yfy
^^
f^
P
'
.|^ ^(-4^1)
4.
Hence, the
Y^ are linearly
[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
Yy
= AXf, =
X^
]'
1,0,1
into
[2,3,-1]',
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
Q-^AX,
Problem
--3
and
,!
^Xg
[Y^.Y^.Yg]X
12,
same transformation
From Problem
is contrary to the
c =
is K, = [l,l,l
+ 2c =
]'+ [3,0,-2]' +
5. If
+ spXp)
independent.
Let
A(Sj^X^+ S2X^+
Q^^Q^r.,
\^
14 -6
14
-9
X,
12,
Chap-
LINEAR TRANSFORMATIONS
98
[CHAP. 12
SUPPLEMENTARY PROBLEMS
Problem
6. In
show:
(b)
Af
= [1,1,2]',
A Y
into
7.
8.
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.
Use Theorem
changed.
to
III
show
....
X^
X whose
image
is
12.
X-^.X^.
if
-2
dependent vectors
^i=[l,l,l]',
JVg =
X.
show
]',
(a) it is singular,
(6) the
in-
3 5
[2.I.2
(c) the
image
of V^{R) is a Vs(R).
13.
2 4 X.
show
(a) it is singular,
(6) the
113
2
V,
14.
(R) spanned by
]'
Yi,
= 1,2
]'
spanned by [5,7.5]'.
Hint.
Y
15. Prove:
16.
and [3.2.0
1.1,1
Let
BZ
(j
AX
^2
12
_1
1_
= [1.0.1]',
Z3
(6)
= [14,10,6]' is the
= [1.1.1]'
image of
new
A^
be Chosen.
Yy =
PX
and Y^ = PY.
Q ^AQX.
AT
=P
basis, say
where P
where Q
Let
new
Show
Z, =
that
Y has coordinates
Y^ = [8,4,2]'
0-1"
(d)
AX
A^
(a)
X^
[1,1,0]',
(c)
Kj.
-1
-1
iZ^,
Zf2,
Z^i
^.
0"
1
17.
7^
1
1
1
1
Xjf
W^= [o,-l,2]',
IK,= [4,1,0]'
CHAP.
IK5
LINEAR TRANSFORMATIONS
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
AX. A
is singular,
123"
(a)
Problem
-4ns.
19.
12.
(6)
Problem
13.
(c)
X.
(a)
(6)
(c)
I/^()
]'
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^
13
spaces.
^ spanned
(c)
10
-14-6
20.
1,
n.
PX
yi
(i
^h
1.2
n)
(a)
(b)
(c)
Prove:
If
(d)
Prove:
If
(e)
(/)
= [s, 1, 4,
/^ is
/1./2
a permuta-
and
P^^P^Pt.
of a
/.
1,2
of
and ^
col-
mentary n-vectors.
PP'
which
in
chapter 13
Vectors Over the Real Field
all
= [ji,
YnY
72, ,
X-Y
(13.1)
Example
1.
x^y^
^2=
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)
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
.
The following
(13.2)
(a)
X^-X^ = X^-X^,
ib)
X^-(X^ + Xs)
(c)
(X^+ X^)
0.
(X^+Xs)-X^
(Xs+ X^)
X^-hX^ = HX^-X^)
X and Y
vectors
X^-X^
^i(R),
of
X^-X^ + X^-X^
=
+
Example
denoted by
\\
are orthogonal.
X\\
is
(13.3)
II
1(c),
^11
\\
X^\\ =
\/
V3
X-X
\/
xl +
xl+
---
X.
See Problems
100
1-2.
CHAP.
13]
(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
If
\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
If
(13-6)
If
X^,
Any
I.
set of
m<
vector
0,
IUII + ||y||
X^
X^
then for
then
are
= 1,2
m<n
m. (c.lX^+ 0^X2+
we have
<
l!^+yll
)/(/?),
3.
is said to be
Ijf(/?).
orthogonal to every
if it is
If
a vector
to the space
is
X^,
it is
orthogonal
spanned by them.
See Problem
which
is orthogonal to
at
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 V^iR), 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
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
See Problem
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
1^11
basis of
Example
F(/?).
3.
Construct, using the Gram-Schmidt process, an orthogonal basis of V2(R). given a basis
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 =
here because
Each vector
X.^
and
[1,-2.1]'
- -^y, - ^[1,1.1]'
[1,2,3]'
G^ = -jpj =
and
Gg
[-1,0,1]'
]',
-ii-
[-I/V2,
0,
1/\A2]'
A^2 a^re
Gj =
orthogonal vectors.
CHAP.
13]
Let
^m be a basis
^2
Zi,
103
X^,
Xg,(l< s<
X^
m),
are
mutually orthogonal.
y^, Yg
}^
VI. If
X-i^,
X2,
, Xs,(l< s<m),
X^^^,
^m
X^^.^-
is an
orthonormal basis.
THE GRAMIAN.
Let X^, X^
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
Problem
if
14,
Chapter
17,
we
ORTHOGONAL MATRICES. A
XpX^
...
XpXp
is diagonal.
shall prove
VII.
A^ A^ A2
A-L
Z^, X^,
....
Xp,
G >0. The
|
AA'
(13.9)
A'A
equality holds
if
and only
if
that is, if
(13.9')
A'
Prom
(13.9)
it
A'
Example
4.
By Examples, A
l/\/3
l/x/6
l/\[Z
-2/\/6
l/\/3
\/\[&
-1/^2
is orthogonal.
l/x/2
If
is orthogonal, its
X.
XI.
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
XII.
A
lem
if
y,
Y,f]
it
preserves inner
Thus,
linear transformation
Y=AX
is called
and only
if
orthogonal
if its
if it
matrix
is orthogonal.
we prove
10,
XIII.
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.
Zj.
(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
2/3
1/3
(6) Write
2/3
-2/3
column of zeros.
3.
Then by
2/3
0_
(3.11)
If
O"
-2/3 -1/3
[A:,y,o]
If
if A"
is
or
orthogonal to both
Vn(R), then
Assume then
A:
\X-Y\ <
that
and K.
||A'||.||y||
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
Since
Xi-Y
= 0, (i = 1,2
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
^^*
-^I'-^s
^h be a basis
of the
FV).
let X^,^^
be a vector
in the
vM)
A"
of
v\R)
"
<*)
The condition
that
if its
m-\\Y\
<
{aiXi + a^X^+... +
and only
<
Any vector
if
a^X^ + a^X^ +
...
Xf,
aj^X,^
+ a^^^Xf,^^
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
""^^
Vn-k'J'^^
^^* -^i-^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,-
system
(i) is
of rank k
Uniqueness follows from the fact that the intersection space of the
V^iR) and the V^'^(R)
space so that the sum space is Xi{R).
7.
V^(R),
given
is the zero-
0,
-IA/2 ]'
1A/3]'
8.
satisfy the
Xk.X=0
X^^
to
complete the
-Y =
set.
^^' -^2
>^n
Y^.
Yr,
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^ =
X3 +
aYr,
Thus.
+ bY^
K, =
Y^-X^ + Y^-aY^
Y X
^Sill
X^ ~ ^1^
Xo-
Since
Y^. K,,
Y^
Y^-X^ + aY^-Y^
yi-Yj,
Y^-
X^ + bY^-Y^
Y^.Y^
Y^-X^ + aY^-Y^
and
Then
(d)
- ^ll^
is obtained.
and
7, =
Z3 -
^ki^ K _
2W^ v
CHAP.
9.
13]
Take
Y^ =
X^
= [2.I.3Y.
K,
X^ -
^^i-i
"2
[1.1.1]'
Fg,
107
Then
=
^3 V
- ii[2,1.3]'
[1,2.3]'
[-6/7.15/14,3/14]'
^1 '"^3
- ^
14
I4J
]',
[_3
3j
-1/^3
]'
10. Prove:
Let
y^. Yq
Suppose A
and only
if its
is
Y^-Y^
YiY^
AX.
Then
/.
(X'^A')(AX^)
X\X^
X^^-X^
are preserved.
and A
matrix is orthogonal.
(1)
and, by
if
X{(A'A)X2
Then
A'A=1
X^X^.
is orthogonal.
SUPPLEIVIENTARY PROBLEMS
H. Given
.Yg =
]'.
]'.
find:
(c)
Ans.
(a) 6, 0,
-3
Vb,
(6)
12.
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).
containing
and y.
(a)
Show
that
(b) Write
vector of
I^(ft) is
vector
orthogonal to itself
if
and only
if it is
and Y.
(6) Prove:
if
Z^ Xg = Xj_-Xq=
0,
then
108
16.
Prove:
vector
P^"(/?) If
and only
[CHAP. 13
if it
17.
18.
Prove:
20.
space is
I^(if).
19.
If
Show
Prove:
that
A"
||
!|Z +
1|
<
(||.Y||
+ \y\
|1a:|1
if
21.
Show
22.
(o)
Show
||y||)^,
and only
if
11.
Y .Z
X^
that if X^.
Problem
of
X^
2 are
]'
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
23.
Prove: (a)
is
orthogonal and
.4
is
24.
Prove Theorems
25.
Prove;
26.
each element of A
If
(6) If
If
.4
1,
equal to
is
of^
its
cofactor in
^4
.4
where A
(or A'A),
is orthogonal,
is n-square. is a
C'BC commute.
diagonal matrix
if
and only
if
are orthogonal.
27.
Prove:
UX
28.
Prove:
If
29.
Prove:
If
is n-square, then
X-(^AY) = {A'X) -Y
(b)X.X
30.
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^ =
[-4/\^, -3/^34,
[2/VI3.
V^i^R)
0,
(6) [3,0,2]'
3/\/34]'
-3/VT3]', [0,1,0]'
in order:
[3.2,1]'
[4,-1,0]',
(a)
[iV^. -iV^,
(b)
[^\/~2, 0,
(c)
[2V5/5, -V5/5,
[4,0,-1]'
0]',
iV2]',
[0,1,0]',
0]',
Take
Ans.
Y^ = X^,
[1,-1,-2]'
[2,-1,-2]',
[1,3,1]',
A^
^=1
[0.l/^^2.l/\f2Y,
0,
if
cl
Ans.
32.
Z^
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
]'.
]'
n);
CHAP.
13]
33-
34.
Show
35.
Prove:
36.
Use Problem 35
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^^APX^
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.
12).
Show
that if
is
orthogonal so also is
B.
and
ZxF
and
= [yi.yQ.ygY
where
25, 23]'
= (I
- A) (I + A)'"^
is
Jl
21 =
3 ys
,
Z2 =
X^, Y^.
^1
.
],
Zg =
yi
estciblish:
of
(6)
(c)
XxY
(d)
(kX)xY
41. If W. X. Y.
(a)
(d)
k(XxY)
XxikY),
k a scalar.
X x{Y+Z)
(b)X-(YxZ)
(c)
-(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
X xY
yi
(a)
be-
skew-symmetric.
= [21,
^3 ys
AX
and con-
Theorem XIV.
versely, to prove
39.
-51
comes
is orthogonal.
10
.4
-A)(I + A)~^
given
"5-14
If
= (I
2-3
-12
37. Prove:
12
s"!
(a)
+ A is non-singular, then
Ans.
]'.
[l.2,3,4]',
"
(a)
X-^ =
109
chapter 14
Vectors Over the Complex Field
COMPLEX NUMBERS.
fy.
are equal if and only if the real and imaginary parts of one are equal
A complex number
The conjugate
x + iy =
of the
if
complex number
It
and only
its
if
x = y =
= x+iy
0.
is given
by
= x+iy = xiy.
(14.1)
VECTORS. Let X
and
> \A
>
\z\
The sum
\J
z-z =
\fW+y^
\y\
The
is to
it
|z|
totality of
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,
72
product
defined as
X-Y
(14.2)
XY
%yi
(14.3)
I-y =
hold.
2?.{X.Y)
where R(X-Y)
X-Y.
(c)
X-(cY) = c(X-Y)
(d)
= 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
X-Y-Y-X
(g)
||Z||
if
\/
X-X =
X-Y = Y-X =
Q.
Triangle Inequality
\\X+Y\\
XnJn
(cX)-Y = c(X-Y)
For vectors of
(14.4)
X-Y+Y.X
(/)
(b)
The length
Two
y^
x^y^
<
||Z||
<
\\X\\-\\Y\\
||i'll
2)
\X-Y\
110
\/%%
%^2
+ XnXn-
1.
CHAP.
14]
I.
Any
set of
111
III.
If
X^,
X^
X^,
then
it
is or-
V^iC) is a subspace of V^C), k>h, then there exists at least one vector
in
J^(C),
m>0, contains m
m mutually
orthog-
onal vectors.
(14.6)
Let
X,. X^
Y,
X,
In
Xn
Yd
Xn
yn
_
~
X^
^2-^3
Y .Y
l2-l2
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
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-
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^
Following Problem
14,
Chapter
17,
is diagonal.
we may prove
\G\
>
0.
The equality
112
UNITARY MATRICES. An
fCHAP. 14
(AyA = A(A)'=
if
that is
if
(i)'=
^.
we have
VII.
I,
The product
IX.
The determinant
X.
linear transformation
(14.8)
where A
1.
AX
XI.
linear transformation preserves lengths (and hence, inner products) if and only
matrix is unitary.
if its
XII.
If
AX
is a transformation of
if
is unitary.
SOLVED PROBLEMS
1.
2+
(a)
X-Y
X'Y
[l-i,f,i;
3/
l-2i
7 4 3j
Y-X
Y'X
+i
[2-3i.l + 2i.-i]
3i
3i
= X-Y.
2.
(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):
As
in the
\X-Y\
<
2R(X-y)
p||
2C(X-Y)
11Y||.
is true if
or
= 0.
When X and Y
a'Wxf
+ 2aR{X-Y) +
are non-zero
Yf
{aX+Y)-{aX+Y)
\\Y\\
>
0.
CHAP.
14]
\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
Hermitian,
Consider B = ~iA.
and S
|UM|y||
forall^andy.
is
(A)A
This
|c||U||-iyl|
Thus,
Since
\\cX\\-\\y\\
\X-Y\.
(A)' =
t(Z)'
-A. Then
i(-A)
-iA
= iA.
SUPPLEMENTARY PROBLEMS
6.
A:2 = [l,
1+
/,
o]',
and Xg =
[i, 1
j,
2]'
(a)
(b)
(c)
show
(d)
Ans.
(a)
that
each vector Xi
2-3i.-i
(b) ^/e
7.
Show
8.
9.
that [l + i.i.lV.
is
X^ and
V^ V^
.
[iA-i.oy, and
10.
Prove Theorems
11.
I-IV.
Jf g
(d)
[l -i.
1. 3j ]'
114
12.
in order, construct
[CHAP. 14
an orthonormal basis
for iy.C)
when
the
vectors are
(a)
(b)
[l + i.i.lY.
Arts,
[O.^V2.-iV2]', [1(1 + 0.
(a)
ri
[2(1+0. 2..2
(&)
13. Prove: If
is
/I
T
J.
[3;^.
I. 2
5t
]'.
[-^i,
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
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.
18.
Prove:
If
the columns of
16.
If
if
the columns of
is
is re-square,
X-AY
then
A'X-Y
19.
Use Problem 18
(6)
"^
i
L-1+.-
_9 + 8i
Ans.
r-l +
|_
-4-2i"|
2j
If
2-4i
4-lOj
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
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
is non-singular, then
(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
(15.1)
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
SYMMETRIC MATRICES.
I.
whose
Example
1.
In
Problem
we prove
1,
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
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
is not unique.
10
0-100
ffgCi) and Kg(^).
for
D by
while the
10
10
transformations
replace
0-900
D by
There
however, no pair of
is,
D by
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.
r is
matrix
P
(15.2)
The
'r-p
p-(r p)
is called the
signature.
Example
2.
Applying the transformations H23. K^a and H^ik), 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
symmetric matrices.
CHAP.
IN
CONGRUENCE
15]
117
have
is
/^
(15.3)
H^ii)
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
ra-square
and only
we have
^&:]
R'AR
and
K^f^i)
If
is
skew-symmetric, then
(FAP)'
FAT
r(-A)P
-FAP
Thus,
VI.
Every matrix
B =
FAP
symmetric.
In
Problem
we prove
4,
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
The set
if
and only
of all matrices of the type (15.4) is a canonical set over congruence for re-square
skew-symmetric matrices.
[^
],
(15.5)
Thus,
or
conjunctive
if
FAP
P such
that
CONGRUENCE
118
Two
IX.
[CHAP.
if
and only
if
15
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
r is
Ip
-Ir-p
(15.6)
The
p-(r-p)
Two
XI.
rank and index or the same rank and the same signature.
The reduction
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.
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
is the rank of
A and
p is the
index of iA.
if
and only
See Problem
if
8.
CHAP.
CONGRUENCE
15]
119
SOLVED PROBLEMS
1.
of rank
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
we have obtained
S+1, s+2
k^
S+2, s+ 1
'^ri
however, some
k^j
say
s+^ /
0,
Suppose then
hss
in
the matrix
= o.
s+i
every
If
we move
4y
*s+l, n
"n,
s->-2
= 0,
we have proved
it
r.
If,
column
we
Since
whose
2.
first t
Reduce
In
'12
.2
to
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
2J
k\f2
-i
2\/2
2i
-i
2-j
Vc\?'\
iV2
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.
B
where
It
D^
= Q.
U :]
then S = ^.
(J
If
= 1,2,
-4
0,
diag(Di, Dg
to a matrix
0)
then some
mj
and
/th and second rows; then Interchange the Jth
oy
a..
j
-a,;,
-Oy
g
2
'S*
i
first
first
first
column by
l/a^^-
CHAP.
CONGRUENCE
15]
121
1
''2
to obtain
-1
and from
it,
3,
obtain
4
1
Di
-1
F4
If
5.
/v=
0.
a^s 4 0,
we need
-2 -1
-4
3
10
10
1-3
10
-4320
0-3
10
10
-4230
-1 -2
to obtain
1
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
-5
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
2i
-l-2j
-i
is
+ 2t
1
1
1 +
123
2j
-i
+i
1,
-l].
10
Then P'AP
= diag[i.
i.
-i
SUPPLEMENTARY PROBLEMS
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
-"
(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
+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.
in
+i
tl+
(a)
(6)
'
1-i
Ans.
13.
[1
(a)
-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-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
(a)
3i"|
10
3i
[CHAP. 15
satisfies
C'DC
14
If
15.
if
and only
|Cl=l.
if
L-i oj
Show
p.
that
>
if
and only
if
-p
is even.
16- Prove:
Hint.
17.
Take P
= I
matrices.
18.
19.
Prove:
It
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 + TT^iS- T) and I+P is non-singular,
Hint.
21.
and only
20.
if
S(/-P)(/ + P)-^
S(I + PrHl-P).
Show
that
chapter 16
Bilinear Forms
AN EXPRESSION
and
(y^, y^
which
Jri)
is linear
(x^, x^
m)
is called a bilinear
XxJi
2x172
ISxija
and
4x271
15x272
^s/s
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
See Problem
Example
1.
Ui.
%.%
X'AY
125
Ti
73
y2
1.
BILINEAR FORMS
126
CANONICAL FORMS.
Let the
[CHAP. 16
new variables
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
Two
and only
if
if
IX,
I.
only
If
if
sire
equivalent over
r,
if
and
P and Q such
that
Ir
FAQ
and C =
Taking B = P'
in (16.2)
(16.4)
UXPAQ)V
reduced
U'
U^V^
^2^2
to
U^V^
Thus,
II.
Any
mations over
of rank
u^^v^
+ u^v^ +
+ u^i'r-
X^AY
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
X-j^
X2
Xg
Ur,
and
U,
Ti
72
Ts
^1
"3
1^3
V^
^3
See Problem 2
bilinear form
1 S
X'AY
ij x
'^iJj
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
III.
where A
If
(yi.ys. .3^)-
and Y =
is called
CU
Y = CV,
and
We have
U\C'AC)V.
the variables.
V.
Two
variables
if
and only
From Theorem
I,
if their
ChapterlS, we have
(16.5)
%%yi
From Theorems
II
15,
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.
formation
Y = CV,
3.
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,
U\C AC)V.
is
if
and
VIII.
[CHAP. 16
Problem 4, we prove
In
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.
Ix^y^ + 2%yi
"ix-^y^
ical form.
2
The
-2
-1/6 -5/6
and
-1 -1
By Problem
6.
Chapter
5.
10
3-2
2-2
7/6
10
/g
PAQ
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.
mations
X'AY = X'
12
10
10
-8
(b) (16.6) in
field.
1
(o)
Prom Example
1,
Chapter
-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
^usfg.
(6)
Prom Example
Chapter
2,
15,
-5 -2 -1
1
-1
u.
-5 -2 -1
2
1-1
1
Prom the
result of
Example
- "3%Chapter
2,
000'
-5
-2
-1
we may obtain
15.
-1
-5202
-2j
-1-110
10
10
10
-5 -2J -1
i
-1
2
u.
-5 -2J -1
-1
2
t
reduce X'AY to
4.
Prove:
if
and only
if its
1.
is factorable so that
aij
aj5
a^bj
aib^
fk^'j
fe*s_
= [a^-], as
bib^
Lfei "ks_
vanishes.
rank is
of
/4
"k
"k
is 1.
(S rijXj)(l s^jyp
jnr
f(x.y)
and
'^jyj
BILINEAR FORMS
130
[CHAP. 16
SUPPLEMENTARY PROBLEMS
5.
Obtain linear transformations which reduce each of the following bilinear forms to canonical form (16.4)
(a)
x^j^
- 2^73 +
"ix^j^
(6)
-11
-4
7 y.
(c)
-2
6.
-2
-2
X'
5-510
^sTs
10
2-510
(rf)
12
10
(a)
14
y and
Ans.
(a)
(6)
X'
10
5_
-2 -7
'l
(6)
-3 -4V3/3
\/3/3
\/3/3
Ir
7. If Bj^,
B2. Ci
B-^A^C^ = B^A^C^ =
= (B'^B^iU,
8. Interpret
Problem
23,
C^C^V
Chapter
5, in
9. Write the
transformation contragredient to
.Y
1
1
11.
is
/.
.4ns.
-2
i
2
10.
221
~2
.Y
PV Y
.
PV.
-1
12. If
X'AY
is a real
XA Y
Show
that
when
reciprocal bilinear forms are transformed cogrediently by the same orthogonal transformation, reciprocal bilinear forms result.
13.
Use Problem
U1V2
14.
W2V-L
+ U3V4.- u^V3 +
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
^It ^2
Example
a^,-
"u
X'AX
are elements of
over
in the variables
^7J'
1.
2
as^^
+ 2*2 -
2
"^x^
'ix^x^
may be
written in various
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,
-2
0-7
4Xj^X2
SXj^Xg
[o^^- ]
X'
7Xg
"-^s
'
If
otherwise, non-singular.
TRANSFORMATIONS. The
(17.2)
(.BYyA(BY)
BY
YXB'AB)Y
Two
The rank
X=BY
we have
variables.
II.
Two
if
and only
if their
matrices are
congruent over F.
Prom Problem
1,
Chapter
15, it
the form
(17.3)
K^i
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
-2
Example
2.
Reduce
X'
Example
of
-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.
Example
3.
1
4x^*2
7%;
3
"*"
'
^*1*3
7^:
Ux^-2xJ
2Sx
(\-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
3.
Let the
(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
Wi
"'i
VSi Zi
'j
= 1,2
(/
=r+l,r+2
r)
71)
or
'
;-
'''4k-
.1.1
J
wl + wl +
(17.5)
w^^
;2^^
wl
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.
Example
4.
In
Example
o
y^
q" =
into
J
reduces
q
"
^12
2,
a^ + 2x^
8xx
1213
- Ix^ - ixx
3
was reduced
to
i.
to
"'
2,2
w^+
w^ -
w^.
Xl
X2
|mJ2 + 5V^;g
X3
3"^
4/3
v^
4/3
2^
1/3
qr
to
W^ +
In
wl-
The quadratic
q
w%.
Problem
we prove
5,
2.
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.
It is
2i
^iVi.
(i
Zj
yj
(7
= r+1, r+2,
1.
r)
...,7i)
QUADRATIC FORMS
134
[CHAP. 17
Ul
1
1, 1,
-'-
''''{/h^'/h;'-
)^
-4--
<
(17.6)
Thus,
VI.
field of rank
singular transformation over the complex field to the canonical form (17.6).
Two complex
VII.
if its
the
q>
a;'s,
be reduced
-yl
i.e.,
y^ + yj
to
=X'AX,
= p<
n.
"^
\a\
=0,
0.
A
i.e.,
-^
i.e.,
j\-^ J^-^
0.
q = X'AX, \a\ f 0, in
Thus, in the real field a
y^ and for any non-trivial set
re,
p =
- y^ Clearly,
X'AX
if
0.
to
=X'AX
VIII. If q
U| >0.
PRINCIPAL MINORS. A
6,
we prove
has
at least
r differ-
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.
singular matrix
XI.
C such
that
A =
exists a matrix
of rank
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.
See Problem
XIII. If
is non-negative.
over F,
[a^,-]
8.
we
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
Example
5.
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
in the
r.
By Theorem
A contains
IX,
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
^^^^
so on, until
is regularly arranged.
Thus,
See Problem
all
X'AX of rank
Pr
if
and only
if its
rank is
re
10.
and
is positive.
QUADRATIC FORMS
136
[CHAP. 17
method of Kronecker
which only the squared terms of the variables appear
XIX.
X'AX
q =
for
is
based on
in n variables of rank
F which reduces
X'BX
r,
then by a non-
q to
if
Pn-i =
re-1)
1, 2.
^nnyn
or
10
a^n
BY
1
<^n-i,
carries q into
HiJiYj + Pn-xPnYn
in
-2
-2
4
1
Example
6.
X'AX
= X'
-^i
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
q =
a,
're,?!-! ^
'
0,
-8
in
-8
Y'
-2
-2
36
squared form
F and
if
o^-i, n-i
"
xi
yi
CLi,n-iyn-i
Xn
^n-i.n yn.
*ra-i
(i
'^inyn,
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
J -1
Example
7.
X'AX
^22
- O'ss =
but
reduces X'AX
xi
X2
xq
*i3ys
*23r3
-1
or
0-10
a 32 72
-1
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
(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
pr-L =
a,-,
r-i ^
2ar,r~ipr{yr-i-yr)
(17.10)
reduced
to
another con-
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
P^,
po, pi, P2
Let X'AX f
0,
ratic form.
(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.
y?
Conversely,
if
!Y^
a^Xx + 02*2 +
72
X2
of rank
1.
or 2
it
Xn
1.
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
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
form (17.3).
to the
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']
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^
- 24x^+
8a:^a=2
2 (*i
2 (*i
(^ + ^ + 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)
*2 +
4*^*2 +
2,
we have
4*^*3 + 4*2 +
tenii in
16*2*3 + **
*2*3
we use
(*^
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
4 2
1
y^
+ ^y^ -
-1
2yi^
-1
ii
1
6"
-4
2
I
X\ hence,
4.
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
,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
X
).
FY
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]
(iii)
*i2*2+- +
*in^n)^ +
...
(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
+ bp^i,rf-J^
q<p
6.
'^g
M*re =
+ c^^x^
(Ot^.ttg
0!^^).
When
this
solutionis
+ c.^n'^nf +
+ '^9+2,2*2 +
+ (<^7ii + ^(?22 +
q<p. A repetition
a contradiction. Hence q=p.
x^f
+ l,/!^^
'^q
+ 6^
x^+- + c
IV,
substituted into
"^g
+ *in*n =
''21*1+^22*2 +
Thus,
q.
(''ii*i
*ii
141
Thus,
has
>an)^
at least
assumption that
different
from zero.
Since A is of rank
r,
it
has
at least
Now
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
pal minor of
7.
Prove:
matrix
A
C
in the
"vv
of rank
of rank
is positive semi-definite if
and only
if
there exists a
Since A is of rank
r,
'--[t:]
QUADRATIC FORMS
142
such that A
of rank
B%B.
and A
Since
Ai = A^
A'i'=
we have A
[CHAP. 17
Set
= A^B; then
is
as required.
= C'C
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.
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.
A^j,
A^^j-^, ...
three,
By Theorem
Ai>
VI,
A^j>
0,
0,
...
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
2.
Moreover,
it is
true for
of order
n>2
?^
After the ith row and the ith column have been
new matrix has p_^ = CC^j^ 0.
moved
to
a
n, n-i
^n-i,n-i
and
at
0.
Move
_i,n =
In the
Vi,n
^n-i,n
\n-i
^-
-a.n-i,n
^^
PnsPn
Vi,n
Pn^^O.
Note that this also proves Theorem XV.
10.
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
?^
pi = P2 =
0,
ps ^
0,
we examine
the matrix
CHAP.
QUADRATIC FORMS
17]
143
A yields
2
2
11
1111
for
11.
which p^ -
1,
p^ =
0,
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 /
0,
XIX
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
to
XCX
= X'
12 10
Since S is of rank
0.
3, it
can be reduced to
2 1
4 3
3 5
for
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
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
yl + 54y|
54^3^
- 243y|
QUADRATIC FORMS
144
14. Prove:
Ai
[CHAP. 17
Xp,
Ai
A]_
A2
;t2-x
A2 ^2
Ag Ai
'
>
Z^ Zi
Xp- X2
and only
if
Xp- Xp
if
P
(a)
0<Z-Z
and
)( 2 X:xA
X-x.
j=i
1=1
X\X':XAX
t-j'
(6)
let
G >
= [i,2
r(A:^-;f,-)^
t
X^x^
i=i
X'GX
J'
0.
Then there
Then Z
xp]' f 0.
exist scalars
k^^^.k^
such that
S
f = i=i
4,,jy.^
X..^
+ k^Xj
kiXjj-Xi
X2 +
+ kpXj
Xp
= 1.2
p)
homogeneous equations
of
Xj-X^xx + Xa-Xqxq +
'J
We have proved
that
k;
G >
1
(i = 1,2
+ Xa-X,
'j
'
p),
P ^P
and \G\
p)
0.
0.
= 1,2
we need only
(b).
assume
to
^=
^.
0,
if
= 0,
;?,
= 0, (/ = 1,2
and
=
|
S kjXy^ =
Thus,
f= l^k^Xi.
where
p)
p
i= 1
SUPPLEMENTARY PROBLEMS
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^ +
(a)
X'
Ans.
6-2 X
(a)
of
Problem
+ 2y^ - 48y|
is
(6)
y,=
a;i
-3
-5
1-13-3
13
012
X
(c)
2^2^
+ 4y^
-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
(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
Show
X'\
Prove:
different ranks.
are equivalent.
field to / (-/).
21.
22.
(a)
Show
(6)
Show
23.
of
Problem 12
is
reduced
to
X'CX hy X
RX.
where
Ks^K^^{-5)K^^a).
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 xs^i. xs+2.
..x^, then g = ?i + 92 is a positive definite form in xx,x2
x^.
that if
Prove: If
Hint: Consider
24.
25.
26.
Theorem
X.)
Hint: Consider
A can be
D'AD=I.
written as 4 =
CC. (Problems
B and C
if
B'AB =1 and A
= C'C
then
27.
28.
Show
29.
30.
that
By Kronecker's
if
and only
if
XX
ac-l^ >0.
and XXI.
of the following
(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
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)
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
by
where
is Hermitian
Hermitian form.
X'HX
(18.1)
v=\
j=i
The rank
of
__
h^jXiXj,
= hj^
hij
we
and
hence,
here are analogous to those of Chapter 17 and their proofs require only minor
changes from
hjiXjXi
Moreover,
for
the pair of
+ hijXiXj
hijXiXj
Thus,
1.
The values
into an-
{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
and
HI.
Two
if
and only
if their
can be reduced
to
diagonal
form
(18.3)
Kjxn
k^y-iLji
'crfryr'
^i^^
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
if
A
a.:e
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.
If
= C'C.
is positive,
is
and conversely.
non-negative, and
conversely.
SOLVED PROBLEM
1
1.
Reduce
X'
\-2i
2 + 3i
-4 +
Prom Problem
7,
1-2/
+ 2i
to
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_
BY
,0
to
0'
HC
2/VlO
(-4 +
4i)/V^
1/VIO
-i/yflQ
-j/VTo
1/v^
HERMITIAN FORMS
148
[CHAP.
18
SUPPLEMENTARY PROBLEMS
2.
Reduce each
+ 2il
l-2i
(c)
+
+
3j
3j
i"
(d)
For
(6), first
Ans.
{a)
(.b)
3i
3i
2i
2-J
2
2J
by
+J
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 +
j_ri
Ans.
X= -^\
that X'
1-t
-1
l/v^
1+i
-1
-3+J
-3-J
5.
6.
Obtain
7.
for
Prove:
BY
Show
Ti
(-i-2i)/v3i
:^'
v/2"Lo
4.
J<iyi
Jb
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
^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:
2/VlO
.0
3.
30/\/^
(-3-2J)/\/lO
1/^2
where
77-
= \h^j\
chapter 19
The Characteristic Equation of a Matrix
THE PROBLEM.
Let Y = AX,
where A = [o^],
(i,j
= 1,2
Z=
X
vectors X
n).
x^Y
[%i,%2
into a vector
Y =
JriY
\.%,J-z
is
possibility of certain
a scalar of
or of
some
We
shall investigate
here the
field 3F of
AX
(19.1)
for
which
XX
\X-AX
(19.2)
From
(XI-A)X
(19.1),
we
obtain
A di^
O^ 2
^21
^ f^2
~
The system
of
\XI-A\
(19.3)
A Oil
a-zi
^271
X a-nn
On2
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.
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
solution is given by
A=5
Xl
_X3
row equivalent
is
-1
xs
-2 -1
-1
[CHAP. 19
is the
When A
Two
A2=1,
(19.2)
becomes
-2 -l'
Xl
-2
-1
X2
-2
-1
_X3_
Xl + 2x2
+^3
Xi = [2,-l,0]'
Every vector hX^+kX^ = [2h+k,-h,-ky is an invariant vector of A.
and X2 = [1,0,-1]'.
is the
See Problems
GENERAL THEOREMS.
In
Problem
3,
(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.
In
Problem
k\
is
re!
4,
(ra
3) of
when k = n, and
As a consequence
III.
If
of
when k>n.
is
greater than
r.
See Problem
5.
In particular
Iir.
is
re
of Ex.
1,
<^(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
A=5
is of
CHAP.
19]
dimension
1.
multiplicity
2,
151
= l, of
is of dimension 2 (see
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
Since any principal minor of A is the conjugate of the corresponding principal minor of A,
we have
V.
of
/I'
roots of A.
By comparing
VI. If
Xi,
scalar, then
VII.
If
Problem
VIII.
A2
AAi,
kX^
we have
scalar, then
In
characteristic equations,
7,
If
is a
if
A;
if
& is a
we prove
is
SOLVED PROBLEMS
1.
If
is re-square,
(19.4)
where s^
We
show
0(A)
(m =
1,
that
|A/-^|
sum
...
+ s_,A + (-l)"|i|
A-aii
0-ai2
0-021
A-a22
0-a2n
O-'^ni
O-Ons
X-
a-in
ar
-A
its
columns numbered h,
ia
are columns
terminant becomes
152
%.ii
[CHAP. 19
"iiA
"is.ii
"imA
"inA
"im.-im
(-1)'"
I
%.fe
"in-k
t^.to
im!
i\
Aj
i
^'2
(-1)"
as
i^
(ii, is
%''''OT
..
where
^n
'i.'2
..
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
Assume
--
(1)
8 + 2
A^'
+ 9A^ -
7A
assume
-1
-4
-2
-1
-1
5-4
+
-2
4-1
2.
+ 0^X2 + 03X3
A^^
(11)
may be
by A and obtain
(III)
we have
Now
-2
(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 =
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
= 0.
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
<^"(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
-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'^AXi) + 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.
2,
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
is that
spanned by
A=
1:
XI -A
-1
=
1
1-4
sion
1.
The associated
[2, 3,
-4
-2
-1
For
-2, -3]'.
1-5
[3, 6,
-4, -5]'
dimen-
154
7.
Prove:
If
n-
[CHAP. 19
By Problem
1,
a"+ .!"-%
(i)
where
s^.
n - 1) is (-1)
= 1. 2
(i
l/xZ-adj/ll
where Sj
= 1, 2
(/
By
and the definitions of
llll-M
A\ = \AY' ^ then
(6.4)
s,-
and
S,-
all
Si/^""^
sum
l'^!
+ S^^^/s +
...
(-if
ladj^l
Sj = (-1)" |-4|s._
Si= (-l)"s^_.,,
-^
To
Jl 1
and
m"
1)
^n-i^ + (-1)
(-if Ml"-^si,
|adj
|;x/-adj A\
+ ._,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-
...
(i)
...
^_,a
(-i)"Ml
8.
AAA
We have
=
</,(A)
|X/-P|
\\PIP'-P\
|-PA(^/-P')|
is a reciprocal equation.
A"|f/-P|
A"c^(i)
SUPPLEMENTARY PROBLEMS
9.
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]';
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)
[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,
[3,0, 1,4]';
XX
X'AX
10.
Prove:
If A" is
H.
Prove:
if
155
then
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'
equation.
t:]
Prove:
If the n-stjuare
matrix A is of rank
Prove:
If
A and B
sum
r.
\(X
+ k)I
diag(.4i,
-A\
A2
r<n, show
then at least n
= (X + k~ X-O (X + k
(X + k -X^^.
that
-r
is non-singular, then
-X2)
NA
AN
and
roots.
18.
X9.
17,
show
that
Write
-A)\
roots.
iAl IA2
lA
are
The characteristic
20. Prove:
Hint.
21.
If
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.
The characteristic
(-i)"M|
<^ (0)
(-1)
(0)
(-1)
0^">(O)
n!
r!
of
of
.4
Problem 23 into
c/3(X)
to obtain (19.4).
1.
II,
0(0)
<P
(?i(0)
+ (^'(0)-X +-^(f^'(0)-X^
...
+^0'^'^'(O)-A"
chapter 20
Similarity
TWO
if
B = R'^-AR
(20.1)
Example
11
2J
R~'^AR
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.
An
invariant vector of
1)^ =
of
B associated with A=
is
Y^ = [l,0,0]' and
it
is readily
= RYi =
that
[1, 1, 1]' is
root
Example
characteristic
1.
Two
I.
shown
same characteristic
roots.
If
A^ of B, then
is an invariant vector of
RY
1.
is an invariant vector of
root A^ of A.
D =
a^)
diag(ai, 02
2.
are
DE^ = a^E^,
(i
Any
The elementary
1, 2, ...,n).
and 4
for proofs)
invariant vectors.
IV.
If
an re-square matrix
A has
re
it
is similar
to a diagonal matrix.
See Problem
In
Problem
V.
\IA
6,
5.
we prove
Over a
field
factors completely in
F and
if
and only
if
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.
to
a triangular matrix
See Problems
7-8.
If
is
orthogonal matrix
any real
n-
P such
that
See Problem
of
Theorem
of
Theorem VHI
11.
DIAGONABLE MATRICES. A
Theorem IV
is
SOLVED PROBLEMS
1.
Prove:
Two
Let A and B
R~^AR
A/-B
(')
same characteristic
roots.
\I -
R'^AR
R-'^UR - 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:
is an invariant vector oi
is an invariant vector of
ARY
A,-,
then
AR; then
=
RBY
A corresponding
RX^Y
X-RY
X^X
has
re
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
If
is similar to a
it
diagonal matrix.
AR
,4A'^ = A^A'^,
AX^]
[AX^.AX^
0'
[Ai^i, ^2^2
Ai
1X1,
R ^AR
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).
Take
charac-
n).
1, 2
and
0"
-1
a diagonal matrix.
6.
Prove:
Over a
field
factors completely in
of
X^I-A.
suppose that
First,
r,
'i-''2
Take
where
Xi^.Xi^.
exist scalars a
.,Xir-
s).
XJ'
Now
= 1,2
(i
such that
(i)
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,
7.
Prove:
(i);
it
is
an
thus, the
are the
characteristic roots of A.
Take
X-i
as the
first
CHAP.
SIMILARITY
20]
Qi^AQ^
(i)
where
A-^
is of order n
Since
|A/
hi
Bil
[o
A,j
1.
-ft
159
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
=
Q2^A^Q^
(il)
where A2
is of order n
2.
n =
If
3,
/I2 = [As],
[^2
52I
[0
A^j
I2
/i
n-2
(iii)
(?2
such that
8.
(J
/1()
is triangular
Q2J
<?3
Qn-i
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
-l]'.
[4, 0,
Take ft
then
-20
o"!
ft A1Q2
and
II
J_
20
-15
-48
64
-11
48j
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''^AP 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
hi
P'lAP^
where Ai
is of order n
fill
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
for
10.
which P~^AP
is triangular
pi
0]
|_o
pj
Pi.
pn-2
"I
P_J
P-^AP
is triangular
2_
Prom Example
to
P""
1"
1,
5, 1, 1
is [1,0, -1]'.
"
We take
Qi
10
-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
[l,
161
r
we
"2=1 1/V3
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.
f/
such that
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)
0,
1-i, 3 +
2i.
For A =
0,
take
[l,
Qi
-1
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
12.
P such
that
P ^AP
+2J
= Ui.
is triangular
The
[1,1,1]',
[1,-2,1]' respectively.
-1
Now
may be taken as
[l,0, -l]'.
Taking
l/V^
P
-1/-/2
find
orthogonal.
we
-1
-1
P~^ AP
= diag(2,3, 6).
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
SUPPLEMENTARY PROBLEMS
13.
Find an orthogonal matrix P such that f'^AP 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"
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.
18.
is real
S
are similar.
Let
is tri-
(/)
\/^[2
\/y[2
Prove:
If
U'^AU
-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
B
to
that
and
VIII.
C)
= diag(Ci, C2
= diagCSj, B^).
Theorem
Show
m).
1, 2
20.
1/^6"
(6)
Ans.
19.
1/^
I/V3"
1/V2"
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
BT^EK
that
B^).
to prove
/i
and
/g are
B and C are
those
similar.
oJ
= diag(Bi,
Bg
B)
Hint.
AE
Q'^BP'^N.
See Problem
15,
Chapter 19
A^ are non-singular and of the same order, show thut A-^A^-.. A^, A2As-.. A^A^, A3... A^A-^^A^.
have the same characteristic equation.
22. If A-i.A^
...
23.
Q'^AQ
Let
of
(a)
B where B
is triangular
Show that
S
A Q
is triangular
roots of A.
(b)
.4.
Show
that
X,-
= trace
24.
Show
25.
Show
that
'2
_1
2_
-1
2-1
and
-3
-2
chapter 21
Similarity to
be combined but
The characteristic
I.
a Diagonal Matrix
See Problem
The
II.
1.
See Problem
When A
is real
9,
Chapter 20, is
0;
hence,
is a real ra-square
III.
Theorem
IV.
III
If
A,
A).
r^
of a real
Theorem
III
r^.
becomes
BY
(21.1)
where
the
...,
implies
2.
Aiyi + X^yl +
r
is the rank of
A and
Ai,
Aj
...
A^ are
+ X^y"^
its
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-
^|
0.
real symmetric matrix is positive definite if and only if all of its characteristic
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.
... ^A.
The totality of
such diagonal matrices constitutes a canonical set for real symmetric matrices under orthogonal
similarity. We have
VIII.
the
Two
same characteristic
163
have
164
AX
IX. If X'
and
X'BX
In
Problem
4,
[CHAP. 21
we prove
(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
+ 72
+ Ay
+ Tn
\\BA\=0.
See also Problems
HERMITIAN MATRICES.
X.
4-5.
we have
See Problem
XI.
The
7.
If
U such
U'HU = U~^HU =
that
U'
HU
A,-
A.) is
r-
Then diag(Ai, A2
A).
...,
A, there
The matrix H
diag(Ai, A2,
be arranged so that Ai i A2 ^
The
totality of
i Ansuch diago-
nal matrices constitutes a canonical set for Hermitian matrices under unitary similarity.
There
follows
Two
XIV.
NORMAL MATRICES. An
diagonal,
and only
if
real
if
and only
if
symmetric,
real
is, if
if
skew-Hermitian, and
unitary matrices.
XV.
In
If
Problem
XVI.
mal
8,
is a
is
a.
U'AU
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,
See Problem
10.
CHAP.
211
MX.
If
r^
165
r^.
if
and only
if
SOLVED PROBLEMS
1.
The
Prove:
which
BX
is real
is a
{(h
+ ik)I-A\[(h-ik)I-A\
{h
+ ik)I
-A
A are
all real.
Consider
(hl-Af+k'^I
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
+ k^X'X
Also,
0.
X'X>0. Thus,
A:
and
2.
Prove:
A'l
As of 4.
AX^
and X^ be invariant vectors associated respectively with the distinct characteristic roots Xi and
Then
=
Ai^i
AX^
and
= Aj'^s.
A^2'4^i =
also
X-iX^X^
and
Xj^AX2 = XqXiX^
Taking transposes
^1 A X^ = A 1 A*! ^2
Then
3.
X^X^X^
and
X[X2
A2^iA^2
XqAX-j^ =
X2X2X1
= 0.
Find an orthogonal matrix P such that P'^AP 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
6, 6, 12.
- -
For A =
6,
we have
-1
Xl
-4
X2
-1
X3
or
0,
Xi
- 2x2 + xs
-l]' and
X2
= [l,
1, l]'.
When A
= 12,
in-
we take X3 =
166
1/VT
P
-i/x/T
is left as an exercise to
It
4.
Prove:
If
show
that
Xiyi + \2y2 +
\\B-A\ =0.
By Theorem
^'B^
^'^'^
^nTn
= diag(l/\/7H, X/sfJT^
l/yjl
6, 12).
...
GV
Then V =
BGH)W
G'AGH)W
(H
HW
carries
(i)
into
+ wt
X^y2 + X^y^ +
...
Y (K~ IK)Y
-'i
Since for
values of
all
KY
which
XK'h'g'BGHK - K'h'g'AGHK
=
=
Prom Problem
3,
+ Xy^
-'2
'n
diag(X,X
diag(X-Xi,
X)
- diag(Xi,X2
X-X2
X-X)
=0
\/^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
+ Ait;^
Y'(K'h'g'BGHK)Y
5.
into
where Xi.Xg
it
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
jjii.jj.^
X/\f&
I/VT
V'(G'BG)V
Let
we have
(i)
where
1/x/y
= (iiag(6,
nite,
Now
P~^AP
of
[CHAP. 21
X'BX =
X'
-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]
6.
CY
KY
167
A = CP
where C
diag(\/^,
VT^
C^
C'^AA'C'^^ C ^C'^C^
is orthogonal as required.
Prove:
pose
Then
C'^A.
The
QBQ-
QB^Q-^
and
= I
Then there
17.
A:^
>
Thus A
0.
Define
symmetric and
is positive definite
PP'
that
8.
QB^Q-^QB^Q-^
Now C
= QB-^Q''^.
Define
7.
and
\fk~^)
that
is positive definite
show
is left as an excersise to
It
= lY.
X'BX
{GH)Y
Prove:
Bi
AA'
is orthogonal.
CP
with C
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^ =
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^=
= (X^I
(W X ^)' {W X ^)
XA
XXI - \ a'
a' A
0;
Prove:
An
re-square matrix
Suppose A
is normal.
By
AA
-A)
then
and
B'X^
9.
(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
Now
Xfi-i
bn-i. n
is
...
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
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 =Xi^i,_I'A'2 = AaATa.
Now ^2/lA'i = Ai^a^i and,
AXi=\iXi, AX^
X[a' X^
11.
XiX2=
Xx iXo.,
x^ -
X[a' X^
But
= X^X'-^X^.
- 4.xi = 40
12xix.2.
X'AX
(i)
of
A-1
(A-5)(A + 8)
-2/\flf\
/v/T3
'VTs
The conic
The transformation
-6~]
40
::]
A is
|A/-^|
Now
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
to
(i)
13"!
Vis'
-\y
3/\/l3.
2/\/l3
3/VT3.
[3, -2]'
40
5yl
-Sj
Lo
is an hyperbola.
axes
in plane analytic
12.
One problem
i.e.,
we show here
Without
at-
"
and
characteristic equation of
1A/-.11
~.
-;=|
-1
-7
-1
-7
-9
is
A-3
1,
Ai =
The
A2
-1
-1
-1
-2
-1
-2
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 +
X + 2y
\^
The rank
of
is 3
0,
z =
C(-l,
or
4; the
The
x = x'
of translation are
di,
y = y',
1,
we have
D^,
111
4F
z = z'
2Z
find
0, 4).
- 4
4.
1-2. i^s-
we
d = -4.
equation is
The equations
Prom
0, 4).
0-1=0
+22-7=0
+1=0
y +
\_v\.v.2,v^.
\_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
SUPPLEMENTARY PROBLEMS
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
-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
/'"^^P = diag(Ai,
If
-^r + i'
'^1
Xi,
A.1
^1 ~^v)
~'^r+2
'^ ^
17.
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
+ /iri(/-4) is orthogonal.
20.
Prove:
If
Is
21.
Prove:
If
is normal, then
22. Prove:
x\ - 312%ia:>, + 17 %|
(c)
Prove:
B =
P"\Ai/ -/1)P
"^S"
^ri)'
'^r-i-2
each locus:
(o)
(c)
X^ + i-
''^"'^
16.
19.
[CHAP.
is similar to
and only
if
A~'^
if it
H + iK, where H
can be expressed as
and
commu-
are
23. If
roots of
ty =
for
Prove:
If
X2A2
^n^rv
where
are XiAi,
C/"^<4(7 =
Write
Hint.
24.
AA
J
T*
= [. ],
then
Xi, A,2
A.,
is unitary
and T is triangular.
_
C/
Now
tr(rT') =
tr(-4/l')
requires
is non-singular, then
AA'
is real
and non-singular.
25.
Prove:
If
if
P'^ AP
(/)
Define by
B.-
and Xj,
(j
= 1, 2
s) the
and
BA
are normal.
A be
(X-x^f^iX-x^f'^ ...a-'Kfs
<^(X)
^B
P such
that
diag(Ai/.ri' '^2-'r2
-^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'^AP
(6)
~-
AiBi
A282+
...
(c)
(d)
E^Ej
(e)
1 + 2 + + E3
=0
for
A5B,
^j.
=
(/)
The
(g)
(Xil-A)Ei
(h)
IfpM
Hint.
= 0,
(J
= 1,2
s)
Establish
^^
p(A)
Aii + A22 +
p(Ai)i + p(A2)2 +
xIe
+ P(^s^^s-
Aii + A22 +
+ X^E^.
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 non-singular, then
A-^
If
and
/^(A)
/(A)/(A-A^),
(i
= l,2
Then
s).
(I)
(m)
171
Al%
A-2%
+ A-^fig
...
A^'^
/f
vTii
VA22
...
VAs^s
Equation
(n)
27. (o)
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:
is non-singular, then
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
f/
= H~'^A.
is
normal
if
and only
if
H~^AH
and only
if
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"
Define
Hint.
35.
is unique.
A= HU
such that
31.
4/9
30.
49
20
--
29.
it
A-^
Obtain
9_
38/9
28.
that
-4/9
lO"
-10
196
(c)
Show
(b)
decomposition of A.
tf
real
A and
(6)
(6)
(c)
If
(d)
If
is real
is real
= tiA.
is unitary.
~iH)
(a)
I's.
(S')"^ commute,
r.
and
of
is unitary.
is
2i
of
U'^AU, where U
symmetric (Hermitian),
values
least and Ais the greatest characteristic root of A.
its field of
Prove:
real.
Ai i A < A, where Ai
is the
chapter 22
to be
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.
If
every
a =
oj;
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
If
(i)
f(X) is of degree
^(X) + g(A)
is of
degree
TO
when m.>n,
F[X\ as elements
of a
number sys-
For example
f(X)-g(X) = g(X)-f(X)
and
= g(X) + f(X)
n,
of degree at most
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
(22.2)
g(X) =
then
Problem
If
degree
/(A)
is either the
h{X)-g{X) +
r(A)
172
If
CHAP.
22]
Let
= h(X)-g(X).
/(A)
When
Example
1,
(22.3)
By
is free of A.
n.
When
(22.3),
f(a) =
/(A) is divided by
h(X)-(X-a) +
/(A)
r
if its
where
g(A.) is
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
is f(a).
m. A
GREATEST COMMON
DIVISOR.
If
X-
a as a factor
and only
if
and
if
f(a) =
0.
g(A), it is called a
common
divisor of
(i)
d(X) is monic,
(ii)
d{X) is a
(Hi)
common
/"(A)
and g(X)
if
Problem
2,
we prove
a unique greatest
common
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)
d(X) =
of
/"(A)
3.
1.
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
V.
not
1,
a (A) -/(A) +
b(X)- g(X)
and conversely.
divisor is
i.
See Problem
if
4.
their greatest
174
If
VIII.
and
UNIQUE FACTORIZATION.
g:(A)
/(A)
if
each divides
one of
/(A)
and
h(X).
soalsodoes /(A)-g(A).
A(A),
Problems, we prove
In
IX.
(22.5)
/(A)
where
is
/(A).
If /"(A)
[CHAP. 22
is a
9i(A)
?2(A)
^^(A)
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
+ a^A + Oq
and
/(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
r(X) = /i(A).
we have proved
~X
and
Otherwise, we form
/(A)
^A"-\(A) - ;^A^-\(A)
/,(A)
f^X) =0 or is of degree less than that of g(A), we have proved the theorem. Otherwise, we repeat
Since in each step, the degree of the remainder (assumed ^ 0) is reduced, we eventually reach
a remainder r(A) = 4(A) which is either the zero polynomial or is of degree less than that of g(A).
Again,
if
the process.
r(A)
h(X)-g(X)
and
+ r(X)
/(A)
k(X)-g(X)
'c(A)-g(A)
s(A)
Then
+ s(X)
and
[k(X) -h(X)]g(X)
Now r(A)-s(A)
is of
to or greater than m.
while, unless
=
0,
r(A)
A:(A)
s(A) =
r(X)
s(X)
h(X) = 0.
[k(X)
- A(A)]g(A)
is of degree equal
Then both
CHAP.
2.
22]
175
Prove:
If f(X) and g(A) are polynomials in F[X], not both zero, they have a unique greatest
divisor d(\) 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^
or is of
and k(X)
6^
or is of
it
or is of
rgCA) =
'2(A)
it
b'^^giX)
and we have
(a)
with
+ r^iX)
/(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,
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
?i(A)-g(A) +
/(A.)
where
we have
is the
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)
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)
's-2(A)
9s(A)-'-s-i(A)
&(A)?^0
r^cA),
and
(^^
's-i(A)
By
(vi), rjCA)
's-3(A)
Prom(l)
If
ri(A)
'2(A)
Prom
(iii),
/3(A)
'3(A)
Continuing,
d(X)
The proof
9s+i(A)-'-s(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),
= e~^'s(A).
A2(A)-/(A)
(iv).
r5_i(A)
then
c,
[l + 9i(A)-92(A)]g(A)
we obtain
Prom
-52(A). /(A)
=
t^(K)
9s-i(A)-'s-2(A)
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)
ksiX)-g(.X)
=
h(X)-f(X) + k(X)-g(X)
as required.
176
3.
[CHAP. 22
/(A)
We
and
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
7A + 10
(^A + p^)(17A+34)
(A'
6A+4) +
4A^ +
+ 7A+10)
(A^
and
(Iv)
A^
+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
(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
< m and
+ 14A^ - 17A
4)-g(A)
re>
is not 1,
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)
then
d(X)-g^(X)
g(A)
<m. Now
=
gi(A)-rf(A)-/i(A)
g(A)-/i(A)
and
gi(A)-/(A)
= gi(A)
we have
(a).
Then, using
[-/i(A)-g(A)]
k(X)-g(X)
(a) holds.
Then by Theorem IV
there exist
(a),
a(X)
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)
if (a)
CHAP.
5.
22]
177
/(A)
where
c-q^(X).q^a)
...
qr(\)
is a
,^
Write
(i)
/(A)
where a
theorem.
(")
and
/!(A)
a-/i(A)
If /^(A)
/(A)
If g(A)
is irreducible,
n-g(A)-A(A)
(ii)
To
"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^(\).
in numbering,
92(A) divides
we have
9i(A)
...
Ps(A)
= 1,2
SUPPLEMENTARY PROBLEMS
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
it
()
/(A)
2A^-A=+2A2-6A-4,
(6)
/(A)
A^-
(c)
/(A)
(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
('^)
III.
Ans.
11.
lo
jL(5A+2)/(A)
^^{-l5X^+4.A>?~55X+ib)sa)
then
g(A) = d{X)-h(X)
or
178
12.
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
Ans.
1,
[CHAP. 22
(A^- l)(A^ + A+
l.c.m.
1)
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 -
(A-c)^g(A).
k>
(a)
of /(A)
if
Prove:
Let
/I.
sA!^
- 9A -
51
by a polynomial domain?
Show
/(c) = 0.
and only
common
if c
if
An n-square matrix A
is normal it
k-l
of /'(A),
(b)
common
/(A) = s(X)-D(X).
'
Show
that c is a
a^A
in
Hint:
20.
A^
A -
(f)(A)
root of multiplicity
19.
and
c is a factor of /(A).
/(A) =
Suppose
- 9A -
g(A) = t(X)-D(X).
+ a^^^A
...
+ aiA + oqI
D (A)
and D(X)
= d(X).
c{X)-d(X).
chapter 23
Lambda
Matrices
mxn
non-zero
(23.1)
^(A)
[ij (A)]
oii(A)
ai2(A)
...
ai(A)
asiCA)
a22(A.)
...
CsnCA)
Omi(A)
a^^iX.)
...
aj^n(>^)
is called a A-matrix.
Ap\^
^(A)
mxn
sp-l
A.^X^-^
...
A,\
+ Ao
matrices over F.
A^+A
1.
A* + 2A= + 3A^ +
5]
'^(A)
.
A=
- 4
A^
- 3A=
J
1
A^ +
A +
-J
-4
L
zero.
The
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).
A(k)
Apk^
+ Ap_.^k^~^ +
179
...
A^k ^ Ao
LAMBDA MATRICES
180
However, when \
[CHAP. 23
We
Ag(C)
(23.5)
A^C^
. ^^
C^A^
.
C^
+ .
A^_,C^
p-i
to the
define
+ A,_C + Ao
...
and
4 (J)
(23.6)
A^
^p
+ ^^^i +
^o
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
Qi_(X), RxiX);
(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)
/?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
'
a^aJ
ii
A-iirA^+i
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
See Problem
is called scalar.
bqX"-!^ + 5,_,A'?"'-/ +
6(A)
(23.9)
trix
3.
B(A) =
...
+ b,X
fc(A)-/
+ bol^
polynomial.
If in
A(X)
(23.10)
6(A) = b(X)
4.
Let
then
Example
I.
2A
B(X)- Q^(X) +
A + ll
and
(A)
LA^^-I
2A+1J
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.
matrix over F.
(23.11)
A(\)
(2i(A)-(A/-S) +
A(\)
(A/-fi)-(?2(A) + R2
/?!
and
(23.12)
m.
If
A.
It
can be shown
re, until
is
ra-
''
A, are obtained,
then
Aj^(B)
A^B^
Ap_^B^~'
...
a,
Aj^(B)
B^Ap
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^Q^i^^-^^
Theorem
III.
'4(A)
/(A)-/
apIX^ + ap.jx''^ +
...
+ ai/A + Oq/
^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
CAYLEY-HAMILTON THEOREM.
trix
A/-B
= f(B).
KI^- B
if
and only
if
f(B) =
0.
(A/-i)-adj(A/-i)
(f>(X)-I
LAMBDA MATRICES
182
Then ^(A)
6.
(^(A) =
by XI - A and, by Theorem V,
/ is divisible
Example
[CHAP. 23
[oj,-]
of
Thus,
0.
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.
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
R^iX)
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)
Ri(X)
and
(ii)
If
p <
q,
then
(I)
and
where C(A)
If
is either zero or of
degree at most p -
q,
fii(A) = A(X).
4(A) - ApB-^B(X)X^''^
C(A)
1.
we have
ApB'q'-^''^
(i)
with
and
R^iX) =
C(A)
CHAP.
LAMBDA MATRICES
23]
C^X +
C(X) =
If
...
where
A(\)
If
s>
q.
form
- ApB-^''B(\)\P-1 -
Qi(\)
we continue
otherwise,
of decreasing degrees,
and we have
To
the process.
we
q,
C^S-1B(A)X^-'?
we have
and
(II),
begin with
- B(X)B-^''ApXP~1
rA^ +
and
A + A'
A"
A^ +
A(X)
left
1,
Chapter 22.
A'-A-l]
2A=-l
^(A)
|_
L
(a)
A=
A"
(?2(A),
S(A)
L-A^ + 2
A^
- A
= Qi(X)-B(X) + Ri(X).
= S(A)-(?2(A) + R^iX)
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
2X +
+ 4
6A +
A^ +
[CHAP. 23
2I
3A + 5J
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.
,^n
of degree p in
x:.
40
-8
Prove that
\h(A)\
= h(\)- h(X2)
h(X^).
We have
(i)
\XI-A\
(A-Ai)(A-Ao)...(A-A^)
Let
(ii)
h(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)!
...
...
(Sp
...
ic(si-A)(s2-A)
...
(s^-A)!
...
-X)!
h(\i_)h(\2) ...h(\^)
(ii).
SUPPLEMENTARY PROBLEMS
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
^i(C)-B^(C)
B^(C)-A^(C)
E
3-
^i(C)
[3
-7J
[::;]
:]
--M
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
LAMBDA MATRICES
186
9.
|> + 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
17A -
-21A + 4
-2A+3
loA +
- 2A -
2A^-4A
-A +
llA +
4A - 4
2A +
16A - 16
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.
(?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)
A^'+SA^+sA
A 1
A
A+lJ
A +
B(A)- C(A)
Fa^'+SA^-VA-
12.
[A - 3
^A
=1fx
C(X)
''*
+ 2j
(a)
.4(A) =
187
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,
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
g(B))
diag(g(Si), g(B2)
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.
20.
Prove:
21.
Let A(t)
If
is an invariant vector of
= [a^,-()]
where the
A of Problem
17, then
polynomials
\\I
and
(xj-Ai)(::i-A2)...(*^-A)
if
= 0.
-1
Take
t.
A(t)
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)
^A
dt
'it).
n
Hint.
use
and differentiate
cv,-()
a,-i,(0 6i,,-().
k=i ^^
"^^
For
(d),
chapter 24
Smith Normal Form
BY AN ELEMENTARY TRANSFORMATION
(7)
The interchange
The multiplication
ff|--
The addition
to the ith
/(X),
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
is an elementary
ma-
trix in F[\].
II.
III.
Two
If
\A(\)\ = k
The rank
0,
re-square X-matrices i(X) and B(X) with elements in F[X] are called equivalent pro-
ff 2
'
B(X)
(24.1)
...
K^ such
that
P(\)-A(X)-Qa)
Thus,
IV.
Equivalent
In
mxn
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
188
to the
CHAP.
24]
189
A (A)
/2(A)
N(\)
(24.2)
...
where each
monic and
/^(A) is
of rank
(i
0_
1, 2, ...
1).
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)
= 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)
h^iX),
(24.3),
gs(A)
Now
we
go(A) =
define
(24.4)
Ai(A)-A2(A)-...-As(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.
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,
+ 3
2A%3A'^+A
3A^ +
6A
common divisor
A(X) is
A^+A'^ + A
for
common
+ 3
=X,
Then, by (24.4),
/2(A)
g2(A)/gi(A)
/3(A)
A,
g3(A)/g2(A)
A^+A
10
A
A'(A)
A^+A
For another reduction, see Problem
polynomials
4.
f^(X),f-2{X)
/V(A)
As
we have
1,
k r,
then
/i(A) = /2(A) =
...
190
VIII.
Two
[CHAP. 24
re-square X-matrices over F[X] are equivalent over F[X\ if and only if they
ELEMENTARY
be expressed as
(24.5)
{p,(X)\'^i^\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
of A(X).
Example
2.
field
10
10
(A.-1)(A^+1)
'
(A-1)(A^ + 1)^A
(A-l)^(X" + l)^A^(A"-3)
The rank
is 5.
The invariant
=
/i(A)
(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)^,
Example
3.
(a)
Over the
it
A^
(A'^+l),
distinct;
A,
A""
- 3
in the listing
each ele-
Example
2 are
ele-
since
(6)
A^-3
A-1,
A-1,
(A^lf,
(A^'
l)''
(A^'
l),
X',
A,
X+x/l
X-\fz,
can be factored.
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.
A"",
A,
(A-l)^
(A-^)^
A-l,
be
(A+l)^
A+
Find the invariant factors and write the Smith canonical form.
To form
a" (A-1)'' (A
if
i.e.,
CHAP.
24]
To form
f^i\),
191
and
common multiple
UiX.)
Repeating, /sCX) =
Now
A(X-l).
A^(X-lf(A+l)
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,
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
Its effect
or S(X) = kR(X).
on ^(A) is either
S(X)
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)
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
for B'(A)
the greatest
common
divisor of all s-
192
3.
Prove:
of rank
[a^j(\)]
[CHAP. 24
/2(A)
N(X)
where each
fi(\) is
L(A)
monic and
(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)
Then by transformations
of type
3,
/i(A)
(i)
B(A)
an (A).
By Theorem
ay (A)
where
ri^-(A) is
I,
Let
ai--(A)
Prom the
/th
(l,l)-position as the
new oii(A).
If
row
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
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
first
not divisible by
On (A).
Now
row
we proceed
to obtain
(i).
Otherwise, suppose
a,-,-(A)
is
"J
Let
aii(X) - qix(X)- a^^iX) 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)
^ij(X) +
it
gij(A)U- ?ii(A)!aii(A)
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
we
(i).
/2(A)
C(A)
Ultimately,
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
CHAP.
4.
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""
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
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
A+A
2A^ + 2A
2A^ + 2A
A" + A
5.
Reduce
A =
-4(A)
A""
+ A
A + 2
A^
A''
- 3A +
A=- 2A
2A
A= + A -
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+
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);
SUPPLEMENTARY PROBLEMS
6.
Show
7.
Prove:
that
H.jK^j
H^{k)K^a/k)
constant.
H^j{f{\))
Kj^{-f(,X))
/.
and only
if |.4(X)| is
a non-zero
194
8.
Prove:
to / by
[CHAP. 24
elementary transformations
if
is a non-zero constant.
9.
Prove:
if
and only
if /1(A) is
a product of
elementary matrices.
10.
A(Xr^
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.
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"
A^A""
+ 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
3A + 2
\^
a""
A= + 3
(e)
12.
'-Xj
- A -
a""
if)
'Xy
2A-2
A+1
A% A
- A=
A+
A^ + 2
A^
-A^A
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
for
each of the
CHAP.
13.
24]
195
Find
its
real field.
(a)
(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,
(A^+lf,
l,
-if,
(A
(A
-if
A-1
(A^'+lf,
l,
+A
A-1
-if, A-1,
A^'
A''
(A^
A+1
+ lf.
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,
(A+2f
(a)
1,
1,
1,
1,
A,
(b)
1,
1,
1,
A,
A^cA-l),
(c)
1,
1,
A-1, (A-
(d)
1,
1,
1,
A^A-lf
A(A + 2f,
A''(A
+ 2f,
A''(A
2f
!Dxx
+ (D + i)x2
(0 + 2)^1
(D + l)x2
where
xj, x^,
unknown
xs are
the polynomials in
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
+1).
Hs^i-kD),
obtain
k(D +
5D + 6
e*
-1
PAQ
at
D+1
:i2(-l),
(D+ 2)X2
system is
AX
Now
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
Use the
linear transformation
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
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
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
Prom Theorems
HI.
An
and
II,
we have
re-square matrix
A over F
\I- A
invariant factors of \I -
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-
j^a))
\P(X)\.\Qa)\4,a)
is the
A(A)-/2(A)-.../na).
and we have
is the
iant factors of
If
(II)
If
(III)
If
4 =
A ^
oq/,
then m{X) =
A^ ^ aA + hi
X-ao:
i^ = a^A +
for all a
aj
and b but
're(A)
A''
A^
then
m(A) =
a^A"^
a^)?
122
1.
196
- a^X - uq
+ a^A + a^I
- 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
'100"
'122"
=
oo
1
1
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,
f^(X), f^(k)
of
Set
we prove
2,
oqI
197
(fi(\)
ra-square matrix
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.
NON-DEROGATORY MATRICES. An
re-square matrix
A whose
An
re-square matrix
is
otherwise, derogatory.
non-derogatory
if
and only
if
minimum
We have
A has
similarity invariant.
It
is also
X.
If
easy
to
show
polynomial
and m2(A).
is the least
Let gi(X),g2(X),
B = diag(^i, ^2
acteristic and
We
^n)
has
matrices.
^(A)
fgi(A)!
ig2(A)P-ig^(A)!
>
= 1,2,
'^
^(A)
C(g)
re
(;
,02
a" +
a-iA"'^
and for
multiple of ^^(A)
minimum polynomial.
g(X)
(25-2)
common
..gnCA)
/rei(A)
= [-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
XII.
istic
its character-
See Problem
XIII.
If /4 is
/ =
(25.4)
[o],
if
ra
and
1,
...
if
ra
>
...
f^{\)
then
O'
...
...
has
5.
easy to show
is
It
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
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)
199
left
member
R^(\I-A)
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)
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
/ =
Prove:
only
If
is
R2R3
Thus, Rg = R^
/?i(A/-^)/?2
F and
if
By
(vii) is of
degree zero
A divides
in
ARi2 - R^AR^
R^: then XI - B
/Ji =
RxUXl - A)
XI - B
2.
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.
gri-i(X)
\XI-A\
cf,(X)
(n
- l)-square minors
g-i(A)-/(A)
and
adj(A/-/l)
g_i(A)-S(A)
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
0.
of XI
-A. Then
200
By Theorem
V, m(\) divides
Suppose
fn(X-)-
(ii)
fn(\)
Since m(A) =
0,
XI
-A
[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)
and
Now
=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.
its
characteristic and
of (25.3) is
-1
-1
^1
"2
an-2
To
A + a^_j
A times the
third column,
,n-l
-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-
5.
minimum polynomial
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
SUPPLEMENTARY PROBLEMS
6.
- 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.
8.
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)
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
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
(A-3)
2)'';
A^-A
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)
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.
VIII.
Prove Theorem X.
Hint.
m(D)
= diag(m(Bi), m(B^)
requires
m(Si) = miB^) =
m^X)
divide m(X).
the
202
12.
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
Hint.
Prove:
is the
P and Q such
matrices
16.
PCQ
that
or the
Problem
in
18.
F
it
with C and
and only
if
non-singular.
R(X) = \C - A
and
S(A.)
XD-B
have the
1,
s,
slmA.
of
Problem
9(h).
20.
PDQ
1.
1.
If
The theorem follows from Theorem VII or assume the contrary and
Then {A -\^I)q(A) + rl =
and A -X^l has an inverse.
Hint.
19.
= A.
or
A:.
17.
[CHAP. 25
^ 0.
Use
-4
Prove:
If
^
to
[o
show
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.
r,
;]
minimum polynomial
that the
m(A) = {\-X^)q{X) +
w:rite
g(.4) is
is
expressible as a polynomial in A of
22.
Prove:
If
the
s in A,
then the
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.
B.m(^S)./l
Let A be of dimension
at
.A
(B,4).m(B.4)
mxn
Let
X.I
and
A-n{BA)-B
m >
n,
25.
AB
and BA.
Show 0(A)
(AB)-n(AB)
Xj^ is
0.
mute, then
AB
Prove:
If
A and B com-
an invariant vector of B.
of
chapter 26
Canonical Forms Under Similarity
THE PROBLEM.
Chapter 25
In
matrices A and
In this chapter,
These matrices,
ical matrix
(ii)
put into view either the invariant factors or the elementary divisors.
'[^
o]
all
mxn
They correspond
matrices of rank
to the
canon-
under equivalence.
<26-l)
(1, 1
1,
f^(\.)
f.(X), f.^^iX),
...,
of degree s^,
(i
(26-2)
To show
D^ = diag
is
f^(X))
= j.j +
n).
We define as the
C(/))
that
l,/i(A))
(1, 1
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
f^iX))
We have proved
I.
Every square matrix A is similar to the direct sum (26.2) of the companion matrices
Example
1.
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.
fi(\)
ipi(X)l^"!p2(A)!^2^..{p,(X)!'^is(pi(A)i'"''iP2(A)i'"''...iPi(
(i
= i,i+'^
dis-
n)
where not every factor need appear since some of the ^'s may be zero.
C{p?^^) of any factor present has
fac-
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.
A^-A,+
l,
(X
-A+1)
1,
.
[-1],
[-1],
[-:
.;]
[.:
\
1
II is
-3
A+1, X+1,
CHAP.
26]
205
-1
-1
1
-1 -2
1
-1
1
1
1
1
-1
-3
If
q =
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
..
Note that
in (26.4), there is a
continuous line of
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
shown
to
[CHAP. 26
C(p^) and
may be substituted
for it in the
II.
We have
Every square matrix A over F
III.
Example
4.
2,
of
\I - A.
1,
(A+l)^is
r-1
10
-11
and that of (A^-A + 1)^ is
0-11
nils
-1
-1
-1
-1
1
-1
1
1
-1
1
1
-1
The use
what misleading.
It
was used
Theorem
is
some-
Theorem
III
is
(26.5)
..
1
1
C^(P)
..
.
!p(A)!
1
<^.,
(A-a^)^
III is
2.
or classical
Let
Example
5.
-A
be:
A-i, X + i, (X-i)
(X+i)
CHAP.
26]
of
207
is
-1
1
-i
-i
An
re-square matrix
and only
divisors oi XI -
if
the elementary
minimum polynomial of
See Problems 2-
4.
will be
shown
We
in
to
Theo-
is
made
in
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
The vectors
Then
(A
Y;
thus,
and
A"
degree
p,
as its leader.
AX
-I)X=0
=[-1,0,1]'
= [2,
1, l]'
(A+I)Y=0
vector X.
non-singular matrix
(26-6)
R'^AR
diag(C.,C.^^
C)
A^i
c^ ^_X'i-
c,.
We
A
-^i2
1
-^i3
Ci
.
shall
assume
that
C,-
the
208
Prom
(26.6),
we have
AR
(26.7)
RS
Rdiag(Cj,Cj^^
R-,R-^
AR
A[R..R.^^
[CHAP. 26
From
RJ
R^
C)
so that R^
n)
(26.7),
IRj.Rj,,
iRjCj.Rj^^C.^^
C)
/?Jdiag(C,,..C^.^^
R,,CJ
and
AR^ = R^Ci.
Denote the
si
(i=j,j +
n)
RiCi
[Rii, Ri2,
, Risi\Ci
st
[Ri2,Ri3, ,Risj,> -
Rik^ik]
^__
Since
ARi
A[Ri^,Ri^, -.Risi]
[ARi^.ARi2,. ..ARisil
RiCi
we have
ARn,
Ri2 =
(26.8)
Risi = A^'-'Ri.
and
-|>i.%
(26.9)
^Ris,
we obtain
-i=
A'iRu
c^kA^-'Ri^
l
or
(A^^ + cis.A^'i'^
(26.10)
Prom the
(26.11)
...
may be
a^A
+ 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 f^iX).
Thus, the column vectors of fi^ consist of the vectors of the chain having X^, belonging to /^(X),
Xj^
as leader.
To summarize:
chains
Xi,AXi
whose leaders belong
A^^"'Xi
(i=j,i +
<
s,-
s- ^^
+1
...
/ (A),
n)
fj.AX.)
/^(X)
^ s.
n
We have
VI.
let
X^ be
(it)
let
(iii)
let
X^^^he
is linearly
F which
and so on.
Then,
for
^_^
maximum length
for all
ra-
vectors over F\
of
and_i;
of
of a chain
vectors over
F which
CHAP.
26]
AR
Example
7.
209
Let
4X
= [l,
Independent while
arly
belongs to
/gCX) =
= Q
and
A'
15
6
we
4A',4=A']
[a;.
^^A: =
{A-bA^+I)X
Thus,
1, l]',
find
-3
6-5
-1
AR
[AX.A'^X^a'^X]
14
25
30
0-1
and
R ^AR
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.
Example
8.
Let
Tak
A"
then ^ A' =
= [l,-l,o]';
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
Y, -4? = [2,
= 5A'^Y +
The polynomial
AT
R'^AR =
1, 2]',
SAY -77.
^^^ =[ 11,
8, s]'
are linearly
Thus, Y belongs
m(A) =
to
in fact,
unless
minimum
R'^AR
11
when
[Y,AY,A'^Y]
See Problems
5-6.
SOLVED PROBLEMS
Prove:
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-
l.f^(\).
=
But /^(A) =
|A/-C^(p)|
ip(A)!'?
since
\XI-C(pf
jp(A)!9
210
2.
(a) is that of
divisor being
1.
-1
(b)
(a)
-1
-4
-1
[CHAP. 26
(a) is that of
-6
-1
-4
Theorem
2,
I,
X+
2,
X+
2,
X-
2,
X-
X+
2,
3.
-2
-2
4
(b)
(a)
12
4.
2,
X+
2,
(X'^
-3
-3
(a) is that of
+2X-lf
(X^
(X + 2)(X^ + 2X -
of
Theorem
is (b)
2)(X^+2X-1)^
(X +
1)^
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
-4
A"
-Z;
-ir.
A''
belongs to
=[-3.1.1,1.
A*-2A^ + 1
1,2]'
We
tentatively
assume
Xg and
The vector
AY
- [-1,0,
Y belongs
Xs
1,
to
A -1.
When
for Y.
-1
[Xs.AXs.Xg.AXe.A^Xe.A^'Xfi
1
-2
-3
"o
-1
-1
-1
write
-1
R~'^AR 1
so that
we
2
1
Note.
and
AZ
The vector Z =
- [3, 0, -2,
1,
-2,
o]' is
linearly independent of
-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^.
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.
SUPPLEMENTARY PROBLEMS
7.
Can any
9,
of these matrices be
I,
n, in
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,
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"
to
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,
a
/S
/3
(b)
X^+
1,
214
11.
Hint.
12.
In
Example
show
6,
X,AX. and Y
that
m(A)
A (A)
g (A) +
"
[CHAP. 26
r (X).
form.
13.
In
Problem
6:
Take 7
(a)
= [o,
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.
X^ = Z
-X^
1.
15.
AR
linearly independent of
1, 0, O]',
(a) to build R.
25, find
R such
R' AR
that
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
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
dY
RY
Xq
t.
dxA'
dxn
-jf. -jf,
^J
AX
carries
R ^ARY +
(i)
into
R^H
dt
choose R so that R~
AR
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 -
Cae*-
5(C3+C4)e"* -
4 + 2
6t
- 4
2+3{-2_
INDEX
Absolute value of a complex number, 110
Addition
of matrices,
2,
of vectors, 67
determinant of, 49
inverse from, 55
rank of, 50
Algebraic complement, 24
Anti-commutative matrices, 11
Associative laws for
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
for addition, 2
for multiplication, 3
Congruent matrices, 115
Canonical form
Conjugate
of a complex number, 12
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
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
Image
1
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
75
system of homogeneous, 78
system of non-homogeneous, 77
Equivalence relation, 9
Equivalent
bilinear forms, 126
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
equal, 2
102, 111
divisor, 173
equivalent, 40
over a
field,
65
INDEX
Matrices (cont.)
product of, 3
Null space, 87
Nullity, 87
sum
of, 2
Matrix
derogatory, 197
diagonable, 157
diagonal, 10
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
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
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
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
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
Rank
of adjoint, 50
of bilinear form, 125
of Hermitian form, 146
of matrix, 39
of product, 43
134
INDEX
218
Right divisor, 180
Right inverse, 63
Root
of polynomial, 178
of scalar matrix polynomial, 187
Row
(cont.)
deiinition of, 12
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 form, 147
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
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)
[^i;]
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
\^ij\
A. A,-
149
23
<^(A)
149
a,.
23
E^, (matrix)
170
39
/(A)
172
h-^ij
39
F[A]
172
Hi(k), K^ik)
39
^(A)
H^.{k). K.j(k)
39
-V.
40
^ii'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
^A
43
87
219