Professional Documents
Culture Documents
Eigen
Eigen
T (av) aT (v)
a =
(av).
T (av) A
=
s a m e eigen value
Thus, also a n eigen
tor of T
vector longing to
belonging to the same
value A is called
is to the same eigen
This shows that a v v e c t o r s belonging o
belonging
ofV.
eigen this eigen space is
that this a subspace
.A. The set
set of
of all such eigen
all such shall ghow that
all show
eigen spacec e of A.
Later o n
we
is called matric
(315)
316 Linear Algebra
Au -àjul =0
(A -Al) u=0, u #0
A-Al is singular
|s4 - ! |=0
a t i s f i e s the equation | A - I | = 0.
To prove that
A -N|=0.
A1 is a latent root of A.
Egn lialawes n i Eigen lecnors 317
By assumption. / A - A,li =0.
of A
corresponding to the latent value 1. The space spanned these
space and this vector is defined as
by vectors is a vector
space
of Acorresponding to the eigen value characteristic subspace or eigen space
11.5. Eigen space.
A.
Definition. Let T be a linear operator on vector
matrix representation of T relative to a space V (K). Let A be the
that 3v# 0 e Vs.t. T (v) lv. any basis of V. Let
à be an eigen value of T so
=
318 Linear Algebra
The eigen space of , denoted by Va, is defined as
VvEV: T(v) =Av
V is a subspace of V (K). (Refer Thcorem 11)
Theorem 2. (Cayley Hamilton Theorem) (Meerut 1997)
Bvery square matrix satisfies its own characteristic equation.
An alternate form of the theorem. Every matrix is a zero of its characteristic
polynomial.
Proof. Let A be a square matrix of order n, I an identity matrix of order n, x an
indeterminate. Then | A xl | 0 is the characteristic equation of A.
- =
Let A =lain
xn
a-X d12 a13 dn
d21 a22 a23 12n
A-xl |=
an1 n2 n3 n-x
= (- 1y" {x"+p*+Po-+... +P,), say
a+a, - 1** + a-2 +... +ag, say
Thus | A xl | is expressible as
-
degree
1 (at most), each cofactor
ordinary
element of adj (A xl) is an polynomial
in x of degree n 1 at most. Hence each -
=
I (40 t ax + ag t
Comparing the coefficients of the like powers ofx on both a,.
+ ...
sides
ABg d,
AB,-Bo a,
AB,- B = (dg,
''*****'**''*'''''**''
AB,1B,, 2 1
alues and Eigen Vectors
Eigu
319
- B-1G,.
Premultiplying these equations successivelv by 1, A, A2,AS, ...,A" and then
adding, we get
0= ag +aA Proved
+
11.6. Expression for the inverse of a non-singular matrix. Let a +agA2+ ..
GyA"
aan-singular matrix of order n | A |#0. Then
so that
A-xl | =4o +ajt +a +a, =0 ... (1)
This is true for every : and hence in particular for x = 0, so that (3) gives
For | A | 0 ,
A-01 | =a0 or | A|=ag or ag#0
So we can divide (2) by ag.
positive if
i) A A* numbers, not
all zero, we have
s e q u e n c e (x) of real
(1) For every
>0
numbers,
field is complex
be removed. If the from the
statement.
then bar may may be dropped
If field is of reals, condition (i)
and therefore,
the
then (ii) = (i) ofA. positive matrix
is
diagonal of a
Here A* conjugate transpose
=
on the
main
. (1)
Now supposo
zero.
scalars,
not all becomes
,
are n
dx= 1.
anda Now (1)
wnere
1 , X2, . . . ,
= 0 v j e x c e p t . j
= i
|'>0
0 or aj | x;
j
320 Linear Algebra
or
a>0a8z~| z |2
From this we conclude that
au>0-vi-1,2,..n.
Hence each entry on the main diagonal ofa positive matrix in positive.
Ex. 3. Find the chaucteristic equation of the matrix
-1
A-
1
.. (2)
2
A=-1 -1 -5 5
6-5
-1 -5 6
6 -5
A-AA =-5 5
6-5
5 6
22
-21
-21
22 -21
22
22-21 22
L.H.S. of (2)
=A3 -6A+9A -41
22 -21 22 -36 30 -30 -9 9
-21 22 -21 30 -36 18 -9
22 -21 22 -30 30 -36 -9 18
0 0
0 4 0
0 -4
0
000-0-R.H.S. of (2).
0 0 Proved.
Problem 4. Determine A for the matrix A
Solution. By Ex. 3, given in Ex. 3.
A6A 9A 4=0,
Eign
Values and Eigen Vectors 321
I-a-6A?.94)
-
6A + 9)
Ans.
roots of a matru
Theorem 3. The eigen vectors corresponding to distinct eigen
are linearly independent.
to distinct roots
Proof. Let uj, u2 Un be eigen vectors of Acorres ponding
...,
Then
A. Ag ,
A
Au, = \u,r=1, 2, ..,n . (1)
*a*r, s, s.t. r * s
set.
To prove that uj, ug, u form a linearly independent . (2)
=0, a1, Gz a, beingscalars.
a11 ta2l42t. + a,4n
...,
Let
A and then using (1), .. (3)
Premultiplying (2) by
1Ai t a2A2l42 t.+Gadnta =
obtain
by A and using (1), we
premultiplying this (4)
Again t... +aiba =0
a1Asu1 t a 2 ua
Repeating this process,
.. (5)
*****
uat
a a u +aza4 and (5) are equivalent to
the single matrix
that
Then the last says
is non-singular.
coefficient matrix
that the
(1) implies
agz
'
This =
0;
r= 1,
a,
322 Linear Algebra
Thus we have shown that every relation of the type (2) implies that a, = 0;
r 1,2,., n.
Consequently the required result follows.
11.7. Annihilating polynomials.
Definition. Suppose 7T is a linear operator over a finite dimensional vector
space V (P). Let f (*) be a polynomial over a field F. Let A be n xn matrix over F.
ffT)=0, then we #ay that the polynomialf (r) annihilates the linear operator
T.
f A)0, then we say that the polynomial f () annihilates the matrix A.
Example. Every linear operator Ton an n-dimensional vector space satisfies
its charucteristic equation.
11.8. Monic Polynomial.
Definition. A polynomial f(x) over a field Fis called monic polynomial if the
coefficient of highest degree term in it is unity.
Examples. (i) x"- 6x* + Tx 11 is a monic polynomial
-
or
m-6=0
mbmfor m=1,2, .., n.
Thisf=g. Hence the uniqueness.
Theorem 5. Let T be
the diagonalizable
distinct characteristic values
a
T.
linear operator
and let c1,
polynomial of Then the minimal Co ., Ch be
polynomial for T is the
p () =(x -c) *-
c2)...-ca).
Values and Eigen Vecton
Eign
Solution. c1. C2 Ch are the characteristic roots of T (1)
p ) (x -c) (x -e)..(
( )
)-0
)
characteristic equation of T. ch
Hfence each
is the for T. FHence e
characteristic root of T is a root of minimal polynomial
polynomial
Also minimnl
ominimal
X -cCk 18 an factorof
is
factor of
-
This (T-Ch) u =0
I(T- c) (T- co)... (T- ch)u =0,
I(T-c1) (T-cg).. (T- ch),.. (T-
clu =0
=0.
l p (T)u =0, by (1) >p (T)
Hence the result
follows. What is the minima
dimensional, vector space.
finite the
Problem 5. Let V be minimal polynomial for
a
the
the identity operator on V? What
is
polynomial for
zero operator ? x - 1 annihilates
polynomial
0. Therefore the monic
Solution. Since I - 1/ =
I. Consequently
the identity operator lowest degree that annihilates
Also x - 1 is a
polynomial of
annihilates the
zero
polynomial for 1.
r- 1 is minimal w e s e e that the
monic polynomial
x
O. Hence
Part. Again that
annihilates
Second of lowest degree
and first is the polynomial
operator O O. What is the
for F.
minimal polynomial
over
vector space
x is the n-dimensional on V?
6. Let V be a operator
Problem on V, (ii) z e r o
the identity operator
polynomial of (i)
characteristic basis for V.
S o l u t i o n . Let B
be any ordered
I unit matrix.
(i) Then lg
=
=
ofI is
polynomial
The characteristic 0
(1 = (1 - x)"
0
(1-x)
det ( - x l ) =
0 0 (1-x) Ans.
matrix.
O lg = 0 = zero
(i) det ( 0 - x )
characteristic
polynomial of Ois
The
0
O () Ans.
vector space
V and
(-1"x". ndimensional
= (- x"= an
linear operator on7 n e n vector a e V is aneigen v e c t o r
T be a a
B is a n eigen
Theorem 6. Let
an
ordered
basta B.
ordinute
vector X relative to
matrix ofTw.rt
A be the
co
its
value c iff
relative to its eigen
c
value
of T
corresponding
to its eigen
Dector ofA
Proof. We are given
324
ordered basis of V (F)
(i) [T]R = A, where B =l01..., a) is an
(ii) T(u)= cu,
where X = (b1, b2 6,).
(ii) X=lulg and sou
=
b;G
Aim. T iff AX = cX
(u) = cu
. (1)
IT-cll= [Tlg - c Mg =A -cl
Using the fact that T (w)g = 7lg lulg (Refer Theorem 2, Chapter 6)
IT-cllg lulp = (T-ch) ulg
We get
(A -c) X =(7 (u)
-
cl (u))g = {T (u) -
cu))g
or
AX-cX = {T (u) - cul
or
AX-cX= 0 iff T (u) - cu = (
This proves that
AX= cX
or
T (u) = cu.
iff
Theorem 7. Similar matrices A and B have the same characteristic polynomial
and hence the same eigen values. if X is an eigen vector of A corresponding to the
where B = P'AP
Proof. (i) A and B are similar matrices
3 non--singular matrix P
s.t. B P'AP
B -xl P'AP - xP-lp
= PAP-PxIP=P (A - x) P
det (B-x) = det [Pl (A - x) Pl
=
(det P) det (A -x) det P =
(detPA) (det P) det (A -
xl
But (detP)(det P= det (PlP) = detI=1
det (B-xl) = det (A -xl).
This shows that A and B have the same characteristic polynomials.
ii) Xis an eigen vector of A relative to eigen value c
AX=cX.
Observe that B(PX)=(PAP) (Px)
(PA) (PP)X= (P'A) IX
= Pl(AX)= Pl (¢X) =cP-lx
OT B (Plx)=c (PX)
P X is eigen vector of B relative to eigen value c.
Theorem 8. Let T be a linear operator on a finite dimensional vector
Then the following are equivalent: space V.
(i) cis a characteristic value of T
(i) The operator T- cl is singular (non-invertible)
(ii) det (T- cl) = 0.
Proof. To prove (i) ^(ii)
c is a characteristic value ofT
non-zero vector (characteristic vector)
Values and Eigen Vectors 325
Eign
uE
Vs.t. T (u) =
cu.
T (u) - cl (u) = 0
u # 0.
with u * 0o.
Now (T-c) (u) 0 = or T (u) = cu
8-x 6
7-x -+182 45x.
A- 2 3-x
Then | A - xl | = 0 gives
+18- 45x =0
- 18+45x =0
or
x ( - 18r +45) =0
or
3) =
0, giving z =
0, 3, 15.
x(x- 15) (x
-
or
3, 15.
Eigen values a r e 0, u.
equation Au
=
6
8
or
8x 6y + 22 =0,
This -6x+ 7y - 42 =0,
2x 4y + 32 =0.
On
solving any
two, we get 2 2 * 1
326
Linear Algebra
Then
is an eigen vector corresponding to the eigen value l =
0, k, being any non-zero real
number. Next consider the case in which^ =3.
Then (1) becomes (A 3) u = 0
8-3 -6
or
-6 7-3
2
This is equivalent to 5x-6y + 22 = 0,
-6x +4y 4z 0, -
=
2x-4y + Oz = 0.
This gives b (say).
2
Then u =b 1
-2
is an eigen vector corresponding to the eigen value ^
number. =
3;b being any non-zero real
Similarly u=c
is an eigen vector corresponding to the eigen value^
Thus the =
15.
subspaces spanned by the vectors
Au ru, =
=
0
and Elgen ectors
Eign alueN 327
o
which is oquivalent to 2y + 32 0, y + 3z = 0,z = 0.
Solving. we get z 0, y = 0, x = a.
1
a
0
is an eigen vector belonging to the eigen value A 1, a being any non-zero number. =
2
i.e
0
This is equivalent to -* + 2y +3z = 0, 32 =0.
On solving, we get = 0 , x =2y,
corresponding to two
distinct eigen values
the eigen spaces of A
are
separately
of the
=0, 2. 0 gives a system of the equations
Remark. In case =
2, if (A - 21)u =
0. ..(2)
type 0, x +2y +2=
+2y +z =
0, x+ 2y +2 =
= 0. .. (3)
or of the type +2 0, -2x + y - z =
4x 2y + - 22 0, 2x-y
=
n this case =
=
-1, z =
1; x =
1, y =0,z
*=1,y
Hence
- value A
=
2,
eigen to the vectors
to the es corresponding
is an eigen vector belonging spaces
three eigen
In this w e have
case
2 f a =2,2, 2.
328
Linear Algebra
In this case we have again three eigen spaces, corresponding to the eigen vectors
, if A= 2, 2, 2
0 0 0
This is possible only if there exists a basis lv1, U2, .., Up} of V s.t.
T
(01) aj'1, T (V2) agU2s, T (0,) G,"n
= = =
values a, a2,
Under such circumstances, T is to be diagonalizable operator. Hence a
diagonalizable operator is defined as follows:
"IfTis a linear operator on a finite dimensional vector space V, then Tis said
tobe diagonalizable operator if 3 a basis B for V s.t.
each vector of B is an eigen
vector of T" If V is n-dimensional, then T is
vectors.
diagonalizable iff T has n L.I. eigen
11.11. Diagonalizable Matrix.
Definition. A matrix A over a field Fis said to be
to a diagonal matrix D over F. It means that A is diagonalizable if it is similar
matrix S such that diagonalizable if3 a non-singular
S AS diagonal matrix
Also A is then said to D.
diagonalize
Theorem 9. A neces8sary and
D or to transform A to diagonal form.
be sufficient condition that an
diagonalizable is that Ahasn linearly independent eigen vectors matrixA over F n xn
in Va ().
Proof. I. Suppose A is
To prove that A has n diagonalizable
n xn matrix over F.
Then it can be verified that the eigen values ofD are precisely its diagonnl
elements d1. da, . d. But A is similar to D. Hence A also has the aame eigen valuen
d,. da d
Pis non-singular = | P |*0 = columns of P are L.I.
Take Pi. Pa Pn as the columns of P.
Then P1,Pas , are L.l. Also (1) =
d d
0
0
A
IP1,Pa., PAl P =
. . .
P,l
0 d
Equating columns on both sides,
AP, = d,P, (for i = 1, 2,..., n). . (2)
Now (2) declares that P is an eigen vector A corresponding to the eigen value
But we have already shown that (P1. P2 Pl is L.I. Thus A has n L.l. eigen
d.
vectors P, P2 .., P, belonging to V, (F).
II. Conversely suppose that A has n L.I. eigen vectors X1, X2, ..., Xn
that
corresponding to the eigen values A. 2,., n
so
AX = ; Xi, (3)
vectors
is the matrix formed by column
X] i.e.,X
Take X =
X1, X2, ..,
An
X BAX =
ndotnt eigon
vectors ofT Hence.
az,..., G
aren
Dheaty
*
Itmeans that d,
by def., i s diagonalizable.
iagonalizable.
be diagonalizable.
(11). Let T
ep that A is
diagonalizable,
lo prove
330 Linear Algebra
Tis diagonalizable
T has n LI. eigen vector a. 2 ., G E V.
Suppose X, is the co-ordinate vector e V, )corresponding to the eigen vector
o;e Vrelative to B.
Linear independence of lal >X1, X2, .., X, are L.I.
Now X1,X2,.,X, are n LI. eigen vectors* of the matrix A. Hence, by the
previous theorem, A is diagonalizable.
Problem 9. Provethatcharacteristic roots ofa diagonal matrix are the diagonal
elements. Hence show that if a matrix A is similar to a diagonal matrix D, then
diagonal elements of D are the characteristic roots of A.
Solution.
a11 0
a22 0 0
i) Let D=0
a33 0
0 nn
Then D is a diagonal matrix of order n x n.
a11-X 0 0
0 a22 0
D-xl=|
0 0 nn-x
Then | D-xl | =
(«11-2)
(a22 -x).. (n
The characteristic equation of D is |D -xl | = 0
i.e., 1 1 - ) (a22 - x) ... (ann -x) = 0.
This * =011,a22, nn
x = diagonal elements of D.
i.e.
It means that characteristic values of a diagonal matrix are its diagonal
element.
i) Let a matrix A be similar to a diagonal matrix D.
To prove that characteristic values of A are the diagonal elements of D.
A and D are similar
A - x l | =| D -xl |
-16
3 x 4=0
8 7-x
Applying C1+ C2+ C3 and then taking common (-1 -x) from C, we get
1- 1
3- 0
1 4
or -(1+x)|0 0 -1-*by R2 -R1, =0, R3 R1
-
4 3-
Expanding along C. - (1 +*) (1) (-1-z) (3-x) = 0.
-9- 4
3-A . (1)
-8
-16 8
where
and
R2 (-1), ie., Ra
-
single
This is equivalent to- 8x equntion
+ 4y + 4z - 0
E (2)
(3)
or 2x-y-z =0. of
li.e., of (2)| is 1. Hence number
coeflicient matrix matrix
The rank of the c0etficient matrix -its rank = 3-1=
dependent solution is number of columns of
2.
332 Linear Algebra
X1nd1-X
Case II. When = 3.
Then (1) becomes
-12 4
-8 0
-16 8
Applying R12 (-1), R32 (-2), we get
4
-8 0 4
8 8
R31-2) gives
4
0
This -4x +4y = 0, -8x + 42 =0
- y = 0, 2x-2 = 0
=y
Takingx 1, we have x 1,y =1,z
= = = 2
Let P be the matrix formed by the eigen vectors X corresponding to the eigen
values A= -1, -1, 3.
Then P=1 -1
10 1