3.

62 Engineering Mathematics
homogeneous systems Ax = 0, where A is
Find all the solutions of the following given as
the
1 2
3 37.
1

4 -7
2
38. 4
3 -11
1
fol owing.
36. 1 -2 -10
-4 |-1 -2 4 9
-6\
-1 -3 1 1 1
|1 2 3 4 1 1 -1 1
1 1 40. 41.
39. 1 -7 -13 -1 -1 -1 1
2 6 12| -1 5 9 1 -1

3 1 1 4 1 -3 2 1 -2 1
-1]
0 4 10 1 2 -1 -2 -3 1 1 -2
43. 44. 3
42. 0 -5 -1 4
1 7 17 3 1 -5 8
2 2 4 3 |5 -1 -7 -4 -7 2 -1

1 -2 -1|
2 1 1 -2
45.
3 2 -1 -3
|4 2 2 -4|
Using the Gauss-Jordan method find the inverses of the following matrices.
1 1 1 2 3
|-1 1 1

46. 1 2
1 -1 1
3 47. 1 3 3 48.
1 -1 1
1 3 4 0 1 2| 1 1 -1
1 1 1 1] 1 1 0
-1 1 1 1 1 1
49. 50.
-1 1 4 4 1
1 -1 1 0 0 1

3.5 Eigenvalue Problems

Let A = (ai) be a square matrix of order n. The matrix A may be singular or
the homogeneous system of equations non-singular. Cos
Ax =Ax or (A -I) x = 0 (3.39)
where 2is a scalar and I is an identity matrix of order n. The homogeneous system of equations (3.39)
always has a trivial solution. We need to find values of 2for which the homogeneous syystem(3.39)
has non-trivial solutions. The values of 2, for which non-trivial solutions of the homogeneous system
(3.39) exist, are called the eigenvalues or the characteristic values of Aand the correspondingnon-
trivial solution vectors x are called the eigenvectors or the characteristic Vectors of A. Ifx is a non-
trivialsolution of the homogeneous system (3.39), then a x, where a is any Constant is also asolution
multiple. The
of the homogeneous system. Hence, an Constant
is eigenvector unique only upto a
 Matrices and Eigenvalue Problems. 3.63
problemofdetermining the eigenvalues and the corresponding eigenvectors of a square matrix Ais
Ped an eigenvalue problem,
15.1 Eigenvalues and Eigenvectors
homogeneous system (3.39) has a
Ifthe non-trivial solution, then the rank of the coefficient matrix
(A-2)isless than n, that is, the coefficient matrix must be singular. Therefore,

din
det (A - AI) = d2n =0. (3.40)

anl an2 ann2

Expandingthe determinant given in Eq. (3.40), we obtain a polynomial of degree nin 2, which is of
theform
P,(4) =|A -1I|=(- 1)"[2" -c + ,22-...+(- 1)°ca] =0
2"-c 2 + ,2-,..+(-1)"c, =0. (3.41)
called
where Cj, C2, C Can be expressed interms of the elements a; of the matrix A. This equation is
which can
the characteristic equation of the matrix A. The polynomial equation P, () =0 has n roots (2) = 0
he real or complex, simple or repeated. The roots 4,, ,.., , of the polynomial equation P,
coefficients, we can write
the
are called the eigenvalues. By using the relation between the roots and

If we set l= 0in Eq. (3.40), we get

(3.42)

Therefore, we get
of eigenvalues = |A
sum of eigenvalues = trace (A) and product
and the largest eigenvalue in magnitude is called
The set of eigenvalues is called the spectrum of A A=0, that is the matrix is singular, then from
the spectral radius ofA and is denoted by p (A). If|
wefind that one of the eigenvalues must be zero. Conversely, if one of the eigenvalues is
Eq. (3.42),
an upper triangular or alower triangular matrix,
zero, then|A=0. Note that if Ais a diagonal or eigenvalues of A.
A are the
then the diagonal elements of the matrix
homogeneous system (A-4;I)x=0for each
Afer determining the eigenvalues Aj's,we solve the
eigenvectors.
corresponding
Ai, i=1,2, ..., n to obtain the
Properties of eigenvalues and eigenvectors
eigenvector. Then we have the following results.
Let 2 be an eigenvalue of Aand x be its corresponding
1. aA has eigenvalue a and the
corresponding eigenvector is x.
Ax = x’ CAX = (a)x.
 3.64 Engineering Mathematics
eigenvector is x for any
2. A" has eigenvalue " and the corresponding positive
Premultiplying both sides of Ax = Ax by A, we get integer m.
AAX= AÂx = AAx = (ax) or A'x = 1lx.
Therefore, A? has the eigenvalue A andthe corresponding eigenvector is y
successively m times, we obtain the result.
3. A-kI has the eigenvalue d - k, for any scalar kand the corresponding
Premultiplying
Ax =Ax ’ Ax -k Ix = x - kx eigenvector is x.
or (A kI)x =(-k) x.
4. A (if it exists) has the eigenvalue 1/2 and the corresponding eigenvector is x.
Premultiplying both sides of Ax = Axby A,we get
A-'Ax =AA-'x or A-'x=(1W)x.
5. (A - kI has the eigenvalue 1/(2 - k) and the corresponding eigenvector is x for any
`calar k.

6. A and A' have the same eigenvalues, sinçe a determinant can be expanded by rows
columns.

7. For a real matrix A, if a +iß is an eigenvalue, then its conjugate a - iß is also an eigenvalue
(since the characteristic equation has real coefficients). When the matrix Ais complex. this
property does not hold.
We now present an important result which gives the relationship of a matrix A and its characteristic
equation.
Theorem 3.9 (Cayley-Hamilton theorem) Every square matrix Asatisfies its own characteristic
equation, that is

A"- cjA +..+-1C-A +(-1)"c,I =0. (3.43)

Proof The cofactors of the elements of the determinant | A -I|are polynomials in of degree
(n -1)or less. Therefore, the elements of the adjoint matrix (transpose of the cofactor
matrix) are also
polynomials in Aof degree (n-1) or less. Hence, we can express the adjoint matrix as a polynomial
in 2 whose coefficients B1, B2,..., B, are square matrices of order n having elements as
of the elements of the matrix A. Ths, we can funcaos
write
adj (A - AI) = B,2"+ B,2 +...+B-d+ B:
We also have

(A - AI) adj (A - AI) =|A - AI|I.

Therefore, we can write for any
(A-1)B,2"+B"-+...+ B-1+ B,) = 1"I -c4=1+...+(- l)'c-dl+(-I)'e,!
Comparing the coefficients of various powers of 2, we obtain
-B, =I
AB -B, =- c|l
AB, - B3 = czI
 Matrices and Eigenvalue Problems 3.65

AB--B, =(-1yG
AB, =(-1y'c,l.
Premultiplyingthese equations by A", An-
.,A, I respectively and adding. we get
A"-c A+...+(- 1y'ch-A +(-1y'c, I = 0
which proves the theorem
Remark 16a We have for any non-zero vector x

Ix = lx
Ax = Ax
A'x = 'x

A"x = 2"x.

Multiplying these equations by (- 1y'e,, (- 1cn-..,(- I)e, Irespectively and adding, we get
(-1'e, Ix +(-1yci Ax +...+(-1)'e,A'x +A'x
=(-1)'e, x+(- 1y'c Ax+...+(- 1)'e,2'x +1'x
[A" - c A- +...+(- 1y-'c A+(- 1)'c, ) x
= [2" -c2+...+ (- 1yc-l+(-1yc]x =0x= 0.
Since x 0, there is a possibility that
A"- c A=+..+ (- 1y'clA +(-1'c, I =0.
Remark 16b

(a) We can use Eq. (3.43) to find A (if it exists) in terms of the powers of the matrix A.
Premultiplying both sides in Eq. (3.43) by A, we get
A-A"-cj A-'A+...+(-1yAA +(-1'c,AI= A0=0
A-1 =-(A-l-c,A-2+... +(-1)=' C-] (3.44)
Cn
lower powers of A as
(b) We can use Eg. (3.43) to obtain A" in terms of
A" = c A7=-cA+...+(-1y'c, I. (3.45)

Example 3.39 Verify Cayley-Hamilton theorem for the matrix

2 01
1 2
2

eigenvalues of A'are squares

Also (i) obtain A-! and A?, (ii) find eigenvalues of A, A» and verify that
3.66 Engineering Mathematics

of those of A, (iii) find the spectral radius of A.

Solution The characteristic equation of A is given by (

|A - AI| = -1 =(l - 2)(1 - 2)² - 4} -

1 2 2(-(1 - ) - 2)
=(1- )a'-22 -3) 2(2 3) = - + 32 -+ 3=0
1 2 0 1 2 0 -1 4 4|
Now, A'=-1 1 2-1 1
1 2
1 2 3
6
4
5

-1 4 4] 1 2 0 -1 10 121
A' =A'A = 0 4|| -1 1 2 = 1 11 10
0 6 5| 1 2 1 -1 16 17
We have

-1 10 12 -1 4 4 1 2 0]
-A+ 3A- A +3I = 1 11 10+3 0 3
[1 0 0
1 2|+3 0 1 0
-1 16 17 0 6 1 2 |0 0
0
0 0 0= 0.
(3.46)
0 0
Hence, A satisfies the characteristic equation -7 +
32-+3= 0.
i)) From Eq. (3.46), we get
1 4 4 3 6 0 1 0 0)
A'=[A- 3A + )=| 0 3 4 -3 3 6+ 0 1 0
0 6 3 6 3 0 0 1)
-3 -2 4
3 1 -2
-3 0 3

From Eq. (3.46), we get

-3 12 12| [-1 10 12
A =3A-A +31= 2 0 3 0 0]
9 12
1 2 +
0 33 o=I11 100
18 15 |-I16 17
1 2 0 0 3
(ii) Eigenvalues of A are the roots of
 Matrices and Eigenyalue Problems 3.67
2-32 + d-3=0 or (a- 3\2 +
The characteristic equation of A' is i)=0 or =3, i, - i.
given by
-1-1 4 4

3-4 4
6 5-2
=(-1-a)[3- A)5 A)-24] =0
of (a+ 1(2'-8À - 9) =0 or (2+
12-9)( + 1) = 0.
The eigenvalues of A are 9, - 1, - 1which are the
squares of the eigenvalues oI A.
(ii) The spectral radius of Ais given by
p(A) =largest eigenvalue in magnitude = m¡x
||=3.
1
o] veify Cayley }amilenThm Find Nese.
Example 3.40 If A=1 then showthat A" =A-2+ A2-Ifor n >3. Hence, find A".
0 1

Solution The characteristic equation of Ais given by

|1-1 0
|A - AI|= 1 -2 1 = (1- A(a-1) =0, or 1- 2?-A+1= 0.
1 -4

Using Cayley-Hamilton theorem, we get

A-A'-A +I= 0, or A'-A= A-I.
Premultiplying both sides successively by A, we obtain
A-A'=A -I
A-A'= A'-A
A=-A-= A -A4
A"- A+l= A2-A3
Adding these equations, we get
A" -A=A-1 or A"=A+A'-I, n23.
Using this equation recursively, we obtain
A"= (A+A?-) + A² -I=A=4+ 2(A2-)
=(A6 4 A- ) +2(A- I)= A-+ 3(A? - I)

= A7-(n-2)
+(n- 2)(4? -I)=A-n-2)1.
Substituting n = 50, we get
3.68 Engineering Mathematics

0 01 0 0
AS0 = 25A2-241 = 25|1 01
1
0-24|0 1
1 |0 0 25
Example 3.41 Find the eigenvalues and the corresponding eigenvectors of
the
fol owing matrices
(ii) A=|0 2
|2 0 3
Solution
() The characteristic equation of A is given by
4
|A- l|= 2-A
=0 or -32- 10 =0, or =-2,5.
Corresponding to the cigenvalue = - 2, we have
3
(A + 21)x =
3
C-0)
Hence, the eigenvector x is given by
or 3x1 +4x)=0 or
X1=-i2
4

Since an eigenvector is unique upto a constant multiple, we

can take the eigenvector as
-4,3)".
Corresponding to the eigenvalue = 5, we have
(A -S1)x = or x - )= 0, or x = ).

Therefore, the eigenvector is given by x = (, x) =*(1, 1)" or simply (1, 1)'.

(ü) The characteristicequation of A is given
by
|A -À|= -1 =0, or 1-22+ 2 = 0, or d= l ti.
Corresponding to the eigenvalue d = 1+i, we have

or
- ix + x) =0 and -x- ixy = 0.
Both the equations reduce to - X -ixn =0. Choosing x) = 1, we get x =-iTherefore,the
eigenvector is x= (-i, 1j".
Corresponding lo the eigenvalue A= |-i, we have
 3.69
Matrices and Eigenvalue Problems

[A -(1- i)l]x :
ix + *= 0 and -x + ix, = 0.
Both the equations reduce to - x + ix, = 0.Choosing x, = 1,we get x = i. Therefore, the
eigenvector is x = [i, 1j.
Remark 17

For a real matrix A, the eigenvalues and the corresponding eigenvectors can be complex.
(ii) The characteristic equation of Ais given by
|1-a 0 0

|A - AI|= 0 2 =0 or (1- )(2- )(3- 2) =0

2 0 3-A
or d= 1, 2, 3.
Corresponding to the eigenvalue 1 = 1, we have
fo 0. 07
2+ *y =0
(A - I)x=0 1 1 X =|0 Or
+ X} =0.
|2 0 0
chosen
We obtain two equations in three unknowns. One of the variables x1, X, X can be
arbitrarily. Taking x, =1,we obtain the eigenvector as [- 1,- 1, 1]".
Corresponding to the eigenvalue =2, we have

(A- 21)x = 0
2 0

the eigenvector as [0, 1,01".

or x = 0, x, = 0 and x arbitrary. Taking x, = 1, we obtain
Corresponding to the eigenvalue = 3, we have
-2 01
(A -3I)x = 1x=|0 or
|*2 + X3 = 0.
2 0 3 0

1].
Choosing x, = 1, we obtain the eigenvector as [0, - 1,
Example 3,42 Find the eigenvalues and the corresponding eigenvectors of the folowing matrices
1 1 0 0
(i) (iii) A=
|0 0
0 0
 Algebraic Multipliatyi Hou ea ines an gen a pears a
Geometrie Mulhplicity: Dimension tha eign space
associe wit
3.70 Engineering Mathematics
Solution In each of the above problems, we obtain the characteristic
Therefore, the eigenvalues are À=1, 1, 1, arepeated value. Since a: 3×3 matrix has equation
it is important to know, whether the given matrix has 3 linearly independent
lesser number of linearly independent eigenvectors.
Corresponding to the cigenvalue A=1, we obtain the following eigenvectors eigenvect3oeirgsenyaues,
(0 0
Geonatie |*2=0,
() (A - I)x =0 1'x= or X3 =0
0
arbitrary.
Choosingxj = 1, we obtain the solution as [1, 0, 0j".
Hence, A has only one independent eigenvector.
AM=3 (a=)
01
(i) (A - I)xX =0 0
0|| =|0 or
0 0
0 X1,X3 arbitrary.
Taking x=0, x3 = 1 and x = 1, x, =0, we obtain two
linearly independent solutione
X= [0,0, 1], x =[1, 0, 0j".
In this case A has two linearly
independent eigenvectors.
0
(ii) (A - I)x=0 0,

This system is satisfied for arbitrary values of all the

three variables. Hence, we obtain three
linearly independent eigenvectors, which can be written as
X=[1,0, 0], X, =[0, 1,0]", x, = [0, 0, 1]'.
We now state some important results regarding the
and the coresponding linearly relationship between the eigenvalues of amari
independent eigenvectors.
Figenvectors corresponding to distinct eigenvalues are linearly independent.
If 2is an eigenvalue of multiplicity mof a square matrix Aof order n, thenthe number
of lincarly independent
eigenvectors associated with is given by
p=n-r, where r=rank (A -I), 1SpSm.
Remark 18
In Example 3.41, all the eigenvalues are distinct and therefore, the corresponding Cigenvectorsarc
linearly independent. In Example 3.42 the eigenvalue = lis of multiplicity 3. We findthatin
(i) Example 3.42(6), the rank of the matrix A - Iis 2 and we obtain one linearly independent
eigenvector.
independent
() Example 3.42(1), the rank of the
eigenvectors. matrix A-Iis l and we obtaintwolinearly
 Matrices and Eigenvalue Problems 3.71
() EXample 3.42(iii), the rank of the matrix A -is0and we obtain three linearly independent
eigenvectors.
3.5.2 Similar and
Diagonalizable Matrices
Similar matrices
Let Aand B be square matrices of the same order. The matrix Ais said to be similar to the matrix B
if there exists an invertible matrix Psuch
that
APBP or PA = BP. (3.47)
Postmultiplying both sides in Eq. (3.47) by P-, we get
PAP- = B.
Theretere, A is similar to Bif and only if B is similar to A. The matrix P is called the similarity
matriz. We now prove a result regarding eigenvalues of similar matrices.
MTheorem 3.10 Similar matrices have the same characteristic equation (and hence the same eigenvalues).
Further, ifx is an eigenvector ofA corresponding to the eigenvalue 2, then P'x is an eigenvector of
Bcorresponding to the eigenvalue , where Pis the similarity matrix.
Proof Let ¦ be an eigenvalue and xbe the corresponding eigenvector of A. That is
Ax = Ax.

Premultiplying both sides by an invertible matrix P, we obtain

Pr'Ax = AP-'x.
Set x = Py. We get
p-'APy = AP-'Py, or (P'AP)y =ly or By =y.
where B =P-'AP. Therefore, B has the same eigenvalues as A, that is the characteristic equation of
Bis same as the characteristic equation of A. Now, Aand Bare similar matrices. Therefore, similar
matrices have the same characteristic equation (and hence the same eigenvalues). Also x= Py, that
is eigenvectors ofA and B are related by x = Py or y= P'x.
Remark 19
(a) Theorem 3.10 states that if two matrices are similar, then they have the. same characteristic
equation and hence the same eigenvalues. However, the converse of this theorem is not true.
Two matrices which have the same characteristic equation need not always be similar.
(b) If A is similar to B and B is similar to C, then A is similar to C.
Let there be two invertible matrices P and Q such that
A= P-'BP and B= Q'CQ.
Then A= P'Q'cQP =RCR, where R= QP.
Example 3.43 Examine whether A is similar to B, where

(1) A and B :
3.72 Engineering Mathematics

and

Solution The given matrices are similar if there exists an invertible matrix P such that
A PBP or PA = BP.

We shall determine a, b, c, dsuch that PA = BP and then

check whether
P is non-singular.
5a -25 5a a+ 2c
)
Sc- 2d Sc-3a +4c b+2d4d
-36+
Equating the corresponding elements, we obtain the system of equations
5a -2b=a+ 2c, or 4a - 2b - 2c = 0
5a = b+ 2d, or 5a -b- 2d =0
Sc- 2d =-3a + 4c, or 3a + c- 2d = 0
5c = -3b + 4d, or 3b + 5c - 4d = 0.
Asolution to this system of equations is a= 1, b= 1, c= l,d= 2.
Therefore, we get which is a non-singular matrix. Hence the matrices Aand B
are similar.

(ii)

Equating the corresponding elements, we get

a = a+ c, b=b +d or c=d= 0.

Therefore, P - which is a singular matrix.

|0 0|
Since an invertible matrix P does not exist. the matrices A and B are
not simlar.
It can be verified that the eigenvalues of Aare 1, 1 whereas the eigenvalues of Bare , 2.
In practice, it is usually difficult to obtaina non-singular matrix P which satisfies the cquationA=
P-BP for any twO matrices Aand B. However, it is possible to obtain the matrix P when Aor Bis
a diagonal matrix. Thus, our interest is to find a similarity matrix P such that for a given matrix A.
we have

D = P'AP or PDP- = A
where D is a diagonal matrix. If such a matrix exists, then we Say that the matrix AIS
diagonalizable.
 Matrices and Eigenvalue Problems 3.73

Diagonalizable matrices
A matriX A is diagonalizable, if it js similar to adiagonal matrix, that is there ex1sts an
matrix P such that PAP=D, where Disa diaoonal matrix Since, similar matrices havenyouoe the same
egenvalues, the diagonal elements of Dare the eigenvalues of A. Anecessary and sufficient condition
for the existence of P is given in the
following theorem.
Pheorem 3.7
A square matrix A of order n is diagonalizable if and only if it has n lincarly
independent eigenvectors.
Proot We shall prove the case that if Ahas nlinearly independent eigenvectors, then Ais diagonalizable.
Let X1, X2 X, be n linearly independent eigenvectors corresponding to the eigenvalues
h, Az., (not necessarily distinct) of the matrix Ain the same order, that the eigenvector x
corresponds to the eigenvalue , j=1, 2, .. .,n. Let
P= (X1, x,..., x] and D diag (2j, ..,)
be the diagonal matrix with eigenvalues of A as its diagonaB elements. The matrix P is called the
modal mnatrix bf A and D is called the/spectral matrix of A. We have
AP = A(X1, X2, ... , x] = (AX], Axz, .. ., Ax)
=(4jx, x,...,,x,) = (x, Xz, , x)D =PD. (3.48)
Since the columns of P are linearly independent, the rank of P is n and therefore the matrix P is
invertible.Premultiplying both sides in Eq. (3.48) by P, we obtain
P'AP = rPD =D (3.49)
which implies that Ais similar toD.Therefore, the matrix of eigenvectors P reduces a matrix Ato its
diagonal form.
Postmultiplying both sides in Eq. (3.48) by P, we obtain
A = PDp-! (3.50)
Remark 20
A square matrix Aof order nhas always nlinearly indendent eigenvectors when its eigenvalues
are distinct. The matrix may also have n linearly independent eigenvectors even when some
eigenvalues are repeated (see Example 3.42(ü)). Therefore, there is no restriction imposed on
the eigenvalues of the matrix A in Theorem 3.11.
(b)Prom Eq. (3.50), we obtain
A'= AA= (PDP-)(PDP) = PD'p-!,
Repeating the pre-multiplication (post-multiplication) mtimes, we get
Am= PD"P for any positive integer m.
Therefore, if A is diagonalizable, so is A",
c If D is a diagonal matrix of ordern, and
 3.74 Engineering Mathematics

0 |
then DM=
D=

we get
for any positive integer m. If Q(D) is a polynomial in D, then

Q(D) =

Now, let a matrix A be diagonalizable. Then, we have

A= PDPl and A" = PD"Pl
for any positive integer m. Hence, we obtain
QlA) = POD)P
For any matrix polynomial Q(A).
Example 3.44 Show that the matrix
3
A =|-2 1
1 2|
is diagonalizable. Hence, findP such that P'AP is a diagonal
matrix. Then, obtain the matrix
B= A+ 5A + 31.

Solution The characteristic equation ofA is given by

3-2 1 -1
|A - AI|=-2 2 = 1'-622 + 112 -6= 0, or =1, 2, 3.
2-1
Since the matrix Ahas three distinct eigenvalues, it has
hence it is diagonalizable. three linearly independent eigenvecios
The eigenvector corresponding to the
eigenvalue =lis the solution of the system
2
(A - I) x 1|
-2 0
The solution is x =-1
The eigenvector coresponding to the
eigenvalue =2is the solutionof the syste
-] 0
(A - 21)x=-2 -]
0 1
2|| *2 - The solution is x2=|0}
 Matrices and Eigenvalue Problems 3.75
The eigenvector corresponding to the eigenvalue A= 3 is the solution of the system
(A - 3I)x = -2 -2 2 .The solution is x3 =
1

Hence, the modal matrix is given by

1 1 01] -1 -1 1
P= (x1, X2, X3] =-1 0 1land P-! = 2 1
1 1 1 -1
It can be verified that P'AP = diag (1, 2, 3).
We have D = diag (1, 2, 3), D' = diag (1,4, 9).
Therefore, A' + 5A + 31 = P(D²+ 5D + 3I) P-1
0 3 01 9 01
Now, D²+ 5D +3I =0 4 0 0 10 0+0 =0 17 0
9 0 15 0 0 3 0 0 27

Hence, we obtain

1 1 0 9 0|-1 -1 25 -8
2 1 9 18
A² +5A +31 =-1 0 1 0 17
1 1 1||0 0 27| 1 0 -2 8 19

Example 3.45 Examine whether the matrix A, where A is given by

2 2 -2 2-31
2 (ii) A = 2 1 -6
(i) A=
-1 -2 0
-1 2
diagonal matrix.
is diagonalizable. If so, obtain the matrix P such that P'AP is a
Solution
A is given by
(i) The characteristic equationof the matrix
1-a 2 2

|A - AI|= 2 -1
-1 2 2-1

=(1-AL(2 - )(2 - ) - 2] - [2- 2(2 - 1)]= (1- Ay2- X2- A) = 0.

the eigenvectors corresponding to the repeated eigenvalue l =2.
or A= 1, 2, 2. We first find
We have the system
 3.76 Engineering Mathematics

-1 2
(A - 21)x = 0

|-1 2

Since the rank of the coefficient matrix is 2, it has one linearly independent
obtain another linearly independent cigenvector corresponding to the eigenvalue eigenvector. We
dieahg=\on.aSliiznabclee.
the matrix Ahas onlytwo linearly independent eigenvectors, the matrix is not
(i) The characteristic cquation of the matrixA is given by
-2-A 2 -3
|A- AI|= 2 -6 =0 or 2' + ' - 212-, 45 =0, or
-1 -2 -2 2=5,-3,3.
Eigenvector corresponding to the eigenvalue A= 5 is the solution of the systerm
-7 2 -3
(A - SI)x= 2 -4 -6|| X2
|-1 -2 -5
A solution of this system is [1, 2, - 1]'.
Eigenvectors corresponding to l=-3 are the solutions of the system
1 2
-3|| 1
(A + 3I)x = 2 4
-6|| X2 or x + 2x,-3x3 =0.
-1 -2 3

The rank of the coefficient matrix is 1. Therefore, the system has

two linearly independent
solutions. We use the equation xj+2x,-3x, =0 to find two linearly independent
Taking x;=0, x)= 1, we obtain the eigenvector [-2, 1, 0j and taking x, =0, X} =eigenvectors.
l, we obtain
the eigenvector [3, 0, 1]'. The given 3 x3 matrix has three linearly
independent eigenvectos.
Therefore, the matrix A is diagonalizable. The modal matrix P is given by
-2 31 1 2 -31
P= 2 0 and P-!=-2 4 6
|-1 0 1 2 5
It can be verified that P'AP =
diag (5, - 3, - 3).
Example 346 The eigenvectors of a3 x 3matrix Acorresponding to the eigenvalues 1,1, 3are
[).0, - 1j', [0, 1, - I)' and [1, 1, 0j
respectively. Find the matrix A.
Solution We have

0 1 0 07
modal matrix P = 0
1| and the spectral matrix D= |0 1
-1 -1 0 3
|0 0
 Matrices and Eigenvalue Problems 3.77

1 -1
We find that
p-=-1 1
1
1

0 1 1 0 -1 1
Therefore, A = PDP-= 0
-1 -1 3
1
7
11 1 -1 -1| 2 1 1
0 1 1|| 1 2 1
-1 -1 0| 3 33 0 1

3.5.3 Special Matrices

In this section we define some special matrices and study the properties of the
eigenvalues and
eigenvectorsof these matrices. These matriçes have applications in many areas. We first give some
definitions.
Let x = (1, 2, " Ap' and y= (O, y2,y)' be two vectors of dimension n in R" or C".
Then we define the following:
Inner Product (dot product) of vectors Let x and y be two vectors in IR". Then

Xy=xy = (3.51)
is called the inner product of the vectors x and yand is ascalar. The inner product is also denoted by
<X, y>. In this case x y =y"x. Note that x x20and x x= 0 if and only if x= 0.
If x and y are in C", then the inner product of these vectors is defined as
n

x"y=x"ù= 2x9, and y x= y'% = 2 yz,

where and ù are complex conjugate vectors of xand yrespectively. Note that x y=y.x. It can
be easily verified that
(ax + By) z= a(x" z) + B(y"z)
for any vectors X, y, Z and scalars a, B.

Length (norm of a vector) Let x be a vector in R" or C". Then

|| x||= xx=x+ x;+...+*?
is called the length or the norm of the vector x.
Unit vector The vector xis called a unit vector if ||x ||=1. If x*0, then the vector xl| x||is
always a unit vector.
Orthogonal vectors The vectors x and y for which xy=0are said to be orthogonal vectors.
3.78 Engineering Mathematics
Orthonormal vectors The vectors of x and y for which
xy=0 and | x ||= 1, || y|| =1
are called orthonormal vectors. If x, y are any vectors and xy= 0, then
orthonormal.
For example, the set of vectors
,y\y|are
form an orthonormal set in IR3

3i -4i\ 3i/5 -4i/5 0

(ii) -4i 3i 0 form an orthogonal set in C and-4i/5 3i/5 0
0 +i)
form an orthonormal set in C.
0)\(1+2)
Orthonormal and unitary system of vectors Let xj, K2, ,Xm be m vectors in R". Then i.
set of vectors forms an orthonormal system of vectors, if

[0,
Xx} =x x, = |1, i#j
i=j.
Let x, X, ..., x, be m vectors in C".Then, this set of vectors forms an
if
unitary system of vectors

0, ij
(1, i=j.
In section 3.2.2, we have defined symmetric, skewW-symmetric, Hermitian and
We now define a few more special matrices. skew-Hermitian matrices.
Orthogonal matrices Areal matrix A is orthogonal if A- = A'.A simple example is
cos -sin 0
A=
sin Cos

Unitary matrices Acomplex matrix A is unitary if A- =(A)T, or (A)-!- A". IfA is rea, uia
unitary matrix is same as orthogonal matrix. We note the
following.
1. IfA and Bare Hermitian matrices, then aA +BB is also Hermitian for any real scalars c, B
since

(aA + BB) = (aA + BB) = aAT + BB? = aA +BB.

2. Eigenvalues and eigenvectors of Aare the conjugates of the eiggenvalues and eigenvectors
A, since

Ax = Ax gives Ax = x.
3. The inverse of a unitary A-=A Let
(orthogonal) matrix is unitary (orthogonal). Wehave
B=A-. Then
 Matrices and Eigenvalue Problems 3.79

B= A=( [( B".
We now establish some important
results.
Theorem 3.12 An orthogonal set of vectors is linearly independent.
Proof Let XË, X Xm be an orthogonal set of vectors, that is x, " X, =0, i#j. Consider tne vov
equation

X= 1 XË+ 2X2 t...+ am Xm=0 (3.52)

where a, . Cm are scalars. Taking the inner product of the vector x in Eq. (3.52) with xË, We
get
X" X1 = (a X+ Q, Xy +...+ am Xm)" x =0 X1 =0
or
a,(x1 x) = 0 or a l X|= 0.
Since || X|*0, we get a = 0. Similarly, taking the inner products of x with x,, XJ.
successively, we find that a =O, =...=am= 0. Therefore, the set of orthogonal vectorsX1, X.
is linearly independent.
Pheorem 3.13 The eigenvalues of
(i) an Hermitian matrix are real.
(iü) a skew-Hermitian matrix are zero or pure imaginary.
(ii) an unitary matrix are of magnitude 1.
Proof Let be an eigenvalue and x be the corresponding eigenvector of the matrix A. We have
Ax = Ax. Premultiplying both sides by , we get
'Ax = Az'x or = Ax (3.53)
x
Note that IAx and Tx are scalars. Also, the denominator "x is always real and positive.
Therefore, the behaviour of Ais governed by the scalar "Ax.
(i) Let Abe an Hermitian matrix,that is A =A". Now,
(RAx) = =xAR =(x"AT)= "Ax
since RTATx is a scalar. Therefore, "Ax isreal. From Eq.(3.53), we conclude that Ais real.
(ii) Let A be a skew-Hermitian matrix, that is A"=A. Now,
(RIAx) = x¤R = -xA=-(x"ATR) =-'AX
since xA is a scalar. Therefore, 'Ax is zero or pure imaginary. From Eq. (3.53), we
conclude that is zero or pure imaginary.
(iii) Let A be an unitary matrix, that is A= (A)'. Now, from
Ax= x or Ax=7x (3.54)

we get ( =(k )" or RA"=ÎT

 3.80 Engineering Mathematics

or R'A-!=.
Using Eqs. (3.54) and (3.55), we can write
(3.55)
(AXAx) =(2z(Ax)=|a|2*x
or
x= |A|'x
Since x # 0, we have 'x #0. Therefore, | 2| = 1, or A= 1.
Remark 21
Hence, the result.
From Theorem 3.13, we conclude that the eigenvalues of
() a symmetric matrix are real.
(ii) a skew-symmetric matrix are zero or pure
imaginary.
(üi) an orthogonal matrix are of
magnitude 1 and are real or complexconjugate pairs
Theorem 3.14 The column vectors (and also row
system of vectors. vectors) of an unitary matrix form
an
Proof Let Abe an unitary matrix of order n, unitary
with column vectors X1, K2, ., X,.
Then

A-'A=A'A= X2,x,]=

Therefore, 0, i#j
|1, i=j.
Hence, the column vectors of A form an
an unitary matrix and the unitary system. Since the inverse of an
columns ofA- are the coniugate of the rows of A, unitary matrix is also
vectors of A also form an unitary we conclude that the row
Remark 22
system.
(a) From Theorem 3.14, we
orthogonal matrix form anconclude that the column vectors (and also the row vectors) of an
orthonormal system of vectors.
(b) Asymmetric matrix of
order nhas nlinearly and hence is diagonalizable.
Example 3.47 Show that the independent eigenvectors
the eigenvectors matrices Aand A" have the same andfor distinct eigenvalucs
Solution We have
corresponding to AandA" are mutually eigenvalues orthogonal.
Since A and A |A-ÀI|=|(A" - A1|=||4- A'| =|A"-
have the same I,
Let characteristic equation,
and u be two distinct eigenvalues of A. Let they
x behave
the the same eigenvalues.
eigenvector Correspondingtothe
 3.84 Engineering Muthematies
positive definite.
Therefore, the given matrixis
diagonally dominant (3>|-24|. 4>|2/) with
Note that Ais Hermitian, strictly
diagonal cntries. Therefore, A is positive
definite (see Remark 23(a).)
()
positive real
(c)
0 3

+ ay+*Ry- ix, Ey +3xyR

Taking a P + iq1, ) P + iqz, Xy = pP, + iqy and simplifying, we obtain

Q= (pi+ai) +(P;+ q) +3(pf+ qi) - 2(P193- Ps41)
-(P)- q9+(py+ q)'+(Pi+ h)+2(pi +4)>0.
Therefore, the matrix A is positive definite. It can be verified that the eigenvalues of A
are 1, 2, 2 which are all positive.
Example 3.50 Let A be a real square matrix. Show that the matrix A'A has real and positiue
cigenvalues.
Solution Since (AA) =AA, the matrix A"A is symmetric. Therefore, the eigenvalues of A'A are
all real. Now,

s'ATAx =(Ax)°(Ax) =y'y, where Ax =y.

Since y'y > 0 for any vector y # 0, the matrix A"A is
of A'A are positive. Therefore, all the eigenvalues of positive definite and hence all the eigenvalues
AA are real and positive.
Exercise 3.4
Verify the Cayley-Hamilton theorem for the matrix A. Find
to 6. A-!, if it exists, where A is as given in Problems l
0
[3 -2
0
2.0 2 3. 2 3 -2
4 4 3
2 3
|3 -1
|1 2 -21|
4. 21|
3 3 -1|
-1
Find all the eigenvalues and the
of the matrices are diagonalizable?corresponding eigenvectors of the matrices given in Problems 7to 18.
Which
2
7. 2 2 3
-I 2 2 9. 3
2 0
 Matrices and Eigenvalue Problems 3.85
1 3 3
10. 1 4 3 i
11.
-1 3 4 12. i
-i -i
0 1 0
i 0
01
0
0 2 -2 0
0 1 1 1
13. 1 0
0 1 14. -1 1 0 1 1
-1 1 15.
-2 1
0 1 1 0 1
1 -2 1 1
1 2 3 41 0 0 0
2 1 4 0 1
16. 0 1
3 4 1 17. 1 0 0 18. 0 0 0 1 0
4 3 2 1
00 0
0 0 1
Show that the matrices given in Problems 19 to 24 are
diagonal matrix. diagonalizable. Find the matrix P such that PAP is a
3 2 1
-3 -2 1 2 1
19. 0 2
20.-2 04 21.203
1 2 3
as-6 -3 5 1-3 0
1 1 -6 -67 2 -2|
22.1 1
23.-13 4 2 24. 1 1
-1 -6 -4! 3 -

Find the matrix A whose eigenvalues and the corresponding eigenvectors are as given in Problems 25 to 30.
25. Eigenvalues: 2, 2, 4; Eigenvectors: (-2, 1, 0)T,.(-1,0, 1)7, (1,0, 1).
26. Eigenvalues: 1, - 1, 2; Eigenvectors: (1, 1, 0)", (1, 0,1)F, (3, 1, 1)".
27. Eigenvalues: 1, 2, 3; Eigenvectors: (1, 2, 1)', (2, 3, 4)7, (1, 4, 9).
28. Eigenvalues: 1, 1, 1; Bigenvectors: (-1, 1, 1)" (1, -1, 1)", (1, 1, 1".
29. Eigenvalues: 0, 1,1; Eigenvectors: (- 1, 1, 0)", (1, 0,--1)", (1, 1,1).
30. Eigenvalues: 0, 0, 3;Eigenvectors: (1, 2, - 1)", (-2, 1, 0)". (3, 0, 1)".
31. Let a 4 x 4 matrix Ahave eigenvalues 1, - 1,2, - 2. Find the value of the determinant of the matrix
B= 2A + A--I.
32. Leta 3 x3 matrix A have eigenvalues 1, 2, -1. Find the trace of the matrix B A-A-1, A2
33. Show that the matrices A and P'APhave the same eigenvalues.
34. Let Aand Bbe square matrices of the same order. Then, show that AB and BA have the same eigenvalues
but different eigenvectors.
35. Show that the matrices A-B and BA" have the same eigenvalues but different eigenvectors
36. An nxn matrix A is nilpotent if for some posthve lneger k, A" =0. Show that all the eigenvalues of a
nilpotent matrix are zero.
= A, then
37. If Ais an nXn diagonalizable matrix and A show that each
eigenvalue of A is 0 or 1.
 3.86 Engineering Mathematics

,a #b, is transformed to adiagonal matrix D = P-'AP

a

38. Show that the matrix A= h

,where Pi
cOs - sin 0 and tan 20=2h
a-b
of the form P = sin 0 cos 0)
show that (i) A-' is similar to B-! .(ii) A" is similar to Bm for any positive
39. LetA be similar to B. Then
integer m, (i) | A =|B|.
the same order. Then, show that AB is
40. LetA and Bbe symmetric matrices of symmetric if and only if
AB = BA.
symmetric.
41. For any square matrix A, showthat A'A is
42. Let A be a non-singular matrix. Show that A"A is symmetric if and only if A2 = (A72
orthogonal matr
43. If A is a symmnetric matrix and P-AP = D, then show that P is an
44. Show that the product of two orthogonal matrices of the same order is also an orthogonal matrix.
n
45. Find the conditions that a matrix A =| l n2 is orthogonal.
m3

46. If Ais an orthogonal matrix, show that| A=t1.

47. Prove that the eigenvectors of a symmetricmatrix corresponding to distinct eigenvalues are orthogonal
48. Amatrix Ais called a normal matrix if AA"= A"A. Show that the Hermitian, skew-Hermitian and
unitary matrices are normal.
49. If amatrix A can be diagonalized using an orthogonal matrix then show that A is symmetric.
50. Suppose that a matrix Ais both unitary and Hermitian. Then, show that A= A-l,
51. If A isa symmetric matrix and xAx> 0for every real vector x# 0, then show that ~'Az is real and
positive for any complex vector z#0.
52. Show that an unitary transfornmation y=Ax, where Ais an unitary matrix preserves the value of the innet
product.
53. Prove that a real 2 x 2 symmetric matrix is positive definite if and only if a>0 (1xl leading
C

minor) and ac - b >0 (2x2leading minor).

2 1 31
54. Showthat the matrix -3 4 -1| is positive definite.
|-1 1 2

[-3 -2 1
55. Show that the matrix-2 4| is not positive definite.
-6 -3 5
Find the symmetric or the Hermitian matrix Afor the quadratic forms given in Problems 56to 60.
56. x- 2x1*2 +4x2Xy -x+*.
57. 3x +2xj42-4xg +8xp; +x.
58. x + 2ixjX) - 8x3 +4ixzx, +4x.
3.92 Engineering Mathematics
28. Inconsistent.
27. [1 + a- 2a, a], a arbitrary.
30. [1, 3, 3). 31. [3/2, 3/2, 1]. 32. (-1, -
29. [1, 1, 1).
33. [(S + a-4ß)/3, (1 + 2a + By3,
a, B], a, Barbitrary. \2,3141.
arbitrary.
35. [- 1/4, 1/4, 1/4, 1/4].
34. [2 - a, 1, a, 1], a
37. [- 15a, 134, o), a arbitrary.
36. (- a, a, a], a arbitrary.
39. [- 2al3, 7al3, 8al3, a], a arbitrary.
38. [0, 0, 0].
40. (2(8- 3, - (5ß + a)3, B a], a, Barbitrary.
42. [(2B- 5a)y4, -(10ß+ a)4, B, a, a, B
41. [0, 0, 0, 0].
43. [(a+ 5ØV3, (4ß - 7a)/3, B, a], a, Barbitrary. arbitrary.
44. [(38- Sc)/3, (3ß- 4a)3, B, a], a, Barbitrary.
45. [a-38, 5B, B, a], a, Barbitrary. -1 1 1
3 -5 6
1 1 -11 1
4 -5 48, -1 1 1
47. -2
46. 1 -3 3
4 1 1 -1
-1 2 -1 1 -2
1 1 1 -1

J1 -1 -1 -17 [-1 -V3 13

1 1 0 -1
50.
49. 1 0 -1 -1 1
1/3 -1/3 0
1 1|

Exercise 3.4

16 -201
1. P() =A -92? 91+ 81 =0; Al= 16 13
-20 4 -5
6 -4 -2
2. P(A) =P - 81? +20A 16 =0; A-' =16 0

-2 -4 6
3. P) = - 3A? + 2l = 0; Inverse does not exist.
4 4 -4|
4. PA) = - -42 +4=0; A-'= -2 -1
-2 1

-3 -7
5. PA) = -52'+94- 13 = 0; A-!=I
5 -1

[i-1
6 PQ) = 1'-32' + 62-4+2i = 0: A-!=-l+3i
10
i-1|
 Matrices and Eigenvalue Problems 3.93
7. A= 1: (1, 1, - 1);
=2,2: (2, 1, 0)': not diagonalizable.
g. A=- 1: (0, 1, 1)'; à= :
(1 + i, 1, 1; à
9. A= 1, 1, 1: (0, 3, - 2]'; not =-i:[|-i, 1, 1,diagonalizable.
10, A= , 1: [0, 1, - 1]; =7: diagonalizable.
(6, 7, 5); not
11. A=0: [i, 0, 1]: A=1+ 3: (1,
diagonalizable.
V3-1, - ij.
A=l-3: [1, -(V3+ 1), -j: diagonalizable.
12 A=-i, - i: [1,0, - 1], 1, - 1,0]: A =
2i: (1, 1, 1;diagonalizable.
13. À= 0, 0, 0, 0: [1, 0, 0, 0]:; not diagonalizable.
14. À=0, 0: [1, 0, 0, 1]. [1,-1, - 1,0j: A= 2: (1, 1, 0, 0j":
A =-2: [1, 0, 1, 1]; diagonalizable.
15. =- 1, -1,- 1: [1, - 1, 0, 0]', [1, 0, - 1,0j', [1, 0, 0, - 1]:
d= 3: [1, 1, 1, 1]; diagonalizable.
16. À=-4: [1, 1, - 1, -1: A=10: (1, 1, 1,1: A=/Z:(W2-1, 1- 2,-1, 1J",
2=-2: [-(1+ \2), 1+ 42, -1, 1]"; diagonalizable.
17. d=-1,- 1: [1, 0, 0,0,- 1],[0, 1, 0, 1, 0j7; A= 1, 1, 1: [1,0, 0, 0, 1], [o, 1,0, 1, 0] (0, 0. 1, 0.,0j:
diagonalizable.
18. 2 = 1, w, w², w, w*, wis fifth root of unity. Let , =w',j =0, 1, 2, 3, 4. d=5; [1..$.3.`)
j= 0, 1, 2, 3, 4; diagonalizable.
19. = 2, 2: [1, 0, - 1j". -2, 1, 0j"; A=4: [1, 0, 1j".
1 -2 1 2 -1]|
P= 1
-1 2

20. 2 =1: ([1, 2, oj; A=- 1: [3, 2, 2]; =2: (- 1,3, 1j'.
|8 -5 71
3 -1| 1
P=-2 -2
2 -4 -2
(3 - i, 13i, - 4]".
21. ) = 0: (3, 1, - 21"; A= 2i: [3 + i, 1 + 3i, 4]"; d= -2:
24 -8 161
3+i 3-i1 2-6i -8
2i-6
P= 1
|-2
1+3i
-4
1-3i
-4
P -2i-6 2+ 6i -8

[1, I,0j.
22. = 0: [1, 0,- 1]: A= 1:(- 1,- 1, 1J, A= 2:
-1

P= 0 -I
1
(2, 1, 0]".
23. 2 =1: (3, - 1, 3]": A= 2, 2: [2, 0, I]'.
3 2
2 2|
3 -6 -5
|-1 0
3 2
3.94 Engineering Mathematics

24. A=1: [1, - 1, - j: d= 2: [0, 1, 11", à=-2: (8 -5, 71".

1 0 81 [12 8 -8|
-s P-= 12 15 -3
|-1 1 7 -1 1

-2 -1 1 0 2 07 |3 2 1
25. P = 1 0 o
-2 1A = PDP- =0 0
1 2 2 3
1 1 3 2 6 -5
26. P= | 0 PDP-!1=|1 0
1
3 -3 -4
1 -11 14 -5] 30 -12 6
27. P = 2 3 4 14 -8
2 ; A = PDP-! = 2 4 14
1 4 -5 2 1 -34 4 38
1 1 |0 21 0 0
|2 2 ;A= PDP-! = 0 1 0
1 1 -1 [2 2
|0 0 1
|-1 1 1 -1 2
29. P 31
1 0 1 1
-1 1 1
1
1
-2;A= PDP = 1
2
1 1
2
1
-2 3] 1 2 -3
30. P = 2 1 18 451
|-1
O}P-=-2
1
4 6;A = PDP-! 1
0
1 2 5 3 6 15
31. Eigenvalues of Bare 22,+ (1/2,) - 1,j= 1, 2, 3, 4
of B = 154. or 2, - 4, 72, 11/2. | B|= product of
eigenvalues
32. Eigenvalues of Bare , + 2} - (1/), j= 1, 2,
B= 15/2. 3 or 1, 11/2, 1. Trace of B
sum of eigenvalues o
33. Premultiply Ax = Ax by P- and
34. Let be an eigenvalue and xbe substitute x = Py.
the corresponding
A and substitute x= Ay.
We get eigenvector of AB, that is ABx = Àx. Premuupy
BAy =y. Therefore,
are related by x = Ay. is also an
eigenvalue of BA and eigeue
35. Let be an eigenvalue and x be
by A and set x =A'y. We the corresponding eigenvector of A-B,that is A-
obtain BA-y= ly, Therefore, Ais also an Bx= AX. rnehe
corresponding eigenvector
36. From Ax = Ax, we
y = Ax. eigenvalue of BA
obtain A'x = x=0.
37. Since Ais a Therefore, 2 = 0ord= 0, since x * 0.
diagonalizable matrix, there
eigenvalues of Aand Dare same.existsWe ahave P-'A'Pmatrix
non-singular Sincethat A'P'AP= =]A, Dweandthe
= D'P, such get
plAP= D'. Therefore, we obtain D² -D =0. Thus D = 0 or 0
or 1. D=I. Hence, the eigenvalues of Aare