Eigen space

[Link]#
A =
I :?) → vertical shear

"¥±÷f
Are there vectors in pf
directions
whose
under
a
preserved
the action of A
?
1197191--19 ]
=② ←%Ttor
vectors
on
y¥☐ At
,
eigenvalue
?⃝
 23-11-2022

Eigenvalues and Eigenvectors

Let A = aij n⇥n

be an n ⇥ n matrix (with real entries).
Recall x 2 Rn , Ax is a transformation from Rn to Rn . Transformations
Definition: If is a constant such that, < Matrices .

Ax = x, (1)

then is called an eigen value of A and x is called corresponding eigenvector,

where x 6= 0.

How to find
Working rule:
eigenvalues
Now, given A, how to find the eigenvalues and eigenvectors?
&
eigenvectors?
Well form (1),

identity
Ax x=0
I isthe

Consequently, matrix of same

Ax Ix = 0 size as A -

(A I)x = 0 XI → diagonal (2)

matrix
A necessary and sufficient condition for (2) to have non-zero solution is:

det(A I) = 0 (3)

Equation (3) is called the characteristic equation of matrix A.

We observe that equation (3) is nth degree polynomial in of the form,

n n 1 n 2
an + an 1 + an 2 + · · · + a1 + a0 = 0
characteristic
1
equation
Atte :

(i) To find eigenvalues of A, we simply solve (3) to get n roots. So A has

at most n real roots
eigenvalues
.

(ii) For a given eigenvalue, , the corresponding eigenvectors are obtained

using (2). One can expect infinitely many eigenvectors corresponding
to an eigenvalue.

(iii) In fact, the eigenvectors of a form subspace of Rn and this is called as

the eigenspace of A corresponding to .

Simple Illustrative Example:

Question #1
0 1
B7 4C
Consider A = @ A
5 2

(i) Find the eigenvalues of A

(ii) Find the eigenvector of A and state the dimensions of eigenspaces of A.

Solution:

The characteristic equation is |A I| = 0

0 1 0 1
B7 4C B 0C
@ A @ A =0
5 2 0

2
That is,

7 4
=0 (4)
5 2

So,

(7 )( 2 ) + 20 = 0
2
14 7 +2 + + 20 = 0
2
5 +6=0

Std ax4bd+c=o
Solving, the eigenvalues are:
.

Eq .

p A. = I
b± b2 4ac
= b= -5
2a p
c. = 6
( 5) ± 52 4 · 6
=
p 2
5 ± 25 24
=
2
5±1
=
2
5+1 5 1
= ,
2 2
6 4
= ,
2 2
= 3, 2 (We have two real eigenvalues)

3
Case 1: =3 0 1
Bx C
Form (4), if x = @ A are the corresponding eigenvector then:
y 1- ✗ = DX
0
Ax -
✗ ✗ =

(A I)x = 0
0 10 1 0 1
B7 3 4 C Bx C B0 C
@ A@ A = @ A
5 2 3 y 0
0 10 1 0 1
B4 4 C Bx C B0 C
@ A@ A = @ A
5 5 y 0

4x 4y = 0

5x 5y = 0
variable )x=y
free
←
Put y = t and x = t
n=y=t
The eigenvectors are:
0 1 01
Bx C Bt C
@ A=@A
y t
0 1 0 1
Bx C B1 C
@ A = t@ A; t 2 R
y 1

The eigenspace is:

80 1 9
>
< x >
=
B C
W1 = @ A 2 R x = y = t
2
>
: y >
;

4
A✗ =
If HE ] =
1¥ : ¥1 1¥ ] = =3
1¥]
3✗
We
=

80 1 9
>
< t >
=
BC
W1 = @ A 2 R t 2 R
2
>
: t >
;
8 0 1 9
>
< 1 >
=

-1--3-1
B C
W1 = t @ A 2 R t 2 R
2
>
: 1 >
;
A 0 1

, ,
B1 C
A basis for W1 is: B1 = @ A and dim(W1 ) = 1.
1
Case 2: = 2
Substitute into the matrix (A I)x = 0

0 10 1 0 1
B7 2 4 C BxC B0C
@ A@ A = @ A
5 2 2 y 0

0 10 1 0 1
B5 4C BxC B0C
@ A@ A = @ A
5 4 y 0
0
hey
{
=
5N
-

We solve, (one row disappears), 0

5h hey
-
= .

Single Eqn
.

5x 4y = 0
two
4
for
Unknowns
.

x= y
5

Put y = t, to get,

Ike
4
variable x= t
5

5
 The eigenvectors are:
0 1 0 1
4t
BxC B 5 C
@ A=@ A
y t
0 1 0 1
BxC t B4C
@ A = @ A; t 2 R
y 5 5

The eigenspace is:

80 1 9
>
< x >
=
B C 4t
W2 = @ A 2 R x =
2
and y = t
>
: y 5 >
;

80 1 9
>
< 4t >
=
B5C
W2 = @ A 2 R t 2 R
2
>
: t >
;
8 0 1 9
>
<t 4 >
=
B C
@ A2R t2R
2
W2 =
>
:5 5 >
;
2X 0 1
A- ✗ =

, ,
B4 C
A basis for W2 is: B2 = @ A and dim(W2 ) = 1.
5

Question #2
0 1
B2 1 C
Consider A = @ A
0 2

(i) Find the eigenvalues of A

(ii) Find the eigenvector of A and state the dimensions of eigenspaces of A.

6
Solution:

The characteristic equation is |A I| = 0

0 1 0 1
B2 1 C B 0C
@ A @ A =0
0 2 0

That is,

2 1
=0 (5)
0 2

So,
(2 )(2 )=0

(2 )2 = 0

) = 2, 2

Eigenvector coresponding
0 1 to = 2

Bx C
Form (5), if x = @ A are the corresponding eigenvector then:
y

(A I)x = 0

7
 0 10 1 0 1
B2 2 1 C BxC B0C
@ A@ A = @ A
0 2 2 y 0
0 10 1 0 1
B0 1 C Bx C B0 C
@ A@ A = @ A
0 0 y 0

0 y=0

y=0

Put y = 0
The eigenvectors are:
0 1 0 1
Bx C Bx C
@ A=@ A
y 0
0 1 0 1
Bx C B1C
@ A = x@ A; x 2 R
y 0

The eigenspace is:

80 1 9
>
< x >
=
B C
W = @ A 2 R x = x and y = 0
2
>
: y >
;

¥-÷
80 1 9
>
< x >
=
B C
W = @ A2R x2R
2
>
: 0 >
;
8 0 1 9
>
< >
=
B1C
W = x@ A 2 R x 2 R
2
>
: 0 >
;

A✗ =
[317/2]=1%1--21%1 = 2x
 0 1
B1 C
A basis for W is: B = @ A and dim(W ) = 1.
0

Question #3
0 1
B1 1C
Consider A = @ A
1 1

(i) Find the eigenvalues of A

(ii) Find the eigenvector of A and state the dimensions of eigenspaces of A.

Solution:

The characteristic equation is |A I| = 0

0 1 0 1
B1 1C B 0C
@ A @ A =0
1 1 0

That is,

1 1
=0 (6)
1 1

9
So,

(1 )(1 )+1=0

(1 )2 + 1 = 0
2
(1 + 2 )+1=0
2
2 +2=0

Solving, the eigenvalues are:

p
b± b2 4ac
=
2a p
( 2) ± ( 2)2 4·2
=
p 2
2± 4 8
=
p2
2± 4
=
2p
2±2 1
=
2
2(1 ± i)
=
2
= 1 ± i (We have two complex eigenvalues)

10
Case 1: = 1 +0
i 1
Bx C
Form (6), if x = @ A are the corresponding eigenvector then:
y

0 10 1 0 1
B1 (1 + i) 1 C BxC B0C
@ A@ A = @ A
1 1 (1 + i) y 0
0 10 1 0 1
B i 1C BxC B0C
@ A@ A = @ A
1 i y 0

ix y=0

x iy = 0

) x = iy

Put y = t and x = it
The eigenvectors are:
0 1 0 1
BxC BitC
@ A=@ A
y t
0 1 0 1
Bx C BiC
@ A = t@ A; t 2 C
y 1

The eigenspace is:

80 1 9
>
< x >
=
B C
W1 = @ A 2 C x = it and y = t
2
>
: y >
;

11
 80 1 9
>
< it >
=
B C
W1 = @ A 2 C t 2 C
2
>
: t >
;
8 0 1 9
>
< >
=
BiC
W1 = t @ A 2 C t 2 C
2
>
: 1 >
;
0 1
BiC
A basis for W1 is: B1 = @ A and dim(W1 ) = 1.
1
Case 2: = 1 i
Substitute into the matrix (A I)x = 0

0 10 1 0 1
B1 (1 i) 1 C Bx C B0 C
@ A@ A = @ A
1 1 (1 i) y 0

0 10 1 0 1
Bi 1 C Bx C B0 C
@ A@ A = @ A
1 i y 0

We solve,
ix y=0

x + iy = 0

)x= iy

Put y = t, to get,
x= it

12
The eigenvectors are:
0 1 0 1
BxC B itC
@ A=@ A
y t
0 1 0 1
BxC B iC
@ A = t@ A; t 2 C
y 1

The eigenspace is:

80 1 9
>
< x >
=
B C
W 2 = @ A 2 C2 x = it and y = t
>
: y >
;

80 1 9
>
< >
=
B itC
W2 = @ A 2 C t 2 C
2
>
: t >
;
8 0 1 9
>
< >
=
B iC
W2 = t @ A 2 C t 2 C
2
>
: 1 >
;
0
1
B iC
A basis for W2 is: B2 = @ A and dim(W2 ) = 1.
1

Question #4
0 1
B 1 2 3C
B C
Consider A = B
B 2 3 6CC
@ A
2 2 4

(i) Find the eigenvalues of A.

13
(ii) Find the linearly independent eigenvectors of A.

(iii) Let P be the matrix where columns are eigenvectors obtained in (ii).
A
Show that /
P is similar to a diagonal matrix D.
That is P 1
AP = D. For your P what is P 1
A3 P ?

(iv) Do the labourers work to show that if f ( ) = 0 is the equation for A,

then f (A) = 0, where 0 is the 3 ⇥ 3 zero matrix.

(v) Use result (iv) to express A 1

linearly in terms of powers of A and I.

Solution:

(i) To find Eigenvalues:

The characteristic equation is: |A I| = 0

1 2 3
i.e., 2 3 6 =0 (7)

2 2 4

Using row one (you may use any row or column), we solve:

3 6 2 6 2 3
(1 ) 2 +3 =0
2 4 2 4 2 2

2
(1 )[ + 7 + 12 12] 2[8 + 2 12] + 3[ 4 + 6 + 2 ] = 0
2 3 2
+7 7 +8 4 +6+6 =0
3 2
6 + 9 + 14 = 0
3
+6 2
9 14 = 0 obtain one of

14
the root
by
inspection .

7=-1 is a root .
 ✗ 3+6×2 92 14--0
-
-

IE 5×-451-14*14=0 +

5×(7-+1) 141×-1-11=0
747-+1)
-

I ✗45×-1 4)
0
(7+1)
=

with the help of the factor theorem, ( + 1) is a factor,

So ,

3 2 2
+6 9 14 = ( + 1)( +5 14)
3 2
+6 9 14 = ( + 1)( + 7)( 2)

So,
( + 1)( + 7)( 2) = 0

The eigenvalues are: 7, 1, 2

(ii) To find Eigenvectors:

Case: 1 = 7

15
 0 1
Bx C
B C
Form (7), if x = B C
By C are the corresponding eigenvector then:
@ A
z

(A I)x = 0

0 10 1 0 1
B1 2 3 C Bx C B0 C
B CB C B C
B 2 3 6 C B y C = B0 C
B CB C B C
@ A@ A @ A
2 2 4 z 0

0 10 1 0 1
B1 + 7 2 3 C Bx C B0 C
B CB C B C
B 2 6 C B C B C
B 3+7 C B y C = B0 C
@ A@ A @ A
2 2 4+7 z 0

16
(I [ g)
2 3
8

Ri
→ 4 Rzt 0 10 1 0 1
Rz Ri
B 8 2 3C BxC B0C
Rz
-

→ 4 B CB C B C
B 2 4 6 C B y C = B0 C
Rz B
@
CB C B C
A@ A @ A

4%04
2 2 3 z 0

0 10 1 0 1
a
6 B8 2 3 C Bx C B0 C
0 B CB C B C
R2 B0 18 27C By C = B0C
B CB C B C
B @ A@ A @ A
-

3 0 6 9 z 0
→
Rz

88¥
0 10 1 0 1

%
(
B8 2 3 C Bx C B0 C
B CB C B C
~ ° B0 18 27C By C = B0C
B CB C B C
O @ A@ A @ A
O 0 0 0 z 0
-
We solve, 8x + 2y + 3z = 0 and 2y + 3z = 0
3t
Put z = t to get y = and x = 0.
2
So,
Reduced
0 1 0 1 p =
No .

of 2
equators
=

Bx C B 0 C
B C B 3t C
By C = B C variables =3
B C B 2C
@ A @ A 9 No
= .

of
z t
variables

0 1 0 1
No .

of free
I
q p
-
=

Bx C 0C =
B
B C tB C
By C = B 3C
B C 2B C
@ A @ A
z 2

(these are the eigenvectors corresponding eigenvalues)

17
Case: 2 = 10 1
Bx C
B C
Form (7), if x = B C
By C are the corresponding eigenvector then:
@ A
z

(A I)x = 0

0 10 1 0 1
B1 2 3 C Bx C B0 C
B CB C B C
B 2 3 6 C B y C = B0 C
B CB C B C
@ A@ A @ A
2 2 4 z 0

0 10 1 0 1
B1 + 1 2 3 C Bx C B0 C
B CB C B C
B 2 6 C B C B C
B 3+1 C B y C = B0 C
@ A@ A @ A
2 2 4+1 z 0

0 10 1 0 1
B 2 2 3 C Bx C B0 C
B CB C B C
B 2 6C B C B C
B 2 C B y C = B0 C
@ A@ A @ A
2 2 3 z 0

0 10 1 0 1
B2 2 3 C Bx C B0 C
B CB C B C
B0 0 9 C B y C = B0 C
B CB C B C
@ A@ A @ A
0 0 6 z 0

i.e., 2x + 2y + 3z = 0 and z = 0

18
Put y = t to get z = 0 and x = t.
So,
0 1 0 1

(1)
B tC B 1C
B C B C
= B t C = t B 1 C ; t 2 R are the eigenvectors
B C B C
@ A @ A
0 0

Case: 3 =2 0 1
Bx C
B C
Form (7), if x = B C
By C are the corresponding eigenvector then:
@ A
z

(A I)x = 0

0 10 1 0 1
B1 2 3 C Bx C B0 C
B CB C B C
B 2 3 6 C B y C = B0 C
B CB C B C
@ A@ A @ A
2 2 4 z 0

0 10 1 0 1
B1 2 2 3 C Bx C B0 C
B CB C B C
B 2 6 C B C B C
B 3 2 C B y C = B0 C
@ A@ A @ A
2 2 4 2 z 0

19
 0 10 1 0 1
B 1 2 3 C Bx C B0 C
B CB C B C
B 2 6C B C B C
B 5 C B y C = B0 C
@ A@ A @ A
2 2 6 z 0

0 10 1 0 1
B 1 2 3 C Bx C B0 C
B CB C B C
B 0 9 0C B C B C
B C B y C = B0 C
@ A@ A @ A
0 6 0 z 0

We solve:

0
x + 2y + 3z = 0 and y = 0
Put z = t we get,
0
117--1:)
+1 :)
0 1 0 1
B3tC B3 C =
B C B C
0 x = 3t and B
@ A
C B C
B 0 C = t B0 C ; t 2 R
@ A
t 1

0 1 0 1
0 1
B 0 C B 1C B3 C
B C B C B C
The eigenvectors B C B C B C
B 3 C , B 1 C , and B0C are linearly independent
@ A @ A @ A
2 0 1
because they are form different eigenspace, the following theorem proves
the same result.

Theorem #1
Eigenvectors from different eigenspace are linearly independent

Proof: ← +2
Let x1 and x2 be two eigenvectors of A corresponding to , and respec-
tively with 1 6= 2

20
It follows that: Ax1 = 1 x1 and A ✗
z= 7211¢
Now suppose there exists k1 k2 2 R such that,
,

k1 x 1 + k2 x 2 = 0 (8)

Then,
pre
multiply by A

i.e., k1 Ax1 + k2 Ax2 = 0

So,
k1 1 x1 + k2 2 x2 =0 (9)

Multiplying both sides by , we get,

ÉÉs ) by 7, we
get
k1 1 x1 + k2 1 x2 =0 (10)

÷Y:¥÷÷÷
Subtracting (9) from (10) gives, ⑨ → kid ,
×, + kiddie
④
k2 1 x2 k2 2 x2 =0

k2 [ 1 2 ]x2 =0 (11)

Since (11) must be true for x 6= 0

i
K2 [ 1 2] =0

But, 1 2 6= 0 (we assume this)

) k2 = 0

21
Plug into (8) to get,

k1 x 1 + 0 = 0

k1 x 1 = 0

k1 = 0 (* x1 6= 0)

In summary,

k 1 x 1 + k2 x 2 = 0

) k1 = k2 = 0

) x1 and x2 are linearly independent.

Three independent eigenvectors are:

⑦ ①1
0 1 0
2
0 1
B 0C B 1C B3 C
B C B C B C
B 3C , B 1C , and B0 C
B C B C B C
@ A @ A @ A
2 0 1
0 1
B 0 1 3C
P =B
B
B 3 1 0C
C
C
modal matrix
@ A

p= 1×1×2×3 ]
2 0 1

9×1%2%3
Finding P 1
:
To find the inverse matrix, augment it with the identity matrix and
perform row operations trying to make the identity matrix to the left.
Then to the right will be inverse matrix

22
 [ ELI ] F [ I 15 ]
'

So, augment the matrix with identity matrix:

0 1
B 0 1 3 1 0 0C
B C
B 3 1 0 0 1 0C
B C
@ A
2 0 1 0 0 1

Since the element at row R1 and column C1 (pivot element) equals 0,

we need to swap rows.
Find the first non-zero element in column C1 under pivot entry.
The first non-zero element is at row R2.

0
(R1 ! R2) 0 1
B 3 1 0 0 1 0C
B C
R ,
→
Rz B 0
B
@
1 3 1 0 0CC
A
2 0 1 0 0 1

✓ ◆
R2
R1 =
3 0 1
1 1
B 1 3
0 0 0C
3
B C
B 0 1 3 1 0 0C
B C
@ A
2 0 1 0 0 1
✓ ◆
R3 = R3 + 2R1

0 1
1 1
B 1 0 3
0 0C
3
B C
B 0 1 3 1 0 0C
B C
@ A
2 2
0 3
1 0 3 1

23
✓ ◆
R2 = R2
0 1
1 1
B 1 0 3
0 0C 3
B C
B 0 1 3 1 0 0C
B C
@ A
0 23 1 0 23 1
✓ ✓ ◆ ◆ ✓ ✓ ◆ ◆
1 2
R1 = R1 R2 and R3 = R3 R2
3 3

0 1
1 1
B1 0 1 3
0C 3
B C
B0 1 3 1 0 0C
B C
@ A
2 2
0 0 3 3 3
1

✓ ◆ ✓ ◆
R3
R2 = R2 + R3 and R3 =
3

0 1
1 1
B1 0 1 3 3
0C
B C
B0 1 0 1 2
1C
B 3 3 C
@ A
2 2 1
0 0 1 9 9 3

✓ ◆
R1 = R1 R3

0 1
1 1 1
B1 0 0 9 9 3C
B C
B0 1 0 1 2
1 C
B 3 3 C
@ A
2 2 1
0 0 1 9 9 3

24
 As can be seen, we have obtained the identity matrix to the left. So,
0 1
1 1 1
B 9 9 3C
B C
P 1
=B
B
1
3
2
3
1 C
C
@ A
2 2 1
9 9 3

0 10 10 1
1 1 1
B 9 9
1
3C B
2
3C B 0 1 3C
B CB CB C
P AP = B
1
B
1
3
2
3
1 C B
CB 2 3 6C B
CB 3 1 0CC
@ A@ A@ A
2 2 1
9 9 3
2 2 4 2 0 1
0 10 1
7 7 7
B 9 9 3C B
0 1 3C
B CB C
=B
B
1
3
2
3
1C B
C B 3 1 0C
C
@ A@ A
4 4 2
9 9 3
2 0 1
0 1
B 7 0 0C
B C
=B
B 0 1 0CC
@ A
0 0 2

0 1
B 7 0 0C
B C
By the way, P 1 AP = D where D = B
B 0 1 0CC
@ A
0 0 2

(iii) To prove that A is similar to diogonal matrix

25
 Theorem #2
Let A = aij n⇥n
with eigenvectors 1, 2, . . . , n (possibly re-
peated). If x1 , x2 , . . . , xn are independent eigenvectors cor-
responding to 1, 2, . . . , n (respectively) if we get P =
(x1 , x2 , . . . , xn )n⇥n , then P 1
AP = D. Where,

D = diag( 1 , 2, . . . , n)
0 1
0 ... 0
B 1 C
B C
B0 ... 0C
B 2 C
B C
D=B B0 0 ... 0C C
B C
B .. .. .. .. C
B . . . .C
@ A
... ... ... n

So where as P is not unique, for a given f , D is unique.

Proof:
We shall show that AP = AD for a specific D and since P must be

26
invertible, the proof will be done.

AP = A(x1 , x2 , . . . , xn )n⇥n

= (Ax1 , Ax2 , . . . , Axn )n⇥n

= ( 1 x1 , 2 x2 , . . . , n xn )n⇥n
0 1
1 ... 0 ... 0
B C
B C
B0 ... ... 0C
B 2 C
B C
= (x1 , x2 , . . . , xn ) B
B0 ... 0 ... 0C C
B C
B .. .. .. .. .. C
B. . . . .C
@ A
0 ... ... ... n

AP = P D

where D = diag( 1 , 2, . . . , n)

Since AP = P D,

1
P AP = D

) D is similar to A.
To find: D3 = P 1
A3 P

27
 From the above result we have, same as P and D We have,

1
P AP = D

D3 = (P 1
AP )3

D3 = P 1
AP P 1
AP P 1
AP

D3 = P 1
A3 P

[
'
D= n
.

Now, .

0 1

:( ) ¥
B 7 0 0C
B C
D3 = B
B 0 2 0CC ☐
@ A
'

0 0 1 -
,

0 1
ii.
B 343 0 0C
B C
D3 = B
B 0 8 0CC
@ A
0 0 1

In summary, 0 1
B 343 0 0C
B C
P 1 A3 P = B
B 0 8 0CC
@ A
0 0 1

# cteristic
(iv) To prove that A staisfies its carastric equation
The characteristic equation is

f( ) = 0

28
 3 2
( +6 9 14 = 0)

We want to show that A satisfies characteristic equation,

i.e., A3 + 6A2 9A 14I = 0
Now consider,

A3 + 6A2 9A 14I

0 13 0 12 0 1 0 1
B 1 2 3C B 1 2 3C B1 2 3C B1 0 0 C
B C B C B C B C
=B
B 2 3 6C B
C + 6B 2 3 6C B
C 9 B2 3 6CC 14 B
B0 1 0 C
C
@ A @ A @ A @ A
2 2 4 2 2 4 2 2 4 0 0 1

0 1 0 1 0 1 0 1
B 1 2 3C is
B 1 2 3C B1 2 3C B14 0 0 C

÷÷÷±÷
B C B .C B C B C
=B
B 2 3 6C B
C + 6B 2 3 6CC 9B
B2 3 6CC
B 0 14 0 C
B C

(
@ A @ A @ A @ A
2 2 4 →
2 →
2 94
4 2 2 4 0 0 14

0 1
B0 0 0 C
B C
=B C
B0 0 0 C = 0
@ A
0 0 0

The above result can be stated as the following theorem.

µ .my , ygyyy
23

y
3
5 6 9
,, , , ,
, , ,
-
to -10 34 2 2 -4
7 8 78 -226
29

-
"
l : :)
0 I 0
 Theorem #3 (Cayley-Hamilton’s Theorem)
A matrix a satisfies its characteristic equation.
That is, if f ( ) = 0 is the characteristic equation of A, then
f (A) = 0.

(v) To find A 1 :
From part (iv),
A3 + 6A2 9A 14I = 0
'
A-
multiply by
A is inevitable and

A 1 (A3 + 6A2 9A 14I) = A 1 0

A2 + 6A 9I 14A 1
=0
A2 6A 9I
)A 1
= +

:-#
14 14 14

¥1 :)
?::÷ If +

:( ÷:) -

31g
I
=

µ% Y
30
-517
'
17
-617
"
14

)
0 14 21
=

¥ ( 4
2
-10

2
-12