Matrix Review

Transpose of a matrix: Exchange rows and columns

Standard operations are:

(A + B)T = AT + B T
(AB)T = B T AT

For symmetric matrices, by definition,

AT = A

Determinants:  
a1 a2
For a 2 × 2 matrix A = :
b1 b2
 
 a1 a2 
|A| = 
  = a1 b2 − b1 a2
b1 b2 

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 1 / 22
 Matrix Review


a1 a2 a3
For a 3 × 3 matrix A =  b1 b2 b3 :
c1 c2 c3
 
 a1 a2 a3 
 
|A| =  b1 b2 b3 
 c1 c2 c3 
   
 b2 b3   b1 b3 
= a1   − a2 
 
c2 c3  c1 c3 
 
 b1 b2 
+a3  
c1 c2 
= a1 b2 c3 − a1 c2 b3 − a2 b1 c3
+a2 c1 b3 + a3 b1 c2 − a3 c1 b2

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 2 / 22
 Matrix Review

Inverse of a matrix:
Note that inverse of a square matrix A exists if |A| =
6 0.
adj(A)
A−1 = |A|
AA = A−1 A
−1 =I
−1
(AB) = B −1 A−1

−1
T
AT = A−1 = A−T
 
a1 a2
For a 2 × 2 matrix A = :
b1 b2
 
−1 1 b2 −a2
A =
|A| −b1 a1

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 3 / 22
 Matrix Review


a1 a2 a3
For a 3 × 3 matrix A =  b1 b2 b3 :
c1 c2 c3
 
b2 c3 − c2 b3 − (a2 c3 − c2 a3 ) a2 b3 − b2 a3
1 
A−1 = − (b1 c3 − c1 b3 ) a1 c3 − c1 a3 − (a1 b3 − b1 a3 ) 
|A|
b1 c2 − c1 b2 − (a1 c2 − c1 a2 ) a1 b2 − b1 a2

x, y and z are linearly independent vectors if αx 6= y and

αx + βy = z for some scalar constants α and β.
Rank of a Matrix:
A matrix A is called rank r if the number of linearly independent
columns is equal to r .
square n × n matrix has rank n if and only if |A| =
6 0.

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 4 / 22
 Matrix Review

Eigenvalues of a Matrix:
Equation for determining eigenvalues |λI − A| = 0.
The above expression always reduces to finding the roots of

λn + a1 λn−1 + · · · + an−1 λ + an = 0

Note that, eigenvalues and poles are equivalent.

Note: The number of eigenvectors = number of eigenvalues = n
Procedure for determining eigenvectors:
Case 1) When all of the eigenvalues are distinct we use

(λI − A) v = 0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 5 / 22
 Matrix Review
Example
 
0 1
Find the eigenvalues of A = .
−6 −5
Let’s find the roots of |λI − A| = 0
    
 1 0 0 1 
|λI − A| = λ  − 
0 1 −6 −5 
 
 λ −1 
=   
6 λ+5 
= λ2 + 5λ + 6
= (λ + 2) (λ + 3)

So the eigenvalues are found as λ1 = −3 and λ2 = −2.

 
0
The eigenvectors, denoted by vi , are found via (A − λi I ) vi = .
0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 6 / 22
 Matrix Review
Example

For λ1 = −3
     
0 0 1 1 0
= +3 v1
0 −6 −5 0 1
  
3 1 v11
=
−6 −2 v12
 
3v11 + v12
=
−6v11 − 2v12

1
So v1 = . Note that any multiple of v1 would be an
−3
eigenvector.

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 7 / 22
 Matrix Review
Example

For λ2 = −2
     
0 0 1 1 0
= +2 v1
0 −6 −5 0 1
  
2 1 v11
=
−6 −3 v12
 
2v11 + v12
=
−6v11 − 3v12

1
So v2 = . Note that any multiple of v2 would be an
−2
eigenvector.

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 8 / 22
 Matrix Review

Case 2) When some eigenvalues are repeated, we use the generalized

procedure:
Step 1: For an eigenvalue repeated m times in an n × n matrix A,
find the smallest p such that

Rank {(A − λI )p } = n − m

Note that n ≥ m.
Step 2: For 1 ≤ k ≤ p, find Nk
n o n o
Nk = Rank (A − λI )k−1 − Rank (A − λI )k

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 9 / 22
 Matrix Review
Step 3: Find a vector, denoted by vp , such that
(A − λI )p vp = 0
but
(A − λI )p−1 vp 6= 0
then find
vj = (A − λI ) vj+1
where j = (p − 1) , (p − 2) , · · · , 1.
All the v vectors we find are generalized eigenvectors.
Step 4: Reduce each Nk by 1. If all Nk = 0 then the procedure is
complete.
Step 5: Find the largest k for which Nk is not zero. Call it κ. Find a
vector vκ that is linearly independent of previous vector v . Then find
vj = (A − λI ) vj+1
where j = (κ − 1) , (κ − 2) , · · · , 1.
Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 10 / 22
 Matrix Review
Example

Find 
the eigenvalues
 and the corresponding eigenvectors of
1 1 2
A =  0 1 3 .
0 0 1
The eignevalues of A are found as λ = 1 with a multiplicity of m = 3.
Step 1: For an eigenvalue repeated 3 times in the 3 × 3 matrix A, find
the smallest p such that

Rank {(A − λI )p } = n − m = 3 − 3 = 0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 11 / 22
 Matrix Review
Example

For p = 1
   
1 1 2 1 0 0
(A − λI ) =  0 1 3  − λ  0 1 0 
0 0 1 0 0 1
 
0 1 2
=  0 0 3 
0 0 0
⇒ Rank {(A − λI )} = 2

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 12 / 22
 Matrix Review
Example

For p = 2
  
0 1 2 0 1 2
(A − λI )2 =  0 0 3  0 0 3 
0 0 0 0 0 0
 
0 0 3
=  0 0 0 
0 0 0
n o
⇒ Rank (A − λI )2 = 1

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 13 / 22
 Matrix Review
Example

For p = 3
  
0 0 3 0 1 2
(A − λI )3 =  0 0 0  0 0 3 
0 0 0 0 0 0
 
0 0 0
=  0 0 0 
0 0 0
n o
⇒ Rank (A − λI )2 = 0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 14 / 22
 Matrix Review
Example

Step 2: For 1 ≤ k ≤ 3, find Nk

n o n o
N1 = Rank (A − λI )0 − Rank (A − λI )1
= 3−2=1

n o n o
N2 = Rank (A − λI )1 − Rank (A − λI )2
= 2−1=1

n o n o
N3 = Rank (A − λI )2 − Rank (A − λI )3
= 1−1=0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 15 / 22
 Matrix Review
Example
 
0
Step 3: v3 =  0  satisfies the following
1

(A − λI )3 v3 = 0
(A − λI )2 v3 6= 0
From the general formula vj = (A − λI ) vj+1 for j = 2, 1, we find
    
0 1 2 0 2
v2 =  0 0 3  0  =  3 
0 0 0 1 0
    
0 1 2 2 3
v1 =  0 0 3  3  =  0 
0 0 0 0 0
Note that v1 , v2 and v3 are the generalized eigenvectors.
Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 16 / 22
 Matrix Review
Example

Step 4: Reducing each Nk by 1 results in N1 = N2 = N3 = 0 so the

procedure is complete.

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 17 / 22
 Matrix Review
Example

Find 
the eigenvalues
 and the corresponding eigenvectors of
1 1 2
A =  0 1 3 .
0 0 2
The eignevalues of A are found as λ = 1 with a multiplicity of m = 2
and λ = 2 with a multiplicity of m = 1.
Step 1: For an eigenvalue repeated 2 times in the 3 × 3 matrix A, find
the smallest p such that

Rank {(A − λI )p } = n − m = 3 − 2 = 1

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 18 / 22
 Matrix Review
Example

For p = 1
   
1 1 2 1 0 0
(A − λI ) =  0 1 3  − λ  0 1 0 
0 0 2 0 0 1
 
0 1 2
=  0 0 3 
0 0 1
⇒ Rank {(A − λI )} = 2

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 19 / 22
 Matrix Review
Example

For p = 2
  
0 1 2 0 1 2
(A − λI )2 =  0 0 3  0 0 3 
0 0 1 0 0 1
 
0 0 5
=  0 0 3 
0 0 1
n o
⇒ Rank (A − λI )2 = 1

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 20 / 22
 Matrix Review
Example

Step 2: For 1 ≤ k ≤ 2, find Nk

n o n o
N1 = Rank (A − λI )0 − Rank (A − λI )1
= 3−2=1
n o n o
N2 = Rank (A − λI )1 − Rank (A − λI )2
= 2−1=1
 
0
Step 3: v2 =  1  satisfies the following
0

(A − λI )3 v2 = 0
(A − λI )2 v2 6= 0

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 21 / 22
 Matrix Review
Example

From the general formula vj = (A − λI ) vj+1 for j = 1, we find

    
0 1 2 0 1
v1 =  0 0 3  1  =  0 
0 0 1 0 0

Note that v1 and v2 are the generalized eigenvectors.

Step 4: Reducing each Nk by 1 results in N1 = N2 = 0 so the
procedure is complete.

Asst. Prof. Enver Tatlicioglu (EEE@IYTE) EE362 Feedback Control Systems May 16, 2010 22 / 22