Mathematics for Business I

Matrix Diagonalization

Vita Zhukova
 Eigenvalues and Eigenvectors. Definition
 a11 a12 ... a1n 
 
 a 21 a 22 ... a 2 n 
Let A be a square matrix of order n: A
 . . . . 
 
 a n1 an2 ... a nn 

We say that   R is the Eigenvalue of A if there exists a non null vector,

 x1 
 
 x2 
x   / x  0 that verifies the following equality: Ax   x
.
 
x 
 n
 Eigenvalues and Eigenvectors. Definition

Notice that if A is of oreder n, the vector x has to have n elements:

A n •n x n •1

The vector , x , that verifies this equality is called Eigenvector associated to 

 Eigenvalues and Eigenvectors

The equality Ax   x
can be also expressed as Ax   x  0 , where ( A  I )x  0

It is about System of Homogeneous Linear Equations,

where (A  I) is the matrix of coefficients
 Eigenvalues and Eigenvectors

 a11 a12 ... a1n   1 0 ... 0   x1   0

       
  a a 22 ... a 2 n   0 1 ... 0   x 2   0
( A  I ) x  0    
21
 . . . .   . . ... .    .   . 
       
 a n1 an2 ... a nn   0 0 ... 1   x n   0
 a11   a12 ... a1n   x1   0
    
 a 21 a 22   ... a 2 n   x 2   0

 . . . .   .   .
    
 a n1 an2 ... a nn    x n   0
  
( A  I ) x 0
 Eigenvalues and Eigenvectors
The system is homogeneous and, therefore, always has solution:

• If the determinant (A  I)  0 , then it is C.D which unique solution is x0

• In order to be the vector x a non null vector, the determinant has to be (A  I)  0
In this case, we will have C.U, that receives the name of “Characteristic Equation”. The
development of the determinant leads to the polinomial in  of degree n that admits n
solutions for  , the n matrix eigenvalues.
 Eigenvalues and Eigenvectors. Example

Let’s look for eigenvalues and eigenvectors associated to the matrix A :

 3 0 1
 
A   0 1 1
 0 0 5
 

The eigenvalues and eigenvectors are those for which the following
expression is satisfied: Ax   x , where x  0
 How to find eigenvalues and eigenvectors
We want the system Ax   x ( A  I )x  0 to be consistent underdetermined,
for which there will be solutions different to the trivial ones.

For that, the matrix ( A  I ), has not to be of the máximum Rank, that is, we look for
( A  I )  0

 3 0 1  1 0 0 3   0 1
   
 0 1 1   0 1 0   0 1   1  3    1   5     0
 0 0 5  0 0 1  5
    0 0
 How to find eigenvalues and eigenvectors

Therefore, the determinant will be equal to 0 and the system will be consistent
underdetermined if: 3    1   5     0

That is, for 3 posible values of  : 1  3, 2  1, 3  5 , that are 3

eigenvalues associated to matrix A
 How to find eigenvalues and eigenvectors
We look for eigenvector associated to 1  3 , that will be all vectors for which
the following expression is satisfied:

A x  3x ( A  3I ) x  0

 3 0 1  x   1 0 0  x   0   0 0 1  x   0 
            
 0  1 1  y   3 0 1 0  y    0    0  4 1  y    0 
 0 0 5  z   0 0 1  z   0   0 0 2  z   0 
            
 How to find eigenvalues and eigenvectors
The matrix rank is equal to 2, we will have to
 0 0 1  x   0 
    
 0  4 1  y    0  parametrize 1 of unknowns.
 0 0 2  z   0 
     The system is the following:

 4 y  z  0 y  0 Therefore, the eigenvectors associated to eigenvalue

 
2z  0  z  0 3 are vectors (x,y,z) where y=0 and z=0, x= α being a
parameter.
 x    1 The eigenvectors associated to eigenvalue 3 are all
     
 y    0     0 vectors proportional to the vector (1,0,0). It is a vector
z 0  0
      subspace of dimention equal to 1.
 How to find eigenvalues and eigenvectors
That is, the matrix A multiplies by the scalar 3 any vector that belongs to a
vectorial subspace generated by (1,0,0). Let´s see:

 3 0 1   3 0 1    3   
        
A   0  1 1    0  1 1  0    0   3 0 
 0 0 5   0 0 5  0   0   0 
        

Ax   x
 How to find eigenvalues and eigenvectors
Now, we look for eigenvectors associated to 2  1 , that will be all vectos for
which the following expression is satisfied:

A x  1x ( A   1I ) x  0

 3 0 1  x   1 0 0  x   0   4 0 1  x   0 
            
 0  1 1  y    1 0 1 0  y    0    0 0 1  y    0 
 0 0 5  z   0 0 1  z   0   0 0 6  z   0 
            
 How to find eigenvalues and eigenvectors
 4 0 1  x   0 
     The matrix rank is equal to 2, we will have to
 0 0 1  y    0  parametrize 1 of unknowns.
 0 0 6  z   0 
     The system is the following:
 4 x  z  0
 x  0 Therefore, the eigenvectors associated to eigenvalue
z0   z  0  -1 are vectors (x,y,z) where y=0 and z=0, x=α being a
 
6z  0  parameter.
 x  0  0
      The eigenvectors associated to eigenvalue -1 are all
 y        1  vectors proportional to the vector (0,1,0). It is a vector
z 0  0
      subspace of dimention equal to 1.
 How to find eigenvalues and eigenvectors
That is, the matrix A multiplies by the scalar -1 any vector that belongs to a
vectorial subspace generated by (0,1,0). Let´s see it:

 3 0 1   3 0 1  0   0  0
        
A   0  1 1    0  1 1           1  
 0 0 5   0 0 5  0   0  0
        

Ax   x
 How to find eigenvalues and eigenvectors
Now, we look for eigenvectors associated to 3  5 , that will be all vectos for
which the following expression is satisfied:

Ax  5x ( A  5I ) x  0

 3 0 1  x   1 0 0  x   0    2 0 1  x   0 
            
 0  1 1  y   5 0 1 0  y    0    0  6 1  y    0 
 0 0 5  z   0 0 1  z   0   0    
         0 0  z   0 
 How to find eigenvalues and eigenvectors
  2 0 1  x   0 
     The matrix rank is equal to 2, we will have to
 0  6 1  y    0  parametrize 1 of unknowns.
 0    
 0 0  z   0 
The system is the following:

   1
x     
2  x  2   3
 2
 2x  z  0       1   
 y   y            1 ,   6
 6 y  z  0 6 z   6 6  
   1  6
z      
    

 How to find eigenvalues and eigenvectors
That is, the matrix A multiplies by the scalar 5 any vector that belongs to a
vectorial subspace generated by (3,1,6). Let´s see:

 3 0 1   3 0 1  3   15   3 
        
A   0  1 1    0  1 1  1    5   5 1 
 0 0 5   0 0 5  6   30   6 
        

Ax   x
 How to find eigenvalues and eigenvectors
Let’s see the case in which we have a matrix with repeated eigenvalues, to
learn what are its associated eigenvectors, let:
 3  1  1
 
A   1 1  1
1 1 1 
 

The eigenvalues and eigenvectors are those for which the following
expression is satisfied: Ax   x , where x  0
 How to find eigenvalues and eigenvectors
We want the system Ax   x ( A  I )x  0 to be consistent undetermined,
for which there will be solutions different to the trivial ones.

For that, the matrix ( A  I ), hasn’t to be of the máximum Rank,that is, we look for

( A  I )  0

 3  1  1  1 0 0  3    1 1
   
 1 1  1    0 1 0   1 1    1    5  8  4  0
3 2

1 1 1   0 0 1  1 1  
    1
 How to find eigenvalues and eigenvectors

Therefore, the determinant will be equal to 0 and the system will be consistent
underdetermined if:    5  8  4  0    1  2  0
3 2 2

That is, if: 1  1, 2  2 (double) , that are 2 eigenvalues associated to the matriz
A where one of them is double.
 How to find eigenvalues and eigenvectors
Let’s look for eigenvalues associated to 1  1 , that will be all vectos for which
the following expression is satisfied: A x  1x ( A  1I ) x  0

 3  1  1 x   1 0 0  x   0   2  1  1 x   0 
            
 1 1  1 y   1 0 1 0  y    0    1 0  1 y    0 
 1  1 1  z   0 0 1  z   0   1  1 0  z   0 
            
 How to find eigenvalues and eigenvectors
 2  1  1 x   0  The matrix rank is equal to 2, we will have to
    
 1 0  1 y    0  parametrize 1 of unknowns.
 1  1 0  z   0  The system is the following:
    

2 x  y  z  0 2  y  z  0 y  2  z 
  
xz 0  z0  z  

 x    1
      The eigenvectors associated to eigenvalue 1 are all
 y        1 vectors proportional to the vector (1,1,1). It is a vector
 z    1
      subspace of dimention equal to 1.
 How to find eigenvalues and eigenvectors
That is, the matrix A multiplies by the scalar 1 any vector that belongs to a
vectorial subspace generated by (1,1,1). Let´s see:

 3  1  1  3  1  1     
        
A   1 1  1   1 1  1      1  
 1  1 1   1  1 1      
        

Ax   x
 How to find eigenvalues and eigenvectors
Now, we look for eigenvectors associated to 2  2 , that will be all vectos for
which the following expression is satisfied:

Ax  2 x ( A  2  I )x  0

 3  1  1 x   1 0 0  x   0  1  1  1 x   0 
            
 1 1  1 y   2 0 1 0  y    0   1  1  1 y    0 
 1  1 1  z   0 0 1  z   0  1  1  1 z   0 
            
 How to find eigenvalues and eigenvectors
1  1  1 x   0  The matrix rank is equal to 1, we will have to
    
1  1  1 y    0  parametrize 2 of unknowns.
1  1  1 z   0 
     The system is the following:

x  y  z  0 x    x        0  1 0
             
x  y  z  0  y     y        0         0    1 

x  y  z  0 
z    z            1   1
           

The eigenvectors associated to eigenvalue 2 are all vectors that belong to

vectorial subspace generated by the vectors (1,0,1) and (0,1,-1). It is a vector
subspace of dimention 2.
 How to find eigenvalues and eigenvectors
That is, the matrix A multiplies by the scalar 2 any vector that belongs to a
vector subspace generated by (1,0,1) and (0,1,-1). Let´s see:

 3  1  1  3  1  1    2    
        
A   1 1  1   1 1  1     2   2  
 1  1 1   1  1 1      2       
        

Ax   x
 Eigenvalues. Properties
1.- The eigenvalue associated to an eigenvector is unique.

Proof: Let 1 2 be two eigenvalues associated to the same eigenvector x( x  0)

If this is true, we have that Ax   1 x , where

Ax   2 x

 1   2
 1 x   2 x  (  1   2 )x  0  
x  0
Being x an eigenvalue, it has to be different to zero.
Hence, the solution has to be 1 2
 Eigenvalues. Properties

2.- Let x1 , x 2 ,... x n be eigenvectors associated to the same eigenvalue  , then

all its linear combination ( x  1 x1   2 x 2 ... n x n ) is also an eigenvector
associated to 
 Eigenvalues. Properties
3.- A sufficient, but not necessary condition for eigenvectors, x1 , x 2 ,... x n ,
to be Linearly Independent is that the respective eigenvalues 1 ,  2 ,...,  n
be different between each other

• If  1   2 ,...,   n  x1 , x 2 ,... x n are Linearly Independent

• But if x1 , x 2 ,... x n are Linearly Independent 
 1  2 ,...,  n
 Eigenvalues. Properties
Proposition:

If  i is the number of times the eigenvalue  i is repeated, a necessary and

sufficient condition for the eigenvectors x1 , x 2 ,... x n to be linearly independent
consists in checking whether rg(A-I) = n-  i (n is the matrix order).

We perform this check only for the repeated eigenvalues, that is

x1 , x 2 ,... x n are linearly independent  rg(A-I) = n-  i ,  i /  i  2
 Eigenvalues and Eigenvectors of a Diagonal Matrix
 d1 0 ... 0 
 
0 d2 ... 0 
Let D be a diagonal matrix of order n: D .

... ... ... ...
 
0 ... d n 
 0

To obtain eigenvalues, we define the system ( A  1I ) x  0 :

 d1 0 ... 0   1 0 ... 0  x1   0   d1   0 ... 0  x1   0 

           
 0 d2 ... 0   0 1 ... 0  x2   0   0 d2   ... 0  x2   0 
     
 ... ... ... ...  ... ... ...  
... ...    ...   ... ... ... ...   ...  ...
           
 0 0 ... d n   0 0 ... 1  xn   0   0 0 ... d n    xn   0 
 Eigenvalues and Eigenvectors of a Diagonal Matrix

d1   0 ... 0
0 d2   ... 0
0
... ... ... ...
0 0 ... d n  

where (d1   )(d2   )...(dn   )  0 is the characteristic equation which solutions

are eigenvalues. In this case   d1 ,  d2,...,  dn
Eigenvalues and Eigenvectors of a Diagonal Matrix

Therefore, the eigenvalues of the diagonal matrix, D, coincide with the

elements of the main diagonal.

Moreover, the associated eigenvectors coincide always with the vectors

of the canonical basis.

Let’s see an example,

 Eigenvalues and Eigenvectors of a Diagonal Matrix
 3 0 0
 
Let D be the diagonal matrix of order 3: D   0  1 0 
 0 0 2
 

To obtain eigenvalues we compose the system:

 3 0 0   1 0 0  x1   0   3   0 0  x1   0 
           
  0  1 0     0 1 0   x2    0    0  1   0   x2    0 
 0 0 2   0 0 1  x3   0   0 0 2    x3   0 
 Eigenvalues and Eigenvectors of a Diagonal Matrix

3 0 0
0 1   0  0  (3   )(1   )(2   )  0
0 0 2

which solutions are Eigenvalues, in this case 1  3, 2  1 y 3  2

 Eigenvalues and Eigenvectors of a Diagonal Matrix

Now we look for eigenvectors associated to 1  3 , that will be all vectors for
which the following expression is satisfied:

A x  3x ( A  3 I )x  0

 3 0 0  x   1 0 0  x   0   0 0 0  x   0 
            
 0  1 0  y   3 0 1 0  y    0    0  1 0  y    0 
 0 0 2  z   0 0 1  z   0   0 0 2  z   0 
            
 Eigenvalues and Eigenvectors of a Diagonal Matrix

 x  x    1
  y0 y0      
  y:    y    0     0
 z  2z  0 z0 z 0  0
       

Therefore, the eigenvector associated with the eigenvalue 1  3 will be the

vector (1,0,0) or any other vector generated from this one. Similarly, we can
easily check that eigenvectors associated to 2  1 will be the vectors generated
by the vector (0,1,0) and the vectors associated to the eigenvalue 3  2 will be
the ones generated by (0,0,1).
 Similar Matrices

The diagonalization is a particular case of the similar matrices.

Two matrices A and B are similar if there exists a transition matrix P such that

A=P-1BP

Let’s see some properties of similar matrices and its application to

diagonalizable matrices:
 Similar Matrices. Properties

1.- Two similar matrices have the same eigenvalues

Let A be a square matrix of order n with eigenvalues 1 ,  2 ,...,  n

If A is diagonalizable, then there exists a Diagonal Matrix D,

1
D  C AC similar to the matrix A

Because A and D are similar matrices, the eigenvalues of the matrix D are also
the eigenvalues of the matrix A: 1 ,  2 ,...,  n

.
 Similar Matrices. Properties

Given that the eigenvalues of the diagonal matrix D coincide with the elements
of the main diagonal, we get:

 1 0 ... 0 
 
 0 2 ... 0 
D
0 0 ... 0 
 
0 0 ...  n 
 Similar Matrices. Properties

2. If two matrices are similar, they have the same determinant and the
same trace.

Consecuently, if A is diagonalizable, A  D and tr(A)=tr(D)

 Similar Matrices. Properties
3. If two matrices are similar, the n-th power matrices are also similar

1
If A is diagoanlizable,  D  C AC such that A and D are similar.
n n
Since A and D are similar, this property tells us that A and D are similar
1
and also D  C A C
n n

This property simpliefies the calculus of the powers of diagonalizable matrices.

 Matrix Diagonalization
Diagonalize a matrix is a process that consists in getting a diagonal matrix D
from a square matrix A, for which there exists a transition matrix C:

A=CDC-1

If it is possible, then the matrix A is diagonalizable

 Matrix Diagonalization
Theorem:
Let A be a square matrix of order n with eigenvalues 1 ,  2 ,...,  n and
eigenvectors x1 , x 2 ,... x n

A is diagonalizable  x1 , x 2 ,... x n are linearly independent

 1 0 ... 0 
 
1 0 2 ... 0 
In this case,  D  C AC = 0  , where C is the Transition Matrix,
0 ... 0
 
0 0 ...  n 

which columns are given by x1 , x 2 ,... x n (eigenvectors of A)

 Example. Matrix Diagonalization
Let’s diagonalize the matrix A:
 3 0 1
 
A   0 1 1
 0 0 5
 

That is, we need to find a diagonal matrix that is similar to A, which is the
same as to find a diagonal matrix D for which there exists a transition matrix C
such that A=CDC-1.
 Example. Matrix Diagonalization

We have to find its eigenvalues, that are 1  3, 2  1, 3  5

The eigenvectors associated to each one are:

1  0 6
     
1  3  v1   0 ; 2  1  v2   1 ; 3  5  v3   2 
 0  0 12 
     
 Example. Matrix Diagonalization
According to the properties of the eigenvalues and eigenvectors, we know that
eigenvectors that correspond to different eigenvalues are linearly independent.

Since this is the case, all eigenvectors are independent and, therefore, the
matrix A is diagonalizable with A=CDC-1, being

 3 0 1  3 0 0 1 0 6 
     
A   0 1 1 D   0 1 0  C  0 1 2 
 0 0 5  0 0 5  0 0 12 
     
 Matrix Power through Diagonalization
One of the properties of the similar matrices is that the power matrices are
also similar, that in case of diagonalizable matrices makes the calculus of
the power matrix easily.

If A is diagonalizable, then Dn  C 1 AnC

1
Premultiplying by C and postmultiplying by C both parts, we have
1
n
CD C 1 1 n 1
 CC A CC , from where we get CD C
n
A n

Let’s see an example:

 Matrix Power through Diagonalization
Let’s compute  3 0 1
 
the matrix n-th power A   0 1 1
 0 0 5
 

We have to check whether it is diagonalizable (that actually is)

The eigenvalues of the given matrix are 1  3, 2  1, 3  5 with the

associated eigenvectors
1  0 6
     
1  3  v1   0 ; 2  1  v2   1 ; 3  5  v3   2 
 0  0 12 
     
 Matrix Power through Diagonalization
Therefore, we know that we can We can also obtain the
construct the transition matrix C, such inverse matrix C, that is
that A=CDC-1, where
 1
1 0  
 3 0 1  3 0 0  2
   
A   0 1 1 D   0 1 0  1 
C  0 1 
1
 0 0 5  0 0 5  6
     1 
0 0 
 12 
1 0 6 
 
C  0 1 2 
 0 0 12 
 
 Matrix Power through Diagonalization
We compute now the matrix A n-th power as An=C-1DnC, where
 1  1
n1 0   1 0  
1 0 6  3 0 0  2 1 0 6  3 n
0 0  2
    1   
0  0
1
A  0
n
1 2  0 1 0  0 1   0 1 2  0 1 n
1  
 6   6
0 0 12  0 0 5   0 
 1   0 12  0 0 5n  1 
0 0  0 0 
 12   12 

 1  n 5 n  3n 
1 0   3 0 
 3n 0 3  5 
n
2  2 
 
n  1  5   1
n n

 0 1 n
25  0 1   0 1 n

0  6 6
0 12  5n  1   
 0 0 5 n

0 0   
 12 
 
 Matrix Power through Diagonalization

Point out that in order to obtain the given result it is almost impossible to
develop the n-th multiplication of
 3 0 1
 
A   0 1 1
 0 0 5
 
 Diagonalization of Symmetric Matrix
The matrix diagonalization has 2 peculiarities:

1. Every symmetric matrix is always diagonalizable, hence there is no

need to check for diagonalizability. If A is symmetic, it is for sure formed by
its eigenvectors (transition matrix, C) and allows to obtain a diagonal
1
matrix, D, similar to A, such that D  C AC
 Diagonalization of Symmetric Matrix

2. We can use the transition matrix obtained from orthonormalization

of column vectors of C (eigenvectors of A). This transition matrix U will be
then an orthonormal matrix, such that, D  U 1
AU  U AU
 Diagonalization of Symmetric Matrix
To get U we procede as follows:

• If A is symmetric with  1   2 ...   n  the eigenvectors of A are orthogonal

two-by-two; then we will get U by transforming them into unitary vectors.

• If A is symmetric with  i not all unequal  the eigenvectors of A

could be orthogonal or not; if they are not, we get U by employing the
Gram – Schmidt orthonormalization.
 Gram- Schmidt’s orthogonalization process
Let v , v ,..., v  / v
1 2 n i R n
be a non-orthonormal system:

We constranct an orhonormal system which is as follows:

y1
y1  v1 u1 
y1
y2
y 2  v 2  (u1  v 2 )  u1 u 2 
y2
y3
y 3  v 3  (u1  v 3 ) u1  (u 2  v 3 ) u 2 u 3 
y3
...
 Example. Diagonalization of Symmetric Matrix
 1 0 0 
 
Let’s see if the matrix A   0  2  1  is diagonalizable.
 0 1  2 
 

For this, we look for the diagonal matrix D, the transition matrix C, and the
transition orthogonal matrix U.

Firstly, observe that the matrix A is symmetric, and then it is diagonlaizable.

 Example. Diagonalization of Symmetric Matrix
We want to find eigenvalues looking for the system ( A  I )x  0 to be
consistent undetermined.
For that, we look for the rank of the matrix ( A  I ) , that is ( A  I )  0

 1 0 0   1 0 0  1   0 0
   
 0  2 1    0 1 0   0  2    1   1    2      1     0
2

 0 1  2   0 0 1  1  2  
    0
 Example. Diagonalization of Symmetric Matrix

 1 0 0   1 0 0 1   0 0
   
 0  2 1    0 1 0   0  2    1   1    2      1     0
2

 0 1  2  0 0 1  1  2  
    0

 1     2   
2

 1   1   3    4    1   1   3     0
2

The eigenvalues are then   3 y   1 (double).

We look for eigenvectors associated to each one.
 Example. Diagonalization of Symmetric Matrix

We firstly look for eigenvectors associated to 1  3 that will be all vectos for
which the following expression is satisfied:

A x  3x ( A  3I ) x  0

 1 0 0  x   1 0 0  x   0   2 0 0  x   0 
            
 0  2  1  y    3 0 1 0  y    0    0 1  1 y    0 
 0  1  2  z   0 0 1  z   0   0  1 1  z   0 
            
Example:How to obtain eigenvalues and eigenvectors
 2 0 0  x   0  The matrix rank is equal to 2, we will have to
    
 0 1  1 y    0  parametrize 1 of unknowns (and eliminate the last
 0  1 1  z   0 
     equation).

x  0  Therefore, the eigenvectors associated to eigenvalue -3

2x  0  
 y    are vectors (x,y,z) where x=0 but y and z are equal to a
y  z  0 
z    parameter.

 x  0  0 The eigenvectors associated to eigenvalue -3 are all

     
 y        1  vectors proportional to the vector (0,1,1). It is a vector
 z    1
      subspace of dimention equal to 1.
 Example. Diagonalization of Symmetric Matrix

We look for eigenvectors associated to 2  1 (that is double), that will be all

vectos for which the following expression is satisfied:

Ax   x ( A  I )x  0

 1 0 0  x   1 0 0  x   0   0 0 0  x   0 
            
 0  2  1  y    1 0 1 0  y    0    0  1  1 y    0 
 0  1  2  z   0 0 1  z   0   0  1  1 z   0 
            
 Example. Diagonalization of Symmetric Matrix
 0 0 0  x   0  The matrix rank is equal to 1, we will have to
    
 0  1  1 y    0  parametrize 2 of unknowns.
 0  1  1 z   0  The system is the following:
    
x    x        0  1 0
 y  z  0             
 y    y        0         0    1 
 y  z  0 
z     z      0      0   1
           

The eigenvectors associated to eigenvalue -1 are all vectors proportional to the

vectors (1,0,0) and (0,1,-1). It is a vector subspace of dimention equal to 2.
 Example. Diagonalization of Symmetric Matrix

Since we have already found eigenvalues of the matrix A and eigenvectors

associated to each eigenvalue, we can diagonalize the matrix A using the
transition matrix that is formed by eigenvectors:

D=C-1AC, where

 1 0 0  3 0 0  0 1 0   0 0.5 0.5 

      1  
A   0  2  1 , D   0  1 0 , C   1 0 1 , C   1 0 0 
 0 1  2   0 0  1    
    1 0  1  0 0.5  0.5 
 Example. Diagonalization of Symmetric Matrix

Moreover, the transition matrix that we will use is an orthogonal matrix U

which column vectors are the othonormal basis of vectors of C. The 3 vectors
are:
 0 1 0
     
u   1 , v   0 , w   1 
1  0   1
     

We can check that given vectors are already orthogonal, because

u  v  0; u  w  0; v  w  0 . Hence, there is no need to apply Gram – Schmidt
orthonormalization.
 Example. Diagonalization of Symmetric Matrix

To be these vectors an orthonormal vector we have to look for its proportional

and unitary vectors:

 0 1 0
         
1    0  1   
1  0  1
 0    0 
u    1  v 
   w   1  1 
1u     ;1v     0 ;1w     
u 1 1  2 
2 2 v 1  
2 w 1 1  2 
2 2

1  0  1
    
 2  2
 Example. Diagonalization of Symmetric Matrix
Once we get orthonormal vectors, we can use the transition orthogonal
matrix by diagonalizing our orthogonal matrix A as D=UtAU, since for an
orthogonal matrix U, its inverse matrix coincides with the transposed matrix:

 
   1 1 
 1 0 0  3 0 0   0 1 0  0 
     1 1  t 1
 2 2 
A   0  2  1 , D   0  1 0 ,U   0 , U  U   1 0 0 
 0 1  2   0 0  1 
2 2
 0 1 1 
 
    1 1  
 0    2 2
 2 2