Matrix Algebra: MTH-174_Beyond Syllabus Prof. U.

Sarkar

Diagonalization: If 𝜆1 , 𝜆2 , … , 𝜆𝑛 be 𝑛 Eigen values and 𝑋1 , 𝑋2 , … , 𝑋𝑛 be the corresponding linearly

independent Eigen vectors of a square matrix 𝐴 of order 𝑛, then 𝑃−1 𝐴𝑃 is a diagonal matrix, where
𝑃 = [𝑋1 , 𝑋2 , … , 𝑋𝑛 ] is called the modal matrix and the diagonal matrix 𝐷 = 𝑃−1 𝐴𝑃 is given by
𝜆1 0 … … 0
0 𝜆 … … 0
𝐷 = … …2 … … …
… … … … …
0 0 … … 𝜆𝑛

5 0 1
Example 1: Let 𝐴 = 0 −2 0 . Find the modal matrix 𝑃 such that 𝑃−1 𝐴𝑃 is a diagonal matrix.
1 0 5
5 0 1
Solution: Let 𝜆 be the Eigen value of the given matrix 𝐴 = 0 −2 0 . Then we have
1 0 5
5−𝜆 0 1
𝐴 − 𝜆𝐼 = 0 ⇒ 0 −2 − 𝜆 0 = 0 ⇒ 𝜆 = −2, 4, 6.
1 0 5−𝜆

So the Eigen values are −2, 4 and 6.

𝑥
Let 𝑋 = 𝑦 be an Eigen vector. Then
𝑧

for 𝜆 = −2 we have

5+2 0 1 𝑥 0
7𝑥 + 𝑧 = 0
0 −2 + 2 0 𝑦 = 0 ⇒ ⇒𝑥=0=𝑧
𝑥 + 7𝑧 = 0
1 0 5+2 𝑧 0
0 0 0
Hence for any non-zero value of 𝑘, the Eigen vector is given by 𝑘 = 𝑘 1 , i.e., 1 = 𝑋1 .
0 0 0

for 𝜆 = 4 we have

5−4 0 1 𝑥 0 𝑥+𝑧 =0
0 −2 − 4 0 𝑦 = 0 ⇒ ⇒ 𝑧 = −𝑥, 𝑦 = 0
6𝑦 = 0
1 0 5−4 𝑧 0
𝑘 1 1
Hence for any non-zero value of 𝑘, the Eigen vector is given by 0 = 𝑘 0 , i.e., 0 = 𝑋2 .
−𝑘 −1 −1

for 𝜆 = 6 we have

5−6 0 1 𝑥 0 𝑥−𝑧 =0
0 −2 − 6 0 𝑦 = 0 ⇒ ⇒ 𝑧 = 𝑥, 𝑦 = 0
8𝑦 = 0
1 0 5−6 𝑧 0
 Matrix Algebra: MTH-174_Beyond Syllabus Prof. U. Sarkar

𝑘 1 1
Hence for any non-zero value of 𝑘, the Eigen vector is given by 0 = 𝑘 0 , i.e., 0 = 𝑋3 .
𝑘 1 1

Therefore, the modal matrix 𝑃 is given by

0 1 1
𝑃 = 𝑋1 , 𝑋2 , … , 𝑋𝑛 = 1 0 0
0 −1 1

And the diagonal matrix 𝑃 −1 𝐴𝑃 will be

−2 0 0
𝐷 = 𝑃−1 𝐴𝑃 = 0 4 0
0 0 6

 Jacobi’s Method (Method for computing all eigen values)

Using this method we can compute all the eigen values of a symmetric matrix. In this method we find
eigen values and corresponding eigen vectors for a given matrix A by performing a series of similarity
transformation of the form

𝐴𝑟+1 = 𝐽𝑟𝑇 𝐴𝑟 𝐽𝑟 (𝑟 = 1,2,3, … )

with the assumption 𝐴1 = 𝐴. Here 𝐽𝑟 is called the Jacobi matrix (or rotation matrix) at the rth step.

Following steps are adopted in the Jacobi method:

 Find the pth and qth row and column which correspond to the off diagonal element having highest
value.
 Compute the Jacobi matrix 𝐽(𝑝, 𝑞, 𝜃) after calculating the angle (𝜃) of similarity rotation using
the formula
2𝑎 𝑝𝑞
tan 2𝜃 = 𝑎 −𝑎 𝑞𝑞
𝑝𝑝

 Apply the Jacobi matrix to the matrix as the

way mentioned above. And do the same till
when the matrix converted completely into a
diagonal matrix. The diagonal element will be
the eigenvalues.
 The eigenvectors will be the columns of the product of the Jacobi matrices.
Example 2: Find all the eigen values and corresponding eigen vectors of the following symmetric matrix.

1 2 2
𝐴= 2 3 2
2 2 1

Solution: We have
 Matrix Algebra: MTH-174_Beyond Syllabus Prof. U. Sarkar

1 2 2
𝐴= 2 3 2 = 𝐴1
2 2 1

Step 1: The numerically largest off-diagonal element of 𝐴1 is 𝑎13 = 𝑎31 = 2 which gives 𝑝 = 1 and
𝑞 = 3. So, from the matrix, we get 𝑎11 = 1 and 𝑎33 = 1. Therefore,

2𝑎13 2×2 𝜋
tan 2𝜃 = = =∞⇒𝜃=
𝑎11 − 𝑎33 1 − 1 4
Thus

1 1
0 −
cos 𝜃 0 − sin 𝜃 2 2
𝐽1 = 0 1 0 = 0 1 0
sin 𝜃 0 cos 𝜃 1 1
0
2 2

Therefore,

1 1 1 1
0 0 −
2 2 1 2 2 2 2 3 2 0
𝐴2 = 𝐽1𝑇 𝐴1 𝐽1 = 0 1 0 2 3 2 0 1 0 = 2 3 0
1 1 2 2 1 1 1 0 0 −1
− 0 0
2 2 2 2

Step 2: The numerically largest off-diagonal element of 𝐴2 is 𝑎12 = 𝑎21 = 2 which gives 𝑝 = 1 and
𝑞 = 2. So, from the matrix, we get 𝑎11 = 3 and 𝑎22 = 3. Therefore,

2𝑎12 2×2 𝜋
tan 2𝜃 = = =∞⇒𝜃=
𝑎11 − 𝑎22 3 − 3 4
Thus

1 1
− 0
cos 𝜃 − sin 𝜃 0 2 2
𝐽2 = sin 𝜃 cos 𝜃 0 = 1 1
0 0 1 0
2 2
0 0 1

Therefore,

1 1 1 1
0 − 0
2 2 3 2 0 2 2 5 0 0
𝐴3 = 𝐽2𝑇 𝐴2 𝐽2 = 1 1 2 3 0 1 1 = 0 1 0
− 0 0 0 −1 0 0 0 −1
2 2 2 2
0 0 1 0 0 1

We see that 𝐴3 is a diagonal matrix and hence the eigen values are 5, 1 and −1.
 Matrix Algebra: MTH-174_Beyond Syllabus Prof. U. Sarkar

Now,

1 1 1
1 1 1 1 − −
0 − − 0 2 2 2
2 2 2 2 1 1
𝐽1 × 𝐽2 = 0 1 0 1 1 = 0
1 1 0 2 2
0 2 2 1 1 1
2 2 0 0 1 −
2 2 2

Therefore, the corresponding eigen vectors are:

1
2
1
𝑋1 = 2
corresponding to the eigen value 𝜆1 = 5
1
2

1
−
2
1
𝑋2 = 2
corresponding to the eigen value 𝜆2 = 1
1
−2

and
1
− 2
𝑋3 = 0 corresponding to the eigen value 𝜆3 = −1.
1
2

 Power Method (Method for computing selected eigen values)

Using power method, one may find the largest or smallest eigen value and corresponding eigen vecor.

 Power method to find the largest Eigen value and Eigen vector:

The method is very simple. First take an initial Eigen vector 𝑋0 . Then

(i) Compute 𝐴𝑋0 = 𝑘1 𝑋1 where 𝑘1 is the numerically largest value of the components of 𝐴𝑋0 .
(ii) Compute 𝐴𝑋1 = 𝑘2 𝑋2 where 𝑘2 is the numerically largest value of the components of 𝐴𝑋1 .
and so on.
𝑘 𝑛 +1 −𝑘 𝑛
The iteration will be stopped if the percentage error 𝑘 𝑛 +1
× 100%, 𝑛 = 1,2, … is very small. Then
𝑘𝑛+1 will be called the largest Eigen value and 𝑋𝑛+1 is the corresponding Eigen vector.

Example 3: Find the largest Eigen value and the corresponding Eigen vector of the matrix 𝐴 =
1 6 1
1 2 0 .
0 0 3
 Matrix Algebra: MTH-174_Beyond Syllabus Prof. U. Sarkar

1
Solution: Let the initial Eigen vector be 𝑋0 = 0 .
0
Now

1 6 1 1 1 1 1
(i) 𝐴𝑋0 = 1 2 0 0 = 1 = 1 × 1 = 𝑘1 𝑋1 where 𝑘1 = 1 and 𝑋1 = 1 .
0 0 3 0 0 0 0
1 6 1 1 7 1 1
(ii) 𝐴𝑋1 = 1 2 0 1 = 3 = 7 × 0.428 = 𝑘2 𝑋2 where 𝑘2 = 7 and 𝑋2 = 0.428 .
0 0 3 0 0 0 0
𝑘 2 −𝑘 1
Percentage error = 𝑘
× 100% = 85.71%, a large quantity.
2
1 6 1 1 3.568 1 1
(iii) 𝐴𝑋2 = 1 2 0 0.428 = 1.856 = 3.568 × 0.52 = 𝑘3 𝑋3 where 𝑘3 = 3.568 and 𝑋3 = 0.52 .
0 0 3 0 0 0 0
𝑘 3 −𝑘 2
Percentage error = × 100% = 96.18%, a large quantity.
𝑘3
1 6 1 1 4.12 1 1
(iv) 𝐴𝑋3 = 1 2 0 0.52 = 2.04 = 4.12 × 0.495 = 𝑘4 𝑋4 where 𝑘4 = 4.12 and 𝑋4 = 0.495 .
0 0 3 0 0 0 0
𝑘 4 −𝑘 3
Percentage error = 𝑘4
× 100% = 13.4%, a moderate quantity.
1 6 1 1 3.97 1 1
(v) 𝐴𝑋4 = 1 2 0 0.495 = 1.99 = 3.97 × 0.5 = 𝑘5 𝑋5 where 𝑘5 = 3.97 and 𝑋5 = 0.5 .
0 0 3 0 0 0 0
𝑘 5 −𝑘 4
Percentage error = × 100% = 3.78%, quite small quantity.
𝑘5
1 6 1 1 4 1 1
(vi) 𝐴𝑋5 = 1 2 0 0.5 = 2 = 4 × 0.5 = 𝑘6 𝑋6 where 𝑘6 = 4 and 𝑋6 = 0.5 .
0 0 3 0 0 0 0
𝑘 6 −𝑘 5
Percentage error = 𝑘6
× 100% = 0.75%, very small quantity.

1
Hence the largest Eigen value is 4 and corresponding Eigen vector is 0.5 .
0

 Power method to find the smallest Eigen value and Eigen vector: The smallest Eigen value is the
reciprocal of the largest Eigen value of 𝐴−1 .

6 5
Example 4: Find the smallest Eigen value and corresponding Eigen vector of the matrix 𝐴 = .
1 2
1 2 −5
Solution: Let 𝐵 = 𝐴−1 = 7
−1 6
1
Let the initial Eigen vector be 𝑋0 = .
1
Now

1 2 −5 1 1 −3 −0.6 −0.6
(i) 𝐵𝑋0 = 7 = = 0.71 × = 𝑘1 𝑋1 where 𝑘1 = 0.71 and 𝑋1 = .
−1 6 1 7 5 1 1
 Matrix Algebra: MTH-174_Beyond Syllabus Prof. U. Sarkar

1 2 −5 −0.6 1 −6.2 −0.94 −0.94

(ii) 𝐵𝑋1 = 7 = = 0.94 × = 𝑘2 𝑋2 where 𝑘2 = 0.94 and 𝑋2 = .
−1 6 1 7 6.6 1 1
𝑘 2 −𝑘 1
Percentage error = 𝑘2
× 100% = 24.47%
1 2 −5 −0.94 1 −6.88 −0.99 −0.99
(iii) 𝐵𝑋2 = 7 = = 0.99 × = 𝑘3 𝑋3 where 𝑘3 = 0.99 and 𝑋3 = .
−1 6 1 7 6.94 1 1
𝑘 3 −𝑘 2
Percentage error = 𝑘3
× 100% = 5.05%
−0.999
(iv) 𝐵𝑋3 = −0.6 0.4 −0.99
=
1 −6.98
= 0.999 ×
−0.999
= 𝑘4 𝑋4 where 𝑘4 = 0.999 and 𝑋4 = .
0.8 −0.2 1 7 6.99 1 1
𝑘 4 −𝑘 3
Percentage error = 𝑘4
× 100% = 0.9%

−1
Hence the largest Eigen value of 𝐴−1 is 0.999, i.e. 1 and corresponding Eigen vector is .
1
−1
Therefore, the smallest Eigen value of 𝐴 is 1/1 = 1 and the corresponding Eigen vector is .
1