Mathematical Methods

in Electrical Engineering
Lecture 𝟎𝟒 – Diagonalization

Dr. Elie Abou Diwan

 Eigenvalues and Eigenvectors
DEFINITION:
Let 𝐴 be an 𝑛 × 𝑛 matrix. A complex number λ is called an eigenvalue of 𝐴 if there exists a
nonzero vector 𝑣 ∈ C 𝑛 such that 𝐴𝑣 = λ𝑣.
The vector 𝑣 is called an eigenvector corresponding to λ.
The eigenspace corresponding to λ is the set of all vectors satisfying 𝐴𝑣 = λ𝑣.

THEOREM:
Let 𝐴 be an 𝑛 × 𝑛 matrix. The characteristic polynomial of 𝐴, denoted by 𝑝(λ), is the polynomial
defined by 𝑝 λ = det⁡(𝐴 − λ𝐼𝑛 ).
The eigenvalues of 𝐴 are the roots of its characteristic polynomial i.e. the roots of the equation
det 𝐴 − λ𝐼𝑛 = 0.
 Eigenvalues and Eigenvectors (cont.)
Proof:
If λ is an eigenvalue, then there exists a vector 𝑣 ≠ 0 such that 𝐴𝑣 = λ𝑣.
⇒ There exists a vector 𝑣 ≠ 0 such that 𝐴𝑣 − λ𝐼𝑛 𝑣 = 0.
⇒ There exists a vector 𝑣 ≠ 0 such that (𝐴 − λ𝐼𝑛 )𝑣 = 0.
⇒ There exists a vector 𝑣 ≠ 0 such that 𝐵𝑣 = 0 (homogeneous linear system).
If 𝐵 is invertible, then the only solution to the homogeneous linear system 𝐵𝑣 = 0 is the trivial
solution 𝑣 = 0. But here 𝑣 ≠ 0, then 𝐵 is not invertible.
Consequently, det 𝐵 = det⁡(𝐴 − λ𝐼𝑛 ) = 0.

THEOREM:
The eigenspace 𝑆λ is the set of all solutions of the system (𝐴 − λ𝐼𝑛 )𝑣 = 0.
 Eigenvalues and Eigenvectors (cont.)
EXAMPLE 𝟏:
0 0 −1
Let 𝐴 = 1 2 −2 . Find the eigenvalues and bases for the eigenspace of 𝐴.
1 0 0

To find the eigenvalues of 𝐴, we solve the equation det⁡(𝐴 − λ𝐼3 ) = 0.

0 0 −1 1 0 0 0 0 −1 λ 0 0 −λ 0 −1
𝐴 − λ𝐼3 = 1 2 −2 − λ 0 1 0 = 1 2 −2 − 0 λ 0 = 1 2 − λ −2
1 0 0 0 0 1 1 0 0 0 0 λ 1 0 −λ
−λ 0 −1
det 𝐴 − λ𝐼3 = 1 2 − λ −2 = −λ 2 − λ −λ + 1 2 − λ = 2 − λ λ2 + 1
1 0 −λ
det 𝐴 − λ𝐼3 = 0 ⇒ 2 − λ λ2 + 1 = 0
2−λ=0⇒λ=2
λ2 + 1 = 0 ⇒ λ = ±𝑖
Finally, λ1 = 2, λ2 = 𝑖, λ3 = −𝑖 are the eigenvalues of 𝐴.
 Eigenvalues and Eigenvectors (cont.)
Eigenvector corresponding to eigenvalue λ1 = 2:
𝐴 − λ1 𝐼3 𝑣⁡ = 0
−2 0 −1 𝑥 0
⇒ 1 0 −2 𝑦 = 0
1 0 −2 𝑧 0
𝑥=0
⇒ 𝑦=𝑦
𝑧=0
𝑥 0 0
⇒𝑣= 𝑦 = 𝑦 =𝑦 1
𝑧 0 0
0
Therefore, 𝑣1 = 1 is a basis for the eigenspace 𝑆1 .
0
 Eigenvalues and Eigenvectors (cont.)
Eigenvector corresponding to eigenvalue λ2 = 𝑖:
𝐴 − λ2 𝐼3 𝑣 = 0
−𝑖 0 −1 𝑥 0
⇒ 1 2 − 𝑖 −2 𝑦 = 0
1 0 −𝑖 𝑧 0
𝑥 = 𝑖𝑧
⇒ 𝑦 = 𝑧⁡
𝑧 = 𝑧⁡
𝑥 𝑖𝑧 𝑖
⇒𝑣= 𝑦 = 𝑧 =𝑧 1
𝑧 𝑧 1
𝑖
Therefore, 𝑣2 = 1 is a basis for the eigenspace 𝑆2 .
1
 Eigenvalues and Eigenvectors (cont.)
Eigenvector corresponding to eigenvalue λ3 = −𝑖:
𝐴 − λ3 𝐼3 𝑣 = 0
𝑖 0 −1 𝑥 0
⇒ 1 2 + 𝑖 −2 𝑦 = 0
1 0 𝑖 𝑧 0
𝑥 = −𝑖𝑧⁡⁡
⇒ 𝑦 = 𝑧⁡⁡⁡⁡⁡⁡
𝑧 = 𝑧⁡⁡⁡⁡⁡⁡⁡
𝑥 −𝑖𝑧 −𝑖
⇒𝑣= 𝑦 = 𝑧 =𝑧 1
𝑧 𝑧 1
−𝑖
Therefore, 𝑣3 = 1 is a basis for the eigenspace 𝑆3 .
1

Remark: 𝑣3 = 𝑣2∗ since λ3 = 𝜆∗2

 Eigenvalues and Eigenvectors (cont.)
EXAMPLE 𝟐:
𝐴 is an 𝑛 × 𝑛 matrix whose eigenvalues are λ1 , λ2 , … , λ𝑛 . Find the eigenvalues of the matrix α𝐴.

The eigenvalues of the matrix α𝐴 are the roots of the equation det⁡(α𝐴 − λ′𝐼𝑛 ) = 0.
λ′
⇒ det α 𝐴 − 𝐼𝑛 = 0
α
λ′
⇒ α𝑛 det 𝐴 − 𝐼𝑛 = 0 because det α𝐵 = α𝑛 det⁡(𝐵)
α
Since λ is an eigenvalue of 𝐴, then det 𝐴 − λ𝐼𝑛 = 0.
By identification:
λ′
λ = ⇒ λ′ = αλ
α
Consequently, the eigenvalues of the matrix α𝐴 are αλ1 , αλ2 , … , αλ𝑛 .
 Eigenvalues and Eigenvectors (cont.)
EXAMPLE 𝟑:
Show that 𝐴−1 exists if and only if λ1 , λ2 , … , λ𝑛 ≠ 0.

Since this is an if and only if statement, then we must prove both implications. That is:
• If 𝐴−1 exists, then λ1 , λ2 , … , λ𝑛 ≠ 0
• If λ1 , λ2 , … , λ𝑛 ≠ 0, then 𝐴−1 exists

If 𝐴−1 exists, then det⁡(𝐴) ≠ 0

⇒ det⁡(𝐴 − 0𝐼𝑛 ) ≠ 0
⇒ λ = 0 cannot be an eigenvalue (because if 0 is an eigenvalue, then we should have
det 𝐴 − 0𝐼𝑛 = 0). Consequently, λ1 , λ2 , … , λ𝑛 ≠ 0.

If λ is an eigenvalue, then det 𝐴 − λ𝐼𝑛 = 0. For λ = 0, det 𝐴 − 0𝐼𝑛 = det 𝐴 = 0. 𝐴 is not

invertible in this case. Consequently, if λ1 , λ2 , … , λ𝑛 ≠ 0, then 𝐴 is invertible.

Remark: 𝐴 is not invertible if and only if 0 is an eigenvalue of 𝐴.

 Similar Matrices & Diagonalization
DEFINITION:
Two 𝑛 × 𝑛 matrices 𝐴 and 𝐵 are said to be similar if there exists an invertible matrix 𝑃 such that
𝐵 = 𝑃−1 𝐴𝑃. (𝐴 is similar to 𝐵 ⇔ 𝐴~𝐵 ⇔ 𝑃−1 𝐴𝑃 = 𝐵)

Remark: Two similar matrices have the same eigenvalues.

Proof:
Let 𝐴 and 𝐵 be two similar matrices. So there exists an invertible matrix 𝑃 such that 𝐵 = 𝑃−1 𝐴𝑃.
If λ is an eigenvalue of 𝐵, then det 𝐵 − λ𝐼𝑛 = 0
⇒ det 𝑃−1 𝐴𝑃 − λ𝐼𝑛 = 0
⇒ det 𝑃−1 𝐴𝑃 − λ𝑃−1 𝐼𝑛 𝑃 = 0
(λ𝑃−1 𝐼𝑛 𝑃 = λ𝑃−1 𝑃 = λ𝐼𝑛 where 𝑃−1 𝑃 = 𝐼𝑛 because 𝑃 is invertible)
⇒ det 𝑃−1 𝐴𝑃 − 𝑃−1 λ𝐼𝑛 𝑃 = 0
⇒ det 𝑃−1 𝐴 − λ𝐼𝑛 𝑃 = 0
 Similar Matrices & Diagonalization (cont.)
⇒ det 𝑃−1 det 𝐴 − λ𝐼𝑛 det⁡(𝑃) = 0 since det 𝐴𝐵 = det 𝐴 det⁡(𝐵)
1 −1
1
⇒ det 𝐴 − λ𝐼𝑛 det 𝑃 = 0⁡since⁡𝑃⁡is⁡invertible, then det 𝑃 =
det⁡(𝑃) det⁡(𝑃)
⇒ det 𝐴 − λ𝐼𝑛 = 0
⇒ λ is an eigenvalue of 𝐴.

DEFINITION:
A 𝑛 × 𝑛 matrix 𝐴 is said to be diagonalizable if it is similar to a diagonal matrix 𝐵. In other words,
if matrix 𝐴 is similar to a diagonal matrix 𝐵, then there exits an invertible matrix 𝑃 such that
𝑃−1 𝐴𝑃 = 𝐵 ⇔ 𝐴 = 𝑃𝐵𝑃−1
 Similar Matrices & Diagonalization (cont.)
Why is this useful?
Suppose you wanted to find 𝐴3 .
If 𝐴 is diagonalizable, then 𝐴3 = 𝑃𝐵𝑃−1 3 where 𝐵 is a diagonal matrix and 𝑃 is an invertible
matrix.
⇒ 𝐴3 = 𝑃𝐵𝑃−1 𝑃𝐵𝑃−1 𝑃𝐵𝑃−1 = 𝑃𝐵 𝑃−1 𝑃 𝐵 𝑃−1 𝑃 𝐵𝑃−1 = 𝑃𝐵𝐼𝑛 𝐵𝐼𝑛 𝐵𝑃−1 = 𝑃𝐵3 𝑃−1
In general, 𝐴𝑘 = 𝑃𝐵𝑘 𝑃−1
7 0 0
Or powers of diagonal matrices are relatively easy to compute. For example, if 𝐵 = 0 −2 0
0 0 3
73 0 0
then 𝐵3 = 0 −2 3 0
0 0 33
This means that finding 𝐴𝑘 involves only two matrix multiplications instead of the 𝑘 matrix
multiplications that would be necessary to multiply 𝐴 by itself 𝑘 times.
 Similar Matrices & Diagonalization (cont.)
REMARK:
The main diagonal entries are the eigenvalues of a diagonal matrix. We can use this property to
get the eigenvalues of a similar matrix because two similar matrices have the same eigenvalues.

THEOREM:
If 𝐴 is an 𝑛 × 𝑛 matrix and 𝐴 has 𝑛 linearly independent eigenvectors, then 𝐴 is diagonalizable
(similar to a diagonal matrix 𝐵).
 Similar Matrices & Diagonalization (cont.)
Proof:
Suppose that 𝑛 × 𝑛 matrix 𝐴 has eigenvalues λ1 , λ2 , … , λ𝑛 with corresponding linearly
independent eigenvectors 𝑣1 , 𝑣2 , … , 𝑣𝑛 .

λ1 0 0 … 0
0 λ2 0 … 0
Let 𝐵 be the diagonal matrix 0 0 λ3 … 0
⋮ ⋮ ⋮ ⋱ ⋮
0 0 0 … λ𝑛

𝑝11 𝑝12 𝑝13 … 𝑝1𝑛

𝑝21 𝑝22 𝑝23 … 𝑝2𝑛
Let 𝑃 be the matrix 𝑝31 𝑝32 𝑝33 … 𝑝3𝑛 = 𝑣1 𝑣2 𝑣3 … 𝑣𝑛
⋮ ⋮ ⋮ ⋱ ⋮
𝑝𝑛1 𝑝𝑛2 𝑝𝑛3 … 𝑝𝑛𝑛
 Similar Matrices & Diagonalization (cont.)
Matrix 𝑃 is invertible because its columns form a linearly independent set.

Furthermore:
𝐴𝑃 = 𝐴 𝑣1 𝑣2 𝑣3 … 𝑣𝑛 = 𝐴𝑣1 𝐴𝑣2 𝐴𝑣3 … 𝐴𝑣𝑛
Or 𝐴𝑣 = λ𝑣, then:
𝐴𝑃 = λ1 𝑣1 λ2 𝑣2 λ3 𝑣3 … λ𝑛 𝑣𝑛

On the other hand:

λ1 0 0 … 0
0 λ2 0 … 0
𝑃𝐵 = 𝑣1 𝑣2 𝑣3 … 𝑣𝑛 0 0 λ3 … 0 = λ1 𝑣1 λ2 𝑣2 λ3 𝑣3 … λ𝑛 𝑣𝑛
⋮ ⋮ ⋮ ⋱ ⋮
0 0 0 … λ𝑛
 Similar Matrices & Diagonalization (cont.)
Consequently, 𝐴𝑃 = 𝑃𝐵
⇒ 𝐴𝑃𝑃−1 = 𝑃𝐵𝑃−1
⇒ 𝐴𝐼𝑛 = 𝑃𝐵𝑃−1 where 𝑃𝑃−1 = 𝐼𝑛 because 𝑃 is invertible
⇒ 𝐴 = 𝑃𝐵𝑃−1
Thus, matrix 𝐴 is diagonalizable because there exists an invertible matrix 𝑃 such that 𝐴 = 𝑃𝐵𝑃−1
where 𝐵 is a diagonal matrix.

Remark: This proof shows us how, in case 𝑛 × 𝑛 matrix 𝐴 has 𝑛 linearly independent
eigenvectors, to find both a diagonal matrix 𝐵 to which 𝐴 is similar and an invertible matrix 𝑃 for
which 𝐴 = 𝑃𝐵𝑃−1 . We state this as a corollary.
 Similar Matrices & Diagonalization (cont.)
COROLLARY:
If 𝐴 is an 𝑛 × 𝑛 matrix and 𝐴 has 𝑛 linearly independent eigenvectors 𝑣1 , 𝑣2 , … , 𝑣𝑛 with
corresponding eigenvalues λ1 , λ2 , … , λ𝑛 , then 𝐴 = 𝑃𝐵𝑃−1 where 𝐵 is the diagonal matrix
λ1 0 0 … 0
0 λ2 0 … 0
0 0 λ3 … 0
⋮ ⋮ ⋮ ⋱ ⋮
0 0 0 … λ𝑛
𝑝11 𝑝12 𝑝13 … 𝑝1𝑛
𝑝21 𝑝22 𝑝23 … 𝑝2𝑛
and 𝑃 is the invertible matrix 𝑝31 𝑝32 𝑝33 … 𝑝3𝑛 = 𝑣1 𝑣2 𝑣3 … 𝑣𝑛
⋮ ⋮ ⋮ ⋱ ⋮
𝑝𝑛1 𝑝𝑛2 𝑝𝑛3 … 𝑝𝑛𝑛
 Similar Matrices & Diagonalization (cont.)
REMARKS:
λ1 0 0 … 0 λ1 𝑚 0 0 … 0
0 λ2 0 … 0 0 λ2 𝑚 0 … 0
REMARK 𝟏: If 𝐵 = 0 0 λ3 … 0 , then 𝐵𝑚 = 0 0 λ3 𝑚 … 0
⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋮ ⋱ ⋮
0 0 0 … λ𝑛 0 0 0 … λ𝑛 𝑚

REMARK 𝟐: If 𝐴 is diagonalizable, then 𝐴𝑚 = 𝑃𝐵𝑚 𝑃−1

 Similar Matrices & Diagonalization (cont.)
EXAMPLE 𝟏:
13 −8
Show that matrix 𝐴 = is diagonalizable.
25 −17

First, we find the eigenvalues of 𝐴:

13 − λ −8
det 𝐴 − λ𝐼2 = 0 ⇒ = 0 ⇒ 13 − λ −17 − λ + 200 = 0
25 −17 − λ
2
⇒ λ + 4λ − 21 = 0
∆′ = 𝑏 ′2 − 𝑎𝑐 = 4 − 1 −21 = 25
−𝑏 ′ − ∆′
λ1 = = −7
⇒ 𝑎
−𝑏 ′ + ∆′
λ2 = = 3⁡⁡⁡
𝑎

Finally, λ1 = −7, and λ2 = 3 are the eigenvalues of 𝐴.

 Similar Matrices & Diagonalization (cont.)
Eigenvector corresponding to eigenvalue λ1 = −7:
𝐴 − λ1 𝐼2 𝑣 = 0

20 −8 𝑥 0
⇒ 𝑦 =
25 −10 0

2
⇒ 𝑥 = 5𝑦
𝑦 = 𝑦⁡⁡⁡
𝑥 2 2
⇒ 𝑣 = 𝑦 = 5𝑦 = 𝑦 5
𝑦 1

2/5
Eigenvector 𝑣1 =
1
 Similar Matrices & Diagonalization (cont.)
Eigenvector corresponding to eigenvalue λ2 = 3:
𝐴 − λ2 𝐼2 𝑣 = 0
10 −8 𝑥 0
⇒ =
25 −20 𝑦 0
4
⇒ 𝑥 = 5𝑦
𝑦 = 𝑦⁡⁡⁡
𝑥 4 4
⇒ 𝑣 = 𝑦 = 5𝑦 = 𝑦 5
𝑦 1
4/5
Eigenvector 𝑣2 =
1

2/5 4/5
Finally, since the vectors and are linearly independent, we conclude that 𝐴 = 𝑃𝐵𝑃−1
1 1
−7 0 2/5 4/5
where 𝐵 is the diagonal matrix and 𝑃 is the invertible matrix .
0 3 1 1
 Similar Matrices & Diagonalization (cont.)
EXAMPLE 𝟐:
0 0 −1
Let 𝐴 = 1 2 −2 . Find 𝐴𝑛 .
1 0 0

STEP 𝟏: Find the eigenvalues and the corresponding eigenvectors:

0
Eigenvalue: 2, eigenvector: 1
0
𝑖
Eigenvalue: 𝑖, eigenvector: 1
1
−𝑖
Eigenvalue: −𝑖, eigenvector: 1
1
 Similar Matrices & Diagonalization (cont.)
0 𝑖 −𝑖
STEP 𝟐: Check if 1 , 1 and 1 are linearly independent:
0 1 1

0 𝑖 −𝑖 0 𝑖 −𝑖
1 1 1 = −2𝑖 ≠ 0 ⇒ 1 , 1 and 1 are linearly independent.
0 1 1 0 1 1

STEP 𝟑: Find the diagonal matrix 𝐵 and the invertible matrix 𝑃:

0 𝑖 −𝑖
Since the vectors 1 , 1 and 1 are linearly independent, we conclude that 𝐴 = 𝑃𝐵𝑃−1
0 1 1
2 0 0 0 𝑖 −𝑖
where 𝐵 is the diagonal matrix 0 𝑖 0 and 𝑃 is the invertible matrix 1 1 1 .
0 0 −𝑖 0 1 1
 Similar Matrices & Diagonalization (cont.)
STEP 𝟒: Find 𝑃−1 :

𝑎𝑑𝑗(𝑃) 𝐶𝑡 𝐶11 𝐶12 𝐶13 0 −1 1

𝑃−1 = = ⁡where⁡𝐶 = 𝐶21 𝐶22 𝐶23 = −2𝑖 0 0
det⁡(𝑃) det⁡(𝑃) 𝐶31 𝐶32 𝐶33 2𝑖 −𝑖 −𝑖

0 −2𝑖 2𝑖
𝐶 𝑡 = −1 0 −𝑖
1 0 −𝑖

0 𝑖 −𝑖
det 𝑃 = 1 1 1 = −2𝑖
0 1 1

0 1 −1
𝑃−1 = −𝑖 2 0 1/2
𝑖/2 0 1/2
 Similar Matrices & Diagonalization (cont.)
STEP 𝟓: Find 𝐴𝑛 :
0 𝑖 −𝑖 2𝑛 0 0 0 1 −1
𝐴𝑛 = 𝑃𝐵𝑛 𝑃−1 = 1 1 1 0 𝑖𝑛 0 − 𝑖 2 0 1/2
0 0 −𝑖 𝑛 𝑖/2 0 1/2
0 1 1

𝑖 𝑛 + −𝑖 𝑛 𝑖(𝑖 𝑛 ) − 𝑖 −𝑖 𝑛
0
0 𝑖(𝑖 𝑛 ) −𝑖 −𝑖 𝑛 0 1 −1 2 2
−𝑖 𝑖 𝑛 + 𝑖 −𝑖 𝑛
𝑖 𝑛
+ −𝑖 𝑛
𝐴𝑛 = 2𝑛 𝑖𝑛 −𝑖 𝑛 − 𝑖 2 0 1/2 = 2𝑛 −2𝑛 +
0 𝑖𝑛 −𝑖 𝑛 𝑖/2 0 1/2 2 2
−𝑖 𝑖 𝑛 + 𝑖 −𝑖 𝑛
𝑖 𝑛 + −𝑖 𝑛
0
2 2
 Cayley-Hamilton Theorem
THEOREM:
Suppose 𝐴 is an 𝑛 × 𝑛 matrix. The Cayley-Hamilton theorem states that 𝐴 satisfies its own
characteristic polynomial. In other words, if 𝑝 λ = det⁡(𝐴 − λ𝐼𝑛 ) is the characteristic polynomial
of 𝐴 then 𝑝 𝐴 = 0.

EXAMPLE:
2 5
Let 𝐴 = . Use the Cayley-Hamilton theorem to find 𝐴−1 .
2 −1
2−λ 5
𝑝 λ = det 𝐴 − λ𝐼2 = = λ2 − λ − 12
2 −1 − λ
By Cayley-Hamilton theorem, 𝑝 𝐴 = 0
2 2
𝐴 − 𝐼2
⇒ 𝐴 − 𝐴 − 12𝐼2 = 0 ⇒ 𝐴 − 𝐴 = 12𝐼2 ⇒ 𝐴 𝐴 − 𝐼2 = 12𝐼2 ⇒ 𝐴 = 𝐼2
12
𝐴 − 𝐼2 1 1 5
⇒ 𝐴−1 = =
12 12 2 −2
 Matrix Differential Equation
SYSTEM OF LINEAR DIFFERENTIAL EQUATIONS (LDE):
𝑥′1 = 𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ + 𝑎1𝑛 𝑥𝑛
𝑥′2 = 𝑎21 𝑥1 + 𝑎22 𝑥2 + ⋯ + 𝑎2𝑛 𝑥𝑛
Consider the system
⋮
𝑥′𝑛 = 𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + ⋯ + 𝑎𝑛𝑛 𝑥𝑛
𝑥1 𝑥′1
𝑥2 𝑥′2
Let 𝑥 = ⋮ and 𝑥 ′ = , then the system can be written in the form 𝑥 ′ = 𝐴𝑥 where
⋮
𝑥𝑛 𝑥′𝑛
𝑎11 𝑎12 … 𝑎1𝑛
𝑎21 𝑎22 … 𝑎2𝑛
𝐴= ⋮ ⋮ ⋱ ⋮
𝑎𝑛1 𝑎𝑛2 … 𝑎𝑛𝑛
 Matrix Differential Equation (cont.)
THEOREM:
Let 𝑣 be an eigenvector corresponding to an eigenvalue λ of matrix 𝐴, then 𝑥 𝑡 = 𝑒 λ𝑡 𝑣 is a
solution to the system 𝑥 ′ = 𝐴𝑥.

Proof:
𝑥 𝑡 = 𝑒 λ𝑡 𝑣 ⇒ 𝑥 ′ 𝑡 = λ𝑒 λ𝑡 𝑣
𝐴𝑥 = 𝐴 𝑒 λ𝑡 𝑣 = 𝑒 λ𝑡 𝐴𝑣 = 𝑒 λ𝑡 λ𝑣 = λ𝑒 λ𝑡 𝑣 = 𝑥′

COROLLARY:
If 𝐴 has 𝑛 linearly independent eigenvectors 𝑣1 , … , 𝑣𝑛 , then the vector functions
𝑥1 = 𝑒 λ1𝑡 𝑣1 , … , 𝑥𝑛 = 𝑒 λ𝑛 𝑡 𝑣𝑛 are 𝑛 linearly independent solutions and the general solution to
the system is 𝑥 = 𝐶1 𝑥1 + ⋯ + 𝐶𝑛 𝑥𝑛 where 𝐶1 , … , 𝐶𝑛 are constants of integration.
 Matrix Differential Equation (cont.)
EXAMPLE:
Use the method of eigenvalues and eigenvectors to solve the system of linear differential equations
𝑥′1 = 6𝑥1 + 5𝑥2
𝑥′2 = 𝑥1 + 2𝑥2 ⁡⁡

𝑥′1 6 5 𝑥1
𝑥′ = 𝐴𝑥 ⇔ =
𝑥′2 1 2 𝑥2
Eigenvalues of 𝐴:
6−λ 5
det 𝐴 − λ𝐼2 = 0 ⇔ = 0 ⇒ 6 − λ 2 − λ − 5 = 0 ⇒ λ2 − 8λ + 7 = 0
1 2−λ
′ ′2
∆ = 𝑏 − 𝑎𝑐 = 16 − 1 7 = 9
−𝑏 ′ − ∆′
λ1 = = 1⁡⁡⁡
⇒ 𝑎
−𝑏 ′ + ∆′
λ2 = = 7⁡⁡⁡
𝑎
Finally, λ1 = 1, and λ2 = 7 are the eigenvalues of 𝐴.
 Matrix Differential Equation (cont.)
Eigenvector corresponding to eigenvalue λ1 = 1:
𝐴 − λ1 𝐼2 𝑣 = 0

5 5 𝑥 0
⇒ 𝑦 =
1 1 0
𝑥 = −𝑦
⇒ 𝑦 = 𝑦⁡⁡⁡

𝑥 −𝑦 −1
⇒𝑣= 𝑦 = 𝑦 =𝑦
1

−1
Eigenvector 𝑣1 =
1
 Matrix Differential Equation (cont.)
Eigenvector corresponding to eigenvalue λ2 = 7:
𝐴 − λ2 𝐼2 𝑣 = 0

−1 5 𝑥 0
⇒ 𝑦 =
1 −5 0

𝑥 = 5𝑦
⇒
𝑦 = 𝑦⁡⁡⁡
𝑥 5𝑦 5
⇒𝑣= 𝑦 = =𝑦
𝑦 1
5
Eigenvector 𝑣2 =
1
𝑥1 −1 𝑡 5 7𝑡
The solution to the original system is 𝑥 = 𝑥 = 𝐶1 𝑒 + 𝐶2 𝑒
2 1 1
 Matrix Differential Equation (cont.)
REMARK:
If 𝑛 × 𝑛 matrix 𝐴 has 𝑛 linearly independent eigenvectors 𝑣1 , … , 𝑣𝑛 , then 𝐴 is diagonalizable. In
other words, there exists an invertible matrix 𝑃 such that 𝐵 = 𝑃−1 𝐴𝑃 or equivalently 𝐴 =
𝑃𝐵𝑃−1 where 𝐵 is a diagonal matrix.
Since 𝑥 ′ = 𝐴𝑥, then 𝑥 ′ = 𝑃𝐵𝑃−1 𝑥
⇒ 𝑃−1 𝑥 ′ = 𝑃−1 𝑃𝐵𝑃−1 𝑥
⇒ 𝑃−1 𝑥 ′ = 𝐼𝑛 𝐵𝑃−1 𝑥 = 𝐵𝑃−1 𝑥
Change of variable: Let 𝑦 = 𝑃−1 𝑥, then 𝑦 ′ = 𝑃−1 𝑥′
Consequently, 𝑦 ′ = 𝐵𝑦. This equation can be used to find 𝑦.
Finally, the general solution of the system 𝑥 ′ = 𝐴𝑥 is 𝑥 = 𝑃𝑦.
(𝑦 = 𝑃−1 𝑥 ⇒ 𝑃𝑦 = 𝑃𝑃−1 𝑥 ⇒ 𝑃𝑦 = 𝐼𝑛 𝑥 ⇒ 𝑃𝑦 = 𝑥)
 Matrix Differential Equation (cont.)
EXAMPLE:
𝑥′1 = 3𝑥1 − 5𝑥2
Use the diagonalization technique to solve the system of linear differential equations
𝑥′2 = 𝑥1 − 𝑥2 ⁡⁡⁡⁡⁡

′ 𝑥′1 3 −5 𝑥1
𝑥 = 𝐴𝑥 ⇔ =
𝑥′2 1 −1 𝑥2
Eigenvalues of 𝐴:
3−λ −5
det 𝐴 − λ𝐼2 = 0 ⇔ = 0 ⇒ 3 − λ −1 − λ + 5 = 0 ⇒ λ2 − 2λ + 2 = 0
1 −1 − λ
∆′ = 𝑏 ′2 − 𝑎𝑐 = 1 − 1 2 = −1 = 𝑖 2
−𝑏 ′ − ∆′
λ1 = = 1 − 𝑖⁡⁡⁡
⇒ 𝑎
−𝑏 ′ + ∆′
λ2 = = 1 + 𝑖⁡⁡⁡
𝑎

Finally, λ1 = 1 − 𝑖, and λ2 = 1 + 𝑖 are the eigenvalues of 𝐴.

 Matrix Differential Equation (cont.)
Eigenvector corresponding to eigenvalue λ1 = 1 − 𝑖:
𝐴 − λ1 𝐼2 𝑣 = 0

2+𝑖 −5 𝑥 0
⇒ 𝑦 =
1 −2 + 𝑖 0

𝑥 = (2 − 𝑖)𝑦
⇒
𝑦 = 𝑦⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡

𝑥 (2 − 𝑖)𝑦 2−𝑖
⇒𝑣= 𝑦 = =𝑦
𝑦 1

2−𝑖
Eigenvector 𝑣1 =
1
 Matrix Differential Equation (cont.)
2+𝑖
The eigenvector corresponding to eigenvalue λ2 = 1 + 𝑖 is 𝑣2 = because λ1 and λ2 are
1
complex conjugates.

Diagonalization of 𝐴:
1 −2 + 𝑖 𝑖 1
− +𝑖
2+𝑖 2−𝑖 1+𝑖 0
𝑃= ,𝐵 = , and⁡𝑃−1 = −1 2 + 𝑖 = 2 2
1 1 0 1−𝑖 2+𝑖 2−𝑖 𝑖 1
1 1 −𝑖
2 2
Now solve 𝑦 ′ = 𝐵𝑦:
𝑦′1 1+𝑖 0 𝑦1
⇒ = 𝑦2
𝑦′2 0 1−𝑖

𝑦1 (𝑡) = 𝐶1 𝑒 1+𝑖 𝑡
𝑦′1 = 1 + 𝑖 𝑦1
⇒ ⇒ 1−𝑖 𝑡
𝑦′2 = 1 − 𝑖 𝑦2 𝑦2 (𝑡) = 𝐶2 𝑒
 Matrix Differential Equation (cont.)
Consequently:

𝑦1 𝐶1 𝑒 1+𝑖 𝑡
𝑦= 𝑦 = 1−𝑖 𝑡
2 𝐶2 𝑒

But 𝑥 = 𝑃𝑦, then:

1+𝑖 𝑡 𝐶1 2 + 𝑖 𝑒 1+𝑖 𝑡 + 𝐶2 (2 − 𝑖)𝑒 1−𝑖 𝑡
2+𝑖 2−𝑖 𝐶1 𝑒
𝑥= 1−𝑖 𝑡
= 1+𝑖 𝑡
1 1 𝐶2 𝑒 𝐶1 𝑒 + 𝐶2 𝑒 1−𝑖 𝑡

Remark: We can use Euler identity 𝑒 𝑖𝑡 = cos 𝑡 + 𝑖 sin 𝑡 to write in terms of cos 𝑡 and sin 𝑡.