Chapter 4 Diagonalization

4.1 Eigenvalues and Eigenvectors

Definition 4.1:
Let A ∈ Mn(R). A scalar λ is called an eigenvalue of A if there exists a nonzero column vector
X ∈ Mn,1(R) such that AX = λ X . X is called an eigenvector of A corresponding to (or associated
with) the eigenvalue λ.

Example 4.1:
⎡1 2 ⎤ ⎡− 1⎤ ⎡1 ⎤
Let A = ⎢ ⎥ , X 1 = ⎢ ⎥ and X 2 = ⎢ ⎥ .
⎣ 4 3⎦ ⎣1⎦ ⎣ 2⎦
⎡1 2⎤ ⎡− 1⎤ ⎡ 1 ⎤ ⎡− 1⎤
AX1 = ⎢ ⎥ ⎢ ⎥ = ⎢ ⎥ = −1⎢ ⎥ = − X 1 .
⎣4 3⎦ ⎣ 1 ⎦ ⎣− 1⎦ ⎣1⎦
∴ X1 is an eigenvector of A corresponds to the eigenvalue λ1 = -1.
⎡1 2 ⎤ ⎡1 ⎤ ⎡ 5 ⎤ ⎡1 ⎤
AX2 = ⎢ ⎥ ⎢ ⎥ = ⎢ ⎥ = 5⎢ ⎥ = 5 X 2 .
⎣4 3⎦ ⎣2⎦ ⎣10⎦ ⎣ 2⎦
∴ X2 is an eigenvector of A corresponds to the eigenvalue λ2 = 5.

Example 4.2: Let A ∈ Mn(R). Show that

(i) If λ ∈ R is an eigenvalue of A, then λn is an eigenvalue of An for every positive integer n.

(ii) If X ∈ Mn,1(R) is an eigenvector of A corresponding to the eigenvalue λ ∈ R, then, for every

nonzero scalar α ∈ R, the vector αX is an eigenvector of A corresponds to the eigenvalue λ.

Solution:
(i) λ is an eigenvalue of A if there exists a nonzero column vector such that AX = λX .
An X = An−1 (AX ) = An−1 (λX) = λ An−2 (AX) = λ An−2 (λX) = λ2 An−2 X = λ2 An−3 (AX) = λ2 An−3 (λX)
= λ3 An−3 X = ⋅⋅⋅ = λn-1 (AX ) = λn-1(λX) = λnX.
Hence, λn is an eigenvalue of An for every positive integer n.

(ii) For every nonzero scalar α ∈ R, AX = λX ⇒ α(AX ) = α(λX) ⇒ A(αX ) = λ(αX).

Hence, the vector αX is an eigenvector of A corresponds to the eigenvalue λ.

Theorem 4.2:
Let A ∈ Mn(R). A scalar λ is an eigenvalue of A if and only if det(A − λIn) = 0.

Proof:
λ is an eigenvalue of A ⇒ there exists a nonzero column vector X ∈ Mn,1 such that AX = λX
⇒ there exists a nonzero column vector X ∈ Mn,1 such that (A − λIn)X = 0
⇒ the homogenous system has a nontrivial solution (∵ X ≠ 0)
⇒ the matrix (A − λIn) is singular (not invertible) ⇒ det(A − λIn) = 0.
Definition 4.3:
Let A ∈ Mn(R). Then, p(λ) = det(A − λIn) is called a characteristic polynomial of A and
p(λ) = det(A − λIn) = 0 is the characteristic equation of A.
1
Procedure to determine eigenvalues and eigenvector of A
1. For a given A ∈ M n , determine the characteristic polynomial
a11 − λ a12 ... a1n
a21 a22 − λ ... a2n
P( λ ) = det(A − λ I n ) = .

an1 an2 ... ann − λ

2. Solve the characteristic equation p(λ) = 0 and determine all the eigenvalues λi.

3. For each λi, solve for ei in equation Ae = λiei , or equivalently, (A − λiIn)ei = 0, and find an
eigenvector ei corresponding to λi.

Example 4.3:
⎡ 1 1 − 2⎤
Determine the characteristic equation of A = ⎢− 1 2 1 ⎥ .
⎢ ⎥
⎣⎢ 0 1 − 1⎥⎦
Then, for each λi, find an eigenvector ei corresponding to λi.

Solution:
⎡ 1 1 −2 ⎤ ⎡ 1 0 0 ⎤ ⎡1 − λ 1 −2 ⎤
⎢ ⎥ ⎢ ⎥
A − λ I 3 = ⎢ −1 2 1 ⎥ − λ ⎢ 0 1 0 ⎥ = ⎢ − 1 2−λ 1 ⎥⎥
⎢
⎢ 0 1 −1 ⎥ ⎢ 0 0 1 ⎥ ⎢⎣ 0 1 − 1 − λ ⎥⎦
⎣ ⎦ ⎣ ⎦

So, the characteristic polynomial is

p(λ) = det(A − λI3)
1− λ 1 −2
= −1 2−λ 1
0 1 −1− λ
2−λ 1 1 −2
= (1 − λ ) − (−1) (Expanding along column 1)
1 −1− λ 1 −1− λ
= (1 − λ )[(2 − λ )(−1 − λ ) − 1] + [(−1 − λ ) + 2]
= λ3 − 2λ2 − λ + 2 .
Let p(λ) = det(A − λIn) = 0.
Then λ3 − 2λ2 − λ + 2 = (λ − 2)(λ − 1)(λ + 1) = 0 .
∴ λ1 = 2 , λ2 = 1 , λ3 = −1.

Next, find eigenvectors ei corresponding to λi by solving the system (A − λiI3)ei = 0.

⎡1 − λ 1 − 2 ⎤⎛ x1 ⎞ ⎛ 0 ⎞
⎜ ⎟ ⎜ ⎟
That is, Solve ⎢ − 1 2 − λ 1 ⎥⎥⎜ x 2 ⎟ = ⎜ 0 ⎟ .
⎢
⎢⎣ 0 1 − 1 − λ ⎥⎦⎜⎝ x3 ⎟⎠ ⎜⎝ 0 ⎟⎠

2
Case (i) λ1 = 2.
Augmented matrix:
⎡1 − 2 1 − 2 0⎤ ⎡ − 1 1 − 2 0⎤ ⎡ − 1 0 1 0⎤
⎢ −1 2 − 2 R2 → R2 − R1 R3 →R3 + R2
⎢ 1 ⎥
0⎥ ⎯⎯ ⎯⎯→ ⎢⎢ 0 − 1 3 0⎥⎥ ⎯⎯ ⎯⎯→ ⎢ 0 − 1 3 0⎥
⎢ ⎥
⎢⎣ 0 1 − 1 − 2 0⎥⎦ ⎢⎣ 0 1 − 3 0⎥⎦ ⎢⎣ 0 0 0 0⎥⎦
∴ 0=0
−x2 + 3x3 = 0 ⇒ x2 = 3x3 and x3 = x1
−x1 + x3 = 0
Let x1 = k. Then x2 = 3k, x3 = k , k ≠ 0.
⎧⎛ ⎫
k ⎞⎟
∴ The set of eigenvectors corresponding to λ1 = 2 is given by ⎨⎜ 3k ⎟ ,k ≠ 0⎪⎬ .
⎪⎜
⎪⎜ k ⎟ ⎪
⎩⎝ ⎠ ⎭

Case (ii) λ2 = 1.
Augmented matrix:
⎡1 − 1 1 − 2 0⎤ ⎡ 0 1 − 2 0⎤ ⎡ 0 1 − 2 0⎤
⎢ −1 2 −1 R3 → R3 − R1 R2 → R2 + R1
⎢ 1 ⎥
0⎥ ⎯⎯ ⎯⎯→ ⎢⎢− 1 − 1 1 0⎥⎥ ⎯⎯ ⎯⎯→ ⎢⎢− 1 0 3 0⎥⎥
⎢⎣ 0 1 − 1 − 1 0⎥⎦ ⎢⎣ 0 0 0 0⎥⎦ ⎢⎣ 0 0 0 0⎥⎦
∴ 0=0
−x1 + 3x3 = 0 ⇒ x1 = 3x3 and x2 = 2x3
x2 − 2x3 = 0
Let x3 = k. Then x1 = 3k, x2 = 2k, k ≠ 0.
⎧⎛ 3k ⎞ ⎫
⎜ ⎟
∴ The set of eigenvectors corresponding to λ2 = 1 is given by ⎪⎨⎜ 2k ⎟, k ≠ 0⎪⎬ .
⎪⎜ k ⎟ ⎪
⎩⎝ ⎠ ⎭

Case (iii) λ3 = −1.

Augmented matrix:
⎡1 + 1 1 − 2 0⎤ ⎡2 1 − 2 0⎤ ⎡ 2 0 − 2 0⎤
⎢ −1 2 +1 R2 →R2 + 12 R1 R2 →R2 − 72 R3
⎢ 1 0⎥ ⎯⎯ ⎯ ⎯→ ⎢⎢0 72 0
⎥ 0⎥⎥ ⎯→ ⎢⎢0 0 0 0⎥⎥
⎯⎯⎯ ⎯
⎢⎣ 0 1 − 1 + 1 0⎥⎦ ⎢⎣0 1 0 0⎥⎦ ⎢⎣0 1 0 0⎥⎦
∴ x2 = 0
0 = 0 ⇒ x2 = 0 and x1 = x3
2x1 − 2x3 = 0
Let x1 = k. Then x3 = k, k ≠ 0.
⎧⎛ k ⎞ ⎫
⎜ ⎟
∴ The set of eigenvectors corresponding to λ2 = −1 is given by ⎨⎜ 0 ⎟, k ≠ 0⎪⎬ .
⎪
⎪⎜ k ⎟ ⎪
⎩⎝ ⎠ ⎭

3
Example 4.4:
⎡cosθ − sin θ ⎤
Determine all the eigenvalues and eigenvectors of A = ⎢ .
⎣ sin θ cosθ ⎥⎦

Solution:
The characteristic polynomial is
p(λ) = det(A − λI2)
cosθ − λ − sin θ
=
sin θ cosθ − λ
= (cos θ − λ ) 2 + sin 2 θ
= λ2 − 2λ cosθ + 1 .
Let p(λ) = det(A − λIn) = 0.
So, λ2 − 2λ cosθ + 1 = 0
2 cosθ ± 4 cos 2 θ − 4
∴ λ= = cosθ ± i sin θ .
2
Next, find eigenvectors ei corresponding to λi by solving the system (A − λiI2)ei = 0.
⎡cosθ − λ − sin θ ⎤⎛ x1 ⎞ ⎛ 0 ⎞
That is, solve ⎢ ⎥⎜⎜ x ⎟⎟ = ⎜⎜ 0 ⎟⎟ .
⎣ sin θ cos θ − λ ⎦⎝ 2 ⎠ ⎝ ⎠

Case (i) λ1 = cosθ + isinθ.

Augmented matrix:
⎡− i sin θ − sin θ 0⎤ R2 →R2 −iR1 ⎡− i sin θ − sin θ 0⎤
⎢ sin θ ⎥ ⎯⎯ ⎯⎯→ ⎢ 0
⎣ − i sin θ 0⎦ ⎣ 0 0⎥⎦
∴ 0=0
(−isinθ)x1 + (−sinθ)x2 = 0 ⇒ x2 = −ix1.
Let x1 = k. Then x2 = −ik, k ≠ 0.
∴ The set of eigenvectors corresponding to λ1 = cosθ + isinθ is given by ⎧⎨⎛⎜ k ⎞⎟, k ≠ 0⎫⎬ .
⎜ ⎟
⎩⎝ − ik ⎠ ⎭

Case (ii) λ2 = cosθ − isinθ.

Augmented matrix:
⎡i sin θ − sin θ 0⎤ R2 →R2 +iR1 ⎡i sin θ − sin θ 0⎤
⎢ sin θ i sin θ 0⎥ ⎯⎯ ⎯⎯→ ⎢ 0 0 0⎥⎦
⎣ ⎦ ⎣
∴ 0=0
(isinθ)x1 −(sinθ)x2 = 0 ⇒ x2 = ix1.
Let x1 = k. Then x2 = ik, k ≠ 0.
∴ The set of eigenvectors corresponding to λ2 = cosθ + isinθ is given by ⎧⎨⎛⎜ k ⎞⎟, k ≠ 0⎫⎬ .
⎜ ⎟
⎩⎝ ik ⎠ ⎭

4
4.2 Diagonalization

Recall that a matrix D = (di,j) ∈ Mn is called a diagonal matrix if di,j = 0 for all i ≠ j. That is,
⎡d 11 0 ! 0 ⎤
⎢0 d ! 0 ⎥⎥
⎢ 22
D= .
⎢ " " # " ⎥
⎢ ⎥
⎣ 0 ! 0 d nn ⎦

Properties of diagonal matrix:

Let D and W are two diagonal matrices of order n.
⎡d 11 0 ! 0 ⎤ ⎡ w11 0 ! 0 ⎤
⎢0 d ! 0 ⎥ ⎥ ⎢ 0 w ! 0 ⎥⎥
That is, D = ⎢ 22
and W = ⎢ 22
.
⎢ " " # " ⎥ ⎢ " " # " ⎥
⎢ ⎥ ⎢ ⎥
⎣ 0 ! 0 d nn ⎦ ⎣ 0 ! 0 wnn ⎦
⎡ w11 d 11 0 ! 0 ⎤
⎢ 0 w22 d 22 ! 0 ⎥⎥
1. DW and WD are also diagonal matrix and DW = WD = ⎢ .
⎢ " " # " ⎥
⎢ ⎥
⎣ 0 ! 0 wnn d nn ⎦

2. The determinant of D = det(D) = d11d22 d33… dnn .

3. If all of the diagonal elements of D are nonzero (that is dii ≠ 0 ∀i = 1, 2, …, n), then D is
⎡ 1 ⎤
⎢d 0 ! 0 ⎥
⎢ 11 ⎥
⎢ 0 1
invertible (or nonsingular) and D −1 ! 0 ⎥.
=⎢ d 22 ⎥
⎢ " " # " ⎥
⎢ 1 ⎥
⎢ 0 ! 0 ⎥
⎣⎢ d nn ⎦⎥

4. The eigenvalues of D are its diagonal elements, that is d11, …, dnn .

Of course, most square matrices are not diagonal, but some are related to diagonal matrices in a
way we will find useful in solving problems.

Definition 4.5:
Let A, B ∈ Mn(R). A matrix A is similar to B, denoted A ~ B, if there exists an invertible matrix P
∈ Mn(R) such that A = P −1BP.

Definition 4.6:
Let A ∈ Mn(R). Then, A is diagonalizable ⇔ there exists an n-square matrix P such that P-1AP is
a diagonal matrix. (That is, P-1AP = D where D is a diagonal matrix.)
When such a matrix P exists, we say that P diagonalizes A.

5
Example 4.5:
⎡3 2 ⎤ ⎡2 1 ⎤
Let A = ⎢ ⎥ . There exists a matrix P = ⎢1 − 3⎥ such that
⎣3 − 2 ⎦ ⎣ ⎦
1⎡ 3 1 ⎤⎡3 2 ⎤⎡ 2 1 ⎤ 1⎡ 3 1 ⎤ ⎡ − 3⎤ ⎡4 0 ⎤
8
P −1 AP = ⎢ = = = diagonal matrix.
7 ⎣1 − 2⎥⎦ ⎢⎣3 − 2⎥⎦ ⎢⎣1 − 3⎥⎦ 7 ⎢⎣1 − 2⎥⎦ ⎢⎣4 9 ⎥⎦ ⎢⎣0 − 3⎥⎦
Therefore, P diagonalizes A.

Theorem 4.7:
Let A ∈ Mn(R). Then, A is diagonalizable ⇔ A has n linearly independent eigenvectors.

Theorem 4.8:
n column vector v1 , v 2 ,..., v n are linearly independent
⇔ The matrix A = ( v1 , v 2 ,..., v n ) is invertible (nonsingular) ⇔ det(A) ≠ 0.

Procedure to determine a matrix P that diagonalizes the matrix A:

1. Determine the characteristic equation of A, that is p(λ) = det(A − λIn) = 0.

2. Solve the characteristic equation p(λ) = 0 and determine all the eigenvalues λi.

3. For each λi, solve (A − λiIn)ei = 0 to find an eigenvector ei corresponding to λi.

4. Form the matrix P = [e1 | e2 | … | en].

5. Determine whether |P| is nonzero.

⎡λ1 0 ! 0⎤
⎢0 λ 2 ! 0 ⎥⎥
6. If |P| ≠ 0, then P-1AP = ⎢ .
⎢" " # "⎥
⎢ ⎥
⎣0 ! 0 λn ⎦

Example 4.6:
⎡− 1 4⎤
Let A = ⎢ . Find a matrix P that diagonalizes A and find the diagonal matrix D such that D =
⎣0 3⎥⎦
-1
P AP.

Solution:
The characteristic equation is p(λ) = det(A − λI2) = 0.
−1− λ 4
det(A − λI2) = =0
0 3−λ
⇒ λ2 − 2λ − 3 = 0
⇒ (λ − 3)(λ + 1) = 0 ⇒ λ = −1, 3 .

Next, find eigenvectors ei corresponding to λi by solving the system (A − λiI2)ei = 0.

⎡− 1 − λ 4 ⎤⎛ x1 ⎞ ⎛ 0 ⎞
That is, solve ⎢ ⎜ ⎟ = ⎜ ⎟.
⎣ 0 3 − λ ⎥⎦⎜⎝ x2 ⎟⎠ ⎜⎝ 0 ⎟⎠
6
Case (i) λ1 = −1.
Augmented matrix:
⎡0 4 0 ⎤ R2 → R2 − R1 ⎡0 4 0 ⎤
⎢0 4 0 ⎥ ⎯ ⎯ ⎯ ⎯ → ⎢0 0 0 ⎥
⎣ ⎦ ⎣ ⎦
∴ 4x2 = 0 ⇒ x2 = 0.
∴ The set of eigenvectors corresponding to λ1 = −1 is given by ⎧⎨⎛⎜ k ⎞⎟, k ≠ 0⎫⎬ .
⎜ ⎟
⎩⎝ 0 ⎠ ⎭
⎛1⎞
∴ When k = 1, e1 = ⎜⎜ ⎟⎟ is a eigenvector corresponding to λ1 = −1.
⎝ 0⎠

Case (ii) λ2 =3.

⎡ − 4 4 0⎤
Augmented matrix is ⎢ ⎥.
⎣ 0 0 0 ⎦
∴ −4x1 + 4x2 = 0 ⇒ x1 = x2.
∴ The set of eigenvectors corresponding to λ2 = 3 is given by ⎧⎨⎛⎜ k ⎞⎟, k ≠ 0⎫⎬ .
⎜ ⎟
⎩⎝ k ⎠ ⎭
⎛1⎞
∴ When k = 1, e2 = ⎜⎜ ⎟⎟ is a eigenvector corresponding to λ2 = 3.
⎝1⎠

⎡1 1⎤ 1 1 ⎡1 − 1⎤
Let P = ⎢ ⎥ . Since | P |= = 1 ≠ 0 , then P −1 exists and P −1 = ⎢ ⎥.
⎣0 1⎦ 0 1 ⎣0 1 ⎦
⎡1 − 1⎤ ⎡− 1 4⎤ ⎡1 1⎤ ⎡1 − 1⎤ ⎡− 1 3⎤ ⎡− 1 0⎤
Since P −1 AP = ⎢ ⎥⎢ ⎥⎢ ⎥=⎢ ⎥⎢ ⎥=⎢ ⎥ = D.
⎣0 1 ⎦ ⎣ 0 3⎦ ⎣0 1⎦ ⎣0 1 ⎦ ⎣ 0 3⎦ ⎣ 0 3⎦
⎡1 1⎤ ⎡ − 1 0⎤
Therefore P = ⎢ ⎥ diagonalizes the matrix A and D = P −1 AP = ⎢ ⎥.
⎣0 1⎦ ⎣ 0 3⎦

Example 4.7:
⎡− 1 0 5 ⎤
Let A = ⎢ 0 1 0 ⎥ . Find a matrix P that diagonalizes A and find the diagonal matrix D such that
⎢ ⎥
⎢⎣ 0 0 − 2⎥⎦
D = P-1AP.

Solution:
The characteristic polynomial of A is p(λ) = det(A − λI3), that is
−1− λ 0 5
p (λ ) = 0 1− λ 0 = (−1 − λ )(1 − λ )(−2 − λ ) ,
0 0 −2−λ
and p(λ) = 0 ⇒ λ = − 1, 1, − 2.
Next, find eigenvectors ei corresponding to λi by solving the system (A − λiI3)ei = 0.

7
 ⎡− 1 − λ 0 5 ⎤⎛ x1 ⎞ ⎛ 0 ⎞
That is, Solve ⎢ 0 1 − λ 0 ⎥⎜ x ⎟ = ⎜ 0 ⎟ .
⎢ ⎥⎜ 2 ⎟ ⎜ ⎟
⎢⎣ 0 0 − 2 − λ ⎥⎦⎜⎝ x3 ⎟⎠ ⎜⎝ 0 ⎟⎠
Case (i) λ1 = −1.
Augmented matrix:
⎡− 1 + 1 0 5 0⎤ ⎡0 2 0 0 ⎤ ⎡0 2 0 0 ⎤
⎢ 0 R1 ↔ R2 R3 →R3 + R2 / 5
⎢ 1+1 0 0⎥ ⎯⎯⎯→ ⎢⎢0 0 5 0⎥⎥ ⎯⎯
⎥ ⎯⎯ ⎯→ ⎢0 0 5 0 ⎥
⎢ ⎥
⎢⎣ 0 0 − 2 + 1 0⎥⎦ ⎢⎣0 0 − 1 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
∴ x2 = 0 and x3 = 0.
Let x1 = k be any nonzero number.
⎧⎛ k ⎞ ⎫
⎜ ⎟
∴ The set of eigenvectors corresponding to λ1 = −1 is given by ⎨⎜ 0 ⎟, k ≠ 0⎪⎬ .
⎪
⎪⎜ 0 ⎟ ⎪
⎩⎝ ⎠ ⎭
⎛1⎞
⎜ ⎟
When k = 1, e1 = ⎜ 0 ⎟ is an eigenvector corresponding to λ1 = −1.
⎜ 0⎟
⎝ ⎠

Case (ii) λ2 = 1.
Augmented matrix:
⎡− 1 − 1 0 5 0⎤ ⎡− 2 0 5 0⎤
⎢ 0 R2 ↔ R3
⎢ 1−1 0 0⎥ ⎯⎯⎯→ ⎢⎢ 0 0 − 3
⎥ 0⎥⎥ .
⎢⎣ 0 0 − 2 − 1 0⎥⎦ ⎢⎣ 0 0 0 0⎥⎦
∴ −2x1 + 5x3 = 0 and −3x3 = 0 ⇒ x3 = 0 and x1 = 0.
Let x2 = k be any nonzero number.
⎧⎛ 0 ⎞ ⎫
∴ ⎜ ⎟
The set of eigenvectors corresponding to λ2 = 1 is given by ⎨⎜ k ⎟, k ≠ 0⎪⎬ .
⎪
⎪⎜ 0 ⎟ ⎪
⎩⎝ ⎠ ⎭
⎛ 0⎞
⎜ ⎟
When k = 1, e2 = ⎜ 1 ⎟ is an eigenvector corresponding to λ2 = 1.
⎜ 0⎟
⎝ ⎠

Case (iii) λ2 = −2.

Augmented matrix:
⎡− 1 + 2 0 5 0⎤ ⎡1 0 5 0⎤
⎢ 0 1+ 2 0 0⎥⎥ = ⎢⎢0 3 0 0⎥⎥ .
⎢
⎢⎣ 0 0 − 2 + 2 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
∴ x1 + 5x3 = 0 and 3x2 = 0 ⇒ x1 = −5x3 and x2 = 0.
Let x3 = k be any nonzero number.
⎧⎛ − 5k ⎞ ⎫
∴ The set of eigenvectors corresponding to λ3= −2 is given by ⎪⎨⎜⎜ 0 ⎟⎟, k ≠ 0⎪⎬ .
⎪⎜ k ⎟ ⎪
⎩⎝ ⎠ ⎭

8
 ⎛ − 5⎞
⎜ ⎟
When k = 1, e3 = ⎜ 0 ⎟ is an eigenvector corresponding to λ3 = −2.
⎜ 1 ⎟
⎝ ⎠

⎡1 0 − 5⎤
Let P = ⎢0 1 0 ⎥ . Since | P |= 1 ≠ 0 , then P −1 exists. .
⎢ ⎥
⎢⎣0 0 1 ⎥⎦
⎡− 1 0 0 ⎤
Therefore P diagonalizes the matrix A and D = P AP = ⎢ 0 1 0 ⎥ .
−1
⎢ ⎥
⎢⎣ 0 0 − 2⎥⎦

Remark 4.1:
If a matrix A has n distinct eigenvalues, then A is diagonalizable.
However, there exist many matrices which are diagonalizable even though they have some repeated
eigenvalues. In these matrices, all we need are n linearly independent eigenvectors of a matrix (that
is, determine a suitable P where |P| ≠ 0).

Example 4.8:
⎡5 −4 4⎤
Let A = ⎢12 − 11 12 ⎥ . Find a matrix P that diagonalizes A and find the diagonal matrix D such
⎢ ⎥
⎢⎣ 4 − 4 5 ⎥⎦
that D = P-1AP.
Solution:
The characteristic polynomial of A is
5−λ −4 4 1− λ 0 −1+ λ 0 0 −1+ λ
p(λ) = det(A − λI3) = 12 − 11 − λ 12 = 12 − 11 − λ 12 = 24 − 11 − λ 12
4 −4 5−λ 4 −4 5−λ 9−λ −4 5−λ
= ! = (−1 + λ )(1 − λ )(3 + λ ) .
p(λ) = 0 ⇒ λ = 1, 1, − 3 .
Next, find eigenvectors ei corresponding to λi by solving the system (A − λiI3)ei = 0.

Case (i) λ1 = λ2 = 1.
Augmented matrix:
⎡5 − 1 −4 4 0⎤ ⎡ 4 − 4 4 0⎤
⎢ 12 − 11 − 1 12 0⎥ ⎯⎯ R2 →R2 −3R1
⎢ ⎥
⎯⎯→ ⎢⎢0 0 0 0⎥⎥ .
⎢⎣ 4 −4 5 − 1 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
∴ x1 − x2 + x3 = 0 ⇒ x1 = x2 − x3.
Let x2 = s, x3 = t be any numbers. Then x1 = s − t.
⎧⎛ s − t ⎞ ⎫
⎜ ⎟
∴ The set of eigenvectors corresponding to λ1 = 1 is given by ⎨⎜ s ⎟, ( s, t ) ≠ (0,0)⎪⎬ .
⎪
⎪⎜ t ⎟ ⎪
⎩⎝ ⎠ ⎭

9
 ⎛1⎞
⎜ ⎟
When s = 1 and t = 0 and, e1 = ⎜ 1 ⎟ is an eigenvector corresponding to λ1 = 1.
⎜ 0⎟
⎝ ⎠
⎛ − 1⎞
When s = 0 and t = 1 and, e2 = ⎜⎜ 0 ⎟⎟ is also an eigenvector corresponding to λ1 = 1.
⎜1⎟
⎝ ⎠
(Note that e1 and e2 are linearly independent eigenvectors.)

Case (ii) λ3 = −3.

Augmented matrix:
⎡5 + 3 −4 4 0⎤ ⎡8 − 4 4 0⎤ ⎡ 2 − 1 1 0⎤
⎢ 12 − 11 + 3 12 0⎥ ⎯⎯ R2 →R2 −3 R1 / 2 ⎢0 − 2 6 0⎥ ⎯⎯ R3 →R3 − R2 ⎢0 − 2 6 0⎥
⎢ ⎥
⎯ ⎯ ⎯ → ⎢ ⎥
⎯⎯→ ⎢ ⎥
⎢⎣ 4 −4 5 + 3 0⎥⎦ ⎢⎣0 − 2 6 0⎥⎦ ⎢⎣0 0 0 0⎥⎦
∴ 2x1 − x2 + x3 = 0
−x2 + 3x3 = 0 ⇒ x2 = 3x3 and x1 = (x2 − x3)/ 2 = x3.
Let x3 = k be any nonzero number.
⎧⎛ k ⎞ ⎫
∴ The set of eigenvectors corresponding to λ3 = −3 is given by ⎪⎜ 3k ⎟, k ≠ 0⎪ .
⎨⎜ ⎟ ⎬
⎪⎜ k ⎟ ⎪
⎩⎝ ⎠ ⎭
⎛ 1⎞
⎜ ⎟
When k = 1, e3 = ⎜ 3 ⎟ is an eigenvector corresponding to λ3 = −3.
⎜ 1⎟
⎝ ⎠

⎡1 − 1 1⎤
Let P = ⎢1 0 3⎥ . Since | P |= −1 ≠ 0 , then P −1 exists.
⎢ ⎥
⎢⎣0 1 1⎥⎦
⎡1 0 0 ⎤
Therefore P diagonalizes the matrix A and D = P AP = ⎢0 1 0 ⎥ .
−1
⎢ ⎥
⎢⎣0 0 − 3⎥⎦

4.3 Cayley-Hamilton Theorem

Theorem 4.9: Cayley-Hamilton Theorem
Every matrix is a zero of its characteristic polynomial. That is, A ∈ Mn(R). Then,
p(λ) = det(A − λIn) = 0 ⇒ p(A) = 0.

⎡3 4⎤
Example 4.9: Verify Cayley-Hamilton Theorem for A = ⎢ ⎥.
⎣1 2⎦
Solution:
⎡3 − λ 4 ⎤
p(λ) = det(A − λIn) = ⎢ = 0 ⇒ (3−λ)(2−λ) − 4 = 0 ⇒ λ2 − 5λ + 2 = 0
⎣ 1 2 − λ ⎥⎦

⎡13 20⎤ ⎡3 4⎤ ⎡1 0⎤
A2 − 5A + 2I2 = ⎢ −5 +2 = 0 . Therefore, p(A) = 0.
⎣5 8 ⎥⎦ ⎢⎣1 2⎥⎦ ⎢⎣0 1⎥⎦
10
Application of Cayley-Hamilton Theorem
1. Cayley-Hamilton Theorem may use to determine Ar, r ≥ 2.
Example 4.10:
From Example 4.9, A2 = 5A − 2I2.
∴ A3 = A2(A) = (5A − 2I2)A = 5A2 − 2A = 5(5A − 2I2) − 2A = 23A − 10I2.
∴ A4 = A3(A) = (23A − 10I2)A = 23(5A − 2I2) − 10A = 115A − 56I2.

In general, Ar = a0In + a1A, r ≥ 2. ------------------------(*)

Each of the eigenvalues λ of A is also satisfies the equation (*).

That is, λr = a0 + a1λ, r ≥ 2.
Determination of a0 and a1:
Let A ∈ M2 . p(λ) = λ2 + c1λ + c2 = 0.
Case 1: λ1≠ λ2 ⇒ λ1r = a0 + a1λ1, (r ≥ 2) -----------------------(1)
and λ2r = a0 + a1λ2. -----------------------(2)
(Solve the equations (1) and (2) for a0 and a1).
Case 2: λ = λ1 = λ2 ⇒ λr = a0 + a1λ ------------------------(3)
r-1
and rλ = a1. (r ≥ 2) ------------------------(4)
(Solve the equations (3) and (4) for a0 and a1).
The technique in Case 2 is followed by the Theorem 3.5.

Theorem 4.10:
If λ = λ0 are two (or more) repeated roots of the polynomial equation f(λ) = 0, then λ0 is also a
root f ʹ(λ ) = 0 .

Proof:
Since λ = λ0 are two repeated roots of f(λ) = 0, then f(λ) = (λ − λ0)(λ − λ0) g(λ) = (λ − λ0)2 g(λ)
Differentiate f with respect to λ, we have f ʹ (λ) = 2(λ − λ0)g(λ) + (λ − λ0)2 gʹ (λ)
⇒ f ʹ (λ0) = 0 ⇒ λ0 is a root f ʹ(λ ) = 0 .

Example 4.11:
⎡0 1⎤
If A = ⎢ ⎥ . Determine A5 and Ar, r ≥ 2. Hence, evaluate A4 − 2A3 + A − I .
⎣− 2 − 3⎦
Solution:
0−λ 1
| A − λI 2 |= = 0 ⇒ (−λ)(−3 −λ) + 2 = 0
− 2 −3−λ
⇒ λ2 + 3λ + 2 = 0 ⇒ (λ + 2)(λ + 1) = 0 ⇒ λ = −1, −2 .
Let A5 = a0I2 + a1A.
For each λ = −1, −2, λ5 = a0 + a1λ.
When λ = −1, (−1)5 = a0 + a1(−1). ----------------- (1)
When λ = −2, (−2)5 = a0 + a1(−2). ----------------- (2)
Solve (1) and (2), we have a1 = 31 and a0 = 30.
⎡1 0⎤ ⎡0 1 ⎤ ⎡ 30 31 ⎤
∴ A5 = 30 ⎢ ⎥ + 31⎢ ⎥ =⎢ ⎥.
⎣0 1⎦ ⎣− 2 − 3⎦ ⎣− 62 − 63⎦
11
In general, λ1r = a0 + a1λ1 ⇒ (−1) r = a0 + a1 (−1) . ----------(3)
λr2 = a0 + a1λ2 ⇒ (−2) r = a0 + a1 (−2) . ----------(4)

(3) − (4), a1 = (−1) r − (−2) r = (−1) r (1 − 2 r ) . ----------(5)

From (3) and (5), a0 = (−1) (1 − 2 ) + (−1) = (−1) (2 − 2 r ) .
r r r r

r
⎡1 0⎤ r ⎡ 0 1⎤ r⎡ 2−2 1 − 2r ⎤
r
( r
∴ A = (−1) (2 − 2 ) ⎢ r
) r
( )
⎥ + (−1) (1 − 2 ) ⎢− 2 − 3⎥ = (−1) ⎢ r +1 r +1 ⎥.
⎣0 1⎦ ⎣ ⎦ ⎣2 − 2 2 − 1⎦

⎡ 2 − 24 1 − 2 4 ⎤ ⎡14 − 15⎤
When r = 4, A 4 = (−1) 4 ⎢ 4+1 4 +1 ⎥=⎢ ⎥.
⎣ 2 − 2 2 − 1⎦ ⎣30 31 ⎦

⎡ 2 − 23 1 − 23 ⎤ ⎡ 6 7 ⎤
When r = 3, A3 = (−1) 3 ⎢ 3+1 3+1 ⎥ = ⎢ ⎥.
⎣ 2 − 2 2 − 1⎦ ⎣ − 14 − 15 ⎦

⎡14 − 15⎤ ⎡ 6 7 ⎤ ⎡0 1 ⎤ ⎡1 0⎤ ⎡− 27 − 28⎤

∴ A4 − 2A3 + A − I = ⎢ ⎥ − 2⎢ ⎥ +⎢ ⎥−⎢ ⎥=⎢ .
⎣30 31 ⎦ ⎣− 14 − 15⎦ ⎣− 2 − 3⎦ ⎣0 1⎦ ⎣ 56 57 ⎥⎦

Alternative method:
Let A4 − 2A3 + A − I = β0I + β1A .
For each λ = −1, −2, λ4 − 2λ3 + λ − 1 = β0 + β1λ.

When λ = −1, (−1)4 − 2(−1)3 + (−1) − 1 = β0 + β1(−1). -----------------(6)

When λ = −2, (−2)4 − 2(−2)3 + (−2) − 1 = β0 + β1(−2). ---------------- (7)

Solve (6) and (7), we have β0 = −27 and β1 = −28.

⎡1 0⎤ ⎡0 1 ⎤ ⎡− 27 − 28⎤
∴ A4 − 2A3 + A − I = −27 ⎢ ⎥ − 28⎢ ⎥=⎢ .
⎣0 1⎦ ⎣− 2 − 3⎦ ⎣ 56 57 ⎥⎦

Example 4.12:
⎡0 1 ⎤
If A = ⎢ ⎥ . Determine A5 and Ar, r ≥ 2. Hence, evaluate A4 − 2A3 + A − I .
⎣ − 1 − 2⎦
Solution:

det(A − λ I 2 ) = 0 − λ 1 = 0 ⇒ (−λ)(−2 −λ) + 1 = 0 ⇒ λ2 + 2λ + 1 = 0

−1 −2 − λ
⇒ (λ + 1)2 = 0 ⇒ λ = −1 (repeated) .
Let A5 = a0I2 + a1A.
For λ = −1, λ5 = a0 + a1λ and
the derivative is 5λ4 = a1.
When λ = −1, (−1)5 = a0 + a1(−1). ----------------- (1)
and 5(−1)4 = a1. ----------------- (2)
Solve (1) and (2), we have a1 = 5 and a0 = 4.
⎡1 0⎤ ⎡ 0 1⎤ ⎡4 5⎤
∴ A5 = 4⎢ ⎥ + 5⎢ ⎥ =⎢ ⎥.
⎣0 1 ⎦ ⎣ − 1 − 2 ⎦ ⎣ − 5 − 6 ⎦
12
Let A4 − 2A3 + A − I = β0I + β1A .

For each λ = −1, λ4 − 2λ3 + λ − 1 = β0 + β1λ.

the derivative is 4λ3 − 6λ2 + 1 = β1.

When λ = −1, (−1)4 − 2(−1)3 + (−1) − 1 = β0 + β1(−1). -----------------(3)

and 4(−1)3 − 6(−1)2 + 1 = β1. ---------------- (4)

Solve (3) and (4), we have β1 = −9 and β1 = −8.

⎡1 0⎤ ⎡ 0 1 ⎤ ⎡ − 8 − 9⎤
∴ A4 − 2A3 + A − I = −8⎢ ⎥ − 9⎢ ⎥=⎢ ⎥.
⎣0 1⎦ ⎣− 1 − 2⎦ ⎣ 9 10 ⎦

2. Cayley-Hamilton Theorem may use to determine A 1, the inverse of A.

−

Example 4.13:
⎡ 1 1 − 2⎤
Let A = ⎢− 1 2 1 ⎥ . Use the Cayley Hamilton Theorem to find the inverse of A.
⎢ ⎥
⎢⎣ 0 1 − 1⎥⎦

Solution:
From Example 4.3, the characteristic equation is given by
p(λ) = det(A − λIn) = λ3 − 2λ2 − λ + 2 = 0.

By the Cayley-Hamilton Theorem, the matrix A satisfies

p(A) = A3 − 2A2 − A + 2I3 = 0.

A3 − 2A2 − A + 2 I3 = 0
⇒ A 1(A3 − 2A2 − A + 2I3) = A 1(0)
− −

⇒ (A 1A)A2 − 2(A 1A)A − A 1A + 2A 1 = 0

− − − −

⇒ A2 − 2A − I3 + 2A 1 = 0−

1
⇒ A 1 = (−A2 + 2A + I3 )
−

2
⎧ ⎡ 1 1 − 2⎤ ⎡ 1 1 − 2⎤ ⎡ 1 1 − 2⎤ ⎡1 0 0⎤ ⎫
1 ⎪ ⎪
⇒ A −1 = ⎨− ⎢⎢− 1 2 1 ⎥⎥ ⎢⎢− 1 2 1 ⎥⎥ + 2⎢⎢− 1 2 1 ⎥⎥ + ⎢⎢0 1 0⎥⎥ ⎬
2⎪
⎩ ⎢⎣ 0 1 − 1⎥⎦ ⎢⎣ 0 1 − 1⎥⎦ ⎢⎣ 0 1 − 1⎥⎦ ⎢⎣0 0 1⎥⎦ ⎪⎭
⎧ ⎡ 0 1 1⎤ ⎡ 1 1 − 2⎤ ⎡1 0 0⎤ ⎫ ⎡3 1 − 5⎤
1⎪ ⎢ ⎪
= ⎨− ⎢− 3 4 3⎥⎥ + 2⎢⎢− 1 2 1 ⎥⎥ + ⎢⎢0 1 0⎥⎥ ⎬ = ⎢⎢1 1 − 1⎥⎥ .
2⎪ ⎪
⎩ ⎢⎣ − 1 1 2⎥⎦ ⎢⎣ 0 1 − 1⎥⎦ ⎢⎣0 0 1⎥⎦ ⎭ ⎢⎣1 1 − 3⎥⎦

13