Chapter V Eigenvalues and Eigenvectors

Definition If 𝐴 is an 𝑛 × 𝑛 matrix, then a non-zero vector 𝐱 in ℝ𝑛 is called an eigenvector of

𝐴 (or of the matrix operator 𝑇𝐴 ) if 𝐴𝐱 is a scalar multiple of 𝐱; i.e 𝐴𝐱 = 𝜆𝐱 for some scalar 𝜆.
The scalar 𝜆 is called an eigenvalue of 𝐴 (or of 𝑇𝐴 ) and 𝐱 is said to be an eigenvector
corresponding to 𝜆.
1 3 0
Example The vector 𝐱 = ( ) is an eigenvector of 𝐴 = ( ) corresponding to
2 8 −1
eigenvalue λ=3, since
3 0 1
𝐴𝐱 = ( ) ( ) = 3𝐱
8 −1 2
𝑎1 0 … 0
0 𝑎2 … 0
Example Let 𝐴 = ( ) be a diagonal matrix, then every unit vector 𝐸 𝑖 (𝑖 =
⋮ ⋮ ⋮
0 … 𝑎𝑛
1, 2, … , 𝑛) is an eigenvector of 𝐴.
We have 𝐴𝐸 𝑖 = 𝑎𝑖 𝐸 𝑖 . Hence, 𝑎𝑖 is eigenvalue and 𝐸 𝑖 is eigenvector.
How to find eigenvalues and eigenvectors?
𝐴𝐱 = 𝜆𝐱 can be written as 𝐴𝐱 = 𝜆𝐼𝐱. i.e (𝜆𝐼 − 𝐴)𝐱 = 𝟎 .
For this equation to have non-trivial solution 𝜆𝐼 − 𝐴 should not be invertible or |𝜆𝐼 − 𝐴 | =
0.
Theorem If 𝐴 is an 𝑛 × 𝑛 matrix, then 𝜆 is an eigenvalue of 𝐴 if and only if |𝜆𝐼 − 𝐴 | = 0.
When |𝜆𝐼 − 𝐴 | is expanded, it takes the form 𝑃(𝜆) = 𝜆 𝑛 + 𝑐1 𝜆 𝑛−1 + ⋯ + 𝑐𝑛 , a polynomial
of degree 𝑛, and is called a Characteristic polynomial.
|𝜆𝐼 − 𝐴 | = 0 is the characteristic equation of 𝐴.
3 0
Example Find eigenvalues and eigenvectors of 𝐴 = ( )
8 −1
𝜆−3 0
Solution |𝜆𝐼 − 𝐴 | = 0 ⇒ | |=0
−8 𝜆 + 1
⇒ (𝜆 − 3)(𝜆 + 1) = 0
⇒ 𝜆 = 3, 𝜆 = −1 are eigenvalues of 𝐴.
To compute the eigenvectors,
𝑥1
(𝜆𝐼 − 𝐴)𝐱 = 𝟎 ⇒ (𝜆 − 3 0 0
) (𝑥 ) = ( )
−8 𝜆+1 2 0
If 𝜆 = 3
0 0 𝑥1 0
( ) (𝑥 ) = ( )
−8 4 2 0
−8𝑥1 + 4𝑥2 = 0 or 𝑥2 = 2𝑥1
Let 𝑥1 = 1. Then 𝑥2 = 2
1
( ) is an eigenvector.
2
1
Of course, all non-zero scalar multiples of ( ) is an eigenvector.
2
1
Hence, {𝑡 ( ) |𝑡 ∈ ℝ\{0}} is the set of eigenvectors (called eigenspace) corresponding to
2
eigenvalue 𝜆 = 3.
0 1 0
Example Find the eigenvalues of 𝐴 = (0 0 1)
4 −17 8
𝜆 −1 0
Solution The characteristic polynomial of 𝐴 is |𝜆𝐼 − 𝐴| = ( 0 𝜆 −1 ) = 𝜆3 −
−4 17 𝜆 − 8
8𝜆2 + 17𝜆 − 4
The eigenvalues of 𝐴 must satisfy the equation 𝜆3 − 8𝜆2 + 17𝜆 − 4 = 0
The roots must be divisors of -4 which are ±1, ±2, and ± 4. 𝜆 = 4 is a solution, hence 𝜆 − 4
is a factor. i.e (𝜆 − 4)(𝜆2 − 4𝜆 + 1) = 0.
The other solutions satisfy 𝜆2 − 4𝜆 + 1 = 0.

4 ± √16 − 4 4 ± 2√3
⇒𝜆= = = 2 ± √3
2 2

i.e 𝜆 = 4, 𝜆 = 2 + √3 and 𝜆 = 2 − √3
Theorem If 𝐴 is an 𝑛 × 𝑛 triangular matrix (upper triangular, lower triangular or diagonal),
then the eigenvalues of 𝐴 are 𝐴 are the entries on the main diagonal of 𝐴.
1
0 0
2
2
Example Find eigenvalues of a lower triangular matrix 𝐴 = −1 0
3
−1
5 −8
( 4 )
1
𝜆− 0 0
2
2 1 2 1
Solution |𝜆𝐼 − 𝐴| = || 1 𝜆−3 0 || = (𝜆 − ) (𝜆 − ) (𝜆 + ) = 0
2 3 4
1
−5 8 𝜆+4

1 2 −1
The eigenvalues are 𝜆 = 2 , 𝜆 = 3 , and 𝜆 = 4

Theorem If 𝐴 is an 𝑛 × 𝑛 matrix, the following statements are equivalent:

 a) 𝜆 is an eigenvalue of 𝐴
b) 𝜆 is a solution of the characteristic equation det(𝜆𝐼 − 𝐴) = 0
c) The system of equations (𝜆𝐼 − 𝐴)𝐱 = 𝟎 has nontrivial solutions
d) There is a nonzero vector 𝐱 such that 𝐴𝐱 = 𝜆𝐱
Eigenspace of 𝐴 corresponding to 𝜆 can also be viewed as:
 The null space of the matrix 𝜆𝐼 − 𝐴
 The kernel of the matrix operator 𝑇𝜆𝐼−𝐴 : ℝ𝑛 → ℝ𝑛
 The set of vectors for which 𝐴𝐱 = 𝜆𝐱
Example (Bases for Eigenspace)
−1 3
Find bases for the eigenspace of the matrix 𝐴 = ( ).
2 0
Solution The characteristic equation of 𝐴 is
𝜆+1 −3
| | = 𝜆(𝜆 + 1) − 6 = (𝜆 − 2)(𝜆 + 3) = 0
−2 𝜆
So the eigenvalues of 𝐴 are 𝜆 = 2 and 𝜆 = −3.

(𝜆𝐼 − 𝐴)𝐱 = 𝟎 ⇒ (𝜆 + 1 −3 𝑥1 0
) (𝑥 ) = ( )
−2 𝜆 2 0
If 𝜆 = 2,
3 −3 𝑥1 0
( ) (𝑥 ) = ( )
−2 2 2 0
3𝑥 − 3𝑥2 = 0
⇒{ 1 ⇒ 𝑥1 = 𝑡, 𝑥2 = 𝑡
−2𝑥1 + 2𝑥2 = 0
𝑥1 𝑡 1
(𝑥 ) = ( ) = 𝑡 ( )
2 𝑡 1
1
⇒ ( ) is a basis for the eigenspace corresponding to 𝜆 = 2.
1
If 𝜆 = −3,
−2 −3 𝑥1 0
( ) (𝑥 ) = ( )
−2 −3 2 0
−3
⇒−2𝑥1 − 3𝑥2 = 0 ⇒ 𝑥1 = 𝑥2
2
−3
Let 𝑥2 = 𝑡, 𝑥1 = 𝑡
2

𝑥1 −3 −3
𝑡
(𝑥 ) = ( 2 ) = 𝑡 ( 2 )
2
𝑡 1
−3
⇒ ( 2 ) is a basis for the eigenspace corresponding to 𝜆 = −3.
1
 0 0 −2
Example Find bases for the eigenspaces of 𝐴 = (1 2 1)
1 0 3
Solution The characteristic equation is 𝜆3 − 5𝜆2 + 8𝜆 − 4 = 0 or (𝜆 − 1)(𝜆 − 2)2 = 0.
𝜆 0 2 𝑥1 0
(𝜆𝐼 − 𝐴)𝐱 = 𝟎 ⇒ (−1 𝜆 − 2 −1 ) (𝑥𝟐 ) = (0)
−1 0 𝜆 − 3 𝑥𝟑 0
If 𝜆 = 2
2 0 2 𝑥1 0
(−1 0 𝑥
−1) ( 𝟐 ) = (0)
−1 0 −1 𝑥𝟑 0
𝑥1 + 𝑥3 = 0
Let 𝑥1 = −𝑠, 𝑥𝟐 = 𝑡, 𝑥3 = 𝑠
−𝑠 −𝑠 0 −1 0
𝐱 = ( 𝑡 ) = ( 0 ) + ( 𝑡 ) = 𝑠 ( 0 ) + 𝑡 (1)
𝑠 𝑠 0 1 0
−1 0
( 0 ) and (1) are linearly independent. They form a basis for the eigenspace
1 0
corresponding to 𝜆 = 2.
If 𝜆 = 1
1 0 2 𝑥1 0
(−1 −1 −1) (𝑥𝟐 ) = (0)
−1 0 −2 𝑥𝟑 0
𝑥1 + 2𝑥3 = 0
𝑥1 = −2𝑥3
⇒ {−𝑥1 − 𝑥2 − 𝑥3 = 0 ⇒ {
𝑥2 = −𝑥1 − 𝑥3 = 𝑥3
−𝑥1 − 2𝑥3 = 0
Let 𝑥3 = 𝑡, 𝑥𝟏 = −2𝑡, 𝑥2 = 𝑡
−2𝑡 −2
𝐱 = ( 𝑡 ) = 𝑡( 1 )
𝑡 1
−2
Hence, ( 1 ) is a basis for the eigenspace corresponding to 𝜆 = 1.
1
Theorem A square matrix 𝐴 is invertible if and only if 𝜆 = 0 is not an eigenvalue of 𝐴.
Proof 𝜆 = 0 is a solution of the characteristic equation
𝜆𝑛 + 𝑐1 𝜆𝑛−1 + ⋯ + 𝑐𝑛 = 0 if and only if 𝑐𝑛 = 0.
|𝜆𝐼 − 𝐴| = 𝜆𝑛 + 𝑐1 𝜆𝑛−1 + ⋯ + 𝑐𝑛
If 𝜆 = 0, det(−A) = (−1)𝑛 det(𝐴) = 𝑐𝑛
𝐴 is invertible if and only if det(𝐴) = (−1)𝑛 𝑐𝑛 ≠ 0
Example (Eigenvalues and Invertibility)
0 0 −2
𝐴 = (1 2 1 ) has eigenvalues 𝜆 = 1 and 𝜆 = 2, none of which is 0. Hence, it is
1 0 3
invertible.
Definition If 𝑇: 𝑽 → 𝑽 is a linear operator on a vector space 𝑽, then a nonzero vector 𝐱 in 𝑽
is called an eigenvector of 𝑇 if 𝑇(𝐱) is a scalar multiple of 𝐱; i.e 𝑇(𝐱) = 𝜆𝐱 for some scalar 𝜆.
The scalar 𝜆 is called an eigenvalue of 𝑇, and 𝐱 is said to be an eigenvector corresponding to
𝜆.
Example If 𝐷: 𝐶 ∞ → 𝐶 ∞ is a differentiation operator on a vector space of functions with
continuous derivatives of all orders on the interval (−∞, ∞), and if λ is a constant, then
𝐷(𝑒 𝜆𝑥 ) = 𝜆𝑒 𝜆𝑥 . Thus, 𝜆 is eigenvalue of 𝐷 and 𝑒 𝜆𝑥 is eigenvector corresponding to 𝜆.

Theorem Let 𝑽 be a vector space and let 𝑇: 𝑽 → 𝑽 be a linear map. Let 𝑉𝜆 be the subspace of
𝑉 generated by all eigenvectors of 𝐴 having 𝜆 as eigenvalue. Then every non-zero element of
𝑉𝜆 is an eigenvector of 𝐴 having 𝜆 as eigenvalue.
Proof Let 𝑣1 , 𝑣2 ∈ 𝑽 be such that 𝑇𝑣1 = 𝜆 𝑣1 and 𝑇𝑣2 = 𝜆 𝑣2 . Then
𝑇(𝑣1 + 𝑣2 ) = 𝑇𝑣1 + 𝑇𝑣2 = 𝜆 𝑣1 + 𝜆 𝑣2 = 𝜆 (𝑣1 + 𝑣2 )
If 𝑐 ∈ 𝑲, then 𝑇(𝑐𝑣1 ) = 𝑐𝑇(𝑣1 ) = 𝑐𝜆 𝑣1 = 𝜆(𝑐𝑣1 )
Thus the subspace 𝑉𝜆 is called the eigenspace of 𝐴 belonging to 𝜆.
Is 𝑉𝜆 a subspace of vector space 𝑽.
Theorem Let V be a vector space and let 𝑇: 𝑽 → 𝑽 be a linear map. Let 𝑣1 , 𝑣2 , … , 𝑣𝑚 be
eigenvectors of 𝑇, with eigenvalues 𝜆1 , 𝜆2 , … , 𝜆𝑚 respectively. Assume that these eigenvalues
are distinct, i.e 𝜆𝑖 ≠ 𝜆𝑗 if 𝑖 ≠ 𝑗. Then 𝑣1 , 𝑣2 , … , 𝑣𝑚 are linearly independent.

Proof By induction on 𝑚. For 𝑚 = 1, an element 𝑣1 ∈ 𝑽, 𝑣1 ≠ 𝟎 is linearly independent.

Assume 𝑚 > 1. Suppose we have a relation
𝑐1 𝑣1 + 𝑐2 𝑣2 + ⋯ + 𝑐𝑚 𝑣𝑚 = 𝟎 … (∗) with scalars 𝑐𝑖 . We must prove all 𝑐𝑖 = 0.
If we multiply (*) by 𝜆1 , we get
𝑐1 𝜆1 𝑣1 + 𝑐2 𝜆1 𝑣2 + ⋯ + 𝑐𝑚 𝜆1 𝑣𝑚 = 𝟎 … (∗∗)
If we apply 𝑇 to (*), we get
𝑐1 𝜆1 𝑣1 + 𝑐2 𝜆2 𝑣2 + ⋯ + 𝑐𝑚 𝜆𝑚 𝑣𝑚 = 𝟎 … (∗∗∗)
Subtracting (**) from (***) gives us,
𝑐2 (𝜆2 − 𝜆1 )𝑣2 + ⋯ + 𝑐𝑚 (𝜆𝑚 − 𝜆1 )𝑣𝑚 = 𝟎
Since 𝜆𝑗 − 𝜆1 ≠ 0 𝑓𝑜𝑟 𝑗 = 2, … , 𝑚, we conclude by induction that 𝑐2 = ⋯ = 𝑐𝑚 = 0.

Going back to (*), we see that 𝑐1 𝑣1 = 𝟎, hence 𝑐1 = 0.