28- Eigenvalues and Eigenvectors VU

Lecture 28

Eigenvalues and Eigenvectors

In this lecture we will discuss linear equations of the form Ax = x and, more generally,
equations of the form Ax = λ x , where λ is a scalar. Such equations arise in a wide
variety of important applications and will be a recurring theme in the rest of this course.

Fixed Points
A fixed point of an n × n matrix A is a vector x in Rn such that Ax = x. Every square
matrix A has at least one fixed point, namely x = 0. We call this the trivial fixed point of
A.
The general procedure for finding the fixed points of a matrix A is to rewrite the equation
Ax = x as Ax = Ix or, alternatively, as
(I – A)x = 0 (1)
Since this can be viewed as a homogeneous linear system of n equations in n unknowns
with coefficient matrix I – A, we see that the set of fixed points of an n × n matrix is a
subspace of Rn that can be obtained by solving (1).

The following theorem will be useful for ascertaining the nontrivial fixed points of a
matrix.

Theorem 1
If A is an n x n matrix, then the following statements are equivalent.
(a) A has nontrivial fixed points.
(b) I – A is singular.
(c) det(I – A) = 0.

Example 1
In each part, determine whether the matrix has nontrivial fixed points; and, if so, graph
the subspace of fixed points in an xy-coordinate system.
3 6  0 2
= (a ) A =  (b) A  
1 2  0 1 

Solution
(a) The matrix has only the trivial fixed point since.

 1 0   3 6   −2 −6 
(I −=A)  − =   
 0 1   1 2   −1 −1 
 −2 −6 
det( I − A) = det  = (−1)(−2) − (−1)(−6) =
− 4≠ 0
 −1 −1 

(b) The matrix has nontrivial fixed points since

©Virtual University Of Pakistan 374

28- Eigenvalues and Eigenvectors VU

 1 0   0 2   1 −2 
( I −=A)  − =   
0 1 0 1 0 0 
 1 −2 
=
det( I − A) det =  0
0 0 

The fixed points x =(x, y) are the solutions of the linear system (I – A)x=0, which we can
express in component form as
1 -2   x  0 
0 0   y  = 0 
    
A general solution of this system is
x = 2t, y = t (2)

which are parametric equations of the line y = 12 x . It follows from the corresponding
vector form of this line that the fixed points are
 x   2t   2 
= x =  =  t  (3)
 y   t  1 
0 2   2t   2t 
As a check, =
Ax   =  =  x
0 1   t   t 
so every vector of form (3) is a fixed point of A.

y = 12 x
(2, 1) •

Figure 1

Eigenvalues and Eigenvectors

In a fixed point problem one looks for nonzero vectors that satisfy the equation Ax = x.
One might also consider whether there are nonzero vectors that satisfy such equations as
Ax = 2x, Ax = –3x, Ax = 2 x
or, more generally, equations of the form Ax = λ x in which λ is a scalar.

Definition If A is an n x n matrix, then a scalar λ is called an eigenvalue of A if there

is a nonzero vector x such that Ax = λ x . If λ is an eigenvalue of A, then every nonzero
vector x such that Ax = λ x is called an eigenvector of A corresponding to λ .

©Virtual University Of Pakistan 375

 28- Eigenvalues and Eigenvectors VU

Example 2
1 6  6 3
=
Let A =  , u =  and v -2  . Are u and v eigenvectors of A?
5 2  -5  
Solution
1 6   6  -24  6
=
Au  =
    = - 4=-5 - 4u
5 2  -5  20   
1 6   3   -9  3
=
Av   =    ≠λ 
5 2  -2  11 -2 
Thus u is an eigenvector corresponding to an eigenvalue – 4, but v is not an eigenvector
of A, because Av is not a multiple of v.

Example 3
1 6 
Show that 7 is an eigenvalue of A =   , find the corresponding eigenvectors.
5 2 
Solution
The scalar 7 is an eigenvalue of A if and only if the equation
Ax = 7x (A)
has a nontrivial solution. But (A) is equivalent to Ax – 7x = 0, or
(A – 7I) x = 0 (B)
To solve this homogeneous equation, form the matrix
1 6  7 0  -6 6 
= A - 7 I = -   
5 2   0 7   5 -5
The columns of A – 7I are obviously linearly dependent, so (B) has nontrivial solutions.
Thus 7 is an eigenvalue of A. To find the corresponding eigenvectors, use row operations:
 −6 6 0 
 5 −5 0 
 
1 −1 0
~ (−1R1 − R2 )
5 −5 0 
1 −1 0
~ ( R2 − 5 R1 )
0 0 0 
1
The general solution has the form x 2   . Each vector of this form with x2 ≠ 0 is an
1
eigenvector corresponding to λ = 7.

The equivalence of equations (A) and (B) obviously holds for any λ in place of λ = 7.
Thus λ is an eigenvalue of A if and only if the equation
(A - λ I)x = 0 (C)
has a nontrivial solution.

©Virtual University Of Pakistan 376

28- Eigenvalues and Eigenvectors VU

Eigen space
The set of all solutions of (A - λ I)x = 0 is just the null space of the matrix A - λ I. So
this set is a subspace of Rn and is called the eigenspace of A corresponding to λ . The
eigenspace consists of the zero vector and all the eigenvectors corresponding to λ .
Example 3 shows that for matrix A in Example 2, the eigenspace corresponding to λ = 7
consists of all multiples of (1, 1), which is the line through (1, 1) and the origin. From
Example 2, one can check that the eigenspace corresponding to λ = -4 is the line through
(6, -5). These eigenspaces are shown in Fig. 1, along with eigenvectors (1, 1) and (3/2, -
5/4) and the geometric action of the transformation x → Ax on each eigenspace.

 4 -1 6 
Example 4 Let A =  2 1 6  .
 
 2 -1 8 
Find a basis for the corresponding eigenspace where eigen value of matrix is 2.

©Virtual University Of Pakistan 377

28- Eigenvalues and Eigenvectors VU

 4 -1 6   2 0 0   2 -1 6 
=
Solution Form A - 2 I =     2 -1 6  and row reduce the
2 1 6 - 0 2 0   
 2 -1 8   0 0 2   2 -1 6 
2 -1 6 0  augmented matrix for (A – 2I) x = 0:
2 -1 6 0 

 2 -1 6 0 
2 -1 6 0 
~  0 0 0 0  R2 − R1
 2 −1 6 0 

 2 -1 6 0 
~  0 0 0 0  R3 − R1
 0 0 0 0 
At this point we are confident that 2 is indeed an eigenvalue of A because the equation
(A – 2I) x = 0 has free variables. The general solution is
2 x1 − x2 + 6 x3 =
0........(a )
Let=
x2 t ,=
x3 s then
2 x1 = t − 6 s
=x1 ( 1 2 ) t − 3s
then
 x1  t / 2 − 3s  t / 2   −3s  1/ 2   −3
 x  = t  = t  + 0  = t 1  + s 0 
 2          
 x3   s  0   s  0  1 

By back substitution the general solution is

 x1  1 2  -3
=     
 x2  x2  1  + x3  0  , x2 and x3 free
 x3   0   1 
The eigenspace, shown in Fig. 2, is a two – dimensional subspace of R3. A basis is
  1   -3 
    
  2  ,  0   is a basis.
0  1  
    

©Virtual University Of Pakistan 378

28- Eigenvalues and Eigenvectors VU

The most direct way of finding the eigenvalues of an n × n matrix A is to rewrite the
equation Ax = λ x as Ax = λ Ix , or equivalently, as
(λ I - A) x = 0 (4)
and then try to determine those values of λ , if any, for which this system has nontrivial
solutions. Since (4) have nontrivial solutions if and only if the coefficient matrix λ I - A is
singular, we see that the eigenvalues of A are the solutions of the equation
det(λ I - A) = 0 (5)
Equation (5) is known as characteristic equation. Also, if λ is an eigenvalue of A, then
equation (4) has a nonzero solution space, which we call the eigenspace of A
corresponding to λ . It is the nonzero vectors in the eigenspace of A corresponding to λ
that are the eigenvectors of A corresponding to λ .
The above discussion is summarized by the following theorem.

Theorem If A is an n × n matrix and λ is a scalar, then the following statements are

equivalent.
(i) λ is an eigenvalue of A.
(ii) λ is a solution of the equation det(λ I - A) = 0 .
(iii) The linear system (λ I - A) x = 0 has nontrivial solutions.

Eigenvalues of Triangular Matrices If A is an n × n triangular matrix with diagonal

entries a11, a22, …, ann, then λ I - A is a triangular matrix with diagonal entries
λ - a11 , λ - a22 , , λ - ann . Thus, the characteristic polynomial of A is
det(λ I - A) = (λ - a11 )(λ - a22 )  (λ - ann )
which implies that the eigenvalues of A are
= λ1 a= 11 , λ2 =
a22 ,  , λn ann
Thus, we have the following theorem.

©Virtual University Of Pakistan 379

 28- Eigenvalues and Eigenvectors VU

Theorem If A is a triangular matrix (upper triangular, lower triangular, or diagonal)

then the eigenvalues of A are the entries on the main diagonal of A.

Example 5 (Eigenvalues of Triangular Matrices)

 12 0 0 0
-1 - 2 0 0 
By inspection, the characteristic polynomial of the matrix A =  3
is
 7 85 6 0
4 
 9 -4 3 6
p (λ ) (λ - 12 )(λ + 23 )(λ - 6) 2 . So the distinct eigenvalues of A are λ = 12 ,
= λ = − 23 , and
λ =6.

Eigenvalues of Powers of a Matrix Once the eigenvalues and eigenvectors of a

matrix A are found, it is a simple matter to find the eigenvalues and eigenvectors of any
positive integer power of A. For example, if λ is an eigenvalue of A and x is a
corresponding eigenvector, then = A2 x A= ( Ax) A= (λ x) λ=( Ax) λ=(λ x) λ 2 x , which
shows that λ 2 is an eigenvalue of A2 and x is a corresponding eigenvector. In general we
have the following result.

Theorem If λ is an eigenvalue of a matrix A and x is a corresponding eigenvector,

and if k is any positive integer, then λ k is an eigenvalue of Ak and x is a corresponding
eigenvector.
Some problems that use this theorem are given in the exercises.

A Unifying Theorem Since λ is an eigenvalue of a square matrix A if and only if

there is a nonzero vector x such that Ax = λ x, it follows that λ = 0 is an eigenvalue of A
if and only if there is a nonzero vector x such that Ax = 0. However, this is true if and
only if det(A) = 0, so we list the following

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

(a) The reduced row echelon form of A is I n .
(b) A is expressible as a product of elementary matrices.
(c) A is invertible.
(d) Ax = 0 has only the trivial solution.
(e) Ax = b is consistent for every vector b in Rn.
(f) Ax = b has exactly one solution for every vector b in Rn
(g) The column vectors of A are linearly independent.
(h) The row vectors of A are linearly independent.
(i) det(A) ≠ 0.
(j) λ = 0 is not an eigenvalue of A.

©Virtual University Of Pakistan 380

 28- Eigenvalues and Eigenvectors VU

Example 6
 6 -3 1 
(1) Is 5 an eigenvalue of A =  3 0 5  ?
 
 2 2 6 
(2) If x is an eigenvector for A corresponding to λ , what is A3x?

Solution
(1) The number 5 is an eigenvalue of A if and only if the equation (A- λ I) x = 0 has a
nontrivial solution. Form
 6 -3 1  5 0 0  1 -3 1
= A - 5 I = 3 0 5  - 0 5 0  
 3 -5 5


 2 2 6  0 0 5   2 2 1
and row reduce the augmented matrix:
1 -3 1 0 
 3 -5 5 0 
 
 2 2 1 0 
1 -3 1 0 
~  0 4 2 0  R2 − 3R1
 2 2 1 0 
1 -3 1 0 
~ 0 4 2 0  R3 − 2 R1
0 8 -1 0 
1 -3 1 0 
~ 0 4 2 0  R3 − 2 R2
0 0 -5 0 
At this point it is clear that the homogeneous system has no free variables. Thus A – 5I is
an invertible matrix, which means that 5 is not an eigenvalue of A.

(2). If x is an eigenvector for A corresponding to λ , then Ax = λ x and so

=
A 2
x A(λ =x ) λ= Ax λ 2 x
Again = A3 x A( A =2
x ) A(λ=2
x ) λ=
2
Ax λ 3 x. The general pattern, Ak x = λ k x , is
proved by induction.

Exercises

3 2 
1. Is λ = 2 an eigenvalue of  ?
3 8 

©Virtual University Of Pakistan 381

 28- Eigenvalues and Eigenvectors VU

 −1 + 2  2 1
2. Is   an eigenvector of 1 4  ? If so, find the eigenvalue.
 1   

4  3 7 9
3. Is  −3 an eigenvector of
 −4 −5 1  ? If so, find the eigenvalue.
   
1   2 4 4 

1 3 6 7 
4. Is  −2  an eigenvector of
3 3 7  ? If so, find the eigenvalue.
   
1  5 6 5 

 3 0 −1
 
5. Is λ = 4 an eigenvalue of 2 3 1 ? If so, find one corresponding eigenvector.
 
 −3 4 5 

1 2 2 
 
6. Is λ = 3 an eigenvalue of 3 −2 1 ? If so, find one corresponding eigenvector.
 
0 1 1 

In exercises 7 to 12, find a basis for the eigenspace corresponding to each listed
eigenvalue.

 4 −2  7 4
=7. A =
−  ,λ 10 =8. A =
− −  ,λ 1 5.
 3 9   3 1

 4 0 1  1 0 −1
 0  ,2,3
9. A = −2 1λ =1,

10. A = 1 −3 0 λ ,=
 2−
  
 −2 0 1   4 −13 1 

3 0 2 0
4 2 3 1
 −1 1λ =−33 , 3 1 0 
11. A = 12. A =  ,λ = 4
  0 1 1 0
 2 4 9   
0 0 0 4

Find the eigenvalues of the matrices in Exercises 13 and 14.

©Virtual University Of Pakistan 382

28- Eigenvalues and Eigenvectors VU

0 0 0  4 0 0 

13. 0 2 5
 
14. 0 0 0

   
0 0 −1 1 0 −3

1 2 3
 
15. For A = 1 2 3 , find one eigenvalue, with no calculation. Justify your answer.
 
1 2 3

16. Without calculation, find one eigenvalue and two linearly independent vectors of
5 5 5 
A = 5 5 5 . Justify your answer.
5 5 5

©Virtual University Of Pakistan 383