You are on page 1of 34

Elementary Linear Algebra

Anton & Rorres, 9th Edition

Chapter 5:

Eigenvalues, Eigenvectors
Chapter Content

 Eigenvalues and Eigenvectors


 Diagonalization
 Orthogonal Digonalization

2018/2/26 Elementary Linear Algebra 2


7-1 Eigenvalue and Eigenvector

 If A is an nn matrix
 a nonzero vector x in Rn is called an eigenvector of A if Ax
is a scalar multiple of x;
 that is, Ax = x for some scalar .
 The scalar  is called an eigenvalue of A, and x is said to be
an eigenvector of A corresponding to .

2018/2/26 Elementary Linear Algebra 3


7-1 Eigenvalue and Eigenvector

 Remark
 To find the eigenvalues of an nn matrix A we rewrite Ax =
x as Ax = Ix or equivalently, (I – A)x = 0.
 For  to be an eigenvalue, there must be a nonzero solution
of this equation. However, by Theorem 6.4.5, the above
equation has a nonzero solution if and only if
det (I – A) = 0.
 This is called the characteristic equation of A; the scalar
satisfying this equation are the eigenvalues of A. When
expanded, the determinant det (I – A) is a polynomial p in
 called the characteristic polynomial of A.

2018/2/26 Elementary Linear Algebra 4


7-1 Example 2
 Find the eigenvalues of
0 1 0
A  0 0 1 
 4 17 8 
 Solution:
 The characteristic polynomial of A is
  1 0 
det( I  A)  det  0  1    3  8 2  17  4
 4 17   8
 The eigenvalues of A must therefore satisfy the cubic equation
3 – 82 + 17 – 4 =0

2018/2/26 Elementary Linear Algebra 5


7-1 Example 3

 Find the eigenvalues of the upper triangular matrix


a11 a12 a13 a14 
0 a a a24 
A 22 23

 0 0 a33 a34 
 
0 0 0 a44 

2018/2/26 Elementary Linear Algebra 6


Theorem 7.1.1

 If A is an nn triangular matrix (upper triangular, low


triangular, or diagonal)
 then the eigenvalues of A are entries on the main diagonal
of A.

 Example 4
 The eigenvalues of the lower triangular matrix
1 
2 0 0 
 2 
A   1 0 
 3 
 5 8  
1
 4 

2018/2/26 Elementary Linear Algebra 7


Theorem 7.1.2 (Equivalent Statements)

 If A is an nn matrix and  is a real number, then the


following are equivalent.
  is an eigenvalue of A.
 The system of equations (I – A)x = 0 has nontrivial solutions.
 There is a nonzero vector x in Rn such that Ax = x.
  is a solution of the characteristic equation det(I – A) = 0.

2018/2/26 Elementary Linear Algebra 8


7-1 Finding Bases for Eigenspaces

 The eigenvectors of A corresponding to an eigenvalue 


are the nonzero x that satisfy Ax = x.

 Equivalently, the eigenvectors corresponding to  are the


nonzero vectors in the solution space of (I – A)x = 0.

 We call this solution space the eigenspace of A


corresponding to .

2018/2/26 Elementary Linear Algebra 9


7-1 Example 5
0 0 2 
 Find bases for the eigenspaces of A  1 2 1 
1 0 3 

 Solution:
 The characteristic equation of matrix A is 3 – 52 + 8 – 4 = 0, or in
factored form, ( – 1)( – 2)2 = 0; thus, the eigenvalues of A are  = 1
and  = 2, so there are two eigenspaces of A.
 0 2   x1  0
 1   2 1   x   0
 (I – A)x = 0    2   (3)
 1 0   3  x3  0

 If  = 2, then (3) becomes  2 0 2  x1  0


 1 0 1  x   0
  2  
 1 0 1  x3  0

2018/2/26 Elementary Linear Algebra 10


7-1 Example 5  2 0 2  x1  0
 1 0 1  x   0
  2  
 1 0 1  x3  0
 Solving the system yield
x1 = -s, x2 = t, x3 = s
 Thus, the eigenvectors of A corresponding to  = 2 are the nonzero
vectors of the form
  s    s  0  1 0
x   t    0    t   s  0   t 1
 s   s  0  1  0
 The vectors [-1 0 1]T and [0 1 0]T are linearly independent and form a
basis for the eigenspace corresponding to  = 2.
 Similarly, the eigenvectors of A corresponding to  = 1 are the nonzero
vectors of the form x = s [-2 1 1]T
 Thus, [-2 1 1]T is a basis for the eigenspace corresponding to  = 1.

2018/2/26 Elementary Linear Algebra 11


Theorem 7.1.3

 If k is a positive integer,  is an eigenvalue of a


matrix A, and x is corresponding eigenvector
 then k is an eigenvalue of Ak and x is a corresponding
eigenvector.

 Example 6 (use Theorem 7.1.3)


0 0 2 
A  1 2 1 
1 0 3 

2018/2/26 Elementary Linear Algebra 12


Theorem 7.1.4

 A square matrix A is invertible if and only if  = 0 is


not an eigenvalue of A.
 (use Theorem 7.1.2)

 Example 7
 The matrix A in the previous example is invertible since it has
eigenvalues  = 1 and  = 2, neither of which is zero.

2018/2/26 Elementary Linear Algebra 13


Theorem 7.1.5 (Equivalent Statements)
 If A is an mn matrix, and if TA : Rn  Rn is multiplication
by A, then the following are equivalent:
 A is invertible.
 Ax = 0 has only the trivial solution.
 The reduced row-echelon form of A is In.
 A is expressible as a product of elementary matrices.
 Ax = b is consistent for every n1 matrix b.
 Ax = b has exactly one solution for every n1 matrix b.
 det(A)≠0.
 The range of TA is Rn.
 TA is one-to-one.
 The column vectors of A are linearly independent.
 The row vectors of A are linearly independent.

2018/2/26 Elementary Linear Algebra 14


Theorem 7.1.5 (Equivalent Statements)

 The column vectors of A span Rn.


 The row vectors of A span Rn.
 The column vectors of A form a basis for Rn.
 The row vectors of A form a basis for Rn.
 A has rank n.
 A has nullity 0.
 The orthogonal complement of the nullspace of A is Rn.
 The orthogonal complement of the row space of A is {0}.
 ATA is invertible.
  = 0 is not eigenvalue of A.

2018/2/26 Elementary Linear Algebra 15


Chapter Content

 Eigenvalues and Eigenvectors


 Diagonalization
 Orthogonal Digonalization

2018/2/26 Elementary Linear Algebra 16


7-2 Diagonalization

 A square matrix A is called diagonalizable


 if there is an invertible matrix P such that P-1AP is a
diagonal matrix (i.e., P-1AP = D);
 the matrix P is said to diagonalize A.

 Theorem 7.2.1
 If A is an nn matrix, then the following are equivalent.
 A is diagonalizable.

 A has n linearly independent eigenvectors.

2018/2/26 Elementary Linear Algebra 17


7-2 Procedure for Diagonalizing a Matrix

 The preceding theorem guarantees that an nn matrix A with n linearly


independent eigenvectors is diagonalizable, and the proof provides the
following method for diagonalizing A.
 Step 1. Find n linear independent eigenvectors of A, say, p1, p2, …, pn.

 Step 2. From the matrix P having p1, p2, …, pn as its column vectors.

 Step 3. The matrix P-1AP will then be diagonal with 1, 2, …, n as
its successive diagonal entries, where i is the eigenvalue
corresponding to pi, for i = 1, 2, …, n.

2018/2/26 Elementary Linear Algebra 18


7-2 Example 1
0 0 2 
 Find a matrix P that diagonalizes A  1 2 1 
 Solution: 1 0 3 

 From the previous example, we have the following bases for the
eigenspaces:
  = 2:  1 0  = 1:   2
p1   0 , p 2  1 p 3   1 
 Thus,  1  0  1 

  1 0  2
P   0 1 1 
 Also,  1 0 1 
 1 0 2  0 0  2  1 0  2 2 0 0
P 1 AP   1 1 1  1 2 1   0 1 1   0 2 0  D
 1 0  1 1 0 3   1 0 1  0 0 1

2018/2/26 Elementary Linear Algebra 19


7-2 Example 2
(A Non-Diagonalizable Matrix)
 Find a matrix P that diagonalizes  1 0 0
A   1 2 0
 Solution:  3 5 2
 The characteristic polynomial of A is
 1 0 0
det( I  A)  1  2 0  (  1)(  2) 2
3 5   2
 The bases for the eigenspaces are
  = 1:  1/ 8   = 2: 0 
p1   1 / 8 p 2  0
 1  1
 Since there are only two basis vectors in total, A is not diagonalizable.

2018/2/26 Elementary Linear Algebra 20


7-2 Theorems

 Theorem 7.2.2
 If v1, v2, …, vk, are eigenvectors of A corresponding to
distinct eigenvalues 1, 2, …, k,
 then {v1, v2, …, vk} is a linearly independent set.

 Theorem 7.2.3
 If an nn matrix A has n distinct eigenvalues
 then A is diagonalizable.

2018/2/26 Elementary Linear Algebra 21


7-2 Example 3

 Since the matrix


0 1 0
A  0 0 1 
 4 17 8 
has three distinct eigenvalues,   4,   2  3,   2  3
 Therefore, A is diagonalizable.
 Further,

4 0 0 
 
P 1 AP   0 2 3 0 
0 2  3 
 0
for some invertible matrix P, and the matrix P can be found using
the procedure for diagonalizing a matrix.

2018/2/26 Elementary Linear Algebra 22


7-2 Example 4 (A Diagonalizable Matrix)
 Since the eigenvalues of a triangular matrix are the entries
on its main diagonal (Theorem 7.1.1).

 Thus, a triangular matrix with distinct entries on the main


diagonal is diagonalizable.

 For example,  1 2 4 0
0 3 1 7 
A
0 0 5 8
 
0 0 0 2 

is a diagonalizable matrix.

2018/2/26 Elementary Linear Algebra 23


7-2 Example 5
(Repeated Eigenvalues and Diagonalizability)
 Whether the following matrices are diagonalizable?

1 0 0
I 3  0 1 0
0 0 1

1 1 0
J 3  0 1 1
0 0 1

2018/2/26 Elementary Linear Algebra 24


7-2 Geometric and Algebraic Multiplicity

 If 0 is an eigenvalue of an nn matrix A


 then the dimension of the eigenspace corresponding to 0 is
called the geometric multiplicity of 0, and

 the number of times that  – 0 appears as a factor in the


characteristic polynomial of A is called the algebraic
multiplicity of A.

2018/2/26 Elementary Linear Algebra 25


Theorem 7.2.4
(Geometric and Algebraic Multiplicity)
 If A is a square matrix, then :

 For every eigenvalue of A the geometric multiplicity is less


than or equal to the algebraic multiplicity.

 A is diagonalizable if and only if the geometric multiplicity


is equal to the algebraic multiplicity for every eigenvalue.

2018/2/26 Elementary Linear Algebra 26


7-2 Computing Powers of a Matrix

 If A is an nn matrix and P is an invertible matrix, then (P-1AP)k = P-


1AkP for any positive integer k.

 If A is diagonalizable, and P-1AP = D is a diagonal matrix, then


P-1AkP = (P-1AP)k = Dk
 Thus,
Ak = PDkP-1
 The matrix Dk is easy to compute; for example, if
 d1 0 ... 0   d1k 0 ... 0 
 0 d ... 0   k 
 , and Dk   0 d ... 0 
D 2 2
: : :  : : : 
   k
 0 0 ... d n
  0 0 ... d n 

2018/2/26 Elementary Linear Algebra 27


7-2 Example 6 (Power of a Matrix)

 Find A13
0 0 2 
A  1 2 1 
1 0 3 

2018/2/26 Elementary Linear Algebra 28


Chapter Content

 Eigenvalues and Eigenvectors


 Diagonalization
 Orthogonal Digonalization

2018/2/26 Elementary Linear Algebra 29


7-3 The Orthogonal Diagonalization
Matrix Form
 Given an nn matrix A, if there exist an orthogonal matrix
P such that the matrix
P-1AP = PTAP
then A is said to be orthogonally diagonalizable and P is
said to orthogonally diagonalize A.

2018/2/26 Elementary Linear Algebra 30


Theorem 7.3.1 & 7.3.2

 If A is an nn matrix, then the following are equivalent.


 A is orthogonally diagonalizable.
 A has an orthonormal set of n eigenvectors.
 A is symmetric.

 If A is a symmetric matrix, then:


 The eigenvalues of A are real numbers.
 Eigenvectors from different eigenspaces are orthogonal.

2018/2/26 Elementary Linear Algebra 31


7-3 Diagonalization of Symmetric Matrices

 The following procedure is used for orthogonally


diagonalizing a symmetric matrix.
 Step 1. Find a basis for each eigenspace of A.
 Step 2. Apply the Gram-Schmidt process to each of these bases
to obtain an orthonormal basis for each eigenspace.
 Step 3. Form the matrix P whose columns are the basis vectors
constructed in Step2; this matrix orthogonally diagonalizes A.

2018/2/26 Elementary Linear Algebra 32


7-3 Example 1
 Find an orthogonal matrix P that diagonalizes  4 2 2
A   2 4 2
 2 2 4
 Solution:
 The characteristic equation of A is
  4 2 2 
det( I  A)  det  2   4 2   (  2) 2 (  8)  0
 2 2   4 1  1
  1  and u   0 
 The basis of the eigenspace corresponding to  = 2 is 1   u 2  
 Applying the Gram-Schmidt process to {u1, u2} yields  0   1 
the following orthonormal eigenvectors:
 1/ 2   1/ 6 
   
v1   1/ 2  and v 2   1/ 6 
 0   
   2 / 6 

2018/2/26 Elementary Linear Algebra 33


7-3 Example 1
1
 The basis of the eigenspace corresponding to  = 8 is u3  1
 Applying the Gram-Schmidt process to {u } yields: 1
3
1/ 3 
 
v 3  1/ 3 
 
1/ 3 
 Thus,
 1 / 2 1/ 6 1/ 3
 
P  v1 v2 v3    1/ 2 1/ 6 1/ 3
 0 2 / 6 1 / 3 

orthogonally diagonalizes A.

2018/2/26 Elementary Linear Algebra 34