8/3/2019 Eigen,Diag,Sumetrix
1/50
Eigenvalues and Eigenvectors
8/3/2019 Eigen,Diag,Sumetrix
2/50
16- 2
Eigenvalues and Eigenvectors
Eigenvalues and Eigenvectors
Diagonalization
Symmetric Matrices and Orthogonal Diagonalization
8/3/2019 Eigen,Diag,Sumetrix
3/50
7.1 Eigenvalues and Eigenvectors
Eigenvalue problem:
IfA is an nvn matrix, do there exist nonzero vectors x in Rn
such thatAx is a scalar multiple ofx
Eigenvalue and eigenvector:Aan nvn matrix
Pa scalar
x a nonzero vector in Rn
xAx P!
Eigenvalue
Eigenvector
Geometrical Interpretation
16- 3
8/3/2019 Eigen,Diag,Sumetrix
4/50
Ex 1: (Verifying eigenvalues and eigenvectors)
-
!
10
02A
-
!
0
11x
11 2012
02
01
1002 xAx !
-!
-!
--
!
Eigenvalue
22 )1(1
01
1
0
1
0
10
02xAx !
-
!
-
!
-
-
!
Eigenvalue
Eigenvector
Eigenvector
-
!
1
02
x
16- 4
8/3/2019 Eigen,Diag,Sumetrix
5/50
Ex 2: (An example of eigenspaces in the plane)
Find the eigenvalues and corresponding eigenspaces of
-
!
10
01A
-
!
-
-
!
y
x
y
xA
10
01v
If ),(v yx!
-
!
-
!
-
-
01
00
10
01 xxx
For a vector on thex-axis Eigenvalue 11!P
Sol:
16- 5
8/3/2019 Eigen,Diag,Sumetrix
6/50
For a vector on they-axis
-
!
-
!
-
-
yyy
01
00
10
01
Eigenvalue 12!P
Geometrically, multiplying a vector (x, y)in R2by the matrixA corresponds to a
reflection in they-axis.
The eigenspace corresponding to is thex-axis.
The eigenspace corresponding to is they-axis.
11 !P
12 !P
16- 6
8/3/2019 Eigen,Diag,Sumetrix
7/50
Thm 7.2: (Finding eigenvalues and eigenvectors of a matrixAMnvn )
0)Idet( !AP(1) An eigenvalue ofA is a scalarP such that .
(2) The eigenvectors ofA corresponding to P are the nonzero
solutions of .
Characteristic polynomial ofAMnvn:
01
1
1)I()Idet( cccAAn
n
n!! PPPPP .
Characteristic equation ofA:
0)Idet( !AP
0)I( ! xAP
LetA is an nvn matrix.
If has nonzero solutions iff .0)I( ! xAP 0)Idet( !AP
0)I( !! xAxAx PP Note:
(homogeneous system)
16- 7
8/3/2019 Eigen,Diag,Sumetrix
8/50
Ex 4: (Finding eigenvalues and eigenvectors)
-
!
51
122A
Sol: Characteristic equation:
0)2)(1(23
51
122)I(
2 !!!
!
PPPP
P
PP A
Eigenvalue: 2,1 21 !! PP
2,1 ! P
16- 8
8/3/2019 Eigen,Diag,Sumetrix
9/50
2)2( 2 !P
0,1
33
0
0
31
124)I(
2
1
2
1
2
{
-
!
-
!
-
-
!
-
-
!
ttt
t
x
x
x
xxAP
1)1( 1 !P
0,1
44
0
0
41
123
)I(
2
1
2
1
1
{
-
!
-
!
-
-
!
-
-
!
ttt
t
x
x
x
x
xAP
16- 9
8/3/2019 Eigen,Diag,Sumetrix
10/50
-
!200
020012
A
Sol: Characteristic equation:
0)2(
200
020012
I 3 !!
! PP
PP
P A
Eigenvalue: 2!P
Ex 5: (Finding eigenvalues and eigenvectors)
Find the eigenvalues and corresponding eigenvectors for
the matrixA. What is the dimension of the eigenspace of
each eigenvalue?
16- 10
8/3/2019 Eigen,Diag,Sumetrix
11/50
The eigenspace of A corresponding to :2!P
-
!
-
-
!
0
0
0
000
000
010
)I(
3
2
1
x
x
x
xAP
0,,
1
00
0
01
0
3
2
1
{
-
-
!
-
!
-
tsts
t
s
x
xx
2toingcorrespondAofeigenspacethe:,
1
0
0
0
0
1
!
-
-
PRtsts
Thus, the dimension of its eigenspace is 2.
16- 11
8/3/2019 Eigen,Diag,Sumetrix
12/50
Notes:
(1) If an eigenvalue P1occurs as a multiple root (ktimes)for
the characteristic polynominal, then P1has multiplicity k.
(2) The multiplicity of an eigenvalue is greater than or equal
to the dimension of its eigenspace.
16- 12
8/3/2019 Eigen,Diag,Sumetrix
13/50
Ex 6Find the eigenvalues of the matrixA and find a basis
for each of the corresponding eigenspaces.
-
!
3001
0201
10510
0001
A
Sol: Characteristic equation:
0)3)(2()1(
3001
0201
10510
0001
I
2 !!
!
PPP
PP
PP
P A
Eigenvalue:3,2,1
321
!!! PPP16- 13
8/3/2019 Eigen,Diag,Sumetrix
14/50
1)1( 1 !P
-
!
-
-
!
0
0
0
0
2001
0101
10500
0000
)I(
4
3
2
1
1
x
x
x
x
xAP
0,,
1
2
02
0
0
10
2
2
4
3
2
1
{
-
-
!
-
!
-
tsts
t
t
st
x
x
xx
1
2
0
2
,
0
0
1
0
-
-
is a basis for the eigenspaceof A corresponding to 1!P
16- 14
8/3/2019 Eigen,Diag,Sumetrix
15/50
2)2( 2 !P
-
!
-
-
!
0
0
0
0
1001
0001
10510
0001
)I(
4
3
2
1
2
x
x
x
x
xAP
0,
0
1
5
0
0
5
0
4
3
2
1
{
-
!
-
!
-
ttt
t
x
x
x
x
0
1
5
0
-
is a basis for the eigenspaceof A corresponding to 2!P
16- 15
8/3/2019 Eigen,Diag,Sumetrix
16/50
3)3( 3 !P
-
!
-
-
!
0
0
0
0
0001
0101
10520
0002
)I(
4
3
2
1
3
x
x
x
x
xAP
0,
1
0
5
0
0
5
0
4
3
2
1
{
-
!
-
!
-
tt
t
t
x
x
x
x
1
0
5
0
-
is a basis for the eigenspace
of A corresponding to 3!P
16- 16
8/3/2019 Eigen,Diag,Sumetrix
17/50
Thm 7.3: (Eigenvalues for triangular matrices)
IfA is an nvn triangular matrix, then its eigenvalues arethe entries on its main diagonal.
Ex 7: (Finding eigenvalues for diagonal and triangular matrices)
-
!
335
011002
)( Aa
-
!
30000040000000000020
00001
)( Ab
Sol:
)3)(1)(2(
335
011
002
I)( !
! PPPP
P
P
P Aa
3,1,2 321 !!! PPP
3,4,0,2,1)( 54321 !!!!! PPPPPb
16- 17
8/3/2019 Eigen,Diag,Sumetrix
18/50
Eigenvalues and eigenvectors of linear transformations:
.ofeigenspacethecalled
isvector)zero(with theofrseigenvectoallsetoftheand
,toingcorrespondofreigenvectoancalledisvectorThe
.)(such thatvectornonzeroaisthereif:
nnsformatiolinear traaofeigenvalueancalledisnumberA
P
P
P
P
P
T
TVVT
x
xxx !p
16- 18
8/3/2019 Eigen,Diag,Sumetrix
19/50
Keywords in Section 7.1:
eigenvalue problem:
eigenvalue:
eigenvector:
characteristic polynomial:
characteristic equation:
eigenspace:
multiplicity:
16- 19
8/3/2019 Eigen,Diag,Sumetrix
20/50
16- 20
Lecture 16: Eigenvalues and Eigenvectors
Today
Eigenvalues and Eigenvectors
Diagonalization
Symmetric Matrices and Orthogonal Diagonalization
8/3/2019 Eigen,Diag,Sumetrix
21/50
7.2 Diagonalization
Diagonalization problem:
For a square matrixA, does there exist an invertible matrixP
such thatP-1APis diagonal?
Diagonalizable matrix:
A square matrixA is called diagonalizable if there exists an
invertible matrixPsuch thatP-1APis a diagonal matrix.
(PdiagonalizesA) Notes:
(1) If there exists an invertible matrixPsuch that ,then two square matricesA andB are called similar.
(2) The eigenvalue problem is related closely to the
diagonalization problem.
APPB 1
!
16- 21
8/3/2019 Eigen,Diag,Sumetrix
22/50
Thm 7.4: (Similar matrices have the same eigenvalues)
IfA andB are similarnvn matrices, then they have the
same eigenvalues.
Pf:
APPBBA1
similarareand
!
A
APPAPPPAP
PAPAPPPPAPPB
!
!!!
!!!
I
III
)I(III
111
1111
P
PPP
PPPP
ThusA andB have the same eigenvalues.
16- 22
8/3/2019 Eigen,Diag,Sumetrix
23/50
Ex 1: (A diagonalizable matrix)
-
!
200
013
031
A
Sol: Characteristic equation:
0)2)(4(
200
013
031
I 2 !!
! PPP
PP
P A
2,2,4seigenvalueThe 321 !!! PPP
reigenvectothe4)1( !P
-
!
0
1
1
1p
16- 23
8/3/2019 Eigen,Diag,Sumetrix
24/50
reigenvectothe2)2( !P
-
!
-
!10
0
,01
1
32 pp
-
!
-
!!
200
020
004
such that
100
011
011
][
1
321
APP
pppP
-
!
-
!
!
200
040
002
100
011
011
][
1
312
APP
pppP Note: If
16- 24
8/3/2019 Eigen,Diag,Sumetrix
25/50
Thm 7.5: (Condition for diagonalization)
An nvn matrixA is diagonalizable if and only if it has n
linearly independent eigenvectors.
Pf:
ablediagonalizis)( A
),,,(and][Let
diagonaliss.t.invertibleanexiststhere
2121
1
nn diagDpppP
APPDP
PPP .. !!
!
][
00
00
00
][
2211
2
1
21
nn
n
n
ppp
pppPD
PPP
P
P
P
.
.
/1//
.
.
.
!
-
!
16- 25
8/3/2019 Eigen,Diag,Sumetrix
26/50
PDAP
ApApApAP n
!@
! ][ 21 .3
)ofrseigenvectoareoftorcolumn vecthe..(
,,2,1,
APpei
nipAp
i
iii -!! P
t.independenlinearlyare,,,invertibleis21 n
pppP .3
rs.eigenvectotindependenlinearlyhas nA@
n
npppnA
PPP .
.
,,seigenvalueingcorrespondhwit
,,rseigenvectotindependenlinearlyhas)(
21
21
nipAp iii ,,2,1, -!! P
][Let 21 npppP .!
16- 26
8/3/2019 Eigen,Diag,Sumetrix
27/50
PDppp
pppApApAppppAAP
n
n
nn
nn
!
-
!
!!!
P
PP
PPP
./1//
.
.
.
.
..
00
00
00
][
][][][
2
1
21
2211
2121
ablediagonalizis
invertibleistindependenlinearlyare,,,
1
11
A
DAPP
Pppp n
!@
.3
16- 27
8/3/2019 Eigen,Diag,Sumetrix
28/50
Ex 4: (A matrix that is not diagonalizable)
-
!
10
21
able.diagonaliznotismatrixfollowingthat theShow
A
Sol: Characteristic equation:
0)1(10
21I 2 !!
! PP
PP A
1eigenvalueThe 1 !P
-
!
-
-
!!0
1reigenvecto
00
10~
00
20I 1pAIAP
A does not have two linearly independent eigenvectors,
soA is not diagonalizable.
16- 28
8/3/2019 Eigen,Diag,Sumetrix
29/50
Steps for diagonalizing an nvn square matrix:
Step 2: Let ][ 21 npppP .!
Step 1: Find n linearly independent eigenvectors
for A with corresponding eigenvalues.nppp .,, 21
Step 3:
-
!!
n
DAPP
P
PP
.
/1//
.
.
00
00
00
2
1
1
nipAp iii ,,2,1,where, -!! P
16- 29
8/3/2019 Eigen,Diag,Sumetrix
30/50
Ex 5: (Diagonalizing a matrix)
diagonal.issuch thatmatrixaFind
113
131
111
1APPP
A
-
!
Sol: Characteristic equation:
0)3)(2)(2(
113
131
111
I !!
! PPP
P
PP
P A
3,2,2seigenvalueThe 321 !!! PPP
16- 30
8/3/2019 Eigen,Diag,Sumetrix
31/50
21 !P
-
-
!
000
010
101
~
313
111
111
I1 AP
-
!
-
!
-
1
0
1
reigenvecto0 1
3
2
1
p
t
t
x
x
x
22 !P
-
-
!
000
10
01
~
113
151
113
I41
41
2 AP
-
!
-
!
-
4
1
1
reigenvecto 241
41
3
2
1
p
t
t
t
x
x
x
16- 31
8/3/2019 Eigen,Diag,Sumetrix
32/50
33 !P
-
-
!
000
110
101
~
413
101
112
I3 AP
-
!
-
!
-
1
1
1
reigenvecto 3
3
2
1
p
t
t
t
x
x
x
-
!
-
!!
300
020
002
s.t.
141
110
111
][
1
321
APP
pppP
16- 32
8/3/2019 Eigen,Diag,Sumetrix
33/50
Notes:
-
!
-
!
k
n
k
k
k
n d
d
d
D
d
d
d
D
.
/1//
.
.
.
/1//
.
.
00
00
00
00
00
00
)1( 2
1
2
1
1
11
)2(
!
!!
PPDA
PAPDAPPD
kk
kk
16- 33
8/3/2019 Eigen,Diag,Sumetrix
34/50
Thm 7.6: (Sufficient conditions for diagonalization)
If an nvn matrixA has n distinct eigenvalues, then the
corresponding eigenvectors are linearly independent and
A is diagonalizable.
Ex 7: (Determining whether a matrix is diagonalizable)
-
!
300
100
121
A
Sol: BecauseA is a triangular matrix, its eigenvalues are
3,0,1 321 !!! PPP
These three values are distinct, so A is diagonalizable.
16- 34
8/3/2019 Eigen,Diag,Sumetrix
35/50
Ex 8: (Finding a diagonalizing matrix for a linear transformation)
diagonal.istorelative
formatrixthesuch thatforbasisaFind
)33()(
bygivennnsformatiolinear trathebeLet
3
321321321321
33
B
TRB
xxx,xx, xxxx,x,xxT
RT:R
!
p
Sol:
-
!
113
131
111
bygivenisformatrixstandardThe
A
T
From Ex.5 you know thatA is diagonalizable.Thus, the three
linearly independent eigenvectors found in Ex.5 can be used
to form the basisB.That is16- 35
8/3/2019 Eigen,Diag,Sumetrix
36/50
)}1,1,1(),4,1,1(),1,0,1{( !B
? A
? A
-
!
!
!
300
020
002][][][
][][][
321
321
BBB
BBB
AvAvAv
vTvTvTD
The matrix forTrelative to this basis is
16- 36
8/3/2019 Eigen,Diag,Sumetrix
37/50
Keywords in Section 7.2:
diagonalization problem:
diagonalization:
diagonalizable matrix:
16- 37
8/3/2019 Eigen,Diag,Sumetrix
38/50
Lecture 16: Eigenvalues and Eigenvectors
Today
Eigenvalues and Eigenvectors
Diagonalization
Symmetric Matrices and Orthogonal Diagonalization
16- 38
8/3/2019 Eigen,Diag,Sumetrix
39/50
7.3 Symmetric Matrices and Orthogonal Diagonalization
Symmetric matrix:
A square matrixA is symmetric if it is equal to its transpose:
TAA !
Ex 1: (Symmetric matrices and nonsymetric
matrices)
-
!
502
031210
A
-
!
13
34B
-
!
501
041
123
C
(symmetric)
(symmetric)
(nonsymmetric)
16- 39
8/3/2019 Eigen,Diag,Sumetrix
40/50
Thm 7.7: (Eigenvalues of symmetric matrices)
IfA is an nvn symmetric matrix, then the following properties
are true.
(1)A is diagonalizable.
(2) All eigenvalues ofA are real.
(3) IfP is an eigenvalue ofA with multiplicity k, then P has k
linearly independent eigenvectors.That is, the eigenspace
ofP has dimension k.
16- 40
8/3/2019 Eigen,Diag,Sumetrix
41/50
Ex 2:
Prove that a symmetric matrix is diagonalizable.
-
!
bc
caA
Pf: Characteristic equation:
0)( 22 !!
! cabba
bc
caAI PP
PP
P
0u22
222
22222
4)(
42
442)(4)(
cba
cbaba
cabbabacabba
!
!
!
As a quadratic in P, this polynomial has a discriminant of
16- 41
8/3/2019 Eigen,Diag,Sumetrix
42/50
04)((1) 22 ! cba
0, !! cba
diagonal.ofmatrixais0
0
-
!
a
aA
04)()2( 22 " cba
The characteristic polynomial ofA has two distinct real roots,
which implies thatA has two distinct real eigenvalues.Thus,
A is diagonalizable.
16- 42
8/3/2019 Eigen,Diag,Sumetrix
43/50
A square matrix P is called orthogonal if it is invertible and
Orthogonal matrix:
TPP !1
Thm 7.9: (Properties of symmetric matrices)
LetAbe an nvn symmetric matrix. IfP1and P2 are distinct
eigenvalues ofA, then their corresponding eigenvectorsx1
andx2 are orthogonal.
Thm 7.8: (Properties of orthogonal matrices)
An nvn matrixPis orthogonal if and only if its column vectors
form an orthogonal set.
16- 43
8/3/2019 Eigen,Diag,Sumetrix
44/50
Ex 5: (An orthogonal matrix)
-
!
53
5
53
4
53
2
5
1
5
2
32
32
31
0P
Sol: IfPis a orthogonal matrix, then I1 !!TT PPPP
I
100
010
001
0
0
53
5
3
253
4
5
132
53
2
5
231
53
5
53
4
53
25
1
5
232
32
31
!
-
!
-
-
!
TPP
-
!
-
!
-
!
535
32
3
534
51
32
2
532
52
31
1 0,,Let ppp
16- 44
8/3/2019 Eigen,Diag,Sumetrix
45/50
1
0
roducesp
321
323121
!!!
!!!
ppp
pppppp
set.lorthonormaanis},,{ 321 ppp
16- 45
8/3/2019 Eigen,Diag,Sumetrix
46/50
Thm 7.10: (Fundamental theorem of symmetric matrices)
LetAbe an nvn matrix.ThenA is orthogonally diagonalizable
and has real eigenvalue if and only ifA is symmetric.
Orthogonal diagonalization of a symmetric matrix:
LetAbe an nvn symmetric matrix.(1) Find all eigenvalues ofA and determine the multiplicity of each.
(2) For each eigenvalue of multiplicity 1, choose a unit eigenvector.
(3) For each eigenvalue of multiplicity ku2, find a set ofk linearly
independent eigenvectors. If this set is not orthonormal, apply Gram-
Schmidt orthonormalization process.
(4) The composite of steps 2 and 3 produces an orthonormal set ofn
eigenvectors. Use these eigenvectors to form the columns ofP.The
matrix will be diagonal.DAPPAPPT
!!1
16- 46
8/3/2019 Eigen,Diag,Sumetrix
47/50
Ex 7: (Determining whether a matrix is orthogonally diagonalizable)
-
!
111
101
111
1A
-
!
081
812
125
2A
-! 1020233A
-
!20
004A
Orthogonally
diagonalizable
Symmetric
matrix
16- 47
8/3/2019 Eigen,Diag,Sumetrix
48/50
Ex 9: (Orthogonal diagonalization)
-
!
142
412
222
A.esdiagonalizthatmatrixorthogonalanFind
A
P
Sol:
0)6()3()1( 2 !! PPP AI
)2oftymultipliciahas(3,6 21 !! PP
),,()2,2,1(,6)2( 32
3
2
3
1
1
1
111
!!!! v
v
uvP
)1,0,2(),0,1,2(,3)3( 322 !!! vvP
Linear Independent
16- 48
8/3/2019 Eigen,Diag,Sumetrix
49/50
Gram-Schmidt Process:
)1,,(),0,1,2(54
52
2
22
233322
!
!!! w
ww
wvvwvw
),,(),0,,(53
5
53
4
53
2
3
335
1
5
2
2
22
!!!!w
wu
w
wu
-
!
53
532
53
4
5
132
53
2
5
231
0
P
-
!
300
030
006
1APP
16- 49
8/3/2019 Eigen,Diag,Sumetrix
50/50
Keywords in Section 7.3:
symmetric matrix:
orthogonal matrix:
orthonormal set:
orthogonal diagonalization:
16 50