實變函數論─應用數學系 吳培元老師
1
Matrix Analysis
第七週課程講義
Nonderogatory matrix:
Thm. , Then the following are equiv.:
(1) ( ) exactly one Jordan block in Jordan form of ;
(2) ( ) for some poly. ;
(3) minimal poly. of =ch
T n n
T T
T M p p
T
2 -1
arac. poly. of ;
(4) dim Alg ;
(5) dim ;
(6) dim ker ( - ) 1 ;
(7) V , , ,..., ( is cycliC);
(8) = ;
(9) is abelian;
(10) Every matr
n n n
T
T n
T n
T I
x x Tx T x T x T
T T
T
11
21
, -1
-1
ix commuting with is a polynomial in T.
*
(11) , with 0 .
0
j j
nn nn
T
a
a
T a j
a a
( ) ( )
1
( )
1
...
( ) ( )
2
Def: is nonderogatory if it satisfies one of the above.
Pf: (3) (1): Jordan form:
charac. poly = ( - )
min. poly = ( - ) (1) ... 0 (3)
i i
ni
i
i
k k
i
i
k i i
i n
i
T
x
x k k i
2
1
(3) (2): Rational form:
charac. poly = (3) ... 1 (2)
min. poly =
(3) (4): (4) deg. of min. poly.= (3) ( min poly. | charac. poly).
(5) (1):
j n
j
p p p
p
n
( ) ( ) ( ) ( ) ( ) ( )
1 2 1 2 2 3
i i i i
(5) + 3 +... + 3 +... = =..=0 (1).i i i i i ik k n k k k k i
(1) (6):
no. of Jordan blocks at = dim ker ( - )= geom. multi of at ( )
(1) ( ), dimker ( - )=1
( ), dimker ( - )=0
n
n
T I T T
T T I
T T I
實變函數論─應用數學系 吳培元老師
2
0
1 -1
-1 1 0
-1
2
(2) (7):
0 -
1 -
( ) ( ) if ( ) ...
0
1 -
1
0
Let .
0
0
0
0
1
Then ( ) , ( ) ,.1
0
0
0
0
T n n
n
n
T T
a
a
T M p M p p x x a x a x a
a
x
M p x M p x
n-1
0
.., ( )
0
1
( ) is cyclic.
But "cyclic" preserved under similarity
cyclic.
T
T
M p x
M p
T
-1 (7) (2): , ,..., basis of .n nx Tx T x
0 0 *
1 0 *
represented , , this basis as 0 1
0 0
0 0 1 *
T w i t
0 * 0 1
1 *
( ) for some poly. .0 1
* 0 1
1
*
T M p p
實變函數論─應用數學系 吳培元老師
3
21
2 2
32 21
1 32 21
1
(11) (7):
1
0
Let .
0
*
**
*
, ...,0
*
0
0
V , , , .
nn
n n
x
a
Tx T x T x
a a
a a a
x Tx T x
( ) ( ) ( ) ( ) ( )
1 2 1 2 3
-1
1 2
(1) (8):
(8) 3 ... = =...=0 (1)
(7) (11): , ,..., indep.
Gram-Schmidt , ,..., o.n. basis of , where
i i i i i
i i i
n
n
n
k k k i k k i
x Tx T x
x x x
1 11 11
2 21 22 22
2
3 31 32 33 33
-1
1 2
, 0
, 0
, 0
... , 0
n
n n n nn nn
x b x b
x b x b Tx b
x b x b Tx b T x b
x b x b Tx b T x b
21
1 11 11 2 1
22 22 11
2
2 21 22
1 1 -
+
bTx b Tx b x x
b b b
Tx b Tx b T x
31 32
22 3
33 33 33
2 1 1
1 =...+ - -
,
b bb x x Tx
b b b
x x x
2 1 , x x
1 2 n
11
22
22
33
, ,...,
* *
*
= 0
0
0 0
T Tx Tx Tx
b
b
b
b
實變函數論─應用數學系 吳培元老師
4
(8) (9) :
Let ,
Then .
(9) (8) :
.
Let .
Then ,
(8)
(8) (10) :
( ) : poly. .
=
A B T
A T AB BA
T T
A T
B T AB BA
A T T T
T p T p
T T
2
1
Generalizations:
(1) , dim Alg( , ) . (M. Garstenhaber, 1961)
Note: Alg( , )=V , 0 , -1 dim Alg( , ) .
Sketch of pf of (1):
A ... with ( )
i j
m i
A B n n AB BA A B n
A B A B i j n A B n
A A A
1
1 1
( ) for .
... (by splitting of commutant).
Alg( , ) Alg( , ) ... Alg( , ).
0 1 0 1
May assume 1 ... 1
0 0
j
m
m m
A i j
B B B
A B A B A B
A
31 2
1 2 3
1
-1 -1-1 -1 2 2 2 -1 -1 -1
...
Check: V 1, ,..., , , ,..., , , ,..., ,..., , ,...,
=Alg
l
l
l
k kk k l l l
k k k k
k k
A A B AB A B B AB A B B AB A B
1
( , )
dim Alg ( , ) +...+ .
Ref: J. Barria & P.R. Holmes, Vector bases for two commuting matrices,
Linear Multilinear Algebra, 27(1990),147-157.
Note1:
l
A B
A B k k n
A M
abelian subalegbran
2
1
dim . (I. Schur, 1905)
(dim 1 & attained)
4
Note2: ,..., , ,m i j j i
A n
hA
A A n n A A A A i j
1dim Alg ( ,..., ) .mA A n
實變函數論─應用數學系 吳培元老師
5
4 4
1 2 3 4
0 0 0
0 0 0
Ex. V . 0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1
, , ,
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
a b
c d
a I M
A A A A
1 4 Then Alg ( ,..., ) & dim 5 4.
A A A A
2
2
1 if even,
4 (2) , , rank( - ) 1 dim Alg( , )
1 ( -1) if odd.
4
Ref: R.M. Guralnick, A note on commuting pairs of matrices, Linear Multilinear
Algeb
n n n
A B n n AB BA A B
n n n
ra, 31(1992), 71-75.
Eigenvalues of Hermitian matrices.
Ref: R.A. Horn & C.R, Johnson, Matrix analysis, Sec. 4.2,4.3.
(1) Hermitian
Variational characterization of ( ), i.e., max., min. of functions change rate
A n n
A
1 n
= 0.
Let ... be eigenvalues of .
Note: Eigenvalues of Hermitian are real.
Pf: Let .
Then , = * , = ,
||
A
A
Ax x
Ax x A x x x Ax
||
, ,
=
real
x x x x x
實變函數論─應用數學系 吳培元老師
6
1
n
1 1 n
Thm1. (Rayleigh-Rite)
Hermitian
= min , : , 1
= max , : , 1
Pf: Let ,..., be unit eigenvectors corresponding to ... , resp.
Note:
n
n
n
A n n
Ax x x x
Ax x x x
y y
's may be chosen to be o.n.
Reason:
(i) Assume
, , ,
||
,
||
, = ,
j
i j
i j i i j i i j
i j
i j j j i j
y
Ay y y y y y
y Ay
y y y y
1 1
1
*
1
*
* *
*
1
*
, =0
(ii) If = , then , may be chosen to be o.n.
Let U= unitary.
Let 1& .
, , ,
n
i j i j
i j i j
n
T
n
n
y y
y y
y y
x y U x y y
Ax x AUy Uy U AUy y
y y
U AU A y y
y y
1 1
1
1 1
*
1
2
' '
1
2 2 2 2
1 1 1 1
' '
' , ,
May assume = .
min , : 1 = min ... : ... 1
n
n n
n n
n
j jj
n
n
n n n
y y
y y
Ax x y
y y
A
Ax x x x x x x
2 2 2 2
1 1 1
1
max ... : ... 1
Note. Hermitian ( ) ,
n n n n
n
x x x x
A n n W A x x
實變函數論─應用數學系 吳培元老師
7
1 -1
1
2
Thm2. = min , : 1 & ,...
= max , : 1 & ,... 1,..., -1, .
Pf:" ":
with 1
, , ,
j j
j n
n n
i i i
i j i j
n n n n
i i k k i i i k k i
i j k j i j k j
Ax x x x y y
Ax x x x y y j n n
x a y a
Ax x A a y a y a y a y a
,
2 2
,
" "
= , RHS
Similarly for 2nd equality.
Note: Not much use: need eigenvectors to deduce eigenvalues.
i k i k
i k
n n
i i j i j
i j i j
j j j
a y y
a a
Ay y
-1
-1
1
1 -1
,...,1
j+1 n,...,1
Thm3. (Courant-Fischer)
Hermitian, eigenvalues: ... .
Then = max min , : 1, ,...,
= min max , : 1, ,..., 1,...,
n
j
n
j
n
j jW W
W W
A n n
Ax x x x W W
Ax x x x W W j
dim - 1
.
Another expression:
= max min , : 1,
||
nj M
M n j
n
Ax x x x M
1 -1
dim
V ,...,
= min max , : 1,
||
n
j
M
M j
W W
Ax x x x M
j+1 n V ,...,W W