EE263 Autumn 2015 S. Boyd and S.

Lall

Jordan canonical form

I Jordan canonical form

I generalized modes

I Cayley-Hamilton theorem

1
Jordan canonical form
any matrix A ∈ Rn×n can be put in Jordan canonical form by a similarity transfor-
mation, i.e.  
J1
T −1 AT = J = 
 .. 
. 
Jq
where
λi 1
 
 .. 
 λi .  ∈ Cni ×ni

Ji = 
 .. 
 . 1 
λi
Pq
is called a Jordan block of size ni with eigenvalue λi (so n = i=1 ni )

I J is upper bidiagonal

I J diagonal is the special case of n Jordan blocks of size ni = 1

I Jordan form is unique (up to permutations of the blocks)

I can have multiple blocks with same eigenvalue

2
Jordan canonical form

note: JCF is a conceptual tool, never used in numerical computations!

X (s) = det(sI − A) = (s − λ1 )n1 · · · (s − λq )nq
hence distinct eigenvalues ⇒ ni = 1 ⇒ A diagonalizable
dim null(λI − A) is the number of Jordan blocks with eigenvalue λ
more generally, X
dim null(λI − A)k = min{k, ni }
λi =λ

so from dim null(λI − A)k for k = 1, 2, . . . we can determine the sizes of the Jordan
blocks associated with λ

3
Jordan canonical form

I factor out T and T −1 , λI − A = T (λI − J)T −1

I for, say, a block of size 3:

   
0 −1 0 0 0 1
2
λi I − Ji =  0 0 −1  (λi I − Ji ) =  0 0 0 (λi I − Ji )3 = 0
0 0 0 0 0 0

I for other blocks (say, size 3, for k ≥ 2)

(λi − λj )k −k(λi − λj )k−1 (k(k − 1)/2)(λi − λj )k−2

 
k
(λi I −Jj ) =  0 (λj − λi )k −k(λj − λi ) k−1 
0 0 (λj − λi )k

4
Generalized eigenvectors

suppose T −1 AT = J = diag(J1 , . . . , Jq )
express T as
T = [T1 T2 · · · Tq ]
n×ni
where Ti ∈ C are the columns of T associated with ith Jordan block Ji
we have ATi = Ti Ji
let Ti = [vi1 vi2 · · · vini ]
then we have:
Avi1 = λi vi1 ,
i.e., the first column of each Ti is an eigenvector associated with e.v. λi
for j = 2, . . . , ni ,
Avij = vi j−1 + λi vij

the vectors vi1 , . . . vini are sometimes called generalized eigenvectors

5
Jordan form LDS

consider LDS ẋ = Ax
by change of coordinates x = T x̃, can put into form x̃˙ = J x̃
system is decomposed into independent ‘Jordan block systems’ x̃˙ i = Ji x̃i

x̃ni x̃ni −1 x̃1

1/s 1/s 1/s

λ λ λ

Jordan blocks are sometimes called Jordan chains

(block diagram shows why)

6
Resolvent, exponential of Jordan block

resolvent of k × k Jordan block with eigenvalue λ:

−1
s−λ −1

 .. 
s−λ .
(sI − Jλ )−1 = 
 

 .. 
 . −1 
s−λ
(s − λ)−1 (s − λ)−2 ··· (s − λ)−k
 
 (s − λ)−1 ··· (s − λ)−k+1 
=
 
.. .. 
 . . 
(s − λ)−1
= (s − λ)−1 I + (s − λ)−2 F1 + · · · + (s − λ)−k Fk−1

where Fi is the matrix with ones on the ith upper diagonal

7
Resolvent, exponential of Jordan block

by inverse Laplace transform, exponential is:

 
etJλ = etλ I + tF1 + · · · + (tk−1 /(k − 1)!)Fk−1
1 t · · · tk−1 /(k − 1)!
 
 1 · · · tk−2 /(k − 2)! 
= etλ 
 
.. .. 
 . . 
1

Jordan blocks yield:

I repeated poles in resolvent

I terms of form tp etλ in etA

8
Generalized modes
consider ẋ = Ax, with
x(0) = a1 vi1 + · · · + ani vini = Ti a

then x(t) = T eJt x̃(0) = Ti eJi t a

I trajectory stays in span of generalized eigenvectors

I coefficients have form p(t)eλt , where p is polynomial

I such solutions are called generalized modes of the system

with general x(0) we can write

q
X
x(t) = etA x(0) = T etJ T −1 x(0) = Ti etJi (SiT x(0))
i=1

where
S1T
 
 . 
T −1 =  .. 
SqT

hence: all solutions of ẋ = Ax are linear combinations of (generalized) modes

9
Cayley-Hamilton theorem

if p(s) = a0 + a1 s + · · · + ak sk is a polynomial and A ∈ Rn×n , we define

p(A) = a0 I + a1 A + · · · + ak Ak

Cayley-Hamilton theorem: for any A ∈ Rn×n we have X (A) = 0, where X (s) =

det(sI − A)
 
1 2
example: with A = we have X (s) = s2 − 5s − 2, so
3 4

X (A) = A2 − 5A − 2I
   
7 10 1 2
= −5 − 2I
15 22 3 4
=0

10
Cayley-Hamilton theorem

corollary: for every p ∈ Z+ , we have

Ap ∈ span I, A, A2 , . . . , An−1


(and if A is invertible, also for p ∈ Z)

i.e., every power of A can be expressed as linear combination of I, A, . . . , An−1
proof: divide X (s) into sp to get sp = q(s)X (s) + r(s)
r = α0 + α1 s + · · · + αn−1 sn−1 is remainder polynomial
then

Ap = q(A)X (A) + r(A) = r(A) = α0 I + α1 A + · · · + αn−1 An−1

11
Cayley-Hamilton theorem

for p = −1: rewrite C-H theorem

X (A) = An + an−1 An−1 + · · · + a0 I = 0

as
I = A −(a1 /a0 )I − (a2 /a0 )A − · · · − (1/a0 )An−1

(A is invertible ⇔ a0 6= 0) so

A−1 = −(a1 /a0 )I − (a2 /a0 )A − · · · − (1/a0 )An−1

i.e., inverse is linear combination of Ak , k = 0, . . . , n − 1

for p = −2, −3, . . . , use induction:

Ap−1 = −(a1 /a0 )Ap − (a2 /a0 )Ap+1 − · · · − (1/a0 )Ap+n

if Ap , . . . , Ap+n are linear combinations of Ak , k = 0, . . . , n − 1, so is Ap−1

12
Proof of C-H theorem

first assume A is diagonalizable: T −1 AT = Λ

X (s) = (s − λ1 ) · · · (s − λn )

since
X (A) = X (T ΛT −1 ) = T X (Λ)T −1
it suffices to show X (Λ) = 0

X (Λ) = (Λ − λ1 I) · · · (Λ − λn I)
= diag(0, λ2 − λ1 , . . . , λn − λ1 ) · · · diag(λ1 − λn , . . . , λn−1 − λn , 0)
=0

13
Proof of C-H theorem

now let’s do general case: T −1 AT = J

X (s) = (s − λ1 )n1 · · · (s − λq )nq

suffices to show X (Ji ) = 0

n
··· i

0 1 0
n1 0
X (Ji ) = (Ji − λ1 I) · · ·  0 1 ···  n
 · · · (Ji − λq I) q = 0
..
.
| {z }
(Ji −λi I)ni

14