• Tidak ada hasil yang ditemukan

Directory UMM :Journals:Journal_of_mathematics:EJQTDE:

N/A
N/A
Protected

Academic year: 2017

Membagikan "Directory UMM :Journals:Journal_of_mathematics:EJQTDE:"

Copied!
21
0
0

Teks penuh

(1)

Electronic Journal of Qualitative Theory of Differential Equations

2007, No.

19, 1-21;

http://www.math.u-szeged.hu/ejqtde/

Null Controllability of Some Impulsive

Evolution Equation in a Hilbert Spae

R. BOUKHAMLA

(1)

& S. MAZOUZI

(2)

(1)CentreUniversitairedeSouk-Ahras,Souk-Ahras41000,Algeria.

(2)DépartementdeMathématiques,Universitéd'Annaba,BP12,ANNABA23000,Algeria.

Abstrat

Weshallestablishaneessaryandsuientonditionunderwhih

we have thenullontrollabilityofsome rstorderimpulsiveevolution

equation inaHilbertspae.

MSC(2000) : 34A37, 93B05, 93C15.

Keywords: Null-ontrollability, impulsive onditions, mild solutions,

evolution equation.

1 Introdution

The problem of exat ontrollability of linear systems represented by

in-nite onservative systemshas beenextensively studiedby several authorsA.

Haraux [8℄

,

R.Triggiani [16℄

,

Z.H. Guan, T.H. Qian, and X.Yu [7℄, see also

the referenes [1,2, 6, 10,15℄. In the sequel, we shall be onerned with the

problem of null ontrollability of some rst order evolution equation

sub-jet toimpulsiveonditions andso weshall derivea neessary and suient

ondition under whih null ontrollability ours. Atually, we shall

estab-lish anequivalene between the null-ontrollabilityand some observability

inequalityinsomehowmoregeneralframeworkthan thatproposed byA

Ha-raux [8℄. Regarding the literature onthe impulsive dierential equationswe

(2)

y

(

t

) +

Ay

(

t

) =

Bu

(

t

)

,

t

(0

, T

)

\ {

t

k

}k

σ

m

1

,

(1)

y

(0) =

y

0

,

y

(

t

k

) =

I

k

y

(

t

k

) +

D

k

v

k

, k

σ

1

m

,

(1

k

)

where the nal time

T

is a positive number,

y

0

is an initial ondition in a

Hilbertspae

H

endowed withaninnerprodut

h

., .

iH

,

y

(

t

) : [0

, T

]

H

isa vetor funtion,

σ

m

1

is a subset of

N

given by

σ

m

1

=

{

1

,

2

, ..., m

}

, and nally,

{

t

k

}k

σ

m

1

is an inreasing sequene of numbers in the open interval

(0

, T

)

,

and

y

(

t

k

)

denotes the jump of

y

(

t

)

at

t

=

t

k

, i.e.,

y

(

t

k

) =

y t

+

k

y t

k

where

y t

+

k

and

y t

k

represent the right and left limits of

y

(

t

)

at

t

=

t

k

respetively

.

On the other hand, the operators

A, B, I

k

, D

k

:

H

H

are

givenlinearboundedoperators. Moreover, weset thefollowingassumptions:

(H1)

A

=

A

,

(H2)

I

k

=

I

k

,

for every

k

σ

m

1

,

and for eah

k

σ

m

1

, the operator

Ik

=

I

k

+

I

is invertible

,

(H3)

B

=

B

0

and there is

d

0

>

0

suhthat

(

Bu, u

)

H

d

0

k

u

k

2

H

,

for all

u

H,

(H4)

D

k

=

D

k

0

,

for every

k

σ

m

1

,

and for eah

k

σ

m

1

there is

d

k

>

0

suh that

(

D

k

u, u

)

H

d

k

k

u

k

2

H

,

for all

u

H.

In the sequel we shall designate by h the funtion

h

(

t

) =

u

(

t

)

,

{

v

k

}k

σ

m

1

,

where

u

(

t

)

L

2

(0

, T

)

\ {

t

k}k

σ

m

1

;

H

and

{

v

k

}k

σ

m

1

l

2

(

σ

m

1

;

H

)

+

n

{

v

k

}k

σ

m

1

, v

k

H

(3)

We pointout that the spae

K

m

=

L

2

(0

, T

)

\ {

t

k

}k

σ

m

1

;

H

×

l

2

(

σ

1

m

;

H

)

is

a Hilbert spae with respet to the inner produt

h

,

e

h

K

m

=

Z

T

0

(

u

(

t

)

,

e

u

(

t

))

H

dt

+

m

X

k

=1

(

v

k

,

e

v

k

)

H

,

dened for allh

= (

u

(

t

)

,

{

v

k

}

m

k

=1

)

and

e

h

= (

u

e

(

t

)

,

{

e

v

k

}

m

k

=1

)

∈ K

m

.

We shall denoteby

B

the ontroloperator given by

B

=

B,

{

D

k

}k

σ

m

1

∈ L

L

2

(0

, T

)

\ {

t

k

}k

σ

m

1

;

H

×

l

2

(

σ

1

m

;

H

)

,

so that

B

h

(

t

) =

Bu

(

t

)

,

{

D

k

v

k

}k

σ

m

1

.

We have for every h

= (

u

(

t

)

,

{

v

k

}

m

k

=1

)

∈ K

m

(

B

h

,

h

)

K

m

=

Z

T

0

(

Bu

(

t

)

, u

(

t

))

H

dt

+

m

X

k

=1

(

D

k

v

k

, v

k

)

H

=

Z

T

0

(

u

(

t

)

, Bu

(

t

))

H

dt

+

m

X

k

=1

(

v

k

, D

k

v

k

)

H

= (

h

,

B

h

)

K

m

,

whih shows that

B

=

B

, that is,

B

is self-adjoint. On the other hand, we

have

(

B

h

,

h

)

K

m

=

Z

T

0

(

Bu

(

t

)

, u

(

t

))

H

dt

+

m

X

k

=1

(

D

k

v

k

, v

k

)

H

d

0

Z

T

0

k

u

(

t

)

k

2

H

dt

+

m

X

k

=1

d

k

k

v

k

k

2

H

δ

k

h

k

2

K

m

,

where

δ

= max

{

d

0

, d

1

, ..., d

m

}

.

Thus, the operator is

B

bounded in

K

m

.

Next,weonsider the homogeneoussystem assoiatedwith (1) :

ϕ

(

t

) +

(

t

) = 0

,

t

(0

, T

)

\ {

t

k

}k

σ

m

1

,

(2)

ϕ

(0) =

ϕ

0

,

(4)

We point out that on eah interval

[

t

k

, t

k

+1

)

,for

k

= 0

, ..., m

, the solution

ϕ

is left ontinuous ateahtime

t

k

.

Consider the orrespondinghomogeneous bakward problem :

ϕ

˜

(

t

) +

A

ϕ

˜

(

t

) = 0

, ,

t

(0

, T

)

\ {

t

k

}k

σ

m

1

,

(3)

˜

ϕ

(

T

) =

ϕ

0

,

∆ ˜

ϕ

(

t

m

−(

k

−1)

) =

I

˜

m

−(

k

−1)

ϕ

˜

(

t

+

m

−(

k

−1)

)

, k

σ

m

1

,

(3

k

)

where

A

=

A

=

A,

I

˜

m

−(

k

−1)

=

I

m

−(

k

−1)

=

I

m

−(

k

−1)

, k

σ

1

m

.

We observe that the problem (3) onthe interval

[

t

m

, T

]

is equivalent to the

lassial bakward problem

ϕ

˜

(

t

) +

A

ϕ

˜

(

t

) = 0

, t

[

t

m

, T

]

,

˜

ϕ

(

T

) =

ϕ

0

.

We introdue the following spae :

PC

([0

, T

] ;

H

) =

{

y, y

: [0

, T

]

H

suh that

y

(

t

)

isontinuous at

t

=

6

t

k

,

and has disontinuities of rst kind at

t

=

t

k

, for every

k

σ

m

1

}

.

Evidently,

PC

([0

, T

] ;

H

)

is a Banahspae with respet to the norm

k

y

k

PC

= sup

t

∈(0

,T

)

k

y

(

t

)

k

.

On the other hand, we dene the subspaes

PLC

, (respetively,

PRC

)

=

{

y, y

∈ PC

suh that

y

(

t

)

is left (respetively, right) ontinuous at

t

=

t

k

,

for every

k

σ

m

1

}

.

Remark 1 1) The spae

PLC

, (respetively,

PRC

) an be identied to a

subspae of

K

m

.

That is, to eah

y

∈ PLC

, (respetively,

y

˜

∈ PRC

) is

assigned the funtion h (respetively,

e

h) dened by

h

(

t

) =

y

(

t

)

,

{

y

(

t

k

)

}k

σ

m

1

,

and

e

h

(

t

) =

˜

y

(

t

)

,

{

y

˜

(

t

k

)

}k

σ

m

1

.

(5)

2) Let

y

e

∈ PRC

, the funtion

y

an be written as :

e

y

(

t

) =

e

y

[0]

(

t

)

if

t

[

t

0

, t

1

)

e

y

[

k

]

(

t

)

if

t

[

t

k

, t

k

+1

)

e

y

[

m

]

(

t

)

if

t

[

t

m

, T

]

.

Next, let

τ

k

=

t

k

t

k

−1

,

we dene the operator

T

:

D

(

T

) =

PRC ⊂ K

m

K

m

by

(

T

y

e

)(

t

) =

e

y

[0]

((

T

t

)

τ

m

τ1

+1

+

t

0

)

if

t

[

t

m

, T

]

,

e

y

[

k

]

((

t

m

−(

k

−1)

t

)

τ

k

+1

τ

m

(

k

1)

+

t

k

)

if

t

t

m

k

, t

m

−(

k

−1)

, k

σ

1

m

−1

,

e

y

[

m

]

((

t

1

t

)

τ

m

+1

τ1

+

t

m

)

if

t

(0

, t

1

]

.

(4)

We note that the range of

T

is exatly

PLC

. The funtion

(

T

e

y

)(

t

)

an be

written asfollows:

(

T

e

y

)(

t

) =

y

[0]

(

t

)

if

t

[

t

0

, t

1

]

,

y

[

k

]

(

t

)

if

t

(

t

k

, t

k

+1

]

, k

σ

m

−1

1

,

y

[

m

]

(

t

)

if

t

(

t

m

, T

]

.

Let

X

(

t

)

be the resolvent solutionof the operator system

X

(

t

) +

AX

(

t

) = 0

,

0 =

t

0

< t < t

m

+1

=

T, t

6

=

t

k

, k

= 1

,

2

, ..., m,

X

(0) =

I,

X

(

t

k

+ 0)

X

(

t

k

0) =

I

k

X

(

t

k

)

, k

= 1

,

2

, ..., m,

where

I

:

H

H

istheidentityoperator. Weshallsupposethattheoperator

I

k

=

I

k

+

I

has abounded inverse.

Denition 1 A funtion

y

∈ PC

([0

, T

] ;

H

)

is a mild solution to the

impul-sive problem (1) if the impulsive onditions are satised and

y

(

t

) =

G

(

t,

0

+

)

y

0

+

R

t

0

G

(

t, s

)

Bu

(

s

)

ds

+

P

0

<t

k≤

t

G

(

t, t

k

)(

D

k

v

k

)

,

for every

t

(0

, T

)

,

where the evolution operator

G

(

t, s

)

isgiven by
(6)

Itisnothardtohek thattheoperator

G

(

t, t

k

)

satisestheoperatorsystem

G

(

t, t

k

) +

AG

(

t, t

k

) = 0

, t

[

t

k

, t

k

+1

)

, k

σ

0

m

,

G

(

t

k

, t

k

) =

I,

G

(

t

+

k

+1

, t

k

)

G

(

t

k

+1

, t

k

) =

I

k

+1

G

(

t

k

+1

, t

k

)

.

It iswell known that (1)has a unique solution

y

suh that

y

∈ PLC

([0

, T

] ;

H

)

C

1

[0

, T

]

\ {

t

k

}k

σ

m

1

;

H

.

Now, we dene the onept of mild solution for the bakward impulsive

system (3) assoiatedwith system (2).

Denition 2 We say that

ϕ

˜

∈ PRC

([0

, T

] ;

H

)

is a mild solution for the bakward impulsive system (3) if

T

ϕ

˜

is a mild solution for the homogeneous

impulsive system (2).

Letus introdue the notion of the null ontrollability of the initialstate

as follows:

Denition 3 Wesaythattheinitialstate

y

0

H

isnullontrollable attime

T

, if there is a ontrol funtion h

∈ K

m

for whih the solution

y

of system

(1) satises

y

(

T

) = 0

.

2 Main Results

First we begin by the following lemma.

Lemma 1 Assume that

ξ

(

t

)

, ζ

(

t

)

L

1

([0

, T

] ;

H

)

and

{

ξ

k

}

m

k

=1

,

{

ζ

k

}

m

k

=1

l

1

(

σ

m

1

, H

)

. Then, for every vetor funtions

γ

(

t

)

∈ PLC

([0

, T

] ;

H

)

C

1

[0

, T

]

\ {

t

k

}k

σ

m

1

;

H

and

η

(

t

)

∈ PRC

([0

, T

] ;

H

)

C

1

[0

, T

]

\ {

t

k}k

σ

m

1

;

H

satisfying the problem

d

dt

h

γ

(

t

)

, η

(

t

)

i

=

h

ξ

(

t

)

, ζ

(

t

)

i

,

t

6

=

t

k

,

for

k

σ

m

1

,

(7)

h

γ

(

t

)

, η

(

t

)

i|

T

0

=

h

γ

(

T

)

, η

(

T

)

i − h

γ

(0)

, η

(0)

i

(5)

=

Z

T

0

h

ξ

(

t

)

, ζ

(

t

)

i

dt

+

m

X

k

=1

h

ξ

k

, ζ

ki

.

Proof. It is straightforward.

We alsoneed the followingLemmas.

Lemma 2

[14]

If

B ∈ L

(

Km

)

is self-adjoint and nonnegative, then

kB

h

k ≤ kBk

1

/

2

(

B

h

,

h

)

1

/

2

K

m

,

h

∈ K

m

.

Lemma 3 If

τ

k

+1

=

τ

m

−(

k

−1)

, k

σ

m

−1

0

, then forthe mildsolution

ϕ

e

of (3),

the identity holds:

Z

T

0

|

B

ϕ

e

|

2

H

dt

+

m

X

k

=1

D

k

ϕ t

e

+

k

2

H

=

Z

T

0

|

|

2

H

dt

+

m

X

k

=1

D

k

ϕ t

m

−(

k

−1)

2

H

.

(6)

Proof. For eah

k

σ

m

0

,

using the hange of variable

t

(

t

m

−(

k

−1)

t

)

τ

k

+1

τ

m

(

k

1)

+

t

k

we have

Z

t

m

(

k

1)

t

m

k

(

[

m

k

]

(

t

)

, Bϕ

[

m

k

]

(

t

))

dt

=

Z

t

m

(

k

1)

t

m

k

(

B

ϕ

e

[

k

]

((

t

m

−(

k

−1)

t

)

τ

k

+1

τ

m

−(

k

−1)

+

t

k

)

, B

ϕ

e

[

k

]

((

t

m

−(

k

−1)

t

)

τ

k

+1

τ

m

−(

k

−1)

+

t

k

))

dt

=

τ

m

−(

k

−1)

τ

k

+1

Z

t

k

t

k

+1

(

B

ϕ

e

[

k

]

(

s

)

, B

ϕ

e

[

k

]

(

s

))

ds

=

Z

t

k

+1

t

k

(

B

ϕ

e

[

k

]

(

s

)

, B

ϕ

e

[

k

]

(

s

))

ds.

Summingup with respet to

k

, weget

m

X

k

=0

Z

t

m

(

k

1)

t

m

k

(

[

m

k

]

((

t

))

, Bϕ

[

m

k

]

(

t

))

dt

=

m

X

k

=0

Z

t

k

+1

t

k

(8)

Thus, we obtain

Z

T

0

|

B

ϕ

e

|

2

H

dt

=

Z

T

0

|

|

2

H

dt.

On the other hand,by virtue of the denition of the funtion

ϕ

e

weget

ϕ

(

t

m

k

) =

ϕ

e

(

t

k

+1

)

, k

σ

m

0

−1

.

Also, we have

ϕ t

m

−(

k

−1)

=

ϕ

e

(

t

k

)

, k

σ

1

m

,

and

e

ϕ

(

t

m

k

) =

ϕ

(

t

k

+1

)

, k

σ

m

0

−1

.

This implies that

m

X

k

=1

|

D

k

ϕ

e

(

t

k

)

|

2

H

=

m

X

−1

k

=0

h

D

m

k

ϕ

e

(

t

m

k

)

, D

m

k

ϕ

e

(

t

m

k

)

iH

=

m

X

−1

k

=0

h

D

m

k

ϕ

(

t

k

+1

)

, D

m

k

ϕ

(

t

k

+1

)

iH

=

m

X

l

=1

D

l

ϕ t

m

−(

l

−1)

, D

l

ϕ t

m

−(

l

−1)

H

=

m

X

k

=1

D

k

ϕ t

m

−(

k

−1)

, D

k

ϕ t

m

−(

k

−1)

H

=

m

X

k

=1

D

k

ϕ t

m

−(

k

−1)

2

H

,

whih gives (6).

Corollary 1 If

τ

k

+1

=

τ

m

−(

k

−1)

,

for

k

σ

m

−1

0

, and

B, D

k

are nonnegative

in

H,

then the followingholds:

Z

T

0

h

B

ϕ

e

(

t

)

,

ϕ

e

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

k

ϕ

e

(

t

k

)

,

ϕ

e

(

t

k

)

i

=

Z

T

0

h

(

t

)

, ϕ

(

t

)

i

dt

+

k

X

=

m

k

=1

(9)

Proof. This follows immediatelyfromLemma 3 if we substitute

B

by

B

1

2

,

and

D

k

by

D

1

2

k

.

Now, we state and establish the followingTheorem.

Theorem 1 Let

y

0

H

bea giveninitial stateforthe system(1)

,

then

y

0

is

null ontrollable at time

T

if and only if there is a positive onstant

C

suh

that

h

y

0

,

ϕ

˜

0

i

H

C

(Z

T

0

|

|

2

H

dt

+

m

X

k

=1

D

k

ϕ t

m

−(

k

−1)

2

H

)

1

/

2

,

ϕ

˜

0

H,

(7)

where

ϕ

∈ PLC

([0

, T

] ;

H

)

istheuniquemild solutionto

(2)

with

ϕ

(

T

) = ˜

ϕ

0

.

Proof. Itsues toprovethisTheoremforthespeialase

τ

k

+1

=

τ

m

−(

k

−1)

,

for

k

σ

m

−1

0

,

beause the norm

k|

.

|k

+

P

m

k

=0

τ

m

−(

k

−1)

τ

k

+1

R

t

k

+1

t

k

|

.

|

2

H

dt

1

/

2

is

equivalent tothe usual norm of

L

2

([0

, T

] ;

H

)

.

We shall proeed inseveral steps.

Step 1: Let

y

and

ϕ

e

be strong solutions to

(1)

and

(3)

, respetively.

Then, for

t

6

=

t

k

,

k

σ

m

1

, wehave

d

dt

h

y

(

t

)

,

ϕ

e

(

t

)

i

=

h

y

(

t

)

,

ϕ

e

(

t

)

i

+

h

y

(

t

)

,

ϕ

e

(

t

)

i

(8)

=

h

y

(

t

)

,

A

ϕ

e

(

t

)

i

+

h−

Ay

(

t

) +

Bu

(

t

)

,

ϕ

e

(

t

)

i

=

h

y

(

t

)

,

A

ϕ

e

(

t

)

i

+

h−

Ay

(

t

)

,

ϕ

e

(

t

)

i

+

h

Bu

(

t

)

,

ϕ

e

(

t

)

i

=

h

Bu

(

t

)

,

ϕ

e

(

t

)

i

.

Multiplying equation(3

k

) in(3) fromthe leftby

y t

m

−(

k

−1)

the solutionof

(1)

,andmultiplyingequation(1

k

)in(1)fromtherightby

ϕ

e

(

t

k

)

the solution

of (3), and nallyaddingmemberwise we get

h

y

(

t

)

,

ϕ

e

(

t

)

i

|

t

=

t

k

=

h

y

(

t

k

)

,

ϕ

e

(

t

k

)

i

+

h

y

(

t

k

)

,

ϕ

e

(

t

k

)

i

(9)

=

h

y

(

t

k

)

, I

k

ϕ

e

(

t

k

)

i

+

h

I

k

y

(

t

k

) +

D

k

v

k

,

ϕ

e

(

t

k

)

i

=

h

y

(

t

k

)

, I

k

ϕ

e

(

t

k

)

i

+

h

I

k

y

(

t

k

)

,

ϕ

e

(

t

k

)

i

+

h

D

k

v

k

,

ϕ

e

(

t

k

)

i

=

h

D

k

v

k

,

ϕ

e

(

t

k

)

i

.

Setting

γ

(

t

) =

y

(

t

)

,

η

(

t

) =

ϕ

e

(

t

)

,

ξ

(

t

) =

Bu

(

t

)

,

ζ

(

t

) =

ϕ

e

(

t

)

,

ξ

k

=

D

k

v

k

,
(10)

h

y

(

T

)

,

ϕ

e

(

T

)

i − h

y

(0)

,

ϕ

e

(0)

i

=

Z

T

0

h

Bu

(

t

)

,

ϕ

e

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

k

v

k

,

ϕ

e

(

t

k

)

i

.

(10)

Sine

B

isbounded,self-adjointand

B ≥

0

,then bydensitythelatteridentity

isstillvalidformildsolutions

y

of(1). Identity(10)anbewrittenasfollows

h

y

(

T

)

,

ϕ

e

(

T

)

i − h

y

(0)

,

ϕ

e

(0)

i

=

Z

T

0

h

u

(

t

)

, B

ϕ

e

(

t

)

i

dt

+

k

X

=

m

k

=1

h

v

k

, D

k

ϕ

e

(

t

k

)

i

.

(11)

Next, if there is a ertainh

(

t

)

∈ Km

suh that the mild solutionof (1)with

y

(0) =

y

0

satises

y

(

T

) = 0

,

then

−h

y

(0)

,

ϕ

e

(0)

i

=

Z

T

0

h

u

(

t

)

, B

ϕ

e

(

t

)

i

dt

+

k

X

=

m

k

=1

h

v

k

, D

k

ϕ

e

(

t

k

)

i

,

and so by Cauhy-Shwarz Inequality we obtain

|h

y

(0)

,

ϕ

e

(0)

i

H

| ≤

(Z

T

0

k

u

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

v

k

k

2

H

)

1

/

2

(12)

×

(Z

T

0

k

B

ϕ

e

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

D

k

ϕ

e

((

t

k

)

k

2

H

)

1

/

2

.

Using Lemma3, and equation (12) we have

|h

y

(0)

,

ϕ

e

(0)

i

H

| ≤

(Z

T

0

k

u

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

v

k

k

2

H

)

1

/

2

×

(Z

T

0

k

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

D

k

ϕ

(

t

m

−(

k

−1)

)

2

H

)

1

/

2

.

Setting

C

=

k

h

(

t

)

k

K

m

=

(Z

T

0

k

u

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

v

k

k

2

H

(11)

|

(

h

y

(0)

,

ϕ

e

(0)

i

H

| ≤

C

(Z

T

0

k

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

D

k

ϕ

(

t

m

−(

k

−1)

)

2

H

)

1

/

2

.

This shows the neessary ondition of the Theorem.

Step 2: Toprove the suieny we need the following result when

B ≥

α >

0

.

Claim 1 Assume that there is

α >

0

suh that

(Z

T

0

k

Bu

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

D

k

v

k

k

2

H

)

α

(Z

T

0

k

u

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

v

k

k

2

H

)

then, for every

y

0

H

there is

ϕ

0

H

suh that the mild solution of (1)

with

h

(

t

) = (

ϕ

e

(

t

)

,

ϕ

e

(

t

1

)

, ..,

ϕ

e

(

t

k

)

..,

ϕ

e

(

t

m

))

∈ Km

and

y

(0) =

y

0

satises

y

(

T

) = 0

.

To prove this Claim, we onsider for every

z

H

the solution

ϕ

of (2)

satisfying

ϕ

(

T

) =

z

and the unique mildsolution

y

tothe problem

y

(

t

) +

Ay

(

t

) =

B

ϕ

e

(

t

)

, t

(0

, T

)

/

{

t

k

}k

σ

m

1

,

y

(

t

k

) =

I

k

y

(

t

k

) +

D

k

ϕ

e

(

t

k

)

,

y

(

T

) = 0

.

Next, we introdue a bounded linear operator

Λ :

H

H

dened by

Λ

z

=

y

(0)

.

Aording to formula(11) and the Corollary1 wehave

|h

Λ

z, z

i|

=

|−h

y

(0)

,

ϕ

e

(0)

i|

=

Z

T

0

h

B

ϕ

e

(

t

)

,

ϕ

e

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

k

ϕ

e

(

t

k

)

,

ϕ

e

(

t

k

)

i

=

Z

T

0

h

(

t

)

, ϕ

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

k

ϕ

(

t

m

−(

k

−1)

)

, ϕ

(

t

m

−(

k

−1)

)

i

ς

(Z

T

0

k

ϕ

(

t

)

k

2

dt

+

k

X

=

m

k

=1

k

ϕ

(

t

k

)

k

2

(12)

ς

= sup

k

σ

m

0

{

d

k

}

<

.

We have

Z

T

0

k

ϕ

(

t

)

k

2

dt

=

Z

t1

0

k

ϕ

(

t

)

k

2

dt

+

Z

t2

t1

k

ϕ

(

t

)

k

2

dt

+

...

+

Z

T

t

m

k

ϕ

(

t

)

k

2

dt.

Sine there is noimpulsein the interval

[

t

k

, t

k

+1

)

we have

k

ϕ

(

t

)

k

=

ϕ

(

t

+

k

)

,

for every

t

[

t

k

, t

k

+1

)

, k

σ

m

0

,

ϕ

(

t

k

+1

)

=

ϕ

(

t

+

k

−1

)

, k

σ

0

m

.

(13)

Therefore, there are

τ

k

+1

=

t

k

+1

t

k

>

0

,

k

σ

m

0

suh that

R

t

k

+1

t

k

k

ϕ

(

t

)

k

2

dt

ρ

k

ϕ

(

t

+

k

)

2

=

τ

k

+1

I

k

ϕ

(

t

k

) +

ϕ

(

t

k

)

2

, k

σ

m

1

.

(14)

On the other hand,the ontinuity of

I

k

impliesthat

ϕ

(

t

+

k

)

2

=

(

I

k

+

I

)

ϕ

(

t

k

)

2

(1 +

L

(

I

k

))

2

ϕ

(

t

k

)

2

, k

σ

m

1

.

(15)

It follows from(14) and (15) that

R

t

k

+1

t

k

k

ϕ

(

t

)

k

2

dt

τ

k

+1

(1 +

L

(

I

k

))

2

ϕ

(

t

k

)

2

, k

σ

1

m

.

(16)

Sine

m

isnite, and due to (13),(16), then there is a onstant

0

< µ <

suh that

h

Λ

z, z

i ≤

µ

k

z

k

2

,

and thus,

Λ

isbounded.

Now, as

B

is nonnegativein

K

m

, we have

kB

ξ

(

t

)

k ≥

α

{

(

ξ

(

t

)

, ξ

(

t

))

Km

}

1

/

2

for all

ξ

∈ K

m

; thus, by virtue of Lemma 2,we have

(Z

T

0

(

Bu

(

t

)

, u

(

t

))

H

dt

+

k

X

=

m

k

=1

(

D

k

v

k

, v

k

)

H

)

(17)

α

kBk

(Z

T

0

k

u

(

t

)

k

2

H

dt

+

k

X

=

m

k

=1

k

v

k

k

2

H

(13)

h

Λ

z, z

i

=

−h

y

(0)

,

ϕ

e

(0)

i

=

Z

T

0

h

(

t

)

, ϕ

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

k

ϕ

(

t

m

−(

k

−1)

)

, ϕ

(

t

m

−(

k

−1)

)

i

α

kBk

(Z

T

0

k

ϕ

(

t

)

k

2

dt

+

k

X

=

m

k

=1

k

ϕ

(

t

k

)

k

2

)

α

kBk

Z

t1

0

k

ϕ

(

t

)

k

2

dt

=

kBk

αt

1

k

z

k

2

=

θ

k

z

k

2

,

beausethere is noimpulse beforetime

t

1

. Therefore,

Λ

isoerive on

H

.

To show that there is a bijetion from

H

onto

H,

it sues to prove that

Λ +

I

is abijetion from

H

onto

H.

Clearly,

Λ +

I

is injetive sine

h

Λ

z

+

z, z

i

=

h

Λ

z, z

i

+

h

z, z

i ≥

(

θ

+ 1)

k

z

k

2

.

On the other hand,let

y

0

H

,asthe form

a

(

f, g

) +

h

f, g

i

=

h

Λ

f, g

i

+

h

f, g

i

is symmetri and oerive, then, by virtue of Lax-Milgram Theorem, there

is anelement

f

H

suh that

a

(

f, g

) +

h

f, g

i

=

h

y

0

, g

i

,for all

g

H.

This implies that

Λ(

H

) =

H

. Thus, for every

y

0

H

, there is a unique

z

H

suhthat

Λ(

z

) =

y

0

, whihompletes the proof of Claim 1.

Step 3: Assume that

B, D

k

0

, then

B ≥

0

,

e

B

2

=

B,

D

e

k

2

=

D

k

.

We denefor

ε >

0

,

β

ε

+

B

e

2

+

εI,

δ

ε

k

+

D

e

2

k

+

εI,

and

B

ε

+

(

β

ε

;

δ

1

ε

, .., δ

ε

m

) = (

B

e

2

+

εI

;

D

e

2

1

+

εI, ..,

D

e

2

m

+

εI

)

.

Aording toClaim1,there is

ϕ

˜

0

H

suhthat the mildsolution

y

ε

of (1)

with

y

ε

(0) =

y

0

satises

y

ε

(

T

) = 0;

where

B

(

h

)

has been replaed by
(14)

−h

y

(0)

,

ϕ

e

ε

(0)

i

=

Z

T

0

h

β

ε

ε

ϕ

e

(

t

)

,

ϕ

e

ε

(

t

)

i

dt

+

k

X

=

m

k

=1

h

δ

ε

k

ϕ

e

ε

(

t

k

)

,

ϕ

e

ε

(

t

k

)

i

,

(18)

and (7) gives

−h

y

(0)

,

ϕ

e

ε

(0)

i ≤

C

(Z

T

0

h

e

B

2

ϕ

ε

(

t

)

, ϕ

ε

(

t

)

i

dt

+

k

X

=

m

k

=1

h

D

e

k

2

ϕ

ε

(

t

m

−(

k

−1)

)

, ϕ

ε

(

t

m

−(

k

−1)

)

i

)

1

/

2

.

(19)

Whene,

−h

y

(0)

,

ϕ

e

ε

(0)

i ≤

C

(Z

T

0

h

β

ε

ϕ

ε

(

t

)

, ϕ

ε

(

t

)

i

dt

+

k

X

=

m

k

=1

h

δ

k

ε

ϕ

ε

(

t

m

−(

k

−1)

)

, ϕ

ε

(

t

m

−(

k

−1)

)

i

)

1

/

2

.

(20)

It follows atone from (18), (19) and (20) that

ε

R

T

0

k

ϕ

ε

(

t

)

k

2

dt

+

k

P

=

m

k

=1

k

ϕ

ε

(

t

k

)

k

2

+

R

0

T

h

e

ε

(

t

)

,

e

ε

(

t

)

i

dt

+

k

P

=

m

k

=1

h

e

D

k

ϕ

ε

(

t

m

−(

k

−1)

)

,

D

e

k

ϕ

ε

(

t

m

−(

k

−1)

)

i

=

R

0

T

(

β

ε

ϕ

ε

(

t

)

, ϕ

ε

(

t

))

dt

+

k

P

=

m

k

=1

h

δ

ε

k

ϕ

ε

(

t

m

−(

k

−1)

)

, ϕ

ε

(

t

m

−(

k

−1)

))

C

2

.

(21)

Step 4: Aording to the estimate(20) the family

b

ε

=

B

ε

(

ϕ

e

ε

(

t

);

ϕ

e

ε

(

t

1

)

...,

ϕ

e

ε

(

t

m

)

= (

B

e

2

ε

ϕ

e

(

t

);

D

e

1

2

ϕ

e

ε

(

t

1

)

...,

D

e

m

ϕ

e

ε

(

t

m

)) +

ε

(

ϕ

e

ε

(

t

);

ϕ

e

ε

(

t

1

)

...,

ϕ

e

ε

(

t

m

))

is ontainedin abounded subset

K

m

.

Thus, both of the families

ε

(

ϕ

e

ε

(

t

);

ϕ

e

ε

(

t

1

)

...,

ϕ

e

ε

(

t

m

))

and

(

B

ϕ

e

ε

(

t

);

D

1

ϕ

e

ε

(

t

1

)

..., D

m

ϕ

e

ε

(

t

m

))

are bounded in

K

m

. Therefore, we may extrat asubfamily, say
(15)

(

B

e

2

ϕ

e

ε

(

t

);

D

e

1

2

ϕ

e

ε

(

t

1

)

...,

D

e

2

m

ϕ

e

ε

(

t

m

))+

ε

(

ϕ

e

ε

(

t

);

ϕ

e

ε

(

t

1

)

...,

ϕ

e

ε

(

t

m

))

B

h

,

weakly in

K

m

.

Step 5: Takingthe limitas

ε

0

,wesee thatthe solution

y

of(1)with initial ondition

y

(0) =

y

0

, h being as in step 4 satises

y

(

T

) = 0

. This

ompletes the proof of Theorem 1.

As an immediate appliation of the foregoing Theorem we give the

fol-lowingexample.

Example. One dimensionalimpulsiveShrödinger equation :

We onsider the problem

∂y

(

t, x

)

∂t

+

i

2

y

∂x

2

(

t, x

) =

χ

ω0

u

(

t, x

)

, t

(0

, T

)

\ {

t

k

}k

σ

m

1

, x

Ω = (0

,

2

π

)

,

y

(

t,

0) =

y

(

t,

2

π

) = 0

,

(22)

y

(0

, x

) =

y

0

,

y

(

t

k

, x

) =

k

y

(

t

k

, x

) +

χ

ω

k

v

k

(

x

)

, k

σ

m

1

,

where

t

k

+1

t

k

>

2

π, ω

k

= (

a

k

1

, a

2

k

)

, k

σ

0

m

,

{

α

k

}k

σ

m

1

R

+

.

Let

H

=

L

2

(Ω

,

C

)

, Aw

(

x

) =

i

2

w

∂x

2

(

x

)

, D

(

A

) =

n

w

H,

∂x

2

w

2

H, w

(0) =

w

(

π

) = 0

o

,

and

I

k

w

(

x

) =

k

w

(

x

)

and the ontrol operator is given by

B

=

χ

ω0

, D

k

=

χ

ω

k

, then the system (22) beomes an abstrat formulation of (1). As a

onsequene of Theorem 1, the initial state

y

0

L

2

(Ω

,

C

) =

H

of the

solution of (22) is null-ontrollable at

t

=

T,

if and only if, there is

C >

0

suh that

Z

y

0

(

x

)

ϕ

e

0

(

x

)

dx

(23)

C

(Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

+

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

)

1

2

(16)

where

ϕ

e

0

(

x

) =

ϕ

(

T, x

)

and

ϕ

is the mildsolutionof

∂ϕ

(

t, x

)

∂t

+

i

2

ϕ

(

t, x

)

∂x

2

= 0

,

t

(0

, T

)

\ {

t

k

}k

σ

m

1

, x

,

ϕ

(

t,

0) =

ϕ

(

t,

2

π

) = 0

,

ϕ

(0

, x

) =

ϕ

0

(

x

)

, x

,

ϕ

(

t

k

, x

) =

k

ϕ

(

t

k

, x

)

, x

, k

σ

1

m

.

Here

ϕ

is given by

ϕ

(

t

) =

ϕ

[0]

(

t

)

,

if

t

[

t

0

, t

1

)

ϕ

[

k

]

(

t

)

,

if

t

t

k

, t

k

+1)

ϕ

[

m

]

(

t

)

,

if

t

[

t

m

, T

]

,

where

ϕ

[

k

]

(

t

)

is asolution of the lassialShrödinger equation

∂ϕ

[

k

]

(

t, x

)

∂t

+

i

2

ϕ

[

k

]

∂x

2

(

t, x

) =

χ

ω0

u

(

t, x

)

,

t

(

t

0

, t

1

)

, x

Ω = (0

,

2

π

)

,

ϕ

[

k

]

(

t,

0) =

ϕ

[

k

]

(

t,

2

π

) = 0

,

ϕ

[0]

(

t

0

, x

) =

ϕ

0

(

x

)

, x

,

and

∂ϕ

[

k

]

(

t, x

)

∂t

+

i

2

ϕ

[

k

]

∂x

2

(

t, x

) =

χ

ω0

u

(

t, x

)

, t

(

t

k

, t

k

+1

)

,

x

Ω = (0

,

2

π

)

,

ϕ

[

k

]

(

t,

0) =

ϕ

[

k

]

(

t,

2

π

) = 0

,

ϕ

[

k

]

(

t

k

, x

) = (1 +

k

)

ϕ

[

k

−1]

(

t

k

, x

)

, x

, k

σ

m

1

.

Then a standard appliation of a variant of Ingham's Inequality [8℄ shows

that

Z

t

k

+1

t

k

Z

w0

ϕ

[

k

]

(

t, x

)

dtdx

c

(

τ

k

, w

0

)

Z

ϕ

[

k

]

(

t

+

k

, x

)

dx,

for some positiveonstants

c

(

τ

k

, w

0

)

>

0

. Summingup we get

m

X

k

=0

Z

t

k

+1

t

k

Z

w0

ϕ

[

k

]

(

t, x

)

dtdx

=

Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

c

1

m

X

k

=1

Z

ϕ

[

k

]

(17)

where

c

1

= min

k

σ

m

0

c

(

τ

k

, w

0

)

>

0

.

On the other hand, there isa positive onstant

c

2

>

0

suh that

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

c

2

m

X

k

=1

Z

ϕ

[

k

]

2

(

t

+

k

, x

)

dx.

It follows that

Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

+

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

(

c

1

+

c

2

)

m

X

k

=1

Z

ϕ

[

k

]

2

(

t

+

k

, x

)

dx

(

c

1

+

c

2

)

Z

ϕ

[

m

]

2

(

t

+

m

, x

)

dx

= (

c

1

+

c

2

)

Z

|

ϕ

|

2

(

T, x

)

dx.

Now, sine

ϕ

e

0

(

x

) =

ϕ

e

(0

, x

) =

ϕ

(

T, x

)

,

then,

Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

+

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

m

(

c

1

+

c

2

)

Z

e

ϕ

0

2

(

x

)

dx,

from whih we get

Z

e

ϕ

0

2

(

x

)

dx

1

m

(

c

1

+

c

2

)

Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

+

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

!

.

We onlude by Cauhy-Shwarz inequality that

Z

y

0

(

x

)

ϕ

e

0

(

x

)

dx

Z

y

0

2

(

x

)

dx

Z

e

ϕ

0

2

(

x

)

dx

1

/

2

(R

|

y

0

|

2

(

x

)

dx

m

(

c

1

+

c

2

)

)

1

/

2

Z

T

0

Z

ω0

|

ϕ

|

2

(

t, x

)

dxdt

+

m

X

k

=1

Z

ω

k

|

ϕ

|

2

(

t

m

−(

k

−1)

, x

)

dx

!

1

/

2

(18)

stated in Theorem 1.

We onlude our paper by a speial ase when our initial state is an

eigensolution of the followinglinear operator

Γ :

H

H

dened by

Γ(

ψ

) =

Z

T

0

X

−1

(

s

)

B

2

X

(

s

)

ψds

+

k

X

=

m

k

=1

X

−1

(

t

k

)

D

2

k

X

(

t

k

)

ψ.

We have the following result ofnull-ontrollability.

Proposition 1 Let

λ >

0

be an eigenvalue of

Γ

with eigenvetor

ψ

H.

Then, the solution

y

to the problem

y

(

t

) +

Ay

(

t

) =

λ

1

B

2

(

X

(

t

)

ψ

)

,

t

(0

, T

)

/

{

t

k

}k

σ

m

1

,

y

(

t

k

) =

I

k

y

(

t

k

)

1

λ

D

2

k

(

X

(

t

k

)

ψ

)

, k

σ

1

m

y

(0) =

ψ,

(24)

satises

y

(

T

) = 0

.

Proof.

Write system (24) intothe form

y

(

t

) +

Ay

(

t

) =

λ

1

B

2

(

X

(

t

)

ψ

)

,

t

(0

, T

)

/

{

t

k

}k

σ

m

1

,

y

(

t

+

k

) =

I

k

y

(

t

k

)

1

λ

D

2

k

(

X

(

t

k

)

ψ

)

, k

σ

1

m

y

(0) =

ψ.

Therefore, this impulsive problem has a solution whih an be represented

expliitlyas follows

y

(

t

) =

X

(

t

)

ψ

+

Z

t

0

G

(

t, s

)

λ

1

B

2

(

X

(

s

)

ψ

ds

+

X

0

<t

k≤

t

G

(

t, t

k

)

λ

1

D

2

k

X

(

t

k

)

ψ

,

where the evolutionoperator

G

(

t, s

)

isgiven by
(19)

y

(

T

) =

X

(

T

)

ψ

+

Z

T

0

G

(

T, s

)

λ

1

B

2

(

X

(

s

)

ψ

ds

+

X

0

<t

k≤

T

G

(

T, t

k

)

λ

1

D

k

2

X

(

t

k

)

ψ

=

X

(

T

)

ψ

+

Z

T

0

X

−1

(

T

)

G

(

T, s

)

λ

1

B

2

(

X

(

s

)

ψ

ds

λ

1

X

0

<t

k≤

T

X

−1

(

T

)

G

(

T, t

k

)

D

k

2

X

(

t

k

)

ψ

#

=

X

(

T

)

ψ

+

Z

T

0

X

−1

(

s

)

λ

1

B

2

(

X

(

s

)

ψ

ds

λ

1

X

0

<t

k≤

T

X

−1

(

t

k

)

D

2

k

X

(

t

k

)

ψ

#

=

X

(

T

))

ψ

1

λ

Γ (

ψ

)

= 0

.

This shows thatthe initialstate

ψ

isnull-ontrollableattime

T

withontrol

h

(

t

) =

u

(

t

)

,

{

v

k}k

σ

m

1

=

1

λ

X

(

t

)

ψ,

1

λ

X

(

t

k

)

ψ

k

σ

m

1

!

,

whih ompletes the proof of the Proposition.

Referenes

[1℄ N.U. Ahmed, Optimal impulse ontrol for impulsive systems in Banah

spaes. J. Math. Anal.Appl. Vol.1 (No.1)(2000), 37-52.

[2℄ M.U. Akhmetov, A. Zafer, The ontrollability of boundary-value

prob-lems for quasilinear impulsive systems, Nonlinear Analysis 34 (1998)

1055-1065.

[3℄ D.D. BainovandP.S.Simeonov,Systems withimpulseeet,theoryand

appliations,EllisHardwoodseriesinMathematisanditsAppliations,

(20)

toti properties of the solutions, World Sienti, Series onAdvanes in

Math. forApplied Sienes, 28 (1995).

[5℄ L. Berezansky and E. Braverman, Boundedness and stability of

impul-sively perturbed systems in a Banah spae. Preprint funtan/9312001.

[6℄ R.K. George,A.K.NandakumaranandA.Arapostathis,Anote on

on-trollability of impulsive systems. J. Math. Anal. Appl. 241, 276-283,

2000.

[7℄ Z.H. Guan, T.H. Qian, and X. Yu, Controllability and observability of

linear time-varyingimpulsivesystems.IEEECiruitsSyst.I,vol.49,pp.

1198-1208, 2002.

[8℄ A.Haraux, Analternativefuntional approahtoexatontrollability of

onservative systems, Portugaliaemathematia61, 4(2004), 399-437.

[9℄ V. Lakshmikantham, D.D.Bainovand P.S. Simeonov, Theory of

impul-sive dierentialequations.WorldSientiseries inModern

Mathemat-is, Vol. 6,Singapore, 1989.

[10℄ S.Leela,F.A.MRae,and S.Sivasundaram,Controllabilityof impulsive

dierential equations, J. Math. Anal.Appl. 177, 1993 , 24-30.

[11℄ X. Liu, Nonlinear boundary value problems for rst order impulsive

in-tegrodierential equations,Appl. Anal. 36(1990), 119-130.

[12℄ J.H. Liu, Nonlinear impulsive evolution equations, Dynamis Contin.

Disr. ImpulsiveSyst., 6(1999), 77-85.

[13℄ A.M. Samoilenko and N.A. Perestyuk, Impulsive dierential equations,

World Sienti, Singapore, 1995.

[14℄ R.E. Showalter, Hilbert spae methods for partial dierential equations,

Eletroni Journal of Dierential Equations,Monograph 01, 1994.

[15℄ M. Slemrod, Feedbak stabilization of a linear ontrol system in Hilbert

spae with an a prioribounded ontrol,Math. ControlSignals Systems,

(21)

bounded operators, SIAM J. Control Optimiz.13(1) (1975),462-490.

Referensi

Dokumen terkait

DIREKTORAT JENDERAL PERHUBUNGAN UDARA UNIT PENYELENGGARA BANDAR UDARA..

Oleh karena itu, akan dibangun aplikasi game yang menggunakan algoritma A*(A star)[13] terhadap musuh agar bisa lebih agresif, menambahkan karakter bos pada akhir

Kamus data yang diusulkan pada sistem informasi akuntansi laporan keuangan. pada YPM Salman ITB adalah

Penelitian yang dilakukan di Arab Saudi misalnya menun- jukkan hubungan antara penyakit periodon- tal dengan tingkat/kadar glukosa darah pada penderita DM tipe II, dimana

Konflik terbuka yang terjadi dikawasan pertambangan nikel di Kecamatan Talaga Raya adalah akibat ganti rugi lahan dan tanaman warga yang tidak sesuai dengan kesepakatan awal,

Pada tingkat pengamatan, hasil pengamatan berupa { ~r, ~ p } pada fisika klasik sama persis dengan prediksi yang diberikan oleh formulasi. Dengan demikian,

Dokumen Pengadaan dapat diambil dalam bentuk softcopy dan diunduh (download) melalui aplikasi Sistem Pengadaan Secara Elektronik (SPSE) pada alamat website LPSE

Kepada penyedia barang/jasa baik sendiri-sendiri maupun bersama-sama yang berkeberatan atas pengumuman ini diberikan waktu untuk mengajukan sanggahan selama 3 (tiga) hari