Directory UMM :Journals:Journal_of_mathematics:EJQTDE:
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 Spa
e
R. BOUKHAMLA(1) & S. MAZOUZI(2)
(1) Centre Universitaire de Souk-Ahras, Souk-Ahras 41000, Algeria.
(2) Département de Mathématiques, Université d'Annaba, BP 12, ANNABA 23000, Algeria.
Abstra
t
We shall establish a ne
essary and su
ient
ondition under whi
h
we have the null
ontrollability of some rst order impulsive evolution
equation in a Hilbert spa
e.
MSC(2000) : 34A37, 93B05, 93C15.
Keywords: Null-
ontrollability, impulsive
onditions, mild solutions,
evolution equation.
1
Introdu
tion
The problem of exa
t
ontrollability of linear systems represented by innite
onservative systems has been extensively studied by several authors A.
Haraux [8℄, R.Triggiani [16℄, Z.H. Guan, T.H. Qian, and X.Yu [7℄, see also
the referen
es [1, 2, 6, 10,15℄. In the sequel, we shall be
on
erned with the
problem of null
ontrollability of some rst order evolution equation subje
t to impulsive
onditions and so we shall derive a ne
essary and su
ient
ondition under whi
h null
ontrollability o
urs. A
tually, we shall establish an equivalen
e between the null-
ontrollability and some observability
inequality in somehow more general framework than that proposed by A Haraux [8℄. Regarding the literature on the impulsive dierential equations we
refer the reader to the works of D.D. Bainov and P.S. Simeonov [3, 4℄ and
EJQTDE, 2007 No. 19, p. 1
the referen
es [5, 9,11, 12, 13℄. We are going to study the following problem
′
y (t) + Ay (t) = Bu (t) ,
t ∈ (0, T ) \ {tk }k∈σm ,
1
0
y (0) = y ,
∆y (tk ) = Ik y (tk ) + Dk vk , k ∈ σ1m ,
(1)
(1k )
where the nal time T is a positive number, y 0 is an initial
ondition in a
Hilbert spa
e H endowed with an inner produ
t h., .iH , y (t) : [0, T ] → H is a
ve
tor fun
tion, σ1m is a subset of N given by σ1m = {1, 2, ..., m}, and nally,
{tk }k∈σm is an in
reasing sequen
e of numbers in the open interval (0, T ) ,
1
and ∆y (tk ) denotes the jump of y (t) at t = tk , i.e.,
−
∆y (tk ) = y t+
k − y tk
where y t+
and y t−
represent the right and left limits of y (t) at t = tk
k
k
respe
tively. On the other hand, the operators A, B, Ik , Dk : H → H are
given linear bounded operators. Moreover, we set the following assumptions:
(H1) A∗ = −A,
(H2) Ik∗ = −Ik , for every k ∈ σ1m , and for ea
h k ∈ σ1m , the operator
Ik = Ik + I is invertible,
(H3) B ∗ = B ≥ 0 and there is d0 > 0 su
h that
(Bu, u)H ≤ d0 kuk2H , for all u ∈ H,
(H4) Dk∗ = Dk ≥ 0,
dk > 0 su
h that
for every k ∈ σ1m , and for ea
h k ∈ σ1m there is
(Dk u, u)H ≤ dk kuk2H , for all u ∈ H.
In the sequel we shall designate by
h
the fun
tion
h (t) = u (t) , {vk }k∈σm ,
1
where u (t) ∈ L2 (0, T ) \ {tk }k∈σ1m ; H and
{vk }k∈σm ∈ l
1
2
(σ1m ; H)
n
o
+ {vk }k∈σm , vk ∈ H .
1
EJQTDE, 2007 No. 19, p. 2
We point out that the spa
e Km = L2 (0, T ) \ {tk }k∈σ1m ; H × l2 (σ1m ; H) is
a Hilbert spa
e with respe
t to the inner produ
t
h, e
h
Km
=
Z
T
0
(u (t) , u
e (t))H dt +
m
X
k=1
(vk , e
vk )H ,
e
dened for all h = (u (t) , {vk }m
u (t) , {e
vk }m
k=1 ) and h = (e
k=1 ) ∈ Km .
We shall denote by B the
ontrol operator given by
B = B, {Dk }k∈σm ∈ L L2 (0, T ) \ {tk }k∈σm ; H × l2 (σ1m ; H) ,
1
1
so that
Bh (t) = Bu (t) , {Dk vk }k∈σm .
1
We have for every h =
(u(t), {vk }m
k=1 )
(Bh, h)Km =
=
Z
∈ Km
T
(Bu (t) , u (t))H dt +
0
Z
m
X
(Dk vk , vk )H
k=1
T
(u (t) , Bu (t))H dt +
0
m
X
(vk , Dk vk )H
k=1
= (h, Bh)Km ,
whi
h shows that B∗ = B, that is, B is self-adjoint. On the other hand, we
have
(Bh, h)Km =
Z
T
0
≤ d0
≤ δ
(Bu (t) , u (t))H dt +
Z
m
X
(Dk vk , vk )H
k=1
T
0
ku (t)k2H dt +
khk2Km
,
m
X
k=1
dk kvk k2H
where δ = max {d0 , d1 , ..., dm } . Thus, the operator is B bounded in Km .
Next, we
onsider the homogeneous system asso
iated with (1) :
′
ϕ (t) + Aϕ (t) = 0, t ∈ (0, T ) \ {tk }k∈σm ,
1
ϕ (0) = ϕ0 ,
∆ϕ (tk ) = Ik ϕ (tk ) , k ∈ σ1m .
(2)
(2k )
EJQTDE, 2007 No. 19, p. 3
We point out that on ea
h interval [tk , tk+1 ), for k = 0, ..., m, the solution ϕ
is left
ontinuous at ea
h time tk .
Consider the
orresponding homogeneous ba
kward problem :
(3)
′
−ϕ˜ (t) + Aϕ(t)
˜
= 0, , t ∈ (0, T ) \ {tk }k∈σm ,
1
0
ϕ(T
˜ ) = ϕ ,
m
∆ϕ(t
˜ m−(k−1) ) = −I˜m−(k−1) ϕ(t
˜ +
m−(k−1) ), k ∈ σ1 ,
(3k )
where
∗
A = A∗ = −A, I˜m−(k−1) = Im−(k−1)
= −Im−(k−1) , k ∈ σ1m .
We observe that the problem (3) on the interval [tm , T ] is equivalent to the
lassi
al ba
kward problem
′
−ϕ˜ (t) + Aϕ(t)
˜ = 0, t ∈ [tm , T ] ,
0
ϕ(T
˜ )=ϕ .
We introdu
e the following spa
e : PC ([0, T ] ; H) = {y, y : [0, T ] → H
su
h that y(t) is
ontinuous at t 6= tk , and has dis
ontinuities of rst kind at
t = tk , for every k ∈ σ1m }.
Evidently, PC ([0, T ] ; H) is a Bana
h spa
e with respe
t to the norm
kykPC = sup ky(t)k .
t∈(0,T )
On the other hand, we dene the subspa
es PLC , (respe
tively, PRC )=
{y, y ∈ PC su
h that y(t) is left ( respe
tively, right)
ontinuous at t = tk ,
for every k ∈ σ1m }.
Remark 1 1) The spa
e PLC , (respe
tively, PRC )
an be identied to a
subspa
e of Km . That is, to ea
h y ∈ PLC , (respe
tively, y˜ ∈ PRC ) is
assigned the fun
tion h (respe
tively, eh) dened by
h (t) = y (t) , {y (tk )}k∈σ1m ,
and
e
h (t) = y˜ (t) , {˜
y (tk )}k∈σm .
1
The mapping y 7→ h (t) (respe
tively, y˜ 7→ he) is a linear inje
tion.
EJQTDE, 2007 No. 19, p. 4
2) Let ye ∈ PRC , the fun
tion y
an be written as :
ye[0] (t)
ye[k] (t)
ye(t) =
ye[m] (t)
if t ∈ [t0 , t1 )
if t ∈ [tk , tk+1 )
if t ∈ [tm , T ] .
Next, let τk = tk − tk−1 , we dene the operator T : D(T ) = PRC ⊂ Km →
Km by
τ1
+ t0 )
ye[0] ((T − t) τm+1
if
t ∈ [tm , T ] ,
τk+1
ye[k]((tm−(k−1) − t) τm−(k−1) + tk ) if t ∈ tm−k , tm−(k−1) , k ∈ σ1m−1 ,
(T ye)(t) =
+ tm )
if
t ∈ (0, t1 ] .
ye[m] ((t1 − t) τm+1
τ1
(4)
We note that the range of T is exa
tly PLC . The fun
tion (T ye)(t)
an be
written as follows:
y[0] (t)
y[k](t)
(T ye)(t) =
y[m] (t)
if t ∈ [t0 , t1 ] ,
if t ∈ (tk , tk+1] , k ∈ σ1m−1 ,
if t ∈ (tm , T ] .
Let X (t) be the resolvent solution of the operator system
′
X (t) + AX(t) = 0, 0 = t0 < t < tm+1 = T, t 6= tk , k = 1, 2, ..., m,
X(0) = I,
X(tk + 0) − X(tk − 0) = Ik X(tk ), k = 1, 2, ..., m,
where I : H → H is the identity operator. We shall suppose that the operator
Ik = Ik + I has a bounded inverse.
Denition 1 A fun
tion y ∈ PC ([0, T ] ; H)is a mild solution to the impul-
sive problem (1) if the impulsive
onditions are satised and
y(t) =
Rt
+ 0
G(t,
0
)y
+
G(t, s)Bu (s) ds
0
P
+ 0 0.
Claim 1 Assume that there is α > 0 su
h that
(Z
T
0
kBu(t)k2H dt +
k=m
X
k=1
kDk vk k2H
)
≥α
(Z
T
0
ku(t)k2H dt +
k=m
X
k=1
kvk k2H
)
then, for every y 0 ∈ H there is ϕ0 ∈ H su
h that the mild solution of (1)
with
e
ϕ(t
e 1 ), .., ϕ(t
e k ).., ϕ(t
e m )) ∈ Km and y(0) = y 0
h(t) = (ϕ(t),
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 mild solution y to the problem
/
′
y (t) + Ay(t) = B ϕ(t),
e
t ∈ (0, T ) {tk }k∈σm ,
1
∆y(tk ) = Ik y(tk ) + Dk ϕ(t
e k ),
y(T ) = 0.
Next, we introdu
e a bounded linear operator Λ : H → H dened by
Λz = −y(0).
A
ording to formula (11) and the Corollary 1 we have
Z
k=m
T
X
|hΛz, zi| = |−hy(0), ϕ(0)i|
e
=
hB ϕ(t),
e
ϕ(t)idt
e
+
hDk ϕ(t
e k ), ϕ(t
e k )i
0
k=1
Z
k=m
T
X
=
hDk ϕ(tm−(k−1) ), ϕ(tm−(k−1) )i
hBϕ(t), ϕ(t)idt +
0
k=1
)
(Z
k=m
T
X
kϕ(tk )k2 ,
≤ ς
kϕ(t)k2 dt +
0
k=1
EJQTDE, 2007 No. 19, p. 11
where
ς = sup {dk } < ∞.
k∈σ0m
We have
Z
T
0
2
kϕ(t)k dt =
Z
t1
2
kϕ(t)k dt +
0
Z
t2
t1
2
kϕ(t)k dt + ... +
Z
T
tm
kϕ(t)k2 dt.
Sin
e there is no impulse in the interval [tk , tk+1 ) we have
, for every t ∈ [tk , tk+1 ) , k ∈ σ m ,
kϕ(t)k =
ϕ(t+
)
0
k
−
+
ϕ(t )
=
ϕ(t )
, k ∈ σ m .
0
k−1
k+1
(13)
Therefore, there are τk+1 = tk+1 − tk > 0, k ∈ σ0m su
h that
R tk+1
tk
2
= τk+1
Ik ϕ(t− ) + ϕ(t− )
2 , k ∈ σ m . (14)
kϕ(t)k2 dt ≤ ρk
ϕ(t+
)
1
k
k
k
On the other hand, the
ontinuity of Ik implies that
+
2
ϕ(t )
=
(Ik + I)ϕ(t− )
2 ≤ (1 + L(Ik ))2
ϕ(t− )
2 , k ∈ σ m .
1
k
k
k
(15)
It follows from (14) and (15) that
R tk+1
tk
2
, k ∈ σ1m .
kϕ(t)k2 dt ≤ τk+1 (1 + L(Ik ))2
ϕ(t−
k)
(16)
Sin
e m is nite, and due to (13),(16), then there is a
onstant 0 < µ < ∞
su
h that hΛz, zi ≤ µ kzk2 , and thus, Λ is bounded.
Now, as B is nonnegative in Km , we have
kBξ (t)k ≥ α {(ξ (t) , ξ (t))Km }1/2
for all ξ ∈ Km ; thus, by virtue of Lemma 2, we have
(Z
T
(Bu(t), u(t))H dt +
0
≥ α kBk
(Z
T
0
k=m
X
ku(t)k2H dt +
(Dk vk , vk )H
k=1
k=m
X
k=1
kvk k2H
)
)
(17)
.
EJQTDE, 2007 No. 19, p. 12
It follows from (11), (17) and Corollary 1 that
hΛz, zi = −hy(0), ϕ(0)i
e
Z T
k=m
X
=
hBϕ(t), ϕ(t)idt +
hDk ϕ(tm−(k−1) ), ϕ(tm−(k−1) )i
0
≥ α kBk
≥ α kBk
(Z
Z
0
0
t1
k=1
T
2
kϕ(t)k dt +
k=m
X
k=1
kϕ(tk )k
2
)
kϕ(t)k2 dt = kBk αt1 kzk2 = θ kzk2 ,
be
ause there is no impulse before time t1 . Therefore, Λ is
oer
ive on H .
To show that there is a bije
tion from H onto H, it su
es to prove that
Λ + I is a bije
tion from H onto H. Clearly, Λ + I is inje
tive sin
e
hΛz + z, zi = hΛz, zi + hz, zi ≥ (θ + 1) kzk2 .
On the other hand, let y 0 ∈ H , as the form a(f, g) + hf, gi = hΛf, gi + hf, gi
is symmetri
and
oer
ive, then, by virtue of Lax-Milgram Theorem, there
is an element f ∈ H su
h that
a(f, g) + hf, gi = hy 0, gi, for all g ∈ H.
This implies that Λ(H) = H . Thus, for every y 0 ∈ H , there is a unique
z ∈ H su
h that Λ(z) = −y 0 , whi
h
ompletes the proof of Claim 1.
Step 3: Assume that B, Dk ≥ 0, then B ≥ 0,
We dene for ε > 0,
e 2 = B, D
e 2 = Dk .
B
k
e 2 + εI,
βε + B
e 2 + εI,
δε + D
k
and
k
ε
e 2 + εI; D
e 2 + εI, .., D
e 2 + εI).
Bε + (β ε ; δ1ε , .., δm
) = (B
1
m
A
ording to Claim 1, there is ϕ˜0,ε ∈ H su
h that the mild solution yε of (1)
with yε (0) = y 0 satises yε (T ) = 0; where B(h) has been repla
ed by
Bε (ϕ(t),
e
ϕ(t
e 1 ), .., ϕ(t
e k ).., ϕ(t
e m )) ∈ Km .
EJQTDE, 2007 No. 19, p. 13
We obtain from (11) and Corollary 1
−hy(0), ϕ
eε (0)i =
Z
T
hβεε ϕ(t),
e
ϕ
eε (t)idt +
0
and (7) gives
−hy(0), ϕ
eε(0)i ≤ C
(Z
T
0
When
e,
−hy(0), ϕ
eε(0)i ≤ C
(Z
e 2 ϕε (t), ϕε (t)idt +
hB
T
ε
0
hβ ϕε (t), ϕε (t)idt +
k=m
X
hδkε ϕ
eε (tk ), ϕ
eε (tk )i,
k=1
k=m
X
k=1
e 2 ϕε (tm−(k−1) ), ϕε (tm−(k−1) )i
hD
k
hδkε ϕε (tm−(k−1) ), ϕε (tm−(k−1) )i
(20)
It follows at on
e from (18), (19) and (20) that
ε
RT
=
0
RT
0
2
kϕε (t)k dt +
+
RT
0
k=m
P
k=1
kϕε (tk )k
k=m
e ε (t), Bϕ
e ε (t)idt + P hD
e k ϕε (tm−(k−1) ), D
e k ϕε (tm−(k−1) )i
hBϕ
(β ε ϕε (t), ϕε (t))dt +
k=1
k=m
P
k=1
Step 4:
2
hδkε ϕε (tm−(k−1) ), ϕε (tm−(k−1) )) ≤ C 2 .
(21)
A
ording to the estimate (20) the family
bε = Bε (ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm )
2
2
e ϕ(t);
e ϕ
emϕ
= (B
e
D
eε (t1 )..., D
eε (tm )) + ε(ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm ))
ε
1
is
ontained in a bounded subset
Thus, both of the families
√
ε(ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm ))
are bounded in
Km .
Km .
and
(B ϕ
eε (t); D1 ϕ
eε (t1 )..., Dm ϕ
eε (tm ))
Therefore, we may extra
t a subfamily, say
(B ϕ
eε (t); D1 ϕ
eε (t1 )..., Dm ϕ
eε (tm )) ⇀ h,
)1/2
(19)
k=m
X
k=1
(18)
weakly in
Km .
EJQTDE, 2007 No. 19, p. 14
)1/2
.
.
Then
learly
e2ϕ
e 2ϕ
e 2 eε (tm ))+ε(ϕ
(B
eε (t); D
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm )) ⇀ Bh, weakly in Km .
1 eε (t1 )..., Dm ϕ
Taking the limit as ε → 0, we see that the 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 appli
ation of the foregoing Theorem we give the following example.
Example. One dimensional impulsive S
hrödinger equation :
We
onsider the problem
Step 5:
∂2y
∂y (t, x)
+ i 2 (t, x)
∂t
∂x
y(t, 0)
y (0, x)
∆y (tk , x)
= χω0 u (t, x) , t ∈ (0, T ) \ {tk }k∈σm , x ∈ Ω = (0, 2π),
1
(22)
= y(t, 2π) = 0,
= y0,
= iαk y (tk , x) + χωk vk (x), k ∈ σ1m ,
where
tk+1 − tk > 2π, ωk = (ak1 , ak2 ) ⊂ Ω, k ∈ σ0m , {αk }k∈σm ⊂ R+ .
1
Let
n
o
2
2
H = L2 (Ω, C), Aw(x) = i ∂∂xw2 (x), D(A) = w ∈ H, ∂∂xw2 ∈ H, w(0) = w(π) = 0 ,
and Ik w(x) = iαk w(x) and the
ontrol operator is given by B = χω0 , Dk =
χωk , then the system (22) be
omes an abstra
t formulation of (1). As a
onsequen
e of Theorem 1, the initial state y 0 ∈ L2 (Ω, C) = H of the
solution of (22) is null-
ontrollable at t = T, if and only if, there is C > 0
su
h that
Z
0
y 0(x)ϕ
e
(x)dx
Ω
(Z Z
T
≤ C
0
2
ω0
|ϕ| (t, x)dxdt +
(23)
m Z
X
k=1
2
ωk
|ϕ| (tm−(k−1) , x)
) 21
, ∀ϕ
e0 ∈ L2 (Ω, C),
EJQTDE, 2007 No. 19, p. 15
where
ϕ
e0 (x) = ϕ(T, x)
and
ϕ
∂ϕ(t, x)
∂ 2 ϕ(t, x)
+i
∂t
∂x2
ϕ(t, 0)
ϕ (0, x)
∆ϕ (tk , x)
Here
ϕ
where
is the mild solution of
= 0,
t ∈ (0, T ) \ {tk }k∈σm , x ∈ Ω,
1
= ϕ(t, 2π) = 0,
= ϕ0 (x), x ∈ Ω,
= iαk ϕ (tk , x) , x ∈ Ω, k ∈ σ1m .
is given by
ϕ[k](t)
ϕ[0] (t) ,
ϕ[k] (t) ,
ϕ(t) =
ϕ[m] (t) ,
if
if
if
t ∈ [t0 , t1 )
t ∈ tk , tk+1)
t ∈ [tm , T ] ,
is a solution of the
lassi
al S
hrödinger equation
∂ 2 ϕ[k]
∂ϕ[k] (t, x)
+i
(t, x) = χω0 u (t, x) , t ∈ (t0 , t1 ) , x ∈ Ω = (0, 2π),
∂t
∂x2
ϕ[k] (t, 0) = ϕ[k] (t, 2π) = 0,
ϕ[0] (t0 , x) = ϕ0 (x), x ∈ Ω,
and
∂ 2 ϕ[k]
∂ϕ[k] (t, x)
+i
(t, x) = χω0 u (t, x) , t ∈ (tk , tk+1 ) , x ∈ Ω = (0, 2π),
∂t
∂x2
ϕ[k] (t, 0) = ϕ[k] (t, 2π) = 0,
ϕ[k] (tk , x) = (1 + iαk )ϕ[k−1] (tk , x) , x ∈ Ω, k ∈ σ1m .
Then a standard appli
ation of a variant of Ingham's Inequality [8℄ shows
that
Z
tk+1
tk
Z
w0
ϕ[k] (t, x)dtdx ≥ c(τk , w0 )
for some positive
onstants
m Z
X
k=0
tk+1
tk
Z
w0
c(τk , w0 ) > 0.
ϕ[k] (t, x)dtdx =
Z
Ω
+
ϕ[k] (t , x)dx,
k
Summing up we get
Z
0
T
Z
ω0
|ϕ|2 (t, x)dxdt
m Z
X
+
ϕ[k] (t , x)dx,
≥ c1
k
k=1
Ω
EJQTDE, 2007 No. 19, p. 16
where
c1 = minm c(τk , w0 ) > 0.
k∈σ0
On the other hand, there is a positive
onstant
m Z
X
ωk
k=1
It follows that
Z
0
T
Z
ω0
c2 > 0
su
h that
m Z
X
2 +
ϕ[k] (t , x)dx.
|ϕ| (tm−(k−1) , x) ≥ c2
k
2
k=1
|ϕ|2 (t, x)dxdt
m Z
X
+
k=1
ωk
Ω
|ϕ|2 (tm−(k−1) , x)
m Z
X
2 +
ϕ[k] (t , x)dx
≥ (c1 + c2 )
k
Ω
Zk=1
2
≥ (c1 + c2 ) ϕ[m] (t+
m , x)dx
Ω
Z
= (c1 + c2 ) |ϕ|2 (T, x)dx.
Ω
ϕ
e (x) = ϕ(0,
e x) = ϕ(T, x), then,
Z
m Z
X
0 2
2
2
|ϕ| (tm−(k−1) , x) ≥ m(c1 + c2 ) ϕ
e (x)dx,
|ϕ| (t, x)dxdt +
Now, sin
e
Z
T
0
Z
ω0
0
k=1
ωk
Ω
from whi
h we get
Z
Ω
0 2
ϕ
e (x)dx ≤
1
m(c1 + c2 )
Z
T
0
Z
ω0
2
|ϕ| (t, x)dxdt +
m Z
X
k=1
ωk
2
|ϕ| (tm−(k−1) , x) .
We
on
lude by Cau
hy-S
hwarz inequality that
Z
1/2
Z
Z
0 2
0 2
0
y 0(x)ϕ
y (x)dx
ϕ
e (x)dx
e (x)dx ≤
Ω
Ω
Ω
(R
)1/2 Z Z
0 2
T
|y
|
(x)dx
Ω
≤
|ϕ|2 (t, x)dxdt
m(c1 + c2 )
0
ω0
!1/2
m Z
X
|ϕ|2 (tm−(k−1) , x)dx
,
+
k=1
!
ωk
EJQTDE, 2007 No. 19, p. 17
whi
h establishes the ne
essary and su
ient
ondition of null
ontrollability
stated in Theorem 1.
We
on
lude our paper by a spe
ial
ase when our initial state is an
eigensolution of the following linear operator Γ : H → H dened by
Γ(ψ) =
Z
T
X
−1
2
(s)B X(s)ψds +
0
k=m
X
X −1 (tk )Dk2 X(tk )ψ.
k=1
We have the following result of null-
ontrollability.
Proposition 1 Let λ > 0 be an eigenvalue of Γ with eigenve
tor ψ ∈ H.
Then, the solution y to the problem
/
′
t ∈ (0, T ) {tk }k∈σm ,
y (t) + Ay(t) = − λ1 B 2 (X(t)ψ),
1
1 2
m
D
(X(t
)ψ),
k
∈
σ
∆y(t
)
=
I
y(t
)
−
k
k
k
k
1
λ k
y(0) = ψ,
(24)
satises
y(T ) = 0.
Proof.
Write system (24) into the form
/
′
1 2
t ∈ (0, T ) {tk }k∈σm ,
y (t) + Ay(t) = − λ B (X(t)ψ),
1
+
1 2
m
D
(X(t
)ψ),
k
∈
σ
y(t
)
=
I
y(t
)
−
k
k
k
1
k
λ k
y(0) = ψ.
Therefore, this impulsive problem has a solution whi
h
an be represented
expli
itly as follows
y(t) = X(t)ψ+
Z
0
t
X
1 2
1 2
G(t, tk ) − Dk X(tk )ψ ,
G(t, s) − B (X(s)ψ ds+
λ
λ
0
2007, No. 19, 1-21; http://www.math.u-szeged.hu/ejqtde/
Null Controllability of Some Impulsive
Evolution Equation in a Hilbert Spa
e
R. BOUKHAMLA(1) & S. MAZOUZI(2)
(1) Centre Universitaire de Souk-Ahras, Souk-Ahras 41000, Algeria.
(2) Département de Mathématiques, Université d'Annaba, BP 12, ANNABA 23000, Algeria.
Abstra
t
We shall establish a ne
essary and su
ient
ondition under whi
h
we have the null
ontrollability of some rst order impulsive evolution
equation in a Hilbert spa
e.
MSC(2000) : 34A37, 93B05, 93C15.
Keywords: Null-
ontrollability, impulsive
onditions, mild solutions,
evolution equation.
1
Introdu
tion
The problem of exa
t
ontrollability of linear systems represented by innite
onservative systems has been extensively studied by several authors A.
Haraux [8℄, R.Triggiani [16℄, Z.H. Guan, T.H. Qian, and X.Yu [7℄, see also
the referen
es [1, 2, 6, 10,15℄. In the sequel, we shall be
on
erned with the
problem of null
ontrollability of some rst order evolution equation subje
t to impulsive
onditions and so we shall derive a ne
essary and su
ient
ondition under whi
h null
ontrollability o
urs. A
tually, we shall establish an equivalen
e between the null-
ontrollability and some observability
inequality in somehow more general framework than that proposed by A Haraux [8℄. Regarding the literature on the impulsive dierential equations we
refer the reader to the works of D.D. Bainov and P.S. Simeonov [3, 4℄ and
EJQTDE, 2007 No. 19, p. 1
the referen
es [5, 9,11, 12, 13℄. We are going to study the following problem
′
y (t) + Ay (t) = Bu (t) ,
t ∈ (0, T ) \ {tk }k∈σm ,
1
0
y (0) = y ,
∆y (tk ) = Ik y (tk ) + Dk vk , k ∈ σ1m ,
(1)
(1k )
where the nal time T is a positive number, y 0 is an initial
ondition in a
Hilbert spa
e H endowed with an inner produ
t h., .iH , y (t) : [0, T ] → H is a
ve
tor fun
tion, σ1m is a subset of N given by σ1m = {1, 2, ..., m}, and nally,
{tk }k∈σm is an in
reasing sequen
e of numbers in the open interval (0, T ) ,
1
and ∆y (tk ) denotes the jump of y (t) at t = tk , i.e.,
−
∆y (tk ) = y t+
k − y tk
where y t+
and y t−
represent the right and left limits of y (t) at t = tk
k
k
respe
tively. On the other hand, the operators A, B, Ik , Dk : H → H are
given linear bounded operators. Moreover, we set the following assumptions:
(H1) A∗ = −A,
(H2) Ik∗ = −Ik , for every k ∈ σ1m , and for ea
h k ∈ σ1m , the operator
Ik = Ik + I is invertible,
(H3) B ∗ = B ≥ 0 and there is d0 > 0 su
h that
(Bu, u)H ≤ d0 kuk2H , for all u ∈ H,
(H4) Dk∗ = Dk ≥ 0,
dk > 0 su
h that
for every k ∈ σ1m , and for ea
h k ∈ σ1m there is
(Dk u, u)H ≤ dk kuk2H , for all u ∈ H.
In the sequel we shall designate by
h
the fun
tion
h (t) = u (t) , {vk }k∈σm ,
1
where u (t) ∈ L2 (0, T ) \ {tk }k∈σ1m ; H and
{vk }k∈σm ∈ l
1
2
(σ1m ; H)
n
o
+ {vk }k∈σm , vk ∈ H .
1
EJQTDE, 2007 No. 19, p. 2
We point out that the spa
e Km = L2 (0, T ) \ {tk }k∈σ1m ; H × l2 (σ1m ; H) is
a Hilbert spa
e with respe
t to the inner produ
t
h, e
h
Km
=
Z
T
0
(u (t) , u
e (t))H dt +
m
X
k=1
(vk , e
vk )H ,
e
dened for all h = (u (t) , {vk }m
u (t) , {e
vk }m
k=1 ) and h = (e
k=1 ) ∈ Km .
We shall denote by B the
ontrol operator given by
B = B, {Dk }k∈σm ∈ L L2 (0, T ) \ {tk }k∈σm ; H × l2 (σ1m ; H) ,
1
1
so that
Bh (t) = Bu (t) , {Dk vk }k∈σm .
1
We have for every h =
(u(t), {vk }m
k=1 )
(Bh, h)Km =
=
Z
∈ Km
T
(Bu (t) , u (t))H dt +
0
Z
m
X
(Dk vk , vk )H
k=1
T
(u (t) , Bu (t))H dt +
0
m
X
(vk , Dk vk )H
k=1
= (h, Bh)Km ,
whi
h shows that B∗ = B, that is, B is self-adjoint. On the other hand, we
have
(Bh, h)Km =
Z
T
0
≤ d0
≤ δ
(Bu (t) , u (t))H dt +
Z
m
X
(Dk vk , vk )H
k=1
T
0
ku (t)k2H dt +
khk2Km
,
m
X
k=1
dk kvk k2H
where δ = max {d0 , d1 , ..., dm } . Thus, the operator is B bounded in Km .
Next, we
onsider the homogeneous system asso
iated with (1) :
′
ϕ (t) + Aϕ (t) = 0, t ∈ (0, T ) \ {tk }k∈σm ,
1
ϕ (0) = ϕ0 ,
∆ϕ (tk ) = Ik ϕ (tk ) , k ∈ σ1m .
(2)
(2k )
EJQTDE, 2007 No. 19, p. 3
We point out that on ea
h interval [tk , tk+1 ), for k = 0, ..., m, the solution ϕ
is left
ontinuous at ea
h time tk .
Consider the
orresponding homogeneous ba
kward problem :
(3)
′
−ϕ˜ (t) + Aϕ(t)
˜
= 0, , t ∈ (0, T ) \ {tk }k∈σm ,
1
0
ϕ(T
˜ ) = ϕ ,
m
∆ϕ(t
˜ m−(k−1) ) = −I˜m−(k−1) ϕ(t
˜ +
m−(k−1) ), k ∈ σ1 ,
(3k )
where
∗
A = A∗ = −A, I˜m−(k−1) = Im−(k−1)
= −Im−(k−1) , k ∈ σ1m .
We observe that the problem (3) on the interval [tm , T ] is equivalent to the
lassi
al ba
kward problem
′
−ϕ˜ (t) + Aϕ(t)
˜ = 0, t ∈ [tm , T ] ,
0
ϕ(T
˜ )=ϕ .
We introdu
e the following spa
e : PC ([0, T ] ; H) = {y, y : [0, T ] → H
su
h that y(t) is
ontinuous at t 6= tk , and has dis
ontinuities of rst kind at
t = tk , for every k ∈ σ1m }.
Evidently, PC ([0, T ] ; H) is a Bana
h spa
e with respe
t to the norm
kykPC = sup ky(t)k .
t∈(0,T )
On the other hand, we dene the subspa
es PLC , (respe
tively, PRC )=
{y, y ∈ PC su
h that y(t) is left ( respe
tively, right)
ontinuous at t = tk ,
for every k ∈ σ1m }.
Remark 1 1) The spa
e PLC , (respe
tively, PRC )
an be identied to a
subspa
e of Km . That is, to ea
h y ∈ PLC , (respe
tively, y˜ ∈ PRC ) is
assigned the fun
tion h (respe
tively, eh) dened by
h (t) = y (t) , {y (tk )}k∈σ1m ,
and
e
h (t) = y˜ (t) , {˜
y (tk )}k∈σm .
1
The mapping y 7→ h (t) (respe
tively, y˜ 7→ he) is a linear inje
tion.
EJQTDE, 2007 No. 19, p. 4
2) Let ye ∈ PRC , the fun
tion y
an be written as :
ye[0] (t)
ye[k] (t)
ye(t) =
ye[m] (t)
if t ∈ [t0 , t1 )
if t ∈ [tk , tk+1 )
if t ∈ [tm , T ] .
Next, let τk = tk − tk−1 , we dene the operator T : D(T ) = PRC ⊂ Km →
Km by
τ1
+ t0 )
ye[0] ((T − t) τm+1
if
t ∈ [tm , T ] ,
τk+1
ye[k]((tm−(k−1) − t) τm−(k−1) + tk ) if t ∈ tm−k , tm−(k−1) , k ∈ σ1m−1 ,
(T ye)(t) =
+ tm )
if
t ∈ (0, t1 ] .
ye[m] ((t1 − t) τm+1
τ1
(4)
We note that the range of T is exa
tly PLC . The fun
tion (T ye)(t)
an be
written as follows:
y[0] (t)
y[k](t)
(T ye)(t) =
y[m] (t)
if t ∈ [t0 , t1 ] ,
if t ∈ (tk , tk+1] , k ∈ σ1m−1 ,
if t ∈ (tm , T ] .
Let X (t) be the resolvent solution of the operator system
′
X (t) + AX(t) = 0, 0 = t0 < t < tm+1 = T, t 6= tk , k = 1, 2, ..., m,
X(0) = I,
X(tk + 0) − X(tk − 0) = Ik X(tk ), k = 1, 2, ..., m,
where I : H → H is the identity operator. We shall suppose that the operator
Ik = Ik + I has a bounded inverse.
Denition 1 A fun
tion y ∈ PC ([0, T ] ; H)is a mild solution to the impul-
sive problem (1) if the impulsive
onditions are satised and
y(t) =
Rt
+ 0
G(t,
0
)y
+
G(t, s)Bu (s) ds
0
P
+ 0 0.
Claim 1 Assume that there is α > 0 su
h that
(Z
T
0
kBu(t)k2H dt +
k=m
X
k=1
kDk vk k2H
)
≥α
(Z
T
0
ku(t)k2H dt +
k=m
X
k=1
kvk k2H
)
then, for every y 0 ∈ H there is ϕ0 ∈ H su
h that the mild solution of (1)
with
e
ϕ(t
e 1 ), .., ϕ(t
e k ).., ϕ(t
e m )) ∈ Km and y(0) = y 0
h(t) = (ϕ(t),
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 mild solution y to the problem
/
′
y (t) + Ay(t) = B ϕ(t),
e
t ∈ (0, T ) {tk }k∈σm ,
1
∆y(tk ) = Ik y(tk ) + Dk ϕ(t
e k ),
y(T ) = 0.
Next, we introdu
e a bounded linear operator Λ : H → H dened by
Λz = −y(0).
A
ording to formula (11) and the Corollary 1 we have
Z
k=m
T
X
|hΛz, zi| = |−hy(0), ϕ(0)i|
e
=
hB ϕ(t),
e
ϕ(t)idt
e
+
hDk ϕ(t
e k ), ϕ(t
e k )i
0
k=1
Z
k=m
T
X
=
hDk ϕ(tm−(k−1) ), ϕ(tm−(k−1) )i
hBϕ(t), ϕ(t)idt +
0
k=1
)
(Z
k=m
T
X
kϕ(tk )k2 ,
≤ ς
kϕ(t)k2 dt +
0
k=1
EJQTDE, 2007 No. 19, p. 11
where
ς = sup {dk } < ∞.
k∈σ0m
We have
Z
T
0
2
kϕ(t)k dt =
Z
t1
2
kϕ(t)k dt +
0
Z
t2
t1
2
kϕ(t)k dt + ... +
Z
T
tm
kϕ(t)k2 dt.
Sin
e there is no impulse in the interval [tk , tk+1 ) we have
, for every t ∈ [tk , tk+1 ) , k ∈ σ m ,
kϕ(t)k =
ϕ(t+
)
0
k
−
+
ϕ(t )
=
ϕ(t )
, k ∈ σ m .
0
k−1
k+1
(13)
Therefore, there are τk+1 = tk+1 − tk > 0, k ∈ σ0m su
h that
R tk+1
tk
2
= τk+1
Ik ϕ(t− ) + ϕ(t− )
2 , k ∈ σ m . (14)
kϕ(t)k2 dt ≤ ρk
ϕ(t+
)
1
k
k
k
On the other hand, the
ontinuity of Ik implies that
+
2
ϕ(t )
=
(Ik + I)ϕ(t− )
2 ≤ (1 + L(Ik ))2
ϕ(t− )
2 , k ∈ σ m .
1
k
k
k
(15)
It follows from (14) and (15) that
R tk+1
tk
2
, k ∈ σ1m .
kϕ(t)k2 dt ≤ τk+1 (1 + L(Ik ))2
ϕ(t−
k)
(16)
Sin
e m is nite, and due to (13),(16), then there is a
onstant 0 < µ < ∞
su
h that hΛz, zi ≤ µ kzk2 , and thus, Λ is bounded.
Now, as B is nonnegative in Km , we have
kBξ (t)k ≥ α {(ξ (t) , ξ (t))Km }1/2
for all ξ ∈ Km ; thus, by virtue of Lemma 2, we have
(Z
T
(Bu(t), u(t))H dt +
0
≥ α kBk
(Z
T
0
k=m
X
ku(t)k2H dt +
(Dk vk , vk )H
k=1
k=m
X
k=1
kvk k2H
)
)
(17)
.
EJQTDE, 2007 No. 19, p. 12
It follows from (11), (17) and Corollary 1 that
hΛz, zi = −hy(0), ϕ(0)i
e
Z T
k=m
X
=
hBϕ(t), ϕ(t)idt +
hDk ϕ(tm−(k−1) ), ϕ(tm−(k−1) )i
0
≥ α kBk
≥ α kBk
(Z
Z
0
0
t1
k=1
T
2
kϕ(t)k dt +
k=m
X
k=1
kϕ(tk )k
2
)
kϕ(t)k2 dt = kBk αt1 kzk2 = θ kzk2 ,
be
ause there is no impulse before time t1 . Therefore, Λ is
oer
ive on H .
To show that there is a bije
tion from H onto H, it su
es to prove that
Λ + I is a bije
tion from H onto H. Clearly, Λ + I is inje
tive sin
e
hΛz + z, zi = hΛz, zi + hz, zi ≥ (θ + 1) kzk2 .
On the other hand, let y 0 ∈ H , as the form a(f, g) + hf, gi = hΛf, gi + hf, gi
is symmetri
and
oer
ive, then, by virtue of Lax-Milgram Theorem, there
is an element f ∈ H su
h that
a(f, g) + hf, gi = hy 0, gi, for all g ∈ H.
This implies that Λ(H) = H . Thus, for every y 0 ∈ H , there is a unique
z ∈ H su
h that Λ(z) = −y 0 , whi
h
ompletes the proof of Claim 1.
Step 3: Assume that B, Dk ≥ 0, then B ≥ 0,
We dene for ε > 0,
e 2 = B, D
e 2 = Dk .
B
k
e 2 + εI,
βε + B
e 2 + εI,
δε + D
k
and
k
ε
e 2 + εI; D
e 2 + εI, .., D
e 2 + εI).
Bε + (β ε ; δ1ε , .., δm
) = (B
1
m
A
ording to Claim 1, there is ϕ˜0,ε ∈ H su
h that the mild solution yε of (1)
with yε (0) = y 0 satises yε (T ) = 0; where B(h) has been repla
ed by
Bε (ϕ(t),
e
ϕ(t
e 1 ), .., ϕ(t
e k ).., ϕ(t
e m )) ∈ Km .
EJQTDE, 2007 No. 19, p. 13
We obtain from (11) and Corollary 1
−hy(0), ϕ
eε (0)i =
Z
T
hβεε ϕ(t),
e
ϕ
eε (t)idt +
0
and (7) gives
−hy(0), ϕ
eε(0)i ≤ C
(Z
T
0
When
e,
−hy(0), ϕ
eε(0)i ≤ C
(Z
e 2 ϕε (t), ϕε (t)idt +
hB
T
ε
0
hβ ϕε (t), ϕε (t)idt +
k=m
X
hδkε ϕ
eε (tk ), ϕ
eε (tk )i,
k=1
k=m
X
k=1
e 2 ϕε (tm−(k−1) ), ϕε (tm−(k−1) )i
hD
k
hδkε ϕε (tm−(k−1) ), ϕε (tm−(k−1) )i
(20)
It follows at on
e from (18), (19) and (20) that
ε
RT
=
0
RT
0
2
kϕε (t)k dt +
+
RT
0
k=m
P
k=1
kϕε (tk )k
k=m
e ε (t), Bϕ
e ε (t)idt + P hD
e k ϕε (tm−(k−1) ), D
e k ϕε (tm−(k−1) )i
hBϕ
(β ε ϕε (t), ϕε (t))dt +
k=1
k=m
P
k=1
Step 4:
2
hδkε ϕε (tm−(k−1) ), ϕε (tm−(k−1) )) ≤ C 2 .
(21)
A
ording to the estimate (20) the family
bε = Bε (ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm )
2
2
e ϕ(t);
e ϕ
emϕ
= (B
e
D
eε (t1 )..., D
eε (tm )) + ε(ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm ))
ε
1
is
ontained in a bounded subset
Thus, both of the families
√
ε(ϕ
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm ))
are bounded in
Km .
Km .
and
(B ϕ
eε (t); D1 ϕ
eε (t1 )..., Dm ϕ
eε (tm ))
Therefore, we may extra
t a subfamily, say
(B ϕ
eε (t); D1 ϕ
eε (t1 )..., Dm ϕ
eε (tm )) ⇀ h,
)1/2
(19)
k=m
X
k=1
(18)
weakly in
Km .
EJQTDE, 2007 No. 19, p. 14
)1/2
.
.
Then
learly
e2ϕ
e 2ϕ
e 2 eε (tm ))+ε(ϕ
(B
eε (t); D
eε (t); ϕ
eε (t1 )..., ϕ
eε (tm )) ⇀ Bh, weakly in Km .
1 eε (t1 )..., Dm ϕ
Taking the limit as ε → 0, we see that the 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 appli
ation of the foregoing Theorem we give the following example.
Example. One dimensional impulsive S
hrödinger equation :
We
onsider the problem
Step 5:
∂2y
∂y (t, x)
+ i 2 (t, x)
∂t
∂x
y(t, 0)
y (0, x)
∆y (tk , x)
= χω0 u (t, x) , t ∈ (0, T ) \ {tk }k∈σm , x ∈ Ω = (0, 2π),
1
(22)
= y(t, 2π) = 0,
= y0,
= iαk y (tk , x) + χωk vk (x), k ∈ σ1m ,
where
tk+1 − tk > 2π, ωk = (ak1 , ak2 ) ⊂ Ω, k ∈ σ0m , {αk }k∈σm ⊂ R+ .
1
Let
n
o
2
2
H = L2 (Ω, C), Aw(x) = i ∂∂xw2 (x), D(A) = w ∈ H, ∂∂xw2 ∈ H, w(0) = w(π) = 0 ,
and Ik w(x) = iαk w(x) and the
ontrol operator is given by B = χω0 , Dk =
χωk , then the system (22) be
omes an abstra
t formulation of (1). As a
onsequen
e of Theorem 1, the initial state y 0 ∈ L2 (Ω, C) = H of the
solution of (22) is null-
ontrollable at t = T, if and only if, there is C > 0
su
h that
Z
0
y 0(x)ϕ
e
(x)dx
Ω
(Z Z
T
≤ C
0
2
ω0
|ϕ| (t, x)dxdt +
(23)
m Z
X
k=1
2
ωk
|ϕ| (tm−(k−1) , x)
) 21
, ∀ϕ
e0 ∈ L2 (Ω, C),
EJQTDE, 2007 No. 19, p. 15
where
ϕ
e0 (x) = ϕ(T, x)
and
ϕ
∂ϕ(t, x)
∂ 2 ϕ(t, x)
+i
∂t
∂x2
ϕ(t, 0)
ϕ (0, x)
∆ϕ (tk , x)
Here
ϕ
where
is the mild solution of
= 0,
t ∈ (0, T ) \ {tk }k∈σm , x ∈ Ω,
1
= ϕ(t, 2π) = 0,
= ϕ0 (x), x ∈ Ω,
= iαk ϕ (tk , x) , x ∈ Ω, k ∈ σ1m .
is given by
ϕ[k](t)
ϕ[0] (t) ,
ϕ[k] (t) ,
ϕ(t) =
ϕ[m] (t) ,
if
if
if
t ∈ [t0 , t1 )
t ∈ tk , tk+1)
t ∈ [tm , T ] ,
is a solution of the
lassi
al S
hrödinger equation
∂ 2 ϕ[k]
∂ϕ[k] (t, x)
+i
(t, x) = χω0 u (t, x) , t ∈ (t0 , t1 ) , x ∈ Ω = (0, 2π),
∂t
∂x2
ϕ[k] (t, 0) = ϕ[k] (t, 2π) = 0,
ϕ[0] (t0 , x) = ϕ0 (x), x ∈ Ω,
and
∂ 2 ϕ[k]
∂ϕ[k] (t, x)
+i
(t, x) = χω0 u (t, x) , t ∈ (tk , tk+1 ) , x ∈ Ω = (0, 2π),
∂t
∂x2
ϕ[k] (t, 0) = ϕ[k] (t, 2π) = 0,
ϕ[k] (tk , x) = (1 + iαk )ϕ[k−1] (tk , x) , x ∈ Ω, k ∈ σ1m .
Then a standard appli
ation of a variant of Ingham's Inequality [8℄ shows
that
Z
tk+1
tk
Z
w0
ϕ[k] (t, x)dtdx ≥ c(τk , w0 )
for some positive
onstants
m Z
X
k=0
tk+1
tk
Z
w0
c(τk , w0 ) > 0.
ϕ[k] (t, x)dtdx =
Z
Ω
+
ϕ[k] (t , x)dx,
k
Summing up we get
Z
0
T
Z
ω0
|ϕ|2 (t, x)dxdt
m Z
X
+
ϕ[k] (t , x)dx,
≥ c1
k
k=1
Ω
EJQTDE, 2007 No. 19, p. 16
where
c1 = minm c(τk , w0 ) > 0.
k∈σ0
On the other hand, there is a positive
onstant
m Z
X
ωk
k=1
It follows that
Z
0
T
Z
ω0
c2 > 0
su
h that
m Z
X
2 +
ϕ[k] (t , x)dx.
|ϕ| (tm−(k−1) , x) ≥ c2
k
2
k=1
|ϕ|2 (t, x)dxdt
m Z
X
+
k=1
ωk
Ω
|ϕ|2 (tm−(k−1) , x)
m Z
X
2 +
ϕ[k] (t , x)dx
≥ (c1 + c2 )
k
Ω
Zk=1
2
≥ (c1 + c2 ) ϕ[m] (t+
m , x)dx
Ω
Z
= (c1 + c2 ) |ϕ|2 (T, x)dx.
Ω
ϕ
e (x) = ϕ(0,
e x) = ϕ(T, x), then,
Z
m Z
X
0 2
2
2
|ϕ| (tm−(k−1) , x) ≥ m(c1 + c2 ) ϕ
e (x)dx,
|ϕ| (t, x)dxdt +
Now, sin
e
Z
T
0
Z
ω0
0
k=1
ωk
Ω
from whi
h we get
Z
Ω
0 2
ϕ
e (x)dx ≤
1
m(c1 + c2 )
Z
T
0
Z
ω0
2
|ϕ| (t, x)dxdt +
m Z
X
k=1
ωk
2
|ϕ| (tm−(k−1) , x) .
We
on
lude by Cau
hy-S
hwarz inequality that
Z
1/2
Z
Z
0 2
0 2
0
y 0(x)ϕ
y (x)dx
ϕ
e (x)dx
e (x)dx ≤
Ω
Ω
Ω
(R
)1/2 Z Z
0 2
T
|y
|
(x)dx
Ω
≤
|ϕ|2 (t, x)dxdt
m(c1 + c2 )
0
ω0
!1/2
m Z
X
|ϕ|2 (tm−(k−1) , x)dx
,
+
k=1
!
ωk
EJQTDE, 2007 No. 19, p. 17
whi
h establishes the ne
essary and su
ient
ondition of null
ontrollability
stated in Theorem 1.
We
on
lude our paper by a spe
ial
ase when our initial state is an
eigensolution of the following linear operator Γ : H → H dened by
Γ(ψ) =
Z
T
X
−1
2
(s)B X(s)ψds +
0
k=m
X
X −1 (tk )Dk2 X(tk )ψ.
k=1
We have the following result of null-
ontrollability.
Proposition 1 Let λ > 0 be an eigenvalue of Γ with eigenve
tor ψ ∈ H.
Then, the solution y to the problem
/
′
t ∈ (0, T ) {tk }k∈σm ,
y (t) + Ay(t) = − λ1 B 2 (X(t)ψ),
1
1 2
m
D
(X(t
)ψ),
k
∈
σ
∆y(t
)
=
I
y(t
)
−
k
k
k
k
1
λ k
y(0) = ψ,
(24)
satises
y(T ) = 0.
Proof.
Write system (24) into the form
/
′
1 2
t ∈ (0, T ) {tk }k∈σm ,
y (t) + Ay(t) = − λ B (X(t)ψ),
1
+
1 2
m
D
(X(t
)ψ),
k
∈
σ
y(t
)
=
I
y(t
)
−
k
k
k
1
k
λ k
y(0) = ψ.
Therefore, this impulsive problem has a solution whi
h
an be represented
expli
itly as follows
y(t) = X(t)ψ+
Z
0
t
X
1 2
1 2
G(t, tk ) − Dk X(tk )ψ ,
G(t, s) − B (X(s)ψ ds+
λ
λ
0