Professional Documents
Culture Documents
such that
and lim
0
. In
section 3 we set up notation and state the initial boundary-value problems. Section
4 establishes the existence, uniqueness and the smoothness with respect to time
variable of generalized solutions of these problems by the approximating boundary
method and results in Fichera [6].
The main idea is to apply Theorem 2.1 from Section 2 to establish the unique
solvability of the mentioned problem in V
m,1
(Q
T
), where V
m,1
(Q
T
) is a closed
subspace of H
m,1
(Q
T
), and Gardings inequality holds in V
m,1
(Q
T
).
2. Approximating boundary theorem
Theorem 2.1. Let be a bounded domain in R
n
. Then there exists a sequence of
smooth domains
such that
and lim
0
= .
2000 Mathematics Subject Classication. 35D05, 35D10, 35L55, 35M10.
Key words and phrases. Initial boundary-value problems; hyperbolic systems; Cusp domain;
approximating boundary method; generalized solution; existence; uniqueness; smoothness.
c 2008 Texas State University - San Marcos.
Submitted May 8, 2008. Published October 14, 2008.
1
2 N. M. HUNG, V. T. LUONG EJDE-2008/138
Proof. For > 0 arbitrary, set S
= x : dist(x, ) ,
= S
and
is the boundary of
and
by J
h
(x) the mollication of J(x), i.e,
J
h
(x) =
_
R
n
h
(x y)J(y)dy,
where
h
is a mollier. If h <
2
, then J
h
(x) has following properties:
(1) J
h
(x) = 0 if x /
/2
;
(2) 0 J
h
(x) 1;
(3) J
h
(x) = 1 in
2
;
(4) J
h
C
(R
n
).
We now x a constant c (0, 1), and set
c
= x : J
h
(x) > c. It is obvious
that
/2
c
2
. Therefore,
c
and lim
0
c
= ,
c
= x :
J
h
(x) = c.
Assume that K is a critical set of J
h
. By Sards theorem (J
h
(K)) = 0, it
implies that there exists a constant c
0
(0, 1) such that x : J
h
(x) = c
0
is
not a critical set of J
h
.
Denote
c0
= x : J
h
(x) > c
0
and F(x) = J
h
(x) c
0
. If x
0
c0
, then
F(x
0
) = J
h
(x
0
)c
0
= 0 and gradJ
h
(x
0
) ,= 0. This implies there exists a
J
h
xi
(x
0
) ,=
0, without loss generality we can suppose that
J
h
xn
(x
0
) ,= 0. Using the implicit
function theorem, there exists a neighbourhood W of (x
0
1
, . . . , x
0
n1
) in R
n1
a
neighbourhood V of x
0
n
in R and an innitely dierentiable function z : W R
such that x U
x
0
c0
, (where U
x
0 = WV ) if and only if x = (x
1
, . . . , x
n
) U
x
0,
x
n
= z(x
1
, . . . , x
n1
). Hence,
c0
is smooth and lim
0
c0
= . The theorem is
proved.
Suppose domain is not smooth at one isolated point. The denition is a formal
description of domains with a cusp point.
Denition 2.2. We call a bounded domain R
n
a exterior cusp domain if
(1) O , O is a smooth (n 1)-dimensional manifold of the class
C
.
(2) Denote x
= (x
1
, x
2
, . . . , x
n1
), then
x : 0 < x
n
< 1 x = (x
, x
n
) R
n
: [x
[ < x
k
n
, k 1.
Denition 2.3. We call a bounded domain R
n
a interior cusp domain if
(1) O , O is a smooth (n 1)-dimensional manifold of the class
C
.
(2) Denote x
= (x
1
, x
2
, . . . , x
n1
), then
x : 0 < x
n
< 1 x = (x
, x
n
) R
n
: [x
[ > x
k
n
, k 1.
Let
(R
+
), 0 < < 1/4, satisfying 0
1 and
(t) = 1, t < ,
(t) = 0, t 2.
When is a exterior cusp domain, we set
= x R
n
: x
k
n
= (1
([x
[))[x
[ + 2
([x
[), 0 < x
n
< 1.
When is a interior cusp domain, we set
= x R
n
: x
n
= (1 +
([x
[))[x
[
1/k
(2)
1/k
([x
[), x
n
< 1
EJDE-2008/138 INITIAL BOUNDARY-VALUE PROBLEMS 3
Denote
0
= x : x
k
n
= [x
[, 0 < x
n
< 1. If [x
[ 2 then
0
else
. We will denote by
(
0
)
then
= .
3. Statement of the problem
Set Q
T
= (0, T), Q
T
=
= t = .
For each multi index = (
1
, . . . ,
n
) N
n
, set [[ =
1
+ +
n
and D
||
/x
1
1
. . . x
n
n
. Let us consider the partial dierential operator of order 2m
L(x, t) =
m
||,||=0
D
_
a
(x, t)D
_
, (3.1)
where a
= (1)
||+||
a
. a
, and
a
(x, t)
c
0
[[
2m
[[
2
, (3.2)
for all R
n
0, C
s
0 and all (x, t) Q
T
.
In this paper, we use the usual functional spaces:
(Q
T
), C
(Q
T
), L
2
(Q
T
), L
2
(), H
m
(),
H
m
()
(see [3, 4, 6, 7] for the precise denitions). We introduce some functional spaces
which will be used in this paper.
H
m,1
(Q
T
) is the space consisting of all functions u = (u
1
, . . . , u
s
) from L
2
(Q
T
)
which have generalized derivatives up to order m by x and up to order 1 by t and
belonging to L
2
(Q
T
). The norm in this space is dened as follows:
|u|
m,1
=
_
m
||=0
_
Q
T
_
[D
u[
2
+[u
t
[
2
_
dxdt
_
1/2
,
where [D
u[
2
=
s
i=1
[D
u
i
[
2
, [u
t
[
2
=
s
i=1
u
i
/t
2
.
H
m,1
(Q
T
) is the closure in H
m,1
(Q
T
) of the set consisting of all functions in
C
(Q
T
), vanish near S
T
= (0, T).
V
m,1
(Q
T
) is a closed subspace of H
m,1
(Q
T
) having the following properties:
(i) V
m,1
(Q
T
)
H
m,1
(Q
T
);
(ii) Denote by
B
T
(u, v) =
m
||,||=0
(1)
||
_
Q
T
a
(x, t)D
uD
v dxdt
and for t [0, T),
B(u, v)(t) =
m
||,||=0
(1)
||
_
(x, t)D
uD
vdx, u, v V
m,1
(Q
T
),
then for all u V
m,1
(Q
T
) satisfy
(1)
m
B(u, u)(t)
0
|u(., t)|
2
H
m
()
,
0
> 0, t in [0, T). (3.3)
4 N. M. HUNG, V. T. LUONG EJDE-2008/138
(iii) Assume that x
0
, U is a neighbourhood of x
0
in R
n
and denote by
I = (U)(0, T). Let
(I) and v V
m,1
(Q
T
), then v V
m,1
(I).
In the case, V
m,1
(Q
T
) =
H
m,1
(Q
T
) or V
m,1
(Q
T
) = H
m,1
(Q
T
), condi-
tion (iii) is obvious.
Suppose that
T
) = u V
m,1
(Q
T
) :
j
u
= 0, j = 0, 1, . . . , m1
where
1
= x
: [x
[ 2. Then V
m,1
0
(Q
T
) is closed subspace of V
m,1
(Q
T
).
By zero extension of u V
m,1
0
(Q
T
) out of Q
T
, we regard that u V
m,1
(Q
T
);
therefore, from (3.3) we get the inequality
(1)
m
B(u, u)(t)
0
|u(., t)|
2
H
m
(
)
,
0
> 0, for all t [0, T), (3.4)
holds for all u V
m,1
0
(Q
T
).
We have the following results obtained in [6].
Theorem 3.1 ([6]). If f C
(Q
T
) and
k
f
t
k
t=0
= 0, for k = 0, 1, . . . , then there
is the unique function u C
(Q
T
) such that
(1)
m1
L(x, t)u u
tt
= f(x, t) in Q
T
(3.5)
satises initial conditions u(x, 0) = u
t
(x, 0) = 0; moreover, u V
m,1
0
(Q
T
) and
boundary conditions
L(x, t)u, v)
L2(
)
= B(u, v)(t) (3.6)
holds for all v V
m,1
0
(Q
T
) and all t [0, T], where the scalar product is
u, v)
L2(
)
=
_
uvdx.
In the cylinder Q
T
, we consider systems
(1)
m1
L(x, t)u u
tt
= f(x, t), (3.7)
where f L
2
(Q
T
).
Denition 3.2. A function u V
m,1
(Q
T
) is a generalized solution of initial
boundary-value problems for systems (3.7) if it satises following equalities:
(1)
m1
B
T
(u, ) +u
t
,
t
)
L2(Q
T
)
= f, )
L2(Q
T
)
(3.8)
for all test function V
m,1
(Q
T
), (x, T) = 0, and initial conditions holds
u(x, 0) = u
t
(x, 0) = 0. (3.9)
In particular, V
m,1
(Q
T
) =
H
m,1
(Q
T
) or V
m,1
(Q
T
) = H
m,1
(Q
T
), then we have
denition of generalized solutions of the st initial boundary-value problem or sec-
ond initial boundary-value problems for system (3.7).
EJDE-2008/138 INITIAL BOUNDARY-VALUE PROBLEMS 5
4. The uniqueness and existence theorems
In this section, we investigate the unique solvability of initial boundary-value
problems for the systems (3.7). We start with studying the uniqueness theorem.
Theorem 4.1. Assume that for a positive constant ,
sup
_
, [a
[ : (x, t) Q
T
, 0 [[, [[ m
_
.
Then the initial boundary-value problems (3.8), (3.9) for systems (3.7) has no more
than one generalized solution in V
m,1
(Q
T
).
Proof. Suppose that problems (3.8), (3.9) has two solutions u
1
, u
2
in V
m,1
(Q
T
).
Put u = u
1
u
2
, (3.8) implies
(1)
m1
B
T
(u, ) +u
t
,
t
)
L2(Q
T
)
= 0 (4.1)
for all V
m,1
(Q
T
), (x, T) = 0. For b (0, T), we set
(x, t) =
_
0 if t (b, T]
_
t
b
u(x, )d if t [0, b].
It is easy to check that (x, t) V
m,1
(Q
T
),
t
(x, t) = u(x, t), (x, T) = 0. Put
(x, t) in (4.1), we have
(1)
m1
_
b
0
B(
t
, )(t)dt +
_
Q
b
tt
t
dxdt = 0.
Adding this equality with its complex conjugate, we obtain
2 Re(1)
m1
_
b
0
B(
t
, )(t)dt +
_
Q
b
t
[
t
[
2
dxdt = 0. (4.2)
We have
2 Re
_
b
0
(1)
m1
B(
t
, )(t)dt
=
_
0
(1)
m1
t
_
B(, )(t)
_
dt + Re
m
||,||=0
_
Q
b
(1)
||+m
a
t
D
dxdt
= (1)
m1
_
B(, )(b) B(, )(0)
+ Re
m
||,||=0
_
Q
b
(1)
||+m
a
t
D
dxdt.
Since B(, )(b) = 0, it implies
2 Re
_
b
0
(1)
m1
B(
t
, )(t)dt
= (1)
m
B(, )(0) + Re
m
||,||=0
_
Q
b
(1)
||+m
a
t
D
dxdt.
(4.3)
and since
t
(x, 0) = u(x, 0) = 0,
_
_
b
0
t
[
t
[
2
dtdx =
_
[
t
(x, b)[
2
dx. (4.4)
6 N. M. HUNG, V. T. LUONG EJDE-2008/138
Put (4.3) and(4.4) in (4.2), we obtain
(1)
m
B(, )(0) +
_
[
t
(x, b)[
2
dx
= Re
m
||,||=0
_
Q
b
(1)
||+m1
a
t
D
dxdt.
(4.5)
Set
v
(x, t) =
_
0
t
D
(x, b) =
_
0
b
D
u(x, )d =
_
0
b
D
t
(x, )d = D
(x, 0).
Using (3.3) we have
(1)
m
B(, )(0)
0
_
||=0
[D
(x, 0)[
2
dx =
0
_
||=0
[v
(x, b)[
2
dx. (4.6)
From equality (4.5), we use (4.6) and Cauchy inequality, we will obtain
0
_
||=0
[v
(x, b)[
2
dx +
_
[v(x, b)[
2
dx K
1
_
Q
b
m
||=0
[v
(x, t)[
2
dxdt
set K = K
1
/
0
is a constant independent of b, then
_
||=0
[v
(x, b)[
2
dx K
_
Q
b
m
||=0
[v
(x, t)[
2
xdt
By setting
y(b) =
_
||=0
[v
(x, b)[
2
dx
we have
y(b) K
_
b
0
y(t)dt,
The Gronwall-Bellmann inequality implies y(b) = u(x, b) = 0, for all b (0, T) and
all x ; hence, u 0 in Q
T
.
Now, we establish the existence of generalized solutions of mentioned problems
by the approximating boundary method. Firstly, we will prove some following needed
propositions.
Proposition 4.2. If f C
(Q
T
),
k
f
t
k
t=0
= 0, for k = 0, 1, . . . and
sup
_
, [a
[ : (x, t) Q
T
, 0 [[, [[ m
_
,
then the generalized solutions u
V
m,1
0
(Q
T
) of problem (3.8), (3.9), in smooth
cylinders Q
T
, satises the estimate
|u
|
2
m,1
C|f|
2
L2(Q
T
)
where C is a constant independent of .
EJDE-2008/138 INITIAL BOUNDARY-VALUE PROBLEMS 7
Proof. By Theorem 3.1, we have u
(Q
T
) V
m,1
0
(Q
T
), satisfying systems
(3.5), boundary conditions (3.6) and initial conditions u
(x, 0) = u
t
(x, 0) = 0, it is
clear that u
T
.
After multiplying (3.5) by u
t
, integrating on Q
Lu
k
u
t
dxdt
_
0
_
k
u
tt
u
t
dxdt =
_
Q
fu
t
dxdt.
From that, and using equality (3.6), we get
(1)
m
_
0
B(u
, u
t
)(t)dt +
_
0
_
u
k
tt
u
t
dxdt =
_
Q
fu
t
dxdt
Adding this equality with its complex conjugate we obtain
(1)
m
2 Re
_
0
B(u
, u
t
)(t)dt +
_
_
0
t
[u
t
[
2
dtdx = 2 Re
_
Q
fu
t
dxdt. (4.7)
We now change the left terms of (4.7)
2(1)
m
Re
_
0
B(u
, u
t
)(t)dt
= (1)
m
Re
_
0
t
_
B(u
, u
)(t)
_
dt
Re
_
Q
,=0
(1)
||+m
a
t
D
dxdt
= (1)
m
Re B(u
, u
)() Re
_
Q
,=0
(1)
||+m
a
t
D
dxdt.
Using the initial conditions, we get
_
_
0
t
[u
t
[
2
dtdx =
_
[u
t
[
2
dx = |u
t
(x, t)|
L2(
)
Therefore, basing on what has been discussed above, equality (4.7) can be rewritten
as the form
(1)
m
B(u
, u
)() +|u
t
(x, t)|
L2(
)
= Re
_
Q
(1)
||+m
m
||,||=0
a
t
D
dxdt 2 Re
_
Q
fu
t
dxdt.
(4.8)
From (4.8), by using inequality (3.4), and Cauchy inequality, we obtain
0
|u
(x, t)|
2
H
m
(
)
+|u
t
(x, t)|
L2(
)
C
1
_
m
||=0
_
Q
_
[D
[
2
+[u
t
[
2
_
dxdt +|f|
2
L2(Q
T
)
_
.
Therefore,
|u
(x, t)|
2
H
m
(
)
+|u
t
(x, t)|
L2(
)
C
2
_
_
0
_
|u
(., t)|
2
H
m
(
t
)
+|u
t
(., t)|
L2(
t
)
_
dt +|f|
2
L2(Q
T
)
_
8 N. M. HUNG, V. T. LUONG EJDE-2008/138
where C
2
= C
1
/min
0
; 1. Denote
Z() = |u
(x, t)|
2
H
m
(
)
+|u
t
(x, t)|
L2(
)
We get
Z() C
2
_
_
0
Z(t)dt +|f|
2
L2(Q
T
)
_
.
The Gronwall - Bellmann inequality implies
Z() C
2
e
C2
|f|
2
L2(Q
T
)
, (0, T)
By integrating with respect to from 0 to T this inequality, we obtain
|u
|
2
m,1
C|f|
2
L2(Q
T
)
, C = C
2
(e
C2T
1),
where C is a absolute constant.
In the next proposition, we prove result of proposition 4.2 without conditions
k
f
t
k
t=0
= 0, for k = 0, 1, . . . .
Proposition 4.3. If f C
(Q
T
) and
sup
_
, [a
[ : (x, t) Q
T
, 0 [[, [[ m
_
,
then generalized solution u
V
m,1
0
(Q
T
) of problem (3.8), (3.9) in smooth cylinders
Q
T
satises the following estimates
|u
|
2
m,1
C|f|
2
L2(Q
T
)
where C is a constant independent of .
Proof. Denote
f
h
(x, t) =
_
_
0 if (x, t) ,= Q
T
f(x, t) if (x, t) Q
T
, t > h
0, if (x, t) Q
T
, t h
for all h > 0. We will denote by gh
2
the mollication of f
h
; i.e.,
gh
2
(x, t) =
_
R
n+1
h
2
(x y, t )f
h
(y, )dyd,
where
h
is a mollier. Then gh
2
C
(Q
T
), gh
2
0, t <
h
2
and gh
2
f in L
2
(Q
T
).
Applying proposition (4.2) to replace f by gh
2
, we get u
h
as a generalized solution
of the problem (3.8), (3.9) in Q
T
and the following estimate holds
|u
h
|
2
m,1
C|gh
2
|
2
L2(Q
T
)
where C is a absolute constant. As gh
2
f in L
2
(Q
T
), u
h
is a Cauchy sequence
in V
m,1
0
(Q
T
). Therefore, u
h
u
is a generalized
solution of the problem and satises the following estimate
|u
|
2
m,1
C|f|
2
L2(Q
T
)
.
(Q
T
).
EJDE-2008/138 INITIAL BOUNDARY-VALUE PROBLEMS 9
Proposition 4.4. Assume that f C
(Q
T
) and
sup
_
, [a
[ : (x, t) Q
T
, 0 [[, [[ m
_
.
Then problem (3.8), (3.9) in cylinder Q
T
has the generalized solution u V
m,1
(Q
T
)
and
|u|
2
m,1
C|f|
2
L2(Q
T
)
(4.9)
where C is a constant independent of u and f.
Proof. By proposition 4.3, the generalized solution u
V
m,1
0
(Q
T
) of problem (3.8),
(3.9) in the smooth cylinder Q
T
satises the following inequality
|u
|
2
m,1
C|f|
2
L2(Q
T
)
(4.10)
where C is a constant independent of .
Since |f|
2
L2(Q
T
)
|f|
2
L2(Q
T
)
, we have
|u
|
2
m,1
C|f|
2
L2(Q
T
)
.
Set
u
(x, t) =
_
u
(x, t) if (x, t) Q
T
0 if (x, t) Q
T
Q
T
then
| u
|
2
m,1
= |u
|
2
m,1
C|f|
2
L2(Q
T
)
. (4.11)
This means that the set u
>0
is uniform bounded in the space V
m,1
(Q
T
).
So we can take a subsequence, denote also by u
(Q
T
) such that
0 in Q
T
Q
T
,
(x, T) = 0, and
|
|
m,1
0 when 0. Since u
T
, we have
(1)
m1
B
T
(u
) +u
t
,
t
)
L2(Q
T
)
= f,
)
L2(Q
T
)
Rewrite this equality in the form
(1)
m1
B
T
( u
) + u
t
,
t
)
L2(Q
T
)
= f,
)
L2(Q
T
)
Passing to the limit when 0, 0 for the weakly convergent sequence, we get
(1)
m1
B
T
( u
0
, ) + u
0t
,
t
)
L2(Q
T
)
= f, )
L2(Q
T
)
Since V
m,1
(Q
T
) is embedded continuously into L
2
(), the trace sequence u
(x, 0)
of u
(x, 0) = 0, so that u
0
(x, 0) = 0. Hence, u
0
(x, t) is a generalized solution of
problem (3.8), (3.9). Moreover, from (4.11) we have
| u
0
|
2
m,1
lim
0
| u
|
2
m,1
C|f|
2
L2(Q
T
)
.
(Q
T
). Using this proposition and properties of
mollication of f L
2
(Q
T
), we obtain the following theorem.
10 N. M. HUNG, V. T. LUONG EJDE-2008/138
Theorem 4.5. If f L
2
(Q
T
), and
sup
_
, [a
[ : (x, t) Q
T
, 0 [[, [[ m
_
,
then problem (3.8), (3.9) in the cylinder Q
T
has generalized solutions u V
m,1
(Q
T
)
and
|u|
2
m,1
C|f|
2
L2(Q
T
)
,
where C is a constant independent of u and f.
The following theorem shows that the generalized solution u V
m,1
(Q
T
) of
problem (3.8), (3.9) is smooth with respect to time variable t, if the right-hand side
f and coecients of the operator (3.1) are smooth enough with respect to t.
Theorem 4.6. Let
(i)
k
f
t
k
L
2
(Q
T
), k h;
(ii)
k
f
t
k
t=0
= 0, x , k h 1;
(iii) sup
_
k
a
t
k
, k < h : (x, t) Q
T
, 0 [[, [[ m
_
.
Then the generalized solution u V
m,1
(Q
T
) of problem (3.8), (3.9) has generalized
derivatives with respect to t up to order h in V
m,1
(Q
T
) and
_
_
h
u
t
h
_
_
2
m,1
C
h
k=0
_
_
k
f
t
k
_
_
2
L2(Q
T
)
where C is a constant independent of u and f.
This theorem is proved by arguments analogous to those in the proof of propo-
sitions 4.2, 4.3, 4.4 and by induction on h.
Acknowledgements. The authors would like to thank the anonymous referee for
his/her helpful comments and suggestions.
References
[1] Adams, R. A.; Sobolev spaces, Academic Press, New York- San Francisco- London 1975.
[2] Egorov, Y. and Kondratiev, V.; On Spectral Theory of Elliptic Operators, Birkhauser Verlag,
Basel- Boston- Berlin 1991.
[3] Hung, N. M.; The rst initial boundary-value problem for Schrodinger systems in non-smooth
domains, Di. Urav., 34 (1998), pp. 1546-1556 (in Russian).
[4] Hung, N. M.; Asymptotics of solutions to the rst boundary-value problem for strongly hy-
perbolic systems near a conical point of the boundary. Mat. Sb., V.190(1999), N7, 103-126.
[5] Hung, N. M. Son, N. T. K.; Existence and smoothness of solutions to second initial boundary-
nvalue problems for Schrodinger systems in cylinders with non-smooth bases, EJDE, Vol.
2008 (2008), No. 35, pp. 1-11.
[6] Fichera, G.; Existence theorems in elasticity, Springer-verlag Berlin - Heidelberg - New York
1972.
[7] Showalter, R. E.; Hilbert Space method for partial dierential equations, Electronic journal
of dierential equations, Monograph 01, 1994.
Nguyen Manh Hung
Department of Mathematics, Hanoi National University of Education, Hanoi, Vietnam
E-mail address: hungnmmath@hnue.edu.vn
Vu Trong Luong
Department of Mathematics, Taybac University, Sonla, Vietnam
E-mail address: vutrongluong@gmail.com