2 (2), 2008, 159-165
©BEYKENT UNIVERSITY
BEHAVIOR OF THE SOLUTION OF
EVOLUTIONARY TYPE EQUATION IN A
NONSMOOTH DOMAIN
R.A.RASULOV
The Sheki Filial of the Institute of Teachers, Sheki, Az5, Republic of Azerbaijan, rasulovr@yahoo.com
Received: 11 July 2008, Accepted: 14 July 2008
ABSTRACT
In this study the estimations of the solution and its derivative with respect to t of the initial boundary value problem for the second order evolutionary type equation in a domain with non smooth boundary are obtained.
EVOLÜSYONAL TÜRDEN DİFERANSİYEL
DENKLEMİN DÜZGÜN OLMAYAN
BÖLGEDEKİ ÇÖZÜMÜNÜN DAVRANIŞI
ÖZET
Makalede ikinci basamaktan evolüsyonal türden denklem için yazılmış başlangıç sınır değer probleminin düzgün olmayan bölgedeki çözümü ve çözümün t 'e göre türevi için bir değerlendirme elde edilmiştir.
Key Words: Evoluationary type equation; Estimations in a nonsmooth domain
1 INTRODUCTION
It is known that the theory of general boundary value problems for a hyperbolic equation or system of equations in cylindrical domain with smooth data has been developed sufficiently well. But, many problems of the mathematical physics are reduced to an evolutionary type equations. However, equations and system of equations of an evolutionary type in a domain with non smooth boundary has not been studied. It should be noted that the similar results for domain with smooth boundary has been obtained in [1].
0
Let W2U ( Qt ) be the Sobolev's space of the functions having the generalized
d2 0
dtdx
ifollowing formula
derivatives ^ ^ — g L2 (QT). Define the scalar product in W~2U (QT ) by the
{u,v)i~u(
ßr) = JQ {u
tv
t+ VuVv + Vu
tVv
t)dxdt.
TWe will consider the class of equation
( a.. ^ (i)
d
2u
^ d . .du
y
— a
i• (t, x ) —
dt , j=! dx
tV
dx. J J d 2 ( 3 ^du
dx
;V
J J dV — — bi J(t,x)—— + —(b(t,x)u)+a(t,x)u = f(t,x)
dtdx
idx
;dt
with following initial and boundary conditions
u(0, x) = p(x), u
t(0, x) = y(x), u i
3Q= 0
(2)Here, QT = Q X ( 0 , T ) where QG Rn, n > 2 , T < ~ and has the
nonsmooth boundary, and the real value functions at J (t, x ) , bi J (t, x ) ,
(i = 1,2,..., n ) , a(t, x) and b(t, x) satisfy the conditions above
n (3)
a, j (t, x) =
aji(t, x), y
l|||
2<
¿ a - , j (t, x ) | | j
<Y$\, Y > 0,
i,
j=1 nda (t x)
|GR
n, X ^ ' '
| |
>0, a
hJ(t,x)
GC^
(QT),a(t, x)
>
0,
i,
j=1 dt nda(t, x)__ . , „1,0,
dt > 0, a(t,X)G C1XU(QT),¿ b , J (t, xllJ > 0, £
D-J
X)| | > 0,
i, J=1 i, j = 1 at bu (t, x) G C1,,X1(QT ), b(t, x) > 0, b(t, x) G C1x0 (QT ). ( 4 ) In addition f (x, t) G L2 (QT ) .The equation (1) is the general form of the equation
d 2
u
Aau
A ( 5 )=
nA— + Audt2
3t
where, n
= const > 0
is a parameter. The eq.(5) describes the distribution of perturbation in viscous medium, the distribution of sound in viscous medium and etc.Now, we will estimate the solution and its derivatives with respect to t in norm W21,1 (QT ) . We approximate the domain Q by domains Q£ with the
smooth boundary such that Q£ c Q , £ > 0 , dQE G Cc , and
lim
£^
c Q£= Q.
Let Q j= Q
£x[0,
T], and let u£ is generalized solution of the problem (1),(2) in domain Q£ .In order to solve the problem (1),(2) we use the following auxiliary results.
Lemma 1. If f G Cc (Qt ), then the problem (1),(2) is a generalized solution
u (x, t) in QT and the following estimates are valid
ll
u(
x,
11
W~21,1(QT ) <C
JI
fllL
2(QT),
I k
( x,
t% % )
< C20
f ( x,
t L2(QT )+1
ft(x,11(6)
(7)
with some constants C1, C2 .
Proof. If we consider the problem (1),(2) in Q , then for the solution of this £
l k (
x, Oil
#2uf
QT>
-c j 4
2 (j
U
« 'IWU(Qj ) -
C4 (ll
f'
1L2 (Qj ) +1 ¿ L (QThen the sequence u£(x, t) has a weak limit in W2U (QT) and, we assume
that for £ ^ 0 the u0 (x, t) is a weak limit of the sequence u£ (x, t) in
W21'1 (QT ). The function will be a generalized solution of the problem (1),(2)
in QT satisfying the estimates (6),(7).
Lemma 2. If
f (x,t), f
t(x,t)e L
2(Q
T),
then the problem (1),(2) is ageneralized solution u(x, t) e W21'1 (QT ) and the following estimates are
valid
l|u(x,
tiW
2U(QT)
-Q |
f (x,
t)|L2(QT),
( 8 )ll
ut
( x,
t I W~U(QT)
- C6(ll
f (x,11L2Q )+l
FT(x,t )l
L2 QT ) )(9)
with some constants C5, C6 .
Proof. Let fh be the average function of the left part of the equation (1), and
uh is solution of the equation (1) responding to fh . The convergence of uh
to u should appear from the estimates (6),(7), and for u (x, t) is valid the inequality (8),(9).
We denote the intersection of the cylinder QT with hyperplane t = const by
.
Assume that the boundary of domain ^ consists of the intersection of two infinitely smooth (n — 1) dimensional hypersurfaces along some infinitly smooth (n — 2) dimensional manifold ae C™ under nonzero angle y ( p ) .
d2
Fix
p e ax[0,T]
and reduce the operator . to canonical^J dxdx. '
J
that these results carry a local character, that these are valid in the case when the DQ consists of finite number surfaces C ™ , in pairs intersecting along manifolds of the class C™ .
Note that at the neighborhood U p of the any point p of manifold
S = ax [0, T ] it is possible to introduce a local system of coordinates in the
special form. If the neighborhood U p is small enough, then there is infinity
differentiable transform translating the domain Q n U to angle. Therefore, for the sake of generality it is possible to assume that at some neighborhood of the point p the boundary surface is the form 0 < ^xf + x?, = r < ^ ,
0 < p < 5 = const, here p is polar angle in plane ( xl, x2) ,
I x ! < - , (i = 3,4,...,n).
Let Vk ( Qt ) is a space of the functions having generalized derivatives with 2
respect to all variables until k. order inclusively in QT with norm
KK (QT ) 2 k
JJL 2>
T s=0 2s—2k d su
dtS0 dx'dz2Here, the function P ( t ) G C ( Q t ) is positive everywhere except on the
manifold S and any neighborhood S coinciding with r( x) .
Theorem 1. Assume that
d
' f
( x:
t;
z )e ^
1(Qt ), 0 < i < i, f = 0
dtldz2 1
at some neighborhood of the
{(x, t), x
e dQ,t = 0}, 0 <a<
2n,
a^n, k
+1^
mn
a
m > 0
set
is
integer. Let for any pair . and ( m2, s2) , here mt > 0, st > 0 are integers,
7m.'
^7IM,' 71m'
such that '- + st — k — 1 < 0, is fulfilled '- + s1 ^ '- + s2.
a a a
2 2Then the generalized solution of the problem (1),(2) has the form
(( x, t )u (x, z, t) =
mn+s (10)
Z(x, z)r
aln
qrKj (z, t
j(p, z, t) + ( ( x , z)v(x, z, t),
j
where
((x, z)
is infinitely differentiable function, withsupp ((x, z) c V
pand the summation is introduced with respect to multi-index
mn
J = (m, S, q, p ) with
a
+ S < k +1,
0
<
q < q0(m, s),
0 < p < p0( m , s), m > 0 , s > 0 , (m, s, p , q) are integer numbers. Here,
q0 (m,0) = 0 , p0 (m,0) = 1, and for all functions Kj (z, t), v ( x , z, t) the
following estimate i—1
Z
j = 0 d jK:dt
J1d
J21
i
< C10
Z
i—1 k —2+ Z Z
H (QT) J-»'-» J—k—2d
J+'K
:dt
11dz
2dx'
(11) L2 (QT ) j = 0d
jf
dtj dz2 vk (QT ) 2is valid, here, Q = ( Q n U ) x [ 0 , T], C10 is constant which depends on
domain QT only, and the functions n j ( p , z, t) are infinitely differentiable
with respect to all variables.
Proof. It is obvious that u(t, x) is the generalized solution of the elliptical type equations from space W / ( Q ) almost all t on Qt which is the
intersection of the cylinder QT with hyperplane t = const. Taking into
account the condition of the theorem and applying the theorem on smoothness of the solution of the Dirichlet's problem for the second order elliptic equation in domain with non smooth boundary [2], we get that u(t, x) e C 5 (Qt),
0 < 5 < 1 . Using the suggested methods in [4] we prove the desired results. The analogous results for linear equation has been obtained in [3].
2 2
REFERENCES
[1] I.V. Suveyka, Differential Equations I. XIX, 2, 1983.
[2] E. Mierzemen, Zur Regular tat verallgemeinerter hosing Ven guesilinearen elliptischen
Differetialgleichungen ZwWEI ter Ordiny in Gebieten mit Eckem. AZ eitschrift fur Analis
und there Anwendungen, Bd 1(4), 1982, pp.69-71.
[3] V.A. Kondratiev, Boundary Problems for Elliptic equations in domain with conical or angle
domains. Notes MMO, pp. 205-292.
[4] I.I. Melnikov, Singularities Solutions of mixed Problems for second order hyperbolic
Equations in Domain with piecewise smooth Boundary. Uspehi. Math. Nauk. V37, 1, 1982,pp.149-150.