Abstract. In this paper, we study the Neumann problem for second order
hyperbolic equations without initial data in nonsmooth domains. Our intension was to prove the existence of a generalized solution to this problem by
applying the results of this problem to the initial data.
7 trang |
Chia sẻ: thanhle95 | Lượt xem: 421 | Lượt tải: 0
Bạn đang xem nội dung tài liệu The neumann problem for second order hyperbolic equations in nonsmooth domains, để tải tài liệu về máy bạn click vào nút DOWNLOAD ở trên
JOURNAL OF SCIENCE OF HNUE
Natural Sci., 2012, Vol. 57, No. 3, pp. 53-59
THE NEUMANN PROBLEM FOR SECOND ORDER HYPERBOLIC
EQUATIONS IN NONSMOOTH DOMAINS
Nguyen Van Trung
Hong Duc University, Thanh Hoa City
E-mail: trungnv2000@gmail.com
Abstract. In this paper, we study the Neumann problem for second order
hyperbolic equations without initial data in nonsmooth domains. Our inten-
sion was to prove the existence of a generalized solution to this problem by
applying the results of this problem to the initial data.
Keywords: Neumann problem, nonsmooth domains, hyperbolic equations,
generalized solution.
1. Introduction
We are concerned with boundary value problems and value for hyperbolic
equations in nonsmooth domains. The problems with the Dirichlet boundary condi-
tions and initial data have been investigated in [2,3,4]. The boundary value problem
without initial condition for parabolic equation has been investigated in [5,6,7,8].
The main goal of this work is to prove the existence of a generalized solution of the
Neumann problem without having initial data.
Let Ω be a nonsmooth domain in Rn (n ≥ 2). For h ∈ R, set Qh = Ω ×
(h,∞), Sh = ∂Ω× (h,∞), S = ∂Ω× R, Q = Ω× R. Let
L(x, t; ∂) = −
n∑
i,j=1
∂i(aij(x, t)∂j) +
n∑
i=1
bi(x, t)∂i + c(x, t),
be a second order partial differential operator, where ∂i = ∂xi , and aij , bi, c are
bounded functions from C∞(Q).
We study the present paper hyperbolic equation
utt + L(x, t;D)u = f in Q, (1.1)
with Neumann boundary conditions
∂u(x, t)
∂νL
= 0, (x, t) ∈ S (1.2)
where ∂u(x,t)
∂νL
=
∑n
j=1 ∂ju(x, t)aij(x, t)νi(x, t).
53
Nguyen Van Trung
Let m, k be non-negative integers. Denote by Hm(Ω) the Sobolev spaces in [1].
By the notation (., .) we mean the inner product in L2(Ω). Denote by H−1(Ω) the
dual space of H1(Ω). The pairing between H1(Ω) and H−1(Ω) is denoted by 〈., .〉.
Let X be a Banach space, γ = γ(t) be a real functions. We denote by
L2(a,+∞, γ;X), the space of functions f : (a,+∞)→ X with the norm
‖f‖Lp(a,+∞,γ;X) =
(∫ +∞
a
‖f(t)‖2Xe
−γ(t)tdt
) 1
2
<∞.
Finally, we introduce the Sobolev spaceH1,1∗ (Qa, γ) which consists all functions
u defined on Qa such that u ∈ L2(a,+∞, γ;H1(G)), ut ∈ L2(a,+∞, γ;L2(Ω)) and
utt ∈ L2(a,+∞, γ;H
−1(Ω)) with the norm
‖u‖2
H
1,1
∗ (Qa,γ)
= ‖u‖2L2(a,+∞,γ;H1(Ω)) + ‖ut‖
2
L2(a,+∞,γ;L2(Ω))
+ ‖utt‖
2
L2(a,+∞,γ;H−1(Ω))
.
To simplify notation, we set L2(Q, γ0) = L2(R, γ0;L2(Ω)).
We assume that the operator L is uniformly strong elliptic, that is, there exists
a constant µ0 > 0 such that
n∑
i,j=1
aij(x, t)ξξ ≥ µ0|ξ|
2, ξ ∈ R, (x, t) ∈ Q. (1.3)
Let us introduce the following bilinear form
B(u, v; t) =
∫
Ω
( n∑
i,j=1
(aij(x, t)∂ju∂iv +
n∑
i=1
bi(x, t)∂iuv + c(x, t)uv
)
dx.
Then we have from the classical Green formula
B(u, v; t)− (L(x, t;D)u, v) =
∫
∂Ω
∂u
∂νL
vds, u, v ∈ C∞(Ω).
Definition 1.1. Let f ∈ L2(Q, γ0), then a function u ∈ H1,1∗ (Q, γ) is called a
generalized solution of problem (1.1) - (1.2) if and only if the equality
〈utt, v〉+B(u, v; t) = (f, v), a.e. t ∈ R, (1.4)
holds for all v ∈ H1(Ω).
2. The main results
Theorem 2.1. Suppose that the coefficients of the operator L satisfy
sup{|aij|, |aijt|, |bi|, |c| : i, j = 1, . . . , n; (x, t) ∈ Q} ≤ µ, µ = const.
54
The Neumann problem for second order hyperbolic equations in nonsmooth domains
Then for each γ(t)t > γ0(t)t, t ∈ R, problem (1.1)-(1.2) has a generalized solution u
in the space H1,1∗ (Q, γ) and the following estimate holds
‖u‖2
H
1,1
∗ (Q,γ)
≤ C‖f‖2L2(Q,γ0) (2.1)
here C is a constant independent of u and f .
Proof. To prove the theorem, we construct a family approximate solution uh of
solution u of the problem (1.1)-(1.2). It is known that there is a smooth function χ(t)
which is equal to 1 on [1,+∞), is equal to 0 on (−∞, 0] and assumes value in [0, 1] on
[0; 1] (see [4, Th. 5.5] for more details). Moreover, we can suppose that all derivatives
of χ(t) are bounded. Let h ∈ (−∞, 0] be an integer. Setting fh(x, t) = χ(t−h)f(x, t)
then
fh =
{
f if t ≥ h + 1
0 if t ≤ h.
Moreover, if f ∈ L2(Q, γ0), fh ∈ L2(h,∞, γ;L2(Ω)), fh ∈ L2(Q; γ0) and
‖fh‖2L2(Qh;γ0) = ‖f
h‖2L2(Q;γ0) ≤ ‖f‖
2
L2(Q;γ0). (2.1)
Fixed f ∈ L2(Q; γ0), we consider the following problem in the cylinder Qh:
utt + L(x, t,D)u = f
h(x, t) in Qh, (2.2)
∂u(x, t)
∂νL
= 0, (x, t) ∈ Sh, (2.3)
u |t=h= 0, ut |t=h= 0 on Ω. (2.4)
This is the initial boundary value problem for hyperbolic equations in cylinders Qh.
A function uh ∈ H1,1∗ (Qh, γ) is called a generalized solution of the problem (2.2)-(2.4)
iff uh(., h) = 0, uht (., h) = 0 and the equality
〈uhtt, v〉+B(u
h, v; t) = (fh, v),
holds for a.e. t ∈ (h,∞) and all v ∈ H1(Ω).
Lemma 2.1. For any h fixed, there exists a solution of the problem (2.2)-(2.4).
Firstly, we will prove the existence by Galerkin’s approximating method. Let
{ωk(x)}
∞
k=1 be an orthogonal basis of H1(Ω) which is orthonormal in L2(Ω). Put
uN(x, t) =
N∑
k=1
CNk (t)ωk(x)
where CNk (t), k = 1, . . . , N, is the solution of the following ordinary differential
system:
(uNtt , ωk) +B(u
N , ωk; t) = (f
h, ωk), k = 1, . . . , N, (2.2)
55
Nguyen Van Trung
with the initial conditions
CNk (h) = 0, C
N
kt(h) = 0, k = 1, . . . , N. (2.3)
Let us multiply (2.2) by CNkt(t), then take the sum with respect to k from 1 to
N to arrive at
(uNtt , u
N
t ) +B(u
N , uNt ; t) = (f
h, uNt ).
Since (uNtt , uNt ) = ddt
(
‖uNt ‖
2
L2(Ω)
)
, we get
d
dt
(
‖uNt ‖
2
L2(Ω)
)
+ 2B(uN , uNt ; t) = 2(f
h, uNt ). (2.4)
By the Cauchy-Schwarz inequality, we have
2|(fh, uNt )| ≤ ‖f
h‖2L2(Ω) + ‖u
N
t ‖
2
L2(Ω). (2.5)
Furthermore, we can write
B(uN , uNt ; t) =
∫
Ω
n∑
i,j=1
Aij(x, t)Dju
NDiu
N
t dx+
∫
Ω
n∑
i=1
Bi(x, t)Diu
NuNt +C(x, t)u
NuNt dx
=: B1 +B2. (2.6)
It is easy to see
B1 =
d
dt
(1
2
A[uN , uN , t]
)
−
1
2
∫
Ω
n∑
i,j=1
AijtDju
NDiu
Ndx, (2.7)
for the symmetric bilinear form
A[uN , uN , t] =
∫
Ω
n∑
i,j=1
AijDju
NDiu
Ndx.
The equality (2.7) implies
B1 ≥
d
dt
(1
2
A[uN , uN , t]
)
− µ‖uN‖2H1(Ω), (2.8)
and we note also
|B2| ≤ µ
(
‖uN‖2H1(Ω) + ‖u
N
t ‖
2
L2(Ω)
)
, (2.9)
Combining estimates (2.4)-(2.9), we obtain
d
dt
(
‖uNt ‖
2
L2(Ω)
+ A[uN , uN , t]
)
≤ 2µ
(
‖uNt ‖
2
L2(Ω)
+ ‖uN‖2H1(Ω)
)
+ ‖fh‖2L2(Ω)
≤ µ1
(
‖uNt ‖
2
L2(Ω)
+ A[uN , uN , t]
)
+ ‖fh‖2L2(Ω) (2.10)
56
The Neumann problem for second order hyperbolic equations in nonsmooth domains
where we used (1.3), µ1 = max{2µ, 2µµ0 }.
Now write
η(t) := ‖uNt (., t)‖
2
L2(Ω)
+ A[uN , uN , t]; ξ(t) := ‖fh(., t)‖2L2(Ω), t ∈ [h,∞).
Then (2.10) implies
η′(t) ≤ µ1η(t) + ξ(t), for a.e. t ∈ [h,∞).
Thus the differential form of Gronwall-Belmann’s inequality yields the esti-
mate
η(t) ≤ C
t∫
h
eµ1(t−s)ξ(s)ds, t ∈ [h,∞). (2.11)
We obtain from (2.11) and (1.3) the following estimate
‖uN(., t)‖2L2(Ω) + ‖u
N‖2H1(Ω) ≤ C
t∫
h
eµ1(t−s)‖fh‖2L2(Ω)ds ≤ Ce
µ1t‖fh‖2L2(Q,γ0),
where γ0(t) ∈ [µ1,+∞) for t 0. Now multiplying both
sides of this inequality by e−γ(t)t, then integrating them with respect to t from h to
∞, we obtain
‖uN(., t)‖2L2(h,∞,γ,L2(Ω)) + ‖u
N‖2L2(h,∞,γ,H1(Ω)) ≤ C‖f
h‖2L2(Q,γ0), (2.12)
where γ(t) ∈ [µ1,+∞) for t > 0 and γ(t) ∈ [0, µ1) for t < 0.
Fix any v ∈ H1(Ω), with ‖v‖2H1(Ω) ≤ 1 and write v = v
1 + v2 where v1 ∈
span{ωk}Nk=1 and (v2, ωk) = 0, k = 1, . . . , N, (v2 ∈ span{ωk}Nk=1
⊥
). We have ‖v1‖H1(Ω) ≤
‖v‖H1(Ω) ≤ 1. Utilizing (2.2), we get
(uNtt , v
1) +B(uN , v1; t) = (fh, v1) for a.e. t ∈ [h,+∞).
From uN(x, t) =
N∑
k=1
CNk (t)ωk, we can see that
(uNtt , v) = (u
N
tt , v
1) = (fh, v1)− B(uN , v1; t).
Consequently,
|(uNtt , v)| ≤ C
(
‖fh‖2L2(Ω) + ‖u
N‖2H1(Ω)
)
.
Since this inequality holds for all v ∈ H˚1(Ω), ‖v‖H1(Ω) ≤ 1, the following inequality
will be inferred
‖uNtt ‖
2
H−1(Ω) ≤ C
(
‖fh‖2L2(Ω) + ‖u
N‖2H1(Ω)
)
. (2.13)
57
Nguyen Van Trung
Multiplying (2.13) by e−γ(t)t, then integrating them with respect to t from h to ∞,
and by using (2.12), we obtain
‖uNtt ‖
2
L2(h,∞,γ,H−1(Ω))
≤ C‖fh‖2L2(Q,γ0). (2.14)
Combining (2.12) and (2.14), we arrive at
‖uN‖2
H
1,1
∗ (Qh,γ)
≤ C‖fh‖2L2(Q,γ0) (2.15)
where C is a absolute constant.
From the inequality (2.15), by standard weakly convergent arguments, we
can conclude that the sequence {uN}∞N=1 possesses a subsequence convergent to a
function uh ∈ H1,1∗ (Qh, γ), which is a generalized solution of problem (2.2)-(2.4).
Let k be a integer less than h, denote uk a generalized solution of the problem
(2.2)-(2.4) when we replaced h by k. We define uh in the cylinder Qk by setting
uh(x, t) = 0 for k ≤ t ≤ h. Putting uhk = uh − uk, fhk = fh − fk, so uhk is the
generalized solution of the following problem
uhktt + L(x, t,D)u
hk = fhk(x, t) in Qk, (2.12)
uhk = 0 on Sk, j = 1, ..., m, (2.13)
uhk |t=k= 0, u
hk
t |t=k= 0 on Ω. (2.14)
We have
‖uhk‖2
H
1,1
∗ (Qk,γ)
≤ C‖fh − fk‖2L2(Q,γ0),
and
‖fh − fk‖2L2(Q,γ0) =
h+1∫
k
e−γ0(t)t‖fh − fk‖2L2(Ω)dt.
≤ 2
h+1∫
k
e−γ0(t)t‖f‖2L2(Ω)dt.
Thus
‖uh − uk‖2
H
1,1
∗ (Q(k,∞),γ)
≤ 2C
h+1∫
k
e−γ0(t)t‖f‖2L2(Ω)dt. (2.15)
Since f ∈ L2(Q, γ0),
h∫
k
e−γ0(t)t‖f‖2L2(Ω)dt → 0 when h, k → −∞. It follows that
{uh}−∞h=0 is a Cauchy sequence. So {uh} is convergent to u in H1,1∗ (Qk, γ) (Consider
uh in the cylinder Q by setting uh(x, t) = 0 for all t < h). Because
‖fh − f‖2L2(Q,γ0) =
h+1∫
h
e−γ0(t)t|χ(t− h)− 1|.‖f‖2L2(Ω)dt+
h∫
−∞
e−γ0(t)t‖f‖2L2(Ω)dt,
58
The Neumann problem for second order hyperbolic equations in nonsmooth domains
so
‖fh − f‖2L2(Q,γ0) ≤ 2
h+1∫
h
e−γ0(t)t‖f‖2L2(Ω)dt+
h∫
−∞
e−γ0(t)t‖f‖2L2(Ω)dt,
{fh} is convergent to f in L2(Q, γ0).
Since uh is a generalized solution of the problem (2.2)-(2.4), we have
〈uhtt, v〉+B(u
h, v; t) = (fh, v),
holds for a.e. t ∈ (h,∞) and all v ∈ H1(Ω). Sending h→ −∞, we obtain (1.4). Thus
u is a generalized solution of the problem (1.2)-(1.3). Using (2.15), letting N →∞,
we gain
‖uh‖2
H
1,1
∗ (Qh,γ)
≤ C‖fh‖2L2(Q,γ0).
Sending h→ −∞ we obtain (2.1). The proof of the Theorem is completed.
REFERENCES
[1] Evans LC, 1998. Partial Differential Equations. Grad. Stud. Math., vol. 19,
Amer. Math. Soc., Providence, RI.
[2] N. M. Hung, 1999. Boundary problems for nonstationary systems in domains
with a non-smooth boundary. Doctor dissertation, Mech. Math. Department MSU,
Moscow.
[3] N. M. Hung, V. T. Luong, 2009. Lp-Regularity of solutions to first initial-
boundary value problem for hyperbolic equations in cusp domains. Electron. J.
Differential Equations, Vol. 2009, No. 151, pp.1-18.
[4] N. M. Hung, V. T. Luong, 2009. Regularity of the solution of the first
initial-boundary value problem for hyperbolic equations in domains with cuspi-
dal points on boundary. Boundary Value Problems, Vol. 2009, Art. ID 135730.
doi:10.1155/2009/135730.
[5] Micheal Renardy, Robert C. Rogers, 2004. An Introduction to Partial Differential
Equations. Springer.
[6] Yu. P. Krasovskii, 1992. An estimate of solutions of parabolic problems without
an initial condition. Math. USSR Izvestiya, Vol. 39, No. 2.
[7] N. M. Bokalo, 1990. Problem without initial conditions for nolinear parabolic
equations. UDC 517. 95.
[8] S. D. Ivasishen, 1983. Parabolic Boundary - Value Problems without initial con-
ditions. UDC 917.946.
59