Volume 2009, Article ID 207873,22pages doi:10.1155/2009/207873
Research Article
Regularity of Parabolic Hemivariational Inequalities with Boundary Conditions
Dong-Gun Park,
1Jin-Mun Jeong,
2and Sun Hye Park
31Mathematics and Materials Physics, Dong-A University, Saha-Gu, Busan 604-714, South Korea
2Division of Mathematical Sciences, Pukyong National University, Busan 608-737, South Korea
3Department of Mathematics, Pusan National University, Busan 609-735, South Korea
Correspondence should be addressed to Jin-Mun Jeong,jmjeong@pknu.ac.kr Received 28 August 2008; Revised 4 December 2008; Accepted 1 January 2009 Recommended by Donal O’Regan
We prove the regularity for solutions of parabolic hemivariational inequalities of dynamic elasticity in the strong sense and investigate the continuity of the solution mapping from initial data and forcing term to trajectories.
Copyrightq2009 Dong-Gun Park et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1. Introduction
In this paper, we deal with the existence and a variational of constant formula for solutions of a parabolic hemivariational inequality of the form:
ux, t Δux, t˙ −divC
εux, t
Ξx, t fx, t inΩ×0,∞, 1.1
ux, t 0 onΓ1×0,∞, 1.2
C ε
ux, t
ν−β·νux, t onΓ0×0,∞, 1.3
Ξx, t∈ϕ ux, t
a.e.x, t∈Ω×0,∞, 1.4
ux,0 u0x inΩ, 1.5
where Ωis a bounded domain in RN with sufficiently smooth boundaryΓ.Let x0 ∈ RN, βx x−x0,Rmaxx∈Ω|x−x0|.The boundaryΓis composed of two piecesΓ0andΓ1, which
are nonempty sets and defined by Γ0:
x∈Γ:βx·ν≥α >0
, Γ1:
x∈Γ:βx·ν≤0
, 1.6
whereν is the unit outward normal vector to Γ. Here ˙u ∂u/∂t,u u1, . . . , uNT is the displacement,εu 1/2{∇u ∇uT} 1/2∂ui/∂xj ∂uj/∂xiis the strain tensor, ϕu ϕ1u1, . . . , ϕNuNT,ϕi is a multi-valued mapping by filling in jumps of a locally bounded functionbi,i1, . . . , N. A continuous mapCfrom the spaceSofN×Nsymmetric matrices into itself is defined by
Cε atrεIbε, forε∈S, 1.7
whereI is the identity ofS, trεdenotes the trace ofε, and a > 0, b > 0. For example, in the caseN 2,Cε E/d1−μ2μtrεI 1−με, whereE >0 is Young’s modulus, 0< μ <1/2 is Poisson’s ratio anddis the density of the plate.
LetH andV be two complex Hilbert spaces. Assume thatV is a dense subspace in Hand the injection ofVintoHis continuous. LetAbe a continuous linear operator fromV intoV∗which is assumed to satisfy G˚arding’s inequality. Namely, we formulated the problem 1.1as
u˙Au−divC εu
Ξ f inΩ×0,∞. 1.8
The existence of global weak solutions for a class of hemivariational inequalities has been studied by many authors, for example, parabolic type problems in1–4, and hyperbolic types in5–7. Rauch8and Miettinen and Panagiotopoulos1,2proved the existence of weak solutions for elliptic one. The background of these variational problems are physics, especially in solid mechanics, where nonconvex and multi-valued constitutive laws lead to differential inclusions. We refer to3,4to see the applications of differential inclusions. Most of them considered the existence of weak solutions for differential inclusions of various forms by using the Faedo-Galerkin approximation method.
In this paper, we prove the existence and a variational of constant formula for strong solutions of parabolic hemivariational inequalities. The plan of this paper is as follows. In Section 2, the main results besides notations and assumptions are stated. In order to prove the solvability of the linear case withΞx, t 0 we establish necessary estimates applying the result of Di Blasio et al.9to1.1–1.5considered as an equation inHas well asV∗. The existence and regularity for the nondegenerate nonlinear systems has been developed as seen in10, Theorem 4.1or11, Theorem 2.6, and the references therein. InSection 3, we will obtain the existence for solutions of1.1–1.5by converting the problem into the contraction mapping principle and the norm estimate of a solution of the above nonlinear equation on L20, T;V∩W1,20, T;V∗⊂C0, T;H. Consequently, ifuis a solution asociated withu0, andf, in view of the monotonicity ofA, we show that the mapping
H×L2
0, T;V∗
u0, f
−→u∈L20, T;V∩C
0, T;H
, 1.9
is continuous.
2. Preliminaries and Linear Hemivariational Inequalities
We denoteξ·ζN
i1ξiζiforξ ξ1, . . . , ξN,ζ ζ1, . . . , ζN∈RNandε·εN
i,j1εijεij for ε,ε∈S. Throughout this paper, we consider
V
u∈
H1ΩN
:u0 onΓ1
, H
L2ΩN , u, v
Ωux·vxdx, u, vΓ0
Γ0
ux·vxdΓ. 2.1
We denoteV∗the dual space ofV,·,·the dual pairing betweenV andV∗.
The norms onV,H, andV∗will be denoted by · ,| · |and · ∗, respectively. For the sake of simplicity, we may consider
u∗≤ |u| ≤ u, u∈V. 2.2
We denote · L2Γ0N by · Γ0. Let Abe the operator associated with a sesquilinear form au, vwhich is defined G˚arding’s inequality
Reau, u≥ω1u2−ω2|u|2, ω1>0, ω2≥0, foru∈V, 2.3
that is,
Au, v au, v, u, v∈V. 2.4
ThenAis a symmetric bounded linear operator fromV intoV∗which satisfies
Au, u≥ω1u2−ω2u2 2.5
and its realization inHwhich is the restriction ofAto
DA {u∈V :Au∈H} 2.6
is also denoted byA. Here, we note thatDAis dense inV. Hence, it is also dense inH. We endow the domainDAofAwith graph norm, that is, foru∈ DA, we defineuDA
|u||Au|. So, for the brevity, we may regard that|u| ≤ u ≤ uDAfor allu∈V. It is known that—Agenerates an analytic semigroupSt t≥0in bothHandV∗.
From the following inequalities
ω1u2≤Reau, u ω2u2≤C|Au||u|ω2|u|2
≤
C|Au|ω2|u|
|u| ≤max C, ω2
uDA|u|, 2.7
it follows that there exists a constantC0>0 such that
u ≤C0u1/2DA|u|1/2. 2.8
So, we may regard asV DA, H1/2,2whereDA, H1/2,2is the real interpolation space betweenDAandH.
Consider the following initial value problem for the abstract linear parabolic type equation:
ut ˙ Aut−divC
εut
ft, t >0, u0 onΓ1×0,∞,
C εu
x, t
ν−β·νux, t onΓ0×0,∞, ux,0 u0x, x∈Ω.
LE
A continuous mapCfrom the spaceSofN×Nsymmetric matrices into itself is defined by Cε atrεIbε, fora >0, b >0, ε∈S. 2.9
It is easily known that
divCεw, v −Cεwν, vΓ0 Cεw, εv, v, w∈V, 2.10 C
ε w1
−C ε
w2
C ε
w1−w2
, w1, w2 ∈V. 2.11
Note that the mapCis linear and symmetric and it can be easily verified that the tensorC satisfies the condition
λ0ε2≤Cε·ε≤λ1ε2, ε∈Sfor someλ0, λ1>0. 2.12
Letλbe the smallest positive constant such that
v2≤λ∇v2 ∀v∈V. 2.13
Simple calculations and Korn’s inequality yield that
λ2|∇u|2≤εu2≤λ3|∇u|2, 2.14 and hence|εu|is equivalent to theH1ΩNnorm onV.Then by virtue of9, Theorem 3.3, we have the following result on the linear parabolic type equationLE.
Proposition 2.1. Suppose that the assumptions stated above are satisfied. Then the following properties hold.
1For anyu0 ∈ V DA, H1/2,2 andf ∈ L20, T;H T > 0, there exists a unique solutionuofLEbelonging to
L2
0, T;DA
∩W1,20, T;H⊂C
0, T;V
2.15
and satisfying
uL20,T;DA∩W1,20,T;H≤C1u0fL20,T;H
, 2.16
whereC1is a constant depending onT.
2Letu0 ∈Handf ∈L20, T;V∗for anyT >0. Then there exists a unique solutionuof LEbelonging to
L20, T;V∩W1,20, T;V∗⊂C
0, T;H
2.17
and satisfying
uL20,T;V∩W1,20,T;V∗ ≤C1u0fL20,T;V∗
, 2.18
whereC1is a constant depending onT.
Proof. 1Letau, v be a bounded sesquilinear form defined inV×V by
au, v Au, v− divC
εu , v
, u, v∈V. 2.19
Noting that by2.10
− divC
εu , u
C
εu , εu
β·νu, u
Γ0, 2.20
and by2.12,2.14, and1.6, λ0λ2u2≤
C εu
, εu
, α|u|2≤
β·νu, u
Γ0, 2.21
it follows that there existω1>0 andω2≥0 such that
Reau, u ≥ω1u2−ω2|u|2, foru∈V. 2.22 LetAbe the operator associated with this sesquilinear form:
Au, v au, v, u, v∈V. 2.23
ThenAis also a symmetric continuous linear operator fromV intoV∗which satisfies Au, u ≥ω1u2−ω2|u|2. 2.24
So we know that—Agenerates an analytic semigroupSt t ≥0in bothHandV∗. Hence, by applying9, Theorem 3.3to the regularity for the solution of the equation:
ut ˙ Aut ft, t >0, u0 onΓ1×0,∞, C
ε
ux, t
ν−β·νux, t onΓ0×0,∞, ux,0 u0x, x∈Ω,
2.25
in the spaceH, we can obtain a unique solutionuofLEbelonging to L2
0, T;DA
∩W1,20, T;H⊂C
0, T;V
2.26
and satisfying the norm estimate2.16.
2It is easily seen that
H
x∈V∗: T
0
AetAx2
∗dt <∞
, 2.27
for the timeT >0. Therefore, in terms of the intermediate theory we can see that
V, V∗1/2,2 H 2.28
and follow the argument of1term by term to deduce the proof of2results.
3. Existence of Solutions in the Strong Sense
This Section is to investigate the regularity of solutions for the following parabolic hemivariational inequality of dynamic elasticity in the strong sense:
ut ˙ Aut−divC ε
ut
Ξx, t ft, t≥0, u0 onΓ1×0,∞,
C ε
ux, t
ν−β·νux, t onΓ0×0,∞ Ξx, t∈ϕ
ux, t
a.e.x, t∈Ω×0,∞, u0 u0.
HIE
Now, we formulate the following assumptions.
HbLetbii1, . . . , N:R → Rbe a locally bounded function verifying
bis≤μi|s| fors∈R, 3.1
whereμi>0.We denote
μmax
μ1, . . . , μN
. 3.2
The multi-valued functionϕi : R → 2Ris obtained by filling in jumps of a function bi:R → Rby means of the functionsbi, bi, bi, bi:R → Ras follows.
bis ess inf
|τ−s|≤biτ, bis ess sup
|τ−s|≤biτ, bis lim
→0bis, bis lim
→0bi
s,
ϕis
bis, bis .
3.3
We denotebξ: b1ξ1, . . . , bNξN,ϕξ: ϕ1ξ1, . . . , ϕNξNforξ ξ1, . . . , ξN∈RN. We will need a regularization ofbidefined by
bnis n ∞
−∞bis−τρnτdτ, 3.4
whereρ∈C0∞−1,1, ρ≥0 and1
−1ρτdτ 1.It is easy to show thatbni is continuous for all n∈Nandbi, bi, bi, bi, bni satisfy the same conditionHbwith possibly different constants ifbi satisfiesHb. It is also known thatbinsis locally Lipschitz continuous ins, that is for anyr >0, there exists a numberLir>0 such that
Hb-1
bni s1
−bn
s2≤Lirs1−s2 3.5
holds for alls1, s2∈Rwith|s1|< r, |s2|< r.We denote
Lr max{L1r, . . . , LNr}. 3.6
The following lemma is from12; Lemma A.5.
Lemma 3.1. Letm∈L10, T;Rsatisfyingmt≥0 for allt∈0, Tanda≥0 be a constant. Letd be a continuous function on0, T⊂Rsatisfying the following inequality:
1
2d2t≤ 1 2a2
t
0
msdsds, t∈0, T. 3.7
Then,
dt≤a t
0
msds, t∈0, T. 3.8
Proof. Let
βt 1
2a2 t
0
msdsds, >0. 3.9
Then
dβt
dt mtdt, t∈0, T, 3.10
and
1
2d2t≤β0t≤βt, t∈0, T. 3.11
Hence, we have
dβt
dt ≤mt√ 2
βt. 3.12
Sincet → βtis absolutely continuous and d
dt
βt 1
2 βt
dβt
dt 3.13
for allt∈0, T, it holds
d dt
βt≤ 1
√2mt, 3.14
that is,
βt≤
β0 1
√2 t
0
msds, t∈0, T. 3.15
Therefore, combining this with3.11, we conclude that
|dt| ≤√ 2
βt≤√ 2
β0
t
0
msds
a t
0
msds, t∈0, T
3.16
for arbitrary >0.
From now on, we establish the following results on the local solvability of the following equation,
ut ˙ Aut−divC ε
ut
−bn ut
ft, t≥0, n∈N, u0 onΓ1×0,∞,
C εu
x, t
ν−β·νux, t onΓ0×0,∞, u0 u0.
HIE-1
Lemma 3.2. Letube a solution of HIE-1andu ∈Br {v∈L20, T;V :||v|| ≤ r}. Then, the following inequality holds, for any 0< t≤T,
ut2u2L20,t;Γ0u2L20,t;V≤c−11 1
2u02f2L20,t;H
eω2Lr1t, 3.17
wherec1min{1/2, α, ω1c0}.
Proof. We remark that from2.11,2.12, it follows that there is a constantc0>0 such that
c0u1t−u2t2≤ C
ε u1t
−C ε
u2t , ε
u1t
−ε u2t
. 3.18
Consider the following equation:
ut ˙ Aut−divC ε
ut
−bn ut
ft, t >0, n∈N. 3.19
Multipying on both sides ofut, we get ut, ut˙
Aut, ut
C ε
ut
, ε
ut
β·νut, ut
bn ut
, ut
ft, ut
, 3.20
and integrating this over0, t, by1.6,2.5,3.18andHb-1, we have 1
2ut2α t
0
uτ2
Γ0dτ ω1c0
t
0
uτ2dτ
≤ 1
2u02
ω2Lrt
0
uτ2dτ t
0
fτ2uτ2 dτ,
3.21
that is,
c1ut2u2L20,t;Γ0u2L20,t;V
≤ 1
2u02f2L20,t;H
ω2Lr 1t
0
uτ2dτ.
3.22
Applying Gronwall lemma, the proof of the lemma is complete.
Theorem 3.3. Assume thatu0∈H,f ∈L20, T;V∗and (Hb). Then, there exists a timeT0>0 such thatHIE-1admits a unique solution
u∈L20, T0;V∩W1,20, T0;V∗∩C
0, T0;H
, 0< T0≤T. 3.23 Proof. Assume that2.5holds forω2/0. Let the constantrsatisfy the following inequality:
c−11 1
2u02f2L20,T;H
eω2Lr1T < r. 3.24
Let us fixT ≥T0>0 such that
max
μ, Lr 4ω2
ω1c0
e2ω2T0−1
<1, 3.25
whereμis given byHb.
InvokingProposition 2.1, for a givenw∈Br{v∈L20, T0;V:v ≤r}, the problem ut ˙ Aut−divC
ε
ut
−bn
wt
ft, t≥0, n∈N, u0 onΓ1×0,∞,
C ε
ux, t
ν−β·νux, t onΓ0×0,∞, u0 u0.
HIE-2
has a unique solutionu∈ L20, T;V∩C0, T;H. To prove the existence and uniqueness of solutions of semilinear typeHIE-1, by virtue ofLemma 3.2, we are going to show that the mapping defined byw→umaps is strictly contractive fromBrinto itself if the condition 3.25is satisfied.
Lemma 3.4. Letu1, u2be the solutions of HIE-2withwreplaced byw1, w2 ∈Br whereBris the ball of radiusrcentered at zero ofL20, T0;V, respectively. Then the following inequality holds:
u1t−u2t≤ t
0
eω2t−sGsds, 3.26
where
Gt Lrw1t−w2t. 3.27
Proof. Let u1, u2 be the solutions of HIE-2 with w replaced by w1, w2 ∈ L20, T0;V, respectively. Then, we have that
d dt
u1t−u2t A
u1t−u2t
− divC
εu1t
−divC ε
u2t −
bn w1t
−bn w2t
, t >0, n∈N.
3.28
Multiplying on both sides ofu1t−u2tand by2.8, we get 1
2 d
dtu1t−u2t2a
u1t−u2t, u1t−u2t
C ε
u1t
−C ε
u2t , ε
u1t
−ε u2t
βx·ν
u1t−u2t
, u1t−u2t
Γ0
− bn
w1t
−bn w2t
, u1t−u2t ,
3.29
and so, by3.18,2.5,Hb, we obtain 1
2 d
dtu1t−u2t2
ω1c0u1t−u2t2
≤ω2u1t−u2t|2Lrw1t−w2tu1t−u2t.
3.30
Putting
Gt Lrw1t−w2t, Ht Gtu1t−u2t 3.31
and integrating3.30over0, t, this yields 1
2|u1t−ut|2 ω1c0
t
0
u1s−u2s2ds≤ω2
t
0
u1s−u2s2ds t
0
Hsds.
3.32
From3.32it follows that d dt
e−2ω2t
t
0
u1s−u2s2ds
2e−2ω2t 1
2u1t−u2t2−ω2
t
0
u1s−u2s2ds
≤2e−2ω2t t
0
Hsds.
3.33
Integrating3.33over0, twe have
e−2ω2t t
0
u1s−u2s2ds≤2 t
0
e−2ω2τ τ
0
Hsds dτ
2 t
0
t
s
e−2ω2τdτHsds2 t
0
e−2ω2s−e−2ω2t
2ω2 Hsds
1 ω2
t
0
e−2ω2s−e−2ω2t
Hsds,
3.34
thus, we get
ω2
t
0
u1s−u2s2ds≤ t
0
e2ω2t−s−1
Hsds. 3.35
From3.32and3.35it follows that 1
2u1t−u2t2 ω1c0
t
0
u1s−u2s2ds
≤ t
0
e2ω2t−sHsds
t
0
e2ω2t−sGsu1s−u2sds,
3.36
which implies 1 2
e−2ω2tu1t−u2t2 ω1c0
e−2ω2t t
0
u1s−u2s2ds
≤ t
0
e−ω2sGse−ω2su1s−u2sds.
3.37
By usingLemma 3.1, we obtain that
e−ω2tu1t−u2t≤ t
0
e−ω2sGsds. 3.38
The proof of lemma is complete.
From3.26and3.36it follows that 1
2u1t−u2t2 ω1c0
t
0
u1s−u2s2ds
≤ t
0
e2ω2t−sGs s
0
eω2s−τGτdτ ds
e2ω2t t
0
e−ω2sGs s
0
e−ω2τGτdτ ds
e2ω2t t
0
1 2
d ds
s
0
e−ω2τGτdτ 2
ds
1 2e2ω2t
t
0
e−ω2τGτdτ 2
≤ 1 2e2ω2t
t
0
e−2ω2τdτ t
0
Gτ2dτ
1
2e2ω2t1−e−2ω2t 2ω2
t
0
Gτ2dτ
Lr2 4ω2
e2ω2t−1t
0
w1s−w2s2ds.
3.39
Starting from the initial valueu0t u0, consider a sequence{un·}satisfying u˙n1t Aun1t−divC
ε
un1t −bn
unt
ft, t≥0 un10 onΓ1×0,∞
Cε
un1x, t
ν−β·νu˙n1x, t, onΓ0×0,∞ un10 u0.
3.40
Then from3.39it follows that 1
2un1t−unt2 ω1c0
t
0
un1s−uns2ds
≤ Lr2 4ω2
e2ω2t−1t
0
uns−un−1s2ds.
3.41
So by virtue of the condition3.25 the contraction principle gives that there exists u·∈L20, T0;Vsuch that
un·−→u· inL20, T0;V, 3.42
and hence, from3.41there existsu·∈C0, T0;Hsuch that un·−→u· inC
0, T0;H
. 3.43
Now, we give a norm estimation of the solution HIE and establish the global existence of solutions with the aid of norm estimations.
Theorem 3.5. Let the assumption (Hb) be satisfied. Assume thatu0 ∈ Handf ∈L20, T;V∗for anyT >0. Then, the solutionuofHIEexists and is unique in
u∈L20, T;V∩W1,20, T;V∗⊂C
0, T;H
. 3.44
Furthermore, there exists a constantC2depending onTsuch that
uL2∩W1,2 ≤C2
1|u0|fL20,T;V∗
. 3.45
Proof. Letw∈Br be the solution of
wt ˙ Awt−divC ε
wt
ft, t≥0, w0 onΓ1×0,∞,
C ε
wx, t
ν−β·νwx, t onΓ0×0,∞, w0 u0.
3.46
Then, since d
dt
ut−wt A
ut−wt
−divC ε
ut
divC ε
wt
−bn ut
, 3.47
by multiplying byut−wt, fromHb,3.18and the monotonicity ofA, we obtain 1
2 d
dtut−wt2
ω1c0ut−wt2≤ω2ut−wt2μutut−wt. 3.48
By integrating on3.48over0, twe have 1
2ut−wt2 ω1c0
t
0
us−ws2ds
≤ω2
t
0
us−ws2dsμ t
0
usus−wsds.
3.49
By the procedure similar to3.39we have 1
2ut−wt ω1c0
t
0
us−ws2ds≤ μ2 4ω2
e2ω2t−1t
0
us2ds. 3.50
Put
M μ2
4ω2
ω1c0
e2ω2t−1
. 3.51
Then it holds
u−wL20,T0;V≤M1/2uL20,T0;V 3.52
and hence, from2.16inProposition 2.1, we have that
uL20,T0;V≤ 1
1−M1/2wL20,T0;V
≤ C0
1−M1/2
1u0fL20,T0;V∗
≤C2
1u0fL20,T0;V∗
3.53
for some positive constantC2. Noting that byHb bnu
L20,T;H≤const·uL20,T;V 3.54 and byProposition 2.1
uW1,20,T;V∗ ≤C1
1u0bnu f
L20,T;V∗
, 3.55
it is easy to obtain the norm estimate ofuinW1,20, T0;V∗satisfying3.45.
Now fromTheorem 3.3it follows that u
T0≤ uC
0,T0,H ≤C2
1u0fL20,T0;V∗
. 3.56
So, we can solve the equation inT0,2T0and obtain an analogous estimate to3.53. Since the condition3.25is independent of initial values, the solution ofHIE-1can be extended the internal0, nT0for a natural numbern, that is, for the initialunT0in the intervalnT0,n 1T0, as analogous estimate3.53holds for the solution in0,n1T0. Furthermore, the estimate3.45is easily obtained from3.53and3.56.
We show thatu,Ξis a solution of the problemHIE.Lemma 3.4andHbgive that bn
ut≤μut≤c, 3.57
and foru0∈H, there exists a unique solutionuofHIEbelonging to L20, T;V∩W1,20, T;V∗⊂C
0, T;H
3.58
and satisfying3.44.
From3.44and3.57, we can extract a subsequence from{un}, still denoted by{un}, such that
un−→uweakly inL20, T;V∩W1,20, T;V∗ 3.59 bn
un
−→Ξweakly inL20, T;H. 3.60
Here, we remark that if V is compactly embedded in H and u ∈ L20, T;V, the following embedding
L20, T;V∩W1,20, T;V∗⊂L20, T;H 3.61
is compact in view13, Theorem 2. Hence, the mapping
un−→ustrongly inL20, T;H. 3.62
By a solution ofHIE-1, we understand a mild solution that has a form
unt Stu0 t
0
St−s divC
ε unt
−bn uns
fs
ds, t≥0, 3.63
so lettingn → ∞and using the convergence results above, we obtain ut ˙ Aut−divC
ε
ut
Ξt ft, 0≤t≤T. 3.64
Now, we show thatΞx, t∈ϕux, ta.e. inQ: Ω×0, T0. Indeed, from3.62we haveuni → ui strongly inL20, T;L2Ωand henceunix, t → uix, ta.e. inQ for each i1,2, . . . , N.Leti∈ {1,2, . . . , N}andη >0.Using the theorem of Lusin and Egoroff, we can choose a subsetω ⊂Qsuch that|ω|< η, ui ∈L2Q\ωanduni → uiuniformly onQ\ω.
Thus, for each >0,there is anM >2/such that unix, t−uix, t<
2 forn > M andx, t∈Q\ω. 3.65 Then, if|unix, t−s|<1/n,we have|uix, t−s|< for alln > Mandx, t∈Q\ω.
Therefore we have bi
uix, t
≤bin
unix, t
≤bi
uix, t
, ∀n > M, x, t∈Q\ω. 3.66
Letφ∈L20, T;L2Ω, φ≥0.Then
Q\ωbi
uix, t
φx, tdx dt≤
Q\ωbni
unix, t
φx, tdx dt
≤
Q\ωbi
uix, t
φx, tdx dt.
3.67
Lettingn → ∞in this inequality and using3.60, we obtain
Q\ωbi
uix, t
φx, tdx dt≤
Q\ωΞix, tφx, tdx dt
≤
Q\ωbi
uix, t
φx, tdx dt,
3.68
whereΞ Ξ1, . . . ,ΞN.Letting → 0in this inequality, we deduce that Ξix, t∈ϕi
uix, t
a.e.inQ\ω, 3.69
and lettingη → 0we get
Ξix, t∈ϕi
uix, t
a.e.inQ. 3.70
This implies thatΞx, t∈ϕux, ta.e. inQ.This completes the proof of theorem.
Remark 3.6. In terms ofProposition 2.1, we remark that ifu0 ∈V DA, H1/2,2 andf ∈ L20, T;Hfor anyT >0 then the solutionuofHIEexists and is unique in
x∈L20, T;DA∩W1,20, T;H⊂C
0, T;V
. 3.71
Futhermore, there exists a constantC2depending onT such that uL2∩W1,2 ≤C2
1u0fL20,T;H
. 3.72
Theorem 3.7. Let the assumption (Hb) be satisfied
1ifu0, f∈V×L20, T;H, then the solutionuofHIEbelongs tou∈L20, T;DA∩ C0, T;Vand the mapping
H×L20, T;H u0, f
−→u∈L2
0, T;DA
∩C
0, T;V
3.73
is continuous.
2letu0, f∈H×L20, T;V∗. Then the solutionuof HIEbelongs tou∈L20, T;V∩ C0, T;Hand the mapping
H×L20, T;V∗u0, f−→u∈L20, T;V∩C
0, T;H
3.74
is continuous.
Proof. 1 It is easy to show that if u0 ∈ V and f ∈ L20, T;H, then u belongs to L20, T;DA∩W1,20, T;H. letu0i, fi ∈ V ×L20, T;Handui ∈ Br be the solution of HIEwithu0n, fiin place ofu0, ffori1,2.Then in view ofProposition 2.1, we have
u1−u2
L20,T;DA∩W1,20,T;H
≤C1u01−u02bnu1−bnu2
L20,T;Hf1−f2
L20,T;H
≤C1u01−u02Lru1−u2
L20,T:Vf1−f2
L20,T;H
.
3.75
Since
u1t−u2t u01−u02 t
0
u˙1s−u˙2s
ds, 3.76
we get
u1−u2
L20,T;H≤√
Tu01−u02 T
√2u1−u2
W1,20,T;H. 3.77
Hence, arguing as in2.8, we get u1−u2
L20,T;V≤C0u1−u21/2
L20,T;DAu1−u21/2
L20,T;H
≤C0u1−u21/2
L20,T;DA
T1/4u01−u021/2 T
√2 1/2
u1−u21/2
W1,20,T;H
≤C0T1/4u01−u021/2u1−u21/2
L20,T;DA0
C0
√T 2
1/2
u1−u2
L20,T;DA∩W1,20,T;H
≤2−7/4C0u01−u022C0
√T 2
1/2
u1−u2
L20,T;DA∩W1,20,T;H.
3.78
Combining3.75with3.78, we obtain u1−u2
L20,T;DA∩W1,20,T;H
≤C1u01−u02f1−f2
L20,T;H
2−7/4C0C1μu01−u02 2C0C1
√T 2
1/2
Lru1−u2
L20,T;DA∩W1,20,T;H.
3.79
Suppose thatu0n, fn → u0, finV ×L20, T;Hand letunandube the solutionsHIE withu0n, fnandu0, f, respectively. Let 0< T1 ≤Tbe such that
2C0C1
T1/√ 21/2
Lr<1. 3.80
Then by virtue of3.79withT replaced byT1we see that
un−→u inL2
0, T1;D A0
∩W1,20, T1;H. 3.81
This implies thatunT1→uT1inV. Hence the same argument shows thatun → uin
L2
T1,min 2T1, T
; D A0
∩W1,2
T1,min 2T1, T
;H
. 3.82
Repeating this process we conclude thatun → uinL20, T;DA0∩W1,20, T;H.
2 If u0, f ∈ H × L20, T;H then u belongs to L20, T;V ∩ C0, T;H from Theorem 3.5. Letu0i, fi∈H×L20, T;Handui ∈Br be the solution ofHIEwithu0i, fi in place ofu0, ffori1,2. MultiplyingHIEbyu1t−u2t, we have
1 2
d
dtu1t−u2t2
ω1c0u1t−u2t2
≤ω2u1t−u2t2bn u1t
−bn
u2tu1t−u2t f1t−f2tu1t−u2t.
3.83
Put
Gt Lru1t−u2tf1t−f2t. 3.84
Then, by the similar argument in3.32, we get 1
2u1t−u2t2 ω1c0
t
0
u1s−u2s2ds
≤ 1
2u01−u022ω2
t
0
u1s−u2s2ds t
0
Gsu1s−u2sds
3.85
and we have that d
dt
e−2ω2t t
0
u1s−u2s2ds
≤2e−2ω2t 1
2u01−u022 t
0
Gsu1s−u2sds
, 3.86
thus, arguing as in3.35we have
ω2
t
0
u1s−u2s2ds
≤ 1 2
e2ω2t−1u01−u022 t
0
e2ω2t−s−1
Gsu1s−u2sds.
3.87
Combining this inequality with3.85it holds that 1
2u1t−x2t2 ω1c0
t
0
u1s−u2s2ds
≤ 1
2e2ω2tu01−u022 t
0
e2ω2t−sGsu1s−u2sds.
3.88
ByLemma 3.1the following inequality 1
2
e−2ω2tu1t−u2t2 ω1c0
e−2ω2t t
0
u1s−u2s2ds
≤ 1
2u01−u022 t
0
e−ω2sGse−ω2su1s−u2sds
3.89
implies that
e−ω2tu1t−u2t≤u01−u02 t
0
e−ω2sGsds. 3.90
Hence, from3.88and3.90it follows that 1
2u1t−u2t2 ω1c0
t
0
u1s−u2s2ds≤ 1
2e2ω2tu01−u022
t
0
e2ω2t−sGseω2su01−u02 s
0
eω2s−τGτds
≤ 1
2e2ω2tu01−u022u01−u02e2ω2t t
0
Gsds Lr2 4ω2
e2ω2t−1t
0
Gs2ds.
3.91
The last term of3.91is estimated as Lr2
e2ω2t−1 4ω2
t
0
2
L2x1s−x2s2k1s−k2s2
ds. 3.92
LetT2< T be such that
ω1c0−Lr2 2ω2
e2ω2T2−1
>0. 3.93
Hence, from3.91and3.92it follows that there exists a constantC >0 such that u1
T2
−u2
T22 T2
0
u1s−u2s2ds≤Cu01−u022 T2
0
f1s−f2s2ds
. 3.94
Suppose u0n, un → u0, fin H×L20, T2;V∗, and let un and ube the solutions HIEwithu0n, fnandu0, f, respectively. Then, by virtue of3.94, we see thatun → u