POLONICI MATHEMATICI LXV.1 (1996)
On a differential inequality for
a viscous compressible heat conducting capillary fluid bounded by a free surface
by Ewa Zadrzy´ nska and Wojciech M. Zaja ¸czkowski (Warszawa)
Abstract. We derive a global differential inequality for solutions of a free bound- ary problem for a viscous compressible heat concluding capillary fluid. The inequality is essential in proving the global existence of solutions.
1. Introduction. The motion of a viscous compressible heat conducting capillary fluid in a bounded domain Ω
t⊂ R
3(which depends on time t ∈ R
1+) is described by the following system with the boundary and initial conditions (see [2], Chs. 2 and 5):
(1)
̺[v
t+ (v · ∇)v] + ∇p − µ∆v − ν∇ div v = ̺f in e Ω
T,
̺
t+ div(̺v) = 0 in e Ω
T,
̺c
v(θ
t+ v · ∇θ) + θp
θdiv v − κ∆θ
− µ 2
X
3 i,j=1(v
i,xj+ v
j,xi)
2− (ν − µ)(div v)
2= ̺r in e Ω
T,
Tn − σHn = −p
0n on e S
T,
v · n = −φ
t/|∇φ| on e S
T,
∂θ/∂n = θ
1on e S
T,
v|
t=0= v
0, ̺|
t=0= ̺
0, θ|
t=0= θ
0in Ω,
where φ(x, t) = 0 describes S
t, n is the outward vector normal to the bound- ary (i.e. n = ∇φ/|∇φ|), e Ω
T= S
t∈(0,T )
Ω
t×{t}, Ω
0= Ω is an initial domain, S e
T= S
t∈(0,T )
S
t×{t}. Moreover, v = v(x, t) (v = (v
1, v
2, v
3)) is the velocity
1991 Mathematics Subject Classification: 35A05, 35R35, 76N10.
Key words and phrases : free boundary, compressible viscous heat conducting fluid, surface tension.
[23]
of the fluid, ̺ = ̺(x, t) the density, θ = θ(x, t) the temperature, f = f (x, t) the external force field per unit mass, r = r(x, t) the heat sources per unit mass, θ
1= θ
1(x, t) the heat flow per unit surface, p = p(̺, θ) the pressure, µ and ν the viscosity coefficients, κ the coefficient of heat conductivity, c
v= c
v(̺, θ) the specific heat at constant volume, and p
0the external (constant) pressure. We assume that the coefficients µ, ν, κ are constants such that κ > 0, ν ≥
13µ > 0 and moreover c
v> 0, which results from thermodynamic considerations. Further, T = T(v, p) denotes the stress tensor of the form T = {T
ij} = {−pδ
ij+ µ(v
i,xj+ v
j,xi) + (ν − µ)δ
ijdiv v} ≡ {−pδ
ij+ D
ij(v)}, where i, j = 1, 2, 3, and D = D(v) = {D
ij} is the deformation tensor.
Finally, we denote by H the double mean curvature of S
twhich is neg- ative for convex domains and can be expressed in the form
Hn = ∆
St(t)x, x = (x
1, x
2, x
3),
where ∆
St(t) is the Laplace–Beltrami operator on S
t. Let S
tbe determined by x = x(s
1, s
2, t), (s
1, s
2) ∈ R
2. Then we have
∆
St(t) = g
−1/2∂
∂s
αg
−1/2b g
αβ∂
∂s
β= g
−1/2∂
∂s
αg
1/2g
αβ∂
∂s
β(α, β = 1, 2),
where the convention summation over repeated indices is assumed, g = det{g
αβ}
α,β=1,2, g
αβ= x
α· x
β, (x
α= ∂x/∂s
α), {g
αβ} is the inverse matrix to {g
αβ} and {bg
αβ} is the matrix of algebraic complements of {g
αβ}.
Assume that the domain Ω is given. Then by (1.1)
5, Ω
t= {x ∈ R
3: x = x(ξ, t), ξ ∈ Ω}, where x = x(ξ, t) is the solution of the Cauchy problem
∂x
∂t = v(x, t), x|
t=0= ξ ∈ Ω, ξ = (ξ
1, ξ
2, ξ
3).
Hence
(1.2) x = ξ +
t
\
0
u(ξ, s) ds ≡ X
u(ξ, t), where u(ξ, t) = v(X
u(ξ, t), t).
Formula (1.2) yields the relation between the Eulerian x and Lagrangian ξ coordinates. Moreover, the kinematic boundary condition (1.1)
5implies that the boundary S
tis a material surface. Thus, if ξ ∈ S = S
0then X
u(ξ, t) ∈ S
tand S
t= {x : x = X
u(ξ, t), ξ ∈ S}.
By the equation of continuity (1.1)
2and (1.1)
5the total mass of the drop is conserved and the following relation between ̺ and Ω
tholds:
\
Ωt
̺(x, t) dx = M.
In this paper we prove a global differential inequality for problem (1.1) (see Theorem 3.13) which we shall use in the next paper to prove the global- in-time existence of a solution to problem (1.1) close to a constant state. The paper is divided into three sections. In Section 2 we introduce some notation.
In Section 3 we formulate a series of lemmas (see Lemmas 3.1–3.12) which are used to derive the differential inequalities (3.46) and (3.47).
Problem (1.1) is also considered in [12]–[17]. In [12] we prove the local existence of a solution to problem (1.1) in Sobolev–Slobodetski˘ı spaces in two cases: σ = 0 and σ > 0. [14] and [15] are devoted to conservation laws for problem (1.1) in two cases: without surface tension and with it, respectively.
In [14] and [15] we prove that we can choose ̺
0, v
0, θ
0, θ
1, p
0, κ, σ (in the case σ > 0) and the form of the internal energy per unit mass ε = ε(̺, θ) in such a way that var
t|Ω
t| is as small as we need. In [16] the global differential inequality in the case σ = 0, analogous to inequality (3.46) is obtained. [17]
is concerned with the global-in-time existence of solutions to problem (1.1) when σ = 0. Finally, [13] contains a review of results from [12], [1]–[17] and this paper.
In order to prove the main result of the paper, i.e. Theorem 3.13, we apply the same method as in [18], [19] and [16], which is very close to the methods used in [10] and [11] (see also [4]–[7] and [8]).
Papers [18] and [19] of W. M. Zaj¸aczkowski and [9] of V. A. Solonnikov and A. Tani refer to the problem corresponding to (1.1) for a compressible barotropic fluid.
In [8] K. Pileckas and W. M. Zaj¸aczkowski proved the existence of sta- tionary motion of a viscous compressible barotropic fluid bounded by a free surface governed by surface tension.
Finally, the motion of a viscous compressible heat conducting fluid in a fixed domain was considered by A. Matsumura and T. Nishida in [3]–[7] and by A. Valli and W. M. Zaj¸aczkowski in [11].
2. Notation. Let Q = Ω
tor Q = S
t(t ≥ 0). We denote by k · k
l,Q(l ≥ 0) and | · |
p,Q(1 ≤ p ≤ ∞) the norms in the usual Sobolev spaces W
2l(Q) and L
p(Q) spaces, respectively.
Next, we introduce the space Γ
kl(Q) of functions u with the norm kuk
Γlk(Q)
= X
i≤l−k
k∂
tiuk
l−i,Q≡ |u|
l,k,Q, where l > 0 and k ≥ 0.
We shall use the following notation for derivatives of u. If u is a scalar- valued function we denote by D
x,tku or u
x...xt...t| {z }
k times
the vector (D
xα∂
itu)
|α|+i=k.
Similarly, if u = (u
1, u
2, u
3) we denote by D
x,tku or u
x...xt...t| {z }
k times
the vector (D
αx∂
tiu
j)
|α|+i=k, j=1,2,3. Hence
|D
kx,tu| = X
|α|+i=k
|D
αx∂
tiu|.
We use the following lemma:
Lemma 2.1. The following imbedding holds:
W
rl(Q) ⊂ L
αp(Q) (Q ⊂ R
3),
where |α| + 3/r − 3/p ≤ l, l ∈ Z, 1 ≤ p, r ≤ ∞; L
αp(Q) is the space of functions u such that |D
xαu|
p,Q< ∞, and W
rl(Q) is the Sobolev space.
Moreover , the following interpolation inequalities are true:
(2.1) |D
αxu|
p,Q≤ cε
1−κ|D
xlu|
r,Q+ cε
−κ|u|
r,Q,
where κ = |α|/l + 3/(lr) − 3/(lp) < 1, ε is a parameter , c > 0 is a constant independent of u and ε;
(2.2) |D
xαu|
q,∂Q≤ cε
1−κ|D
lxu|
r,Q+ cε
1−κ|u|
r,Q,
where κ = |α|/l + 3/(lr) − 2/(lq) < 1, ε is a parameter , c > 0 is a constant independent of u and ε.
Lemma 2.1 follows from Theorem 10.2 of [1].
3. Global differential inequality. In this section we assume that ν >
1
3
µ. Further, assume that the existence of a sufficiently smooth local solution of problem (1.1) has been proved. To prove the desired differential inequality we assume that Ω
t(t ≤ T , T is the time of local existence) is diffeomorphic to a ball, so S
tcan be described by
|x| ≡ r = R(ω, t), ω ∈ S
1, where S
1is the unit sphere.
Moreover, we consider the motion near the constant state v
e= 0, p
e= p
0+2σ/R
e, θ
e= (1/|Ω|)
T
Ω
θ
0dξ, ̺
e= M/((4π/3)R
3e), where R
eis a solution of the equation
p
M
(4π/3)R
3e, θ
e= p
e.
(Obviously, we assume that the above equation is solvable with respect to R
e> 0.)
Let
p
σ= p − p
0− q
0, ̺
σ= ̺ − ̺
e, ϑ
0= θ − θ
e, ϑ = θ − θ
Ωt,
where q
0= 2σ/R
eand θ
Ωt= (1/|Ω
t|)
T
Ωt
θ dx. Then problem (1.1) takes the form
(3.1)
̺[v
t+ (v · ∇)v] − div T(v, p
σ) = ̺f in Ω
t, t ∈ [0, T ],
̺
t+ div(̺v) = 0 in Ω
t, t ∈ [0, T ],
̺c
v(̺, θ)(ϑ
0t+ v · ∇ϑ
0) + θp
θ(̺, θ) div v
− κ∆ϑ
0− µ 2
X
i,j
(∂
xiv
j+ ∂
xjv
i)
2− (ν − µ)(div v)
2= ̺r in Ω
t, t ∈ [0, T ], T(v, p
σ)n = σ∆
Stx · n n + q
0n on S
t, t ∈ [0, T ],
∂ϑ
0/∂n = θ
1on S
t, t ∈ [0, T ],
where T(v, p
σ) = {µ(∂
xiv
j+ ∂
xjv
i) + (ν − µ)δ
ijdiv v − p
σδ
ij} and T is the time of local existence.
In the sequel we shall use the following Taylor formula for p
σ: p
σ= p(̺, θ) − p(̺
e, θ
e)
(3.2)
= p(̺, θ) − p(̺
e, θ) + p(̺
e, θ) − p(̺
e, θ
e)
= (̺ − ̺
e)
1\
0
p
̺(̺
e+ s(̺ − ̺
e), θ) ds
+ (θ − θ
e)
1\
0
p
θ(̺
e, θ
e+ s(θ − θ
e)) ds
≡ p
1̺
σ+ p
2ϑ
0. We shall also use the formula:
p
σ= p(̺, θ) − p(̺
Ωt, θ
Ωt) (3.3)
= (̺ − ̺
Ωt)
1\
0
p
̺(̺
Ωt+ s(̺ − ̺
Ωt), θ) ds
+ (θ − θ
Ωt)
1\
0
p
θ(̺
Ωt, θ
Ωt+ s(θ − θ
Ωt)) ds
≡ p
3̺
Ωt+ p
4ϑ,
where the function ̺
Ωt= ̺
Ωt(t) is a solution of the problem (3.4) p(̺
Ωt, θ
Ωt) = p
e, ̺
Ωt|
t=0= ̺
eand
̺
Ωt= ̺ − ̺
Ωt.
The functions p
i(i = 1, 2, 3, 4) in (3.4) and (3.5) are positive.
Set
̺
∗= min e
ΩT
̺(x, t), ̺
∗= max e
ΩT
̺(x, t), θ
∗= min
e
ΩT
θ(x, t), θ
∗= max e
ΩT
θ(x, t).
Now we point out the following facts concerning the estimates in Lemmas 3.1–3.12 and Theorem 3.13:
1. We denote by ε small constants and for simplicity we do not distinguish them.
2. We denote by C
1and C
2constants which depend on ̺
∗, ̺
∗, θ
∗, θ
∗, T ,
TT
0
kvk
23,Ωt′
dt
′, kSk
4+1/2, on the parameters which guarantee the existence of the inverse transformation to x = x(ξ, t) and also on the constants of imbedding theorems and Korn inequalities. C
1is always the coefficient of a linear term, while C
2is the coefficient of a nonlinear term. For simplicity we do not distinguish different C
1’s and C
2’s.
3. We denote by c absolute constants which may depend on such param- eters as µ, ν, κ, and by c
0< 1 positive constants which may depend on µ, ν, κ, ̺
∗, ̺
∗, θ
∗, θ
∗. For simplicity we do not distinguish different C’s and C
0’s.
4. We underline that all the estimates are obtained under the assump- tion that there exists a local-in-time solution of (1.1), so all the quantities
̺
∗, ̺
∗, θ
∗, θ
∗, T ,
TT
0
kvk
23,Ωt′
dt
′, kSk
4+1/2are estimated by the data func- tions. Moreover, the existence of the inverse transformation to x = x(ξ, t) is guaranteed by the estimates for the local solutions (see [12]).
Lemma 3.1. Let v, ̺, ϑ
0be a sufficiently smooth solution of (3.1).
Then (3.6) 1
2 d dt
\
Ωt
̺v
2+ p
1̺ ̺
2σ+ ̺
2Ωt+ p
2̺c
vp
θθ ϑ
20dx
+ σ 2
d dt
\
St
g
αβn ·
t
\
0
v
sαdt
′n ·
t
\
0
v
sβdt
′ds + c
0kvk
21,Ωt+ (ν − µ)kdiv vk
20,Ωt+ c
0kϑ
0xk
20,Ωt≤ ε
kp
σk
20,Ωt+ kϑ
0txk
20,Ωt+
t
\
0
v
sdt
′2 0,St
+ kH(·, 0) + 2/R
ek
20,S1+ C
1(kvk
20,Ωt+ krk
20,Ωt+ krk
0,Ωt+ kθ
1k
21,Ωt+ kθ
1k
1,Ωt+ kf k
20,Ωt) + C
2X
1Y
1,
where ε > 0 is sufficiently small, s = (s
1, s
2) and
X
1= kvk
22,Ωt+ k̺
σk
22,Ωt+ kϑ
0k
22,Ωt+ k̺
Ωtk
20,Ωt, Y
1= X
1+
1
\
0
v dt
′2 2,St
.
P r o o f. Multiplying (3.1)
1by v, integrating over Ω
tand using the con- tinuity equation (3.1)
2, boundary condition (3.1)
4and (3.2) we obtain
1 2
d dt
\
Ωt
̺v
2dx + µ
2 E
Ωt(v) + (ν − µ)kdiv vk
20,Ωt(3.7)
−
\
Ωt
p
1̺
σdiv vdx −
\
Ωt
p
2ϑ
0div v dx
− σ
\
St
(∆
Stx · n + 2/R
e)n · v ds =
\
Ωt
̺f v dx,
where E
Ωt(v) =
T
Ωt
P
3i,j=1
(∂
xiv
j+ ∂
xjv
i)
2dx.
First, we consider the sum of the second and third terms on the left-hand side of (3.7). We have
(3.8) µ
2 E
Ωt(v) + (ν − µ)kdiv vk
20,Ωt= µ 2
\
Ωt
(v
i,xj+ v
j,xi)
2dx + (ν − µ)
\
Ωt
(div v)
2dx
= µ 2
X
i6=j
\
Ωt
(v
i,xj+ v
j,xi)
2dx + µ 2
X
i=j
\
Ωt
(v
i,xj+ v
j,xi)
2dx
+ (ν − µ)
\
Ωt
(div v)
2dx
= µ 2
X
i6=j
\
Ωt
(v
i,xj+ v
j,xi)
2dx + µ 2 ε
1X
i=j
\
Ωt
(v
i,xj+ v
j,xi)
2dx
+ µ
2 (1 − ε
1) · 4 X
i
\
Ωt
(v
i,xj)
2dx + (ν − µ)
\
Ωt
(div v)
2dx ≡ I,
where ε
1∈ (0, 1). Since (ξ
1+ ξ
2+ ξ
3)
2≤ 3(ξ
12+ ξ
22+ ξ
32) the last two terms in I are estimated from below by
[ν − (1 + 2ε
1)µ/2]
\
Ωt
(div v)
2dx.
Assuming that ν = (1 + 2ε
1)µ/3 we obtain ε
1=
2µ3(ν − µ/3), so (3.9) I ≥ µ
2 ε
1\
Ωt
(v
i,xj+ v
j,xi)
2dx = 3 4
ν − µ
3
\Ωt
(v
i,xj+ v
j,xi)
2dx.
By the continuity equation (3.1)
2, energy equation (3.1)
3and boundary condition (3.1)
5we have
−
\
Ωt
p
1̺
σdiv v dx =
\
Ωt
p
1̺ ̺
σ(̺
σt+ v · ∇̺
σ) dx (3.10)
= 1 2
d dt
\
Ωt
p
1̺
2σ̺ dx + I
1, where
|I
1| ≤ ε(kv
xk
20,Ωt+ kϑ
0xk
20,Ωt) + C
1(krk
20,Ωt+ kθ
1k
21,Ωt) (3.11)
+ C
2(k̺
σk
41,Ωt+ kvk
21,Ωtk̺
σk
22,Ωt+ kvk
21,Ωtkϑ
0k
22,Ωt+ kvk
22,Ωtk̺
σk
21,Ωt+ k̺
σk
22,Ωtk̺
σk
21,Ωt).
Now, we consider the boundary term in (3.7). In the same way as in Lemma 4.1 of [19] we obtain
(3.12) −
\
St
(∆
Stx · n + 2/R
e)v · n ds
= 1 2
d dt
\
St
g
αβn ·
t
\
0
v
sαdt
′n ·
t
\
0
v
sβdt
′ds + I
1, where
|I
1| ≤ ε
t
\
0
v
sdt
′2 0,St
+ kH(·, 0) + 2/R
ek
20,S1+ kvk
21,Ωt(3.13)
+ C
1kvk
20,Ωt+ C
2t
\
0
v dt
′2 2,St
kvk
22,Ωt.
Next, dividing (3.1)
3by θp
θ, multiplying the result by p
2ϑ
0and inte- grating over Ω
twe get
(3.14)
\
Ωt
p
2̺c
vθp
θ∂
tϑ
202 + v · ∇ ϑ
202
dx +
\
Ωt
p
2ϑ
0div v dx
−
\
Ωt
p
2κ∆ϑ
0θp
θϑ
0dx −
\
Ωt
p
2µ 2θp
θX
i,j
(∂
xjv
i+ ∂
xiv
j)
2ϑ
0dx
−
\
Ωt
p
2(ν − µ) θp
θ(div v)
2ϑ
0dx =
\
Ωt
p
2̺r θp
θϑ
0dx.
Therefore, using the same argument as in Lemma 3.1 of [16], by (3.7)–(3.14) we have
(3.15) 1 2
d dt
̺v
2+ p
1̺
2σ̺ + p
2̺c
vθp
θϑ
20dx
+ σ 2
d dt
\
St
g
αβn ·
t
\
0
v
sαdt
′n ·
t
\
0
v
sβdt
′ds + c
0kvk
21,Ωt+ (ν − µ)kdiv vk
20,Ωt+ c
0kϑ
0xk
20,Ωt≤ ε
t
\
0
v
sdt
′2 0,St
+ kH(·, 0) + 2/R
0k
20,S1+ C
1(kvk
20,Ωt+ krk
20,Ωt+ krk
0,Ωt+ kθ
1k
21,Ωt+ kθ
1k
1,Ωt+ kf k
20,Ωt) + C
2X
1Y
1. Finally, by (3.1)
2and (3.5) we have
(3.16) ∂
t̺
Ωt+ v · ∇̺
Ωt+ ̺ div v + ∂
t̺
Ωt= 0, where in view of (3.4) we get
(3.17) ∂
t̺
Ωt= − p
θΩtp
̺Ωt∂
tθ
Ωt. Using the definition of θ
Ωtwe calculate
∂
tθ
Ωt= 1
|Ω
t|
\
Ωt
ϑ
0tdx + 1
|Ω
t|
\
Ωt
θ div v dx (3.18)
− 1
|Ω
t|
2 \Ωt
θ dx
\Ωt
div v dx . Equation (3.16) and formulas (3.17), (3.18) yield
1 2
d dt
\
Ωt
̺
2Ωtdx ≤ ε(k̺
Ωtk
20,Ωt+ kϑ
0txk
20,Ωt) (3.19)
+ C
1(kv
xk
20,Ωt+ kϑ
0xk
20,Ωt+ krk
20,Ωt+ kθ
1k
21,Ωt) + C
2(kvk
21,Ωtkϑ
0k
22,Ωt+ kvk
42,Ωt+ k̺
Ωtk
41,Ωt+ k̺
σk
22,Ωtkϑ
0k
22,Ωt+ kϑ
0k
42,Ωt).
By (3.3) and the Poincar´e inequality
(3.20) kϑk
0,Ωt≤ kϑ
0xk
0,Ωtwe get
(3.21) k̺
Ωtk
0,Ωt≤ C
1(kϑ
0xk
0,Ωt+ kp
σk
0,Ωt).
The estimates (3.15), (3.19) and (3.21) yield (3.6).
Lemma 3.2. Let v, ̺, ϑ
0be a sufficiently smooth solution of (3.1). Then (3.22) 1
2 d dt
\
Ωt
̺v
t2+ p
σ̺̺ ̺
2σt+ ̺c
vθ ϑ
20tdx
+ σ 2
d dt
\
St
g
αβv
sα· nv
sβ· n ds + c
0kv
tk
21,Ωt+ (ν − µ)kdivv
tk
20,Ωt+ c
0kϑ
0tk
21,Ωt≤ ε(kvk
21,Ωt+ kv
xxk
20,Ωt+ kϑ
0xk
20,Ωt)
+ C
1(|f |
21,0,Ωt+ |r|
21,0,Ωt+ krk
0,Ωt+ |θ
1|
22,1,Ωt+ kθ
1k
1,Ωt) + C
2X
22(1 + X
2),
where X
2= |v|
22,1,Ωt
+ |̺
σ|
22,1,Ωt+ |ϑ
0|
22,1,Ωt.
P r o o f. By the same argument as in Lemma 3.2 of [16] we have (3.23) 1
2 d dt
\
Ωt
̺v
t2+ p
σ̺̺ ̺
2σt+ ̺c
vθ ϑ
20tdx + kv
tk
21,Ωt+ (ν − µ)kdivv
tk
20,Ωt+ kϑ
0tk
21,Ωt−
\
St
[T(v, p
σ)]
,tn · v
tds
≤ ε(k̺
σtk
20,Ωt+ kv
tk
20,Ωt+ kϑ
0tk
20,Ωt+ kϑ
0txk
20,Ωt) + C
1(krk
20,Ωt
+ kr
tk
20,Ωt+ |θ
1|
22,1,Ωt) + C
2X
22(1 + X
2).
By the boundary condition (3.1)
4and the same argument as in Lemma 4.2 of [19] we get
(3.24) −
\
St
[T(v, p
σ)]
,tn · v
tds = σ 2
d dt
\
St
g
αβv
sα· nv
sβ· n ds + I
2, where
(3.25) |I
2| ≤ ε(kv
tk
21,Ωt+ kv
xxk
20,Ωt) + C
1kvk
20,Ωt+ C
2X
22. From the continuity equation (3.1)
2it follows that
(3.26) k̺
σtk
20,Ωt≤ C
1kvk
21,Ωt+ C
2kvk
21,Ωtk̺
σk
22,Ωtand equation (3.1)
3yields
kϑ
0tk
20,Ωt≤ εkϑ
0xtk
20,Ωt(3.27)
+ C
1(kv
xk
20,Ωt+ kϑ
0xk
20,Ωt+ krk
20,Ωt+ kθ
1k
21,Ωt) + C
2(kvk
21,Ωtkϑ
0k
22,Ωt+ kvk
41,Ωt+ k̺
σk
22,Ωtkϑ
0k
22,Ωt+ kϑ
0k
42,Ωt).
Therefore, taking into account (3.23)–(3.27) we obtain (3.22).
Lemmas 3.1 and 3.2 imply
Lemma 3.3. Let v, ̺, ϑ
0be a sufficiently smooth solution of (3.1). Then 1
2 d dt
\
Ωt
̺(v
2+ v
2t) + 1
̺ (p
1̺
2σ+ p
σ̺̺
2σt) + ̺
2Ωt+ ̺c
vθ
p
2p
θϑ
20+ ϑ
20tdx
+ σ 2
d dt
\
St
g
αβn ·
t
\
0
v
sαdt
′n ·
t
\
0
v
sβdt
′+ n · v
sαn · v
sβds + c
0(kvk
21,Ωt+ kv
tk
21,Ωt) + (ν − µ)(kdiv vk
20,Ωt+ kdiv v
tk
20,Ωt) + c
0(kϑ
0xk
20,Ωt+ kϑ
0tk
21,Ωt)
≤ ε
k̺
σk
20,Ωt+ kv
xxk
20,Ωt+
t
\
0
v
sdt
′2 0,St
+ kH(·, 0) + 2/R
ek
20,S1+ C
1(kvk
20,Ωt+ |r|
21,0,Ωt+ krk
0,Ωt+ |θ
1|
22,1,Ωt+ kθ
1k
1,Ωt+ |f |
21,0,Ωt) + C
2[X
1Y
1+ X
22(1 + X
2)].
In order to obtain an inequality for derivatives with respect to x we rewrite problem (3.1) in the Lagrangian coordinates and next we introduce a partition of unity in the fixed domain Ω. Thus we have
(3.28)
ηu
it− ∇
ujT
iju(u, p
σ) = ηg
i, i = 1, 2, 3, η
σt+ η∇
u· u = 0,
ηc
v(η, Γ )γ
0t− κ∇
2uγ
0= ηk − Γ p
Γ(η, Γ )∇
u· u + µ
2 X
3 i,j=1(ξ
kxi∂
ξku
j+ ξ
kxj∂
ξku
i)
2+ (ν − µ)(∇
u· u)
2,
T
u(u, p
σ)n = σ∆
Stx(ξ, t) · n n + q
0n, n · ∇
uγ
0= Γ
1,
where η(ξ, t) = ̺(x(ξ, t), t), u(ξ, t) = v(x(ξ, t), t), g(ξ, t) = f (x(ξ, t), t), Γ (ξ, t) = θ(x(ξ, t), t), γ
0(ξ, t) = ϑ
0(x(ξ, t), t), Γ
1(ξ, t) = θ
1(x(ξ, t), t), n = n(ξ, t) and
T
u(u, p
σ) = {T
uij(u, p
σ)} = {−p
σδ
ij+ µ(∇
uiu
j+ ∇
uju
i) + (ν − µ)δ
ij∇
u· u},
∇
ui= ξ
kxi∂
ξkand div T
u(u, p
σ) = ∇
u· T
u(u, p
σ).
By (3.4) and (3.5) we have respectively
p
σ= p
1η
σ+ p
2γ
0and
p
σ= p
3η
Ωt+ p
4γ,
where η
σ= η − ̺
e, γ
0= Γ − θ
e, η
Ωt= η − ̺
Ωt, p
1=
T1
0
p
η(̺
e+ sη
σ, Γ ) ds, p
2=
T1
0
p
Γ(̺
e, θ
e+ sγ
0) ds, p
3=
T1
0
p
η(̺
Ωt+ sη
Ωt, Γ ) ds, p
4=
T1
0
p
Γ(̺
Ωt, θ
Ωt+ sγ) ds, p
i> 0 (i = 1, 2, 3, 4).
Let us introduce a partition of unity ({ e Ω
i}, {ζ
i}), Ω = S
i
Ω e
i. Let e Ω be one of the e Ω
i,sand ζ(ξ) = ζ
i(ξ) be the corresponding function. If e Ω is an interior subdomain then let e ω be a set such that e ω ⊂ e Ω and ζ(ξ) = 1 for ξ ∈ e ω. Otherwise we assume that e Ω ∩ S = ∅, e ω ∩ S 6= ∅, e ω ⊂ e Ω. Take any β ∈ e ω ∩ S ⊂ e Ω ∩ S = e S∂ and introduce local coordinates {y} associated with {ξ} by the relation
(3.29) y
k= α
kl(ξ
l− β
l), α
3k= n
k(β), k = 1, 2, 3,
where {α
kl} is a constant orthogonal matrix such that e S is described by the equation y
3= F (y
1, y
2), F ∈ W
24−1/2and
Ω = {y : |y e
i| < d, i = 1, 2, F (y
′) < y
3< F (y
′) + d, y
′= (y
1, y
2)}.
Next introduce functions u
′, η
′, Γ
′, γ
′0, γ
′, Γ
1′by means of the formulas u
′i(y) = α
iju
j(ξ)|
ξ=ξ(y), η
′(y) = η(ξ)|
ξ=ξ(y),
Γ
′(y) = Γ (ξ)|
ξ=ξ(y), γ
0′(y) = γ
0(ξ)|
ξ=ξ(y), γ
′(y) = γ(ξ)|
ξ=ξ(y), Γ
1′(y) = Γ
1(ξ)|
ξ=ξ(y),
where ξ = ξ(y) is the inverse transformation to (3.29). Further, we introduce new variables by
z
i= y
i(i = 1, 2), z
3= y
3− e F (y), y ∈ e Ω,
which will be denoted by z = Φ(y), where e F is an extension of F , so e F ∈ W
24. Let b Ω = Φ( e Ω) = {z : |z
i| < d, i = 1, 2, 0 < z
3< d} and b S = Φ( e S).
Define
b
u(z) = u
′(y)|
y=Φ−1(z), η(z) = η b
′(y)|
y=Φ−1(z), Γ (z) = Γ b
′(y)|
y=Φ−1(z), b γ
0(z) = γ
0′(y)|
y=Φ−1(z),
bγ(z) = γ
′(y)|
y=Φ−1(z), Γ b
1(z) = Γ
1(y)|
y=Φ−1(z).
Set b ∇
k= ξ
lxk(ξ)z
iξl∇
zi|
ξ=χ−1(z), where χ(ξ) = Φ(ψ(ξ)) and y = ψ(ξ) is described by (3.29). We also introduce the following notation:
u(ξ) = u(ξ)ζ(ξ), e η(ξ) = η(ξ)ζ(ξ), e Γ (ξ) = Γ (ξ)ζ(ξ), e e γ
0(ξ) = γ
0(ξ)ζ(ξ),
e
γ(ξ) = γ(ξ)ζ(ξ), Γ e
1(ξ) = Γ
1(ξ)ζ(ξ)
for ξ ∈ e Ω, e Ω ∩ S = ∅ and
e u(z) = b u(z)b ζ(z), η(z) = b e η(z)b ζ(z), Γ (z) = b e Γ (z)b ζ(z), e γ
0(z) = bγ
0(z)b ζ(z),
eγ(z) = b γ(z)b ζ(z), Γ e
1(z) = b Γ
1(z)b ζ(z) for z ∈ b Ω = Φ( e Ω), e Ω ∩ S 6= ∅.
Using the above notation we can rewrite problem (3.28) in the following form in an interior subdomain :
ηe u
it− ∇
ujT
uij(e u, e p
σ) = ηeg
i− ∇
ujB
uij(u, ζ) − T
uij(u, p
σ)∇
ujζ
≡ ηeg
i+ k
1, i = 1, 2, 3, η e
σt+ η∇
u· e u = ηu · ∇
uζ ≡ k
2,
ηc
v(η, Γ )eγ
t− κ∇
2ueγ + Γ p
Γ(η, Γ )∇
u· e u
= ηe k +
µ 2
X
3 i,j=1(ξ
kxi∂
ξku
j+ ξ
kxj∂
ξku
i)
2+ (ν − µ)(∇
u· u)
2ζ + Γ p
Γ(η, Γ )u · ∇
uζ
− κ(∇
2uζγ + 2∇
uζ · ∇
uγ) − ηc
v(η, Γ )ζ∂
tθ
Ωt≡ ηe k + k
3, where e p
σ= p
σζ and
B
u(u, ζ) = {B
uij(u, ζ)} = {µ(u
i∇
ujζ + u
j∇
uiζ) + (ν − µ)δ
iju · ∇
uζ}.
In boundary subdomains we have
ηe b u
it− b ∇
jT b
ij(e u, e p
σ) = b ηeg
i− b ∇
jB b
ij(b u, b ζ) − b T
ij(b u, p
σ) b ∇
jζ b (3.30)
≡ b ηeg
i+ k
4i, e
η
σt+ b η b ∇ · e u = b ηb u · b ∇b ζ ≡ k
5, b
ηc
v(b η, b Γ )eγ
t− κ b ∇
2e γ + b Γ p
Γˆ(b η, b Γ ) b ∇ · e u
= b ηe k +
µ 2
X
3 i,j=1( b ∇
ib u
j+ b ∇
jb u
i)
2+ (ν − µ)( b ∇ · b u)
2ζ b
+ b Γ p
Γˆ(b η, b Γ )b u · b ∇b ζ − κ( b ∇
2ζbγ + b b ∇b ζ · b ∇bγ)
− b ηc
v(b η, b Γ )∂
tθ
Ωtb ζ ≡ b ηe k + k
6, T(e b u, e p
σ)b n − σ b ∆
Sˆξ · b b nb nb ζ − σ b ∆
Sˆt
\
0
e u dt
′· bnbn = 2σ R
0ζ b b n + k
7+ k
8, b
n · b ∇eγ = e Γ
1+ k
9,
where k
7i= b B
ij(b u, b ζ)b n
j, k
8= −σ(2 b ∇
Tt
0
b u dt
′∇b b ζ +
Tt
0
b u dt
′∇ b
2ζ) · b b nb n, k
9= b
n · b ∇b ζbγ and b T, b B indicate that the operator ∇
uis replaced by b ∇.
In the considerations below we denote z
1, z
2by τ and z
3by n.
Lemma 3.4. Let v, ̺, ϑ
0be a sufficiently smooth solution of (3.1). Then 1
2 d dt
\
Ωt
̺v
x2+ p
σ̺̺ ̺
2x+ ̺c
vθ ϑ
20xdx
+ σ 2
d dt
\
St
1 2 δ e
αβn ·
t
\
0
v
ppαdt
′n ·
t
\
0
v
ppβdt
′ds
+ σ 2
d dt
\
St
n ·
t
\
0
v
p1p2dt
′2
ds
+ σ 2
d dt
\
St
X
2 i=11 2 n ·
t
\
0
v
pipidt
′+ 2(H(·, 0) + 2/R
e)
2ds
+ c
0(kv
xk
21,Ωt+ k̺
Ωtk
20,Ωt+ k̺
σxk
20,Ωt+ k̺
σtk
20,Ωt+ kϑ
0xxk
20,Ωt)
≤ ε
kv
xtk
20,Ωt+ kϑ
0xtk
20,Ωt+
t
\
0
v dt
′2 0,Ωt
+ kH(·, 0) + 2/R
ek
20,S1+ kR(·, t) − R(·, 0)k
22,S1+ C
1(|v|
21,0,Ωt+ k̺
Ωtk
20,Ωt+ kϑk
20,Ωt+ kϑ
0xk
20,Ωt+ kϑ
0tk
20,Ωt+ kf k
21,Ωt+ krk
21,Ωt+ kθ
1k
22,Ωt) + C
2(X
3Y
3+ kH(·, 0) + 2/R
ek
40,S1),
where the summation over the repeated indices (α, β = 1, 2) and coordinates (x, p = (p
1, p
2)) is assumed, e δ
αβon each boundary part Σ
t= S
t∩{ζ(x) 6= 0}
(ζ belongs to a partition of unity of Ω
t) is of the form e δ
αβ= δ
αβ+ 2ε
αβ, ε
αβ= −F
pαF
pβ(1 + F
p21+ F
p22)
−1, F is the function such that in the local coordinates {y}, P
t
is described by the formula
(3.31) y
i= p
i(i = 1, 2), y
3= F (p
1, p
2, t) and supp ζ is so small that |F
p| ≤ 1/2. Moreover ,
X
3= |v|
22,1,Ωt
+ |̺
σ|
22,1,Ωt+ |ϑ
0|
22,1,Ωt+ k̺
Ωtk
20,Ωt, Y
3= X
3+ kvk
23,Ωt
+ kϑ
0xk
22,Ωt+ kϑk
20,Ωt
+ k̺
Ωtk
20,Ωt+
t
\
0
kvk
23,Ωt′
dt
′.
P r o o f. Similarly to [16] (see the proof of Lemma 3.4) we obtain the following estimate for interior subdomains:
(3.32) 1 2
d dt
\
e
Ω
ηe u
2ξ+ p
σηη eη
2Ωtξ+ ηc
vΓ γ e
ξ2A dξ
+ µ
2 ke u
ξk
21,Ωe + κ
θ
∗keγ
ξξk
20,Ωe + ke η
Ωtk
21,Ωe
≤ ε(ke u
ξξk
20,Ω
e + kη
σξk
20,Ω
e + keγ
ξξk
20,Ω
e ) + C
1(|u|
21,0,Ω
e + kvk
21,Ωt+ kγ
0ξk
20,Ωe + kγk
20,Ω
e + kϑ
0tk
20,Ωt+ kη
Ωt
k
20,e
Ω
+ kegk
20,Ωe + ke kk
20,Ω
e ) + C
2h
X
3( e Ω) +
t
\
0
kuk
23,Ω
e dt
′Y
3( e Ω) + kγk
22,Ω
e (kϑ
0tk
20,Ωt+ kvk
21,Ωt) i , where
X
3( e Ω) = |u|
22,1,Ω
e + |̺
σ|
22,1,Ωe + |γ
0|
22,1,Ωe + kη
Ωtk
20,Ωe , Y
3( e Ω) = X
3( e Ω) + kuk
23,Ω
e + kγk
23,Ω
e + kη
Ωtk
20,Ωe +
t
\
0
kuk
23,Ωe dt
′. Now, we consider subdomains near the boundary. Differentiate (3.30)
1with respect to τ , multiply the result by e u
τJ and integrate over b Ω (J is the Jacobian of the transformation x = x(z)). Next, divide (3.30)
3by b Γ , differentiate the result with respect to τ , multiply by eγ
τJ and integrate over Ω. Hence using Lemma 5.1 of [18] we get b
1 2
d dt
\
Ωˆ
b ηe u
2τ+ p
σ ˆηb
η eη
2Ωtτ+ ηc b
vΓ b eγ
τ2J dz + µ
2 ke u
τk
21,Ωb + κ
θ
∗keγ
τ zk
20,Ω
b −
\
Sˆ
(b nb T(e u, e p
σ))
,τu e
τJ dz
′− κ
\
Sˆ
b n 1
Γ b ∇eγ b
,τ
eγ
τJ dz
′≤ ε(ke u
zzk
20,Ω
b + kb η
σzk
20,Ω
b + kbγ
0zzk
20,Ω
b ) + C
1(|b u|
21,0,Ω
b + kvk
21,Ωt+ kbγ
0τk
20,Ω
b
+ kbγk
20,Ωb + kϑ
0tk
20,Ωt+ kb η
Ωtk
20,Ωb + kegk
21,Ωb + ke kk
21,Ω
b ) + C
2h
X
2( b Ω) +
t
\
0
kb uk
23,Ω
b dt
′Y
2( b Ω) + kbγk
22,Ωb (kϑ
0tk
20,Ωt+ kvk
21,Ωt
) i
,
where X
2( b Ω) and Y
2( b Ω) are defined analogously to X
2( e Ω) and Y
2( e Ω).
Using the boundary condition (3.30)
4we have (3.33) −
\
b
S
(b nb T(e u, e p
σ))
,τu e
τJ dz
′≤ − σ 2
d dt
\
b
S
g
αβn · b
t
\
0
u e
ppαdt
′b n ·
t
\
0
u e
ppβdt
′J dz
′− σ
\
b
S
( b H(·, 0) + 2/R
e)b ζ · e u
pp· bnJ dz
′+ ε
t
\
0
e u dt
′2
2,S
b + ke u
zzk
20,Ω
b + k( b H(·, 0) + 2/R
e)b ζk
20,S
b + kR(·, t) − R(·, 0)k
22,S1+ C
2kb uk
20,Ω
b + kb uk
22,Ω
b
t
\
0
u dt e
′2 3,Ω
b
.
By the boundary condition (3.30)
5we get (3.34) − κ
\
b
S
b n · 1
Γ b ∇bγ b
,τ
e γ
τJ dz
′≤ εkbγ
0zzk
20,Ω
b + C
1(kbγk
20,Ωb + kbγ
0zk
20,Ω
b + k e Γ
1k
22,Ω
b ) + C
2kbγk
22,Ωb
kbγ
0k
22,Ω
b + kbγk
22,Ωb + kb η
σk
22,Ω
b +
t
\
0
b u dt
′2 3,Ω
b
. To obtain (3.33) and (3.34) we have applied the interpolation inequality (2.2) (see Lemma 2.1).
For the quantities 1
2 d dt
\
Ωˆ
p
σ ˆηb
η eη
2ΩtnJ dz + c
0ke η
Ωtnk
20,Ω
b , 1
2 d dt
\
Ωˆ
ηe b u
23nJdz + c
0ke u
3nnk
20,Ωb ,
ke η
Ωtk
20,Ωt, ke u
′zτk
20,Ωb , ke η
Ωtτk
20,Ωb , ke u
′nnk
20,Ωb , 1 2
d dt
\
Ωˆ
ηe b u
2nJ dz, 1
2 d dt
\
Ωˆ
b ηc
vΓ b γ e
n2Jdz + κ
θ
∗keγ
nnk
20,Ωb
we obtain the same estimates as in the proof of Lemma 3.4 of [16]. Therefore, we have
(3.35) 1 2
d dt
\
Ωˆ
ηe b u
2z+ p
σb
ηb
η eη
2Ωtz+ ηc b
vΓ b e γ
z2J dz
+ σ 2
d dt
\
Sˆ
h g
αβb n ·
t
\
0
e
u
ppαdt
′n · b
t
\
0
e u
ppβdt
′+ 2( b H(·, 0) + 2/R
e)b ζ b n ·
t
\
0
e u
ppdt
′i
J dz
′+ µ 2 ke u
zk
21,Ω
b + κ
θ
∗keγ
zzk
20,Ωb + c
0ke η
Ωtk
21,Ωb
≤ ε ke u
zzk
20,Ω
b + kb η
σzk
20,Ω
b + kbγ
0zzk
20,Ω
b + ke u
ztk
20,Ω
b + keγ
0ztk
20,Ω
b +
t
\
0
u dt e
′2
2,S
b + k( b H(·, 0) + 2/R
e)b ζk
20,S
b + kR(·, t) − R(·, 0)k
22,S1+ C
1(|b u|
21,0,Ω
b + kvk
21,Ωt+ kbγ
0τk
20,Ω
b + kbγk
20,Ωb + kϑ
0tk
20,Ωt+ kb η
Ωtk
20,Ωb + kegk
21,Ωb + ke kk
21,Ω
b ) + C
2h
X
2( b Ω) +
t
\
0
kb uk
23,Ω
b dt
′Y
2( b Ω) + kbγk
22,Ωb (kϑ
0tk
20,Ωt+ kvk
21,Ωt) i . We estimate the second term on the left-hand side of (3.35) in the same way as in the proof of Lemma 4.4 of [19]. Going back to the variables ξ in (3.35), next from the resulting estimate and (3.32), after summing over all neighbourhoods of the partition of unity and finally going back to the variables x and using (3.26) we get
(3.36) 1 2
d dt
\
Ωt
̺v
x2+ p
σ̺̺ ̺
2σx+ ̺c
vθ ϑ
20xdx
+ σ 2
d dt
\
St
1 2 e δ
αβn ·
t
\
0
v
ppαdt
′n ·
t
\
0
v
ppβdt
′ds
+ σ 2
d dt
\
St
n ·
t
\
0
v
p1p2dt
′2
ds
+ σ 2
d dt
\
St
X
2 i=11 2 n ·
t
\
0