Introduction
In this paper, we show how a Newtonian viscous fluid can develop long-range memory effects. When a large number of tiny elastic inclusions are periodically suspended in a viscous fluid, the interplay between the elastic vibrations of the particles and the viscous dissipation of the surrounding medium generates convolution-in-time terms in the effective equations. Such emergent behaviors are central to the design of advanced complex fluids, such as locally resonant acoustic metamaterials, where heavy and highly deformable inclusions are used to achieve specific wave-damping or shock-absorbing properties. The resulting macroscopic model is no longer a standard Stokes system, but a Stokes system with memory. Remarkably, the character of this memory—oscillatory and persistent versus decaying—depends on a single dimensionless parameter that encodes the relative size of the inclusions. These phenomena, invisible in any single-phase model, emerge naturally through the mathematical framework of homogenization theory .
The homogenization of Stokes flow in perforated domains with rigid obstacles is by now a classical subject. In a landmark series of papers, Allaire [2 , 3 ] established that, depending on the relative size of the obstacles, the effective model can range from the Stokes equations with a Brinkman correction to those with a Darcy term. Closely related is the work of Cioranescu and Murat [14 ] , who identified the appearance of “strange terms” in the homogenization of elliptic problems in perforated domains. When the obstacles are rigid, the energy stored in the interaction is entirely dissipative, and the effective models remain local in time.
A fundamentally different situation arises when the inclusions are elastic : they can deform, store potential energy, and vibrate. The interplay between the viscous dissipation of the surrounding fluid and the elastic oscillations of the inclusions creates a mechanism for memory effects in the effective medium. The emergence of non-local terms in time from homogenization of systems that are local in time was first identified by Sanchez-Palencia [25 ] , who showed that periodic mixtures of elastic solids and viscous fluids lead to viscoelastic effective media with exponentially decaying memory kernels. Unlike the setting in [25 ] , where both phases are connected, our work considers isolated elastic inclusions. Francfort and Suquet [17 ] demonstrated that composite materials with instantaneous viscous dissipation at the microscale can homogenize into effective media with fading memory, showing how microscopic friction transforms into macroscopic long-range relaxation. In a related direction, Briane and Casado-Díaz [7 , 8 ] proved that the homogenization of magneto-elastodynamic systems with strongly varying magnetic fields and with soft inclusions gives rise to a viscoelastic effective behavior with memory terms, further illustrating how the elastic-viscous coupling at the microscale generates non-local-in-time effects in the macroscopic model. For further developments on the homogenization of viscoelastic models and the rigorous derivation of memory effects, we also refer the reader to [1 , 15 ] .
The problem addressed in the present paper is most closely related to recent works on the homogenization of fluid-structure interaction systems in periodic media.
In [11 ] , Casado-Díaz studied a coupled linearized elastodynamic–Stokes system in a periodic domain where the fluid phase consists of isolated inclusions surrounded by a connected elastic matrix. This geometry models, in a simplified way, biological media such as skin, where cells can be idealized as fluid-filled compartments embedded in an extracellular matrix (see, e.g., [21 ] for related spatial models in mathematical biology, and [12 ] for recent advances in the homogenization of liquid-filled elastomers). The main result of [11 ] is the derivation of a macroscopic viscoelastic equation with a convolution memory kernel acting on the strain history. The memory kernel decays exponentially to zero, the exponential decay being a consequence of the viscous dissipation inside the fluid inclusions.
In the present work we consider the complementary geometry. We study an incompressible viscous fluid filling a three-dimensional bounded domain Ω ⊂ R 3 \Omega\subset\mathbb{R}^{3} Ω ⊂ R 3 , in which a large number of tiny elastic inclusions are periodically distributed. Specifically, the inclusions occupy a family of sets of the form ε k + r ε O \varepsilon k + r_{\varepsilon} O ε k + r ε O with k ∈ Z 3 k \in\mathbb{Z}^{3} k ∈ Z 3 , where ε \varepsilon ε is the period and r ε ≪ ε r_{\varepsilon} \ll\varepsilon r ε ≪ ε is the size of each inclusion. Now it is the fluid phase that is connected and the elastic phase that is disconnected. This inversion of the topology has deep consequences for the character of the memory. We note that this setting was already studied by Fleury [16 ] in the context of acoustic wave propagation. However, under the standard scaling assumptions of [16 ] (densities and compressibilities of both phases of the same order), no resonance-driven memory effects arise and the effective model reduces to a Darcy-type law with a complex permeability (for other rigorous derivations of Darcy’s law in non-periodic suspensions of solid particles, we also refer to [9 ] ). The contrast with our work is precisely the highly contrasted scaling, which triggers a resonant regime leading to qualitatively different memory terms.
The microscopic model couples the linearized elasticity system in the solid inclusions with the Stokes equations in the surrounding fluid, with the standard interface conditions of velocity continuity and stress equilibrium on the boundary of each inclusion. However, the density and the elastic coefficients of the solid must be rescaled in order for the inclusions—whose total volume fraction tends to zero—to have a non-trivial effect in the limit. Specifically, we impose the density of order ε 3 / r ε 3 \varepsilon^{3}/r_{\varepsilon}^{3} ε 3 / r ε 3 and the elasticity tensor of order ε 3 / r ε \varepsilon^{3}/r_{\varepsilon} ε 3 / r ε . This critical scaling ensures that the kinetic energy and the elastic energy stored in each vanishingly small inclusion remain of the same order as the kinetic energy of the fluid.
While this scaling may appear exotic at first sight, it is physically well-motivated by the theory of locally resonant acoustic metamaterials. In the pioneering work of Liu et al. [20 ] , it was shown experimentally that composite materials consisting of heavy metallic cores coated by a thin layer of soft silicone rubber and embedded in an epoxy matrix can exhibit sonic band gaps at unexpected frequencies. The physical mechanism is local resonance: the heavy core, with its large inertia, vibrates against the restoring force of the soft coating, creating internal oscillators that interact with the propagating wave. From the theoretical point of view, Auriault and Boutin [6 ] used multi-scale asymptotic homogenization to show rigorously that in highly contrasted elastic composites—including three-component systems with hard inclusions coated by soft material—long-wavelength macroscopic waves induce local resonances in the soft phase, generating a frequency dependent effective tensorial density different from the real scalar density. At the macroscopic scale, non local effect in time appears. The key feature of these metamaterials, from our perspective, is precisely that the resonant units are heavy (high density) and enclosed in soft elastic coatings (low effective stiffness), which matches our scaling regime. In this way, our critical regime can be understood as a continuum-mechanics idealization of locally resonant metamaterial inclusions: the parameter ε 3 / r ε 3 \varepsilon^{3}/r_{\varepsilon}^{3} ε 3 / r ε 3 represents the high effective density of the composite inclusion, while ε 3 / r ε \varepsilon^{3}/r_{\varepsilon} ε 3 / r ε represents its low effective stiffness.
Our main results are the following. After establishing the well-posedness and energy estimates for the ε \varepsilon ε -dependent system, we pass to the limit using the periodic unfolding method and the techniques developed in [10 ] . The asymptotic behavior depends crucially on the limit parameter γ = lim ε → 0 r ε / ε 3 ∈ [ 0 , + ∞ ] \gamma= \lim_{\varepsilon\to0} r_{\varepsilon}/\varepsilon^{3} \in[0,+\infty] γ = lim ε → 0 r ε / ε 3 ∈ [ 0 , + ∞ ] . We extend our study to all possible values of γ \gamma γ , the most interesting cases being γ ∈ ( 0 , + ∞ ) \gamma\in(0,+\infty) γ ∈ ( 0 , + ∞ ) and γ = ∞ \gamma= \infty γ = ∞ .
For γ = 0 \gamma= 0 γ = 0 , the solid particles are too small to affect the macroscopic flow, and the fluid and solid equations uncouple in the limit. However, for γ ∈ ( 0 , + ∞ ] \gamma\in(0,+\infty] γ ∈ ( 0 , + ∞ ] , we show that the limit behavior can be described by a single closed integro-differential equation for the macroscopic fluid velocity. This equation takes the form of a Stokes system with memory : the standard Stokes equations are modified by a time-convolution term of the velocity.
For the particular case γ = ∞ \gamma= \infty γ = ∞ , the macroscopic velocity v v v of the fluid and the microscopic deformation u ^ \hat{u} u ^ of the inclusions satisfy a coupled two-scale system with the fluid velocity as Dirichlet boundary data for the elastic deformation. For this case γ = ∞ \gamma= \infty γ = ∞ , the explicit form of the convolution kernels can be obtained via spectral decomposition of the Dirichlet elastic operator, illustrating an oscillatory behavior of the memory kernel, which differs fundamentally from the exponentially decaying kernels obtained in the model with a connected solid phase [11 ] . Since the elastic inclusions are isolated and the microscopic equation has no internal viscous dissipation, the microscopic elastic oscillations are purely conservative. As a result, the memory kernel is oscillatory and almost-periodic —it does not decay at infinity. This represents a qualitatively new type of effective behavior: a Newtonian viscous fluid whose macroscopic response acquires a persistent, non-decaying memory due to the purely conservative elastic vibrations of the suspended particles.
In contrast, for the intermediate regime γ ∈ ( 0 , + ∞ ) \gamma\in(0,+\infty) γ ∈ ( 0 , + ∞ ) , the microscopic elastodynamic equation is coupled to a local microscopic Stokes problem. The viscous dissipation of the local Stokes problem damps the elastic oscillations, and we show that the memory kernels decay to zero at infinity (Theorem 5.8 and Remark 5.10 ). Thus, in contrast with the persistent, non-decaying memory of the case γ = ∞ \gamma= \infty γ = ∞ , the effective memory is of fading type.
Organization of the paper
The paper is organized as follows. In Section 2, we collect the notation used throughout the paper. In Section 3, we introduce the microscopic model and the scaling assumptions. Section 4 establishes uniform a priori estimates for the deformation, velocity, and pressure. In Section 5, we use these estimates together with the periodic unfolding method to pass to the limit and derive the homogenized system, first as a coupled two-scale problem and then as a closed integro-differential equation with memory for the macroscopic fluid velocity.
Notation
For two vectors a , b ∈ R 3 a,b \in\mathbb{R}^{3} a , b ∈ R 3 , we write a ⋅ b = ∑ i = 1 3 a i b i a \cdot b = \sum_{i=1}^{3} a_i b_i a ⋅ b = ∑ i = 1 3 a i b i . For two matrices ξ , η ∈ R 3 × 3 \xi,\eta\in\mathbb{R}^{3 \times3} ξ , η ∈ R 3 × 3 , we write ξ : η = trace ( ξ η T ) = ∑ i , j = 1 3 ξ i j η i j \xi: \eta= \operatorname{trace}(\xi\eta^{T}) = \sum_{i,j=1}^{3} \xi_{ij}\eta_{ij} ξ : η = trace ( ξ η T ) = ∑ i , j = 1 3 ξ ij η ij .
For a vector function u u u , we denote by e ( u ) e(u) e ( u ) the symmetric part of its gradient, i.e., e ( u ) = ( D u + D u T ) / 2 e(u) = (Du + Du^{T})/2 e ( u ) = ( D u + D u T ) /2 . If there is an ambiguity on the variable of derivation, we use e x ( u ) e_x(u) e x ( u ) or e y ( u ) e_y(u) e y ( u ) .
For a measurable set U ⊂ R 3 U \subset\mathbb{R}^{3} U ⊂ R 3 , we denote by χ U \chi_U χ U its characteristic function. For a measurable function u u u in U U U , we denote by u χ U u\chi_U u χ U its extension by zero to R 3 \mathbb{R}^{3} R 3 .
For two normed spaces X 1 , X 2 X_1, X_2 X 1 , X 2 , we denote by L ( X 1 ; X 2 ) \mathcal{L}(X_1; X_2) L ( X 1 ; X 2 ) the space of continuous linear functions from X 1 X_1 X 1 into X 2 X_2 X 2 . When X 1 = X 2 X_1 = X_2 X 1 = X 2 , we simply write L ( X 1 ) \mathcal{L}(X_1) L ( X 1 ) .
For g ∈ H − 1 / 2 ( Γ ) g \in H^{-1/2}(\Gamma) g ∈ H − 1/2 ( Γ ) and u ∈ H 1 / 2 ( Γ ) u \in H^{1/2}(\Gamma) u ∈ H 1/2 ( Γ ) we use the notation ∫ Γ g u d σ \int_{\Gamma} gu\,d\sigma ∫ Γ g u d σ to mean the duality product ⟨ g , u ⟩ H − 1 / 2 ( Γ ) , H 1 / 2 ( Γ ) \langle g,u\rangle_{H^{-1/2}(\Gamma),H^{1/2}(\Gamma)} ⟨ g , u ⟩ H − 1/2 ( Γ ) , H 1/2 ( Γ ) .
For an open set U ⊂ R 3 U \subset\mathbb{R}^{3} U ⊂ R 3 , we denote by D ( U ) \mathcal{D}(U) D ( U ) the space of C ∞ C^{\infty} C ∞ -functions with compact support in U U U , and by D ′ ( U ) \mathcal{D}'(U) D ′ ( U ) its topological dual, i.e., the space of distributions in U U U .
D 1 , 2 ( R 3 ∖ O ‾ ) D^{1,2}(\mathbb{R}^{3} \setminus\overline{O}) D 1 , 2 ( R 3 ∖ O ) denotes the space of functions w ∈ L 6 ( R 3 ∖ O ‾ ) w \in L^{6}(\mathbb{R}^{3} \setminus\overline{O}) w ∈ L 6 ( R 3 ∖ O ) such that ∇ w ∈ L 2 ( R 3 ∖ O ‾ ) \nabla w \in L^{2}(\mathbb{R}^{3} \setminus\overline{O}) ∇ w ∈ L 2 ( R 3 ∖ O ) .
We define κ : R 3 → Z 3 \kappa: \mathbb{R}^3 \to\mathbb{Z}^3 κ : R 3 → Z 3 by the condition κ ( x ) ∈ Z 3 \kappa(x) \in\mathbb{Z}^3 κ ( x ) ∈ Z 3 , x − κ ( x ) ∈ Y x-\kappa(x) \in Y x − κ ( x ) ∈ Y , a.e. x ∈ R 3 x \in\mathbb{R}^3 x ∈ R 3 . For ε > 0 \varepsilon> 0 ε > 0 , ε κ ( x / ε ) \varepsilon\kappa(x/\varepsilon) ε κ ( x / ε ) is the center of the cube C k ε C_k^\varepsilon C k ε containing x x x .
Problem setting
Our goal is to study the asymptotic behavior of an incompressible viscous fluid in which small elastic particles of size r ε r_\varepsilon r ε are immersed, arranged periodically with a small period ε ≫ r ε \varepsilon\gg r_\varepsilon ε ≫ r ε . Assuming that the density and the viscosity of the fluid are constants of order one, we impose the density of the solid particles of order ε 3 / r ε 3 \varepsilon^3/r_\varepsilon^3 ε 3 / r ε 3 and the coefficients of the corresponding elasticity tensor of order ε 3 / r ε \varepsilon^3/r_\varepsilon ε 3 / r ε . Note that this implies that the density is of a much higher order of magnitude than the elastic coefficients. Thus, the elastic particles are highly deformable.
Assuming that the elastic deformations are sufficiently small, the problem can be modeled by the linearized system
ε 3 r ε 3 ρ s ∂ t t 2 u ε − ε 3 r ε div x ( A e x ( u ε ) ) = f ε in ( 0 , T ) × Ω ε s , ρ l ∂ t v ε − 2 μ div x ( e x ( v ε ) ) + ∇ x p ε = f ε in ( 0 , T ) × Ω ε l , div x v ε = 0 in ( 0 , T ) × Ω ε l , ∂ t u ε = v ε , ε 3 r ε A e x ( u ε ) ν = ( 2 μ e x ( v ε ) − p ε I ) ν on ( 0 , T ) × ∂ Ω ε s , v ε = 0 on ( 0 , T ) × ∂ Ω , u ε ∣ t = 0 = u ε 0 , ∂ t u ε ∣ t = 0 = u ε 1 in Ω ε s , v ε ∣ t = 0 = v ε 0 in Ω ε l . (1) \begin{aligned}
\frac{\varepsilon^3}{r_\varepsilon^3}\rho^s\partial_{tt}^2u_\varepsilon-\frac{\varepsilon^3}{r_\varepsilon}\operatorname{div}_x\left(Ae_x(u_\varepsilon)\right)&=f_\varepsilon&&\text{in }(0,T)\times\Omega_\varepsilon^s,\\
\rho^l\partial_tv_\varepsilon-2\mu\operatorname{div}_x\left(e_x(v_\varepsilon)\right)+\nabla_xp_\varepsilon&=f_\varepsilon&&\text{in }(0,T)\times\Omega_\varepsilon^l,\\
\operatorname{div}_xv_\varepsilon&=0 &&\text{in }(0,T)\times\Omega_\varepsilon^l,\\
\partial_tu_\varepsilon=v_\varepsilon,\quad\frac{\varepsilon^3}{r_\varepsilon}Ae_x(u_\varepsilon)\nu&=(2\mu e_x(v_\varepsilon)-p_\varepsilon I)\nu&&\text{on }(0,T)\times\partial\Omega_\varepsilon^s,\\
v_\varepsilon&=0 &&\text{on }(0,T)\times\partial\Omega,\\
u_\varepsilon|_{t=0}=u_\varepsilon^0,\quad\partial_tu_\varepsilon|_{t=0}=u_\varepsilon^1&&\text{in }\Omega_\varepsilon^s,\\
v_\varepsilon|_{t=0}=v_\varepsilon^0&&\text{in }\Omega_\varepsilon^l. \tag*{(1)}
\end{aligned} r ε 3 ε 3 ρ s ∂ tt 2 u ε − r ε ε 3 div x ( A e x ( u ε ) ) ρ l ∂ t v ε − 2 μ div x ( e x ( v ε ) ) + ∇ x p ε div x v ε ∂ t u ε = v ε , r ε ε 3 A e x ( u ε ) ν v ε u ε ∣ t = 0 = u ε 0 , ∂ t u ε ∣ t = 0 = u ε 1 v ε ∣ t = 0 = v ε 0 = f ε = f ε = 0 = ( 2 μ e x ( v ε ) − p ε I ) ν = 0 in Ω ε s , in Ω ε l . in ( 0 , T ) × Ω ε s , in ( 0 , T ) × Ω ε l , in ( 0 , T ) × Ω ε l , on ( 0 , T ) × ∂ Ω ε s , on ( 0 , T ) × ∂ Ω , (1)
where Ω ⊂ R 3 \Omega\subset\mathbb{R}^3 Ω ⊂ R 3 is a connected bounded open set whose boundary can be locally written as the graph of a Lipschitz function and ε , r ε > 0 \varepsilon,r_\varepsilon>0 ε , r ε > 0 , such that
ε → 0 , r ε ε → 0. (2) \varepsilon\to0,\qquad\frac{r_\varepsilon}{\varepsilon}\to0. \tag*{(2)} ε → 0 , ε r ε → 0. (2)
The solid part Ω ε s \Omega_\varepsilon^s Ω ε s and the fluid part Ω ε l \Omega_\varepsilon^l Ω ε l are defined as
Ω ε s = ⋃ k ∈ Z 3 ε k + r ε O ⊂ Ω ( ε k + r ε O ) , Ω ε l = Ω ∖ Ω ε s ‾ . (3) \Omega_\varepsilon^s=\bigcup_{\substack{k\in\mathbb{Z}^3\\ \varepsilon k+r_\varepsilon\mathcal{O}\subset\Omega}}(\varepsilon k+r_\varepsilon\mathcal{O}),\qquad
\Omega_\varepsilon^l=\Omega\setminus\overline{\Omega_\varepsilon^s}. \tag*{(3)} Ω ε s = k ∈ Z 3 ε k + r ε O ⊂ Ω ⋃ ( ε k + r ε O ) , Ω ε l = Ω ∖ Ω ε s . (3)
where
O open, with O ‾ ⊂ Y : = ( − 1 2 , 1 2 ) 3 . (4) \mathcal{O}\ \text{open, with}\ \overline{\mathcal{O}}\subset Y:=\left(-\frac{1}{2},\frac{1}{2}\right)^3. \tag*{(4)} O open, with O ⊂ Y := ( − 2 1 , 2 1 ) 3 . (4)
The boundary of the open set O O O is assumed to be locally the graph of a Lipschitz function. Moreover Y ∖ O ‾ Y \setminus\overline{O} Y ∖ O is connected.
Through translation we can also assume that the center of mass of ∂ O \partial O ∂ O coincides with the origin, that is,
∫ ∂ O y d σ ( y ) = 0. (5) \int_{\partial O} y\,\mathrm{d}\sigma(y)=0. \tag*{(5)} ∫ ∂ O y d σ ( y ) = 0. (5)
We will also introduce
Ω ε ∗ = ⋃ ε ( k + Y ) ⊂ Ω ε ( k + Y ) . (6) \Omega_\varepsilon^*=\bigcup_{\varepsilon(k+Y)\subset\Omega}\varepsilon(k+Y). \tag*{(6)} Ω ε ∗ = ε ( k + Y ) ⊂ Ω ⋃ ε ( k + Y ) . (6)
The coefficients ρ s \rho^s ρ s , ρ l \rho^l ρ l and μ \mu μ in (3.1) are assumed to be positive constants, the tensor A A A belongs to the space L ( R s 3 × 3 ) \mathcal{L}(\mathbb{R}^{3\times3}_s) L ( R s 3 × 3 ) and satisfies the ellipticity and symmetry conditions
{ ∃ α > 0 such that A ξ : ξ ≥ α ∣ ξ ∣ 2 ∀ ξ ∈ R s 3 × 3 , A ξ : η = A η : ξ ∀ ξ , η ∈ R s 3 × 3 . (7) \left\{
\begin{aligned}
&\exists\ \alpha>0\ \text{such that } A\xi:\xi\geq\alpha|\xi|^2 &&\forall\xi\in\mathbb{R}^{3\times3}_s,\\
&A\xi:\eta=A\eta:\xi&&\forall\xi,\eta\in\mathbb{R}^{3\times3}_s.
\end{aligned}
\right. \tag*{(7)} { ∃ α > 0 such that A ξ : ξ ≥ α ∣ ξ ∣ 2 A ξ : η = A η : ξ ∀ ξ ∈ R s 3 × 3 , ∀ ξ , η ∈ R s 3 × 3 . (7)
The vector ν \nu ν in (3.1) refers to the outer unit normal vector on ∂ Ω ε s \partial\Omega_\varepsilon^s ∂ Ω ε s that point outwards O O O .
With respect to the right-hand side and the initial conditions let us assume
f ε ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , u ε 0 ∈ H 1 ( Ω ε s ) 3 , u ε 1 ∈ L 2 ( Ω ε s ) 3 , f_\varepsilon\in L^1(0,T;L^2(\Omega))^3,\quad u_\varepsilon^0\in H^1(\Omega_\varepsilon^s)^3,\quad u_\varepsilon^1\in L^2(\Omega_\varepsilon^s)^3, f ε ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , u ε 0 ∈ H 1 ( Ω ε s ) 3 , u ε 1 ∈ L 2 ( Ω ε s ) 3 ,
v ε 0 ∈ L 2 ( Ω ε l ) 3 , div x v ε 0 = 0 in Ω ε l , v ε 0 ⋅ ν = 0 on ∂ Ω , (8) v_\varepsilon^0\in L^2(\Omega_\varepsilon^l)^3,\quad\operatorname{div}_x v_\varepsilon^0=0\ \text{in }\Omega_\varepsilon^l,\quad v_\varepsilon^0\cdot\nu=0\ \text{on }\partial\Omega, \tag*{(8)} v ε 0 ∈ L 2 ( Ω ε l ) 3 , div x v ε 0 = 0 in Ω ε l , v ε 0 ⋅ ν = 0 on ∂ Ω , (8)
where now ν \nu ν is the outer unit normal vector on ∂ Ω \partial\Omega ∂ Ω .
From the physical point of view, the unknown functions u ε u_\varepsilon u ε , v ε v_\varepsilon v ε and p ε p_\varepsilon p ε respectively represent the deformation of the solid particles, the velocity of the fluid and its pressure. The coefficients ε 3 ρ s / r ε 3 \varepsilon^3\rho^s/r_\varepsilon^3 ε 3 ρ s / r ε 3 , ρ l \rho^l ρ l and μ \mu μ are the density of the elastic particles, the density of the fluid and the viscosity coefficient. The tensor ε 3 A / r ε \varepsilon^3A/r_\varepsilon ε 3 A / r ε is the elasticity tensor and f ε f_\varepsilon f ε is the exterior force acting on Ω \Omega Ω .
The following theorem proves the existence of a unique solution for (3.1). It does not use the particular structure of our set Ω ε s \Omega_\varepsilon^s Ω ε s . A similar result where, contrary to the case presented here, it is the solid part which surrounds the liquid part has been proved in [11 ] .
Theorem 3.1 (Existence and uniqueness). Let ω , Ω \omega,\Omega ω , Ω be two bounded open sets in R 3 \mathbb{R}^3 R 3 such that ω ‾ ⊂ Ω \overline{\omega}\subset\Omega ω ⊂ Ω , whose boundaries can be locally written as the graph of a Lipschitz function. We also consider three positive constants ρ s , ρ l , T \rho^s,\rho^l,T ρ s , ρ l , T , and a tensor A ∈ L ( R s 3 × 3 ) A\in\mathcal{L}(\mathbb{R}^{3\times3}_s) A ∈ L ( R s 3 × 3 ) which satisfies (3.7). Then, for every f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 f\in L^1(0,T;L^2(\Omega))^3 f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , u 0 ∈ H 1 ( ω ) 3 u^0\in H^1(\omega)^3 u 0 ∈ H 1 ( ω ) 3 , u 1 ∈ L 2 ( ω ) 3 u^1\in L^2(\omega)^3 u 1 ∈ L 2 ( ω ) 3 and v 0 ∈ L 2 ( Ω ∖ ω ‾ ) 3 v^0\in L^2(\Omega\setminus\overline{\omega})^3 v 0 ∈ L 2 ( Ω ∖ ω ) 3 such that
div x v 0 = 0 in Ω ∖ ω ‾ , v 0 ⋅ ν = 0 on ∂ ω , \operatorname{div}_x v^0=0\quad\text{in }\Omega\setminus\overline{\omega},\qquad v^0\cdot\nu=0\quad\text{on }\partial\omega, div x v 0 = 0 in Ω ∖ ω , v 0 ⋅ ν = 0 on ∂ ω ,
there exists a unique solution ( u , v , p ) (u,v,p) ( u , v , p ) of
{ ρ s ∂ t u − div x ( A e x ( u ) ) = f in ( 0 , T ) × ω , ρ ℓ ∂ t v − 2 μ div x e x ( v ) + ∇ x p = f in ( 0 , T ) × ( Ω ∖ ω ‾ ) , div x v = 0 in ( 0 , T ) × ( Ω ∖ ω ‾ ) , ∂ t u = v , A e x ( u ) v = ( 2 μ e x ( v ) − p I ) v on ( 0 , T ) × ∂ ω , v = 0 on ( 0 , T ) × ∂ Ω , u ∣ t = 0 = u 0 , ∂ t u ∣ t = 0 = u 1 in ω , v ∣ t = 0 = v 0 in Ω ∖ ω ‾ . (9) \begin{cases}
\rho^{s}\partial_t u-\operatorname{div}_{x}(A e_x(u))=f & \text{in }(0,T)\times\omega,\\
\rho^{\ell}\partial_t v-2\mu\operatorname{div}_{x}e_x(v)+\nabla_xp=f & \text{in }(0,T)\times(\Omega\setminus\overline{\omega}),\\
\operatorname{div}_{x}v=0 & \text{in }(0,T)\times(\Omega\setminus\overline{\omega}),\\
\partial_tu=v,\quad A e_x(u)v=(2\mu e_x(v)-pI)v & \text{on }(0,T)\times\partial\omega,\\
v=0 & \text{on }(0,T)\times\partial\Omega,\\
u|_{t=0}=u^0,\quad\partial_tu|_{t=0}=u^1 & \text{in }\omega,\\
v|_{t=0}=v^0 & \text{in }\Omega\setminus\overline{\omega}. \tag*{(9)}
\end{cases} ⎩ ⎨ ⎧ ρ s ∂ t u − div x ( A e x ( u )) = f ρ ℓ ∂ t v − 2 μ div x e x ( v ) + ∇ x p = f div x v = 0 ∂ t u = v , A e x ( u ) v = ( 2 μ e x ( v ) − p I ) v v = 0 u ∣ t = 0 = u 0 , ∂ t u ∣ t = 0 = u 1 v ∣ t = 0 = v 0 in ( 0 , T ) × ω , in ( 0 , T ) × ( Ω ∖ ω ) , in ( 0 , T ) × ( Ω ∖ ω ) , on ( 0 , T ) × ∂ ω , on ( 0 , T ) × ∂ Ω , in ω , in Ω ∖ ω . (9)
with
u ∈ C 1 ( [ 0 , T ] ; L 2 ( ω ) 3 ) ∩ C 0 ( [ 0 , T ] ; H 1 ( ω ) 3 ) , u\in C^1([0,T];L^2(\omega)^3)\cap C^0([0,T];H^1(\omega)^3), u ∈ C 1 ([ 0 , T ] ; L 2 ( ω ) 3 ) ∩ C 0 ([ 0 , T ] ; H 1 ( ω ) 3 ) ,
v ∈ C 0 ( [ 0 , T ] ; L 2 ( Ω ∖ ω ‾ ) 3 ) , p = d P d t with P ∈ C 0 ( [ 0 , T ] ; L 2 ( Ω ∖ ω ‾ ) ) , P ∣ t = 0 = 0 in Ω ∖ ω ‾ . v\in C^0([0,T];L^2(\Omega\setminus\overline{\omega})^3),\quad p=\frac{dP}{dt}\ \text{with }P\in C^0([0,T];L^2(\Omega\setminus\overline{\omega})),\ P|_{t=0}=0\ \text{in }\Omega\setminus\overline{\omega}. v ∈ C 0 ([ 0 , T ] ; L 2 ( Ω ∖ ω ) 3 ) , p = d t d P with P ∈ C 0 ([ 0 , T ] ; L 2 ( Ω ∖ ω )) , P ∣ t = 0 = 0 in Ω ∖ ω .
Moreover, such solution satisfies the energy identity
1 2 d d t ( ρ s ∫ ω ∣ ∂ t u ∣ 2 d x + ∫ ω A e x ( u ) : e x ( u ) d x + ρ ℓ ∫ Ω ∖ ω ‾ ∣ v ∣ 2 d x ) + 2 μ ∫ Ω ∖ ω ‾ ∣ e x ( v ) ∣ 2 d x = ∫ ω f ⋅ ∂ t u d x + ∫ Ω ∖ ω ‾ f ⋅ v d x in ( 0 , T ) . (10) \frac{1}{2}\frac{d}{dt}\left(\rho^s\int_{\omega}|\partial_tu|^2\,dx+\int_{\omega}A e_x(u):e_x(u)\,dx+\rho^\ell\int_{\Omega\setminus\overline{\omega}}|v|^2\,dx\right)
+2\mu\int_{\Omega\setminus\overline{\omega}}|e_x(v)|^2\,dx
=\int_{\omega}f\cdot\partial_tu\,dx+\int_{\Omega\setminus\overline{\omega}}f\cdot v\,dx\quad\text{in }(0,T). \tag*{(10)} 2 1 d t d ( ρ s ∫ ω ∣ ∂ t u ∣ 2 d x + ∫ ω A e x ( u ) : e x ( u ) d x + ρ ℓ ∫ Ω ∖ ω ∣ v ∣ 2 d x ) + 2 μ ∫ Ω ∖ ω ∣ e x ( v ) ∣ 2 d x = ∫ ω f ⋅ ∂ t u d x + ∫ Ω ∖ ω f ⋅ v d x in ( 0 , T ) . (10)
Proof. As in [11 ] , Theorem 3.2, let us use the Lumer-Philips theorem which is equivalent to the Hille-Yosida theorem (see e.g. [24 ] , theorem 4.3). Introducing w = ∂ t u w=\partial_tu w = ∂ t u , we rewrite (3.9) as the first-order system
{ ∂ t u = w in ( 0 , T ) × ω , ∂ t w = 1 ρ s div x ( A e x ( u ) ) + 1 ρ s f in ( 0 , T ) × ω , ∂ t v = 2 μ ρ ℓ div x e x ( v ) − 1 ρ ℓ ∇ x p + 1 ρ ℓ f in ( 0 , T ) × ( Ω ∖ ω ‾ ) , div x v = 0 in ( 0 , T ) × ( Ω ∖ ω ‾ ) , w = v , A e x ( u ) v = ( 2 μ e x ( v ) − p I ) v on ( 0 , T ) × ∂ ω , v = 0 on ( 0 , T ) × ∂ Ω , u ∣ t = 0 = u 0 , w ∣ t = 0 = u 1 in ω , v ∣ t = 0 = v 0 in Ω ∖ ω ‾ . (11) \begin{cases}
\partial_tu=w & \text{in }(0,T)\times\omega,\\
\partial_tw=\frac{1}{\rho^s}\operatorname{div}_{x}(A e_x(u))+\frac{1}{\rho^s}f & \text{in }(0,T)\times\omega,\\
\partial_tv=\frac{2\mu}{\rho^\ell}\operatorname{div}_{x}e_x(v)-\frac{1}{\rho^\ell}\nabla_xp+\frac{1}{\rho^\ell}f & \text{in }(0,T)\times(\Omega\setminus\overline{\omega}),\\
\operatorname{div}_{x}v=0 & \text{in }(0,T)\times(\Omega\setminus\overline{\omega}),\\
w=v,\quad A e_x(u)v=(2\mu e_x(v)-pI)v & \text{on }(0,T)\times\partial\omega,\\
v=0 & \text{on }(0,T)\times\partial\Omega,\\
u|_{t=0}=u^0,\quad w|_{t=0}=u^1 & \text{in }\omega,\\
v|_{t=0}=v^0 & \text{in }\Omega\setminus\overline{\omega}. \tag*{(11)}
\end{cases} ⎩ ⎨ ⎧ ∂ t u = w ∂ t w = ρ s 1 div x ( A e x ( u )) + ρ s 1 f ∂ t v = ρ ℓ 2 μ div x e x ( v ) − ρ ℓ 1 ∇ x p + ρ ℓ 1 f div x v = 0 w = v , A e x ( u ) v = ( 2 μ e x ( v ) − p I ) v v = 0 u ∣ t = 0 = u 0 , w ∣ t = 0 = u 1 v ∣ t = 0 = v 0 in ( 0 , T ) × ω , in ( 0 , T ) × ω , in ( 0 , T ) × ( Ω ∖ ω ) , in ( 0 , T ) × ( Ω ∖ ω ) , on ( 0 , T ) × ∂ ω , on ( 0 , T ) × ∂ Ω , in ω , in Ω ∖ ω . (11)
For
H v : = { v ∈ L 2 ( Ω ∖ ω ‾ ) 3 : div x v = 0 in Ω ∖ ω ‾ , v ⋅ ν = 0 on ∂ Ω } , H_v:=\{v\in L^2(\Omega\setminus\overline{\omega})^3:\operatorname{div}_{x}v=0\ \text{in }\Omega\setminus\overline{\omega},\ v\cdot\nu=0\ \text{on }\partial\Omega\}, H v := { v ∈ L 2 ( Ω ∖ ω ) 3 : div x v = 0 in Ω ∖ ω , v ⋅ ν = 0 on ∂ Ω } ,
we introduce the Hilbert space
H = H 1 ( ω ) 3 × L 2 ( ω ) 3 × H v , \mathcal{H}=H^1(\omega)^3\times L^2(\omega)^3\times H_v, H = H 1 ( ω ) 3 × L 2 ( ω ) 3 × H v ,
endowed with the scalar product
( ( u 1 , w 1 , v 1 ) , ( u 2 , w 2 , v 2 ) ) = ∫ ω ( A e x ( u 1 ) : e x ( u 2 ) + ρ s u 1 ⋅ u 2 + ρ s w 1 ⋅ w 2 ) d x + ρ f ∫ Ω ∖ ω ‾ v 1 ⋅ v 2 d x , ∀ ( u 1 , w 1 , v 1 ) , ( u 2 , w 2 , v 2 ) ∈ H . \begin{aligned}
((u_1,w_1,v_1),(u_2,w_2,v_2))={}&\int_{\omega}\left(Ae_x(u_1):e_x(u_2)+\rho^s u_1\cdot u_2+\rho^s w_1\cdot w_2\right)\,dx\\
&+\rho^f\int_{\Omega\setminus\overline{\omega}}v_1\cdot v_2\,dx,\qquad\forall(u_1,w_1,v_1),(u_2,w_2,v_2)\in\mathcal H.
\end{aligned} (( u 1 , w 1 , v 1 ) , ( u 2 , w 2 , v 2 )) = ∫ ω ( A e x ( u 1 ) : e x ( u 2 ) + ρ s u 1 ⋅ u 2 + ρ s w 1 ⋅ w 2 ) d x + ρ f ∫ Ω ∖ ω v 1 ⋅ v 2 d x , ∀ ( u 1 , w 1 , v 1 ) , ( u 2 , w 2 , v 2 ) ∈ H .
In this space we define the unbounded operator A : D ( A ) ⊂ H → H \mathcal A:D(\mathcal A)\subset\mathcal H\to\mathcal H A : D ( A ) ⊂ H → H by
D ( A ) = { ( u , w , v ) ∈ H : div x ( A e x ( u ) ) ∈ L 2 ( ω ) 3 , w ∈ H 1 ( ω ) 3 , ∃ p ∈ L 2 ( Ω ∖ ω ‾ ) such that 2 μ div x e x ( v ) + ∇ x p ∈ H v , w = v on ∂ ω , A e x ( u ) ν − 2 μ e x ( v ) ν + p ν = 0 on ∂ ω } . (12) \begin{aligned}
D(\mathcal A)=\biggl\{(u,w,v)\in\mathcal H:\ &\operatorname{div}_x(Ae_x(u))\in L^2(\omega)^3,\ w\in H^1(\omega)^3,\\
&\exists p\in L^2(\Omega\setminus\overline{\omega})\ \text{such that }2\mu\operatorname{div}_x e_x(v)+\nabla_xp\in H_v,\\
&w=v\ \text{on }\partial\omega,\ Ae_x(u)\nu-2\mu e_x(v)\nu+p\nu=0\ \text{ on }\partial\omega\biggr\}. \tag*{(12)}
\end{aligned} D ( A ) = { ( u , w , v ) ∈ H : div x ( A e x ( u )) ∈ L 2 ( ω ) 3 , w ∈ H 1 ( ω ) 3 , ∃ p ∈ L 2 ( Ω ∖ ω ) such that 2 μ div x e x ( v ) + ∇ x p ∈ H v , w = v on ∂ ω , A e x ( u ) ν − 2 μ e x ( v ) ν + p ν = 0 on ∂ ω } . (12)
and
A ( u , w , v ) = ( w , 1 ρ s div x ( A e x ( u ) ) , 2 μ ρ f div x e x ( v ) − 1 ρ f ∇ x p ) . \mathcal A(u,w,v)=\left(w,\frac{1}{\rho^s}\operatorname{div}_x(Ae_x(u)),\frac{2\mu}{\rho^f}\operatorname{div}_x e_x(v)-\frac{1}{\rho^f}\nabla_xp\right). A ( u , w , v ) = ( w , ρ s 1 div x ( A e x ( u )) , ρ f 2 μ div x e x ( v ) − ρ f 1 ∇ x p ) .
The operator A \mathcal A A is well defined since the function p p p in the definition of D ( A ) D(\mathcal A) D ( A ) is unique. Namely, if p 1 , p 2 p_1,p_2 p 1 , p 2 satisfy such conditions, then p 1 − p 2 p_1-p_2 p 1 − p 2 solves
{ Δ ( p 1 − p 2 ) = 0 in Ω ∖ ω ‾ p 1 − p 2 = 0 on ∂ ω , ∇ ( p 1 − p 2 ) ⋅ ν = 0 on ∂ Ω , \begin{cases}
\Delta(p_1-p_2)=0 & \text{in }\Omega\setminus\overline{\omega}\\
p_1-p_2=0 & \text{on }\partial\omega,\qquad\nabla(p_1-p_2)\cdot\nu=0\ \text{on }\partial\Omega,
\end{cases} { Δ ( p 1 − p 2 ) = 0 p 1 − p 2 = 0 in Ω ∖ ω on ∂ ω , ∇ ( p 1 − p 2 ) ⋅ ν = 0 on ∂ Ω ,
and thus p 1 = p 2 p_1=p_2 p 1 = p 2 in Ω ∖ ω ‾ \Omega\setminus\overline{\omega} Ω ∖ ω .
In order to prove that A − I \mathcal A-I A − I is in the conditions of the Lumer-Philips theorem and then that A \mathcal A A defines a semigroup, we have to prove that A − I \mathcal A-I A − I dissipative and surjective.
That A − I \mathcal A-I A − I is dissipative, just follows by an integration by parts, taking into account the definition of A \mathcal A A . Namely, we have
( ( A − I ) ( u , w , v ) , ( u , w , v ) ) H = ∫ ω ( A e x ( w ) : e x ( u ) + ρ s w ⋅ u + div x ( A e x ( u ) ) ⋅ w ) d x + ∫ Ω ∖ ω ‾ ( 2 μ div x e x ( v ) − ∇ x p ) ⋅ v d x − ∥ ( u , w , v ) ∥ H 2 = ρ s ∫ ω w ⋅ u d x − 2 μ ∫ Ω ∖ ω ‾ ∣ e x ( v ) ∣ 2 d x − ∥ ( u , w , v ) ∥ H 2 ≤ − 1 2 ∥ ( u , w , v ) ∥ H 2 . (13) \begin{aligned}
((\mathcal A-I)(u,w,v),(u,w,v))_{\mathcal H}
={}&\int_{\omega}\left(Ae_x(w):e_x(u)+\rho^s w\cdot u+\operatorname{div}_x(Ae_x(u))\cdot w\right)\,dx\\
&+\int_{\Omega\setminus\overline{\omega}}\left(2\mu\operatorname{div}_x e_x(v)-\nabla_xp\right)\cdot v\,dx-\|(u,w,v)\|_{\mathcal H}^2\\
={}&\rho^s\int_{\omega}w\cdot u\,dx-2\mu\int_{\Omega\setminus\overline{\omega}}|e_x(v)|^2\,dx-\|(u,w,v)\|_{\mathcal H}^2\\
\le{}&-\frac{1}{2}\|(u,w,v)\|_{\mathcal H}^2. \tag*{(13)}
\end{aligned} (( A − I ) ( u , w , v ) , ( u , w , v ) ) H = = ≤ ∫ ω ( A e x ( w ) : e x ( u ) + ρ s w ⋅ u + div x ( A e x ( u )) ⋅ w ) d x + ∫ Ω ∖ ω ( 2 μ div x e x ( v ) − ∇ x p ) ⋅ v d x − ∥ ( u , w , v ) ∥ H 2 ρ s ∫ ω w ⋅ u d x − 2 μ ∫ Ω ∖ ω ∣ e x ( v ) ∣ 2 d x − ∥ ( u , w , v ) ∥ H 2 − 2 1 ∥ ( u , w , v ) ∥ H 2 . (13)
For the surjectivity, given ( f , g , h ) ∈ H (f,g,h)\in\mathcal H ( f , g , h ) ∈ H , we have to prove the existence of ( u , w , v ) ∈ D ( A ) (u,w,v)\in D(\mathcal A) ( u , w , v ) ∈ D ( A ) such that
( A − I ) ( u , w , v ) = ( f , g , h ) . (14) (\mathcal A-I)(u,w,v)=(f,g,h). \tag*{(14)} ( A − I ) ( u , w , v ) = ( f , g , h ) . (14)
This is a system of three equations, the first of them reads as w − u = f w-u=f w − u = f . Eliminating u u u from this equation and recalling the definition of D ( A ) D(\mathcal A) D ( A ) , we deduce that (3.14) is equivalent to find ( w , v , p ) ∈ H 1 ( ω ) 3 × H 1 ( Ω ∖ ω ‾ ) 3 × L 2 ( Ω ∖ ω ‾ ) (w,v,p) \in H^{1}(\omega)^{3} \times H^{1}(\Omega\setminus\overline{\omega})^{3} \times L^{2}(\Omega\setminus\overline{\omega}) ( w , v , p ) ∈ H 1 ( ω ) 3 × H 1 ( Ω ∖ ω ) 3 × L 2 ( Ω ∖ ω ) such that
{ − div x ( A e x ( w ) ) + ρ s w = − ρ s g − div x ( A e x ( f ) ) in ω , − 2 μ div x e x ( v ) + ∇ x p + ρ l v = − ρ l h in Ω ∖ ω ‾ , div x v = 0 in Ω ∖ ω ‾ , w = v , A e x ( w − f ) ν = ( 2 μ e x ( v ) − p I ) ν on ∂ ω , v = 0 on ∂ Ω . (15) \left\{
\begin{aligned}
-\operatorname{div}_{x}(A e_{x}(w))+\rho^{s}w&=-\rho^{s}g-\operatorname{div}_{x}(A e_{x}(f)) &&\text{in }\omega,\\
-2\mu\operatorname{div}_{x}e_{x}(v)+\nabla_{x}p+\rho^{l}v&=-\rho^{l}h &&\text{in }\Omega\setminus\overline{\omega},\\
\operatorname{div}_{x}v&=0 &&\text{in }\Omega\setminus\overline{\omega},\\
w=v,\quad A e_{x}(w-f)\nu&=(2\mu e_{x}(v)-pI)\nu&&\text{on }\partial\omega,\\
v&=0 &&\text{on }\partial\Omega.
\end{aligned}
\right. \tag*{(15)} ⎩ ⎨ ⎧ − div x ( A e x ( w )) + ρ s w − 2 μ div x e x ( v ) + ∇ x p + ρ l v div x v w = v , A e x ( w − f ) ν v = − ρ s g − div x ( A e x ( f )) = − ρ l h = 0 = ( 2 μ e x ( v ) − p I ) ν = 0 in ω , in Ω ∖ ω , in Ω ∖ ω , on ∂ ω , on ∂ Ω. (15)
Let us apply the Lax-Milgram theorem to prove the existence of solution for (3.15). We define
H ^ = { ( w , v ) ∈ H 1 ( ω ) 3 ∩ H 1 ( Ω ∖ ω ‾ ) 3 : w = v on ∂ ω , v = 0 on ∂ Ω } . \widehat{H}=\left\{(w,v)\in H^{1}(\omega)^{3}\cap H^{1}(\Omega\setminus\overline{\omega})^{3}:w=v\text{ on }\partial\omega,\ v=0\text{ on }\partial\Omega\right\}. H = { ( w , v ) ∈ H 1 ( ω ) 3 ∩ H 1 ( Ω ∖ ω ) 3 : w = v on ∂ ω , v = 0 on ∂ Ω } .
H ~ = { ( w , v ) ∈ H ^ : div x v = 0 in Ω ∖ ω ‾ } . \widetilde{H}=\left\{(w,v)\in\widehat{H}:\operatorname{div}_{x}v=0\text{ in }\Omega\setminus\overline{\omega}\right\}. H = { ( w , v ) ∈ H : div x v = 0 in Ω ∖ ω } .
and a : H ^ × H ^ → R \mathbf{a}:\widehat{H}\times\widehat{H}\to\mathbb{R} a : H × H → R the continuous bilinear form given by
a ( ( w 1 , v 1 ) , ( w 2 , v 2 ) ) = ∫ ω ( A e x ( w 1 ) : e x ( w 2 ) + ρ s w 1 ⋅ w 2 ) d x + ∫ Ω ∖ ω ‾ ( 2 μ e x ( v 1 ) : e x ( v 2 ) + ρ l v 1 ⋅ v 2 ) d x , ∀ ( w 1 , v 1 ) , ( w 2 , v 2 ) ∈ H ^ . \mathbf{a}((w_{1},v_{1}),(w_{2},v_{2}))=\int_{\omega}\left(Ae_{x}(w_{1}):e_{x}(w_{2})+\rho^{s}w_{1}\cdot w_{2}\right)dx+\int_{\Omega\setminus\overline{\omega}}\left(2\mu e_{x}(v_{1}):e_{x}(v_{2})+\rho^{l}v_{1}\cdot v_{2}\right)dx,\qquad\forall(w_{1},v_{1}),(w_{2},v_{2})\in\widehat{H}. a (( w 1 , v 1 ) , ( w 2 , v 2 )) = ∫ ω ( A e x ( w 1 ) : e x ( w 2 ) + ρ s w 1 ⋅ w 2 ) d x + ∫ Ω ∖ ω ( 2 μ e x ( v 1 ) : e x ( v 2 ) + ρ l v 1 ⋅ v 2 ) d x , ∀ ( w 1 , v 1 ) , ( w 2 , v 2 ) ∈ H .
Reasoning as in (3.13) we have that a \mathbf{a} a is coercive and thus, by the Lax-Milgram theorem there exists a unique solution of
{ ( w , v ) ∈ H ~ a ( ( w , v ) , ( w ~ , v ~ ) ) = ∫ ω ( − ρ s g ⋅ w ~ + A e x ( f ) : e x ( w ~ ) ) d x − ρ l ∫ Ω ∖ ω ‾ h ⋅ v ~ d x , ∀ ( w ~ , v ~ ) ∈ H ~ . \left\{
\begin{aligned}
(w,v)&\in\widetilde{H}\\
\mathbf{a}((w,v),(\widetilde{w},\widetilde{v}))&=\int_{\omega}\left(-\rho^{s}g\cdot\widetilde{w}+Ae_{x}(f):e_{x}(\widetilde{w})\right)dx-\rho^{l}\int_{\Omega\setminus\overline{\omega}}h\cdot\widetilde{v}\,dx,\quad\forall(\widetilde{w},\widetilde{v})\in\widetilde{H}.
\end{aligned}
\right. ⎩ ⎨ ⎧ ( w , v ) a (( w , v ) , ( w , v )) ∈ H = ∫ ω ( − ρ s g ⋅ w + A e x ( f ) : e x ( w ) ) d x − ρ l ∫ Ω ∖ ω h ⋅ v d x , ∀ ( w , v ) ∈ H .
Then, defining E : H ^ → R E:\widehat{H}\to\mathbb{R} E : H → R by
E ( w ~ , v ~ ) = a ( ( w , v ) , ( w ~ , v ~ ) ) − ∫ ω ( − ρ s g ⋅ w ~ + A e x ( f ) : e x ( w ~ ) ) d x + ρ l ∫ Ω ∖ ω ‾ h ⋅ v ~ d x , E(\widetilde{w},\widetilde{v})=\mathbf{a}((w,v),(\widetilde{w},\widetilde{v}))-\int_{\omega}\left(-\rho^{s}g\cdot\widetilde{w}+Ae_{x}(f):e_{x}(\widetilde{w})\right)dx+\rho^{l}\int_{\Omega\setminus\overline{\omega}}h\cdot\widetilde{v}\,dx, E ( w , v ) = a (( w , v ) , ( w , v )) − ∫ ω ( − ρ s g ⋅ w + A e x ( f ) : e x ( w ) ) d x + ρ l ∫ Ω ∖ ω h ⋅ v d x ,
we have that E ( w ~ , v ~ ) = 0 E(\widetilde{w},\widetilde{v})=0 E ( w , v ) = 0 for every ( w ~ , v ~ ) ∈ H ^ (\widetilde{w},\widetilde{v})\in\widehat{H} ( w , v ) ∈ H with div x v = 0 \operatorname{div}_{x}v=0 div x v = 0 in Ω ∖ ω ‾ \Omega\setminus\overline{\omega} Ω ∖ ω . In particular this holds for w ~ = 0 \widetilde{w}=0 w = 0 , v ~ ∈ H 0 1 ( Ω ∖ ω ‾ ) 3 \widetilde{v}\in H_{0}^{1}(\Omega\setminus\overline{\omega})^{3} v ∈ H 0 1 ( Ω ∖ ω ) 3 with null divergence. This allows us to use De Rham’s theorem to deduce the existence of p ∈ L 2 ( Ω ∖ ω ‾ ) p\in L^{2}(\Omega\setminus\overline{\omega}) p ∈ L 2 ( Ω ∖ ω ) , unique, up to an additive constant such that
E ( 0 , v ~ ) = − ∫ Ω ∖ ω ‾ p div x v ~ d x , ∀ v ~ ∈ H 0 1 ( Ω ∖ ω ‾ ) 3 . E(0,\widetilde{v})=-\int_{\Omega\setminus\overline{\omega}}p\operatorname{div}_{x}\widetilde{v}\,dx,\qquad\forall\widetilde{v}\in H_{0}^{1}(\Omega\setminus\overline{\omega})^{3}. E ( 0 , v ) = − ∫ Ω ∖ ω p div x v d x , ∀ v ∈ H 0 1 ( Ω ∖ ω ) 3 .
Since for every w ^ ∈ H 1 ( ω ) 3 \widehat{w}\in H^{1}(\omega)^{3} w ∈ H 1 ( ω ) 3 with
∫ ∂ ω w ^ ⋅ ν d s ( x ) = 0 , (16) \int_{\partial\omega}\widehat{w}\cdot\nu\,ds(x)=0, \tag*{(16)} ∫ ∂ ω w ⋅ ν d s ( x ) = 0 , (16)
there exists v ^ ∈ H 0 1 ( Ω ∖ ω ‾ ) 3 \hat{v} \in H_0^1(\Omega\setminus\overline{\omega})^3 v ^ ∈ H 0 1 ( Ω ∖ ω ) 3 with null divergence in Ω ∖ ω ‾ \Omega\setminus\overline{\omega} Ω ∖ ω such that v ^ = w ^ \hat{v} = \hat{w} v ^ = w ^ on ∂ ω \partial\omega ∂ ω , we get that w , v , p w,v,p w , v , p satisfy
∫ ω ( A e x ( w ) : e x ( w ^ ) + ρ s w ⋅ w ^ ) d x + ∫ Ω ∖ ω ‾ ( 2 μ e x ( v ) : e x ( v ^ ) + ρ f v ⋅ v ^ − p div x v ^ ) d x = ∫ ω ( − ρ s g ⋅ w ^ + A e x ( f ) : e x ( w ^ ) ) d x − ρ f ∫ Ω ∖ ω ‾ h ⋅ v ^ d x (17) \int_{\omega} \left(A e_x(w) : e_x(\hat{w}) + \rho^s w \cdot\hat{w}\right) dx + \int_{\Omega\setminus\overline{\omega}} \left(2\mu e_x(v) : e_x(\hat{v}) + \rho^f v \cdot\hat{v} - p \operatorname{div}_x \hat{v}\right) dx
=
\int_{\omega} \left(-\rho^s g \cdot\hat{w} + A e_x(f) : e_x(\hat{w})\right) dx - \rho^f \int_{\Omega\setminus\overline{\omega}} h \cdot\hat{v} \, dx \tag*{(17)} ∫ ω ( A e x ( w ) : e x ( w ^ ) + ρ s w ⋅ w ^ ) d x + ∫ Ω ∖ ω ( 2 μ e x ( v ) : e x ( v ^ ) + ρ f v ⋅ v ^ − p div x v ^ ) d x = ∫ ω ( − ρ s g ⋅ w ^ + A e x ( f ) : e x ( w ^ ) ) d x − ρ f ∫ Ω ∖ ω h ⋅ v ^ d x (17)
for every ( w ^ , v ^ ) ∈ H ^ (\hat{w},\hat{v}) \in\hat{H} ( w ^ , v ^ ) ∈ H ^ which satisfies (16) . Therefore ( w ^ , v ^ ) ∈ H ^ (\hat{w},\hat{v}) \in\hat{H} ( w ^ , v ^ ) ∈ H ^ is solution of a problem similar to (15) that arises from replacing the second condition in the fourth line by
A e x ( w − f ) ν − ( 2 μ e x ( v ) − p I ) ν = c ν on ∂ ω , A e_x(w-f)\nu- (2\mu e_x(v)-pI)\nu= c\nu\quad\text{on } \partial\omega, A e x ( w − f ) ν − ( 2 μ e x ( v ) − p I ) ν = c ν on ∂ ω ,
with c c c constant. Since p p p was defined up to an additive constant, we can choose it in such way that c = 0 c=0 c = 0 . This proves the existence of a unique solution for (15) .
The Lumer-Philips theorem then proves the existence of a continuous semigroup S ( t ) : H → H S(t): H \to H S ( t ) : H → H such that for every ( u 0 , u 1 , v 0 ) ∈ H (u^0,u^1,v^0) \in H ( u 0 , u 1 , v 0 ) ∈ H , and every f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 f \in L^1(0,T;L^2(\Omega))^3 f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 with
div x f = 0 in ( 0 , T ) × ( Ω ∖ ω ‾ ) , f ⋅ ν = 0 on ( 0 , T ) × ∂ Ω , (18) \operatorname{div}_x f = 0 \quad\text{in } (0,T) \times(\Omega\setminus\overline{\omega}), \qquad f \cdot\nu= 0 \quad\text{on } (0,T) \times\partial\Omega, \tag*{(18)} div x f = 0 in ( 0 , T ) × ( Ω ∖ ω ) , f ⋅ ν = 0 on ( 0 , T ) × ∂ Ω , (18)
i.e. such that ( 0 , f χ ω , f χ Ω ∖ ω ‾ ) ∈ L 1 ( 0 , T ; H ) (0,f\chi_{\omega},f\chi_{\Omega\setminus\overline{\omega}}) \in L^1(0,T;H) ( 0 , f χ ω , f χ Ω ∖ ω ) ∈ L 1 ( 0 , T ; H ) , there exist a unique solution ( u , v , w , p ) (u,v,w,p) ( u , v , w , p ) of (11) , where ( u , v , w ) ∈ L ∞ ( 0 , T ; H ) (u,v,w) \in L^\infty(0,T;H) ( u , v , w ) ∈ L ∞ ( 0 , T ; H ) and it is given by
( u ( t ) , w ( t ) , v ( t ) ) = S ( t ) ( u 0 , w 0 , v 0 ) + ∫ 0 t S ( t − r ) ( 0 , 1 ρ s f ( r ) , 1 ρ f f ( r ) ) d r . (u(t),w(t),v(t)) = S(t)(u^0,w^0,v^0) + \int_0^t S(t-r)\left(0,\frac{1}{\rho^s}f(r),\frac{1}{\rho^f}f(r)\right) dr. ( u ( t ) , w ( t ) , v ( t )) = S ( t ) ( u 0 , w 0 , v 0 ) + ∫ 0 t S ( t − r ) ( 0 , ρ s 1 f ( r ) , ρ f 1 f ( r ) ) d r .
The general case follows by noting that every f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 f \in L^1(0,T;L^2(\Omega))^3 f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 can be decomposed as f 1 + f 2 f_1+f_2 f 1 + f 2 where f 1 f_1 f 1 satisfies (18) and f 2 f_2 f 2 is such that
f 2 = 0 in ω , ∫ Ω ∖ ω ‾ f 2 ⋅ v d x = 0 , ∀ v ∈ H v . f_2=0 \quad\text{in } \omega, \qquad\int_{\Omega\setminus\overline{\omega}} f_2 \cdot v \, dx = 0, \quad\forall v \in H_v. f 2 = 0 in ω , ∫ Ω ∖ ω f 2 ⋅ v d x = 0 , ∀ v ∈ H v .
De Rham’s Theorem then proves the existence of a distribution p ∈ L 1 ( 0 , T ; L 2 ( ω ) ) p \in L^1(0,T;L^2(\omega)) p ∈ L 1 ( 0 , T ; L 2 ( ω )) such that
∇ x p = f in ( 0 , T ) × ( Ω ∖ ω ‾ ) , p = 0 on ( 0 , T ) × ∂ ω . \nabla_x p = f \quad\text{in } (0,T) \times(\Omega\setminus\overline{\omega}), \qquad p=0 \quad\text{on } (0,T) \times\partial\omega. ∇ x p = f in ( 0 , T ) × ( Ω ∖ ω ) , p = 0 on ( 0 , T ) × ∂ ω .
A solution of (11) with f = f 2 f=f_2 f = f 2 and null initial data is given for such p p p and u , v , w u,v,w u , v , w the null functions. By linearity we then have the existence of solution for (11) .
The semigroup theory also provides the energy identity (10) , which formally can be obtained by multiplying the first equation in (9) by ∂ t u \partial_t u ∂ t u , the second one by v v v and then integrating by parts.
The fact that p p p can be written as the time derivative of a function in C 0 ( [ 0 , T ] ; L 2 ( ω ) ) C^0([0,T];L^2(\omega)) C 0 ([ 0 , T ] ; L 2 ( ω )) vanishing at t = 0 t=0 t = 0 can be deduced integrating in time in (9) and then using De Rham’s theorem. We will use a similar reasoning in the proof of Proposition 4.1 to get some a priori estimates for the solution of (1) .
A priori estimates
Thanks to Theorem 3.1 we know that (1) has a unique solution for every ε > 0 \varepsilon>0 ε > 0 . Our aim in this section is to provide some a priori estimates for such solution which will be used in the next one to carry out the homogenization. This is given by the following proposition (see also Remark 4.2 below)
Proposition 4.1. Let ( u ε , v ε , p ε ) (u_\varepsilon,v_\varepsilon,p_\varepsilon) ( u ε , v ε , p ε ) be the solution of (1) , where Ω ε s , Ω ε l \Omega_\varepsilon^s,\Omega_\varepsilon^l Ω ε s , Ω ε l are defined by (3) , ρ s , ρ l , μ \rho^s,\rho^l,\mu ρ s , ρ l , μ are positive, A ∈ L ( R 3 × 3 ) A\in L(\mathbb{R}^{3\times3}) A ∈ L ( R 3 × 3 ) satisfies (7) and the right-hand side f ε f_\varepsilon f ε and the initial conditions u ε 0 , u ε 1 , v ε 0 u_\varepsilon^0,u_\varepsilon^1,v_\varepsilon^0 u ε 0 , u ε 1 , v ε 0 satisfy (8) . Then, denoting D ε D_\varepsilon D ε by
D ε ( t ) = ε 3 r ε 3 ∥ u ε 1 ∥ L 2 ( Ω ε s ) 3 2 + ε 3 r ε ∥ e x ( u ε 0 ) ∥ L 2 ( Ω ε s ) 3 × 3 2 + ∥ v ε 0 ∥ L 2 ( Ω ε l ) 3 2 + ( ∫ 0 t ∥ f ε ∥ L 2 ( Ω ε l ) 3 d s ) 2 , D_\varepsilon(t)=\frac{\varepsilon^3}{r_\varepsilon^3}\lVert u_\varepsilon^1\rVert_{L^2(\Omega_\varepsilon^s)^3}^2+\frac{\varepsilon^3}{r_\varepsilon}\lVert e_x(u_\varepsilon^0)\rVert_{L^2(\Omega_\varepsilon^s)^{3\times3}}^2+\lVert v_\varepsilon^0\rVert_{L^2(\Omega_\varepsilon^l)^3}^2+\left(\int_0^t\lVert f_\varepsilon\rVert_{L^2(\Omega_\varepsilon^l)^3}\,ds\right)^2, D ε ( t ) = r ε 3 ε 3 ∥ u ε 1 ∥ L 2 ( Ω ε s ) 3 2 + r ε ε 3 ∥ e x ( u ε 0 ) ∥ L 2 ( Ω ε s ) 3 × 3 2 + ∥ v ε 0 ∥ L 2 ( Ω ε l ) 3 2 + ( ∫ 0 t ∥ f ε ∥ L 2 ( Ω ε l ) 3 d s ) 2 , we have:
There exists C > 0 C>0 C > 0 , which only depends on ρ s \rho^s ρ s , ρ l \rho^l ρ l , α \alpha α and ∣ A ∣ |A| ∣ A ∣ such that
ε 3 r ε 3 ∥ ∂ t u ε ( t ) ∥ L 2 ( Ω ε s ) 3 2 + ε 3 r ε ∥ e x ( u ε ) ( t ) ∥ L 2 ( Ω ε s ) 3 × 3 2 + ∥ v ε ( t ) ∥ L 2 ( Ω ε l ) 3 2 + ∫ 0 t ∥ e x ( v ε ) ∥ L 2 ( Ω ε l ) 3 × 3 2 d s ≤ C D ε ( t ) , ∀ t ∈ [ 0 , T ] . (19) \frac{\varepsilon^3}{r_\varepsilon^3}\lVert\partial_tu_\varepsilon(t)\rVert_{L^2(\Omega_\varepsilon^s)^3}^2+\frac{\varepsilon^3}{r_\varepsilon}\lVert e_x(u_\varepsilon)(t)\rVert_{L^2(\Omega_\varepsilon^s)^{3\times3}}^2+\lVert v_\varepsilon(t)\rVert_{L^2(\Omega_\varepsilon^l)^3}^2+\int_0^t\lVert e_x(v_\varepsilon)\rVert_{L^2(\Omega_\varepsilon^l)^{3\times3}}^2\,ds\le CD_\varepsilon(t),\qquad\forall t\in[0,T]. \tag*{(19)} r ε 3 ε 3 ∥ ∂ t u ε ( t ) ∥ L 2 ( Ω ε s ) 3 2 + r ε ε 3 ∥ e x ( u ε ) ( t ) ∥ L 2 ( Ω ε s ) 3 × 3 2 + ∥ v ε ( t ) ∥ L 2 ( Ω ε l ) 3 2 + ∫ 0 t ∥ e x ( v ε ) ∥ L 2 ( Ω ε l ) 3 × 3 2 d s ≤ C D ε ( t ) , ∀ t ∈ [ 0 , T ] . (19) Moreover, there exist P ε ∈ C 0 ( [ 0 , T ] ; L 2 ( Ω ) ) P_\varepsilon\in C^0([0,T];L^2(\Omega)) P ε ∈ C 0 ([ 0 , T ] ; L 2 ( Ω )) , ζ ε ∈ C 0 ( [ 0 , T ] ) \zeta_\varepsilon\in C^0([0,T]) ζ ε ∈ C 0 ([ 0 , T ]) such that
p ε = ∂ t P ε in ( 0 , T ) × Ω ε l , (20) p_\varepsilon=\partial_tP_\varepsilon\quad\text{in }(0,T)\times\Omega_\varepsilon^l, \tag*{(20)} p ε = ∂ t P ε in ( 0 , T ) × Ω ε l , (20) P ε ( t , x ) = ζ ε ( t ) a.e. in Ω ε s , (21) P_\varepsilon(t,x)=\zeta_\varepsilon(t)\quad\text{a.e. in }\Omega_\varepsilon^s, \tag*{(21)} P ε ( t , x ) = ζ ε ( t ) a.e. in Ω ε s , (21) P ε ( 0 , x ) = 0 a.e. in Ω , ∫ Ω P ε ( t , x ) d x = 0 , ∀ t ∈ [ 0 , T ] , (22) P_\varepsilon(0,x)=0\quad\text{a.e. in }\Omega,\qquad\int_\Omega P_\varepsilon(t,x)\,dx=0,\quad\forall t\in[0,T], \tag*{(22)} P ε ( 0 , x ) = 0 a.e. in Ω , ∫ Ω P ε ( t , x ) d x = 0 , ∀ t ∈ [ 0 , T ] , (22) ∫ Ω P ε div x φ d x = ε 3 r ε 3 ρ s ∫ Ω ε s ( ∂ t u ε − u ε 1 ) ⋅ φ d x + ρ l ∫ Ω ε l ( v ε − v ε 0 ) ⋅ φ d x + ε 3 r ε ∫ Ω ε s A e x ( ∫ 0 t u ε d s ) : e x ( φ ) d x + 2 μ ∫ Ω ε l e x ( ∫ 0 t v ε d s ) : e x ( φ ) d x − ∫ Ω ∫ 0 t f ε d s ⋅ φ d x , ∀ φ ∈ H 0 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] . (23) \int_\Omega P_\varepsilon\operatorname{div}_x\varphi\,dx=\frac{\varepsilon^3}{r_\varepsilon^3}\rho^s\int_{\Omega_\varepsilon^s}(\partial_tu_\varepsilon-u_\varepsilon^1)\cdot\varphi\,dx+\rho^l\int_{\Omega_\varepsilon^l}(v_\varepsilon-v_\varepsilon^0)\cdot\varphi\,dx+\frac{\varepsilon^3}{r_\varepsilon}\int_{\Omega_\varepsilon^s}Ae_x\left(\int_0^t u_\varepsilon\,ds\right):e_x(\varphi)\,dx+2\mu\int_{\Omega_\varepsilon^l}e_x\left(\int_0^t v_\varepsilon\,ds\right):e_x(\varphi)\,dx-\int_\Omega\int_0^t f_\varepsilon\,ds\cdot\varphi\,dx,\quad\forall\varphi\in H_0^1(\Omega)^3,\ \forall t\in[0,T]. \tag*{(23)} ∫ Ω P ε div x φ d x = r ε 3 ε 3 ρ s ∫ Ω ε s ( ∂ t u ε − u ε 1 ) ⋅ φ d x + ρ l ∫ Ω ε l ( v ε − v ε 0 ) ⋅ φ d x + r ε ε 3 ∫ Ω ε s A e x ( ∫ 0 t u ε d s ) : e x ( φ ) d x + 2 μ ∫ Ω ε l e x ( ∫ 0 t v ε d s ) : e x ( φ ) d x − ∫ Ω ∫ 0 t f ε d s ⋅ φ d x , ∀ φ ∈ H 0 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] . (23) There exists C > 0 C>0 C > 0 which only depends on Ω \Omega Ω , O \mathcal{O} O , ρ s \rho^s ρ s , ρ l \rho^l ρ l , α \alpha α , and ∣ A ∣ |A| ∣ A ∣ such that
∥ P ε ( t ) ∥ L 2 ( Ω ) ≤ C ( 1 + ε 3 r ε ) 1 2 D ε ( t ) 1 2 , ∣ ζ ε ( t ) ∣ ≤ C [ ε 3 r ε 3 ( 1 + ε 3 r ε ) ] 1 2 D ε ( t ) 1 2 , ∀ t ∈ [ 0 , T ] . (24) \lVert P_\varepsilon(t)\rVert_{L^2(\Omega)}\le C\left(1+\frac{\varepsilon^3}{r_\varepsilon}\right)^{\frac12}D_\varepsilon(t)^{\frac12},\qquad|\zeta_\varepsilon(t)|\le C\left[\frac{\varepsilon^3}{r_\varepsilon^3}\left(1+\frac{\varepsilon^3}{r_\varepsilon}\right)\right]^{\frac12}D_\varepsilon(t)^{\frac12},\quad\forall t\in[0,T]. \tag*{(24)} ∥ P ε ( t ) ∥ L 2 ( Ω ) ≤ C ( 1 + r ε ε 3 ) 2 1 D ε ( t ) 2 1 , ∣ ζ ε ( t ) ∣ ≤ C [ r ε 3 ε 3 ( 1 + r ε ε 3 ) ] 2 1 D ε ( t ) 2 1 , ∀ t ∈ [ 0 , T ] . (24)
Proof. Defining E ε : [ 0 , T ] → R E_\varepsilon: [0,T] \to\mathbb{R} E ε : [ 0 , T ] → R by
E ε ( t ) : = ρ ε s ∫ Ω ε s ∣ ∂ t u ε ∣ 2 d x + ∫ Ω ε s A ε e x ( u ε ) : e x ( u ε ) d x + ρ ε l ∫ Ω ε l ∣ v ε ∣ 2 d x , ∀ t ∈ [ 0 , T ] , E_\varepsilon(t) := \rho_\varepsilon^s \int_{\Omega_\varepsilon^s} |\partial_t u_\varepsilon|^2 \,dx + \int_{\Omega_\varepsilon^s} A_\varepsilon e_x(u_\varepsilon) : e_x(u_\varepsilon) \,dx + \rho_\varepsilon^l \int_{\Omega_\varepsilon^l} |v_\varepsilon|^2 \,dx, \qquad\forall t \in[0,T], E ε ( t ) := ρ ε s ∫ Ω ε s ∣ ∂ t u ε ∣ 2 d x + ∫ Ω ε s A ε e x ( u ε ) : e x ( u ε ) d x + ρ ε l ∫ Ω ε l ∣ v ε ∣ 2 d x , ∀ t ∈ [ 0 , T ] ,
we can use the energy identity (10) , applied to (1) , and the Cauchy-Schwarz inequality to deduce
1 2 d E ε d t + 2 μ ∫ Ω ε l ∣ e x ( v ε ) ∣ 2 d x ≤ ∥ f ε ∥ L 2 ( Ω ) 3 E ε a.e. in ( 0 , T ) , (25) \frac{1}{2}\frac{dE_\varepsilon}{dt} + 2\mu\int_{\Omega_\varepsilon^l} |e_x(v_\varepsilon)|^2 \,dx \le\|f_\varepsilon\|_{L^2(\Omega)^3}\sqrt{E_\varepsilon} \qquad\text{a.e. in }(0,T), \tag*{(25)} 2 1 d t d E ε + 2 μ ∫ Ω ε l ∣ e x ( v ε ) ∣ 2 d x ≤ ∥ f ε ∥ L 2 ( Ω ) 3 E ε a.e. in ( 0 , T ) , (25)
which implies
E ε ( t ) ≤ E ε ( 0 ) + ∥ f ε ∥ L 1 ( 0 , t ; L 2 ( Ω ) 3 ) , in ( 0 , T ) , \sqrt{E_\varepsilon(t)} \le\sqrt{E_\varepsilon(0)} + \|f_\varepsilon\|_{L^1(0,t;L^2(\Omega)^3)}, \qquad\text{in }(0,T), E ε ( t ) ≤ E ε ( 0 ) + ∥ f ε ∥ L 1 ( 0 , t ; L 2 ( Ω ) 3 ) , in ( 0 , T ) ,
and then (19) .
In order to estimate the pressure p ε p_\varepsilon p ε , we take φ ∈ H 0 1 ( Ω ) 3 \varphi\in H_0^1(\Omega)^3 φ ∈ H 0 1 ( Ω ) 3 with div x φ = 0 \operatorname{div}_x\varphi= 0 div x φ = 0 in Ω ε l \Omega_\varepsilon^l Ω ε l , as test function in the first and second equation in (1) . Integrating in ( 0 , t ) × Ω (0,t) \times\Omega ( 0 , t ) × Ω , with t ∈ ( 0 , T ) t \in(0,T) t ∈ ( 0 , T ) and taking into account the second boundary condition in the fourth line, we deduce that G ε ∈ C 0 ( [ 0 , T ] ; H − 1 ( Ω ) 3 ) G_\varepsilon\in C^0([0,T];H^{-1}(\Omega)^3) G ε ∈ C 0 ([ 0 , T ] ; H − 1 ( Ω ) 3 ) , defined as
⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 : = ε 3 r ε 3 ρ ε s ∫ Ω ε s ( ∂ t u ε ( t ) − u ε 1 ) ⋅ φ d x + ρ ε l ∫ Ω ε l ( v ε ( t ) − v ε 0 ) ⋅ φ d x + ε 3 r ε ∫ Ω ε s A ε e x ( ∫ 0 t u ε d s ) : e x ( φ ) d x + 2 μ ∫ Ω ε l e x ( ∫ 0 t v ε d s ) : e x ( φ ) d x − ∫ Ω ∫ 0 t f ε d s ⋅ φ d x , (26) \begin{aligned}
\langle G_\varepsilon(t),\varphi\rangle_{H^{-1}(\Omega)^3,H_0^1(\Omega)^3}
:= {}& \frac{\varepsilon^3}{r_\varepsilon^3}\rho_\varepsilon^s \int_{\Omega_\varepsilon^s} \left(\partial_t u_\varepsilon(t)-u_\varepsilon^1\right)\cdot\varphi\,dx \\
&+\rho_\varepsilon^l\int_{\Omega_\varepsilon^l}\left(v_\varepsilon(t)-v_\varepsilon^0\right)\cdot\varphi\,dx
+\frac{\varepsilon^3}{r_\varepsilon}\int_{\Omega_\varepsilon^s} A_\varepsilon e_x\left(\int_0^t u_\varepsilon\,ds\right):e_x(\varphi)\,dx \\
&+2\mu\int_{\Omega_\varepsilon^l} e_x\left(\int_0^t v_\varepsilon\,ds\right):e_x(\varphi)\,dx
-\int_\Omega\int_0^t f_\varepsilon\,ds\cdot\varphi\,dx, \tag*{(26)}
\end{aligned} ⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 := r ε 3 ε 3 ρ ε s ∫ Ω ε s ( ∂ t u ε ( t ) − u ε 1 ) ⋅ φ d x + ρ ε l ∫ Ω ε l ( v ε ( t ) − v ε 0 ) ⋅ φ d x + r ε ε 3 ∫ Ω ε s A ε e x ( ∫ 0 t u ε d s ) : e x ( φ ) d x + 2 μ ∫ Ω ε l e x ( ∫ 0 t v ε d s ) : e x ( φ ) d x − ∫ Ω ∫ 0 t f ε d s ⋅ φ d x , (26)
for every φ ∈ H 0 1 ( Ω ) 3 \varphi\in H_0^1(\Omega)^3 φ ∈ H 0 1 ( Ω ) 3 , and every t ∈ [ 0 , T ] t \in[0,T] t ∈ [ 0 , T ] , satisfies
⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = 0 , ∀ φ ∈ H 0 1 ( Ω ) 3 , with div x φ = 0 in Ω ε l . (27) \langle G_\varepsilon(t),\varphi\rangle_{H^{-1}(\Omega)^3,H_0^1(\Omega)^3}=0,\qquad\forall\varphi\in H_0^1(\Omega)^3,\qquad\text{with }\operatorname{div}_x\varphi=0\text{ in }\Omega_\varepsilon^l. \tag*{(27)} ⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = 0 , ∀ φ ∈ H 0 1 ( Ω ) 3 , with div x φ = 0 in Ω ε l . (27)
Since in particular (27) holds true for every φ ∈ H 0 1 ( Ω ) 3 \varphi\in H_0^1(\Omega)^3 φ ∈ H 0 1 ( Ω ) 3 with div x φ = 0 \operatorname{div}_x\varphi=0 div x φ = 0 in Ω \Omega Ω , De Rham’s theorem proves the existence of P ε ∈ C 0 ( [ 0 , T ] ; L 2 ( Ω ) ) P_\varepsilon\in C^0([0,T];L^2(\Omega)) P ε ∈ C 0 ([ 0 , T ] ; L 2 ( Ω )) , with P ε ( 0 ) = 0 P_\varepsilon(0)=0 P ε ( 0 ) = 0 such that
⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = ∫ Ω P ε ( t ) div x φ d x , ∀ φ ∈ H 0 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] , (28) \langle G_\varepsilon(t),\varphi\rangle_{H^{-1}(\Omega)^3,H_0^1(\Omega)^3}
=\int_\Omega P_\varepsilon(t)\operatorname{div}_x\varphi\,dx,\qquad\forall\varphi\in H_0^1(\Omega)^3,\ \forall t\in[0,T], \tag*{(28)} ⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = ∫ Ω P ε ( t ) div x φ d x , ∀ φ ∈ H 0 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] , (28)
∫ Ω P ε ( t ) d x = 0 , ∥ P ε ( t ) ∥ L 2 ( Ω ) ≤ C ∥ G ε ( t ) ∥ H − 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] . (29) \int_\Omega P_\varepsilon(t)\,dx=0,\qquad\|P_\varepsilon(t)\|_{L^2(\Omega)}\le C\|G_\varepsilon(t)\|_{H^{-1}(\Omega)^3},\qquad\forall t\in[0,T]. \tag*{(29)} ∫ Ω P ε ( t ) d x = 0 , ∥ P ε ( t ) ∥ L 2 ( Ω ) ≤ C ∥ G ε ( t ) ∥ H − 1 ( Ω ) 3 , ∀ t ∈ [ 0 , T ] . (29)
Let us prove that (27) and (28) imply the existence of ζ ε ∈ C 0 ( [ 0 , T ] ) \zeta_\varepsilon\in C^0([0,T]) ζ ε ∈ C 0 ([ 0 , T ]) , such that P ε ( t ) = ζ ε ( t ) P_\varepsilon(t)=\zeta_\varepsilon(t) P ε ( t ) = ζ ε ( t ) a.e. in Ω ε s \Omega_\varepsilon^s Ω ε s , for every t ∈ [ 0 , T ] t\in[0,T] t ∈ [ 0 , T ] . For this purpose, we take k 1 , k 2 ∈ Z 3 k_1,k_2\in\mathbb{Z}^3 k 1 , k 2 ∈ Z 3 , k 1 ≠ k 2 k_1\ne k_2 k 1 = k 2 such that ε ( k 1 + Y ) , ε ( k 2 + Y ) ⊂ Ω \varepsilon(k_1+Y),\varepsilon(k_2+Y)\subset\Omega ε ( k 1 + Y ) , ε ( k 2 + Y ) ⊂ Ω , and we use that for every ψ 1 ∈ L 2 ( ε k 1 + r ε O ) \psi_1\in L^2(\varepsilon k_1+r_\varepsilon O) ψ 1 ∈ L 2 ( ε k 1 + r ε O ) , ψ 2 ∈ L 2 ( ε k 2 + r ε O ) \psi_2\in L^2(\varepsilon k_2+r_\varepsilon O) ψ 2 ∈ L 2 ( ε k 2 + r ε O ) , with
∫ ε k 1 + r ε O ψ 1 d x + ∫ ε k 2 + r ε O ψ 2 d x = 0 , (30) \int_{\varepsilon k_1+r_\varepsilon O}\psi_1\,dx+\int_{\varepsilon k_2+r_\varepsilon O}\psi_2\,dx=0, \tag*{(30)} ∫ ε k 1 + r ε O ψ 1 d x + ∫ ε k 2 + r ε O ψ 2 d x = 0 , (30)
there exists φ ∈ H 0 1 ( Ω ) 3 \varphi\in H^1_0(\Omega)^3 φ ∈ H 0 1 ( Ω ) 3 such that
div x φ = ψ 1 in ϵ k 1 + r ϵ O , div x φ = ψ 2 in ϵ k 2 + r ϵ O , div x φ = 0 otherwise . \operatorname{div}_{x}\varphi=\psi_1\ \text{in }\epsilon k_1+r_{\epsilon}\mathcal{O},\qquad\operatorname{div}_{x}\varphi=\psi_2\ \text{in }\epsilon k_2+r_{\epsilon}\mathcal{O},\qquad\operatorname{div}_{x}\varphi=0\ \text{otherwise}. div x φ = ψ 1 in ϵ k 1 + r ϵ O , div x φ = ψ 2 in ϵ k 2 + r ϵ O , div x φ = 0 otherwise .
Then, by (28)
∫ ϵ k 1 + r ϵ O P ϵ ( t ) ψ 1 d x + ∫ ϵ k 2 + r ϵ O P ϵ ( t ) ψ 2 d x = ⟨ G ϵ , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = 0. \int_{\epsilon k_1+r_{\epsilon}\mathcal{O}}P_{\epsilon}(t)\psi_1dx+\int_{\epsilon k_2+r_{\epsilon}\mathcal{O}}P_{\epsilon}(t)\psi_2dx=\langle G_{\epsilon},\varphi\rangle_{H^{-1}(\Omega)^3,H^1_0(\Omega)^3}=0. ∫ ϵ k 1 + r ϵ O P ϵ ( t ) ψ 1 d x + ∫ ϵ k 2 + r ϵ O P ϵ ( t ) ψ 2 d x = ⟨ G ϵ , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 = 0.
Since this holds for every ψ 1 , ψ 2 \psi_1,\psi_2 ψ 1 , ψ 2 satisfying (30) we deduce the existence of ζ ϵ \zeta_{\epsilon} ζ ϵ such that
P ϵ ( t ) = ζ ϵ ( t ) in ( ϵ k 1 + r ϵ O ) ∪ ( ϵ k 2 + r ϵ O ) , ∀ t ∈ [ 0 , T ] . P_{\epsilon}(t)=\zeta_{\epsilon}(t)\quad\text{in }(\epsilon k_1+r_{\epsilon}\mathcal{O})\cup(\epsilon k_2+r_{\epsilon}\mathcal{O}),\quad\forall t\in[0,T]. P ϵ ( t ) = ζ ϵ ( t ) in ( ϵ k 1 + r ϵ O ) ∪ ( ϵ k 2 + r ϵ O ) , ∀ t ∈ [ 0 , T ] .
By the arbitrariness of k 1 , k 2 k_1,k_2 k 1 , k 2 we get (21) .
In order to estimate ∥ P ϵ ( t ) ∥ L 2 ( Ω ) \|P_{\epsilon}(t)\|_{L^2(\Omega)} ∥ P ϵ ( t ) ∥ L 2 ( Ω ) let us estimate ∥ G ϵ ( t ) ∥ H − 1 ( Ω ) 3 \|G_{\epsilon}(t)\|_{H^{-1}(\Omega)^3} ∥ G ϵ ( t ) ∥ H − 1 ( Ω ) 3 and ∣ ζ ϵ ( t ) ∣ |\zeta_{\epsilon}(t)| ∣ ζ ϵ ( t ) ∣ . Since G ϵ ( t ) G_{\epsilon}(t) G ϵ ( t ) is defined as the sum of several terms, we need to estimate the norm in H − 1 ( Ω ) 3 H^{-1}(\Omega)^3 H − 1 ( Ω ) 3 of each one. The unique which presents some difficulties is the first one. We use that for every φ ∈ H 0 1 ( Ω ) 3 \varphi\in H^1_0(\Omega)^3 φ ∈ H 0 1 ( Ω ) 3 , (19) gives
ϵ 3 r ϵ 3 ∣ ∫ Ω ϵ ( ∂ t u ϵ ( t ) − u ϵ 1 ) ⋅ φ d x ∣ ≤ C ϵ 3 / 2 r ϵ 3 / 2 D ϵ ( t ) 1 2 ( ∫ Ω ϵ ∣ φ ∣ 2 d x ) 1 2 . (31) \frac{\epsilon^3}{r_{\epsilon}^3}\left|\int_{\Omega_{\epsilon}}(\partial_tu_{\epsilon}(t)-u_{\epsilon}^1)\cdot\varphi\,dx\right|\le C\frac{\epsilon^{3/2}}{r_{\epsilon}^{3/2}}D_{\epsilon}(t)^{\frac12}\left(\int_{\Omega_{\epsilon}}|\varphi|^2dx\right)^{\frac12}. \tag*{(31)} r ϵ 3 ϵ 3 ∫ Ω ϵ ( ∂ t u ϵ ( t ) − u ϵ 1 ) ⋅ φ d x ≤ C r ϵ 3/2 ϵ 3/2 D ϵ ( t ) 2 1 ( ∫ Ω ϵ ∣ φ ∣ 2 d x ) 2 1 . (31)
In the right-hand side we use that thanks to Hölder’s and Sobolev’s inequalities, for every k ∈ Z 3 k\in\mathbb{Z}^3 k ∈ Z 3 , we have
∫ ϵ k + r ϵ O ∣ φ − 1 ϵ 3 ∫ ϵ k 3 φ d y ∣ 2 d x ≤ C ( ∫ ϵ k 3 ∣ φ − 1 ϵ 3 ∫ ϵ k 3 φ d y ∣ 6 d x ) 1 3 r ϵ 2 ≤ C r ϵ 2 ∫ ϵ k 3 ∣ D x φ ∣ 2 d x , \int_{\epsilon k+r_{\epsilon}\mathcal{O}}\left|\varphi-\frac{1}{\epsilon^3}\int_{\epsilon^3_k}\varphi\,dy\right|^2dx\le C\left(\int_{\epsilon^3_k}\left|\varphi-\frac{1}{\epsilon^3}\int_{\epsilon^3_k}\varphi\,dy\right|^6dx\right)^{\frac13}r_{\epsilon}^2\le Cr_{\epsilon}^2\int_{\epsilon^3_k}|D_x\varphi|^2dx, ∫ ϵ k + r ϵ O φ − ϵ 3 1 ∫ ϵ k 3 φ d y 2 d x ≤ C ∫ ϵ k 3 φ − ϵ 3 1 ∫ ϵ k 3 φ d y 6 d x 3 1 r ϵ 2 ≤ C r ϵ 2 ∫ ϵ k 3 ∣ D x φ ∣ 2 d x ,
and then
∫ ϵ k + r ϵ O ∣ φ ∣ 2 d x ≤ C r ϵ 3 ϵ 3 ∫ ϵ k 3 ∣ φ ∣ 2 d x + C r ϵ 2 ∫ ϵ k 3 ∣ D x φ ∣ 2 d x . \int_{\epsilon k+r_{\epsilon}\mathcal{O}}|\varphi|^2dx\le C\frac{r_{\epsilon}^3}{\epsilon^3}\int_{\epsilon^3_k}|\varphi|^2dx+Cr_{\epsilon}^2\int_{\epsilon^3_k}|D_x\varphi|^2dx. ∫ ϵ k + r ϵ O ∣ φ ∣ 2 d x ≤ C ϵ 3 r ϵ 3 ∫ ϵ k 3 ∣ φ ∣ 2 d x + C r ϵ 2 ∫ ϵ k 3 ∣ D x φ ∣ 2 d x .
Adding in k k k , and replacing in (31) , we obtain
ϵ 3 r ϵ 3 ∣ ∫ Ω ϵ ( ∂ t u ϵ ( t ) − u ϵ 1 ) ⋅ φ d x ∣ ≤ C D ϵ ( t ) 1 2 ( ∫ Ω ∣ φ ∣ 2 d x + C ϵ 3 r ϵ ∫ Ω ∣ D x φ ∣ 2 d x ) 1 2 . \frac{\epsilon^3}{r_{\epsilon}^3}\left|\int_{\Omega_{\epsilon}}(\partial_tu_{\epsilon}(t)-u_{\epsilon}^1)\cdot\varphi\,dx\right|\le CD_{\epsilon}(t)^{\frac12}\left(\int_{\Omega}|\varphi|^2dx+C\frac{\epsilon^3}{r_{\epsilon}}\int_{\Omega}|D_x\varphi|^2dx\right)^{\frac12}. r ϵ 3 ϵ 3 ∫ Ω ϵ ( ∂ t u ϵ ( t ) − u ϵ 1 ) ⋅ φ d x ≤ C D ϵ ( t ) 2 1 ( ∫ Ω ∣ φ ∣ 2 d x + C r ϵ ϵ 3 ∫ Ω ∣ D x φ ∣ 2 d x ) 2 1 .
Using this estimate in the definition of G ϵ G_{\epsilon} G ϵ , and taking into account Poincaré’s inequality, we deduce
∥ G ϵ ( t ) ∥ H − 1 ( Ω ) 3 ≤ C ( 1 + ϵ 3 r ϵ ) 1 2 D ϵ ( t ) 1 2 , ∀ t ∈ [ 0 , T ] , (32) \|G_{\epsilon}(t)\|_{H^{-1}(\Omega)^3}\le C\left(1+\frac{\epsilon^3}{r_{\epsilon}}\right)^{\frac12}D_{\epsilon}(t)^{\frac12},\quad\forall t\in[0,T], \tag*{(32)} ∥ G ϵ ( t ) ∥ H − 1 ( Ω ) 3 ≤ C ( 1 + r ϵ ϵ 3 ) 2 1 D ϵ ( t ) 2 1 , ∀ t ∈ [ 0 , T ] , (32)
which combined with (29) proves the estimate for ∥ P ϵ ( t ) ∥ L 2 ( Ω ) \|P_{\epsilon}(t)\|_{L^2(\Omega)} ∥ P ϵ ( t ) ∥ L 2 ( Ω ) in (24) .
Let us now estimate ∣ ζ ϵ ( t ) ∣ |\zeta_{\epsilon}(t)| ∣ ζ ϵ ( t ) ∣ . For this purpose, we define ψ ∈ H 1 ( Y ) 3 \psi\in H^1(Y)^3 ψ ∈ H 1 ( Y ) 3 by
ψ = 1 in Y ∖ r ϵ ϵ O ‾ , ψ = ( 1 − ϵ 3 r ϵ 3 ∣ O ∣ ) in r ϵ ϵ O . \psi=1\ \text{in }Y\setminus\frac{r_{\epsilon}}{\epsilon}\overline{\mathcal{O}},\qquad\psi=\left(1-\frac{\epsilon^3}{r_{\epsilon}^3|\mathcal{O}|}\right)\ \text{in }\frac{r_{\epsilon}}{\epsilon}\mathcal{O}. ψ = 1 in Y ∖ ϵ r ϵ O , ψ = ( 1 − r ϵ 3 ∣ O ∣ ϵ 3 ) in ϵ r ϵ O .
Then, since ψ \psi ψ has null integral in Y Y Y , there exists ϕ ∈ H 0 1 ( Y ) 3 \phi\in H^1_0(Y)^3 ϕ ∈ H 0 1 ( Y ) 3 such that
div x ϕ = ψ in Y , ∥ ϕ ∥ H 0 1 ( Y ) 3 ≤ C ∥ ψ ∥ L 2 ( Y ) ≤ C ε 3 / 2 r ε 3 / 2 . \operatorname{div}_x\phi=\psi\text{ in }Y,\qquad\|\phi\|_{H^1_0(Y)^3}\le C\|\psi\|_{L^2(Y)}\le C\frac{\varepsilon^{3/2}}{r_\varepsilon^{3/2}}. div x ϕ = ψ in Y , ∥ ϕ ∥ H 0 1 ( Y ) 3 ≤ C ∥ ψ ∥ L 2 ( Y ) ≤ C r ε 3/2 ε 3/2 .
From this function ϕ \phi ϕ we define φ \varphi φ as
φ ( x ) = { ε ϕ ( x − ε k ε ) if ε ( k + Y ) ⊂ Ω , 0 otherwise . \varphi(x)=
\begin{cases}
\varepsilon\phi\left(\dfrac{x-\varepsilon k}{\varepsilon}\right) & \text{if }\varepsilon(k+Y)\subset\Omega,\\
0 & \text{otherwise}.
\end{cases} φ ( x ) = ⎩ ⎨ ⎧ εϕ ( ε x − ε k ) 0 if ε ( k + Y ) ⊂ Ω , otherwise .
Using it in (4.10) we get
( 1 − ε 3 r ε 3 ∣ O ∣ ) ζ ε ( t ) ∣ Ω ε l ∣ + ∫ Ω ε l P ε ( t , x ) d x = ⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 , \left(1-\frac{\varepsilon^3}{r_\varepsilon^3|\mathcal{O}|}\right)\zeta_\varepsilon(t)|\Omega_\varepsilon^l|+\int_{\Omega_\varepsilon^l}P_\varepsilon(t,x)\,dx=\langle G_\varepsilon(t),\varphi\rangle_{H^{-1}(\Omega)^3,H^1_0(\Omega)^3}, ( 1 − r ε 3 ∣ O ∣ ε 3 ) ζ ε ( t ) ∣ Ω ε l ∣ + ∫ Ω ε l P ε ( t , x ) d x = ⟨ G ε ( t ) , φ ⟩ H − 1 ( Ω ) 3 , H 0 1 ( Ω ) 3 ,
which recalling that P ε ( t ) P_\varepsilon(t) P ε ( t ) has null integral in Ω \Omega Ω , and then that
∫ Ω ε l P ε ( t , x ) d x = − ∫ Ω ε l P ε ( t , x ) d x = − ζ ε ( t ) ∣ Ω ε l ∣ , \int_{\Omega_\varepsilon^l}P_\varepsilon(t,x)\,dx=-\int_{\Omega_\varepsilon^l}P_\varepsilon(t,x)\,dx=-\zeta_\varepsilon(t)|\Omega_\varepsilon^l|, ∫ Ω ε l P ε ( t , x ) d x = − ∫ Ω ε l P ε ( t , x ) d x = − ζ ε ( t ) ∣ Ω ε l ∣ ,
provides
∣ ζ ε ( t ) ∣ ≤ C [ ε 3 r ε 3 ( 1 + ε 3 r ε ) ] 1 2 D ε ( t ) 1 2 . |\zeta_\varepsilon(t)|\le C\left[\frac{\varepsilon^3}{r_\varepsilon^3}\left(1+\frac{\varepsilon^3}{r_\varepsilon}\right)\right]^{\frac12}D_\varepsilon(t)^{\frac12}. ∣ ζ ε ( t ) ∣ ≤ C [ r ε 3 ε 3 ( 1 + r ε ε 3 ) ] 2 1 D ε ( t ) 2 1 .
Remark 4.2. By (4.2) and (4.6), we get the following estimate for the pressure
∥ p ε ∥ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ε l ) ) ≤ C ( 1 + ε 3 r ε 3 ) 1 2 ∥ D ε ∥ L ∞ ( 0 , T ) 1 2 . (33) \|p_\varepsilon\|_{W^{-1,\infty}(0,T;L^2(\Omega_\varepsilon^l))}\le C\left(1+\frac{\varepsilon^3}{r_\varepsilon^3}\right)^{\frac12}\|D_\varepsilon\|_{L^\infty(0,T)}^{\frac12}. \tag*{(33)} ∥ p ε ∥ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ε l )) ≤ C ( 1 + r ε 3 ε 3 ) 2 1 ∥ D ε ∥ L ∞ ( 0 , T ) 2 1 . (33)
Another estimate can be obtained using the results in [3 ] where it is proved that if L ε ∈ H − 1 ( Ω ε l ) 3 L_\varepsilon\in H^{-1}(\Omega_\varepsilon^l)^3 L ε ∈ H − 1 ( Ω ε l ) 3 satisfies
⟨ L ε , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = 0 , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 with div x ϕ = 0 in Ω ε l , \langle L_\varepsilon,\phi\rangle_{H^{-1}(\Omega_\varepsilon^l)^3,H^1_0(\Omega_\varepsilon^l)^3}=0,\qquad\forall\phi\in H^1_0(\Omega_\varepsilon^l)^3\text{ with }\operatorname{div}_x\phi=0\text{ in }\Omega_\varepsilon^l, ⟨ L ε , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = 0 , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 with div x ϕ = 0 in Ω ε l ,
then there exists q ε ∈ L 2 ( Ω ) q_\varepsilon\in L^2(\Omega) q ε ∈ L 2 ( Ω ) such that
⟨ L ε , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = ∫ Ω ε l q ε div x ϕ d x , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 , \langle L_\varepsilon,\phi\rangle_{H^{-1}(\Omega_\varepsilon^l)^3,H^1_0(\Omega_\varepsilon^l)^3}=\int_{\Omega_\varepsilon^l}q_\varepsilon\operatorname{div}_x\phi\,dx,\qquad\forall\phi\in H^1_0(\Omega_\varepsilon^l)^3, ⟨ L ε , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = ∫ Ω ε l q ε div x ϕ d x , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 ,
∥ q ε ∥ L 2 ( Ω ) ≤ C ( 1 + r ε ε ) 1 2 ∥ L ε ∥ H − 1 ( Ω ε l ) 3 , ∫ Ω q ε d x = 0. \|q_\varepsilon\|_{L^2(\Omega)}\le C\left(1+\frac{r_\varepsilon}{\varepsilon}\right)^{\frac12}\|L_\varepsilon\|_{H^{-1}(\Omega_\varepsilon^l)^3},\qquad\int_\Omega q_\varepsilon\,dx=0. ∥ q ε ∥ L 2 ( Ω ) ≤ C ( 1 + ε r ε ) 2 1 ∥ L ε ∥ H − 1 ( Ω ε l ) 3 , ∫ Ω q ε d x = 0.
Applying this result to the distribution G ε ( t ) G_\varepsilon(t) G ε ( t ) in the proof of Proposition 4.1, we deduce the existence of Q ε ∈ C 0 ( [ 0 , T ] ; L 2 ( Ω ) ) Q_\varepsilon\in C^0([0,T];L^2(\Omega)) Q ε ∈ C 0 ([ 0 , T ] ; L 2 ( Ω )) , with Q ε ( 0 ) = 0 Q_\varepsilon(0)=0 Q ε ( 0 ) = 0 , such that
⟨ G ε ( t ) , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = ∫ Ω ε l Q ε ( t , x ) div x ϕ d x , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 , \langle G_\varepsilon(t),\phi\rangle_{H^{-1}(\Omega_\varepsilon^l)^3,H^1_0(\Omega_\varepsilon^l)^3}=\int_{\Omega_\varepsilon^l}Q_\varepsilon(t,x)\operatorname{div}_x\phi\,dx,\qquad\forall\phi\in H^1_0(\Omega_\varepsilon^l)^3, ⟨ G ε ( t ) , ϕ ⟩ H − 1 ( Ω ε l ) 3 , H 0 1 ( Ω ε l ) 3 = ∫ Ω ε l Q ε ( t , x ) div x ϕ d x , ∀ ϕ ∈ H 0 1 ( Ω ε l ) 3 ,
∫ Ω Q ε d x = 0 , ∥ Q ε ∥ L 2 ( Ω ) ≤ C ( 1 + r ε ε 3 ) 1 2 ∥ G ε ∥ H − 1 ( Ω ε l ) 3 a.e. in ( 0 , T ) , (34) \int_{\Omega} Q_{\varepsilon}\,dx=0,\qquad\|Q_{\varepsilon}\|_{L^{2}(\Omega)}\le C\left(1+\frac{r_{\varepsilon}}{\varepsilon^{3}}\right)^{\frac{1}{2}}\|G_{\varepsilon}\|_{H^{-1}(\Omega_{\varepsilon}^{l})^{3}}\quad\text{a.e. in }(0,T), \tag*{(34)} ∫ Ω Q ε d x = 0 , ∥ Q ε ∥ L 2 ( Ω ) ≤ C ( 1 + ε 3 r ε ) 2 1 ∥ G ε ∥ H − 1 ( Ω ε l ) 3 a.e. in ( 0 , T ) , (34)
where using the definition of G ε G_{\varepsilon} G ε in the proof of Proposition 4.1 it is simple to check
∥ G ε ( t ) ∥ H − 1 ( Ω ε l ) 3 ≤ C D ε ( t ) 1 2 , ∀ t ∈ [ 0 , T ] . \|G_{\varepsilon}(t)\|_{H^{-1}(\Omega_{\varepsilon}^{l})^{3}}\le C D_{\varepsilon}(t)^{\frac{1}{2}},\qquad\forall t\in[0,T]. ∥ G ε ( t ) ∥ H − 1 ( Ω ε l ) 3 ≤ C D ε ( t ) 2 1 , ∀ t ∈ [ 0 , T ] .
Comparing with (28) we deduce the existence of η ε ∈ C 0 ( [ 0 , T ] ) \eta_{\varepsilon}\in C^{0}([0,T]) η ε ∈ C 0 ([ 0 , T ]) , such that the function P ε P_{\varepsilon} P ε in Proposition 4.1 also satisfies
P ε ( t , x ) = Q ε ( t , x ) + η ε ( t ) a.e. in Ω ε l , ∀ t ∈ [ 0 , T ] . (35) P_{\varepsilon}(t,x)=Q_{\varepsilon}(t,x)+\eta_{\varepsilon}(t)\quad\text{a.e. in }\Omega_{\varepsilon}^{l},\qquad\forall t\in[0,T]. \tag*{(35)} P ε ( t , x ) = Q ε ( t , x ) + η ε ( t ) a.e. in Ω ε l , ∀ t ∈ [ 0 , T ] . (35)
Combining (20) , (24) , (34) and (35) we deduce in particular the estimate
∥ ∇ x p ε ∥ W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ε l ) 3 ) ≤ C ∥ D ε ∥ L ∞ ( 0 , T ) 1 2 , (36) \|\nabla_{x}p_{\varepsilon}\|_{W^{-1,\infty}(0,T;H^{-1}(\Omega_{\varepsilon}^{l})^{3})}\le C\|D_{\varepsilon}\|_{L^{\infty}(0,T)}^{\frac{1}{2}}, \tag*{(36)} ∥ ∇ x p ε ∥ W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ε l ) 3 ) ≤ C ∥ D ε ∥ L ∞ ( 0 , T ) 2 1 , (36)
where C C C only depends on Ω \Omega Ω , O \mathcal{O} O , ρ s \rho^{s} ρ s , ρ l \rho^{l} ρ l , α \alpha α , and ∣ A ∣ |A| ∣ A ∣ .
Remark 4.3. Taking in (26) φ = φ ( t , x ) \varphi=\varphi(t,x) φ = φ ( t , x ) with φ ∈ L ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 \varphi\in L^{\infty}(0,T;H_{0}^{1}(\Omega))^{3} φ ∈ L ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 , using (28) and Fubini’s theorem, we get
∫ 0 T ∫ Ω P ε div x φ d x d t = − ε 3 r ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ ∫ 0 T φ d t d x + ε 3 r ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ φ d x d t − ρ l ∫ Ω ε l v ε 0 ⋅ ∫ 0 T φ d t d x + ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ φ d x d t + ε 3 r ε ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( ∫ t T φ d s ) d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( ∫ t T φ d s ) d x d t − ∫ 0 T ∫ Ω f ε ⋅ ∫ t T φ d s d x d t . \begin{aligned}
\int_{0}^{T}\int_{\Omega}P_{\varepsilon}\operatorname{div}_{x}\varphi\,dxdt
={}&-\frac{\varepsilon^{3}}{r_{\varepsilon}^{3}}\rho^{s}\int_{\Omega_{\varepsilon}^{s}}u_{\varepsilon}^{1}\cdot\int_{0}^{T}\varphi\,dtdx
+\frac{\varepsilon^{3}}{r_{\varepsilon}^{3}}\rho^{s}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{s}}\partial_{t}u_{\varepsilon}\cdot\varphi\,dxdt\\
&-\rho^{l}\int_{\Omega_{\varepsilon}^{l}}v_{\varepsilon}^{0}\cdot\int_{0}^{T}\varphi\,dtdx
+\rho^{l}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{l}}v_{\varepsilon}\cdot\varphi\,dxdt\\
&+\frac{\varepsilon^{3}}{r_{\varepsilon}}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{s}}A e_{x}(u_{\varepsilon}):e_{x}\left(\int_{t}^{T}\varphi\,ds\right)\,dxdt\\
&+2\mu\int_{0}^{T}\int_{\Omega_{\varepsilon}^{l}}e_{x}(v_{\varepsilon}):e_{x}\left(\int_{t}^{T}\varphi\,ds\right)\,dxdt
-\int_{0}^{T}\int_{\Omega}f_{\varepsilon}\cdot\int_{t}^{T}\varphi\,ds\,dxdt.
\end{aligned} ∫ 0 T ∫ Ω P ε div x φ d x d t = − r ε 3 ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ ∫ 0 T φ d t d x + r ε 3 ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ φ d x d t − ρ l ∫ Ω ε l v ε 0 ⋅ ∫ 0 T φ d t d x + ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ φ d x d t + r ε ε 3 ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( ∫ t T φ d s ) d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( ∫ t T φ d s ) d x d t − ∫ 0 T ∫ Ω f ε ⋅ ∫ t T φ d s d x d t .
that by rearranging the terms and renaming φ \varphi φ to the function ∫ t T φ d s \int_{t}^{T}\varphi\,ds ∫ t T φ d s , in such way that now φ \varphi φ is a function in W 1 , ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 W^{1,\infty}(0,T;H_{0}^{1}(\Omega))^{3} W 1 , ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 satisfying φ ( T , x ) = 0 \varphi(T,x)=0 φ ( T , x ) = 0 a.e. x ∈ Ω x\in\Omega x ∈ Ω , proves that the solution of the system (3.1) satisfies the variational equation
− ε 3 r ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t φ d x d t + ε 3 r ε ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( φ ) d x d t − ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω P ε div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ε ⋅ φ d x d t + ε 3 r ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ φ ( 0 , x ) d x + ρ l ∫ Ω ε l v ε 0 ⋅ φ ( 0 , x ) d x , (37) \begin{aligned}
-\frac{\varepsilon^{3}}{r_{\varepsilon}^{3}}\rho^{s}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{s}}\partial_{t}u_{\varepsilon}\cdot\partial_{t}\varphi\,dxdt
+\frac{\varepsilon^{3}}{r_{\varepsilon}}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{s}}A e_{x}(u_{\varepsilon}):e_{x}(\varphi)\,dxdt\\
-\rho^{l}\int_{0}^{T}\int_{\Omega_{\varepsilon}^{l}}v_{\varepsilon}\cdot\partial_{t}\varphi\,dxdt
+2\mu\int_{0}^{T}\int_{\Omega_{\varepsilon}^{l}}e_{x}(v_{\varepsilon}):e_{x}(\varphi)\,dxdt\\
+\int_{0}^{T}\int_{\Omega}P_{\varepsilon}\operatorname{div}_{x}\partial_{t}\varphi\,dxdt
=\int_{0}^{T}\int_{\Omega}f_{\varepsilon}\cdot\varphi\,dxdt\\
+\frac{\varepsilon^{3}}{r_{\varepsilon}^{3}}\rho^{s}\int_{\Omega_{\varepsilon}^{s}}u_{\varepsilon}^{1}\cdot\varphi(0,x)\,dx
+\rho^{l}\int_{\Omega_{\varepsilon}^{l}}v_{\varepsilon}^{0}\cdot\varphi(0,x)\,dx, \tag*{(37)}
\end{aligned} − r ε 3 ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t φ d x d t + r ε ε 3 ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( φ ) d x d t − ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω P ε div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ε ⋅ φ d x d t + r ε 3 ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ φ ( 0 , x ) d x + ρ l ∫ Ω ε l v ε 0 ⋅ φ ( 0 , x ) d x , (37)
∀ φ ∈ W 1 , ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 \forall\varphi\in W^{1,\infty}(0,T;H_{0}^{1}(\Omega))^{3} ∀ φ ∈ W 1 , ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 , such that φ ( T , x ) = 0 \varphi(T,x)=0 φ ( T , x ) = 0 a.e. in Ω \Omega Ω .
The homogenization results
The objective of this section is to carry out the homogenization of (1) . To this end we will adapt the ideas in [10 ] , which are closely related to the unfolding and two-scale convergence methods ([4 ] , [5 ] , [13 ] , [22 ] ). Our main results are given by Theorems 5.4, 5.6 and 5.8.
Taking into account Proposition 4.1, let us assume that the initial data and the right-hand side f ε f_\varepsilon f ε in (9) satisfy
∃ C > 0 such that for every ε > 0 , we have ε 3 r ε ∥ u ε 0 ∥ H 1 ( Ω ε s ) 3 2 + ε 3 r ε 3 ∥ u ε 1 ∥ L 2 ( Ω ε s ) 3 2 + ∥ u ε 0 ∥ L 2 ( Ω ε l ) 3 2 + ∥ f ε ∥ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 2 ≤ C . (38) \exists C>0\text{ such that for every }\varepsilon>0,\text{ we have}
\qquad
\frac{\varepsilon^3}{r_\varepsilon}\lVert u_\varepsilon^0\rVert_{H^1(\Omega_\varepsilon^s)^3}^2+
\frac{\varepsilon^3}{r_\varepsilon^3}\lVert u_\varepsilon^1\rVert_{L^2(\Omega_\varepsilon^s)^3}^2+
\lVert u_\varepsilon^0\rVert_{L^2(\Omega_\varepsilon^l)^3}^2+
\lVert f_\varepsilon\rVert_{L^1(0,T;L^2(\Omega))^3}^2\le C. \tag*{(38)} ∃ C > 0 such that for every ε > 0 , we have r ε ε 3 ∥ u ε 0 ∥ H 1 ( Ω ε s ) 3 2 + r ε 3 ε 3 ∥ u ε 1 ∥ L 2 ( Ω ε s ) 3 2 + ∥ u ε 0 ∥ L 2 ( Ω ε l ) 3 2 + ∥ f ε ∥ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 2 ≤ C . (38)
Moreover, to avoid a limit force given by a measure, we will assume that f ε f_\varepsilon f ε satisfies the equi-integrability condition
∀ γ > 0 , ∃ δ > 0 such that ∫ E ∥ f ε ∥ L 2 ( Ω ) 3 ≤ γ , ∀ ε > 0 , ∀ E ⊂ [ 0 , T ] , ∣ E ∣ < δ . (39) \forall\gamma>0,\quad\exists\delta>0\text{ such that }\int_E\lVert f_\varepsilon\rVert_{L^2(\Omega)^3}\le\gamma,\quad\forall\varepsilon>0,\quad\forall E\subset[0,T],\quad|E|<\delta. \tag*{(39)} ∀ γ > 0 , ∃ δ > 0 such that ∫ E ∥ f ε ∥ L 2 ( Ω ) 3 ≤ γ , ∀ ε > 0 , ∀ E ⊂ [ 0 , T ] , ∣ E ∣ < δ . (39)
In these conditions Proposition 4.1 and Remark 4.2 prove the existence of C > 0 C>0 C > 0 such that
ε 3 r ε 3 ∥ ∂ t u ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε s ) 3 ) 2 + ε 3 r ε ∥ u ε ∥ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 × 3 2 + ∥ v ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε l ) 3 ) 2 + ∥ e x ( v ε ) ∥ L 2 ( 0 , T ; L 2 ( Ω ε l ) ) 3 × 3 2 + ( 1 + r ε ε 3 ) ∥ p ε ∥ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 2 + ∥ ∇ x p ε ∥ W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ) ) 3 2 ≤ C . (40) \frac{\varepsilon^3}{r_\varepsilon^3}\lVert\partial_tu_\varepsilon\rVert_{L^\infty(0,T;L^2(\Omega_\varepsilon^s)^3)}^2+
\frac{\varepsilon^3}{r_\varepsilon}\lVert u_\varepsilon\rVert_{L^\infty(0,T;H^1(\Omega_\varepsilon^s))^{3\times3}}^2+
\lVert v_\varepsilon\rVert_{L^\infty(0,T;L^2(\Omega_\varepsilon^l)^3)}^2
+\lVert e_x(v_\varepsilon)\rVert_{L^2(0,T;L^2(\Omega_\varepsilon^l))^{3\times3}}^2
+\left(1+\frac{r_\varepsilon}{\varepsilon^3}\right)\lVert p_\varepsilon\rVert_{W^{-1,\infty}(0,T;L^2(\Omega))}^2
+\lVert\nabla_xp_\varepsilon\rVert_{W^{-1,\infty}(0,T;H^{-1}(\Omega))^3}^2\le C. \tag*{(40)} r ε 3 ε 3 ∥ ∂ t u ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε s ) 3 ) 2 + r ε ε 3 ∥ u ε ∥ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 × 3 2 + ∥ v ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε l ) 3 ) 2 + ∥ e x ( v ε ) ∥ L 2 ( 0 , T ; L 2 ( Ω ε l ) ) 3 × 3 2 + ( 1 + ε 3 r ε ) ∥ p ε ∥ W − 1 , ∞ ( 0 , T ; L 2 ( Ω )) 2 + ∥ ∇ x p ε ∥ W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ) ) 3 2 ≤ C . (40)
As usual, we want to use these estimates to extract a subsequence for which the solutions u ε u_\varepsilon u ε , v ε v_\varepsilon v ε , p ε p_\varepsilon p ε converge to some limits that we want to identify. The first difficulty arises from the fact that these sequences are defined in spaces that vary with ε \varepsilon ε . In this sense, we will use the following lemma which is proved in Theorems 4.2 and 4.4 in [23 ] .
Lemma 5.1. There exists a linear extension operator E ε : H ∂ Ω 1 ( Ω ε l ) 3 → H 0 1 ( Ω ) 3 \mathcal{E}_\varepsilon:H^1_{\partial\Omega}(\Omega_\varepsilon^l)^3\to H^1_0(\Omega)^3 E ε : H ∂ Ω 1 ( Ω ε l ) 3 → H 0 1 ( Ω ) 3 such that
E ε v = v in Ω ε l , ∀ v ∈ H ∂ Ω 1 ( Ω ε l ) 3 , \mathcal{E}_\varepsilon v=v\text{ in }\Omega_\varepsilon^l,\qquad\forall v\in H^1_{\partial\Omega}(\Omega_\varepsilon^l)^3, E ε v = v in Ω ε l , ∀ v ∈ H ∂ Ω 1 ( Ω ε l ) 3 , and
∥ E ε v ∥ H 0 1 ( Ω ) 3 ≤ C ( ∥ v ∥ L 2 ( Ω ε l ) 3 + ∥ e x ( v ) ∥ L 2 ( Ω ε l ) 3 × 3 ) , ∀ v ∈ H ∂ Ω 1 ( Ω ε l ) 3 , (41) \lVert\mathcal{E}_\varepsilon v\rVert_{H^1_0(\Omega)^3}\le C\left(\lVert v\rVert_{L^2(\Omega_\varepsilon^l)^3}+\lVert e_x(v)\rVert_{L^2(\Omega_\varepsilon^l)^{3\times3}}\right),\qquad\forall v\in H^1_{\partial\Omega}(\Omega_\varepsilon^l)^3, \tag*{(41)} ∥ E ε v ∥ H 0 1 ( Ω ) 3 ≤ C ( ∥ v ∥ L 2 ( Ω ε l ) 3 + ∥ e x ( v ) ∥ L 2 ( Ω ε l ) 3 × 3 ) , ∀ v ∈ H ∂ Ω 1 ( Ω ε l ) 3 , (41) with C C C independent of ε \varepsilon ε and v v v .
Thanks to this lemma, we can extend v ε 0 v_\varepsilon^0 v ε 0 to the whole of Ω \Omega Ω and thus, using (38) , assume that v ε 0 v_\varepsilon^0 v ε 0 is a bounded sequence in H 0 1 ( Ω ) 3 H^1_0(\Omega)^3 H 0 1 ( Ω ) 3 . Analogously, we can extend v ε v_\varepsilon v ε to Ω \Omega Ω to get a sequence which is bounded in W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 W^{1,\infty}(0,T;L^2(\Omega))^3\cap L^\infty(0,T;H^1_0(\Omega))^3 W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L ∞ ( 0 , T ; H 0 1 ( Ω ) ) 3 . Now the question is what can we do with u ε 0 u_\varepsilon^0 u ε 0 , u ε 1 u_\varepsilon^1 u ε 1 and u ε u_\varepsilon u ε . It is here where we can apply the unfolding method [5 ] , [10 ] , [13 ] , which will be also useful to study the behavior of v ε v_\varepsilon v ε near the interface ∂ Ω ε s \partial\Omega_\varepsilon^s ∂ Ω ε s . The idea is as follows:
In order to study the behavior of a sequence of measurable functions w ε w_\varepsilon w ε defined in the varying spaces Ω ε s \Omega_\varepsilon^s Ω ε s , let us study the behavior of the sequence w ^ ε : Ω × O → R \widehat{w}_\varepsilon:\Omega\times\mathcal{O}\to\mathbb{R} w ε : Ω × O → R defined as
w ^ ε ( x , y ) = w ε ( ε κ ( x ε ) + r ε y ) , a.e. ( x , y ) ∈ Ω ε ∗ × O . \widehat{w}_\varepsilon(x,y)=w_\varepsilon\left(\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_\varepsilon y\right),\qquad\text{a.e. }(x,y)\in\Omega_\varepsilon^*\times\mathcal{O}. w ε ( x , y ) = w ε ( ε κ ( ε x ) + r ε y ) , a.e. ( x , y ) ∈ Ω ε ∗ × O .
Note that for x ∈ C k ε ⊂ Ω x\in C_k^\varepsilon\subset\Omega x ∈ C k ε ⊂ Ω , the sequence w ^ ε ( x , y ) \widehat{w}_\varepsilon(x,y) w ε ( x , y ) does not depend on x x x while as function of y y y it is just the result of using the change of variables
y = x − ε k r ε , y=\frac{x-\varepsilon k}{r_\varepsilon}, y = r ε x − ε k ,
which transforms the set ε k + r ε O \varepsilon k+r_\varepsilon\mathcal{O} ε k + r ε O into O \mathcal{O} O . The variable x x x represents the macroscopic variable and y y y the microscopic one.
Analogously, we can use the same idea to study the behavior near Ω ε l \Omega_\varepsilon^l Ω ε l of a sequence w ε w_\varepsilon w ε defined in the whole of Ω \Omega Ω .
We will use the following compactness lemma
Lemma 5.2. Let u ε u_\varepsilon u ε , v ε v_\varepsilon v ε be two functions in W 1 , ∞ ( 0 , T ; L 2 ( Ω ε s ) ) 3 ∩ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 W^{1,\infty}(0,T;L^2(\Omega_\varepsilon^s))^3\cap L^\infty(0,T;H^1(\Omega_\varepsilon^s))^3 W 1 , ∞ ( 0 , T ; L 2 ( Ω ε s ) ) 3 ∩ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 and W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 W^{1,\infty}(0,T;L^2(\Omega))^3\cap L^2(0,T;H_0^1(\Omega))^3 W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 respectively, such that
ε 3 r ε 3 ∥ ∂ t u ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε s ) ) 3 2 + ε 3 r ε ∥ u ε ∥ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 2 + ∥ v ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 2 + ∥ D x v ε ∥ L 2 ( 0 , T ; L 2 ( Ω ) ) 3 × 3 2 ≤ C , (42) \frac{\varepsilon^3}{r_\varepsilon^3}\|\partial_tu_\varepsilon\|_{L^\infty(0,T;L^2(\Omega_\varepsilon^s))^3}^2+\frac{\varepsilon^3}{r_\varepsilon}\|u_\varepsilon\|_{L^\infty(0,T;H^1(\Omega_\varepsilon^s))^3}^2+\|v_\varepsilon\|_{L^\infty(0,T;L^2(\Omega))^3}^2+\|D_xv_\varepsilon\|_{L^2(0,T;L^2(\Omega))^{3\times3}}^2\leq C, \tag*{(42)} r ε 3 ε 3 ∥ ∂ t u ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ε s ) ) 3 2 + r ε ε 3 ∥ u ε ∥ L ∞ ( 0 , T ; H 1 ( Ω ε s ) ) 3 2 + ∥ v ε ∥ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 2 + ∥ D x v ε ∥ L 2 ( 0 , T ; L 2 ( Ω ) ) 3 × 3 2 ≤ C , (42) div x v ε = 0 in ( 0 , T ) × Ω ε l , (43) \operatorname{div}_xv_\varepsilon=0\quad\text{in }(0,T)\times\Omega_\varepsilon^l, \tag*{(43)} div x v ε = 0 in ( 0 , T ) × Ω ε l , (43) ∂ t u ε = v ε on ( 0 , T ) × ∂ Ω ε s , (44) \partial_tu_\varepsilon=v_\varepsilon\quad\text{on }(0,T)\times\partial\Omega_\varepsilon^s, \tag*{(44)} ∂ t u ε = v ε on ( 0 , T ) × ∂ Ω ε s , (44) then, defining u ^ ε : ( 0 , T ) × Ω ε ∗ × O → R 3 \widehat{u}_\varepsilon:(0,T)\times\Omega_\varepsilon^*\times\mathcal{O}\to\mathbb{R}^3 u ε : ( 0 , T ) × Ω ε ∗ × O → R 3 by
u ^ ε ( t , x , y ) = u ε ( t , ε κ ( x ε ) + r ε y ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω ε ∗ × O , (45) \widehat{u}_\varepsilon(t,x,y)=u_\varepsilon\left(t,\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_\varepsilon y\right)\quad\text{a.e. }(t,x,y)\in(0,T)\times\Omega_\varepsilon^*\times\mathcal{O}, \tag*{(45)} u ε ( t , x , y ) = u ε ( t , ε κ ( ε x ) + r ε y ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω ε ∗ × O , (45) there exist two functions u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 \widehat{u}\in W^{1,\infty}(0,T;L^2(\Omega;L^2(\mathcal{O})))^3\cap L^\infty(0,T;L^2(\Omega;H^1(\mathcal{O})))^3 u ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 and v ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 v\in W^{1,\infty}(0,T;L^2(\Omega))^3\cap L^2(0,T;H_0^1(\Omega))^3 v ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , with
div x v = 0 in ( 0 , T ) × Ω , (46) \operatorname{div}_xv=0\quad\text{in }(0,T)\times\Omega, \tag*{(46)} div x v = 0 in ( 0 , T ) × Ω , (46) d d t ∫ ∂ O u ^ ⋅ ν d s ( y ) = 0 in ( 0 , T ) × Ω , (47) \frac{d}{dt}\int_{\partial\mathcal{O}}\widehat{u}\cdot\nu\,ds(y)=0\quad\text{in }(0,T)\times\Omega, \tag*{(47)} d t d ∫ ∂ O u ⋅ ν d s ( y ) = 0 in ( 0 , T ) × Ω , (47) such that for a subsequence, we have
u ^ ε ⇀ ∗ u ^ in W 1 , ∞ ( 0 , T ; L 2 ( K ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( K ; H 1 ( O ) ) ) 3 \widehat{u}_\varepsilon\mathrel{\overset{*}{\rightharpoonup}}\widehat{u}\quad\text{in }W^{1,\infty}(0,T;L^2(K;L^2(\mathcal{O})))^3\cap L^\infty(0,T;L^2(K;H^1(\mathcal{O})))^3 u ε ⇀ ∗ u in W 1 , ∞ ( 0 , T ; L 2 ( K ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( K ; H 1 ( O )) ) 3 ∀ K ⊂ Ω , compact . (48) \forall K\subset\Omega,\ \text{compact}. \tag*{(48)} ∀ K ⊂ Ω , compact . (48)
v ε ⇀ v in L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 . (49) v_\varepsilon\rightharpoonup v \quad\text{in } L^2(0,T;H_0^1(\Omega))^3. \tag*{(49)} v ε ⇀ v in L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 . (49)
Moreover, if we assume that
∃ lim ε → 0 r ε ε 3 = γ ∈ ( 0 , ∞ ] , (50) \exists\lim_{\varepsilon\to0}\frac{r_\varepsilon}{\varepsilon^3}=\gamma\in(0,\infty], \tag*{(50)} ∃ ε → 0 lim ε 3 r ε = γ ∈ ( 0 , ∞ ] , (50)
then, defining v ^ ε \widehat{v}_\varepsilon v ε as (v ε v_\varepsilon v ε is assumed to be extended by zero outside Ω \Omega Ω )
v ^ ε ( t , x , y ) = v ε ( t , ε k ( x ε + r ε y ) ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω × ( ε Y / r ε ) , (51) \widehat{v}_\varepsilon(t,x,y)=v_\varepsilon\left(t,\varepsilon k\left(\frac{x}{\varepsilon}+r_\varepsilon y\right)\right)\quad\text{a.e. }(t,x,y)\in(0,T)\times\Omega\times(\varepsilon Y/r_\varepsilon), \tag*{(51)} v ε ( t , x , y ) = v ε ( t , ε k ( ε x + r ε y ) ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω × ( ε Y / r ε ) , (51)
there exists v ^ : ( 0 , T ) × Ω × R 3 → R 3 \widehat{v}:(0,T)\times\Omega\times\mathbb{R}^3\to\mathbb{R}^3 v : ( 0 , T ) × Ω × R 3 → R 3 such that, for a subsequence, we have
v ^ ε ⇀ v ^ in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B ( 0 , R ) ∖ O ) ) ) 3 , ∀ R > 0 , (52) \widehat{v}_\varepsilon\rightharpoonup\widehat{v}\quad\text{in }L^2(0,T;L^2(\Omega;H^1(B(0,R)\setminus\mathcal{O})))^3,\quad\forall R>0, \tag*{(52)} v ε ⇀ v in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B ( 0 , R ) ∖ O )) ) 3 , ∀ R > 0 , (52)
∂ t u ^ = v ^ on ( 0 , T ) × Ω × ∂ O . (53) \partial_t\widehat{u}=\widehat{v}\quad\text{on }(0,T)\times\Omega\times\partial\mathcal{O}. \tag*{(53)} ∂ t u = v on ( 0 , T ) × Ω × ∂ O . (53)
If γ = ∞ \gamma=\infty γ = ∞ , then
v ^ = v a.e. in ( 0 , T ) × Ω × R 3 . (54) \widehat{v}=v\quad\text{a.e. in }(0,T)\times\Omega\times\mathbb{R}^3. \tag*{(54)} v = v a.e. in ( 0 , T ) × Ω × R 3 . (54)
If γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , then
v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ) ) ) 3 . (55) \widehat{v}-v\in L^2(0,T;L^2(\Omega;D^{1,2}(\mathbb{R}^3)))^3. \tag*{(55)} v − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 )) ) 3 . (55)
div y v ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) (56) \operatorname{div}_y\widehat{v}=0\quad\text{in }(0,T)\times\Omega\times(\mathbb{R}^3\setminus\mathcal{O}) \tag*{(56)} div y v = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) (56)
Remark 5.3. For γ ∈ ( 0 , ∞ ] \gamma\in(0,\infty] γ ∈ ( 0 , ∞ ] conditions (5.17) or (5.18), (5.19) imply (5.10). Thus, this condition only acts for γ = 0 \gamma=0 γ = 0 .
Proof of Lemma 5.2. The existence of v v v such that (5.12) holds for a subsequence just follows from v ε v_\varepsilon v ε bounded in W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 W^{1,\infty}(0,T;L^2(\Omega))^3\cap L^2(0,T;H_0^1(\Omega))^3 W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 and having null divergence in Ω ε l \Omega_\varepsilon^l Ω ε l where the measure of Ω ∖ Ω ε l \Omega\setminus\Omega_\varepsilon^l Ω ∖ Ω ε l goes to zero.
Now, using the change of variables y = ( x − ε k ) / r ε y=(x-\varepsilon k)/r_\varepsilon y = ( x − ε k ) / r ε in each cube ε ( k + y ) \varepsilon(k+y) ε ( k + y ) , k ∈ Z 3 k\in\mathbb{Z}^3 k ∈ Z 3 , we have
∫ Ω ε l ∫ O ( ∣ ∂ t u ^ ε ∣ 2 + ∣ e y ( u ^ ε ) ∣ 2 ) d y d x = ε 3 r ε ∫ Ω ε l ∫ O ( 1 r ε 2 ∣ ∂ t u ε ∣ 2 + ∣ e x ( u ε ) ∣ 2 ) d y d x ≤ C a.e. in ( 0 , T ) . \int_{\Omega_\varepsilon^l}\int_{\mathcal{O}}\left(|\partial_t\widehat{u}_\varepsilon|^2+|e_y(\widehat{u}_\varepsilon)|^2\right)\,dy\,dx
=\frac{\varepsilon^3}{r_\varepsilon}\int_{\Omega_\varepsilon^l}\int_{\mathcal{O}}\left(\frac{1}{r_\varepsilon^2}|\partial_tu_\varepsilon|^2+|e_x(u_\varepsilon)|^2\right)\,dy\,dx\leq C\quad\text{a.e. in }(0,T). ∫ Ω ε l ∫ O ( ∣ ∂ t u ε ∣ 2 + ∣ e y ( u ε ) ∣ 2 ) d y d x = r ε ε 3 ∫ Ω ε l ∫ O ( r ε 2 1 ∣ ∂ t u ε ∣ 2 + ∣ e x ( u ε ) ∣ 2 ) d y d x ≤ C a.e. in ( 0 , T ) .
This implies the existence of u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 \widehat{u}\in W^{1,\infty}(0,T;L^2(\Omega;L^2(\mathcal{O})))^3\cap L^\infty(0,T;L^2(\Omega;H^1(\mathcal{O})))^3 u ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 such that (5.11) holds for a subsequence of ε \varepsilon ε .
Taking into account the boundary condition ∂ t u ε = v ε \partial_tu_\varepsilon=v_\varepsilon ∂ t u ε = v ε on ( 0 , T ) × ∂ Ω ε l (0,T)\times\partial\Omega_\varepsilon^l ( 0 , T ) × ∂ Ω ε l , where v ε v_\varepsilon v ε vanishes on ∂ Ω \partial\Omega ∂ Ω and has null divergence in Ω ε l \Omega_\varepsilon^l Ω ε l , we have that
∫ ε k + r ε ∂ O ∂ t u ε ⋅ ν d s = 0 in ( 0 , T ) , ∀ k ∈ R 3 , k + ε Y ⊂ Ω . \int_{\varepsilon k+r_\varepsilon\partial\mathcal{O}}\partial_tu_\varepsilon\cdot\nu\,ds=0\quad\text{in }(0,T),\quad\forall k\in\mathbb{R}^3,\ k+\varepsilon Y\subset\Omega. ∫ ε k + r ε ∂ O ∂ t u ε ⋅ ν d s = 0 in ( 0 , T ) , ∀ k ∈ R 3 , k + ε Y ⊂ Ω.
As a consequence the function u ^ \widehat{u} u must satisfy (5.10).
Now, we introduce
m ε ( t , x ) : = ∫ ε Y / r ε v ^ ε ( t , x , y ) d y = ∫ ε ( κ ( x / ε ) + Y ) v ε ( t , z ) d z , a.e. ( t , x ) ∈ ( 0 , T ) × Ω , m_\varepsilon(t,x):=\int_{\varepsilon Y/r_\varepsilon}\hat{v}_\varepsilon(t,x,y)\,dy=\int_{\varepsilon(\kappa(x/\varepsilon)+Y)}v_\varepsilon(t,z)\,dz,\qquad\text{a.e. }(t,x)\in(0,T)\times\Omega, m ε ( t , x ) := ∫ ε Y / r ε v ^ ε ( t , x , y ) d y = ∫ ε ( κ ( x / ε ) + Y ) v ε ( t , z ) d z , a.e. ( t , x ) ∈ ( 0 , T ) × Ω ,
which can be easily seen to satisfy (see e.g. [10 ] )
m ε ⇀ v in L 2 ( 0 , T ; L 2 ( R 3 ) ) . (57) m_\varepsilon\rightharpoonup v\quad\text{in }L^2(0,T;L^2(\mathbb{R}^3)). \tag*{(57)} m ε ⇀ v in L 2 ( 0 , T ; L 2 ( R 3 )) . (57)
Then, for
w ^ ε = v ^ ε − m ε , (58) \hat{w}_\varepsilon=\hat{v}_\varepsilon-m_\varepsilon, \tag*{(58)} w ^ ε = v ^ ε − m ε , (58)
we have
D y v ^ ε = D y w ^ ε a.e. in ( 0 , T ) × R 3 × ( ε Y / r ε ) , (59) D_y\hat{v}_\varepsilon=D_y\hat{w}_\varepsilon\quad\text{a.e. in }(0,T)\times\mathbb{R}^3\times(\varepsilon Y/r_\varepsilon), \tag*{(59)} D y v ^ ε = D y w ^ ε a.e. in ( 0 , T ) × R 3 × ( ε Y / r ε ) , (59)
r ε ε 3 ∫ 0 T ∫ R 3 ∫ ε r ε Y ∣ D y w ^ ε ∣ 2 d y d x d t = ∫ 0 T ∫ Ω ∣ D x v ε ∣ 2 d x d t ≤ C , (60) \frac{r_\varepsilon}{\varepsilon^3}\int_0^T\int_{\mathbb{R}^3}\int_{\frac{\varepsilon}{r_\varepsilon}Y}|D_y\hat{w}_\varepsilon|^2\,dy\,dx\,dt=\int_0^T\int_\Omega|D_xv_\varepsilon|^2\,dx\,dt\le C, \tag*{(60)} ε 3 r ε ∫ 0 T ∫ R 3 ∫ r ε ε Y ∣ D y w ^ ε ∣ 2 d y d x d t = ∫ 0 T ∫ Ω ∣ D x v ε ∣ 2 d x d t ≤ C , (60)
and then, by the Sobolev-Wirtinger inequality
r ε ε 3 ∫ 0 T ∫ R 3 ( ∫ ε r ε Y ∣ w ^ ε ∣ 6 d y ) 1 3 d x d t ≤ r ε ε 3 ∫ 0 T ∫ R 3 ∫ ε r ε Y ∣ D y w ^ ε ∣ 2 d y d x d t ≤ C . \frac{r_\varepsilon}{\varepsilon^3}\int_0^T\int_{\mathbb{R}^3}\left(\int_{\frac{\varepsilon}{r_\varepsilon}Y}|\hat{w}_\varepsilon|^6\,dy\right)^{\frac{1}{3}}dx\,dt\le\frac{r_\varepsilon}{\varepsilon^3}\int_0^T\int_{\mathbb{R}^3}\int_{\frac{\varepsilon}{r_\varepsilon}Y}|D_y\hat{w}_\varepsilon|^2\,dy\,dx\,dt\le C. ε 3 r ε ∫ 0 T ∫ R 3 ( ∫ r ε ε Y ∣ w ^ ε ∣ 6 d y ) 3 1 d x d t ≤ ε 3 r ε ∫ 0 T ∫ R 3 ∫ r ε ε Y ∣ D y w ^ ε ∣ 2 d y d x d t ≤ C .
By (5.7) we also have
∂ t u ^ ε = v ^ ε in ( 0 , T ) × Ω × ∂ O . \partial_t\hat{u}_\varepsilon=\hat{v}_\varepsilon\quad\text{in }(0,T)\times\Omega\times\partial\mathcal{O}. ∂ t u ^ ε = v ^ ε in ( 0 , T ) × Ω × ∂ O .
Therefore, for a subsequence of ε \varepsilon ε , we have:
If γ = ∞ \gamma=\infty γ = ∞ , then
v ^ ε ⇀ v in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B R ) ) ) 3 , ∀ R > 0 , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O . \hat{v}_\varepsilon\rightharpoonup v\quad\text{in }L^2(0,T;L^2(\Omega;H^1(B_R)))^3,\quad\forall R>0,\qquad\partial_t\hat{u}=v\quad\text{on }(0,T)\times\Omega\times\partial\mathcal{O}. v ^ ε ⇀ v in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B R )) ) 3 , ∀ R > 0 , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O .
If γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , then there exists v ^ : ( 0 , T ) × Ω × R 3 → R 3 \hat{v}:(0,T)\times\Omega\times\mathbb{R}^3\to\mathbb{R}^3 v ^ : ( 0 , T ) × Ω × R 3 → R 3 such that
v ^ ε ⇀ v ^ in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B R ) ) ) 3 , ∀ R > 0 , ∂ t u ^ = v ^ on ( 0 , T ) × Ω × ∂ O , (61) \hat{v}_\varepsilon\rightharpoonup\hat{v}\quad\text{in }L^2(0,T;L^2(\Omega;H^1(B_R)))^3,\quad\forall R>0,\qquad\partial_t\hat{u}=\hat{v}\quad\text{on }(0,T)\times\Omega\times\partial\mathcal{O}, \tag*{(61)} v ^ ε ⇀ v ^ in L 2 ( 0 , T ; L 2 ( Ω ; H 1 ( B R )) ) 3 , ∀ R > 0 , ∂ t u ^ = v ^ on ( 0 , T ) × Ω × ∂ O , (61)
v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ) ) ) 3 . (62) \hat{v}-v\in L^2(0,T;L^2(\Omega;D^{1,2}(\mathbb{R}^3)))^3. \tag*{(62)} v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 )) ) 3 . (62)
Since div y v ^ ε ( t , x , y ) = 0 \operatorname{div}_y\hat{v}_\varepsilon(t,x,y)=0 div y v ^ ε ( t , x , y ) = 0 for y ∉ O y\notin\mathcal{O} y ∈ / O , it is also clear that (5.19) is satisfied.
Theorem 5.4. Assume
∃ lim ε → 0 r ε ε 3 = γ ∈ [ 0 , ∞ ] , (63) \exists\lim_{\varepsilon\to0}\frac{r_\varepsilon}{\varepsilon^3}=\gamma\in[0,\infty], \tag*{(63)} ∃ ε → 0 lim ε 3 r ε = γ ∈ [ 0 , ∞ ] , (63) and that the right-hand sides and the initial data in (3.1) are such that there exist
f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 f\in L^1(0,T;L^2(\Omega))^3 f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , u ^ 0 ∈ L 2 ( Ω ; H 1 ( O ) ) 3 \hat{u}^0\in L^2(\Omega;H^1(\mathcal{O}))^3 u ^ 0 ∈ L 2 ( Ω ; H 1 ( O ) ) 3 , u ^ 1 ∈ L 2 ( Ω ; L 2 ( O ) ) 3 \hat{u}^1\in L^2(\Omega;L^2(\mathcal{O}))^3 u ^ 1 ∈ L 2 ( Ω ; L 2 ( O ) ) 3 and v 0 ∈ L 2 ( Ω ) 3 v^0\in L^2(\Omega)^3 v 0 ∈ L 2 ( Ω ) 3 such that defining
u ^ ε 0 ( x , y ) = u ε 0 ( ε κ ( x ε ) + r ε y ) , u ^ ε 1 ( x , y ) = u ε 1 ( ε κ ( x ε ) + r ε y ) , a.e. ( x , y ) ∈ Ω × O , \hat{u}_\varepsilon^0(x,y)=u_\varepsilon^0\left(\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_\varepsilon y\right),\qquad\hat{u}_\varepsilon^1(x,y)=u_\varepsilon^1\left(\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_\varepsilon y\right),\qquad\text{a.e. }(x,y)\in\Omega\times\mathcal{O}, u ^ ε 0 ( x , y ) = u ε 0 ( ε κ ( ε x ) + r ε y ) , u ^ ε 1 ( x , y ) = u ε 1 ( ε κ ( ε x ) + r ε y ) , a.e. ( x , y ) ∈ Ω × O ,
we have
f ε → f in L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , (64) f_{\varepsilon} \to f \quad\text{in } L^{1}(0,T;L^{2}(\Omega))^{3}, \tag*{(64)} f ε → f in L 1 ( 0 , T ; L 2 ( Ω ) ) 3 , (64)
u ^ ε 0 → u ^ 0 in L 2 ( Ω ; H 1 ( O ) ) 3 , (65) \hat{u}_{\varepsilon}^{0} \to\hat{u}^{0} \quad\text{in } L^{2}(\Omega;H^{1}(\mathcal{O}))^{3}, \tag*{(65)} u ^ ε 0 → u ^ 0 in L 2 ( Ω ; H 1 ( O ) ) 3 , (65)
u ^ ε 1 → u ^ 1 in L 2 ( Ω ; L 2 ( O ) ) 3 , (66) \hat{u}_{\varepsilon}^{1} \to\hat{u}^{1} \quad\text{in } L^{2}(\Omega;L^{2}(\mathcal{O}))^{3}, \tag*{(66)} u ^ ε 1 → u ^ 1 in L 2 ( Ω ; L 2 ( O ) ) 3 , (66)
v ε 0 → v 0 in L 2 ( Ω ) 3 . (67) v_{\varepsilon}^{0} \to v^{0} \quad\text{in } L^{2}(\Omega)^{3}. \tag*{(67)} v ε 0 → v 0 in L 2 ( Ω ) 3 . (67)
Then, there exist u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) 3 ) ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) 3 ) \hat{u} \in W^{1,\infty}(0,T;L^{2}(\Omega;L^{2}(\mathcal{O}))^{3}) \cap L^{\infty}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O}))^{3}) u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) 3 ) ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) 3 ) , with
u ^ ( 0 , x , y ) = u ^ 0 ( x , y ) , a.e. ( x , y ) ∈ Ω × O , (68) \hat{u}(0,x,y)=\hat{u}^{0}(x,y), \quad\text{a.e. }(x,y)\in\Omega\times\mathcal{O}, \tag*{(68)} u ^ ( 0 , x , y ) = u ^ 0 ( x , y ) , a.e. ( x , y ) ∈ Ω × O , (68)
v ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 v \in W^{1,\infty}(0,T;L^{2}(\Omega))^{3}\cap L^{2}(0,T;H_{0}^{1}(\Omega))^{3} v ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , and p ∈ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ) ) p\in W^{-1,\infty}(0,T;L^{2}(\Omega)) p ∈ W − 1 , ∞ ( 0 , T ; L 2 ( Ω )) such that the solutions u ε u_{\varepsilon} u ε , v ε v_{\varepsilon} v ε , p ε p_{\varepsilon} p ε of (3.1) satisfy (5.9), (5.11), (5.12), and
∇ x p ε ⇀ ∇ x p in W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ) ) 3 , (69) \nabla_{x}p_{\varepsilon}\rightharpoonup\nabla_{x}p \quad\text{in } W^{-1,\infty}(0,T;H^{-1}(\Omega))^{3}, \tag*{(69)} ∇ x p ε ⇀ ∇ x p in W − 1 , ∞ ( 0 , T ; H − 1 ( Ω ) ) 3 , (69)
where u ^ ε \hat{u}_{\varepsilon} u ^ ε is defined by (5.8).
The functions u ^ \hat{u} u ^ , v v v , p p p are characterized by (p p p is defined up to a distribution in W − 1 , ∞ ( 0 , T ) W^{-1,\infty}(0,T) W − 1 , ∞ ( 0 , T ) ):
If γ = 0 \gamma=0 γ = 0 ,
{ ρ s ∂ t t 2 u ^ − div y ( A e y ( u ^ ) ) = 0 in ( 0 , T ) × Ω × O , ∂ t ∫ ∂ O u ^ d s ( y ) = 0 , A e y ( u ^ ) ν ^ = c ( t , x ) ν on ( 0 , T ) × Ω × ∂ O , u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in Ω × O . (70) \left\{
\begin{aligned}
\rho^{s}\partial_{tt}^{2}\hat{u}-\operatorname{div}_{y}(A e_{y}(\hat{u}))&=0 &&\text{in }(0,T)\times\Omega\times\mathcal{O},\\
\partial_{t}\int_{\partial\mathcal{O}}\hat{u}\,ds(y)&=0,\quad A e_{y}(\hat{u})\hat{\nu}=c(t,x)\nu&&\text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
\hat{u}|_{t=0}&=\hat{u}^{0},\quad\partial_{t}\hat{u}|_{t=0}=\hat{u}^{1} &&\text{in }\Omega\times\mathcal{O}.
\end{aligned}
\right. \tag*{(70)} ⎩ ⎨ ⎧ ρ s ∂ tt 2 u ^ − div y ( A e y ( u ^ )) ∂ t ∫ ∂ O u ^ d s ( y ) u ^ ∣ t = 0 = 0 = 0 , A e y ( u ^ ) ν ^ = c ( t , x ) ν = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in ( 0 , T ) × Ω × O , on ( 0 , T ) × Ω × ∂ O , in Ω × O . (70)
{ ρ l ∂ t v − 2 μ div x e x ( v ) + ∇ x p = f in ( 0 , T ) × Ω , div x v = 0 in ( 0 , T ) × Ω , v = 0 on ( 0 , T ) × ∂ Ω , v ∣ t = 0 = v 0 in Ω . (71) \left\{
\begin{aligned}
\rho^{l}\partial_{t}v-2\mu\operatorname{div}_{x}e_{x}(v)+\nabla_{x}p&=f &&\text{in }(0,T)\times\Omega,\\
\operatorname{div}_{x}v&=0 &&\text{in }(0,T)\times\Omega,\\
v&=0 &&\text{on }(0,T)\times\partial\Omega,\\
v|_{t=0}&=v^{0} &&\text{in }\Omega.
\end{aligned}
\right. \tag*{(71)} ⎩ ⎨ ⎧ ρ l ∂ t v − 2 μ div x e x ( v ) + ∇ x p div x v v v ∣ t = 0 = f = 0 = 0 = v 0 in ( 0 , T ) × Ω , in ( 0 , T ) × Ω , on ( 0 , T ) × ∂ Ω , in Ω. (71)
If γ = ∞ \gamma=\infty γ = ∞ , then
{ ρ s ∂ t t 2 ∫ O u ^ d y + ρ l ∂ t v − 2 μ div x ( e x ( v ) ) + ∇ x p = f in ( 0 , T ) × Ω , div x v = 0 in ( 0 , T ) × Ω , ρ s ∂ t t 2 u ^ − div y ( A e y ( u ^ ) ) = 0 in ( 0 , T ) × Ω × O , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , v = 0 on ( 0 , T ) × ∂ Ω , v ∣ t = 0 = v 0 , u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in Ω × O . (72) \left\{
\begin{aligned}
\rho^{s}\partial_{tt}^{2}\int_{\mathcal{O}}\hat{u}\,dy+\rho^{l}\partial_{t}v-2\mu\operatorname{div}_{x}(e_{x}(v))+\nabla_{x}p&=f &&\text{in }(0,T)\times\Omega,\\
\operatorname{div}_{x}v&=0 &&\text{in }(0,T)\times\Omega,\\
\rho^{s}\partial_{tt}^{2}\hat{u}-\operatorname{div}_{y}(A e_{y}(\hat{u}))&=0 &&\text{in }(0,T)\times\Omega\times\mathcal{O},\\
\partial_{t}\hat{u}&=v &&\text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
v&=0 &&\text{on }(0,T)\times\partial\Omega,\\
v|_{t=0}&=v^{0},\quad\hat{u}|_{t=0}=\hat{u}^{0},\quad\partial_{t}\hat{u}|_{t=0}=\hat{u}^{1} &&\text{in }\Omega\times\mathcal{O}.
\end{aligned}
\right. \tag*{(72)} ⎩ ⎨ ⎧ ρ s ∂ tt 2 ∫ O u ^ d y + ρ l ∂ t v − 2 μ div x ( e x ( v )) + ∇ x p div x v ρ s ∂ tt 2 u ^ − div y ( A e y ( u ^ )) ∂ t u ^ v v ∣ t = 0 = f = 0 = 0 = v = 0 = v 0 , u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in ( 0 , T ) × Ω , in ( 0 , T ) × Ω , in ( 0 , T ) × Ω × O , on ( 0 , T ) × Ω × ∂ O , on ( 0 , T ) × ∂ Ω , in Ω × O . (72)
If γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , there also exist v ^ \hat{v} v ^ , with v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O ) ) 3 ) \hat{v}-v\in L^{2}(0,T;L^{2}(\Omega;D^{1,2}(\mathbb{R}^{3}\setminus\mathcal{O}))^{3}) v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O ) ) 3 ) , and p ^ ∈ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( R 3 ∖ O ) ) ) \hat{p}\in W^{-1,\infty}(0,T;L^{2}(\Omega;L^{2}(\mathbb{R}^{3}\setminus\mathcal{O}))) p ^ ∈ W − 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( R 3 ∖ O ))) such that defining v ^ ε \hat{v}_{\varepsilon} v ^ ε by (5.14) and p ^ ε \hat{p}_{\varepsilon} p ^ ε by
p ^ ε ( t , x , y ) = p ε ( t , ε κ ( x ε ) + r ε y ) in ( 0 , T ) × Ω × ( ε Y / r ε ) , (73) \hat{p}_{\varepsilon}(t,x,y)=p_{\varepsilon}\left(t,\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_{\varepsilon}y\right)\quad\text{in }(0,T)\times\Omega\times(\varepsilon Y/r_{\varepsilon}), \tag*{(73)} p ^ ε ( t , x , y ) = p ε ( t , ε κ ( ε x ) + r ε y ) in ( 0 , T ) × Ω × ( ε Y / r ε ) , (73)
we have (5.24), (5.25) and
∇ y p ^ ε ⇀ ∇ y p ^ in W − 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( B ( 0 , R ) ∖ O ) ) ) 3 . (74) \nabla_{y}\hat{p}^{\varepsilon} \rightharpoonup\nabla_{y}\hat{p}\ \text{in}\ W^{-1,\infty}(0,T;L^{2}(\Omega;L^{2}(B(0,R)\setminus\mathcal{O})))^{3}. \tag*{(74)} ∇ y p ^ ε ⇀ ∇ y p ^ in W − 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( B ( 0 , R ) ∖ O )) ) 3 . (74)
with u ^ , v , p , v ^ , p ^ \hat{u},v,p,\hat{v},\hat{p} u ^ , v , p , v ^ , p ^ defined by
ρ s ∂ t t 2 ∫ O u ^ d y + ρ l ∂ t v − 2 μ div x e x ( v ) + ∇ x p = f in ( 0 , T ) × Ω , div x v = 0 in ( 0 , T ) × Ω , ρ s ∂ t t 2 u ^ − div y ( A e y ( u ^ ) ) = 0 in ( 0 , T ) × Ω × O , − 2 μ div y e y ( v ^ ) + ∇ y p ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , div y v ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , v = 0 on ( 0 , T ) × ∂ Ω , ∂ t u ^ = v ^ , A e y ( u ^ ) ν = γ ( 2 μ e y ( v ^ ) − p ^ I ) ν on ( 0 , T ) × Ω × ∂ O , v ∣ t = 0 = v 0 , u ^ ∣ t = 0 = u 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in Ω × O . (75) \begin{aligned}
\rho^{s}\partial_{tt}^{2}\int_{\mathcal{O}}\hat{u}\,dy+\rho^{l}\partial_{t}v-2\mu\operatorname{div}_{x}e_{x}(v)+\nabla_{x}p&=f &&\text{in }(0,T)\times\Omega,\\
\operatorname{div}_{x}v&=0 &&\text{in }(0,T)\times\Omega,\\
\rho^{s}\partial_{tt}^{2}\hat{u}-\operatorname{div}_{y}(Ae_{y}(\hat{u}))&=0 &&\text{in }(0,T)\times\Omega\times\mathcal{O},\\
-2\mu\operatorname{div}_{y}e_{y}(\hat{v})+\nabla_{y}\hat{p}&=0 &&\text{in }(0,T)\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
\operatorname{div}_{y}\hat{v}&=0 &&\text{in }(0,T)\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
v&=0 &&\text{on }(0,T)\times\partial\Omega,\\
\partial_{t}\hat{u}&=\hat{v},\quad Ae_{y}(\hat{u})\nu=\gamma(2\mu e_{y}(\hat{v})-\hat{p}I)\nu&&\text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
v|_{t=0}&=v^{0},\quad\hat{u}|_{t=0}=u^{0},\quad\partial_{t}\hat{u}|_{t=0}=\hat{u}^{1} &&\text{in }\Omega\times\mathcal{O}. \tag*{(75)}
\end{aligned} ρ s ∂ tt 2 ∫ O u ^ d y + ρ l ∂ t v − 2 μ div x e x ( v ) + ∇ x p div x v ρ s ∂ tt 2 u ^ − div y ( A e y ( u ^ )) − 2 μ div y e y ( v ^ ) + ∇ y p ^ div y v ^ v ∂ t u ^ v ∣ t = 0 = f = 0 = 0 = 0 = 0 = 0 = v ^ , A e y ( u ^ ) ν = γ ( 2 μ e y ( v ^ ) − p ^ I ) ν = v 0 , u ^ ∣ t = 0 = u 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in ( 0 , T ) × Ω , in ( 0 , T ) × Ω , in ( 0 , T ) × Ω × O , in ( 0 , T ) × Ω × ( R 3 ∖ O ) , in ( 0 , T ) × Ω × ( R 3 ∖ O ) , on ( 0 , T ) × ∂ Ω , on ( 0 , T ) × Ω × ∂ O , in Ω × O . (75)
Remark 5.5. Reasoning similarly to Lemma 5.2 it can be proved that (5.1) implies that, for a subsequence of ε \varepsilon ε , there exist u ^ 0 ∈ L 2 ( Ω ; H 1 ( O ) ) 3 \hat{u}^{0}\in L^{2}(\Omega;H^{1}(\mathcal{O}))^{3} u ^ 0 ∈ L 2 ( Ω ; H 1 ( O ) ) 3 , u ^ 1 ∈ L 2 ( Ω ; L 2 ( O ) ) 3 \hat{u}^{1}\in L^{2}(\Omega;L^{2}(\mathcal{O}))^{3} u ^ 1 ∈ L 2 ( Ω ; L 2 ( O ) ) 3 and v 0 ∈ L 2 ( Ω ) 3 v^{0}\in L^{2}(\Omega)^{3} v 0 ∈ L 2 ( Ω ) 3 such that (5.28), (5.29), (5.30) hold. Therefore, these assumptions are not restrictive.
We also remark that due to div x v 0 = 0 \operatorname{div}_{x}v^{0}=0 div x v 0 = 0 in Ω ℓ ε \Omega_{\ell}^{\varepsilon} Ω ℓ ε , we have
div x v 0 = 0 in Ω . (76) \operatorname{div}_{x}v^{0}=0\quad\text{in }\Omega. \tag*{(76)} div x v 0 = 0 in Ω. (76)
The existence of f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 f\in L^{1}(0,T;L^{2}(\Omega))^{3} f ∈ L 1 ( 0 , T ; L 2 ( Ω ) ) 3 satisfying (5.27) for a subsequence of ε \varepsilon ε , also follows from (5.1) and (5.2).
Proof of Theorem 5.4. Thanks to (5.3), we can apply the compactness Lemma 5.2 to deduce the existence of u ^ \hat{u} u ^ , v v v and v ^ \hat{v} v ^ the functions given by this lemma, which exist for a subsequence.
By (5.11), the function u ^ \hat{u} u ^ satisfies
u ^ ( 0 , x , y ) = u ^ 0 ( x , y ) a.e. in Ω × O . (77) \hat{u}(0,x,y)=\hat{u}^{0}(x,y)\quad\text{a.e. in }\Omega\times\mathcal{O}. \tag*{(77)} u ^ ( 0 , x , y ) = u ^ 0 ( x , y ) a.e. in Ω × O . (77)
Let us consider φ ∈ C 1 ( [ 0 , T ] ; C c 1 ( Ω ) ) 3 \varphi\in C^{1}([0,T];C_{c}^{1}(\Omega))^{3} φ ∈ C 1 ([ 0 , T ] ; C c 1 ( Ω ) ) 3 and ϕ ∈ C 1 ( [ 0 , T ] × Ω ‾ × R 3 ) 3 \phi\in C^{1}([0,T]\times\overline{\Omega}\times\mathbb{R}^{3})^{3} ϕ ∈ C 1 ([ 0 , T ] × Ω × R 3 ) 3 , such that
φ ( T , x ) = ϕ ( T , x , y ) = 0 , ∀ ( x , y ) ∈ Ω × R 3 , (78) \varphi(T,x)=\phi(T,x,y)=0,\quad\forall(x,y)\in\Omega\times\mathbb{R}^{3}, \tag*{(78)} φ ( T , x ) = ϕ ( T , x , y ) = 0 , ∀ ( x , y ) ∈ Ω × R 3 , (78)
∫ ∂ O ϕ ⋅ ν d s ( y ) = 0 in [ 0 , T ] × Ω . (79) \int_{\partial\mathcal{O}}\phi\cdot\nu\,ds(y)=0\quad\text{in }[0,T]\times\Omega. \tag*{(79)} ∫ ∂ O ϕ ⋅ ν d s ( y ) = 0 in [ 0 , T ] × Ω. (79)
{ If γ = ∞ ⇒ ϕ ( t , x , y ) = φ ( t , x ) , ∀ ( t , x , y ) ∈ [ 0 , T ] × Ω × ( R 3 ∖ O ) , If γ < ∞ ⇒ ∃ R > 0 s.t. ϕ ( t , x , y ) = φ ( t , x ) , ∀ ( t , x , y ) ∈ [ 0 , T ] × Ω × { ∣ y ∣ > R } . (80) \left\{
\begin{aligned}
\text{If }\gamma=\infty&\Rightarrow\phi(t,x,y)=\varphi(t,x),\quad\forall(t,x,y)\in[0,T]\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
\text{If }\gamma<\infty&\Rightarrow\exists R>0\ \text{s.t. }\phi(t,x,y)=\varphi(t,x),\quad\forall(t,x,y)\in[0,T]\times\Omega\times\{|y|>R\}.
\end{aligned}
\right. \tag*{(80)} { If γ = ∞ If γ < ∞ ⇒ ϕ ( t , x , y ) = φ ( t , x ) , ∀ ( t , x , y ) ∈ [ 0 , T ] × Ω × ( R 3 ∖ O ) , ⇒ ∃ R > 0 s.t. ϕ ( t , x , y ) = φ ( t , x ) , ∀ ( t , x , y ) ∈ [ 0 , T ] × Ω × { ∣ y ∣ > R } . (80)
For such φ \varphi φ and ϕ \phi ϕ , we define Φ ε : [ 0 , T ] × Ω → R \Phi_\varepsilon: [0,T] \times\Omega\to\mathbb{R} Φ ε : [ 0 , T ] × Ω → R as
Φ ε ( t , x ) = ϕ ( t , x , x − ε k r ε ) a.e. in ( 0 , T ) × ε ( k + Y ) , ∀ k ∈ Z 3 . \Phi_\varepsilon(t,x)=\phi\left(t,x,\frac{x-\varepsilon k}{r_\varepsilon}\right)\quad\text{a.e. in }(0,T)\times\varepsilon(k+Y),\ \forall k\in\mathbb{Z}^{3}. Φ ε ( t , x ) = ϕ ( t , x , r ε x − ε k ) a.e. in ( 0 , T ) × ε ( k + Y ) , ∀ k ∈ Z 3 .
Using Φ ε \Phi_\varepsilon Φ ε as test function in (37) we get
− ε 3 r ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t Φ ε d x d t + ε 3 r ε ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( Φ ε ) d x d t − ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t Φ ε d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( Φ ε ) d x d t + ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω f ε ⋅ Φ ε d x d t + ε 3 r ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ Φ ε ( 0 , x ) d x + ρ l ∫ Ω ε l v ε 0 ⋅ Φ ε ( 0 , x ) d x . (81) -\frac{\varepsilon^{3}}{r_\varepsilon^{3}}\rho^{s}\int_{0}^{T}\int_{\Omega_\varepsilon^{s}}\partial_tu_\varepsilon\cdot\partial_t\Phi_\varepsilon\,dxdt+\frac{\varepsilon^{3}}{r_\varepsilon}\int_{0}^{T}\int_{\Omega_\varepsilon^{s}}A e_x(u_\varepsilon):e_x(\Phi_\varepsilon)\,dxdt
\\
-\rho^{l}\int_{0}^{T}\int_{\Omega_\varepsilon^{l}}v_\varepsilon\cdot\partial_t\Phi_\varepsilon\,dxdt+2\mu\int_{0}^{T}\int_{\Omega_\varepsilon^{l}}e_x(v_\varepsilon):e_x(\Phi_\varepsilon)\,dxdt
\\
+\int_{0}^{T}\int_{\Omega}P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dxdt=\int_{0}^{T}\int_{\Omega}f_\varepsilon\cdot\Phi_\varepsilon\,dxdt
\\
+\frac{\varepsilon^{3}}{r_\varepsilon^{3}}\rho^{s}\int_{\Omega_\varepsilon^{s}}u_\varepsilon^{1}\cdot\Phi_\varepsilon(0,x)\,dx+\rho^{l}\int_{\Omega_\varepsilon^{l}}v_\varepsilon^{0}\cdot\Phi_\varepsilon(0,x)\,dx. \tag*{(81)} − r ε 3 ε 3 ρ s ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t Φ ε d x d t + r ε ε 3 ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( Φ ε ) d x d t − ρ l ∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t Φ ε d x d t + 2 μ ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( Φ ε ) d x d t + ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω f ε ⋅ Φ ε d x d t + r ε 3 ε 3 ρ s ∫ Ω ε s u ε 1 ⋅ Φ ε ( 0 , x ) d x + ρ l ∫ Ω ε l v ε 0 ⋅ Φ ε ( 0 , x ) d x . (81)
Let us pass to the limit in each of the terms that appear in (81) .
Terms 1 and 2. Performing the change of variable y = ( x − ε k ) / r ε y=(x-\varepsilon k)/r_\varepsilon y = ( x − ε k ) / r ε in each solid inclusion ε k + r ε O \varepsilon k+r_\varepsilon\mathcal{O} ε k + r ε O , k ∈ Z 3 k\in\mathbb{Z}^{3} k ∈ Z 3 , and using (48) , we get
ε 3 r ε 3 ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ε ⋅ ∂ t ϕ d y d x d t + O ε → ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t ϕ d y d x d t . (82) \begin{aligned}
\frac{\varepsilon^{3}}{r_\varepsilon^{3}}\int_{0}^{T}\int_{\Omega_\varepsilon^{s}}\partial_tu_\varepsilon\cdot\partial_t\Phi_\varepsilon\,dxdt
&=\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}\partial_t\widehat{u}_\varepsilon\cdot\partial_t\phi\,dydxdt+O_\varepsilon\\
&\to\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}\partial_t\widehat{u}\cdot\partial_t\phi\,dydxdt. \tag*{(82)}
\end{aligned} r ε 3 ε 3 ∫ 0 T ∫ Ω ε s ∂ t u ε ⋅ ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ∫ O ∂ t u ε ⋅ ∂ t ϕ d y d x d t + O ε → ∫ 0 T ∫ Ω ∫ O ∂ t u ⋅ ∂ t ϕ d y d x d t . (82)
ε 3 r ε ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( Φ ε ) d x d t = ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ε ) : e y ( ϕ ) d y d x d t + O ε → ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( ϕ ) d y d x d t . (83) \begin{aligned}
\frac{\varepsilon^{3}}{r_\varepsilon}\int_{0}^{T}\int_{\Omega_\varepsilon^{s}}A e_x(u_\varepsilon):e_x(\Phi_\varepsilon)\,dxdt
&=\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}A e_y(\widehat{u}_\varepsilon):e_y(\phi)\,dydxdt+O_\varepsilon\\
&\to\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}A e_y(\widehat{u}):e_y(\phi)\,dydxdt. \tag*{(83)}
\end{aligned} r ε ε 3 ∫ 0 T ∫ Ω ε s A e x ( u ε ) : e x ( Φ ε ) d x d t = ∫ 0 T ∫ Ω ∫ O A e y ( u ε ) : e y ( ϕ ) d y d x d t + O ε → ∫ 0 T ∫ Ω ∫ O A e y ( u ) : e y ( ϕ ) d y d x d t . (83)
Term 3. By (80) we have that Φ ε ( t , x ) = φ ( t , x ) \Phi_\varepsilon(t,x)=\varphi(t,x) Φ ε ( t , x ) = φ ( t , x ) except in a subset of Ω \Omega Ω , independent of t ∈ ( 0 , T ) t\in(0,T) t ∈ ( 0 , T ) , whose measure tends to zero with ε \varepsilon ε . Thus, using the Lebesgue dominated convergence theorem, we get
Φ ε → φ in H 1 ( 0 , T ; L 2 ( Ω ) ) 3 . (84) \Phi_\varepsilon\to\varphi\quad\text{in }H^{1}(0,T;L^{2}(\Omega))^{3}. \tag*{(84)} Φ ε → φ in H 1 ( 0 , T ; L 2 ( Ω ) ) 3 . (84)
By (49) , we then have
∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t . (85) \int_{0}^{T}\int_{\Omega_\varepsilon^{l}}v_\varepsilon\cdot\partial_t\Phi_\varepsilon\,dxdt\to\int_{0}^{T}\int_{\Omega}v\cdot\partial_t\varphi\,dxdt. \tag*{(85)} ∫ 0 T ∫ Ω ε l v ε ⋅ ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t . (85)
Term 4. If γ = ∞ \gamma=\infty γ = ∞ , then by (80) , Φ ε = φ \Phi_\varepsilon=\varphi Φ ε = φ in ( 0 , T ) × Ω ε l (0,T)\times\Omega_\varepsilon^{l} ( 0 , T ) × Ω ε l and thus, taking into account that ∣ Ω ε l ∣ → 0 |\Omega_\varepsilon^{l}|\to0 ∣ Ω ε l ∣ → 0 , we get
∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( Φ ε ) d x d t → ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t . (86) \int_{0}^{T}\int_{\Omega_\varepsilon^{l}}e_x(v_\varepsilon):e_x(\Phi_\varepsilon)\,dxdt\to\int_{0}^{T}\int_{\Omega}e_x(v):e_x(\varphi)\,dxdt. \tag*{(86)} ∫ 0 T ∫ Ω ε l e x ( v ε ) : e x ( Φ ε ) d x d t → ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t . (86)
For γ < ∞ \gamma< \infty γ < ∞ , taking into account (80) , we also have for every k ∈ Z 3 k \in\mathbb{Z}^{3} k ∈ Z 3
D x Φ ε = ( D x φ ( t , x , x − ε k r ε ) + 1 r ε D y φ ( t , x , x − ε k r ε ) ) χ { ∣ x − ε k ∣ < R r ε } + D x φ χ { ∣ x − ε k ∣ > R r ε } a.e. in ε ( k + Y ) . (87) D_x\Phi_\varepsilon= \left(D_x\varphi\left(t,x,\frac{x-\varepsilon k}{r_\varepsilon}\right)+\frac{1}{r_\varepsilon}D_y\varphi\left(t,x,\frac{x-\varepsilon k}{r_\varepsilon}\right)\right)\chi_{\{|x-\varepsilon k|<Rr_\varepsilon\}}+D_x\varphi\chi_{\{|x-\varepsilon k|>Rr_\varepsilon\}}\quad\text{a.e. in }\varepsilon(k+Y). \tag*{(87)} D x Φ ε = ( D x φ ( t , x , r ε x − ε k ) + r ε 1 D y φ ( t , x , r ε x − ε k ) ) χ { ∣ x − ε k ∣ < R r ε } + D x φ χ { ∣ x − ε k ∣ > R r ε } a.e. in ε ( k + Y ) . (87)
If γ = 0 \gamma=0 γ = 0 , then
1 r ε 2 ∑ k ∈ Z 3 ∫ C k ε ∩ Ω ε ∣ D y Φ ( x − ε k r ε ) ∣ 2 d x ≤ C r ε ε 3 → 0 , (88) \frac{1}{r_\varepsilon^2}\sum_{k\in\mathbb{Z}^3}\int_{C_k^\varepsilon\cap\Omega_\varepsilon}\left|D_y\Phi\left(\frac{x-\varepsilon k}{r_\varepsilon}\right)\right|^2\,dx\leq C\frac{r_\varepsilon}{\varepsilon^3}\to0, \tag*{(88)} r ε 2 1 k ∈ Z 3 ∑ ∫ C k ε ∩ Ω ε D y Φ ( r ε x − ε k ) 2 d x ≤ C ε 3 r ε → 0 , (88)
and therefore, similarly to (84) we have
D x Φ ε → D x φ in L 2 ( 0 , T ; L 2 ( Ω ) ) 3 × 3 , D_x\Phi_\varepsilon\to D_x\varphi\quad\text{in }L^2(0,T;L^2(\Omega))^{3\times3}, D x Φ ε → D x φ in L 2 ( 0 , T ; L 2 ( Ω ) ) 3 × 3 ,
which implies that (86) also holds true for γ = 0 \gamma=0 γ = 0 .
In the case γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , taking into account (87) , performing the change of variables y = ( x − ε k ) / r ε y=(x-\varepsilon k)/r_\varepsilon y = ( x − ε k ) / r ε in each cube ε ( k + Y ) \varepsilon(k+Y) ε ( k + Y ) , and using (5.43, (49) and (52) , we get
∫ 0 T ∫ Ω ε e x ( v ε ) : e x ( Φ ε ) d x d t = ∑ k ∈ Z 3 ∫ 0 T ∫ Ω ε ∩ ( ε k + B R r ε ) e x ( v ε ) : e x ( Φ ε ) d x d t + ∑ k ∈ Z 3 ∫ 0 T ∫ Ω ε ∩ ( C k ε ∖ ( ε k + B R r ε ) ) e x ( v ε ) : e x ( Φ ε ) d x d t = r ε ε 3 ∫ 0 T ∫ Ω × ( B R ∖ O ) e y ( v ^ ε ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω ε e x ( v ε ) : e x ( φ ) d x d t + O ε → γ ∫ 0 T ∫ Ω × ( B R ∖ O ) e y ( v ^ ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t . (89) \begin{aligned}
\int_0^T\int_{\Omega_\varepsilon}e_x(v_\varepsilon):e_x(\Phi_\varepsilon)\,dx\,dt
={}&\sum_{k\in\mathbb{Z}^3}\int_0^T\int_{\Omega_\varepsilon\cap(\varepsilon k+B_{Rr_\varepsilon})}e_x(v_\varepsilon):e_x(\Phi_\varepsilon)\,dx\,dt\\
&+\sum_{k\in\mathbb{Z}^3}\int_0^T\int_{\Omega_\varepsilon\cap(C_k^\varepsilon\setminus(\varepsilon k+B_{Rr_\varepsilon}))}e_x(v_\varepsilon):e_x(\Phi_\varepsilon)\,dx\,dt\\
={}&\frac{r_\varepsilon}{\varepsilon^3}\int_0^T\int_{\Omega\times(B_R\setminus O)}e_y(\widehat v_\varepsilon):e_y(\varphi)\,dx\,dy\,dt+\int_0^T\int_{\Omega_\varepsilon}e_x(v_\varepsilon):e_x(\varphi)\,dx\,dt+O_\varepsilon\\
\to{}&\gamma\int_0^T\int_{\Omega\times(B_R\setminus O)}e_y(\widehat v):e_y(\varphi)\,dx\,dy\,dt+\int_0^T\int_\Omega e_x(v):e_x(\varphi)\,dx\,dt. \tag*{(89)}
\end{aligned} ∫ 0 T ∫ Ω ε e x ( v ε ) : e x ( Φ ε ) d x d t = = → k ∈ Z 3 ∑ ∫ 0 T ∫ Ω ε ∩ ( ε k + B R r ε ) e x ( v ε ) : e x ( Φ ε ) d x d t + k ∈ Z 3 ∑ ∫ 0 T ∫ Ω ε ∩ ( C k ε ∖ ( ε k + B R r ε )) e x ( v ε ) : e x ( Φ ε ) d x d t ε 3 r ε ∫ 0 T ∫ Ω × ( B R ∖ O ) e y ( v ε ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω ε e x ( v ε ) : e x ( φ ) d x d t + O ε γ ∫ 0 T ∫ Ω × ( B R ∖ O ) e y ( v ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t . (89)
Term 5. If γ = 0 \gamma=0 γ = 0 , using (21) , (35) and (79) we have
∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ε Q ε div x ∂ t Φ ε d x d t \int_0^T\int_\Omega P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dx\,dt=\int_0^T\int_{\Omega_\varepsilon}Q_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dx\,dt ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ε Q ε div x ∂ t Φ ε d x d t
where thanks to (34) , and γ = 0 \gamma=0 γ = 0 , we can assume the existence of Q ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) Q\in L^\infty(0,T;L^2(\Omega)) Q ∈ L ∞ ( 0 , T ; L 2 ( Ω )) with null integral in Ω \Omega Ω such that
Q ε ⇀ ∗ Q in L ∞ ( 0 , T ; L 2 ( Ω ) ) . (90) Q_\varepsilon\overset{*}{\rightharpoonup}Q\quad\text{in }L^\infty(0,T;L^2(\Omega)). \tag*{(90)} Q ε ⇀ ∗ Q in L ∞ ( 0 , T ; L 2 ( Ω )) . (90)
Using then (87) and (88) , applied to ∂ t Φ ε \partial_t\Phi_\varepsilon ∂ t Φ ε instead of Φ ε \Phi_\varepsilon Φ ε , we get
∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω Q div x ∂ t φ d x d t . (91) \int_0^T\int_\Omega P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dx\,dt\to\int_0^T\int_\Omega Q\operatorname{div}_x\partial_t\varphi\,dx\,dt. \tag*{(91)} ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω Q div x ∂ t φ d x d t . (91)
If γ > 0 \gamma>0 γ > 0 , (21) and (79) imply
∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ε P ε div x ∂ t Φ ε d x d t . \int_0^T\int_{\Omega} P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dxdt=\int_0^T\int_{\Omega_\varepsilon} P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dxdt. ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t = ∫ 0 T ∫ Ω ε P ε div x ∂ t Φ ε d x d t .
where thanks to (4.6), and extracting a subsequence if necessary, we can assume the existence of P ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) P\in L^\infty(0,T;L^2(\Omega)) P ∈ L ∞ ( 0 , T ; L 2 ( Ω )) with null integral in Ω \Omega Ω such that
P ε ⇀ ∗ P in L ∞ ( 0 , T ; L 2 ( Ω ) ) . (92) P_\varepsilon\overset{*}{\rightharpoonup} P\quad\text{in }L^\infty(0,T;L^2(\Omega)). \tag*{(92)} P ε ⇀ ∗ P in L ∞ ( 0 , T ; L 2 ( Ω )) . (92)
In the case γ = ∞ \gamma=\infty γ = ∞ , due to (5.43), we then have
∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω P div x ∂ t ϕ d x d t . (93) \int_0^T\int_{\Omega}P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dxdt\to\int_0^T\int_{\Omega}P\operatorname{div}_x\partial_t\phi\,dxdt. \tag*{(93)} ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → ∫ 0 T ∫ Ω P div x ∂ t ϕ d x d t . (93)
In the case γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , taking into account that the sequence P ^ ε \widehat{P}_\varepsilon P ε defined as
P ^ ε ( t , x , y ) = P ε ( t , ε κ ( x ε ) + r ε y ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω × ε r ε Y , (94) \widehat{P}_\varepsilon(t,x,y)=P_\varepsilon\left(t,\varepsilon\kappa\left(\frac{x}{\varepsilon}\right)+r_\varepsilon y\right)\quad\text{a.e. }(t,x,y)\in(0,T)\times\Omega\times\frac{\varepsilon}{r_\varepsilon}Y, \tag*{(94)} P ε ( t , x , y ) = P ε ( t , ε κ ( ε x ) + r ε y ) a.e. ( t , x , y ) ∈ ( 0 , T ) × Ω × r ε ε Y , (94)
satisfies
r ε 3 ε 3 ∫ Ω × Y / r ε ∣ P ^ ε ∣ 2 d x d y = ∫ Ω ∣ P ε ∣ 2 d x , a.e. in ( 0 , T ) , \frac{r_\varepsilon^3}{\varepsilon^3}\int_{\Omega\times Y/r_\varepsilon}|\widehat{P}_\varepsilon|^2\,dxdy=\int_{\Omega}|P_\varepsilon|^2\,dx,\quad\text{a.e. in }(0,T), ε 3 r ε 3 ∫ Ω × Y / r ε ∣ P ε ∣ 2 d x d y = ∫ Ω ∣ P ε ∣ 2 d x , a.e. in ( 0 , T ) ,
we can assume the existence of P ^ ∈ L ∞ ( 0 , T ; L 2 ( Ω × R 3 ) ) \widehat{P}\in L^\infty(0,T;L^2(\Omega\times\mathbb{R}^3)) P ∈ L ∞ ( 0 , T ; L 2 ( Ω × R 3 )) such that
r ε P ^ ε χ ε Y / r ε ⇀ ∗ P ^ in L ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( R 3 ) ) ) . (95) r_\varepsilon\widehat{P}_\varepsilon\chi_{\varepsilon Y/r_\varepsilon}\overset{*}{\rightharpoonup}\widehat{P}\quad\text{in }L^\infty(0,T;L^2(\Omega;L^2(\mathbb{R}^3))). \tag*{(95)} r ε P ε χ ε Y / r ε ⇀ ∗ P in L ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( R 3 ))) . (95)
Then, reasoning similarly to the fourth term, we get
∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → γ ∫ 0 T ∫ Ω × ( B R ∖ O ) P ^ div y ∂ t ϕ d x d y d t + ∫ 0 T ∫ Ω P div x ∂ t ϕ d x d t . (96) \int_0^T\int_{\Omega}P_\varepsilon\operatorname{div}_x\partial_t\Phi_\varepsilon\,dxdt
\to\gamma\int_0^T\int_{\Omega\times(B_R\setminus\mathcal{O})}\widehat{P}\operatorname{div}_y\partial_t\phi\,dxdy dt+\int_0^T\int_{\Omega}P\operatorname{div}_x\partial_t\phi\,dxdt. \tag*{(96)} ∫ 0 T ∫ Ω P ε div x ∂ t Φ ε d x d t → γ ∫ 0 T ∫ Ω × ( B R ∖ O ) P div y ∂ t ϕ d x d y d t + ∫ 0 T ∫ Ω P div x ∂ t ϕ d x d t . (96)
Terms 6, 7, 8. By (5.27), (5.29) and (5.30), we have
∫ 0 T ∫ Ω f ε ⋅ Φ ε d x d t → ∫ 0 T ∫ Ω f ⋅ ϕ d x d t , (97) \int_0^T\int_{\Omega}f_\varepsilon\cdot\Phi_\varepsilon\,dxdt\to\int_0^T\int_{\Omega}f\cdot\phi\,dxdt, \tag*{(97)} ∫ 0 T ∫ Ω f ε ⋅ Φ ε d x d t → ∫ 0 T ∫ Ω f ⋅ ϕ d x d t , (97)
ε 3 r ε 3 ∫ Ω ε u ε 1 ⋅ Φ ε ( 0 , x ) d x → ∫ Ω × O u ^ 1 ⋅ ϕ ( 0 , x , y ) d y d x , (98) \frac{\varepsilon^3}{r_\varepsilon^3}\int_{\Omega_\varepsilon}u_\varepsilon^1\cdot\Phi_\varepsilon(0,x)\,dx\to\int_{\Omega\times\mathcal{O}}\widehat{u}^1\cdot\phi(0,x,y)\,dydx, \tag*{(98)} r ε 3 ε 3 ∫ Ω ε u ε 1 ⋅ Φ ε ( 0 , x ) d x → ∫ Ω × O u 1 ⋅ ϕ ( 0 , x , y ) d y d x , (98)
∫ 0 T ∫ Ω v ε 0 ⋅ Φ ε ( 0 , x ) d x d t → ∫ Ω v 0 ⋅ ϕ ( 0 , x ) d x . (99) \int_0^T\int_{\Omega}v_\varepsilon^0\cdot\Phi_\varepsilon(0,x)\,dxdt\to\int_{\Omega}v^0\cdot\phi(0,x)\,dx. \tag*{(99)} ∫ 0 T ∫ Ω v ε 0 ⋅ Φ ε ( 0 , x ) d x d t → ∫ Ω v 0 ⋅ ϕ ( 0 , x ) d x . (99)
Reasoning by density, we have then proved that the functions u ^ \widehat{u} u , v v v and v ^ \widehat{v} v satisfy the variational problems
If γ = 0 \gamma= 0 γ = 0
{ u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 , v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , Q ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) div x v = 0 in ( 0 , T ) × Ω , u ^ ∣ t = 0 = u ^ 0 in Ω × O , ∂ t ∫ ∂ O u ^ ⋅ v d s ( y ) = 0 − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t ϕ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( ϕ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω Q div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω ∫ O u 1 ⋅ ϕ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ ( ϕ , φ ) ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 × W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 ϕ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O , ∂ t ∫ ∂ O ϕ ⋅ v d s ( y ) = 0. (100) \left\{
\begin{aligned}
&\hat{u} \in W^{1,\infty}(0,T;L^{2}(\Omega;L^{2}(\mathcal{O})))^{3} \cap L^{\infty}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O})))^{3},\\
&v \in L^{\infty}(0,T;L^{2}(\Omega))^{3} \cap L^{2}(0,T;H_{0}^{1}(\Omega))^{3},\quad Q \in L^{\infty}(0,T;L^{2}(\Omega))\\
&\operatorname{div}_{x}v=0\ \text{in }(0,T)\times\Omega,\quad\hat{u}|_{t=0}=\hat{u}^{0}\ \text{in }\Omega\times\mathcal{O},\quad\partial_{t}\int_{\partial\mathcal{O}}\hat{u}\cdot v\,ds(y)=0\\
&-\rho^{s}\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}\partial_{t}\hat{u}\cdot\partial_{t}\phi\,dydxdt+\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}A e_{y}(\hat{u}):e_{y}(\phi)\,dydxdt\\
&-\rho^{l}\int_{0}^{T}\int_{\Omega}v\cdot\partial_{t}\varphi\,dxdt+2\mu\int_{0}^{T}\int_{\Omega}e_{x}(v):e_{x}(\varphi)\,dxdt\\
&+\int_{0}^{T}\int_{\Omega}Q\operatorname{div}_{x}\partial_{t}\varphi\,dxdt=\int_{0}^{T}\int_{\Omega}f\cdot\varphi\,dxdt\\
&+\rho^{s}\int_{\Omega}\int_{\mathcal{O}}u^{1}\cdot\phi(0,x,y)\,dydx+\rho^{l}\int_{\Omega}v^{0}\cdot\varphi(0,x)\,dx,\\
&\forall(\phi,\varphi)\in W^{1,1}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O})))^{3}\times W^{1,1}(0,T;H^{1}(\Omega))^{3}\\
&\phi(T,x,y)=\varphi(T,x)=0\ \text{in }\Omega\times\mathcal{O},\quad\partial_{t}\int_{\partial\mathcal{O}}\phi\cdot v\,ds(y)=0.
\end{aligned}
\right. \tag*{(100)} ⎩ ⎨ ⎧ u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 , v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , Q ∈ L ∞ ( 0 , T ; L 2 ( Ω )) div x v = 0 in ( 0 , T ) × Ω , u ^ ∣ t = 0 = u ^ 0 in Ω × O , ∂ t ∫ ∂ O u ^ ⋅ v d s ( y ) = 0 − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t ϕ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( ϕ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω Q div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω ∫ O u 1 ⋅ ϕ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ ( ϕ , φ ) ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 × W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 ϕ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O , ∂ t ∫ ∂ O ϕ ⋅ v d s ( y ) = 0. (100)
If γ = ∞ \gamma= \infty γ = ∞
{ u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 , v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , P ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) , div x v = 0 in ( 0 , T ) × Ω , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , u ^ ∣ t = 0 = u ^ 0 in Ω × O , − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t ϕ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( ϕ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω P div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω × O u 1 ⋅ ϕ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ ϕ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 , ∀ φ ∈ W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 , ϕ ( t , x , y ) = φ ( t , x ) on ( 0 , T ) × Ω × ∂ O , ∫ ∂ O ϕ ⋅ v d s ( y ) = 0 in ( 0 , T ) × Ω , ϕ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O . (101) \left\{
\begin{aligned}
&\hat{u} \in W^{1,\infty}(0,T;L^{2}(\Omega;L^{2}(\mathcal{O})))^{3} \cap L^{\infty}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O})))^{3},\\
&v \in L^{\infty}(0,T;L^{2}(\Omega))^{3} \cap L^{2}(0,T;H_{0}^{1}(\Omega))^{3},\quad P \in L^{\infty}(0,T;L^{2}(\Omega)),\\
&\operatorname{div}_{x}v=0\ \text{in }(0,T)\times\Omega,\\
&\partial_{t}\hat{u}=v\ \text{on }(0,T)\times\Omega\times\partial\mathcal{O},\quad\hat{u}|_{t=0}=\hat{u}^{0}\ \text{in }\Omega\times\mathcal{O},\\
&-\rho^{s}\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}\partial_{t}\hat{u}\cdot\partial_{t}\phi\,dydxdt+\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}A e_{y}(\hat{u}):e_{y}(\phi)\,dydxdt\\
&-\rho^{l}\int_{0}^{T}\int_{\Omega}v\cdot\partial_{t}\varphi\,dxdt+2\mu\int_{0}^{T}\int_{\Omega}e_{x}(v):e_{x}(\varphi)\,dxdt\\
&+\int_{0}^{T}\int_{\Omega}P\operatorname{div}_{x}\partial_{t}\varphi\,dxdt=\int_{0}^{T}\int_{\Omega}f\cdot\varphi\,dxdt\\
&+\rho^{s}\int_{\Omega\times\mathcal{O}}u^{1}\cdot\phi(0,x,y)\,dydx+\rho^{l}\int_{\Omega}v^{0}\cdot\varphi(0,x)\,dx,\\
&\forall\phi\in W^{1,1}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O})))^{3},\ \forall\varphi\in W^{1,1}(0,T;H^{1}(\Omega))^{3},\\
&\phi(t,x,y)=\varphi(t,x)\ \text{on }(0,T)\times\Omega\times\partial\mathcal{O},\quad\int_{\partial\mathcal{O}}\phi\cdot v\,ds(y)=0\ \text{in }(0,T)\times\Omega,\\
&\phi(T,x,y)=\varphi(T,x)=0\ \text{in }\Omega\times\mathcal{O}.
\end{aligned}
\right. \tag*{(101)} ⎩ ⎨ ⎧ u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 , v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , P ∈ L ∞ ( 0 , T ; L 2 ( Ω )) , div x v = 0 in ( 0 , T ) × Ω , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , u ^ ∣ t = 0 = u ^ 0 in Ω × O , − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t ϕ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( ϕ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + ∫ 0 T ∫ Ω P div x ∂ t φ d x d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω × O u 1 ⋅ ϕ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ ϕ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 , ∀ φ ∈ W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 , ϕ ( t , x , y ) = φ ( t , x ) on ( 0 , T ) × Ω × ∂ O , ∫ ∂ O ϕ ⋅ v d s ( y ) = 0 in ( 0 , T ) × Ω , ϕ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O . (101)
If γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ )
u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O ) ) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O ) ) ) 3 v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , P ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) , v ^ ∈ L 2 ( 0 , T ; L 2 ( Ω ; H l o c 1 ( R 3 ∖ O ) ) ) 3 , P ^ ∈ L ∞ ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O ) ) ) , div x v = 0 in ( 0 , T ) × Ω , div y v ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , u ^ ∣ t = 0 = u ^ 0 in Ω × O , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O ) ) ) 3 , ∂ t u ^ = v ^ on ( 0 , T ) × Ω × ∂ O , − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t φ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( φ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + 2 μ γ ∫ 0 T ∫ Ω × ( R 3 ∖ O ) e y ( v ^ ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω P div x φ d x d t + γ ∫ 0 T ∫ Ω × ( R 3 ∖ O ) P ^ div y φ d x d y d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω × O u 1 ⋅ φ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ φ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H l o c 1 ( R 3 ) ) ) 3 , ∀ φ ∈ W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 , φ − φ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ) ) ) 3 , ∫ O φ ⋅ y d s ( y ) = 0 , φ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O . (102) \begin{aligned}
&\hat{u} \in W^{1,\infty}(0,T;L^{2}(\Omega;L^{2}(\mathcal{O})))^{3} \cap L^{\infty}(0,T;L^{2}(\Omega;H^{1}(\mathcal{O})))^{3}\\
&v \in L^{\infty}(0,T;L^{2}(\Omega))^{3} \cap L^{2}(0,T;H_{0}^{1}(\Omega))^{3},\quad P \in L^{\infty}(0,T;L^{2}(\Omega)),\\
&\hat{v} \in L^{2}(0,T;L^{2}(\Omega;H_{\mathrm{loc}}^{1}(\mathbb{R}^{3}\setminus\mathcal{O})))^{3},\quad\hat{P} \in L^{\infty}(0,T;L^{2}(\Omega;D^{1,2}(\mathbb{R}^{3}\setminus\mathcal{O}))),\\
&\operatorname{div}_{x}v=0\text{ in }(0,T)\times\Omega,\quad\operatorname{div}_{y}\hat{v}=0\text{ in }(0,T)\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
&\hat{u}|_{t=0}=\hat{u}^{0}\text{ in }\Omega\times\mathcal{O},\quad\partial_{t}\hat{u}=v\text{ on }(0,T)\times\Omega\times\partial\mathcal{O},\\
&\hat{v}-v\in L^{2}(0,T;L^{2}(\Omega;D^{1,2}(\mathbb{R}^{3}\setminus\mathcal{O})))^{3},\quad\partial_{t}\hat{u}=\hat{v}\text{ on }(0,T)\times\Omega\times\partial\mathcal{O},\\
&-\rho^{s}\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}\partial_{t}\hat{u}\cdot\partial_{t}\varphi\,dy\,dx\,dt+\int_{0}^{T}\int_{\Omega}\int_{\mathcal{O}}A e_{y}(\hat{u}):e_{y}(\varphi)\,dy\,dx\,dt\\
&-\rho^{l}\int_{0}^{T}\int_{\Omega}v\cdot\partial_{t}\varphi\,dx\,dt+2\mu\int_{0}^{T}\int_{\Omega}e_{x}(v):e_{x}(\varphi)\,dx\,dt\\
&+2\mu\gamma\int_{0}^{T}\int_{\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O})}e_{y}(\hat{v}):e_{y}(\varphi)\,dx\,dy\,dt+\int_{0}^{T}\int_{\Omega}P\operatorname{div}_{x}\varphi\,dx\,dt\\
&+\gamma\int_{0}^{T}\int_{\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O})}\hat{P}\operatorname{div}_{y}\varphi\,dx\,dy\,dt=\int_{0}^{T}\int_{\Omega}f\cdot\varphi\,dx\,dt\\
&+\rho^{s}\int_{\Omega\times\mathcal{O}}u^{1}\cdot\varphi(0,x,y)\,dy\,dx+\rho^{l}\int_{\Omega}v^{0}\cdot\varphi(0,x)\,dx,\\
&\forall\varphi\in W^{1,1}(0,T;L^{2}(\Omega;H_{\mathrm{loc}}^{1}(\mathbb{R}^{3})))^{3},\ \forall\varphi\in W^{1,1}(0,T;H^{1}(\Omega))^{3},\\
&\varphi-\varphi\in W^{1,1}(0,T;L^{2}(\Omega;D^{1,2}(\mathbb{R}^{3})))^{3},\quad\int_{\mathcal{O}}\varphi\cdot y\,ds(y)=0,\\
&\varphi(T,x,y)=\varphi(T,x)=0\text{ in }\Omega\times\mathcal{O}. \tag*{(102)}
\end{aligned} u ^ ∈ W 1 , ∞ ( 0 , T ; L 2 ( Ω ; L 2 ( O )) ) 3 ∩ L ∞ ( 0 , T ; L 2 ( Ω ; H 1 ( O )) ) 3 v ∈ L ∞ ( 0 , T ; L 2 ( Ω ) ) 3 ∩ L 2 ( 0 , T ; H 0 1 ( Ω ) ) 3 , P ∈ L ∞ ( 0 , T ; L 2 ( Ω )) , v ^ ∈ L 2 ( 0 , T ; L 2 ( Ω ; H loc 1 ( R 3 ∖ O )) ) 3 , P ^ ∈ L ∞ ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O ))) , div x v = 0 in ( 0 , T ) × Ω , div y v ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , u ^ ∣ t = 0 = u ^ 0 in Ω × O , ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , v ^ − v ∈ L 2 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 ∖ O )) ) 3 , ∂ t u ^ = v ^ on ( 0 , T ) × Ω × ∂ O , − ρ s ∫ 0 T ∫ Ω ∫ O ∂ t u ^ ⋅ ∂ t φ d y d x d t + ∫ 0 T ∫ Ω ∫ O A e y ( u ^ ) : e y ( φ ) d y d x d t − ρ l ∫ 0 T ∫ Ω v ⋅ ∂ t φ d x d t + 2 μ ∫ 0 T ∫ Ω e x ( v ) : e x ( φ ) d x d t + 2 μ γ ∫ 0 T ∫ Ω × ( R 3 ∖ O ) e y ( v ^ ) : e y ( φ ) d x d y d t + ∫ 0 T ∫ Ω P div x φ d x d t + γ ∫ 0 T ∫ Ω × ( R 3 ∖ O ) P ^ div y φ d x d y d t = ∫ 0 T ∫ Ω f ⋅ φ d x d t + ρ s ∫ Ω × O u 1 ⋅ φ ( 0 , x , y ) d y d x + ρ l ∫ Ω v 0 ⋅ φ ( 0 , x ) d x , ∀ φ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; H loc 1 ( R 3 )) ) 3 , ∀ φ ∈ W 1 , 1 ( 0 , T ; H 1 ( Ω ) ) 3 , φ − φ ∈ W 1 , 1 ( 0 , T ; L 2 ( Ω ; D 1 , 2 ( R 3 )) ) 3 , ∫ O φ ⋅ y d s ( y ) = 0 , φ ( T , x , y ) = φ ( T , x ) = 0 in Ω × O . (102)
From these variational systems, taking p = ∂ t P = ∂ t Q p=\partial_{t}P=\partial_{t}Q p = ∂ t P = ∂ t Q , and recalling (76) we conclude the thesis of the theorem. □
Theorem 5.4 provides some weak convergences for the solutions of (1) . The following theorem proves that these convergences hold in fact in the strong sense if the data converges strongly.
Theorem 5.6. Assume that the convergences in (64) , (65) , (66) and (67) hold in the strong sense. Then in Theorem 5.4 we have
∫ Ω ε ∗ ∫ O ( ∣ ∂ t ( u ^ ε − u ^ ) ∣ 2 + ∣ e y ( u ^ ε − u ^ ) ∣ 2 ) d y d x + ∫ Ω ε ∣ v ε − v ∣ 2 d x → 0 in L ∞ ( 0 , T ) . (103) \int_{\Omega_{\varepsilon}^{*}}\int_{\mathcal{O}}\left(|\partial_{t}(\hat{u}_{\varepsilon}-\hat{u})|^{2}+|e_{y}(\hat{u}_{\varepsilon}-\hat{u})|^{2}\right)\,dy\,dx+\int_{\Omega_{\varepsilon}}|v_{\varepsilon}-v|^{2}\,dx\to0\quad\text{in }L^{\infty}(0,T). \tag*{(103)} ∫ Ω ε ∗ ∫ O ( ∣ ∂ t ( u ^ ε − u ^ ) ∣ 2 + ∣ e y ( u ^ ε − u ^ ) ∣ 2 ) d y d x + ∫ Ω ε ∣ v ε − v ∣ 2 d x → 0 in L ∞ ( 0 , T ) . (103) For γ = 0 \gamma=0 γ = 0 or γ = ∞ \gamma=\infty γ = ∞ , we also have
∫ 0 T ∫ Ω ε l ∣ e x ( v ε − v ) ∣ 2 d x d t → 0. (104) \int_{0}^{T}\int_{\Omega_{\varepsilon}^{l}}|e_{x}(v_{\varepsilon}-v)|^{2}\,dx\,dt\to0. \tag*{(104)} ∫ 0 T ∫ Ω ε l ∣ e x ( v ε − v ) ∣ 2 d x d t → 0. (104) In the case γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , taking R ε > 0 R_{\varepsilon}>0 R ε > 0 such that
R ε r ε → ∞ , ε R ε → 0 , (105) \frac{R_{\varepsilon}}{r_{\varepsilon}}\to\infty,\qquad\frac{\varepsilon}{R_{\varepsilon}}\to0, \tag*{(105)} r ε R ε → ∞ , R ε ε → 0 , (105)
we have
∫ 0 T ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε − v ) ∣ 2 d x d s + ∫ 0 T ∫ Ω ∫ B R ε / r ε ∣ r ε ε 3 e y ( v ^ ε ) − γ e y ( v ^ ) ∣ 2 d y d x d s → 0. (106) \int_{0}^{T}\int_{\Omega\setminus B(\varepsilon k,R_{\varepsilon})}|e_x(v_{\varepsilon}-v)|^2\,dx\,ds+\int_{0}^{T}\int_{\Omega}\int_{B_{R_{\varepsilon}/r_{\varepsilon}}}\left|\frac{r_{\varepsilon}}{\varepsilon^3}e_y(\hat{v}_{\varepsilon})-\gamma e_y(\hat{v})\right|^2\,dy\,dx\,ds\to0. \tag*{(106)} ∫ 0 T ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε − v ) ∣ 2 d x d s + ∫ 0 T ∫ Ω ∫ B R ε / r ε ε 3 r ε e y ( v ^ ε ) − γ e y ( v ^ ) 2 d y d x d s → 0. (106)
Proof. The result easily follows from the convergence of the energies. We just prove it in the case γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) . Integrating in time (3.10) applied to (3.1), passing to the limit in the right-hand side thanks to the strong convergence of the right-hand side and the initial data, and using the energy identity for (5.38) we have
∫ Ω ε s ( ρ s ε 3 2 r ε 3 ∣ ∂ t u ε ∣ 2 + ε 3 2 r ε A e x ( u ε ) : e x ( u ε ) ) d x + ∫ Ω ε l ( ρ l 2 ∣ v ε ∣ 2 + 2 μ ∫ 0 t ∣ e x ( v ε ) ∣ 2 d s ) d x = ∫ 0 t ( ∫ Ω ε s f ε ⋅ ∂ t u ε d x + ∫ Ω ε l f ε ⋅ v ε d x ) d s + ∫ Ω ε ∗ ∫ O ( ρ s 2 ∣ u ^ ε 1 ∣ 2 + 1 2 A e y ( u ^ ε 0 ) : e y ( u ^ ε 0 ) ) d y d x + ρ l 2 ∫ Ω ε l ∣ v ε 0 ∣ 2 d x \begin{aligned}
\int_{\Omega_{\varepsilon}^{s}}\left(\frac{\rho^{s}\varepsilon^{3}}{2r_{\varepsilon}^{3}}|\partial_tu_{\varepsilon}|^2+\frac{\varepsilon^{3}}{2r_{\varepsilon}}A_{ex}(u_{\varepsilon}):e_x(u_{\varepsilon})\right)\,dx+\int_{\Omega_{\varepsilon}^{l}}\left(\frac{\rho^{l}}{2}|v_{\varepsilon}|^2+2\mu\int_{0}^{t}|e_x(v_{\varepsilon})|^2\,ds\right)\,dx \\
&=\int_{0}^{t}\left(\int_{\Omega_{\varepsilon}^{s}}f_{\varepsilon}\cdot\partial_tu_{\varepsilon}\,dx+\int_{\Omega_{\varepsilon}^{l}}f_{\varepsilon}\cdot v_{\varepsilon}\,dx\right)\,ds \\
&+\int_{\Omega_{\varepsilon}^{*}}\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\hat u_{\varepsilon}^{1}|^2+\frac{1}{2}A_{ey}(\hat u_{\varepsilon}^{0}):e_y(\hat u_{\varepsilon}^{0})\right)\,dy\,dx+\frac{\rho^{l}}{2}\int_{\Omega_{\varepsilon}^{l}}|v_{\varepsilon}^{0}|^2\,dx
\end{aligned} ∫ Ω ε s ( 2 r ε 3 ρ s ε 3 ∣ ∂ t u ε ∣ 2 + 2 r ε ε 3 A e x ( u ε ) : e x ( u ε ) ) d x + ∫ Ω ε l ( 2 ρ l ∣ v ε ∣ 2 + 2 μ ∫ 0 t ∣ e x ( v ε ) ∣ 2 d s ) d x = ∫ 0 t ( ∫ Ω ε s f ε ⋅ ∂ t u ε d x + ∫ Ω ε l f ε ⋅ v ε d x ) d s + ∫ Ω ε ∗ ∫ O ( 2 ρ s ∣ u ^ ε 1 ∣ 2 + 2 1 A ey ( u ^ ε 0 ) : e y ( u ^ ε 0 ) ) d y d x + 2 ρ l ∫ Ω ε l ∣ v ε 0 ∣ 2 d x
⟶ ∫ Ω ( ∫ 0 t f ⋅ v d s + ∫ O ( ρ s 2 ∣ u ^ 1 ∣ 2 + A e y ( u ^ 0 ) : e y ( u ^ 0 ) ) d y + ρ l 2 ∣ v 0 ∣ 2 ) d x = ∫ Ω ( ∫ O ( ρ s 2 ∣ ∂ t u ^ ∣ 2 + 1 2 A e y ( u ^ ) : e y ( u ^ ) ) d y + ρ l 2 ∣ v ∣ 2 ) d x + 2 μ ∫ 0 t ∫ Ω ( γ ∫ R 3 ∖ O ∣ e y ( v ^ ) ∣ 2 d y + ∣ e y ( v ) ∣ 2 ) d x d s . \begin{aligned}
\longrightarrow\int_{\Omega}\left(\int_{0}^{t}f\cdot v\,ds+\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\hat u^{1}|^2+A_{ey}(\hat u^{0}):e_y(\hat u^{0})\right)\,dy+\frac{\rho^{l}}{2}|v^{0}|^2\right)\,dx \\
&=\int_{\Omega}\left(\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\partial_t\hat u|^2+\frac{1}{2}A_{ey}(\hat u):e_y(\hat u)\right)\,dy+\frac{\rho^{l}}{2}|v|^2\right)\,dx+2\mu\int_{0}^{t}\int_{\Omega}\left(\gamma\int_{\mathbb R^3\setminus\mathcal O}|e_y(\hat v)|^2\,dy+|e_y(v)|^2\right)\,dx\,ds.
\end{aligned} ⟶ ∫ Ω ( ∫ 0 t f ⋅ v d s + ∫ O ( 2 ρ s ∣ u ^ 1 ∣ 2 + A ey ( u ^ 0 ) : e y ( u ^ 0 ) ) d y + 2 ρ l ∣ v 0 ∣ 2 ) d x = ∫ Ω ( ∫ O ( 2 ρ s ∣ ∂ t u ^ ∣ 2 + 2 1 A ey ( u ^ ) : e y ( u ^ ) ) d y + 2 ρ l ∣ v ∣ 2 ) d x + 2 μ ∫ 0 t ∫ Ω ( γ ∫ R 3 ∖ O ∣ e y ( v ^ ) ∣ 2 d y + ∣ e y ( v ) ∣ 2 ) d x d s .
Thus, taking into account that for R ε > 0 R_{\varepsilon}>0 R ε > 0 satisfying (5.68), we get
∫ Ω ε s ( ρ s ε 3 2 r ε 3 ∣ ∂ t u ε ∣ 2 + ε 3 2 r ε A e x ( u ε ) : e x ( u ε ) ) d x + ∫ Ω ε l ( ρ l 2 ∣ v ε ∣ 2 + 2 μ ∫ 0 t ∣ e x ( v ε ) ∣ 2 d s ) d x = ∫ Ω ε ∗ ∫ O ( ρ s 2 ∣ ∂ t u ^ ε ∣ 2 + 1 2 A e y ( u ^ ε ) : e y ( u ^ ε ) ) d y d x + ρ l 2 ∫ Ω ε l ∣ v ε ∣ 2 d x + 2 μ ∫ 0 t ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε ) ∣ 2 d x d s + 2 μ r ε ε 3 ∫ 0 t ∫ Ω ∫ B R ε / r ε ∣ e y ( v ^ ε ) ∣ 2 d y d x d s , \begin{aligned}
\int_{\Omega_{\varepsilon}^{s}}\left(\frac{\rho^{s}\varepsilon^{3}}{2r_{\varepsilon}^{3}}|\partial_tu_{\varepsilon}|^2+\frac{\varepsilon^{3}}{2r_{\varepsilon}}A_{ex}(u_{\varepsilon}):e_x(u_{\varepsilon})\right)\,dx+\int_{\Omega_{\varepsilon}^{l}}\left(\frac{\rho^{l}}{2}|v_{\varepsilon}|^2+2\mu\int_{0}^{t}|e_x(v_{\varepsilon})|^2\,ds\right)\,dx \\
&=\int_{\Omega_{\varepsilon}^{*}}\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\partial_t\hat u_{\varepsilon}|^2+\frac{1}{2}A_{ey}(\hat u_{\varepsilon}):e_y(\hat u_{\varepsilon})\right)\,dy\,dx+\frac{\rho^{l}}{2}\int_{\Omega_{\varepsilon}^{l}}|v_{\varepsilon}|^2\,dx \\
&+2\mu\int_{0}^{t}\int_{\Omega\setminus B(\varepsilon k,R_{\varepsilon})}|e_x(v_{\varepsilon})|^2\,dx\,ds+\frac{2\mu r_{\varepsilon}}{\varepsilon^{3}}\int_{0}^{t}\int_{\Omega}\int_{B_{R_{\varepsilon}/r_{\varepsilon}}}|e_y(\hat v_{\varepsilon})|^2\,dy\,dx\,ds,
\end{aligned} ∫ Ω ε s ( 2 r ε 3 ρ s ε 3 ∣ ∂ t u ε ∣ 2 + 2 r ε ε 3 A e x ( u ε ) : e x ( u ε ) ) d x + ∫ Ω ε l ( 2 ρ l ∣ v ε ∣ 2 + 2 μ ∫ 0 t ∣ e x ( v ε ) ∣ 2 d s ) d x = ∫ Ω ε ∗ ∫ O ( 2 ρ s ∣ ∂ t u ^ ε ∣ 2 + 2 1 A ey ( u ^ ε ) : e y ( u ^ ε ) ) d y d x + 2 ρ l ∫ Ω ε l ∣ v ε ∣ 2 d x + 2 μ ∫ 0 t ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε ) ∣ 2 d x d s + ε 3 2 μ r ε ∫ 0 t ∫ Ω ∫ B R ε / r ε ∣ e y ( v ^ ε ) ∣ 2 d y d x d s ,
we have proved the following convergence result for the energies
∫ Ω ε ∗ ∫ O ( ρ s 2 ∣ ∂ t u ^ ε ∣ 2 + 1 2 A e y ( u ^ ε ) : e y ( u ^ ε ) ) d y d x + ρ l 2 ∫ Ω ε l ∣ v ε ∣ 2 d x + 2 μ ∫ 0 t ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε ) ∣ 2 d x d s + 2 μ r ε ε 3 ∫ 0 t ∫ Ω ∫ B R ε / r ε ∣ e y ( v ^ ε ) ∣ 2 d y d x d s \begin{aligned}
\int_{\Omega_{\varepsilon}^{*}}\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\partial_t\hat u_{\varepsilon}|^2+\frac{1}{2}A_{ey}(\hat u_{\varepsilon}):e_y(\hat u_{\varepsilon})\right)\,dy\,dx+\frac{\rho^{l}}{2}\int_{\Omega_{\varepsilon}^{l}}|v_{\varepsilon}|^2\,dx \\
&+2\mu\int_{0}^{t}\int_{\Omega\setminus B(\varepsilon k,R_{\varepsilon})}|e_x(v_{\varepsilon})|^2\,dx\,ds+\frac{2\mu r_{\varepsilon}}{\varepsilon^{3}}\int_{0}^{t}\int_{\Omega}\int_{B_{R_{\varepsilon}/r_{\varepsilon}}}|e_y(\hat v_{\varepsilon})|^2\,dy\,dx\,ds
\end{aligned} ∫ Ω ε ∗ ∫ O ( 2 ρ s ∣ ∂ t u ^ ε ∣ 2 + 2 1 A ey ( u ^ ε ) : e y ( u ^ ε ) ) d y d x + 2 ρ l ∫ Ω ε l ∣ v ε ∣ 2 d x + 2 μ ∫ 0 t ∫ Ω ∖ B ( ε k , R ε ) ∣ e x ( v ε ) ∣ 2 d x d s + ε 3 2 μ r ε ∫ 0 t ∫ Ω ∫ B R ε / r ε ∣ e y ( v ^ ε ) ∣ 2 d y d x d s
⟶ ∫ Ω ( ∫ O ( ρ s 2 ∣ ∂ t u ^ ∣ 2 + 1 2 A e y ( u ^ ) : e y ( u ^ ) ) d y + ρ l 2 ∣ v ∣ 2 ) d x + 2 μ ∫ 0 t ∫ Ω ( ∣ e y ( v ) ∣ 2 + γ ∫ R 3 ∖ O ∣ e y ( v ^ ) ∣ 2 d y ) d x d s . \begin{aligned}
\longrightarrow\int_{\Omega}\left(\int_{\mathcal O}\left(\frac{\rho^{s}}{2}|\partial_t\hat u|^2+\frac{1}{2}A_{ey}(\hat u):e_y(\hat u)\right)\,dy+\frac{\rho^{l}}{2}|v|^2\right)\,dx \\
&+2\mu\int_{0}^{t}\int_{\Omega}\left(|e_y(v)|^2+\gamma\int_{\mathbb R^3\setminus\mathcal O}|e_y(\hat v)|^2\,dy\right)\,dx\,ds.
\end{aligned} ⟶ ∫ Ω ( ∫ O ( 2 ρ s ∣ ∂ t u ^ ∣ 2 + 2 1 A ey ( u ^ ) : e y ( u ^ ) ) d y + 2 ρ l ∣ v ∣ 2 ) d x + 2 μ ∫ 0 t ∫ Ω ( ∣ e y ( v ) ∣ 2 + γ ∫ R 3 ∖ O ∣ e y ( v ^ ) ∣ 2 d y ) d x d s .
From this convergence it is simple to check that (5.66), (5.67) and (5.69) hold.
Remark 5.7. Theorem 5.4 provides a two-scale limit problem for (3.1) depending on the macroscopic variables ( t , x ) (t,x) ( t , x ) and the microscopic on y y y .
For γ = 0 \gamma=0 γ = 0 the equations describing the behavior of the elastic and fluid components are uncoupled. The macroscopic behavior of the fluid is just given by (5.34). In this case the solid particles are so small that they do not affect the movement of the fluid.
In the case γ = ∞ \gamma=\infty γ = ∞ and γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) , the behavior of the solid particles and the fluid are interrelated. In order to get a macroscopic equation for the velocity and pressure of the fluid it is necessary to eliminate the function u ^ \hat{u} u ^ in (5.35) and the functions ( u ^ , v ^ , p ^ ) (\hat{u},\hat{v},\hat{p}) ( u ^ , v ^ , p ^ ) in (5.38). The corresponding result is given by Theorem 5.8 below, which shows that the presence of solid particles leads to the appearance of a memory term in the Stokes system that governs the behavior of the fluid.
Theorem 5.8. Assume γ ∈ ( 0 , ∞ ] \gamma\in(0,\infty] γ ∈ ( 0 , ∞ ] , then there exist R γ ∈ L l o c 2 ( [ 0 , ∞ ) ; H 1 ( O ) ′ ) 3 \mathcal{R}_{\gamma}\in L^{2}_{\mathrm{loc}}([0,\infty);H^{1}(\mathcal{O})')^{3} R γ ∈ L loc 2 ([ 0 , ∞ ) ; H 1 ( O ) ′ ) 3 and S γ ∈ L l o c 2 ( [ 0 , ∞ ) ; L 2 ( O ) ) 3 \mathcal{S}_{\gamma}\in L^{2}_{\mathrm{loc}}([0,\infty);L^{2}(\mathcal{O}))^{3} S γ ∈ L loc 2 ([ 0 , ∞ ) ; L 2 ( O ) ) 3 such that v v v and ∇ x p \nabla_{x}p ∇ x p in Theorem 5.4 are characterized as the unique solution of the following Stokes problem with memory terms
{ ∂ t ( ( 2 ρ s ∣ O ∣ + ρ f ) v + ∫ 0 t S γ ( t − s ) v d s ) + R γ ( t ) u ^ 0 + S γ ( t ) u ^ 1 − 2 μ div x e x ( v ) + ∇ x p = f in ( 0 , T ) × Ω , div x v = 0 in ( 0 , T ) × Ω , v = 0 on ( 0 , T ) × ∂ Ω , v ∣ t = 0 = v 0 in Ω . (107) \left\{
\begin{aligned}
\partial_{t}\left((2\rho^{s}|\mathcal{O}|+\rho^{f})v+\int_{0}^{t}\mathcal{S}_{\gamma}(t-s)v\,ds\right)+\mathcal{R}_{\gamma}(t)\hat{u}^{0}+\mathcal{S}_{\gamma}(t)\hat{u}^{1}-2\mu\operatorname{div}_{x}e_{x}(v)+\nabla_{x}p&=f &&\text{in }(0,T)\times\Omega,\\
\operatorname{div}_{x}v&=0 &&\text{in }(0,T)\times\Omega,\\
v&=0 &&\text{on }(0,T)\times\partial\Omega,\\
v|_{t=0}&=v^{0} &&\text{in }\Omega.
\end{aligned}
\right. \tag*{(107)} ⎩ ⎨ ⎧ ∂ t ( ( 2 ρ s ∣ O ∣ + ρ f ) v + ∫ 0 t S γ ( t − s ) v d s ) + R γ ( t ) u ^ 0 + S γ ( t ) u ^ 1 − 2 μ div x e x ( v ) + ∇ x p div x v v v ∣ t = 0 = f = 0 = 0 = v 0 in ( 0 , T ) × Ω , in ( 0 , T ) × Ω , on ( 0 , T ) × ∂ Ω , in Ω. (107) Moreover, the operators S γ \mathcal{S}_{\gamma} S γ and R γ \mathcal{R}_{\gamma} R γ satisfy the positivity condition
1 2 ∣ 2 ρ s ∣ O ∣ w + u 1 + ∫ 0 t ( S γ ( t − s ) w + R γ u 0 + S γ u 1 ) d s ∣ 2 ≤ ρ s ∣ O ∣ ( ρ s ( ∣ w ∣ 2 − ∣ w 0 ∣ 2 ) + ∫ 0 t ( ∂ t ∫ 0 r S γ ( r − s ) w d s + R γ u 0 + S γ u 1 ) w d r ) \frac{1}{2}\left|2\rho^{s}|\mathcal{O}|w+u^{1}+\int_{0}^{t}\left(\mathcal{S}_{\gamma}(t-s)w+\mathcal{R}_{\gamma}u^{0}+\mathcal{S}_{\gamma}u^{1}\right)\,ds\right|^{2}
\leq\rho^{s}|\mathcal{O}|\left(\rho^{s}(|w|^{2}-|w^{0}|^{2})+\int_{0}^{t}\left(\partial_{t}\int_{0}^{r}\mathcal{S}_{\gamma}(r-s)w\,ds+\mathcal{R}_{\gamma}u^{0}+\mathcal{S}_{\gamma}u^{1}\right)w\,dr\right) 2 1 2 ρ s ∣ O ∣ w + u 1 + ∫ 0 t ( S γ ( t − s ) w + R γ u 0 + S γ u 1 ) d s 2 ≤ ρ s ∣ O ∣ ( ρ s ( ∣ w ∣ 2 − ∣ w 0 ∣ 2 ) + ∫ 0 t ( ∂ t ∫ 0 r S γ ( r − s ) w d s + R γ u 0 + S γ u 1 ) w d r ) + ρ s ∣ O ∣ 2 ∫ O ( ρ s ∣ u 1 ∣ 2 + A e y ( u 0 ) : e y ( u 0 ) ) d y , ∀ t ∈ ( 0 , T ) , (108) +\frac{\rho^{s}|\mathcal{O}|}{2}\int_{\mathcal{O}}\left(\rho^{s}|u^{1}|^{2}+A e_{y}(u^{0}):e_{y}(u^{0})\right)\,dy,\qquad\forall t\in(0,T), \tag*{(108)} + 2 ρ s ∣ O ∣ ∫ O ( ρ s ∣ u 1 ∣ 2 + A e y ( u 0 ) : e y ( u 0 ) ) d y , ∀ t ∈ ( 0 , T ) , (108) for every w ∈ H 1 ( R ) w\in H^{1}(\mathbb{R}) w ∈ H 1 ( R ) , every u 0 ∈ H 1 ( O ) 3 u^{0}\in H^{1}(\mathcal{O})^{3} u 0 ∈ H 1 ( O ) 3 and every u 1 ∈ L 2 ( O ) 3 u^{1}\in L^{2}(\mathcal{O})^{3} u 1 ∈ L 2 ( O ) 3 .
For γ < ∞ \gamma<\infty γ < ∞ , the operators R γ \mathcal{R}_{\gamma} R γ and S γ \mathcal{S}_{\gamma} S γ are in fact in L 2 ( 0 , ∞ ; H 1 ( O ) ′ ) 3 L^{2}(0,\infty;H^{1}(\mathcal{O})')^{3} L 2 ( 0 , ∞ ; H 1 ( O ) ′ ) 3 and S γ ∈ L 2 ( 0 , ∞ ; L 2 ( O ) ) 3 \mathcal{S}_{\gamma}\in L^{2}(0,\infty;L^{2}(\mathcal{O}))^{3} S γ ∈ L 2 ( 0 , ∞ ; L 2 ( O ) ) 3 respectively and there exists C > 0 C>0 C > 0 independent of γ \gamma γ such that
∥ R ∥ L 2 ( 0 , ∞ ; H 1 ( O ) ′ ) 3 + ∥ S ∥ L 2 ( 0 , ∞ ; L 2 ( O ) ) 3 ≤ C γ . (109) \|\mathcal{R}\|_{L^{2}(0,\infty;H^{1}(\mathcal{O})')^{3}}+\|\mathcal{S}\|_{L^{2}(0,\infty;L^{2}(\mathcal{O}))^{3}}\leq C\sqrt{\gamma}. \tag*{(109)} ∥ R ∥ L 2 ( 0 , ∞ ; H 1 ( O ) ′ ) 3 + ∥ S ∥ L 2 ( 0 , ∞ ; L 2 ( O ) ) 3 ≤ C γ . (109) Proof. Let us first assume γ ∈ ( 0 , ∞ ) \gamma\in(0,\infty) γ ∈ ( 0 , ∞ ) .
The functions ( u ^ , v ^ , p ^ ) (\hat{u},\hat{v},\hat{p}) ( u ^ , v ^ , p ^ ) in (75) can be obtained from v v v by solving the system
ρ s ∂ t t 2 u ^ − div y ( A e y ( u ^ ) ) = 0 in ( 0 , T ) × Ω × O , − 2 μ div y e y ( v ^ ) + ∇ y p ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , div y v ^ = 0 in ( 0 , T ) × Ω × ( R 3 ∖ O ) , ∂ t u ^ = v ^ , A e y ( u ^ ) ν = γ ( 2 μ e y ( v ^ ) − p ^ I ) ν on ( 0 , T ) × Ω × ∂ O , v ^ − v ∈ D 1 , 2 ( R 3 ∖ O ) 3 , u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in Ω × O . (110) \begin{aligned}
\rho^{s}\partial_{tt}^{2}\hat{u}-\operatorname{div}_{y}(A e_{y}(\hat{u}))&=0 &&\text{in }(0,T)\times\Omega\times\mathcal{O},\\
-2\mu\operatorname{div}_{y}e_{y}(\hat{v})+\nabla_{y}\hat{p}&=0 &&\text{in }(0,T)\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
\operatorname{div}_{y}\hat{v}&=0 &&\text{in }(0,T)\times\Omega\times(\mathbb{R}^{3}\setminus\mathcal{O}),\\
\partial_{t}\hat{u}&=\hat{v},\quad A e_{y}(\hat{u})\nu=\gamma(2\mu e_{y}(\hat{v})-\hat{p}I)\nu&&\text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
\hat{v}-v&\in D^{1,2}(\mathbb{R}^{3}\setminus\mathcal{O})^{3},\\
\hat{u}|_{t=0}&=\hat{u}^{0},\quad\partial_{t}\hat{u}|_{t=0}=\hat{u}^{1} &&\text{in }\Omega\times\mathcal{O}. \tag*{(110)}
\end{aligned} ρ s ∂ tt 2 u ^ − div y ( A e y ( u ^ )) − 2 μ div y e y ( v ^ ) + ∇ y p ^ div y v ^ ∂ t u ^ v ^ − v u ^ ∣ t = 0 = 0 = 0 = 0 = v ^ , A e y ( u ^ ) ν = γ ( 2 μ e y ( v ^ ) − p ^ I ) ν ∈ D 1 , 2 ( R 3 ∖ O ) 3 , = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in ( 0 , T ) × Ω × O , in ( 0 , T ) × Ω × ( R 3 ∖ O ) , in ( 0 , T ) × Ω × ( R 3 ∖ O ) , on ( 0 , T ) × Ω × ∂ O , in Ω × O . (110)
Let us use the semigroup theory to solve this problem, for such purpose we define A γ : D ( A γ ) ⊂ H 1 ( O ) 3 × L 2 ( O ) 3 → H 1 ( O ) 3 × L 2 ( O ) 3 A_{\gamma}:D(A_{\gamma})\subset H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3}\to H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3} A γ : D ( A γ ) ⊂ H 1 ( O ) 3 × L 2 ( O ) 3 → H 1 ( O ) 3 × L 2 ( O ) 3 by
D ( A γ ) = { ( u ~ , z ~ ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 : div y A e y ( u ~ ) ∈ L 2 ( O ) 3 , z ~ ∈ H 1 ( O ) 3 , D(A_{\gamma})=\left\{(\widetilde{u},\widetilde{z})\in H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3}:\ \operatorname{div}_{y}A e_{y}(\widetilde{u})\in L^{2}(\mathcal{O})^{3},\ \widetilde{z}\in H^{1}(\mathcal{O})^{3},
\right. D ( A γ ) = { ( u , z ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 : div y A e y ( u ) ∈ L 2 ( O ) 3 , z ∈ H 1 ( O ) 3 ,
∃ ( v ~ , p ~ ) ∈ D 1 , 2 ( R 3 ∖ O ) 3 × L 2 ( R 3 ∖ O ) such that \left.
\exists(\widetilde{v},\widetilde{p})\in D^{1,2}(\mathbb{R}^{3}\setminus\mathcal{O})^{3}\times L^{2}(\mathbb{R}^{3}\setminus\mathcal{O})\text{ such that}
\right. ∃ ( v , p ) ∈ D 1 , 2 ( R 3 ∖ O ) 3 × L 2 ( R 3 ∖ O ) such that
− 2 μ div y e y ( v ~ ) + ∇ y p ~ = div y z ~ = 0 in R 3 ∖ O , -2\mu\operatorname{div}_{y}e_{y}(\widetilde{v})+\nabla_{y}\widetilde{p}=\operatorname{div}_{y}\widetilde{z}=0\text{ in }\mathbb{R}^{3}\setminus\mathcal{O}, − 2 μ div y e y ( v ) + ∇ y p = div y z = 0 in R 3 ∖ O ,
z ~ = v ~ , A e y ( u ~ ) ν = γ ( 2 μ e y ( v ~ ) − p ~ I ) ν on ∂ O } , \left.
\widetilde{z}=\widetilde{v},\quad A e_{y}(\widetilde{u})\nu=\gamma(2\mu e_{y}(\widetilde{v})-\widetilde{p}I)\nu\text{ on }\partial\mathcal{O}\right\}, z = v , A e y ( u ) ν = γ ( 2 μ e y ( v ) − p I ) ν on ∂ O } ,
A γ ( u ~ z ~ ) = ( z ~ 1 ρ s div y A e y ( u ~ ) ) , ∀ ( u ~ , z ~ ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 A_{\gamma}\begin{pmatrix}\widetilde{u}\\\widetilde{z}\end{pmatrix}
=\begin{pmatrix}\widetilde{z}\\ \dfrac{1}{\rho^{s}}\operatorname{div}_{y}A e_{y}(\widetilde{u})\end{pmatrix},
\qquad\forall(\widetilde{u},\widetilde{z})\in H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3} A γ ( u z ) = z ρ s 1 div y A e y ( u ) , ∀ ( u , z ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3
in such way that taking
w ~ = u ^ − ∫ 0 t v d s , z ~ = ∂ t u ^ − v , v ~ = v ^ − v , (111) \widetilde{w}=\hat{u}-\int_{0}^{t}v\,ds,\qquad\widetilde{z}=\partial_{t}\hat{u}-v,\qquad\widetilde{v}=\hat{v}-v, \tag*{(111)} w = u ^ − ∫ 0 t v d s , z = ∂ t u ^ − v , v = v ^ − v , (111)
we have that (110) is equivalent to
d d t ( w ~ z ~ ) = A γ ( w ~ z ~ ) − ( 0 ∂ t v ) in ( 0 , T ) , ( w ~ ( 0 ) z ~ ( 0 ) ) = ( u ^ 0 u ^ 1 − v ( 0 ) ) . (112) \frac{d}{dt}\begin{pmatrix}\widetilde{w}\\\widetilde{z}\end{pmatrix}
=A_{\gamma}\begin{pmatrix}\widetilde{w}\\\widetilde{z}\end{pmatrix}
-\begin{pmatrix}0\\\partial_{t}v\end{pmatrix}
\quad\text{in }(0,T),\qquad
\begin{pmatrix}\widetilde{w}(0)\\\widetilde{z}(0)\end{pmatrix}
=\begin{pmatrix}\hat{u}^{0}\\\hat{u}^{1}-v(0)\end{pmatrix}. \tag*{(112)} d t d ( w z ) = A γ ( w z ) − ( 0 ∂ t v ) in ( 0 , T ) , ( w ( 0 ) z ( 0 ) ) = ( u ^ 0 u ^ 1 − v ( 0 ) ) . (112)
Taking in H 1 ( O ) 3 × L 2 ( O ) 3 H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3} H 1 ( O ) 3 × L 2 ( O ) 3 the norm
∥ ( u ~ , z ~ ) ∥ 2 = ∫ O ( A e y ( u ~ ) : e y ( u ~ ) + ∣ u ~ ∣ 2 + ρ s ∣ z ~ ∣ 2 ) d y , ∀ ( u ~ , z ~ ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 , \|(\widetilde{u},\widetilde{z})\|^{2}
=\int_{\mathcal{O}}\left(A e_{y}(\widetilde{u}):e_{y}(\widetilde{u})+|\widetilde{u}|^{2}+\rho^{s}|\widetilde{z}|^{2}\right)\,dy,
\qquad\forall(\widetilde{u},\widetilde{z})\in H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3}, ∥ ( u , z ) ∥ 2 = ∫ O ( A e y ( u ) : e y ( u ) + ∣ u ∣ 2 + ρ s ∣ z ∣ 2 ) d y , ∀ ( u , z ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 ,
and using that
( A γ ( u ~ z ~ ) , ( u ~ z ~ ) ) = ∫ O z ~ ⋅ u ~ d y − 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ~ ) ∣ 2 d y , ∀ ( u ~ , z ~ ) ∈ D ( A γ ) , \left(A_{\gamma}\begin{pmatrix}\widetilde{u}\\\widetilde{z}\end{pmatrix},
\begin{pmatrix}\widetilde{u}\\\widetilde{z}\end{pmatrix}\right)
=\int_{\mathcal{O}}\widetilde{z}\cdot\widetilde{u}\,dy
-2\mu\gamma\int_{\mathbb{R}^{3}\setminus\mathcal{O}}|e_{y}(\widetilde{v})|^{2}\,dy,
\qquad\forall(\widetilde{u},\widetilde{z})\in D(A_{\gamma}), ( A γ ( u z ) , ( u z ) ) = ∫ O z ⋅ u d y − 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ) ∣ 2 d y , ∀ ( u , z ) ∈ D ( A γ ) ,
we get that A γ − I A_{\gamma}-I A γ − I is dissipative. Moreover, reasoning similarly to the proof of Theorem 3.1 , the Lax-Milgram’s theorem proves that for every ( g , h ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 (g,h)\in H^{1}(\mathcal{O})^{3}\times L^{2}(\mathcal{O})^{3} ( g , h ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 and every λ > 0 \lambda>0 λ > 0 problem
A γ ( u ~ z ~ ) − λ ( u ~ z ~ ) = ( g h ) , A_{\gamma}\begin{pmatrix}\widetilde{u}\\\widetilde{z}\end{pmatrix}
-\lambda\begin{pmatrix}\widetilde{u}\\\widetilde{z}\end{pmatrix}
=\begin{pmatrix}g\\h\end{pmatrix}, A γ ( u z ) − λ ( u z ) = ( g h ) ,
or equivalently
{ z ~ − λ u ~ = g in O , div ( A e y ( u ~ ) ) − λ ρ s z ~ = ρ s h in O , 2 μ div y e y ( v ~ ) − ∇ y p ~ = 0 in R 3 ∖ O , div y v ~ = 0 in R 3 ∖ O , z ~ = v ~ , A e y ( u ~ ) ν = γ ( 2 μ e y ( v ~ ) − p ~ I ) ν on ∂ O , v ~ ∈ D 1 , 2 ( R 3 ∖ O ) 3 , \begin{cases}
\widetilde{z}-\lambda\widetilde{u}=g & \text{in }\mathcal{O},\\
\operatorname{div}(A e_y(\widetilde{u}))-\lambda\rho^s\widetilde{z}=\rho^s h & \text{in }\mathcal{O},\\
2\mu\operatorname{div}_y e_y(\widetilde{v})-\nabla_y\widetilde{p}=0 & \text{in }\mathbb{R}^3\setminus\mathcal{O},\\
\operatorname{div}_y\widetilde{v}=0 & \text{in }\mathbb{R}^3\setminus\mathcal{O},\\
\widetilde{z}=\widetilde{v},\quad A e_y(\widetilde{u})\nu=\gamma(2\mu e_y(\widetilde{v})-\widetilde{p}I)\nu& \text{on }\partial\mathcal{O},\\
\widetilde{v}\in D^{1,2}(\mathbb{R}^3\setminus\mathcal{O})^3,
\end{cases} ⎩ ⎨ ⎧ z − λ u = g div ( A e y ( u )) − λ ρ s z = ρ s h 2 μ div y e y ( v ) − ∇ y p = 0 div y v = 0 z = v , A e y ( u ) ν = γ ( 2 μ e y ( v ) − p I ) ν v ∈ D 1 , 2 ( R 3 ∖ O ) 3 , in O , in O , in R 3 ∖ O , in R 3 ∖ O , on ∂ O ,
has a unique solution, such that ( u ~ , v ~ ) ∈ D ( A γ ) (\widetilde{u},\widetilde{v})\in D(\mathcal{A}_\gamma) ( u , v ) ∈ D ( A γ ) , ( v ~ , p ~ ) ∈ D 1 , 2 ( R 3 ∖ O ) 3 × L 2 ( R 3 ∖ O ) (\widetilde{v},\widetilde{p})\in D^{1,2}(\mathbb{R}^3\setminus\mathcal{O})^3\times L^2(\mathbb{R}^3\setminus\mathcal{O}) ( v , p ) ∈ D 1 , 2 ( R 3 ∖ O ) 3 × L 2 ( R 3 ∖ O ) . This proves that A γ − I \mathcal{A}_\gamma-I A γ − I is in the conditions of the Lumer-Phillips theorem and then that A γ \mathcal{A}_\gamma A γ generates a continuous semigroup G ( t ) \mathcal{G}(t) G ( t ) in H 1 ( O ) 3 × L 2 ( O ) 3 H^1(\mathcal{O})^3\times L^2(\mathcal{O})^3 H 1 ( O ) 3 × L 2 ( O ) 3 . Writing
G ( t ) = ( G 11 ( t ) G 12 ( t ) G 21 ( t ) G 22 ( t ) ) , \mathcal{G}(t)=
\begin{pmatrix}
\mathcal{G}_{11}(t) & \mathcal{G}_{12}(t)\\
\mathcal{G}_{21}(t) & \mathcal{G}_{22}(t)
\end{pmatrix}, G ( t ) = ( G 11 ( t ) G 21 ( t ) G 12 ( t ) G 22 ( t ) ) ,
where the definition of A γ \mathcal{A}_\gamma A γ gives
G 21 ( t ) u ~ = d G 11 d t ( t ) u ~ , G 22 ( t ) z ~ = d G 12 d t ( t ) z ~ , ∀ ( u ~ , z ~ ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 , (113) \mathcal{G}_{21}(t)\widetilde{u}=\frac{d\mathcal{G}_{11}}{dt}(t)\widetilde{u},\qquad
\mathcal{G}_{22}(t)\widetilde{z}=\frac{d\mathcal{G}_{12}}{dt}(t)\widetilde{z},\qquad
\forall(\widetilde{u},\widetilde{z})\in H^1(\mathcal{O})^3\times L^2(\mathcal{O})^3, \tag*{(113)} G 21 ( t ) u = d t d G 11 ( t ) u , G 22 ( t ) z = d t d G 12 ( t ) z , ∀ ( u , z ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 , (113)
we get by (5.74) and (5.75) that the function u ^ \widehat{u} u in (5.73) is given by
u ^ = ∫ 0 t v d s + G 11 ( t ) u ^ 0 + G 12 ( t ) ( u ^ 1 − v ( 0 ) ) − ∫ 0 t G 12 ( t − s ) ∂ s v d s = ∫ 0 t v d s + G 11 ( t ) u ^ 0 + G 12 ( t ) u ^ 1 − ∫ 0 t d G 12 d t ( t − s ) v d s . (114) \begin{aligned}
\widehat{u}&=\int_0^t v\,ds+\mathcal{G}_{11}(t)\widehat{u}^{0}+\mathcal{G}_{12}(t)(\widehat{u}^{1}-v(0))-\int_0^t\mathcal{G}_{12}(t-s)\partial_s v\,ds\\
&=\int_0^t v\,ds+\mathcal{G}_{11}(t)\widehat{u}^{0}+\mathcal{G}_{12}(t)\widehat{u}^{1}-\int_0^t\frac{d\mathcal{G}_{12}}{dt}(t-s)v\,ds. \tag*{(114)}
\end{aligned} u = ∫ 0 t v d s + G 11 ( t ) u 0 + G 12 ( t ) ( u 1 − v ( 0 )) − ∫ 0 t G 12 ( t − s ) ∂ s v d s = ∫ 0 t v d s + G 11 ( t ) u 0 + G 12 ( t ) u 1 − ∫ 0 t d t d G 12 ( t − s ) v d s . (114)
Let us obtain some regularity properties for G \mathcal{G} G . We have that for every ( u ~ 0 , u ~ 1 ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 (\widetilde{u}^{0},\widetilde{u}^{1})\in H^1(\mathcal{O})^3\times L^2(\mathcal{O})^3 ( u 0 , u 1 ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 the function
u ~ = G 11 ( t ) u ~ 0 + G 12 ( t ) u ~ 1 \widetilde{u}=\mathcal{G}_{11}(t)\widetilde{u}^{0}+\mathcal{G}_{12}(t)\widetilde{u}^{1} u = G 11 ( t ) u 0 + G 12 ( t ) u 1
is characterized by the existence of ( v ~ , p ~ ) ∈ L 2 ( 0 , ∞ ; D 1 , 2 ( R 3 ∖ O ) 3 ) ∩ L 2 ( 0 , ∞ ; L 2 ( R 3 ∖ O ) ) (\widetilde{v},\widetilde{p})\in L^2(0,\infty;D^{1,2}(\mathbb{R}^3\setminus\mathcal{O})^3)\cap L^2(0,\infty;L^2(\mathbb{R}^3\setminus\mathcal{O})) ( v , p ) ∈ L 2 ( 0 , ∞ ; D 1 , 2 ( R 3 ∖ O ) 3 ) ∩ L 2 ( 0 , ∞ ; L 2 ( R 3 ∖ O )) such that ( u ~ , v ~ , p ~ ) (\widetilde{u},\widetilde{v},\widetilde{p}) ( u , v , p ) solves
{ ρ s ∂ t t 2 u ~ − div y ( A e y ( u ~ ) ) = 0 in ( 0 , ∞ ) × O , − 2 μ div y ( e y ( v ~ ) ) + ∇ y p ~ = 0 in ( 0 , ∞ ) × ( R 3 ∖ O ) , div y v ~ = 0 in ( 0 , ∞ ) × ( R 3 ∖ O ) , ∂ t u ~ = v ~ , A e y ( u ~ ) ν = γ ( 2 μ e y ( v ~ ) − p ~ I ) ν on ( 0 , ∞ ) × ∂ O , u ~ ∣ t = 0 = u ~ 0 , ∂ t u ~ ∣ t = 0 = u ~ 1 in O . (115) \begin{cases}
\rho^s\partial_{tt}^2\widetilde{u}-\operatorname{div}_y(Ae_y(\widetilde{u}))=0 & \text{in }(0,\infty)\times\mathcal{O},\\
-2\mu\operatorname{div}_y(e_y(\widetilde{v}))+\nabla_y\widetilde{p}=0 & \text{in }(0,\infty)\times(\mathbb{R}^3\setminus\mathcal{O}),\\
\operatorname{div}_y\widetilde{v}=0 & \text{in }(0,\infty)\times(\mathbb{R}^3\setminus\mathcal{O}),\\
\partial_t\widetilde{u}=\widetilde{v},\quad Ae_y(\widetilde{u})\nu=\gamma(2\mu e_y(\widetilde{v})-\widetilde{p}I)\nu& \text{on }(0,\infty)\times\partial\mathcal{O},\\
\widetilde{u}|_{t=0}=\widetilde{u}^{0},\quad\partial_t\widetilde{u}|_{t=0}=\widetilde{u}^{1} & \text{in }\mathcal{O}. \tag*{(115)}
\end{cases} ⎩ ⎨ ⎧ ρ s ∂ tt 2 u − div y ( A e y ( u )) = 0 − 2 μ div y ( e y ( v )) + ∇ y p = 0 div y v = 0 ∂ t u = v , A e y ( u ) ν = γ ( 2 μ e y ( v ) − p I ) ν u ∣ t = 0 = u 0 , ∂ t u ∣ t = 0 = u 1 in ( 0 , ∞ ) × O , in ( 0 , ∞ ) × ( R 3 ∖ O ) , in ( 0 , ∞ ) × ( R 3 ∖ O ) , on ( 0 , ∞ ) × ∂ O , in O . (115)
For this system, the conservation of the energy reads as
1 2 d d t ∫ O ( ρ s ∣ ∂ t u ~ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y + 2 γ μ ∫ R 3 ∖ O ∣ e y ( v ~ ) ∣ 2 d y = 0. (116) \frac{1}{2}\frac{d}{dt}\int_{\mathcal{O}}\left(\rho^s|\partial_t\widetilde{u}|^2+Ae_y(\widetilde{u}):e_y(\widetilde{u})\right)\,dy+2\gamma\mu\int_{\mathbb{R}^3\setminus\mathcal{O}}|e_y(\widetilde{v})|^2\,dy=0. \tag*{(116)} 2 1 d t d ∫ O ( ρ s ∣ ∂ t u ∣ 2 + A e y ( u ) : e y ( u ) ) d y + 2 γ μ ∫ R 3 ∖ O ∣ e y ( v ) ∣ 2 d y = 0. (116)
On the other hand, for φ ∈ C c ∞ ( R 3 ) \varphi\in C^\infty_c(\mathbb{R}^3) φ ∈ C c ∞ ( R 3 ) such that φ = 1 \varphi= 1 φ = 1 in O \mathcal{O} O , (5.78) gives
ρ s ∂ t t 2 ∫ O u ~ d y = ∫ ∂ O A e y ( v ~ ) d s ( y ) = γ ∫ ∂ O ( 2 μ e y ( v ~ ) − p I ) ν d s ( y ) = − 2 γ μ ∫ R 3 ∖ O e y ( v ~ ) ∇ y φ d y + γ ∫ R 3 ∖ O p ~ ∇ y φ d y . (117) \rho^s\partial_{tt}^2\int_{\mathcal{O}}\widetilde{u}\,dy=\int_{\partial\mathcal{O}}A e_y(\widetilde{v})\,ds(y)=\gamma\int_{\partial\mathcal{O}}(2\mu e_y(\widetilde{v})-pI)\nu\,ds(y)
=-2\gamma\mu\int_{\mathbb{R}^3\setminus\mathcal{O}}e_y(\widetilde{v})\nabla_y\varphi\,dy+\gamma\int_{\mathbb{R}^3\setminus\mathcal{O}}\widetilde{p}\nabla_y\varphi\,dy. \tag*{(117)} ρ s ∂ tt 2 ∫ O u d y = ∫ ∂ O A e y ( v ) d s ( y ) = γ ∫ ∂ O ( 2 μ e y ( v ) − p I ) ν d s ( y ) = − 2 γ μ ∫ R 3 ∖ O e y ( v ) ∇ y φ d y + γ ∫ R 3 ∖ O p ∇ y φ d y . (117)
In the last term of this equality we can use that thanks to De Rham theorem and ∇ p ~ = 2 μ div e y ( v ~ ) \nabla\widetilde{p}=2\mu\operatorname{div}e_y(\widetilde{v}) ∇ p = 2 μ div e y ( v ) in R 3 ∖ O \mathbb{R}^3\setminus\mathcal{O} R 3 ∖ O , there exists a constant C > 0 C>0 C > 0 and a measurable function c ~ : ( 0 , ∞ ) → R \widetilde{c}:(0,\infty)\to\mathbb{R} c : ( 0 , ∞ ) → R such that
∥ p − c ~ ∥ L 2 ( R 3 ∖ O ) ≤ C ∥ e y ( v ~ ) ∥ L 2 ( R 3 ∖ O ) 3 × 3 a.e. in ( 0 , ∞ ) . \|p-\widetilde{c}\|_{L^2(\mathbb{R}^3\setminus\mathcal{O})}\leq C\|e_y(\widetilde{v})\|_{L^2(\mathbb{R}^3\setminus\mathcal{O})^{3\times3}}\quad\text{a.e. in }(0,\infty). ∥ p − c ∥ L 2 ( R 3 ∖ O ) ≤ C ∥ e y ( v ) ∥ L 2 ( R 3 ∖ O ) 3 × 3 a.e. in ( 0 , ∞ ) .
Using then that φ ∈ C c ∞ ( R 3 ∖ O ) \varphi\in C^\infty_c(\mathbb{R}^3\setminus\mathcal{O}) φ ∈ C c ∞ ( R 3 ∖ O ) , φ = 1 \varphi=1 φ = 1 in O \mathcal{O} O implies
∫ R 3 ∖ O ∇ y φ d y = − ∫ ∂ O ν d s ( y ) = 0 , \int_{\mathbb{R}^3\setminus\mathcal{O}}\nabla_y\varphi\,dy=-\int_{\partial\mathcal{O}}\nu\,ds(y)=0, ∫ R 3 ∖ O ∇ y φ d y = − ∫ ∂ O ν d s ( y ) = 0 ,
the last term in (5.80) satisfies
∫ R 3 ∖ O p ~ ∇ y φ d y = ∫ R 3 ∖ O ( p ~ − c ~ ) ∇ y φ d y . \int_{\mathbb{R}^3\setminus\mathcal{O}}\widetilde{p}\nabla_y\varphi\,dy=\int_{\mathbb{R}^3\setminus\mathcal{O}}(\widetilde{p}-\widetilde{c})\nabla_y\varphi\,dy. ∫ R 3 ∖ O p ∇ y φ d y = ∫ R 3 ∖ O ( p − c ) ∇ y φ d y .
Therefore (5.80) implies for another constant C > 0 C>0 C > 0
∣ ∫ O ∂ t t 2 u ~ d y ∣ ≤ C γ ∥ e y ( v ~ ) ∥ L 2 ( R 3 ∖ O ) 3 × 3 a.e. in ( 0 , ∞ ) , \left|\int_{\mathcal{O}}\partial_{tt}^2\widetilde{u}\,dy\right|\leq C\gamma\|e_y(\widetilde{v})\|_{L^2(\mathbb{R}^3\setminus\mathcal{O})^{3\times3}}\quad\text{a.e. in }(0,\infty), ∫ O ∂ tt 2 u d y ≤ C γ ∥ e y ( v ) ∥ L 2 ( R 3 ∖ O ) 3 × 3 a.e. in ( 0 , ∞ ) ,
which thanks to (5.79) proves
∫ R ∣ ∫ O ∂ t t 2 u ~ d y ∣ 2 d y ≤ C γ ( ∥ e y ( u ~ 0 ) ∥ L 2 ( O ) 3 × 3 2 + ∥ u ~ 1 ∥ L 2 ( O ) 3 2 ) . \int_{\mathbb{R}}\left|\int_{\mathcal{O}}\partial_{tt}^2\widetilde{u}\,dy\right|^2dy\leq C\gamma\left(\|e_y(\widetilde{u}^0)\|_{L^2(\mathcal{O})^{3\times3}}^2+\|\widetilde{u}^1\|_{L^2(\mathcal{O})^3}^2\right). ∫ R ∫ O ∂ tt 2 u d y 2 d y ≤ C γ ( ∥ e y ( u 0 ) ∥ L 2 ( O ) 3 × 3 2 + ∥ u 1 ∥ L 2 ( O ) 3 2 ) .
This inequality and (5.79) prove that the linear operators R ( t ) ∈ ( H 1 ( O ) 3 ) ∗ \mathcal{R}(t)\in(H^1(\mathcal{O})^3)^* R ( t ) ∈ ( H 1 ( O ) 3 ) ∗ , S ( t ) ∈ L 2 ( O ) 3 \mathcal{S}(t)\in L^2(\mathcal{O})^3 S ( t ) ∈ L 2 ( O ) 3 defined by
⟨ R ( t ) , u ~ 0 ⟩ = ρ s ∂ t t 2 ∫ O G 11 ( t ) u ~ 0 d y , ∫ O S ( t ) u ~ 1 d y = ρ s ∂ t t 2 ∫ O G 12 ( t ) u ~ 1 d y , (118) \langle\mathcal{R}(t),\widetilde{u}^0\rangle=\rho^s\partial_{tt}^2\int_{\mathcal{O}}\mathcal{G}_{11}(t)\widetilde{u}^0\,dy,\qquad\int_{\mathcal{O}}\mathcal{S}(t)\widetilde{u}^1\,dy=\rho^s\partial_{tt}^2\int_{\mathcal{O}}\mathcal{G}_{12}(t)\widetilde{u}^1\,dy, \tag*{(118)} ⟨ R ( t ) , u 0 ⟩ = ρ s ∂ tt 2 ∫ O G 11 ( t ) u 0 d y , ∫ O S ( t ) u 1 d y = ρ s ∂ tt 2 ∫ O G 12 ( t ) u 1 d y , (118)
for every u ~ 0 ∈ H 1 ( O ) 3 \widetilde{u}^0\in H^1(\mathcal{O})^3 u 0 ∈ H 1 ( O ) 3 , u ~ 1 ∈ L 2 ( O ) 3 \widetilde{u}^1\in L^2(\mathcal{O})^3 u 1 ∈ L 2 ( O ) 3 , satisfy R ∈ L 2 ( 0 , ∞ ; H 1 ( O ) 3 ) ∗ \mathcal{R}\in L^2(0,\infty;H^1(\mathcal{O})^3)^* R ∈ L 2 ( 0 , ∞ ; H 1 ( O ) 3 ) ∗ , S ∈ L 2 ( 0 , ∞ ; L 2 ( O ) 3 ) \mathcal{S}\in L^2(0,\infty;L^2(\mathcal{O})^3) S ∈ L 2 ( 0 , ∞ ; L 2 ( O ) 3 ) . Moreover, there exists C > 0 C>0 C > 0 such that (5.72) is satisfied.
Deriving twice with respect to t t t in (5.77) and using the first equation in (5.38), we deduce that v v v , p p p solve (5.70).
On the other hand, replacing in (5.73) v v v by a function w ∈ H 1 ( 0 , ∞ ) w\in H^1(0,\infty) w ∈ H 1 ( 0 , ∞ ) and u ^ 0 \widehat{u}^0 u 0 , u ^ 1 \widehat{u}^1 u 1 as functions u 0 ∈ H 1 ( O ) 3 u^0\in H^1(\mathcal{O})^3 u 0 ∈ H 1 ( O ) 3 , u 1 ∈ L 2 ( O ) 3 u^1\in L^2(\mathcal{O})^3 u 1 ∈ L 2 ( O ) 3 , multiplying the first equation by ∂ t u ^ − w \partial_t\widehat{u}-w ∂ t u − w , the second one by v ^ − v \widehat{v}-v v − v , integrating and adding, we get the energy identity
1 2 d d t ∫ O ( ρ s ∣ ∂ t u ^ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y + 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ^ ) ∣ 2 d y = ρ s ∫ O ∂ t t 2 u ^ d y ⋅ v in ( 0 , T ) , \frac{1}{2}\frac{d}{dt}\int_{\mathcal{O}}\left(\rho^s|\partial_t\widehat{u}|^2+A e_y(\widetilde{u}):e_y(\widetilde{u})\right)\,dy+2\mu\gamma\int_{\mathbb{R}^3\setminus\mathcal{O}}|e_y(\widehat{v})|^2\,dy=\rho^s\int_{\mathcal{O}}\partial_{tt}^2\widehat{u}\,dy\cdot v\quad\text{in }(0,T), 2 1 d t d ∫ O ( ρ s ∣ ∂ t u ∣ 2 + A e y ( u ) : e y ( u ) ) d y + 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ) ∣ 2 d y = ρ s ∫ O ∂ tt 2 u d y ⋅ v in ( 0 , T ) ,
which using (5.77) as above can be written as
1 2 d d t ∫ O ( ρ s ∣ ∂ t u ^ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y + 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ~ ) ∣ 2 d y = ( ∂ t ( 2 ρ s ∣ O ∣ w + ∫ 0 t S ( t − s ) w d s ) + R ( t ) u 0 + S ( t ) u 1 ) v . \frac{1}{2}\frac{d}{dt}\int_{\mathcal{O}}\left(\rho^{s}|\partial_t\hat{u}|^2+A e_y(\tilde{u}):e_y(\tilde{u})\right)\,dy+2\mu\gamma\int_{\mathbb{R}^3\setminus\mathcal{O}}|e_y(\tilde{v})|^2\,dy
=
\left(\partial_t\left(2\rho^{s}|\mathcal{O}|w+\int_0^t S(t-s)w\,ds\right)+\mathcal{R}(t)u^0+\mathcal{S}(t)u^1\right)v. 2 1 d t d ∫ O ( ρ s ∣ ∂ t u ^ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y + 2 μ γ ∫ R 3 ∖ O ∣ e y ( v ~ ) ∣ 2 d y = ( ∂ t ( 2 ρ s ∣ O ∣ w + ∫ 0 t S ( t − s ) w d s ) + R ( t ) u 0 + S ( t ) u 1 ) v .
Integrating in ( 0 , T ) (0,T) ( 0 , T ) , using (5.77) and the definitions of R \mathcal{R} R and S \mathcal{S} S , and taking into account that
∣ ∫ O ∂ t u ^ ( 0 , y ) d y + ∫ 0 t ∫ O ρ s ∂ t t 2 u ^ d y d s ∣ 2 = ∣ ∫ O ∂ t u ^ ( t , y ) d y ∣ 2 ≤ ∣ O ∣ ∫ O ∣ ∂ t u ^ ∣ 2 d y , \left|\int_{\mathcal{O}}\partial_t\hat{u}(0,y)\,dy+\int_0^t\int_{\mathcal{O}}\rho^{s}\partial_{tt}^2\hat{u}\,dy\,ds\right|^2=\left|\int_{\mathcal{O}}\partial_t\hat{u}(t,y)\,dy\right|^2\leq|\mathcal{O}|\int_{\mathcal{O}}|\partial_t\hat{u}|^2\,dy, ∫ O ∂ t u ^ ( 0 , y ) d y + ∫ 0 t ∫ O ρ s ∂ tt 2 u ^ d y d s 2 = ∫ O ∂ t u ^ ( t , y ) d y 2 ≤ ∣ O ∣ ∫ O ∣ ∂ t u ^ ∣ 2 d y ,
we deduce in particular (5.71)
Let us now consider the case γ = ∞ \gamma=\infty γ = ∞ . The reasoning is similar (somewhat simpler). Now, u ^ \hat{u} u ^ is obtained from v v v , u ^ 0 \hat{u}^0 u ^ 0 and u ^ 1 \hat{u}^1 u ^ 1 by solving the system
{ ρ s ∂ t t 2 u ^ − div y ( A e y ( u ^ ) ) = 0 in ( 0 , T ) × Ω × O ∂ t u ^ = v on ( 0 , T ) × Ω × ∂ O , u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in Ω × O , (119) \begin{cases}
\rho^{s}\partial_{tt}^2\hat{u}-\operatorname{div}_y(Ae_y(\hat{u}))=0 & \text{in }(0,T)\times\Omega\times\mathcal{O}\\
\partial_t\hat{u}=v & \text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
\hat{u}|_{t=0}=\hat{u}^0,\quad\partial_t\hat{u}|_{t=0}=\hat{u}^1 & \text{in }\Omega\times\mathcal{O}, \tag*{(119)}
\end{cases} ⎩ ⎨ ⎧ ρ s ∂ tt 2 u ^ − div y ( A e y ( u ^ )) = 0 ∂ t u ^ = v u ^ ∣ t = 0 = u ^ 0 , ∂ t u ^ ∣ t = 0 = u ^ 1 in ( 0 , T ) × Ω × O on ( 0 , T ) × Ω × ∂ O , in Ω × O , (119)
which implies that u ^ \hat{u} u ^ is still given by (5.77), where now for every ( u ~ 0 , u ~ 1 ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 (\tilde{u}^0,\tilde{u}^1)\in H^1(\mathcal{O})^3\times L^2(\mathcal{O})^3 ( u ~ 0 , u ~ 1 ) ∈ H 1 ( O ) 3 × L 2 ( O ) 3 the function
u ~ = G 11 ( t ) u ~ 0 + G 12 ( t ) u ~ 1 \tilde{u}=\mathcal{G}_{11}(t)\tilde{u}^0+\mathcal{G}_{12}(t)\tilde{u}^1 u ~ = G 11 ( t ) u ~ 0 + G 12 ( t ) u ~ 1
is defined as the solution of
{ ρ s ∂ t t 2 u ~ − div y ( A e y ( u ~ ) ) = 0 in ( 0 , T ) × Ω × O ∂ t u ~ = 0 on ( 0 , T ) × Ω × ∂ O , u ~ ∣ t = 0 = u ~ 0 , ∂ t u ~ ∣ t = 0 = u ~ 1 in Ω × O , (120) \begin{cases}
\rho^{s}\partial_{tt}^2\tilde{u}-\operatorname{div}_y(Ae_y(\tilde{u}))=0 & \text{in }(0,T)\times\Omega\times\mathcal{O}\\
\partial_t\tilde{u}=0 & \text{on }(0,T)\times\Omega\times\partial\mathcal{O},\\
\tilde{u}|_{t=0}=\tilde{u}^0,\quad\partial_t\tilde{u}|_{t=0}=\tilde{u}^1 & \text{in }\Omega\times\mathcal{O}, \tag*{(120)}
\end{cases} ⎩ ⎨ ⎧ ρ s ∂ tt 2 u ~ − div y ( A e y ( u ~ )) = 0 ∂ t u ~ = 0 u ~ ∣ t = 0 = u ~ 0 , ∂ t u ~ ∣ t = 0 = u ~ 1 in ( 0 , T ) × Ω × O on ( 0 , T ) × Ω × ∂ O , in Ω × O , (120)
The conservation of the energy now reads as
1 2 d d t ∫ O ( ρ s ∣ ∂ t u ~ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y = 0 , (121) \frac{1}{2}\frac{d}{dt}\int_{\mathcal{O}}\left(\rho^{s}|\partial_t\tilde{u}|^2+A e_y(\tilde{u}):e_y(\tilde{u})\right)\,dy=0, \tag*{(121)} 2 1 d t d ∫ O ( ρ s ∣ ∂ t u ~ ∣ 2 + A e y ( u ~ ) : e y ( u ~ ) ) d y = 0 , (121)
while thanks to the hidden regularity theorem for the elasticity system ([18 , 19 ] ), we deduce that for every t > 0 t>0 t > 0 , there exists C t > 0 C_t>0 C t > 0 such that
∣ ρ s ∫ 0 t ∫ O ∂ t t 2 u ~ d y d s ∣ 2 = ∣ ∫ 0 t ∫ ∂ O A e y ( u ~ ) d s ( y ) ∣ 2 ≤ t ∣ ∂ O ∣ ∫ 0 t ∫ ∂ O ∣ A e y ( u ~ ) ∣ 2 d s ( y ) ≤ C t ( ∥ u ~ 0 ∥ H 1 ( O ) 3 2 + ∥ u ~ 1 ∥ L 2 ( O ) 3 2 ) . \left|\rho^{s}\int_0^t\int_{\mathcal{O}}\partial_{tt}^2\tilde{u}\,dy\,ds\right|^2=\left|\int_0^t\int_{\partial\mathcal{O}}Ae_y(\tilde{u})\,ds(y)\right|^2
\leq t|\partial\mathcal{O}|\int_0^t\int_{\partial\mathcal{O}}|Ae_y(\tilde{u})|^2\,ds(y)\leq C_t\left(\|\tilde{u}^0\|_{H^1(\mathcal{O})^3}^2+\|\tilde{u}^1\|_{L^2(\mathcal{O})^3}^2\right). ρ s ∫ 0 t ∫ O ∂ tt 2 u ~ d y d s 2 = ∫ 0 t ∫ ∂ O A e y ( u ~ ) d s ( y ) 2 ≤ t ∣ ∂ O ∣ ∫ 0 t ∫ ∂ O ∣ A e y ( u ~ ) ∣ 2 d s ( y ) ≤ C t ( ∥ u ~ 0 ∥ H 1 ( O ) 3 2 + ∥ u ~ 1 ∥ L 2 ( O ) 3 2 ) .
This allows us to derive twice in (5.77) to deduce that u u u is still the solution of (5.70) where now R \mathcal{R} R , S \mathcal{S} S are just in L l o c 2 ( [ 0 , ∞ ) ; H 1 ( O ′ ) 3 ) L^2_{\mathrm{loc}}([0,\infty);H^1(\mathcal{O}')^3) L loc 2 ([ 0 , ∞ ) ; H 1 ( O ′ ) 3 ) and L l o c 2 ( [ 0 , ∞ ) ; L 2 ( O ) 3 ) L^2_{\mathrm{loc}}([0,\infty);L^2(\mathcal{O})^3) L loc 2 ([ 0 , ∞ ) ; L 2 ( O ) 3 ) respectively. Thanks to (5.84) they are such that there exists C > 0 C>0 C > 0 satisfying
∥ ∫ 0 t R d s ∥ ( H 1 ( O ′ ) 3 ) ∗ + ∥ ∫ 0 t S d s ∥ L 2 ( O ) 3 ≤ C , ∀ t ≥ 0. (122) \left\|\int_0^t\mathcal{R}\,ds\right\|_{(H^1(\mathcal{O}')^3)^*}+\left\|\int_0^t\mathcal{S}\,ds\right\|_{L^2(\mathcal{O})^3}\leq C,\qquad\forall t\geq0. \tag*{(122)} ∫ 0 t R d s ( H 1 ( O ′ ) 3 ) ∗ + ∫ 0 t S d s L 2 ( O ) 3 ≤ C , ∀ t ≥ 0. (122)
Inequality (5.71) follows as above, using the conservation of the energy for (5.82). ∎
Similarly, R ∞ ( t ) \mathcal{R}_{\infty}(t) R ∞ ( t ) is almost-periodic in t t t and, in general, does not decay as t → ∞ t \to\infty t → ∞ . This oscillatory, non-decaying character of the memory kernel is in stark contrast with the exponentially decaying kernels obtained in [11 ] , where the viscous dissipation inside the fluid inclusions acts as a damping mechanism. In the present case γ = ∞ \gamma= \infty γ = ∞ , the inclusions are purely elastic and isolated, so there is no microscopic dissipation and the memory persists indefinitely.
Acknowledgments: This work has been partially supported by FEDER project PID2023-149186NB-I00 and the IMUS-Maria de Maeztu grant CEX2024-001517-M - Apoyo a Unidades de Excelencia María de Maeztu. Both projects are funded by the Ministerio de Ciencia, Innovación y Universidades of the government of Spain.
References
[1 ] Z. Abdessamad, I. Kostin, G. Panasenko, V.P. Smyshlyaev. Memory effect in homogenization of a viscoelastic Kelvin-Voigt model with time-dependent coefficients. Math. Models Methods Appl. Sci. 19 (2009), 1603-1630.
[2 ] G. Allaire. Homogenization of the Navier-Stokes equations in open sets perforated with tiny holes I. Abstract framework, a volume distribution of holes. Arch. Rat. Mech. Anal. 113 (1991), 209–259.
[3 ] G. Allaire. Homogenization of the Navier–Stokes equations in open sets perforated with tiny holes II: Non-critical sizes of the holes for a volume distribution and a surface distribution of holes, Arch. Ration. Mech. Anal. 113 (1991), 261–298.
[4 ] G. Allaire. Homogenization and two-scale convergence. SIAM J. Math. Anal. 23 (1992), 1482-1518.