Research article

Local existence of a solution to a free boundary problem describing migration into rubber with a breaking effect

  • Received: 31 December 2021 Revised: 27 July 2022 Accepted: 26 September 2022 Published: 24 October 2022
  • We consider a one-dimensional free boundary problem describing the migration of diffusants into rubber. In our setting, the free boundary represents the position of the front delimitating the diffusant region. The growth rate of this region is described by an ordinary differential equation that includes the effect of breaking the growth of the diffusant region. In this specific context, the breaking mechanism is should be perceived as a non-dissipative way of describing eventual hyperelastic response to a too fast diffusion penetration. In recent works, we considered a similar class of free boundary problems modeling diffusants penetration in rubbers, but without attempting to deal with the possibility of breaking or accelerating the occurring free boundaries. For simplified settings, we were able to show the global existence and uniqueness as well as the large time behavior of the corresponding solutions to our formulations. Since here the breaking effect is contained in the free boundary condition, our previous results are not anymore applicable. The main mathematical obstacle in ensuring the existence of a solution is the non-monotonic structure of the free boundary. In this paper, we establish the existence and uniqueness of a weak solution to the free boundary problem with breaking effect and give explicitly the maximum value that the free boundary can reach.

    Citation: Kota Kumazaki, Toyohiko Aiki, Adrian Muntean. Local existence of a solution to a free boundary problem describing migration into rubber with a breaking effect[J]. Networks and Heterogeneous Media, 2023, 18(1): 80-108. doi: 10.3934/nhm.2023004

    Related Papers:

    [1] Avner Friedman . PDE problems arising in mathematical biology. Networks and Heterogeneous Media, 2012, 7(4): 691-703. doi: 10.3934/nhm.2012.7.691
    [2] Ciro D'Apice, Olha P. Kupenko, Rosanna Manzo . On boundary optimal control problem for an arterial system: First-order optimality conditions. Networks and Heterogeneous Media, 2018, 13(4): 585-607. doi: 10.3934/nhm.2018027
    [3] Peter V. Gordon, Cyrill B. Muratov . Self-similarity and long-time behavior of solutions of the diffusion equation with nonlinear absorption and a boundary source. Networks and Heterogeneous Media, 2012, 7(4): 767-780. doi: 10.3934/nhm.2012.7.767
    [4] Naoki Sato, Toyohiko Aiki, Yusuke Murase, Ken Shirakawa . A one dimensional free boundary problem for adsorption phenomena. Networks and Heterogeneous Media, 2014, 9(4): 655-668. doi: 10.3934/nhm.2014.9.655
    [5] Jésus Ildefonso Díaz, Tommaso Mingazzini, Ángel Manuel Ramos . On the optimal control for a semilinear equation with cost depending on the free boundary. Networks and Heterogeneous Media, 2012, 7(4): 605-615. doi: 10.3934/nhm.2012.7.605
    [6] Bendong Lou . Self-similar solutions in a sector for a quasilinear parabolic equation. Networks and Heterogeneous Media, 2012, 7(4): 857-879. doi: 10.3934/nhm.2012.7.857
    [7] Brahim Amaziane, Leonid Pankratov, Andrey Piatnitski . An improved homogenization result for immiscible compressible two-phase flow in porous media. Networks and Heterogeneous Media, 2017, 12(1): 147-171. doi: 10.3934/nhm.2017006
    [8] Anya Désilles, Hélène Frankowska . Explicit construction of solutions to the Burgers equation with discontinuous initial-boundary conditions. Networks and Heterogeneous Media, 2013, 8(3): 727-744. doi: 10.3934/nhm.2013.8.727
    [9] Linglong Du . Characteristic half space problem for the Broadwell model. Networks and Heterogeneous Media, 2014, 9(1): 97-110. doi: 10.3934/nhm.2014.9.97
    [10] Dilip Sarkar, Shridhar Kumar, Pratibhamoy Das, Higinio Ramos . Higher-order convergence analysis for interior and boundary layers in a semi-linear reaction-diffusion system networked by a $ k $-star graph with non-smooth source terms. Networks and Heterogeneous Media, 2024, 19(3): 1085-1115. doi: 10.3934/nhm.2024048
  • We consider a one-dimensional free boundary problem describing the migration of diffusants into rubber. In our setting, the free boundary represents the position of the front delimitating the diffusant region. The growth rate of this region is described by an ordinary differential equation that includes the effect of breaking the growth of the diffusant region. In this specific context, the breaking mechanism is should be perceived as a non-dissipative way of describing eventual hyperelastic response to a too fast diffusion penetration. In recent works, we considered a similar class of free boundary problems modeling diffusants penetration in rubbers, but without attempting to deal with the possibility of breaking or accelerating the occurring free boundaries. For simplified settings, we were able to show the global existence and uniqueness as well as the large time behavior of the corresponding solutions to our formulations. Since here the breaking effect is contained in the free boundary condition, our previous results are not anymore applicable. The main mathematical obstacle in ensuring the existence of a solution is the non-monotonic structure of the free boundary. In this paper, we establish the existence and uniqueness of a weak solution to the free boundary problem with breaking effect and give explicitly the maximum value that the free boundary can reach.



    Both natural rubber and synthetic elastomers are widely used in engineering applications, such as connecting components for offshore wind farms that are nowadays intensively used to harvesting energy. Due to the growing importance of good theoretical estimations of durability properties of elastomeric parts, a well-trusted modelling of the material behaviour is an essential prerequisite so that numerical prediction is available for large times. Rubber-like materials exhibit a strongly non-linear behaviour characterized by a joint large strain and a non-linear stress-strain response [3]. In the presence of aggressive environments (like those where large bodies of water are present), the nonlinear behavior of elastomers, which can be in fact studied in controlled laboratory conditions, alters non-intuitively [2]. The main reason for this situation is that the porous structure of rubber-like materials allows small particles to diffuse inside. Ingressing particles, often ionically charged, accumulate and interact with the internal solid fabric. Such interactions lead to unwanted alterations of the originally designed mechanical behavior; see e.g., [8] for more context on this physical problem. From the modeling and mathematical upscaling point of view, the derivation of the correct model equations for such a problem setting is open.

    In our recent works [1,5], and [6], we considered a class of one-dimensional free boundary problems to model the diffusants penetration in dense and foam rubbers. The use of kinetic-type free boundary conditions is meant to avoid the explicit description of the mechanics of the involved materials and how the constitutive laws (in particular, the stress tensor) are affected by the presence of diffusants. One can use such free boundary formulations to derive theoretically-grounded practical estimations of the position of the diffusants penetration front and compare them with laboratory experiments; e.g., [7] where a parameter identification exercise has been performed in this context.

    The main novelty introduced in the present paper is that the growth rate of the region filled with diffusants is described by an ordinary differential equation that includes the effect of breaking its growth. The breaking mechanism should be seen here as the hyperelastic response to a too fast diffusion penetration. Such a mechanism introduces non-monotonicity in our problem formulation. This makes us wonder whether this type of problems has a chance to be well-posed and, if true, in which sense?

    To address the question, we consider the following one-dimensional free boundary problem show in Eqs (1.1)–(1.5). Let t[0,T] be the time variable for T>0. The problem is to find a curve z=s(t) on [0,T] and a function u defined on the set Qs(T):={(t,z)|0<t<T, 0<z<s(t)} such that the following system of equations is satisfied:

    utuzz=0 for (t,z)Qs(T), (1.1)
    uz(t,0)=β(b(t)γu(t,0)) for t(0,T), (1.2)
    uz(t,s(t))=σ(u(t,s(t)))st(t) for t(0,T), (1.3)
    st(t)=a0(σ(u(t,s(t)))αs(t)) for t(0,T), (1.4)
    s(0)=s0,u(0,z)=u0(z) for z[0,s0], (1.5)

    where β, γ, a0 and α are given positive constants, b is a given function on [0,T] and s0 and u0 are the initial data. In Eq (1.3) and Eq (1.4), σ is the function on R called the positive part, i.e. it is given by Eq (1.6),

    σ(r)={r if r0,0 if r<0. (1.6)

    The problem in Eqs (1.1)–(1.5), which is denoted here by (P)(u0,s0,b), was proposed by Nepal S et al., in [7] as a mathematical model describing the migration of diffusants into rubber. The set [0,s(t)] represents the region occupied by a solvent occupying the one-dimensional pore [0,), where s=s(t) is the position of moving interface of the region and u=u(t,z) represents the content of the diffusant at the place z[0,s(t)] at time t>0.

    From the mathematical point of view, (P)(u0,s0,b) is a one-phase free boundary problem with Robin-type boundary conditions at both boundaries. Such kind of problem structure has been considered in [5,6], and [1]. In our recent work [1], as a simplified setting for (P)(u0,s0,b), we consider the case α=0 and prove the existence and uniqueness of a globally-in-time solution. Furthermore, what concerns the large time behavior of a solution, we show that the free boundary s goes to infinity as time elapses.

    In this paper, we consider the case α>0. As already anticipated, the difficulty in this case is the lack of the monotonicity of the boundary condition imposed on the moving boundary. Indeed, similarly to the proof of the existence result in [1], by introducing a variable ˜u(t,y)=u(t,ys(t)) for (t,y)Q(T):=(0,T)×(0,1) we transform (P)(u0,s0,b) into a problem on the fixed domain Q(T). Let denote this problem by (PC)(˜u0,s0,b), where ˜u0(y)=u0(s0y) for y[0,1]. Moreover, to find a solution (PC)(˜u0,s0,b) we first consider the following auxiliary problem (AP)(˜u0,s,b): For a given function s(t) on [0,T], find ˜u(t,y) satisfying:

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)˜uy(t,y) for (t,y)Q(T),1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T),1s(t)˜uy(t,1)=a0σ(˜u(t,1))(σ(˜u(t,1))αs(t)) for t(0,T),˜u(0,y)=˜u0(y) for y[0,1].

    Here, we set gα(r)=a0σ(r)(σ(r)αs(t)) for rR and t(0,T). In the case α=0, due to the function σ, g0(r)=a0(σ(r))2 is monotonically increasing with respect to r. Due to this fact, we can use the theory of evolution equations governed by subdifferentials of convex functions (cf. [4] and references therein) and find a solution ˜u of (AP)(˜u0,s,b) in the case α=0. However, looking now at the case α>0, since gα(r) is not monotonic anymore with respect to r, we cannot apply the theory of evolution equations directly. Hence, the existence of a solution to (AP)(˜u0,s,b) in the case α>0 is not at all clear. As far as we are aware, the type of free boundary problems is novel. We refer the reader to [9] (and related references) for explanations of how and why free-boundary problems can be used to model fast transitions in materials.

    The purpose of this paper is to establish a methodology to deal with the existence of locally-in-time solutions to (PC)(u0,s0,b) in the case α>0. For this to happen, we consider weak solutions to (PC)(˜u0,s0,b). The definition of our concept of weak solutions is explained in Section 2. As next step, we proceed in the following way: For a given function η on Q(T) and ε>0, we consider in Section 3 a smooth approximation ηε of η and construct a solution ˜u to the following auxiliary problem: For given s on [0,T], find ˜u(t,y) such that it holds

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)ηy(t,y) for (t,y)Q(T),1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T),1s(t)˜uy(t,1)=a0(σ(˜u(t,1)))2ασ(ηε(t))s(t)) for t(0,T),˜u(0,y)=˜u0(y) for y[0,1].

    The plan is to obtain uniform estimates of solutions with respect to ε. After that, by the limiting process ε0 and benefitting of Banach's fixed point theorem, we wish to construct a weak solution ˜u of (AP)(˜u0,s,b). In Section 4, we define a solution mapping ΓT between s and the weak solution ˜u of (AP)(˜u0,s,b). We show that, for some TT, the mapping ΓT is a contraction mapping on a suitable function space. Finally, using Banach's fixed point theorem, we prove the existence and uniqueness of a weak solution to the coupled problem (s,˜u) of (PC)(˜u0,s0,b) on [0,T]. Additionally, we show that the maximal length of the free boundary s is a priori determined by given parameters, the time derivative of the function b, and initial data s0. Moreover, we guarantee that the solution ˜u is non-negative and bounded on Q(T).

    In this paper, we use the following notations. We denote by ||X the norm for a Banach space X. The norm and the inner product of a Hilbert space H are denoted by ||H and (,)H, respectively. Particularly, for ΩR, we use the notation of the usual Hilbert spaces L2(Ω), H1(Ω), and H2(Ω). Throughout this paper, we assume the following parameters and functions:

    (A1) a0, α, γ, β and T are positive constants.

    (A2) s0>0 and u0L(0,s0) such that u00 on [0,s0].

    (A3) bW1,2(0,T) with b0 on (0,T). Also, we set

    b=max{max0tTb(t),γ|u0|L(0,s0)}.

    It is convenient to consider (P)(u0,s0,b) transformed into a non-cylindrical domain. Let T>0. For given sW1,2(0,T) with s(t)>0 on [0,T], we introduce the following new function obtained by the change of variables and fix the moving domain:

    ˜u(t,y)=u(t,ys(t)) for (t,y)Q(T):=(0,T)×(0,1). (2.1)

    By using the function ˜u, (P)(u0,s0,b) becomes the following problem (PC)(˜u0,s0,b) posed on the non-cylindrical domain Q(T):

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)˜uy(t,y) for (t,y)Q(T), (2.2)
    1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T), (2.3)
    1s(t)˜uy(t,1)=σ(˜u(t,1))st(t) for t(0,T), (2.4)
    st(t)=a0(σ(˜u(t,1))αs(t)) for t(0,T), (2.5)
    s(0)=s0, (2.6)
    ˜u(0,y)=u0(ys(0))(:=˜u0(y)) for y[0,1]. (2.7)

    Here, we introduce the following function space: For T>0, we put H=L2(0,1), X=H1(0,1), V(T)=L(0,T;H)L2(0,T;X) and |z|V(T)=|z|L(0,T;H)+|zy|L2(0,T;H) for zV(T). Note that V(T) is a Banach space with the norm ||V(T). Also, we denote by X and ,X the dual space of X and the duality pairing between X and X, respectively.

    We define now our concept of solutions to (PC)(˜u0,s0,b) on [0,T] in the following way:

    Definition 2.1. For T>0, let s be a function on [0,T] and ˜u be a function on Q(T), respectively. We call that a pair (s,˜u) is a solution of (P)(˜u0,s0,b) on [0,T] if the next conditions (S1)–(S4) hold:

    (S1) sW1,(0,T), s>0 on [0,T], ˜uW1,2(0,T;X)V(T).

    (S2)

    T0˜ut(t),z(t)Xdt+Q(T)1s2(t)˜uy(t)zy(t)dydt+T01s(t)σ(˜u(t,1))st(t)z(t,1)dtT01s(t)β(b(t)γ˜u(t,0))z(t,0)dt=Q(T)yst(t)s(t)˜uy(t)z(t)dydt for zV(T).

    (S3) st(t)=a0(σ(˜u(t,1))αs(t)) for a.e. t(0,T).

    (S4) s(0)=s0 and ˜u(0,y)=˜u0(y) for a.e. y[0,1].

    The main result of this paper is the existence and uniqueness of a locally-in-time solution of (PC)(˜u0,s0,b). We state this result in the next theorem.

    Theorem 2.2. Let T>0. If (A1)–(A3) hold, then there exists TT such that (PC)(˜u0,s0,b) has a unique solution (s,˜u) on [0,T]. Moreover, the function ˜u is non-negative and bounded on Q(T).

    In this section,

    we consider the following auxiliary problem (AP1)(˜u0,s,b): For T>0 and sW1,2(0,T) with s>0 on [0,T]

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)˜uy(t,y) for (t,y)Q(T), (3.1)
    1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T), (3.2)
    1s(t)˜uy(t,1)=a0σ(˜u(t,1))(σ(˜u(t,1))αs(t)) for t(0,T), (3.3)
    ˜u(0,y)=˜u0(y) for y[0,1], (3.4)

    where σ is the same function as in Eq (1.6).

    Definition 3.1. For T>0, let ˜u be a function on Q(T), respectively. We call that a function ˜u is a solution of (AP1)(˜u0,s,b) on [0,T] if the conditions (S'1)–(S'3) hold:

    (S'1) ˜uW1,2(0,T;X)V(T).

    (S'2)

    T0˜ut(t),z(t)Xdt+Q(T)1s2(t)˜uy(t)zy(t)dydt+T0a0s(t)σ(˜u(t,1))(σ(˜u(t,1))αs(t))z(t,1)dtT01s(t)β(b(t)γ˜u(t,0))z(t,0)dt=Q(T)yst(t)s(t)˜uy(t)z(t)dydt for zV(T).

    (S'3) ˜u(0,y)=˜u0(y) for y[0,1].

    Now, we introduce the following problem (AP2)(˜u0,s,η,b): For sW1,(0,T) with s>0 on [0,T] and ηV(T)

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)ηy(t,y) for (t,y)Q(T), (3.5)
    1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T), (3.6)
    1s(t)˜uy(t,1)=a0((σ(˜u(t,1))2ασ(η(t,1))s(t)) for t(0,T), (3.7)
    ˜u(0,y)=˜u0(y) for y[0,1], (3.8)

    The definition of solutions of (AP2)(˜u0,s,η,b) is Definition 3.1 with (S'2) replaced by (S''2), which now reads

    (S''2):

    T0˜ut(t),z(t)Xdt+Q(T)1s2(t)˜uy(t)zy(t)dydt+T0a0s(t)((σ(˜u(t,1))2ασ(η(t,1))s(t))z(t,1)dtT01s(t)β(b(t)γ˜u(t,0))z(t,0)dt=Q(T)yst(t)s(t)ηy(t)z(t)dydt for zV(T).

    First, we construct a solution ˜u of (AP2)(˜u0,s,η,b) on [0,T]. To do so, for each ε>0 we solve the following problem (AP2)ε(˜u0ε,s,η,b):

    ˜ut(t,y)1s2(t)˜uyy(t,y)=yst(t)s(t)ηy(t,y) for (t,y)Q(T),1s(t)˜uy(t,0)=β(b(t)γ˜u(t,0)) for t(0,T),1s(t)˜uy(t,1)=a0((σ(˜u(t,1))2ασ((ρεη)(t,1))s(t)) for t(0,T),˜u(0,y)=˜u0ε(y) for y[0,1].

    Here ρε is a mollifier with support [ε,ε] in time and ρεη is the convolution of ρε with η:

    (ρεη)(t,1)=ρε(ts)¯η(s,1)ds for t[0,T], (3.9)

    where ¯η(t,1)=η(t,1) for t(0,T) and vanishes otherwise. Also, ˜u0ε is an approximation function of ˜u0 such that {˜u0ε}X, |˜u0ε|H|˜u0|H+1 and ˜u0ε˜u0 in H as ε0.

    Now, we define a family {ψt}t[0,T] of time-dependent functionals ψt:HR{+} for t[0,T] as follows:

    ψt(u):={12s2(t)10|uy(y)|2dy+1s(t)u(1)0a0((σ(ξ))2dξa0αu(1)σ((ρεη)(t,1))1s(t)u(0)0β(b(t)γξ)dξ if uD(ψt),+otherwise,

    where D(ψt)=X for t[0,T]. Here, we show the property of ψt.

    Lemma 3.2. Let sW1,2(0,T) with s>0 on [0,T], ηV(T) and assume (A1)–(A3). Then the following statements hold:

    (1) There exists positive constants C0, C1 and C such that the following inequalities hold:

    (i) |u(y)|2Cy(ψt(u)+1)foruD(ψt)andy=0,1,(ii) 12s2(t)|uy|2HC(ψt(u)+1)foruD(ψt),

    (2) For t[0,T], the functional ψt is proper, lower semi-continuous, and convex on H.

    Proof. We fix ε>0 and let t[0,T], uD(ψt) and put l=max0tT|s(t)| and ηε(t)=(ρεη)(t,1) for t[0,T]. If u(1)<0, 1s(t)u(1)0a0(σ(ξ))2dξ=0. If u(1)0, then it holds

    1s(t)u(1)0a0(σ(ξ))2dξa0αu(1)σ(ηε(t))=a0s(t)13u3(1)a0αu(1)σ(ηε(t))a03s(t)u3(1)13δ3u3(1)23δ3/2(a0ασ(ηε(t)))32,

    where δ is an arbitrary positive number. By using the fact that σ(r)|r| for rR and taking a suitable δ=δ0 we have

    1s(t)u(1)0a0(σ(ξ))2dξa0αu(1)ηε(t)23δ3/20(a0αηε)32, (3.10)

    where ηε=max0tT|ηε(t)|. Moreover, for both cases u(0)<0 and u(0)0, we observe that

    1s(t)u(0)0β(b(t)γξ)dξ=βs(t)[γ2u2(0)b(t)u(0)]βγ2lu2(0)βbau(0)βγ4lu2(0)βlγ(ba)2. (3.11)

    Accordingly, if u(1)<0, then we have that,

    ψt(u)12s2(t)10|uy(y)|2dy+βγ4lu2(0)βlγ(ba)2. (3.12)

    If u(1)0, then, by Eq (3.10) and Eq (3.11) we also have that:

    ψt(u)12s2(t)10|uy(y)|2dy23δ3/20(a0αηε)32+βγ4lu2(0)βlγ(ba)2. (3.13)

    In Eq (3.12) and Eq (3.13), since the first term in the right-hand side is non-negative we can find a positive constant C0 that (i) of Lemma 3.2 for y=0 holds. In addition, by βγ4lu2(0)0 we also see that (ii) of Lemma 3.2 holds. Moreover, it holds that:

    |u(1)|2=|10uy(y)dy+u(0)|22(10|uy(y)|2dy+|u(0)|2)2(2l22s2(t)10|uy(y)|2dy+|u(0)|2). (3.14)

    Therefore, by Eq (3.14) and the result for |u(0)|2 and |uy|2H we see that there exists a positive constant C1 such that (i) of Lemma 3.2 hold for y=1.

    Next, we prove statement (2). For t[0,T] and rR, put

    g1(s(t),ηε(t),r)=1s(t)r0a0(σ(ξ))2dξa0αrσ(ηε(t)),g2(s(t),b(t),r)=1s(t)r0β(b(t)γξ)dξ.

    Then, by g1(s(t),ηε(t),r)=a0αrσ(ηε(t)) for r0, g1(s(t),ηε(t),r) is a linear decreasing function for r0. Also, by s(t)>0 we see that,

    2r2g1(s(t),ηε(t),r)=2a0s(t)r>0 for r>0,2r2g2(s(t),b(t),r)=βγs(t)>0 for rR.

    This means that ψt is convex on H. By using (i) and (ii) of Lemma 3.2 together with Sobolev's embedding XC([0,1]) in one dimensional case, it is easy to prove that the level set of ψt is closed in H, fact which ensures to the lower semi-continuity of ψt. Thus, we see that statement (2) holds.

    Lemma 3.3 guarantees the existence of a solution to (AP2)ε(˜u0ε,s,η,b).

    Lemma 3.3. Let T>0 and ε>0. If (A1)–(A3) hold, then, for given sW1,(0,T) with s>0 on [0,T] and ηV(T), the problem (AP2)ε(˜u0ε,s,η,b) admits a unique solution ˜u on [0,T] such that ˜uW1,2(0,T;H)L(0,T;X). Moreover, the function tψt(˜u(t)) is absolutely continuous on [0,T].

    Proof. Let ε>0 be arbitrarily fixed. By Lemma 3.2, for t[0,T] ψt is a proper lower semi-continuous convex function on H. From the definition of the subdifferential of ψt, for t[0,T], zψt(u) is characterized by u, zH,

    z=1s2(t)uyy on (0,1),1s(t)uy(0)=β(b(t)γu(0)), 1s(t)uy(1)=a0((σ(u(1))2ασ((ρεη)(t,1))s(t)).

    Namely, ψt is single-valued. Also, we see that there exists a positive constant C such that for each t1, t2[0,T] with t1t2, and for any uD(ψt1), there exists ˉuD(ψt2) satisfying the inequality

    |ψt2(ˉu)ψt1(u)|C(|s(t1)s(t2)|+|b(t1)b(t2)|+|ηε(t1)ηε(t2)|(1+|ψt1(u)|), (3.15)

    where ηε(t)=(ρεη)(t). Indeed, we take ˉu:=u. Then, ˉuD(ψt2), and by the definition of ψt, it holds

    ψt2(ˉu)ψt1(u)=(12s2(t2)12s2(t1))10|uy(y)|2dy+(1s(t2)1s(t1))u(1)0a0(σ(ξ))2dξ+a0αu(1)(ηε(t1)ηε(t2))(1s(t2)1s(t1))u(0)0β(b(t2)γξ)dξβs(t1)(b(t2)b(t1))u(0). (3.16)

    We denote each term in the right-hand side by I1, I2, I3, I4 and I5. Let put l=max0tT|s(t)| and a=min0tT|s(t)|. For other than I2, by the fact that |u(y)|12(1+u2(y)) for y=0, 1 and Lemma 3.2 it easy to see that,

    |I1|C1|s(t1)s(t2)|(|ψt1(u)|+1), (3.17)
    5k=3|Ik|C2(|ηε(t1)ηε(t2)|+|s(t2)s(t1)|+|b(t2)b(t1)|)(|ψt1(u)|+1), (3.18)

    where C1 and C2 are positive constants. For I2 by the definition of ψt it holds that,

    1s(t1)u(1)0a0(σ(ξ))2dξ=ψt1(u)12s2(t1)10|uy(y)|2dy+a0αu(1)ηε(t1)+1s(t1)u(0)0β(b(t1)γξ)dξ.

    Here, we note that

    |a0αu(1)ηε(t1)|a0αηε2(1+u2(1)),|1s(t1)u(0)0β(b(t1)γξ)dξ|1a(βb|u(0)|+γ2u2(0))1a(βb2+(βb2+γ2)u2(0)).

    where ηε=max0tT|ηε(t)|. Hence, by Lemma 3.2 we have that,

    |I2|C3|s(t1)s(t2)|(|ψt1(u)|+1), (3.19)

    where C3 is a positive constant. As a consequence, by Eqs (3.17)–(3.19) we infer that there exists a positive constant C such that Eq (3.15) holds.

    Now, (AP2)ε(˜u0ε,s,η,b) can be written into the following Cauchy problem (CP)ε:

    ˜ut+ψt(˜u(t))=yst(t)s(t)ηy(t) in H,˜u(0,y)=˜u0ε(y) for y[0,1].

    Since ystsηyL2(0,T;H), by the general theory of evolution equations governed by time dependent subdifferentials (cf. [4]) we see that (CP)ε has a solution ˜u on [0,T] such that ˜uW1,2(Q(T)), ψt(˜u(t))L(0,T) and tψt(˜u(t)) is absolutely continuous on [0,T]. This implies that ˜u is a unique solution of (AP2)ε(˜u0ε,s,η,b) on [0,T].

    As next step, we provide an uniform estimate with respect to ε on a solution ˜u of (AP2)ε(˜u0ε,s,η,b).

    Lemma 3.4. Let T>0, sW1,(0,T) with s>0 on [0,T], ηV(T) and ˜uε be a solution of (AP2)ε(˜u0ε,s,η,b) on [0,T] for each ε>0. Then, it holds that

    |˜uε(t)|2H+t0|˜uεy(τ)|2HdyM(1+|η|2V(T))fort[0,T]andε(0,1], (3.20)

    where M=M(a0,a,β,b,T) is a positive constant which is independent of ε and depends on a0, a, β, b and T, a=min0tTs(t).

    Proof. Let ˜uε be a solution of (AP2)ε(˜u0ε,s,η,b) on [0,T] for each ε>0.

    First, it holds that

    12ddt|˜uε(t)|2H101s2(t)˜uεyy(t)˜uε(t)dy=10yst(t)s(t)ηy(t)˜uε(t)dy. (3.21)

    The second term on the left-hand side is as follows:

    101s2(t)˜uεyy(t)˜uε(t)dy=a0s(t)((σ(˜uε(t,1))2ασ(ηε(t))s(t))˜uε(t,1)1s(t)β(b(t)γ˜uε(t,0))˜uε(t,0)+1s2(t)10|˜uεy(t)|2dya0α|ηε(t)||˜uε(t,1)|1s(t)β(b(t)γ˜uε(t,0))˜uε(t,0)+1s2(t)10|˜uεy(t)|2dy,

    where ηε(t)=(ρεη)(t,1). From the above, we obtain that

    12ddt|˜uε(t)|2H+1s2(t)10|˜uεy(t)|2dy10yst(t)s(t)ηy(t)˜uε(t)dy+a0α|ηε(t)||˜uε(t,1)|+1s(t)β(b(t)γ˜uε(t,0))˜uε(t,0) for t[0,T]. (3.22)

    We estimate the right-hand side of Eq (3.22). First, by Young's inequality we have that,

    10yst(t)s(t)ηy(t)˜uε(t)dy14s2(t)10|ηy(t)|2dy+|st(t)|210|˜uε(t)|2dy. (3.23)

    Here, by Sobolev's embedding theorem in one dimension, we note that it holds that,

    |z(y)|2Ce|z|X|z|H for zX and y[0,1], (3.24)

    where Ce is a positive constant defined from Sobolev's embedding theorem. By Eq (3.24) and sa on [0,T], we obtain

    1s(t)β(b(t)γ˜uε(t,0))˜uε(t,0)βbs(t)|˜uε(t,0)|βbCe2s(t)(|˜uεy(t)|H|˜uε(t)|H+|˜uε(t)|2H)+βb2s(t)14s2(t)|˜uεy(t)|2H+((βbCe)24+βbCe2a)|˜uε(t)|2H+βb2a, (3.25)

    and

    a0α|ηε(t)||˜uε(t,1)|a0α2(Ce(|˜uεy(t)|H|˜uε(t)|H+|˜uε(t)|2H)+|ηε(t)|2)14s2(t)|˜uεy(t)|2H+(s2(t)(a0α2Ce)2+a0α2Ce)|˜uε(t)|2H+a0α2|ηε(t)|2. (3.26)

    From Eqs (3.21)–(3.26), we have that for t[0,T]

    12ddt|˜uε(t)|2H+12s2(t)10|˜uεy(τ)|2dy(|st(t)|2+(βbCe)24+βbCe2a)|˜uε(t)|2H+βb2a+(s2(t)(a0α2Ce)2+a0α2Ce)|˜uε(t)|2H+14s2(t)|ηy(t)|2H+a0α2|ηε(t)|2. (3.27)

    Denote F(t) the coefficient of |˜uε|2H in the right-hand side. As sW1,2(0,T), we observe that FL1(0,T). Then, by Gronwall's inequality, we have for t1[0,T],

    12|˜uε(t1)|2H+12l2t10|˜uεy(τ)|2Hdy(Gε(T)t10F(t)dt)et10F(t)dt (3.28)

    where Gε(t)=12(|˜u0|2H+1+βbaT+12a2t0|ηy(t)|2Hdt+a0αt0|ηε(t)|2dt) for t[0,T]. We note that it holds that

    T0|ηε(τ,1)|2dτT0|η(τ,1)|2dτCeT0(|ηy(τ)|H|η(τ)|H+|η(τ)|2H)dtCe(|η|L(0,T;H)T1/2(T0|ηy(τ)|2Hdτ)1/2+T|η|2L(0,T;H))CeT1/2(1+T1/2)|η|2V(T1). (3.29)

    Therefore, by Eq (3.28) and Eq (3.29) we see that there exists a positive constant M=M(a0,a,β,b,T) such that Lemma 3.4 holds.

    Lemma 3.5. Let T>0, sW1,(0,T) with s>0 on [0,T] and ηV(T). If (A1)–(A3) hold, then, (AP2)(˜u0,s,η,b) has a unique solution ˜u on [0,T].

    Proof. Let sW1,(0,T) with s>0 on [0,T]. Then, we already have a solution ˜uε of (AP2)ε(˜u0ε,s,η,b) on [0,T] for each ε>0. By letting ε0 we show the existence of a solution ˜u of (AP2)(˜u0,s,η,b) on [0,T]. First, by Lemma 3.4 we see that {˜uε} is bounded in L(0,T;H)L2(0,T;X). Next, for zX, it holds that,

    |10˜uεt(t)zdy|=|1s2(t)(10˜uεy(t)zydy)a0s(t)((σ(˜uε(t,1))2ασ((ρεη)(t,1))s(t))z(1)+1s(t)β(b(t)γ˜uε(t,0))z(0)+10yst(t)s(t)ηy(t)zdy|1a2|˜uεy(t)|H|zy|H+a0a|˜uε(t,1)|2|z(1)|+a0α|(ρεη)(t,1)||z(1)|+βba|z(0)|+βγa|˜uε(t,0)||z(0)|+|st(t)|a|ηy(t)|H|z|H for a.e. t[0,T]. (3.30)

    By the estimate Eq (3.30) and Eq (3.24) we infer that {˜uεt} is bounded in L2(0,T;X). Therefore, we take a subsequence {εi}{ε} such that for some ˜uW1,2(0,T;X)L(0,T;H)L2(0,T;X), ˜uεi˜u weakly in W1,2(0,T;X)L2(0,T;X), weakly-* in L(0,T;H) as i. Also, by Aubin's compactness theorem, we see that ˜uεi˜u in L2(0,T;H) as i.

    Now, we prove that the limit function ˜u is a solution of (AP2)(˜u0,s,η,b) on [0,T] satisfying ˜uW1,2(0,T;X)V(T), (S''2) and (S3). Let zV(T). Then, it holds that,

    T010˜uεt(t)z(t)dydt+T01s2(t)(10˜uεy(t)zy(t)dy)dt+T0a0s(t)((σ(˜uε(t,1))2ασ((ρεη)(t,1))s(t))z(t,1)dtT01s(t)β(b(t)γ˜uε(t,0))z(t,0)dt=T010yst(t)s(t)ηy(t)z(t)dydt. (3.31)

    From the weak convergence, it is easy to see that,

    T010˜uεit(t)z(t)dydtT0˜ut(t),z(t)Xdt,T01s2(t)(10˜uεiy(t)zy(t)dy)dtT01s2(t)(10˜uy(t)zy(t)dy)dt as i.

    The third term of the left-hand side of Eq (3.31) is as follows:

    |T0a0s(t)((σ(˜uε(t,1))2ασ((ρεη)(t,1))s(t))z(t,1)dtT0a0s(t)((σ(˜u(t,1))2ασ(η(t,1))s(t))z(t,1)dt|a0a(T0|˜uε(t,1)˜u(t,1)|2dt)1/2(T02(|˜uε(t,1)|2+|˜u(t)|2)|z(t,1)|2dt)1/2.+a0αT0|(ρεη)(t,1)η(t,1)||z(t,1)|dt

    Here, by Eq (3.24) we note that it holds that,

    T0|˜uε(t,z)˜u(t,z)|2dtCe(T0|˜uε(t)˜u(t)|2X)1/2(T0|˜uε(t)˜u(t)|2H)1/2 for z=0,1, (3.32)

    and

    T0|˜uε(t,1)|2|z(t,1)|2dtC2eT0|˜uε(t)|X|˜uε(t)|H|z(t)|X|z(t)|HdtC2e|˜uε|L(0,T;H)|z|L(0,T;H)(T0|˜uε(t)|2X)1/2(T0|z(t)|2X)1/2.

    Since (ρεη)(t,1)η(t,1) in L2(0,T) as ε0, by the boundedness of ˜uε in L2(0,T;X), ˜uL2(0,T;X) and the strong convergence in L2(0,T;H) we see that

    T0a0s(t)((σ(˜uεi(t,1))2ασ((ρεiη)(t,1))s(t))z(t,1)dtT0a0s(t)((σ(˜u(t,1))2ασ(η(t,1))s(t))z(t,1)dt as i.

    What concerns the forth term on the left-hand side of Eq (3.31), by Eq (3.32) it follows that,

    T01s(t)β(b(t)γ˜uεi(t,0))z(t,0)dtT01s(t)β(b(t)γ˜u(t,0))z(t,0)dt as i.

    Therefore, by the limiting process i in Eq (3.31) we see that the limit function ˜u is a solution of (AP2)(˜u0,s,η,b) on [0,T]. Also, the solution ˜u is unique. Indeed, let ˜u1 and ˜u2 be a solution of (AP2)(˜u0,s,η,b) on [0,T] and put ˜u=˜u1˜u2. Then, (S''2) implies that

    ˜ut,zX+1s2(10˜uyzydy)+a0s((σ(˜u1(,1))2ασ(η(,1))s)((σ(˜u2(,1))2ασ(η(,1))s))z(1)+1sβγ(˜u1(,0)˜u2(,0))z(0)=0 for zX a.e. on [0,T]. (3.33)

    We take z=˜u in Eq (3.33). Then, by the monotonicity of σ we have

    12ddt|˜u(t)|2H+1s2(t)10|˜uy(t)|2dy0 for a.e. t[0,T]. (3.34)

    By Eq (3.34), we have the uniqueness of a solution ˜u to (AP2)(˜u0,s,η,b) on [0,T].

    Lemma 3.6. Let T>0 and sW1,(0,T) with s>0 on [0,T]. Then, (AP1)(˜u0,s,b) has a unique solution ˜u on [0,T].

    Proof. From Lemma 3.5, we see that (AP2)(˜u0,s,η,b) has a solution ˜u on [0,T] such that ˜uW1,2(0,T;X)L(0,T;H)L2(0,T;X). Define a solution operator δT(η)=˜u, where ˜u is a unique solution of (AP2)(˜u0,s,η,b) for given ηV(T). Let put δT(ηi)=˜ui for i=1,2 and η=η1η2 and ˜u=˜u1˜u2. Then, by (S''2) it holds that

    12ddt|˜u(t)|2H+1s210|˜uy(t)|2dy+a0s(t)((σ(˜u1(t,1))2ασ(η1(t,1))s(t))((σ(˜u2(t,1))2ασ(η2(t,1))s(t)))˜u(t,1)+βγs(t)|˜u(t,0)|2=10yst(t)s(t)ηy(t)˜u(t)dy for a.e. t[0,T]. (3.35)

    For the third term I3 in the left-hand side of Eq (3.35), by the monotonicity of σ we have that I3a0α|η(t,1)||˜u(t,1)|. From Eq (3.24) it holds that,

    |η(t,1)||˜u(t,1)|C1/2e|η(t,1)|(|˜uy(t)|1/2H|˜u(t)|1/2H+|˜u(t)|H)

    and

    t0|η(τ,1)||˜u(τ,1)|dτC1/2e(|˜u|1/2L(0,T1;H)t0|η(t,1)||˜uy(τ)|1/2Hdτ+|˜u|L(0,T1;H)t0|η(t,1)|dτ)C1/2e(|˜u|1/2L(0,T1;H)T1/41(t0|˜uy(τ)|2H)1/4+|˜u|L(0,T1;H)T1/21)(t0|η(t,1)|2)1/2. (3.36)

    Let T1(0,T] and we integrate Eq (3.35) over [0,t] for any t[0,T1]. Then, by Eq (3.36) we obtain,

    δ(|˜u(t)|2H+t0|˜uy(τ)|2Hdτ)a0αC1/2eT1/41(1+T1/41)(t0|η(τ,1)|2dτ)1/2|˜u|V(T1)+|st|L(0,T1)aT1/21(t0|ηy(τ)|2Hdτ)1/2|˜u|V(T1) (3.37)

    and δ=min{1/2,1/l2}, where l=max0tT|s(t)|. Finally, by Eq (3.37) we have

    δ|˜u|V(T1)[a0αC1/2eT1/41(1+T1/41)+|st|L(0,T1)aT1/21]|η|V(T1). (3.38)

    From Eq (3.38) we see that there exists T1T such that δT1 is a contraction on V(T1). Hence, Banach's fixed point theorem guarantees that there exists ˜uV(T1) such that δT1(˜u)=˜u. Thus, we see that (AP1)(˜u0,s,b) has a solution ˜u on [0,T1]. Here, T1 is independent of the choice of the initial data. Therefore, by repeating the local existence argument, we have a unique solution ˜u of (AP1)(˜u0,s,b) on the whole interval [0,T]. Thus, we see that Lemma 3.6 holds.

    Here, for given sW1,(0,T) with s>0 on [0,T] we show that a solution ˜u of (AP1)(˜u0,s,b) is non-negative and bounded on Q(T).

    Lemma 3.7. Let T>0, sW1,(0,T) with s>0 on [0,T] and ˜u be a solution of (AP1)(˜u0,s,b) on [0,T]. Then, it holds that

    0˜u(t)u(T):=max{αl(T),bγ}on[0,1]fort[0,T],

    where l(T)=max0tT|s(t)|.

    Proof. By (S'2), we note that it holds that,

    ˜ut,zX+101s2˜uyzydy+a0sσ(˜u(,1))(σ(˜u(,1))αs)z(1)1sβ(b()γ˜u(,0))z(0)=10ysts˜uyzdy for zX a.e. on [0,T]. (3.39)

    First, we prove that ˜u(t)0 on [0,1] for t[0,T]. By taking z=[˜u]+ in Eq (3.39) we have,

    12ddt|[˜u(t)]+|2H+1s2(t)10|[˜u(t)]+y|2dya0s(t)σ(˜u(,1))(σ(˜u(,1))αs(t))[˜u(t,1)]++1s(t)β(b(t)γ˜u(,0))[˜u(t,0)]+=10yst(t)s(t)˜uy(t)[˜u(t)]+dy a.e. on [0,T]. (3.40)

    Here, the third term in the left-hand side of Eq (3.40) is equal to 0 and the forth term in the left-hand side of Eq (3.40) is non-negative. Also, we obtain that

    10yst(t)s(t)[˜u(t)]+y[˜u(t)]+dy12s2(t)|[˜u(t)]+y|2H+(st(t))22|[˜u(t)]+|2H,

    Then, we have

    12ddt|[˜u(t)]+|2H+12s2(t)10|[˜u(t)]+y|2dy(st(t))22|[˜u(t)]+|2H for a.e. t[0,T].

    Therefore, by Gronwall's inequality and the assumption that ˜u00 on [0,1], we conclude that ˜u(t)0 on [0,1] for t[0,T].

    Next, we show that a solution ˜u of (AP1)(˜u0,s,b) has a upper bound u(T). Put U(t,y)=[˜u(t,y)u(T)]+ for y[0,1] and t[0,T]. Then, it holds that

    12ddt|U(t)|2H+1s2(t)10|Uy(t)|2dy+a0s(t)σ(˜u(,1))(σ(˜u(,1))αs(t))U(t,1)1s(t)β(b(t)γ˜u(,0))U(t,0)=10yst(t)s(t)˜uy(t)U(t)dy for a.e. t[0,T]. (3.41)

    Here, by ˜u(t)0 on [0,1] for t[0,T] we note that a0σ(˜u(,1))(σ(˜u(t,1))αs(t))=a0˜u(t,1)(˜u(t,1)αs(t)). Then, by u(T)αl(T)αs(t) for t[0,T], it holds that

    a0s(t)˜u(t,1)(˜u(t,1)αs(t))U(t,1)a0s(t)˜u(t,1)(u(T)αs(t))U(t,1)0.

    Also, by Eq (1.3) and bb, we observe that,

    1s(t)β(b(t)γ˜u(t,0))U(t,0)=1s(t)β(γ˜u(t,0)b+bb(t))U(t,0)βγs(t)|U(t,0)|2+βs(t)(bb(t))U(t,0)0. (3.42)

    By applying the above two results to Eq (3.41) we obtain that,

    12ddt10|U(t)|2dy+12s2(t)10|Uy(t)|2dy(st(t))22|U(t)|2H for a.e. t[0,T].

    This result and the assumption that ˜u0b/γ on [0,1] implies that ˜u(t)u(T) on [0,1] for t[0,T]. Thus, Lemma 3.7 is proven.

    At the end of this section, we relax the condition sW1,(0,T), namely, for given sW1,2(0,T) with s>0 on [0,T], we construct a solution to (AP1)(˜u0,s,b).

    Lemma 3.8. Let T>0 and sW1,2(0,T) with s>0 on [0,T]. If (A1)–(A3) hold, then, (AP1)(˜u0,s,b) has a unique solution ˜u on [0,T].

    Proof. For given sW1,2(0,T) with s>0 on [0,T], we choose a sequence {sn}W1,(0,T) and l, a>0 satisfying asnl on [0,T] for each nN, sns in W1,2(0,T) as n. By Lemma 3.6 we can take a sequence {˜un} of solutions to (AP1)(˜u0,sn,b) on [0,T]. Let zX. Then, it holds that

    ˜unt,zX+1s2n(10˜unyzydy)+a0snσ(˜un(,1))(σ(˜un(,1))αsn)z(1)1snβ(bγ˜un(,0))z(0)=10ysntsn˜unyzdy a.e. on [0,T]. (3.43)

    We take z=˜un in Eq (3.43). Then, similarly to the proof of Lemma 3.4 we derive

    12ddt|˜un(t)|2H+14s2n(t)10|˜uny(τ)|2dy(|snt(t)|2+(βbCe)24+βbCe2a)|˜un(t)|2H+βb2a+(s2n(t)(a0αCe)2+a0αCe)|˜un(t)|2H for a.e. t[0,T].

    From this, we infer that {˜un} is bounded in L(0,T;H)L2(0,T;X). Also, by referring to the derivation of Eq (3.30) we obtain from Eq (3.43) that

    |10˜unt(t)zdy|1a2|˜uny(t)|H|zy|H+a0a|˜un(t,1)|2|z(1)|+a0α|˜un(t,1)||z(1)|+βba|z(0)|+βγa|˜un(t,0)||z(0)|+|snt(t)|a(|˜un(t,1)||z(1)|+|˜un(t)|H(|zy|H+|z|H)).

    Here, we note that

    10ysnt(t)sn(t)˜uny(t)zdy=snt(t)sn(t)(˜un(t,1)z(1)10˜un(t)(yzy+z)dy).

    Here, by Lemma 3.7, we note that 0˜unu on Q(T) for each nN, where u=max{αl,bγ}. Using this result and the boundedness of {˜un} in V(T), we see that {˜unt} is bounded in L2(0,T;X). Therefore, we take a subsequence {nj}{n} such that for some ˜uW1,2(0,T;X)V(T), ˜unj˜u strongly in L2(0,T;H), weakly in W1,2(0,T;X)L2(0,T;X), weakly-* in L(0,T;H), and ˜unj(,x)˜u(,x) in L2(0,T) at x=0, 1 as j.

    Now, we consider the limiting process j in the following way:

    T0˜unt(t),z(t)Xdt+T01s2n(t)(10˜uny(t)zy(t)dy)dt+T0a0sn(t)σ(˜un(t,1))(σ(˜un(t,1))αsn(t))z(t,1)dtT01sn(t)β(b(t)γ˜un(t,0))z(t,0)dt=T010ysnt(t)sn(t)˜uny(t)z(t)dydt for zV(T). (3.44)

    Note that by snjs in W1,2(0,T) as j, it holds that snjs in C([0,T]) as j. From the convergence of ˜unj and snj, Eq (3.32) and asnjl on [0,T], it is clear that

    T0˜unjt(t),z(t)XdtT0˜ut(t),z(t)Xdt,T0a0snj(t)σ(˜unj(t,1))(σ(˜unj(t,1))αsnj(t))z(t,1)dtT0a0s(t)σ(˜u(t,1))(σ(˜u(t,1))αs(t))z(t,1)dt

    and

    T01snj(t)β(b(t)γ˜unj(t,0))z(t,0)dtT01s(t)β(b(t)γ˜u(t,0))z(t,0)dt

    as j. For the second term in the left-hand side of Eq (3.44), it follows that,

    |T01s2nj(t)(10˜unjy(t)zy(t)dy)dtT01s2(t)(10˜uy(t)zy(t)dy)dt|T0|1s2nj(t)1s2(t)||˜unjy(t)|H|zy(t)|Hdt+|T01s2(t)(10(˜unjy(t)˜uy(t))zy(t)dy)dt|2la2|snjs|C([0,T])|˜unjy|L2(0,T;H)|zy|L2(0,T;H)+|T0(˜unjy(t)˜uy(t),1s2(t)zy(t))Hdt|

    Hence, we observe that

    T01s2nj(t)(10˜unjy(t)zy(t)dy)dtT01s2(t)(10˜uy(t)zy(t)dy)dt as j.

    Also, the right-hand side of Eq (3.44) is as follows:

    |T010ysnjt(t)snj(t)˜unjy(t)z(t)dydtT010yst(t)s(t)˜uy(t)z(t)dydt|T0|snjt(t)snj(t)st(t)s(t)||˜unjy(t)|H|z(t)|Hdt+|T0(˜unjy(t)˜uy(t),yst(t)s(t)z(t))Hdt|1a|snjtst|L2(0,T)|˜unjy|L2(0,T;H)|z|L(0,T;H)+1a2|snjs|C([0,T])|st|L2(0,T)|˜unjy|L2(0,T;H)|z|L(0,T;H)+|T0(˜unjy(t)˜uy(t),yst(t)s(t)z(t))Hdt|

    From this, we have that

    T010ysnjt(t)snj(t)˜unjy(t)z(t)dydtT010yst(t)s(t)˜uy(t)z(t)dydt as j.

    Finally, by letting j we see that ˜u is a solution of (AP)(˜u0,s,b) on [0,T]. Uniqueness is proved by the same argument of the proof of Lemma 3.5.

    In this section, using the results obtained in Section 3, we establish the existence of a locally-in-time solution (PC)(˜u0,s0,b). Throughout of this section, we assume (A1)–(A3). First, for T>0, l>0 and a>0 such that a<s0<l we set

    M(T,a,l):={sW1,2(0,T)|asl on [0,T],s(0)=s0}.

    Also, for given sM(T,a,l), we define two solution mappings as follows: Ψ:M(T,a,l)W1,2(0,T;X)L(0,T;H)L2(0,T;X) by Ψ(s)=˜u, where ˜u is a unique solution of (AP1)(˜u0,s,b) on [0,T] and ΓT:M(T,a,l)W1,2(0,T) by ΓT(s)=s0+t0a0(σ(Ψ(s)(τ,1))αs(τ))dτ for t[0,T]. Moreover, for any K>0 we put

    MK(T):={sM(T,a,l)| |s|W1,2(0,T)K}.

    Now, we show that for some T>0, ΓT is a contraction mapping on the closed set of MK(T) for any K>0.

    Lemma 4.1. Let K>0. Then, there exists a positive constant TT such that the mapping ΓT is a contraction on the closed set MK(T) in W1,2(0,T).

    Proof. For T>0, a>0 and l>0 such that a<s0<l, let sM(T,a,l) and ˜u=Ψ(s). First, from the proof of Lemma 3.8 we note that it holds

    |Ψ(s)|W1,2(0,T;X)+|Ψ(s)|L(0,T;H)+|Ψ(s)|L2(0,T;X)C for sMK(T), (4.1)

    where C=C(T,˜u0,K,l,b,β,s0) is a positive constant depending on T, ˜u0, K, l, b, β and s0.

    First we show that there exists T0T such that ΓT0:MK(T0)MK(T0) is well-defined. Let K>0 and sMK(T). By the definition of σ and Ψ(s)=˜u is a solution of (AP1)(˜u0,s,b), we observe that

    ΓT(s)(t)=s0+t0a0(σ(Ψ(s)(τ,1))αs(τ))dτs0a0αlt for t[0,T]. (4.2)

    Also, by Eq (3.24) and Eq (4.1), it holds that

    ΓT(s)(t)s0+a0T1/2(t0|˜u(τ,1)|2dτ)1/2s0+a0T1/2(Ce|˜u|L(0,t;H)t0|˜u(τ)|X)1/2s0+a0T1/2(T1/4C1/2eC),t0|ΓT(s)(τ)|2dτ2s20T+4a20Tt0(|˜u(τ,1)|2+(αs(t))2)dτ2s20T+4a20T(Ce|˜u|L(0,t;H)t0|˜u(τ)|X+(αl)2T)2s20T+4a20T(T1/2CeC2+(αl)2T), (4.3)

    and

    t0|ΓT(s)(τ)|2dτa20t0|σ(Ψ(s)(τ,1))αs(τ)|2dτ2a20(T1/2CeC2+(αl)2T), (4.4)

    where C is the same positive constant as in (4.1). Therefore, by Eqs (4.2)–(4.4) we see that there exists T0T such that ΓT0(s)MK(T0).

    Next, for s1 and s2MK(T0), let ˜u1=Ψ(s1) and ˜u2=Ψ(s2) and set ˜u=˜u1˜u2, s=s1s2. Then, it holds that

    ˜ut,zX+10(1s21˜u1y1s22˜u2y)zydy+a0(1s1σ(˜u1(,1))(σ(˜u1(,1))αs1)1s2σ(˜u2(,1))(σ(˜u2(,1))αs2))z(1)(1s1β(b(t)γ˜u1(,0))1s2β(b(t)γ˜u2(,0)))z(0)=10(ys1ts1˜u1yys2ts2˜u2y)zdy for zX a.e. on [0,T]. (4.5)

    By taking z=˜u in Eq (4.5) we have

    12ddt|˜u|2H+10(1s21˜u1y1s22˜u2y)˜uydy+a0(1s1σ(˜u1(,1))(σ(˜u1(,1))αs1))1s2σ(˜u2(,1))(σ(˜u2(,1))αs2))˜u(,1)(1s1β(b(t)γ˜u1(,0))1s2β(b(t)γ˜u2(,0))˜u(,0)=10(ys1ts1˜u1yys2ts2˜u2y)˜udy a.e. on [0,T]. (4.6)

    For the second term of the left-hand side of Eq (4.6), we observe that

    10(1s21(t)˜u1y(t)1s22(t)˜u2y(t))˜uy(t)dy=1s21(t)|˜uy(t)|2H+10(1s21(t)1s22(t))˜u2y(t)˜uy(t)dy1s21(t)|˜uy(t)|2H2l|s(t)|a3s1(t)|˜u2y(t)|H|˜uy(t)|H(1η2)1s21(t)|˜uy(t)|2H12η(2la3)2|s(t)|2|˜u2y(t)|2H, (4.7)

    where η is arbitrary positive number. Next, the third term in the left-hand side of Eq (4.6) is as follows:

    a0(σ(˜u1(t,1))(σ(˜u1(t,1))αs1(t))s1(t)σ(˜u2(t,1))(σ(˜u2(t,1))αs2(t))s2(t))˜u(t,1)=a0[1s1(t)(σ(˜u1(t,1))(σ(˜u1(t,1))αs1(t))σ(˜u2(t,1))(σ(˜u2(t,1))αs2(t)))+(1s1(t)1s2(t))σ(˜u2(t,1))(σ(˜u2(t,1))αs2(t))]˜u(t,1)=a0[1s1(t)(σ(˜u1(t,1))σ(˜u2(t,1)))(σ(˜u1(t,1))αs1(t))]˜u(t,1)+σ(˜u2(t,1))s1(t)(σ(˜u1(t,1))αs1(t)(σ(˜u2(t,1))αs2(t)))˜u(t,1)+(1s1(t)1s2(t))σ(˜u2(t,1))(σ(˜u2(t,1))αs2(t))˜u(t,1):=I1+I2+I3.

    By the monotonicity of σ(r) and Eq (3.24) we have that

    I1a0s1(t)(|˜u1(t,1)|+αl)|Ce|˜u(t)|X|˜u(t)|H (4.8)

    and

    I2=a0s1(t)σ(˜u2(t,1))(σ(˜u1(t,1))αs1(t)(σ(˜u2(t,1))αs2(t)))˜u(t,1)a0|˜u2(t,1)|s1(t)α|s(t)||˜u(t,1)|Ce(a0α˜u2(t,1))22s21(t)|˜u(t)|X|˜u(t)|H12|s(t)|2. (4.9)

    Also, using the fact that σ(r)|r| for rR and Eq (3.24), we have the following estimate:

    |I3|(|s(t)|s1(t)s2(t))a0σ(˜u2(t,1))(σ(˜u2(t,1))+αl)|˜u(t,1)|Ce(a0˜u2(t,1))22a2s21(t)|˜u(t)|X|˜u(t)|H+12˜u22(t,1)|s(t)|2+Ce(a0αl˜u2(t,1))22a2s21(t)|˜u(t)|X|˜u(t)|H+12|s(t)|2. (4.10)

    Here, we put L(1)s1(t)=a0(|˜u1(t,1)|+αl)Ce and L(1)s2(t)=Ce(a0α˜u2(t,1))2/2+Ce(a0˜u2(t,1))2/2a2+Ce(a0αl˜u2(t,1))2/2a2. Next, by using Eq (3.24) and (A3), we have

    β(1s1(t)1s2(t))b(t)˜u(t,0)+βγ(1s1(t)˜u1(t,0)1s2(t)˜u2(t,0))˜u(t,0)=β(1s1(t)1s2(t))b(t)˜u(t,0)+βγ1s1(t)|˜u(t,0)|2+βγ(1s1(t)1s2(t))˜u2(t,0)˜u(t,0)((βb)2Ce2a2s21(t)+(βγ|˜u2(t,0)|)2Ce2a2s21(t))|˜u(t)|X|˜u(t)|H|s(t)|2 for t[0,T0]. (4.11)

    For the right-hand side of Eq (4.6), we separate as follows:

    10(ys1t(t)s1(t)˜u1y(t)ys2t(t)s2(t)˜u2y(t))˜u(t)dy=10ys1t(t)s1(t)˜uy(t)˜u(t)dy+10yst(t)s1(t)˜u2y(t)˜u(t)dy+10(1s1(t)1s2(t))ys2t(t)˜u2y(t)˜u(t)dy:=I4+I5+I6.

    Then, the three terms are estimated in the following way:

    I4η2s21(t)|˜uy(t)|2H+12η|s1t(t)|2|˜u(t)|2H,I512a(|st(t)|2+|˜u2y(t)|2H|˜u(t)|2H),I612a2(|s(t)|2|˜u2y(t)|2H+|s2t(t)|2|˜u(t)|2H).

    From Eq (4.6) and all estimates we derive the following inequality:

    12ddt|˜u(t)|2H+(1η)1s21(t)|˜uy(t)|2H1s1(t)L(1)s1(t)|˜u(t)|X|˜u(t)|H+1s21(t)(L(1)s2(t)+(βγ|˜u2(t,0)|)2Ce2a2+(βb)2Ce2a2)|˜u(t)|X|˜u(t)|H+(12η|s1t(t)|2+12a|˜u2y(t)|2H+12a2|s2t(t)|2)|˜u(t)|2H+(12a2|˜u2y(t)|2H+12η(2la3)2|˜u2y(t)|2H+12˜u22(t,1)+2)|s(t)|2+12a|st(t)|2. (4.12)

    We put C5(t)=((βγ|˜u2(t,0)|)2Ce)/2a2+((βb)2Ce)/2a2. By Young's inequality we have that

    1s1(t)L(1)s1(t)|˜u(t)|X|˜u(t)|H1s1(t)L(1)s1(t)(|˜uy(t)|H|˜u(t)|H+|˜u(t)|2H)η2s21(t)|˜uy(t)|2H+((L(1)s1(t))22η+1aL(1)s1(t))|˜u(t)|2H,

    and

    (L(1)s2(t)+C5(t))1s21(t)|˜u(t)|X|˜u(t)|H(L(1)s2(t)+C5(t))1s21(t)(|˜uy(t)|H|˜u(t)|H+|˜u(t)|2H)1s21(t)η2|˜uy(t)|2H+1a2((L(1)s2(t)+C5(t))22η+(L(1)s2(t)+C5(t)))|˜u(t)|2H.

    Hence, by applying these results to Eq (4.12) and taking a suitable η=η0, we obtain

    12ddt|˜u(t)|2H+121s21(t)|˜uy(t)|2H((L(1)s1(t))22η0+1aL(1)s1(t))|˜u(t)|2H+1a2((L(1)s2(t)+C5(t))22η0+(L(1)s2(t)+C5(t)))|˜u(t)|2H.+(12η0|s1t(t)|2+12a|˜u2y(t)|2H+12a2|s2t(t)|2)|˜u(t)|2H+(12a2|˜u2y(t)|2H+12η0(2la3)2|˜u2y(t)|2H+12˜u22(t,1)+2)|s(t)|2+12a|st(t)|2. (4.13)

    Now, we put the summation of all coefficients of |˜u(t)|2H by L(2)s(t) for t[0,T0] and L(3)s2(t)=|˜u2y(t)|2H/2a2+(4l2|˜u2y(t)|2H)/2η0a6+˜u22(t,1)/2+2. Then, we have

    12ddt|˜u(t)|2H+121s21(t)|˜uy(τ)|2HL(2)s(t)|˜u(t)|2H+L(3)s2(t)|s(t)|2+12a|st(t)|2 for t[0,T0]. (4.14)

    By Eq (3.24) and Eq (4.1) we note that ˜u2i(,1),˜u4i(,1)L1(0,T0) for i=1,2. From this and the fact that siMK(T0) for i=1,2, we see that L(2)sL1(0,T0) and L(3)s2L1(0,T0). Also, it holds that

    L(3)s2(t)|s(t)|2L(3)s2(t)T0|st|2L2(0,t) for t[0,T0].

    Therefore, Gronwall's inequality guarantees that

    12|˜u(t)|2H+121l2t0|˜uy(τ)|2Hdτ[(2|L(3)s2|L1(0,T0)T0+1a)|st|2L2(0,t)]e2t0L(2)s(τ)dτ for t[0,T0]. (4.15)

    By using Eq (4.15) we show that there exists TT0 such that ΓT is a contraction mapping on the closed subset of MK(T). To do so, from the subtraction of the time derivatives of ΓT0(s1) and ΓT0(s2) and relying on Eq (3.24) and Eq (4.15), we have for T1T0 the following estimate:

    |(ΓT1(s1))t(ΓT1(s2))t|L2(0,T1)a0(|σ(˜u1(,1))σ(˜u2(,1))|L2(0,T1)+α|s1(t)s2(t)|L2(0,T1))a0Ce(T10(|˜uy(t)|H|˜u(t)|H+|˜u(t)|2H)dt)1/2+a0αT1|s|W1,2(0,T1)a0Ce(|˜u|12L(0,T1;H)(T10|˜uy(t)|Hdt)12+T1|˜u|L(0,T1;H))+a0αT1|s|W1,2(0,T1). (4.16)

    Using Eq (4.15) and Eq (4.16), we obtain

    |ΓT1(s1)ΓT1(s2)|L2(0,T1)T1C6(T141|s|W1,2(0,T1)+T1|s|W1,2(0,T1)+T1|s|W1,2(0,T1)), (4.17)

    where C6 is a positive constant obtained by Eq (4.15). Therefore, by Eq (4.16) and Eq (4.17) we see that there exists TT0 such that ΓT is a contraction mapping on a closed subset of MK(T).

    From Lemma 4.1, by applying Banach's fixed point theorem, there exists sMK(T), where T is the same as in Lemma 4.1 such that ΓT(s)=s. This implies that (PC)(˜u0,s0,b) has a unique solution (s,˜u) on [0,T].

    At the end of this section, we show the boundedness of a solution ˜u to (PC)(˜u0,s0,b) which completes Theorem 2.2. By (S2) it holds that

    ˜ut,zX+101s2˜uyzydy+1sσ(˜u(,1))stz(1)1sβ(b()γ˜u(,0))z(0)=10ysts˜uyzdy for zX a.e. on [0,T]. (4.18)

    First, we note that the solution ˜u of (PC)(˜u0,s0,b) is non-negative.

    Lemma 4.2. Let T>0 and (s,˜u) be a solution of (PC)(˜u0,s0,b) on [0,T]. Then, ˜u(t)0 on [0,1] for t[0,T].

    Proof. Lemma 4.2 is proved by taking z=[˜u]+ in Eq (4.18) and using the argument of the proof of Lemma 3.7. Here, by st(t)=a0(σ(˜u(t,1))αs(t)) we note that it holds that

    10yst(t)s(t)[˜u(t)]+y[˜u(t)]+dy=st(t)s(t)1012(ddy(y|[˜u(t)]+|2)|[˜u(t)]+|2)=st(t)2s(t)|[˜u(t,1)]+|2st(t)2s(t)|[˜u(t)]+|2H=a0(σ(˜u(t,1))αs(t))2s(t)|[˜u(t,1)]+|2a0(σ(˜u(t,1))αs(t))2s(t)|[˜u(t)]+|2Ha0α2|[˜u(t)]+|2H.

    From this, we derive

    12ddt|[˜u(t)]+|2H+12s(t)10|[˜u(t)]+y|2dya0α2|[˜u(t)]+|2H for a.e. t[0,T].

    This implies that ˜u(t)0 on [0,1] for t[0,T].

    Next, we show the boundedness of the solution (s,˜u) of (PC)(˜u0,s0,b).

    Lemma 4.3. Let T>0 and (s,˜u) be a solution of (PC)(˜u0,s0,b) on [0,T]. Then, it holds that

    (i)s(t)Mfort[0,T],(ii)0˜u(t)u:=max{αM,bγ}on[0,1]fort[0,T],

    where M is a positive constant which depends on β, γ, α, b, |˜u0|H, s0, |bt|L2(0,T) and |bt|L1(0,T).

    Proof. First, we prove (i). By taking z=s(˜ubγ) in Eq (4.18) it holds that:

    s(t)2ddt|˜u(t)b(t)γ|2H+(bt(t)γ,s(t)(˜u(t)b(t)γ))H+1s(t)10|˜uy(t)|2dy+σ(˜u(t,1))st(t)(˜u(t,1)b(t)γ)β(b(t)γ˜u(t,0))(˜u(t,0)b(t)γ)=10yst(t)˜uy(t)(˜u(t)b(t)γ)dy for a.e. t[0,T]. (4.19)

    The second term of the left-hand side of Eq (4.19) is follows:

    |(bt(t)γ,s(t)(˜u(t)b(t)γ))H|=|bt(t)γs(t)10(˜u(t,y)˜u(t,0)+˜u(t,0)b(t)γ)dy|1γ|bt(t)||s(t)|(|˜uy(t)|H+|˜u(t,0)b(t)γ|). (4.20)

    Also, by Lemma 4.2 we note that σ(˜u(t,1))st(t)=˜u(t,1)st(t). Moreover, we observe that

    10yst(t)˜uy(t)(˜u(t)b(t)γ)dy=st(t)1012(y(y(˜u(t)b(t)γ)2)(˜u(t)b(t)γ)2)dy=st(t)2(˜u(t,1)b(t)γ)2st(t)210(˜u(t)b(t)γ)2dy. (4.21)

    Hence, by Eq (4.20) and Eq (4.21) we have

    12ddt(s(t)|˜u(t)b(t)γ|2H)+1s(t)10|˜uy(t)|2dy+˜u(t,1)st(t)(˜u(t,1)b(t)γ)st(t)2(˜u(t,1)b(t)γ)2+βγ|˜u(t,0)b(t)γ|21γ|bt(t)||s(t)|(|˜uy(t)|H+|˜u(t,0)b(t)γ|) for a.e. t[0,T]. (4.22)

    Here, the third and forth terms of the left-hand side of Eq (4.22) is as follows:

    ˜u(t,1)st(t)(˜u(t,1)b(t)γ)st(t)2(˜u(t,1)b(t)γ)2=st(t)˜u2(t,1)st(t)˜u(t,1)b(t)γst(t)2(˜u2(t,1)2˜u(t,1)b(t)γ+(b(t)γ)2)=st(t)2˜u2(t,1)st(t)2(b(t)γ)2. (4.23)

    Since st(t)=a0(˜u(t,1)αs(t)), it holds that

    st(t)2˜u2(t,1)=˜u(t,1)2(|st(t)|2a0+αs(t)st(t))=12a0˜u(t,1)|st(t)|2+αs(t)2(|st(t)|2a0+αs(t)st(t))=12a0˜u(t,1)|st(t)|2+αs(t)2|st(t)|2a0+α26ddts3(t). (4.24)

    Note that the first and second terms in the last of Eq (4.24) are non-negative. Accordingly, by Eq (4.23) and Eq (4.24) we have that

    ˜u(t,1)st(t)(˜u(t,1)b(t)γ)st(t)2(˜u(t,1)b(t)γ)2α26ddts3(t)st(t)2(b(t)γ)2=α26ddts3(t)12γ2(ddt(s(t)b2(t))2s(t)b(t)bt(t)). (4.25)

    Also, for the right-hand side of Eq (4.22) we have

    1γ|bt(t)||s(t)|(|˜uy(t)|H+|˜u(t,0)b(t)γ|)12s(t)|˜uy(t)|2H+12s3(t)(|bt(t)|γ)2+βγ2|˜u(t,0)b(t)γ|2+12βγ|s(t)|2(|bt(t)|γ)2. (4.26)

    By combining Eq (4.25) and Eq (4.26) with Eq (4.22) we have

    12ddt(s(t)|˜u(t)b(t)γ|2H)+12s(t)10|˜uy(t)|2dy+α26ddts3(t)12γ2ddt(s(t)b2(t))+βγ2|˜u(t,0)b(t)γ|21γ2s(t)b(t)|bt(t)|+12s3(t)(|bt(t)|γ)2+12βγ|s(t)|2(|bt(t)|γ)2 for a.e. t[0,T]. (4.27)

    Then, by integrating Eq (4.27) over [0,t] for t[0,T] we obtain that

    12s(t)|˜u(t)b(t)γ|2H+t012s(τ)|˜uy(τ)|2Hdτ+α26s3(t)12γ2(s(t)b2(t))12s0|˜u0b(0)γ|2H+α26s30+12γ2t0s3(τ)|bt(τ)|2dτ+bγ2t0s(τ)|bt(τ)|dτ+12βγ3t0|s(τ)|2|bt(τ)|2dτ for t[0,T]. (4.28)

    Here, by Young's inequality it follows that

    α26s3(t)12γ2(s(t)b2(t))α26s3(t)(η33s3(t)+23η3/2(12γ2(b)2)3/2), (4.29)

    where η is an arbitrary positive number. Therefore, by Eq (4.28) and Eq (4.29) we obtain

    (α26η33)s3(t)23η3/2(12γ2(b)2)3/2+12s0|˜u0b(0)γ|2H+α26s30+12γ2t0s3(τ)|bt(τ)|2dτ+bγ2t0s(τ)|bt(τ)|dτ+12βγ3t0|s(τ)|2|bt(τ)|2dτ for t[0,T]. (4.30)

    We put M(η)=(s0|˜u0b(0)γ|2H)/2+(α2s30)/6+(2(12γ2(b)2)3/2))/3η3/2. Then, by taking a suitable η=η0 and J1(t)=t0s3(τ)|bt(τ)|2dτ, J2(t)=t0s(τ)|bt(τ)|dτ and J3(t)=t0|s(τ)|2|bt(τ)|2dτ for t[0,T] we derive that

    α212s3(t)M(η0)+12γ2J1(t)+bγ2J2(t)+12βγ3J3(t) for t[0,T]. (4.31)

    Then, we have that

    J1(t)12M(η0)α2|bt(t)|2+6(γα)2|bt(t)|2J1(t)+12α2(bγ2J2(t)+12βγ3J3(t))|bt(t)|2 for t[0,T].

    Hence, by Gronwall's inequality we obtain that

    J1(t)[12M(η0)α2|bt|2L2(0,T)+12α2(bγ2J2(t)+12βγ3J3(t))|bt|2L2(0,T)]e6(γα)2|bt|2L2(0,T) for t[0,T]. (4.32)

    Here, we put N(T)=12α2|bt|2L2(0,T)e6(γα)2|bt|2L2(0,T). Then, by Eq (4.31) and Eq (4.32) we obtain that

    α212s3(t)M(η0)+12γ2M(η0)N(T)+bγ2(1+N(T)2γ2)J2(t)+12βγ3(1+N(T)2γ2)J3(t) for t[0,T]. (4.33)

    Now, we put l(T)=max0tT|s(t)|. Then, we have

    bγ2(1+N(T)2γ2)J2(t)bγ2(1+N(T)2γ2)l(T)|bt|L1(0,T)η33l3(T)+23η3/2(bγ2(1+N(T)2γ2)|bt|L1(0,T))3/2,

    and

    12βγ3(1+N(T)2γ2)J3(t)12βγ3(1+N(T)2γ2)l2(T)|bt|2L2(0,T)2η3/23l3(T)+13η3(12βγ3(1+N(T)2γ2)|bt|2L2(0,T))3.

    Hence, by adding these estimates to Eq (4.33) and taking a suitable η=η0 we see that there exists a positive constant M which depends on β, γ, α, b, s0, |bt|L2(0,T) and |bt|L1(0,T) such that s(t)M for t[0,T].

    Next, we show (ii). Put U(t,y)=[˜u(t,y)u]+ for y[0,1] and t[0,T], and then take z=U(t,y) in Eq (4.18). Here, by ˜u(t)0 on [0,1] for t[0,T] we note that st(t)=a0(σ(˜u(t,1))αs(t))=a0(˜u(t,1)αs(t)). Then, we observe that:

    10yst(t)s(t)˜uy(t)U(t)dy=st(t)2s(t)10(ddy(yU2(t))U2(t))dy=st(t)2s(t)|U(t,1)|2st(t)2s(t)|U(t)|2Hst(t)2s(t)|U(t,1)|2+a0α2|U(t)|2H.

    Then, by using the argument of the proof of Lemma 3.7 we derive

    12ddt|U(t)|2H+1s2(t)10|Uy(t)|2dy+1s(t)˜u(t,1)st(t)U(t,1)st(t)2s(t)|U(t,1)|2a0α2|U(t)|2H for a.e. t[0,T]. (4.34)

    Here, by uαMαs(t) for t[0,T], it holds that

    st(t)s(t)˜u(t,1)U(t,1)st(t)2s(t)|U(t,1)|2=st(t)2s(t)|U(t,1)|2+st(t)s(t)uU(t,1)a0(uαs(t))s(t)(|U(t,1)|22+uU(t,1))0.

    By applying the result to Eq (4.34) we obtain that

    12ddt10|U(t)|2dy+12s2(t)10|Uy(t)|2dya0α2|U(t)|2H for a.e. t[0,T].

    This implies that ˜u(t)u on [0,1] for t[0,T]. Thus, Lemma 4.3 is now proven.

    Combining the statements of Lemma 4.2 and Lemma 4.3, we can conclude that Theorem 2.2 holds.

    The first author is supported by Grant-in-Aid No. 20K03704, JSPS. The second author is supported by Grant-in-Aid No. 19K03572, JSPS

    The authors declare there is no conflict of interest.



    [1] T. Aiki, K. Kumazaki, A. Muntean, A free boundary problem describing migration into rubbers-quest of the large time behavior, Z. Angew. Math. Mech., 102 (2022), e202100134. https://doi.org/10.1002/zamm.202100134 doi: 10.1002/zamm.202100134
    [2] E. Brunier, G. Antonini, Experimental and numerical description of the diffuision of a liquid in a swelling elastomer, In: C. A. Brebbia, G. A. Karamidas, Eds., Computational Methods and Experimental Measurements, Berlin: Springer, 1984,623–632.
    [3] G. Chagnon, E. Verron, L. Gornet, G. Marckmann, P. Charrier, On the relevance of Continuum Damage Mechanics as applied to the Mullins effect in elastomers, J. Mech. phys. Solids, 52 (2004), 1627–1650. https://doi.org/10.1016/j.jmps.2003.12.006 doi: 10.1016/j.jmps.2003.12.006
    [4] N. Kenmochi, Solvability of nonlinear evolution equations with time-dependent constraints and applications, Bull. Fac. Education, Chiba Univ., 30 (1981), 1–87.
    [5] K. Kumazaki, A. Muntean, Local weak solvability of a moving boundary problem describing swelling along a halfline, Netw. Heterog. Media, 14 (2019), 445–496. https://doi/10.3934/nhm.2019018 doi: 10.3934/nhm.2019018
    [6] K. Kumazaki, A. Muntean, Global weak solvability, continuous dependence on data, and large time growth of swelling moving interfaces, Interfaces Free Bound., 22 (2020), 27–49. https://doi.org/10.4171/ifb/431 doi: 10.4171/ifb/431
    [7] S. Nepal, R. Meyer, N. H. Kröger, T. Aiki, A. Muntean, Y. Wondmagegne, U. Giese, A moving boundary approach of capturing diffusants penetration into rubber: FEM approximation and comparison with laboratory measurements, Kautsch. Gummi, Kunsts., 5 (2020), 61–69.
    [8] Z. Ren, A. S. Verma, Y. Li, J. J.E. Teuwen, Z. Jiang, Offshore wind turbine operations and maintenance: A state-of-the-art review, Renew. Sust. Energ. Rev., 144 (2021), 110886. https://doi.org/10.1016/j.rser.2021.110886 doi: 10.1016/j.rser.2021.110886
    [9] A. Visintin, Introduction to the models of phase transitions, Bollettino dell'Unione Matematica Italiana, 1 (1998), 1–47.
  • This article has been cited by:

    1. Kota Kumazaki, Adrian Muntean, A two-scale model describing swelling in porous materials with elongated internal structures, 2025, 0033-569X, 10.1090/qam/1705
  • Reader Comments
  • © 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1432) PDF downloads(53) Cited by(1)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog