This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Global solutions and blow-up for the wave equation with variable coefficients: II. boundary supercritical source

Tae Gab Ha Department of Mathematics, and Institute of Pure and Applied Mathematics, Jeonbuk National University, Jeonju 54896, Republic of Korea [email protected]
Abstract.

In this paper, we consider the wave equation with variable coefficients and boundary damping and supercritical source terms. The goal of this work is devoted to prove the local and global existence, and classify decay rate of energy depending on the growth near zero on the damping term. Moreover, we prove the blow-up of the weak solution with positive initial energy as well as nonpositive initial energy.

Key words and phrases:
wave equation with variable coefficients; supercritical source; existence of solutions; energy decay rates; blow-up
2020 Mathematics Subject Classification:
35L05; 35L20; 35A01; 35B40; 35B44

1. Introduction

In this paper, we are concerned with the local and global existence, energy decay rates and finite time blow-up of the solution for the following wave equation

{uttμ(t)Lu=f(x,t)inΩ×(0,+),u=0onΓ0×(0,+),μ(t)uνL+q(ut)=h(u)onΓ1×(0,+),u(x,0)=u0(x),ut(x,0)=u1(x),\begin{cases}\vspace{3mm}u_{tt}-\mu(t)Lu=f(x,t)&\hskip 14.22636pt\text{in}\hskip 14.22636pt\Omega~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u=0&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{0}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}\mu(t)\frac{\partial u}{\partial\nu_{L}}+q(u_{t})=h(u)&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{1}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u(x,0)=u_{0}(x),\hskip 14.22636ptu_{t}(x,0)=u_{1}(x),\end{cases} (1.1)

where Lu=div(A(x)u)=i,j=1nxi(aij(x)uxj)Lu=div(A(x)\nabla u)=\sum^{n}_{i,j=1}\frac{\partial}{\partial x_{i}}\bigl{(}a_{ij}(x)\frac{\partial u}{\partial x_{j}}\bigr{)}, where A(x)=(aij(x))A(x)=(a_{ij}(x)) is a symmetric and positive matrix, and uνL=i,j=1naij(x)uxjνi\frac{\partial u}{\partial\nu_{L}}=\sum^{n}_{i,j=1}a_{ij}(x)\frac{\partial u}{\partial x_{j}}\nu_{i}, where ν=(ν1,,νn)\nu=(\nu_{1},\cdots,\nu_{n}) is the outward unit normal to Γ\Gamma. Ω\Omega is a bounded domain of n\mathbb{R}^{n}(n3n\geq 3) with smooth boundary Γ=Γ0Γ1\Gamma=\Gamma_{0}\cup\Gamma_{1}. Here, Γ0\Gamma_{0} and Γ1\Gamma_{1} are closed and disjoint with meas(Γ0)0meas(\Gamma_{0})\neq 0.

During the past decades, the problem (1.1) has been widely studied. The condition h(s)s0h(s)s\leq 0 means that hh represents an attractive force. When h(s)s0h(s)s\geq 0 as in the present case, hh represents a source term. This situation is more delicate than attractive force, since the solution of (1.1) can blow up. The damping-source interplay in system (1.1) arise naturally in many contexts, for instance, in classical mechanics, fluid dynamics and quantum field theory (cf. [29, 41]). The interaction between two competitive force, that is damping term and source term, make the problem attractive from the mathematical point of view.

For the present case when hh is polynomial nonlinear source term such as h(u)=|u|γuh(u)=|u|^{\gamma}u, the stability of (1.1) has been studied by many authors (see [7, 8, 14, 15, 16, 17, 18, 20, 21, 22, 24, 25, 26, 27, 28, 30, 31, 36, 37, 38, 39, 43] and a list on references therein), where h(u)h(u) is subcritical or critical source. However, very few results addressed wave equations influenced by supercritical sources (cf. [1, 3, 11, 12, 42]). For example, [42] proved the local and global existence, uniqueness and Hadamard well-posedness for the wave equation when source terms can be supercritical or super-supercritical. However, the author do not considered the energy decay and blow-up of the solutions. [11] considered a system of nonlinear wave equations with supercritical sources and damping terms. They proved global existence and exponential and algebraic uniform decay rates of energy, moreover, blow-up result for weak solutions with nonnegative initial energy. But as far as I know, the only problem with considering supercritical source is the constant coefficients case, that is, A=IA=I and dimension n=3n=3.

In the case of variable coefficients, that is AIA\neq I, boundary stability of the wave equation was considered in [4, 9, 15, 18, 22]. The wave equations with variable coefficients arise in mathematical modeling of inhomogeneous media in solid mechanics, electromagnetic, fluid flows through porous media. For the variable coefficients problem, the main tool is Riemannian geometry method, which was introduced by [46] and has been widely used in the literature, see [6, 9, 32, 33, 34, 44, 45] and a list of references therein. However, there were very few results considered the source term. For example, [4] proved the uniform decay rate of the energy to the viscoelastic wave equation with variable coefficients and acoustic boundary conditions without damping term. Recently, [26] studied the general decay rate of the energy for the wave equation with variable coefficients and Balakrishnan-Taylor damping and source term without imposing any restrictive growth near zero on the damping term. However, above mentioned examples were not considered Riemannian geometry, and only treated a subcritical source. More recently, [19] proved the uniform energy decay rates of the wave equation with variable coefficients applying the Riemannian geometry method and modified multiplier method. But, it was considered a subcritical source. [23] studied local and global existence, energy decay rate and blow-up of solution for the wave equation with variable coefficients, however it was not considered Reimannian geometry, and was treated the interior supercritical source. There is none, to my knowledge, for the variable coefficients problem having both damping and source terms on Riemannian geometry as well as considering boundary supercritical source.

Our main motivation is constituted by three dimensional case, in which the source term can be supercritical on variable coefficient problem. The differences from previous literatures are as follows:

  1. (i)

    Supercritical source for n3n\geq 3.

  2. (ii)

    Variable coefficient problem having source term on Riemannian geometry.

  3. (iii)

    Blow-up result with positive initial energy as well as nonpositive initial energy.

In order to overcome difficulties to prove above statements, first, we refine the energy space and a constant used in potential well method, because we do not guarantee H01(Ω)Lγ+2(Γ1)H^{1}_{0}(\Omega)\hookrightarrow L^{\gamma+2}(\Gamma_{1}) since the source term is supercritical. Also we have a hypothesis on damping term for proving existence of solutions and energy decay rates (see Remark 2.2). Second, we use the Faedo-Galerkin method because nonlinear semigroup arguments considered in the previous literatures cannot be used since this paper deal with an operator μ(t)L-\mu(t)L, which depends on tt. Third, we refine the key point constants used to prove blow-up result. So, this paper has improved and generalized previous literatures.

The goal of this paper is to prove the existence result using the Faedo-Galerkin method and truncated approximation method, and classify the energy decay rate applying the method developed in [46] and [35]. Moreover, we prove the blow-up of the weak solution with positive initial energy as well as nonpositive initial energy. This paper is organized as follows: In Section 2, we recall the notation, hypotheses and some necessary preliminaries and introduce our main result. In Section 3, we prove the local existence of weak solutions, and show the global existence of weak solution in each two conditions in Section 4. In Section 5, we prove the uniform decay rate under suitable conditions on the initial data and boundary damping by the differential geometric approach. In Section 6, we prove the blow-up of the weak solution with positive initial energy as well as nonpositive initial energy. by using contradiction method.

2. Preliminaries

We begin this section by introducing some notations and our main results. Throughout this paper, we define the Hilbert space H01(Ω)={uH1(Ω);u=0onΓ0}H^{1}_{0}(\Omega)=\{u\in H^{1}(\Omega);u=0~{}~{}\text{on}~{}~{}\Gamma_{0}\} with the norm uH01(Ω)=uL2(Ω)||u||_{H^{1}_{0}(\Omega)}=||\nabla u||_{L^{2}(\Omega)} and ={uH01(Ω);uLγ+2(Γ1)}\mathcal{H}=\{u\in H^{1}_{0}(\Omega);u\in L^{\gamma+2}(\Gamma_{1})\} with the norm u=uH01(Ω)+uLγ+2(Γ1)||u||_{\mathcal{H}}=||u||_{H^{1}_{0}(\Omega)}+||u||_{L^{\gamma+2}(\Gamma_{1})}. ||||p||\cdot||_{p} and ||||p,Γ||\cdot||_{p,\Gamma} are denoted by the Lp(Ω)L^{p}(\Omega) norm and the Lp(Γ)L^{p}(\Gamma) norm, respectively, and u,v=Ωu(x)v(x)𝑑x\langle u,v\rangle=\int_{\Omega}u(x)v(x)dx and u,vΓ=Γu(x)v(x)𝑑Γ\langle u,v\rangle_{\Gamma}=\int_{\Gamma}u(x)v(x)d\Gamma. Moreover, we need some notations on Riemannian geometry, it is mentioned in [46] and reference therein. For the reader’s comprehension, we will repeat them here.

Let A(x)=(aij(x))A(x)=(a_{ij}(x)) be a symmetric and positive definite matrix for all xnx\in\mathbb{R}^{n} (n3)(n\geq 3) and aij(x)a_{ij}(x) be smooth functions on n\mathbb{R}^{n} satisfying

c1i=1nξi2i,j=1naij(x)ξiξj,xn,0ξ=(ξ1,,ξn)Tn,c_{1}\sum^{n}_{i=1}\xi^{2}_{i}\leq\sum^{n}_{i,j=1}a_{ij}(x)\xi_{i}\xi_{j},\hskip 8.53581pt\forall x\in\mathbb{R}^{n},\hskip 8.53581pt0\neq\xi=(\xi_{1},\cdots,\xi_{n})^{T}\in\mathbb{R}^{n}, (2.1)

where c1c_{1} is a positive constants. Set

G(x)=(gij(x))=A1(x).G(x)=\bigl{(}g_{ij}(x)\bigr{)}=A^{-1}(x).

For each xnx\in\mathbb{R}^{n}, we define the inner product g(,)=,gg(\cdot,\cdot)=\langle\cdot,\cdot\rangle_{g} and the norm ||g|\cdot|_{g} on the tangent space xn=n\mathbb{R}^{n}_{x}=\mathbb{R}^{n} by

g(X,Y)=X,Yg=i,j=1ngij(x)αiβj,|X|g=X,Xg12,X=i=1nαixi,Y=i=1nβixixn.g(X,Y)=\langle X,Y\rangle_{g}=\sum^{n}_{i,j=1}g_{ij}(x)\alpha_{i}\beta_{j},\hskip 8.53581pt|X|_{g}=\langle X,X\rangle^{\frac{1}{2}}_{g},\hskip 8.53581pt\forall X=\sum^{n}_{i=1}\alpha_{i}\frac{\partial}{\partial x_{i}},~{}~{}Y=\sum^{n}_{i=1}\beta_{i}\frac{\partial}{\partial x_{i}}\in\mathbb{R}^{n}_{x}. (2.2)

Then (n,g)(\mathbb{R}^{n},g) is a Riemannian manifold with Riemann metric gg. gu\nabla_{g}u and DgD_{g} are denoted by the gradient of u and Levi-Civita connection in the Riemannian metric gg, respectively. It follows that

gu=i=1n(j=1naij(x)uxj)xi=A(x)u,|gu|g2=i,j=1naij(x)uxiuxj.\nabla_{g}u=\sum^{n}_{i=1}\Bigl{(}\sum^{n}_{j=1}a_{ij}(x)\frac{\partial u}{\partial x_{j}}\Bigr{)}\frac{\partial}{\partial x_{i}}=A(x)\nabla u,\hskip 8.53581pt|\nabla_{g}u|^{2}_{g}=\sum^{n}_{i,j=1}a_{ij}(x)\frac{\partial u}{\partial x_{i}}\frac{\partial u}{\partial x_{j}}. (2.3)

Let HH be a vector field on (n,g)(\mathbb{R}^{n},g). Then the covariant differential DgHD_{g}H of HH determines a bilinear form on xn×xn\mathbb{R}^{n}_{x}\times\mathbb{R}^{n}_{x}, for each xnx\in\mathbb{R}^{n}, by

DgH(X,Y)=DgXH,Yg,X,Yxn,D_{g}H(X,Y)=\langle D_{gX}H,Y\rangle_{g},\hskip 8.53581pt\forall X,Y\in\mathbb{R}^{n}_{x}, (2.4)

where DgXHD_{gX}H is the covariant derivative of the vector field HH with respect to XX.

(𝐇𝟏)\bf{(H_{1})} Hypothesis on 𝛀\bf\Omega.

Let Ωn\Omega\subset\mathbb{R}^{n} be a bounded domain, n3n\geq 3, with smooth boundary Γ=Γ0Γ1\Gamma=\Gamma_{0}\cup\Gamma_{1}. Here Γ0\Gamma_{0} and Γ1\Gamma_{1} are closed and disjoint with meas(Γ0)0meas(\Gamma_{0})\neq 0. There exists a vector field HH on the Riemannian manifold (n,g)(\mathbb{R}^{n},g) such that

DgH(X,X)σ|X|g2,Xxn,xΩ¯,D_{g}H(X,X)\geq\sigma|X|^{2}_{g},\hskip 8.53581pt\forall X\in\mathbb{R}^{n}_{x},\hskip 8.53581ptx\in\overline{\Omega}, (2.5)

where σ\sigma is a positive constant and

Hν0onΓ0andHνδ>0onΓ1,H\cdot\nu\leq 0\hskip 8.53581pt\text{on}\hskip 8.53581pt\Gamma_{0}\hskip 8.53581pt\text{and}\hskip 8.53581ptH\cdot\nu\geq\delta>0\hskip 8.53581pt\text{on}\hskip 8.53581pt\Gamma_{1}, (2.6)

where ν\nu represents the unit outward normal vector to Γ\Gamma. Moreover we assume that

μ(0)u0νL+g(u1)=h(u0)onΓ1.\mu(0)\frac{\partial u_{0}}{\partial\nu_{L}}+g(u_{1})=h(u_{0})\hskip 8.53581pt\text{on}\hskip 8.53581pt\Gamma_{1}. (2.7)

(𝐇𝟐)\bf{(H_{2})} Hypothesis on μ\bf\mu, 𝐟\bf f.

Let μW1,(0,)W1,1(0,)\mu\in W^{1,\infty}(0,\infty)\cap W^{1,1}(0,\infty) satisfying following conditions:

μ(t)μ0>0andμ(t)0a.e. in[0,),\hskip 14.22636pt\mu(t)\geq\mu_{0}>0\hskip 14.22636pt\text{and}\hskip 14.22636pt\mu^{\prime}(t)\leq 0\hskip 8.53581pt\text{a.e. in}\hskip 8.53581pt[0,\infty), (2.8)

where μ0\mu_{0} is a positive constant. We assume that

fH1(0,;L2(Ω)).f\in H^{1}(0,\infty;L^{2}(\Omega)). (2.9)

(𝐇𝟑)\bf(H_{3}) Hypothesis on 𝐪\bf q.

Let q:q:\mathbb{R}\rightarrow\mathbb{R} be a nondecreasing C1C^{1} function such that q(0)=0q(0)=0 and suppose that there exist positive constants c3c_{3}, c4c_{4}, and a strictly increasing and odd function β\beta of C1C^{1} class on [1,1][-1,1] such that

|β(s)||q(s)||β1(s)|if|s|1,|\beta(s)|\leq|q(s)|\leq|\beta^{-1}(s)|\hskip 8.53581pt\text{if}\hskip 8.53581pt|s|\leq 1, (2.10)
c3|s|ρ+1|q(s)|c4|s|ρ+1if|s|>1,c_{3}|s|^{\rho+1}\leq|q(s)|\leq c_{4}|s|^{\rho+1}\hskip 8.53581pt\text{if}\hskip 8.53581pt|s|>1, (2.11)

where β1\beta^{-1} denotes the inverse function of β\beta, and ρ2(n2)γ2n(n2)γ\rho\geq\frac{2(n-2)\gamma-2}{n-(n-2)\gamma}.

(𝐇𝟒)\bf{(H_{4})} Hypothesis on γ\bf\gamma.

Let γ\gamma be a constant satisfying the following condition:

1n2<γn1n2.\frac{1}{n-2}<\gamma\leq\frac{n-1}{n-2}. (2.12)
Lemma 2.1.

([46]) Let u,vC1(Ω¯)u,v\in C^{1}(\overline{\Omega}) and HH, XX be vector fields on (n,g)(\mathbb{R}^{n},g). Then

  1. (i)
    H(u)=gu,Hg,H(x),A(x)X(x)g=H(x)X(x).H(u)=\langle\nabla_{g}u,H\rangle_{g},\hskip 8.53581pt\langle H(x),A(x)X(x)\rangle_{g}=H(x)\cdot X(x).
  2. (ii)
    div(uH)=udiv(H)+H(u),Ω𝑑iv(H)𝑑x=ΓHν𝑑Γ.div(uH)=u~{}div(H)+H(u),\hskip 8.53581pt\int_{\Omega}div(H)~{}dx=\int_{\Gamma}H\cdot\nu~{}d\Gamma.
  3. (iii)
    Ωu𝑑iv(gv)𝑑x=ΓuvνL𝑑ΓΩgu,gvg𝑑x.\int_{\Omega}u~{}div(\nabla_{g}v)~{}dx=\int_{\Gamma}u\frac{\partial v}{\partial\nu_{L}}~{}d\Gamma-\int_{\Omega}\langle\nabla_{g}u,\nabla_{g}v\rangle_{g}~{}dx.
  4. (iv)
    gu,g(H(u))g=DgH(gu,gu)+12div(|gu|g2H)12|gu|g2div(H).\langle\nabla_{g}u,\nabla_{g}(H(u))\rangle_{g}=D_{g}H(\nabla_{g}u,\nabla_{g}u)+\frac{1}{2}div(|\nabla_{g}u|^{2}_{g}H)-\frac{1}{2}|\nabla_{g}u|^{2}_{g}div(H).
Remark 2.1.

Hypothesis (2.5) was introduced by Yao [46] for the exact controllability of the wave equation with variable coefficients. The existence of such a vector field depends on the sectional curvature of the Riemannian manifold (n,g)(\mathbb{R}^{n},g). There are several methods and examples in [46] to find out a vector field HH that is satisfied the hypothesis (2.5). Specially, if A=IA=I, the constant coefficient case, the condition (2.5) is automatically satisfied by choosing H=xx0H=x-x_{0} for any x0nx_{0}\in\mathbb{R}^{n}.

Remark 2.2.

In view of the critical Sobolev imbedding H12(Γ1)L2(n1)n2(Γ1)H^{\frac{1}{2}}(\Gamma_{1})\hookrightarrow L^{\frac{2(n-1)}{n-2}}(\Gamma_{1}), the map k(u)=|u|γuk(u)=|u|^{\gamma}u is not locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) into L2(Γ1)L^{2}(\Gamma_{1}) for the supercritical values 1n2<γn1n2\frac{1}{n-2}<\gamma\leq\frac{n-1}{n-2}. However, by the hypothesis on ρ\rho (ρ2(n2)γ2n(n2)γ)\Bigl{(}\rho\geq\frac{2(n-2)\gamma-2}{n-(n-2)\gamma}\Bigr{)}, k(u)k(u) is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) into Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}).

Remark 2.3.

For n>3n>3, if 2n2γn1n2\frac{2}{n-2}\leq\gamma\leq\frac{n-1}{n-2}, then the inequality ργ\rho\geq\gamma always holds true under the assumption ρ2(n2)γ2n(n2)γ\rho\geq\frac{2(n-2)\gamma-2}{n-(n-2)\gamma} (see Figure 1).

Refer to caption
Figure 1. The admissible range of the damping parameter ρ\rho and the exponent of the source γ\gamma.
Definition 2.1.

(Weak solution). u(x,t)u(x,t) is called a weak solution of (1.1) on Ω×(0,T)\Omega\times(0,T) if uC(0,T;)C1(0,T;L2(Ω))u\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)), ut|Γ1Lρ+2(0,T;Γ1)u_{t}|_{\Gamma_{1}}\in L^{\rho+2}(0,T;\Gamma_{1}) and satisfies (1.1) in the distribution sense, i.e.,

0TΩutϕtdxdt+Ωutϕ𝑑x|0T+0Tμ(t)Ωgu,gϕg𝑑x𝑑t+0TΓ1q(ut)ϕ𝑑Γ𝑑t0TΓ1h(u)ϕ𝑑Γ𝑑t=0TΩf(x,t)ϕ𝑑x𝑑t,\int^{T}_{0}\int_{\Omega}-u_{t}\phi_{t}dxdt+\int_{\Omega}u_{t}\phi dx\Bigr{|}^{T}_{0}+\int^{T}_{0}\mu(t)\int_{\Omega}\langle\nabla_{g}u,\nabla_{g}\phi\rangle_{g}dxdt\\ +\int^{T}_{0}\int_{\Gamma_{1}}q(u_{t})\phi d\Gamma dt-\int^{T}_{0}\int_{\Gamma_{1}}h(u)\phi d\Gamma dt=\int^{T}_{0}\int_{\Omega}f(x,t)\phi dxdt,

for any ϕC(0,T;)C1(0,T;L2(Ω))\phi\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)), ϕ|Γ1Lρ+2(0,T;Γ1)\phi|_{\Gamma_{1}}\in L^{\rho+2}(0,T;\Gamma_{1}) and u(x,0)=u0(x)u(x,0)=u_{0}(x)\in\mathcal{H}, ut(x,0)=u1(x)L2(Ω)u_{t}(x,0)=u_{1}(x)\in L^{2}(\Omega).

Remark 2.4.

One easily check that =H01(Ω)\mathcal{H}=H^{1}_{0}(\Omega) when 1n2<γ2n2\frac{1}{n-2}<\gamma\leq\frac{2}{n-2}. Moreover, if n=3n=3, then we can replace \mathcal{H} by H01(Ω)H^{1}_{0}(\Omega), since H01(Ω)Lγ+2(Γ1)H^{1}_{0}(\Omega)\hookrightarrow L^{\gamma+2}(\Gamma_{1}) (see Figure 2).

Refer to caption
Figure 2. The admissible range of parameters aa and bb with respect to the trace imbedding H01(Ω)Laγ+b(Γ1)H^{1}_{0}(\Omega)\hookrightarrow L^{a\gamma+b}(\Gamma_{1}).

The energy associated to the problem (1.1) when h(u)=|u|γuh(u)=|u|^{\gamma}u is given by

E(t)=12ut(t)22+12μ(t)|gu(t)|g221γ+2u(t)γ+2,Γ1γ+2.E(t)=\frac{1}{2}||u_{t}(t)||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

We now state our main results.

Theorem 2.1.

Suppose that (H1)(H4)(H_{1})-(H_{4}) hold and let h(u)=|u|γuh(u)=|u|^{\gamma}u. Then given the initial data (u0,u1)×L2(Ω)(u_{0},u_{1})\in\mathcal{H}\times L^{2}(\Omega), there exist T>0T>0 and a weak solution of problem (1.1). Moreover, the following energy identity holds for all 0tT0\leq t\leq T:

E(t)+0tΓ1q(us)us𝑑Γ𝑑s120tμ(s)|gu(s)|g22𝑑s0tΩf(x,s)us𝑑x𝑑s=E(0).E(t)+\int^{t}_{0}\int_{\Gamma_{1}}q(u_{s})u_{s}d\Gamma ds-\frac{1}{2}\int^{t}_{0}\mu^{\prime}(s)||~{}|\nabla_{g}u(s)|_{g}||^{2}_{2}ds-\int^{t}_{0}\int_{\Omega}f(x,s)u_{s}dxds=E(0). (2.13)

Furthermore, if one of the assumptions hold: ργ\rho\geq\gamma or

E(0)<d0and|gu0|g2<λ0E(0)<d_{0}\hskip 8.53581pt\text{and}\hskip 8.53581pt||~{}|\nabla_{g}u_{0}|_{g}||_{2}<\lambda_{0} (2.14)

with f=0f=0, where

λ0=(μ0K0γ+2)1/γandd0=γμ02(γ+2)λ02,K0=supu,u0(uγ+2,Γ1|gu|g2).\lambda_{0}=\Bigl{(}\frac{\mu_{0}}{K^{\gamma+2}_{0}}\Bigr{)}^{1/\gamma}\hskip 8.53581pt\text{and}\hskip 8.53581ptd_{0}=\frac{\gamma\mu_{0}}{2(\gamma+2)}\lambda^{2}_{0},\hskip 8.53581ptK_{0}=\sup_{u\in\mathcal{H},u\neq 0}\Bigl{(}\frac{||u||_{\gamma+2,\Gamma_{1}}}{||~{}|\nabla_{g}u|_{g}||_{2}}\Bigr{)}.

Then the solution u(x,t)u(x,t) of (1.1) is global.

Theorem 2.2.

Suppose that the hypotheses in Theorem 2.1 and (2.14) with ργ\rho\leq\gamma and f=0f=0 hold. Then we have following energy decay rates:

  1. (i)

    Case 1 : β\beta is linear. Then we have

    E(t)C1eωt,E(t)\leq C_{1}e^{-\omega t},

    where ω\omega is a positive constant.

  2. (ii)

    Case 2 : β\beta has polynomial growth near zero, that is, β(s)=sρ+1\beta(s)=s^{\rho+1}. Then we have

    E(t)C2(1+t)2ρ.E(t)\leq\frac{C_{2}}{(1+t)^{\frac{2}{\rho}}}.
  3. (iii)

    Case 3 : β\beta does not necessarily have polynomial growth near zero. Then we have

    E(t)C3(F1(1t))2,E(t)\leq C_{3}\Bigl{(}F^{-1}\Bigl{(}\frac{1}{t}\Bigr{)}\Bigr{)}^{2},

    where F(s)=sβ(s)F(s)=s\beta(s) and CiC_{i} (i=1,2,3i=1,2,3) are positive constants that depends only on E(0)E(0).

Theorem 2.3.

Suppose that hypotheses (H1)(H4)(H_{1})-(H_{4}) with ρ<γ\rho<\gamma and f=0f=0 hold. Moreover, assume that

(u0,u1){(u0,u1)×L2(Ω);|gu0|g2>λ0,1<E(0)<d0}(u_{0},u_{1})\in\{(u_{0},u_{1})\in\mathcal{H}\times L^{2}(\Omega);||~{}|\nabla_{g}u_{0}|_{g}||_{2}>\lambda_{0},~{}-1<E(0)<d_{0}\}

and

β1(1)((γ+2)(μ0γλ022(γ+2)E1)28(γ+1)meas(Γ1)(μ0λ022E1))γ+1γ+2,\beta^{-1}(1)\leq\Bigl{(}\frac{(\gamma+2)(\mu_{0}\gamma\lambda^{2}_{0}-2(\gamma+2)E_{1})^{2}}{8(\gamma+1)meas(\Gamma_{1})(\mu_{0}\lambda^{2}_{0}-2E_{1})}\Bigr{)}^{\frac{\gamma+1}{\gamma+2}}, (2.15)

where

E1={0ifE(0)<0,positive constant satisfyingE(0)<E1<d0andE1<E(0)+1ifE(0)0.E_{1}=\begin{cases}0\hskip 8.53581pt&\text{if}\hskip 8.53581ptE(0)<0,\\ \text{positive constant satisfying}\hskip 8.53581ptE(0)<E_{1}<d_{0}\hskip 8.53581pt\text{and}\hskip 8.53581ptE_{1}<E(0)+1\hskip 8.53581pt&\text{if}\hskip 8.53581ptE(0)\geq 0.\end{cases}

Then the weak solution of the problem (1.1) blows up in finite time.

3. Proof of Theorem 2.1 : local existence

This section is devoted to prove the local existence in Theorem 2.1. The proof is based on three steps according to the following condition of the source term hh:

  1. (1)

    Existence of the global solution when hh is globally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) to L2(Γ1)L^{2}(\Gamma_{1}).

  2. (2)

    Existence of the local solution when hh is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) to L2(Γ1)L^{2}(\Gamma_{1}).

  3. (3)

    Existence of the local solution when hh is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) to Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}).

Then since the mapping |u|γu|u|^{\gamma}u is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) to Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) (see Remark 2.2), the existence of the local solution can be guaranteed even if h(u)=|u|γuh(u)=|u|^{\gamma}u.

3.1. Globally Lipschitz source

We first deal with the case where the source hh is globally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) to L2(Γ1)L^{2}(\Gamma_{1}). In this case, we have the following result.

Proposition 3.1.

Assume that (H1)(H3)(H_{1})-(H_{3}) hold. In addition, assume that (u0,u1)×L2(Ω)(u_{0},u_{1})\in\mathcal{H}\times L^{2}(\Omega) and h:H01(Ω)L2(Γ1)h:H^{1}_{0}(\Omega)\rightarrow L^{2}(\Gamma_{1}) is globally Lipschitz continuous. Then problem (1.1) has a unique global solutions uC(0,T;)C1(0,T;L2(Ω))u\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)) for arbitrary T>0T>0.

Our goal in this subsection is to show the local existence result for problem (1.1). We construct an approximate solution by using the Faedo-Galerkin method. Let {wj}j\{w_{j}\}_{j\in\mathbb{N}} be a basis in H01(Ω)H^{1}_{0}(\Omega) and define Vm=span{w1,w2,,wm}V_{m}=span\{w_{1},w_{2},\cdots,w_{m}\}. Let u0mu^{m}_{0} and u1mu^{m}_{1} be sequences of VmV_{m} such that u0mu0u^{m}_{0}\rightarrow u_{0} strongly in H01(Ω)H^{1}_{0}(\Omega) and u1mu1u^{m}_{1}\rightarrow u_{1} strongly in L2(Ω)L^{2}(\Omega). We search for a function, for each η(0,1)\eta\in(0,1) and mm\in\mathbb{N},

uηm(t)=j=1mδjm(t)wju^{\eta m}(t)=\sum^{m}_{j=1}\delta^{jm}(t)w_{j}

satisfying the approximate perturbed equation

{Ωuttηmw𝑑x+μ(t)Ωguηm,gwg𝑑x+ηΓ1utηmw𝑑Γ+Γ1q(utηm)w𝑑ΓΓ1h(uηm)w𝑑Γ=Ωf(t)w𝑑x,wVm,u0ηm=j=1mu0,wjwj,u1ηm=j=1mu1,wjwj.\begin{cases}\vspace{3mm}\int_{\Omega}u^{\eta m}_{tt}wdx+\mu(t)\int_{\Omega}\langle\nabla_{g}u^{\eta m},\nabla_{g}w\rangle_{g}dx+\eta\int_{\Gamma_{1}}u^{\eta m}_{t}wd\Gamma\\ \vspace{3mm}\hskip 28.45274pt+\int_{\Gamma_{1}}q(u^{\eta m}_{t})wd\Gamma-\int_{\Gamma_{1}}h(u^{\eta m})wd\Gamma=\int_{\Omega}f(t)wdx,\hskip 8.53581ptw\in V_{m},\\ \vspace{3mm}u^{\eta m}_{0}=\sum^{m}_{j=1}\langle u_{0},w_{j}\rangle w_{j},\hskip 8.53581ptu^{\eta m}_{1}=\sum^{m}_{j=1}\langle u_{1},w_{j}\rangle w_{j}.\end{cases} (3.1)

Since (3.1) is a normal system of ordinary differential equations, there exist uηmu^{\eta m}, solutions to problem (3.1). A solution uu to problem (1.1) on some internal [0,tm)[0,t_{m}), tm(0,T]t_{m}\in(0,T] will be obtain as the limit of uηmu^{\eta m} as mm\rightarrow\infty and η0\eta\rightarrow 0. Next, we show that tm=Tt_{m}=T and the local solution is uniformly bounded independent of mm, η\eta and tt. For this purpose, let us replace ww by utηmu^{\eta m}_{t} in (3.1) we obtain

ddt[12utηm22+12μ(t)|guηm|g22+1γ+2uηmγ+2,Γ1γ+2]+ηutηm2,Γ12+Γ1q(utηm)utηm𝑑Γ\displaystyle\frac{d}{dt}\Biggl{[}\frac{1}{2}||u^{\eta m}_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u^{\eta m}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\Biggr{]}+\eta||u^{\eta m}_{t}||^{2}_{2,\Gamma_{1}}+\int_{\Gamma_{1}}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma (3.2)
=12μ(t)|guηm|g22+Γ1h(uηm)utηm𝑑Γ+Γ1|uηm|γuηmutηm𝑑Γ+Ωf(t)utηm𝑑x.\displaystyle=\frac{1}{2}\mu^{\prime}(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+\int_{\Gamma_{1}}h(u^{\eta m})u^{\eta m}_{t}d\Gamma+\int_{\Gamma_{1}}|u^{\eta m}|^{\gamma}u^{\eta m}u^{\eta m}_{t}d\Gamma+\int_{\Omega}f(t)u^{\eta m}_{t}dx.

We will now estimate Γ1q(utηm)utηm𝑑Γ\int_{\Gamma_{1}}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma, Γ1|uηm|γuηmutηm𝑑Γ\int_{\Gamma_{1}}|u^{\eta m}|^{\gamma}u^{\eta m}u^{\eta m}_{t}d\Gamma and Γ1h(uηm)utηm𝑑Γ\int_{\Gamma_{1}}h(u^{\eta m})u^{\eta m}_{t}d\Gamma. From the hypotheses on qq, we have

Γ1q(utηm)utηm𝑑Γ\displaystyle\int_{\Gamma_{1}}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma =|utηm|1q(utηm)utηm𝑑Γ+|utηm|>1q(utηm)utηm𝑑Γ\displaystyle=\int_{|u^{\eta m}_{t}|\leq 1}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma+\int_{|u^{\eta m}_{t}|>1}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma (3.3)
|utηm(t)|>1q(utηm)utηm𝑑Γ\displaystyle\geq\int_{|u^{\eta m}_{t}(t)|>1}q(u^{\eta m}_{t})~{}u^{\eta m}_{t}d\Gamma
c3Γ1|utηm|ρ+2𝑑Γc3|utηm|1|utηm|ρ+2𝑑Γ\displaystyle\geq c_{3}\int_{\Gamma_{1}}|u^{\eta m}_{t}|^{\rho+2}d\Gamma-c_{3}\int_{|u^{\eta m}_{t}|\leq 1}|u^{\eta m}_{t}|^{\rho+2}d\Gamma
c3utηmρ+2,Γ1ρ+2c3meas(Γ1).\displaystyle\geq c_{3}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}-c_{3}meas(\Gamma_{1}).

From the assumption ρ2(n2)γ2n(n2)γ\rho\geq\frac{2(n-2)\gamma-2}{n-(n-2)\gamma}, we have the imbedding H01(Ω)L2(n1)n2(Γ1)L(γ+1)ρ+2ρ+1(Γ1)H^{1}_{0}(\Omega)\hookrightarrow L^{\frac{2(n-1)}{n-2}}(\Gamma_{1})\hookrightarrow L^{(\gamma+1)\frac{\rho+2}{\rho+1}}(\Gamma_{1}), so that by Young’s inequality with ρ+1ρ+2+1ρ+2=1\frac{\rho+1}{\rho+2}+\frac{1}{\rho+2}=1 we deduce that

Γ1|uηm|γuηmutηm𝑑Γ\displaystyle\int_{\Gamma_{1}}|u^{\eta m}|^{\gamma}u^{\eta m}u^{\eta m}_{t}d\Gamma C(ϵ0)uηm(γ+1)ρ+2ρ+1,Γ1γ+1+ϵ0utηmρ+2,Γ1ρ+2\displaystyle\leq C(\epsilon_{0})||u^{\eta m}||^{\gamma+1}_{(\gamma+1)\frac{\rho+2}{\rho+1},\Gamma_{1}}+\epsilon_{0}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}} (3.4)
C(ϵ0)|guηm|g2γ+1+ϵ0utηmρ+2,Γ1ρ+2\displaystyle\leq C(\epsilon_{0})||~{}|\nabla_{g}u^{\eta m}|_{g}||^{\gamma+1}_{2}+\epsilon_{0}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}
C(ϵ)(1+|guηm|g2)2+ϵ0utηmρ+2,Γ1ρ+2\displaystyle\leq C(\epsilon)\bigl{(}1+||~{}|\nabla_{g}u^{\eta m}|_{g}||_{2}\bigr{)}^{2}+\epsilon_{0}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}
C(ϵ)(1+|guηm|g22)+ϵ0utηmρ+2,Γ1ρ+2.\displaystyle\leq C(\epsilon)\bigl{(}1+||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}\bigr{)}+\epsilon_{0}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}.

Under the assumption that hh is globally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) into L2(Γ1)L^{2}(\Gamma_{1}) we have

h(uηm)2,Γ1h(uηm)h(0)2,Γ1+h(0)2,Γ1Lhuηm2+h(0)2,Γ1C4(uηm2+1),||h(u^{\eta m})||_{2,\Gamma_{1}}\leq||h(u^{\eta m})-h(0)||_{2,\Gamma_{1}}+||h(0)||_{2,\Gamma_{1}}\leq L_{h}||\nabla u^{\eta m}||_{2}+||h(0)||_{2,\Gamma_{1}}\leq C_{4}(||\nabla u^{\eta m}||_{2}+1),

where LhL_{h} is the Lipschitz constant and C4C_{4} is for some positive constant, so that by Hölder’s and Young’s inequalities and from the fact (2.1) and the imbedding Lρ+2(Γ1)L2(Γ1)L^{\rho+2}(\Gamma_{1})\hookrightarrow L^{2}(\Gamma_{1}), we deduce that

|Γ1h(uηm)utηm𝑑Γ|\displaystyle\Bigl{|}\int_{\Gamma_{1}}h(u^{\eta m})u^{\eta m}_{t}d\Gamma\Bigr{|} C(ϵ)h(uηm)2,Γ12+ϵutηm2,Γ12\displaystyle\leq C(\epsilon)||h(u^{\eta m})||^{2}_{2,\Gamma_{1}}+\epsilon||u^{\eta m}_{t}||^{2}_{2,\Gamma_{1}} (3.5)
C(ϵ)(1c1|guηm|g22+1)+ϵutηm2,Γ12,\displaystyle\leq C(\epsilon)\Bigl{(}\frac{1}{c_{1}}||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+1\Bigr{)}+\epsilon||u^{\eta m}_{t}||^{2}_{2,\Gamma_{1}},
C(ϵ)(1c1|guηm|g22+1)+ϵCρ+2,22utηmρ+2,Γ12\displaystyle\leq C(\epsilon)\Bigl{(}\frac{1}{c_{1}}||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+1\Bigr{)}+\epsilon C^{2}_{\rho+2,2}||u^{\eta m}_{t}||^{2}_{\rho+2,\Gamma_{1}}
C(ϵ)(1c1|guηm|g22+1)+ϵ2ρ+1Cρ+2,22(1+utηmρ+2,Γ1ρ+2),\displaystyle\leq C(\epsilon)\Bigl{(}\frac{1}{c_{1}}||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+1\Bigr{)}+\epsilon 2^{\rho+1}C^{2}_{\rho+2,2}\bigl{(}1+||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}\bigr{)},

where Cρ+2,2C_{\rho+2,2} is an imbedding constant.

Replacing (3.3), (3.4) and (3.5) in (3.2) we get

ddt[12utηm22+12μ(t)|guηm|g22+1γ+2uηmγ+2,Γ1γ+2]\displaystyle\frac{d}{dt}\Biggl{[}\frac{1}{2}||u^{\eta m}_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u^{\eta m}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\Biggr{]} (3.6)
+ηutηm2,Γ12+(c3ϵ0ϵ2ρ+1Cρ+2,22)utηmρ+2,Γ1ρ+2\displaystyle\hskip 14.22636pt+\eta||u^{\eta m}_{t}||^{2}_{2,\Gamma_{1}}+\Bigl{(}c_{3}-\epsilon_{0}-\epsilon 2^{\rho+1}C^{2}_{\rho+2,2}\Bigr{)}||u^{\eta m}_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}
c3meas(Γ1)+C(ϵ0)+C(ϵ)+ϵ2ρ+1Cρ+2,22\displaystyle\leq c_{3}meas(\Gamma_{1})+C(\epsilon_{0})+C(\epsilon)+\epsilon 2^{\rho+1}C^{2}_{\rho+2,2}
+(12μ(t)+C(ϵ0)+C(ϵ)c1)|guηm|g22+12f(t)22+12utηm22.\displaystyle\hskip 14.22636pt+\Bigl{(}\frac{1}{2}\mu^{\prime}(t)+C(\epsilon_{0})+\frac{C(\epsilon)}{c_{1}}\Bigr{)}||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+\frac{1}{2}||f(t)||^{2}_{2}+\frac{1}{2}||u^{\eta m}_{t}||^{2}_{2}.

By integrating (3.6) over (0,t)(0,t) with t(0,tm)t\in(0,t_{m}) we have

12utηm22+12μ(t)|guηm|g22+1γ+2uηmγ+2,Γ1γ+2+η0tusηm2,Γ12𝑑s\displaystyle\frac{1}{2}||u^{\eta m}_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u^{\eta m}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\eta\int^{t}_{0}||u^{\eta m}_{s}||^{2}_{2,\Gamma_{1}}ds (3.7)
+(c3ϵ0ϵ2ρ+1Cρ+2,22)0t||usηm||ρ+2,Γ1ρ+2ds\displaystyle\hskip 14.22636pt+\Bigl{(}c_{3}--\epsilon_{0}-\epsilon 2^{\rho+1}C^{2}_{\rho+2,2}\Bigr{)}\int^{t}_{0}||u^{\eta m}_{s}||^{\rho+2}_{\rho+2,\Gamma_{1}}ds
(c3meas(Γ1)+C(ϵ0)+C(ϵ)+ϵ2ρ+1Cρ+2,22)T\displaystyle\leq\Bigl{(}c_{3}meas(\Gamma_{1})+C(\epsilon_{0})+C(\epsilon)+\epsilon 2^{\rho+1}C^{2}_{\rho+2,2}\Bigr{)}T
+12u122+12μ(0)|gu0|g22+1γ+2u0γ+2,Γ1γ+2\displaystyle\hskip 14.22636pt+\frac{1}{2}||u_{1}||^{2}_{2}+\frac{1}{2}\mu(0)||~{}|\nabla_{g}u_{0}|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u_{0}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}
+(12μL(0,T)+C(ϵ0)+C(ϵ)c1)0t|guηm(s)|g22𝑑s\displaystyle\hskip 14.22636pt+\Bigl{(}\frac{1}{2}||\mu^{\prime}||_{L^{\infty}(0,T)}+C(\epsilon_{0})+\frac{C(\epsilon)}{c_{1}}\Bigr{)}\int^{t}_{0}||~{}|\nabla_{g}u^{\eta m}(s)|_{g}||^{2}_{2}ds
+120tf(s)22𝑑s+120tusηm(s)22𝑑s.\displaystyle\hskip 14.22636pt+\frac{1}{2}\int^{t}_{0}||f(s)||^{2}_{2}ds+\frac{1}{2}\int^{t}_{0}||u^{\eta m}_{s}(s)||^{2}_{2}ds.

Therefore, choosing ϵ0=14c3\epsilon_{0}=\frac{1}{4}c_{3} and ϵ=c32ρ+3Cρ+2,22\epsilon=\frac{c_{3}}{2^{\rho+3}C^{2}_{\rho+2,2}} and then by Gronwall’s lemma we obtain

utηm22+|guηm|g22+uηmγ+2,Γ1γ+2+0tusηm(s)2,Γ12𝑑s+0tusηm(s)ρ+2,Γ1ρ+2𝑑sC5,||u^{\eta m}_{t}||^{2}_{2}+||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}+||u^{\eta m}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}||u^{\eta m}_{s}(s)||^{2}_{2,\Gamma_{1}}ds+\int^{t}_{0}||u^{\eta m}_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds\leq C_{5}, (3.8)

where C5C_{5} is a positive constant which is independent of mm, η\eta and tt. The estimate (3.8) implies that

uηmis uniformly bounded inL(0,T;)u^{\eta m}\hskip 8.53581pt\text{is uniformly bounded in}\hskip 8.53581ptL^{\infty}(0,T;\mathcal{H}) (3.9)

and

utηmis uniformly bounded inL(0,T;L2(Ω)).u^{\eta m}_{t}\hskip 8.53581pt\text{is uniformly bounded in}\hskip 8.53581ptL^{\infty}(0,T;L^{2}(\Omega)). (3.10)

We note that from (3.8), taking the hypotheses on qq into account we also obtain

0tΓ1|q(usηm(s))|2𝑑Γ𝑑sC6,\int^{t}_{0}\int_{\Gamma_{1}}|q(u^{\eta m}_{s}(s))|^{2}d\Gamma ds\leq C_{6}, (3.11)

where C6C_{6} is a positive constant independent of mm, η\eta and tt.

From (3.8)-(3.11), there exists a subsequence of {uηm}\{u^{\eta m}\}, which we still denote by {uηm}\{u^{\eta m}\}, such that

uηmuηweakly star inL(0,T;),u^{\eta m}\rightarrow u^{\eta}\hskip 8.53581pt\text{weakly star in}\hskip 8.53581ptL^{\infty}(0,T;\mathcal{H}),
utηmutηweakly star inL(0,T;L2(Ω)),u^{\eta m}_{t}\rightarrow u^{\eta}_{t}\hskip 8.53581pt\text{weakly star in}\hskip 8.53581ptL^{\infty}(0,T;L^{2}(\Omega)),
uηmuηweakly inL2(0,T;),u^{\eta m}\rightarrow u^{\eta}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;\mathcal{H}),
utηmutηweakly inL2(0,T;L2(Ω)),u^{\eta m}_{t}\rightarrow u^{\eta}_{t}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Omega)),
utηmutηweakly inL2(0,T;L2(Γ1)),u^{\eta m}_{t}\rightarrow u^{\eta}_{t}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})),
uttηmuttηweakly inL2(0,T;H1(Ω)),u^{\eta m}_{tt}\rightarrow u^{\eta}_{tt}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;H^{-1}(\Omega)),
q(utηm)ψweakly inL2(0,T;L2(Γ1)).q(u^{\eta m}_{t})\rightarrow\psi\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})).

Since H1/2(Γ)L2(Γ)H^{1/2}(\Gamma)\hookrightarrow L^{2}(\Gamma) is compact, we have, thanks to Aubin-Lions Theorem that

uηmuηstrongly inL2(0,T;L2(Γ1)),u^{\eta m}\rightarrow u^{\eta}\hskip 8.53581pt\text{strongly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})),

and consequently, by making use of Lions lemma, we deduce

h(uηm)h(uη)weakly inL2(0,T;L2(Γ1)).h(u^{\eta m})\rightarrow h(u^{\eta})\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})).

The above convergences permit us to pass to the limit in the (3.1). Since {wj}\{w_{j}\} is a basis of H01(Ω)H^{1}_{0}(\Omega) and VmV_{m} is dense in H01(Ω)H^{1}_{0}(\Omega), after passing to the limit we obtain

0TΩuttη,vdxθ(t)dt+0Tμ(t)Ωguη,gvg𝑑xθ(t)𝑑t+η0TΓ1utηv𝑑Γθ(t)𝑑t\displaystyle\int^{T}_{0}\int_{\Omega}u^{\eta}_{tt},vdx\theta(t)dt+\int^{T}_{0}\mu(t)\int_{\Omega}\langle\nabla_{g}u^{\eta},\nabla_{g}v\rangle_{g}dx\theta(t)dt+\eta\int^{T}_{0}\int_{\Gamma_{1}}u^{\eta}_{t}vd\Gamma\theta(t)dt (3.12)
+0TΓ1ψv𝑑Γθ(t)𝑑t0TΓ1h(uη)v𝑑xθ(t)𝑑t=0TΩf(x,t)v𝑑xθ(t)𝑑t,\displaystyle+\int^{T}_{0}\int_{\Gamma_{1}}\psi vd\Gamma\theta(t)dt-\int^{T}_{0}\int_{\Gamma_{1}}h(u^{\eta})vdx\theta(t)dt=\int^{T}_{0}\int_{\Omega}f(x,t)vdx\theta(t)dt,

for all θD(0,T)\theta\in D(0,T) and vH01(Ω)v\in H^{1}_{0}(\Omega).

Since estimates (3.8) and (3.11) are also independent of η\eta, we can pass to the limit when η0\eta\rightarrow 0 in uηu^{\eta} obtaining a function uu by the same argument used to obtain uηu^{\eta} from uηmu^{\eta m}, such that

uηuweakly inL2(0,T;),u^{\eta}\rightarrow u\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;\mathcal{H}), (3.13)
utηutweakly inL2(0,T;L2(Ω)),u^{\eta}_{t}\rightarrow u_{t}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Omega)), (3.14)
utηutweakly inL2(0,T;L2(Γ1)),u^{\eta}_{t}\rightarrow u_{t}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})), (3.15)
uttηuttweakly inL2(0,T;H1(Ω)),u^{\eta}_{tt}\rightarrow u_{tt}\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;H^{-1}(\Omega)), (3.16)
q(utη)ψweakly inL2(0,T;L2(Γ1)),q(u^{\eta}_{t})\rightarrow\psi\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})), (3.17)
h(uη)h(u)weakly inL2(0,T;L2(Γ1)).h(u_{\eta})\rightarrow h(u)\hskip 14.22636pt\text{weakly in}\hskip 14.22636ptL^{2}(0,T;L^{2}(\Gamma_{1})). (3.18)

By above convergences in (3.12), we have

0TΩutt,vdxθ(t)dt+0Tμ(t)Ωgu,gvg𝑑xθ(t)𝑑t\displaystyle\int^{T}_{0}\int_{\Omega}u_{tt},vdx\theta(t)dt+\int^{T}_{0}\mu(t)\int_{\Omega}\langle\nabla_{g}u,\nabla_{g}v\rangle_{g}dx\theta(t)dt (3.19)
+0TΓ1ψv𝑑Γθ(t)𝑑t0TΓ1h(u)v𝑑xθ(t)𝑑t=0TΩf(x,t)v𝑑xθ(t)𝑑t.\displaystyle+\int^{T}_{0}\int_{\Gamma_{1}}\psi vd\Gamma\theta(t)dt-\int^{T}_{0}\int_{\Gamma_{1}}h(u)vdx\theta(t)dt=\int^{T}_{0}\int_{\Omega}f(x,t)vdx\theta(t)dt.

From the (3.19) and taking vD(Ω)v\in D(\Omega), we conclude that

uttμ(t)Lu=finD(Ω×(0,T))u_{tt}-\mu(t)Lu=f\hskip 8.53581pt\text{in}\hskip 8.53581ptD^{\prime}(\Omega\times(0,T)) (3.20)

and since (3.17) and (3.18), it holds that

μ(t)uν+ψ=h(u)inL2(0,T;L2(Γ1)).\mu(t)\frac{\partial u}{\partial\nu}+\psi=h(u)\hskip 8.53581pt\text{in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})).

Our goal is to show that ψ=q(ut)\psi=q(u_{t}). Indeed, considering w=uηmw=u^{\eta m} in (3.1) and then integrating over (0,T)(0,T), we have

0Tuttηm,uηm𝑑t+0Tμ(t)|guηm|g22𝑑t+η0Tutηm,uηmΓ1𝑑t+0Tq(utηm),uηmΓ1𝑑t0Th(uηm),uηmΓ1𝑑t=0Tf,uηm𝑑t.\int^{T}_{0}\langle u^{\eta m}_{tt},u^{\eta m}\rangle dt+\int^{T}_{0}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}dt+\eta\int^{T}_{0}\langle u^{\eta m}_{t},u^{\eta m}\rangle_{\Gamma_{1}}dt\\ +\int^{T}_{0}\langle q(u^{\eta m}_{t}),u^{\eta m}\rangle_{\Gamma_{1}}dt-\int^{T}_{0}\langle h(u^{\eta m}),u^{\eta m}\rangle_{\Gamma_{1}}dt=\int^{T}_{0}\langle f,u^{\eta m}\rangle dt.

Then from convergences (3.13)-(3.18) we obtain

limm,η00Tμ(t)|guηm|g22𝑑t=0Tutt,u𝑑t0Tψ,uΓ1𝑑t+0Th(u),uΓ1𝑑t+0Tf,u𝑑t.\lim_{m\rightarrow\infty,\eta\rightarrow 0}\int^{T}_{0}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}dt=-\int^{T}_{0}\langle u_{tt},u\rangle dt-\int^{T}_{0}\langle\psi,u\rangle_{\Gamma_{1}}dt+\int^{T}_{0}\langle h(u),u\rangle_{\Gamma_{1}}dt+\int^{T}_{0}\langle f,u\rangle dt. (3.21)

By combining (3.20) and (3.21), we have

limm,η00Tμ(t)|guηm|g22𝑑t=0Tμ(t)|gu|g22𝑑t,\lim_{m\rightarrow\infty,\eta\rightarrow 0}\int^{T}_{0}\mu(t)||~{}|\nabla_{g}u^{\eta m}|_{g}||^{2}_{2}dt=\int^{T}_{0}\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}dt,

which implies that

|guηm|g|gu|gstrongly inL2(0,T;L2(Ω)).|\nabla_{g}u^{\eta m}|_{g}\rightarrow|\nabla_{g}u|_{g}\hskip 8.53581pt\text{strongly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Omega)). (3.22)

Next, considering w=utηmw=u^{\eta m}_{t} in (3.1) and then integrating over (0,T)(0,T), we have

0Tuttηm,utηm𝑑t+0Tμ(t)Ωguηm,gutηmg𝑑x𝑑t+η0Tutηm2,Γ12𝑑t+0Tq(utηm),utηmΓ1𝑑t0Th(uηm),utηmΓ1𝑑t=0Tf,utηm𝑑t.\int^{T}_{0}\langle u^{\eta m}_{tt},u^{\eta m}_{t}\rangle dt+\int^{T}_{0}\mu(t)\int_{\Omega}\langle\nabla_{g}u^{\eta m},\nabla_{g}u^{\eta m}_{t}\rangle_{g}dxdt+\eta\int^{T}_{0}||u^{\eta m}_{t}||^{2}_{2,\Gamma_{1}}dt\\ +\int^{T}_{0}\langle q(u^{\eta m}_{t}),u^{\eta m}_{t}\rangle_{\Gamma_{1}}dt-\int^{T}_{0}\langle h(u^{\eta m}),u^{\eta m}_{t}\rangle_{\Gamma_{1}}dt=\int^{T}_{0}\langle f,u^{\eta m}_{t}\rangle dt.

From (3.14)-(3.18) and (3.22), we arrive at

limm,η00Tq(utηm),utηmΓ1𝑑t=0Tψ,utΓ1𝑑t.\lim_{m\rightarrow\infty,\eta\rightarrow 0}\int^{T}_{0}\langle q(u^{\eta m}_{t}),u^{\eta m}_{t}\rangle_{\Gamma_{1}}dt=\int^{T}_{0}\langle\psi,u_{t}\rangle_{\Gamma_{1}}dt. (3.23)

On the other hand, since qq is a nondecreasing monotone function, we get

0Tq(utηm)q(φ),utηmφΓ1𝑑t0\int^{T}_{0}\langle q(u^{\eta m}_{t})-q(\varphi),u^{\eta m}_{t}-\varphi\rangle_{\Gamma_{1}}dt\geq 0

for all φL2(Γ1)\varphi\in L^{2}(\Gamma_{1}). Thus, it implies that

0Tq(utηm),φΓ1𝑑t+0Tq(φ),utηmφΓ1𝑑t0Tq(utηm),utηmΓ1𝑑t.\int^{T}_{0}\langle q(u^{\eta m}_{t}),\varphi\rangle_{\Gamma_{1}}dt+\int^{T}_{0}\langle q(\varphi),u^{\eta m}_{t}-\varphi\rangle_{\Gamma_{1}}dt\leq\int^{T}_{0}\langle q(u^{\eta m}_{t}),u^{\eta m}_{t}\rangle_{\Gamma_{1}}dt.

By considering (3.15), (3.17) and (3.23), we obtain

0Tψq(φ),utφΓ1𝑑t0,\int^{T}_{0}\langle\psi-q(\varphi),u_{t}-\varphi\rangle_{\Gamma_{1}}dt\geq 0, (3.24)

which implies that ψ=q(ut)\psi=q(u_{t}).

We now show the uniqueness of the solution. Let u1u^{1} and u2u^{2} be two solutions of problem (1.1). Then z=u1u2z=u^{1}-u^{2} verifies

Ωzttwdx+μ(t)Ωgz,gwgdx+Γ1(q(ut1)q(ut2))wdΓ=Γ1(h(u1)h(u2)wdΓdt,\int_{\Omega}z_{tt}wdx+\mu(t)\int_{\Omega}\langle\nabla_{g}z,\nabla_{g}w\rangle_{g}dx+\int_{\Gamma_{1}}(q(u^{1}_{t})-q(u^{2}_{t}))wd\Gamma=\int_{\Gamma_{1}}(h(u^{1})-h(u^{2})wd\Gamma dt,

for all ww\in\mathcal{H}. By replacing w=ztw=z_{t} in above identity and observing that qq is monotonously nondecreasing and h:H01(Ω)L2(Γ1)h:H^{1}_{0}(\Omega)\rightarrow L^{2}(\Gamma_{1}) is globally Lipschitz, it holds that

ddt[12zt22+12μ(t)|gz|g22]C7|gz|g22,\frac{d}{dt}\Biggl{[}\frac{1}{2}||z_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}z|_{g}||^{2}_{2}\Biggr{]}\leq C_{7}||~{}|\nabla_{g}z|_{g}||^{2}_{2},

Where C7C_{7} is for some positive constant. By integrating from 0 to tt and using Gronwall’s Lemma, we conclude that zt2=|gz|g2=0||z_{t}||_{2}=||~{}|\nabla_{g}z|_{g}||_{2}=0.

3.2. Locally Lipschitz source

In this subsection, we loosen the globally Lipschitz condition on the source by allowing hh to be locally Lipschitz continuous. More precisely, we have the following result.

Proposition 3.2.

Assume that (H1)(H4)(H_{1})-(H_{4}) hold. In addition, assume that (u0,u1)×L2(Ω)(u_{0},u_{1})\in\mathcal{H}\times L^{2}(\Omega) and h:H01(Ω)L2(Γ1)h:H^{1}_{0}(\Omega)\rightarrow L^{2}(\Gamma_{1}) is locally Lipschitz continuous satisfying c5|s|γ+1|h(s)|c6|s|γ+1c_{5}|s|^{\gamma+1}\leq|h(s)|\leq c_{6}|s|^{\gamma+1}, where c5c_{5}, c6c_{6} are for some positive constants. Then problem (1.1) has a unique local solution uC(0,T;)C1(0,T;L2(Ω))u\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)) for some T>0T>0.

Proof.

Define

hK(u)={h(u)if|gu|g2K,h(Ku|gu|g2)if|gu|g2>K,h_{K}(u)=\begin{cases}h(u)\hskip 8.53581pt&\text{if}\hskip 8.53581pt||~{}|\nabla_{g}u|_{g}||_{2}\leq K,\\ h\Bigl{(}\frac{Ku}{||~{}|\nabla_{g}u|_{g}||_{2}}\Bigr{)}\hskip 8.53581pt&\text{if}\hskip 8.53581pt||~{}|\nabla_{g}u|_{g}||_{2}>K,\end{cases}

where KK is a positive constant. With this truncated hKh_{K}, we consider the following problem:

{uttμ(t)Lu=f(x,t)inΩ×(0,+),u=0onΓ0×(0,+),μ(t)uνL+q(ut)=hK(u)onΓ1×(0,+),u(x,0)=u0(x),ut(x,0)=u1(x).\begin{cases}\vspace{3mm}u_{tt}-\mu(t)Lu=f(x,t)&\hskip 14.22636pt\text{in}\hskip 14.22636pt\Omega~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u=0&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{0}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}\mu(t)\frac{\partial u}{\partial\nu_{L}}+q(u_{t})=h_{K}(u)&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{1}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u(x,0)=u_{0}(x),\hskip 14.22636ptu_{t}(x,0)=u_{1}(x).\end{cases} (3.25)

Since hK:H01(Ω)L2(Γ1)h_{K}:H^{1}_{0}(\Omega)\rightarrow L^{2}(\Gamma_{1}) is globally Lipschitz continuous for each KK (see [10]), then by Proposition 3.1, the truncated problem (3.25) has a unique global solution uKC(0,T;)C1(0,T;L2(Ω))u_{K}\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)) for any T>0T>0. Moreover by [30] there exists a sequence of functions uKlu^{l}_{K}, which converges to uKu_{K} in the class C(0,T;H2(Ω))C1(0,T;H01(Ω))C(0,T;H^{2}(\Omega))\cap C^{1}(0,T;H^{1}_{0}(\Omega)). For simplifying the notation in the rest of the proof, we shall express uKlu^{l}_{K} as uu.

By the regularity of uu, we can multiply (3.25)\eqref{321*} by utu_{t} and integrate on Ω×(0,t)\Omega\times(0,t), where 0<t<T0<t<T. Then we obtain by using the fact μ(s)<0\mu^{\prime}(s)<0 for all s>0s>0,

12(ut22+μ(t)|gu|g22)+1γ+2uγ+2,Γ1γ+2+0tΓ1q(us(x,s))us(x,s)𝑑Γ𝑑s\displaystyle\frac{1}{2}(||u_{t}||^{2}_{2}+\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2})+\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}\int_{\Gamma_{1}}q(u_{s}(x,s))u_{s}(x,s)d\Gamma ds (3.26)
12(u122+μ(0)|gu0|g22)+1γ+2u0γ+2,Γ1γ+2+0tΩf(x,s)us(x,s)𝑑x𝑑s\displaystyle\leq\frac{1}{2}(||u_{1}||^{2}_{2}+\mu(0)||~{}|\nabla_{g}u_{0}|_{g}||^{2}_{2})+\frac{1}{\gamma+2}||u_{0}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}\int_{\Omega}f(x,s)u_{s}(x,s)dxds
+(1+1c5)0tΓ1|hK(u(x,s))||us(x,s)|𝑑Γ𝑑s.\displaystyle\hskip 14.22636pt+\Bigl{(}1+\frac{1}{c_{5}}\Bigr{)}\int^{t}_{0}\int_{\Gamma_{1}}|h_{K}(u(x,s))|~{}|u_{s}(x,s)|d\Gamma ds.

We note that hK:H01(Ω)Lρ+2ρ+1(Γ1)h_{K}:H^{1}_{0}(\Omega)\rightarrow L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) is globally Lipschitz with Lipschitz constant Lh(K)L_{h}(K) (see [10, 13]). Hence we estimate the last term on the right-hand side of (3.26) as follows:

(1+1c5)0tΓ1|hK(u(x,s))||us(x,s)|𝑑Γ𝑑s\displaystyle\Bigl{(}1+\frac{1}{c_{5}}\Bigr{)}\int^{t}_{0}\int_{\Gamma_{1}}|h_{K}(u(x,s))|~{}|u_{s}(x,s)|d\Gamma ds (3.27)
(1+1c5)0thK(u(s))ρ+2ρ+1,Γ1usρ+2,Γ1𝑑s\displaystyle\leq\Bigl{(}1+\frac{1}{c_{5}}\Bigr{)}\int^{t}_{0}||h_{K}(u(s))||_{\frac{\rho+2}{\rho+1},\Gamma_{1}}||u_{s}||_{\rho+2,\Gamma_{1}}ds
ϵ10tus(s)ρ+2,Γ1ρ+2𝑑s+C(ϵ1)0thK(u(s))ρ+2ρ+1,Γ1ρ+2ρ+1𝑑s\displaystyle\leq\epsilon_{1}\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds+C(\epsilon_{1})\int^{t}_{0}||h_{K}(u(s))||^{\frac{\rho+2}{\rho+1}}_{\frac{\rho+2}{\rho+1},\Gamma_{1}}ds
ϵ10tus(s)ρ+2,Γ1ρ+2𝑑s+C(ϵ1)(2c1Lh(K))ρ+2ρ+10t|gu|g22𝑑s\displaystyle\leq\epsilon_{1}\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds+C(\epsilon_{1})\Bigl{(}\frac{2}{\sqrt{c_{1}}}L_{h}(K)\Bigr{)}^{\frac{\rho+2}{\rho+1}}\int^{t}_{0}||~{}|\nabla_{g}u|_{g}||^{2}_{2}ds
+tC(ϵ1)((2c1Lh(K))ρ+2ρ+1+2(ρ+1)|h(0)|ρ+2ρ+1meas(Γ1)).\displaystyle\hskip 14.22636pt+tC(\epsilon_{1})\Bigl{(}\Bigl{(}\frac{2}{\sqrt{c_{1}}}L_{h}(K)\Bigr{)}^{\frac{\rho+2}{\rho+1}}+2^{-(\rho+1)}|h(0)|^{\frac{\rho+2}{\rho+1}}meas(\Gamma_{1})\Bigr{)}.

From the hypothesis on qq, we have

0tΓ1q(us(x,s))us(x,s)𝑑Γ𝑑sc30tus(s)ρ+2,Γ1ρ+2𝑑stc3meas(Γ1).\int^{t}_{0}\int_{\Gamma_{1}}q(u_{s}(x,s))u_{s}(x,s)d\Gamma ds\geq c_{3}\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds-tc_{3}meas(\Gamma_{1}). (3.28)

By replacing (3.27) and (3.28) in (3.26) and choosing ϵ1c3\epsilon_{1}\leq c_{3}, we get

ut22+|gu|g22+uγ+2,Γ1γ+2+0tus(s)ρ+2,Γ1ρ+2𝑑sC9+C1(Lh(K))T+C2(Lh(K))0tus(s)22+|gu(s)|g22ds||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds\\ \leq C_{9}+C_{1}(L_{h}(K))T+C_{2}(L_{h}(K))\int^{t}_{0}||u_{s}(s)||^{2}_{2}+||~{}|\nabla_{g}u(s)|_{g}||^{2}_{2}ds (3.29)

for all t[0,T]t\in[0,T], where

C9=12C8(u122+μ(0)|gu0|g22+u0γ+2,Γ1γ+2+fL2(0,T;L2(Ω))),C_{9}=\frac{1}{2C_{8}}(||u_{1}||^{2}_{2}+\mu(0)||~{}|\nabla_{g}u_{0}|_{g}||^{2}_{2}+||u_{0}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+||f||_{L^{2}(0,T;L^{2}(\Omega))}),
C1(Lh(K))=1C8C(ϵ1)((2c1Lh(K))ρ+2ρ+1+2(ρ+1)|h(0)|ρ+2ρ+1meas(Γ1))+1C8c3meas(Γ1),C_{1}(L_{h}(K))=\frac{1}{C_{8}}C(\epsilon_{1})\Bigl{(}\Bigl{(}\frac{2}{\sqrt{c_{1}}}L_{h}(K)\Bigr{)}^{\frac{\rho+2}{\rho+1}}+2^{-(\rho+1)}|h(0)|^{\frac{\rho+2}{\rho+1}}meas(\Gamma_{1})\Bigr{)}+\frac{1}{C_{8}}c_{3}meas(\Gamma_{1}),
C2(Lh(K))=1C8(C(ϵ1)(2c1Lh(K))ρ+2ρ+1+12),C_{2}(L_{h}(K))=\frac{1}{C_{8}}\Bigl{(}C(\epsilon_{1})\Bigl{(}\frac{2}{\sqrt{c_{1}}}L_{h}(K)\Bigr{)}^{\frac{\rho+2}{\rho+1}}+\frac{1}{2}\Bigr{)},

for C8=min{1γ+2,μ02,c3ϵ1}C_{8}=\min\{\frac{1}{\gamma+2},\frac{\mu_{0}}{2},c_{3}-\epsilon_{1}\}. Thus by Gronwall’s inequality, (3.29) becomes

ut22+|gu|g22+uγ+2,Γ1γ+2+0tus(s)ρ+2,Γ1ρ+2𝑑s(C9+C1(Lh(K))T)eC2(Lh(K))t,||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds\leq(C_{9}+C_{1}(L_{h}(K))T)e^{C_{2}(L_{h}(K))t},

for all t[0,T]t\in[0,T]. If we choose

T=min{1C1(Lh(K)),1C2(Lh(K))ln2},T=\min\Bigl{\{}\frac{1}{C_{1}(L_{h}(K))},\frac{1}{C_{2}(L_{h}(K))}\ln 2\Bigr{\}}, (3.30)

then

ut22+|gu|g22+uγ+2,Γ1γ+2+0tus(s)ρ+2,Γ1ρ+2𝑑s2(C9+1)K2for allt[0,T],||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\int^{t}_{0}||u_{s}(s)||^{\rho+2}_{\rho+2,\Gamma_{1}}ds\leq 2(C_{9}+1)\leq K^{2}\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\in[0,T], (3.31)

provided we choose K22(C9+1)K^{2}\geq 2(C_{9}+1). Consequently, (3.31) gives us that |gu|g2K||~{}|\nabla_{g}u|_{g}||_{2}\leq K for all t[0,T]t\in[0,T]. Therefore, by the definition of hKh_{K}, we have that hK(u)=h(u)h_{K}(u)=h(u) on [0,T][0,T]. By the uniqueness of solutions, the solution of the truncated problem (3.25) accords with the solution of the original, non-truncated problem (1.1) for t[0,T]t\in[0,T], which means that the proof of Proposition 3.2 is completed.

3.3. Completion of the proof for the local existence

In order to establish the existence of solutions, we need to extend the result in Proposition 3.2 where the source hh is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) into Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}). For the construction of the Lipschitz approximation for the source, we employ another truncated function introduced in [40]. Let δnC0()\delta_{n}\in C^{\infty}_{0}(\mathbb{R}) be a cut off function such that

{0δn1,δn(s)=1,if|s|n,δn(s)=0,if|s|2n,\begin{cases}0\leq\delta_{n}\leq 1,\\ \delta_{n}(s)=1,\hskip 8.53581pt&\text{if}\hskip 8.53581pt|s|\leq n,\\ \delta_{n}(s)=0,\hskip 8.53581pt&\text{if}\hskip 8.53581pt|s|\geq 2n,\end{cases}

and |δn(s)|Cn|\delta^{\prime}_{n}(s)|\leq\frac{C}{n} for some constant CC independent from nn and define

hn(u)=h(u)δn(u).h_{n}(u)=h(u)\delta_{n}(u). (3.32)

Then the truncated function hnh_{n} is satisfied the following lemma. The proof of this lemma is a routine series of estimates as in [1, 13], so we omit it here.

Lemma 3.1.

The following statements hold.

  1. (1)

    hn:H01(Ω)L2(Γ1)h_{n}:H^{1}_{0}(\Omega)\rightarrow L^{2}(\Gamma_{1}) is globally Lipschitz continuous.

  2. (2)

    hn:H01ϵ(Ω)Lρ+2ρ+1(Γ1)h_{n}:H^{1-\epsilon}_{0}(\Omega)\rightarrow L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) is locally Lipschitz continuous with Lipschitz constant independent of nn.

With the truncated source hnh_{n} defined in (3.32), by Proposition 3.2 and Lemma 3.1, we have a unique local solution unC(0,T;)C1(0,T;L2(Ω))u^{n}\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)) satisfying the following approximation of (1.1)

{uttμ(t)Lu=f(x,t)inΩ×(0,+),u=0onΓ0×(0,+),μ(t)uνL+q(ut)=hn(u)onΓ1×(0,+),u(x,0)=u0(x),ut(x,0)=u1(x).\begin{cases}\vspace{3mm}u_{tt}-\mu(t)Lu=f(x,t)&\hskip 14.22636pt\text{in}\hskip 14.22636pt\Omega~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u=0&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{0}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}\mu(t)\frac{\partial u}{\partial\nu_{L}}+q(u_{t})=h_{n}(u)&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{1}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u(x,0)=u_{0}(x),\hskip 14.22636ptu_{t}(x,0)=u_{1}(x).\end{cases} (3.33)

From Lemma 3.1, the life span TT of each solution unu^{n}, given in (3.30), is independent of nn since the local Lipschitz constant of the mapping hn:H01(Ω)Lρ+2ρ+1(Γ1)h_{n}:H^{1}_{0}(\Omega)\rightarrow L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) is independent of nn. Also we known that TT depends on KK, where K22(C9+1)K^{2}\geq 2(C_{9}+1), however, since u1n22+|gu0n|g22+u0nγ+2,Γ1γ+2u122+|gu0|g22+u0γ+2,Γ1γ+2||u^{n}_{1}||^{2}_{2}+||~{}|\nabla_{g}u^{n}_{0}|_{g}||^{2}_{2}+||u^{n}_{0}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\rightarrow||u_{1}||^{2}_{2}+||~{}|\nabla_{g}u_{0}|_{g}||^{2}_{2}+||u_{0}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}, we can choose KK sufficiently large so that KK is independent of nn. By (3.31),

utn22+|gun|g22+unγ+2,Γ1γ+2K2||u^{n}_{t}||^{2}_{2}+||~{}|\nabla_{g}u^{n}|_{g}||^{2}_{2}+||u^{n}||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\leq K^{2} (3.34)

for all t[0,T]t\in[0,T]. Therefore, there exists a function uu and a subsequence of {un}\{u^{n}\}, which we still denote by {un}\{u^{n}\}, such that

unuweak star inL(0,T;),u^{n}\rightarrow u\hskip 8.53581pt\text{weak star in}\hskip 8.53581ptL^{\infty}(0,T;\mathcal{H}), (3.35)
utnutweak star inL(0,T;L2(Ω)).u^{n}_{t}\rightarrow u_{t}\hskip 8.53581pt\text{weak star in}\hskip 8.53581ptL^{\infty}(0,T;L^{2}(\Omega)). (3.36)

By (3.34), (3.35) and (3.36), we infer

ut22+|gu|g22+uγ+2,Γ1γ+2K2||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\leq K^{2} (3.37)

for all t[0,T]t\in[0,T]. Moreover, by Aubin-Lions Theorem, we have

unustrongly inL(0,T;H1ϵ(Ω)),u^{n}\rightarrow u\hskip 8.53581pt\text{strongly in}\hskip 8.53581ptL^{\infty}(0,T;H^{1-\epsilon}(\Omega)), (3.38)

for 0<ϵ<10<\epsilon<1. Since unu^{n} is a solution of (3.33), it holds that

0TΩuttn,ϕdxdt+0Tμ(t)Ωgun,gϕg𝑑x𝑑t+0TΓ1q(utn)ϕ𝑑Γ𝑑t=0TΩf(x,t)ϕ𝑑x𝑑t+0TΓ1hn(un)ϕ𝑑Γ𝑑t,\int^{T}_{0}\int_{\Omega}u^{n}_{tt},\phi dxdt+\int^{T}_{0}\mu(t)\int_{\Omega}\langle\nabla_{g}u^{n},\nabla_{g}\phi\rangle_{g}dxdt+\int^{T}_{0}\int_{\Gamma_{1}}q(u^{n}_{t})\phi d\Gamma dt\\ =\int^{T}_{0}\int_{\Omega}f(x,t)\phi dxdt+\int^{T}_{0}\int_{\Gamma_{1}}h_{n}(u^{n})\phi d\Gamma dt, (3.39)

for any ϕC(0,T;)C1(0,T;L2(Ω))\phi\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)), ϕLρ+2(0,T;Γ1)\phi\in L^{\rho+2}(0,T;\Gamma_{1}).

Now we will show that

limn0TΓ1hn(un)ϕ𝑑Γ𝑑t=0TΓ1h(u)ϕ𝑑Γ𝑑t.\lim_{n\rightarrow\infty}\int^{T}_{0}\int_{\Gamma_{1}}h_{n}(u^{n})\phi d\Gamma dt=\int^{T}_{0}\int_{\Gamma_{1}}h(u)\phi d\Gamma dt. (3.40)

Indeed, we have

|0TΓ1(hn(un)h(u))ϕ𝑑Γ𝑑t|0TΓ1|hn(un)hn(u)||ϕ|𝑑Γ𝑑t+0TΓ1|hn(u)h(u)||ϕ|𝑑Γ𝑑t.\Bigl{|}\int^{T}_{0}\int_{\Gamma_{1}}(h_{n}(u^{n})-h(u))\phi d\Gamma dt\Bigr{|}\leq\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u^{n})-h_{n}(u)|~{}|\phi|d\Gamma dt+\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u)-h(u)|~{}|\phi|d\Gamma dt. (3.41)

By (2)(2) in Lemma 3.1 and (3.38), we obtain

0TΓ1|hn(un)hn(u)||ϕ|𝑑Γ𝑑t\displaystyle\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u^{n})-h_{n}(u)|~{}|\phi|d\Gamma dt (0TΓ1|hn(un)hn(u)|ρ+2ρ+1𝑑Γ𝑑t)ρ+1ρ+2(0TΓ1|ϕ|ρ+2𝑑Γ𝑑t)1ρ+2\displaystyle\leq\Bigl{(}\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u^{n})-h_{n}(u)|^{\frac{\rho+2}{\rho+1}}d\Gamma dt\Bigr{)}^{\frac{\rho+1}{\rho+2}}\Bigl{(}\int^{T}_{0}\int_{\Gamma_{1}}|\phi|^{\rho+2}d\Gamma dt\Bigr{)}^{\frac{1}{\rho+2}} (3.42)
C(K)ϕLρ+2(0,T;Γ1)(0TunuH1ϵ(Ω)ρ+2ρ+1𝑑t)ρ+1ρ+20.\displaystyle\leq C(K)||\phi||_{L^{\rho+2}(0,T;\Gamma_{1})}\Bigl{(}\int^{T}_{0}||u^{n}-u||^{\frac{\rho+2}{\rho+1}}_{H^{1-\epsilon}(\Omega)}dt\Bigr{)}^{\frac{\rho+1}{\rho+2}}\rightarrow 0.

Since δn(u(x))1\delta_{n}(u(x))\rightarrow 1 a.e. in Ω\Omega, we have hn(u)h(u)h_{n}(u)\rightarrow h(u) a.e. Then we also have |hn(u)h(u)|ρ+2ρ+12ρ+2ρ+1|h(u)|ρ+2ρ+1|h_{n}(u)-h(u)|^{\frac{\rho+2}{\rho+1}}\leq 2^{\frac{\rho+2}{\rho+1}}|h(u)|^{\frac{\rho+2}{\rho+1}} and h(u)Lρ+2ρ+1(Γ1)h(u)\in L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}), for uH01(Ω)u\in H^{1}_{0}(\Omega). Thus by the Lebesgue Dominated Convergence Theorem, we have

0TΓ1|hn(u)h(u)||ϕ|𝑑Γ𝑑t\displaystyle\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u)-h(u)|~{}|\phi|d\Gamma dt (0TΓ1|hn(u)h(u)|ρ+2ρ+1𝑑Γ𝑑t)ρ+1ρ+2(0TΓ1|ϕ|ρ+2𝑑Γ𝑑t)1ρ+2\displaystyle\leq\Bigl{(}\int^{T}_{0}\int_{\Gamma_{1}}|h_{n}(u)-h(u)|^{\frac{\rho+2}{\rho+1}}d\Gamma dt\Bigr{)}^{\frac{\rho+1}{\rho+2}}\Bigl{(}\int^{T}_{0}\int_{\Gamma_{1}}|\phi|^{\rho+2}d\Gamma dt\Bigr{)}^{\frac{1}{\rho+2}} (3.43)
ϕLρ+2(0,T;Γ1)(0TΓ1|h(u)|ρ+2ρ+1|δn(u)1|ρ+2ρ+1𝑑Γ𝑑t)ρ+1ρ+20.\displaystyle\leq||\phi||_{L^{\rho+2}(0,T;\Gamma_{1})}\Bigl{(}\int^{T}_{0}\int_{\Gamma_{1}}|h(u)|^{\frac{\rho+2}{\rho+1}}|\delta_{n}(u)-1|^{\frac{\rho+2}{\rho+1}}d\Gamma dt\Bigr{)}^{\frac{\rho+1}{\rho+2}}\rightarrow 0.

From convergences (3.42) and (3.43), (3.41) gives us (3.40).

On the other hand, by using similar arguments from (3.21) to (3.24), we get

q(utn)q(ut)weakly inL2(0,T;L2(Γ1)).q(u^{n}_{t})\rightarrow q(u_{t})\hskip 8.53581pt\text{weakly in}\hskip 8.53581ptL^{2}(0,T;L^{2}(\Gamma_{1})). (3.44)

Convergences (3.36), (3.37), (3.40) and (3.44) permit us to pass to the limit in (3.39) and conclude the following result.

Proposition 3.3.

Assume that (H1)(H4)(H_{1})-(H_{4}) hold. In addition, assume that (u0,u1)×L2(Ω)(u_{0},u_{1})\in\mathcal{H}\times L^{2}(\Omega) and h:H01(Ω)Lρ+2ρ+1(Γ1)h:H^{1}_{0}(\Omega)\rightarrow L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) is locally Lipschitz continuous. Then problem (1.1) has a local solution uC(0,T;)C1(0,T;L2(Ω))u\in C(0,T;\mathcal{H})\cap C^{1}(0,T;L^{2}(\Omega)) for some T>0T>0.

Let h(u)=|u|γuh(u)=|u|^{\gamma}u, then h:H01(Ω)Lρ+2ρ+1(Γ1)h:H^{1}_{0}(\Omega)\rightarrow L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}) is locally Lipschitz continuous (see Remark 2.2). Thus by Proposition 3.3, the proof of the local existence statement in Theorem 2.1 is completed.

3.4. Energy identity

It is well known that to prove the uniqueness of weak solutions, we will justify the energy identity (2.13). The energy identity can be derived formally by multiplying (1.1) by utu_{t}. But, such a calculation is not justified, since utu_{t} is not sufficiently regular to be the test function in as required in Definition 2.1. To overcome this problem, we employ the operator Tϵ=(IϵL)1T^{\epsilon}=(I-\epsilon L)^{-1} to smooth function in space, which is mentioned in Appendix A of [13]. We recall important properties of TϵT^{\epsilon} which play an essential role when establishing the energy identity.

Lemma 3.2.

([13]) Let uϵ=Tϵuu^{\epsilon}=T^{\epsilon}u. Then following statements hold.

  1. (1)

    If uL2(Ω)u\in L^{2}(\Omega), then uϵ2u2||u^{\epsilon}||_{2}\leq||u||_{2} and uϵuu^{\epsilon}\rightarrow u in L2(Ω)L^{2}(\Omega) as ϵ0\epsilon\rightarrow 0.

  2. (2)

    If uH01(Ω)u\in H^{1}_{0}(\Omega), then uϵ2u2||\nabla u^{\epsilon}||_{2}\leq||\nabla u||_{2} and uϵuu^{\epsilon}\rightarrow u in H01(Ω)H^{1}_{0}(\Omega) as ϵ0\epsilon\rightarrow 0.

  3. (3)

    If uLp(Γ1)u\in L^{p}(\Gamma_{1}) with 1<p<1<p<\infty, then uϵp,Γ1up,Γ1||u^{\epsilon}||_{p,\Gamma_{1}}\leq||u||_{p,\Gamma_{1}} and uϵuu^{\epsilon}\rightarrow u in Lp(Γ1)L^{p}(\Gamma_{1}) as ϵ0\epsilon\rightarrow 0.

We will now justify the energy identity (2.13). We play the operator TϵT^{\epsilon} on every term of (1.1) and multiply by utϵu^{\epsilon}_{t}. Then we obtain by integrating in space and time

0tΩussϵusϵ𝑑x𝑑s+0tμ(s)Ωguϵ,gusϵg𝑑x𝑑s+0tΓ1Tϵ(q(us))usϵ𝑑Γ𝑑s=0tΩf(x,s)usϵ𝑑x𝑑s+0tΓ1Tϵ(h(u))usϵ𝑑Γ𝑑s.\int^{t}_{0}\int_{\Omega}u^{\epsilon}_{ss}u^{\epsilon}_{s}dxds+\int^{t}_{0}\mu(s)\int_{\Omega}\langle\nabla_{g}u^{\epsilon},\nabla_{g}u^{\epsilon}_{s}\rangle_{g}dxds+\int^{t}_{0}\int_{\Gamma_{1}}T^{\epsilon}(q(u_{s}))u^{\epsilon}_{s}d\Gamma ds\\ =\int^{t}_{0}\int_{\Omega}f(x,s)u^{\epsilon}_{s}dxds+\int^{t}_{0}\int_{\Gamma_{1}}T^{\epsilon}(h(u))u^{\epsilon}_{s}d\Gamma ds. (3.45)

Since uH01(Ω)u\in H^{1}_{0}(\Omega) and utL2(Ω)u_{t}\in L^{2}(\Omega), we have by Lemma 3.2, uϵuu^{\epsilon}\rightarrow u in H01(Ω)H^{1}_{0}(\Omega) and utϵutu^{\epsilon}_{t}\rightarrow u_{t} in L2(Ω)L^{2}(\Omega). Therefore using this convergences, we have

limϵ0(0tΩussϵusϵ𝑑x𝑑s+0tμ(s)Ωguϵ,gusϵg𝑑x𝑑s)=12(ut22+|gu|g22u122|gu0|g22).\lim_{\epsilon\rightarrow 0}\Bigl{(}\int^{t}_{0}\int_{\Omega}u^{\epsilon}_{ss}u^{\epsilon}_{s}dxds+\int^{t}_{0}\mu(s)\int_{\Omega}\langle\nabla_{g}u^{\epsilon},\nabla_{g}u^{\epsilon}_{s}\rangle_{g}dxds\Bigr{)}=\frac{1}{2}\bigl{(}||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}-||u_{1}||^{2}_{2}-||~{}|\nabla_{g}u_{0}|_{g}||^{2}_{2}\bigr{)}. (3.46)

Since ut,q(ut)L2(Γ1)u_{t},q(u_{t})\in L^{2}(\Gamma_{1}), we easily check that

limϵ00tΓ1Tϵ(q(us))usϵ𝑑Γ𝑑s=0tΓ1(q(us))us𝑑Γ𝑑s.\lim_{\epsilon\rightarrow 0}\int^{t}_{0}\int_{\Gamma_{1}}T^{\epsilon}(q(u_{s}))u^{\epsilon}_{s}d\Gamma ds=\int^{t}_{0}\int_{\Gamma_{1}}(q(u_{s}))u_{s}d\Gamma ds. (3.47)

Recall that utLρ+2(Γ1)u_{t}\in L^{\rho+2}(\Gamma_{1}) and h(u)Lρ+2ρ+1(Γ1)h(u)\in L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}). By Lemma 3.2, we have utϵutu^{\epsilon}_{t}\rightarrow u_{t} in Lρ+2(Γ1)L^{\rho+2}(\Gamma_{1}) and Tϵ(h(u))h(u)T^{\epsilon}(h(u))\rightarrow h(u) in Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}). Thus by Lebesgue Dominated Convergence Theorem, we obtain

limϵ00tΓ1Tϵ(h(u))usϵ𝑑Γ𝑑s=0tΓ1(h(u))us𝑑Γ𝑑s.\lim_{\epsilon\rightarrow 0}\int^{t}_{0}\int_{\Gamma_{1}}T^{\epsilon}(h(u))u^{\epsilon}_{s}d\Gamma ds=\int^{t}_{0}\int_{\Gamma_{1}}(h(u))u_{s}d\Gamma ds. (3.48)

Convergences (3.46)-(3.48) permit us to pass to the limit in (3.45), consequently, the energy identity (2.13) holds.

4. Proof of Theorem 2.1 : global existence

In this section we prove that a local weak solution uu on [0,T][0,T] can be extended to [0,)[0,\infty). From the standard continuation argument of ODE theory, it suffices to show that ut22+|gu|g22+uγ+2,Γ1γ+2||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}} is bounded independent of tt. We now consider the following two cases:

4.1. ργ\rho\geq\gamma

Using the energy identity (2.13), we obtain

ddt[12ut22+12μ(t)|gu|g22+1γ+2uγ+2,Γ1γ+2]\displaystyle\frac{d}{dt}\biggl{[}\frac{1}{2}||u_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\biggr{]} (4.1)
=Γ1q(ut)ut𝑑Γ+12μ(t)|gu|g22+2Γ1|u|γuut𝑑Γ+Ωf(t)ut𝑑x.\displaystyle=-\int_{\Gamma_{1}}q(u_{t})u_{t}d\Gamma+\frac{1}{2}\mu^{\prime}(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}+2\int_{\Gamma_{1}}|u|^{\gamma}uu_{t}d\Gamma+\int_{\Omega}f(t)u_{t}dx.

By the same argument as (3.3), we have

Γ1q(ut)ut𝑑Γc3utρ+2,Γ1ρ+2c3meas(Γ1).\int_{\Gamma_{1}}q(u_{t})u_{t}d\Gamma\geq c_{3}||u_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}-c_{3}meas(\Gamma_{1}). (4.2)

Using the Hölder and Young inequalities with γ+1γ+2+1γ+2=1\frac{\gamma+1}{\gamma+2}+\frac{1}{\gamma+2}=1 and the imbedding Lρ+2(Γ1)Lγ+2(Γ1)L^{\rho+2}(\Gamma_{1})\hookrightarrow L^{\gamma+2}(\Gamma_{1}), we deduce that

2Γ1|u|γuut𝑑Γ\displaystyle 2\int_{\Gamma_{1}}|u|^{\gamma}uu_{t}d\Gamma C(ϵ2)uγ+2,Γ1γ+2+ϵ2Cρ+2,γ+2γ+2utρ+2,Γ1γ+2\displaystyle\leq C(\epsilon_{2})||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\epsilon_{2}C^{\gamma+2}_{\rho+2,\gamma+2}||u_{t}||^{\gamma+2}_{\rho+2,\Gamma_{1}} (4.3)
C(ϵ2)uγ+2,Γ1γ+2+ϵ22ρ+1Cρ+2,γ+2γ+2(1+utρ+2,Γ1ρ+2),\displaystyle\leq C(\epsilon_{2})||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\epsilon_{2}2^{\rho+1}C^{\gamma+2}_{\rho+2,\gamma+2}\bigl{(}1+||u_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}\bigr{)},

where Cρ+2,γ+2C_{\rho+2,\gamma+2} is an imbedding constant. By replacing (4.2) and (4.3) in (4.1) and using the Young inequality and (2.8), we get

ddt[12ut22+12μ(t)|gu|g22+1γ+2uγ+2,Γ1γ+2]\displaystyle\frac{d}{dt}\biggl{[}\frac{1}{2}||u_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\biggr{]} (4.4)
12ut22+C(ϵ2)uγ+2,Γ1γ+2+(c3meas(Γ1)+ϵ22ρ+1Cρ+2,γ+2γ+2+12f22)\displaystyle\leq\frac{1}{2}||u_{t}||^{2}_{2}+C(\epsilon_{2})||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+(c_{3}meas(\Gamma_{1})+\epsilon_{2}2^{\rho+1}C^{\gamma+2}_{\rho+2,\gamma+2}+\frac{1}{2}||f||^{2}_{2})
+(ϵ22ρ+1Cρ+2,γ+2γ+2c3)utρ+2,Γ1ρ+2.\displaystyle\hskip 14.22636pt+(\epsilon_{2}2^{\rho+1}C^{\gamma+2}_{\rho+2,\gamma+2}-c_{3})||u_{t}||^{\rho+2}_{\rho+2,\Gamma_{1}}.

Let

E~(t)=12ut22+12μ(t)|gu|g22+1γ+2uγ+2,Γ1γ+2.\widetilde{E}(t)=\frac{1}{2}||u_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

Choosing ϵ2=c32ρ+1Cρ+2,γ+2γ+2\epsilon_{2}=\frac{c_{3}}{2^{\rho+1}C^{\gamma+2}_{\rho+2,\gamma+2}}, we rewrite (4.4) as

E~(t)C10+C11E~(t),\widetilde{E}^{\prime}(t)\leq C_{10}+C_{11}\widetilde{E}(t),

where C10C_{10} and C11C_{11} are positive constants. Now applying Gronwall’s inequality, we have that E~(t)(C12E~(0)+C13)eC12t\widetilde{E}(t)\leq(C_{12}\widetilde{E}(0)+C_{13})e^{C_{12}t}, where C12C_{12} and C13C_{13} are positive constants. Consequently, since E~(0)\widetilde{E}(0) is bounded we conclude that ut22+|gu|g22+uγ+2,Γ1γ+2||u_{t}||^{2}_{2}+||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}} is bounded.

4.2. The potential well

First of all, we will find a stable region. We set

0<K0:=supu,u0(uγ+2,Γ1|gu|g2)<0<K_{0}:=\sup_{u\in\mathcal{H},u\neq 0}\Bigl{(}\frac{||u||_{\gamma+2,\Gamma_{1}}}{||~{}|\nabla_{g}u|_{g}||_{2}}\Bigr{)}<\infty

and the functional

J(u)=μ02|gu|g221γ+2uγ+2,Γ1γ+2,u.J(u)=\frac{\mu_{0}}{2}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}},\hskip 14.22636ptu\in\mathcal{H}. (4.5)

We also define the function, for λ>0\lambda>0,

j(λ)=μ02λ21γ+2K0γ+2λγ+2,j(\lambda)=\frac{\mu_{0}}{2}\lambda^{2}-\frac{1}{\gamma+2}K^{\gamma+2}_{0}\lambda^{\gamma+2}, (4.6)

then

λ0=(μ0K0γ+2)1/γ\lambda_{0}=\Bigl{(}\frac{\mu_{0}}{K^{\gamma+2}_{0}}\Bigr{)}^{1/\gamma}

is the absolute maximum point of jj and

j(λ0)=γμ02(γ+2)λ02=d0.j(\lambda_{0})=\frac{\gamma\mu_{0}}{2(\gamma+2)}\lambda^{2}_{0}=d_{0}.

The energy associated to the problem (1.1) is given by

E(t)=12ut(t)22+12μ(t)|gu(t)|g221γ+2u(t)γ+2,Γ1γ+2,E(t)=\frac{1}{2}||u_{t}(t)||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}, (4.7)

for uu\in\mathcal{H}. By (2.8) and (4.5)-(4.7), we deduce

E(t)J(u(t))μ02|gu(t)|g22K0γ+2γ+2|gu(t)|g2γ+2=j(|gu(t)|g2).E(t)\geq J(u(t))\geq\frac{\mu_{0}}{2}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-\frac{K^{\gamma+2}_{0}}{\gamma+2}||~{}|\nabla_{g}u(t)|_{g}||^{\gamma+2}_{2}=j(||~{}|\nabla_{g}u(t)|_{g}||_{2}). (4.8)
Lemma 4.1.

Let uu be a weak solution for problem (1.1). Suppose that

E(0)<d0and|gu0|g2<λ0.E(0)<d_{0}\hskip 8.53581pt\text{and}\hskip 8.53581pt||~{}|\nabla_{g}u_{0}|_{g}||_{2}<\lambda_{0}.

Then

|gu(t)|g2<λ0for allt0.||~{}|\nabla_{g}u(t)|_{g}||_{2}<\lambda_{0}\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0.
Proof.

It is easy to verify that jj is increasing for 0<λ<λ00<\lambda<\lambda_{0}, decreasing for λ>λ0\lambda>\lambda_{0}, j(λ)j(\lambda)\rightarrow-\infty as λ+\lambda\rightarrow+\infty. Then since d0>E(0)j(|gu0|g2)j(0)=0d_{0}>E(0)\geq j(||~{}|\nabla_{g}u_{0}|_{g}||_{2})\geq j(0)=0, there exist λ0<λ0<λ0~\lambda^{\prime}_{0}<\lambda_{0}<\tilde{\lambda_{0}}, which verify

j(λ0)=j(λ0~)=E(0).j(\lambda^{\prime}_{0})=j(\tilde{\lambda_{0}})=E(0). (4.9)

Considering that E(t)E(t) is nonincreasing, we have

E(t)E(0)for allt0.E(t)\leq E(0)\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0. (4.10)

From (4.8) and (4.9), we deduce that

j(|gu0|g2)E(0)=j(λ0).j(||~{}|\nabla_{g}u_{0}|_{g}||_{2})\leq E(0)=j(\lambda^{\prime}_{0}). (4.11)

Since |gu0|g2<λ0||~{}|\nabla_{g}u_{0}|_{g}||_{2}<\lambda_{0}, λ0<λ0\lambda^{\prime}_{0}<\lambda_{0} and jj is increasing in [0,λ0)[0,\lambda_{0}), from (4.11) it holds that

|gu0|g2λ0.||~{}|\nabla_{g}u_{0}|_{g}||_{2}\leq\lambda^{\prime}_{0}. (4.12)

Next, we will prove that

|gu(t)|g2λ0for allt0.||~{}|\nabla_{g}u(t)|_{g}||_{2}\leq\lambda^{\prime}_{0}\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0. (4.13)

We argue by contradiction. Suppose that (4.13) does not hold. Then there exists time tt^{*} which verifies

|gu(t)|g2>λ0.||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}>\lambda^{\prime}_{0}. (4.14)

If |gu(t)|g2<λ0||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}<\lambda_{0}, from (4.8), (4.9) and (4.14) we can write

E(t)j(|gu(t)|g2)>j(λ0)=E(0),E(t^{*})\geq j(||~{}|\nabla_{g}u(t^{*})|_{g}||_{2})>j(\lambda^{\prime}_{0})=E(0),

which contradicts (4.10).

If |gu(t)|g2λ0||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}\geq\lambda_{0}, then we have, in view of (4.12), that there exists λ0¯\bar{\lambda_{0}} which verifies

|gu0|g2λ0<λ0¯<λ0|gu(t)|g2.||~{}|\nabla_{g}u_{0}|_{g}||_{2}\leq\lambda^{\prime}_{0}<\bar{\lambda_{0}}<\lambda_{0}\leq||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}. (4.15)

Consequently, from the continuity of the function |gu()|g2||~{}|\nabla_{g}u(\cdot)|_{g}||_{2} there exists t¯(0,t)\bar{t}\in(0,t^{*}) verifying

|gu(t¯)|g2=λ0¯.||~{}|\nabla_{g}u(\bar{t})|_{g}||_{2}=\bar{\lambda_{0}}. (4.16)

Then from (4.8), (4.9), (4.15) and (4.16), we get

E(t¯)j(|gu(t¯)|g2)=j(λ0¯)>j(λ0)=E(0),E(\bar{t})\geq j(||~{}|\nabla_{g}u(\bar{t})|_{g}||_{2})=j(\bar{\lambda_{0}})>j(\lambda^{\prime}_{0})=E(0),

which also contradicts (4.10). This completes the proof of Lemma 4.1.

From (4.8) and Lemma 4.1, we arrive at

E(t)J(u(t))>|gu(t)|g22(μ02K0γ+2γ+2λ0γ)=μ0|gu(t)|g22(121γ+2)E(t)\geq J(u(t))>||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}\Bigl{(}\frac{\mu_{0}}{2}-\frac{K^{\gamma+2}_{0}}{\gamma+2}\lambda^{\gamma}_{0}\Bigr{)}=\mu_{0}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}\Bigl{(}\frac{1}{2}-\frac{1}{\gamma+2}\Bigr{)} (4.17)

and, consequently,

J(t)0(J(t)=0iffu=0)and|gu(t)|g222(γ+2)μ0γE(t).J(t)\geq 0~{}~{}(J(t)=0~{}~{}\text{iff}~{}~{}u=0)\hskip 8.53581pt\text{and}\hskip 8.53581pt||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}\leq\frac{2(\gamma+2)}{\mu_{0}\gamma}E(t). (4.18)

By virtue of (4.17), we get

J(u(t))>μ0γ2(γ+2)|gu(t)|g22.J(u(t))>\frac{\mu_{0}\gamma}{2(\gamma+2)}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}. (4.19)

Hence

12ut(t)22+μ0γ2(γ+2)|gu(t)|g22<12||ut(t)||22+J(u(t))E(t)E(0).\frac{1}{2}||u_{t}(t)||^{2}_{2}+\frac{\mu_{0}\gamma}{2(\gamma+2)}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}<\frac{1}{2}||u_{t}(t)||^{2}_{2}+J(u(t))\leq E(t)\leq E(0).

Therefore, there exists a positive constant C14C_{14} independent of tt such that

ut(t)22+|gu(t)|g22C14E(0).||u_{t}(t)||^{2}_{2}+||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}\leq C_{14}E(0). (4.20)

Moreover, if we define the functional I(u(t))I(u(t)) by

I(u(t))=μ0|gu(t)|g22u(t)γ+2,Γ1γ+2,I(u(t))=\mu_{0}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}},

then from the relationship I(u(t))=(γ+2)J(u(t))μ0γ2|gu(t)|g22I(u(t))=(\gamma+2)J(u(t))-\frac{\mu_{0}\gamma}{2}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2} and the strict inequality (4.19), we obtain

I(u(t))>0for allt0.I(u(t))>0\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0. (4.21)

Consequently, from (4.20) and (4.21) we have

ut(t)22+|gu(t)|g22+u(t)γ+2,Γ1γ+2(1+μ0)C14E(0).||u_{t}(t)||^{2}_{2}+||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}+||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\leq(1+\mu_{0})C_{14}E(0).

This it the completion of the proof of the global existence of solutions of (1.1).

5. Proof of Theorem 2.2 : energy decay

In this section we prove the uniform decay rates for the solution of the following problem:

{uttμ(t)Lu=0inΩ×(0,+),u=0onΓ0×(0,+),μ(t)uνL+q(ut)=|u|γuonΓ1×(0,+),u(x,0)=u0(x),ut(x,0)=u1(x),\begin{cases}\vspace{3mm}u_{tt}-\mu(t)Lu=0&\hskip 14.22636pt\text{in}\hskip 14.22636pt\Omega~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u=0&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{0}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}\mu(t)\frac{\partial u}{\partial\nu_{L}}+q(u_{t})=|u|^{\gamma}u&\hskip 14.22636pt\text{on}\hskip 14.22636pt\Gamma_{1}~{}~{}\times~{}~{}(0,+\infty),\\ \vspace{3mm}u(x,0)=u_{0}(x),\hskip 14.22636ptu_{t}(x,0)=u_{1}(x),\end{cases} (5.1)

We consider the following additional hypothesis on HH:

σdiv(H)σ(γ+4)γ+2.\sigma\leq div(H)\leq\frac{\sigma(\gamma+4)}{\gamma+2}. (5.2)

Unless otherwise stated, the constant CC is a generic positive constant, different in various occurrences. We define the energy associated to problem (5.1):

E(t)=12ut22+12μ(t)|gu|g221γ+2uγ+2,Γ1γ+2.E(t)=\frac{1}{2}||u_{t}||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

Then

E(t)=12μ(t)|gu|g22Γ1q(ut)ut𝑑Γ0,E^{\prime}(t)=\frac{1}{2}\mu^{\prime}(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\int_{\Gamma_{1}}q(u_{t})u_{t}d\Gamma\leq 0,

it follows that E(t)E(t) is a nonincreasing function.

First of all, we recall technical lemmas which will play an essential role when establishing the asymptotic behavior.

Lemma 5.1.

([35]) Let E:++E:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+} be a nonincreasing function and ϕ:++\phi:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+} a strictly increasing function of class C1C^{1} such that

ϕ(0)=0andϕ(t)+ast+.\phi(0)=0\hskip 14.22636pt\text{and}\hskip 14.22636pt\phi(t)\rightarrow+\infty\hskip 14.22636pt\text{as}\hskip 14.22636ptt\rightarrow+\infty.

Assume that there exists σ0\sigma\geq 0 and ω>0\omega>0 such that

S+E1+σ(t)ϕ(t)𝑑t1ωEσ(0)E(S)\int^{+\infty}_{S}E^{1+\sigma}(t)\phi^{\prime}(t)dt\leq\frac{1}{\omega}E^{\sigma}(0)E(S)

for all S0S\geq 0. Then EE has the following decay property:

ifσ=0,thenE(t)E(0)e1ωϕ(t),for allt0,\text{if}\hskip 14.22636pt\sigma=0,\hskip 14.22636pt\text{then}\hskip 14.22636ptE(t)\leq E(0)e^{1-\omega\phi(t)},\hskip 14.22636pt\text{for all}\hskip 14.22636ptt\geq 0,
ifσ>0,thenE(t)E(0)(1+σ1+ωσϕ(t))1σ,for allt0.\text{if}\hskip 14.22636pt\sigma>0,\hskip 14.22636pt\text{then}\hskip 14.22636ptE(t)\leq E(0)\Bigl{(}\frac{1+\sigma}{1+\omega\sigma\phi(t)}\Bigr{)}^{\frac{1}{\sigma}},\hskip 14.22636pt\text{for all}\hskip 14.22636ptt\geq 0.
Lemma 5.2.

([35]) Let E:++E:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+} be a nonincreasing function and ϕ:++\phi:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+} a strictly increasing function of class C1C^{1} such that

ϕ(0)=0andϕ(t)+ast+.\phi(0)=0\hskip 14.22636pt\text{and}\hskip 14.22636pt\phi(t)\rightarrow+\infty\hskip 14.22636pt\text{as}\hskip 14.22636ptt\rightarrow+\infty.

Assume that there exists σ>0\sigma>0, σ0\sigma^{\prime}\geq 0 and C>0C>0 such that

S+E1+σ(t)ϕ(t)𝑑tCE1+σ(S)+C(1+ϕ(S))σEσ(0)E(S),0S<+.\int^{+\infty}_{S}E^{1+\sigma}(t)\phi^{\prime}(t)dt\leq CE^{1+\sigma}(S)+\frac{C}{(1+\phi(S))^{\sigma^{\prime}}}E^{\sigma}(0)E(S),\hskip 14.22636pt0\leq S<+\infty.

Then, there exists C>0C>0 such that

E(t)E(0)C(1+ϕ(t))(1+σ)/σ,t>0.E(t)\leq E(0)\frac{C}{(1+\phi(t))^{(1+\sigma^{\prime})/\sigma}},\hskip 14.22636pt\forall t>0.

Let us now multiply equation (5.1) by Ep(t)ϕ(t)uE^{p}(t)\phi^{\prime}(t)\mathcal{M}u, where u\mathcal{M}u is given by

u=2H(u)+(div(H)σ)u,\mathcal{M}u=2H(u)+(div(H)-\sigma)u,

p0p\geq 0 and ϕ:\phi:\mathbb{R}\rightarrow\mathbb{R} is a concave nondecreasing function of class C2C^{2}, such that ϕ(t)+\phi(t)\rightarrow+\infty as t+t\rightarrow+\infty, and then integrate the obtained result over Ω×[S,T]\Omega\times[S,T]. Then we have

0\displaystyle 0 =STEp(t)ϕ(t)Ωu(uttμ(t)Lu)𝑑x𝑑t\displaystyle=\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}\mathcal{M}u\Bigl{(}u_{tt}-\mu(t)Lu\Bigr{)}dxdt (5.3)
=STEp(t)ϕ(t)Ωuttu𝑑x𝑑tSTEp(t)ϕ(t)Ω(div(H)σ)uμ(t)Lu𝑑x𝑑t\displaystyle=\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{tt}\mathcal{M}udxdt-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}(div(H)-\sigma)u\mu(t)Ludxdt
2STEp(t)ϕ(t)ΩH(u)μ(t)Lu𝑑x𝑑t.\displaystyle\hskip 14.22636pt-2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}H(u)\mu(t)Ludxdt.

We note that

STEp(t)ϕ(t)Ωuttu𝑑x𝑑t\displaystyle\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{tt}\mathcal{M}udxdt
=[Ep(t)ϕ(t)Ωutu𝑑x]STST(pEp1(t)E(t)ϕ(t)+Ep(t)ϕ′′(t))Ωutu𝑑x𝑑t\displaystyle=\Bigl{[}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{t}\mathcal{M}udx\Bigr{]}^{T}_{S}-\int^{T}_{S}(pE^{p-1}(t)E^{\prime}(t)\phi^{\prime}(t)+E^{p}(t)\phi^{\prime\prime}(t))\int_{\Omega}u_{t}\mathcal{M}udxdt
2STEp(t)ϕ(t)ΩutH(ut)𝑑x𝑑tSTEp(t)ϕ(t)Ω(div(H)σ)|ut|2𝑑x𝑑t\displaystyle\hskip 14.22636pt-2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{t}H(u_{t})dxdt-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}(div(H)-\sigma)|u_{t}|^{2}dxdt
STEp(t)ϕ(t)Ω(div(H)σ)uμ(t)Lu𝑑x𝑑t\displaystyle-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}(div(H)-\sigma)u\mu(t)Ludxdt
=STEp(t)ϕ(t)μ(t)Γ1(div(H)σ)uuνL𝑑Γ𝑑t+STEp(t)ϕ(t)μ(t)Ω(div(H)σ)|gu|g2𝑑x𝑑t\displaystyle=-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Gamma_{1}}(div(H)-\sigma)u\frac{\partial u}{\partial\nu_{L}}d\Gamma dt+\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Omega}(div(H)-\sigma)|\nabla_{g}u|^{2}_{g}dxdt

and using Lemma 2.1 and the fact H(u)uνL=|gu|g2H(u)\frac{\partial u}{\partial\nu_{L}}=|\nabla_{g}u|^{2}_{g} on Γ0\Gamma_{0},

2STEp(t)ϕ(t)ΩH(u)μ(t)Lu𝑑x𝑑t\displaystyle-2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}H(u)\mu(t)Ludxdt
=STEp(t)ϕ(t)μ(t)Γ1(2uνLH(u)|gu|g2(Hν))𝑑Γ𝑑t\displaystyle=-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Gamma_{1}}\biggl{(}2\frac{\partial u}{\partial\nu_{L}}H(u)-|\nabla_{g}u|^{2}_{g}(H\cdot\nu)\biggr{)}d\Gamma dt
STEp(t)ϕ(t)μ(t)Γ0|gu|g2(Hν)𝑑Γ𝑑t+2STEp(t)ϕ(t)μ(t)ΩDgH(gu,gu)𝑑x𝑑t\displaystyle\hskip 14.22636pt-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Gamma_{0}}|\nabla_{g}u|^{2}_{g}(H\cdot\nu)d\Gamma dt+2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Omega}D_{g}H(\nabla_{g}u,\nabla_{g}u)dxdt
STEp(t)ϕ(t)μ(t)Ω|gu|g2𝑑iv(H)𝑑x𝑑t.\displaystyle\hskip 14.22636pt-\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Omega}|\nabla_{g}u|^{2}_{g}div(H)dxdt.

By replacing above identities in (5.3), we obtain

σSTEp(t)ϕ(t)Ω|ut|2𝑑x𝑑t+2STEp(t)ϕ(t)μ(t)ΩDgH(gu,gu)𝑑x𝑑t\displaystyle\sigma\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}|u_{t}|^{2}dxdt+2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Omega}D_{g}H(\nabla_{g}u,\nabla_{g}u)dxdt (5.4)
σSTEp(t)ϕ(t)μ(t)Ω|gu|g2𝑑x𝑑t(div(H)σ)STEp(t)ϕ(t)Γ1|u|γ+2𝑑Γ𝑑t\displaystyle\hskip 14.22636pt-\sigma\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\mu(t)\int_{\Omega}|\nabla_{g}u|^{2}_{g}dxdt-(div(H)-\sigma)\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u|^{\gamma+2}d\Gamma dt
=[Ep(t)ϕ(t)Ωutu𝑑x]ST+ST(pEp1(t)E(t)ϕ(t)+Ep(t)ϕ′′(t))Ωutu𝑑x𝑑t\displaystyle=-\Bigl{[}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{t}\mathcal{M}udx\Bigr{]}^{T}_{S}+\int^{T}_{S}(pE^{p-1}(t)E^{\prime}(t)\phi^{\prime}(t)+E^{p}(t)\phi^{\prime\prime}(t))\int_{\Omega}u_{t}\mathcal{M}udxdt
+2STEp(t)ϕ(t)Γ1|u|γuH(u)𝑑Γ𝑑t\displaystyle\hskip 14.22636pt+2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u|^{\gamma}uH(u)d\Gamma dt
+STEp(t)ϕ(t)Γ1q(ut)u+(|ut|2μ(t)|gu|g2)(Hν)dΓdt\displaystyle\hskip 14.22636pt+\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}q(u_{t})\mathcal{M}u+\bigl{(}|u_{t}|^{2}-\mu(t)|\nabla_{g}u|^{2}_{g}\bigr{)}(H\cdot\nu)d\Gamma dt
:=I1+I2+I3+I4.\displaystyle:=I_{1}+I_{2}+I_{3}+I_{4}.

Now we are going to estimate terms on the right hand side of (5.4).

EstimateforI1:=[Ep(t)ϕ(t)Ωutu𝑑x]STEstimate~{}~{}for~{}~{}I_{1}:=-\Bigl{[}E^{p}(t)\phi^{\prime}(t)\int_{\Omega}u_{t}\mathcal{M}udx\Bigr{]}^{T}_{S} ;

Using the Young inequality and the inequality

Ω|u|2𝑑xcΩΩ|gu|g2𝑑x,cΩ>0,uH01(Ω),\int_{\Omega}|u|^{2}dx\leq c^{*}_{\Omega}\int_{\Omega}|\nabla_{g}u|^{2}_{g}dx,\hskip 8.53581ptc^{*}_{\Omega}>0,\hskip 8.53581pt\forall u\in H^{1}_{0}(\Omega),

we obtain

|Ωutu𝑑x|CE(t),\Bigl{|}\int_{\Omega}u_{t}\mathcal{M}udx\Bigr{|}\leq CE(t), (5.5)

consequently,

I1C[Ep(t)ϕ(t)E(t)]STCEp+1(S).I_{1}\leq-C\Bigl{[}E^{p}(t)\phi^{\prime}(t)E(t)\Bigr{]}^{T}_{S}\leq CE^{p+1}(S). (5.6)

EstimateforI2:=ST(pEp1(t)E(t)ϕ(t)+Ep(t)ϕ′′(t))Ωutu𝑑x𝑑tEstimate~{}~{}for~{}~{}I_{2}:=\int^{T}_{S}\Bigl{(}pE^{p-1}(t)E^{\prime}(t)\phi^{\prime}(t)+E^{p}(t)\phi^{\prime\prime}(t)\Bigr{)}\int_{\Omega}u_{t}\mathcal{M}udxdt ;

From (5.5), we have

|I2|CST|pEp1(t)E(t)ϕ(t)+Ep(t)ϕ′′(t)|E(t)𝑑tCEp(S)STE(t)dt+CEp+1(S)STϕ′′(t)dtCEp+1(S).|I_{2}|\leq C\int^{T}_{S}|pE^{p-1}(t)E^{\prime}(t)\phi^{\prime}(t)+E^{p}(t)\phi^{\prime\prime}(t)|E(t)dt\\ \leq CE^{p}(S)\int^{T}_{S}-E^{\prime}(t)dt+CE^{p+1}(S)\int^{T}_{S}-\phi^{\prime\prime}(t)dt\leq CE^{p+1}(S). (5.7)

EstimateforI3:=2STEp(t)ϕ(t)Γ1|u|γuH(u)𝑑Γ𝑑tEstimate~{}~{}for~{}~{}I_{3}:=2\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u|^{\gamma}uH(u)d\Gamma dt ;

By the Young inequality with ρ+1ρ+2+1ρ+2=1\frac{\rho+1}{\rho+2}+\frac{1}{\rho+2}=1 and from the fact k(u)=|u|γuk(u)=|u|^{\gamma}u is locally Lipschitz from H01(Ω)H^{1}_{0}(\Omega) into Lρ+2ρ+1(Γ1)L^{\frac{\rho+2}{\rho+1}}(\Gamma_{1}), (2.1) and (4.18) we get

Γ1|u|γuH(u)𝑑Γ\displaystyle\int_{\Gamma_{1}}|u|^{\gamma}uH(u)d\Gamma C(ϵ3)Γ1|u|(γ+1)(ρ+2)ρ+1𝑑Γ+ϵ3Γ1|H(u)|ρ+2𝑑Γ\displaystyle\leq C(\epsilon_{3})\int_{\Gamma_{1}}|u|^{\frac{(\gamma+1)(\rho+2)}{\rho+1}}d\Gamma+\epsilon_{3}\int_{\Gamma_{1}}|H(u)|^{\rho+2}d\Gamma (5.8)
C(ϵ3)Lγρ+2ρ+1u2ρ+2ρ+1+ϵ3supxΩ¯|H|gρ+2Γ1|gu|gρ+2𝑑Γ\displaystyle\leq C(\epsilon_{3})L^{\frac{\rho+2}{\rho+1}}_{\gamma}||\nabla u||^{\frac{\rho+2}{\rho+1}}_{2}+\epsilon_{3}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma
C(ϵ3)Lγρ+2ρ+1c11|gu|g22+ϵ3supxΩ¯|H|gρ+2Γ1|gu|gρ+2𝑑Γ\displaystyle\leq C(\epsilon_{3})L^{\frac{\rho+2}{\rho+1}}_{\gamma}c^{-1}_{1}||~{}|\nabla_{g}u|_{g}||^{2}_{2}+\epsilon_{3}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma
C(ϵ3)Lγρ+2ρ+12(γ+2)μ0γc1E(t)+ϵ3supxΩ¯|H|gρ+2Γ1|gu|gρ+2𝑑Γ,\displaystyle\leq C(\epsilon_{3})L^{\frac{\rho+2}{\rho+1}}_{\gamma}\frac{2(\gamma+2)}{\mu_{0}\gamma c_{1}}E(t)+\epsilon_{3}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma,

consequently,

I3C(ϵ3)Ep+1(S)+ϵ3supxΩ¯|H|gρ+2STEp(t)ϕ(t)Γ1|gu|gρ+2𝑑Γ𝑑t.I_{3}\leq C(\epsilon_{3})E^{p+1}(S)+\epsilon_{3}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma dt. (5.9)

EstimateforI4:=STEp(t)ϕ(t)Γ1q(ut)u+(|ut|2μ(t)|gu|g2)(Hν)dΓdtEstimate~{}~{}for~{}~{}I_{4}:=\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}q(u_{t})\mathcal{M}u+\bigl{(}|u_{t}|^{2}-\mu(t)|\nabla_{g}u|^{2}_{g}\bigr{)}(H\cdot\nu)d\Gamma dt ;

From the Young inequality with ρ+1ρ+2+1ρ+2=1\frac{\rho+1}{\rho+2}+\frac{1}{\rho+2}=1, we have

2Γ1q(ut)H(u)𝑑Γ\displaystyle 2\int_{\Gamma_{1}}q(u_{t})H(u)d\Gamma C(ϵ4)Γ1|q(ut)|ρ+2ρ+1𝑑Γ+ϵ4Γ1|H(u)|ρ+2𝑑Γ\displaystyle\leq C(\epsilon_{4})\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma+\epsilon_{4}\int_{\Gamma_{1}}|H(u)|^{\rho+2}d\Gamma
C(ϵ4)Γ1|q(ut)|ρ+2ρ+1𝑑Γ+ϵ4supxΩ¯|H|gρ+2Γ1|gu|gρ+2𝑑Γ.\displaystyle\leq C(\epsilon_{4})\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma+\epsilon_{4}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma.

Similar arguments as (5.8) we have

(div(H)σ)Γ1q(ut)u𝑑ΓCΓ1|q(ut)|αα1𝑑Γ+CΓ1|u|α𝑑ΓCΓ1|q(ut)|αα1𝑑Γ+CE(t),(div(H)-\sigma)\int_{\Gamma_{1}}q(u_{t})ud\Gamma\leq C\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma+C\int_{\Gamma_{1}}|u|^{\alpha}d\Gamma\leq C\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma+CE(t),

where α=(γ+1)(ρ+2)ρ+1\alpha=\frac{(\gamma+1)(\rho+2)}{\rho+1}. Hence we obtain

I4\displaystyle I_{4} CEp+1(S)+C(ϵ4)STEp(t)ϕ(t)Γ1|q(ut)|ρ+2ρ+1𝑑Γ𝑑t+CSTEp(t)ϕ(t)Γ1|q(ut)|αα1𝑑Γ𝑑t\displaystyle\leq CE^{p+1}(S)+C(\epsilon_{4})\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt+C\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt (5.10)
+CSTEp(t)ϕ(t)Γ1|ut|2𝑑Γ𝑑t+ϵ4supxΩ¯|H|gρ+2STEp(t)ϕ(t)Γ1|gu|gρ+2𝑑Γ𝑑t\displaystyle\hskip 14.22636pt+C\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u_{t}|^{2}d\Gamma dt+\epsilon_{4}\sup_{x\in\overline{\Omega}}|H|^{\rho+2}_{g}\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|\nabla_{g}u|^{\rho+2}_{g}d\Gamma dt
δμ0STEp(t)ϕ(t)Γ1|gu|g2𝑑Γ𝑑t.\displaystyle\hskip 14.22636pt-\delta\mu_{0}\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|\nabla_{g}u|^{2}_{g}d\Gamma dt.

By replacing (5.6), (5.7), (5.9) and (5.10) in (5.4) and choosing ϵ3\epsilon_{3}, ϵ4\epsilon_{4} small enough, we obtain from (5.2)

STEp+1(t)ϕ(t)𝑑t\displaystyle\int^{T}_{S}E^{p+1}(t)\phi^{\prime}(t)dt CEp+1(S)+CSTEp(t)ϕ(t)Γ1|ut|2𝑑Γ𝑑t:=I5\displaystyle\leq CE^{p+1}(S)+C\underbrace{\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u_{t}|^{2}d\Gamma dt}_{:=I_{5}} (5.11)
+CSTEp(t)ϕ(t)Γ1|q(ut)|ρ+2ρ+1𝑑Γ𝑑t:=I6+CSTEp(t)ϕ(t)Γ1|q(ut)|αα1𝑑Γ𝑑t:=I7.\displaystyle\hskip 14.22636pt+C\underbrace{\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt}_{:=I_{6}}+C\underbrace{\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt}_{:=I_{7}}.

Now we are going to estimate the last three terms on the right hand side of (5.11).

5.1. Case 1 : β\beta is linear.

Since β\beta is linear, we can rewrite the hypothesis of qq as follows:

c7|s||q(s)|c8|s|if|s|1,c_{7}|s|\leq|q(s)|\leq c_{8}|s|\hskip 8.53581pt\text{if}\hskip 8.53581pt|s|\leq 1,
c3|s|c3|s|ρ+1|q(s)|c4|s|ρ+1if|s|>1,c_{3}|s|\leq c_{3}|s|^{\rho+1}\leq|q(s)|\leq c_{4}|s|^{\rho+1}\hskip 8.53581pt\text{if}\hskip 8.53581pt|s|>1,

for some positive constants c7c_{7}, c8c_{8}. Hence we get

|ut|1|ut|2𝑑Γc71|ut|1utq(ut)𝑑Γc71E(t),\int_{|u_{t}|\leq 1}|u_{t}|^{2}d\Gamma\leq c^{-1}_{7}\int_{|u_{t}|\leq 1}u_{t}q(u_{t})d\Gamma\leq-c^{-1}_{7}E^{\prime}(t), (5.12)
|ut|>1|ut|2𝑑Γc31|ut|>1utq(ut)𝑑Γc31E(t),\int_{|u_{t}|>1}|u_{t}|^{2}d\Gamma\leq c^{-1}_{3}\int_{|u_{t}|>1}u_{t}q(u_{t})d\Gamma\leq-c^{-1}_{3}E^{\prime}(t), (5.13)
|ut|1|q(ut)|ρ+2ρ+1𝑑Γ|ut|1|q(ut)|2𝑑Γc8|ut|1utq(ut)𝑑Γc8E(t),\int_{|u_{t}|\leq 1}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma\leq\int_{|u_{t}|\leq 1}|q(u_{t})|^{2}d\Gamma\leq c_{8}\int_{|u_{t}|\leq 1}u_{t}q(u_{t})d\Gamma\leq-c_{8}E^{\prime}(t), (5.14)
|ut|>1|q(ut)|ρ+2ρ+1𝑑Γ=|ut|>1|q(ut)|1ρ+1|q(ut)|𝑑Γc41ρ+1|ut|>1utq(ut)𝑑Γc41ρ+1E(t),\int_{|u_{t}|>1}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma=\int_{|u_{t}|>1}|q(u_{t})|^{\frac{1}{\rho+1}}|q(u_{t})|d\Gamma\leq c^{\frac{1}{\rho+1}}_{4}\int_{|u_{t}|>1}u_{t}q(u_{t})d\Gamma\leq-c^{\frac{1}{\rho+1}}_{4}E^{\prime}(t), (5.15)
|ut|1|q(ut)|αα1𝑑Γc8|ut|1utq(ut)𝑑Γc8E(t)\int_{|u_{t}|\leq 1}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma\leq c_{8}\int_{|u_{t}|\leq 1}u_{t}q(u_{t})d\Gamma\leq-c_{8}E^{\prime}(t) (5.16)

and, since ργ\rho\leq\gamma, it holds that ρ+1α11\frac{\rho+1}{\alpha-1}\leq 1. Consequently,

|ut|>1|q(ut)|αα1𝑑Γ=|ut|>1|q(ut)|1α1|q(ut)|𝑑Γc41α1|ut|>1utq(ut)𝑑Γc41α1E(t).\int_{|u_{t}|>1}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma=\int_{|u_{t}|>1}|q(u_{t})|^{\frac{1}{\alpha-1}}|q(u_{t})|d\Gamma\leq c^{\frac{1}{\alpha-1}}_{4}\int_{|u_{t}|>1}u_{t}q(u_{t})d\Gamma\leq-c^{\frac{1}{\alpha-1}}_{4}E^{\prime}(t). (5.17)

From (5.12)-(5.17), we obtain

I5+I6+I7CEp+1(S).I_{5}+I_{6}+I_{7}\leq CE^{p+1}(S). (5.18)

Combining (5.11) and (5.18), it follows that

STEp+1(t)ϕ(t)𝑑tCEp(0)E(S),\int^{T}_{S}E^{p+1}(t)\phi^{\prime}(t)dt\leq CE^{p}(0)E(S),

which implies by Lemma 5.1 with p=0p=0

E(t)E(0)e1ϕ(t)C.E(t)\leq E(0)e^{1-\frac{\phi(t)}{C}}.

Let us set ϕ(t):=mt\phi(t):=mt, where mm is for some positive constant, then ϕ(t)\phi(t) satisfies all the required properties and we obtain that the energy decays exponentially to zero.

5.2. Case 2 : β\beta has polynomial growth near zero.

Assume that β(s)=sρ+1\beta(s)=s^{\rho+1}. Let p=ρ2p=\frac{\rho}{2}, then we rewrite (5.11) as

STEρ2+1(t)ϕ(t)𝑑t\displaystyle\int^{T}_{S}E^{\frac{\rho}{2}+1}(t)\phi^{\prime}(t)dt CE(S)+CSTEρ2(t)ϕ(t)Γ1|ut|2𝑑Γ𝑑t+CSTEρ2(t)ϕ(t)Γ1|q(ut)|ρ+2ρ+1𝑑Γ𝑑t\displaystyle\leq CE(S)+C\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u_{t}|^{2}d\Gamma dt+C\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt (5.19)
+CSTEρ2(t)ϕ(t)Γ1|q(ut)|αα1𝑑Γ𝑑t.\displaystyle\hskip 14.22636pt+C\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt.

By hypotheses on qq and the Hölder inequality with 2ρ+2+ρρ+2=1\frac{2}{\rho+2}+\frac{\rho}{\rho+2}=1, we have

|ut|1|ut|2𝑑Γ|ut|1(utq(ut))2ρ+2𝑑ΓC(|ut|1utq(ut)𝑑Γ)2ρ+2C(E(t))2ρ+2\int_{|u_{t}|\leq 1}|u_{t}|^{2}d\Gamma\leq\int_{|u_{t}|\leq 1}(u_{t}q(u_{t}))^{\frac{2}{\rho+2}}d\Gamma\leq C\biggl{(}\int_{|u_{t}|\leq 1}u_{t}q(u_{t})d\Gamma\biggr{)}^{\frac{2}{\rho+2}}\leq C\bigl{(}-E^{\prime}(t)\bigr{)}^{\frac{2}{\rho+2}}

and

|ut|>1|ut|2𝑑Γc32(ρ+2)|ut|>1(utq(ut))2ρ+2𝑑ΓC(E(t))2ρ+2.\int_{|u_{t}|>1}|u_{t}|^{2}d\Gamma\leq c^{2(\rho+2)}_{3}\int_{|u_{t}|>1}(u_{t}q(u_{t}))^{\frac{2}{\rho+2}}d\Gamma\leq C\bigl{(}-E^{\prime}(t)\bigr{)}^{\frac{2}{\rho+2}}.

Hence

STEρ2(t)ϕ(t)Γ1|ut|2𝑑Γ𝑑t\displaystyle\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|u_{t}|^{2}d\Gamma dt (5.20)
=STEρ2(t)ϕ(t)|ut|1|ut|2𝑑Γ𝑑t+STEρ2(t)ϕ(t)|ut|>1|ut|2𝑑Γ𝑑t\displaystyle=\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{|u_{t}|\leq 1}|u_{t}|^{2}d\Gamma dt+\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{|u_{t}|>1}|u_{t}|^{2}d\Gamma dt
CSTEρ2(t)ϕ(t)(E(t))2ρ+2𝑑t\displaystyle\leq C\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\bigl{(}-E^{\prime}(t)\bigr{)}^{\frac{2}{\rho+2}}dt
ϵ5STEρ2+1(t)ϕ(t)𝑑t+C(ϵ5)E(S).\displaystyle\leq\epsilon_{5}\int^{T}_{S}E^{\frac{\rho}{2}+1}(t)\phi^{\prime}(t)dt+C(\epsilon_{5})E(S).

Similarly as (5.14) and (5.15) we have

|ut|1|q(ut)|ρ+2ρ+1𝑑Γ|ut|1|q(ut)|2𝑑Γ|ut|1(utq(ut))2ρ+2𝑑ΓC(E(t))2ρ+2\int_{|u_{t}|\leq 1}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma\leq\int_{|u_{t}|\leq 1}|q(u_{t})|^{2}d\Gamma\leq\int_{|u_{t}|\leq 1}(u_{t}q(u_{t}))^{\frac{2}{\rho+2}}d\Gamma\leq C\bigl{(}-E^{\prime}(t)\bigr{)}^{\frac{2}{\rho+2}}

and

|ut|>1|q(ut)|ρ+2ρ+1𝑑Γc41ρ+1E(t).\int_{|u_{t}|>1}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma\leq-c^{\frac{1}{\rho+1}}_{4}E^{\prime}(t).

Hence

STEρ2(t)ϕ(t)Γ1|q(ut)|ρ+2ρ+1𝑑Γ𝑑tϵ6STEρ2+1(t)ϕ(t)𝑑t+C(ϵ6)E(S).\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt\leq\epsilon_{6}\int^{T}_{S}E^{\frac{\rho}{2}+1}(t)\phi^{\prime}(t)dt+C(\epsilon_{6})E(S). (5.21)

Similarly as (5.16) and (5.17) we have

|ut|1|q(ut)|αα1𝑑Γ|ut|1|q(ut)|2𝑑ΓC(E(t))2ρ+2\int_{|u_{t}|\leq 1}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma\leq\int_{|u_{t}|\leq 1}|q(u_{t})|^{2}d\Gamma\leq C\bigl{(}-E^{\prime}(t)\bigr{)}^{\frac{2}{\rho+2}}

and

|ut|>1|q(ut)|αα1𝑑Γc41α1E(t).\int_{|u_{t}|>1}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma\leq-c^{\frac{1}{\alpha-1}}_{4}E^{\prime}(t).

Hence

STEρ2(t)ϕ(t)Γ1|q(ut)|αα1𝑑Γ𝑑tϵ7STEρ2+1(t)ϕ(t)𝑑t+C(ϵ7)E(S).\int^{T}_{S}E^{\frac{\rho}{2}}(t)\phi^{\prime}(t)\int_{\Gamma_{1}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt\leq\epsilon_{7}\int^{T}_{S}E^{\frac{\rho}{2}+1}(t)\phi^{\prime}(t)dt+C(\epsilon_{7})E(S). (5.22)

By replacing (5.20)-(5.22) in (5.19) and choosing ϵ5,ϵ6,ϵ7\epsilon_{5},\epsilon_{6},\epsilon_{7} sufficiently small, we get

STEρ2+1(t)ϕ(t)𝑑tCE(S),\int^{T}_{S}E^{\frac{\rho}{2}+1}(t)\phi^{\prime}(t)dt\leq CE(S),

which implies by Lemma 5.1 and choosing ϕ(t)=mt\phi(t)=mt,

E(t)CE(0)(1+t)2ρ.E(t)\leq\frac{CE(0)}{(1+t)^{\frac{2}{\rho}}}.

5.3. Case 3 : β\beta does not necessarily have polynomial growth near zero.

We will use the method of partitions of boundary modified the arguments in [35]. For every t1t\geq 1, we consider the following partitions of boundary depending ϕ(t)\phi^{\prime}(t):

Γ11={xΓ1;|ut(t)|ϕ(t)},Γ12={xΓ1;ϕ(t)<|ut(t)|1},Γ13={xΓ1;|ut(t)|>1}\Gamma_{1}^{1}=\{x\in\Gamma_{1};|u_{t}(t)|\leq\phi^{\prime}(t)\},\hskip 8.53581pt\Gamma_{1}^{2}=\{x\in\Gamma_{1};\phi^{\prime}(t)<|u_{t}(t)|\leq 1\},\hskip 8.53581pt\Gamma_{1}^{3}=\{x\in\Gamma_{1};|u_{t}(t)|>1\}

if ϕ(t)1\phi^{\prime}(t)\leq 1, or

Γ14={xΓ1;|ut(t)|1<ϕ(t)},Γ15={xΓ1;1<|ut(t)|ϕ(t)},Γ16={xΓ1;|ut(t)|>ϕ(t)>1}\Gamma_{1}^{4}=\{x\in\Gamma_{1};|u_{t}(t)|\leq 1<\phi^{\prime}(t)\},\Gamma_{1}^{5}=\{x\in\Gamma_{1};1<|u_{t}(t)|\leq\phi^{\prime}(t)\},\Gamma_{1}^{6}=\{x\in\Gamma_{1};|u_{t}(t)|>\phi^{\prime}(t)>1\}

if ϕ(t)>1\phi^{\prime}(t)>1. Then Γ1=Γ11Γ12Γ13\Gamma_{1}=\Gamma_{1}^{1}\cup\Gamma_{1}^{2}\cup\Gamma_{1}^{3} (or Γ1=Γ14Γ15Γ16\Gamma_{1}=\Gamma_{1}^{4}\cup\Gamma_{1}^{5}\cup\Gamma_{1}^{6}). Let us estimate I5I_{5}, I6I_{6} and I7I_{7} on these partitions.

(I) Part on Γ1i\Gamma_{1}^{i}, i=3,5,6i=3,5,6.

By same arguments as (5.13), (5.15) and (5.17) we get

STEp(t)ϕ(t)Γ1i|ut|2𝑑Γ𝑑tCEp+1(S),\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|u_{t}|^{2}d\Gamma dt\leq CE^{p+1}(S), (5.23)
STEp(t)ϕ(t)Γ1i|q(ut)|ρ+2ρ+1𝑑Γ𝑑tCEp+1(S)\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt\leq CE^{p+1}(S) (5.24)

and

STEp(t)ϕ(t)Γ1i|q(ut)|αα1𝑑Γ𝑑tCEp+1(S).\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt\leq CE^{p+1}(S). (5.25)

(II) Part on Γ12\Gamma_{1}^{2}.

Using the fact β\beta is increasing, ϕ\phi^{\prime} is nonincreasing and (2.10), we obtain

STEp(t)ϕ(t)Γ12|ut|2𝑑Γ𝑑t\displaystyle\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{2}}|u_{t}|^{2}d\Gamma dt STEp(t)β(ϕ(t))Γ12utq(ut)𝑑Γ𝑑t\displaystyle\leq\int^{T}_{S}\frac{E^{p}(t)}{\beta(\phi^{\prime}(t))}\int_{\Gamma_{1}^{2}}u_{t}q(u_{t})d\Gamma dt (5.26)
1β(ϕ(T))STEp(t)(E(t))𝑑t\displaystyle\leq\frac{1}{\beta(\phi^{\prime}(T))}\int^{T}_{S}E^{p}(t)\bigl{(}-E^{\prime}(t)\bigr{)}dt
1β(ϕ(T))Ep+1(S),\displaystyle\leq\frac{1}{\beta(\phi^{\prime}(T))}E^{p+1}(S),
STEp(t)ϕ(t)Γ12|q(ut)|ρ+2ρ+1𝑑Γ𝑑t\displaystyle\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{2}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt STEp(t)Γ12|u(t)||q(ut)|2𝑑Γ𝑑t\displaystyle\leq\int^{T}_{S}E^{p}(t)\int_{\Gamma_{1}^{2}}|u^{\prime}(t)|~{}|q(u_{t})|^{2}d\Gamma dt (5.27)
β1(1)STEp(t)Γ12utq(ut)𝑑Γ𝑑t\displaystyle\leq\beta^{-1}(1)\int^{T}_{S}E^{p}(t)\int_{\Gamma_{1}^{2}}u_{t}q(u_{t})d\Gamma dt
β1(1)Ep+1(S)\displaystyle\leq\beta^{-1}(1)E^{p+1}(S)

and

STEp(t)ϕ(t)Γ12|q(ut)|αα1𝑑Γ𝑑tSTEp(t)Γ12|u(t)||q(ut)|2𝑑Γ𝑑tβ1(1)Ep+1(S).\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{2}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt\leq\int^{T}_{S}E^{p}(t)\int_{\Gamma_{1}^{2}}|u^{\prime}(t)|~{}|q(u_{t})|^{2}d\Gamma dt\leq\beta^{-1}(1)E^{p+1}(S). (5.28)

(III) Part on Γ1i\Gamma_{1}^{i}, i=1,4i=1,4.

Using the fact E(t)E(t) is nonincreasing and β1\beta^{-1} is increasing, we have

STEp(t)ϕ(t)Γ1i|ut|2𝑑Γ𝑑t\displaystyle\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|u_{t}|^{2}d\Gamma dt Ep(S)(β1(ϕ(T)))2STΓ1iϕ(t)(β1(ϕ(t)))2𝑑Γ𝑑t\displaystyle\leq\frac{E^{p}(S)}{(\beta^{-1}(\phi^{\prime}(T)))^{2}}\int^{T}_{S}\int_{\Gamma_{1}^{i}}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}d\Gamma dt (5.29)
meas(Γ1)(β1(ϕ(T)))2Ep(S)STϕ(t)(β1(ϕ(t)))2𝑑t.\displaystyle\leq\frac{meas(\Gamma_{1})}{(\beta^{-1}(\phi^{\prime}(T)))^{2}}E^{p}(S)\int^{T}_{S}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt.

From (2.10), we obtain

STEp(t)ϕ(t)Γ1i|q(ut)|ρ+2ρ+1𝑑Γ𝑑t\displaystyle\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma dt STEp(t)ϕ(t)Γ1i|q(ut)|2𝑑Γ𝑑t\displaystyle\leq\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|q(u_{t})|^{2}d\Gamma dt (5.30)
STEp(t)ϕ(t)Γ1i(β1(|u(t)|))2𝑑Γ𝑑t\displaystyle\leq\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}(\beta^{-1}(|u^{\prime}(t)|))^{2}d\Gamma dt
meas(Γ1)Ep(S)STϕ(t)(β1(ϕ(t)))2𝑑t\displaystyle\leq meas(\Gamma_{1})E^{p}(S)\int^{T}_{S}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt

and

STEp(t)ϕ(t)Γ1i|q(ut)|αα1𝑑Γ𝑑tmeas(Γ1)Ep(S)STϕ(t)(β1(ϕ(t)))2𝑑t.\int^{T}_{S}E^{p}(t)\phi^{\prime}(t)\int_{\Gamma_{1}^{i}}|q(u_{t})|^{\frac{\alpha}{\alpha-1}}d\Gamma dt\leq meas(\Gamma_{1})E^{p}(S)\int^{T}_{S}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt. (5.31)

Therefore from (5.23)-(5.31), we deduce that

I5+I6+I7CEp+1(S)+CEp(S)STϕ(t)(β1(ϕ(t)))2𝑑t.I_{5}+I_{6}+I_{7}\leq CE^{p+1}(S)+CE^{p}(S)\int^{T}_{S}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt. (5.32)

To estimate the last term of the right hand side of (5.32), we need the following additional assumption over ϕ\phi (see [35], p.434):

1ϕ(t)(β1(ϕ(t)))2𝑑tconverges.\int^{\infty}_{1}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt\hskip 8.53581pt\text{converges}.

Then by replacing (5.32) in (5.11) we obtain

STEp+1(t)ϕ(t)𝑑t\displaystyle\int^{T}_{S}E^{p+1}(t)\phi^{\prime}(t)dt CEp+1(S)+CEp(S)S+ϕ(t)(β1(ϕ(t)))2𝑑t\displaystyle\leq CE^{p+1}(S)+CE^{p}(S)\int^{+\infty}_{S}\phi^{\prime}(t)(\beta^{-1}(\phi^{\prime}(t)))^{2}dt (5.33)
CEp+1(S)+CEp(S)ϕ(S)+(β1(1(ϕ1)(s)))2𝑑s.\displaystyle\leq CE^{p+1}(S)+CE^{p}(S)\int^{+\infty}_{\phi(S)}\Bigl{(}\beta^{-1}(\frac{1}{(\phi^{-1})^{\prime}(s)})\Bigr{)}^{2}ds.

Define ψ(t)=1+1t1β(1s)𝑑s\psi(t)=1+\int^{t}_{1}\frac{1}{\beta(\frac{1}{s})}ds, t1t\geq 1. Then ψ\psi is strictly increasing and convex (cf. [35], [39]). We now take ϕ(t)=ψ1(t)\phi(t)=\psi^{-1}(t), then we can rewrite (5.33) as

STEp+1(t)ϕ(t)𝑑tCEp+1(S)+Cϕ(S)Ep(S),\int^{T}_{S}E^{p+1}(t)\phi^{\prime}(t)dt\leq CE^{p+1}(S)+\frac{C}{\phi(S)}E^{p}(S),

which implies, by applying Lemma 5.2 with p=1p=1,

E(t)Cϕ2(t)t>0.E(t)\leq\frac{C}{\phi^{2}(t)}\hskip 14.22636pt\forall t>0.

Let s0s_{0} be a number such that β(1s0)1\beta(\frac{1}{s_{0}})\leq 1. Since β\beta is nondecreasing, we have

ψ(s)1+(s1)1β(1s)1F(1s)ss0,\psi(s)\leq 1+(s-1)\frac{1}{\beta(\frac{1}{s})}\leq\frac{1}{F(\frac{1}{s})}\hskip 14.22636pt\forall s\geq s_{0},

where F(s)=sβ(s)F(s)=s\beta(s), consequently, having in mind that ϕ=ψ1\phi=\psi^{-1}, the last inequality yields

sϕ(1F(1s))=ϕ(t)witht=1F(1s).s\leq\phi\Bigl{(}\frac{1}{F(\frac{1}{s})}\Bigr{)}=\phi(t)\hskip 14.22636pt\text{with}\hskip 14.22636ptt=\frac{1}{F(\frac{1}{s})}.

Then we conclude that

1ϕ(t)F1(1t).\frac{1}{\phi(t)}\leq F^{-1}(\frac{1}{t}).

Therefore the proof of Theorem 2.2 is completed.

6. Proof of Theorem 2.3 : blow-up

This section is devoted to prove the blow-up result. First of all, we introduce a following lemma that is essential role for proving the blow-up.

Lemma 6.1.

Under the hypotheses given in Theorem 2.3 the weak solution to problem (1.1) verifies

|gu(t)|g2>λ0for all0<t<Tmax.||~{}|\nabla_{g}u(t)|_{g}||_{2}>\lambda_{0}\hskip 8.53581pt\text{for all}\hskip 8.53581pt0<t<T_{\max}.
Proof.

We recall the function, for λ>0\lambda>0,

j(λ)=μ02λ21γ+2K0γ+2λγ+2,j(\lambda)=\frac{\mu_{0}}{2}\lambda^{2}-\frac{1}{\gamma+2}K^{\gamma+2}_{0}\lambda^{\gamma+2},

where K0=supu,u0(uγ+2,Γ1|gu|g2)K_{0}=\sup_{u\in\mathcal{H},u\neq 0}\Bigl{(}\frac{||u||_{\gamma+2,\Gamma_{1}}}{||~{}|\nabla_{g}u|_{g}||_{2}}\Bigr{)}. Then

λ0=(μ0K0γ+2)1/γ\lambda_{0}=\Bigl{(}\frac{\mu_{0}}{K^{\gamma+2}_{0}}\Bigr{)}^{1/\gamma}

is the absolute maximum point of jj and

j(λ0)=γμ02(γ+2)λ02=d0.j(\lambda_{0})=\frac{\gamma\mu_{0}}{2(\gamma+2)}\lambda^{2}_{0}=d_{0}.

The energy associated to problem (1.1) is given by

E(t)=12ut(t)22+12μ(t)|gu(t)|g221γ+2u(t)γ+2,Γ1γ+2.E(t)=\frac{1}{2}||u_{t}(t)||^{2}_{2}+\frac{1}{2}\mu(t)||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

We observe that from the definition of jj, we have

E(t)μ02|gu(t)|g221γ+2u(t)γ+2,Γ1γ+2j(|gu(t)|g2)for allt0.E(t)\geq\frac{\mu_{0}}{2}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}-\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\geq j(||~{}|\nabla_{g}u(t)|_{g}||_{2})\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0. (6.1)

Note that jj is increasing for 0<λ<λ00<\lambda<\lambda_{0}, decreasing for λ>λ0\lambda>\lambda_{0}, j(λ)j(\lambda)\rightarrow-\infty as λ+\lambda\rightarrow+\infty.

We will now consider the initial energy E(0)E(0) divided into two cases: E(0)0E(0)\geq 0 and E(0)<0E(0)<0.

Case 1 : E(0)0E(0)\geq 0.

There exist λ1<λ0<λ1\lambda^{\prime}_{1}<\lambda_{0}<\lambda_{1} such that

j(λ1)=j(λ1)=E(0).j(\lambda_{1})=j(\lambda^{\prime}_{1})=E(0). (6.2)

By considering that E(t)E(t) is nonincreasing, we have

E(t)E(0)for allt>0.E(t)\leq E(0)\hskip 8.53581pt\text{for all}\hskip 8.53581ptt>0. (6.3)

From (6.1) and (6.2) we deduce

j(|gu0|g2)E(0)=j(λ1).j(||~{}|\nabla_{g}u_{0}|_{g}||_{2})\leq E(0)=j(\lambda_{1}). (6.4)

Since |gu0|g2>λ0||~{}|\nabla_{g}u_{0}|_{g}||_{2}>\lambda_{0}, λ0<λ1\lambda_{0}<\lambda_{1} and j(λ)j(\lambda) is decreasing for λ0<λ\lambda_{0}<\lambda, from (6.4) we get

|gu0|g2λ1.||~{}|\nabla_{g}u_{0}|_{g}||_{2}\geq\lambda_{1}. (6.5)

Now we will prove that

|gu(t)|g2λ1for all0<t<Tmax||~{}|\nabla_{g}u(t)|_{g}||_{2}\geq\lambda_{1}\hskip 8.53581pt\text{for all}\hskip 8.53581pt0<t<T_{\max} (6.6)

by using the contradiction method. Suppose that (6.6) does not hold. Then there exists t(0,Tmax)t^{*}\in(0,T_{\max}) which verifies

|gu(t)|g2<λ1.||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}<\lambda_{1}. (6.7)

If |gu(t)|g2>λ0||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}>\lambda_{0}, from (6.1), (6.2) and (6.7) we can write

E(t)j(|gu(t)|g2)>j(λ1)=E(0),E(t^{*})\geq j(||~{}|\nabla_{g}u(t^{*})|_{g}||_{2})>j(\lambda_{1})=E(0),

which contradicts (6.3).

If |gu(t)|g2λ0||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}\leq\lambda_{0}, we have, in view of (6.5), that there exists λ¯\bar{\lambda} which verifies

|gu(t)|g2λ0<λ¯<λ1|gu0|g2.||~{}|\nabla_{g}u(t^{*})|_{g}||_{2}\leq\lambda_{0}<\bar{\lambda}<\lambda_{1}\leq||~{}|\nabla_{g}u_{0}|_{g}||_{2}. (6.8)

Consequently, from the continuity of the function |gu()|g2||~{}|\nabla_{g}u(\cdot)|_{g}||_{2} there exists t¯(0,t)\bar{t}\in(0,t^{*}) verifying |gu(t¯)|g2=λ¯||~{}|\nabla_{g}u(\bar{t})|_{g}||_{2}=\bar{\lambda}. Then from the last identity and taking (6.1), (6.2) and (6.8) into account we deduce

E(t¯)j(|gu(t¯)|g2)=j(λ¯)>j(λ1)=E(0),E(\bar{t})\geq j(||~{}|\nabla_{g}u(\bar{t})|_{g}||_{2})=j(\bar{\lambda})>j(\lambda_{1})=E(0),

which also contradicts (6.3).

Case 2 : E(0)<0E(0)<0.

There is λ2>λ0\lambda_{2}>\lambda_{0} such that

j(λ2)=E(0),j(\lambda_{2})=E(0),

consequently, by (6.1) we have

j(|gu0|g2)E(0)=j(λ2).j(||~{}|\nabla_{g}u_{0}|_{g}||_{2})\leq E(0)=j(\lambda_{2}).

From the fact j(λ)j(\lambda) is decreasing for λ0<λ\lambda_{0}<\lambda, we get

|gu0|g2λ2.||~{}|\nabla_{g}u_{0}|_{g}||_{2}\geq\lambda_{2}.

By the same argument as Case 1, we obtain

|gu(t)|g2λ2for all0<t<Tmax.||~{}|\nabla_{g}u(t)|_{g}||_{2}\geq\lambda_{2}\hskip 8.53581pt\text{for all}\hskip 8.53581pt0<t<T_{\max}.

Thus the proof of Lemma 6.1 is completed.

Now we will prove the blow-up result. In order to prove that TmaxT_{\max} is necessarily finite, we argue by contradiction. Assume that the weak solution u(t)u(t) can be extended to the whole interval [0,)[0,\infty).

Let E1E_{1} be a real number such that

E1={0ifE(0)<0,positive constant satisfyingE(0)<E1<d0andE1<E(0)+1ifE(0)0.E_{1}=\begin{cases}0\hskip 8.53581pt&\text{if}\hskip 8.53581ptE(0)<0,\\ \text{positive constant satisfying}\hskip 8.53581ptE(0)<E_{1}<d_{0}\hskip 8.53581pt\text{and}\hskip 8.53581ptE_{1}<E(0)+1\hskip 8.53581pt&\text{if}\hskip 8.53581ptE(0)\geq 0.\end{cases}

By setting G(t):=E1E(t)G(t):=E_{1}-E(t), we have

G(t)=E(t)0,G^{\prime}(t)=-E^{\prime}(t)\geq 0, (6.9)

which implies that G(t)G(t) is nondecreasing, consequently,

0<G0:=E1E(0)<10<G_{0}:=E_{1}-E(0)<1 (6.10)

and from Lemma 6.1, (2.8) and the definition of d0d_{0},

G0G(t)\displaystyle G_{0}\leq G(t) E1μ02|gu(t)|g22+1γ+2u(t)γ+2,Γ1γ+2\displaystyle\leq E_{1}-\frac{\mu_{0}}{2}||~{}|\nabla_{g}u(t)|_{g}||^{2}_{2}+\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}} (6.11)
<d0μ02λ02+1γ+2||u(t)||γ+2,Γ1γ+2\displaystyle<d_{0}-\frac{\mu_{0}}{2}\lambda^{2}_{0}+\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}
1γ+2u(t)γ+2,Γ1γ+2.\displaystyle\leq\frac{1}{\gamma+2}||u(t)||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

We define

M(t)=G1χ¯(t)+τN(t),N(t)=Ωutu𝑑x,M(t)=G^{1-\overline{\chi}}(t)+\tau N(t),\hskip 8.53581ptN(t)=\int_{\Omega}u_{t}udx, (6.12)

where χ¯\overline{\chi} and τ\tau are small positive constants to be chosen later. Then we have

M(t)=(1χ¯)Gχ¯(t)G(t)+τN(t).M^{\prime}(t)=(1-\overline{\chi})G^{-\overline{\chi}}(t)G^{\prime}(t)+\tau N^{\prime}(t). (6.13)

We are now going to analyze the last term on the right-hand side of (6.13).

Lemma 6.2.
N(t)\displaystyle N^{\prime}(t) C15(ut22+uγ+2,Γ1γ+2+G(t)G(t)G0χ¯χGχ¯(t))\displaystyle\geq C_{15}\bigl{(}||u_{t}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+G(t)-G^{\prime}(t)G^{\overline{\chi}-\chi}_{0}G^{-\overline{\chi}}(t)\bigr{)} (6.14)
+μ0(θ21)|gu|g22θE1ζϵ8,\displaystyle\hskip 14.22636pt+\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}},

where C15C_{15} is for some positive constant, 0<χ<γρ(ρ+2)(γ+2)0<\chi<\frac{\gamma-\rho}{(\rho+2)(\gamma+2)}, θ=γ+2ϵ8\theta=\gamma+2-\epsilon_{8} with 0<ϵ8<min{1,γ}0<\epsilon_{8}<\min\{1,\gamma\} and ζ=(γ+1)meas(Γ1)(β1(1))γ+2γ+1γ+2\zeta=\frac{(\gamma+1)meas(\Gamma_{1})\bigl{(}\beta^{-1}(1)\bigr{)}^{\frac{\gamma+2}{\gamma+1}}}{\gamma+2}.

Proof.

Using Eq. (1.1), we obtain

N(t)\displaystyle N^{\prime}(t) =ut22μ(t)|gu|g22+uγ+2,Γ1γ+2Γ1q(ut)u𝑑Γ\displaystyle=||u_{t}||^{2}_{2}-\mu(t)||~{}|\nabla_{g}u|_{g}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}-\int_{\Gamma_{1}}q(u_{t})ud\Gamma (6.15)
(1+θ2)ut22+μ0(θ21)|gu|g22+(1θγ+2)uγ+2,Γ1γ+2+θG(t)θE1\displaystyle\geq\Bigl{(}1+\frac{\theta}{2}\Bigr{)}||u_{t}||^{2}_{2}+\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}+\Bigl{(}1-\frac{\theta}{\gamma+2}\Bigr{)}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\theta G(t)-\theta E_{1}
Γ1q(ut)u𝑑Γ,\displaystyle\hskip 14.22636pt-\int_{\Gamma_{1}}q(u_{t})ud\Gamma,

where θ=γ+2ϵ8\theta=\gamma+2-\epsilon_{8} with 0<ϵ8<min{1,γ}0<\epsilon_{8}<\min\{1,\gamma\}.

We will now estimate the last term on the right-hand side of (6.15). We note that

|Γ1q(ut)u𝑑Γ|Γ1|q(ut)||u|𝑑Γ=|ut|1|q(ut)||u|𝑑Γ+|ut|>1|q(ut)||u|𝑑Γ.\Bigl{|}\int_{\Gamma_{1}}q(u_{t})ud\Gamma\Bigr{|}\leq\int_{\Gamma_{1}}|q(u_{t})|~{}|u|d\Gamma=\int_{|u_{t}|\leq 1}|q(u_{t})|~{}|u|d\Gamma+\int_{|u_{t}|>1}|q(u_{t})|~{}|u|d\Gamma.

By using (2.10) and the imbedding Lγ+2(Γ1)Lρ+2(Γ1)L^{\gamma+2}(\Gamma_{1})\hookrightarrow L^{\rho+2}(\Gamma_{1}), we have

|ut|1|q(ut)||u|𝑑Γ\displaystyle\int_{|u_{t}|\leq 1}|q(u_{t})|~{}|u|d\Gamma (|ut|1|q(ut)|ρ+2ρ+1𝑑Γ)ρ+1ρ+2(|ut|1|u|ρ+2𝑑Γ)1ρ+2\displaystyle\leq\Bigl{(}\int_{|u_{t}|\leq 1}|q(u_{t})|^{\frac{\rho+2}{\rho+1}}d\Gamma\Bigr{)}^{\frac{\rho+1}{\rho+2}}~{}\Bigl{(}\int_{|u_{t}|\leq 1}|u|^{\rho+2}d\Gamma\Bigr{)}^{\frac{1}{\rho+2}} (6.16)
(|ut|1|β1(1)|ρ+2ρ+1𝑑Γ)ρ+1ρ+2uρ+2,Γ1\displaystyle\leq\Bigl{(}\int_{|u_{t}|\leq 1}|\beta^{-1}(1)|^{\frac{\rho+2}{\rho+1}}d\Gamma\Bigr{)}^{\frac{\rho+1}{\rho+2}}~{}||u||_{\rho+2,\Gamma_{1}}
β1(1)(meas(Γ1))γ+1γ+2u|γ+2,Γ1\displaystyle\leq\beta^{-1}(1)\bigl{(}meas(\Gamma_{1})\bigr{)}^{\frac{\gamma+1}{\gamma+2}}~{}||u|||_{\gamma+2,\Gamma_{1}}
(γ+1)meas(Γ1)(β1(1))γ+2γ+1ϵ8(γ+2)+ϵ8γ+1γ+2uγ+2,Γ1γ+2.\displaystyle\leq\frac{(\gamma+1)meas(\Gamma_{1})\bigl{(}\beta^{-1}(1)\bigr{)}^{\frac{\gamma+2}{\gamma+1}}}{\epsilon_{8}(\gamma+2)}+\frac{\epsilon^{\gamma+1}_{8}}{\gamma+2}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}.

On the other hand, by using (2.11), we obtain

|ut|>1|q(ut)||u|𝑑Γ\displaystyle\int_{|u_{t}|>1}|q(u_{t})|~{}|u|d\Gamma c4|ut|>1|ut|ρ+1|u|𝑑Γ\displaystyle\leq c_{4}\int_{|u_{t}|>1}|u_{t}|^{\rho+1}|u|d\Gamma (6.17)
c4(|ut|>1|ut|ρ+2𝑑Γ)ρ+1ρ+2uρ+2,Γ1\displaystyle\leq c_{4}\Bigl{(}\int_{|u_{t}|>1}|u_{t}|^{\rho+2}d\Gamma\Bigr{)}^{\frac{\rho+1}{\rho+2}}||u||_{\rho+2,\Gamma_{1}}
(C(ϵ9)|ut|>1|ut|ρ+2𝑑Γ+ϵ9uγ+2,Γ1(γ+2)(χ+1γ+2)(ρ+2))uγ+2,Γ1(γ+2)χ,\displaystyle\leq\Bigl{(}C(\epsilon_{9})\int_{|u_{t}|>1}|u_{t}|^{\rho+2}d\Gamma+\epsilon_{9}||u||^{(\gamma+2)(\chi+\frac{1}{\gamma+2})(\rho+2)}_{\gamma+2,\Gamma_{1}}\Bigr{)}~{}||u||^{-(\gamma+2)\chi}_{\gamma+2,\Gamma_{1}},

where 0<χ<γρ(ρ+2)(γ+2)0<\chi<\frac{\gamma-\rho}{(\rho+2)(\gamma+2)} and C(ϵ9)C(\epsilon_{9}), ϵ9\epsilon_{9} are for some positive constants. Moreover χ<γρ(ρ+2)(γ+2)\chi<\frac{\gamma-\rho}{(\rho+2)(\gamma+2)} implies that (χ+1γ+2)(ρ+2)<1(\chi+\frac{1}{\gamma+2})(\rho+2)<1. Hence we get

uγ+2,Γ1(γ+2)(χ+1γ+2)(ρ+2){uγ+2,Γ1γ+2ifuγ+2,Γ1γ+2>1,G01G0ifuγ+2,Γ1γ+21.||u||^{(\gamma+2)(\chi+\frac{1}{\gamma+2})(\rho+2)}_{\gamma+2,\Gamma_{1}}\leq\begin{cases}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}&\hskip 8.53581pt\text{if}\hskip 8.53581pt||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}>1,\\ G^{-1}_{0}G_{0}&\hskip 8.53581pt\text{if}\hskip 8.53581pt||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\leq 1.\end{cases}

From (6.10) and (6.11) we have

uγ+2,Γ1(γ+2)(χ+1γ+2)(ρ+2)G01uγ+2,Γ1γ+2||u||^{(\gamma+2)(\chi+\frac{1}{\gamma+2})(\rho+2)}_{\gamma+2,\Gamma_{1}}\leq G^{-1}_{0}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}

and, consequently, from (2.11), (6.9), (6.11) and (6.17),

|ut|>1|q(ut)||u|𝑑Γ\displaystyle\int_{|u_{t}|>1}|q(u_{t})|~{}|u|d\Gamma (C(ϵ9)|ut|>1|ut|ρ+2𝑑Γ+ϵ9G01uγ+2,Γ1γ+2)uγ+2,Γ1(γ+2)χ\displaystyle\leq\Bigl{(}C(\epsilon_{9})\int_{|u_{t}|>1}|u_{t}|^{\rho+2}d\Gamma+\epsilon_{9}G^{-1}_{0}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\Bigr{)}~{}||u||^{-(\gamma+2)\chi}_{\gamma+2,\Gamma_{1}} (6.18)
(C(ϵ9)G(t)+ϵ9G01uγ+2,Γ1γ+2)Gχ(t)\displaystyle\leq\Bigl{(}C(\epsilon_{9})G^{\prime}(t)+\epsilon_{9}G^{-1}_{0}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}\Bigr{)}~{}G^{-\chi}(t)
C(ϵ9)G(t)G0χ¯χGχ¯(t)+ϵ9G0(χ+1)uγ+2,Γ1γ+2,\displaystyle\leq C(\epsilon_{9})G^{\prime}(t)G^{\overline{\chi}-\chi}_{0}G^{-\overline{\chi}}(t)+\epsilon_{9}G^{-(\chi+1)}_{0}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}},

for 0<χ¯<χ0<\overline{\chi}<\chi. From (6.16) and (6.18), we get that

|Γ1q(ut)u𝑑Γ|C(ϵ9)G(t)G0χ¯χGχ¯(t)+(ϵ8γ+1γ+2+ϵ9G0(χ+1))uγ+2,Γ1γ+2+ζϵ8,\Bigl{|}\int_{\Gamma_{1}}q(u_{t})ud\Gamma\Bigr{|}\leq C(\epsilon_{9})G^{\prime}(t)G^{\overline{\chi}-\chi}_{0}G^{-\overline{\chi}}(t)+\Bigl{(}\frac{\epsilon^{\gamma+1}_{8}}{\gamma+2}+\epsilon_{9}G^{-(\chi+1)}_{0}\Bigr{)}||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+\frac{\zeta}{\epsilon_{8}}, (6.19)

where ζ=(γ+1)meas(Γ1)(β1(1))γ+2γ+1γ+2\zeta=\frac{(\gamma+1)meas(\Gamma_{1})\bigl{(}\beta^{-1}(1)\bigr{)}^{\frac{\gamma+2}{\gamma+1}}}{\gamma+2}.

By replacing (6.19) in (6.15) and choosing ϵ9\epsilon_{9} small enough we obtain

N(t)C16(ut22+uγ+2,Γ1γ+2+G(t)G(t)G0χ¯χGχ¯(t))+μ0(θ21)|gu|g22θE1ζϵ8,N^{\prime}(t)\geq C_{16}\bigl{(}||u_{t}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+G(t)-G^{\prime}(t)G^{\overline{\chi}-\chi}_{0}G^{-\overline{\chi}}(t)\bigr{)}+\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}},

where C16C_{16} is a positive constant. Therefore (6.14) follows.

The following Lemma estimates the last three terms on the right-hand side of (6.14).

Lemma 6.3.
μ0(θ21)|gu|g22θE1ζϵ8>0for24ϱζ2ϱϵ8+24ϱζ2ϱ,\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}}>0\hskip 8.53581pt\text{for}\hskip 8.53581pt\frac{\ell-\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}\leq\epsilon_{8}\leq\frac{\ell+\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}, (6.20)

where ϱ=μ0λ022E1\varrho=\frac{\mu_{0}\lambda^{2}_{0}}{2}-E_{1} and =γμ0λ022(γ+2)E1\ell=\frac{\gamma\mu_{0}\lambda^{2}_{0}}{2}-(\gamma+2)E_{1}.

Proof.

From Lemma 5.1 and the definition of θ\theta, we have

μ0(θ21)|gu|g22θE1ζϵ8>μ0(θ21)λ02θE1ζϵ8=(E1μ0λ022)ϵ8ζϵ8+γμ0λ022(γ+2)E1=ϱϵ82ϵ8+ζϵ8:=P(ϵ8).\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}}>\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}\lambda^{2}_{0}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}}\\ =\Bigl{(}E_{1}-\frac{\mu_{0}\lambda^{2}_{0}}{2}\Bigr{)}\epsilon_{8}-\frac{\zeta}{\epsilon_{8}}+\frac{\gamma\mu_{0}\lambda^{2}_{0}}{2}-(\gamma+2)E_{1}=-\frac{\varrho\epsilon^{2}_{8}-\ell\epsilon_{8}+\zeta}{\epsilon_{8}}:=P(\epsilon_{8}). (6.21)

We note that

ϱ=μ0λ022E1>μ0λ022d0=1γ+2K0γ+2λ0γ+2>0\varrho=\frac{\mu_{0}\lambda^{2}_{0}}{2}-E_{1}>\frac{\mu_{0}\lambda^{2}_{0}}{2}-d_{0}=\frac{1}{\gamma+2}K^{\gamma+2}_{0}\lambda^{\gamma+2}_{0}>0

and

=γμ0λ022(γ+2)E1>γμ0λ022(γ+2)d0=0.\ell=\frac{\gamma\mu_{0}\lambda^{2}_{0}}{2}-(\gamma+2)E_{1}>\frac{\gamma\mu_{0}\lambda^{2}_{0}}{2}-(\gamma+2)d_{0}=0.

Since (2.15) holds, we get

24ϱζ0.\ell^{2}-4\varrho\zeta\geq 0.

Therefore, P(ϵ8)P(\epsilon_{8}) represents a curve connecting horizontal axis points 24ϱζ2ϱ\frac{\ell-\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho} and +24ϱζ2ϱ\frac{\ell+\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}, and

P(ϵ8)0for24ϱζ2ϱϵ8+24ϱζ2ϱ.P(\epsilon_{8})\geq 0\hskip 8.53581pt\text{for}\hskip 8.53581pt\frac{\ell-\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}\leq\epsilon_{8}\leq\frac{\ell+\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}.
Refer to caption
Figure 3. The figure of P(ϵ8)P(\epsilon_{8})

Thus we obtain

μ0(θ21)|gu|g22θE1ζϵ8>0for24ϱζ2ϱϵ8+24ϱζ2ϱ.\mu_{0}\Bigl{(}\frac{\theta}{2}-1\Bigr{)}||~{}|\nabla_{g}u|_{g}||^{2}_{2}-\theta E_{1}-\frac{\zeta}{\epsilon_{8}}>0\hskip 8.53581pt\text{for}\hskip 8.53581pt\frac{\ell-\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}\leq\epsilon_{8}\leq\frac{\ell+\sqrt{\ell^{2}-4\varrho\zeta}}{2\varrho}.

Combining (6.13), (6.14), (6.20) and then choosing 0<χ¯<min{12,χ}0<\overline{\chi}<\min\{\frac{1}{2},\chi\} and τ\tau small enough, we obtain

M(t)C17(ut22+uγ+2,Γ1γ+2+G(t)),M^{\prime}(t)\geq C_{17}\bigl{(}||u_{t}||^{2}_{2}+||u||^{\gamma+2}_{\gamma+2,\Gamma_{1}}+G(t)\bigr{)},

where C17C_{17} is a positive constant, which implies that M(t)M(t) is a positive increasing function. By same arguments as p.333 in [16], we have

M(t)C18M11χ¯(t)for allt0,M^{\prime}(t)\geq C_{18}M^{\frac{1}{1-\overline{\chi}}}(t)\hskip 8.53581pt\text{for all}\hskip 8.53581ptt\geq 0,

where C18C_{18} is a positive constant and 1<11χ¯<21<\frac{1}{1-\overline{\chi}}<2. Hence we conclude that M(t)M(t) blows up in finite time and uu also blows up in finite time. Thus this is a contradiction, consequently, the proof of Theorem 2.3 is completed.

Acknowledgments

This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education (2022R1I1A3055309).

References

  • [1] L. Bociu,  Local and global wellposedness of weak solutions for the wave equation with nonlinear boundary and interior sources of supercritical exponents and damping, Nonlinear Anal., 71 (2009) e560–e575.
  • [2] L. Bociu, I. Lasiecka,  Uniqueness of weak solutions for the semilinear wave equations with supercritical boundary/interior sources and damping, Discrete Contin. Dyn. Syst., 22 (2008) 835–860.
  • [3] L. Bociu, M. Rammaha, D. Toundykov,  On a wave equation with supercritical interior and boundary sources and damping terms, Math. Nachr., 284 (16) (2011) 2032–2064.
  • [4] Y. Boukhatem, B. Benabderramane,  Existence and decay of solutions for a viscoelastic wave equation with acoustic boundary conditions, Nonlinear Anal., 97 (2014) 191–209.
  • [5] S. Brenner and L. R. Scott,  The Mathematical Theory of Finite Element Methods, Springer Verlag, 1994.
  • [6] X. Cao and P. Yao  General decay rate estimates for viscoelastic wave equation with variable coefficients, J. Syst. Sci. Complex, 27 (2014) 836–852.
  • [7] M. M. Cavalcanti, V. N. Domingos Cavalcanti, I. Lasiecka,  Well-posedness and optimal decay rates for the wave equation with nonlinear boundary damping-source interaction, J, Differential Equations, 236 (2007) 407–459.
  • [8] M. M. Cavalcanti, V. N. Domingos Cavalcanti, P. Martinez,  Existence and decay rate estimates for the wave equation with nonlinear boundary damping and source term, J. Differential Equations, 203 (2004) 119–158.
  • [9] M. M. Cavalcanti, A. Khemmoudj, M. Medjden,  Uniform stabilization of the damped Cauchy-Ventcel problem with variable coefficients and dynamic boundary conditions, J. Math. Anal. Appl. 328 (2007) 900–930.
  • [10] I. Chueshov, M. Eller and I. Lasiecka,  On the attractor for a semilinear wave equation with critical exponent and nonlinear boundary dissipation, Comm. Partial Differential Equations, 27 (9-10) (2002) 1901–1951.
  • [11] Y. Guo and M. A. Rammaha,  Global existence and decay of energy to systems of wave equations with damping and supercritical sources, Z. Angew. Math. Phys., 64 (3) (2013) 621–658.
  • [12] Y. Guo, M. A. Rammaha and S. Sakuntasathien  Blow-up of a hyperbolic equation of viscoelasticity with supercritical nonlinearities, J. Differential Equations, 262 (2017) 1956–1979.
  • [13] Y. Guo, M. A. Rammaha, S. Sakuntasathien, E. S. Titi and D. Toudykov,  Hadamard well-posedness for a hyperbolic equation of viscoelasticity with supercritical sources and damping, J. Differetinal Equations, 257 (2014) 3778–3812.
  • [14] T. G. Ha,  Asymptotic stability of the semilinear wave equation with boundary damping and source term, C. R. Math. Acad. Sci. Paris Ser. I, 352 (2014) 213–218.
  • [15] T. G. Ha,  Asymptotic stability of the viscoelastic equation with variable coefficients and the Balakrishnan-Taylor damping, Taiwanese J. Math., 22 (4) (2018) 931–948.
  • [16] T. G. Ha,  Blow-up for semilinear wave equation with boudnary damping and source terms, J. Math. Anal. Appl., 390 (2012) 328-334.
  • [17] T. G. Ha,  Blow-up for wave equation with weak boundary damping and source terms, Appl. Math. Lett., 49 (2015) 166-172.
  • [18] T. G. Ha,  Energy decay for the wave equation of variable coefficients with acoustic boundary conditions in domains with nonlocally reacting boundary, Appl. Math. Lett., 76 (2018) 201–207.
  • [19] T. G. Ha,  Energy decay rate for the wave equation with variable coefficients and boundary source term, Appl. Anal., 100 (11) (2021) 2301–2314.
  • [20] T. G. Ha,  General decay estimates for the wave equation with acoustic boundary conditions in domains with nonlocally reacting boundary, Appl. Math. Lett., 60 (2016) 43–49.
  • [21] T. G. Ha,  General decay rate estimates for viscoelastic wave equation with Balakrishnan-Taylor damping, Z. Angew. Math. Phys., 67 (2) (2016) Art. 32, 17 pp.
  • [22] T. G. Ha,  Global existence and general decay estimates for the viscoelastic equation with acoustic boundary conditions, Discrete Contin. Dyn. Syst., 36 (2016) 6899–6919.
  • [23] T. G. Ha,  Global solutions and blow-up for the wave equation with variable coefficients: I. interior supercritical source, Appl. Math. Optim. 84 (suppl. 1) (2021) 767–803.
  • [24] T. G. Ha,  On the viscoelastic equation with Balakrishnan-Taylor damping and acoustic boundary conditions, Evol. Equ. Control Theory, 7 (2) (2018) 281–291.
  • [25] T. G. Ha,  On viscoelastic wave equation with nonlinear boundary damping and source term, Commun. Pur. Appl. Anal., 9 (6) (2010) 1543–1576.
  • [26] T. G. Ha,  Stabilization for the wave equation with variable coefficients and Balakrishnan-Taylor damping, Taiwanese J. Math., 21 (2017) 807–817.
  • [27] T. G. Ha, D. Kim and I. H. Jung,  Global existence and uniform decay rates for the semi-linear wave equation with damping and source terms, Comput. Math. Appl., 67 (2014) 692–707.
  • [28] V. Komornik and E. Zuazua,  A direct method for boundary stabilization of the wave equation, J. Math. Pures Appl., 69 (1990) 33–54.
  • [29] I. Lasiecka,  Mathematical Control Theory of Coupled PDE’s, CBMS-SIAM Lecture Notes, SIAM, Philadelphia, 2002.
  • [30] I. Lasiecka and D. Tataru,  Uniform boundary stabilization of semilinear wave equations with nonlinear boundary damping, Differential and Integral Equations, 6 (3) (1993) 507–533.
  • [31] I. Lasiecka, D. Toundykov,  Energy decay rates for the semilinear wave equation with nonlinear localized damping and source terms, Nonlinear Anal., 64 (2006) 1757–1797.
  • [32] I. Lasiecka, R. Triggiani, P. F. Yao,  Inverse/observability estimates for second-order hyperbolic equations with variable coefficients, J. Math. Anal. Appl., 235 (1999) 13–57.
  • [33] J. Li, S. Chai,  Energy decay for a nonlinear wave equation of variable coefficients with acoustic boundary conditions and a time-varying delay in the boundary feedback, Nonlinear Anal., 112 (2015) 105–117.
  • [34] L. Lu, S. Li,  Higher order energy decay for damped wave equations with variable coefficients, J. Math. Anal. Appl., 418 (2014) 64–78.
  • [35] P. Martinez,  A new method to obtain decay rate estimates for dissipative systems, ESAIM: Control, Optimisation Calc. Var., 4 (1999) 419–444.
  • [36] J. Y. Park, T. G. Ha,  Energy decay for nondissipative distributed systems with boundary damping and source term, Nonlinear Anal., 70 (2009) 2416–2434.
  • [37] J. Y. Park and T. G. Ha,  Existence and asymptotic stability for the semilinear wave equation with boundary damping and source term, J. Math. Phys., 49 (2008) 053511.
  • [38] J. Y. Park, T. G. Ha,  Well-posedness and uniform decay rates for the Klein-Gordon equation with damping term and acoustic boundary conditions, J. Math. Phys., 50 (2009) 013506.
  • [39] J. Y. Park, T. G. Ha, Y. H. Kang,  Energy decay rates for solutions of the wave equation with boundary damping and source term, Z. Angew. Math. Phys., 61 (2010) 235–265.
  • [40] P. Radu,  Weak solutions to the Cauchy problem of a semilinear wave equation with damping and source terms, Adv. Differential Equations, 10 (11) (2005) 1261–1300.
  • [41] I. E. Segal,  Non-linear semigroups, Ann. Math., 78 (1963) 339–364.
  • [42] E. Vitillaro,  On the wave equation with hyperbolic dynamical boundary conditions, interior and boundary damping and supercritical sources, J. Differential Equations, 265 (2018) 4873–4941.
  • [43] E. Vitillaro,  On the wave equation with hyperbolic dynamical boundary conditions, interior and boundary damping and source, Arch. Ration. Mech. Anal., 223 (2017) (3) 1183–1237.
  • [44] J. Wu,  Uniform energy decay of a variable coefficient wave equation with nonlinear acoustic boundary conditions, J. Math. Anal. Appl., 399 (2013) 369–377.
  • [45] P. Yao,   Energy decay for the Cauchy problem of the linear wave equation of variable coefficients with dissipation, Chin. Ann. Math. Ser. B, 31 (1) (2010) 59–70.
  • [46] P. F. Yao,  On the observability inequality for exact controllablility of wave equations with variable coefficients, SIAM J. Control Optim., 37 (1999) 1568–1599.