\documentclass[reqno]{amsart} \usepackage{hyperref} \AtBeginDocument{{\noindent\small \emph{Electronic Journal of Differential Equations}, Vol. 2012 (2012), No. 92, pp. 1--13.\newline ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu \newline ftp ejde.math.txstate.edu} \thanks{\copyright 2012 Texas State University - San Marcos.} \vspace{9mm}} \begin{document} \title[\hfilneg EJDE-2012/92\hfil Blow-up results] {Blow-up results for systems of nonlinear Klein-Gordon equations with arbitrary positive initial energy} \author[S.-T. Wu\hfil EJDE-2012/92\hfilneg] {Shun-Tang Wu} \address{Shun-Tang Wu \newline General Education Center\\ National Taipei University of Technology\\ Taipei, 106 Taiwan} \email{stwu@ntut.edu.tw} \thanks{Submitted February 24, 2012. Published June 7, 2012.} \subjclass[2000]{35L05, 35L15, 35L70} \keywords{Blow-up; life span; damping; nonlinear wave equations; \hfill\break\indent Klein-Gordon equation} \begin{abstract} The initial boundary value problem for a system of nonlinear Klein-Gordon equations in a bounded domain is considered. We prove the existence of local solutions by using a successive approximation method. Then, we show blow-up results with arbitrary positive initial energy by a concavity method. Also estimates for the lifespan of solutions are given. \end{abstract} \maketitle \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{example}[theorem]{Example} \allowdisplaybreaks \section{Introduction} In this article we study the existence and blow-up of local solutions for the system of nonlinear Klein-Gordon equations $$(u_i)_{tt}-\Delta u_i+m_i^2u_i+(u_i)_t=f_i(u)\quad\text{in } \Omega\times[0,T),\; i=1,2, \label{e1.1}$$ with initial conditions $$u(x,0) =\phi(x) ,\quad u_t(x,0) =\varphi(x) ,\quad x\in\Omega, \label{e1.2}$$ and boundary conditions $$u(x,t) =0,\quad x\in\partial\Omega\times(0,T), \label{e1.3}$$ where $u=(u_1,u_2)$, $\phi=(\phi_1,\phi_2)$, $\varphi=(\varphi_1,\varphi_2)$, and $\Omega\subset R^N$, $N\geq1$, is a bounded domain with smooth boundary $\partial\Omega$ so that Divergence theorem can be applied and $T>0$. Let $\Delta=\sum_{j=1}^N\frac{\partial^2}{\partial x_{j}^2}$ be the Laplace operator, $m_i\neq0$ is a real constant and $f_i(u)$ is a nonlinear function of $u$, $i=1,2$. Before stating our results, we first recall the existing results about the initial boundary value problem for a single wave equation $$u_{tt}-\Delta u+a| u_t| ^{m-1}u_t=b|u| ^{p-1}u,\label{e1.4}$$ where $a>0$, $b>0$, $m\geq1$, and $p\geq1$. There are numerous results about the global existence, asymptotic behavior and blow-up of solutions for \eqref{e1.4}. Levine \cite{l1} firstly showed that the solutions with negative initial energy blow up in finite time for equation \eqref{e1.4} with linear damping ($m=1$). Georgiev and Todorova \cite{g2} extended Levine's result to nonlinear case ($m>1$). They showed that solutions with negative initial energy continue to exist globally in time if $m\geq p$ and blow-up in finite time if $p>m$ and the initial energy is sufficiently negative. Later, Levine and Serrin \cite{l4} and Levine, Park, and Serrin \cite{l3} generalized this result to an abstract setting and to unbounded domains. By combining the arguments in \cite{g2} and \cite{l4}, Vitillaro \cite{v1} extended these results to nonlinear damping ($m>1$) and the solution has positive initial energy. Messaoudi \cite{m2} improved the work of \cite{g2} without imposing the condition that energy is sufficiently negative. Similar results have also been established by Todorova \cite{t1,t3} for different Cauchy problems. For related results on a single wave equation, we refer the reader to \cite{m3,p1,z1} and the references therein. On the other hand, Levine and Todorova \cite{l2} proved the local solution blows up in finite time for some initial data with arbitrary high initial energy. Then this result was improved by Todorova and Vitillaro \cite{t2}. However, they did not give a sufficient condition for the initial data such that the corresponding solutions blow up in finite time with arbitrary positive initial energy. Recently, Wang \cite{w1} discussed the blow-up phenomena for equation \eqref{e1.4} with $a=0$. They obtained a sufficient condition of the initial data such that the solution of \eqref{e1.4} blows up in finite time when the positive initial energy is arbitrarily large. Now, we return to the initial boundary problem for the system of nonlinear wave equations as follows $$\begin{gathered} (u_i)_{tt}-\Delta u_i+m_i^2u_i+| ( u_i)_t| ^{p_i-1}(u_i)_t =f_i(u)\quad\text{in }\Omega\times[0,T),\; i=1,2,\\ u(x,0) =\phi(x) ,\quad u_t(x,0) =\varphi(x) ,\quad x\in\Omega,\\ u(x,t) =0,\quad x\in\partial\Omega\times(0,T), \end{gathered}\label{e1.5}$$ where $p_1$, $p_2\geq1$ and $\Omega$ is a bounded domain with smooth boundary. Reed \cite{r1} proposed this interesting problem without imposing damping terms $| (u_i)_t| ^{p-1}(u_i)_t$ in \eqref{e1.5} to describe the interaction of scalar fields $u_1$, $u_2$ of mass $m_1$, $m_2$ respectively. As in the case of a single wave equation, it is worth noting that when the damping terms $| (u_i)_t| ^{p_i-1}( u_i)_t$ is absent, then the force term $f_i(u)$ causes finite blow-up of solution for \eqref{e1.5}. In this direction, Wang \cite{w2} studied \eqref{e1.5} with $f_1(u_1,u_2)=a_1| u_2| ^{q_2+1}| u_1| ^{q_1-1}u_1$ and $f_2(u_1,u_2)=a_2| u_1| ^{q_1+1}| u_2| ^{q_2-1}u_2$ and obtained that the solutions blow up in finite time with arbitrary positive initial energy. On the other hand, if the source term $f_i(u)$ is removed from the equation, then the damping terms should assure global existence and decay of solutions. However, when both damping and source terms are present, then the analysis of their interaction and their influence on the behavior of solutions becomes more difficult. Agre and Rammaha \cite{a1} considered \eqref{e1.5} with $$\begin{gathered} f_1(u_1,u_2)=(r+1)\Big[ a| u_1+u_2| ^{r-1}(u_1+u_2)+b| u_1| ^{\frac{r-3}{2}}| u_2| ^{\frac{r+1}{2}}u_1\Big] ,\\ f_2(u_1,u_2)=(r+1)\Big[ a| u_1+u_2| ^{r-1}(u_1+u_2)+b| u_2| ^{\frac{r-3}{2}}| u_1| ^{\frac{r+1}{2}}u_2\Big] , \end{gathered}\label{e1.6}$$ where $r\geq3$, $a>1$ and $b>0$. They showed the existence of global solutions if $r\leq\min\{ p_1,p_2\}$ and proved the blow-up of solutions if $r>\min\{ p_1,p_2\}$ and initial energy is negative. Later, Alves et al \cite{a2} improved these results and they obtained several results on the global, uniform decay rates, and blow up of solutions in finite time when the initial energy is nonnegative by involving the Nehari manifold. Recently, Li and Tsai \cite{l5} considered a class of nonlinear terms which includes \eqref{e1.6} in a bounded domain where the global existence and blow-up behavior of solutions \ without imposing damping terms were discussed. However, on considering the blow-up properties, the initial energy can not be arbitrarily large in that paper. This motivates us to consider the problem of how to obtain the blow-up of solutions when the initial energy is arbitrarily large. Inspired by these previous works \cite{l5,w1,w2}, in this present paper, we would like to investigate the local existence and then establish a sufficient condition of the initial data with arbitrarily high initial energy such that the corresponding local solution of the system for the nonlinear Klein-Gordon equations \eqref{e1.1}-\eqref{e1.3} blows up in finite time. The method used here are the successive approximation method and the concavity method. In this way, we can extend the result of \cite{w1} to a system with linear damping terms and the result of \cite{l5} without setting any restriction on upper bound of the initial energy. The paper is organized as follows. In section 2, we first introduced some notations used throughout this paper and then state the local existence Theorem \ref{thm2.4}. In section 3, we prove the main result Theorem \ref{thm3.4} which shows blow-up properties of solutions with highly positive initial energy. \section{Exitance of local solutions} In this section we shall discuss the existence of local solutions for \eqref{e1.1}-\eqref{e1.3} by the method of successive approximations. First we give the notation which will be used throughout the paper. Let $W^{m,p}(\Omega)$ be the usual Sobolev space. Specially, $W^{m,2}(\Omega)$ and $W^{0,p}(\Omega)$ will be marked by $H^{m}(\Omega)$ and $L^p(\Omega)$, respectively. And we denote $\| \cdot\|_p$ to be $L^p$-norm for $1\leq p\leq\infty$. $H_0^1(\Omega)$ is the closure of $C_0^{\infty}(\Omega)$ with respect to the norm $\| u\|_{H_0^1}=\| \nabla u\|_2$. Define \begin{gather*} H1=C^1([0,T];L^2(\Omega))\cap C^{0}([0,T];H_0^1 (\Omega)), \\ H2=C^2([0,T];L^2(\Omega))\cap C^1([0,T];H_0^1 (\Omega)),\text{ for }T>0. \end{gather*} Now, we make the following assumptions: \begin{itemize} \item[(A1)] $f_i :R^2\to \mathbb{R}$ is continuously differentiable such that for each $u=(u_1,u_2)\in H_0^1(\Omega)\times H_0^1(\Omega)$, we have $u_if_i\in L^1(\Omega)$, $i=1,2$ and $F(u)\in L^1(\Omega)$, where $F(u)=\int_0^{u_1}f_1(s,u_2)ds+\int_0^{u_2}f_2(0,s)ds.$ \item[(A2)] $f_i(0)=0$ and for any $\rho>0$ there exists a constant $k(\rho)>0$ such that $\| f_i(u)-f_i(v)\|_2\leq k(\rho)\|u-v\|_{H_0^1\times H_0^1},\quad i=1,2,$ where $u,v\in H_0^1(\Omega) \times H_0^1( \Omega)$ with $\| u\|_{H_0^1\times H_0^1 },\| v\|_{H_0^1\times H_0^1}$. \item[(A3)] $\frac{\partial f_1}{\partial u_2}=\frac{\partial f_2}{\partial u_1}.$ \end{itemize} Note that the function of the form $f_1(u_1,u_2)=u_1^{s-1}u_2 ^{s}+u_1^p$, $f_2(u_1,u_2)=u_2^{s-1}u_1^{s}+u_2^{q}$ satisfy the assumptions (A1)-(A3) where $11$ for $N=1,2$. \begin{lemma}[Sobolev-Poincar\'{e} \cite{m1}] \label{lem2.1} Let $2\leq p\leq\frac{2N}{N-2}$. then the inequality $\| u\|_p\leq c_{s}\| \nabla u\| _2,\text{\ \textit{for} }u\in H_0^1(\Omega),$ holds for some positive constant $c_{s}$. \end{lemma} \begin{lemma}[\cite{g1}] \label{lem2.2} Let $\delta\geq0$, $T>0$ and $h$ be a Lipschitizan function over $[0,T)$. Assume that $h(0)\geq0$ and $h'(t)+\delta h(t)>0$ for a.e. $t\in(0,T)$. Then $h(t)>0$ for all $t\in(0,T)$. \end{lemma} Before proving the existence theorem for nonlinear equations \eqref{e1.1}-\eqref{e1.3}, we need the existence result for a linear wave equation which is given in \cite{h1}. \begin{lemma} \label{lem2.3} Assume that $f\in W^{1,1}([0,T];L^2(\Omega))$ and that $u_0\in H^2(\Omega)\cap H_0^1(\Omega)$ and $u_1\in H_0^1 (\Omega)$, then the linear problem with damping \begin{gather*} u_{tt}-\Delta u+u_t =f(t,x),\\ u(0) =u_0,\quad u_t(0) =u_1, \quad x\in\Omega,\\ u(x,t) =0,\quad x\in\partial\Omega\times(0,T), \end{gather*} has a unique solution $u\in H2$. \end{lemma} \begin{theorem} \label{thm2.4} Assume that the assumptions {\rm (A1)--(A3)} hold and let $(\phi_1,\phi_2) \in H_0^1(\Omega)\times H_0^1(\Omega)$ and $(\varphi_1,\varphi_2) \in L^2(\Omega)\times L^2(\Omega)$. Then problem \eqref{e1.1}--\eqref{e1.3} admits a unique solution $(u_{1,}u_2)$ in $H1\times H1$. \end{theorem} \begin{proof} Since $H^2(\Omega)\cap H_0 ^1(\Omega)$ is dense in $H_0^1(\Omega)$ and $H_0^1 (\Omega)$ is dense in $L^2(\Omega)$, it suffices to consider problem \eqref{e1.1}--\eqref{e1.3} for $\phi_i\in H^2(\Omega)\cap H_0^1(\Omega)$ and $\varphi_i\in H_0^1(\Omega)$, $i=1,2$. Let $\{ u^{m}=(u_1^{m},u_2^{m})\}_{m\geq1}$ be a sequence of solutions obtained by considering the approximation problem $$\label{e2.1} \begin{gathered} (u_i^{m+1})_{tt}-\Delta u_i^{m+1}+(u_i ^{m+1})_t =-m_i^2u_i^{m}+f_i(u^{m}),\quad i=1,\quad 2,\\ u^{m+1}(x,0) =\phi(x),\quad u_t^{m+1}(x,0)=\varphi(x),\quad x\in\Omega,\\ u^{m+1}(x,t) =0,\quad x\in\partial\Omega\times(0,T), \end{gathered}$$ with the initial function $u^1(x,0)=\phi(x)$. Using Lemma \ref{lem2.3} and (A1)--(A2), we see that \eqref{e2.1} has a unique solution $u^{m}\in H2\times H2$. In the following, we would like to estimate the solution obtained above. Multiplying by $(u_i^{m+1})_t$ on both sides of \eqref{e2.1} and then integrating it over $\Omega$, we have \begin{align*} & \int_{\Omega}(u_i^{m+1})_t[(u_i ^{m+1})_{tt}-\Delta u_i^{m+1}+(u_i^{m+1})_t]dx\\ & =\int_{\Omega}(u_i^{m+1})_t[-m_i^2u_i ^{m}+f_i(u^{m}) ]dx. \end{align*} Using the Divergence theorem and H\"{o}lder inequality, we obtain $$\frac{d}{dt}\| Du_i^{m+1}\|_2\leq\| m_i ^2u_i^{m}+f_i(u^{m}) \|_2, \label{e2.2}$$ where $D\equiv(\partial_t,\nabla_{x})$ and $\|Du_i\|_2^2=\int_{\Omega}(| (u_i)_t| ^2+| \nabla u_i|^2) dx$. Integrating \eqref{e2.2} from 0 to $t$, we obtain $$\| Du_i^{m+1}\|_2(t)\leq\| Du_i^{m+1} \|_2(0)+\int_0^{t}\| m_i^2u_i^{m}+f_i( u_1^{m},u_2^{m}) \|_2(r)dr. \label{e2.3}$$ For simplicity, we denote \begin{gather} \beta_i =\| Du_i^{m+1}\| (0)=(\| \varphi_i\|_2^2+\| \phi_i\|_2 ^2) ^{1/2},\quad i=1,2,\nonumber \\ \beta =\beta_1+\beta_2,\label{e2.4}\\ G_{m,i} =m_i^2\| u_i^{m}\|_2+\| f_i(u_1^{m},u_2^{m}) \|_2,\quad i=1,2,\; m\geq1,\label{e2.5}\\ H^{k}(t) =\| Du^{k}\|_2(t) =(\|Du_1^{k}\|_2+\| Du_2^{k}\|_2) (t),\quad k\geq1, \label{e2.6} \end{gather} where $Du^{k}=(Du_1^{k},Du_2^{k})$. Then using Lemma \ref{lem2.1} and (A2), we have $$G_{m,1}+G_{m,2}\leq c\| Du^{m}\|_2(t), \label{e2.7}$$ here $c$ is some positive constant. It follows from \eqref{e2.3}-\eqref{e2.5} that $$\| Du_i^2\|_2(t) \leq\beta_i+\int_0^{t} m_i^2\| \phi_i\|_2+\| f_i(\phi) \|_2dt \leq\beta_i+G_{1,i}t. \label{e2.8}$$ Thus by \eqref{e2.6} and \eqref{e2.8}, we obtain $$H^2(t)\leq\beta+ct\| Du^1\|_2(t). \label{e2.9}$$ Define $$K_{\infty,\tau}(u^{i}) =\sup\{ \| Du^{i} \|_2(t)\mid0\leq t\leq\tau\} , \label{e2.10}$$ and take a constant $M>\beta$. Then $H^1(t)\leq M$, and hence $K_{\infty ,\tau}(u^1) \leq M$. Therefore, from \eqref{e2.9}, we see that $H^2(t)\leq\beta+ctM\leq M,$ provided that $\tau=(M-\beta)/(cM)$. That is, $K_{\infty,\tau}(u^2) \leq M$. Suppose that $K_{\infty,\tau}(u^{m}) \leq M$, then, using \eqref{e2.3}, \eqref{e2.5}, \eqref{e2.7} and \eqref{e2.10}, we obtain \begin{aligned} H^{m+1}(t) & \leq\beta+\int_0^{t}(G_{m,1}+G_{m,2})(r)dr\\ & \leq\beta+\int_0^{t}c\| Du^{m}\|_2(r)dr\\ & \leq\beta+cK_{\infty,\tau}(u^{m}) t\leq M,\quad 0\leq t\leq\tau. \end{aligned}\label{e2.12} Thus $K_{\infty,\tau}(u^{m+1}) \leq M$. Hence, we have $$K_{\infty,\tau}(u^{m}) \leq M,\text{ for all }m\geq1. \label{e2.13}$$ Below we shall show that $\{ u^{m}\}_{m\geq1}$ is a Cauchy sequence in $H1\times H1$. Let $z^{m}=u^{m+1}-u^{m}$. From \eqref{e2.1}, for $i=1,2$, we have $$\label{e2.14} \begin{gathered} (z_i^{m})_{tt}-\Delta z_i^{m}+(z_i^{m}) _t =-m_i^2z_i^{m-1}+f_i(u^{m}) -f_i( u^{m-1}) ,\\ z^{m}(x,0) =0,\quad z_t^{m}(x,0)=0,\quad x\in\Omega,\\ z^{m}(x,t) =0,\quad x\in\partial\Omega\times(0,T). \end{gathered}$$ As in the previous arguments, we obtain $$\begin{split} &\| Dz^{m}\|_2(t)\\ &\leq\| Dz^{m}\|_2(0)+\sum_{i=1}^2 \int_0^{t}(m_i^2\| z_i^{m-1}\|_2+\| f_i(u^{m}) -f_i(u^{m-1}) \|_2)dr. \end{split}\label{e2.15}$$ From \eqref{e2.14}, we obtain $\| Dz^{m}\|_2(0)=0$. Then, by \eqref{e2.13}, Lemma \ref{lem2.1} and (A2), we have $\| Dz^{m}\|_2(t)\leq L\int_0^{t}\| Dz^{m-1}\|_2(r)dr,\quad 0\leq t\leq\tau,$ where $L$ is a constant depending on $m_1$, $m_2$ and Sobolev constant. Thus by induction, we obtain $$K_{\infty,\tau}(z^{m}) \leq L\tau K_{\infty,\tau}( z^{m-1}) \leq\cdots\leq(L\tau) ^{m-1}K_{\infty,\tau }(z^1) . \label{e2.16}$$ Therefore, for any positive integer $p$ and $L\tau\in(0,1)$, we see that \begin{align*} K_{\infty,\tau}(u^{m+p}-u^{m}) & \leq((L\tau) ^{m+p-2}+\cdots+(L\tau) ^{m-1}) K_{\infty,\tau}(u^2-u^1) \\ & \leq\frac{(L\tau) ^{m-1}}{1-L\tau}K_{\infty,\tau}( u^2-u^1) \to0 \quad\text{ as }m\to\infty. \end{align*} Hence, the Cauchy sequence $\{ u^{m}\}_{m\geq1}$ converges in $H1\times H1$ and the limit function $u=\lim_{m\to\infty}u^{m}$ in $H1\times H1$ is a solution defined on $[0,\tau)$ for problem \eqref{e1.1}--\eqref{e1.3}. \noindent\textbf{Uniqueness.} Let $u$ and $\widehat{u}$ be two solutions defined on $[0,T)$ of problem \eqref{e1.1}-\eqref{e1.3}. Set $w=u-\widehat{u}$. From \eqref{e1.1}, we have \begin{gather*} (w_i)_{tt}-\Delta w_i+(w_i)_t =-m_i^2w_i+f_i(u) -f_i(\widehat{u}) ,\quad i=1, 2\\ w(x,0) =0,\quad w_t(x,0)=0,\quad x\in\Omega,\\ w(x,t) =0,\quad x\in\partial\Omega\times(0,T). \end{gather*} Similar to \eqref{e2.15}, we obtain $\| Dw\|_2^2(t)\leq\| Dw\|_2 ^2(0)+c\int_0^{t}\| Dw\|_2^2(r)dr.$ The Gronwall's inequality implies $\| Dw\|_2^2(t)=0,\text{ for }0\leq t0 such that \[ u_1f_1(u)+u_2f_2(u)\geq(2+4\delta)F(u),\quad \text{for all }u_1,u_2\in \mathbb{R},$ where $F(u)$ is given in (A1). \end{itemize} \noindent\textbf{Definition.} A solution $(u_1(t),u_2(t))$ of \eqref{e1.1}-\eqref{e1.3} is said to blow up if there exists a finite time $T$ such that $$\lim_{t\to T^{-}}(\Vert u_1(t)\Vert_2^2+\Vert u_2(t)\Vert_2^2) =\infty. \label{e3.1}$$ Let $(u_1(t),u_2(t))$ be the solution of \eqref{e1.1}-\eqref{e1.3}, we define the energy function $$E(t)=\frac{1}{2} \sum_{i=1}^2 [\| (u_i)_t\|_2^2+\Vert\nabla u_i\Vert_2^2+\Vert u_i\Vert_2^2]-\int_{\Omega}F(u)dx,\quad t\geq0 \label{e3.2}$$ and $$I(u(t))\equiv I(t)= \sum_{i=1}^2 [\Vert\nabla u_i\Vert_2^2+\Vert u_i\Vert_2^2]-\int_{\Omega} \sum_{i=1}^2 u_if_i(u)dx. \label{e3.3}$$ \begin{lemma} \label{lem3.1} Let $u$ be a solution of \eqref{e1.1}-\eqref{e1.3}. Then $E(t)$ is a nonincreasing function and $$E(t) =E(0) -\int_0^{t} \sum_{i=1}^2 \| (u_i)_t\|_2^2dt. \label{e3.4}$$ \end{lemma} \begin{proof} By differentiating \eqref{e3.2} and using \eqref{e1.1}-\eqref{e1.3}, (A1) and (A3), we obtain $\frac{dE(t)}{dt}=- \sum_{i=1}^2 \| (u_i)_t\|_2^2.$ Thus, the result of Lemma \ref{lem3.1} follows. \end{proof} \begin{lemma} \label{lem3.2} Assume {\rm (A4)} and that $(\phi_1,\varphi_1), (\phi_2,\varphi_2)\in H_0^1(\Omega)\times L^2 (\Omega)$ satisfy $E(0)>0$, $I(0)<0$, \begin{gather} \Vert\phi_1\Vert_2^2+\Vert\phi_2\Vert_2^2 >\frac{1+2\delta }{\delta}E(0),\label{e3.5}\\ \int_{\Omega}(\phi_1\varphi_1+\phi_2\varphi_2) dx >0. \label{e3.6} \end{gather} Then $\Vert u_1(t)\Vert_2^2+\Vert u_2(t)\Vert_2^2>\frac{1+2\delta }{\delta}E(0)\quad \text{and}\quad I(t)<0,$ for all $t\in[0,T)$. \end{lemma} \begin{proof} First, we prove that $I(t)<0$, for all $t\in[0,T)$. Suppose not, then there exists $T^{\ast}>0$ such that $T^{\ast}=\min\{ t\in[0,T);I(t)=0\}$. We define $G(t)=\int_{\Omega}(u_1^2(x,t)+u_2^2(x,t)) dx.$ Using \eqref{e1.1}, we have \begin{gather*} G'(t)=2\int_{\Omega} \sum_{i=1}^2 u_i(u_i)_tdx, \\ G''(t)=2\int_{\Omega} \sum_{i=1}^2 ((u_i)_t^2-| \nabla u_i| ^2-u_i^2+u_if_i(u)) dx-2\int_{\Omega} \sum_{i=1}^2 u_i(u_i)_tdx. \end{gather*} Then, from \eqref{e3.3} it follows that $$G''(t)+G'(t)=2[ \sum_{i=1}^2 \int_{\Omega}(u_i)_t^2dx-I(t)]>0, \label{e3.7}$$ for all $t\in[0,T^{\ast})$. By Lemma \ref{lem2.2} and \eqref{e3.6}, we obtain $G'(t)>0$, for all $t\in[0,T^{\ast})$. This implies $G(t)$ is strictly increasing on $[0,T^{\ast})$. Thus, from \eqref{e3.5}, we have $G(t)>G(0)>\frac{1+2\delta}{\delta}E(0),$ for all $t\in(0,T^{\ast})$. From the continuity of $u(t)$ at $t=T^{\ast}$, we see that $$G(T^{\ast})=\sum_{i=1}^2 \| u_i(T^{\ast})\|_2^2>\frac{1+2\delta}{\delta}E(0). \label{e3.8}$$ On the other hand, from \eqref{e3.2} and Lemma \ref{lem3.1}, we have $$\begin{split} &\sum_{i=1}^2 \big(\Vert\nabla u_i(T^{\ast})\Vert_2^2+\| u_i(T^{\ast})\|_2^2\big) -2\int_{\Omega}F(u_1(T^{\ast}),u_2(T^{\ast}))dx\\ &\leq 2E(T^{\ast})\leq2E(0). \end{split} \label{e3.9}$$ Noting that from the assumption $I(T^{\ast})=0$ and (A4) give us $$\sum_{i=1}^2 \big(\Vert\nabla u_i(T^{\ast})\Vert_2^2+\| u_i(T^{\ast})\|_2^2\big) \geq (2+4\delta)\int_{\Omega}F(u_1(T^{\ast}),u_2(T^{\ast}))dx, \label{e3.10}$$ which together with \eqref{e3.9} implies $\sum_{i=1}^2 \big(\Vert\nabla u_i(T^{\ast})\Vert_2^2+\| u_i(T^{\ast })\|_2^2\big) \leq\frac{1+2\delta}{\delta}E(0).$ It is a contradiction to \eqref{e3.8}. Hence, $I(t)<0$, for all $t\in[0,T)$. Therefore, following the same arguments as above, we deduce that $G(t)$ is strictly increasing on $[0,T)$ and $\Vert u_1(t)\Vert_2^2+\Vert u_2(t)\Vert_2^2>\frac{1+2\delta}{\delta}E(0),$ for all $t\in[0,T)$. \end{proof} Now, let $$a(t) =\sum_{i=1}^2 \Big(\int_{\Omega}u_i^2dx+\int_0^{t}\Vert u_i\Vert_2 ^2dt\Big) ,\quad t\geq0. \label{e3.11}$$ We need the following lemma to derive our result. \begin{lemma} \label{lem3.3}. Assume that {\rm (A1), (A3) (A4)} hold. Then $$a''(t) \geq4(\delta+1) \int_{\Omega} \sum_{i=1}^2 (u_i)_t^2dx+(4+8\delta) \int_0^{t} \sum_{i=1}^2 \Vert(u_i)_t\Vert_2^2dt. \label{e3.12}$$ \end{lemma} \begin{proof} Form \eqref{e3.11} and using \eqref{e1.1}, we have $$a'(t) =\sum_{i=1}^2 \Big(\int_{\Omega}2u_i(u_i)_tdx+\Vert u_i\Vert _2^2\Big) , \label{e3.13}$$ and $$a''(t) =2 \sum_{i=1}^2 \Big(\int_{\Omega}(u_i)_t^2dx-\| \nabla u_i\|_2^2-\| u_i\|_2^2\Big) +2\int_{\Omega} \sum_{i=1}^2 u_if_i(u)dx. \label{e3.14}$$ Employing \eqref{e3.2} , \eqref{e3.4} and (A4), we obtain \begin{align*} a''(t) & =4\int_{\Omega} \sum_{i=1}^2 (u_i)_t^2dx-4E(t)+2\int_{\Omega}(u_1 f_1(u)+u_2f_2(u)-2F(u)) dx\\ & \geq 4\int_{\Omega} \sum_{i=1}^2 (u_i)_t^2dx-4E(0)+4\int_0^{t} \sum_{i=1}^2 \Vert(u_i)_t\Vert_2^2dt+8\delta\int_{\Omega}F(u)dx. \end{align*} Then, using \eqref{e3.2} and \eqref{e3.4} again, we see that \begin{align*} a''(t) & \geq4(1+\delta)\int_{\Omega} \sum_{i=1}^2 (u_i)_t^2dx+4\delta \sum_{i=1}^2 \Vert\nabla u_i\Vert_2^2+4\delta( \sum_{i=1}^2 \Vert u_i\Vert_2^2-\frac{1+2\delta}{\delta}E(0)) \\ & +4(1+2\delta)\int_0^{t} \sum_{i=1}^2 \Vert(u_i)_t\Vert_2^2dt. \end{align*} Therefore, from Lemma \ref{lem3.2}, we obtain \eqref{e3.12}. \end{proof} Now, we are in a position to state and prove our main result. \begin{theorem} \label{thm3.4} Assume that {\rm (A1)-(A4)} hold. Also assume that $(\phi_1,\varphi_1)$, $(\phi_2,\varphi_2)\in H_0 ^1(\Omega)\times L^2(\Omega)$ satisfy the assumptions of Lemma \ref{lem3.2}. Then the local solution $(u_1(t)$, $u_2(t))$ of \eqref{e1.1}-\eqref{e1.3} blows up at finite time $T^{\ast}$ in the sense of \eqref{e3.1}. Moreover, if $2\delta\int_{\Omega}(\phi_1\varphi_1+\phi_2\varphi_2) dx>\| \phi_1\|_2^2+\| \phi_2\|_2^2,$ then the finite time $T^{\ast}$ is estimated by $$T^{\ast}\leq\frac{\| \phi_1\|_2^2+\| \phi _2\|_2^2}{2\delta\int_{\Omega}(\phi_1\varphi_1 +\phi_2\varphi_2) dx-(\| \phi_1\| _2^2+\| \phi_2\|_2^2) }.\label{e3.15}$$ \end{theorem} \begin{proof} We first note that $$2\int_0^{t}\int_{\Omega}u_i(u_i)_t\,dx\,dt=\Vert u_i \Vert_2^2-\Vert\phi_i\Vert_2^2.\label{e3.16}$$ By H\"{o}lder inequality and Young's inequality,from \eqref{e3.16} we have $$\Vert u_i\Vert_2^2\leq\Vert\phi_i\Vert_2^2+\int_0^{t}\Vert u_i\Vert_2^2dt+\int_0^{t}\Vert(u_i)_t\Vert_2 ^2dt,\quad i=1,2.\label{e3.17}$$ Next, we will find the estimate for the life span of $a(t)$. Let $$J(t) =\big[a(t) +(T_1-t) \sum_{i=1}^2 \Vert\phi_i\Vert_2^2\big]^{-\delta},\quad \text{for }t\in[0,T_1], \label{e3.18}$$ where $T_1>0$ is a certain constant which will be specified later. Then we have \begin{gather} J'(t) =-\delta J(t) ^{1+\frac{1}{\delta}}(a'(t) - \sum_{i=1}^2 \Vert\phi_i\Vert_2^2) , \label{e3.19}\\ J''(t) =-\delta J(t) ^{1+\frac {2}{\delta}}V(t) , \label{e3.20} \end{gather} where $$V(t) =a''(t) \big[a(t) +(T_1-t) \sum_{i=1}^2 \Vert\phi_i\Vert_2^2\big]-(1+\delta) \Big(a'(t) - \sum_{i=1}^2 \Vert\phi_i\Vert_2^2\Big) ^2. \label{e3.21}$$ For simplicity of calculation, for $i=1,2$, we denote $P_i =\int_{\Omega}u_i^2dx,\quad Q_i=\int_0^{t}\| u_i\|_2^2dt,\quad R_i =\int_{\Omega}(u_i)_t^2dx,\quad S_i =\int_0^{t}\| (u_i)_t\|_2^2dt.$ From \eqref{e3.13} \eqref{e3.16}, and H\"{o}lder inequality, we obtain \label{e3.22} \begin{aligned} a'(t) & = \sum_{i=1}^2 \Big(\int_{\Omega}2u_i(u_i)_tdx+\Vert\phi_i\Vert _2^2\Big) +2 \sum_{i=1}^2 \int_0^{t} \int_{\Omega} u_i(u_i)_t \,dx\,dt\\ & \leq2(\sqrt{R_1P_1}+\sqrt{Q_1S_1}+\sqrt{R_2P_2}+\sqrt {Q_2S_2})+ \sum_{i=1}^2 \Vert\phi_i\Vert_2^2. \end{aligned} By \eqref{e3.12}, we have $$a''(t) \geq4(1+\delta) (R_1+S_1+R_2+S_2) . \label{e3.23}$$ Thus, from \eqref{e3.22}, \eqref{e3.23}, \eqref{e3.21} and \eqref{e3.18}, we obtain \begin{align*} V(t) & \geq[4(1+\delta) (R_1+S_1+R_2+S_2) ]J(t) ^{-1/\delta}\\ &\quad -4(1+\delta) (\sqrt{R_1P_1}+\sqrt{Q_1S_1}+\sqrt{R_2P_2}+\sqrt{Q_2S_2}) ^2. \end{align*} Further, by \eqref{e3.18}$and \eqref{e3.11}$, we deduce that $V(t) \geq4(1+\delta) \big[(R_1 +S_1+R_2+S_2) (T_1-t) \sum_{i=1}^2 \Vert\phi_i\Vert_2^2+\Theta(t)\big],$ where \begin{align*} \Theta(t) & =(R_1+S_1+R_2+S_2) (P_1 +Q_1+P_2+Q_2) \\ & \quad -(\sqrt{R_1P_1}+\sqrt{Q_1S_1}+\sqrt{R_2P_2}+\sqrt {Q_2S_2}) ^2. \end{align*} By Schwartz inequality, $\Theta(t)$ is nonnegative. Hence, we have $$V(t) \geq0,\quad \text{for }t\geq0. \label{e3.24}$$ Therefore by \eqref{e3.20} and \eqref{e3.24}, we obtain $J''(t) \leq0$ for $t\geq0$, and then $$J(t)\leq J(0)+J'(0)t, \quad \text{for }t\geq0.\label{e3.25}$$ Also, we note that $J(0)>0\quad\text{and}\quad J'(0)<0$ due to \eqref{e3.18}, \eqref{e3.19} and \eqref{e3.6}. Hence, if we choose $T_1\geq-J(0)/J'(0)$, from \eqref{e3.25}, there exists a finite time $T^{\ast}\leq T_1$ such that $\lim_{t\to T^{\ast-}}J(t) =0.$ Then, it follows from the definition on $J(t)$ by \eqref{e3.18} that $\lim_{t\to T^{\ast-}} \sum_{i=1}^2 \Big(\Vert u_i\Vert_2^2+\int_0^{t}\Vert u_i(s)\Vert_2 ^2ds\Big) =\infty,$ which implies that $\lim_{t\to T^{\ast-}} \sum_{i=1}^2 \Vert u_i\Vert_2^2=\infty.$ Moreover, if $2\delta\int_{\Omega}(\phi_1\varphi_1+\phi_2\varphi_2) dx>\| \phi_1\|_2^2+\| \phi_2\|_2^2,$ the upper bound $T^{\ast}$ can be estimated as $T^{\ast}\leq\frac{\| \phi_1\|_2^2+\| \phi _2\|_2^2}{2\delta\int_{\Omega}(\phi_1\varphi_1 +\phi_2\varphi_2) dx-(\| \phi_1\| _2^2+\| \phi_2\|_2^2) }.$ This completes the proof. \end{proof} \begin{example} \label{examp3.5} \rm Consider the system \eqref{e1.1}-\eqref{e1.3} with $f_1(u_1,u_2)=u_1^2u_2,\quad f_2(u_1,u_2)=u_1u_2^2;$ that is, we consider the problem $$\begin{gathered} (u_1)_{tt}-\Delta u_1+u_1+(u_1) _t=u_1^2u_2\quad \text{in }\Omega\times[0,T),\\ (u_2)_{tt}-\Delta u_2+u_2+(u_2) _t=u_2^2u_1\quad \text{in }\Omega\times[0,T),\\ u_1(x,0)=\phi_1,\quad u_2(x,0)=\phi_2,\quad x\in\Omega,\\ (u_1)_t(x,0)=\varphi_1,\quad (u_2)_t(x,0)=\varphi_2,\quad x\in\Omega,\\ u_1(x,t)=0,\quad u_2(x,t)=0,\quad x\in\partial\Omega\times(0,T). \end{gathered} \label{e3.26}$$ By \eqref{e3.2} and \eqref{e3.3}, we have \begin{gather*} E(t) =\frac{1}{2}\sum_{i=1}^2 \big[\| (u_i)_t\|_2^2+\Vert\nabla u_i\Vert_2^2+\Vert u_i\Vert_2^2\big]-\frac{1}{2}\| u_1 ^2u_2^2\|_2^2, \\ I(t) =\sum_{i=1}^2 \big[\Vert\nabla u_i\Vert_2^2+\Vert u_i\Vert_2^2\big] -2\int_{\Omega} u_1^2u_2^2dx, \end{gather*} and assumption (A4) is satisfied with $\delta=1/2$. To apply Theorem \ref{thm3.4}, we need to check that the initial data set that satisfies conditions $E(0)>0$, $I(0)<0$ and $$\Vert\phi_1\Vert_2^2+\Vert\phi_2\Vert_2^2>4E(0), \label{e3.27}$$ by \eqref{e3.5} is not empty. Setting $$\begin{gathered} \alpha=\| \phi_1\|_2^2+\| \phi_2\|_2^2,\quad \beta=\| \nabla\phi_1\|_2 ^2+\| \nabla\phi_2\|_2^2,\\ \gamma=\| \phi_1\phi_2\|_2^2,\quad \lambda =\| \varphi_1\|_2^2+\| \varphi_2\|_2^2. \end{gathered}\label{e3.28}$$ Then the above conditions $E(0)>0$, $I(0)<0$ and \eqref{e3.27} read as follows \begin{gather} E(0) =\frac{1}{2}(\alpha+\beta+\lambda) -\frac{1}{2}\gamma>0,\label{e3.29}\\ I(0) =\alpha+\beta-2\gamma<0,\label{e3.30}\\ \alpha >2(\alpha+\beta+\lambda) -2\gamma. \label{e3.31} \end{gather} Having \eqref{e3.30} in mind, we choose $\phi_1$ and $\phi_2$ such that $$\alpha+\beta=2\gamma-\varepsilon\gamma, \label{e3.32}$$ with $0<\varepsilon<2$. Thus \eqref{e3.30} is satisfied. At this moment, we consider two cases: (i) $0<\varepsilon\leq1$ and (ii) $1<\varepsilon<2$. Case (i) $0<\varepsilon\leq1$. In this case, we further require $\phi_1$ and $\phi_2$ to satisfy $\alpha>-2(\varepsilon-1) \gamma$, and then, select $\lambda$ such that $$0<\lambda<\frac{\alpha}{2}+(\varepsilon-1) \gamma. \label{e3.33}$$ Substituting \eqref{e3.32} into \eqref{e3.29} and $0<\varepsilon\leq1$, we see that $2E(0) =\alpha+\beta+\lambda-\gamma =\lambda-(\varepsilon-1) \gamma >0,$ this implies that \eqref{e3.29} is achieved. Since $\lambda<\frac{\alpha} {2}+(\varepsilon-1) \gamma$ by \eqref{e3.33}, we deduce that $\alpha >2\lambda-2(\varepsilon-1) \gamma =2(\alpha+\beta+\lambda) -2\gamma,$ where the last equality is derived due to \eqref{e3.32}. Thus \eqref{e3.31} is obtained. Case (ii) $1<\varepsilon<2$. In this case, we select $\lambda$ such that $$(\varepsilon-1) \gamma<\lambda<\frac{\alpha}{2}+( \varepsilon-1) \gamma. \label{e3.34}$$ Similarly as in part (i), we see that the conditions \eqref{e3.29}-\eqref{e3.31} are satisfied. Therefore, from above arguments, the set of all initial data which satisfy the conditions $E(0)>0$, $I(0)<0$ and \eqref{e3.27} is not empty. Furthermore, although $\Vert\phi_1\Vert_2^2+\Vert\phi_2\Vert_2 ^2>4E(0)$ gives an upper bound of the initial energy $E(0)$. $E(0)$ can be chosen to be arbitrary positive provided that $\alpha=\| \phi _1\|_2^2+\| \phi_2\|_2^2$ is large enough and $\beta,\gamma$ can be also larger accordingly to make sure \eqref{e3.29}-\eqref{e3.31} is still satisfied. Next, we give an example to illustrate the above discussion is workable. Consider the problem \eqref{e3.26} with $\Omega=(0,4)$, $\phi_1(x)=\begin{cases} x,& 00, \\ I(0)=\alpha+\beta-2\gamma=-932<0, \\ 2(\alpha+\beta+\lambda) -2\gamma=2(\lambda-348.8)<\alpha. \end{gather*} Thus Theorem \ref{thm3.4} is applicable. \end{example} \begin{example} \label{examp3.6} \rm Consider the system \eqref{e1.1}-\eqref{e1.3} in \mathbb{R}^3 with \[ f_1(u_1,u_2)=4\lambda(u_1+\alpha u_2) ^{3}+2\beta u_1u_2^2,\quad f_2(u_1,u_2)=4\alpha\lambda(u_1+\alpha u_2) ^{3}+2\beta u_1^2u_2.$ Assume that $\lambda>0$, $\beta>0$ and $\alpha$ is any real number. Now we have $F(u_1,u_2)=\lambda(u_1+\alpha u_2) ^{4}+2\beta u_1^2u_2^2.$ We see that (A4) is satisfied if $0<\delta\leq 1/2$. Thus Theorem \ref{thm3.4} is applicable. \end{example} \subsection*{Acknowledgements} The author would like thank the anonymous referees for their comments. He is also grateful to Professor Liu Wenjun for his comments on Example \ref{examp3.5}. \begin{thebibliography}{00} \bibitem{a1} K. Agre and M. A. Rammaha; Systems of nonlinear wave equations with damping and source terms, \textit{Differential Integral Equations} 19(2006) 1235-1270. \bibitem{a2} C. O. Alves, M. M. Cavalcanti, V. N. Domingos Cavalcanti, M. A. Rammaha and D. Toundykov; On existence, uniform decay rates and blow up for solutions of systems of nonlinear wave equations with damping and source terms, \textit{Discrete Contin. Dyn. Syst. Ser. S} 2(2009) 583-608. \bibitem{g1} F. Gazzola and M. Squassina; Global solutions and finite time blow up for damped semilinear wave equations, \textit{Ann. Inst. H. Poincar\'{e} Anal. Non Lin\'{e}aire} 3(2006) 185-207. \bibitem{g2} V. Georgiev and D. Todorova; Existence of solutions of the wave equations with nonlinear damping and source terms, \textit{J. Differential Equations} 109(1994) 295-308. \bibitem{h1} A. Haraux; Semi-Linear Hyperbolic Problems in Bounded Domains, C. N. S. R., Laboratoire associ\'{e}, Universit\'{e} Paris VI, France. \bibitem{l1} H. A. Levine; Some additional remarks on the nonexistence of global solutions to nonlinear wave equations, \textit{SIAM J. Math. Anal.} 5(1974) 138-146. \bibitem{l2} H. A. Levine and D. Todorova; Blow up of solutions of the Cauchy problem for a wave equation with nonlinear damping and source terms and positive initial energy, \textit{Proceedings of The American Mathematical Society} 129(2001) 793-805. \bibitem{l3} H. A. Levine, S. Ro Park and J. Serrin; Global existence and global nonexistence of solutions of the Cauchy problem for a nonlinearly damped wave equation, \textit{Journal of Mathematical Analysis and Applications} 228(1998) 181--205. \bibitem{l4} H. A. Levine and J. Serrin; A global non-existence theorem for quasilinear evolution equation with dissipation, \textit{Archives for Rational Mechanics and Analysis} 137(1997) 341--361. \bibitem{l5} M. R. Li and Long-Yi Tsai; Existence and nonexistence of global solutions of some system of semilinear wave equations, \textit{Nonlinear Analysis TMA} 54(2003) 1397-1415. \bibitem{m1} T. Matsuyama and R. Ikehata; On global solutions and energy decay for the wave equations of Kirchhoff type with nonlinear damping term, \textit{Journal of Mathematical Analysis and Applications} 204(1996) 729-753. \bibitem{m2} S. A. Messaoudi; Blow up in a non-linearly damped wave equation, \textit{Mathematische Nachrichten} 231(2001) 1--7. \bibitem{m3} S. A. Messaoudi; Blow up in the Cauchy problem for a nonlinearly damped wave equation, \textit{Comm. On Applied Analysis}, 7(2003) 379-386. \bibitem{p1} P. Pucci and J. Serrin; Global nonexistence for abstract evolution equations with positive initial energy, \textit{J. Differential Equations} 150(1998) 203-214. \bibitem{r1} M. Reed; Abstract Nonlinear Wave Equations, Springer-verlag, 1976. \bibitem{t1} G. Todorova; Stable and unstable sets for the Cauchy problem for a non-linear wave with non-linear damping and source terms, \textit{Journal of Mathematical Analysis and Applications,} 239(1999) 213 --226. \bibitem{t2} D. Todorova and E. Vitillaro; Blow-up for nonlinear dissipative wave equations in $R^{n}$, \textit{Journal of Mathematical Analysis and Applications} 303(2005) 242-257.\ \bibitem{t3} G. Todorova; Cauchy problem for a non-linear wave with non-linear damping and source terms, \textit{Comptes Rendus de Academie des Sciences Paris Serie} I, 326(1998) 191--196. \bibitem{v1} E. Vitillaro; Global non-existence theorems for a class of evolution equations with dissipation, \textit{Archives for Rational Mechanics Analysis,} 149(1999) 155--182. \bibitem{w1} Y. Wang; A sufficient condition for finite time blow up of the nonlinear Klein-Gordon equations with arbitrarily positive infinite initial energy, \textit{Proceedings of The American Mathematical Society} 136(2008) 3477-3482. \bibitem{w2} Y. Wang; Non-existence of global solutions of a class of coupled non-linear Klein-Gordon equations with non-negative potentials and arbitrary initial energy, \textit{IMA Journal of Applied Mathematics}, 74(2009) 392-415. \bibitem{z1} Y. Zhou; A blow-up result for a nonlinear wave equation with damping and vanishing initial energy in $R^N$, \textit{Applied Math. Letters}, 18(2005) 281-286. \end{thebibliography} \end{document}