American Institute of Mathematical Sciences

$\frac{\partial u_{j}}{\partial t}=d_{j}$Δ $u_{j}- a_{j}u_{j} +g_{j}(x,u) \ \ \text{in}\ \ \Omega\times[0,T) $ ,
$\frac{\partial u_{j}}{\partial\nu}=0\ \ \text{on}\ \ \partial\Omega\times[ 0,T) $,
$u_{j}(x,0) =\varphi_{j}(x)\ \ \text{in}\ \ \Omega$

where $\Omega$ is a smooth bounded domain in $\mathbb{R}^{n}$ with $\nu$ its unit outer normal, $j=1,2$, $u=(u_1,u_2)$ and

$g_{1}(x,u) =\rho_{1}(x,u) \frac{u_{1}^{p}}{u_{2}^{q}}+\sigma_{1}(x) $ ,
$g_{2}(x,u) =\rho_{2}(x,u) \frac{u_{1}^{r}}{u_{2}^{s}}+\sigma_{2}(x) $.

Here $d_{j}, a_{j}$ are positive constants, $\rho_{1}\geq0$, $\rho _{2}>0,\sigma_{j}\geq0$ are bounded smooth functions and $p,q,r,s$ are nonnegative constants satisfying $0<\frac{p-1}{r}<\frac{q}{s+1}$.
We show that there is a unique global solution when p-1 < r, which improves 1987 result of K. Masuda and K. Takahashi [10]. Asymptotic bounds of global solutions are also established which yield new a priori estimates of stationary solutions.

$\frac{d^2x}{dt^2}+\arctan x=p(t),$

where $p(t+1)=p(t)$ is a smooth function.

$\A(u'(t))+ \B(u(t))-\lambda u(t)$ ∋ $f \mbox{in } \H \mbox{ for a.e. }t\in (0,+\infty)$
$u(0)=u_{0},$

where $\A$ is a maximal (possibly multivalued) monotone operator from the Hilbert space $\H$ to itself, while $\B$ is the subdifferential of a proper, convex and lower semicontinuous function φ:$\H\rightarrow (-\infty,+\infty]$ with compact sublevels in $\H$ satisfying a suitable compatibility condition. Finally, $\lambda$ is a positive constant. The existence of solutions is proved by using an approximation-a priori estimates-passage to the limit procedure. The main result of this paper is that the set of all the solutions generates a Generalized Semiflow in the sense of John M. Ball [8] in the phase space given by the domain of the potential φ. This process is shown to be point dissipative and asymptotically compact; moreover the global attractor, which attracts all the trajectories of the system with respect to a metric strictly linked to the constraint imposed on the unknown, is constructed. Applications to some problems involving PDEs are given.