The evolution to a steady state for a porous medium model (Q2497408)

This paper concerns the asymptotic behavior of the nonnegative solution \(T = T(x,t)\) of the initial-boundary value problem for the degenerate diffusion equation: \[ \left\{\begin{alignedat}{2} \frac{\partial T}{\partial t} &= \text{ div }(k(T)\nabla T)\quad&&\text{in }\Omega \times (0,\infty), \\ T(x,0) &= f(x) \quad&&\text{in }\Omega, \\ T(x,t) &= \overline{T}(x)\quad&&\text{on }\partial\Omega\times (0,\infty), \end{alignedat}\right. \] where \(\Omega\) is a bounded smooth domain in Euclidean space and \(k(\tau) = e^\tau-1\). Let \(U = U(x)\) be the solution of the corresponding steady state problem. By putting \(u = T-U\) and \(\Phi(u,U) = \int_0^u d\overline{u}\int_0^{\overline{u}} k(\tau + U)\,d\tau\), one may have \[ \left\{\begin{alignedat}{2} \frac{\partial u}{\partial t} &= \Delta \Phi_u\quad&&\text{in } \Omega \times (0,\infty), \\ u(x,0) &= f(x)-U(x) \quad&&\text{in }\;\Omega, \\ u(x,t) &= 0\quad&&\text{on }\;\partial\Omega\times (0,\infty). \end{alignedat}\right. \] Under the assumption that \(\max \overline{T} > 0\), by replacing \(f\) and \(\overline{T}\) with \(f+\varepsilon\) and \(\overline{T}+\varepsilon\) for \(\varepsilon > 0\), respectively, one may deal with positive classical solutions \(T\) and \(U\). Then, by applying the energy method to the functional \(E(t) = \int_\Omega \Phi(u,U)\,dx\), the author obtains the inequality \[ \frac {dE}{dt} \leq -\lambda_1 K | \Omega| ^{-1} E^2, \] where \(K\) is a positive constant depending only on \(\max \overline{T}\), \(| \Omega| \) is the volume of \(\Omega\), and \(\lambda_1\) is the first eigenvalue of \(-\Delta\) on \(\Omega\) with homogeneous Dirichlet boundary condition. Thus integration of this inequality yields \[ E(t) \leq E(0)[1+\lambda_1 K\{E(0) | \Omega| ^{-1}\}t]^{-1}, \] which concludes that \(E(t) \to 0\) as \(t \to \infty\), and hence \(T \to U\) as \(t \to \infty\). Moreover, the author considers the backwards in time version of the initial-boundary value problem: \[ \left\{\begin{alignedat}{2} \frac{\partial T}{\partial t} &= \text{ div }(k(T)\nabla T)\quad&&\text{in }\;\Omega \times (-t_0,0), \\ T(x,0) &= f(x) \quad&&\text{in } \Omega, \\ T(x,t) &= \overline{T}(x)\quad&&\text{on } \partial\Omega\times (-t_0,0), \end{alignedat}\right. \] and by using the same inequality concludes that there exists no solution provided\break \(t_0 > [\{E(0)| \Omega| ^{-1}\}\lambda_1 K]^{-1}\). When \(k(\tau) = \tau^{n-1}\) for some \(n > 1\), the corresponding results were already obtained by \textit{J. N. Flavin} and \textit{S. Rionero} [J. Math. Anal. Appl. 281, 221--232 (2003; Zbl 1027.35054)].