Asymptotic behavior of positive solutions of random and stochastic parabolic equations of Fisher and Kolmogorov types (Q5957846)

The authors study long-term behavior of positive solutions for the following random parabolic equation \[ \begin{cases} u_t=\Delta u+m (\theta_t \omega,x,u)g(u),\;x\in D\\ {\partial u\over \partial n}=0,\;x\in \partial D,\end{cases} \tag{1} \] where \(D\subset\mathbb{R}^N\) has an open, bounded and connected smooth boundary \(\partial D\), \(\omega\in\Omega\), \(((\Omega,{\mathcal F}, P)\), \((\theta_t)_{t\in R})\) is an ergodic metric dynamic system, \(m(\omega,x,u)\) is measurable and \(m\cdot g\) is of Fisher type. Moreover, the authors study long-term behavior of positive solutions for the stochastic parabolic Stratonovich problem \[ \begin{cases} du=\bigl(\Delta u+m(x,u)g(u)\bigr) dt+g(u) \circ dW(t,u),\;x \in D\\ {\partial u\over\partial n}=0,\;x\in\partial D,\end{cases} \tag{2} \] where \(D\) is as in (1), \(\omega\in\Omega\), \((W(t,\cdot))_{t\in\mathbb{R}}\) is a two-sided scalar Wiener process on the probability space \((\Omega,{\mathcal F},P)\), \(\circ dW (t,\omega)\) denotes Stratonovich's differential, and \(m\cdot g\) is either of Fisher or Kolmogorov type. The authors prove two main theorems of the following form: Theorem A. Consider (1). Then one and only one of following alternatives is valid: (i) There is one trivial equilibrium solution such that for each strictly positive random variable \(\varphi,\pi_t (\varphi(\omega), \omega)\) converges as \(t\to\infty\) to this equilibrium solutions for a.e. \(\omega \in\Omega\). (ii) For every strictly positive random variable \(\varphi, \pi_t(\varphi (\omega),\omega)\) is neither bounded away from the trivial equilibrium solutions nor does \(\omega\mapsto \pi_t(\varphi (\omega), \omega)\) converge pointwise a.e. on \(\Omega\) to any of the trivial equilibrium solutions as \(t\to\infty\). (iii) For every strictly positive random variable \(\varphi,\pi_t (\varphi (\omega),\omega)\) is bounded away from the trivial equilibrium solutions. Moreover, if alternative (iii) holds and there is a strictly positive random variable \(\varphi_0\) such that \(\pi_t(\varphi_0 (\omega), \omega)\) is uniformly bounded away from the trivial equilibrium solutions, then (iv) There is a one parameter family of positive equilibrium solutions \(\{\varphi_\lambda\}_{\lambda\in I} (I\subset\mathbb{R}\) is an interval) such that for any strictly positive random variable \(\varphi\), for a.e. \(\omega\in\Omega\) there is \(\lambda(\omega)\in I\) such that \(\pi_t(\varphi (\omega),\omega)\) approaches \(\varphi_{\lambda (\omega)}(\theta_t (\omega))\) as \(t\to\infty\). Theorem B. Consider (2). Then one and only one of the following alternatives is met: (i) There is one trivial equilibrium solution such that for every strictly positive random variable \(\varphi,\pi_t(\varphi(\omega),\omega)\) converges as \(t\to\infty\) to this trivial equilibrium solution for a.e. \(\omega \in\Omega\). (ii) For every strictly positive random variable \(\varphi,\pi_t (\varphi (\omega),\omega)\) is neither bounded away from the trivial equilibrium solutions nor does \(\omega\mapsto \pi_t(\varphi(\omega),\omega)\) converge pointwise a.e. on \(\Omega\) to a trivial equilibrium solution as \(t\to\infty\). (iii) For every positive random variable \(\varphi,\pi_t (\varphi(\omega), \omega)\) is bounded away from the trivial equilibrium solutions. However, on uniformly positive equilibrium solutions and under an additional condition on \(m(x,u)\), alternative (iii) never occurs.