Large time monotonicity of solutions of reaction-diffusion equations in \(\mathbb R^N\) (Q1707303)

Motivated by a problem from tumor modelling and medical imaging the authors study solutions to the reaction-diffusion equation \[ u_t = \mathrm{div}_x(A(x)\nabla u) + f(x, u), \quad t>0,\, x\in\mathbb{R}^N \] on the whole space. The initial condition \(u(0,x)=u_0(x)\) is assumed to satisfy \(0\leq u_0(x)\leq 1\) and either has Gaussian decay, i.e. \[ u_0(x) = \mathcal{O}(e^{-\beta|x|^2}) \] as \(|x|\to\infty\) for some \(\beta>0\) or decays with a precise rate, i.e. \[ \gamma e^{-\lambda |x|} \leq u_0(x) \leq \delta e^{-\lambda |x|} \] for some \(\lambda>0\), \(0<\gamma\leq\delta\) if \(|x|\) is sufficiently large. The first case includes in particular the case of compactly supported initial data. The symmetric matrix field \(A\) is assumed to be locally asymptotically homogeneous at infinity, i.e. \(|\nabla A_{ij}(x)| \to 0\) as \(|x| \to\infty\) which allows for weak oscillations in the coefficients of \(A\). For the nonlinearity \(f(x,u)\) a set of conditions generalizing the classical Fisher-KPP-type nonlinearity \(f(x,u)=u(1-u)\) is imposed. These assumptions imply in particular that the only stationary solutions are \(u\equiv 0\) and \(u\equiv 1\) and that solutions converge to \(+1\) locally uniformly in \(x\in\mathbb{R}^N\). The main result states that solutions are eventually monotonically increasing in time in the sets where they are bounded away from zero. In the one-dimensional case under some additional hypotheses the authors show that monotonicity in time \(u_t(t,x)>0\) even holds for all \(x\in\mathbb{R}\) at sufficiently large times. The main part of the proof consists of showing finite time monotonicity properties \(u(t + \tau, x) \geq u(t, x)\) that hold for some fixed \(\tau>0\), all \(x\in\mathbb{R}^N\) and all sufficiently large times \(t\geq t_0\).