A convolution model for interfacial motion: The generation and propagation of internal layers in higher space dimensions (Q1977666)

This paper analyzes the properties of the solutions of \[ u_t= J* u-u- f(u),\quad x\in\mathbb{R}^N,\quad t\geq 0\tag{1} \] with \(*\) denoting convolution and \(J\), \(f\) satisfying \(J\in C^2(\mathbb{R}^N)\), \(J(-x)= J(x)\), \(J(x)\geq 0\), \(\int J dx= 1\), \(\int x^2J dx< \infty\), \(f\in C^2(\mathbb{R})\) vanishes only in \(-1\), \(+1\), \(\alpha\in(- 1,+1)\), \(f'(\pm 1)> 0\). The principal aim is to study the spread of regions in which \(u\) is close to extreme values \(u= \pm 1\) corresponding to the stable phases. The paper generalizes many results previously known for the reaction-diffusion equation \(u=\Delta u- f(u)\). It consists of 4 sections (plus a concise and clear introduction), each devoted to a particular aspect. The first topic presented is the onset of interfaces. This refers to the slow diffusion/fast reaction case, introducing the rescaled operator \(J_\varepsilon(x)= \varepsilon^{-N}J({x\over\varepsilon})\) and the corresponding problem \[ \varepsilon u_t= J_\varepsilon* u-u- f(u),\quad x\in\mathbb{R}^N,\quad t\geq 0,\tag{2} \] \[ u(x,0; \varepsilon)= \phi(x),\quad -1\leq\phi(x)\leq 1,\quad \phi\in C^2.\tag{3} \] The result is that (for small \(\varepsilon>0\)) at those points \(x\) where \(\phi(x)\geq \alpha+ M_0\sqrt\varepsilon|\ln\varepsilon|\) we have \(u(x, \tau_0 \varepsilon|\ln\varepsilon|)\geq 1- M_1\varepsilon\), where the constants \(M_0\), \(M_1\), \(\tau_0\) depend on \(f\) only. The symmetric property is valid for points such that \(\phi(x)\leq \alpha- M_0\sqrt\varepsilon \ln|\varepsilon|\). Roughly speaking, in the regions \(|\phi- \alpha|> O(\sqrt\varepsilon \ln|\varepsilon|)\) \(u\) becomes close to \(\pm 1\) in time \(O(\varepsilon|\ln \varepsilon|)\) and the interfacial region has a thickness \(O(\sqrt\varepsilon|\ln \varepsilon|)\). The next session deals with the spread of one phase at the expense of the other. The statement of the theorem is rather technical, but its physical meaning is clear and relates the behaviour of the system to that of the 1D travelling wave (t.w.). Namely, if the 1D t.w. has a velocity \(c<0\), than \(u\) approaches the radial 1D t.w. for large \(t\) (with a shift in phase), provided \(\phi(x)\geq \alpha+\eta\) in a suitable ball. Then the authors reconsider the family of problems (2), (3) with the aim of studying the limit \(\varepsilon\to 0\). Again the behaviour of the system is influenced by that of the 1D t.w. The assumption that \(J\) is radial is added. The result is the following: let the 1D t.w. has a velocity \(c<0\) and define \(\Omega^+_0= \{x\in\mathbb{R}^N: \phi(x)> \alpha\}\). Then for all \(t>0\), \(u(x,t,\varepsilon)\) tends to \(+1\) if \(\text{dist}(x,\Omega^+_0)< -ct\) and to \(-1\) if \(\text{dist}(x, \Omega^+_0)> -ct\). If instead \(c= 0\), then \(u\) tends to \(+1\) or \(-1\) if \(\phi(x)> \alpha\) or \(\phi(x)< \alpha\), respectively. The last section considers again equation (2), with \(J(x)x^4\leq C\), and \(f'(u)>-1\), \(\int^1_{-1} f du= 0\). This idea is to study the structure of \(u\) near its zero level set by means of a (formal) asymptotic analysis. It is established that in the case of vanishing speed of the 1D t.w. the interface motion is driven by curvature.