Singular limit for stochastic reaction-diffusion equation and generation of random interfaces (Q1806300)

Let \(f\in C^\infty (R)\) have exactly three zeros \(u_1= -1< u_2< u_3=1\) such that \(f'(\pm 1)< 0\), \(f'(u_2)> 0\), and \(\int_{-1}^1 f(u) du=0\). Let \(D\subset R^2\) be a bounded region with smooth boundary \(\partial D\) and let \(n\) denote the unit award normal to \(\partial D\). The author studies the limit behaviour, as \(\varepsilon\to 0\), of the solution \(u^\varepsilon (t,x)\) of the stochastic reaction-diffusion equation \[ \partial_t u^\varepsilon (t,x)= \Delta u^\varepsilon+ \varepsilon^{-2} f(u^\varepsilon)+ \varepsilon^{-1} \xi_t^\varepsilon \quad (t>0,\;x\in D), \] \[ u^\varepsilon (0,x)= u_0^\varepsilon(x) \quad (x\in D), \qquad \partial_n u^\varepsilon (t,x)= 0\quad (t>0,\;x\in \partial D), \] where the driving noise \(\xi_t^\varepsilon\) is smooth in \(t\) and behaves like a white noise in the limit \(\varepsilon \downarrow 0\). The author shows that \(u^\varepsilon (t,\cdot)\to I_{\Gamma_t} (\cdot)- I_{D\setminus \Gamma_t} (\cdot)\) in distributional sense where the curvature of the closed curve \(\Gamma\subset D\) is the solution of a certain nonlinear stochastic PDE. The problem behind is to prove the convergence of solutions of nonlinear PDEs with rapidly oscillating coefficients to that of a nonlinear stochastic PDE. In former work the author has studied already the one-dimensional case. The present paper generalizes earlier works on the linear case by \textit{R. Bouc} and \textit{E. Pardoux} [Stochastic Anal. Appl. 2, 369-442 (1984; Zbl 0574.60066)], \textit{H. Watanabe} [Appl. Math. Optimization 20, No. 1, 81-96 (1989; Zbl 0678.60046)], and others.