Diffusion approximation for hyperbolic stochastic differential equations (Q1382465)

A diffusion approximation for the solutions to the Goursat problem for the random nonlinear hyperbolic equation \[ \frac {\partial ^{2}X^{\varepsilon }_{s,t}}{\partial s\partial t} = \frac 1{\varepsilon }F\left (\frac {s}{\varepsilon },\frac {t}{\varepsilon } \right )\sigma (X^{\varepsilon }_{s,t}) + b(X^{\varepsilon }_{s,t}), \;X^{\varepsilon }_{0,t} =X^{\varepsilon }_{s,0} = x, \quad (s,t)\in [0,S]\times [0,T], \] is studied. The functions \(b\) and \(\sigma \) are assumed to have continuous and bounded derivatives up to the third order; the random field \(F\) is of the form \[ F(s,t) = \sum ^{\infty }_{k=1}\sum ^{\infty }_{l=1} Z_{k,l}\mathbf{1}_{[k-1,k[\times [l-1,l[}(s,t), \] where \(\{Z_{k,l}\}\) is a family of independent, identically distributed centered random variables such that \(|Z_{k,l}|\leq M\) for a constant \(M\geq 0\). Using a martingale problem approach to two-parameter processes the authors prove that the law of the solution \(X^{\varepsilon }_{s,t}\) converges weakly on the Banach space \(\mathcal C([0,S]\times [0,T])\) as \(\varepsilon \downarrow 0\) to the law of the unique solution to the Stratonovich equation \[ \roman dX_{s,t} = \gamma \sigma (X_{s,t})\circ \roman dW_{s,t} + b(X_{s,t}), \;X_{0,t}=X_{s,0}=x, \quad (s,t)\in [0,S]\times [0,T], \] where \(W\) denotes the Brownian sheet and \(\gamma = E(Z^{2}_{k,l})\). Analogous results in the linear case (\(b=0\), \(\sigma (y) = y\)) were obtained recently by \textit{R. A. Carmona} and \textit{J. P. Fouque} [Probab. Theory Relat. Fields 98, No. 3, 277-298 (1994; Zbl 0794.60060)].