Normal approximation for a random elliptic equation (Q398777)

The author studies the statistical behaviour of the random variable \[ \Gamma_{L,\beta} = \frac{1}{| D_L| } \int_{D_L} a(x)| \nabla\phi + e_1| ^2 + \beta \phi^2 dx, \] where \(\phi(x)\) is the solution of the random elliptic equation \[ -\nabla\cdot(a(x)(\nabla \phi(x)+e_1))+\beta\phi(x) = 0,\;\; x\in D_L\subset{\mathbb R}^d. \] Here, \(a(x)\) is a \(L^{\infty}({\mathbb R}^d)\) stationary random field satisfying an ellipticity condition, \(\beta\) is a nonnegative constant, the unit vector \(e_1\) is deterministic, and \(D_L = [0,\, L)^d\) is a domain. This equation is considered here together with a suitable periodic boundary condition. The main result of the paper is that the distribution law of the above random variable is approximately normal when the domain size \(L\) is sufficiently large. This approximation is shown by using an error estimate on the total variation distance between two random variables. The error estimate is derived using a second-order Poincaré inequality.