1: \begin{abstract}

2: In this paper, we propose and analyze a new stochastic homogenization method for diffusion equations

3: with random and fast oscillatory coefficients. In the proposed method, the homogenized solutions are

4: sought through a two-stage procedure. In the first stage, the original oscillatory diffusion equation is approximated, for each fixed random sample $\omega$, by a spatially homogenized diffusion equation with

5: piecewise constant coefficients, {\color{black}resulting in} a random diffusion equation.

6:  In the second stage,

7: the {\color{black}resulting} random diffusion equation is approximated and computed by using an efficient

8: multi-modes Monte Carlo method  which only requires to solve a  diffusion equation

9: with a constant diffusion coefficient and a random right-hand side.

10: The main advantage of the proposed method is that it separates the

11: computational difficulty caused by the spatial fast oscillation of the solution and that caused by the randomness of the solution, so they can be overcome separately using different strategies.

12: The convergence of the solution of the spatially homogenized equation (from the first stage) to the solution of the original random diffusion equation is established and the optimal rate of convergence is also obtained for the proposed multi-modes Monte Carlo method.

13: Numerical experiments on some benchmark test problems for random composite materials are also

14: presented to gauge the efficiency and accuracy of the proposed two-stage stochastic homogenization method.

15: \end{abstract}

16: