1: \begin{abstract}

2: Numerical modeling of wave propagation in heterogeneous media is important

3: in many applications. Due to the complex nature, direct

4: numerical simulations on the fine grid are prohibitively expensive. It is therefore

5: important to develop efficient and accurate methods that allow the

6: use of coarse grids. In this paper, we present a multiscale finite

7: element method for  wave propagation on a coarse grid.

8: The proposed method is based on the Generalized Multiscale Finite Element

9: Method (GMsFEM) (see \cite{egh12}).

10: To construct

11: multiscale basis functions, we start with two snapshot spaces in each

12: coarse-grid block where

13: one represents the degrees of freedom on the boundary and the other

14: represents the degrees of freedom in the interior.

15:  We use local spectral problems to identify important

16: modes in each snapshot space.

17: These local spectral problems are different from each other and their

18: formulations are based on the analysis. To our best knowledge,

19: this is the first time where

20: multiple snapshot spaces

21: and multiple spectral problems are used and

22: necessary for efficient computations.

23: Using the dominant modes from local spectral problems,

24: multiscale basis functions are constructed to

25: represent the solution space locally within each coarse block.

26: These multiscale basis functions

27: are coupled via the symmetric interior penalty discontinuous Galerkin

28: method which provides a block diagonal mass matrix, and, consequently,

29: results in fast computations in an explicit time discretization.

30: Our methods' stability

31: and spectral convergence  are rigorously analyzed. Numerical examples

32: are presented to show our methods' performance. We also test oversampling

33: strategies. In particular,

34: we discuss how the modes from different snapshot spaces can affect the proposed methods' accuracy.

35: \end{abstract}

36: