1: \begin{abstract}

2: This paper presents a wavelet Galerkin method for solving 2D elliptic interface problems $-\nabla\cdot(a\nabla u)=f$ in $\Omega\backslash \Gamma$, where $\Gamma$ is a smooth interface inside $\Omega$ while the variable scalar coefficient $a>0$ and source $f$ are possibly discontinuous across $\Gamma$. Using a biorthogonal wavelet basis which is derived from bilinear finite elements and forms a Riesz basis of $H^1_0(\Omega)$, we devise a simple strategy to attain nearly optimal convergence rates

3: $\bo(h^2 |\log(h)|^2)$ in the $L_2(\Omega)$-norm and $\bo(h |\log(h)|)$ in the $H^1(\Omega)$-norm with respect to the approximation order.

4: Along the interface $\Gamma$,

5: 	we incorporate extra wavelet elements to effectively capture the geometry of $\Gamma$ and singularities of the solution $u$ whose gradient is discontinuous across $\Gamma$ (and hence, low regularity of $u$).

6: Using the dual part of the biorthogonal wavelet basis, we prove these nearly optimal convergence rates by developing Bessel properties of wavelets and various inequalities in fractional Sobolev spaces, and by employing finite element method arguments.

7: To achieve higher accuracy and to overcome high contrast coefficients $a$, similar to meshfree methods, our method essentially boils down to adding more wavelet elements throughout the domain and along the interface, instead of re-meshing the domain with a smaller mesh size as in finite element methods.

8: In contrast to all other methods for the 2D elliptic interface problems, since our 2D wavelet basis is a Riesz basis of the Sobolev space $H^1_0(\Omega)$, the coefficient matrices obtained from the discretization have relatively small and uniformly bounded condition numbers irrespective of their sizes.

9: 	\end{abstract}

10: