A Wavelet Based Numerical Method for Nonlinear Partial Differential Equations

This paper is concerned with the numerical treatment of quasilinear elliptic partial differential equations. In order to solve the given equation we propose to use a Galerkin approach, but, in contrast to conventional finite element discretizations, we work with trial spaces that, not only exhibit the usual approximation and good localization properties, but, in addition, lead to expansions of any element in the underlying Hilbert spaces in terms in multiscale or wavelet bases with certain stability properties. Specifically, we select as trial spaces a nested sequence of spaces from an appropriate biorthogonal multiscale analysis. This gives rise to a nonlinear discretized system. To overcome the problems of nonlinearity, we make use of the machinery of interpolating wavelets to obtain knot oriented quadrature rules. Finally, Newton's method is applied to approximate the solution in the given ansatz space. The results of some numerical experiments with different biorthogonal systems, confirming the applicability of our scheme, are presented.