Efficient and accurate spectral methods for PDEs with singular solutions
主 题: Efficient and accurate spectral methods for PDEs with singular solutions
报告人: Shen Jie (Xiamen University and Purdue University)
时 间: 2017-01-03 16：00--17：00
地 点: 威廉希尔理科一号楼1479室
The usual spectral methods will provide high-order accuracy for problems with smooth solutions. However, they may not work well for problems with singular solutions due to various facts such as corner singularities, non-matching boundary conditions, non-smooth coefficients. If the form of the singular expansion for the solution is known, we develop a Muntz Galerkin method which is based on specially tuned Muntz polynomials to deal with the singular behaviors of the underlying problems, and show that it provide optimal error estimates. On the other hand, if the Muntz Galerkin method is not applicable or efficient,we present a new extended spectral-Galerkin method which allows us to split it into two separate problems: one is to find an approximationfor the smooth part by a usual spectral method, the other is to determine an approximation to the singular part with $k$ terms by solving a $k\times k$ system. So the new method is very easy to implement, very efficient and is capable of providing very accurate approximations for a class of singular problems. We will present ample numerical results for a variety of problems with singular solutions, including fractional PDEs, to demonstrate the effectiveness of our approaches.
