From Additive Average Schwarz Methods to Non-overlapping Spectral Additive Schwarz Methods

12/25/2020 ∙ by Yi Yu, et al. ∙ 0

In this paper, we design and analyze new methods based on Additive Average Schwarz–AAS Method introduced in [3]. The new methods are designed for elliptic problems with highly heterogeneous coefficients. The methods are of non-overlapping type and the subdomain interactions are obtained via the coarse space. The first method is the Minimum Energy Schwarz–MES Method, it has the best coarse space with one degree of freedom per subdomain. The condition number of MES is always smaller than AAS methods. The second class of methods is the Non-overlapping Spectral Additive Schwarz–NOSAS methods, they are based in low-rank discrete energy harmonic extension in each subdomain. We prove that the condition number of NOSAS methods do not depend on the coefficients, and with a specific choice of the threshold of the eigenvalues, we can obtain condition numbers for the preconditioned system similar to the AAS method with constant coefficients. Additionally, the NOSAS methods have good parallelization properties, and the size of the coarse problem is equal to the total number of eigenvalues chosen in each subdomain and which might be related to the number of islands or channels with high coefficient that touch the boundary of the subdomains.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.