Published 2003 | Version Accepted Version
Technical Report Open

An Odyssey into Local Refinement and Multilevel Preconditioning II: Stabilizing Hierarchical Basis Methods

Abstract

The concept of a stable Riesz basis plays a crucial role in the design of efficient multilevel preconditioners. In this article, we present a thorough analysis of the relationship between Riesz bases, matrix conditioning, and multilevel stability criteria, and the impact of local adaptive mesh refinement on these concepts. Wavelet-like modi cations have recently been successful in optimally stabilizing hierarchical basis methods in the setting of quasiuniform meshes. We closely examine the wavelet modi ed hierarchical basis (WMHB) methods by Vassilevski and Wang, and we extend the existing two- and three-dimensional optimality results for the quasiuniform setting to local refinement settings. Such modi cations rely primarily on establishing an optimal BPX preconditioner for the refinement procedures under consideration. The local refinement procedures we consider are the well-known red-green and red refinements in two dimensions, and their natural extensions to to three dimensions. The first article in this series established the fundamental assumption for the analysis of the WMHB preconditioner, together with a number of supporting results pertaining to BPX preconditioner. With these supporting tools in place, we prove the optimality of WMHB preconditioner. In the presence of continuously differentiable partial differential equation coefficients, we extend the optimality results for multiplicative WMHB method to locally re ned two-dimensional meshes by using two different red refinement strategies. Without such smoothness assumptions on the coefficients, we show that the early suboptimal results can also be extended to locally refined meshes. An interesting implication of the optimality of WMHB preconditioner is the H¹-stability of the linear operators used. In the limiting case where such operators reduce to L₂-projection, one can guarantee an a priori H1-stability of L₂-projection without verifying somewhat cumbersome mesh conditions after refinement has taken place. The existing a posteriori approaches in the literature dictate a reconstruction of the mesh if such conditions cannot be satis ed. In search of optimal results, we prove that the optimality can still be achieved for the whole class of the mentioned local refinements by choosing an additive version of WMHB with coefficients in L_∞. The proof techniques allow extensions of the optimality to arbitrary spatial dimensions d ≥ 1.

Additional Information

The first author was supported in part by the Burroughs Wellcome Fund through the LJIS predoctoral training program at UC San Diego, in part by NSF (ACI-9721349, DMS-9872890), and in part by DOE (W-7405-ENG-48/B341492). Other support was provided by Intel, Microsoft, Alias|Wavefront, Pixar, and the Packard Foundation. The second author was supported in part by NSF CAREER Award 9875856 and in part by a UCSD Hellman Fellowship. The authors thank R. Bank and P. Vassilevski for many enlightening discussions.

Attached Files

Accepted Version - cit-asci-tr161.pdf

Files

cit-asci-tr161.pdf

Files (1.1 MB)

Name Size
md5:e792c8ac02e09054b144ba7bdd5bf0f3
1.1 MB Preview Download

Additional details

Identifiers

Eprint ID
119190
Resolver ID
CaltechAUTHORS:20230210-175634271

Funding

Burroughs Wellcome Fund
University of California, San Diego
NSF
ACI-9721349
NSF
DMS-9872890
Department of Energy (DOE)
W-7405-ENG-48
Department of Energy (DOE)
B341492
Intel
Microsoft
Alias|wavefront
Pixar
David and Lucile Packard Foundation
NSF
DMS-9875856
Hellman Fellowship

Dates

Created
2023-02-11
Created from EPrint's datestamp field
Updated
2023-02-11
Created from EPrint's last_modified field

Caltech Custom Metadata

Caltech groups
Accelerated Strategic Computing Initiative
Series Name
ASCI Technical Report
Series Volume or Issue Number
ASCI-TR161