Skip to main navigation Skip to search Skip to main content

A perturbation-method-based post-processing for the planewave discretization of Kohn-Sham models

  • Eric Cancès
  • , Geneviève Dusson
  • , Yvon Maday
  • , Benjamin Stamm
  • , Martin Vohralík
  • Sorbonne Université
  • Institut de Calcul et de la Simulation
  • Institut Universitaire de France
  • Women and Infants Hospital of Rhode Island-Warren Alpert Medical School of Brown University
  • INRIA Institut National de Recherche en Informatique et en Automatique

Research output: Contribution to journalArticlepeer-review

Abstract

In this article, we propose a post-processing of the planewave solution of the Kohn-Sham LDA model with pseudopotentials. This post-processing is based upon the fact that the exact solution can be interpreted as a perturbation of the approximate solution, allowing us to compute corrections for both the eigenfunctions and the eigenvalues of the problem in order to increase the accuracy. Indeed, this post-processing only requires the computation of the residual of the solution on a finer grid so that the additional computational cost is negligible compared to the initial cost of the planewave-based method needed to compute the approximate solution. Theoretical estimates certify an increased convergence rate in the asymptotic convergence range. Numerical results confirm the low computational cost of the post-processing and show that this procedure improves the energy accuracy of the solution even in the pre-asymptotic regime which comprises the target accuracy of practitioners.

Original languageEnglish
Pages (from-to)446-459
Number of pages14
JournalJournal of Computational Physics
Volume307
DOIs
Publication statusPublished - 15 Feb 2016

Keywords

  • Density-functional theory
  • Nonlinear eigenvalue problem
  • Perturbation method
  • Planewave approximation
  • Post-processing

Fingerprint

Dive into the research topics of 'A perturbation-method-based post-processing for the planewave discretization of Kohn-Sham models'. Together they form a unique fingerprint.

Cite this