Trace Conserving Purification for Linear Scaling [O(N)] Methods: A First Enhancement to CP2K

Report No. ARL-CR-0746
Authors: Jonathan Mullin
Date/Pages: September 2014; 20 pages
Abstract: Implementation of the low-order purification method as an alternative to diagonalization for self-consistent density function theory calculations lowers the memory and time to solution for analysis of chemical systems. There is some improvement for low-sparsity systems, but the highest speed-up is seen in low band gap and highly sparse matrix problems. This work has been conducted and tested in the CP2K program.
Distribution: Approved for public release
  Download Report ( 0.784 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: September 1, 2014