Premium
Acceleration of self‐consistent‐field convergence by combining conventional diagonalization and a diagonalization‐free procedure
Author(s) -
Baldes Alexander,
Klopper Wim,
Šimunek Ján,
Noga Jozef,
Weigend Florian
Publication year - 2011
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.21877
Subject(s) - convergence (economics) , acceleration , matrix (chemical analysis) , mathematics , field (mathematics) , scheme (mathematics) , computer science , algorithm , mathematical optimization , density matrix , mathematical analysis , physics , quantum mechanics , pure mathematics , materials science , economics , composite material , economic growth , quantum
Abstract A new scheme that combines conventional matrix diagonalization with the recently proposed diagonalization‐free algorithm has been developed to obtain the density matrix for the next self‐consistent‐field iteration from the Fock matrix of the current iteration. In this manner, the advantages of the two methods are combined. The more rapid convergence of the diagonalization‐free algorithm for density matrices rather close to self consistence and the more robust convergence of the conventional matrix diagonalization further away from self‐consistence. The scheme has been implemented in the one‐ and two‐component self‐consistent‐field procedures in the program system TURBOMOLE. The number of iterations is typically reduced by about 10%, but savings are usually much larger for slowly converging cases. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011