Histogram filtering: A technique to optimize wave functions for use in Monte Carlo simulations

Wave functions are optimized using a histogram-based technique that deals with the statistical error plaguing traditional Monte Carlo optimizations. Following a sensitivity study on H2+, we variance- and energy-optimize explicitly correlated wave functions for He (up to 18 variational parameters), H...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The Journal of chemical physics 1999-12, Vol.111 (22), p.9971-9981
Hauptverfasser: Snajdr, Martin, Dwyer, Jason R., Rothstein, Stuart M.
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Wave functions are optimized using a histogram-based technique that deals with the statistical error plaguing traditional Monte Carlo optimizations. Following a sensitivity study on H2+, we variance- and energy-optimize explicitly correlated wave functions for He (up to 18 variational parameters), H2 (up to 10 parameters), and LiH (up to 32 parameters). To gauge the convergence of the variational energy as the quality of the wave functions improves, we adopt some simple ones from the literature in addition to more sophisticated ones unique to this paper. One for LiH has the lowest variational energy for a compact, explicitly correlated wave function to date. For the molecules we determine the optimal bond distance at the same time as we optimize either the variational energy or the variance of the local energy, but agreement with experiment is reasonable only for the energy optimizations. The energy of variance-optimized molecular wave functions appears to converge slowly to the energy optimization results as the wave function quality improves. Variance optimizations done keeping the bond distance fixed equal to the exact value improves the energy somewhat.
ISSN:0021-9606
1089-7690
DOI:10.1063/1.480351