The Markovian Multiagent Monte-Carlo method as a differential evolution approach to the SCF problem for restricted and unrestricted Hartree-Fock and Kohn-Sham-DFT
In this paper we present the Markovian Multiagent Monte-Carlo Second Order Self-Consistent Field Algorithm (M3-SOSCF). This algorithm provides a highly reliable methodology for converging SCF calculations in single-reference methods using a modified differential evolution approach. Additionally, M3...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article (Journal) |
| Language: | English |
| Published: |
7 October 2023
|
| In: |
The journal of chemical physics
Year: 2023, Volume: 159, Issue: 13, Pages: 1-14 |
| ISSN: | 1089-7690 |
| DOI: | 10.1063/5.0159737 |
| Online Access: | Verlag, lizenzpflichtig, Volltext: https://doi.org/10.1063/5.0159737
|
| Author Notes: | Linus Bjarne Dittmer and Andreas Dreuw |
| Summary: | In this paper we present the Markovian Multiagent Monte-Carlo Second Order Self-Consistent Field Algorithm (M3-SOSCF). This algorithm provides a highly reliable methodology for converging SCF calculations in single-reference methods using a modified differential evolution approach. Additionally, M3 is embarrassingly parallel and modular in regards to Newton-Raphson subroutines. We show that M3 is able to surpass contemporary SOSCFs in reliability, which is illustrated by a benchmark employing poor initial guesses and a second benchmark with SCF calculations which face difficulties using standard SCF algorithms. Furthermore, we analyse inherent properties of M3 and show that in addition to its robustness and efficiency, it is more user-friendly than current SOSCFs. |
|---|---|
| Item Description: | Gesehen am 06.05.2024 |
| Physical Description: | Online Resource |
| ISSN: | 1089-7690 |
| DOI: | 10.1063/5.0159737 |