Accurate and Fast Geometry Optimization with Time Estimation and Method Switching
Geometry optimization is an important task in quantum chemical calculations to analyze the characteristics of molecules. A top concern on it is a long execution time because time-consuming energy and gradient calculations are repeated across several to tens of steps. In this work, we present a schem...
Gespeichert in:
Hauptverfasser: | , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Geometry optimization is an important task in quantum chemical calculations
to analyze the characteristics of molecules. A top concern on it is a long
execution time because time-consuming energy and gradient calculations are
repeated across several to tens of steps. In this work, we present a scheme to
estimate the execution times of geometry optimization of a target molecule at
different accuracy levels (i.e., the combinations of ab initio methods and
basis sets). It enables to identify the accuracy levels where geometry
optimization will finish in an acceptable time. In addition, we propose a
gradient-based method switching (GMS) technique that reduces the execution time
by dynamically switching multiple methods during geometry optimization. Our
evaluation using 46 molecules in total shows that the geometry optimization
times at 20 accuracy levels are estimated with a mean error of 29.5%, and GMS
reduces the execution time by up to 42.7% without affecting the accuracy of
geometry optimization. |
---|---|
DOI: | 10.48550/arxiv.2404.12842 |