Rigorous force field optimization principles based on statistical distance minimization
- Chemical Sciences Division, Geochemistry & Interfacial Sciences Group, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6110 (United States)
We use the concept of statistical distance to define a measure of distinguishability between a pair of statistical mechanical systems, i.e., a model and its target, and show that its minimization leads to general convergence of the model’s static measurable properties to those of the target. We exploit this feature to define a rigorous basis for the development of accurate and robust effective molecular force fields that are inherently compatible with coarse-grained experimental data. The new model optimization principles and their efficient implementation are illustrated through selected examples, whose outcome demonstrates the higher robustness and predictive accuracy of the approach compared to other currently used methods, such as force matching and relative entropy minimization. We also discuss relations between the newly developed principles and established thermodynamic concepts, which include the Gibbs-Bogoliubov inequality and the thermodynamic length.
- OSTI ID:
- 22489697
- Journal Information:
- Journal of Chemical Physics, Vol. 143, Issue 14; Other Information: (c) 2015 AIP Publishing LLC; Country of input: International Atomic Energy Agency (IAEA); ISSN 0021-9606
- Country of Publication:
- United States
- Language:
- English
Similar Records
Algorithmic randomness and physical entropy
New perspectives in thermodynamics