Bayesian Calibration of Transferable, Coarse-Grained Force Fields

Published

April 15, 2016

Author(s)

Thomas W. Rosch, Frederick R. Phelan Jr., Paul Patrone

Abstract

Generating and calibrating forces that are transferable across a range of state-points remains a challenging problem in coarse-grained (CG) molecular dynamics (MD). In this work, we present a Bayesian correction algorithm, inspired by ideas from uncertainty quantification and numerical analysis, for addressing this problem. The main idea behind our algorithm is to estimate functional derivatives (i.e. linear sensitivities) of CG simulations relative to small changes in the forces. Using these sensitivities, we construct a probabilistic update model in which we constrain the output of a CG method to yield target predictions to within statistical significance. Using density- temperature relationships as a running example, we demonstrate that this correction algorithm is robust to various choices the modeler can make when coarse-graining. Moreover, we show that our approach can speed up coarse-graining by reducing the number of atomistic simulations needed to construct initial estimates of CG forces.

Citation

Journal of Chemical Theory and Computation

Pub Type

Journals

Download Paper

Local Download

Keywords

Coarse-graining, Uncertainty quantification, Bayesian probability

Citation

Rosch, T. , Phelan Jr., F. and Patrone, P. (2016), Bayesian Calibration of Transferable, Coarse-Grained Force Fields, Journal of Chemical Theory and Computation, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=919663 (Accessed October 21, 2025)