Molecular modeling (i.e., Monte Carlo and molecular dynamics methods) are becoming computationally-accessible for large-scale simulations of various thermophysical and mechanical properties of substances.

Relatively simple, semiempirical force field-based models appear to provide sufficient framework for quantitative predictions of a range of properties. The key issue with predictive molecular modeling is the parametrization of the force field. There are a large number of force fields in existence and significant efforts are spent on their development and improvement. However, to-date, development of force field models is not performed in a systematic manner. The parameters are usually obtained to reproduce limited experimental observations, often of questionable or unknown quality. Furthermore, the parameter adjustments are usually performed in a brute-force manner, without considerations of nonlinear coupling between multiple parameters. As a result, the models (that reproduce the data that they were calibrated against) often poorly extrapolate to different conditions or different properties. Uncertainties associated with such extrapolations are not easily quantifiable. Our goal is to use the extensive experimental data archives available at NIST/TRC and advanced multidimensional optimization techniques to generate self-consistent, predictive force-filed models applicable over wide range of conditions.

 

key words

Molecular dynamics; Molecular simulation; Monte Carlo; Force field; Thermodynamic properties; Transport properties;

Citizenship:  Open to U.S. citizens

Level:  Open to Postdoctoral applicants