Extending the Time and Length Scale of Electronic Structure Methods Through Force Matching

点击翻译按钮获取中文摘要

WIth support from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry, Professor Feng Wang of the University of Arkansas at Fayetteville is developing a new method to connect forces from quantum mechanics to the understanding chemical properties. While any property, in principle, can be computed with the fundamental equations of quantum mechanics, such computations quickly become too complex to be practical. The new method, adaptive force matching (AFM), maps such first principles calculations to simple mathematical expressions and thus allow macroscopic properties of materials to be computed with a tractable computational cost. Specfic systems to be studied by the Wang research team will involve predictions of solubility, of minerology processes under cryogenic conditions, and of properties of microporous crystals. The research is expected to have long range scientific broader impacts in functional materials, and even potentially drug design and planetary science. Dr. Feng Wang will also train next generation scientists with a collaboration between the departments of chemistry and theatre, where students in chemistry will be mentored by master of fine arts students for engaging presentations of chemistry concepts. Educational videos will be produced as part of this collaboration.Under this award, Professor Feng Wang and his research group are developing methods that will extend the time and length scale of electronic structure calculations to enable predictive simulations of ensemble properties and of slow dynamics. The adaptive force matching method maps potential energy surfaces of electronic structure methods computed in the condensed phase to simple molecular mechanics energy expressions. The advantage of this technique is to enable molecular dynamics simulations at larger length and time scales on a potential energy surface that remains accurate to the reference electronic structure. In the proposed work, further development of AFM will incorporate better regularization of parameters through cross validations and Baysesian inference. A new method to treat polarization is also being developed. The method will be used to predict solubilities of molecular liquids, to study mineralogical processes under cryogenic conditions, and to compute properties of microporous crystals.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.