Nanoscience and Microsystems ETDs

Publication Date

Fall 11-16-2020

Abstract

Classical potentials that are capable of describing charge transfer and charge polarization in complex systems are of central importance for classical atomistic simulation of biomolecules and materials. Current potentials—regardless of the system—do not generalize well, and, with the exception of highly-specialized empirical potentials tuned for specific systems, cannot describe chemical bond formation and breaking. The charge-transfer embedded atom method (CT-EAM), a formal, DFT-based extension to the original EAM for metals, has been developed to address these issues by modeling charge distortion and charge transfer in interacting systems using pseudoatom building blocks instead of the electron densities of isolated atoms. CT-EAM incorporates the concept of the atom-in-molecule (or pseudoatom) as a key feature of its potential design. This enables the electron density to serve as a mechanism for transmitting the underlying quantum mechanical behavior of the system’s electrons directly into the potential. It is this feature that opens the prospect for CT-EAM to accurately describe reactive interactions in complex biophysical and materials systems. In addition, CT-EAM has the important advantage of being DFT-based. It thus has a formal theoretical foundation, allowing it to be fitted with a small number of parameter, and giving it the potential for good generalizability and transferability. In this dissertation, we focus on the first major step in successfully parametrizing a CT-EAM potential to describe amino acid interactions, and ultimately the dynamics of larger biomolecular systems such as proteins. We describe two important accomplishments toward this end. First, we validate the physics of the ensemble-of-ensemble atom-in-molecule density decomposition (“density deconstruction”) consistent with the statistical ensemble design of the CT-EAM energy functional. We develop a nested, grid-based methodology for fitting a weighted sum of basis densities to a given molecular density, with all densities computed at a high level of quantum mechanical theory. This procedure is applied to two exemplar systems: LiF and CO. LiF is studied because it is a simple ionic system with clear charge transfer states. CO is of particular interest since it contains two of the five atoms (C, N, O, H, S) appearing in the twenty naturally-occurring amino acids and historically presented a challenge to quantum chemistry due to its small dipole moment. As a metric for assessing the quality of our density decompositions, we compute the effective charge transfer in these systems, obtaining surprising and remarkable agreement with the results of topological decompositions using the very different QTAIM methodology of R. F. W. Bader. The second contribution of this work is the development of a systematic methodology for fitting analytic forms to the pseudoatom basis densities, which consist of neutral, integer-charge, and excited states of the isolated atoms of the given molecular system being modeled. We develop and implement a constrained-fitting strategy for accurately modeling sphericalized versions of these basis densities, exploring a series of analytic models to capture short-range, long-range, and intermediate-range shell structure behavior, subject to known, formal constraints. Results are presented for H, Li, C, N, O, and F, which include the principal elements relevant to the study of biophysical molecules.

Keywords

potentials, force fields, reactive bond formation, biomolecular, charge distortion, charge transfer, charge-transfer embedded atom method, pseudoatom, atom-in-molecule, basis densities

Document Type

Dissertation

Language

English

Degree Name

Nanoscience and Microsystems

Level of Degree

Doctoral

Department Name

Nanoscience and Microsystems

First Committee Member (Chair)

Susan R. Atlas

Second Committee Member

Andrew P. Shreve

Third Committee Member

David H. Dunlap

Fourth Committee Member

Marek Osinski

Comments

Corrected title page per review.

Download

Share

COinS