Theses and Dissertations

Issuing Body

Mississippi State University

Advisor

Rai, Neeraj

Committee Member

Toghiani, Hossein

Committee Member

Kundu, Santanu

Committee Member

Bharadwaj, Vivek

Date of Degree

8-7-2025

Original embargo terms

Visible MSU Only 1 year

Document Type

Dissertation - Campus Access Only

Major

Engineering (Chemical Engineering)

Degree Name

Doctor of Philosophy (Ph.D.)

College

James Worth Bagley College of Engineering

Department

Dave C. Swalm School of Chemical Engineering

Abstract

It is crucial to harness energy from sustainable sources to combat the dwindling nonrenewable feedstocks used to produce everyday commodities and fuels. Liquid-phase heterogeneous catalysis has demonstrated remarkable product selectivity and yields for reactions relevant to biomass valorization under mild operating conditions. However, active site characterization is challenging due to the complex solvated environment and interface. Ground-state first-principles (FP) molecular simulations have provided insights into reactive systems in catalysis. These calculations, however, scale poorly with the number of electrons. Observing atomic motion with molecular dynamics (MD) or sampling static properties with Monte Carlo (MC) simulations at statistically relevant lengths- and timescales are not computationally tractable from first principles. A faster approach to FP calculations that retains the underlying potential energy surface is necessary to enable a high throughput understanding of reactive phenomena at an atomic resolution and to search for optimal catalyst properties and operating conditions. Machine-learned interatomic potentials (MLIPs) can serve as a surrogate model for first principles (FP) calculations, enabling large-scale simulations of complex systems. In this research, we first explore the interaction between methanol and water with the surface of a Brønsted acid MWW zeolite nanosheet, using FP-MD to illustrate the need for MLIPs. Next, we show that MLIPs can be utilized to elucidate the nature of transition metal surface active sites and capture environmental effects by simulating hydrogen dissociation over molybdenum carbide polymorphs. We also apply these methods to study solvent interactions with open and closed-site Sn-BEA zeolites. Finally, we integrate MC subroutines into the Atomic Simulation Environment to enable FP Grand Canonical Monte Carlo (MC) and MC with machine-learned interatomic potentials. The research presented here illustrates that MLIPs can be utilized to simulate complex catalytic systems, retaining the underlying ab initio potential energy surface. We also provide the scientific community access to an open-source framework for FP Grand Canonical Monte Carlo simulations.

Share

COinS