03/12/2025
By Faranak Hatami

Candidate Name: Faranak Hatami
Defense Date &Time: Wednesday, March 26, 11 a.m. EST
Location: Olney Hall, Room 218; please contact Faranak Hatami (faranak_hatami@student.uml.edu) for a zoom link.

Dissertation Title: Transport Property Analysis and Multi-Objective Optimization of Force Fields Parameters for Tri-Butyl-Phosphate

Committee Members: 
Valmor F. de Almeida, Ph.D., Department of Chemical Engineering, Nuclear Engineering, University of Massachusetts Lowell
Erin Rachel Bertelsen, Ph.D., Department of Physics and Applied Physics, University of Massachusetts Lowell
Ofer Cohen, Ph.D., Department of Physics and Applied Physics, University of Massachusetts Lowell

Abstract:

Tri-n-Butyl Phosphate (TBP) is a critical component in hydrometallurgical solvent extraction, widely used for separating uranium, plutonium, and zirconium in nuclear fuel reprocessing. Despite its importance, the current status of atomistic simulations aimed at predicting thermodynamic and transport properties on par to experimental measurements is unsatisfactory. This work was a step towards improving the capability of predicting properties of this industrially relevant solvent molecule in the liquid form, which serves as framework for applications to other proposed solvent extraction agents. The investigation carried out here was aimed at demonstrating the performance of existing and newly proposed force fields for accurately predicting properties of interest for TBP using classical Molecular Dynamics (MD) simulations. This was done in two parts.

In the first part, using and developing various force fields with and without polarization, an extensive comparison of predictive capability of representative properties (mass density, heat of vaporization, electric dipole moment, self-diffusion coefficient, and shear viscosity) was made. These predictions involved both equilibrium and non-equilibrium simulations with a total of 20 force fields (10 force fields without polarization and their polarized counterparts). Some of the non-polarized force fields were available in the literature, some were proposed in this work based on intuition, and all polarized fields were first described here. This first part of the work established that the best performing force field, Polarized AMBER-MNDO, predicts all properties within a combined 74.3% deviation from experimental data. This value is largely skewed by the prediction of shear viscosity at a deviation from the experimental value of -65%. This is one but many aspects of the investigation in this first part described later in this thesis, and one of the reasons motivating the development of part two.

In the second part, guided by the results of the first part, a Genetic-MD optimization method was developed to reduce the error of property predictions as compared to experimental values. The optimization approach developed and tested here was a multi-objective minimization of force field parameters (Lennard-Jones and electric charge) using the Pareto approach, specifically tailored for TBP. The study tackles the challenges of computing TBP's thermophysical properties by optimizing LJ and atomic charge parameters through a Non-dominating Sorting Genetic Algorithms (NSGA) and Neural Network (NN) optimization model.
This study demonstrates that integrating Multi-Objective Optimization (MOO) with MD and NN enhances force field parameterization.