This presentation focuses on the development of theoretical and simulation methods aimed at optimizing the conversion of a variety of coals into high-value carbon fibers.  The complexity of coal structure and associated chemical conversion processes demands accurate, yet computationally highly efficient methodologies for predictive simulations to guide experimental efforts.  Our computational chemistry- and machine learning-based methods address two major challenges: 1) generate accurate, atomic-scale structural models for the various types of coals through inverse structural design based on theoretical and experimental spectroscopic characterization, and 2) accurately describe the chemical breakdown processes occurring during the solvation extraction processes to produce isotropic pitch, and the subsequent heat soaking process to produce the mesophase pitch from which high-quality carbon fibers are spun.   The ultimate goal of our development work is a “virtual reactor” which would inform which type of coals/solvent mixtures and environmental parameters (temperature, pressure, reaction time) are needed for producing carbon products, such as carbon fibers, with tunable physical and mechanical properties.  Our multiscale method development involves accurate coupled cluster calculations to generate high-level reaction energy data, over density-functional tight-binding (DFTB) calculations to produce large molecular training sets for neural network-based reactive simulations of the coal breakdown process.

Computational chemistry:  Machine learning-based methods to optimize the conversion of coals to carbon fibers and other value-added products

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.