Large-Scale Computational Screening of Aluminosilicate Zeolites for Molecular Capture, Storage and Separation
Author(s)
Daou, Alan S. S.
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Industrial alkane separations traditionally rely on energy-intensive distillation processes. To mitigate this, adsorption-based separations using porous materials, specifically zeolites, offer a non-thermal alternative. However, large combinations of exchanged cations, framework topologies, and aluminum compositions exist, making it challenging to identify the optimal candidates. Large scale computational studies present an efficient approach to screen candidates and design processes before lab experimentation. These studies rely on accurate results from classical simulation techniques such as Grand Canonical Monte Carlo or Molecular Dynamics and an efficient workflow. The accuracy and/or transferability of the available forcefields often limit the scale of these studies.
This thesis aimed to address these limitations by developing a suite of tools for the high-throughput screening of silica and aluminosilicate zeolites for separations. We first studied the impact of intrinsic flexibility on adsorption properties in zeolites and confirmed the viability of rigid frameworks, an essential step for the development of computationally efficient force fields.
We then developed a fully transferable force field based on first-principles quantum mechanical methods that can accurately describe both the adsorption and diffusion properties of alkanes and some small adsorbates in siliceous and cationic zeolites. By fitting these force fields to DFT/CC energies, we retain the accuracy of QM methods. To streamline screening, we developed an algorithm based on DFT methods to efficiently generate computationally ready cationic zeolite structures with accurate Si/Al ratio dependent lattice constants. These allowed us to obtain simulation results that are quantitatively accurate to experimental measurements.
This led to the creation of a computational screening workflow for adsorption-based methane/butane separation in zeolites using the algorithmically generated structures and DFT-parametrized force fields. We then developed machine learning tools for the screening of real and hypothetical zeolites to facilitate high throughput screening. These machine learning models, trained on accurate data from our force fields, is a valuable complement to classical zeolite screening simulations.
Overall, the work presented here serves to present more accurate, faster methods to screen zeolites for separation purposes.
Sponsor
Date
2024-04-27
Extent
Resource Type
Text
Resource Subtype
Dissertation