Machine Learning Assisted Study of Early-Stage MOF Self-Assembly: Structural Characterization and Multi-Scale Modeling
Date
2024-01-25
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Metal-Organic Frameworks (MOFs) are an important class of materials with a broad range of applications (e.g., gas storage and catalysis). While a great number of MOF compounds are reported in the literature, where these studies have described a multitude of ways of synthesizing MOFs, there have been few studies probing the self-assembly process. Mechanisms of MOF self-assembly are amenable to Molecular Dynamic (MD) simulations, although there have been very few such studies to date, and even they suffer from unreasonable assumptions or fall short due to the limited timescales available to all-atom simulations. To allow simulations to reach much longer timescales, this thesis project looks to find an optimal way to develop effective potentials. The all-atom molecular interactions have been redefined so that the explicit solvent molecules could be omitted from the system without losing the solvent effect on the self-assembly processes. Since a large majority of atoms in the explicit system are the solvent molecules, we expect the effective potentials would result in a dramatic acceleration of the simulations. The complexity of the potential space makes the optimization slow and sensitive to small perturbations. We have explored the challenges associated with the iterative optimization of effective potentials and worked especially on the transferability of these potentials from a small MOF-like system to an ever-evolving environment in MOF self-assembly. Another important aspect of the project is to characterize the structures seen in the self-assembly process and to quantify the structural order, which can be further used to bias the simulations and further speed up the sampling, thus bridging the gap between experimental and computational studies.
Description
Keywords
Metal-organic framework, Self-assembly, Molecular simulation, Multi-scale modeling, Machine learning, Effective potentials, Particle swarm optimization
Citation
Shukla, R. (2024). Machine learning assisted study of early-stage MOF self-assembly: structural characterization and multi-scale modeling (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.