Loading...
Combining free energy calculations with machine learning predictions of adsorption to design metal-organic frameworks for chemical separations
Anderson, Ryther M.
Anderson, Ryther M.
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2021
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
2022-06-25
Abstract
The separation of chemical mixtures into their sufficiently pure components is a necessary step in essentially all industrial chemical processes. Presently, the predominant method for industrial chemical separation is distillation, an energy-intensive process. Therefore, the development of less energy-intensive separation technologies can reduce industrial energy usage by a consequential margin. The development of these technologies hinges on the discovery and design of materials (i.e. separation agents) that can selectively interact with mixture components on the molecular level. Metal-organic frameworks (MOFs) are a type of nanoporous crystal that, due to their uniquely tunable pore geometry and chemistry, have particular promise for use as separation agents. MOF tunability results from the modularity of their construction. However, this modularity also results in an intractably large number of candidate MOFs for any given separation. Furthermore, many candidate MOFs are likely to be "hypothetical", meaning they have not been synthesized. A key challenge, then, is to identify, for a given separation, high-performing MOFs that can also be synthesized. Considering the aforementioned structural diversity of MOFs and the variety of conditions encountered in industrial separations, screening MOFs for use as separation agents requires efficient computational prediction of their performance and synthetic likelihood. In this thesis, we explore strategies for the identification of high-performing and thermodynamically accessible MOFs for use in various gas separations. Integral to this work is the development of increasingly universal machine learning models capable of predicting the performance of diverse MOFs in multiple separations. We can use these models both to rapidly identify high-performing MOFs and to derive generalized design rules that can be applied to existing MOF platforms. In addition, we develop methods for the large-scale calculation of MOF free energy, allowing us to derive criteria for MOF synthetic likelihood, at least from a thermodynamic perspective. When used in concert, these methods create a unified approach for the efficient computational discovery and design of MOFs for chemical separations.
Associated Publications
Rights
Copyright of the original work is retained by the author.