Designing Mxene Catalysts For Clean Energy Chemistries Using High-Throughput First-Principles Calculations And Data-Driven Methods
The field of heterogeneous catalysis has been prompted to shift toward designing catalysts that can perform beneficial chemistries at ambient conditions. Such materials should have high activity and stability and avoid issues prevalent in other traditional catalysts that are not earth-abundant, chemically efficient, or with high selectivity to carry out these reactions. Two coupled electrochemical reactions - hydrogen evolution (HER) and nitrogen reduction (NRR) - embody this problem, where expensive materials are highly active but not selective for the chemistry of interest. Therefore, we continue searching to explore new classes of materials that can tune the active sites or reactivity and bring activity to maximum chemical rates.
In this thesis, we study a new class of materials called 2D MXenes that have intriguing electronic and mechanical properties. These anisotropic materials consisting of basal planes between interlayer spacing are derived from 3D bulk materials and have significantly higher reactivity, conductivity, and surface area when compared to their MAX phase precursors. Therefore MXenes have been of interest for catalysis applications. Past experimental characterization confirmed MXenes to have surface functionalization. However, previous theoretical and computational studies of MXenes as HER and NRR catalysts have modeled the basal plane functionalization to be pristine. To counter this, we model MXenes with different functional groups that demonstrate the wide range of reactivity of the basal planes. Also, we form new catalytic trends and compare them to previously computed scaling relations of existing metals like metals and metal compounds. However, just altering the basal plane functionalization does not fully demonstrate the tunability of MXenes for improving their catalytic activity.
Therefore, we study the effect of physicochemical changes to MXenes as catalysts for the electrochemical HER and NRR. We perform density functional theory calculations to predict the material properties and their interactions with H* and *NxHy intermediates. Such changes include altering chemical structure, doping, straining, and supporting of MXene materials. We find that all these materials modifications can impact the adsorption energies of intermediates and hence the NRR/HER activity. The sulfidation and biaxial straining of MXenes also increased the HER activity of terminated MXenes. We then compile and categorize data from the above physicochemical studies into a database and design a machine learning study where we featurize the data to predict the adsorption energies for these coupled reactions. One feature alone is infeasible to predict adsorption energies, prompting the development of an extra tree regressor. We demonstrate promising predictability in our ML model with an error of 0.24 eV and R2 of 0.95. Electronic structure features of the functional group on the basal plane show that sulfidation of MXenes improves NRR thermodynamics. This thesis pushes forward the catalysis field by elucidating the effect of tuning 2D materials to enhance their chemical activity and the usefulness of data analytics and machine learning to assist materials discovery of novel catalysts for the future clean energy economy.