Date of Award
January 2025
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Energy Engineering
First Advisor
Olusegun Stanley Tomomewo
Abstract
The rising demand for clean and sustainable energy sources has underscored the critical role of hydrogen in the global energy transition. However, the widespread adoption of green hydrogen remains constrained by the high cost and scarcity of platinum-based electrocatalysts used in Proton Exchange Membrane (PEM) electrolyzers. This dependence inflates production costs and poses long-term risks related to supply security and scalability. While numerous studies have explored alternative materials, the discovery of efficient and stable non-platinum catalysts remains a major bottleneck due to the vastness of chemical space and the resource-intensive nature of experimental screening. To address this challenge, this study applied a machine learning approach to systematically explore and classify transition metal–ligand complexes based on electronic and structural features relevant to Hydrogen Evolution Reaction (HER) performance. Using the tmQM dataset, which contains over 108,000 transition metal complexes from the Cambridge Structural Database (CSD), we selected five key descriptors: HOMO–LUMO gap, dipole moment, molecular size, metal node degree, and total charge, to represent catalytic activity and structural robustness. Three unsupervised clustering algorithms: K-Means, DBSCAN, and Gaussian Mixture Models (GMM), were applied to partition the dataset into chemically and electronically distinct classes. By comparing their clustering performance, the most robust model was selected to guide the identification of promising candidates that align with desirable HER-relevant features, including low HOMO–LUMO gaps, high dipole moments, and moderate coordination numbers. The resulting clusters revealed multiple classes of metal–ligand complexes with favorable catalytic and structural properties. One cluster was enriched with candidates exhibiting strong electronic reactivity and solvation potential, coupled with neutral charge and manageable size, indicative of both catalytic efficiency and synthetic feasibility. Importantly, all shortlisted complexes are structurally documented in crystallographic databases, which enables direct experimental validation and minimizes trial-and-error in material design.
Recommended Citation
Benarbia, Achouak, "Machine Learning-Guided Discovery Of Metal-Ligand Catalysts For Affordable Green Hydrogen Produced Via Proton Exchange Membrane Electrolyzers" (2025). Theses and Dissertations. 7098.
https://commons.und.edu/theses/7098