Application of density functional theory and machine learning in heterogenous-based catalytic reactions for hydrogen production

Various feedstocks such as natural gas, glycerol, biomass, methanol, ethane, and other hydrocarbons can be reformed to generate hydrogen as a viable alternative source of renewable energy. Also, hydrogen is generated via other processes not associated with reforming including electrolysis, thermolysis, and photolysis. The reforming of the different feedstock for hydrogen reaction generally requires the utilization of heterogeneous catalysts to speed up the reactions and reduce the energy used up in the reaction. Experimental studies provide an understanding of the reaction mechanism and the nature of the reactants and products of the reaction. Computational studies involving density functional theory provide even greater insights into these reactions. Its combination with machine learning provides huge potentials for the study and discovery of technologies for hydrogen production but remains underutilized. The use of both computational techniques has widely been adjudged as the most economical and precise means of screening multiple catalysts in the heterogeneous reactions involved in hydrogen production processes. This paper reviews the application of density functional theory and machine learning in thermochemical reactions associated with the production of hydrogen. It also highlights the state-of-the-art computational methodologies employed in the design of hydrogen production technologies such as methane pyrolysis, steam methane reforming, dry reforming of methane, and other reforming processes for hydrogen production. The current progress and knowledge gaps in the research and development of hydrogen production technologies from a computational point of view are also discussed.