Two-way liquid organic hydrogen carriers (LOHCs) – organic molecules that can store hydrogen as reversible hydrogen-containing chemical bonds – offer a transformative solution to the problem of energy storage and transportation, thereby addressing critical challenges in developing new energy and hydrogen sources. In this technology, a hydrogen-lean molecule (such as toluene) is catalytically hydrogenated to a hydrogen-rich molecule (the hydrogen carrier) which can then be stored and transported to the destination inexpensively. At the demand site, the molecules are dehydrogenated on-demand to produce hydrogen/energy and the “spent” molecule is recycled back to the hydrogen source for subsequent hydrogenation. Key to the large-scale deployment of this technology is the identification of the optimal LOHCs that satisfy a slew of physical, thermodynamic, synthetic, and kinetic constraints. Designing an LOHC bearing specific structure and function optimized for H2 delivery is intrinsically a multiscale problem of efficiently exploring the enormous organic molecular space, reliably assessing if the molecule can be synthesized, linking atomic scale LOHC information with process-level performance metrics and global-level supply chain considerations. In this context, a specific focus of our research is in the computational discovery of these molecules that is cognizant of chemistry-specific information such as synthetic feasibility and reaction kinetics.
This talk will describe our initial forays towards this end. First, I will discuss a new molecule discovery workflow based on cheminformatics and data-driven property modeling that we have developed to screen millions of organic molecules to identify promising LOHCs. In particular, we employ an automated network generation tool, called RING, to identify the specific pairs of molecules (i.e. hydrogen-rich and lean forms of the LOHCs) and the dehydrogenation pathways connecting them; subsequently, a set of machine learned models are leveraged to quickly and reliably predict physical and thermodynamic properties and the synthetic feasibility of these molecules. This process results in hundreds of potentially new synthetically accessible molecule pairs that have high hydrogen storage capacities, low melting and high boiling points, and reasonable heats of hydrogenation.
One property that this workflow misses is the kinetics, i.e. the rate and conditions at which these hydrogenation and dehydrogenation steps can be carried out selectively (i.e. with minimal degradation of molecules). Rapid methods to estimate kinetic properties are, however, currently unavailable to include in our discovery workflow. In the second part of this talk, I will discuss the complex reaction landscape of dehydrogenation of tetrahydropyrrole, which is one of the simplest hydrogen carriers. Using a combination of density functional theory (DFT) and microkinetic modeling, we develop a mechanistic model of tetrahydropyrrole dehydrogenation on platinum to understand the flux-carrying pathways, rate-determining steps, the most abundant reaction intermediate, the effect of molecular substitution, and directions for designing active and selective catalysts. Since such a methodology is challenging to scale up to screen millions of molecules in a high-throughput fashion, we finally propose a new mathematical framework based on informatics, optimization, and machine learning to rapidly estimate kinetics of potential hydrogen carriers.