RESEArCH

Heterogeneous catalysis, vital for industrial processes, drives over 20% of chemical conversions. Despite its prevalence, understanding surface phenomena and atomic-level reactions remains challenging. The application of Density Functional Theory (DFT) calculations, especially in the context of reaction and activation energies, offers a window into the intricate workings of catalyst activity and stability. The imperative to develop computational tools capable of deciphering complex reactions involved in biomass chemistry is more pressing than ever. One such reaction is reductive amination of 5-HMF, where we tried to understand the detailed mechanistic insights on a metal catalyst, turnovers and rate-determining step using ab initio microkinetic model (MKM). Our findings led us to propose a bimetallic catalyst that could mitigate deactivation from strong nitrogen bindings.

Further, difference in morphology of nanoparticles can lead to different exposed surface facets and ultimately a change in catalytical activity. To understand this, structure-dependent activity and selectivity for furfural acetalization reaction in presence of alcohols (methanol, ethanol, propanol and butanol) as solvents was studied over well-defined supported Pd nanostructures. From the experimental findings, it was discerned that furfural conversion on Pd nanostructures followed the trend: cube > spheres > octahedra. The effect of different alcohols was also tested, and the reactivity trends followed the order: methanol > ethanol > propanol > butanol for furfural dialkyl acetal formation. DFT calculations elucidated the role of hydrogen bonding network between the solvent molecules and adsorbate in proton transfer which resulted in the reduction of the activation barriers and the stabilization of the transition-state structures.

Moreover, screening of active and selective catalyst from DFT simulations is difficult, due to a wide choice of catalytic materials and huge computational cost associated with them. Therefore, we used DFT energetics and readily available periodic properties of elements to train and build ML models to predict the binding energies for major reaction descriptors such as CO and OH. We found xGBR to be the best model in our case since it gave the lowest RMSEs. These predicted binding energies were later utilized by ab intio MKM to predict the turnovers for CO production in reverse water gas shift reaction to screen cheaper Cu-based bimetallics.

Overall, from this study an effective design of metal and bimetallic catalyst can be envisaged for guiding different types of reactions which forms an effective strategy in catalytic transformation of bio-renewable substrates.