Chemical Pressure-Driven Enhancement of the Hydrogen Evolving Activity of Ni₂P from Nonmetal Surface Doping Interpreted via Machine Learning

Abstract

The activity of NiP catalysts for the hydrogen evolution reaction (HER) is currently limited by strong H adsorption at the Ni-hollow site. We investigate the effect of surface nonmetal doping on the HER activity of the NiP termination of NiP(0001), which is stable at modest electrochemical conditions. Using density functional theory (DFT) calculations, we find that both 2 p nonmetals and heavier chalcogens provide nearly thermoneutral H adsorption at moderate surface doping concentrations. We also find, however, that only chalcogen substitution for surface P is exergonic. For intermediate surface concentrations of S, the free energy of H adsorption at the Ni-hollow site is -0.11 eV, which is significantly more thermoneutral than the undoped surface (-0.45 eV). We use the regularized random forest machine learning algorithm to discover the relative importance of structure and charge descriptors, extracted from the DFT calculations, in determining the HER activity of NiP(0001) under different doping concentrations. We discover that the Ni-Ni bond length is the most important descriptor of HER activity, which suggests that the nonmetal dopants induce a chemical pressure-like effect on the Ni-hollow site, changing its reactivity through compression and expansion.

A Critical Review of Machine Learning of Energy Materials

“…In contrast, the data‐driven approach starts from an initial large set of candidate features and down‐selects a subset of features. This down‐selection process can be automatic, e.g., using L 1 or L 0 regularization (least absolute shrinkage and selection operator, LASSO), feature importance, genetic algorithms, etc. However, a drawback of data‐driven feature selection is that the selected features do not imply causality with respect to the target and will be highly dependent on the chosen hyperparameters of the model .…”

“…For example, the coefficients in simple linear regression model as well as the logistic regression can provide an indication of the relative importance of different features. Feature importance analysis in tree‐based models has been routinely used to assess the important physical parameters that are related to the predictive targets . For deep learning‐based models, visualizing the hidden layer activations has been found to give interpretable chemical intuition and accurate mapping of structural space .…”

“…ML models can also provide useful catalyst design insights. Wexler et al have predicted the hydrogen evolution reaction (HER) activity on nonmetal doped Ni 3 P 2 termination of Ni 2 P(0001) surfaces. The H binding energy on the pristine surfaces was too strong for catalysis and the authors found that by doping the surface with nonmetal elements (As, B, C, N, O, S, Se, Si, and Te) the binding energy of H can be tuned.…”

“…d) The correlation between average Ni–Ni bond distance and the H binding free energy induced by chemical and mechanical pressure. Reproduced with permission . Copyright 2018 American Chemical Society.…”

Machine Learning for Computational Heterogeneous Catalysis

“…By relating these descriptors to experimental oxygen evolution activities of transition‐metal oxide catalysts, the authors unveiled the importance of the transition metal e g / t 2 g electron occupancy in determining the catalytic activity. Wexler et al . used a regularized random forest machine‐learning algorithm to investigate the relative importance that structure and charge descriptors have on the HER activity of Ni 2 P(0001).…”