Adsorption energies are necessary but not sufficient to identify good catalysts

TL;DR

The paper interrogates the practicality of using thermodynamic overpotential $\eta$, derived from adsorption free energies $\Delta G_{ads}$, as a primary high-throughput metric for catalyst discovery. By leveraging large OC20/OC22 datasets for HER and OER, it systematically quantifies uncertainties in $\Delta G_{ads}$ and propagates them to $\eta$, revealing substantial overlap among surfaces and widespread false positives. The analysis shows that a large fraction of random materials appear to be ideal catalysts within the estimated uncertainty, challenging the selectivity of overpotential as a screening criterion and highlighting the potential value of Pourbaix stability as a more selective filter. The authors advocate shifting objective functions toward stability, lifetime, and affordability, while acknowledging that OC datasets remain invaluable for training ML force fields and enabling broader catalyst discovery efforts.

Abstract

As a core technology for green chemical synthesis and electrochemical energy storage, electrocatalysis is central to decarbonization strategies aimed at combating climate change. In this context, computational and machine learning driven catalyst discovery has emerged as a major research focus. These approaches frequently use the thermodynamic overpotential, calculated from adsorption free energies of reaction intermediates, as a key parameter in their analysis. In this paper, we explore the large-scale applicability of such overpotential estimates for identifying good catalyst candidates by using datasets from the Open Catalyst Project (OC20 and OC22). We start by quantifying the uncertainty in predicting adsorption energies using \textit{ab initio} methods and find that $\sim$0.3-0.5 eV is a conservative estimate for a single adsorption energy prediction. We then compute the overpotential of all materials in the OC20 and OC22 datasets for the hydrogen and oxygen evolution reactions. We find that while the overpotential allows the identification of known good catalysts such as platinum and iridium oxides, the uncertainty is large enough to misclassify a broad fraction of the datasets as ``good'', which limits its value as a screening criterion. These results question the reliance on overpotential estimation as a primary evaluation metric to sort through catalyst candidates and calls for a shift in focus in the computational catalysis and machine learning communities towards other metrics such as synthesizability, stability, lifetime or affordability.

Figures (4)

• Figure 1: Overpotential as a function of calculated hydrogen adsorption free energy (${\Delta G}_{*H}$) in the OC20 dataset. Experimental data is from danilovic_enhancing_2012
• Figure 2: Overpotential as a function of the typical rate limiting step (${\Delta G}_{*O}$ - ${\Delta G}_{*OH}$) in the OC22 dataset. The dotted line is the distribution of calculated $\eta$'s for the best known catalyst under the assumption that adsorption free energy calculations are normally distributed with a standard deviation of 0.5 eV.
• Figure 3: Decomposition energy for materials in OC22 that have at least one surface with all three OER adsorption energies ($\Delta G_{*OH}$, $\Delta G_{*O}$, $\Delta G_{*OOH}$) and are part of the Materials Project database
• Figure S1: Side view of the IrO$_2$$\{$110$\}$ surface showing the surface terminations, the axis perpendicular to the surface points up. Blue: Ir, red: O; the starred O are the adsorbed O.