Revisiting the universal principle for the rational design of single-atom electrocatalysts

The notion of descriptors has been widely used for assessing structure–activity relationships for many types of heterogenous catalytic reaction, as well as in searching for highly active single-atom catalysts (SACs). Here, with the aid of a machine-learning model for identifying key intrinsic properties of SACs, we revisit our previous descriptor φ [ Nat. Catal.1, 339–348 (2018) ] and present φ ′ to correlate the activity of graphene-based SACs for the oxygen reduction reaction, oxygen evolution reaction and hydrogen evolution reaction. The descriptor φ ′ not only captures the activity trend among experimentally reported SACs, but can also help with the search for SACs to replace precious-metal-based commercial catalysts (for example Pt/C and IrO 2 ), including Fe-pyridine/pyrrole-4N for the oxygen reduction reaction and Co-pyridine/pyrrole-4N for the oxygen evolution reaction (discovered in previous experimental studies). More importantly, we show that the descriptor φ ′ can be broadly applicable to correlate SACs embedded in small-, mid- and large-sized macrocyclic complexes, so long as the active metal centre has the same local coordination environment.