Assessment of the Accuracy of Density Functionals for Calculating Oxygen Reduction Reaction on Nitrogen-Doped Graphene

Experimental studies of the oxygen reduction reaction (ORR) at nitrogen doped graphene electrodes have reported a remarkably low overpotential, on the order of 0.5 V, similar to Pt based electrodes. Theoretical calculations using density functional theory have lent support for this claim. However, other measurements have indicated that transition metal impurities are actually responsible for the ORR activity, thereby raising questions about the reliability of both the experiments and the calculations. In order to assess the accuracy of the theoretical calculations, various generalized gradient approximation (GGA), meta-GGA and hybrid functionals are employed here and calibrated against high-level wave function based coupled cluster calculations (CCSD(T)) of the overpotential as well as self-interaction corrected density functional calculations and published quantum Monte Carlo calculations of O adatom binding to graphene. The PBE0 and HSE06 hybrid functionals are found to give more accurate results than the GGA and meta-GGA functionals, as would be expected, and for low dopant concentration, 3.1%, the overpotential is calculated to be 1.0 V. The GGA and meta-GGA functionals give a lower estimate by as much as 0.4 V. When the dopant concentration is doubled, the overpotential calculated with hybrid functionals drops, while it increases in GGA functional calculations. The opposite trends result from different potential determining steps, the *OOH species being of central importance in the hybrid functional calculations while the reduction of *O determines the overpotential obtained in GGA and meta-GGA calculations. The results presented here are mainly based on calculations of periodic representations of the system, but a comparison is also made with molecular flake models which are found to give erratic results.