Predicting protic redox potentials with computational error bars

Design of redox-active species for applications such as sensing and energy storage requires reliable prediction of redox potentials, frequently in protic environments. However, this is complicated by the fact that multiple reaction pathways are possible, which cannot be easily elucidated experimentally. Computational assignment of redox reaction mechanisms by matching computed and measured redox potentials is an enticing possibility, but demands careful treatment of computational uncertainties because these may be of the same order of magnitude as mechanism-dependent differences. Here, we introduce a new scheme for estimating computational “error bars” by identifying density functionals whose predicted redox potentials systematically bound experimental data. We demonstrate that the bounding behaviour of hybrid functionals may be tuned by altering the proportion of Hartree–Fock exchange, while generalised gradient functionals are more sensitive to the nature and form of the specific gradient corrections applied to correct for the local density approximation. • B3LYP and TPSS functionals accurately predict protic redox potentials. • Related functionals systematically under/over-predict redox potentials. • Bounding behaviour of hybrid functionals tuned by altering % HF exchange. • Bounding behaviour of GGAs and mGGAs determined by form of GGA/mGGA corrections. • Recommended sequences of functionals: BLYP-B3LYP-BHHLYP, PZKB-TPSS-PBE0.