Forecasted trends in fossil fuel consumption are predicted to exacerbate global CO2 emissions in the coming decades. Conversion of atmospheric CO2 to value-added fuels and chemicals via electrochemical CO2 reduction (CO2R) offers one solution to create a circular carbon economy. While traditional metallic catalysts have been limited in their reactivity and selectivity to produce commercialized rates of hydrocarbons, graphene-based single-atom catalysts (SACs) have emerged as candidates due to their tunable electronic structure and high atomic efficiency. This new class of catalysts provides the blueprint for a large chemical space, of which the catalytic properties can be tuned with dopants and metal centers. Despite this potential, the relationship between SAC compositions, structures, and catalytic performance remains unclear. To understand such relationships, density functional theory (DFT) is widely used, yet it suffers from electron self-interaction error, lack of a derivative discontinuity, and a single-reference description of the electronic structure. These limitations are problematic in describing transition metal chemistry and reaction kinetics. Alternatively, multireference wavefunction methods offer a more reliable description of the electronic structure and charge transfer processes. Here, we first performed a systematic study to understand the role of metal centers and dopants in tuning reactivity of SACs for electrochemical CO2R. We then benchmarked a finite molecular flake model in simulating graphene-based SACs. By employing this model, we performed multireference wavefunction methods, i.e., complete active space self-consistent field (CASSCF) method coupled with complete active space second-order perturbation theory (CASPT2), in predicting key catalytic properties of SACs. This work advances fundamental understanding of electronic and catalytic properties of SACs at a higher level of accuracy than ever before, which is crucial and foundational in catalyst design. The developed methodology in this work can be extended to study other heterogeneous catalysis which are problematic to model with DFT.
