Self-interaction error (SIE) is a well-known limitation of standard density functional theory (DFT), yet its significance for single-atom catalysts (SACs) remains largely overlooked. Here we show that correcting SIE via $\mathrm{DFT}+U$ is not merely a methodological refinement but a prerequisite for reliably predicting the electronic structure and catalytic activity of $3d$ metal-${\mathrm{N}}_4$ SACs, especially for near-half-filled $d$-shell systems (Mn, Fe, Co, Ni). These SACs exhibit pronounced $U$-sensitivity, manifesting as $d$-level shifting and spin-state transitions driven by the interplay between on-site Coulomb repulsion and the quasi-$D_{4h}$ ligand field. Crucially, this sensitivity translates directly to catalysis. Using the oxygen reduction reaction as a model reaction, we find that $\mathrm{DFT}+U$, in agreement with experiments, successfully predicts $\mathrm{Fe}\text{−}{\mathrm{N}}_4$ to be the more active catalyst than $\mathrm{Co}\text{−}{\mathrm{N}}_4$, whereas standard DFT fails. This reversal originates from a downshift of occupied $d$ orbitals that weakens hybridization with ${}^{*}\mathrm{OOH}$ and ${}^{*}\mathrm{OH}$. Interestingly, for ${\mathrm{Fe}}^{*}\mathrm{O}, {\mathrm{Co}}^{*}\mathrm{O}$, and ${\mathrm{Mn}}^{*}\mathrm{OH}$, a spin-state transition partially offsets this weakening via Hund's exchange, leading to a less pronounced decrease in adsorption energy. These findings establish that SIE correction is essential for reliable SAC modeling, particularly for near-half-filled $d$-shell systems, and our work provides a mechanistic framework to understand its impact on catalytic predictions.

Close

PrevNext

View figure in article

• Local density of states
• Spin state transition
• Graphene
• DFT+U
• Hybrid functionals