Abstract Copper (Cu) undergoes significant surface reconstruction during CO₂ electroreduction, which is strongly modulated and accelerated by reaction intermediates, yet the atomic-scale mechanism remains far behind the experimental observations. By integrating machine learning interatomic potentials with large-scale grand canonical Monte Carlo simulations, we systematically investigated *CO- and *H-induced surface roughening across various Cu facets. Our simulations demonstrate that high surface *H (*Hsur) coverage facilitates subsurface hydrogen (*Hsub) incorporation on (100)-dominated Cu facets under typical working conditions (−1 V vs. reversible hydrogen electrode), while such penetration is negligible on (111)-like surfaces. This facet dependence is primarily attributed to a *H-induced hexagonal surface reconstruction observed on (100)-dominated facets, a process driven primarily by geometric rather than electronic effects. Specifically, high *H coverage triggers a partial transition of Cu atoms from ideal 4-fold hollow sites to more closely packed 3-fold arrangements. The local densification expands the spacing at the remaining 4-fold sites, thereby reducing the energy barrier for *Hsub migration into the subsurface. Further analysis reveals that *Hsub alone is sufficient to induce Cu adatom formation, even in the absence of nearby *CO, uncovering a revised structural evolution paradigm for Cu surface roughening. We propose an alloying strategy using low hydrogen affinity metals (Zn, Al, Ga) to effectively suppress *Hsur incorporation, offering a promising pathway for designing Cu-based catalysts with long-term stability.
Peng et al. (Thu,) studied this question.