ArXiv · 2026
We compare the migration of a single Cr atom on monolayer graphene, between AB-stacked layers and between AA-stacked layers, with one first-principles protocol (PBE+D3(BJ), one cell family, registry anchors and spin-state search). We find three main results. First, on the surface and in the AB gallery Cr stays magnetic and moves almost freely. On the surface the C-C bridge is the minimum in 4x4 and 5x5 cells, and the highest point we sample on a hop lies 10 meV above it (5x5). In the AB gallery Cr sits on the inversion centres of the bilayer, which form a kagome net, and hops between them over 15 meV with a moment of about 4.32 μ_B. Second, in the AA gallery the minimum is a nonmagnetic η⁶-η⁶ sandwich that pulls the layers together. Leaving it requires opening the gallery and switching Cr to a moment of 4-5 μ_B, so we map the path with separately relaxed spin branches. The barrier estimate is 636 meV in a 3x3 cell but 260 and 185 meV in 5x5 cells with the far-field spacing pinned and free, and it is not yet converged. At 300 K, with equal prefactors and if the spin change is not rate-limiting, Cr hops between AA layers 3-4 orders of magnitude more slowly than between AB layers, using the 5x5 AA estimates and the 3x3 AB band. These AA numbers are specific to PBE+D3: an on-site U of 1.0 eV on Cr d already makes the magnetic states the AA ground state. Third, Cr binds 246-666 meV more strongly in AA than in AB registry, so the fast AB hop requires that the layers keep their Bernal registry around Cr.
Try inveni