Nanostructured techniques are intrinsically metastable and topic to coarsening. For supported 3D metallic nanoclusters (NCs), coarsening can contain NC diffusion throughout the help and subsequent coalescence (as an alternative choice to Ostwald ripening). When used as catalysts, this results in deactivation. The dependence of diffusivity, D_N, on NC dimension, N (in atoms), controls coarsening kinetics. Conventional mean-field (MF) idea for D_Nversus N assumes that NC diffusion is mediated by unbiased random hopping of floor adatoms with low coordination, and predicts that D_N ∼ hN^−4/3n_eq. Right here, h = ν exp[−E_d/(k_BT)] denotes the hop fee, and n_eq = exp[−E_form/(k_BT)] the density of these adatoms. The adatom formation power, E_kind, approaches a finite large-N restrict, as does the efficient barrier, E_eff = E_d + E_kind, for NC diffusion. This MF idea is critically assessed for a sensible stochastic atomistic mannequin for diffusion of faceted fcc metallic NCs with a {100} aspect epitaxially connected to a (100) help floor. First, the MF formulation is refined to account for distinct densities and hop charges for floor adatoms on totally different aspects and alongside the bottom contact line, and to include the precise values of E_kind and n_eqversus N for our mannequin. MF idea then captures the prevalence of native minima in D_Nversus N at closed-shell sizes, as proven by KMC simulation. Nonetheless, the MF therapy additionally shows basic shortcomings because of the characteristic that diffusion of faceted NCs is definitely dominated by a cooperative multi-step course of involving disassembling and reforming of outer layers on facet aspects. This mechanism results in an E_eff which is effectively above MF values, and which will increase with N, options captured by a beyond-MF therapy.