| 초록 |
Interaction of an alloy metal surface with reactants depends on the surface metal composition. Thus, the thermodynamics on the segregation of impurity metals toward the host metals’ surface can be critical to design highly reactive alloy surface. Using uniquely-collected 1,366 density functional theory (DFT)-calculated segregation energies (Esegr) from previous studies, we constructed a deep neural network (segDNN) model correlating the Esegr with a numerical fingerprint composed of 19 features for facet-, site-, and elemental-dependencies, which are easily accessible from literature. Reasonable interpolative and extrapolative prediction by the segDNN was clearly demonstrated using principal component analysis (PCA). Feature importance analysis based on normalized sensitivity coefficient (NSC) and example applications of the segDNN for alloy surface design also will be presented. |
