| 초록 |
Interaction of an alloy catalyst with reactants, closely related with the reactivity, can be changed by the surface metal composition. Thus, information on the segregation of impurity metals towards the surface can be critical to design highly reactive alloy catalysts. Using density functional theory (DFT)-calculated segregation energies (Esegr) of transition metal alloys from a previous study, neural network models correlating numerical fingerprint of a segregation site with the Esegr was constructed with cross-validation mean absolute error (MAE) of 0.06 eV. Even with only tabulated features, the models reasonably predicted the Esegr well. Insights extracted by feature importance analysis also will be presented. |
