A datamining approach to predict the formation enthalpy for rare-earth dihydrides REH2 (RE = Ce,Pr,Dy)

• High correlations were found between a, R, χ and ΔH from PCA analysis.
• The formation enthalpy of CeH2, PrH2, DyH2 was estimated by ANN model.
• The formation enthalpy of CeH2, PrH2, DyH2 was calculated by DFT+U.
• The results are in excellent agreement between ANN, DFT+U and experimental/theoretical data reported in literature.

Different theoretical methods estimating the formation enthalpy for hydrides have been adopted, among them the density functional theory (DFT) calculations. In this paper, an attempt has been made to study and estimate the formation enthalpy for CeH2, PrH2 and DyH2 hydrides, by mean of a datamining approach. Based on the data-sets selected from different published work, a principal component analysis (PCA) is applied to select the significant modeling inputs and simplify the model structure. An artificial neural network (ANN) with 3-13-1 architecture structure has been developed to estimate ΔH. The structural parameter and elastic properties of Ce, Pr, Dy, CeH2, PrH2 and DyH2 were calculated by the density functional theory (DFT+U). Also, the formation enthalpy of CeH2, PrH2 and DyH2 are determined. The results of ΔH calculated with both ANN and DFT+U are basically consistent and are in good agreement with the experimental data. The adopted ANN structure provided an advantageous technique in predicting the non-linear relationship between the formation enthalpy and its influences factors (a, R, χ) which were previously selected by PCA.

Figure optionsDownload as PowerPoint slide