Two-dimensional Janus material MoS2(1-x)Se2x (0 < x < 1) for photovoltaic applications: A machine learning and density functional study

Janus transition-metal dichalcogenides MoSSe has been attracted much attention due to its excellent electronic properties induced by mirror symmetry breaking. In this work, based on machine learning and density functional theory, the photoelectric conversion coefficient (PCE) along with the variation of Se concentration in MoS2(1-x)Se2x (0 < x < 1) are explored. Ten most important features are sorted out by calculating the Pearson correlation coefficient matrices to identify the linear relationship between any two features and their correlation. The coefficient of determination (R-2), root mean square error (RMSE) and mean absolute relative error (MARE) are evaluated for the built machine learning models of random forest (RF) and multiple linear regression (MLR). For the prediction of the PCE, the RF algorithm using structural information and computational features obtained from DFT calculations is validated possessing high efficiency. In a wide range of doping concentrations (n = 32%-96%), we predict that the PCE of MoS2(1-x)Se2x has a high value larger than 18%. High optical absorption intensity in an order of 10(5) cm(-1) is obtained in MoS0.89Se1.11, which has potential application in solar cell.

Automated summary

This study explores the impact of Se concentration variation on the photoelectric conversion coefficient in MoS2(1-x)Se2x using machine learning and density functional theory. By calculating correlation coefficient matrices, ten key features are identified and random forest and multiple linear regression models are built. The research suggests that the PCE of MoS2(1-x)Se2x may exceed 18% in a wide range of doping concentrations. High optical absorption intensity is obtained in MoS0.89Se1.11, showing potential application in solar cells.