研究目的
Investigating the prediction of point defect properties in common semiconductors to identify potential deep lying impurity states.
研究成果
The DFT+ML approach can lead to a significant acceleration in the estimation of stable deep lying impurity in semiconductors, and transform the design of novel materials for improved photovoltaics.
研究不足
The PBE level of theory is used despite the band gap underestimation, but trends in defect energetics and energy levels can transform to knowledge that is useful for a first stage of defect screening.
1：Experimental Design and Method Selection:
Density functional theory (DFT) was used to compute defect formation energies and charge transition levels of hundreds of impurities in CdX chalcogenide compounds. Machine learning techniques were applied on the DFT data to develop predictive models.
2：Sample Selection and Data Sources:
Impurity atoms from 63 elements were simulated at the Cd, X (Te/Se/S) or interstitial sites in a 64 atom 2x2x2 CdX supercell in different charged states.
3：List of Experimental Equipment and Materials:
Vienna ab-initio Simulation Package (VASP) was used with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional and projector-augmented wave (PAW) pseudopotentials.
4：Experimental Procedures and Operational Workflow:
An energy cut-off of 400 eV was applied and all atoms were relaxed until forces on each were less than 0.05 eV/?. A 3x3x3 Monkhorst-Pack meshes were used for Brillouin zone integration.
5：05 eV/?. A 3x3x3 Monkhorst-Pack meshes were used for Brillouin zone integration.
Data Analysis Methods:
5. Data Analysis Methods: Random forest regression was used to map the descriptors to the formation energies and transition levels using 5-fold cross-validation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
