2019/06/05 by Arun Mannodi‐Kanakkithodi, Mannodi-Kanakkithodi, Arun, Michael Y. Toriyama +9 · 1 citation
Engineering · Materials Science · #Advanced Semiconductor Detectors and Materials #Chalcogenide Semiconductor Thin Films #FOS: Physical sciences #Machine Learning in Materials Science #Materials Science (cond-mat.mtrl-sci)
paper · pdf · doi:10.48550/arxiv.1906.02244
openalex publication_date 2019/06/05 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
The ability to predict the likelihood of impurity incorporation and their\nelectronic energy levels in semiconductors is crucial for controlling its\nconductivity, and thus the semiconductor's performance in solar cells,\nphotodiodes, and optoelectronics. The difficulty and expense of experimental\nand computational determination of impurity levels makes a data-driven machine\nlearning approach appropriate. In this work, we show that a density functional\ntheory-generated dataset of impurities in Cd-based chalcogenides CdTe, CdSe,\nand CdS can lead to accurate and generalizable predictive models of defect\nproperties. By converting any semiconductor + impurity system into a set of\nnumerical descriptors, regression models are developed for the impurity\nformation enthalpy and charge transition levels. These regression models can\nsubsequently predict impurity properties in mixed anion CdX compounds (where X\nis a combination of Te, Se and S) fairly accurately, proving that although\ntrained only on the end points, they are applicable to intermediate\ncompositions. We make machine-learned predictions of the Fermi-level dependent\nformation energies of hundreds of possible impurities in 5 chalcogenide\ncompounds, and suggest a list of impurities which can shift the equilibrium\nFermi level in the semiconductor as determined by the dominant intrinsic\ndefects. These dominating impurities as predicted by machine learning compare\nwell with DFT predictions, revealing the power of machine-learned models in the\nquick screening of impurities likely to affect the optoelectronic behavior of\nsemiconductors.\n