Parts | Days | Selection | Search | Updates | Downloads | Help
HL: Fachverband Halbleiterphysik
HL 7: Poster I
HL 7.38: Poster
Monday, March 27, 2023, 13:00–15:00, P2/EG
Defect tolerance of halide perovskites solar absorbers via machine learning — •Anoop K. Chandran1, Christoph Friedrich1, Uwe Rau2, Stefan Blügel1, Thomas Kirchartz2, and Irene Aguilera3 — 1Peter Grünberg Institute and Institute for Advanced Simulation, Forschungszentrum Jülich, Germany — 2IEK5-Photovoltaik, Forschungszentrum Jülich, Germany — 3Institute of Physics, University of Amsterdam, The Netherlands
The deformation potential measures the changes in the bandgap of materials upon compression of the bonds, which helps to identify the bonding or anti-bonding nature of the valence and conduction bands. Previous works indicate that the bonding or anti-bonding nature impacts a material's tendency to exhibit shallow or deep intrinsic defect levels. Our high-throughput search is based on all-electron density functional theory calculations. Among 1173 studied halide perovskites, only 18% present a favourable anti-bonding valence band. We also establish the calculation of the deformation potential as an effective way to determine the bonding or anti-bonding nature near the band edges. However, subsequent supercell calculations reveal no explicit connection between the nature of the band edges and the defect tolerance. Deep learning neural networks require large datasets. To overcome this limitation, we devised a novel approach for the one-shot binary classification of the deformation potentials. Instead of learning to classify a single material, the network learns the similarity or the difference between two materials.