|
Abstract
: Point defects and impurities critically govern the performance, stability, and dopability of II–VI semiconductors like CdTe, CdSe, ZnTe, and their alloys.
However, accurate defect modeling remains computationally expensive and methodologically fragmented. In this webinar, I will present DefectDB, a unified informatics platform for data-driven defect modeling in Cd/Zn–S/Se/Te semiconductors that integrates high-throughput
first-principles calculations, machine-learning force fields (MLFF), and interactive analysis tools. The platform encompasses a curated database of DFT calculations (PBEsol and HSE06+SOC) for bulk and defect structures across multiple charge states, covering
native defects, substitutional dopants, and defect complexes in binary, ternary, and quaternary compositions. To overcome the computational bottleneck of conventional supercell calculations, DefectDB incorporates crystal graph neural network-based MLFF for
rapid configuration sampling, geometry optimization, and finite-temperature molecular dynamics with orders-of-magnitude speedup. I will demonstrate the platform's five integrated modules spanning defect formation energy analysis, MLFF-accelerated defect simulation,
diffusion barrier calculation via nudged elastic band methods, grain boundary and dislocation core modeling, and heterointerface construction. Together, these capabilities enable systematic defect discovery, thermodynamic screening, and transport property
prediction, establishing a scalable and reproducible framework for defect physics in technologically important photovoltaic semiconductors.
Habibur Rahman is a PhD candidate in Materials Engineering at Purdue University, working under the supervision of Prof. Arun Kumar Mannodi-Kanakkithodi. His research focuses on computational defect physics in thin-film photovoltaic materials, combining
density functional theory with machine learning approaches to accelerate defect modeling in II–VI semiconductors and related chalcogenide systems.
|
