Doped: A Python Package for Solid-State Defect and Dopant Calculations

When and Where

Apr 25, 2024
11:15am - 11:30am

Room 335, Level 3, Summit

Presenter

Seán Kavanagh

David Scanlon

Aron Walsh

Co-Author(s)

Seán Kavanagh¹,David Scanlon²,Aron Walsh³

Harvard University¹,University of Birmingham²,Imperial College London³

Abstract

Seán Kavanagh¹,David Scanlon²,Aron Walsh³

Harvard University¹,University of Birmingham²,Imperial College London³

Defect-induced non-radiative recombination typically represents the dominant limiting factor in the efficiencies of emerging inorganic solar cells / photocatalysts.1,2 Computational methods are widely used to predict defect behavior in solar materials, before combining and comparing theoretical predictions with experimental measurements. However, there are many critical stages in the computational workflow for defects, which, when performed manually, not only leave room for human error but also consume significant researcher time and effort. Moreover, there are growing efforts to perform high-throughput defect investigations,3–5 necessitating robust, user-friendly and efficient software implementing this calculation workflow. Here we report doped, our python package for the full generation, calculation setup, post-processing and analysis of defect supercell calculations.2,6–9 The generation and thermodynamic analysis (i.e. defect formation energy diagrams, chemical potentials, doping analysis etc.) are agnostic to the underlying first-principles software, while input file generation is supported for several of the most widely-used DFT codes, including VASP, FHI-aims, CP2k, Quantum Espresso and CASTEP. A defect charge state prediction algorithm is implemented, which is shown to significantly outperform previous oxidation-state approaches in terms of both efficiency and completeness. Moreover, doped is built to be compatible with other computational toolkits for advanced defect characterisation, including ShakeNBreak10 for defect structure-searching, py-sc-fermi11 for in-depth concentration, doping and Fermi level analysis, and CarrierCapture.jl12/nonrad13 for non-radiative recombination calculations. Its object-oriented python framework make it readily-usable in high-throughput architectures such as atomate(2) or AiiDA, with examples included in the documentation. We will discuss the key features of doped for computational defect workflows, exemplified with relevant solar cell materials (CdTe, Sb2Se3, t-Se). We anticipate that doped will serve as a highly useful tool for computational defect researchers, being an efficient platform for conducting reproducible calculations of solid-state defect properties. 1 Y.-T. Huang et al, Nanotechnology, 2021, 32, 132004. 2 S. R. Kavanagh et al, ACS Energy Lett., 2021, 6, 1392–1398. 3 D. Dahliah et al, Energy Environ. Sci., 2021, 14, 5057–5073. 4 pymatgen-analysis-defects, https://materialsproject.github.io/pymatgen-analysis-defects/intro.html 5 D. Broberg et al, npj Comput Mater, 2023, 9, 1–12. 6 Y.-T. Huang & S. R. Kavanagh et al Nat Commun, 2022, 13, 4960. 7 S. R. Kavanagh et al Faraday Discuss., 2022, 239, 339–356. 8 A. Nicolson et al, Journal of Materials Chemistry A 2023. 9 S. R. Kavanagh, doped v.2.0.5 Zenodo 2023. 10 I. Mosquera-Lois & S. R. Kavanagh et al, Journal of Open Source Software, 2022, 7, 4817. 11 A. G. Squires et al, Journal of Open Source Software, 2023, 8, 4962. 12 S. Kim et al, Journal of Open Source Software, 2020, 5, 2102. 13 M. E. Turiansky et al, Computer Physics Communications, 2021, 267, 108056.