doped: a python package for solid-state defect and dopant calculations

Slides for my talk on our comprehensive defect modelling package: doped, at MRS Spring 2024, Seattle.
YouTube video recording: https://youtu.be/-Z-R9sedeqY
Also see the Twitter thread highlighting the key features of doped here: https://x.com/Kavanagh_Sean_/status/1780667455297458185

References:
doped docs: https://doped.readthedocs.io
ShakeNBreak docs: https://shakenbreak.readthedocs.io
doped paper: https://joss.theoj.org/papers/10.21105/joss.06433
ShakeNBreak theory paper: https://www.nature.com/articles/s41524-023-00973-1
ShakeNBreak code paper: https://joss.theoj.org/papers/10.21105/joss.04817

Questions welcome! For other computational photovoltaics, defects and disorder talks, have a look at my YouTube channel!
https://www.youtube.com/SeanRKavanagh
For other research articles see:
https://bit.ly/3pBMxOG

Abstract:
Defects are a universal feature of crystalline solids, dictating the key properties and performance
of many functional materials. Given their crucial importance yet inherent difficulty in measuring
experimentally, computational methods (such as DFT and ML/classical force-fields) are widely
used to predict defect behaviour at the atomic level and the resultant impact on macroscopic
properties. Here we report doped, a Python package for the generation, pre-/post-processing,
and analysis of defect supercell calculations. doped has been built to implement the defect
simulation workflow in an efficient and user-friendly – yet powerful and fully-flexible – manner,
with the goal of providing a robust general-purpose platform for conducting reproducible
calculations of solid-state defect properties.

Transcript

1. 2 Seán R. Kavanagh doped: Python toolkit for robust and repeatable charged defect supercell calculations
2. doped.readthedocs.io Scan for docs/paper: (Also shown later) Kavanagh* et al. Journal of Open Source Software 2024
3. Defect Simulation Workflow 5
4. Design Philosophy • Wide-spanning functionality • Reasonable defaults, but with full control/flexibility for the user • User-friendly, extensive tutorials/documentation, automated compatibility/sanity checks • Aid and encourage reproducibility • Computational efficiency through intelligent algorithms • Publication-ready outputs 6
5. Defect Simulation Workflow 7
6. Defect Simulation Workflow 8
7. 9 • Candidate interstitial sites generated with Voronoi tessellation (most reliable approach)1 – or user-specified. • Auto-generated optimal supercell, or user-provided. • Generates all intrinsic (and extrinsic if specified) point defects. • Fully flexible 1. Kononov et al 2023 J. Phys.: Condens. Matter 35 334002
8. 10 1. Kononov et al 2023 J. Phys.: Condens. Matter 35 334002 • Candidate interstitial sites generated with Voronoi tessellation (most reliable approach)1 – or user-specified. • Auto-generated optimal supercell, or user-provided. • Generates all intrinsic (and extrinsic if specified) point defects. • Fully flexible
9. 11 doped: Supercell Generation Goal: Maximise the image distance for the minimum number of atoms (electrons) in our supercell • In certain cases, may want to optimize dielectric-weighted minimum image distance (e.g. 2D/low-dimensional materials), or adjust to optimize kpoint sampling 11 *Dielectric-weighted “optimal” supercell may be desired for highly-anisotropic systems
10. 12 Efficient, direct optimization of image distance over all possible supercell transformations, and accounting for rotational invariances 12 doped: Supercell Generation Example: Cubic Structure ➡ minimal DFT/computational cost Mean improvements of ~10% compared to custom ASE scans -> ~35% DFT cost reduction (Averaged over several crystal systems)
11. 13 doped: Charge State Estimation • Defects can adopt various charge states in materials – but which ones will actually be stable? • False positives = Charge states included, but not stable (bad, but inevitable) • False negatives = Not included, but actually stable (very bad)
12. 14 doped: Charge State Estimation • Defects can adopt various charge states in materials – but which ones will actually be stable? • False positives = Charge states included, but not stable (bad, but inevitable) • False negatives = Not included, but actually stable (very bad) • As always, fully tunable (probability threshold to control lean/completeness) & flexible 14 * False negatives underestimated for PyCDT due to bias in dataset (Intermediate charge states which are metastable should still be calculated and are not considered false positives)
13. Defect Simulation Workflow 15
14. Defect Simulation Workflow 16 Automatic charge/spin state setting, consistency checks for calculation settings, k-point grid generation, fully flexible…
15. Calculation I/O 17 • Fully supported for VASP • Supercell file generation supported for CP2K, FHI-aims, Quantum Espresso… (via pymatgen) • Fully customizable, hybrid or GGA DFT, SOC settings, magnetic moments, +U… • Automatically set appropriate charge and spin states
16. 18 Calculation Parsing • Native, automated parsing support for VASP • Thermodynamic/plotting/analysis agnostic of underlying DFT/ML/FF code • Automatic consistency checks for: • k-point settings • supercell sizes • pseudopotentials (POTCARs) • input file / basis set (INCAR) settings • charge corrections • Automatic determination of defect site, charge, point symmetry, strain field, charge correction, charge delocalization, configurational/spin degeneracy, formation energy… aiding analysis
17. Defect Simulation Workflow 20
18. 21 21 Finite-Size Charge Corrections As always, automated but fully flexible/controllable (sampling region, excluded atoms – useful for low-dimensional systems…) Automated estimation of error in charge correction energy, based on variance of site/electrostatic potential in sampling region Kumagai-Oba (eFNV); anisotropic Freysoldt (FNV); isotropic
19. 22 Formation Energy Diagrams 22 • With/out metastable states • Controllable grouping of same-type defects by distance between equiv. sites • Auto labelling of inequivalent sites by point symmetry & neighbour distances • Easy looping through chemical potential space (just set “X-rich/poor” etc) • Fully customizable (as w/correction plots) • … :
20. 23 • Efficient competing phase selection based on chosen error tolerances and database phase diagrams. • Automated grid interpolation for multi-dimensional property optimization (doping, defect concentration, luminescence, recombination…)* • Automated plotting, analysis tools… Chemical Potentials / Competing Phases Cu-Si-Se Phase Diagram (Materials Project) *Work by Dr Alex Squires
21. 24 Chemical Potentials / Competing Phases Cu-Si-Se Phase Diagram (Materials Project) • Efficient competing phase selection based on chosen error tolerances and database phase diagrams. • Automated grid interpolation for multi-dimensional property optimization (doping, defect concentration, luminescence, recombination…)* • Automated plotting, analysis tools… *Work by Dr Alex Squires
22. 25 Chemical Potentials / Competing Phases Cu-Si-Se Phase Diagram (Materials Project) • Efficient competing phase selection based on chosen error tolerances and database phase diagrams. • Automated grid interpolation for multi-dimensional property optimization (doping, defect concentration, luminescence, recombination…)* • Automated plotting, analysis tools… *Work by Dr Alex Squires
23. 26 Chemical Potentials / Competing Phases • Efficient competing phase selection based on chosen error tolerances and database phase diagrams. • Automated grid interpolation for multi-dimensional property optimization (doping, defect concentration, luminescence, recombination…)* • Automated plotting, analysis tools… *Work by Dr Alex Squires
24. Defect Simulation Workflow 27
25. Defect Simulation Workflow 28
26. 29 Further Thermodynamics 29 1. Mosquera-Lois Chem Soc Rev 2023 2. Kavanagh et al. Faraday Discussions 2022 • Automatic determination of relaxed point symmetries and corresponding orientational & spin degeneracy factors g – important!1,2 • Rapidly scan through chemical potential space, (non-)equilibrium (annealed & quenched) Fermi level / carrier concentrations • Native support for temperature-dependent electronic structure & chemical potentials • Dopability limits & doping windows • Transition levels (incl. metastable states)… !! = #exp −Δ) *" + …
27. 30 Further Thermodynamics 30 • Automatic determination of relaxed point symmetries and corresponding orientational & spin degeneracy factors g – important!1,2 • Rapidly scan through chemical potential space, (non-)equilibrium (annealed & quenched) Fermi level / carrier concentrations • Native support for temperature-dependent electronic structure & chemical potentials • Dopability limits & doping windows • Transition levels (incl. metastable states)… 1. Mosquera-Lois Chem Soc Rev 2023 2. Kavanagh et al. Faraday Discussions 2022 !! = #exp −Δ) *" +
28. 31 Further Thermodynamics 31 1. Mosquera-Lois Chem Soc Rev 2023 2. Kavanagh et al. Faraday Discussions 2022 • Automatic determination of relaxed point symmetries and corresponding orientational & spin degeneracy factors g – important!1,2 • Rapidly scan through chemical potential space, (non-)equilibrium (annealed & quenched) Fermi level / carrier concentrations • Native support for temperature-dependent electronic structure & chemical potentials • Dopability limits & doping windows • Transition levels (incl. metastable states)… !! = #exp −Δ) *" +
29. 32 Further Thermodynamics 32 1. Mosquera-Lois Chem Soc Rev 2023 2. Kavanagh et al. Faraday Discussions 2022 • Automatic determination of relaxed point symmetries and corresponding orientational & spin degeneracy factors g – important!1,2 • Rapidly scan through chemical potential space, (non-)equilibrium (annealed & quenched) Fermi level / carrier concentrations • Native support for temperature-dependent electronic structure & chemical potentials • Dopability limits & doping windows • Transition levels (incl. metastable states)… !! = #exp −Δ) *" +
30. 33 Reproducibility • All objects & info directly output to JSON/yaml • Small files with all necessary info to reproduce, encouraged! • Includes generated/parsed defects, calculation settings, competing phases, corrections, thermodynamics Python framework -> plug and play with other packages (e.g. CarrierCapture/nonrad/TLC for electron-hole recombination, pymatgen, ShakeNBreak, easyunfold, atomate2, AiiDA…)
31. Acknowledgements Contributors: Alex Squires, Adair Nicolson, Irea Mosquera- Lois, Alex Ganose, Bonan Zhu, Katarina Brlec PhD Supervisors: Profs David Scanlon & Aron Walsh & user feedback, often from Walsh & Scanlon group members Current Funding & Position: Postdoc w/ Prof. Boris Kozinsky @ Harvard (Materials Intelligence Research (MIR) Group) Code docs: doped.readthedocs.io Kavanagh* et al. Journal of Open Source Software 2024 Scan for docs/paper: @Kavanagh_Sean_ Solar cells, batteries, thermoelectrics, transparent conductors, ferroelectrics, photocatalysts, phosphors…
36. 39 Automated Local Strain / Displacement Analysis
37. 40 Automated Shallow Defects / Perturbed Host States Analysis
38. 41 Automated Dopability Analysis
39. 42 Automated Transition Level Analysis
40. • Automated charge correction error analysis • High-throughput compatibility (AiiDA, atomate(2)…) • Simple interface with CarrierCapture.jl, nonrad, ShakenBreak etc via pymatgen API • Multiprocessing & numerical optimization for expedited generation, parsing & analysis (even for highly-complex / disordered systems) • Concentration / formation energy / symmetry / … tabulation • Gas molecule generation for competing phases • … 43
41. 45 ShakeNBreak 45 Run calculations (semi-automated w/snb-run) shakenbreak.readthedocs.io
42. 46 Chemical Potentials / Competing Phases • Efficient competing phase selection based on error tolerances and database phase diagrams. • Automated grid interpolation for multi- dimensional property optimization (doping, defect concentration, luminescence, recombination…) • Automated plotting, analysis tools…
43. 47 Automated Symmetry & Orientational/Spin Degeneracy Analysis