Goals - Installation - Usage - Milestones
PseudopotentialStudio (PStudio) aims to become a full fledged Python package to generate atomic pseudopotentials for electronic structure codes like Quantum-Espresso.
Why am I coding this package? There exist several pseudopotential generation codes: ld1, uspp, opium, atom, ape, oncvpsp, atompaw, fhi98pp... Their input files are difficult to write and poorly documented. They implement different pseudization methods and produce pseudopotentials with different file formats that cannot be used directly by every electronic structure code.
All existing codes are non interactive. Typically you have to repeat the following steps: (1) write the input file (2) run the code (3) extract data from the output (4) plot the results, until you are satisfied with the result. PStudio can be run in a Jupyter notebook and will provide utility functions to plot and analyze the results interactively.
Finally, the mathematics behind pseudopotential generation is seldom explained in the details, in the original papers. In most codes, the numerics (fitting, derivative matching, non-linear system, special functions) is implemented in Fortran. PStudio instead relies on NumPy and SciPy for the numerical part and on Matplotlib for visualization. Additionally PStudio uses LibXC for exchange and correlation and few Fortran routines from the ONCVPSP code.
- generate norm-conserving, ultrasoft and PAW pseudopotentials
- generate pseudo wavefunctions, GIPAW reconstruction, two-electron ERIs
- provide a didactic code to atomic/pseudopotential calculations
- provide a Python library to develop quickly new pseudopotential schemes
PStudio require a Fortran compiler like gfortran. After downloading the code, the Fortran routines can be built in-place:
git clone https://github.com/dceresoli/PStudio
cd PStudio
./build.sh
Create a new Python3 notebook with Jupyter or nteract and:
from pstudio import AE
ae = AE('Ti', xcname='LDA', relativity='SR')
ae.run(verbose=True)
...See the directory notebooks for further examples.
- Norm conserving TM and RRKJ methods
- MetaGGA functionals, GIPAW reconstruction, ERIs
- Norm conserving HSCV, ONCVPSP methods
- Ultrasoft Vanderbilt, RRKJUS methods
- Tables of PP radii from existing libraries
- PAW pseudopotentials
- other methods: MRPP, direct fitting (VBC-type)
PStudio includes routines from ONCVPSP and from GPAW. The pyxclib module was slightly modified from LibXC. PStudio is distributed under the GNU GPL v3 license.