HPRO: Hamiltonian Projection and Reconstruction to atomic Orbitals

HPRO is a Python package for converting plane-wave DFT calculation results to Hamiltonian matrices in atomic orbital basis. It serves as a bridge between first-principles calculations and tight-binding models or machine learning applications like DeepH.

Features

• Two conversion methods: Wavefunction projection and real-space reconstruction
• Multiple DFT interfaces: Quantum ESPRESSO, BerkeleyGW, GPAW, SIESTA
• MPI parallelization: Scalable calculations on HPC clusters
• DeepH format output: Direct compatibility with DeepH-pack

Installation

Basic Installation

pip install hpro

Or install from source:

pip install git+https://github.com/Xiaoxun-Gong/HPRO.git

With MPI Support

pip install hpro[mpi]

With Full Parallel Support (MPI + SLEPc)

pip install hpro[slepc]

Note: SLEPc requires PETSc to be installed with MKL Pardiso. See the SLEPc documentation for details.

Quick Start

from HPRO import PW2AOkernel

# Create kernel for real-space reconstruction
kernel = PW2AOkernel(
    aodata_interface='siesta',      # SIESTA orbital files
    aodata_root='./aobasis',
    hrdata_interface='qe-bgw',      # QE + BerkeleyGW
    vscdir='./scf/VSC',
    upfdir='./pseudos',
    ecutwfn=30
)

# Run reconstruction
kernel.run_pw2ao_rs('./output')

For diagonalization and band structure calculation:

from HPRO.aodiag import AODiagKernel

diag = AODiagKernel()
diag.setk(kpts, weights, symbols)
diag.load_deeph_mats('./output')
diag.diag(nbnd=36)
diag.write('./output')

Supported DFT Codes

Documentation

• AGENTS.md - Development guidelines
• design/ - Module design documentation

Requirements

• Python >= 3.9
• numpy >= 1.7
• scipy
• h5py
• tqdm
• matplotlib

Citation

If you use HPRO in your research, please cite:

@software{hpro2024,
  author = {Gong, Xiaoxun},
  title = {HPRO: Hamiltonian Projection and Reconstruction to atomic Orbitals},
  year = {2024},
  url = {https://github.com/Xiaoxun-Gong/HPRO}
}

License

GNU General Public License v3.0 or later (GPL-3.0-or-later)

Author

Xiaoxun Gong (xiaoxun.gong@berkeley.edu)