1_data_prepare/README.md

Diamond – Step 1: Data Preparation

Generate displaced supercell configurations, run QE DFT, transform to AO basis, and verify band structure reconstruction.

Prerequisites

1. QE pseudopotential: Place C.upf in ../pseudos/
2. Siesta pseudopotential: Place C.psml (or appropriate file) in ../aobasis/
3. QE on PATH: Source ~/scripts/tools/qe.sh or set manually
4. Conda environments: hpro (HPRO), deeph (DeepH)

Workflow

# 1. Create input files and directory structure
python prepare.py

# 2. Run QE calculations (SCF + pw2bgw + bands)
python run.py

# 3. Transform Hamiltonians to AO basis via HPRO
python reconstruct.py

# 4. Compare QE and reconstructed band structures
python compare_bands.py

Output

data/
  disp-01/ .. disp-50/    displaced 2×2×2 supercell SCF + reconstruction
  bands/
    uc/                   pristine unit cell bands
    sc/                   pristine 2×2×2 supercell bands
    kpath.json            k-path used for band calculations
band_compare_uc.png       QE vs HPRO comparison for unit cell
band_compare_sc.png       QE vs HPRO comparison for supercell

Parameters (params.json)

Key	Description
supercell_size	Supercell dimensions (default 2×2×2)
n_displacements	Number of displaced configs (default 50)
displacement_groups	Amplitude groups in Angstrom
reconstruction.method	"original" (real-space H(R))
reconstruction.nbnd	Bands to include in reconstruction comparison
cluster.run_sh	Path to cluster submission script (null = local)