EFT-vdW Prototype

This repository prototypes an EFT-style atomic channel pipeline for computing imaginary-frequency polarizabilities, C6 coefficients, and long-range pairwise dispersion tails.

Setup

Install the Python dependencies and run the test suite:

python3 -m pip install -r requirements.txt
python3 -m unittest discover

Minimal Ar Reproduction

Reproduce the current Ar TDHF benchmark and Ar2 long-range tail:

python3 pyscf_export_ar_tdhf_oscillators.py --basis aug-cc-pVQZ --nstates 200 --output ar_tdhf_channels.csv
python3 run_alpha_table.py --input ar_tdhf_channels.csv --output ar_tdhf_alpha_c6_table.csv
python3 run_c6_table.py --input ar_tdhf_channels.csv --output ar_tdhf_c6_table.csv
python3 compare_alpha_c6.py --eft ar_tdhf_alpha_c6_table.csv
python3 run_ar2_tail.py

Current Ar Prediction Baseline

The PySCF 3D MO oscillator route is the current prediction baseline for Ar. Its summary is stored in results/ar/ar_mo_basis_convergence.csv.

The radial shell-average route is diagnostic only until the per-shell TRK overcounting is fixed. In particular, diffuse radial shell-average runs can violate the oscillator-strength sum rule even when the corresponding 3D MO dipole-integral route does not.

Ar Basis Convergence Caveat

results/ar/ar_basis_convergence.csv includes three standard PySCF basis sets:

• cc-pVTZ
• aug-cc-pVTZ
• aug-cc-pVQZ

It also includes d-aug-cc-pVTZ-local. This is not an official Basis Set Exchange d-aug-cc-pVTZ basis. It is a locally generated, even-tempered second diffuse augmentation built from PySCF's aug-cc-pVTZ basis by adding one extra diffuse primitive per angular momentum channel. Treat it as a reproducible local sensitivity test, not as a published basis-set benchmark.

Status And Limitations

Implemented:

• Oscillator-channel alpha(i xi) and C6 backend
• Calibrated, MO, TDHF, and radial diagnostic routes
• Ar TDHF benchmark and Ar2 long-range tail
• Ar basis, TDHF nstates, and oscillator-strength audit tables
• TDHF core/valence/semicore decomposition for Ar, Kr, Mg, and Ca
• PSP+EFT-core benchmark candidates:
  • Mg q2 is the strongest clean matched-q2 benchmark
  • Sn q4 is the strongest official-matched secondary candidate found so far
  • Ge q4 is a weaker official-matched control
• Generated-basis freeze/audit protocol for future Ca/Sr/Ba q2 work

Not yet implemented:

• Valence-screened W_v
• RPA/log-determinant many-body energy
• Periodic systems and forces

Current Benchmark Pointers

• Project status: PROJECT_STATUS.md
• Mg q2 benchmark: results/mg_q2_clean_benchmark.md
• Sn q4 reference repair: results/official_matched_candidate_scan/sn/reference_repair/summary.md
• Official-matched candidate scan: results/official_matched_candidate_scan/summary.md
• Generated basis protocol: docs/generated_basis_protocol.md