cpp-hf 1.0

Hartree–Fock (k-grid) with FFTW + Eigen via pybind11

cpp_hf — Fast Hartree–Fock on k‑grids (FFTW + Eigen + pybind11)

cpp_hf provides a compiled extension module exposing a fast Hartree–Fock self‑consistent field (SCF) loop for uniform 2D k‑grids. It uses FFTW (batched 2D complex DFTs), Eigen (Hermitian eigendecompositions), and pybind11 to bind the C++ implementation to Python.

• Batched 2D FFTs (FFTW guru interface)
• Dense Hermitian diagonalization (Eigen)
• EDIIS + Broyden mixing (robust convergence)
• Simple, NumPy‑friendly API

This package is designed to interoperate with ContiMod, but it can be used on its own as a lightweight HF kernel.

Project links:

• PyPI: https://pypi.org/project/cpp-hf/
• Source: https://github.com/skilledwolf/cpp_hf

Installation

Quick install (prebuilt wheel if available):

pip install cpp-hf

Notes:

• The import name is cpp_hf (underscore), even though the package name uses a hyphen.
• Wheels may not be available for all platforms/Python versions yet; in that case pip will build from source.

Build from source (requirements):

• Python 3.8+
• A C++17 compiler
• FFTW and headers (threads variant optional)
• CMake ≥ 3.21

Build via pip (with optional CMake flags):

pip install . \
  -C cmake.define.HF_USE_OPENMP=ON \
  -C cmake.define.HF_USE_FFTW_THREADS=ON

macOS (Homebrew) dependencies:

brew install fftw eigen cmake

If FFTW is not found, set one of PKG_CONFIG_PATH, FFTW3_DIR, or CMAKE_PREFIX_PATH to point to your installation.

The repo includes a convenience script that activates a conda env and builds:

sh build_install.sh

Quick start

import numpy as np
import cpp_hf

# Grid and shapes
nk = 128; d = 2
weights = np.ones((nk, nk)) * ( (2/nk)*(2/nk) / (2*np.pi)**2 )  # scalar mesh measure
H = np.zeros((nk, nk, d, d), dtype=np.complex128)
Vq = (1.0/np.sqrt((np.linspace(-1,1,nk)[:,None]**2 + np.linspace(-1,1,nk)[None,:]**2) + 0.1)).astype(np.complex128)[...,None,None]
P0 = np.zeros_like(H)

# Target electron density (half‑filling)
ne_target = 0.5 * d * weights.sum()

P_fin, F_fin, E_fin, mu_fin, n_iter = cpp_hf.hartreefock_iteration_cpp(
    weights, H, Vq, P0,
    float(ne_target),  # electron density target
    0.5,               # temperature
    50,                # max_iter
    1e-3,              # commutator tolerance
    6,                 # mixing history (DIIS/EDIIS/Broyden)
    1.0,               # mixing alpha (kept for compatibility)
)
print("iters:", n_iter, "mu:", mu_fin, "E:", E_fin)

API

hartreefock_iteration_cpp(
    weights:      np.ndarray[(nk,nk), float64],
    hamiltonian:  np.ndarray[(nk,nk,d,d), complex128],
    v_coulomb:    np.ndarray[(nk,nk,1,1) or (nk,nk,d,d), complex128],
    p0:           np.ndarray[(nk,nk,d,d), complex128],
    electron_density0: float,
    T: float,
    max_iter: int,
    comm_tol: float,
    diis_size: int,
    mixing_alpha: float,
) -> tuple[P, F, E, mu, n_iter]

• weights is the uniform k‑mesh measure (e.g., dkx*dky/(2π)^2).
• v_coulomb can be scalar per‑k ([...,1,1]) or a full matrix per‑k ([...,d,d]).
• Returns the converged density P, mean‑field F, total energy, chemical potential, and the iteration count.

License

GPLv2+ — see LICENSE.

Third‑party notices: see THIRD_PARTY_NOTICES.md.

Acknowledgments

• FFTW — GPLv2+ (www.fftw.org)
• Eigen — MPL2
• pybind11 — BSD‑style