This repository adds utility for PySCF to include SOMF corrections within spinor style and GHF style calculations. To do a spinor (j-adapted) style calculation, build a spinor SCF and attach an X2CAMF spin-orbit Hamiltonian with the .x2camf() shortcut (the spinor analogue of PySCF's scf.RHF(mol).x2c()):

from pyscf import gto
from socutils.scf import spinor_hf

mol = gto.M(atom=[["O", (0., 0., 0.)],
                  ["H", (0., -0.757, 0.587)],
                  ["H", (0.,  0.757, 0.587)]],
            basis='ccpvdz', verbose=4)

mf = spinor_hf.SCF(mol).x2camf()          # Gaunt + Breit on by default
e_spinor = mf.kernel()

# turn the Gaunt/Breit two-electron SOC corrections off:
e_dc = spinor_hf.SCF(mol).x2camf(with_gaunt=False, with_breit=False).kernel()

Note: the C/C++ libraries these features need are bundled with socutils (under socutils/lib) -- no external packages to install. They are compiled once with a single make at the repo root (needs a BLAS/LAPACK):

make            # builds libx2camf_c (X2CAMF SOC integrals),
                #        libzquatev   (Kramers-restricted spinor SCF), and
                #        libccsdt_clib (spinor CCSDT kernels) into socutils/lib

• x2camf (the SOC integrals) ships as a pure-C reimplementation exposed as import x2camf; an external upstream warlocat/x2camf pybind11 build is optional (only for A/B comparison, selected via X2CAMF_BACKEND / SOCUTILS_X2CAMF).
• zquatev (Kramers-restricted spinor SCF) is the bundled quaternion eigensolver -- the former xubwa/zquatev pip package is no longer needed.

See docs/source/install.rst for details (BLAS/LAPACK vendor selection, environment overrides). To do a GHF (spin-orbital) style calculation, use ghf.GHF with the same .x2camf() shortcut; it runs in a spin-orbital rather than a spinor basis and agrees with the spinor result to numerical precision:

from pyscf import gto
from socutils.scf import ghf

mol = gto.M(atom=[["O", (0., 0., 0.)],
                  ["H", (0., -0.757, 0.587)],
                  ["H", (0.,  0.757, 0.587)]],
            basis='ccpvdz', verbose=4)

gmf = ghf.GHF(mol).x2camf()               # Gaunt + Breit on by default
e_ghf = gmf.kernel()