oqp.library.openqp_dftb

OpenQP runtime adapter for the optional OpenQP-DFTB backend.

Attributes

`HA_TO_WAVENUMBER`
`FINE_STRUCTURE`

Classes

OpenQPDFTBAdapter

Functions

dftb_soc(mol)

Module Contents

class OpenQPDFTBAdapter(mol)

Make OpenQP-DFTB look like a normal OpenQP energy/gradient provider.

mol

config

dftb

natom

nstate

energy()

Return an OpenQP-style state-energy array and update molecule data.

reference()

Return the DFTB reference energy in the same shape as SCF reference().

excitation(_ref_energy=None)

Return DFTB total state energies after a reference-style call.

gradient(states)

Return an OpenQP-style gradient array indexed by state number.

states_overlap(xyz_old, xyz_new, mo_old, mo_new, x_old, x_new, *, noca, nocb, multiplicity, tlf_order=2)

Cross-geometry MRSF/SF state overlap via libopenqp_dftb_c.

All matrix inputs are FORTRAN-FLAT 1-D arrays (i.e. tag.ravel() of the
native-convention tags). Returns (overlap_mo_tag, states_overlap_tag)
in the same native tag layout (transpose view of the Fortran matrices).

soc_matrix(mo, x_singlet, x_triplet, *, noca, nocb)

Raw one-center MRSF SOC matrix (no alpha^2/2) via libopenqp_dftb_c.

Inputs are Fortran-flat 1-D arrays; returns (hsoc_re, hsoc_im) as
(dim, dim) numpy arrays in the GAMESS state ordering, where the
row/column index layout matches the native soc_mrsf convention.

HA_TO_WAVENUMBER = 219474.6313708

FINE_STRUCTURE = 0.0072973525693

dftb_soc(mol)

Standalone SOC driver for method=dftb (mirror of the native compute_soc).

Runs the MRSF response for both target multiplicities from the same ROKS
reference, builds the one-center SOC matrix, diagonalizes
diag((E - e0) * ha2wn) + hsoc * (alpha^2/2 * ha2wn) in numpy, and fills the
same OQP:: tags the native soc_mrsf produces (soc_eval in cm^-1 relative to
the lowest MCH excitation; soc_hsoc_* raw, without the alpha^2/2 prefactor).