PySCF, SAD, and SAP Initial Guesses

OpenQP can use PySCF to prepare molecular orbitals or initial density guesses and then import them through OpenQP's native JSON restart format.

Supported [guess] type values are:

• pyscf: run a PySCF SCF calculation and import the converged orbitals.
• sad: build a superposition-of-atomic-density initial guess using PySCF's atomic-density guess.
• sap: build a superposition-of-atomic-potential initial guess using PySCF's SAP guess.

PySCF is installed with the OpenQP Python package dependencies. MOKIT is not required for pyscf, sad, or sap; MOKIT remains optional only for broader external wavefunction conversion workflows.

Basic Examples

SAD guess

[input]
system=
   O  0.0000000000   0.0000000000  -0.0410615540
   H -0.5331943294   0.5331943294  -0.6144692230
   H  0.5331943294  -0.5331943294  -0.6144692230
charge=0
runtype=energy
basis=6-31g*
functional=bhhlyp
method=hf

[guess]
type=sad

[scf]
type=rhf
multiplicity=1

SAP guess

[guess]
type=sap

Converged PySCF guess

[guess]
type=pyscf

Notes

• sad and sap generate a PySCF initial density and export it to OpenQP without requiring an external MOKIT conversion step.
• pyscf performs a PySCF SCF calculation first, then exports the converged orbitals to OpenQP.
• The exported restart data include molecular orbitals, orbital energies, density matrices, atoms, coordinates, SCF type, basis, and library information in OpenQP JSON format.
• If SCF convergence is difficult, compare final SCF energies and not only iteration counts; different initial guesses can converge to different ROHF roots.
• For neutral mid-sized organic ROHF references, Huckel remains a good first/default attempt. SAP is often a useful retry when Huckel stalls, falls back to TRAH, or reaches a suspicious high-energy ROHF reference.

Back