CSF truncation

[NastaMauger]

Hello,

Is there a way to truncate the final wavefunction at different thresholds in CSF space after generating it using the following code?

from mrh.my_pyscf.fci import csf_solver

norb = 10
nelec = 4 

casci = mcscf.casci_symm.CASCI(rhf, norb, nelec)
casci.verbose = 9 
casci.wfnsym = 'A' 
casci.fcisolver = csf_solver(mol, smult=1)

casci.kernel()
csf = casci.fcisolver.transformer.ncsf
print(f'Number of CSF is: {csf}')

I would like to truncate the wavefunction in CSF space at various thresholds to explore how properties change as a function of the number of determinants.

So far, I have used a naive truncation procedure that sets a threshold based on the CI coefficients, but I would prefer to do the truncation based on the CSFs instead.

Thank you for your help!

[MatthewRHermes]

There is no truncation system built into the solver.

To access the CI vector in the CSF basis so that you can script it yourself, do this:

ci_csf = casci.fcisolver.transformer.vec_det2csf (casci.ci)

The meaning of the ith element of the vector ci_csf can be obtained by

print (casci.fcisolver.transformer.printable_csfstring (i))

[MatthewRHermes]

IDK what workflow you're trying to build here, but if you have some list of CSFs that you want to forbid for whatever reason, then I think you can accomplish this by patching fcisolver.transformer.vec_det2csf and fcisolver.transformer.vec_csf2det to set forbidden CSF elements to zero and either using fcisolver.davidson_only = True or additionally patching fcisolver.pspace like

idx_allow = # a boolean array of shape (ncsf,) which is True where a CSF is allowed and False where a CSF is forbidden
def new_pspace (*args, **kwargs):
    addr, h0 = old_pspace (*args, **kwargs)
    i = idx_allow[addr]
    return addr[i], h0[np.ix_(i,i)]