Generate SAD guess with PSI4

#1

Hello everyone!

I am interested in using PSI4 to generate initial guess densities from atomic superpositions, ideally that I could then print as a .fchk, .wfn or any other structured file format. However, I am not very knowledgeable about the source code, and have not been able to find the SAD bit.

On the other hand, as it is not something particularly useful, I dont really know whether this is possible in PSI4 at all! Is there a guess=only equivalent in PSI4? Other options I can think about involve limiting the number of SCF cycles to 0, but if I am not wrong this will usually perform at least one SCF cycle (maybe only a diagonalization or some integrals, but same applies), which in a large system should be much more expensive than generating the SAD guess. Which is the only thing I really want.

Hence, this topic. Anyone has any info on this?

0 Likes

#2

You want to access this object:
http://psicode.org/psi4manual/master/psi4api.html#psi4.core.SADGuess

This tutorial shows you how to set the molecule/wavefunction object and calculate the SAD:

Most easily you can convert the data (D for density, C for orbital coeff) to a numpy matrix.
Check out the other Tutorials for the basics.
Formats like FCHK or molden might be possible with some tricks, but I cannot say from memory.

0 Likes