Introduction to Hartree-Fock Molecular Orbital Theory Part 3

We discuss the Hartree-Fock equations, the Hartree-Fock algorithm, orbital energies, and practical considerations

[@runakovacs4759]

I love your channel. I wish it was years back - I always lamented how theoretical chemistry had no real youtube presence. I'd have probably not understood too well, but who knows!

[@kenethcoleaudencial8981]

Thank you for the lecture. It really means a lot

[@afborro]

Hello. Thanks once again for the lectures :) Do you have any tips or good reference how to implement SAD. I can sort of see one might do a Hartree Fock on a per atom basis in a small basis set (in UHF) to serve as guess for a molecule, I think I could get that far with the code I have already. Currently I have core and GWH only. My main question: Lets say you have a large basis set for a molecule, how would you map the atomic guess density (in the small basis) to the density matrix for the molecule (in the larger basis). I know there is probably too much here to answer in a few sentences, :) anything would be appreciated. A psi4numpy tutorial would be a nice addition :D while I code in C++. A quick eyeball at psi4 is pretty hard to follow on this part, besides I want to figure out by myself. My personal hack/thought was one could smear out the density and sort of average it out over the new orbitals.

[@DavidSherrill1]

yeah, this is too complicated for me to answer quickly or easily :) But a couple of comments: (1) in my opinion it is better to do like you say, and compute the atomic densities on the fly, then stitch them together. The alternative is to pre-compute them. That seems good, except the pre-computation would have to be done in certain pre-defined basis sets... which is annoying when users then want to use a different basis. (2) A tricky problem is what do to about fractional occupations in the atomic computations, and how to project that onto molecular computations where you might have closed shells, etc. We have a way of handling this in Psi4 but I forget offhand exactly how it works.

[@afborro]

@@DavidSherrill1 Thanks. Food for thought. Not high priority but would be nice to have .. one day. I currently already have a hacky version I cooked up myself, definitely converges faster than core, but only works for a minimal basis, when it is actually more important in the larger basis. I'll follow the smearing idea one day .. if that is a fail I'll have just accept I'll have to figure it out from psi4 source one day with a strong coffee :)