ExternalPotential in Bohr

[loriab]

Description

The psi4.core.ExternalPotential object has long stored its charge locations in the same units as its BasisSet's Molecule. Also, the usual way of initializing it is by creating a psi4.driver.qmmm.QMMM() object and then doing the unusual psi4.set_global_option_python("EXTERN", qmmmobj.extern) to a field of that object. That's the situation for "plain" external charges calcs, where there's only one potential and both charges and locations are specified.

Additionally, there's two related categories. F-SAPT can put different external potentials on any of fragments A, B, C, which have been passed in as energy("fisapt", external_potentials={"A": extern, ...}, ...). Also nbody can take an array of charges (not locations) that replace fragments when the fragments are ghosted. These are passed as energy(..., bsse_type=..., embedding_charges=[monoAchg, ...])

Problems:

• formation of Python objects in the input file and special options setting doesn't translate to QCSchema, which is the sole means of communication for many calcs in DDD.
• the most common calc, the simple single extern has no kwarg-to-energy way to set
• the uncertainty in units of ExternalPotential can be confusing and has led to bugs. Also, in DDD, the spec order of mol and extern may not be so clear to set the units.

This is No. 2 of the DDD series, #1351.

Todos

• Replace QMMM() with QMMMbohr() and issue an update guide if the former is called.
• For the common single-extern case, switch to energy(..., external_potentials=array instead of QMMM() object. This tests whether the external_potentials value is an array for a single extern, which gets processed and set immediately, or a dict of externs, which gets handled by the run_fsapt later on.
• relaxed the array spec by which externs are initialized. instead of q, [x, y, z] also allow q, x, y, z, which makes for very easy numpy processing for units transofmation.
• removed all the units handling around ExternalPotential
• docs

Questions

• ok to use external_potentials kwarg for both simple and fsapt multi frags use cases?

Checklist

• Tests added for any new features
• All or relevant fraction of full tests run

Status

• Ready for review
• Ready for merge

[loriab]

looks like MDI is the first casualty of QMMM and ecosystem GHA. This PR can have a warning for QMMM, then kill it off for DDD.

[JonathonMisiewicz]

LGTM now, many thanks for breaking DDD down.

[hokru]

lgtm. I think this will lead to less confusion about units.

[andysim]

Excellent - great to show the user how to get bohr from Angstrom easily :)

[andysim]

LGTM - this will be a little easier to generalize to other input formats. Thanks for taking care of it. This will simplify the blurred multipole specification, methinks.