Ewald energies and energy differences from EwaldTerm do not match EwaldSummation for charged cells

[lbluque]

Expected Behavior

The portions of the Ewald energy (total, real, reciprocal, and point) should match when computed using the "expanded" matrices created with the EwaldTerm or directly with an EwaldSummation for an ordered structure.

For the case of charged cells all should match except the charged cell correction since the "expanded" Ewald structure has the charge of the disordered structure and not the particular ordered structure.

Current Behavior

The computed Ewald interaction for charged cells using the "expanded" ewald matrices created by the EwaldTerm, do not match the corresponding Ewald energies computed directly for a structure using pymatgen.EwaldSummation when the structure is charged.

The real and reciprocal energies do not match, and as a result the total energy (even excluding the charged cell correction) does not match. The point energies do match.

In MC we should decide if it is necessary to add charge corrections when going off charge, because that is currently not being considered.

Everything works well for charge neutral structures.

Possible Solution

1. Need to find out what is changing the values! Currently I do not know what is causing this.
2. Need to decide if and how to add the charged cell correction in MC sampling.

Steps to Reproduce

For the real space energy for example:

1. Create a disordered structure and an ordered structure with the same supercell.
2. Obtain ewald energy directly from EwaldSummation(ordered structure).real_space_energy
3. Obtain ewald energy by using EwaldTerm(use_term='real').value_from_occupancy(occu for ordered structure, disordered structure)