Allow unitcell to contain non-magnetic atoms

[drjbarker]

For beginners it is often not intuitive that we should only specify magnetic atoms for the unit cell in the config. Moreover the space group of the crystal may actually require the non-magnetic atoms to be correctly determined. We should therefore allow materials with moment=0.0 to be specified and used in JAMS only when determining symmetry.

[drjbarker]

We need to be careful here. At the moment vacancy impurities are represented as sites with zero moment but still exist in the spin array because they may be needed for the FFT in spectral methods.

jams/src/jams/core/lattice.cc

Lines 718 to 724 in affd0d1

	// lattice vacancies
	if (material.moment == 0.0 || material.spin == Vec3{0.0, 0.0, 0.0}) {
	globals::mus(i) = 0.0;
	for (auto n = 0; n < 3; ++n) {
	globals::s(i, n) = 0.0;
	}
	}

[SeanStansill]

Proposed solution:

Create a new optional material parameter is_magnetic. Non-magnetic sites should be set to False, the default for materials where this is not defined will be True (for magnetic sites and spin vacancies).

Within lattice.cc an extra if statement should be added to ensure only is_magnetic==True materials are added to globals::mus and globals::s. In this way, the validity of the FTT methods is preserved and non-magnetic sites can be ignored globally by the hamiltonian, solver and monitors.

This would also require a change in how globals::num_spins is initialized. Currently it is set to the number of spins in the system but this would have to be = atom_counter - non_magnetic_atoms;, = num_magnetic_atoms; or similar

[drjbarker]

It's a good proposal but for new users might be confusing. For example Robert the other day included oxygen sites and just gave them moment of zero because that seemed natural.

It might be that it is better to alter the impurity code (which is currently rarely used). So by default sites with a moment of zero are considered non-magnetic and a flag is needed to identify impurity materials.

Including non magnetic sites in crystals is probably best to be the default rather than assuming impurities because then if needed we can build support for directly reading unit cells in some formats like vasp, vesta, cif etc...

I imagine that num_spins would not need changing because the lattice code would still only generate data arrays for magnetic sites, the non-magnetic sites would exist purely for symmetry calculations.