Michael F. Herbst

See https://github.com/aiidateam/aiida-quantumespresso/blob/main/src/aiida_quantumespresso/workflows/protocols/magnetization.yaml and the discussion in https://github.com/aiidateam/aiida-quantumespresso/issues/1069.

As I mentioned on zoom, I think avoiding to muddle up physical structure and modelling details makes a lot of sense. But in line with that I would not call...

I don't think we can do without it. I agree a `system` argument should not be type-annotated unless needed, but the two examples above from DFTK are cases where removing...

Let me start by also issuing a big **thank you** to @cortner. I think as part of this discussion my general idea of what AtomsBase is for has already sharpened...

> Is there a better terminology? @niklasschmitz also suggested "particle_kind" as an option. Sounds like type but rings different bells for CS people. > There is some value in enforcing...

> how does species or particle_species work in relation to the concept of "chemical species" I think for the standard use case it works fine. I think only for coarse...

I also think a canonical implementation *outside* AtomsBase would be good. For me important would be: - Arbitrary properties - A good set of basic properties (e.g. like FlexibleSystem) I...

For reference: For now I added a workaround, manually calling the respective `rocSOLVER` routine: https://github.com/JuliaMolSim/DFTK.jl/blob/907fbf50d03aecda70e4a30fdb6c59f8e00c9537/ext/DFTKAMDGPUExt.jl#L10

> Looking into this more it really looks like we should have had a geometry field of Model, that would more closely match AtomsBase's System We could just store a...

About the return values, I think this would be a very good initial set: - `ham` - `basis` - `energies` (the first result of `energy_hamiltonian`) - `converged` (`true` or `false`)...