Spin-polarized quantities return order

Describe the issue

#500 brought about this issue. (@tfrederiksen )

The problem is the handling of the spin-densities is that the current read_data (for polarized) returns [up, down], whereas NC/SOC returns [total, x, y, z].

This issue is to discuss the option to change the polarized return order to be [total, z].

POSITIVE:

• this would be conforming to the NC/SOC calculations
• the code would always be the same if users just wanted to plot the total charge ([0]) SIESTA:
• Return values would not be the same as the output data (say Mulliken charges)

Comments would be much appreciated!

[juijan]

I would prefer [total, z] for consistency.

Typically the spin and the total charge are what is most relevant, in my opinion, not what the charge in either spin channel is on its own.

Regarding the index order, I would say first spin then orbitals.

[pfebrer]

I have never done a (meaningful) SOC calculation 😃

From my experience handling NC spins in the visualization module, I agree that it is a bit of a pain to have a completely different way of processing NC and diagonal spins.

But from my experience as a user, up/down components feel very intuitive. It is not easy for me to understand what total and z would store and how would you convert to up/down. Could you perhaps clarify it with an example of an array storing eigenvalues for a given k point?