Individual attributes for different species

[n01r]

While implementing Ionization in PIConGPU it became obvious that the attribute chargeState will not be sufficient in the future. For precise information especially on ion species it will become necessary to introduce another independent attribute:

• coreCharge or protonNumber or maxNumOfElectrons
  That way it will be clear if the species is e.g. Hydrogen or Carbon. Also it will be very native with respect to the underlying physics.

The issue #108 already suggested that e.g. plugins should be able to activate specific attributes. Ideally though, some attributes shouldn't even exist for some species. Setting a coreCharge for each electron does not only look like a waste of memory on a first glance but with the run time creation of electrons in #595 one might encounter more serious memory problems.

It should therefore be possible to assign attributes differently for different species. Memory should also be dedicated to that species accordingly and maybe even automatically.

Example

• we simulate initially neutral Carbon atoms
• they will eventually be stripped off of their 6 electrons
• our simulation is f**king awesome and therefore automatically notices that the amount of electrons increases and gives them more memory

[ax3l]

I am against coreCharge, protonNumber and neutronNumber is something I could relate to, they will be static frame attributes anyway (only static memory /registers used).

[ax3l]

after discussions today, we decided to rather store the # of bound electrons instead of the charge state (since it is easier to be expanded to electron configurations later on).

[n01r]

Should it still be float like you suggested with chargeState? I've never heard of an Effective Number of Electrons 😛 BUT there is an Effective Atomic Number.

[ax3l]

yes, float_X please. since a chargeState Q could in principle be an effective charge state, N_e in Q=Z-N_e is of course, too. (and even more important: Z is a float_X, too)

it's not super-obvious in the first place but likely to be useful in the future.

[n01r]

I already have one branch ready for protonNumber and neutronNumber which is based on my ionization feature branch. I am refactoring chargeState now ... trying to, let's see if it works out. 😉

[n01r]

The only thing left in this issue is automatic memory assignment based on how many particles may exist during the simulation, not only before. But that could be handled later, separately.

[ax3l]

sounds great, do you want to open a fresh issue for that (with a longer description of the memory assignment task)?

[n01r]

Yeah, I'd rather do that.