Fix lightcone neutrino mass maps

Fixes #827 (closed). First the easy part: adding the missing factor of 1/3 in the HEALPix pixel volume. This affects the NeutrinoMass and TotalMass lightcone maps. We also need to accumulate the correct value to each pixel when using use_delta_f_mesh_only.

assigned to @jch

changed the description

We should also add the neutrino weights to the lightcone particle output.

added 2 commits

• 16aaca3f - Include weight in neutrino maps where necessary
• 0b2766c0 - Add neutrino weights to the lightcone particle output

Compare with previous version

The weight calculation is copied from power_spectrum.c. Maybe the struct neutrino_model should be stored somewhere so that it's not regenerated every time a particle crosses the lightcone.

This would improve data locality I think.

Do the contents of the struct change over time?

They do not.

@jch do you want to make that change then?

Yes, I just need to figure out where to keep a copy of the struct.

I was going to add it in the lightcone_props struct but I think that causes a circular include because neutrino.h includes engine.h which includes lightcone.h which would need neutrino.h.

Does it make sense to have it into engine.h?

added bug label

Are there updates on this?

Not yet. I'll try to do a bit of testing this week.

I've run a FLAMINGO L1000N0900/DMO_FIDUCIAL with and without this fix. The mean density of neutrinos as a function of radius in the maps looks ok - we no longer have four times too many!

Great!

@dc-elbe1 any last minute comments?

marked this merge request as ready

requested review from @dc-elbe1

Here's what happens if I run the old (labelled "master") and new maps for one shell through healpy.anafast to get the angular power spectrum.

Excellent, thanks very much John! The plot looks as expected.

changed the description

Are there any news on this?

The remaining issue is whether we want to try to avoid calling gather_neutrino_consts() every time a particle crosses the lightcone, and if so where can we store a struct nu_model so that it doesn't cause circular dependencies. It might not matter very much. I think the overhead will be small compared to the rest of the lightcone calculation.

ok, then we can go ahead with the current version.

approved this merge request

added 1 commit

• 0b61c01b - Applied formatting script

Compare with previous version