Satellites
Fig 1. Correction from last week. Adding kappas is much better at reconstructing kappa_hilbert than anything I've done before. (But note, kappa_hilbert isn't necessarily the quantity that is degenerate with time delay distances, we need to look into this. This is an important concern even if halo reconstruction doesn't work out in the end.). (There is now a concern over the smooth component but I will postpone that discussion till next time.)
Fig. 2. Proportion of kappa, mass and number contributed by galaxies different types of galaxy (central or satellite) for 1000 LoS of reconstruction radius 3 arcmin. These are for all for all LoS, but they are very similar if you make high and low kappa cut. A 3rd type of galaxy is not shown here "halos with no central galaxy".
- Satellites don't contribute much mass, but theysometimes contribute a significant amount of the convergence. Why? Satellites aren't that common in the simulation, they aren't very massive, so how can they still be important? Could it be because kappa is dominated by a small number of halos, and the rest is incremental?
If you don't include satellites, the scatter in figure 1 goes up from 0.014 to 0.021. So we do need to reconstruct them.
So how many halos are genuinely important?
Fig. 3. Histogram of log(kappa) from each of the halos along 1000 LoS, within 3 arcmin. Most halos in the cone don't contribute anything to the kappa. (Assuming truncation at 3 R_vir.)
Fig. 4. Histogram of number of 'important' halos along 1000 LoS. Important is defined as contributing an individual kappa greater than the threshold shown in the legend.
Scatter in Halo Mass Last time we discussed issues with reconstructing halo mass from stellar mass. Tommaso asked to see what happens if we take the stellar masses and use the Behroozi relation to infer halo mass:
Fig. 5. Reconstruction of kappa, using Stefan's stellar masses, and the Behroozi relation to go from stellar mass to halo mass. No scatter is used on the Behroozi relation, or on the observed stellar masses. The are few halos where the mass is significantly overestimated by the Behroozi relation, which in turn overestimates kappa_add. (I haven't shown the values of the scatter - they fluctuate significantly if any of the 1000 LoS get close to the few halos that are overestimated by 2 orders of magnitude)
But the Behroozi relation isn't valid for satellites, and it's the most (stellar) massive of these satellites that are causing the problem. As we saw before, satellites are sometimes important, though rarely dominant. Perhaps we can make a better reconstruction if we treat satellites separately to centrals? If I only reconstruct the centrals I get this:
Fig. 6. As figure 5, but only reconstructing centrals - satellites are ignored. It's better to ignore satellites than to reconstruct if you're going to get their masses badly wrong.
Conclusions. We need to use the same abundance matching relation as was used to create the stellar masses in the catalogues, and satellites have to be treated separately from central galaxies. We'll probably need a group finder.
Fig 7. A combination of Figures 1 and 5 .