Mapping the MW's UFDs

16 Nov, 2025

Today I'm thinking about some in-progress work that builds off of my very first paper.

The original paper

The gist is that the things we can observe about the Milky Way's satellites (their luminosities, velocity dispersions) have to come from the baryons, but I'd like to know properties of their DM halos. To bridge the gap between observed quantities and the DM, you can forward-model a huge number of MW satellite systems, trusting that you are getting DM halos that host physically sensible galaxies. These galaxies won't necessarily look like the MW's satellites directly, but we can extract conditional distributions that hopefully preserve the physical correlations.

Concretely, say we observe a dwarf galaxy with dispersion $σ$ , or rather a distribution of possible dispersions $f_{o b s .} (σ)$ . From the simulated model, we can extract a distribution of halo masses that provide that dispersion, $f_{p r e d .} (M_{v i r} | σ)$ . Then, by Bayes, a reasonable estimate for the dwarf's halo mass is the distribution

f (M_{v i r}) = \int f_{p r e d .} (M_{v i r} | σ) f_{o b s .} (σ) d σ,

which is essentially weighing the forward-modeled halo distribution by how well it matches the observed value of $σ$ . This is especially nice with an efficient forward model where you can vary over a bunch of systematics, like how baryonic feedback affects things, the stellar mass-halo mass relation, etc.

What I'm doing now

In the original paper, I applied this method to the classical satellites. I'm now working with the Safdi group (UC Berkeley) to extend it to the MW's ultra-faint population. An interesting thing has come of this: lots of the UFDs have kinematics consistent with very dense halos (specifically, a lot of mass packed into their half-light radii). In fact, it seems like there aren't even enough dense satellite halos in the model to accommodate them all.

There are, of course, things to be concerned about:

The stellar masses of our simulated dwarfs are super simplistic; we just take a stellar mass-halo mass relation to populate a satellite at infall and evolve it to the present day, tracking tidal mass loss. We don't have any fancy galaxy formation physics to really get good initial conditions. This should be okay, though, since we can make a statement about counting halos that doesn't require them to be populated with galaxies.
Using the kinematics for ultra-faints is a little tricky. The intrinsic dispersion is often only O(km/s), which is comparable to what stellar binaries can introduce in the data. In fact, most observational systematics would push the dispersion toward being overestimated rather than underestimated, which could definitely contribute to this effect.

Things I'm thinking about today

While the stellar masses are a minor concern, it'd be interesting to have a more sophisticated galaxy formation model from the perspective of metallicities. That's one angle of things that our model doesn't capture at all, and if we could use it to condition on quenching/infall times, we might pick out qualitatively different halos (e.g. early infall time -> early formation time -> denser halo). In my original paper, the infall time didn't seem to have too much of an effect on the inferred halo profiles, but thinking back on it now, I think the $f_{o b s} (t_{i n f a l l})$ distributions I used were pretty wide. Metallicity distributions might(?) be narrower and therefore more constraining.

I'm also thinking about MW-specific weirdnesses that the model may not capture. For example, there are lots of streams in the inner regions of the MW (Shipp et al. 2022), which is maybe tied to the MW having a particularly-compact mass distribution itself?¹ If the MW is compact and has lots of low-pericenter streams, maybe we should also expect a large population of low-pericenter intact satellites and we've only observed the densest.

If the low-pericenter objects are contributed by the LMC, this physics should already be captured by our model, but I'm still a little suspicious of the LMC. It seems possible that the low-pericenter objects weren't directly contributed by the LMC, but its gravitational influence shifted things around, which wouldn't be appropriately modeled here. This is very speculative on my part though; I'm not very familiar with the literature on the LMC's effects on the MW's other satellites.

Much to think about!

This is something I learned for myself in a recent project. Interestingly, this might be connected to a retrograde GSE merger (Funakoshi et al. 2025)!↩

#MW satellites #semianalytics #ultrafaints