Sister blog of Physicists of the Caribbean. Shorter, more focused posts specialising in astronomy and data visualisation.

Thursday 3 October 2019

It's not a feature, it's a bug

A couple of weeks ago I mentioned a paper that tries to quantify how weird galaxies without dark matter are in the latest, cutting-edge simulations. The properties of galaxies which look like this in reality depend strongly on their distance, so the authors took the nice approach of quantifying how rare such objects are depending on their true distance. If they're close, then their peculiar velocity would be unusual but their dark matter content would be normal; if they're far away, then their velocities would be normal but their dark matter content would be unusually low.

What annoyed me was that they found objects in the simulations matching those criteria, but didn't describe how such objects (lacking dark matter or moving at weirdly high velocities) form in the simulations. If their formation strongly depends on environment, then the global numbers for how rare they are might be woefully misleading. At least one of those authors has a nasty habit of doing that.

The authors in the current paper, however, have indeed gone and found dark matter deficient galaxies (they do not consider the possibility of strong peculiar velocities here) in simulations and looked at them very, very carefully. And it turns out that there is a strong environmental dependence - but it's a bug, not a feature.

You can't simulate the entire visible Universe, and even if you could, you'd still have to make assumptions about its edge. The standard approach is that the boundary conditions are periodic, so that a galaxy which happens to exit* on one side simply re-appears with the same velocity on the other side. And all of the identified "oddball" galaxies without dark matter are found very close to the edge of the simulation.

* Pursued by a bear ?

Given that the near-edge volume in question is a tiny fraction (0.1%) of the whole simulation, that's already a massive red flag. But it's not quite enough to say for sure if this is the cause. For these weirdos share another common feature of experiencing recent mergers. Although galaxy mergers are common events, it's possible that the galaxy-finding algorithms could occasionally get very confused and start misidentifying which particles belong to which galaxy. So the galaxy could only appear to lose mass because its associated dark matter has been wrongly identified.

But that's not the case. When they looked at the oddballs in detail, they found that as they approached the edge of the simulation they slowed to a crawl, and they do indeed lose much of their dark matter. They're still not sure what causes the bug, as plenty of galaxies cross the boundaries without incident, but it appears to be related to the galaxies having unusually dense inner cores where acceleration is high. It's highly unlikely that this bug has caused any other problems - it's only affected a handful of galaxies out of many thousands - but a bug it very much is.

What does this mean for the earlier paper ? Well, if all simulated galaxies without dark matter are actually due to numerical artifacts, then it strengthens their claim that they're incompatible with the standard model. But this is by no means clear : selection criteria may mean the two teams have identified completely different objects. Since the current team's criteria are quite strict, there could still be dark matter deficient galaxies in the simulations which are a feature, not a bug. And it should also be a cautionary note that the simulations are not yet the be-all and end-all of the standard model of cosmology.

Isolated dark matter deprived galaxies in hydrodynamical simulations: real objects or artefacts?

We searched for isolated dark matter deprived galaxies within several state-of-the-art hydrodynamical simulations: Illustris, IllustrisTNG, EAGLE, and Horizon-AGN and found a handful of promising objects in all except Horizon-AGN. While our initial goal was to study their properties and evolution, we quickly noticed that all of them were located at the edge of their respective simulation boxes.

No comments:

Post a Comment

Back from the grave ?

I'd thought that the controversy over NGC 1052-DF2 and DF4 was at least partly settled by now, but this paper would have you believe ot...