When astronomers attack (each other) by MLR

I'd like to talk a bit about an astronomy story some folks may have seen floating around the news some time ago, and have likely forgotten about. There's a nearby galaxy, catalogued as KKS[2000]04 (after the Karachentsev catalog), which a team led by one Pieter van Dokkum recently argued has no dark matter content. They dubbed it NGC1052-DF2, after the galaxy they feel it's associated with, and their telescope (Dragonfly), and this is the name you'll usually see it listed under.

So why is this interesting? Well, the galaxy, at the distance they measured of ~20 megaparsecs (something like 60 million light years), looks to be extremely diffuse and low in density. It's so low density, in fact, that were some alternative gravity theories correct (e.g. MOND), the stars should be firmly in the non-Newtonian, non-GR framework and therefore should be moving accordingly. But van Dokkum's team found their motions to be fully explainable by the starlight alone -- not in the MOND regime, despite the low density. So ironically, this lack of dark matter provides strong evidence for the existence of dark matter: such galaxies could not exist in a dark-matter free universe, because if gravity was modified, ALL ultra-diffuse galaxies should look like they're dark matter dominated.

Very cool. There are, however, two major sticking points here.

1.) The motions of stars are notoriously difficult to measure. Individual stars are faint when they are far away, and in order to measure motions, one has to split their already faint light into a spectrum to measure redshift. So the amount of light you get at each wavelength is the total light (faint!) of the star divided by the number of wavelength bins (thousands, let's say). Point is, it's tough.

So van Dokkum's team did not use individual stars in this galaxy to measure the kinematics. They used globular clusters. Theoretically this is fine; globular clusters in most galaxies are randomly distributed around their hosts, so they serve as good tracers of the whole galaxy's gravitational field. The problem here is, rather, statistical.

In van Dokkum's paper, they measured the redshifts of 10 globular clusters. They then rejected one as an outlier. What they are trying to measure is something called the velocity dispersion -- the standard deviation of the velocities. Anyone with a passing familiarity with statistics and normal distributions would know that measuring a dispersion accurately from only 9 data points is very hard. Many other authors took issue with this analysis, leading to a subsequent revision by the original authors, which was still controversial.

To date, this issue has not been resolved. It's a bit dry in terms of reading, but anyone interested in small number statistics could learn a lot from following this discussion. But there's another issue that popped up recently.

2.) One of the most notoriously difficult things to measure in astronomy is distance. There is, in fact, only one way to directly measure distances (parallax), and everything else is built from what we call a distance ladder -- theoretical or empirical relations derived from known distances using parallax or other distance metrics. We've gotten pretty good at this, but there are still problems.

Case in point: this particular galaxy. Now, a sister article was published by van Dokkum's team alongside the original, discussing its globular cluster population, which appeared to consist of mostly unusually bright clusters. Fair enough: the galaxy seems unusual in many respects, and it has an unusual GC population. This is, of course, assuming the distance to the galaxy is ~20 Mpc as they claimed (things look fainter the farther away they are).

But the sticking point comes from this: it turns out that populations of GCs do not vary much from galaxy to galaxy (hence why this GC population was described as being unusually bright). The globular cluster luminosity function (GCLF), that which describes the distribution of brightnesses of GCs for a given galaxy can (and often is!) used to measure distance. So, if one assumes the GC population for this galaxy is normal, and uses it to derive a distance, you instead get a number like 13 Mpc.

Things that are closer to us appear bigger on the sky, obviously. Therefore, if this galaxy is really at 13 Mpc, not 20, its actual physical size is much smaller than previously assumed. This means its density is higher -- so high, in fact, it is no longer in the MOND or modified gravity regime, and is instead a perfectly normal dwarf elliptical galaxy, with dark matter. EVERYTHING changes.

So where does this number 20 Mpc come from? Surface brightness fluctuations: a distance metric that relies on the fact that more stars can be crammed into one pixel on your camera the farther away a galaxy is. Unfortunately, different kinds of stars produce different amounts of light in different colors (e.g. massive young stars are bluer than and can outshine old red stars by a factor of 10--100), so this needs to be calibrated. van Dokkum's team used a well-established calibration, but the galaxy's broadband color fell outside the range of the calibration they used, so they just extrapolated.

Again, people took issue with this. A rebuttal paper was soon published from a team in the Canary Islands led by Ignacio Trujillo, who found using five different distance metrics that the galaxy is, in fact, not at 20 Mpc but at 13 Mpc. One of their strongest arguments comes from the tip of the red giant branch (TRGB) method, which is one of the most accurate and well-calibrated distance metrics for galaxies within about 20 Mpc. This method involves, more or less, counting stars, which means that the stars must be resolved (obviously separated from each other, not all mashed into the same pixel), so to ensure this was the case they derived this distance using stars both in the center of the galaxy (where there are more stars, hence better counting statistics) and in the outskirts (where there are fewer stars but they are more spread out, hence less likely to fall into the same pixel).

Needless to say, van Dokkum's team takes issue with this revised distance -- they argue that the stars are not in fact resolved in the images used to derive the red giant branch distance, and that in such cases one can find a "fake" TRGB that's brighter (more stars mashed into the same pixel) and therefore gives a closer distance (which is a fair point, and well-established in the literature). Trujillo's team state that the TRGB distance from the galaxy outskirts resolves that issue either way given the lower density of stars. And so the debate goes on -- is the galaxy actually at 13 Mpc and perfectly normal? Even if it is at 20 Mpc, do its 9 globular clusters accurately trace the underlying mass distribution? Does it have an unusual GC population or not? Is it dark-matter free or not? And so on and so forth.

So there you have it. A little bit of how the astronomy sausage is made. I'm leaving some aspects of this out, still, since this journal entry is already quite long, but I think you guys get the gist by now. For what it's worth, my money is on the 13 Mpc distance: otherwise we fall on the wrong side of Occam's Razor in two ways (apparent dark matter content and globular cluster luminosity function). But hey, this is what they call "peer review".

When astronomers attack (each other)


2 September 2018 at 01:29:28 MDT

Journal Information


Tags Modify

Edit Tags


  • Link

    About a third of the way through, I was honestly lost. The context and terminology went completely over my head. And this is why I bailed on college and went to a trade school.

    • Link

      Probably just means I'm terrible at writing sciencey stuff for general audiences.

  • Link

    Where measurement metrics are subject to very thick accuracy bars, this is where crossing one's data T's and dotting those assumption 'i's (Like using other measurement methods) is important. Else you leave your paper open to ripping by your collegues.

    Plus astronmers are human, or doggish, in that smelling the blood spoor of inconsistency may result in a big dog fight.

    • Link

      Also doesn't help when you publish your shoddy distance measurement in a high-profile journal like Nature.

      By the way, 'spoor' is a great word that I'll have to put in my personal dictionary.