Galaxy X: The dark galaxy that doesn’t exist

Astronomers almost detected an enormous satellite galaxy made of dark matter. Almost.

Ten years ago, a pair of astronomers at the University of California, Berkeley, put forth a bold idea. Based on n-body simulations and radio observations of the Milky Way’s disk, they proposed that almost 300,000 light-years away lies a dwarf galaxy, now known as Galaxy X. This satellite of the Milky Way would be the first confirmed member of a hypothetical class of objects called dark galaxies, made up almost entirely of dark matter.

While the existence of dark galaxies would have major implications for most models of the universe, there was only indirect evidence of Galaxy X, and most of the astronomical community was not convinced. In 2015, however, the group claimed to have discovered four stars, sitting right in the middle of where they predicted the galaxy was. Suddenly, there was compelling reason to believe that this dark galaxy was real.

Things took an interesting turn later that year, however, and as of 2019, we no longer have direct observational evidence for any dark galaxy orbiting the Milky Way. Therefore, for the first time, I find myself writing about a really fascinating object that doesn’t exist. I hope you’ll stay with me as I explore what dark galaxies are, why Galaxy X was such an appealing idea, and an important lesson about studying variable stars.

We live in a galaxy of scallops

Figure 2, Levine et al. 2006. A map of the mean height of the Milky Way’s gas disk. The white areas are on either side of the galactic center, where we can’t get reliable distance measurements. Notice the asymmetry in the warping, as well as small-scale perturbations.

Since the 1950s, astronomers have known that the disk of the Milky Way isn’t flat, but warped. In some places, this warping is quite dramatic, with effects on the order of a kiloparsec. Much of our understanding of the shape of the disk comes from observations of neutral hydrogen (HI) using the 21-cm line (see e.g. Levine et al. 2006). In addition to this warping, there are also localized perturbations in some regions, sometimes called scallops. It’s possible that intergalactic magnetic fields or the intergalactic medium cause these disturbances, but another attractive option involves tidal forces from satellite galaxies.

One group at Berkeley (Chakrabarti et al. 2009) was interested in studying the sort of satellite that would be needed to explain some of the small-scale warping features. They modeled a Milky Way-like galaxy with an exponential disk of gas and stars, as well as a dark matter halo. They added in a smaller dark matter lump — a subhalo — and sent it on a parabolic orbit. It turned out that, assuming a subhalo mass about 1% of the Milky Way and a pericentric distance of 5 kpc, they could reproduce the HI scallops fairly well.

Figure 1, Chakrabarti & Blitz 2009. This simulation, designated 100E0R5, shows surface density perturbations during and after the passage of a dark matter subhalo with a pericentric distance of 5 kpc. The satellite makes its closest approach at 0.299 Gyr, in the second frame, and the present day is at 0.600 Gyr.

In particular, the effects of the subhalo should extend well after the close approach. Running the simulations forward in time showed that the simulated disturbance should agree with what we see today if the perturber was now about 90 kpc away. The precise location ruled out both the Large Magellanic Cloud and the Sagittarius Dwarf Spheroidal Galaxy unless previous calculations of their orbits are inaccurate, and as these are the only satellites large enough, there must be a dim satellite galaxy lurking somewhere: a dark galaxy.

Why are dark galaxies so dark?

One of the pillars of contemporary cosmology is the ΛCDM model of the universe. It suggests that in addition to the regular matter all around us, the cosmos is dominated by dark energy (Λ) and cold dark matter (CDM). The ΛCDM model has been extraordinarily successful in explaining phenomena such as the expansion of the universe and the rotation curves of galaxies, but it isn’t perfect.

Figure 2, Klypin et al. 1999. A frame from a simulation of dark matter in a galaxy group like the Local Group, dominated by two massive halos — the Andromeda galaxy and the Milky Way.

One of its main pitfalls is something called the missing satellites problem, which became apparent in the 1990s. Simulations of galaxy groups like our own (see Klypin et al. 1999) predicted the formation of satellite galaxies like the Magellanic Clouds, as plenty of dark matter subhalos. While we’ve found many satellite galaxies in the Local Group and beyond, simulations predict there should be many more — sometimes by an order of magnitude.

A Green Bank Telescope image of Smith’s Cloud, a high-velocity cloud colliding with the Milky Way. Image credit: Bill Saxton, NRAO/AUI/NSF, under the Creative Commons Attribution 3.0 Unported license.

Several solutions have been put forward. For instance, it’s possible that some satellite galaxies are torn apart by tidal forces, transformed into high-velocity clouds that repeatedly collide with the galactic disk. Another explanation is that the missing subhalos still exist, but only as so-called dark galaxies, objects dominated by dark matter and dim gas. Their star formation would have been quenched, either by supernovae winds that expelled gas into intergalactic space, or by ionizing radiation that prevented molecular clouds from collapsing in the first place. Either scenario should have formed small populations of stars, but not in significant quantities, and dark matter should still dominate over gas, dust, and baryonic matter in general.

Searching for dark galaxies is not an easy task. Astronomers can try to detect them indirectly, by looking for gravitational lensing or perturbations on other galaxies, but these effects aren’t likely to be obvious. Fortunately, dark galaxies aren’t purely dark matter, and by observing gas clouds or the few lonely stars they host, astronomers can try to find these dim objects. Searches have yielded a few unconfirmed candidates in recent years, notably including Dragonfly 44, in the Coma Cluster.

Figure 1, van Dokkum et al. 2016. Dragonfly 44, a candidate dark galaxy, is very faint, even in these images combining g- and i- band data.

Galaxy X is supposedly a mere 90 kpc away, which makes it a prime target for this kind of technique. Even if its mass is only 1% that of the Milky Way, and even if most of that mass is in the form of dark matter, there should still be some stars scattered here and there. All that remains is to find them. Against the odds, after years of modeling, Chakrabarti, Blitz, and collaborators did just that, claiming to have discovered four variable stars where they thought their dark galaxy should be.

But there was a catch.

How they (almost) found it

Paradoxically, some of the best objects for measuring distance in the universe are those that change in brightness: Cepheid variables, a type of variable star that swells and contracts over timescales of days or weeks. Their periods are directly related to their luminosities, so if you can measure a Cepheid’s period and its apparent magnitude, you can figure out how far away it is. Cepheids are incredibly handy in extragalactic astronomy, and have been used to prove that the universe is expanding and that the Andromeda galaxy is far from the Milky Way.

The Galaxy X group (Chakrabarti et al. 2015) decided to look for Cepheids at places in the sky close to the galactic disk. They used Ks-band infrared data from the European Southern Observatory’s VISTA Variables of the Via Lactea (VVV) survey to find red clump stars, hot red giants. One tile of images showed not one but four Cepheid variables clustered within 1 square degree — an unlikely change alignment. With periods ranging from 3.4 to 13.9 days, the group was able to calculate distance of 92, 100, 73 and 91 kpc — roughly the distance of their hypothesized dark galaxy.

Figure 2, Chakrabarti et al. 2015. Ks-band light curves of the four variable stars the ground observed. Note that the data points only last for about a single period; they’re simply plotted multiple times. Additionally, the fourth star’s curve isn’t very smooth or sinusoidal.

It would be highly unusual to see a single lone Cepheid variable this far from the galactic disk, let alone four. It was also unlikely that the stars were tidal debris from, say, the Sagittarius Dwarf Spheroidal Galaxy or the Large Magellanic Cloud, because while these objects interact with the Milky Way quite dramatically, they aren’t close enough in the sky. If the data was correct, then even without spectroscopic measurements, the stars might be part of a new galaxy.

This could be an enormous step towards confirming the existence of Galaxy X. The problem was, the group was relying only on a single set of infrared observations, and only over short timescales, possibly too short to be reliable. What if the luminosity variations were only a temporary phenomenon — and what if the stars weren’t Cepheids at all?

Those standard candles aren’t so standard

Later in 2015, a different group of astronomers (Pietrukowicz et al. 2015) cast doubt on the distance measurements by claiming that the stars were not, in fact, classical Cepheids after all. They turned to I-band observations from the OGLE Galaxy Variability Survey (OGLE GVS). OGLE GVS captured three of the four stars, designated S1, S2 and S3. After supplementing the data with additional I-band imaging of S2 and S3, the team came to a different conclusion: None of the stars were Cepheids.

Figure 2, Pietrukowicz et al. 2015. New I-band light curves for the first three stars show no periodicity for two candidates and un-Cepheid-like variations in the third.

S1 and S2 showed no significant variation whatsoever, and while S3 did appear to oscillate in brightness in a period of 5.695 or 11.39 days, the amplitude of the oscillations was larger than had those measured by the first group. Pietrukowicz et al. argued that S3 is actually a different kind of star: an RS Canum Venaticorum variable, a binary star in which one component displays prominent starspots that change over time as the star rotates. A number of confirmed RS Cvn variables have similar light curves, and the group showed that they can mimic Cepheid pulsations. The amplitude and period of the S3’s variations fit perfectly.

The starspot explanation for S3 was bolstered by the fact that its I-band light curve didn’t look much like a Cepheid’s — a discrepancy also displayed by the original S4 observations. In short, the group concluded, the four stars were almost certainly not Cepheid variables, and therefore the measurements of the dark galaxy’s distance should be thrown out entirely.

Figure 7, Pietrukowicz et al. 2015. Even the amplitudes of S3’s variations don’t match those of Cepheids seen in the Milky Way or the Magellanic Clouds, and its period is suspiciously long.

The new paper nullified the only supposed direct observations of Galaxy X. While this wasn’t proof that the object doesn’t exist, it was a serious blow to the theory — which was supported now only by the earlier simulations. Even these were not necessarily convincing, because they were only one of a number of potential explanations for the warping of the Milky Way’s disk.

Since 2015, no group has produced observational evidence of the putative dark galaxy. Barring the discovery of any objects at that distance and galactic longitude, it seems unlikely that it exists at all. Astronomers have plenty of other dark galaxy candidates to study, of course — Dragonfly 44, the HI region VIRGOHI21, and several others — but none have yielded clinching evidence. For now, dark galaxies remain hypothetical, important but so far undetected denizens of the universe.