The week of June 5, 2017, we held a workshop on dwarf galaxies and the dark matter problem. The workshop was attended by many leaders in the field – giants of dwarf galaxy research. It was held on the campus of Case Western Reserve University and supported by the John Templeton Foundation. It resulted in many fascinating discussions which I can’t possibly begin to share in full here, but I’ll say a few words.

Dwarf galaxies are among the most dark matter dominated objects in the universe. Or, stated more properly, they exhibit the largest mass discrepancies. This makes them great places to test theories of dark matter and modified gravity. By the end, we had come up with a few important tests for both ΛCDM and MOND. A few of these we managed to put on a white board. These are hardly a complete list, but provide a basis for discussion.

First, ΛCDM.

A few issues for ΛCDM identified during the workshop.

UFDs in field: Over the past few years, a number of extremely tiny dwarf galaxies have been identified as satellites of the Milky Way galaxy. These “ultrafaint dwarfs” are vaguely defined as being fainter than 100,000 solar luminosities, with the smallest examples having only a few hundred stars. This is absurdly small by galactic standards, having the stellar content of individual star clusters within the Milky Way. Indeed, it is not obvious to me that all of the ultrafaint dwarfs deserve to be recognized as dwarf galaxies, as some may merely be fragmentary portions of the Galactic stellar halo composed of stars coincident in phase space. Nevertheless, many may well be stellar systems external to the Milky Way that orbit it as dwarf satellites.

That multitudes of minuscule dark matter halos exist is a fundamental prediction of the ΛCDM cosmogony. These should often contain ultrafaint dwarf galaxies, and not only as satellites of giant galaxies like the Milky Way. Indeed, one expects to see many ultrafaints in the “field” beyond the orbital vicinity of the Milky Way where we have found them so far. These are predicted to exist in great numbers, and contain uniformly old stars. The “old stars” portion of the prediction stems from the reionization of the universe impeding star formation in the smallest dark matter halos. Upcoming surveys like LSST should provide a test of this prediction.

From an empirical perspective, I do expect that we will continue to discover galaxies of ever lower luminosity and surface brightness. In the field, I expect that these will be predominantly gas rich dwarfs like Leo P rather than gas-free, old stellar systems like the satellite ultrafaints. My expectation is an extrapolation of past experience, not a theory-specific prediction.

No Large Cores: Many of the simulators present at the workshop showed that if the energy released by supernovae was well directed, it could reshape the steep (‘cuspy’) interior density profiles of dark matter halos into something more like the shallow (‘cored’) interiors that are favored by data. I highlight the if because I remain skeptical that supernova energy couples as strongly as required and assumed (basically 100%). Even assuming favorable feedback, there seemed to be broad (in not unanimous) consensus among the simulators present that at sufficiently low masses, not enough stars would form to produce the requisite energy. Consequently, low mass halos should not have shallow cores, but instead retain their primordial density cusps. Hence clear measurement of a large core in a low mass dwarf galaxy (stellar mass < 1 million solar masses) would be a serious problem. Unfortunately, I’m not clear that we quantified “large,” but something more than a few hundred parsecs should qualify.

Radial Orbit for Crater 2: Several speakers highlighted the importance of the recently discovered dwarf satellite Crater 2. This object has a velocity dispersion that is unexpectedly low in ΛCDM, but was predicted by MOND. The “fix” in ΛCDM is to imagine that Crater 2 has suffered a large amount of tidal stripping by a close passage of the Milky Way. Hence it is predicted to be on a radial orbit (one that basically just plunges in and out). This can be tested by measuring the proper motion of its stars with Hubble Space Telescope, for which there exists a recently approved program.

DM Substructures: As noted above, there must exist numerous low mass dark matter halos in the cold dark matter cosmogony. These may be detected as substructure in the halos of larger galaxies by means of their gravitational lensing even if they do not contain dwarf galaxies. Basically, a lumpy dark matter halo bends light in subtly but detectably different ways from a smooth halo.

No Wide Binaries in UFDs: As a consequence of dynamical friction against the background dark matter, binary stars cannot remain at large separations over a Hubble time: their orbits should decay. In the absence of dark matter, this should not happen (it cannot if there is nowhere for the orbital energy to go, like into dark matter particles). Thus the detection of a population of widely separated binary stars would be problematic. Indeed, Pavel Kroupa argued that the apparent absence of strong dynamical friction already excludes particle dark matter as it is usually imagined.

Short dynamical times/common mergers: This is related to dynamical friction. In the hierarchical cosmogony of cold dark matter, mergers of halos (and the galaxies they contain) must be frequent and rapid. Dark matter halos are dynamically sticky, soaking up the orbital energy and angular momentum between colliding galaxies to allow them to stick and merge. Such mergers should go to completion on fairly short timescales (a mere few hundred million years).


A few distinctive predictions for MOND were also identified.


Tangential Orbit for Crater 2: In contrast to ΛCDM, we expect that the `feeble giant’ Crater 2 could not survive a close encounter with the Milky Way. Even at its rather large distance of 120 kpc from the Milky Way, it is so feeble that it is not immune from the external field of its giant host. Consequently, we expect that Crater 2 must be on a more nearly circular orbit, and not on a radial orbit as suggested in ΛCDM. The orbit does not need to be perfectly circular of course, but is should be more tangential than radial.

This provides a nice test that distinguishes between the two theories. Either the orbit of Crater 2 is more radial or more tangential. Bear in mind that Crater 2 already constitutes a problem for ΛCDM. What we’re discussing here is how to close what is basically a loophole whereby we can excuse an otherwise unanticipated result in ΛCDM.

EFE: The External Field Effect is a unique prediction of MOND that breaks the strong equivalence principle. There is already clear if tentative evidence for the EFE in the dwarf satellite galaxies around Andromeda. There is no equivalent to the EFE in ΛCDM.

I believe the question mark was added on the white board to permit the logical if unlikely possibility that one could write a MOND theory with an undetectably small EFE.

Position of UFDs on RAR: We chose to avoid making the radial acceleration relation (RAR) a focus of the meeting – there was quite enough to talk about as it was – but it certainly came up. The ultrafaint dwarfs sit “too high” on the RAR, an apparent problem for MOND. Indeed, when I first worked on this subject with Joe Wolf, I initially thought this was a fatal problem for MOND.

My initial thought was wrong. This is not a problem for MOND. The RAR applies to systems in dynamical equilibrium. There is a criterion in MOND to check whether this essential condition may be satisfied. Basically all of the ultrafaints flunk this test. There is no reason to think they are in dynamical equilibrium, so no reason to expect that they should be exactly on the RAR.

Some advocates of ΛCDM seemed to think this was a fudge, a lame excuse morally equivalent to the fudges made in ΛCDM that its critics complain about. This is a false equivalency that reminds me of this cartoon:

I dare ya to step over this line!

The ultrafaints are a handful of the least-well measured galaxies on the RAR. Before we obsess about these, it is necessary to provide a satisfactory explanation for the more numerous, much better measured galaxies that establish the RAR in the first place. MOND does this. ΛCDM does not. Holding one theory to account for the least reliable of measurements before holding another to account for everything up to that point is like, well, like the cartoon… I could put an NGC number to each of the lines Bugs draws in the sand.

Long dynamical times/less common mergers: Unlike ΛCDM, dynamical friction should be relatively ineffective in MOND. It lacks the large halos of dark matter that act as invisible catchers’ mitts to make galaxies stick and merge. Personally, I do not think this is a great test, because we are a long way from understanding dynamical friction in MOND.

Non-evolution with redshift: If the Baryonic Tully-Fisher relation and the RAR are indeed the consequence of MOND, then their form is fixed by the theory. Consequently, their slope shouldn’t evolve with time. Conceivably their normalization might (e.g., the value of a0 could in principle evolve). Some recent data for high redshift galaxies place constraints on such evolution, but reports on these data are greatly exaggerated.

These are just a few of the topics discussed at the workshop, and all of those are only a few of the issues that matter to the bigger picture. While the workshop was great in every respect, perhaps the best thing was that it got people from different fields/camps/perspectives talking. That is progress.

I am grateful for progress, but I must confess that to me it feels excruciatingly slow. Models of galaxy formation in the context of ΛCDM have made credible steps forward in addressing some of the phenomenological issues that concern me. Yet they still seem to me to be very far from where they need to be. In particular, there seems to be no engagement with the fundamental question I have posed here before, and that I posed at the beginning of the workshop: Why does MOND get any predictions right?

5 thoughts on “Dwarf Galaxies on the Shoulders of Giants

  1. Thank you for organizing this conference, Stacy, and for bringing together this particular group of scientists. It is a rare such event. Yes, the data on the relative galaxy distribution and velocities definitely rule out particle dark matter to exist, as elaborated in my presentation. The reason is that the huge and massive dark matter haloes (predicted to be present around every galaxy in the mathematically well defined LCDM theory) would lead to too much efficient dynamical friction for the observations to be consistent with theory. This point is elaborated (for those interested) here: .
    We thus already have an important and essentially direct test for the presence of particle dark matter, and it negates its presence. Note that past numerical experiments have all too often been faulty, as e.g. here:…518A..61C , the dark matter halos are assumed to be only as massive as the baryonic galaxies and they are too concentrated thus making-believe that galaxy encounters with dark matter look like what we see on the real sky. This is wrong!! These simulations artificially (compared to the LCDM theory) reduce dynamical friction by having too light and thus low-density dark matter halos. A researcher needs to be aware of the assumptions made, which too often invalidate the results in the sense the these results have nothing whatsoever to do with a theory, therewith being not useful.

    Xavier Hernandez has pointed out, in the meeting, that wide binaries cannot exist in the low mass satellite galaxies if they are immersed in their dark matter halos because dynamical friction which an individual star experiences as it moves through the particle dark matter halo in a low-mass dwarf galaxy will cause the binary orbit to shrink. He published a paper on this recently: Thus, dwarf satellite galaxies, those which are supposed to be very dark matter dominated, should not have wide binary systems.

    Dynamical friction is stronger in Milgromian gravitation (i.e. in MOND): That is, when a baryonic dwarf galaxy is moving through the stellar body of a much larger galaxy then it will suffer a stronger dynamical friction deceleration than in Newtonian gravitation, because Milgromian gravitation is effectively a stronger gravitational effect. This has been calculated by Nipoti & Binney here:
    HOWEVER: In Milgromian gravitation there is no huge and massive dark matter halo, and that is why in MOND galaxies can pass by each other, interact strongly, and separate again, perhaps becoming bound due to dissipation of kinetic energy due to the tidal effects as well as the dynamical friction on the baryonic components if the galaxies come sufficiently close. But they can nevertheless separate to large distances to then again fall towards each other and have another encounter and so on, with the relative orbit slowly decaying. In MOND galaxies can thus orbit and interact with each other many times and mergers are rare. This solves the galaxy-compact-group problem I identified in my talk. A recent paper related to this particular problem is found in the work of Wolfgang Oehm here:
    A direct comparison of galaxy encounters in the dark matter vs MOND theories has been computed to very high resolution including star formation by Florent Renaud et al.:


  2. Thank you very much for organizing this workshop. I followed it via twitter, where some of the attendees posted quite a few science snippets. I’ve recently become interested in tests of LCDM and MOND, so I found those twitter tidbits very informative. May I ask a couple of questions?

    – You mentioned that in LCDM very low mass dwarfs would retain their cores due to insufficient feedback. From an observational point of view, how well constrained are the cusps/cores in such low mass systems? Would the presence of, say, central clusters, rule out central cusps, given that said clusters would be presumably disrupted? (I nevertheless agree with you that even in more massive systems feedback can’t work *that* well to yield the incredibly tight scaling laws we measure).

    – You point out that since UFDs are not in equilibrium, their offset position in the RAR doesn’t necessarily falsify MOND. What’s the current status of tidal dwarf galaxies in this context? Are they in equilibrium? (I guess not?) Do they fall on the RAR?

    Thanks again. Now I’m off to digest all the pdfs of the talks!


  3. We usually observe cores in low mass systems, but the uncertainties in the lowest mass systems where the test is strongest remain an issue. Too many surviving clusters would be a problem; this is observed in Fornax, but Fornax is big enough to have formed a core.

    Tidal galaxies are, sadly, ambiguous at the moment. At first they seemed to favor MOND, then dark matter, and then we realized that the cases we’ve measured haven’t had time to complete a single rotation since they were formed. That’s a pretty strong hint that they haven’t settled into equilibrium. Perhaps there are older systems to be found and measured.


Comments are closed.