A Primer on Dark Matter Physics, Part 2

In part 1 of this article, I talked about the evidence for the existence of “dark matter” – that is, some kind of matter that we can’t see, and that likely is not made of the familiar protons, neutrons, and electrons, but that makes up the majority of the mass of the universe.  I then talked about WIMPs, a class of hypothetical particles that are a leading candidate for dark matter, and gave an overview of the three ways we can try to detect WIMPs – direct searches, indirect searches, and collider searches.  Today I want to delve more deeply into direct WIMP searches, and in particular, I’d like to talk about backgrounds, which are something that physicists spend a lot of time thinking (and worrying) about, but that doesn’t get much press.  I’ll then give a very quick whirlwind tour of some direct detection experiments, and then talk about the results of these searches.  Finally, I’ll wrap things up by talking a bit about the issue of dark matter vs. modified gravity.

If you haven’t read part 1, I suggest that you do so before continuing: https://the-avocado.org/2019/02/18/a-primer-on-dark-matter-physics-part-1/

Direct WIMP Searches and Backgrounds

Whereas indirect and collider searches are things that can be done opportunistically, using detectors that are primarily interested in other kinds of physics, direct searches typically require dedicated detectors, and I think it’s fair to say that most of the effort and money going into looking for dark matter is going into direct searches.

Remember, even though WIMPs mostly ignore other matter and pass right through the Earth without even noticing it, given enough of them, and enough time, a few will collide with atomic nuclei.  The idea with a direct search is that you have a target mass surrounded by instrumentation; if a WIMP collides with a nucleus in the target, the nucleus will recoil, depositing energy, and you use your instrumentation to detect that energy.

The thing is, though, that WIMPs are far from the only thing that can deposit energy in your detector.  For one thing, there’s radiation.  Pretty much everything is radioactive at some level – that is, every material has some unstable nuclei in it, which are liable to decay, emitting a photon, or an electron, or a neutron, or some other particle, and all of these things will deposit energy in your detector if they get in there.  There are also other sources of energetic particles all over the place.  The sun is, obviously, sending out a lot of photons, but it’s also emitting a lot of electrons, protons, and alpha particles in what’s known as the “solar wind”, and some of these particles reach the earth.  And then there are cosmic rays – very energetic particles (mostly protons and alphas) that probably mostly come from distant supernovae.  When these particles hit the Earth’s atmosphere, they produce showers of other particles.  If any of this stuff gets into your detector, the instrumentation will light up.  These detector events caused by things other than the particles you’re looking for (in this case, WIMPs) are called backgrounds, and in this kind of physics, dealing with those backgrounds is the whole game.

I like to think of the strategy for dealing with backgrounds as consisting of three lines of defense.  The first line, and the best option, is to eliminate the backgrounds; the second line is to identify remaining background events and distinguish them from signal events (i.e. the events you’re looking for), and the last defense, if there are still backgrounds that are indistinguishable from signal, is to predict how many such events you should see, so that if you see more than that, that excess is a signal.  Let’s talk about each of these in turn.

Eliminating backgrounds simply means making sure that those particles don’t get into your detector.  A big part of this is choosing the right components for your detector – you want to build it out of low-radioactivity materials to minimize the number of radioactive decays that you will see in it.  For everything from the nuts and bolts that hold it together to the solder on the electrical components, you’ll want to find out the exact composition and manufacturing process to be sure that you’re not unwittingly putting something radioactive in.  And it’s not just about choosing the right parts; it’s also about cleaning them thoroughly.  That stainless steel bolt might have very low radioactivity, but if there’s dirt in the threads, that could be a source of unwanted events.  There have been times where as a physicist I’ve spent weeks on end just cleaning parts.  Ideally, the experiment will be in a “clean room” or “clean lab”, with strict protocols about what can come in.1

So that takes care of the inside of your detector – but there are also, as mentioned, all kinds of particles accosting your detector from the outside as well.  These can be eliminated to some extent through shielding.  And at long last, this allows me to answer the question, “Why are many of these detectors located deep underground?”  It’s because all that earth above the detector acts as a shield against stuff like cosmic rays and solar wind.  Of course, there are also sources of radiation underground – the bedrock itself, for example, contains significant quantities of radioactive isotopes that will be shooting protons and neutrons at the detector.  So you’ll also want to surround the detector itself with some shielding material.  Water actually works very well for this, as it’s cheap and readily available.  That’s why many detectors will be placed inside a big water tank; the water is acting as a shield against radiation coming in from outside.

But no amount of cleaning or shielding will protect you completely.  The deeper underground you go, and the more shielding you put up, the more particles you’ll stop – but there will always be some high-energy particles that manage by chance to penetrate all your defenses.  And even if you build everything out of low-radioactivity components, that’s still not no-radioactivity.  Indeed, sometimes materials that you select for your target mass or instrumentation will inherently have some radioactive component.  For example, there are a lot of good reasons to use argon as a target mass – it has good scintillation properties, it’s non-reactive, and it’s fairly dense – but natural argon contains a radioactive isotope, argon-392.  Or, as another example, one of the devices typically used to instrument a detector is called a photomultiplier tube, or PMT.  It looks like a lightbulb, but it’s actually kind of the opposite – it’s a very sensitive light detector.  When a photon hits it, it generates an electrical signal that can be read out and saved as data on a computer.  But the outside of a PMT is made of glass, and glass contains radioactive isotopes.

A photomultiplier tube (PMT) in its support assembly. The glass in these PMTs is a major source of backgrounds in many experiments.

Since you can’t eliminate all of these backgrounds, the next best thing you can do is identify background events.  In other words, when your instrumentation lights up, you want to be able to look at that event and say with as much certainty as possible, “That was a background event,” or, “That was a signal event.”  You can do this because many backgrounds will produce a different signature in the detector than the signal.  Detectors vary considerably, but typically one is able to measure things like the amount of energy, the time over which it’s deposited, and the precise location in the detector where the collision happened.  So you might be able to say, “This event had more total energy than we’d expect if it were a WIMP event,” for instance.  Typically, in the analysis of the data, you’ll come up with a list of “cuts”; only events passing all of those cuts are considered signal events.  So, you might cut events where the total energy deposited is greater than some value, because you know WIMP events will tend to be lower in energy.  Or, if you know that you have sources of radiation on the outside edges of your detector (like, say, PMT glass), you might cut events that were located near the edges.  There’s a chance, of course, that you’ll throw away a small number of outlier signal events; there’s also a chance that some outlier background events will happen to pass your cuts.  So it’s important to run simulation and optimize those cuts.  Generally, in a dark matter search, where you’re looking for maybe a few signal events per year, but where there are often many background events per second, you’ll want the cuts to be quite strict – it’s better to have a small chance of throwing away a signal event or two than to have a decent chance of a handful of background events masquerading as signal.

In many detectors, there’s a special hardware system in place to identify and reject events whose origins are outside the detector.  This is a veto system, a region surrounding the main detector and separately instrumented – the main detector instrumentation can’t see the veto region, and the veto instrumentation can’t see the main detector.  If a WIMP collision occurs in the main detector, only the main instrumentation will pick it up – the veto won’t see it, because the WIMP, being a WIMP, will have gone right through it.  On the other hand, if, say, a neutron penetrates your shielding and enters the detector, it will probably deposit some energy in the veto region as it passes through as well.  So one of the cuts you use to select the signal events will be to reject any event where there was simultaneous activity in the inner detector and in the veto.

Sometimes, though, there are background events that cannot be distinguished from signal with any kind of reliability.  In this case, the best you can do is to predict those backgrounds as well as possible.  If you know exactly how many background events you expect to pass your cuts, then if you end up with more than that number passing cuts, the excess constitutes a signal.  Of course, you can’t ever predict this exactly, it’s all statistical; but if you expect 5 +/-2 background events to pass your cuts, and you see 20 events passing cuts, then that’s clearly a statistically significant excess.3

In a lot of areas of physics, this latter method of predicting the number of background events and then looking for any excess above that prediction is the norm, because in those areas we’re usually dealing with very large numbers of events, and the backgrounds are usually very well understand and can be predicted very precisely.  In dark matter physics, though, most searches at least aim to be “zero background” experiments, meaning that all backgrounds are either eliminated or identified and cut.

So, if we perform an experiment and see signal events that remain after we’ve cut out the background, that means we’ve discovered dark matter, right?  Not so fast.  It could be that we’ve discovered dark matter, but it could also be that there was some background that we didn’t anticipate, or a background that we did anticipate ended up looking different in our detector than we expected.  Maybe our simulations were not detailed enough; maybe some critical piece of equipment wasn’t cleaned properly; maybe any one of a thousand little things went wrong.  If an experiment were to see a potential dark matter signal, the first thing they’d do would be to try to come up with any plausible explanation for the signal other than dark matter.  They’d likely strive to improve their simulations, re-run their calibrations, and test all possible hypotheses to explain the signal.  If, after all that, the signal still holds up, they’d publish their results – but again, that would hardly be the end of the story.  Other detectors would search for the same signal, hopefully using as wide a variety of approaches as possible; if the signal can’t be reproduced in those experiments, then we can conclude that the first experiment likely did have some unknown background.  Only after the signal was confirmed independently, using different techniques, would we feel comfortable saying that we had probably discovered dark matter.  Indeed, this is not an esoteric point, but something that has already happened.  Several experiments, notably DAMA, CRESST, and CoGeNT, have seen signals that, to the best of those experiments’ discernment, to not look like any of the expected backgrounds.  However, in all these cases, other experiments have been unable to reproduce those signals.  It’s worth noting, also, that the signals seen by these experiments do not agree with each other, either, in terms of what kind of signal is seen.  The consensus, then, is that these experiments are not seeing dark matter, and work is ongoing to try to understand what the source of those signal events is.

A Quick Survey of Detectors

To repeat a bit, any detector looking for WIMPs must include two things: a target and instrumentation.  The target is the material that you’re looking for a WIMP to collide with a nucleus in.  Ideally, you want your target to be as large and as dense as possible – the more nuclei in the target, the better your chance of a WIMP colliding with one.  If a collision does happen, the nucleus will recoil, and it will deposit energy as it travels.  That energy could be deposited in many forms – light, ionization, heat, or even sound.  The instrumentation is whatever you use to detect that deposited energy.

Let’s take what I think is one of the simpler approaches, conceptually: a single-phase noble liquid detector.  In a typical example, the target is a volume of argon.  To increase the density of the target, liquid, rather than gaseous, argon is used – this means that the detector has to be kept at cryogenic temperatures, or the argon will boil.  If a nucleus recoils and travels through the argon, it will deposit some of its energy to electrons in the other argon atoms.  This will put the electrons in an unstable “excited state”.  Soon afterward, the electrons will revert to the stable “ground state” by emitting the excess energy in the form of a photon – that is, in the form of light.  This is called “scintillation”, and it’s something that lots of materials (not just argon) do, though things like the wavelength of the scintillation light and the time it takes for the light to be emitted vary.  Noble liquids like neon, argon, and xenon are a good choice for WIMP searches because of their scintillation properties, density, and chemical intertness.

So, the idea is that a WIMP collides with an argon nucleus, that nucleus recoils through the argon and deposits energy in the electrons of other argon atoms, and those electrons then give off light.  To detect that light, the instrumentation used is an array of photomultiplier tubes, or PMTs.  As I mentioned above, these are essentially just very sensitive light detectors; whenever a photon of the right wavelength hits one, it sends a signal up to your electronics, which you can record onto a hard drive.4

Left: The DEAP-3600 inner vessel looks like some kind of anemone. Right: The MiniCLEAN inner vessel has a tight squeeze. Both of these are single-phase liquid argon detectors.

In addition to producing scintillation light, the recoiling nucleus can also ionize electrons – that is, transfer enough energy to an electron that it breaks free from its atom.  Some detectors are designed to detect this ionization, either by itself or in addition to the scintillation.  To do this, they generally apply an electric field to the target volume.  Any ionized electrons, no longer bound to their atoms, will then be carried by this electric field toward one side of the detector, where instrumentation is deployed that allows those electrons to produce an electric signal.  If you take this approach, you’ll want to choose a target medium that is easily ionizable, and one where the ionized electrons won’t easily be recaptured by other atoms, but will be free to drift all the way to the readout area.  An example is the DMTPC experiment, which uses CF4 gas as the target.

The field cage of the DMTPC experiment creates an electric field that is used to collect ionized electrons.

Other detectors literally try to hear WIMP collisions.  Some of the energy of a recoiling nucleus will be deposited in the form of vibrations in the medium.  In many materials, that vibrational energy will be minimal, and will dissipate quickly; but if you use something like cryogenic germanium and silicon crystals, and instrument it with extremely sensitive microphones, then you can record the sound waves produced by a nuclear recoil.

The CRESST detector, which searches for WIMP collisions via temperature.

Yet another approach relies on the fact that some of a recoiling nucleus’s energy will be deposited as heat, raising the temperature of the medium slightly.  The CRESST experiment, for example, uses calcium tungstate crystals that are cooled to just above absolute zero.  Very sensitive thermometers can then be used to detect nuclear recoils by the tiny temperature increases they produce.

A somewhat different approach is represented by the PICO experiment (and its predecessors, PICASSO and COUPP).  These also rely on thermal excitations – that is, on heat being generated by the recoiling nucleus, but they use droplets of superheated liquid suspended in a gel as the target.  When a nuclear recoil occurs within one of these droplets, it heats the droplet enough to make it undergo a phase change – to boil, essentially, and expand into a gas bubble.  A very fast camera simply takes pictures of the target medium to look for such bubbles.

A background event in one of the PICO detectors. The bubbles are the locations where a particle, in this case a neutron, has collided with atomic nuclei.

Many detectors combine multiple approaches.  This can be very useful for identifying and rejecting backgrounds; a dark matter event might, for instance, produce a different combination of scintillation and ionization than a background event.  Thus, several noble liquid detectors look for scintillation light with PMTs but also apply an electric field to collect ionized electrons5.  Experiments like CDMS using cryogenic germanium crystals can detect both acoustic vibrations and ionization caused by nuclear collisions.  PICO, mentioned above, supplements its photographs of superheated bubbles with sensors that detect acoustic vibrations.  Each approach has its own advantages and disadvantages, and each comes with a host of technological challenges.  Sometimes, in working on these detectors, one feels like one is constantly inventing new and weird problems to solve, often ones that are ridiculously specific.6  But all of these technologies also tie back into each other in sometimes surprising ways, and even failed experiments are often useful in providing technical knowledge that can then be applied elsewhere.7

Results

We’ve been searching for dark matter for quite a while now.  So, what have we found?  Well, the short answer is that we haven’t found dark matter yet.  But, as I noted in part 1, that shouldn’t be interpreted as meaning that these experiments have been failures.  Instead, each search that doesn’t find WIMPs tells us more about what WIMPs aren’t and thus, by inference, what they could be.

Let’s be more quantitative about that.  If a WIMP exists, it will have some particular mass and will also have another property called a “cross section” for nuclear interactions.  The cross section can be thought of as a measure of how likely a particle is to interact with another particle.  (To see why it’s called a “cross section”, imagine the particle as a billiard ball; the bigger the ball, the more likely it is to hit another ball).  A given experiment will only be sensitive to WIMPs of certain masses and cross-sections.  If that search doesn’t see a signal, then we say that it “rules out” WIMPs with those properties.  Well, let’s be more precise – we say that it rules out WIMPs at those masses and cross-sections at some confidence level; in other words we have to consider the possibility that there are WIMPs at that mass and cross-section, but that through a statistical fluke none of them collided within the detector.  But this possibility can be reduced by running experiments for long periods of time and collecting a lot of statistics.

To visualize this, we typically make a plot of hypothetical WIMP cross-section vs. mass, and we can show on this plot which possible WIMPs have been ruled out by experiment8.  Here’s an example:

A plot of WIMP cross-section vs. mass, showing the regions excluded by direct detection experiments, as well as the expected sensitivities of some upcoming experiments.

The solid lines on this plot correspond to various WIMP searches, each of which ruled out WIMPs in the region above the curve.  The dashed lines show the expected sensitivity of several upcoming experiments – that is, if the WIMP mass and cross-section fall above that line, we expect that experiment to be able to detect it.  As we build bigger and more sophisticated detectors, we’re able to push the sensitivity to lower cross-sections, so we’ve gradually ruled out more and more of that space (that is, more and more possible WIMPs) over time.

Here, incidentally, you can also see the signals that I mentioned earlier seen by experiments such as DAMA and CRESST, shown as closed contours.  The blob labelled “DAMA”, for example, means that if the DAMA signal were interpreted as a WIMP, rather than as some unknown background, then the WIMP’s mass and cross-section would lie within those bounds.  As you can see, not only are these signals all in areas ruled out by other experiments, but they also don’t agree with one another, which is ample reason to strongly believe that they are spurious signals rather than real detection of dark matter.

As you can see from the plot, in addition to searching for lower cross-sections, we can also push to higher or lower WIMP mass, though the mass region that has been tested so far is the one that most theoretical predictions for WIMPs seem to lean toward.  If the dark matter particle is significantly less massive, it would be not a WIMP but something like a WISP (“Weakly Interacting Slim Particle”), a hidden photon, or an axion, which generally require a different approach to search for.

However, while there’s clearly still some space for a WIMP to be hiding, it’s fair to say that things are looking a bit less favorable for WIMPs than they did ten or fifteen years ago.  This is not just because direct searches haven’t found WIMPs, but also because indirect and collider searches have come up empty.  Moreover, we haven’t yet seen any evidence for supersymmetry at the LHC (the best particle collider out there at the moment).  Remember from part 1 that supersymmetry is an idea that comes up in several hypothetical extensions of the Standard Model, in which every particle we know has a “superpartner” of much higher mass.  The simplest supersymmetric models tend to predict that the lightest of those superpartners should have masses right around the top of the range that the LHC would be able to produce.  So the fact that we haven’t produced such particles yet doesn’t necessarily mean that supersymmetry is not true, but it is starting to look less likely.  And if supersymmetry isn’t true, then our prime theoretical candidates for WIMPs disappear.  Indeed, the sneutrino has now been ruled out as a WIMP in the simplest models, though there is still some room for the neutralino or the gravitino.  One of the reasons that WIMPs initially looked so promising is that simple supersymmetric models readily predicted a WIMP at just about the expected mass and cross-section; this was called the “WIMP miracle”.  But these simple models look like they’ve now been pretty well ruled out.  Indeed, just the other day I saw an article proclaiming that “the WIMP miracle is dead.”

So we’re in a very ambiguous position right now.  It’s as if we’re looking for a friend we thought would be at a party, and we’ve checked maybe half of the rooms without finding her, and we’ve asked a few people if they’ve seen her and they’ve said no.  She still might be there – I mean, she told us she’d be here, right? – but it’s starting to look like maybe we got the date or the address wrong, or maybe she was held up in traffic or something.

There are several WIMP searches that will come online over the next decade that will push our sensitivity still further, and while I’d love to say that they should clear things up, in truth if they fail to find WIMPs, things will still be somewhat ambiguous.  These experiments will soon start to face a new challenge: the coherent neutrino background.  Essentially, when we push things to a certain sensitivity, we are now looking for WIMPs with such a small “cross-section” that they look a lot like neutrinos.  This means that neutrino-nucleus collisions will start to become an irreducible background – one that can’t be eliminated or identified and cut.  This doesn’t mean that direct WIMP searches can’t probe beyond this level, but it does mean that they will no longer be “zero background” experiments.  The whole game will then be to very precisely predict the neutrino background and look for an excess of signal events above that background.  And it should be noted that this background is an issue only for direct searches.

Dark Matter vs. Modified Gravity

I want to finally come back to something I mentioned in part 1, which is also one of the most common questions I get about dark matter.  As I noted there, given our observations of galactic rotation curves, the Cosmic Microwave Background, and so on, one of two things must, logically, be true: either there is much more matter in the universe than what we can see, or our theory of gravity is wrong.  The first option is what leads to dark matter, but what about the second option?

I should note at this point that a lot of physicists take a fairly dim view of modified gravity theories to begin with, and I don’t think this is entirely fair.  The reason for this is that modified gravity is just one of those ideas that seems to attract a lot of crackpots.  It sometimes seems like everyone who has read a book or two on cosmology fancies himself a genius who will prove Einstein wrong and overthrow the scientific orthodoxy.  As anyone who works in physics knows, crackpot e-mails are a pretty common phenomenon, and one of the most popular types are those expounding some new theory of gravity (other popular topics are entropy and perpetual motion machines).

But it’s not fair to dismiss modified gravity out of hand.  My view is that even if you think dark matter is a far more likely explanation than modified gravity, you should still try to understand modified gravity theories; all possibilities should be explored.  There are, in fact, some serious physicists exploring this possibility and, at least in terms of alternatives to dark matter, the bigshot of this field has for some time been a theory called “MOdified Newtonian Dynamics”, or MOND.

MOND is a theory that mimics Newtonian gravity when the accelerations involved are large, but increasingly differs from it at smaller and smaller accelerations.  Because our direct tests of gravity involve fairly high accelerations, MOND reproduces these results well, but on the larger scale of galaxies its predictions diverge significantly from those of Newtonian gravity.9  And as it turns out (or rather, as MOND was constructed to do), if you plug the distribution of visible matter in a galaxy into MOND, the prediction it spits out for the velocities of the stars is more or less what we see.  In other words, if MOND is the correct theory of gravity, then the galactic rotation curves (and also the velocities of galaxies within galaxy clusters) make perfect sense without assuming any dark matter.

So this is pretty interesting, right?  Well, at first sight, yes it is, and honestly, fifteen years ago I would have said that MOND, which so easily reproduces the galactic rotation curves that we see, looked pretty good as an alternative to dark matter.  There is, prima facie, the problem that the density fluctuations in the CMB don’t really match what MOND would predict, at least not without some finagling – but it’s not hard to believe that such finagling could do the trick, and that a version of MOND could be produced that would get the density fluctuations right.

But there is one very big problem, not just for MOND but for modified gravity as an alternative to dark matter in general.  And I must admit that I have been holding back one major piece of evidence for the existence of dark matter, so that I could dramatically reveal it here.  And that is the Bullet Cluster.

“The Bullet Cluster” is actually the name loosely given to two clusters of galaxies that recently (in cosmological time) collided with each other10.  As I mentioned in part 1, galaxy clusters are groups of galaxies that are gravitationally bound together (our galaxy, the Milky Way, resides within a small cluster known as the Local Group).  The distances between galaxies, even within a cluster, are generally very large, so when two clusters collide, the individual galaxies themselves rarely hit each other.  But the clusters also contain a diffuse gas, and as the two gigantic clouds of gas collide, they slow each other down, and this gas then creates a small drag effect on the galaxies.  The net result is that the two clusters pass through each other, but are significantly slowed down by the collision.  Depending on the exact conditions, this could eventually lead the clusters to merge into a single cluster, or they could each go on their way, slowed down and somewhat deformed by their encounter.

At least, that’s what would happen to the normal matter in the clusters.  If the clusters are, in fact, each embedded within a halo of dark matter, then you’d expect that dark matter not to notice the collision at all; without even slowing down, the two dark matter halos would pass right through each other.  The result would be that the dark matter would race out ahead while the visible matter gets left slightly behind; the dark and visible matter would be somewhat separated from each other.

Using gravitational lensing, we can look at the Bullet Cluster and determine where most of its mass is.  Remember, gravitational lensing is a phenomenon where the force of gravity bends light waves.  By looking at objects behind the Bullet Cluster, and measuring how much the light from those objects has been distorted, we can make a sort of map – the places where that light is the most distorted are the places within the Cluster that have the greatest density.  And this is what we find:

The Bullet Cluster collision, showing X-ray emissions in color and a density map based on gravitational lensing in contour.

The colors in this picture are an x-ray photograph of the Bullet Cluster, which picks up not just the stars but also the diffuse gas.  In this view we see two bright blobs, which are the two clusters moving away from each other.  The contours drawn on the picture, on the other hand, are a density map constructed from gravitational lensing.  As you can see, the gravitational lensing map shows that the densest parts of the clusters are displaced from the visible matter.  This is exactly what we’d expect given the dark matter hypothesis – the visible parts of the clusters were slowed down by the collision, but the dark matter (which, remember, makes up most of the mass of the clusters) was not slowed down.

If, on the other hand, the MOND hypothesis were true, and there were no dark matter, then we would not expect to see the center of mass displaced from the visible matter the way that we do.  Remember, MOND doesn’t add any additional sources of gravity; it just proposes that the sources that do exist generate gravity differently.  MOND, or modified gravity theories in general, might predict that we’d see more, or different, gravitational lensing from the visible matter, but not that we’d see the gravitational lensing in a different location.

Now, some proponents of modified gravity have retorted that “it can be two things”, and proposed that maybe modified gravity is true and also that there’s some dark matter – and in particular, maybe if MOND is true, we can do with little enough dark matter that it could all be accounted for by protons and neutrons, rather than some unknown particle like a WIMP.  But if you ask me, it is a pretty blatant violation of Occam’s Razor to propose two extraordinary hypotheses when just one will suffice.  We should, of course, still keep an open mind about modified gravity; and it’s true that even if dark matter exists, it may also end up being the case that General Relativity needs some corrections.11  But it’s very hard for me to see how modified gravity can be rescued as an alternative to dark matter at this point – especially since, in the years since the first observations of the Bullet Cluster, the same effect has been observed in several other galaxy cluster collisions.  Any attempt to do away with dark matter by invoking modified gravity would have to explain why the centers of mass of these clusters look like they are displaced from the visible matter, and it’s just very difficult to see how they could do that.

Where does all of that leave us?  I’d say it leaves us with a little bit of a mystery.  We now have very convincing evidence that dark matter exists, and yet we’ve ruled out many of the more obvious candidates for that dark matter.  But there are still lots of possibilities.  It could still be a supersymmetric WIMP, even if some of the simpler versions of that hypothesis have been ruled out; it could be a WIMP in some other extension of the Standard Model; it could be an axion, or a hidden photon, or some other more exotic particle.  It should, perhaps, be noted that the dark matter searches we’ve done so far have been: a) searches for what we thought the most obvious candidates were and b) the searches that were easiest to do with our current technology.  These searches come nowhere near exhausting all the kinds of particles that dark matter could be, and there is still a lot to do.

If you’ve stayed with me to the end, thanks for reading, and I hope this was at least somewhat interesting.  It was fun to write!