Check Out My Kinky Curves

As usual, I'm following up my latest paper with a human-readable blog post. In this one I examine yet again why some clouds of gas in the Virgo cluster might be starless "dark galaxies" and how we can distinguish them from gas clouds ripped out of normal galaxies during close encounters. I also explore just why it's seriously crazily unlikely that some clouds are just interaction debris, with the help of several dogs and a lot of long kinky streams.


What are these "galaxy" thingies, anyway ?

Galaxies come in all shapes and sizes. Some are spectacular....

...while others are deadly dull.

Both of these images mainly show you the stars. You can also see some of their gas and dust, but not all, because gas and dust shine at different wavelengths than you can see with an ordinary telescope. What you can't see at all is the dark matter. The reason we think dark matter is there is from looking at the motions of the stars and gas - they're going too fast to be held together by their own mass, so there must* be something else holding them together that we can't see.

* I'm simplifying massively here, in an effort to keep this post shorter than War & Peace.

From the measurements, dark matter typically makes up 90% of the mass of a galaxy. There's far more dark matter than normal matter in the Universe. To the dark matter, stars are unimportant. If every star in the galaxy exploded, the dark matter cloud (or "halo") holding them in their orbits wouldn't notice. So could there be "dark galaxies" in the Universe made entirely of dark matter ?

An old image of mine, symbolically depicting a dark galaxy against a galaxy cluster full of hot gas. In fact the standard models says that dark galaxies should be smooth, boring spheroids without nice spiral arms like in a proper galaxy.

Well, maybe. If they did exist they'd solve a major problem in astrophysics : cosmological models get the structure of the Universe right, but they predict around ten times as many galaxies as observed (most of these "missing galaxies" should be small companions of larger galaxies). If most of these didn't contain stars that would solve the problem pretty neatly.

Real dark galaxies wouldn't need some terrifying stellar Armageddon to explain why they're dark. It could be that some dark halos just never accumulated enough gas to from any stars (or at least too few for us to detect), or the first supernovae may have blasted all the gas out and prevented further star formation. So there are at least vaguely-plausible reasons why these things both should and could exist.

But do they ? Perhaps. If they contain no gas and stars at all then they'd be bloody hard to find, but it's possible they have just a little bit of gas - enough for us to detect, but not so much that they'd form (m)any stars. They'd be discs of gas rotating just like normal galaxies, but optically dark. The only way to see them would be with a radio telescope, which can detect the gas from its own emission.

It never hurts to remind people that a) radio telescopes make images, they generally do not "listen" to the sky; b) the sky would look completely different if we could see the neutral hydrogen gas in our Galaxy, as in the above example.

Do dark galaxies really exist or are you just making this up because galaxies without stars is just a bloody daft idea

Actually, over the years quite a few candidate dark galaxies have been detected. The ones I'm most interested in are - because of my massive ego, obviously - the ones I found in my thesis project, a survey of the Virgo cluster. They seem to fit the bill pretty well : they (apparently) rotate as quickly as massive galaxies (about 100 km/s) but they have just a little bit of gas and no stars whatsoever - nothing else. The only way they could avoid flying apart is if they were embedded in dark matter halos, because their detectable gas mass is very small while their rotation speed is very fast. I'll call them the AGES clouds from here on in, since they were found with the Arecibo Galaxy Environment Survey.

So what's the problem ? If they were far away from any other galaxies, it would be very tempting to declare Game Over - dark galaxies found, cosmology problem solved, tea and biscuits for all, huzzah ! But they're not so isolated. A galaxy cluster is a realm of chaos, with thousands of galaxies flying past each other like bees on crack. These encounters can rip gas out of galaxies and leave bits of hydrogen fluff lying around all over the place.

Dog owners will know what I'm talking about.

It's possible that some or all of these dark hydrogen clouds are just "tidal debris" from galaxy interactions, with the messy encounters creating the illusion of rotation. We don't really measure rotation directly, just how fast the material is moving along our line of sight. Unfortunately our observations aren't very high resolution - we only get one pixel per cloud*. So we don't know how big the gas clouds are either, just their maximum size.

*When we have higher resolution observations of normal galaxies, we invariably find that this "velocity width" measurement really does correspond to rotation.

Previous papers have claimed that it's entirely possible to produce things that look like galaxies but are actually nothing of the sort. Instead of stable, long-lived rotating objects, they're actually short-lived objects in the process of disintegrating and we've just happened to catch them at a bad moment right before they fly apart.

Those results seem to hold pretty well for large objects. Unfortunately a lot of people seem to have read the older results and decided that they can also explain smaller ones. Regular readers will remember that we've previously found pretty strong evidence against that. It's easy to get a large change of velocity (that is, to fake the appearance of rotation) across a large feature, but much harder to get the same change across smaller ones. Specifically, it's damn near impossible to make objects as small as the AGES Virgo clouds with velocity widths anywhere near as high as their measured 100-200 km/s.


Are you really, really sure these clouds aren't just tidal debris ?

Our previous simulations dropped a long gas stream into a model of a galaxy cluster so we could watch the luckless thing get harassed into itty-bitty pieces by the cluster's gravity. This time we've gone one better and dropped a simulated galaxy into the cluster, so we can watch the stream's formation as well as what happens to the parent galaxy. Much more realistic (but technically harder to do and slower to compute, which is why we didn't do it originally).

Just like last time, we dropped our galaxy on a whole bunch of different trajectories (26) to make sure we didn't just happen to select one unusual path where the galaxy got obliterated or something daft like that. But we didn't just pick any-old galaxy to drop through the cluster, oh no. The target galaxy was modelled after this rather photogenic spiral :

M99, a.k.a. NGC 4254 (do forgive me if I switch between the names), as seen in the Sloan Digital Sky Survey.

And you might well think, "that galaxy is a bit funny-looking, why did you choose that one ?". And you'd be right - that prominent spiral arm is rather unusual. It's an even more peculiar galaxy when you look at its gas. In fact, M99 is something of a celebrity in dark galaxy circles because there's a long gas stream coming off it, which contains the infamous VIRGOHI21 :

Comparison of the optical and gas data. If you look closely, you can see hints of a longer gas stream above VIRGOHI21.

VIRGOHI21 is the densest part of the gas stream, and it's broadly similar to the other Virgo clouds (about the same gas mass and "rotation" speed) except for that long stream. So you might well now be convinced that NGC 4254 is a very weird galaxy indeed. Well, it is and it isn't. Its stars have a bit of a funny shape and it's got this weird gas stream. Even ignoring the gas stream, its gas disc is a bit lop-sided too, and it's very gas rich : in fact it has just about as much gas as a galaxy this size is ever likely to have.

And yet... it's not that odd. The overall distribution of stars is quite normal, as is the gas in the disc. It's a galaxy that got up on the wrong side of bed one morning and forgot to shower, not a total monstrous freak of nature that should have been killed at birth. Crucially, since it's got an awful lot of gas in the disc (and since that gas disc looks more-or-less normal), if you didn't know about the gas stream you'd never suspect it once had even more gas than it does now. If anything you'd wonder if maybe some interaction had given it extra gas, not ripped some of its gas away from it.

So there are two natural but mutually exclusive interpretations of NGC 4254 : it's pretty normal if you only consider its overall gas and star contents within the disc, or it's really weird if you look at its precise stellar distribution and the long gas stream. That's an interesting circle to square.

All this makes NGC 4254 a great target galaxy, which is why earlier studies used it as a target too. Our model starts off with an idealised version : we create smooth discs of gas and stars which have the same overall distribution, but don't reproduce the one-armed spiral. We also tried varying the gas content and distribution, as well as the mass of the dark matter since that's not so well-determined from the observations. Then we drop this into our galaxy cluster and see if the interactions reproduce any of those peculiar features, or things resembling the isolated clouds. Our cluster includes 400 other galaxies but only as simple point-masses : we don't model their gas and stars (that's far too computationally demanding), just their gravitational effects on the target.

So using M99 as our target may seem odd initially, but actually it lets us tackle several questions all at once :
1) Can we produce isolated gas clouds that look like dark galaxies by harassing a fairly normal spiral galaxy ?
2) What about fake dark galaxies in streams like VIRGOHI21 and the other strange features specific to the M99 system ?
3) Do the properties of the spiral galaxy make a big difference ?

Feel free to skip the next section if you want to get straight to the answers. Keep reading if it want to put these results in a bit more context.


Interlude : The Unnecessary Prequel

Time for a little backstory. One thing the referee and a slightly irate co-author made me tone down in the published paper was the criticism of the previous models. Originally the paper went into great detail about this, and while this was a bit excessive, I still would have preferred to state a few things more explicitly. Well, now I can ! And I will. Watch me.

Until we started re-investigating this, there were basically two explanations for the VIRGOHI21 system :
1) A simulation of Davies 2008 showed that if VIRGOHI21 was a giant dark galaxy, it could have created the stream largely from its own gas as it flew past M99. This could also explain the one prominent spiral arm. Unfortunately this model put VIRGOHI21 very much at the end of the gas stream, not in the middle as is actually the case. Bugger.
2) A simulation by Duc & Bournaud 2008 in which a normal galaxy tears gas out of M99, and VIRGOHI21 is just a kink in the stream rather than a dark galaxy. The "kink" is in the sense that the velocity of the gas changes very sharply as you go from the stream to VIRGOHI21 itself, rather than the stream suddenly changing direction.

But there are several problems with this second (much more popular) option. As mentioned, the gas content of M99 is pretty high, and its distribution is quite normal - which to my mind is strong evidence that the gas didn't come from M99, even without knowing what the true origin was. We can even quantify this, and although there isn't all that much gas in the stream, there's enough that if it had come from the spiral galaxy that would have made it truly exceptionally gas rich.

And yet long gas streams are very rare. So it's still entirely sensible to postulate that maybe this is indeed what happened : an unusually gas-rich galaxy had an encounter which tore off the gas into this long stream. An unusual explanation for an unusual object is not a daft idea by any means.

But VIRGOHI21 isn't just part of the stream which is a little bit denser than the rest : not only is it quite a lot denser, the velocity structure makes a very sharp "kink" - a sudden change, not a smooth one. Despite claims to the contrary, previous models actually have great difficulty reproducing this feature (more on that in a minute). They can make the long gas stream, sure, but not the feature that marked it out as weird in the first place. Long streams with smooth velocity changes are completely normal and therefore boring.

What's even worse is that unlike this dog they're not even cute.

And the Duc model had other debatable points - nothing outrageous, but things that would be cause for concern. Their model of M99 had it being not only exceptionally gas rich (fair enough if you think the gas really did originate in the galaxy) but with a really weird gas distribution that's not at all typical of a spiral galaxy. They also gave it the absolute minimum amount of dark matter conceivable based on the observations of how fast the galaxy spins.

It's perfectly possible the target galaxy was weird. But this model has a weird galaxy also having a weird encounter, which makes it very unlikely indeed. A weird explanation for a weird object is fine... but don't forget, we also have those quite similar clouds from AGES. So, arguably, VIRGOHI21 isn't that unusual after all. And the conditions of the target galaxy in the Duc model made it so that it was as vulnerable to gas removal as possible - yet still the model did not reproduce the distinctive velocity kink.

Given all this, it seemed to be a good idea to re-visit the Duc model. Not precisely, because setting up a situation which could give a really good match to the M99/VIRGOHI21 system (as Duc did) is tricky and really another project. Instead, we wanted to see if features like that velocity kink (and the AGES clouds) would be more common using a galaxy like Duc's than our preferred "normal" version of M99.

Here I have to say that the referee, although otherwise a level-headed individual, did something I found quite irritating. In an earlier draft I had a (too) long paragraph quoting papers where people had seen the Duc results and decided (in no uncertain terms) that that meant, decisively, that all dark galaxy candidates could be explained as tidal debris. I pointed out in the paper that the success of one model does not preclude the success of others; the referee said this shouldn't be mentioned since any scientists should know this - despite the evidence of the papers where this was clearly not the case at all. Then in a later draft I pointed out the Davies model, which got the odd response that these new results "convincingly exclude" that hypothesis. That's a bit bizarre, because they most certainly don't do anything of the sort ! We didn't examine the Davies model in any way, so how can these results possible have any bearing on them ? Answer : they can't.

Without further ado, then, here's what the new results actually say.


So what happened ?

I expect you'd like to actually see some of the simulations, wouldn't you ? Well, here you go !

Each postage-stamp follows the galaxy through the cluster. You can't see the harassing point-mass galaxies because those are very hard to visualise.

This isn't all of them, but visually they all look something like these. Have a look in the appendix if you want to see every one of the little buggers, but it's when you run the numbers you find out the interesting stuff.


1) Isolated clouds

The results were almost identical to our previous simulations. Isolated clouds up to the size of those seen in Virgo are really common in our simulations, but only at low velocity widths (< 50 km/s). They're pretty rare at higher widths (50 - 100 km/s) and utterly negligible at the high widths we're interested in (the real clouds are ~150 km/s). Harassment can certainly tear up a stream into itty-bitty pieces, but it does a lousy job of making it look like the ittiest-bittiest pieces are rotating. Works quite well for larger pieces, mind you.

2) Fake dark galaxies in streams

Maybe !

In fact our simulations reproduced the kinky velocity curves much better than the Duc model. The figure below compares the real VIRGOHI21, the Duc model, and ours.

Note that the vertical axis shows velocity. Left : the real VIRGOHI21. Middle : the simulation from Duc & Bournaud 2008. Right. Simulation from the latest paper.

Ours is clearly much kinkier than Duc's, though it's easy to understand why people might look at Duc's result and say, "sure, it's not quite kinky enough, but it's pretty close". I was expecting to have to come up with some way to measure exactly how kinky the streams were - thereby inventing a "kinkiness" parameter, but fortunately for me (and unfortunately for those with a sense of humour) this turned out to be unnecessary. Really kinky streams like the one on the right turned out to be so common there's not really any point in measuring them.

In some ways it's easy to see why our streams are so kinky (joke over, I promise) : we have 400 interacting galaxies, Duc had just two. But does this mean we can now explain VIRGOHI21 ? The referee thought so, despite my abject protestations. In fact I don't think we can. Our simulations were certainly kinky enough, but that's about it. They didn't reproduce that one prominent spiral arm, or the lop-sided gas disc, and again that model would require the galaxy to be really exceptionally gas rich. And the disparity between the gas and the stars, with that prominent spiral arm seen only in the distribution of stars and not the gas... well that's just odd.

OK, maybe that was the case. But the sharp kinks tend to be quite short-lived features, so what are the chances that we're seeing an intrinsically unusual galaxy during a very unusual phase of its evolution ? I don't know, but they're not high.


3) Does it matter what the galaxy was like ?

Sort of. OK, a lot. Well... not really. It depends what you're interested in.

We used three different initial galaxies : one is very similar to Duc's (lots of very extended gas and not so much dark matter), one is what we believe is the best match to the observations of M99, and the third is as massive as M99 can possibly be given the observational limits.

Using our standard model - which we believe is closest to the real M99 - we got a gas disc which remained pretty similar to the observations but sometimes had long gas streams pulled out of it. The gas distribution remained very similar to what it was like at the start, i.e. just like the observations.

Using our Duc-like model, which had less dark matter but more gas that was more extended, we got rather more gas stripping, as we'd expect. It tended to lose about the right amount of gas, so by the end the galaxy had about the same gas content as the real M99. But the distribution of gas was all wrong, and never resembled what we see in reality.

Both the standard and Duc-like models gave similar results for fake dark galaxies : kinky streams are very common, but isolated AGES-like clouds are rarer than an albino dodo.

I.e. pretty frickin' rare.

Our most massive model was so massive it basically sat there like a lemon and didn't do anything. Sending it flying through 400 galaxies was like throwing peanuts at a rhinoceros.


Conclusions

More example simulations. Here the black shows the gas. Red highlights gas that could be detected with AGES (the galaxy is the blob in the middle, which always remains detectable). Green shows isolated clouds, similar to the ones found with AGES but usually with much lower velocity widths.

Some dark hydrogen clouds are almost certainly just bits of tidal fluff pulled out of galaxies - at least according to our simulations. Little clouds of gas with low velocity widths are as irrepressible as cockroaches and nearly as common. I suppose it's not impossible that some of them are actually small dark galaxies, but there's no reason to invoke that more radical interpretation. We know galaxies have gas, we know the gas can be removed, we know it can form clouds with certain properties. In contrast the dark galaxy hypothesis has far less evidence to back it up in this case.

We can also say that some features in hydrogen gas streams could well be just due to galaxy interactions - the appearance of rotation can be faked. But does this mean that VIRGOHI21 is itself now definitively not a dark galaxy after all ? Not by a long shot. The "kinky" interpretation doesn't explain all the features of the system, and the alternative hypothesis that it's a dark galaxy hasn't been explored in nearly as much detail. So it's still entirely possible that might do a better job of explaining where all that extra gas came from and why M99 is generally funny-looking. The success of one model does not preclude the success of another !

Likewise, while some clouds are almost certainly tidal debris, we can be equally confident that some are not. It's hardly a revelation that some clouds are debris - it would be positively bonkers to suggest otherwise - but what's more interesting is we can quantify which ones are probably debris and for which ones this explanation looks a bit silly. High velocity width features can be easily formed in streams but not in isolated clouds.

We can also answer why this is. Our simulations showed that clouds are always born in long streams. Small gas clouds are never torn out of galaxies directly. So the only way you can make a small isolated cloud is by ripping up a stream. But... high velocity width features are intrinsically more difficult to detect, so if you can detect something that's ~200 km/s wide you'll also be able to detect the lower-width rest of the stream as well, almost by definition. And high-width features are also the shortest lived, since they're flying apart. To disperse the rest of the stream while keeping the highest-width features as isolated clouds is bloody difficult : the dark galaxy hypothesis seems far more likely. And we already showed that dark galaxies could reproduce the observations very well in these cases.

The above points might make you question why we needed a simulation for this. In principle, yes, we could have deduced them from pure theory. But it's very much easier when you can actually watch it happen. Moreover, the model lets us quantify things far more precisely - especially the rate at which the clouds are produced, which can't be predicted without a simulation.

What's next ? Good news, everyone ! We have another 200 hours of Arecibo time awarded to do another survey area in Vrigo so with any luck we'll find more clouds. That will give us much better information about where they are in the cluster, if they're generally near to other galaxies and if their velocity widths are always so high. And we have 16 hours of time on the VLA to get better resolution observations, so we'll have a very much better idea of whether these things are rotating or not. Lastly, we have more simulations in preparation to test even more hypothesis about the origins and evolution of the clouds. It may take a while, but we'll get to the bottom of this eventually...