Follow the reluctant adventures in the life of a Welsh astrophysicist sent around the world for some reason, wherein I photograph potatoes and destroy galaxies in the name of science. And don't forget about my website,

Monday, 20 June 2016

Into Darkness (I)

My latest paper - sixth as first author - is now online. It's been in draft for about 18 months (as long as two pregnancies, so you decide which was the better choice...), taken over 200 simulations, and the resulting monster is 27 pages long. What great quest could possibly have warranted such fanaticism ? Find out in this and the next blog post. Or if you just don't care that much but are vaguely curious, read the super-short version here. If you don't want the background information, you can do directly to part 2 here.

To start, we have to go much further back than the last 18 months.

Everyone Should Have Their Own Life-Defining GIF

Back in the heady carefree days of 2006, I was a fresh-faced* and fancy free** Masters student about to embark on a PhD course. The direction this took was largely thanks to this, the single most important gif of my life so far :

** Hapless.

I have, of course, explained such 3D data cubes many times before, so let's keep this simple. What you're looking at is a map of part of the Virgo galaxy cluster as seen through a radio telescope. Instead of looking at the stars, you're seeing the hydrogen gas. It's easier to explain in a 2D image. Let's compare what we'd see in "normal" visible light with these gas-sensitive radio wavelengths  :

The really bright big blob you see in the top gif spinning around is the gas in the spiral galaxy, NGC 4254. You can also see it has a long stream of gas, which apparently ends in a bit of rather denser gas that's the object known as VIRGOHI21. Pronounced, "Virgo H one twenty one". "HI" is really H Roman numeral 1, and is the symbol for neutral atomic hydrogen (as opposed to H2, which is molecular hydrogen, and HII, also pronounced "H two", which is ionised hydrogen). "21" is just the object's catalogue number in the survey.

If this is confusing, don't worry. Forget the molecular and ionised hydrogen, they're a bunch of smelly pants. We're only looking at HI . And the visible light from stars, but that's not so important either. The important point is that by measuring the gas, we can measure how fast everything's moving. If we're clever, we can use this to work out if things are rotating and the speed of rotation if they are.

The interesting thing about VIRGOHI21 is that it appears to be rotating. And if it's stable, just like ordinary galaxies it would also need a lot of dark matter to hold it together. So when I started my PhD, it was being widely touted as the first detection of a so-called "dark galaxy" - a dark matter cloud (or "halo" as they are usually known) containing a rotating disc of gas, but no stars.

Actually it wasn't the first time such a claim had been made at all, but I'll get back to that.

You might well be thinking, "Hang on a minute, you four-eyed beardy twerp. There's a great big complicated stream of gas coming out of NGC 4254, so how do you know this VIRGOHI21 thingy isn't just a kink in the stream ? Eh ? How do you know it's really a massive rotating disc ?". Well, we don't. To understand why we'd look at this object and say, "dark galaxy" and not "weird kink", we need a bit of backstory.

Galaxies Don't Need Stars, That's Just RACIST !

The theoretical basis for dark galaxies was that they could solve the "missing satellite" problem. Models predict about ten times more dwarf galaxies than we actually observe, so the idea was that maybe some galaxies just never have enough gas to form much (or any) stars. An elegant solution to a complicated problem that I'm grossly over-simplifying. It's one of the biggest problems in contemporary cosmology, so shouldn't be taken lightly. Any proposed solution has to be treated with strong skepticism*.

* And no, fellow Brits, I refuse point blank to spell it "scepticism". That would be pronounced like in septic, which would be unpleasant for all concerned.

The observational measurements just say that VIRGOHI21 could be a massive rotating disc. Or it could be a much less massive kink in the stream, which might have formed when another galaxy flew past and pulled the gas out of NGC 4254. The idea here is that it looks like it's rotating, but isn't really. But there was some pretty intriguing other evidence, at the time :
  • Unusually, NGC 4254 has one particularly prominent spiral arm. This could be the result of interacting with a massive object.
  • There's no obvious normal galaxy visible nearby that could be responsible for the prominent spiral arm or the gas stream.
  • The kink in the gas stream is very sharp compared to the rest of the stream. Why should only this one feature be so strange ? Few (if any) other streams show features like this.
If VIRGOHI21 really is rotating, it would be massive enough to explain the disturbance to NGC 4254. Galaxies this massive and dark would be controversial, but some models said they were possible. The very sharp kink in the velocity would be because VIRGOHI21 is a compact object, which is perfectly consistent with the observations. And based on what was known about galaxy interactions at the time, it was very difficult to see how an interacting galaxy could produce a stream with all these particular features.

My Masters project had been all about simulating the formation of a spiral galaxy without dark matter. Since this hadn't worked, my natural skepticism of dark matter had taken something of a blow. Not a mortal wound by any means, but enough to give pause for thought. In any case, VIRGOHI21 had turned up in a relatively low sensitivity survey of the cluster. Having access to much deeper AGES data (the Arecibo Galaxy Environment Survey, one of the most sensitive surveys of the cluster ever undertaken) as part of my PhD, the prospects for detecting other dark galaxy candidates looked good. Even if they turned out not to be real dark galaxies, it seemed pretty likely that we'd find lots of cool streams and other neat, weird objects.

It didn't really work out like that.

Where Are All These "Dark Galaxies" Then ?

The AGES data I was given charge of for my thesis consisted of two parts of the Virgo cluster. Well, it seemed like such an interesting place that surveying it twice was a sensible enough idea (observing the whole cluster would take much too long).

VC1 is the larger area in red while the smaller area is VC2.
VC2 contains pretty much bugger all. More accurately, it contains precisely bugger all that hadn't already been discovered. To be really accurate, the "bugger all"-ness is confined to new discoveries from the hydrogen data, but that's another story.

Fortunately, VC1 turned out to be a nice happy data cube full of lots and lots of interesting detections. But did it contain lots of hydrogen streams and dark galaxy candidates ? No, no it did not.

Whoa, steady on there, me ! It's true it didn't contain any streams. But the dark galaxy candidates statement is, like, totally way melodramatic. In fact we detected eight hydrogen clouds without any obvious galaxies present. Six of them seemed to be rotating very nearly as fast as VIRGOHI21. So why do I seem reluctant to say that we didn't find any dark galaxies ?

Regular readers will be screaming that this is the most boring visualisation I've ever done. Where are the glass cubes ? Where are the 3D movies ? Unfortunately, while HI is often stunningly beautiful, these particular detections are just too faint  and too small to show any details. 

There's No Way A Bunch Of Poxy Lines Are Going To Convince Me You've Solved The Missing Satellite Problem.

By this point the overall situation had changed. A rather convincing simulation had been published demonstrating that VIRGOHI21 could, in fact, just be the result of an ordinary encounter after all. You may well have been right to doubt my four-eyed beardy twerpiness proclaiming it to be a rotating disc, because this model said that it was indeed just a kink in the stream. It seemed that it's just an illusion of rotation and therefore totally uninteresting.

I'll save the details of that simulation for next time, though most people were rather swept away by the result. Cooler heads noted that the proposed model was far from perfect, but at the time, most people thought these flaws were minor details that needed ironing out - not gaping holes that meant the whole sorry thing should be burned and the ashes spat on and scattered to the four winds.

A perhaps even more serious problem was the extremely low numbers of these candidate dark galaxies. My supervisor's model had predicted that as many as 23% of our survey's detections should be dark. Even with these objects it was at most 1%, so that particular model was decisively ruled out. Which meant that dark galaxies didn't seem likely to explain the missing satellite problem, which was the reason they were proposed to exist in the first place.

There's a third reason. The Tully-Fisher relation is a very well-established correlation between how fast a galaxy is rotating and how much ordinary matter (gas and stars) it contains. In fact it's so well-established that if you wanted to define a galaxy, you could almost say, "anything that obeys the Tully Fisher relation". While even the faintest (but still optically visible) little smudges we'd found in the Virgo cluster with our survey seemed to obey this relation perfectly, these objects which were apparently totally dark do not.

The faster a galaxy appears to be rotating, the more massive it is. The "Tully-Fisher Relation" just refers to the precise numbers : knowing one, you can calculate the other. That's the red line in the picture. It works pretty well for regular galaxies, but doesn't work for the dark galaxy candidates.

And when I say faint smudges, I mean really faint stuff like this :

All of these pathetic little scraps of starlight sit on the TF relation just fine - even VC2_30 (bottom centre), which looks more like someone's sneezed on the lens than a proper galaxy. They have at least a hundred times less mass than giant objects, but they all obey the same relation. The fact that the truly dark objects don't follow this correlation, whereas the merely extremely dim ones do, marks them out as categorically different.

Taken together - the simulation showing VIRGOHI21 was tidal, the insufficient numbers of dark clouds, and the deviation from the Tully Fisher relation - the case against dark galaxies seemed pretty darn damning.

So Galaxies Have Stars After All Then. So Much For Galaxy Justice Warriors.

And yet... these clouds remained unexplained. Yes, there was one model showing how they might be formed - but only one. Public outreach articles will tell you that science values repetition. Well, sometimes this is if anything excessively true, but not always. In this case - for whatever reason - nobody saw fit to go back and re-examine the model and test it using other codes or explore its implications more fully. And while more clouds continued to be discovered from time to time, this one particular model was almost invariably cited to explain them.

Almost invariably. There were a few for which the basic idea of "tidal debris" (i.e. the gas was pulled out of a normal galaxy during an interaction) just didn't make any sense at all. I've previously described a couple of examples - the Smith Cloud and Keenan's Ring - in detail. At least one other example deserves a look in : HI1225+01, sometimes known as the Giovanelli & Haynes cloud.

Gas distribution shown by the white outline. I'd make a better image but I don't have the raw data to play with.
This enormous cloud is at least 650,000 light years long with a mass of around 4 billion Suns (the stupid poxy little clouds I found are at most 55,000 light years across and have masses no more than around 30 million Suns - more than ten times smaller and a hundred times less massive). True, HI1225+01 has a small galaxy at one end - but it's very small indeed. Far too small to be a credible source of all that gas. It only remains to choose which confused animal meme seems most appropriate.

This dog is exactly like the gas in HI1225+01. It has no idea where it is or what's going on. Also it's a bit damp and surrounded by adorable otters, which is probably where the similarities end.

Yes Alright, But The Universe Is A Jolly Big Place, And There Are Bound To Be A Few Weird Things We Can't Explain.

Indeed. But there seems to me to be enough weird objects that we should stop and take a good hard look at what's going on. To try and make sense of all those otters, I decided to catalogue them. That is, I did a literature search for all the really long hydrogen streams and dark galaxy candidates. Then, as best as I was able, I took all the available measurements and made a table of their properties.

The numbers aren't large - around 40 for both the streams and dark galaxy candidates. Even so, finding all of the damn things, combined with the need to re-measure just about everything, took several months. Still, the awkward fact remained that the numbers are thousands of times too low to explain the missing satellite problem.

But the world turned, and turned again. There were hints that there could be much larger populations of some objects in some regions - not as many as models predicted, but enough to make one wonder. A few new objects seemed to have signs of ordered motions - a really strong signature of rotation (the data for the clouds I found wasn't high enough resolution to show this even if present). Much larger cosmological simulations began reviving the idea that maybe not all dark matter halos contain enough gas to form stars. And then there were of course the discoveries of huge numbers of very large but very faint galaxies - not quite dark galaxies, but certainly very dim indeed.

So the mood of the community had apparently done a complete 360. Maybe several. It had gone from originally thinking that dark galaxies should be very common to dismissing them as unlikely when they weren't immediately found by the first large-area HI surveys. It had turned to outright hostility when they continued to not be found, to the point where alternative explanations for any weird objects became demanded rather than requested. Now it's swung back, at least in some quarters, into a tentative sort of toleration.

Make Up Your Bloody Minds !

If you look hard enough, you'll be able to find a qualified expert with a strong but ridiculous opinion about anything. There were and are certainly some people in this field fervently convinced that dark galaxies do and do not exist. And there's more than a small element of, "my dark galaxy candidate is much better than your dark galaxy candidate" at work. But the more I looked, the more and more it seemed to me that no-one had a really good idea of what was going on. This must, I imagine, be an area that's particularly difficult for journalists - lots of very strong opinions from confident-sounding experts, but no consensus.

Anyway all my cataloguing had left me convinced that my little AGES clouds were indeed particularly weird objects*. None of their nearest galaxies looked disturbed or showed any signs of extended Hstreams. Very few other clouds showed a similar deviation from the Tully-Fisher relation. If their velocity width wasn't due to stable rotation, they ought to have disintegrated so quickly we'd be unlikely to detect them at all. They could be held together by their gas alone... but that would mean they'd have to be incredibly small and dense and form stars like nobody's business. At a more reasonable size, they probably couldn't survive for very long in the chaotic Virgo cluster without dark matter.

* Obviously, I'm not guilty of the my-candidate-is-better-than-your-candidate problem at all, because I'm a special snowflake.

The thing is, none of the original problems with dark galaxies were by themselves fatal. True, not enough had been detected. But we don't really have a great understanding of how gas gets into halos, so maybe most of the dark matter clouds don't even contain detectable amounts of gas. It's a bit like saying, "you told me there were three Easter eggs in this room, but I only found one, so whatever the heck it is I've found, this big chocolatey egg-shaped thing can't possibly be an Easter egg".

Likewise we don't really understand the Tully-Fisher relation all that well either. Actually the fact that faint galaxies lie on the TFR is surprising ! And these dark little clouds don't, so that becomes very interesting indeed. At least one of the optically faint but not totally dark galaxies also deviates in the same way. So defining galaxies as "anything that obeys the TFR" may be a teensy-weensy bit over-zealous. And here's the really interesting thing : lots of new, extremely faint galaxies have recently been discovered in clusters, and at least of them (probably more !) seems to be deviating from the TFR just like the clouds do.

Which only leaves that pesky simulation showing that objects like VIRGOHI21 could just be tidal debris. Now here I have to tread a very careful line indeed, because the authors of that study are certainly vastly more experienced than me when it comes to galaxy simulations. And yet the more I examined their findings, the more doubts I had about their conclusions. Their own figures don't show anything resembling VIRGOHI21. Worse, they were so preoccupied with whether or not they could explain VIROGHI21 they never stopped to think if their mechanism would. That is, a possible explanation does not mean it's a very likely one, let alone correct.


For a topic this controversial, even this long-winded post is a simplification. But it boils down to a few simple points :
  • If galaxy formation models are correct, then there must be a lot of dark matter halos which don't contain any stars and gas. Some might contain just enough gas to be detectable but not enough to form stars. 
  • But on the other hand, the models aren't that sophisticated. Although all models predict lots of dark matter halos, the numbers predicted to have just enough gas to be detectable but no stars vary wildly depending on the exact values used.
  • Gas-rich galaxies with no stars appear to be extremely rare... but, tentatively, a few might exist. That means the models of galaxy formation might be basically correct after all. So even discovering just a few dark galaxies could solve the missing satellite problem.
  • But we don't have a great model for the origin of galaxies and alternatives say that what appear to be dark galaxies could just be the relics of old interactions. So using a very few detections to infer huge numbers of truly dark, undetectable halos is rather dodgy, at best.
  • Galaxies with very few stars and no gas do exist in large numbers with certainty. Those stars must have formed from gas at some point, so truly dark galaxies should exist.
  • But we have absolutely no idea when they converted their gas into stars, or what caused some to remain dark and others to light up. 
  • A very few objects are known which are rotating more quickly than expected based on their observed mass - meaning they have extremely high amounts of dark matter. That would explain why they're not on the Tully-Fisher relation.
  • Yes, but we don't know if those AGES clouds are rotating because the observations don't have enough resolution. Tidal debris is at least as good an explanation. And even really dim galaxies follow the Tully-Fisher relation (despite predictions that they shouldn't), which suggests that totally dark galaxies should as well.
Let me put it another way :

We don't have a good idea of what's going on here. Anyone who says otherwise is guilty of wishful thinking. The only way to proceed is incrementally, point by point, slowly and very carefully tweaking the nipples of knowledge. Or something. But clearly, understanding these objects relies as much on theory as it does observation. So next time we'll head deeper into darkness and look at one of those key points : the popular idea that the objects could be some form of tidal debris.

No comments:

Post a Comment