IAU Symposium 355 : The Realm of the Low Surface Brightness Universe

I've decided to do everyone a favour and not combine the science and pretty pictures of landscapes from the latest conference. You can read about the travel experience of Tenerife here. Although I've already mentioned some of the outcomes, I thought I'd also give a more executive summary of the stuff I found most interesting.

The theme of this conference was the low surface brightness universe - the faint fuzzy stuff. It covered the whole range of such features, focusing on galaxies but also including zodiacal light, the challenges of observing (including a very nice talk from an "amateur") and data processing, philosophy of science issues and reproducibility (it's tricky to get really deep images, and sometimes they give remarkably different results), and was just generally awesome. So here are my highlights.


Mike Disney's introductory talk 

Mike is essentially the conference godfather, who predicted the existence of large numbers of low surface brightness galaxies way back when. He began by noting just how vicious the field can be - and he's right. The conference in Cardiff back in 2007 was the first I ever went to, and by far the most bitter and acrimonious. I'm not sure why it went down that way, but this one didn't. People said plenty of controversial things (especially Mike !) without ever seeming like they were about to start beating each other up.

Mike's major points were that there should be a large population of hitherto unseen faint galaxies, though he tried earnestly to present arguments both in favour and against this. While everyone agrees that galaxies certainly get a lot more interesting the deeper you look (in particular elliptical galaxies tend to look much more interesting), it's more controversial as to how many brand new galaxies this will pick up (more on that later). I've covered some of these arguments before, but I'll be revising some of them thanks to this conference.

One of the issues I don't think people quite got nailed down was the significance of the low surface brightness galaxies. Many good arguments were presented that such hidden galaxies cannot contribute much to the total amount of starlight. This is probably true, but misses the point that such galaxies could still be dynamically massive and dark matter dominated. So perhaps both sides are right, depending on whether one thinks of stars or dark matter as being more important.

Mike's other controversial point didn't get so much attention, probably because the conference wasn't really focused on it : the galaxies seen in the spectacular Hubble Deep Field, he says, cannot possibly be the progenitors of today's galaxies because they're too bright. I'm skeptical, but there wasn't time to go into this much.


Johan Knapen's data talk

This was essentially a philosophy of science talk from the perspective of the sheer data size that's coming our way very soon. This is a very real and serious challenge, with the SKA expected to produce exobytes of data per second. While Mike raised the point about defining the scientific method being very difficult, Johan took this a but further. He listed four possible paradigms of science that have changed over time :

1. Experiment-driven, as in the days of Newton and Gallileo
2. Theory-driven, as in Einstein and other analytic theoreticians
3. Numerical simulations
4. Data exploration

Other suggested that the fifth paradigm could be A.I. while the sixth would be letting Facebook do everything.

Johan's point (if I remember correctly) was that astronomy was going to have to move away from the traditional hypothesis-testing method we learn in schools and towards this "fourth paradigm" of being a data-driven approach. He's not wrong, but those who I talked to seemed to agree that this already the case - and maybe always has been. I've made the point before at length : there's more than one right way of doing science, and the data usually tell you something very interesting but utterly unrelated to what you were interested in. I mean this very literally. And I believe it was Simon Driver who, in discussions afterwards, said that in experiment proposals it should be absolutely legitimate to describe which area of parameter space you wanted to explore and why it was new, without needing to say exactly what you expect to find there. With this I fully agree. It's largely a waste of time describing observational results before you've got them.

Mohammad Akhlaghi's reproducibility talk

Mohammad doesn't like the fact that papers often contain highly vague instructions as to how experiments were carried out and how data was analysed. He's trying to tackle the latter issue by developing a package that makes it very easy to document the full details of the software used, by including the exact software name, version, and all its associated dependencies (libraries etc.), in a way that makes it simple to include in a paper. It'll also let you automatically update any numbers if you change the software without having to redo the calculations or edit the paper yourself, and it doesn't require any special software modules to install : the point is that the user has full control over what packages they use.

I was a bit skeptical listening to this, being acutely aware that many aspects are entirely subjective, but I came around to it afterwards. You'll have to take care that if you make changes, your new numbers are still consistent with your original conclusions. And I'm a bit doubtful that there are that many cases where changing software actually changes the answer. But the basic idea that papers should be as reproducible as possible is something I can definitely get behind - provided we remember that objective, repeatable measurements can still be absolutely wrong. Reproducibility means you can find the errors, not say, "my method is objective and therefore objectively correct", which is an easy trap to fall in to.

Thomas Sedgwick's hunt for dark galaxies with supernovae

This was one of the most interesting and novel methods proposed for finding very faint galaxies. As long as the galaxy is forming some stars, it will have a few supernovae, and these can be detected. This is not something I would ever have thought possible - even though supernovae surveys are now decades old, I just don't think of stochastically exploding stars as something you can do a survey of. But you can. And you can even work out how many galaxies your survey implies, given some very reasonable, justified assumptions about the survey completeness and star formation rate.

Interestingly, it turns out that these corrections imply a galaxy distribution that's fully compatible with the standard model : that there are indeed large numbers of very faint galaxies out there, just as models have predicted but observations failed to find. This is a really cool result, but it relies on an enormously large statistical extrapolation, so it's probably safe to assume the problem isn't solved just yet (and kudos to the speaker for saying as much).

A somewhat similar talk was given by Raja Guhathakurta on looking for faint galaxies by searching for their globular clusters. The difference here is it should be possible to get much better completeness of the sample. We're planning to do a search for such features for the Virgo clouds I've been working on.

Nushkia Chamba's new definition of the size of galaxies

Nushkia knew me as "that guy with the hilarious blog", which absolutely made my day. She's come up with a new parameter for the size of galaxies, which is extremely interesting. If correct, it'll dramatically change how we think about ultra diffuse galaxies. But she asked that this not go on twitter, and I assume that includes other social media so I'll say no more about it. Expect to hear a lot more when it's published.

Daniel Prole's talk on the abundance of UDGs in the field

Are ultra diffuse galaxies a very common galaxy component that until recently went largely undetected, or are they just a smattering of exotic objects ? Several people made the point that UDGs were already know but it's their abundance in new surveys that's got people excited about them again. Daniel (who has the same PhD supervisor as I did) is attempting to estimate how abundant they are in the field, which is much harder than in clusters and has a far large volume. He's got a number, but rather surprisingly hasn't compared this to other numbers so it doesn't mean much yet. I would have thought there are already numbers for more typical galaxies, but perhaps getting a fair comparison (e.g. correcting for survey biases) is harder than you might think.

Pavel Mancera-Piña's talk on the dynamics of UDGs

I was so glad to see this talk. You may remember that I've commented several times on the weird line widths of UDGs with HI detections, which tend to be much less than expected (scroll to "things are getting weird" in that link for a plot). I emailed a couple of people about it - I got a cautious response from one and nothing from another. I've shown several people and they all think it's interesting, but I never have time to work on this myself. Thankfully Pavel does, and he's done a much better job than me of demonstrating that this weird result probably isn't due to observational constraints : these galaxies do seem to be weird. While it looks unlikely that certain famous candidates are not actually galaxies without dark matter, some of these UDGs might soon resurrect this possibility.

My one major concern, which I think should be relatively easy to address, is survey incompleteness. At any given mass, galaxies of low line width are easier to detect. If galaxies are rotating discs, it so happens that means we'll preferentially detect ones which are close to face-on from our perspective, which makes it hard to estimate their true line width (details here). So it might be that the survey is biased towards nearly face-on galaxies - and because they're so damn faint, it's hard to measure their inclination angle directly. In principle one could test this by calculating the line width they'd need to escape detection and the corresponding inclination angle required to reach this. However, I very much doubt this will explain all the objects. Many of those galaxies with reasonably clear detections look quite convincingly close to edge-on, implying a negligible velocity correction.

Freeke van de Vort's talk on simulations the circumgalactic medium

Freeke has done some spectacular simulations of the gas structures around galaxies. Normally I think of this as probably very diffuse, fluffy stuff, interesting but not particularly photogenic. But Freeke's simulations really are spectacular - they look like the sort of thing you'd get if you told a Marvel CGI artist to "make some pretty gas - really go nuts with this". She notes that the simulations aren't yet converged. While they get the major galactic structures right, increasing the resolution keeps changing the results for the CGM. She's also found some gas clouds without stars or dark matter, but of course the resolution dependence makes the significance of this hard to assess.

Honourable mentions

There are too many to mention properly but I can't avoid a few others :

• Eva Grebel made the point that there are a few red, isolated UDGs known in the field, while several people (but particularly Anna Ferré-Mateu) noted that at least some UDGs may indeed be giants even if they aren't the majority. Most people seemed happy with the notion that there may be several different ways to make a UDG. 
• The Dragonfly team defended their data reduction procedures in the face of their failure to detect the double arms of NGC 5907, although no-one seems to know what happened. Another talk showed us that the double arms had been detected independently so they're almost certainly real.
• Gaspar Galaz said that there's a large linear stream extending from Malin 1, which is just weird. How you get a linear stream intersecting a stellar disc, I just don't know.
• Everyone agreed that galaxies look much nicer with deeper imaging but that it's jolly hard to do. 
• Bärbel Koribalski gave a very nice overview of extended optically dark gas features. Very nice to know that it's not just me working on this !
• Sarah Pearson showed how we'll soon have detections of large numbers of globular cluster streams around galaxies. I don't think of streams in statistical terms, but this could be an interesting way to constrain the behaviour of galaxies and their dark matter content.
• Anna Saburova, a collaborator of mine, sounded like she was about to kill everyone (it's the Russian accent) but described how difficult it is to explain giant low surface brightness discs. They most likely have different formation mechanisms - some by catastrophic collisions, others through slow accretion.
• Sebastiano Cantalupo explained dark galaxies at high redshift, which I was surprised to hear may be not all that dissimilar to the candidates at low redshift, with similar masses and dynamics. Definitely one I need to read up on more as I'd assumed the high redshift objects would be very different.

Which just about wraps it up. Plenty of background reading to do until the conference proceedings are released.