The gas that just won't die

This is a paper I really thought I'd blogged ages ago, but after searching high and low I think I might have dreamed it. Still, better late than never.

During my PhD it seemed to be common knowledge that atomic hydrogen couldn't exist below a certain density because it would all just be ionised by the cosmic UV background. Where exactly this assumption/calculation came from I don't know, but according to today's paper there's probably no need to worry about this because this threshold doesn't exist.

On the opposite end of the spectrum, it's more well-established that HI densities rarely exceed a certain threshold, about 10 solar masses per square parsec (10²¹ atoms cm^-2). This seems to be because at this point the gas is so dense that it's pretty much inevitable that it starts to form molecules, and H₂ doesn't radiate in the same way that HI does. Gas in general doesn't really have an upper density limit, it's just that its atomic and molecular states require different detection techniques.

There's not much difficulty in establishing this upper limit for the HI. But the lower limit is much more challenging, because getting observations sensitive down to the predicted threshold for ionisation (10¹⁹ atoms cm^-2) is technically difficult : you need a big telescope and lots of observing time. It's also well below the threshold for star formation, and since most people would say that any gas which isn't forming stars is boring, there hasn't been quite so much pressure to study it. Not that people wouldn't like to, it's just too difficult for a relatively small reward.

Two different kinds of telescopes are available for HI, and they haven't helped matters. Single dishes can be incredibly sensitive but have low resolution. This means they can detect very small masses of gas but they can't accurately estimate how large an area it spans - it's hard for them to distinguish genuinely big, fluffy clouds from a series of small dense clouds all close together. To a single dish, a very low mass but very dense cloud would be "smoothed out", and this beam smearing would make it appear as though it was much less dense than it actually was. Interferometers, on the other hand, are much less sensitive but can have far better resolution : they can accurately measure the size of any detected gas clouds, but they have trouble detecting masses as low as those that single dishes can find.

The upshot is that interferometers might not be able to see very diffuse gas because they're just not sensitive to it, while single dishes smooth everything out and make it appear more diffuse than it really is. Hence the lower limit on gas density remains difficult to establish.

Today's paper attempts to overcome this. They use interferometry data, but instead of trying to detect the faintest gas directly, they use a variant of our old friend stacking. In this case they make radial profiles of the gas in a number of different galaxies, essentially averaging the gas over many different annuli. This part's easy, but the novel approach they use is to extrapolate the rotation curve so they can average over approximately the correct velocity range in each bin. This is necessarily a bit crude, because rotation curves aren't always smooth, but it does seem to work : they show convincingly how this recovers signal well beyond the nominal edge of the gas disc.

The bottom line is that most of these improved radial profiles don't show any evidence of a break. If the HI was lop-sided then these circular averages might wash out a break if it existed, but it doesn't appear that that's the case. The only examples where there do seem to be breaks are galaxies which have unusual structures like rings, and aren't representative cases.

I can't say I'm surprised to find there's no break. Of the little data we do have with single dishes where the gas is well resolved (because the galaxy is particularly close), there's no such break visible : the profile continues right down to our sensitivity limit (~10¹⁷ atoms cm^-2). Now for low-density structures outside of galaxies, where the gas has likely been recently removed from its parent, this might not be a problem at all, since it might take some time for it to become ionised after removal. But it's very hard to believe this is the case for the gas discs of galaxies themselves.

It's still an open question as to whether the gas is really at these very low measured densities or actually is in the form of smaller, denser clouds that are smoothed out, but to me that would seem like a very odd coincidence : surely we wouldn't expect such a neat, smooth profile in that case. Again, possible for the displaced gas, but it seems somehow unlikely for the galaxy discs. 

What are the implications of these ? Well, it's not that ionisation in general doesn't happen : gas tails in Virgo are awfully hard to explain without this, and in some cases having even deeper HI observations doesn't get you any new information. There's probably not much point in going all that much deeper than ~10¹⁷ atoms cm^-2, though I'd certainly like to see someone try. 

But with interferometers generally giving out at much higher densities (typically ~10¹⁹ atoms cm^-2 , though some of the new ones are pushing 10¹⁸ ), there's a lot of scope for more sensitive upcoming instruments to tell us an awful lot about the outermost regions of galaxies. Those regions can tell us both about how galaxies assemble (e.g. as the authors suggest, if it's not ionisation then maybe it's the assembly process that sets the lower density limit) and interact (e.g. by finding elongated gas structures invisible at optical wavelengths).

Presumably it's also bad news for the photoionization models that predicted the density cutoff, but since I never paid them too much attention, I'll worry about them another time.