My Chemical Enviromance

Yay, science post !

It's always nice when you wake up and see an email saying, "Your paper has now been accepted" but you're still half-asleep so you wonder "Dafuq ? I haven't written any papers, bloody spam journals again" but you're curious enough to keep reading the email just in case and realise "Ooooohh yeah, I didn't write that one but I'm a co-author, yay for me cos I didn't have to do any real work this time !" and then you fall back into a satisfying slumber.

Obviously we all sleep at our desks the whole time.

Funnily enough, that's more or less what happened to me last week.

This paper is by long-term collaborator and all-round thoroughly good egg Robert Minchin. It's all about my second favourite topic in my first favourite place : that is, it's about how galaxies lose gas in the Virgo Cluster (my favourite topic being little gas clouds that don't do anything).

I've covered gas loss in Virgo before, but since it's been some considerable time (read : the whole pandemic) since previous posts about this, I won't expect you to go trawling through old posts for details. Oh, you can if you want, but there's no need, I'll just bring you up to speed right now.

Some background : when galaxies get naked

A typical galaxy consists of a disc of stars and some gas all bound together in a dark matter halo. Even outside galaxies space is never truly empty, and especially in dense clusters, the external gas can be significant. It tends to be substantially less dense but also a lot hotter. If a galaxy moves through it fast enough, it can experience a pressure strong enough to dislodge its own gas in a process worryingly known as ram pressure stripping. 

Sometimes, though rarely, the stripped gas can be so dense that you get a star-forming "wake" behind the galaxy as it moves through the cluster. Normally you can't see this gas using optical observations, but have to use other wavelengths like radio to see the stripped gas trails.

While this might happen to some extent in all galaxy environments, clusters are where it really matters : here the external gas is relatively dense and galaxy motions are (by far) the most rapid. Other processes like gravitational encounters between individual galaxies become much less important compared to the enormously strong ram pressure, which can completely strip a galaxy of all its gas in less than an orbit. 

After that... the galaxy is doomed to a slow, lingering death. With no remaining gas it simply can't form any new stars. Its youngest, hottest, bluest stars soon die off, leaving behind only the smaller, dimmer, red stars. And without the mass of the physically thin but dense gas disc helping to hold them together, the random motions of the surviving stars eventually destroy any hints of structure in its stellar disc. Eventually, it turns from a magnificent blue sparkly spiral into a pathetic red elliptical. Everybody hates ellipticals so it becomes a social pariah and never gets invited to parties anymore. Not even during lockdown. No, not even at Downing Street.

The details of the process are controversial, but the basics are accepted well enough :

• Ram pressure is strong enough in clusters to cause even massive spirals to rapidly lose all their gas
• Spiral galaxies in clusters typically have much less gas than those elsewhere
• Other mechanisms can't seem to explain the gas loss.

Which all leads to the main conclusion that ram pressure plays a dominant role in galaxy evolution in clusters.

Some time ago I was co-author on a paper that attempted to model this process using nice, simple analytical formulae. The gold standard is to do full-on numerical simulations that includes all the complex gas dynamics and stuff, but it seems that the simple formulae are actually plenty good enough to predict the basics. This saves an awful lot of time and, more importantly, effort, because running simulations is annoying. More recently, I was able to show that our model does quite well at matching which specific galaxies are predicted to be currently losing gas and which actually show the long gas streams expected when stripping occurs.

Now I like looking for gas streams very much - looking at data is just inherently a good idea, and it also lets you see what's happening as directly as possible. So you might think that this is just a cunning ploy to let me do more data visualisation.... not so ! For we've also shown that sometimes the gas streams are inherently hard to spot, and the gas disperses quite rapidly. So could there be a different signature of stripping we could look for to test which galaxies have been affected ?

In our latest paper it seems the answer is a tentative but enthusiastic "yes !". We found a radically different test for ram pressure that seems to provide a very pleasing confirmation of the model completely independent of its original formulation.

How to hunt for farting galaxies

The way I mentally group my astronomy knowledge is into three basic categories : atomic hydrogen, stars, and everything else. While stars are what get all the glory, and atomic hydrogen is the largest component of the gas, I'm vaguely aware that there's really quite a lot of stuff contained in the "everything else" category.

You get the idea.

Fortunately, Robert is much more acutely aware of this than I am. For example, we can estimate star formation rates by looking at how much light galaxies emit in different wavelengths. The bluer the light, the more it's dominated by hot, short-lived stars* and so the higher the current rate of star formation must be. We can do this just by looking at broad-band optical filters, much like the RGB components you'd see in an ordinary digital image, or we can user similar filters at shorter wavelengths than visible light (for example, ultra-violet emission is an excellent way to look for some of the hottest stars of all). 

* I am simplifying quite a lot here. Chemical composition also affects colour, as does the total intensity. But these factors can be accounted for.

But we can also do something quite a bit different. Broad filters take in emission from a wide range of wavelengths, but some processes produce photons only over very narrow windows. These "spectral lines" provide another way of testing for the high energies associated with hot young stars. One of these lines, the [CII] ("C two") line, from singly ionised carbon*, has become popular in recent years as another tool in the arsenal of available methods of estimating star formation rate. 

*Any sort of astro-chemistry is something I'd normally steer well clear of, on the grounds that even feckin' hydrogen isn't properly understood, but Robert is a much braver man than I. Personally I think of astro-chemistry some sort of advanced alchemy.

However, [CII] is not emitted directly by stars, but from the interstellar gas. And it turns out that injecting energy into the gas from other sources besides star formation, even mechanical energy, can also trigger [CII] emission... and ram pressure stripping might just be a very good way to do that. 

See, the world of galaxy evolution is anything but woke (though there are some very odd attempts to claim that its offensive language is all due to colonial oppression), and stripping is a violent process*. So slamming the galaxy into the intracluster medium might be indeed be a means for causing it to emit at the [CII] frequency. It's already though that other sorts of gas collision can induce this. Indeed, a paper from as far back as 1999 noted that a spiral galaxy in Virgo had a weird excess of [CII], and this later turned out to be one of the best examples of a galaxy experiencing stripping !

* To the authors suggesting we need to use less violent terms I say OH GOD NO, this is one area in which astronomy is at least still able to give things decent names (unlike new telescopes, which are always called the Very Large Something Or Other). I want my silly jokes about naked bestiality, dammit !

So this bodes well. If we can find galaxies with an excess of [CII], more than predicted from their star formation rates, we can compare this with our model. Since we already predicted which galaxies are currently stripping, we can potentially use this as a completely independent test on whether or not our model is any good.

The results

Spectral line observations are always more challenging than broad-band observations, and the [CII] line is technically difficult. So instead of doing our own observations, we mined the Herschel archive. Of the ~2,000 galaxies in the Virgo Cluster, just 14 had suitable observations. 

So right from the start it was clear that this could only be a limited pilot study. But even given this, initially the results looked confusing at best, and at worst, disappointing. Here's our plot of the excess of [CII] emission as a function of distance from the cluster centre :

Yeah, not exactly the clearest trend in the world... take away the two outliers (VCC 737 and 841) and arguably the "trend" goes away completely. Hmm.

What we might naively expect to see is a clear decline with cluster-centric distance, as the corresponding ram pressure should decrease because the cluster's gas is less dense at greater distances. But it's hard to argue we see anything more than a hint of that, and it's not at all convincing.

You can also see we divided our already tiny sample into two even more miniscule samples : some of our galaxies lie close to the centre of the main cluster (the "northern" sample) while the others are all significantly further away ("southern"). The big blue and purple symbols on the right show the means of the two samples. The southern sample is a control group since ram pressure should be less effective at these distances, while the big green symbol shows another set of galaxies of similar masses but found completely outside the cluster - this forms a second control.

At best, there's a bit of a difference. The [CII] emission is a little bit higher in the northern sample than the others. This is what we'd expect, but it's not exactly an edge-of-your-seat result.

But using cluster-centric distance as a proxy for ram pressure may be too simple. As we'd shown in a previous paper, assuming clusters are nicely symmetrical is about as bad as the proverbial cow :

Or graduate students, according to Fritz Zwicky.

Which is where our earlier, analytical model for ram pressure comes in. Surely what we should do, instead of using that old-fashioned crude approach of cluster-centric distance, is check whether our fancy model tells us if the galaxies are currently stripping or not. With this we should have a much more accurate proxy than simple distance.

Our model calculates two things. First, it estimates how much ram pressure a galaxy should be currently experiencing. This is derived from an earlier model of the gas density within the cluster and assuming the galaxy is moving at about the local escape velocity at its current position. We call this parameter P_loc (pressure at the local point). Secondly, given the mass of gas within the galaxy, we calculate the well-known parameter of deficiency, which just means how much gas it's lost compared to a similar galaxy found in isolation. From this we can compute the parameter P_def, the pressure needed to reach its current deficiency.

There's a lot of simplifying assumptions in all this, but it gives a nice, simple result : the higher the pressure ratio, the more likely a galaxy is to be currently losing gas. In contrast, galaxies with low pressures may have lost gas in the past but can't be doing so any more. Sounds great ! And remember, this result agreed well when we looked at which galaxies do seem to be losing gas based on their gas streams. But the result when looking at the [CII] excess ?

This is the most boring plot about strippers I've EVER seen.

Yeah, not so much... This was a bit disappointing considering how well the model had worked when looking for gas streams. What was going on ? Does the [CII] just not tell us anything at all about ram pressure but only star formation ? Had we messed up somehow ? Should we go and hang our heads in shame ?

Probably not. Actually, we'd probably over-complicated the situation. The model was constructed in the framework of the usual way of looking for for stripping, by directly searching for lost gas : either just be measuring how much gas a galaxy had, and/or by seeing if it had any detectable gas streams due to stripping. The pressure ratio works well for that scenario because that's the very thing it was based on. But for just injecting energy into the galaxy, which is what the [CII] is sensitive to... that won't work. The pressure ratio is almost irrelevant here : much more important is simply the current pressure. That, not whether a galaxy is losing gas or not, is what dictates the injected energy - which is what might provoke the [CII]*. And when we plot that :

* For example a galaxy which is currently not actually losing gas, i.e. having a low pressure ratio, might still have a high absolute value of ram pressure. The ideal would be to work out how much energy is being injected and how this relates to [CII] emission, but this is a much bigger task. The point is that pressure should be a much better proxy, and doesn't relate linearly to cluster-centric distance.

Bingo ! Now that's a nice clear trend, especially considering the tiny sample size - and the difference between the two sub-samples is stark. Granted there's one weird outlier, which we're unable to explain, but it'd be surprising if there wasn't. If you don't have one weird outlier in observational astronomy, everyone laughs at you or calls you a liar. Or both.

Actually we were a bit surprised by just how clear this trend is. The model is by design simple, and subject to many uncertainties. One of the biggest is that it still does have to consider projected distance (that is, distance on the sky) from the cluster centre when calculating pressure - it can't use true 3D distance, because we just don't know it. This means the true pressure can always be lower, since the galaxy might actually be a bit in front or behind the bulk of the cluster gas. Yet it works even so.

And the trend seems to continue down to very low ram pressures indeed. So ram pressure might be having an effect even in much less dense environments than clusters, like groups. This is perfectly possible, it's just surprising to see such striking evidence of it. Yet as far as we can tell, there is no good reason to expect this trend to be due to anything mundane : there is no selection effect artificially restricting us to galaxies which only appear to follow the trend but actually do so only by chance. 

We also accounted for the fact that some of the [CII] excess will be due to star formation. The thing about ram pressure is that although on sufficiently large and long scales it becomes very simple (all gas gone => no more stars !), on small and short scales it becomes fiendishly complex. As the pressure builds, it can initially compress the gas disc, temporarily triggering an increase in star formation. Then you get all kinds of terribly turbulent structures developing, which look nice but can't be modelled without proper simulations. But we can account for how much the star formation - whatever its cause - increases the [CII] emission, and find that it isn't enough. The simplest explanation is that we're seeing the direct impact of the ram pressure on the gas - we're not seeing galaxies with an excess only because they happened to have high star formation activity.

Conclusion

I wouldn't want to oversell this, because it's still based on a very small sample. But there's an almost startlingly nice trend between our prediction for how much ram pressure a galaxy is experiencing and the intensity of its excess [CII] emission. So our happy-go-lucky, simple model of ram pressure is vindicated for a second time by a totally different method.

It also means that this could be a new way for looking for the signatures of the effects of environment. Even if the ram pressure isn't actually strong enough to cause gas removal, it seems we can see its effects using the [CII] line. And since this appears to happen even at modest pressures, this could apply not just in clusters (where it's a bit of a case of "big bloody deal, we knew ram pressure was happening anyway"), but also in groups and filaments, where the situation is much less clear. That gives us a new way to examine the external influences acting on galaxies.

But again, 14 galaxies ! That's a sample size less than a full-strength rugby team, for crying out loud. So let's not go nuts : it remains a pilot study, a case of, "hey, this seems to work, let's try this some more and see where it goes", not, "everybody just go home now because we're done."

There are two important caveats to end on. First, an earlier study didn't find any correlation between environment and [CII], though this is probably because the authors (a) had less sensitive observations; (b) targeted larger galaxies, which would require more injected energy; (c) didn't consider the pressure parameter. Second, more problematically, getting [CII] observations is difficult. This is something that the SOFIA telescope would be very good for, so let's just hope NASA don't decide to do anything daft like cancelling it.

(Oh, and an apology. It has only just struck me that the title, "Environmental effects in Herschel observations of the ionized carbon content of star forming dwarf galaxies in the Virgo cluster" is probably one of the most boring titles we've ever done. Can't win 'em all.)