This week my @ERC_Research Starting Grant CSINEUTRONSTAR draws to a close. Over the next few days I'll be tweeting about the research that we've done, what we've learned and (of course) what we're still puzzling over.

Why CSI? The subtitle of the proposal was "The physics and forensics of neutron star explosions"...

Let's start with the goals of the project, written back in the mists of time ~6 years ago. There were two. (1) Try to figure out the thermonuclear burst oscillation mechanism, and (2) Ensure we could use burst oscillations to study dense matter using pulse profile modelling.

Background - accreting neutron stars build up oceans of hydrogen and helium that periodically go boom thanks to thermonuclear reactions going nuts. Bright bursts of X-rays ensue. We observe these with space telescopes.

Sometimes, part of the exploding ocean gets hotter than the rest. The surface is "spotty". As the star spins we therefore see the X-rays vary, leading to them being called 'burst oscillations'. @xrayastroh discovered them in '96 and we're still stumped as to the cause.

Lots of ideas had been proposed - thermonuclear hurricanes, large-scale waves in the burning ocean, and so on. All had come up short in the collision of data and theory. Which was a shame because people kept on finding burst oscillations from different stars.

So plan one was to solve this. Noting that lots of smart people had worked on this already, I did at least have the nous to note that this part was high risk and that we might not succeed. And Reader, I am sorry to say that we have not solved the burst oscillation mechanism.

But did we make progress towards understanding what's going on? Well yes, I think we did. We used the Traditional Astrophysical Attack Technique of Data+Theory.

Let's start with data. By the time we started the project, there was a huge archive of burst and burst oscillation data courtesy of the Rossi X-ray Timing Explorer. So we decided to go hunt through the archive to see if there were any more clues to be found.

Rossi X-ray Timing Explorer was my first X-ray telescope. Fantastic instrument. Will always have special place in my heart...

First thing we did was to go and really pick over the data set to see if there were any more sources to be found. One of the hallmarks of burst oscillations is that they show up at the same frequency in different bursts from the same source.

I had a hunch that if we looked carefully and modelled the noise better we might find a bunch of "sub-threshold" signals at repeating frequencies.

We found...nothing.

Well, that's not quite true.

We found one *strong* signal that had been utterly missed. And we found a strange type of variability that seems to jump around in frequency. We suspect something astrophysical, to do with [insert weird burning physics woo]. We called it "glimmer" because we love My Little Pony.

We delved into the apparent dependence of burst oscillation amplitude on accretion rate. I was hoping that with a bigger data sample we'd find some nice transitions or jumps in behaviour to give a theory steer. No such luck. There are trends but no Rosetta Stone™ moments.

So you know what, data? Fine. Be all enigmatic. See if I care. Because we like a challenge, don't we, over on Team Theory?

Back to it! And Team Theory takes up the challenge of trying to figure out what the heck causes burst oscillations - basically the fact that part of the burning ocean appears hotter and brighter than the rest.

One of the neatest (IMHO) ideas thrown around in the early to mid 2000s was the idea that the spreading thermonuclear flame front could excite large-scale (star-sized) waves in the burning ocean. A sort of thermonuclear El Niño, if you will, causing bright spots.

A particular type of wave was singled out as the best candidate - the buoyant r-mode (where r = Carl-Gustaf Arvid Rossby), meteorology crossovers FTW.

But when theorists calculated how the frequency of the wave should shift as the ocean cooled, post-explosion, they ran into a problem - the frequency should drift much more than we observe. So the model went back in the cupboard.

But burst modelling had moved on! So we wondered if, by including more up to date nuclear burning physics, the drift problem might go away. And it seems like it could! Not definitive yet (small model sample) but worth pursuing (especially given lack of alternative models).