10 inverse microbarns later: The first ATLAS heavy ion run

15 December 2010


The ATLAS heavy ion programme officially began in the ATLAS control room on an unseasonably warm night in early November. The first collisions, just after midnight on Sunday, were already quite impressive, with the ATLAS TRT showing a huge track density in the lower multiplicity events, and a featureless white disc in the higher ones. The stable beams which arrived Monday morning were even better, with the fireworks of thousands of rainbow-colored tracks appearing just seconds after first collisions with the full detector turned on. At that point, we could reflect back on the months of preparations: optimizing the tracking for the enormous number of expected particles, redesigning the complex menu of trigger items, studying the calorimeter response to jets even with these enormous backgrounds, and implementing all of this into the comprehensive reconstruction and analysis software.

And those months seemed to fade away as we watched the show on the wall of the control room, replayed every few seconds with a surprising variety. We saw dense events with thousands of particles followed by events that looked barely different than proton-proton collisions. We saw enormous jets in the calorimeters, jutting clearly above the energy that filled the entire detector. We saw a few muons emerge blithely from the fireball into otherwise-quiescent muon detectors. More than once, I and other colleagues would just find ourselves staring at the wall for hours, just watching the data pour in -- at rates of up to 700 MB/sec.

From the data taking, we moved right into analyzing the data to survey the landscape, relative to our expectations from many years spent on the continuing RHIC program at Brookhaven National Laboratory in the US. And here again, years of preparation, and patient and capable software support from ATLAS were essential to a smooth transition to running on heavy ion collisions. We had found from our simulations that the event size was not particularly large compared with what was found in proton-proton, even though the number of tracks per event was often 100's of times greater than seen previously. And it was a nice surprise to find that the overall multiplicities were in fact smaller than our simulations would have suggested. This is something you never really know until the data is available -- predictions based on RHIC data varied by more than a factor of 2!

Many were surprised how ‘low’ the luminosities seemed, by comparison to proton-proton. While the p+p program had just seen luminosities exceeding 10 32/cm2s, the heavy ion beams were providing roughly 5x1023 in the earliest days (1 colliding bunch), and reaching 3x1025 (approximately 128 colliding bunches). The key factors to keep in mind to compare to proton-proton collisions are 1) lead nuclei are more than 100 times more likely to interact with each other than protons, simply because of their larger size, 2) when they interact, calculations suggest that there are 400 sub-collisions of the nucleons within the nucleus. This gives an ‘effective luminosity’ of 40,000 times the luminosity reported by the machine, which is relevant for the production of high pT objects such as jets or Z bosons. Thus, our effective integrated luminosity for the run (10 inverse microbarns) should be seen as more like 400 inverse nanobarns -- similar to the dataset ATLAS had going into the 2010 conferences!

The first surprises at the LHC energies were noticed in the very first data. Heavy ion physics has deep connections to minimum-bias proton-proton physics in the sense that the basic properties of the collision -- such as the multiplicity distributions or the average transverse momenta of the emitted particles -- are not known ahead of time. Even more interestingly, the ‘soft’ particles which are produced the most copiously, actually carry an enormous amount of information about the time history of the quark gluon plasma. At RHIC, we discovered that the best way to describe the data involved treating the whole plasma as a drop of liquid using the laws of relativistic hydrodynamics, which work just as well for cosmology as they do for us (i.e. they do not depend on the distance scales involved!). We immediately noticed various features of the soft particles -- in particular, the strong ‘elliptic flow’ signal -- that were quite intriguing. We were also discovering that we were leaving the ‘soft’ regime quite quickly, with high rates of particles with transverse momenta exceeding 20 GeV (and reaching far, far beyond that with the larger data set). Simply because of the factor of 14 in beam energy, it took RHIC years to reach even this regime, which we were seeing in the first couple of days of data.

And within a week, we were already starting to look at fully-reconstructed jets in heavy ion collisions. The preparations for this also involved several years of hard work from the heavy ion group. While the procedures for jet reconstruction were essentially the same as with protons, entirely new ones had to be developed for subtracting the enormous overall fluctuations of the heavy ion event, and for transforming the signals in the ATLAS calorimeter into a calibrated energy. Of course, while these methods had been tested extensively on simulated events with jets embedded into this, there was no way to know ahead of time if the physics of either the jets or the underlying event would behave the same way as our expectations. Even worse, most expectations going into the LHC era were that jets would be modified by the hot, dense medium formed in the collision, but only in relatively subtle ways, with a redistribution of energy within the jet cone.

However even the first event displays showed quickly that something interesting was going on. While there were quite a few events with two jets emitted back-to-back in the transverse direction, as observed from proton-proton collisions, there were a surprising number of events where a jet that resembled a towering skyscraper on the ‘lego plot’ was surrounded a spiky, but essentially featureless plain (I named them ‘Burj Dubai’ events). Closer inspection revealed a clear enhancement of the energy in the opposite direction, but spread out at huge angles relative to where it was expected to be. It also turned out that the simulations described the data surprisingly well. The story has been told in other e-News articles but only 10 days later, the ATLAS paper on ‘asymmetric dijets’ was accepted by Physical Review Letters.

Of course, while the paper preparations were certainly exciting, we never lost sight of the fact that there was a run still under full swing at the same time, which would eventually quadruple the data set included in the PRL. Still, after the LHC brought the machine to more than 120 colliding bunches, the run proceeded quite smoothly and ATLAS took data steadily, accumulating 9.2 inverse microbarn to tape -- including 70 million heavy ion events -- more than three times the integrated luminosity expected before the run. The ATLAS operations team, from run coordination to the subsystems, and finally the trigger crew, worked very coherent with the physics group to navigate the changing landscape of the machine conditions, as the steadily increasing luminosity required careful modifications of the running and trigger setup.

So near 10 inverse microbarns later, ATLAS now has enough data for a very productive look at heavy ion physics in the coming year. We have already seen events with Z bosons decaying both to muons and electrons, events with jets and a recoiling photon, and we expect to have nearly 10,000 jets above 100 GeV, enough to perform quantitative studies of jet properties in the heavy ion environment. We will of course also be able to look at the ‘bulk’ properties of the collision in detail: particle multiplicities, momentum spectra, and elliptic flow. What we can't yet predict is what the ‘LHC style’ of heavy ion physics will be -- but all signs point to an exciting few years ahead.

   

Peter Steinberg
Brookhaven National Laboratory