Piles of pile-up data turned to good use

One common misconception about science at the Large Hadron Collider LHC, CERN’s flagship particle accelerator, is that as soon as scientists find something interesting in the proton-proton collisions, their work is done. Another one is that a proton-proton collision is what it says: a collision of a single proton with another single proton. The real picture is more complicated. While protons do collide with other protons they’re only ever one of many protons in a bunch. One of 1·1011 to be precise. So that means it’s never just one on one, but it’s dozens of times one on one when bunches collide. Out of these up to 30 million events per second, only a selection makes it past the ATLAS experiment’s event gamekeeper, the trigger system, which stores it (and all the other collisions that occurred at the same time – the so-called pile-up) away for future analysis. The bunches are so packed full of protons to enhance the probability of something interesting occurring because – and this brings us back to misconception number 1 – the more data you have, the more you know about the processes that occur during the collisions.

University of Geneva researcher Steven Schramm has made it his task to look at what everybody else considers a nuisance: the pile-up. These are the collisions that happened in the same bunch of protons as the collision that was found to be interesting by the trigger system. In a paper posted on arXiv, he and his team have summed up all their research of the last three years; this research formed the basis for Schramm’s ERC grant and his professorship at the University of Geneva. Taking data from the LHC’s Run 2, i.e. data taken between 2015 to 2018, they removed the one collision that triggered the recording and used the rest as their abundant trigger-unbiased dataset which, they think, could be a treasure trove for all kinds of analyses.

“In one sweep we suddenly have fifty times more data than we had before of this kind, which we can use to look at different kinds of interactions,” Schramm says. For example, there are processes in collisions that the trigger cannot efficiently pick up on, but that colaterally make their way into the pile-up. Alternatively, there are processes that occur so often that triggering on them (and thus recording data from every single bunch crossing in which this process occurs, including its pile-up) would lead to uncontrollable amounts of data. The pile-up collisions are all there already, without any additional cost or requirements on the trigger system, readout, or storage.

The team demonstrated that they can extract the jet energy resolution from the pile-up, which is a reference of any detector’s ability to resolve the difference in energy between two jets. A jet is a shower of particles in the detector with all particles flying in the same direction. By comparing with a special trigger-unbiased dataset recorded by ATLAS, they actually managed to reduce the statistical uncertainty for jet transverse momenta below 65 GeV. Transverse at LHC means transverse to the beam direction.

Schramm is relieved to have passed the full ATLAS review after some initial doubt in the community about the feasibility: “It feels wonderful to finally show the world that yes, it works!” The jet energy resolution result is only the beginning. “At the moment, we’re at the stage of technology demonstration, i.e. showing that it works,” says Schramm. The next step would be using it for physics analyses. “We think our approach will be interesting for many low-energy hadronic physics analyses or for ones with a very inefficient trigger,” he says. It also opens the door to even more precise measurements of known parameters of the Standard Model of particle physics.

Usage of this concept is likely to grow with datasets for Run 3, which will include a larger dataset with higher pile-up conditions – up to 200 collisions simultaneously. Moreover, the ATLAS trigger system will record a larger fraction of the delivered data, thus more pile-up collisions will be available for study. “Pile-up will keep growing with every upgrade and future accelerator projects, Schramm predicts. “We hope to have made a useful long-term contribution to the community with our approach.”

Barbara Warmbein