Trigger-level searches for low-mass dijet resonances

The LHC is not only the highest-energy collider ever built, it also delivers proton–proton collisions at a much higher rate than any machine before. The LHC detectors measure each of these events in unprecedented detail, generating enormous volumes of data. To cope, the experiments apply tight online filters (triggers) that identify events of interest for subsequent analysis. Despite careful trigger design, however, it is inevitable that some potentially interesting events are discarded.

The LHC-experiment collaborations have devised strategies to get around this, allowing them to record much larger event samples for certain physics channels. One such strategy is the ATLAS trigger-object level analysis (TLA), which consists of a search for new particles with masses below the TeV scale decaying to a pair of quarks or gluons. The analysis uses selective readout to reduce the event size and therefore allow more events to be recorded, increasing the sensitivity to new physics in domains where rates of Standard Model (SM) background processes are very large.

Dijet searches look for a resonance in the two-jet invariant mass spectrum. The strong-interaction multi-jet background is expected to be smoothly falling, thus a bump-like structure would be a clear sign of a deviation from the SM prediction. As the invariant mass decreases, the rate of multi-jet events increases steeply – to the point where, in the sub-TeV mass range, the data-taking system of ATLAS cannot handle the full rate due to limited data-storage resources. Instead, the ATLAS trigger system discards most of the events in this mass range, reducing the sensitivity to low-mass dijet resonances.

By recording only the final-state objects used to make the trigger decision, however, this limitation can be bypassed. For a dijet-resonance search, the only necessary ATLAS detector signals are the calorimeter information used to reconstruct the jets. This compact data format records far less information for each event, about 1% of the usual amount, allowing ATLAS to record dijet events at a rate 20 times larger than what is possible with standard data-taking (figure, left).

While the TLA technique gives access to physics at lower thresholds, the ATLAS detector information for these events is incomplete. Dedicated reconstruction and calibration techniques had to be developed to deal with the partial event information and, as a result, the invariant mass computed from TLA jets is comparable to that using jets reconstructed from the full detector readout within 0.05%.

The data recorded by ATLAS in 2015 and 2016 at a centre-of-mass energy of 13 TeV did not reveal any bump-like structure in the TLA dijet spectrum. The unprecedented statistical precision allowed ATLAS to set its strongest limits on resonances decaying to quarks in the mass range between 450 GeV and 1 TeV (figure, right). The analysis is sensitive to new particles that could mediate interactions between the SM particles and a dark sector, and to other new resonances at the electroweak scale. This analysis probes an important mass region that could not otherwise be explored in this final state with comparable sensitivity.

ATLAS joins CMS and LHCb with an analysis technique that requires fewer storage resources to collect more LHC data. The technique will be extended in the future, with upgraded trigger farms and detectors making tracking information available at early trigger levels. It will thus play an important role at LHC Run 3 and at the high-luminosity LHC upgrade.