Quark and Gluon Jet Substructure

TL;DR

This paper addresses the challenge of distinguishing quark-initiated from gluon-initiated jets on an event-by-event basis at the LHC, aiming to enhance searches for new physics. It conducts a comprehensive Monte Carlo study across multiple jet pT windows, evaluating thousands of observables that fall into discrete (counts/subjets) and continuous (jet shapes and radial moments) categories, using ROC and multivariate techniques to identify high-impact discriminants. The key finding is that a small set of 1–3 observables largely captures the quark/gluon differences, with discrete counts offering strong power at high quark efficiency and continuous radial moments contributing at lower efficiency; combining observables provides incremental gains, and results depend on the event generator used. The study also discusses operating-point selection, validation with impure samples, and practical implications for detector-level use and MC tuning, highlighting that ~80–90% gluon rejection at 50% quark efficiency is feasible under certain conditions, though generator differences warrant data-driven calibration.

Abstract

Distinguishing quark-initiated jets from gluon-initiated jets has the potential to significantly improve the reach of many beyond-the-standard model searches at the Large Hadron Collider and to provide additional tests of QCD. To explore whether quark and gluon jets could possibly be distinguished on an event-by-event basis, we perform a comprehensive simulation-based study. We explore a variety of motivated and unmotivated variables with a semi-automated multivariate approach. General conclusions are that at 50% quark jet acceptance efficiency, around 80%-90% of gluon jets can be rejected. Some benefit is gained by combining variables. Different event generators are compared, as are the effects of using only charged tracks to avoid pileup. Additional information, including interactive distributions of most variables and their cut efficiencies, can be at http://jets.physics.harvard.edu/qvg.

Figures (28)

• Figure 1: Gluino decay as an of of a quark-heavy signal, in this case with 8 quark jets and no gluon jets produced. Multi-jet events in standard model backgrounds are extremely unlikely to have so many quark jets.
• Figure 2: Fraction of jets which are a light quark jet (up, down or strange) rather than a gluon jet Here all jets have the minimum $p_T$ cut indicated, but photons have a minimum $p_T$ of only 20 GeV.
• Figure 3: Jets are formed by grouping together collinear radiation.
• Figure 4: Parton showers produce quark jets whose properties are largely determined by the emitted gluons, as indicated in the left diagram. On the right, the same configuration is produced when a third hard parton, in this case a gluon, splits into two gluons with momenta equal to the showered gluons. Since the two amplitudes interfere, it might not make sense to describe this final state configuration as having two quark jets. In this case, however, the amplitude for the shower diagram is much larger than the hard-gluon-splitting diagram for the same final-state kinematics. In fact, as the gluons become more collinear with the quarks, the first amplitude is divergent.
• Figure 5: The $p_T$ distributions for two quark and gluon jet samples from Pythia8 with arbitrary normalization. Our samples at each $p_T$ were chosen such that anti-$k_T$ R=0.5 jets had $p_T$ values within 10% of the nominal value. For all QCD jets, this distribution is falling, but within our window, the $p_T$ itself cannot be used distinguish quark from gluon jets. On the left is the 50 GeV sample, and on the right is the 800 GeV sample.