Quirk SUEP

TL;DR

The paper addresses the challenge of discovering physics beyond the Standard Model at the LHC by exploiting Soft-Unclustered-Energy Patterns in low-$p_\mathrm{T}$ tracks that accompany TeV-scale resonances. It introduces a benchmark quirk model with a dark QCD sector, simulates signal and background using MG5_aMC@NLO and Pythia, and evaluates three strategies—rectangular track-multiplicity cuts, a supervised neural network, and the CATHODE weakly supervised anomaly detector—under 140 fb$^{-1}$ at 13 TeV. The results show that track multiplicity is the dominant discriminator, with supervised learning and anomaly detection providing significant but complementary gains over an inclusive resonance search, enabling discovery or exclusion across wide regions of parameter space. This work demonstrates a viable path to extend resonance searches with soft track information and motivates further exploration of different resonance types and dark-radiation patterns.

Abstract

We propose searching for physics beyond the Standard Model in the low-transverse-momentum tracks accompanying hard-scatter events at the LHC. TeV-scale resonances connected to a dark QCD sector could be enhanced by selecting events with anomalies in the track distributions. As a benchmark, a quirk model with microscopic string lengths is developed, including a setup for event simulation. For this model, strategies are presented to enhance the sensitivity compared to inclusive resonance searches: a simple cut-based selection, a supervised search, and a model-agnostic weakly supervised anomaly search with the CATHODE method. Expected discovery potentials and exclusion limits are shown for 140 fb$^{-1}$ of 13 TeV proton-proton collisions at the LHC.

Figures (6)

• Figure 1: Cross section of quirk$-$antiquirk production in $pp$ collisions at 13 $\text{Te V}$ as a function of the quirk mass (top). Branching ratio for the diquirk system to decay into two SM jets, depending on the quirk mass and $\Lambda'$ (bottom). The dashed red line indicates where quirk annihilation to two dark glueballs is kinematically forbidden, invalidating the presented simplified estimate. Close to the dashed red line, the quirks dominantly or entirely annihilate to SM jets.
• Figure 2: Reconstruction-level distributions, comparing the shape of the simulated dijet background (black) with the quirk signal (blue), where $m_{q'}$ is chosen to be 1.5 $\text{Te V}$ and $E_\mathrm{frac}$ is varied between 0.2 and 1.0. Also shown are the dijet background distributions estimated by a CATHODE run (pink). The invariant mass of the dijet system is shown on the top, while the other figures show track distributions in the $m_{jj}$ signal window.
• Figure 3: Sensitivity $S/\sqrt{B}$ for different signal strengths $\mu$ after the preselection, presented for $\Lambda'$ = 10 $\text{Ge V}$, different quirk masses and two values of $E_\mathrm{frac}$.
• Figure 4: Significance improvement curves for a quirk mass of 1.5 $\text{Te V}$ as a function of the false positive rate, for the different selection strategies and two values of $E_\mathrm{frac}$. The significance improvement for CATHODE and IAD depends on the signal strength; results for $S/\sqrt{B} = 1$ and 2 at preselection are presented. The uncertainty bands are derived from ten different samplings of the background and signal samples. The vertical line indicates the chosen false positive rate.
• Figure 5: Required signal strengths for a 5$\sigma$ discovery for $\Lambda'$ = 10 $\text{Ge V}$, different quirk masses and different values of $E_\mathrm{frac}$, for different selection strategies (lower is better). For CATHODE and IAD, the uncertainty bands are dominated by the 1$\sigma$ spread of ten trial runs. For the other curves, they correspond to the expected statistical uncertainties on $S$ and $B$ (as the trial spread is negligible). The dashed horizontal line indicates a signal strength of 1.