Learning to Reconstruct Quirky Tracks

TL;DR

This work tackles the limited sensitivity of traditional tracking to non-helical signatures by applying a graph neural network–based tracker (Exa.TrkX) to reconstruct oscillatory quirk tracks in LHC-like detectors. Quirk dynamics are modeled with a central infracolor force and external magnetic field, and training uses time-ordered simulation data to define true edges, enabling the pipeline to find non-standard trajectories without bespoke algorithms. Across 8- and 25-layer detector geometries, the method achieves 10–90% efficiency depending on quirk parameters, dramatically outperforming SM-based trackers in the cm-scale regime, and remains reasonably robust to background and hit-resolution uncertainties. The results suggest a path toward general, model-agnostic non-helical track reconstruction at the LHC, with implications for broader searches and potential real-time triggering; future work includes pile-up studies and extending generalization to broader classes of unusual tracks.

Abstract

Analysis of data from particle physics experiments traditionally sacrifices some sensitivity to new particles for the sake of practical computability, effectively ignoring some potentially striking signatures. However, recent advances in ML-based tracking allow for new inroads into previously inaccessible territory, such as reconstruction of tracks which do not follow helical trajectories. This paper presents a demonstration of the capacity of ML-based tracking to reconstruct the oscillating trajectories of quirks. The technique used is not specific to quirks, and opens the door to a program of searching for many kinds of non-standard tracks.

Figures (9)

• Figure 1: Example trajectories of quirk pairs in simulated events with a simplified detector geometry described in the text. The top row shows example trajectories for $\Lambda$ = 100 eV, and increases in mass from left to right, with $m_Q = 500,1000,5000$ GeV, all with small oscillation distances. The center (bottom) row shows examples for the same mass range but for $\Lambda = 500 (1000)$ eV, both with visible oscillations. Hits from SM background tracks (see Sec \ref{['sec:qbg']} for details) are in grey, and from quirks are in blue. The true quirk trajectory is also shown, in red.
• Figure 2: Distributions of kinematic quantities for simulated quirk pairs, for several mass values, independent of $\Lambda$. Shown are the $\gamma$ factor, transverse momentum $p_\textrm{T}$ and opening angle $\Delta \phi$.
• Figure 3: Example truth and reconstructed well-behaved quirk tracks with $(m_Q=500$ GeV, $\Lambda=500$ eV).
• Figure 4: Examples of truth and reconstructed tracks for not well-behaved quirks with $m_Q=500$ GeV, $\Lambda=500$ eV.
• Figure 5: Tracking performance for quirks without background for $m_Q=500$ GeV, $\Lambda=500$ eV in the eight-layer geometry. Shown is efficiency versus track $p_\textrm{T}$ (left), quirk opening angle (center), and number of true hits (right) for well-behaved quirks or all quirks. The error bars indicate statistical errors.