Jet Substructure Probe on Scalar Leptoquark Models via Top Polarization

TL;DR

This paper investigates scalar leptoquark models using top polarization and jet substructure techniques at the HL-LHC, achieving significant signal detection and model discrimination.

Contribution

It introduces a novel analysis combining jet substructure and multivariate methods to distinguish scalar leptoquark models via top polarization at high-energy colliders.

Findings

Achieves up to 5.3 sigma signal significance for 1250 GeV leptoquarks.

Demonstrates effective discrimination between $S_3$ and $R_2$ models with 3.2 sigma separation.

Outperforms traditional variables in model distinction using BDT classifier scores.

Abstract

The study of leptoquarks and their couplings to fermions with different chiralities provides a powerful tool for distinguishing among different leptoquark models. As a case study, we focus on two specific third-generation scalar leptoquark models, $S_3$ and $R_2$, which differ in their electroweak quantum numbers and chiral structures of couplings to the top quark, leading to distinct top-quark polarization states. To enhance the efficacy of the analysis, we employ jet substructure techniques like Soft Drop, $N$-subjettiness, and our custom $b$-tagging method, along with other event variables. The analysis has been performed using both fixed radius and dynamic radius jet clustering algorithms. A multivariate analysis using a boosted decision tree (BDT) is performed to isolate signal from the Standard Model background. For a leptoquark mass of 1250 GeV, the analysis achieves a signal significance of up to $5.3\,\sigma$ at the 14 TeV HL-LHC. Furthermore, a $CL_s$-based profile likelihood estimator is applied to polarization-sensitive variables to discriminate between the two models. To enhance separation between the two models, an additional BDT classifier score is obtained by training a BDT network to distinguish between the $S_3$ and $R_2$ models. In the chosen signal region, the BDT classifier score provides a separation score of up to $3.2\,\sigma$, outperforming traditional variables such as $E_b/E_t$ and $\cos\theta_b$.