Enhancing di-jet resonance searches via a final-state radiation jet tagging algorithm

TL;DR

The paper addresses the challenge of di-jet resonance searches being limited by poor mass resolution and background sculpting due to final-state radiation. It introduces an event-level deep neural network that labels jets as sig-isr, sig-fsr, bkg-isr, or bkg-fsr using only the leading three jets, and uses the identified sig-fsr jet to replace the di-jet mass with the tri-jet invariant mass when appropriate. The approach yields about 40% efficiency in correctly tagging FSR jets with roughly 20% misidentification for other categories, and provides a 10–14% improvement in the maximum signal significance across a broad mass range (1.5–3 TeV for $m_{Y_0}$), while preserving smooth background shapes. The method demonstrates a practical, scalable enhancement for di-jet searches at the LHC and HL-LHC, with potential extensions to other hadronic final-state analyses and to alternative jet-clustering configurations.

Abstract

In this article, we investigate the possibility of enhancing the di-jet resonance searches by tagging the final state radiation (FSR) jet, using an event-level deep neural network. It is found that solely relying on the 4-momenta of the leading three jets allows the algorithm to achieve good discriminating power that can identify the hardest FSR jet in signal, while rejecting other soft jets. Once the invariant mass is corrected with the tagged FSR jet, the mass resolution of the signal is greatly enhanced, and the sensitivity of the search is also improved by more than 10%. By crafting the input variables carefully, the algorithm introduces minimal mass sculpting for the background, and its applicability extends to a broad mass range. This work proves that FSR jet tagging can potentially enhance the di-jet resonance searches, suiting various stages of the physics programmes at the Large Hadron Collider (LHC) and High-Luminosity LHC (HL-LHC).

Figures (16)

• Figure 1: Feynman diagrams for a heavy $Y_0$ particle production in $s$-channel with an ISR gluon (left) and an FSR gluon (right). The $Y_0$ particle is a spin-0 mediator in the simplified dark matter model dm_simpdm_simp1dm_simp2.
• Figure 2: Comparison of the $Y_0$ mass reconstructed using the leading two jets (shaded area), the leading three jets (dotted-dashed line) and the leading four jets (dashed line), with the ISR showering switch turned off (left) and on (right). The FSR showering switch is turned on for both. The vertical line indicates the actual $Y_0$ mass (1.5 TeV).
• Figure 3: Selected kinematic distributions of the third jet for the $m_{Y_0\xspace}$ = 1.5 $~\text{TeV}$ and 3 $~\text{TeV}$ samples. The third jets taken from the showering control samples with the FSR/ISR showering switch turned on/off and off/on, are the FSR and ISR jets, respectively. Four quantities are shown: the third jet $\eta$ (upper left), $\Delta \phi$ between the third jet and the (sub-)leading jet (upper right), ratio of the third jet mass (lower left) and $p_{\mathrm{T}}$ (lower right) to the leading jet $p_{\mathrm{T}}$.
• Figure 4: Selected kinematic distributions of the third jet for the $m_{Y_0\xspace}$ = 1.5 $~\text{TeV}$ signal and multi-jet background. The third jets taken from the showering control samples with the FSR/ISR showering switch turned on/off and off/on, are the FSR and ISR jets, respectively. Three quantities are shown: $\Delta \phi$ between the third jet and the (sub-)leading jet (left), ratio of the third jet mass (middle) and $p_{\mathrm{T}}$ (right) to the leading jet $p_{\mathrm{T}}$.
• Figure 5: The number of FSR particles (dotted-dashed line) and ISR particles (solid line) associated with the third jet (left). The ratio of the number of FSR particles (dotted-dashed line) and ISR particles (solid line) to the total number of particles, including those not from ISR/FSR, associated with the third jet (right). The nominal $m_{Y_0\xspace}$ = 1.5 $~\text{TeV}$ signal (dark orange) and multi-jet background (light grey) samples are used.