Machine-Learning Analysis of Radiative Decays to Dark Matter at the LHC (2410.13799v2)

Abstract: The search for weakly interacting matter particles (WIMPs) is one of the main objectives of the High Luminosity Large Hadron Collider (HL-LHC). In this work we use Machine-Learning (ML) techniques to explore WIMP radiative decays into a Dark Matter (DM) candidate in a supersymmetric framework. The minimal supersymmetric WIMP sector includes the lightest neutralino that can provide the observed DM relic density through its co-annihilation with the second lightest neutralino and lightest chargino. Moreover, the direct DM detection cross section rates fulfill current experimental bounds and provide discovery targets for the same region of model parameters in which the radiative decay of the second lightest neutralino into a photon and the lightest neutralino is enhanced. This strongly motivates the search for radiatively decaying neutralinos which, however, suffers from strong backgrounds. We investigate the LHC reach in the search for these radiatively decaying particles by means of cut-based and ML methods and estimate its discovery potential in this well-motivated, new physics scenario.

• The paper demonstrates that ML approaches leveraging kinematic correlations significantly improve signal discrimination in radiative decays.
• It compares traditional cut-based methods with ML techniques, achieving enhanced discovery significance (Z ≥ 5) in targeted parameter spaces.
• The findings suggest that ML-enhanced strategies can complement existing electroweakino searches, advancing dark matter discovery efforts.

A Machine Learning Approach to Radiative Decays and Dark Matter Searches at the LHC

The search for weakly interacting massive particles (WIMPs) is a critical aspect of current particle physics research. This paper investigates radiative decays into dark matter candidates within a supersymmetric framework at the High Luminosity Large Hadron Collider (HL-LHC), employing advanced ML techniques. It addresses a specific region in the parameter space of the Minimal Supersymmetric Standard Model (MSSM) where a neutralino, representing the lightest supersymmetric particle (LSP), is a dark matter candidate.

Introduction to the Problem

At the LHC, the quest to identify WIMPs involves understanding two major questions in particle physics: electroweak symmetry breaking and the nature of dark matter. The MSSM provides a natural framework, positing the lightest neutralino as a dark matter candidate. This neutralino achieves the observed dark matter density through co-annihilation with other particles. This work explores new avenues of detection via radiative decays, specifically, the second lightest neutralino decaying into the lightest neutralino and a photon. Traditional searches face challenges due to significant background noise, an issue potentially mitigated by employing ML techniques.

Theoretical Framework

The theoretical basis involves a compressed supersymmetric electroweak sector, focusing on the neutralino as the LSP. In certain regions of parameter space, specifically where there are small mass differences between the neutralinos, radiative decays become significant. The paper thoroughly examines these parameter spaces, identifying scenarios where direct detection and collider searches intersect to provide meaningful constraints on dark matter candidates.

Methodology and Machine Learning Application

The authors employ both traditional cut-based methods and machine learning to distinguish signal from background. The ML approach captures correlations in kinematic variables that cut-based strategies overlook. Specifically, the application of ML methods, including the Binned-Likelihood (BL) and Machine-Learned Likelihood (MLL) methods, enhances the potential discovery significance by providing better delineation of signal against background noise.

1. Event Simulation: Signal and background events are generated through MadGraph5, Pythia8, and Delphes, simulating the detection processes at the LHC.
2. Feature Analysis: The analysis identifies key kinematic variables that contribute most significantly to signal detection, such as transverse momentum and energy signatures.
3. Statistical Methods: Two primary statistical approaches are employed:
   • Binned-Likelihood (BL) Method: Utilizes binned data, reducing event-specific detail but simplifying statistical analysis.
   • Machine-Learned Likelihood (MLL) Method: Employs Kernel Density Estimators to maintain event-level detail without binning, enhancing detection sensitivity.

Results and Implications

The analysis reveals that ML techniques significantly boost detection capabilities compared to traditional methods. The incorporation of ML shows potential discovery significance in unexplored parameter spaces, especially in the region where electroweakinos are expected to decay radiatively:

• Discovery Potential: The paper identifies regions with potential discovery significance ($Z \ge 5$) in mass ranges that have not been exhaustively explored by traditional methods.
• Complementary Searches: These results suggest that ML-based radiative decay searches could complement or enhance existing electroweakino searches that focus on multilepton signatures.

Future Directions

The integration of ML into the search strategies outlined in this paper emphasizes the adaptability of particle physics to incorporate emerging computational techniques. Looking forward, experimental collaborations may adopt these methods to improve search strategies, further probing the challenging parameter spaces of the MSSM.

Conclusion

This research expands the toolkit available for discovering dark matter candidates at the LHC. By leveraging machine learning, the paper opens new perspectives on radiative decays in supersymmetric models, offering a promising path for future exploration and potential discovery of dark matter signatures.