Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Explaining Data Anomalies over the NMSSM Parameter Space with Deep Learning Techniques

A. Hammad, Raymundo Ramos, Amit Chakraborty, Pyungwon Ko, Stefano Moretti

TL;DR

This work investigates the NMSSM as a compelling BSM framework capable of addressing several experimental anomalies while remaining consistent with current theoretical and experimental constraints. By employing a deep-learning–assisted scanning strategy (DLScanner) over a 12-parameter NMSSM space, the authors identify a golden region where the $2\sigma$–level explanations of the 95 GeV and 650 GeV Higgs-like anomalies, EWino excesses, and the muon $g-2$ measurement co-exist with DM and collider bounds. They further predict accessible mono-$H$ and mono-$Z$ signals at the HL-LHC, detailing a multi-modal DL architecture that fuses graph-based color-flow information with kinematic features to enhance signal–background discrimination. The study demonstrates a concrete path to test the NMSSM explanation of these anomalies and provides open-source tools to reproduce and extend the analysis. Overall, the work highlights DL-assisted parameter scanning as a powerful approach for exploring high-dimensional BSM theories and guiding future collider searches.

Abstract

Motivated by recent results from particle physics analyses, we investigate the Next-to-Minimal Supersymmetric Standard Model (NMSSM) as a framework capable of accommodating a range of current data anomalies across low- and high-energy experiments. These include the so-called 95GeV and 650GeV excesses from Higgs studies, the Electro-Weakino excess from Supersymmetry searches, the latest $(g-2)_μ$ measurements as well as potential deviations from Standard Model (SM) predictions that would appear as a consequence in mono-$H$ (where $H=h_{\rm SM}$) and -$Z$ signatures of Dark Matter. Our analysis demonstrates that viable NMSSM parameter regions exist where all these features can be accommodated at the $2σ$ level while remaining consistent with the most up-to-date theoretical and experimental constraints. To identify such regions, we employ an efficient numerical scanning strategy assisted by deep learning techniques. We further present several benchmark points that realize these scenarios, offering promising directions for future phenomenological studies.

Explaining Data Anomalies over the NMSSM Parameter Space with Deep Learning Techniques

TL;DR

This work investigates the NMSSM as a compelling BSM framework capable of addressing several experimental anomalies while remaining consistent with current theoretical and experimental constraints. By employing a deep-learning–assisted scanning strategy (DLScanner) over a 12-parameter NMSSM space, the authors identify a golden region where the –level explanations of the 95 GeV and 650 GeV Higgs-like anomalies, EWino excesses, and the muon measurement co-exist with DM and collider bounds. They further predict accessible mono- and mono- signals at the HL-LHC, detailing a multi-modal DL architecture that fuses graph-based color-flow information with kinematic features to enhance signal–background discrimination. The study demonstrates a concrete path to test the NMSSM explanation of these anomalies and provides open-source tools to reproduce and extend the analysis. Overall, the work highlights DL-assisted parameter scanning as a powerful approach for exploring high-dimensional BSM theories and guiding future collider searches.

Abstract

Motivated by recent results from particle physics analyses, we investigate the Next-to-Minimal Supersymmetric Standard Model (NMSSM) as a framework capable of accommodating a range of current data anomalies across low- and high-energy experiments. These include the so-called 95GeV and 650GeV excesses from Higgs studies, the Electro-Weakino excess from Supersymmetry searches, the latest measurements as well as potential deviations from Standard Model (SM) predictions that would appear as a consequence in mono- (where ) and - signatures of Dark Matter. Our analysis demonstrates that viable NMSSM parameter regions exist where all these features can be accommodated at the level while remaining consistent with the most up-to-date theoretical and experimental constraints. To identify such regions, we employ an efficient numerical scanning strategy assisted by deep learning techniques. We further present several benchmark points that realize these scenarios, offering promising directions for future phenomenological studies.

Paper Structure

This paper contains 23 sections, 15 equations, 13 figures, 6 tables.

Table of Contents

  1. Introduction
  2. The NMSSM
  3. Parameter Space Scan and Constraints
  4. Setup for DL Scanning of the NMSSM Parameter Space
  5. Theoretical Constraints
  6. Experimental Constraints
  7. Higgs Measurements
  8. DM Measurements
  9. Low Energy Observables
  10. Collider constraints
  11. Anomalies
  12. The 95 GeV Measurement
  13. The EWino Measurement
  14. The 650 GeV Measurement
  15. The $(g-2)_\mu$ Measurement
  16. ...and 8 more sections

Figures (13)

  • Figure 1: Reduced coupling correlations for the SM Higgs boson. The color bar represents the SM Higgs boson mass while the shaded areas represent the allowed region within $2\sigma$ away from the measured central values.
  • Figure 2: Left(Right): SI(SD) measurement versus the mass of the LSP. The color bar represents the relic abundance.
  • Figure 3: Left(Right): NMSSM allowed scan points following $B^\pm \to \tau^\pm\nu$($B\to S\gamma$) constraints. The color bar represents $\tan\beta$ value.
  • Figure 4: Distribution of points in the 2D plane of relevant signal strengths (left) and BRs (right) in order to explain the deviation at 95 GeV. The color gauge represents the $h_{95}$ mass.
  • Figure 5: Left: Expected and observed ATLAS limits with scanned points overlaid in blue, which are demonstrating clear NMSSM compatibility with the EWino anomaly. ATLAS data are extracted from Figure 16 of ATLAS:2021moa. Right: cross-section for chargino--neutralino pair production.
  • ...and 8 more figures