Holistic approach and Advanced Color Singlet Identification for physics measurements at high energy frontier

Abstract

To enhance the discovery power of high-energy colliders, we propose a holistic approach and Advanced Color Singlet Identification (ACSI), both of which utilize inclusive reconstructed information as input. The holistic approach is designed to simultaneously classify physics events, while ACSI focuses on associating final-state particles with their parent massive bosons. Implemented using state-of-the-art artificial intelligence architectures and applied to benchmark analyses with simulated data from a future Higgs factory, these new concepts significantly improve the accuracy of H->bb/cc/ss/gg measurements by up to a factor of two to six.

Figures (5)

• Figure 1: Geometry of the AURORA detector (left) and an event display of a reconstructed $e^+e^- \to Z(\to q\bar{q})H(\to b\bar{b})$ event at the center-of-mass energy of 240 GeV (right). Different particles are depicted with colored curves and straight lines: red for $e^{\pm}$, cyan for $\mu^{\pm}$, blue for $\pi^{\pm}$, orange for photons, and magenta for neutral hadrons.
• Figure 2: Model architectures. Left: ParticleNet architecture and the EdgeConv block for the holistic approach. Right: Particle Transformer architecture for ACSI composed of eight particle–attention blocks followed by a multilayer perceptron for particle-level classification.
• Figure 3: ACSI performance. Left: Visualization of particle-level parentage inference with ACSI for a fully hadronic $ZH$ event at 240 GeV. Each circle represents a reconstructed final-state particle, with the area proportional to its energy. Colors indicate the parent assignment inferred by ACSI. Particles whose parent is misidentified are highlighted using open-circle markers. Right: Scaling behavior of ACSI performance as a function of the training data volume.
• Figure 4: Measurement performance of $H \to c\bar{c}$ and $H \to s\bar{s}$ decays with the holistic approach and ACSI. Signal score distributions for $H \to c\bar{c}$ and $H \to s\bar{s}$ in $\nu\bar{\nu} H$ (upper panels) and $q\bar{q} H$ (lower panels) events. For the $\nu\bar{\nu} H$ channel, the optimized cut (vertical dashed line) yields signal strength precisions of 0.72% and 29% for $H \to c\bar{c}$ and $H \to s\bar{s}$, respectively. For the $q\bar{q} H$ channel, using the holistic approach combined with ACSI, the corresponding precisions are 1.03% and 114%.
• Figure 5: Left: Comprehensive summary of signal strength measurements. A comprehensive summary of signal strength measurements of $q\bar{q}H(j\bar{j})$ and $H \to b\bar{b}/c\bar{c}/gg/s\bar{s}$ in $\nu\bar{\nu}H$ and $q\bar{q}H$ channels. Results for the $H \to b\bar{b}$, $c\bar{c}$, and $gg$ using cut-based and BDT methods, which include a more comprehensive set of backgrounds, are taken from Ref. Zhu:2022lzv, while the result for $H \to s\bar{s}$ with cut-based and BDT is taken from Ref. PhysRevLett.132.221802. Right: Theoretical uncertainty from hadronization models. The $H \to b\bar{b}/c\bar{c}/gg/s\bar{s}$ measurement in $\nu\bar{\nu}H$ channel with different hadronization models, where P6, P8, and H7 represent Pythia-6.4 Pythia6, Pythia-8.313 Bierlich:2022pfr, and Herwig-7.2.2 Bahr:2008pvBellm:2015jjp, respectively.