Reconstruction of boosted and resolved multi-Higgs-boson events with symmetry-preserving attention networks (2412.03819v2)

Abstract: The production of multiple Higgs bosons at the CERN LHC provides a direct way to measure the trilinear and quartic Higgs self-interaction strengths as well as potential access to beyond the standard model effects that can enhance production at large transverse momentum $p_{\mathrm{T}}$. The largest event fraction arises from the fully hadronic final state in which every Higgs boson decays to a bottom quark-antiquark pair ($b\bar{b}$). This introduces a combinatorial challenge known as the \emph{jet assignment problem}: assigning jets to sets representing Higgs boson candidates. Symmetry-preserving attention networks (SPA-Nets) have been been developed to address this challenge. However, the complexity of jet assignment increases when simultaneously considering both $H\rightarrow b\bar{b}$ reconstruction possibilities, i.e., two "resolved" small-radius jets each containing a shower initiated by a $b$-quark or one "boosted" large-radius jet containing a merged shower initiated by a $b\bar{b}$ pair. The latter improves the reconstruction efficiency at high $p_{\mathrm{T}}$. In this work, we introduce a generalization to the SPA-Net approach to simultaneously consider both boosted and resolved reconstruction possibilities and unambiguously interpret an event as "fully resolved'', "fully boosted", or in between. We report the performance of baseline methods, the original SPA-Net approach, and our generalized version on nonresonant $HH$ and $HHH$ production at the LHC. Considering both boosted and resolved topologies, our SPA-Net approach increases the Higgs boson reconstruction purity by 57--62\% and the efficiency by 23--38\% compared to the baseline method depending on the final state.

• The paper introduces a SPA-Net framework that integrates both boosted and resolved jet topologies for multi-Higgs-boson event reconstruction.
• The method achieved a 57–62% improvement in purity and a 23–38% increase in efficiency for nonresonant HH and HHH productions compared to baseline techniques.
• This advancement enhances LHC experimental sensitivity and paves the way for hybrid machine learning approaches in particle physics.

Enhanced Reconstruction of Multi-Higgs-Boson Events Using Symmetry-Preserving Attention Networks

The paper "Reconstruction of boosted and resolved multi-Higgs-boson events with symmetry-preserving attention networks" addresses a significant challenge in the experimental high-energy physics domain: optimizing the reconstruction of events involving multiple Higgs bosons at the Large Hadron Collider (LHC). This research is particularly focused on enhancing the detection and reconstruction capabilities of decay events where Higgs bosons decay into bottom quark-antiquark pairs (bb̅), a dominant decay mode. Such events are critical for probing various aspects of Higgs boson self-interactions and potential extensions beyond the Standard Model.

Overview and Methodology

The paper introduces a sophisticated approach using Symmetry-Preserving Attention Networks (SPA-Nets) to tackle the inherent jet assignment problem in multi-Higgs productions. The problem lies in efficiently associating jets to their source Higgs bosons, which is exacerbated by the presence of multiple possible event topologies. The approach posits two primary configurations: "resolved" jets, characterized by distinct small-radius jet pairs, and "boosted" jets, represented by a single large-radius jet.

The authors have generalized the SPA-Nets framework to simultaneously handle both resolved and boosted topologies within the same model, thus improving the flexibility and accuracy of event interpretation as either fully resolved, fully boosted, or an intermediary state. This adaptation is necessary to exploit the complex kinematic regimes present in Higgs boson pair (HH) and triple (HHH) productions. By incorporating both topological approaches, SPA-Nets can enhance the precision of the jet assignments and ultimately improve reconstruction purity and efficiency.

Key Results and Implications

The generalized SPA-Net model demonstrated a marked improvement in the performance metrics over traditional baseline methods. Specifically, the proposed architecture achieved an increase in reconstruction purity by 57–62% and efficiency by 23–38% for nonresonant HH and HHH productions, compared to the baseline depending on the final state.

These findings not only underscore the advancement in reconstruction performance but also highlight the potential to significantly enhance the sensitivity of experimental analyses at the LHC. Improved event reconstruction in multi-Higgs scenarios directly benefits the precision measurement of the Higgs trilinear and quartic couplings, which are essential probes of the electroweak symmetry breaking mechanism and potential new physics phenomena.

Future Prospects and Theoretical Implications

The successful integration of machine learning, particularly via SPA-Nets, into high-energy physics experiments signifies a broader trend towards leveraging advanced computational techniques to address complex data challenges. The ability to accurately discern event topologies and improve jet tagging is crucial for forthcoming experiments, particularly in high-density data environments expected at future LHC runs or planned future colliders.

Moreover, the research could stimulate further explorations into hybrid machine learning architectures for particle physics, leading to more generalized frameworks capable of autonomously adapting to diverse experimental conditions and datasets. As analysis complexity continues to grow, the extension of such methods to other event types or facilities could catalyze more refined explorations into the fundamental forces governing particle interactions.

In conclusion, the paper offers a substantive contribution to the field by presenting a novel application of neural network architectures, showing how symmetry-preserving networks can be effectively used to solve the intricate challenges associated with Higgs boson event reconstruction. This advancement sets a new benchmark for future explorations in both theoretical and practical implementations within particle physics.