Tagging ultra-boosted jets at FCC-hh using machine learning techniques

TL;DR

This paper explores machine learning methods like XGBoost and CNNs to identify ultra-boosted $W$ bosons and top quarks at the FCC-hh collider, addressing challenges in high-energy object reconstruction for new physics searches.

Contribution

It introduces the application of advanced ML techniques to distinguish ultra-boosted particles from background in future high-energy collider environments.

Findings

ML techniques effectively differentiate $W$ and top jets from QCD background at high $p_T$

XGBoost and CNNs show promising performance for FCC-hh analyses

Potential to enhance new physics searches at future colliders

Abstract

The Future Circular Hadron Collider (FCC-hh) will probe unprecedented energy regimes, enabling direct searches for new elementary particles at a scale of tens of TeV. FCC-hh is currently in the planning stage, and one of its primary physics goals is to search for physics beyond the Standard Model by exploring a previously inaccessible kinematic domain. While venturing into uncharted high-energy territories promises excitement, reconstructing objects with enormous transverse momenta will require overcoming major experimental challenges. This work investigates the identification of boosted $W$ bosons and boosted top quarks in the context of three beyond the Standard Model scenarios: heavy vector-like quark ($B'$), heavy neutral gauge boson ($Z'$), and heavy neutral Higgs boson ($H$). We employ machine learning techniques, including eXtreme Gradient Boosting (XGBoost) and convolutional neural networks (CNN), to identify these ultra-boosted objects in the collider from their SM background counterpart. We evaluate the performance of these techniques in distinguishing $W$ jets and top jets from QCD jets at extremely high transverse momenta ($p_{T}$) values, demonstrating their potential for future FCC-hh analyses.