Inclusive, prompt and non-prompt $\rm{J}/\psi$ identification in proton-proton collisions at the Large Hadron Collider using machine learning

TL;DR

This paper introduces machine learning techniques to distinguish prompt and non-prompt J/ψ mesons in proton-proton collisions, enhancing identification accuracy and enabling detailed differential measurements.

Contribution

It presents the first application of machine learning models like XGBoost and LightGBM for separating prompt and non-prompt J/ψ signals in simulated pp collision data.

Findings

Achieved up to 99% prediction accuracy in classification.

Performed differential measurements of J/ψ production modes.

Compared results with experimental data where available.

Abstract

Studies related to $\rm{J}/\psi$ meson, a bound state of charm and anti-charm quarks ($c\bar{c}$), in heavy-ion collisions, provide genuine testing grounds for the theory of strong interaction, quantum chromodynamics (QCD). To better understand the underlying production mechanism, cold nuclear matter effects, and influence from the quark-gluon plasma, baseline measurements are also performed in proton-proton ($pp$) and proton-nucleus ($p$--A) collisions. The inclusive $\rm{J}/\psi$ measurement has contributions from both prompt and non-prompt productions. The prompt $\rm{J}/\psi$ is produced directly from the hadronic interactions or via feed-down from directly produced higher charmonium states, whereas non-prompt $\rm{J}/\psi$ comes from the decay of beauty hadrons. In experiments, $\rm{J}/\psi$ is reconstructed through its electromagnetic decays to lepton pairs, in either $e^{+}+e^{-}$ or $\mu^{+}+\mu^{-}$ decay channels. In this work, for the first time, machine learning techniques are implemented to separate the prompt and non-prompt dimuon pairs from the background to obtain a better identification of the $\rm{J}/\psi$ signal for different production modes. The study has been performed in $pp$ collisions at $\sqrt{s} = 7$ and 13 TeV simulated using PYTHIA8. Machine learning models such as XGBoost and LightGBM are explored. The models could achieve up to 99\% prediction accuracy. The transverse momentum ($p_{\rm T}$) and rapidity ($y$) differential measurements of inclusive, prompt, and non-prompt $\rm{J}/\psi$, its multiplicity dependence, and the $p_{\rm T}$ dependence of fraction of non-prompt $\rm{J}/\psi$ ($f_{\rm B}$) are shown. These results are compared to experimental findings wherever possible.