Leptoquark Searches at TeV Scale Using Neural Networks at Hadron Collider

Abstract

Several discrepancies in the decay of B-meson decay have drawn a lot of interest in the leptoquarks (LQ), making them an exciting discovery. The current research aims to discover the pair-production of leptoquarks that links strongly to the third generation of quarks and leptons at the center of mass energy $\sqrt{s}$=14 TeV, via proton-proton collisions at the Large Hadron Collider (LHC). Based on the lepton-quark coupling parameters and branching fractions, we separated our search into various benchmark points. The leading order (LO) signals and background processes are generated, while parton showering and hadronization is also performed to simulate the detector effects. The Boosted Decision Trees (BDTs), Multilayer Perceptron (MLP), and Likelihood (LH) methods are effective in improving signal-background discrimination compared to traditional cut-based analysis. The results indicate that these machine learning methods can significantly enhance the sensitivity in probing for new physics signals, such as LQs, at two different integrated luminosities. Specifically, the use of BDTs, MLP, and LH has led to higher signal significances and improved signal efficiency in both hadronic and semi-leptonic decay modes. The results suggest that the LQ masses of 500 GeV and 2.0 TeV in fully hadronic decay modes can be accurately probed with signal significance 176.70 (17.6) and 184.27 (0.01) for MVA (cut-based) at 1000 $fb^{-1}$, respectively. Similarly, in semi-leptonic decay mode the signal significance values are 168.56 and 181.89 at lowest and highest selected LQ masses respectively for MVA method only. The enhanced numbers by a factor of 2 are also reported at 3000 $fb^{-1}$.