Probing Heavy Neutrinos at the LHC from Fat-jet using Machine Learning

We explore the potential to use machine learning methods to search for heavy neutrinos, from their hadronic final states including a fat-jet signal, via the processes $pp \rightarrow W^{\pm *}\rightarrow μ^{\pm} N \rightarrow μ^{\pm} μ^{\mp} W^{\pm} \rightarrow μ^{\pm} μ^{\mp} J$ at hadron colliders. We use either the Gradient Boosted Decision Tree or Multi-Layer Perceptron methods to analyse the observables incorporating the jet substructure information, which is performed at hadron colliders with $\sqrt{s}=$ 13, 27, 100 TeV. It is found that, among the observables, the invariant masses of variable system and the observables from the leptons are the most powerful ones to distinguish the signal from the background. With the help of machine learning techniques, the limits on the active-sterile mixing have been improved by about one magnitude comparing to the cut-based analyses, with $V_{μN}^2 \lesssim 10^{-4}$ for the heavy neutrinos with masses, 100 GeV$~<m_N<~$1 TeV.