Search for light charged Higgs bosons via t->H+ b, with H+ -> cb in lepton+jets final state at center-of-mass energy 13 TeV with the ATLAS detector

Abstract

An analysis for light charged Higgs bosons in $t\to bH^+$ decays, with $H^+$ decays to $c\bar{b}$ is presented. The analysis is focused on searches for top quark pair production in which one top quark decays to $Wb$, with the $W$ boson decaying leptonically, and the other top quark decays to $H^{+}b$. This analysis is based on proton-proton collision data at $\sqrt{s}= 13$ TeV with an integrated luminosity at 139 $\mathrm{fb}^{-1}$ recorded in Run 2 with ATLAS detector at the Large Hadron Collider. Data are analysed in the lepton-plus-jets final state, characterised by an isolated electron or muon and at least four jets.

A fully connected neural network designed from Keras framework is used to build the final discriminant. Hyper-parameters including number of hidden layers, number of nodes in each hidden layer, learning rate, dropout fraction, and batch size are set and optimised using a random search method within a self-defined hyper-parameter space. The neural network classifiers are parameterised as a function of the Higgs boson masses for light charged Higgs signals in order to improve the performance of isolated classifiers by increasing statistics.

Fits are performed on the neural network discriminant across all fit regions by maximising a likelihood function to correct simulation predictions and reduce the impact of systematic uncertainties. The observed (expected) 95\% confident level upper limits are set on $t\to H^+b$ branching ratio as a function of the assumed mass of the charged Higgs boson ($m_{H^{\pm}}$). Observed (expected) exclusion limits of the branching ratio range from 0.15% (0.09%) to 0.42% (0.24%). The largest excess in data is observed at $m_{H^{\pm}} = 130$ GeV with an about 3 $\sigma$ significance.