Hunting and identifying coloured resonances in four top events with machine learning

Abstract

We study prospects to search for pair or singly produced colour octet or colour sextet scalars which decay into two top quarks at the LHC. We focus on the same-sign lepton final state. We train a neural network comprising a simple multilayer perceptron combined with a convolutional neural network to optimize the separation of signal and background events. For LHC operated at 14 TeV and a luminosity of 3 ab$^{-1}$ we find an expected discovery reach of $m_8=1.8$ TeV and $m_6=1.92$ TeV for pair produced colour octets and sextets, respectively, and an expected exclusion reach of $m_8=2.02$ TeV and $m_6=2.14$ TeV. In a second step, we retrain the same network architecture to discriminate between signal processes. The network can clearly distinguish between the different colour representations. Moreover, we can also determine whether there is a significant contribution from single production to pair production for the same final state. The methodology can be applied to BSM candidates of different spin and colour representations.