A Measurement of the Recently Discovered Higgs Boson in the Decay into Two Photons with Associated Jets, Using the ATLAS Detector at the LHC

A new boson has been discovered and measurements are under way using the 7 TeV and 8 TeV proton-proton collision data from the Large Hadron Collider to determine whether or not this is the Higgs boson as predicted by the Standard Model of Particle Physics (SM). Experimentally measuring the nature of this particle’s couplings to other particles will help determine this. The Standard Model Higgs boson is expected to be produced by a variety of production mechanisms. The SM prediction is that the gluon-gluon fusion (ggF) and vector boson fusion (VBF) production mechanisms are the two production processes with the highest and second-highest, rates respectively. This thesis concentrates on the study of the Higgs boson via its decay into two photons, which was one of the key discovery channels. Part of this analysis is to measure the ratio (R) of these rates using 13 fb−1 of ATLAS √ s = 8 TeV proton-proton collision data and determine if R is consistent with the SM prediction. Using the diphoton decay channel, events were selected to form a category of data events which is enriched in VBF events with little gluon-gluon fusion contamination. The selection procedure was optimised using a boosted decision tree (BDT) multivariate classifier. The distinguishing feature of this analysis was that the BDT was trained using background events from the data sample, so as to reduce the dependency on the modelling of the background processes. It was shown that using a BDT classifier, the VBF signal significance improves by 24.0% relative to the standard cut-based analysis and suffers from 12.0% less ggF signal contamination. Using this event classification R was measured as R = σV BF/(σggF +σV BF) = 0.037±0.067(stat)±0.035(syst) where σV BF and σggF are the respective cross sections of the vector boson fusion process and the gluon-gluon fusion process. The SM prediction is R = 0.075 Although the uncertainty on the current measurement is large, it is shown using pseudodata, that this choice of categorisation will help reduce the uncertainty on R when more data are available.