Search for a Charged Higgs from Top Decays

Jenna Lane, Fred Loebinger, Paul S. Miyagawa, Jo Pater, Un-ki Yang

  1. Phase I analysis: A study of the simplifed top mass fitter

  2. Phase II analysis: Implementation of the real top mass fitter

  3. Phase III analysis: Implementation of ATLAS specific jet corrections and di-jet mass fitter

  4. Phase IV analysis: Maximum likelihood fitting

  5. Phase V analysis: Branching ratio calculations



  1. A study of the simplified top mass fitter (see talk)

    Reconstruction of semi-leptonic ttbar events, with the aim of reconstructing the top and W masses.
    A simplified chi squared fitter was used to select the best reconstructed ttbar pair for each event.

    1. Top mass with / without chi squared cut

    2. W mass with / without chi squared cut

    3. Other plots: Neutrino pz solutions , top pT with / without chi squared cut, W pT with / without chi squared cut, ttbar pT with / without chi squared cut.


  2. Implementation of the real topmass fitter into main analysis code (see talk)

    A multi-parameter fitting tool was used to fit a more complex chi squared function to the data.
    In the fitting process, measured momenta etc. were allowed to be varied within their experimental resolutions.

    1. Top mass with / without chi squared cut

    2. W mass with / without chi squared cut


  3. Implementation of ATLAS-specific jet corrections and di-jet mass fitter

    The light jet correction, CalibMapJetRaw, provided by the ATLAS Top working group was applied. (See Wiki page)
    The Standard Model di-jet mass distribution is shown before / after the corrections were applied.
    The fitter was modified, removing the W mass constraint, and introducing a top mass constraint of 175 GeV. This allows the dijet mass to be fitted as a free parameter. (See plots). Interesting issues:

    1. Different distributions for Hadronic W from truth, and Hadronic W found by adding the two true light quarks from the W decay. These should be identical, but the light quark distribution has a tail on the low mass side. Need to know how light quarks are defined in truth, i.e. from the hard scatter, or after FSR.
    2. Running the fitter in its current form makes the W mass distribution worse.

    Reconsider the top mass fitter to find which components are constributing most to the chi squared. Considering events which pass the standard cuts, have at least one b tag and exactly 2 jets matched to the 2 true light quarks within delta R = 0.2.

    1. With ATLAS light jet corrections applied, and no CDF top specific corrections (Plots)
    2. Without ATLAS light jet corrections, CDF top specific corrections are applied (Plots)

    Dominant chi squared contributions appear to be from jets and lepton.

    Study of the number of jets in each event. Plots show the number of jets passing Pt and eta cuts in events containing either 1 or 2 btags (Plots)
    Current Pt cut of 20GeV seems suitable to select events where 4 jets pass the cuts.

    W mass distribution before and after fitting, when all 4 leading jets are matched to the correct quarks and top width set to 1.5Gev (Plots)

    W mass distribution before and after fitting, when all 4 leading jets are matched to the correct quarks and top width set to 25Gev (Plots)

    Event Record

    Study of matching. Apply lepton cut ( 1 tight lepton Pt > 20GeV, |eta| < 2.5), MET cut ( > 20GeV), jet cut ( 4 jets with Pt > 20 GeV |eta| < 2.5, 1 or 2 btags). Require the event to have 4 hard scattered quarks ( 2 b from top, 2 light quarks from W). All quarks must have Pt > 30 GeV, |eta| < 2.5. Jets used for matching have Pt>10GeV, no eta cut. (Plots)

  4. Implementation of CDF maximum likelihood fitting code into the ATLAS framework

    A binned maximum likelihood fit is used to extract the higgs mass from the SM ttbar plus higgs di-jet mass distribution
    Until the fitter is working correctly, this is being tested on the un-fitted di-jet mass distribution. Require events with exactly 2 btags in the 4 leading jets. Di-jet is assumed to be the 2 un-tagged leading jets.

    Di-jet mass distributions selected by fitter btag info

    1. Test the ML fitter by generating a set of pseudodata. Take the di-jet mass distribution as a template, with Nhiggs, Nw For each pseudodata sample, smear each bin content randomly according to a Poisson distribution.

    2. Extract the number of H+, W using the ML fitter. The fitter provides an error on the calculated numbers. Note, for ML fitting, need to specify the errors in the MINUIT fit, as the default setting does not apply here.

    3. Plot the pull distribution (Nhiggs-NpseudodataH)/Fitter error. This should peak at zero

    4. Find the pull width, i.e. the width (sigma)of the distribution (Nhiggs-NpseudodataH)

    5. Check that the fitter errors are not biassed by considering sigma/Fitter error. This should be a gaussian with a peak at 1 if the fitter is estimating the errors correctly. Plots

    Next step is to introduce a non-ttbar background. i.e. W + 4 jets. Have a sample of 14200 events (W->e/mu/tau + nu). As for the H+ and W case, run the analysis code, taking the di-jet to be the two jets that the fitter selects as coming from the W. But work with the un-fitted distribution until the fitter is working properly. Non-ttbar distribution Find that only a small number of events pass the event selection cuts. This plot requires either 1 or 2 btagged jets. Requiring 2 tagged jets resulted in only 17 events passing the cuts

    Find that the number of W + Jets events passing the event selection cuts is much lower than for ttH or ttW (0.8% compared with 5.5%). The biggest difference is the number of events passing the jet cut.(Table). Events reaching the jet cut fail mainly because jets are not btagged (Plots).



Manchester Charged Higgs webpage

Manchester ATLAS webpage

Latest tex file for csc note