Semileptonic ttH(Hbb) studies at UCL Lily Asquith, Catrin Bernius, Sebastian Boeser, Simon Dean, Nikos Konstantinidis Atlas UK Meeting, Higgs Session 11/01/2008
Atlas UK Meeting, Higgs Session Cut-Based Analysis Preselection 1 isolated lepton, pTe>25GeV, pTμ>20GeV, |η|<2.5 At least 6 jets, pT>15GeV, |η|<5 At least 4 tagged as b-jets Wl constrain invariant mass of l to W mass, solve for pzv Wjj Pairs of non b-jets with mjj=mW±25GeV Reconstruction of 2 top quarks Minimise 2 with top masses Reconstructed top masses must be within ±20GeV of mt Reconstruction of Higgs Use the two highest-pT remaining b-jets Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session Studies Shown Here Jets in the crack region Background invariant mass shape Fast simulation jets Generator samples Simon J.H. Dean Atlas UK Meeting, Higgs Session
Jets in the Crack Region (Lily Asquith) Uninstrumented angular region = "crack" Considered 1.4 < || < 1.6 Hadronic tile calorimeters EM accordion calorimeters Hadronic LAr endcap calorimeters Forward LAr calorimeters Simon J.H. Dean Atlas UK Meeting, Higgs Session
Jets in the Crack Region Abundance of jets in ttH final state Poor jet reconstruction may seriously affect overall event reconstruction Study uses cone 0.4 jets in release 12.0.6 ttH and ttbb samples (5870, 5871 & 5855) Matched to truth-jets using dR < 0.2 In ||-bins of 0.2, Gaussian fits made to ET(reco) / ET (true) Linearity = mean Resolution = width / mean Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session Light Jet Performance Low and high energy jet samples ~4% linearity dip in crack region Resolution worse at lower pT Deviations at high eta due to low fit stats Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session B-Jet Performance No b-tagging outside Inner Detector (|| < 2.5) ~4% dip again observed in linearity Resolution also worse at lower pT More work to show impact on hadronic W, tops and Higgs Simon J.H. Dean Atlas UK Meeting, Higgs Session
Background Shape (Catrin Bernius) Sinead M. Farrington, "B Physics Lessons from the Tevatron", YETI'08 signal J. Drohan Broad bb mass peak Kinematic selection produces similar signal and background shape Would like better discrimination Started by considering ttH vs. ttbb (QCD) Simon J.H. Dean Atlas UK Meeting, Higgs Session
Background Shape Formulated "tight selection" on Higgs b-jets at low pt more jets in ttbb than ttH Tight selection: pT(b1Higgs)> 40 GeV, pT(b2Higgs)> 30 GeV Simon J.H. Dean Atlas UK Meeting, Higgs Session
Tight Selection ET(b2higgs) vs ET(b1higgs) in mass window 90GeV<mbb<150GeV to check cut correlation Cut out Tight selection: pT(b1Higgs)> 40 GeV, pT(b2Higgs)> 30 GeV Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session Shape Comparison Normalised to 30 fb-1 ("combined sample") Simon J.H. Dean Atlas UK Meeting, Higgs Session
Correct/Incorrect Combinations in ttH Signal Correct combinations: b-jets (from Higgs) correctly matched Fit Gaussian to correct combinations, Landau to incorrect combinations Simon J.H. Dean Atlas UK Meeting, Higgs Session
Fitting Background & Signal shapes Fit gaussian + landau to ttH signal sample Landau also describes ttbb background very well Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session Combined Fit Fit background + signal shape to the combined sample, keeping parameters fixed but with scaling factors applied Standard Selection Tight Selection Mass window: 90-150 GeV: signal (gauss+landau) = 437.9 ttbb(QCD) (landau) = 3089.1 --------------------------------------------- signal/(signal+ttbb) = 0.12 signal(gauss)/(all signal) = 0.28 Mass window: 90-150 GeV: signal (gauss+landau) = 377.9 ttbb(QCD) (landau) = 1923.2 -------------------------------------------- signal/(signal+ttbb) = 0.15 signal(gauss)/(all signal) = 0.31 Simon J.H. Dean Atlas UK Meeting, Higgs Session
Next Steps For Shape Study Include ttbb(EW) and ttjj backgrounds Explore behaviour of combined fit when varying analysis cuts eg... b-weight cut at different values Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session Fast Simulation Jets Fast simulation uses different cone jet finder than accessed by JetRec Atlfast currently only supplies the results of cone 0.4 jet finder in AOD output Can also run using kT algorithm Done via KtEvent interface, same as JetRec In principle, re-running Atlfast using kT and comparing with full-chain kT jets should give closer agreement Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session 12.0.6 cone jets Full Fast 7000 signal events considered Cone 0.4 Included for both full and fast in 12.0.6 AOD files Noticeably higher number of jets from Atlfast Overestimation of jet efficiency at low pT, as seen in other analyses Consistent with Bill Murray (RAL) Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session 12.0.6 kT jets Full Fast Full chain kT jets included in 12.0.6 AODs, fast not Atlfast run directly from evgen files used in standard production R-parameter = 0.4 Shift to higher pT in fast Cone jets should be more consistent from r14 as FastKtJet interface will be adopted by Atlfast (deliverable) Simon J.H. Dean Atlas UK Meeting, Higgs Session
Generator Studies (Sebastian Boeser) ttHTruthCollector tool used to inspect signal and background truth samples identify the hard scatter vertex, i.e. the one with either an outgoing higgs or two outgoing tops identify any additional b-partons at the hard scatter vertex follow the top-branches to identify the b-partons and W's follow the W-branches to identify the jets or lepton-neutrino pairs Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session tWs decays 38950 ttH events Missing b-quarks from top decays compensated by finding s-quarks The relative number of tWs/tWb is in good agreement with |Vts|2 as given by the PDG tWs decays supported in PYTHIA sample, neglected in other generators Simon J.H. Dean Atlas UK Meeting, Higgs Session
Atlas UK Meeting, Higgs Session W decays 5000 ttjj events 7.4% missing qq pairs from W decays Equally high number of -pairs is found in their place. W being accepted as "hadronic" decay in the semi-leptonic sample. (In some cases there are extra b-quarks at the hard-scatter vertex. Happens if the hard scatter is connected to the underlying event with a b-quark) Simon J.H. Dean Atlas UK Meeting, Higgs Session
Next Steps in Generators Study Understand effect on analysis of additional W decays Understand correct way of dealing with ttbb component in ttjj sample, to avoid double-counting events (not covered here) Simon J.H. Dean Atlas UK Meeting, Higgs Session