Dilepton tth Chris Hill, Joe Incandela, Sue Ann Koay UCSB 17-Nov-2005
A little Introduction UCSB currently heavily involved in Tracker construction However, increasingly becoming interested in physics analysis tth group at UCSB J. Incandela (Prof - 25%), C. Hill (Postdoc - 50%), S.A. Koay (PhD student - 100%) Familiar w/ tth signal Hill, Incandela (with UCSB colleague D. Stuart) authored Phys.Rev.Lett.86:1694-1697,2001 which explored signal at Tevatron if enough luminosity were delivered We had ideas about extracting signal w/o reconstructing the top quarks Curious whether these will work at CMS
Why Dilepton tth? As you know, detection of a low mass Higgs will be difficult tth is an important complement to other modes (e.g. H γγ) But, as you know, tth is not an easy signal Dilepton mode has small branching fraction (4/81 for e,µ) but may help Moreover, if one did not have to reconstruct ttbar it could help even more Somewhat simpler dilepton events (e.g all jets will be b-jets) are the best place to test our kinematic signal separation ideas We are also familiar with top dileptons having performed recent measurements in this channel @ CDF (Phys.Rev.Lett.93:142001,2004) Finally, Susanna mentioned this mode was not yet studied in the PTDR We hope to contribute to the draft if there is time
Signal Sample Using the CMSUAF @ FNAL About 75,000 Pythia tth events generated with a Higgs mass of 115 with CMKIN 4.3.1 We forced W->lν (e,µ only) These events were fully simulated with OSCAR 3.6.5 and reconstructed with ORCA 8.7.4 The analysis was done using the ExRootAnalysis framework
Dilepton tth Selection Events are top dilepton events with two additional bjets from H->bb Thus, as a starting point, we modeled our event selection criteria after the CMS top dilepton PTDR analysis (talk by J. Cuevas at the last CMS week) Two OS leptons (e, µ only) pt > 20, η < 2.4 At least 2 jets with ET > 20, η < 2.5 At least one of which is b-tagged MET > 40
Dilepton Validation Using the same sample (jm03b_ttbar_leptonic from FNAL) of 10 6 tt events with W->lν forced, as J. Cuevas we reproduced his results to < 1%. We then adapted our selection to tth by simply requiring two additional b-tagged jets with ET > 20 and η < 2.5 (After this, however, we uncovered some aspects of their dilepton analysis which didn t make sense to us which we no longer implement)
Muon Selection Muon transverse momentum PT > 20. GeV η < 2.4 Number of events 14000 12000 10000 8000 all reconstructed µ generator µ from W tracks Track Isolation w/in a cone of R=0.3 Should we add calorimeter isolation? p T p µ T < 2.0 0 0 50 100 150 200 250 P T (GeV) Muon track isolation Arbitrary units 6000 4000 2000 0.22 0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 all reconstructed muons 0-1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 Σ [P (R < 0.3 tracks)] - P (µ) T T
Electron ID Electron transverse momentum PT > 20. GeV η < 2.5 Number of events 9000 8000 7000 6000 5000 4000 all reconstructed e generator e from W E/P < 3.0 3000 2000 Ehad/Eem < 0.05 1000 0 0 50 100 150 200 250 Electron track isolation P T (GeV) tracks For now, tried to use a similar track Isolation p T p e T < 2.0 Arbitrary units 0.25 0.2 0.15 0.1 0.05 all reconstructed electrons 0-2 -1 0 1 2 3 4 Σ [P (R < 0.3 tracks)] - P (e) T T
Lepton Questions We have so far only used track isolation, would like to add calorimeter isolation - is there a standard method? In tth analyses we have seen presented at CMS, lepton ID based on a combination of several variables into a likelihood. Presumably, this allows for increased reconstruction efficiency - if so, are these likelihood algorithms documented and/or publicly available?
Missing ET We start with EcalPlusHcalTower in ExRootAnalysis We correct the MET for jet calibrations We correct for the momentum of any identified muons ET miss (e cal + h cal ) Σ(ET calib Require Corrected MET > 40 ET raw ) Σp µ T Arbitrary units 0.035 0.03 0.025 0.02 0.015 0.01 0.005 Missing transverse energy Arbitrary units 0.022 0.02 0.018 0.016 0.014 0.012 0.01 0.008 0.006 0.004 0.002 (e + H tower MET) cal cal - Σ jet correction - Σ 0 0 50 100 150 200 250 MET (GeV) MET(reconstructed) / MET(generator ν) - 1 tower MET + H cal e cal (e + H tower MET) cal cal - Σ jet correction - Σ 0-1 -0.5 0 0.5 1 1.5 2 E µ E µ
Jets Iterative cone of 0.5 (cone size s/b fine since our signal has only 4 jets) ET > 20 b-tagged jet pseudorapidity Number of events 5000 4000 3000 2000 1000 all b-tagged gen. b η < 2.5 0-3 -2-1 0 1 2 3 η R between jets and electrons b-jets identified w/ combined b-tag algorithm Arbitrary units -1 10 jet and closest electron Remove electrons which match with R < 0.4-2 10 R = φ 2 + η 2-3 10 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 R
Jet Calibrations We were not which calibrations to use Tried γ-jet, and MC calibrations Did not see much difference between the two in resolution P T (b-tagged jet)/p (generator b) - 1 T Arbitrary units 0.07 0.06 0.05 0.04 0.03 0.02 0.01 selected raw IC 0.5 jets within R < 0.4 selected b-tagged γ-jet calibrated jets selected b-tagged MonteCarlo calibrated jets 0-1 -0.5 0 0.5 1 1.5 2 We used γ-jet for now
Summary of Event Selection Criteria Two oppositely charged leptons (e,mu) passing all cuts Missing Transverse Energy > 40 GeV At least four IC 0.5 jets passing all cuts At least three of which have been b- tagged
Estimated Signal Yields σtth 500 fb BR = 4/81 3127 events selected out of 76000 703 ee, 899 µµ, 1525 eµ ε = 4.1% Yield @ 30 fb -1 = 30.3 events
Backgrounds So far have only used Pythia (LO) jm03b_ttbar_leptonic (e,µ,τ) as a source of ttjj background events Simply because only high statistic dileptonic sample we could find at FNAL Intend to replace with CompHep+Pythia ttjj, ttbb samples that Susanna has pointed us to Must get relevant GRID permissions etc, first! Have not considered other background sources, physics (e.g. ttz) or instrumental, yet Susanna has also pointed us to ttz, bbz samples on the GRID so we will analyze these soon as well Would also like to look into rates for instrumental backgrounds - e.g. lepton+jet ttjj where W->cs, charm jet is tagged, and jet fakes a lepton
ttjj background yields σttjj 500,000 fb BR = 4/81 156 events selected out of 957959 33 ee, 30 µµ, 67 eµ, 26 (eτ, ττ, µτ) ε = 0.016 % Yield @ 30 fb-1 = 120.6 events Probably an underestimate since we only used LO MC
Signal-to-Background Preliminary S/B 1:4 Will only get worse as we do more realistic treatments of backgrounds But we have not optimized cuts for signal efficiency Also Background #s included taus, signal #s did not Neither include feed-down from mis-reconstructed lepton+jet events And have not yet tried to implement kinematic strategies for separating S&B w/o reconstructing top
Status & Plans Get GRID access, use background (and signal) samples Susanna pointed us to ttjj vs ttbb, W->cs with a fake, etc Look at signal samples with varied Higgs mass Write up some documentation Explore S/B ideas Meet w/ people in two weeks @ CERN