1 Single top in e+jets channel Outline : - Data and MC samples - Overview of the analysis - Loose and topological cuts - MC efficiencies and expected number of background events - Cross checks - Summary - Issues Emmanuel Busato, Bernard Andrieu and Marumi Kado Goal : Give a first look at single top analysis in the e+jets channel with soft muon tagging Take RunI analysis as a starting point. Use results from very recent ttbar analysis (trigger efficiencies, ID efficiencies, soft muon tag efficiencies) This analysis is described in D0Note 4274
1 Data and MC sample Data (great thanks to Lukas !) : Root-tuples from Lukas : 92 pb -1 p and p calo corrections from Jan (shared energy problem) is applied new jet seeding applied QCD control sample from Lukas (23900 events) : signal sample with no requirements on the electron (not even emIDs) MC : single top events generated with ONETOP (p14) Wgfusion (2395 events) schannel (4291 events) Wjj (37950 events), Wjjj (44158 events) and Wjjjj (47500 events) generated with Alpgen (p14) tt (11000 events) generated with Pythia (p14).
1 Overview of the analysis Main backgrounds : W + jets QCD Other background considered : tt pair production We define 4 set of cuts : 1. preselection cuts 2. topological cuts 3. soft muon tagging 4. topological cuts + soft muon tagging For each cut, define a loose and a tight sample : tight loose + electron likelihood (D) Use matrix method to separate Wjets tt singletop contributions from QCD contribution (use qcd and w estimated by Lukas) Expected number of background events : Estimate efficiencies to pass cuts in Monte Carlo (except for QCD) normalize to the number of events found with Matrix Method in preselection sample (except for tt, where we normalize to the measured cross section)
1 Preselection sample definition Trigger : em15_2jt15 (used for tt e jets analysis) MET 15 GeV 1 electron : pT 20 GeV | det | 1.1 (and [1.1,2.5]) 2 nd em object veto number of jets 2 leading jet : pT 15 GeV | det | 2.5 second jet : pT 15 GeV (pT 10 GeV) | det | 2.5 (electron, MET) > 0.5 (not used in RunI) "triangle cuts" : (20/ ). (electron, MET) - MET 0 (20/ ). (jets, MET) - MET 0 (20/ ). (jets, MET) - MET 20 nb tracks vertex 3 vertex |z| 60 Differences with RunI red
1 W transverse mass matrix method njets 1 njets 2 njets 3njets 4 njets 1njets 2 njets 3njets 4
1 Data-Monte Carlo comparison jet variables
1 electron and MET
1 Additional cuts Topological cuts : pT(jet1) + pT(jet2) + pT(electron) + MET 125 GeV pT(jet1) + 4 MET 155 GeV pT(jet3) + 5 pT(jet4) 47 GeV !! ( njets 4) b-tagging : soft muon tagging (muon : medium ; 2 100) R(muon,jet) 0.5 pT(muon) 4GeV | (muon)| 2 (RunI : | (muon)| 1.7) pTrel 0 Require at least one tagged jet
1 Topological variables
1 Some comments Very good agreement for jet multiplicity Looking at W transverse mass, we see that the resolution is better in MC. than in data (jacobian peak sharper) second and leading jet pTs are well reproduced for high pT (>40 – 50 GeV). The agreement is not very good for low pT : data harder than MC most likely due to hadronic energy reconstruction which needs to be improved electron is harder in MC most likely due to estimation of QCD background. Should take into account dependence of qcd on pT(electron), eta(electron)
1 Event yields loose 4 1.87 Tagged 2.1 Tagged topological cuts 77 Topological cuts 269 Preselection N QCD N W+tt+single top tight njets 2 : qcd w 0.021
1 Trigger efficiencies in MC : fold turn on curves for each em15_2jt15 term (v11) in Monte Carlo events MC efficiencies and expected background (single top eff at RunI : % depending on eta of the electron) Correction factor : We want to apply objects ID efficiencies from data to our MC objects Compute efficiencies in reconstructed MC without electron likelihood Apply a correction factor these efficiencies, using objects efficiencies found by several tt analyzers : correction factor =
1 Expected number of signal events Number of single top events : N Wgfusion = 0.24 events (RunI : 0.28 events with 90pb - 1 ) N schannel = 0.15 events (RunI : 0.18 events with 90pb - 1 )
1 Expected number of QCD events Estimation uses only real data. We take the number of QCD events in the preselection and apply topological and tagging efficiencies to it number of QCD events in the final sample eed a control sample to estimate topological and tagging efficiencies obtained by inverting H matrix cut N QCD in final sample = events
1 Expected number of W and tt events topological efficiencies (and tagging efficiency for tt) are obtained from proportion of events that pass each cut in Monte Carlo W jets and tt. tagging efficiency for W jets : Use tag rate functions from Florian (thanks !) to estimate tagging efficiency in Monte Carlo W jets. Florian Beaudette
1 Normalize tt background to the measured cross section : tt in preselection sample = 17.5 6.1 events tt in topological sample = 4.8 1.6 events tt in final sample = 0.6 0.2 events Normalize W+jets background to the number of events in preselection sample : N W in final sample = 4.4 1.2 events
1 Cross checks To any number of events found by the matrix method after a given set of cuts, the various efficiencies determined previously are applied ( column ”Predicted”). The result is compared to the numbers found with the matrix method ( column ”Observed”). QCD : W : Agreement between predicted and observed numbers is good
1 Summary Total background : b= 7.7 events Observed number of events : 4 Limit : 10.2 pb a more conservative limit is found using all lower bounds of b, Luminosity, signal efficiency Limit : 29.7 pb The expected limit is : 44.6 pb Estimated number of background events :
1 Issues (1) number of single top events expected after all cuts : 0.4 We can maybe gain in the selection of Ws (em likelihood efficiency low 70%) change the position of the likelihood cut ? use another discriminant ? QCD background : even with a tight cut on the em likelihood (>0.4), there is 35% of QCD events in the final sample Is there a way to remove QCD events further ? (without degrading signal efficiency !) An interesting variable to look at is the MET significance
1 MET Significance in the e+jets sample QCD W W MET sig. cut is powerfull to remove QCD events (more than the standard MET cut) Would allow to replace the 3 triangle cuts and MET cut by only one cut Drawback : MET significance is a complicated variable hard to understand
1 Data- Monte Carlo comparison after presel : no MET cut, no triangle cuts but METSig>5
1 Issues (2) Another place where we can gain a lot is b-tagging : use of lifetime taggers to increase sensitivity to the signal (cf Mathieu's talk 10/06/03) We can probably gain a lot in topological selection too