HZZ4l isolation study

Introduction

Lepton isolation has a particular importance in Higgs → ZZ → 4 leptons analysis. It is one of the most powerfull criteria when fighting against ttbar → 4 leptons and Zbbbar → 4 leptons backgrounds. Leptons from the signal are expected to be isolated, unlike leptons coming from t or b decay chain. Isolation variable is very sensitive to underlying event activity, which is known with large uncertanty.Therefore, it's very important to control the efficiency of the isolation variable. The most common technique to control the isolation efficiency losses is Random cone technique. In this TWiki we present a random cone study for the HZZ4leptons channel, as well as the tools used.

1. HZZ4e Isolation Study plan for 2010.

February

• Isolation variable and efficiency calculation algorithms on the market - what is currently used (HZZ4l, EGamma, A. Drozdetskiy, A. Graziano, M. LeBourgois, D. Evans)
  • parameters used in algorithms:
    • cone opening,
    • pT normalization, only pT sum,
    • usage of TRK, ECAL and HCAL (Jurrasic method)
    • random cone technique - module to be produced

March, April, May

• Low Lumi (< 5 Z events ~ 0.1 pb-1)
  • random cone technique on 2009(10) data (and try on MC@7TeV) - (in)efficiency check using existing but modified Egamma code
  • commissioning of isolation variables with QCD MinBias data (all charged tracks) - iterative process

May, June

• As soon as we get some electrons from Tag and Probe Zs:
  • commissioning of isolation variables on electrons (could be different from charged track commissioning because of e.g. material budget)
  • propagation of isolation efficiencies from Z to ZZ (even if there is no ZZ events)
  • producing "unreasonable" limits on H→ZZ→4e (could be published by the end of the year - 100 pb-1)

By the end of the year (100 pb-1)

• Deducing isolation efficiencies from Zbb, ttbar events
  • tagging , as example, on Z and b, then probing on other b to check differencies between fake electrons and electrons coming from Z.
  • Isolation vs. pT control of fake electron isolation

2. Important talk, papers and software references

• EGammaIsolationAnalyzer by DaveEvans - uses random cones
• Electron isolation efficiency talk by FedericoFerri

3. Tools currently used

• adopted and modified code from EGamma random cone study (D. Evans)
• to try it out do:

     scramv1 project CMSSW CMSSW_3_3_6_patch6
     cd CMSSW_3_3_6_patch6/src/
     cvs co -r TAG -d IsoDepositAnalyzer/EgammaIsolationAnalyzer/ UserCode/RokoPlestina/IsoDepositAnalyzer/EgammaIsolationAnalyzer
     scram b 
     cmsenv
     cd IsoDepositAnalyzer/EgammaIsolationAnalyzer/test
     cmsRun egammaisolationanalyzer_cfg.py
     

Tags

• roko_0_0_0 is a first publication tag with flat random cone distributions and with some problems in HZZ4l tracker isolation producer

4. Results

• Figures produced with flat eta and phi distribution of random cone - HzzFlatRandomCones
• Figures produced with Zee-like eta and phi distribution of random cones - HzzZeeLikeRandomCones
• 7TeV results V0
  • 7TeV run with data from 30/3/2010 to 12/4/2010
  • flat eta-phi distribution of random cones
  • data and MC skimmed for electron studies (require 1 supercluster OR 1 GsfTrack)
• HZZ4l Isolation observable commissioning for 7 TeV with electrons HzzElectronIsolationObservableCommisioning

• Comparision of EGamma isolation (left) and HZZ4l isolation (right)

Random Cone Distributions

Tracker Isolation

HCal Isolation

Tracker Isolation vs. Eta

Ecal Isolation vs. Eta (only EGamma)

HCal Isolation vs. Eta

5. To do and understand

• RandCone_TowerEmEt.png - why do we obtain zero always? Different from RandCone_TowerEmEtVsEta.png
• understand differences in tracker isolation between EGamma and HZZ4l isolation
• understand the gap in tracker and HCal isolation: Only due to cut on min pT of tracks or something more
• understand where is the problem with plotting hTrackIsoHZZ4lVsEta
• apply spike removal to data sample (provide code)
• consider the EGamma tracker isolation bug for electrons and photons. The bug is fixed in tags RecoEgamma/EgammaIsolationAlgos V00-03-18 and RecoEgamma/PhotonIdentification V01-00-03
  • to see the description of the bug visit this presentation by MatthewLeBourgeois
  • random cone isolation is not affected by this bug since it is throwing cones from (0,0,0) and not from the vertex of a candidate track
  • possible improvements are throwing cones either from primary vertex (PV) or from beam spot (BS)

6. Samples used for the study

Running on MC

• Minimum bias samples used with global tag STARTUP3X_V8P

  /MinBias/charlot-MC_STARTUP3X_V8P_29Jan_900GeV_BSCPhyDecTrigSkim_rereco-a541b356a3eca87648310afd2ce7cf09/USER 

• sample used to extract the electron eta distribution from Zee events:

  /Zee/Summer09-MC_31X_V3-v1/GEN-SIM-RECO 

Running on data

• Minimum bias samples used with global tag GR09_R_V6A

  /MinimumBias/charlot-BeamCommissioning09-Jan29thReReco_BSCnoBeamHaloBptxPhysDecTrigSkim_rereco-d9a3a544833f2aa608ea083761abcd9c/USER

• run selection provided by RobertoSalerno from ElectronFirstData - note checkout of run selection code is needed
• using standard (official) electron salection setting input tag to gsfElectrons::RERECOANDSKIM

7. Little things learned on the way

• hadd newname.root *.root - command used to make single root file from many. Very useful when you have many rootples as output of crab jobs
• to find which global tag should be used look at SWGuideFrontierConditions
• when running on data, make sure to apply run selection and to save events that have passed it only. The output module has to be defined like this:

Show... Hide

   process.out = cms.OutputModule("PoolOutputModule",
   fileName = cms.untracked.string('output.root'),
   # save only events passing the full path 'p'
   SelectEvents   = cms.untracked.PSet( SelectEvents= cms.vstring('p') ),
   # define output
   outputCommands = cms.untracked.vstring('drop *',
                  'keep *_myCollection_*_*' )
   )
   

• to commit and tag your code in UserCode area follow instructions in HowToUserCode

-- RokoPlestina - 30-Apr-2010