ZZllnn Analysis with 12.0.6

1
ZZ?llnn Analysis with 12.0.6

• Thomas Barber
• University of Cambridge
• Diboson Meeting, August 2007

2
Outline

• Present results of ZZ?llnn analysis with 11.0.4
  and 12.0.6 data.
• Analysis procedure using the grid.
• Description of CSC datasets used.
• Electron and muon identification in Eventview.
• Parameters used in cut-based selection.
• Comparison of full-simulation yields to previous
  fast sim results.
• Signal and background yields using tuned cuts.
• Summary and outlook.

3
Analysis Procedure

• Athena analysis using EventView 12.0.6
• Using Distributed Analysis framework with
• Ganga for LCG datasets.
• pAthena for BNL datasets.
• Event Generators mostly either Pythia or MC_at_NLO,
  one AcerMC sample.
• Event Weights in MC_at_NLO are handled by
  subtracting negative events from positive
  distributions and yields.
• Cross Sections calculated by running appropriate
  DC3 joboptions transform to generate 10,000
  events.
• For MC_at_NLO, combine production CS with branching
  ratio from PDG.

4
Datasets

• Analysis done on both CSC11 and CSC12 signal and
  background datasets.
• Two different signal datasets used
• 11.0.4 Signal (blue)
• csc11.005981.Pythiazzllnunu.recon.AOD.v11004213
• Pythia generator leading order
• 48700 events.
• 12.0.6 Signal (red)
• user.HongMa.trig1_misal1_mc12.005932.McAtNlo0310_J
  IMMY_ZZ_llnunu.recon.AOD.v12000601_pUM02203
• MC_at_NLO generator next-to-leading order.
• 118000 events
• Dataset currently unofficial, official sample in
  production.
• In case of v12 datasets, 1mm Bug-fix applied
  where needed.

5
Version 11.0.4 Datasets
6
Version 12.0.6 Datasets
7
EventView Inserters

• EventView used to create NTuples and remove
  overlap (?R 0.1).
• Insertion order
• Electrons ( loose electrons)
• Photons
• Muons ( loose muons)
• TauJets
• TauJetCollection pT gt 5 GeV, either TauRec or
  1p3p
• Particle Jets
• ConeTowerParticleJets, ?R 0.5, pT gt 5 GeV

8
Electron ID

• Electrons from egamma
• (isEM 0x7FF) 0, ie no TRT
• Isolation Et(?R0.45) lt 8 GeV
• pT gt 5 GeV, ? lt 2.5
• 11.0.4 Efficiency 75.0
• 12.0.6 Efficiency 62.6
• Much lower in 12.0.6

9
Muon ID

• Muon from MuID
• Require a best match, with
• ?2(match)/ndof lt 10
• ?2(fit)/ndof lt 5
• Isolation Et(?R0.45) lt 5 GeV
• pT gt 5 GeV, ? lt 2.5

10
Muon Efficiency

• 11.0.4
• Efficiency 86.3

• 12.0.6
• Efficiency 87.7

11
Lepton Kinematics

• Electrons and muons show good agreement.
• Treat them as leptons for rest of analysis.
• pT shows good agreement between two versions.
• Z-width not yet implemented in MC_at_NLO (planned
  for future)

12
Analysis Cuts

• ZZ?llnn signal characterised by 2 high-pT leptons
  and ET(miss).
• Atlantis event displays
• Define cuts to separate signal from background
  channels.
• Main backgrounds from channels with
• Large C.S. (ttbar and Z?ll)
• Similar topology (WZ and WW)
• Following plots are
• normalised to unit area to compare shapes
• before any cuts applied
• Show 12.0.6 distributions (unless specified)

13
Lepton Cuts

• ZZ?llnn signal channel shown in red.
• Lepton pT gt 20 GeV
• Reduces soft electron background, eg ttbar
  (magenta) Z??? (turquoise)
• mll 91.2 GeV lt 10 GeV
• Reduces non-Z background, ttbar (magneta), Z???
  (turquoise) WW?l?l? (purple)

20 GeV
10 GeV
14
Third Lepton Veto

• Define Loose Leptons is anything left in
  Electron/Muon container that does not pass
  initial lepton ID, with pT gt 5 GeV.
• N(leptons) N(quality) N(loose)
• Veto events with N(lepton) gt 2 to cut down WZ
  background.

15
Missing Energy

• Using AOD keys
• METFinal (11.0.4)
• METRefFinal_1mmCorrection (12.0.6.1)
• METRefFinal (gt12.0.6.4)
• Absolute MET cut gt 50 GeV removes Z?ll Drell Yan
  (blue), Z?4l (orange).
• Also require magnitude and direction to match
  that of the Z(ll).
• MET-Zpt/Zpt lt 0.35
• f(z)-f(met) lt 35 deg
• Both cuts are set at 2s from the centre of the
  signal peak.
• Helps to remove background from the WZ channel.

16
Jet Veto

• Require events to have no jets with
• pT gt 30 GeV and ? lt 3.0
• Reduces background from ttbar and Z?ll (high pT
  Z)
• Discrepancy between jet multiplicity 11.0.4 and
  12.0.6 signal Due to Generator, Pythia (LO) vs.
  MC_at_NLO (NLO)
• In the truth, average number of jets per event
• 11.0.4 1.8 12.0.6 5.3

17
Final Cut

• Cut on pT of the Z.
• In fast simulation, pT(Z) gt 150 GeV
• For full simulation, only require pT(Z) gt 100 GeV
• Final value can be set to obtain a required S/B
  ratio.
• Anomalous coupling enhance cross section at high
  Z pT, so this cut will not harm anomalous
  coupling studies.

18
Fast Simulation Comparison

• Run analysis with cuts identical to fast
  simulation study, by Hassani (ATL-COM-PHYS-2002-01
  2)
• Top table reproduction of fast sim results, lower
  table full simulation with 12.0.6 data. (except
  Wt channel, which is 11.0.4)
• Now signal is only 60 of fast simulation
  prediction.
• Larger backgrounds from Zjets and WZ channels.
• Other channels, no events pass, low statistics.

Events expected in 100fb-1 of data
19
Full Simulation Cuts

• Now apply MET matching cuts and reduce pT(Z) cut.
• Plot shows events after cuts, normalised to
  100fb-1
• Signal efficiency increased, background reduced
• 11.0.4 12.0.6
• ? 3.2 ? 2.6
• S/B 2.25 S/B 1.96
• Errors on ttbar Z?ll large due to low
  statistics.

Events expected in 100fb-1 of data
20
Signal Selection Efficiency

• Plots show number of reco events passing cuts
  compared to number of true events passing cuts.
• No ZpT cut is applied.
• Shows a drop in efficiency at high Z pt.
• Possibly due to Jet Veto removing high pT(Z)
  events that also have a high pT jet recoiling.
• Needs further investigation possibly tune the
  jet veto.

21
Conclusions

• Presented results of ZZ?llnn analysis with 11.0.4
  and 12.0.6 data.
• Good agreement between lepton properties in two
  versions, except
• Lower electron efficiency
• No Z width in MC_at_NLO (12.0.6 dataset)
• Cuts defined to separate signal/background.
• Comparison to Fast Simulation shows lower signal
  efficiency and higher backgrounds.
• Requiring Missing Et to match the Z pT and ?
  gives much better yields.

22
Future Outlook

• Perform analysis using official CSC signal
  dataset.
• Investigate and optimise the jet veto (to reduce
  ttbar) and 3rd lepton veto (to reduce WZ).
• Apply results to anomalous coupling analysis to
  put limits on NTGC parameters.
• Trigger Aware Analysis still needs to be
  performed.
• Investigate Multi-variate analysis techniques, eg
  ANN, BDT etc