DWPhysicsAnalysis

Welcome to my analysis issues page. Here I've posted some of the issues I'm looking at and some of the questions I have. Feel free to leave comments if you have any.

High-p_T top/anti-top identification and reconstruction

I'm looking to understand various issues of triggering and reconstruction performance in high-p_T t-tbar production, working in Athena version 14. I'm focusing right now on the semi-leptonic (electron) decay mode, t t -> b e nu, b q q.

Most recently I've been trying to compare the trigger and reco performance in version 14 with what was reported in the ATLAS CSC NOTE "Top quark physics at ATLAS" (using version 12).

Datasets

I've taken official evgen datasets and re-simulated, re-digitized, and re-reconstructed them in version 14 of Athena. I ended up using a couple of the AtlasTier0-14.2.10.Y releases, since those were collections of tags that I could find in 14.2.10.Y that would actually complete the simulation/digitization/reconstruction tasks, though I'm open to re-running these datasets in the version used for official production (whenever that is up and running).

• ~28,000 SM t-tbar events from 5200 evgen dataset
• ~20,000 1 TeV Z' t-tbar events from 6231 evgen dataset
• Simulation & Digitization performed with AtlasTier0-14.2.10.2
• Reconstruction (AOD) performed with AtlasTier0-14.2.10.8

Monte Carlo Event Acceptance Cuts

• Only events with the decays t -> b W, W -> q q and t -> b W, W -> e nu are accepted, based on MC truth
• Detector acceptance cuts applied on truth b q q b e nu:
  • Electron: |η| < 2.5, with 1.37 < |η| < 1.52 excluded (calorimeter crack veto)
  • b-quarks, quarks from W: |η| < 2.5

Commissioning Objects

I've attempted to make object selections which are as close as possible to those described in the Top CSC note.

• Electron (ElectronAODCollection)
  • IsEM = ElectronMedium
  • |η| < 2.5 with 1.37 < |η| < 1.52 excluded (calorimeter crack veto)
  • Calorimeter E_T in cone (ΔR = 0.2) < 6 GeV
• Muon (StacoMuonCollection)
  • |η| < 2.5
  • Calorimeter E_T in cone (ΔR = 0.2) < 6 GeV
• Jet (Kt4LCTopoJets)
  • |η| < 2.5
  • ΔR to any electron > 0.2
• Missing E_T (MET_RefFinal)

Commissioning Event Selection

• One reconstructed electron with p_T > 20 GeV
• No reconstructed muons with p_T > 20 GeV
• E_T > 20 GeV
• ≥ 4 Jets with p_T > 20 GeV
  • AND ≥ 3 Jets with p_T > 40 GeV
• Trigger: EF_e25i_loose

Trigger Decision

I used the isPhysicsPassed() method of the TrigDecisionTool to get a yes/no answer in each event for each trigger of interest.

Event Selection Efficiencies

I've duplicated the efficiency calculations for the various event selection criteria listed above. Each column contains the efficiency for events in the tt->benubqq MC subset to pass the indicated event selection criterion. The final column represents a logical AND of the five selection criteria. I'm seeing a marked reduction in the efficiency of the EF_e25i_loose trigger (which I take to be the closest to the CSC note's e25i trigger) which I do not understand. There is no matching done of the object which fires the trigger to either truth objects or objects reconstructed by the offline algorithms; this is simply the efficiency for the event to pass the EF_e25i_loose trigger in whatever way it pleases.

	Number of Events	Trigger EF_e25i_loose	Electron	MET	4 Jets 20	4 Jets 20 & 3 Jets 40	Combined Efficiency (logical AND)
CSC Note		52.9%	52.0%	91.0%	70.7%	61.9%	18.2%
SM 5200	4343	28.4%	72.5%	90.1%	60.1%	45.1%	9.8%

-- DenverWhittington - 03 Sep 2008