new ntuple production is summarised on the following pages:

• https://twiki.cern.ch/twiki/bin/view/Atlas/TopNtupleAvailability#v1214
• https://twiki.cern.ch/twiki/bin/view/Atlas/CSCObjectSelection
• https://twiki.cern.ch/twiki/bin/view/Atlas/TopCSCDatasets
• https://twiki.cern.ch/twiki/bin/view/Atlas/TopGroupCSCObjectSelection

Recommendations from T&P week about cross-sections, reconstruction & PID (19 Nov 2007)

• summarized the new object selection following the recommendation by Karl. Please see this page https://twiki.cern.ch/twiki/bin/view/Atlas/CSCObjectSelection
• Please consult Karl's talk for more details. http://indico.cern.ch/getFile.py/access?contribId=13&sessionId=14&resId=0&materialId=slides&confId=20504
  • (1) W+jet Currently top uses LO. The recommendation is to normalize to NLO cross-section. For Alpgen samples, this corresponds to an overall normalization factor of 1.15
  • (2) The proposal of the Top group for ttbar has been accepted: use the NLO+NLL cross-section 833 +- 100 pb. The 12% uncertainty covers scale and structure functions ttbar+jets should also use NLO.
  • (3) QCD: use LO Pythia or Herwig with large error
  • (4) Diboson cross-section: use MC@NLO cross-sections
  • (5) bb cross-section: keep Pythia 6.4 value of 6.1 microbarns with large error

• Addition relevant uncertainties are:
  • (1) Luminosity: The Top group has used so far 10%. Numbers quoted at T&P week are 20-30% at start-up (2008) then (2009) decreasing progressively to 5-3%.
  • (2) PDF: The proposal of the Top group to use the approximate reweighting method of CTEQ6M using (x,flavor,Q) has been accepted (see Karl's transparencies for more details). Cross-check with the full procedure, regenerating events with different PDF, should be done for a few cases.

• Reconstruction - PID

The STRICT recommendations on reconstruction and PID are summarised here below.

In general, follow the recommendations from the Combined Performance groups for the variables and the cuts. They will provide efficiencies and corresponding uncertainties for 3 sets of collected data: 100 pb-1, 1 fb-1, 10 fb-1

• • e/gamma: Use tight/medium/loose electrons definitions and the photon definition. There is a note with the details ATL-PHYS-INT-2007-006. The top group used so far 1% for 100 pb-1 . The note says <1% for 100 pb-1.
    • - What crack cut, if any, should be used for electrons?
      • This depends whether we will use calorimeter isolation cut or not. Egamma group insists that we must not use etcone variables as they are buggy, in case we won't cut on isolation, crack cut is not necessary. If we will, then the cut is 1.35-1.65 In topview. Crack cuts for electrons and photons. We have been asked by the SM group conveners about this and the official answer was as Andrei says 1.37 < abseta < 1.52 which is a precisely defined region for photon ID (0.05 wider than the exact region without strips in the EM calorimeter). For electrons, there is no real reason to have a crack cut except perhaps the same one with early data and an as yet poorly understood detector.
    • - What is the deltaR cut for electron-jet overlap removal?
      • 0.3 as used in TopView
    • I use Tru_DecayIDs in my code and it works find (*(TruthTree.Tru_DecayIDs))[i][k], which return the k decay of particle i in your event (TruthTree is a pointer to class Truth made from make class on Truth0 tree). However, I had some problem using Tru_ParentIDs. In fact, if you look at Tru_nParent it always return 1 and the ID stored in (*(TruthTree.Tru_ParentIDs))[i][0] is the particle pdgID of the i particle, meaning that its only parent is itself...
  • Muon: Use STACO (reason: currently less fakes for data with pile-up). Here are some of the numbers quoted for efficiency and resolution: for pT<100 Gev: efficiency 3% ; pT resolution measured from Z with 12% accuracy for PT >100 GeV: 3% + 5% extrapolation error to 1 TeV pT resolution at 1 TeV: 10+-2%. We are not removing muon-jet overlap. The muon-jet deltaR is to define isolated muons, not to remove the overlapping jet. that is 0.2 (etcone20).
  • Tau: TauRec and Tau1P3P are complementary. Use most appropriate depending on the case
  • Jet: Jet energy scale uncertainty will be provided by the Jet CombPerf group
  • MissingET: Use the RefMET algorithm, We are using tower in topview.
  • b-tagging: Use the IP3D+SV1 algorithm (also IP2D tagger for early data). Cut on the weight is analysis dependent.

So, for the Top group the most relevant change concerns the use of STACO, which will mean some reprocessing. We will address that issue soon.