1. Ivo van Vulpen Summary ATLAS Trigger and Physics week November 2006 http://indico.cern.ch/conferenceDisplay.py?confId=6824#6

2. Overview - Software validation & CSC samples - Issues from combined performances - Trigger & Streaming test - Data preparation, data quality and luminosity - Data format - Issues from physics sessions & CSC Notes - Conclusions

3. Software validation & CSC samples

4. Software status 12.0.3 - Several geometries available … and used - New magnetic field map

5. Misaligned geometry … realistic geometry Impact on physics channels under study

6. J/psi invariant mass magnetic field: old geometry: ideal magnetic field: old geometry: misaligned magnetic field: new geometry: misaligned magnetic field: new geometry: misaligned+distrortions

7. Castillo Higgs to 4 muons

8. Timeline Bulk CSC samples Will show example for electrons Bulk reconstruction CSC sample will be done using 12.0.4 CSC Analysis

9. Different releases Release January 2007 Used for first data. 12.0.31 12.0.4 13.x Not full trigger Full trigger info MANY Trigger talks !

10. Monte-Carlo production Expect 2M events/week - CSC sample is limited by storage: will be deleted: Rome data(12-2006), 11.042 RDO’s (01-2007) CSC bulk physics groups extra 1.5 million Alpgen (Martijn/Wouter)

11. Issues from combined performance

12. E/gamma L1: 10/20 constants/channel: 7200 channels Stable, check timing Calibration: - Energy corrections (~10%) - Position (S-shape correction) (track matching, see later) Offline (cell-level): - Physics channels: Corrections at AOD level ? Unal Needed for 12.0.4  still in time Many calibrations required (per electron definition) Data preparation session

13. Track matching efficiency IsEm efficiency Extra material, wider shower shapes, new tracking, new magnetic field Big impact Castillo Differences do not come from detector descriptions

14. Clement A tool for electron calibration at AOD level Uses TMVA. New buzzword: Toolkit for Multi-Variate Analysis (like NN) Train on Z  ee, W  ev, tt in different eta regions IsEM (standard) New AOD study Eta Pt Conclusions: - Better than IsEM on the AOD (not so strange, not calibrated yet) - Used to study electron reconstruction efficiency in different event topologies Eff_elec does not depend on topology

15. Timing & Memory leaks: - Requirement: EventFilter: < 1kbyte/event Why: EF uses offline tools/algorithms (‘normal users’ who are not as strict on memory leaks) Testbeam: - photon energy scale differs by 1% (test using e+e- pairs) - Difference in E1/E2 etc well described by MC  0.1% uncertainty expected Paganis

16. Jets: Calibration procedure: 1] Cell-based 2] Local hadron  Validation using testbeam data  Flexibility on AOD to redo calibrations etc. Jet-electron overlap: lost ‘isolated’ electrons & influence on jet algorithms Analysis level or already at reconstruction level E T -miss: Cranmer Use of Muon Chamber information to improve muon term in EtMiss (also need in AOD) Manouk/Marcel ?

17. Muons – Recent activites MuTag – extends Inner Det track into Spectrometer using also Middle stations segments to cover efficiency gap from 1< eta < 1.3  MuGirl : extrapolates InDet tracks to combined MDT+TGC/RPC segments. Use NN to make decision on match. 50% eff. For pT=2.5 GeV  pT/MeV Efficiency Single  pT=2GeV InnerDet + Calorimeter Muon Tagging (CaloTrackMuid) good for early physics Efficiency 60% Roda Gustavo Reconstructing low P T muons

18. Trigger & streaming tests

19.  Data written out in various streams - Per trigger ? / per … ? - How much data is duplicated ? - How to handle TAG/streams in ongoing analyses (see homework)  LBL: streaming test Use test trigger table to get data on inclusive versus exclusive streaming (Jan 2007) 3.6 M events (5 hours at 200 Hz) Use 1 minute luminosity blocks

20. Data preparation, data quality and luminosity blocks

21. Calibration Data Challenge (CDC) Hulsbergen Hawkings ID alignment (SCT): - FSI (< 1 micron) - tracks … various samples - How: global, local X 2  Dedicated alignment stream (24 h feedback) 900 GeV data less than 1 year away

22. Luminosity blocks Event header Trigger info Lumi info Luminosity estimate: - Handling in TDAQ & streaming - User analyses (Homework) Cluster events per lumi block Dedicate taskforce in ATLAS (of course)

23. Data quality (Mc Pherson) Important, but (too) big scope: - 1000000 histograms - data quality (software validation/hardware/physics ?) - detector QA (already there) - software validation (Wouter) McPherson

24. Gianotti Wouter

25. Data format Farbin

26. Works fine, but has some downsides …

27. For CSC Notes: - Need consistent set of ATLAS numbers Requires common DPD definition (what is an electron) ? … exchange numbers from combine performance groups ? - Not duplicate efforts  common Ntuples / tools Top Group … lots of discussion on common Ntuple Use TopView (Shibata) - How often is top sample processed? Single Ntuple ?

28. Probably the solution, but not for CSC note analysis Release 13 Analyse AOD in Root or Athena

29. Top: - Many talks on CSC notes - Wouter: Estimate QCD multi-jet bkg due to fake (non-prompt)-leptons Top peak without b-tag and ET-miss requirement - Akira: Neutrino solution in tt events Et-miss resolution Lepton resolution 2% Use W-mass constraint: 2 solutions for Pz: - 13% events ‘mathematically lost’ (can be used) - Other events 60% correct choice for Pz Improvement ?

30. Higgs (H  4 leptons and trigger) Rosati LVL1 LVL2 EF MUONS: ~80% Muon acceptance Channel trigger 442e2 1616 99.694.6 110 99.292.2 120 94.168.5 2626 95.657.6 210 91.750.2 220 65.825.3 4 muons 2 muons Trigger (multiple muons) has major impact on analysis. Nice game: Interplay trigger / reconstruction

31. Conclusions

32. Ian 1) Luminosity blocks, tags, trigger 2) New format, distributed analysis, atlfast