ATLAS ZCounting twiki - https://twiki.cern.ch/twiki/bin/viewauth/Atlas/ZCountingLumiATLAS Luminosity Group - https://twiki.cern.ch/twiki/bin/view/Atlas/LuminosityGroupSource code - https://gitlab.cern.ch/atlas/athena/tree/21.0/DataQuality/ZLumiScripts/scriptsLuminosity for Physics - https://twiki.cern.ch/twiki/bin/viewauth/Atlas/LuminosityForPhysics ATLAS Lumi Calculator - https://atlas-lumicalc.cern.ch/ Activate Cernbox. Then accessible via /eos/user/m/miokeefe/ for example - https://cernbox.cern.ch/
Example job can be found in:
See README for description of directory and code structure.
Special high mu Zmumu
Standard MC16a samples (matching 2015+2016 data)
Standard MC16c samples (matching 2017 data) [Now Redundant]
Standard MC16e samples
Running over MC
Same as on ZCounting twiki above, however only the dqt_zlumi_alleff.py script can be run. The luminosity cannot be computed, only the efficiencies.
Long term goal is:
- Use the 2016 example run (304008) and monitor all steps (e.g. T&P efficiency) versus LB.
Tasks / ongoing investigations:
- Understand syst uncertainties in Z counting related to muons, which are:
- T&P: stats dominated, need proper treatment to avoid double counting of stats errors.
- Muon calibration.
- Alignment L time-dependent? Some alignment is done already in calibration loop before full data processing - sufficient?
- Last update: January 29, 2018 (after discussion with Peter Onysi at CERN)
- Understand the cuts applied: isolation, pT, eta, impact parameter cuts - nominal selection so far should be pT>27 GeV, |eta|<2.4 for both muons (can we find these selections in the code?), may need to raise pT cuts for 2018 depending on lowest unprescaled triggers available
- Which efficiencies are applied: trigger, muon reconstruction; should aim to monitor also track efficiency and consider making T&P efficiency differential
- How are those efficiencies calculated: Monitor and plot those efficiencies for an example run for learning - see slides by Harry, excursions are likely low statistics bins where also background subtraction carries large uncertainties (and may lead to efficiency > 1? - to be checked)
- Where is MC used for corrections:
- this should be part of the mu-dependent "MC non-closure" (this encapsulates all effects we cannot determine purely from data, thus "non-closure" may be somewhat misleading) -- however seems this is "correcting" back to total cross section mll>60 GeV, not useful for fiducial cross sections (and maybe not ideal to compare "fiducial Zs with CMS")
- To investigate: QED FSR is properly taken into account, find place where truth-level selection is performed, run MC (technical hurdle: many "duplicated" histograms created when running on MC, need to run small pieces and learn how to merge this efficiently - Michael with test the command "DQ histogram merge")
- How (if at all) are mu corrections applied: implicitly in T&P efficiencies, mu-dependend "MC non-closure"
- Where is official lumi coming from in the plot, and is it really the latest one?
- Implement procedure to control that all events (lumi blocks) were processed (or alternatively correct expected luminosity accordingly) - missing pieces in Michael's plots got filled on rerunning jobs on the grid
- Usage of GRL: typically will run with a very preliminary GRL from the beginning and then later with the official GRL for the final result
- Understand the used "acceptance" (0.3323) and theoretical Z cross section
- Fine tuning:
- understand if assumption SameSign events = Opposite Sign background is justified
- is there a correlation of statistical errors of reconstruction and trigger efficiencies?
available for meeting
- Mon - 1-3 pm
- Tue - 2-5 pm
- Wed -
- Thur - 11-1 pm
- Fri - 9-11 am
- Tue 9am - 5 pm most weeks
- Wed 1.30pm-4pm most weeks
- Fri 1pm-3pm
- Tuesday 9am-12
- Thursday 10am-5pm (Starting Week 7)
Contributors: Michael O'Keefe, Harry Lyons, Uta Klein, Jan Kretzschmar, Max Klein
- 12 Jan 2018