Scottish Universities Physics Alliance

1
ATLAS Comparator
Scottish Universities Physics Alliance
ATLAS at the LHC
ATLAS, located at CERN in Geneva, is due to begin
data-taking in 2007 Currently, simulated events
are produced using two packages Full
simulation/reconstruction and ATLFAST. Full
simulation/reconstruction with Geant4 takes
2000s per event, whereas ATLFAST is able to
process an event in 0.05s, because it smears the
generated truth information It is important that
each approach is validated in an ongoing manner
if they are to be used for physics analysis.
Comparator Design

• The Comparator allows direct comparison of
  physics quantities for the same generated events
  pushed through both full simulation and ATLFAST.
  Running and re-running ATLFAST with different
  values of the smearing parameters allows optimal
  values to be found
• Full/Fast histograms compared using ?2 or
  Kolmogorov tests
• ?2 or Kolmogorov allows Minuit to alter Fast
  smearing parameters and feed them back in to
  ATLFAST
• Comparator also allows selection of different
  parameterisation schemas for individual
  variation and use in the smearing.

AMI
ESD
AOD
Combined Histograms (ATLFAST and Full data)
Standard Comparisons, Goodness-of-fit etc.
Reportof values (e.g. a plot)
From reco
Run ATLFAST
Make AOD
Run Comparison
Register in AMI
Repeat for range of ATLFAST param values
Register in AMI
Comparator Sequence Diagram
A C job uses Minuit to call the Athena
analysis package (which runs ATLFAST
repeatedly) Using the kdevelop Integrated
Development Environment to speed coding and
prototyping
Results
ATLAS Barrel Inner Detector h-gtbb
Feynman Diagram
Example comparison of electron distributions
using 20,000 ttH events ATLFAST smears the truth
information
Electron Results

• Too few reconstructed electrons in the forward
  regions of the detector (due to low track
  matching efficiency), and too many reconstructed
  electrons in the central portion (due to
  secondary electrons)
• The core output of the comparator will be
• vertex distributions,track multiplicities
• particle distributions (e,m etc.), Jet
  distributions
• Pt, E, h, f
• b-tagging quantities
• impact parameter resolution
• tagging efficiency/purity
• specialised parameterisations
• e.g. parameterise secondary electrons and track
  matching efficiency as a function of h