AtlfastC: Improvements to the Atlas Fast Simulation

1
Atlfast-C Improvements to the Atlas Fast
Simulation
- Catherine Wright - First Year Summary
Presentation
04 December 2006, Mon
2
Contents
Aim To introduce the topics I have been working
on for the past 15 months!
? Justify my
grant!!!

• Topics
• Introduce Atlas and its fast simulation,
  ATLfast.
• Highlight developments that lead to Atlfast-C.
• Discuss the Atlfast-C and the Efficiency-Matrix.
• Conclude, and briefly discuss future plans for
  PhD.

3
ATLAS _at_ LHC
This is ATLAS!
25m
40m
4
ATLAS _at_ LHC

• ATLAS is a massive, and very complicated
  detector.
• It has over 100 million parts (easily) and is
  the size of a ship.
• Reconstructing the output from such a detector
  is incredibly difficult, as the interactions of
  all the particles in a collision must be
  simulated. This is done at ATLAS in two ways

• ? Using the GEANT4 tool to simulate the
  interaction of particles with matter in the
  detector.
• ? Using ATLfast a FAST parameterisation of
  detector effects on particle observables.

5
ATLfast A tool for Fast Simulation of the Atlas
Detector
Taking generated data input and using
parameterisations, such as those shown below,
ATLfast simulates the CALORIMETER response and
thus smears such measurable values, as energy,
transverse momentum and position in the detector
(pseudorapidity, ?).
Low ?
High ?
Representative of the physical resolution of the
ATLAS Calorimeters when detecting electron energy
deposits
Reference ? http//www.hep.ucl.ac.uk/atlas/atlfast
/
6
Atlfast (What it can do)
Low ?
High ?
---- Atlfast ---- Atlas Full Reconstruction
E(GeV)
E (GeV)
These photon energy distributions are similar at
both high and low pseudorapidity.

So why change it?
i.e. Atlfast already produces output that is very
similar to the full reconstruction.
7
Atlfast Current Thinking
One often chooses to add a FLAT reconstruction
efficiency of 90 to Atlfast output why not
include the ACTUAL reconstruction efficiency,
thus producing a fast simulation that evolves
with changes to the detector and the
reconstruction algorithms?
Effectively, there are two camps of thought for
this improvement of Atlfast

• Official Atlfast Validation group
• Adding reconstruction and calorimeter layers to
  the fast simulation.
• Not quite as fast.
• Work not completed yet.

https//twiki.cern.ch/twiki/bin/view/Atlas/Atlfast
ValidationGroup
8
Comparator Atlfast vs. Full Sim
Original proposal for improvement to Atlfast
arose from development of the Comparator
software. ? Reference https//uimon.cern.ch/twik
i/bin/view/Atlas/AtlasComparator

• The Comparator used a ?2 minimisation tool,
  MINUIT to vary the parameters in the Atlfast
  smearing formulae to produce the most similar
  full and fast distributions.
• Results in reasonably large changes to the
  smearing parameters ? this is UNACCEPTAPLE as the
  parameters defined are physical variables
  representative of the abilities of specific parts
  of the detector.
• Comparator was using a hammer to break a nut.

Figure provided by C. Collins-Tooth.
9
Full vs. Fast

• For photons
• Deficiency of reconstructed photons in high
  pseudorapidity region of the detector. WHY?

• Atlfast cannot see the effects of the detector,
  it can only mimic the resolution of each part of
  the detector ? as a result, Atlfast cannot wholly
  mimic the loss of particles which occurs in the
  full simulation.

10
A Fast Solution?
Solution
Include fully reconstructed efficiencies in
Atlfast.
Thus detector effects observed in full sim can be
mimicked by Atlfast.
Atlfast-C
Schematic detailing the position of Atlfast-C
within the current data-production framework.
11
Solution Efficiency Matrix

• efficiency removes particles that would not
  be seen in the full reconstruction.
• Cxy contamination (fakes/misidentification)
  adds particles

12
Efficiency Matrix

• Designed to kill particles that (in the full
  chain) would not be reconstructed.
• And to add particles that are misidentified
  i.e. fakes, contamination and migrations.
• May result in the necessity to have new Atlfast
  collections.
• There are a few PROBLEMS Definition of
  Efficiency Contamination

reco in bin gen in bin
Efficiency
Not generic introduces a dependency on bin
size so
matched reco generated
Efficiency
13
Efficiency Matrix

generated (i) and reconstructed (i)

hj

generated (i)
Definition of the Efficiency
This avoids the danger of making the calculations
bin dependent, however it introduces a second
bias, in the MATCHING CRITERIA.
?R0.01
?R0.02
?R0.05
?R0.1
?R0.2
Conclude on Matching Criteria
?R 0.2 and ?E/E lt 0.4
?R0.05
?R0.1
?R0.02
?R0.01
?R0.2
deltaE fixed deltaR varying
deltaR fixed deltaE varying
1.4lt?lt2.5
14
Efficiency Matrix Electrons
15
Efficiency Matrix Photons
Contamination from electrons
Photons efficiency
Contamination from muons
Contamination from jets
16
Efficiency Matrix Muid Muons
Contamination from electrons
Contamination from photons
Muid Muon efficiency
Contamination from jets
17
Efficiency Matrix C7 Jets
Contamination from electrons
Contamination from photons
Contamination from muid muons
Jet reconstruction efficiency
18
Efficiency Matrix Kt Jets
Contamination from electrons
Contamination from photons
Contamination from muid muons
Jet reconstruction efficiency
19
Efficiency Matrix A simple result
Corrected Atlfast distributions
Reconstructed
Corrected Fast
Original Fast
H ? ?? sample
Plots with default values deltaR0.2
deltaE0.4
One can see from the corrected plot that the
efficiency alone is over zealous need to include
contamination from fakes migrations.
20
Conclusions

• Atlfast-C is due to be added to the Atlas CVS
  Repository tomorrow 5th December 2006.
• Active use of the efficiencies will not be seen
  immediately needs to be promoted to the Working
  Groups within the experiment.
• Some small issues, such as updating to the
  latest Atlas Software Release and changes jet and
  muon results to reflect correct detector regions
  must be dealt with.
• Overall, this has been a time consuming project,
  and leaves me looking forward to tackling future
  projects such as a statically combined Higgs
  analysis.