LTracker Progress

1
L-Tracker Progress Plans

• J.Popp
• University of California, Irvine
• Brookhaven National Lab
• MECO Collaboration Meeting
• May 14-15, 2004

2
Outline

• Motivation
• Cursory test of likelihood fitter by itself
• No event selection (trigger energy cut, pitch
  angle cut, etc.)
• No helix/detector crossing pre-selection for
  helix recognition
• Example from study of finite efficiency and
  acceptance of straws
• Mostly resolvable cluster type ambiguity
• Multiple parents for clusters
• Sketch of how we do cluster recognition
• State of our L-tracker software tools

3
Motivation

• Goal Obtain sufficient information about the
  L-tracker to make an acceptable performance
  comparison to the T-tracker.
• Desirable select detector design SOON say, six
  months or less(?)
• Acceptance of likelihood fitter must be
  understood.
• Assuming a favorable outcome we must raise the
  level of sophistication of L-tracker simulation
• Simulate detector operation at the raw data level
  (TDC times, etc.)
• Finish cluster recognition reconstruction
  inefficiencies, etc.
• Expand helix recognition pre-selection methods
• Fully utilize cluster information local angle
  measurements
• Add more event types beyond two orbits with (3,3)
  and (4,4) crossings

4
Simulation GMC Version 077

• Detector solenoid simulation
• 10,000 conversion events higher statistics,
  later
• Electron energy 105 MeV in stopping target
• Magnetic field MIT design calculation
• Calorimeter crystal
• Tracker
• 2848 straws
• Wall thickness 0.0025 cm
• Outer radius 0.25 cm
• Length 260 cm
• Modules are tilted to avoid helix crossing a
  straw more than once
• Test TJs maximum likelihood fitter by itself

5
Energy Spectra Smearing Off 10000 Events

• MECO memo 125, Fig. 2
• Entries 18.2 vs 21.9
• Mean 0.283 vs 0.303 MeV
• RMS 0.354 vs 0.370 MeV
• Chi2/ndf 283/25 vs 270/127
• Mean 0.327 vs 0.329 MeV
• Sigma 0.230 vs 0.209 MeV
• MECO memo 125, Fig. 4
• Entries 18.2 vs 21.9
• Mean 104.0 vs 104.0 MeV
• RMS 0.681 vs 0.695 MeV
• Chi2/ndf 30.4/27 vs 423/166
• Mean 104.4 vs 104.1 MeV
• Sigma 0.329 vs 0.568 MeV

6
Energy Spectra Smearing On 3500 Events

• Smearing off vs on
• E(fitter) E(entry to tracker)
• Entries 21.9 vs 19.5
• Mean 0.303 vs 0.310 MeV
• RMS 0.370 vs 0.336 MeV
• Chi2/ndf 270/127 vs 85.5/80
• Mean 0.329 vs 0.324 MeV
• Sigma 0.209 vs 0.248 MeV
• E(fitter) energy spectrum
• Entries 18.2 vs 19.5
• Mean 104.0 vs 104.1 MeV
• RMS 0.695 vs 0.629 MeV
• Chi2/ndf 423/166 vs 201/139
• Mean 104.1 vs 104.3 MeV
• Sigma 0.568 vs 0.545 MeV

7
Whats missing

• Pre-selection cuts in MECO memo 125 make little
  difference?
• At least 6 helix/detector crossings
• Trigger energy gt 75 MeV
• Pitch angle cut 45-60 degrees
• Likelihood gt 0.001
• Fitting uncertainty lt 600 MeV
• Scattering angle at crossing lt 0.08 radians

8
Example Finite Acceptance 1-hit Inefficiency

• Finite Acceptance
• Close-packed tubes, finite thickness
• Tube geometry distortion later
• Finite single-tube efficiency
• Roy Lees dissertation, p.38 E871 average
  efficiency 0.96
• Given an N-hit cluster the probability there are
  n hits missing is binomially distributed.
• Our software can be easily generalized to
  distinguish clusters with any number of missing
  hits that we may require.

9
Resolvable Cluster Type Ambiguity Example
10
Multiple Parents for Distinct Cluster Types

• Distinct cluster type, i.e.. geometry
• Not a problem, just feature of terrain
• Resolved by nonlinear fits to the observed set of
  time and wire positions
• Could also test each hypothesis

11
Example Ideal Cluster Recognition

• Distribution of cluster candidates per event,
  before reconstruction
• Form list of straw hits
• Form groups of contiguous hits
• Give each hit a group label
• Require hits in all 3 layers
• Finite acceptances and efficiencies will require
  us to relax these conditions later
• Form all combinations of 3- to 9-hits
• Maximum distance between hits
• TDC times have maximum separation
• Reject combinations with no c-type
• Reconstruct each cluster (DTD)
• Reject if no tracks returned by cluster
  reconstruction subroutine

12
State of L-Tracker Tool Development

• Helix pre-selection methods have been developed
  at UCI NYU
• Maximum likelihood fitter acceptance is being
  examined
• Could take up to a week, depending on what we
  find
• Finite cluster efficiencies and acceptances
  60-70 complete
• DTD reconstruction underway
• FDT straight-forward once time differences are
  done
• DTD could be completed in 1-2 weeks/ FDT should
  take a week
• Cluster recognition 50 complete
• Test effectiveness of cluster reconstruction on
  eliminating candidates
• Cluster recognition may take 2-4 weeks, possibly
  less.
• Pre-selection for helix recognition is most
  difficult task
• TJs FORTRAN is very hard to read and follow
• Maybe faster to use NYU pre-selection code for
  L-tracker.

13
Old Pattern Recognition Plan

• Previous detector simulation plan Jan. 2002
• Simulate raw detector signals
• Straw clusters time differences
• Pad clusters time differences
• Crystals
• Cluster recognition
• Straws local position direction in plane
  perpendicular to wires
• Pads local position along wires
• Crystals local position on sensitive surface
• Timing information in all cases
• Helix recognition global track finding
• Detector information pre-selection
• Maximum likelihood method helix fitter

14

15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
Event Selection

• 1000 muon-electron conversions
• Crystal calorimeter E gt 85 MeV
• Helix pitch angle 45-60 degrees
• Ideal clusters with 3 9 hits
• Trajectory crosses all 3 planes
• Hits in cluster are contiguous
• Events with 6-9 clusters

27
Multiplicity Ideal Clusters
28
(No Transcript)