Adventures In Calorimeter Assisted Tracking

1
Adventures In Calorimeter Assisted Tracking

• Chris Meyer, Tyler Rice
• UC Santa Cruz
• October 16, 2007

2
Tools

• The goal is to put together a combination Track
  Finding Driver for the central region of the
  detector. This includes
• VXD Cheating
• AxialBarrelTrackFinderZ
• GarfieldTrackFinder
• These are run through sequentially, eliminating
  hits found on tracks after each step.

3
VXD Cheater

• We started with code written by Eric Wallace
  (formerly of UCSC) and modified it to cheat away
  not only Tracker Barrel Hits, but Vertex Barrel
  Hits as well.
• All associated SimTrackerHits from MCParticles
  originating within 2 cm of the origin are removed
  from hit banks.

4
AxialBarrel Z

• Code originally written by Tim Nelson (SLAC), it
  takes three hit seeds from the Tracker Barrel and
  attaches hits to find 4 or 5 hit Tracks.
• Optimized for non-prompt tracks by Tyler Rice
  (UCSC) Some improvements added (phi-region
  check).
• The code used was modified by Lori Stevens
  (UCSC) introduced virtual segmentation and an
  algorithm to check for linear z-consistency using
  only end-position information from z-modules.
• Z-Modules are user defined, changeable in the
  Driver.

5
GarfieldTrackFinder

• Written by Dima Onoprienko (KSU),
  GafieldTrackFinder employs outside-in track
  finding, beginning with EmCal Stubs.
• Finds tracks with 3, 4 or 5 hits in Barrel
  Detectors.
• Also uses modular z information when finding
  possible hits to attach to track.
• Changes made to Garfield
• GarfieldTracks/Hits have associated MC truth
  information
• Virtual Z-Seg now customizable

6
Whats a Good Track/Particle?

• Findable MC Truth Particle
• FINAL or INTERMEDIATE State Particles
• Charged
• Path Length 50 cm
• Pt 0.75 GeV
• Radius Origin
• cos ?
• Acceptable Reconstructed Track
• cos ?
• Pt 0.75 GeV
• DCA
• Successfully Fit

7
Whats a Good Track/Particle?

• Good Tracks
• Purity 0.8
• Associated MCParticle is findable
• Passes acceptable criteria
• Fake Tracks
• Purity
• Passes acceptable criteria
• Ignored Tracks
• Purity 0.8
• Associated MCParticle is UNfindable
• Purity is calculated strictly using Barrel Hits
  (Calorimeter stub excluded from calculation)

8
Initial Results

• Axial Barrel 4 hit Tracks, Garfield 3 Hit
  Tracks
• Z Pole qqbar Event

5 Hit Tracks
9
Initial Results

• Axial Barrel 4 hit Tracks
• Z Pole qqbar Event

ABTFZ 4 Hit Tracks (Garfield Will Be Run Next)
10
Initial Results

• Axial Barrel 4 hit Tracks, Garfield 3 Hit
  Tracks
• Z Pole qqbar Event

MCPs are those left after running AxialBarrel
11
Initial Results

• Axial Barrel 4 hit Tracks, Garfield 3 Hit
  Tracks
• Z Pole qqbar Event

VERY Low Purity for Reasonable Segmentation
12
What Are We Missing?

• If we change to 5 hit minimum for Axial Barrel
  Purity is increased, but efficiency is decreased
  (Tyler Rice presentation, 8/21/07).
• The main concern is the inability of the
  calorimeter to distinguish good 3 hit tracks from
  fakes.
• Can we reliably reconstruct tracks originating
  outside the second Tracker Barrel layer?
• SeedExtend Algorithm, Tyler Rice, Chris Meyer
• Since Barrel information is precise, make a helix
  from 3 hit seeds (with seeds based on Tim
  Nelsons code) then check for matching
  calorimeter stubs (found using Dima Onoprienkos
  NN Cluster Driver and EmCalStub Driver).

13
Stub Matching

• Muon 1-50 GeV Events

XY Miss Distance between Seed and Stub at
Calorimeter Face
14
Stub Matching

• Muon 1-50 GeV Events

Phi Difference between Seed and Stub at
Calorimeter Face
15
Stub Matching

• Muon 1-50 GeV Events

Curvature Ratio of Seed and Stub
16
Cuts on Matching

• ZPole sidaug05 cuts for SeedExtender
• Seeds
• Phi Separation consecutive layers)
• Helix Stub Matching
• Base Difference
• Phi Difference
• Kappa Ratio ( (?seed - ?stub)/ ?seed )
• Note All values are absolute values

17
Approach

• Loop through list of stubs looking for possible
  seeds to attach.
• Calorimeter Stubs
• Refit stub to get better helix (base position,
  base phi, kappa)
• Use all CalorimeterHits, give more weight to
  closest two, farthest two, and middle two
• Tracking Seeds
• Only look at outer 3 tracking layers
• Require max phi-separation between any two hits
  of 300 miliradians
• If multiple matches are found, seed with smallest
  phi difference from stub is used.

18
Seed Extender
20 total Findable 3 Hit tracks. Binned by Radius
from Origin (mm)
Radius of Origin
19
Seed Extender
12 Found 3 Hit tracks. Binned by Purity
Include stub as 4th hit in purity calculation
20
Seed Extender
2 Found 4 Hit tracks. Binned by Purity (5th Hit
from Calorimeter assumed to be pure)
21
Seed Extender
4 Fake 3 Hit Tracks. Binned by pt (GeV)
22
Conclusions

• ABTFZ and Garfield
• Efficient and Pure for 4 and 5 Hit Tracks
• Cant find 3 Hit Tracks
• Developed SeedExtend for 3 hit Tracks
• On Z-Pole events we get 60 efficiency for 3 Hit
  Tracks
• On Z-Pole events we get 70 purity for 3 Hit
  Tracks
• This can be optimized further by changing cuts,
  but the optimization will depend on the physics
  signatures (isolated signature vs. dense jet)
• Plan to look at meta-stable SUSY signature
  consulted with Jonathan Feng (UC Irvine)
• With this in mind Chris Betancourt (UCSC) is
  looking into ISAJET generation with Jonathans
  guidance.