Tracking Photon Conversions

1
Tracking Photon Conversions
2
Existing Track Seeding

• From pixels
• Widely used, but not useful here
• From stereo silicon layers
• Uses layers 5 and 8 (barrel), not enough
  acceptance

3
First QuestionInside-Out or Outside-In?
Supercluster loses particles more likely for
early conversions, asymmetric, bremming.
Both go into same supercluster
Might be cleaner to work inside-out in a narrow
f-window
Need to work outside-in Maybe look for 2nd track
inside-out
4
Outside-InFinding the starting trajectory state

• Do separately for each basic cluster
• Or should I only do once for each supercluster?
• Assume the conversion happened 1/3rd of the way
  in the tracker
• Hand-wavingly Bayesian, since we cant track the
  outer conversions
• Assume the track had the full energy of the
  cluster
• Calculate where in f to start the track
• From the formula for the intersection of two
  circles
• Propagate this trajectory to outer layers and
  look for consistent hits (Df lt 0.015, Dz
  consistent with IP spread)

5
Outside-InCompleting the Seed

• Next, create a trajectory state from the calo
  cluster position, the hit position, and the
  cluster energy.
• Propagate inwards and look for second hits in the
  seed (still optimizing f and z windows)
• If a second point is found, create a new
  trajectory from a helix of the two tracker points
  and the calo cluster point.
• Use the Kalman Filter Updator to add the points
  to the trajectory, to get combined errors
  correct.
• Send seeds off to Kalman track reconstructor.
• Demand four-hit tracks, hit c2lt5, one lost hit

6
Outside-In SeedingCurrent Results
7
Inside-Out Seeding

• Look for first hits within a narrow f window
  (0.006) along the supercluster centroid
• Once a first hit is found, look for 2nd hits on
  the next two layers, assuming the track starts at
  the first point, and has a curvature of half the
  supercluster energy
• If a second point is found, make a seed using the
  curvature that was found, assuming the track was
  going radially at the first hit.
• Try to find tracks
• If only one track is found, try looser cuts to
  find a second seed with the same vertex

8
Inside-Out SeedingTwo-Way Trajectory Building?
Brem showers degrade the measurement Of the calo
cluster f What if the cluster f is mismeasured,
so I miss the first few points? Tried to make
a TrajectoryBuilder that works In both
directions, but isnt quite working yet
9
Inside-Out Results
10
Next Steps

• Not nearly enough efficiency from either
  algorithm. Keep refining/debugging the
  algorithms
• Find a final combined algorithm
• Different algorithm depending on calo cluster
  shape?
• Implement an outside-in for the first track, then
  inside-out for the second
• Use stereo silicon seeding for early conversions,
  inside-out for later ones
• How to reject p0s? Track h-matching? Can I get
  two tracks and get an energy match constraint, or
  a vertex constraint?