===========================================
Minutes of Tracking Meeting, August 1, 2007
===========================================
taken by U. Husemann


Max Goncharov: Space-Time Vertexing
===================================

- Motivated by search for delayed photons, see CDF Note 8015
- Vertexing is a clustering problem
- Z vertexing: search window too wide (problem at high luminosity)
- Observe tails in difference between t0 from electron track and 
  event t0 (from ZVertex): background from different collision
- Time resolution: 1.28 ns RMS
- Time adds another (almost orthogonal) dimension to vertexing problem
- Algorithm:
  1. Find number of vertices (overestimate), tracks sorted by pT
  2. Merge closeby vertices
- Probabilistic approach to vertex merging: "expectation normalization"
  -> each track is assigned a probability to belong to a given vertex
- Validation: split vertices with many tracks in two vertices and compare
- Efficiency: remove single tracks from vertex with many tracks
- Time resolution: 1.27 ns RMS
- Vertex z resolution has small but long tail from merged vertices 
  (splitting code available but not used)
- Average number of vertices grows linearly with luminosity, 
  except for highest luminosities (while ZVertexFinder is linear)

- Discussion: questions and concerns (probably incomplete)
  o There could be biases from shifted vertices
  o Algorithm uses COT-only tracks, need to add silicon to improve resolution
  o Track resolution information not used to determine vertex position
  o Resolutions in z and t: background and peak have different means
  o There is no d0 cut on tracks (ZVertexFinder has 1 cm d0 cut): 
    could the algorithm be sensitive to Kshorts?
  o Algorithm gets good time resolution from per-wire calibration 
    (needs proper calibration data from additional DB)
  o How does the performance compare with ZVertexFinder?
    . Is new algorithm superior to ZVertexFinder's primary vertex 
      assignment by largest sum of transverse momenta?
    . Jet corrections rely on ZVertexFinder's linear dependence of
      average number of vertices on luminosity
  o ZVertexFinder requires re-tuning for Gen7
    . Treatment of new track categories
    . Criteria for class-12 vertices



All: Status of int10 Validation
===============================

- Checks done so far:
  o Koji Sato: standard COT plots
  o Matt Herndon: silicon forward tracking efficiency
  o Ben, Antonio, Ulrich: started detailed studies

- More thorough validation required:
  o Oliver: 10k tests
  o Ben, Antonio, Ulrich: efficiencies for all tracking algorithms, 
    low level track parameters, silicon hit charge, ...