Vertex Reconstruction Status

Current Status Highlights

Since last Collaboration Meeting
- Independent tests by UPC/LFS and HF groups with real data (Thanks!)
- No major changes in vertex finding algorithms
Highlighted modifications introduced under the S&C's watch
- Beam line constraint is available and treated the same way in both PPV and MinuitVF
- With beamline3D option the full track and the beam line error matrices are used in the fit
- New functionality does not affect past behavior, i.e. the previously existed options are backward compatible
- No change in primary track candidate selection
- Re-vertexing is now possible with muDst tracks

naive 1D fit: Position of local peaks in binned (and weighted) track distribtion along \( z \)
1D fit: By TMinuit with one free parameter—vertex position along \( z \). Forced to stay on beam line
3D fit: By TMinuit with vertex coordinates \( (x, y, z) \) as free parameters

naive 1D fit: Position of local peaks in binned (and weighted) track distribtion along \( z \)
1D fit: By TMinuit with one free parameter—vertex position along \( z \). Forced to stay on beam line
3D fit: By TMinuit with vertex coordinates \( (x, y, z) \) as free parameters

Many thanks to S. Radhakrishnan and X. Dong for testing beamline3D option in Run 16 dAu 200 GeV data. See S&C Meeting, April 26, 2017

Observed expected features including

Sharper vertex peak in \( x-y \) plane due to vertex "attraction" to beam line and similar distributions without the beam line (HFT tracks!)
Wider DCA distribution with beamline

Many thanks to L. Adamczyk, L. Fulek, and R. Sikora for performing embedding and data studies with PPV. Tested recent tunes of beamline options and re-vertexing with muDst. See S&C Meeting, September 25, 2017

Consistent results with beamline and beamline3D. 3D fit improves resolution (although not used directly in this analysis)
Beam line helps to recover low multiplicity (1-2 tracks) vertices
Using only BTOF-matched tracks in vertex fit leads to two-fold increase in signal

In STAR fast detectors are used to mitigate effect on vertex position by pile-up tracks in slow TPC
Track weight is based on the hit/signal and status of detector cell to which track is projected
- Track weights then translate to overall vertex rank
Track Points to Cell Cell Status Cell Signal Track weight

✔ ✔ ✔ higher
✔ ✔ ✗ lower
✔ ✗ — no change
✗ — — no change

Track Points to Cell	Cell Status	Cell Signal	Track weight
✔	✔	✔	higher
✔	✔	✗	lower
✔	✗	—	no change
✗	—	—	no change

Tests by UPS/LFS uncovered inconsistencies in BTOF matching. See details in G. Van Buren's blog
BTOF track matching is available in three variants:
- StGenericVertexFinder: Module-based matching used in vertex ranking
- StTrack2FastDetectorMatcher: Aggregate vertex flags based on cell-based match
- StBTofMatchMaker: Sets matching flags for StMuTracks
BTOF matching flags for muDst tracks do not reflect track matching in vertex finders
Similarly aggregate vertex counters for daughter tracks are not based on StMuTrack flags
The above may lead to inconsistent track selections depending on fields used in muDst trees

Independent tests by PWGs added confidence in recent Vertex Finder modifications
Re-vertexing is now possible with muDst tracks
Moving towards solutions for remaining issues
- Ongoing discussion with TOF team on a possibility to converge to a single BTOF matching algorithm
- A related issue with BTOF is its geometry based on Geant volumes instead of the ROOT's TGeoManager
- Common approach to misaligned TGeo allow for better matching accuracy
- Considering an option of splitting StiMaker into a global and primary track makers with detector-matching and vertex finding run in between
- Lessons learned from BTOF should be extended to ETOF and calorimeters