3 May 2003,

1
Tracking performance in LHCb
IV INTERNATIONAL SYMPOSIUM ON LHC PHYSICS AND
DETECTORS FermiLab, Chicago May 1-3 2003
Tracking Performance
Jeroen van Tilburg NIKHEF On behalf of the LHCb
collaboration

• Overview
• LHCb setup
• Velo/ST/OT performance
• Track finding
• Track fitting

2
LHCb setup
VELO
3
Vertex Locator

• Vertex Locator
• 21 stations.
• R/f sensors (single sided, 45º sectors).
• Pitch ranges from 37 µm to 103 µm.
• 220 µm thin silicon.
• Sensitive area starts at only 8 mm from beam
  axis.

4
Silicon Tracker
The Silicon Tracker Two parts 1. IT (Inner
Tracker) 2. TT (Trigger Tracker)
5
Inner Tracker

• Inner Tracker
• 3 stations (T1-T3) with 4 layers each
  (0,5,-5,0).
• 320 µm thin silicon.
• 198 µm readout pitch.
• 130 k readout channels.

After clustering
30
65
6
Outer Tracker

• Outer Tracker
• 3 stations (T1-T3) with 4 double layers
  (0,5,-5,0).
• 5 mm straws.
• Fast drift gas (Ar(75)/CF4(15)/CO2(10)) ? Signal
  collection lt 50 ns.
• 25 ns beam crossing ? spillover from previous
  and next spills.
• Straws are 4.7 m with readout on top and below
  (long modules).
• 50k readout channels.

Short prototype module
OT double layer cross section
Track
5mm straws
e-
e-
e-
pitch 5.25 mm
e-
e-
7
Outer Tracker
Average occupancy in OT 4 (hottest region 7
)
Cross shape determined by restricting the OT
occupancy
Core s 200 µm
Resolution Tails due to low momentum secondaries
p gt 2 GeV
8
Track finding

• Track finding challenges in LHCb
• High density of hits and tracks.
• Track pattern recognition must be fast (in
  trigger and offline).
• High track efficiency important (especially for
  many-prong decays).

9
Track finding
Zoom of OT station (hits in red)

• Track finding challenges in LHCb
• High density of hits and tracks.
• Track pattern recognition must be fast (in
  trigger and offline).
• High track efficiency important (especially for
  many-prong decays).

10
Track finding algorithms

• Velo tracks
• Find straight line segments in Velo.
• Start search for triplets in R-z projection.
• Then add f hits and extend track to other
  sensors.
• Important for finding primary vertex.
• Efficiency 97, ghost rate lt 5.

Velo tracks
11
Track finding algorithms

• Forward tracks
• Starts with Velo track and find continuations in
  TT and T1-T3.
• Uses optical model.
• Accurate measurement of momentum.
• Long track important for most physics studies.
  B decay products.
• Efficiency 90.
• Afterwards the used hits are discarded for use
  in remaining algorithms.

Velo tracks
Forward tracks
12
Track finding algorithms

• Seed tracks
• Stand-alone track finding in stations T1-T3.
• Tracks almost straight lines (parameterized as
  parabola).
• First search for x-hits then add stereo hits.
• Improves RICH2 performance.

Velo tracks
Forward tracks
Seed tracks
13
Track finding algorithms

• Matched tracks
• Try to match Seed tracks with Velo tracks.
• First, estimate momentum from deflection of Seed
  track.
• Then extrapolate Seed track to Velo. Match with
  Velo track.
• Finds remaining long tracks additional to
  Forward tracks.
• Adds 2 to efficiency for long tracks.

Velo tracks
Forward tracks
Seed tracks
Matched tracks
14
Track finding algorithms

• Velo ? TT (VTT)
• Finds tracks without hits after the magnet
  (momentum too low).
• Start with unused Velo tracks.
• Find a continuation of at least 3 hits in TT.
• Magnetic deflection before TT moderate momentum
  estimate ?p/p20.
• Improve RICH1 performance, slow pions, kaon
  tagging. Efficiency 75.

Velo tracks
Forward tracks
Seed tracks
Matched tracks
VTT tracks
15
Track finding algorithms

• T ? TT (or Upstream)
• Find tracks without hits in Velo.
• Start with unused Seed tracks and try to add
  hits in TT.
• Estimate momentum from deflection Seed track.
• Final momentum estimate ?p/p0.4
• Enhance KS finding. Pion efficiency 74.

Velo tracks
Forward tracks
Seed tracks
Matched tracks
VTT tracks
T ? TT tracks
16
Track finding algorithms

• Finally apply clone killing algorithm.
• Select the best candidate among tracks that
  share many hits.

Velo tracks
Forward tracks
Seed tracks
Matched tracks
VTT tracks
T ? TT tracks
Many track types, many algorithms
17
Event display
Average efficiency 92 Efficiency for B
daughters 95
Event withaverage occupancy
Red measurements (hits)
Blue reconstructed tracks
18
Tracking performance
Efficiency vs p
Ghost rate vs pcut
Ghost rate vs pTcut
Long tracks
Long tracks
95
8
8
Long tracks
Total ghost rate 16 Ghost rate pTgt0.5 GeV
8. Large event to event fluctuations.
Average efficiency 92 Efficiency for pgt5GeV gt
95
19
Robustness tests
Tracking is robust against number of interactions

• Tracking is also robust against
• Lower hit efficiencies,
• Decreased hit resolutions,
• More noise hits.

• Note
• Luminosity adjusted to have the maximum number
  of single interactions.
• Pile-up veto trigger (L0) rejects multiple
  interaction events.

20
Track fit
The tracks are fitted using the Kalman Filter.

• The Kalman Fit
• The prediction step.
• The filter step. Adds measurements one-by-one.
• The smoother step.

direction of the filter
track prediction
filtered track

• The Kalman Fit properties
• Adds measurements recursively.
• Mathematically equivalent to least ?2 method.
• Needs as input initial track estimate.
• Multiple scattering and energy loss are
  naturally included.

21
Track fit resolution
LHCb provides an excellent momentum estimate at
the vertex.
Momentum resolution core s 0.35 2nd s 1.0
(fraction 0.1)
Note Fitted with single Gaussian in each bin.
percent
?p/p
22
Mass and vertex resolutions
Good tracking performance essential input for
physics analysis

• 4-prong prompt decay ? all long tracks.
• Total tracking efficiency 84.60.5 ( 95.6
  per track)
• Good resolutions

Decay channel Bs ? Ds-(?KKp) p
23
Ks reconstruction

• Reconstruction of Ks(? pp-) challenging
• Long decay lengths ( 1 m) many decay outside
  Velo.
• Tracks dont point to interaction point.
• Leave less hits in detector.

Where does the Ks decay? 25 in Velo ? Long
tracks 50 between Velo and TT ? T?TT
tracks 25 after TT ? Lost
pp- mass of selected B ? J/? Ks

• Example B ? J/? Ks
• Combine oppositely charged T?TT tracks.
• pT gt 250 MeV.
• Common vertex.
• Tracking efficiency for Ks 54 (74 per track).

s10.90.6 MeV/c2
24
Conclusions

• Tracking system provides good spatial and
  momentum resolutions. Vertex IP resolution
  (1732/pT) µm, Cluster resolutions 45 µm (ST),
  200 µm (OT),
• Momentum resolution 0.35.
• Many algorithms developed for finding tracks in
  optimised setup. Tracking efficiency 95 for
  B-daughters, For pions from Ks 74. (54 both
  tracks), Ghost rate 8 (for pTgt0.5GeV).
• Tracking is robust against worse conditions.
• Provides excellent input for physics
  analysis e.g. Bs ? Ds-p Mass resolution 13
  MeV/c2 Proper time resolution 42 fs.

25
(No Transcript)
26
BACKUP SLIDES
27
Event display
Average efficiency 90.6 Efficiency for B
daughters 95
Event with average occupancy
Red measurements (hits) Blue reconstructed
tracks
28
Tracking performance
Long tracks
Mean momentum 13 GeV On average 37 measurements
(Velo, IT, OT)
29
Efficiency definition

• Efficiencies are normalised to a sample of
  reconstructable particles
• in VELO at least 3 r and 3 f hits,
• in T stations at least 1 x and 1 stereo hit in
  each station T1-T3.
• Long tracks must be reconstructable in VELO and
  T.
• VTT tracks must be reconstructable VELO and at
  least 3 TT hits.
• TTT tracks must be reconstructable in T and have
  at least 1 hit in TT.
• Successfully reconstructed track at least 70 of
  hits from one MC particle.
• Long tracks must be successfully reconstructed
  in VELO and T,
• VTT tracks must be successfully reconstructed in
  VELO and have at least correct 1 hit in TT,
• TTT tracks must be successfully reconstructed in
  T and have at least 1 correct hit in TT.

30
LHCb classic setup
LHCb light setup
VELO