Track reconstruction and physics analysis in LHCb

1
Track reconstruction and physics analysis in LHCb

• Outline
• Introduction to the LHCb experiment
• Track reconstruction ? finding and fitting
• Physics analysis ? event selection and
  sensitivity study
• More details in my thesis Track simulation
  and reconstruction in LHCb

Seminar Particle and Astrophysics U Zürich,
Physik Institut 07 December 2005, Jeroen van
Tilburg, NIKHEF
2
Reminder CP violation
CKM matrix connects the quark mass eigenstates
with the weak interaction eigenstates
e-i?
CKM matrix
e-iß
ei?
Complex phases in matrix elements ? CP violation
3
The Large Hadron Collider
The LHCb detector
The LHC tunnel
CERN, Geneva
4
The LHCb detector
VELO
5
Track types
Velo tracks used to find primary vertex.
Long tracks used for most physics studies B
decay products.
T tracks improve RICH2 performance.
Upstream tracks improve RICH1 performance,
moderate p estimate
Downstream tracks enhance KS finding.
Different track types, different algorithms
6
Track event display
Outer tracker station
VELO
TT
T2
T3
T1
7
Example Matching algorithm
Matches T tracks with VELO tracks to find long
tracks ? estimate momentum of T track ?
extrapolate T track through magnet to the VELO ?
find best match (based on ?2 cut). ? add TT hits
8
Matching algorithm estimate p
Estimate momentum of the T track with p-kick
method ? Magnetic field is an instant kick at
focal plane zzmagnet. ? Assume track originates
from interaction point. ? Re-evaluate center of
magnet (zc).
p-kick method
p-kick
zmagnet
zc
VELO
T seed
T stations
Bdl 4.2 Tm
9
Matching ?2
Efficiency 91.2 Wrong combinations 4.8
p gt 5 GeV
10
Adding TT hits for matched tracks
? Extrapolate matched tracks to TT stations. ?
Group the hits depending on distance to track. ?
Find best group of TT hits.
Group the hits Distance d to track lt 10 mm ?d in
same station lt 1 mm ?d in other station lt 2
mm Group has at least 3 hits Hit can belong gt 1
group
11
Adding TT hits
Select the group with the lowest q2.
q2 d2 w2spread sd2
Distance deviation of group
Average distance of group
12
Long track performance
13
Long track performance
ghost rate
efficiency
g 7.7 (pgt5 GeV)
e 94.3 (pgt5 GeV)
14
Tracking robustness
relative multiplicity
Tracking is robust against number of interactions
15
Track fit
The tracks are fitted using the Kalman Filter.

• The Kalman Fit properties
• Adds measurements recursively.
• Mathematically equivalent to least ?2 method.
• Multiple scattering and energy loss can be
  naturally included.

16
Outlier removal
Outliers (hits with high ?2 contribution) can be
removed. ? requires a refit ? remove only 1 hit
per iteration
17
Outlier removal (long tracks)
Number of iterations
Improves ?2 distribution
18
Fit quality (long tracks)
19
Momentum resolution
LHCb provides an excellent momentum estimate at
the vertex.
Reconstructed tracks
Ideal tracks
Note Fitted with single Gaussian in each bin.
20
Impact parameter _at_ vertex
21
Physics analysis

• Two benchmark decay channels of LHCb
• Bs ? Ds p measures ?ms (Bs oscillation
  frequency)
• Bs ? Ds K measures ?-2? (CP violation)

• For my thesis I studied the
• event selection for these decays, and the
• final sensitivity on ?ms and ?-2?

22
Branching fractions
Decay channel Branching fraction Annual
production Bs ? Ds p 1.2 10-4 26 M
events Bs ? Ds K 1.0 10-5 2.1 M
events
Event topology
23
Bs ? Ds K and Bs ? Ds K
Bs ? Ds K
and Bs ? Ds K
Included two similar channels
K ? K0 p (67) ? half decays to Ks0
K p0 (33)
Ds ? Ds ? (94)
Event topology
24
Selection strategy

• Preselection to reduce background ? using
  standard LHCb applications (DaVinci and LoKi)
• Remove specific backgrounds ? using a single
  cut
• Tune remaining cuts against generic background?
  using an optimisation tool

25
1. Preselection
Loose cuts
26
2. Specific background
Bs?Dsp background in Bs?DsK selection
? cut on RICH likelihood
27
2. Specific background
For instance, cut at ?lnLKp3 gives
Fit both mass distributions simultaneously to
find the number of signal events (S) and its
error (sS).
50 MeV
28
2. Specific background
Vary ?lnLKp cut to find the optimum with respect
to the statistical significance of the signal
29
3. Generic background

• Optimisation tool
• Optimise remaining cuts simultaneously
• Divide each selection variable into equidistant
  bins.
• Scan the total selection space.
• Find the combination of cuts for which is
  maximal.

30
Final selection cuts
31
Efficiencies and yield
Efficiencies quoted in .
Lower detection efficiency
Low yield
Need to cut harder due to high background
32
Decay time resolution and pull
Pull distribution
Resolution
33
Acceptance function
Selection and trigger cuts reduce efficiency at
zero decay time
After selection and trigger
34
Sensitivity study
Matter
Antimatter
35
Sensitivity study

• Use Toy Monte Carlo and Fitting Program
• Generate events according to expected annual
  yield and with realistic time errors from full
  simulation.
• Account for acceptance function.
• Perform an unbinned likelihood fit to observed
  decay time distribution.
• Fit both Bs?Dsp and Bs?DsK events
  simultaneously.

36
Observed decay times
Bs?DsK 3 years
37
Default parameters
38
Computing power
Submitted 10k jobs (experiments) on the
DataGrid
39
Oscillation frequency
Sensitivity on ?ms
?ms deviation for 100 experiments
Amplitude method
After 1 year
40
Sensitivity on weak phase
Weak phase ?-2?
Error bars represent RMS fluctuation.
Sensitivity for 100 experiments after 3 years.
1 year s 15.2º
41
Conclusions

• Different track reconstruction algorithms
  developed for the different track types (e.g. the
  matching algorithm).
• The LHCb experiment provides an efficient track
  reconstruction of 94 with a ghost rate of 8
  (pgt5 GeV).
• LHCb has an excellent spatial (42 um) and
  momentum resolutions (0.35) at the interaction
  point.
• Three-step event selection for Bs?Dsp and Bs?DsK
  provides a sufficient background reduction.
• After 1 year of running LHCb can measure ?ms up
  to 88 ps-1 and ?-2? with an error of 15.2º.