Vincenzo Vagnoni

1

• Vincenzo Vagnoni
• Umberto Marconi
• INFN Bologna
• Joint Physics Meeting with CERN Theory Group
• 28th LHCb Week
• February 28th 2003

2
Overview

• B0(s) ? h?h? decay selection
• Selection strategy
• Determination of the optimal cuts
• Annual yields and background over signal ratios
• CP sensitivities
• Toy Monte Carlo to generate event by event proper
  time distributions
• Evaluation of ?Adir and ?Amix after one year of
  LHCb data taking
• Two different approaches
• ?2 fit of the binned asymmetry
• Unbinned maximum likelihood estimation
• Consistency checks
• Conclusions

3
VELO 21stations (Rmin 8mm) Si 220 mm, strips R
e f
6?5 m2
3 Tracking stations IT Si strips OT straw tubes
TT1.4?1.2 m2 Si microstrips
4
B decay selection strategy

• MC samples used size (number of events)
• B0 ? pi pi- 60k
• Bs ? K K- 65k
• B0 ? K pi- 62k
• Bs ? K- pi 24k
• ?b ? p K- 20k
• ?b ? p pi- 21k
• inclusive bb background 1.144M
• Procedure followed to tune selection
• Determine offline cuts which maximize the signal
  efficiency but a) reject all the background
  from the (limited) inclusive bb sample b)
  provide good rejection of the specific background
  events
• Multivariate brute force approach to establish
  the best cut values
• grid scan over 11-dimension selection parameter
  space
• reject all the bb background events in 600
  MeV/c2 mass window
• maximize S / sqrt(SB) with B ? channel specific
  background
• Determine trigger cuts which maximize the
  efficiency for offline-selected events at fixed
  trigger output rate (determined with min. bias
  sample)

• only 20 times the signal statistics, but
  BR(signal) ltlt 1/20
• only 1/2 minute of data-taking at nominal LHCb
  luminosity !

5
B0(s) ? h?h?Selection cuts

• Charged tracks
• Each leg identified as a K, or a particle lighter
  than K (using RICH)
• Cuts on
• p
• max. pT
• min. pT
• max. IP/?IP
• min. IP/?IP
• ?2 of common vertex

• B candidate
• Cuts on
• pT
• IP/?IP
• L/?L
• mass

6
Breakup of the optimal cuts
Parameters used for selection Parameters used for selection Parameters used for selection B0 ? pi pi- B0 ? pi pi- B0 ? K pi- B0 ? K pi- Bs ? K K- Bs ? K K-
pi, K smallest pt (GeV/c) gt 0.65 gt 0.15 gt 0.3
pi, K largest pt (GeV/c) gt 3.1 gt 2.5 gt 2.25
pi, K p (GeV/c) ? 2.5, 100 ? 2.75, 200 ? 2.75, 125
pi, K smallest IP/?IP gt 5.75 gt 7 gt 4
pi, K largest IP/?IP gt 6 gt 7.25 gt 4.5
pi, K vertex fit ?2 lt 6.75 lt 6 lt 20
B pt (GeV/c) gt 1.6 gt 2.25 gt 0.6
B IP/?IP lt 2.5 lt 3.25 lt 2.75
B L/?L gt 7.25 gt 8 gt 7.25
B ?m (MeV/c2) lt 45 lt 55 lt 45
7
Rejection of physics background
B0(s) ? h?h? selections

• Relying on
• RICH PID
• mass resolution (18 MeV/c2 single Gaussian)

B0 ? pi pi-
B0 ? K pi-
BS ? K K-
8
Rejection of combinatorial background (B0? pi
pi-)
incl. bb
signal

• Assume combinatorial background dominated by bb
  events
• Can reject all generated bb background
• also when relaxing B mass cut
• Preliminary estimate

B/S lt0.7

• Contribution of ghost tracks to the combinatorial
  background is negligible

9
Inclusive bb background mass spectra
B0 ? pi pi-
B0 ? K pi-
BS ? K K-
Before cuts
After cuts
GeV/c2
GeV/c2
10
Annual yields and B/S ratios
Event type MC sample Efficiency Efficiency B/S B/S Untagged annual yield
Event type MC sample Selection Trigger specific bb inclusive (CL 90) Untagged annual yield
B0 ? pi pi- 60000 0.073 0.31 0.088 lt 0.7 27000
B0 ? pi pi-
B0 ? K pi- 62000 0.084 0.29 0.046 lt 0.2 115000
B0 ? K pi-
Bs ? K K- 65000 0.10 0.27 0.049 lt 0.5 35000
Bs ? K K-
11
Tagging

• LHCb TP (using only opposite side tagging)
• ?tag 0.4
• ?tag 0.3
• ?tag (1-2?tag )2 6.4
• Preliminary studies with the current simulation
  indicate a similar result
• Other methods to be added (analysis on going)
• Bs tag with Kaon same side
• Bd tag with Pion same side
• Jet charge
• In the following we will assume the TP numbers

12
B0 ? pi pi- BS ? K K- CP sensitivities

• Two approaches used
• ?2 fit of the binned asymmetries
• Unbinned maximum likelihood

• Tagged event samples generated by means of a toy
  MC taking into account
• Annual event yield S
• Background over signal ratio B/S
• Tagging efficiency ?tag
• Wrong tag fraction ?tag
• Proper time resolution function R(?-t)
• Acceptance ?reco(?)

B0 ? pi pi- BS ? K K-
S (1 year) 27000 35000
B/S 0.8 0.55
xq 0.755 20
?tag 0.4 0.4
?tag 0.3 0.3
Adir 0.19 -0.17
Amix 0.62 -0.27
13
B0 ? pi pi-
Proper time resolution
Proper time acceptance
Single Gaussian fit ?411 fs
ps
14
CP Asymmetryrelevant formulae
15
Example of fitted asymmetries

• All the following plots correspond to one year of
  data taking

16
Check of fit consistency ?ACP / ?(ACP)
distributions

• Fit of 500 asymmetries generated with different
  random seeds

• Distributions consistent with N(0, 1)

17
Adir vs Amix

• Scatter plots obtained with a multiquadric
  smoothing due to limited statistics (500 trials)

18
Dependence of ?Adir and ?Amix on B/S
Current estimate of B/S

• Expected error dependence on B/S reproduced

19
Unbinned maximum likelihood approach

• Maximize log likelihood with respect to Adir and
  Amix

20
log likelihood plots
21
log likelihooderror ellipses and 68.27
confidence regions
22
Check of log likelihood approach consistency

• 500 event samples generated with different random
  seeds

23
log likelihooddependence of ?Adir and ?Amix on
B/S
24
Dependence of ?Adir and ?Amix on the years of
data taking
25
Conclusions

• ?Adir and ?Amix after one year of LHCb data
  taking have been estimated
• Least ?2 fit of the binned asymmetry and unbinned
  maximum likelihood estimation give consistent
  results

B0 ? pi pi-
BS ? K K-