1. Studies of EPR-type flavor entangled states in Y(4s)->B0B0
We want to provide a Time Dependent Asymmetry (TDA)
fully corrected for experimental effects (background, wrong
tags, resolution) to test alternative models to QM.
AQM(Dt)=cos(Dmd Dt)
AB 17-Feb-06

2. Steps of this analysis selection of the events background subtraction
correction for wrong tag events
data deconvolution
checks
- 5.1 OF+SF=B0 lifetime
- 5.2 second analysis with selection on Dz resolution
6) comparison with models
MC heavily needed to do the subtractions/corrections of
point 1 to 3 and to produce the deconvolution matrices!
AB 17-Feb-06

3. Selection of D*ln events and tagging
"B0 side":
B0 D* l n
D*  D0 + slow pion
D0  Kp Kpp0 K3p
The used particles are then discarded and the
remaining particles are the ”Tag side".
These remaining particles are sent to standard BELLE
procedure to tag the B0 flavour.
We select only the lepon tagged subset with highest
purity parameter r>0.875.
AB 17-Feb-06

4. Event selection summary
AB 17-Feb-06

5. EVENT SELECTION : A few plots... Plepton in cms K,p momenta for D0K p
Data
distributions shown
are obtained with
all the “other” cuts applied MC
pre-analysis cut MC
AB 17-Feb-06

6. A few plots... Momenta for K3p
shifted in opposite
direction
(vertex fit ?)
±0.1 GeV
MC-data
relative
uncertainty
AB 17-Feb-06

7. A few plots... M(D0) and M(D*)-M(D0)
MC shifted by ~0.12 MeV
Data width larger by 0.14 MeV
adjust cuts to get equivalent
configurations
M(D*)-M(D0)
AB 17-Feb-06

8. A few plots... impact parameters
not perfect
but cuts are
large...
±0.01 cm
AB 17-Feb-06

9. A few plots... chi2 from vertex fit
± 1bin = ± 5...
AB 17-Feb-06

10. A few plots... s(Vz) from vertex fit
Artificially rescaled
by 1.125
Errors from vtx fit
do not agree.
N.B.: s(Vz) cut only used
in control analysis
with “high resolution”
± 2bin = ± cm
AB 17-Feb-06

11. A few plots... cos(qB,D*l) Fit using MC shapes: D*ln B0D**ln
cut: |cos(qB,D*l)|<1.1
AB 17-Feb-06

12. Background subtraction Wrong tag evts correction
- Continuum
analyzed 8.3 1/fb of off resonance data => 0 evts
Fake D*
Wrong D* lepton combinations
B± to D0**
Wrong tags
AB 17-Feb-06

13. Raw time dependent asymmetry1 -1
After event selection
syst
stat
Background subtracted ps
AB 17-Feb-06

14. Fake D* control sideband nominal sideband
From sideband : background evts 126±6 and 54±4 OF and SF
Control sideband ±6 and 54±4
AB 17-Feb-06

15. Fake D* from sideband control sideband Effect on the asymmetry
MC truth
data with
systematics
change distributions
by 1 sigma + move cuts
False D* and/or false slow pion
AB 17-Feb-06

16. Wrong D*-lepton combinations
Reversed lepton technique:
in the CMS take Plepton := - Plepton and redo the analysis
78 evts OF and 237 SF
Effect
on the
asymmetry
data with
systematics
change distributions
by 1 sigma + move cuts
main effect from large number of SF...
AB 17-Feb-06

17. Wrong D*-lepton combinations
Reversed lepton technique: method check with MC events
reversed lepton
MC truth
prediction from MC:
20 evts OF and 200 SF
consistent
with data
AB 17-Feb-06

18. Background subtraction B±
Fit using MC shapes:
D*ln
B0D**ln
B±D0**ln
shapes
~identical !
- Need to correct only for B±,
because B0 has mixing
MC gives relative efficiency
Assume MC BRs ratio...
Systematics:
~8% from fit
~20% from BR(YB±  D0**)/BR(Y  B0  D**)
AB 17-Feb-06

19. Background subtraction B±MC
normalized
254 OF
1 SF
Systematics:
~8% from fit
~20% from BR(YB±  D0**)/BR(Y  B0  D**)
AB 17-Feb-06

20. Wrong tag correction Systematics: ±0.010 ±1 sigma in corrections
High purity events: MC predicts 0.015±0.001(stat)
Similar analysis (Phys. Rev. D ) gives 0.020±0.005
We consider 0.015±0.005.
Expect
(1-2w)-1 = 1.033±0.010
attenuation of the TDA
Systematics:
±0.010
±1 sigma in corrections
~3% attenuation
double
counting...
AB 17-Feb-06

21. Construction of the raw TDA
AB 17-Feb-06

22. Raw time dependent asymmetry1 -1
After event selection
syst
stat
Background subtracted ps
AB 17-Feb-06

23. Data deconvolution (see appendix A)
Data correction is done by a program based on a Singular
Value Decomposition (SVD) of the deconvolution matrices.
The matrices are constructed from a MC set of events
which are supposed to reproduce detector effects.
Problems:
MC is not perfect (main problem: reproduction of Dz resolution
(we add an extra smearing of 46 ± 40 mm)
2) Unfolding contains regularization/filtering to get rid of
elements which are statistically non significant, with the
MC shapes used as a priori => can discard Data information
3) We have MC events generated only with QM hypothesis,
given DM, B0 lifetime (solution: mix OF and SF sets)
AB 17-Feb-06

24. Corrected time dependent asymmetryQM
for QM we use fitted value Dmd = h/s
AB 17-Feb-06

25. Discussion on systematic errors
Sytematic errors on the TDA can originate from
1) errors in background subtraction (in the note)
2) intrinsic behaviour of deconvolution procedure (appendix A)
3) imprecision of the Belle full MC (note and appendix B)
2) will be discussed next presentation
AB 17-Feb-06

26. Systematics: selection, bckgrd, wrong tag
fit errors in
B± fraction
and BRs ratio
cuts (appendix B)
 M(D*)-M(D0)
 cos(qB,D*l)
±1s variation
in SF/OF distrib
w = 1.5±0.5%
±1 %
 ±1s variation
in SF/OF distrib
M(D*)-M(D0)
different sidebands
AB 17-Feb-06

27. Method used to get systematic errors in evt selection
for each observable X
1) we first estimate D-MC relative precision on X
then, in a way which should reduce the effect of statistical fluctuations
2) we evaluate syst. error from the integrated asymmetry distribution as a function of X
AB 17-Feb-06

28. Exemple: slow pion vertex chi2±5
Asymmetry vs chi2
Dt>3ps
linear fit
AB 17-Feb-06

29. Exemple: slow pion chi2 comparison DATA & MC
Dt>3ps
data
Dt>3ps MC
AB 17-Feb-06

30. Error estimates (~low Dt)
total ±0.001
AB 17-Feb-06

31. Error estimates (Dt>3 ps)
our choice
of syst errors:
total ±0.005
0-1 ps ± 0.001
1-6 ps ± 0.005
6-20 ps ± 0.010
AB 17-Feb-06

32. Analysis Checks
AB 17-Feb-06

33. Alternative analysis with s(Vz)<0.01 cm
Select events with better resolution in Dz
=> deconvolution procedure has less to do ...
rescale by 1.125
AB 17-Feb-06

34. Alternative analysis with s(Vz)<0.01 cm
Dt>3ps
Dt>3ps
0-1 ps ± 0.003
1-6 ps ± 0.019
6-20 ps ± 0.04
syst errors:
AB 17-Feb-06

35. Analysis with s(Vz)<0.01 cm
N of events reduced by ~18%
s(Vz)<0.01
Errors shown are ~ unfolding  d[s(Vz)]
AB 17-Feb-06

36. Other control: B0 lifetime
Plot of Deconvoluted(SF) + Deconvoluted(OF)
fit: 1.532±0.017 ps
AB 17-Feb-06

37. Results 1
c2=5/11bins
for QM we use pre-BABAR&Belle Dmd = 0.495± h/s
AB 17-Feb-06

38. Results 2 LR SD c2=74/11bins c2=229/11bins
Using pre-BABAR&Belle Dmd = 0.495± h/s
AB 17-Feb-06

39. Results 3 Let free Dmd QM: Dmd = 0.503±0.010(stat) c2 = 5/11bins
LR : Dmd = 0.434±0.009(stat) c2 = 25/11bins
SD : Dmd = 0.400±0.010(stat) c2 = 111/11bins
to be compared with pre-BABAR&Belle
Dmd = 0.495± h/s
AB 17-Feb-06

40. Results 4 Possible contamination of QM events by a fraction l of SD
i.e. Data = (1-l)QM + lSD
gives l=0.028± => compatibe with zero
AB 17-Feb-06

41. Deconvolution Goal: get resolution corrected data.
Constraint: we need to be fair with a “large” class of theoretical models.
Studied with a toy MC with parametrized resolution in Dz
Set of 400 runs, each run consists of:
Generation of ~35000 "MC" events based on QM
Generation of ~7000 "Data" events based on QM or LR or SD
Production of the 2 unfolding matrices for SF, OF events from "MC"
Deconvolution of the "Data" independently for the SF and OF subsets
1) Main problem: need to populate event region where QM does not
produce events, ex: SF~0 when Dt~0. Trick: add some OF !
2) Some residual systematic effects observed. Add correction
and get systematic errors.
AB 17-Feb-06

42. Deviations wrt generated model
stat QM LR SD
syst from deconvolution
AB 17-Feb-06

43. Sensitivity to models Generated: QM (analysis errors not included)
lambda from fit LR SD
chi2/11
lambda generated
AB 17-Feb-06

44. Results QM: chi2 = 5 for 11 bins LR: 112 " SD: 330 "
lambda = 0.02±0.05
AB 17-Feb-06