CP violation in B0s decays at CDF

1
CP violation in B0s decays at CDF
ICHEP08 Philadelphia
29 Jul 5 Aug
Diego Tonelli Fermilab for the CDF Collaboration
2
New Physics in the B sector
B-factories large NP disfavored in tree
processes. If any, look in loops
STANDARD MODEL
NON STANDARD MODEL
W
B0, K0, B0s
B0, K0, B0s
u, c, t
u, c, t
B0, K0, B0s
B0, K0, B0s
W
New Physics factorized into a complex amplitude
Bottom line to constrain NP need to measure
strength and phase
3
Experimental picture (Dec 2007)
B0 mixing
K0 mixing
B0s mixing
LQCD-dominated uncertainty
Experimentally-dominated uncertainty. This
measurement is todays topic
4
Role of b? ccs transitions

W, ?
b
b
s
c

?
u, c, t
u, c, t
c
W
B0s
B0s
B0s
s
F
s
s
s
b
W , ?
Mixing phase sensitive to NP
Tree b?ccs phase 0
Time-evolution
5
Analysis overview
Combine everything in a Maximum Likelihood fit
Sensitivity to phase increases if mesons of
different production flavor are treated
separately
6
Signal extraction and CP-determination
1.4/fb, 2000 decays, S/B2
B0s? J/ ?(?µµ-)F(?KK-) NN maximizes S/v(SB).
Trained on MC for signal and mass-sidebands for
background.
Determine CP of final state from angular
correlations.
7
Flavor-tagging performance
Same tagging used successfully for
mixing-frequency measurement b-quarks mainly
produced in b/bbar-pairs at the Tevatron
Opposite Side looks at decay of the other
b-hadron in the event Same Side exploits the
charge/species correlations with associated
particles produced in hadronization of
reconstructed B0s meson
OST efficiency 96 /- 1 OST dilution 11 /-
2 SST efficiency 50 /- 1 SST dilution 27
/- 4 Total eD2 4
Output decision (b-quark or antib-quark) and the
probab. of being correct
8
Wrapping up all together in a fit
9
Data-driven checks
Angles
Flavor tagging
Mass-lifetime
OST tuned on B SST tuned on MC, checked on
mixing measurement a posteriori
Measurement w/o flavor tagging ?Gs and ts See
talk by Sinead Farrington, here at 11.21 AM
Measured polarization of B0??K consistent w/
B-factories (and competitive!)
10
Non-Gaussian Likelihood
1s and 2s Likelihood sections in the (?Gs, ßs)
plane. All samples below generated with same
true values of parameters!
Non-Gaussian Likelihood (a) symmetries dependent
on choice of strong phases (undetermined from
data) (b) sensitivity to some parameters vanish
for values of other parameters
Not quote central values and their uncertainties.
Use interval estimation (confidence regions)
instead
11
Ensuring coverage
Coverage guaranteed at quoted CL Remap the
observed 2?logL distribution to obtain coverage
accounts for non-Gaussian and non-asymptotic
behavior of Likelihood. E.g. to get the 68 CL,
climb the likelihood by 3.4 units (instead of 2.3
of the asymptotic case)
1s
1-CL
2s
ideal
reality
Include systematics by varying nuisance
parameters within 5s of their estimates on data
and choosing worst case to define the region
2?log(L)
12
Results
2D projection of the multidimensional region in
the space of all (27) fit parameters a specific
value of ?G and ßs is excluded only if it can be
excluded for any assumed values of the nuisance
parameters (within 5s from their nominal
values). No assumptions on strong phases just
data! PRL100, 161802(2008)
Assuming the SM, the probability of observing a
fluctuation as large or larger than observed in
data is 15 (1.5s) One dimensional 0.16 lt ßs lt
1.41 at 68 CL
13
Tevatron combination
ICHEP update
D0 observes a fluctuation consistent with CDF
(see J. Ellison just after me) Combine CDF and D0
iso-CL regions previously checked for coverage
2.2s consistency with
SM. 0.24 lt ßs lt 0.57 OR 0.99 lt ßs
lt 1.33 at 68 CL
hep.physics.indiana.edu/rickv/hfag/combine_dGs.ht
ml
14
Hot off the press2.8/fb update!
Same-side tagger NOT yet used in second half of
sample. PID calibrations still to be
finalized. Equivalent to reduced sample size
2.8/fb ? 2/fb
ICHEP update
3200 decays, S/B2
www-cdf.fnal.gov/physics/new/bottom/080724.blessed
-tagged_BsJPsiPhi_update_prelim/
Once the SST will be calibrated have 20 signal
events by using PID info in selection x3
tagging power in second-half of the sample
15
ICHEP update
Increased dataset still hints at larger than SM
values! Consistency with SM decreased 15 ? 7
(1.8s)
ICHEP update
0.28 lt ßs lt 1.29 at 68 CL
-pi/2 lt ßs lt -1.45 OR -1.01 lt ßs lt -0.57 OR -0.13
lt ßs lt pi/2 at 95 CL
www-cdf.fnal.gov/physics/new/bottom/080724.blessed
-tagged_BsJPsiPhi_update_prelim/
Will shrink further with PID in the whole dataset
16
Summary and outlook
First direct determination of NP phase in B0s
mixing using flavor-tagged decays. PRL100, 161802
(2008). 1.5s agreement with SM.
Halved allowed space of NP parameters.
SM agreement reduces to 2.2s when combined with
D0, which observes consistent result with CDF.
First update on larger dataset confirms old
result and provides tighter constraints 1.5s ?
1.8s wrt SM, although several powerful tools
still being finalized.
CDF on its way to 5-6/fb by 2009. 50 signal
from extension to other triggers. Keep improving
the analysis. L behaves better with higher stat.
17
Backup
Collider Detector at Fermilab
www-cdf.fnal.gov
18
Neutral flavored mesons
Extremely rich phenomenology. (Approximated)
time-evolution
Hamiltonian eigenstates (definite mass and
lifetime) are mixtures of flavor eigenstates
Experimentally accessible quantities
Oscillation frequency
measured 18 ps-1
measured 5-15 x G
Decay-width difference
???
CP-violating phase
19
Fs and ßs are not the same phase
ßs arg - VtbVts/VcbVcs 2.2o (SM) phase
of b?ccs transition that accounts for decay and
mixingdecay. Fs arg-M12/G12 0.24o (SM)
argM12 arg(VtbVts)2 matrix element that
connects matter to antimatter through
oscillation. argG12 arg(VcbVcs)2
VcbVcsVubVus (VubVus)2 width of matter and
antimatter into common final states.
Both SM values are experimentally unaccessible by
current experiments (assumed zero).
If NP occurs in mixing Fs
FsSM FsNP 2ßs 2ßsSM FsNP ? standard
approximation Fs -2ßs
20
Experimental requirements
CDF strengths.
High muon acceptance (84 azimuthal at etalt1.5)
and precise muon ID
Vertex position known with 25 µm uncertainty
Calorimeter for electron ID used in flavor
tagging
dE/dx in drift chamber (1.5s _at_pgt2 GeV/c) and TOF
(2s _at_plt1.6 GeV/c) provide pion/kaon ID crucial in
flavor tagging
Excellent vertexing to resolve fast oscillations
(silicon detector) and momentum resolution for
improving S/B (large radius drift chamber)
immersed in 1.4 T B field.
21
Trigger/NN selection

• ONLINE
• Use 1.3-1.7/fb of dimuon trigger data
• Muons with pTgt1.5 in 0lt?lt0.6 and with pTgt2.2
  in 0.6 lt?lt1.0
• DImuon 2.7lt mµµ lt4.0 GeV/c2

B0s use pT and vertex quality J/? use pT and
vertex prob. F use mass and vertex qual. PID
(dE/dx TOF) K from F
22
Transversity basis
Two different reference frames
J/? at rest
F at rest
State at time t decomposed in polarizations
longitudinal to direction of motion (CP-even),
polarizations transverse and each other
(CP-even), polarizations transverse and - each
other (CP-odd). PLB 369, 144 (1996)
23
Quality of the angular fit
24
Cross-check sample
B0?J/?K0 high-statistics test of angular
efficiencies and fitter
ANN removes K0 mis-reconstructed with swapped
Kp assignment
Agrees with latest Babar results. PRD 76,031102
(2007) Actuallycompetitive -)
25
Tagging calibration and performance
26
More on SST
27
More on OST
28
SST calibration
29
Angular model cross-check
MC
MC
MC
Sideband data
Sideband data
Sideband data
30
Signal probability density
B0s term
Flavor tagging
anti-B0s term
Angular sculpting from MC. Deviations from flat
indicate detector effects. Cross-checked in data
(more later)
31
Flavor-specific time evolutions
32
CP-violating terms
Knowledge of B0s mixing frequency needed
Strong phases d- argA-A0, d argAA0,
33
Pathologic likelihood
Biases and non-Gaussian estimates in
pseudo-experiments. Strong dependence on true
values for biases on some fit parameters.
fits on simulated samples
a) Dependence on one parameter in the likelihood
vanishes for some values of other parameters L
looses degrees of freedom
e.g., if ?G0, d- is undetermined
b) Likelihood symmetries. L invariant under there
are 4 equivalent minima
34
Procedure
Define a grid in (?Gs ßs) space and for any
point -fit data with ?Gs ßs fixed _at_
point -generate toys with all true values from
above fit -fit each toy twice -determine
local p-value by counting fraction of toys with
LR larger than in data.
parameters that maximize L ? nuisance
parameters that maximize L at fixed DGs,bs
35
Dilution asymmetries
Insert an artificial 20 asymmetry in dilution
between matter and antimatter and check the
effect on confidence region
36
Visual effect of constraints
Constraining the strong phases greatly increases
the regularity of the Likelihood and improves the
result. However, no robust theoretical prediction
exist for the strong phases. Typical choice is to
relate them to the B0?J/?K0 phases assuming
SU(3) symmetry
37
Externally-constrained results
1s intervals
1s interval
Add constrain on strong phases from Babars
J/?K0, hep-ex/0411016 and on equal B0s and B0
widths. 2ßs ? 0.40,1.20 _at_ 68 CL
NP-independent assumption.
Constrain G12 0.048 /- 0.018 in
?G
2G12cos(Fs). 2ßs? 0.24,1.36U1.78, 2.9 _at_ 68
CL
38
Contours New vs Old
w/o tagging Likelihood gets sensitive to sin2ßs
instead of sin2ßs. Solutions increase 2? 4 due
to additional likelihood symmetries.
39
Strong phases matter
cos(d-)lt0 and cos(d- - d)gt0
cos(d-)gt0 and cos(d- - d)lt0
40
UTfit claim
Arxiv0803.0658v1hep-ex March, 5, 2008
Some caveats Do not account for non-Gaussian
tails. Some guesswork to remove from D0 results
the assumptions they put in. I do not believe
the 3s significance figure is rigorously derived.
41
A complementary approach
If NP is present, observed ßs -F/2. Accessible
through inclusive measurement of semileptonic
decay asymmetries. Measure asymmetry between
and - - dimuon pairs originated in semileptonic B
decays.
Select same-sign dimuons Select those from b by
looking at impact parameter distributions Extract
asymmetry by correcting for bck/detector
effects Subtract from it the B0
contribution Obtain the B0s contribution