Rare B Decays for BSM
at Tevatron and LHC
S. Uozumi , Kyungpook National University
15-21th July 2012 International Conference on
Beyond The Standard Model @ Quy Nhon, Vietnam
NP?
NP?
Bs(d)  m+m- for New Physics
• FCNC with EW loop diagram -> Highly suppressed in SM
• SM prediction (A.J. Buras arXiv:1012.1447):
- BR(Bs0 g μ+μ-) = (3.2 ± 0.2)×10-9
- BR(Bd0 g μ+μ-) = (1.0 ± 0.1)×10-10
-
Note : SM prediction could need additional scaling by -9%
• Experimentally clean & simple di-muon signal
• Sensitive for several NP contributions :
•MSSM: BR~tan6β
•2HDM : BR~tan4β with H+
•Br. Fraction could even be reduced than SM by unthought
NP effects
BSM 2012 @ Quy Nhon
2
Bs(d) g m+m- Method : Relative normalization counting
Normalization mode, from PDG
–
–
–
Measure the rate of Bs(d) → μ+μ- decays relative to norm. mode B+ g J/y K+
LHCb uses additional normalization modes Bs g J/y f, B0 g K+pApply same sample pre-selection criteria
Further purification of Bs(d) → μ+μ- sample is done by
cut-based / Neural Network (NN) / Boosted Decision Tree (BDT)/ event selection
• Count events in signal regions
– Signal region defined according to mass resolution
– Note : Dm(Bs-Bd) = 86.8 MeV
• BG estimation done by sideband events
BSM 2012 @ Quy Nhon
MC simulation of
Bs and Bd→ m+mfor CDF case3
Bs(d)g
+
mm
Backgrounds
Dominant Backgrounds
Signal
m
Sequential semi-leptonic decay: b → cμ-X → μ+μ-X
Double semileptonic decay: bb → μ+μ-X
Continuum μ+μ- , μ + fake, fake+fake
Bs
Peaking Background in signal region (Bhh)
double
Sequential
m
m
b
c
b
BSM 2012 @ Quy Nhon
m
b
m
c
m
b
4
S/B discrimination parameters
Several kinematic parameters used to purify Bs(d) -> mm signal
- PT(B), PT(m)
-2D/3D proper decay lengths
and those significance
- Isolation
- 2D/3D Pointing angles (Da3d)
- Impact parameter of Bs(0) / muons
and those significance
- Vertex Fitting c2
Most of analyses use multivariable discrimination
parameter
(BDT or NN) for S/B
discrimination
pTm m
I  mm
pT + i pTi
d 0 , Lxy , L3d , l
l
Impact parameter
5
Mmm for Bd g m+m- @CDF-II 9.7 fb-1
•Events are binned according to NeuroBayes NN discriminant output
•Count events in each bin.
• Br. upper limit
Br(Bd g m+m−) < 4.6 (3.8) × 10−9 at 95% (90%) C.L.
SM : (0.10 ± 0.01)×10-9
•Observed events are fairly consistent with background.
-> Works as an important check on BG estimation also for Bs g m+m case
6
Bs g m+m- Double-sided Br. Limit @ CDF-II 9.7 fb-1
CDF II Preliminary 9.7 fb-1
Central-Central
muons
Central-Forward
muons
Double-sided limit :
SM
prediction
Single-sided upper limit (95% CL) :
p-values assuming
Bs background only : 0.94%
Bs SM+background : 7.1%
-> ~2s away from SM (3.2 + 0.2 x 10-9)7
ATLAS Bs(d) g m+m- Result (PLB 713 (2012), pp. 387-407)
–
–
–
–
2.4 fb-1 @ 7 TeV pp collision data
Utilize BDT analysis with 14 variables
Events are binned in 3 rapidity regions
Signal counting is done in each bin
Median expected (95% CL):
SM : (3.2 ± 0.2)×10-9
8
CMS Bs(d) g m+m- Result (JHEP 04 (2012) 033)
–
–
5 fb-1 @ 7 TeV, cut-based analysis optimized
for barrel and endcap regions separately
B0 and Bs are separated thanks to good
mass resolution
Median expected (95% CL):
SM (Bd->mm) : (0.10 ± 0.01)×10-9
(Bs->mm) : (3.2 ± 0.2)×10-9 9
LHCb Bs(d) g m+m–
–
–
–
(PRL 108, 231801 (2012))
Data is from 1 fb-1 in 2011 runs
RICH detector available for fake Kaon ID
BDT analysis separated into 2 stages :
– 1st stage … based on 6 variables
– 2nd stage … Event classification using BDT output with 9 variable
Count number of events in each bin of BDT output value
10
LHCb Bs(d)g m+m- Result
Median expected (95% CL):
Unbinned likelihood fit to Mmm
Projection with 8 BDT bins give
s
SM (Bd->mm) : (0.10 ± 0.01)×10-9
(Bs->mm) : (3.2 ± 0.2)×10-9
Most stringent limit achieved
by one experiments
11
Bs g m+m- LHC combined Result
• Three LHC results are combined
as of June-2012 (CMS PAS BPH-12-009)
LHC combined upper limit at 95% CL :
SM (Bd->mm) : (0.10 ± 0.01)×10-9
(Bs->mm) : (3.2 ± 0.2)×10-9
12
Bs g m+m- Results summary
PLB 713 (2012), pp. 387-407
JHEP 04 (2012) 033
LHC combined
(<4.2 x 10-9)
PRL 108, 231801 (2012))
• Measured Upper limits are very close to SM prediction value
• clear “observation” expected soon followed by “measurement”.
• Precise comparison with SM will be strong tool for BSM search!
13
b sm+m- for New Physics
• Strongly suppressed in SM due to 2nd order weak interaction
• New physics in a loop can change event shapes from SM
NP?
• Many channels with different spectator quark flavors available
to explore b g sm+m- decays for BSM
• CDF, LHCb and B factories have measurements of
several parameters in b sm+m- decays
• differential Br. Fraction (dBr/dq2)
• Forward-backward asymmetry (AFB)
• Angular parameters (FL, AT(2), Aim)
• Isospin asymmetry (AI)
•Another exciting stage to search BSM !
14
b sm+m- decays @ CDF 9.6 fb-1
World’s
First
Observation!
15
b sm+m- diff. Br fraction @ CDF
J/y, y’ veto
• Measurement of Br.
fraction as a function
of q2=Mmm2
• SM prediction well
matches to data
16
b sm+m- diff. Br fraction @LHCb
B0→K*μμ
Bs→φμμ
• Beautiful consistency with SM (LHCb-CONF-2012-008/003)
(except slight deficit in B0 g K0mm low q2 region,
correlating with discrepancy on low-q2 isospin asymmetry)
17
b sm+m- angular analysis
•There are various event-shape parameters in b sm+m- angular analysis to
access physics beyond the SM (arXiv:1108.0695).
Forward-backward asymmetry AFB :
K* polarization parameter FL
Transverse polarization asymmetry AT(2)
T-odd CP asymmetry Aim
Isospin asymmetry AI
18
b sm+m- AFB results
• CDF (9.6 fb-1), LHCb (1 fb-1) and B factories have B0->K(*)mm AFB results
• Every result is consistent with SM prediction.
LHCbCONF2012-008
• Also interesting measurement is
crossing point at AFB=0, since theory
predicts the crossing point accurately.
•LHCb performed world’s first
measurement :
Again consistent with SM prediction
(arXiv:1105.0376).
19
b sm+m- other angular analysis result
• CDF full 9.6 fb-1 analysis
• No significant deviation from SM
found in every parameters…
20
b sm+m- other angular analysis result
S3 ∝ AT(2)(1−FL) :
the asymmetry in K∗0
transverse polarization
• LHCb 1 fb-1 analysis
• Again, No significant deviation
from SM.
21
b sm+m- Isospin asymmetry
SM expectation is zero for any q2 value
LHCbPAPER2012-011
• LHCb B g K mm channel shows
4.4s deviation from zero.
(B factories have also shown asymmetry)
• CDF result is consistent with
SM (=0).
22
Summary
• Heavy flavor sector has been taken an important rule for
various SM measurements for a long time.
• Also to explore BSM, Rare B decays are very interesting stage
to see possible NP contribution through SM-suppressed loop
diagrams.
• Now Tevatron passes the torch to LHC experiments.
• Rapidly growing LHC data is remarkably improving several
B rare decay measurements.
- b g dmm decay : LHCb already has 5s measurement of Br. Fraction
(LHCb-CONF-2012-006)
- Measurement of CP asymmetry in radiative B decays b g s g :
LHCb reported current world”s best result
• As well as discovery of the Higgs-like boson, heavy flavor
sector is also entering to very exciting era. Stay tuned !!
23
Backups
24
Bs g m+m- Results summary
25
D0 Bs(0)g m+m- Result
Use 6.1 fb-1 in D0 runII data
– Final event selection done by Bayesian
Neural Network
– Mass resolution 120 MeV > Dm(Bs-B0)
-> measure admixture of Bs/B0 signal
–
Red : SM x 100
Median expected (95% CL):
Update with full 10 fb-1 will come soon !
BSM 2012 @ Quy Nhon
26
Bs/d g m+m- Method: Relative normalization counti
ng
–
–
•
Same method with previous 6.9 fb-1 results last summer (PRL 107, 191801)
This analysis updates results with whole 9.7 fb-1 of CDF Run-II data
Measure the rate of Bs/d → μ+μ- decays relative to B+ J/yK+ control sample
1. Apply same event pre-selection to both sample
2. Bs → μ+μ- sample is highly purified by Neural Network (NN) event selection
3. NN takes 14 kinematic variables and output value ranging 0 (BG like)~1 (signal like)
4.Bs → μ+μ- candidates are categorized into 8 different NN value bins for counting
5. BG shape is determined from di-muon mass sideband fitting
Final Bs → μ+μ- sample
Original Bs → μ+μ- and B+ J/yK+ samples
NN out=0.700-0.760
..
...
Pre-selection
Pre-selected
B+
J/yK+
Pre-selected
Bs → μ+μ-
NN out=0.987-0.995
Neural Network
Typial BG fitting
For one NN output
bin
NN out=0.995-1.000
Bs/Bd
Signal
region
Mmm for Bs g m+m- with CDF-II 9.7 fb-1
CDF II Preliminary 9.7 fb-1
Central-Central
muons
Central-Forward
muons
•Best fit estimates a 2s excess of signal over background
yielding
28
CDF Bs g m+m- Double-sided Br. Limit with 9.7fb-1
SM
prediction
with SM prediction
Br (3.2  0.2) 10-9
p-values assuming
Bs background only : 0.94%
Bs SM+background : 7.1% … ~2s away from SM
Single-sided upper limit : Br(Bs g m+m−) < 27 × 10−9 at 95% C.L.
29
Bs g m+m- Prospects at LHC
LHCb alone
CMS alone
–
–
LHC is entering to era of Br. measurement
order of 10-9 !
What we will see ?
– Br. Enhancement in order of 10-9 ?
– Consistent Br. with SM?
– Decrease of Br. by NP?
BSM 2012 @ Quy Nhon
Exciting stage to search NP in
heavy flavor sector !
30
Bs  m+m- : an MSSM case
Rouzbeh Allahverdi, Bhaskar Dutta, Yudi Santoso
arXiv:0912.4329
SM
Expectation
Br : (3.2  0.2) 10-9
+
Possible large Br
enhancement
by tan6b
Slepton/sqqark
mass (GeV)
CDMS II
Excluded by
a
1)Rare B decay b  sg Gaugino mass (GeV)
2)No CDM candidate
b
BSM 2012 @ Quy Nhon
31
c
3)Muon
magnetic moment
Reconstructed b sm+m- events and Br. fractions
CDF II 6.7 fb-1, Phys. Rev. Lett 107, 201802 (2011)
Br(Λb0→Λμ+μ-)
=[1.73±0.42(stat)±0.55(syst)]×10-6
Br(Bs0→fμ+μ-) =[1.47±0.24±0.46]×10-6
Br(B+→K+μ+μ-) =[0.46±0.04±0.02]×10-6
Br(B0→K*0μ+μ-) =[1.02±0.10±0.06]×10-6
Br(B0→K0μ+μ-) =[0.32±0.10±0.02]×10-6
Br(B+→K*+μ+μ-) =[0.95±0.32±0.08]×10-6
World’s first observation !!
Other B0/+g K m+m- channels have comparably large statistics
with B factories & LHC.
32
Summary & Conclusions
• B rare decays are excellent stage to search for New Physics effect.
•Using full 9.7 fb-1 of CDF Run-II data sample, we measured :
• Latest CDF result stays ~2s from SM
• Various b g sm+m- decay channels are observed with 6.8 fb-1 of data :
Br(Λb0→Λμ+μ-) =[1.73±0.42(stat)±0.55(syst)]×10-6… world’s first!!
Br(Bs0→fμ+μ-) =[1.47±0.24±0.46]×10-6
Br(B+→K+μ+μ-) =[0.46±0.04±0.02]×10-6
Br(B0→K*0μ+μ-) =[1.02±0.10±0.06]×10-6
Br(B0→K0μ+μ-) =[0.32±0.10±0.02]×10-6
Br(B+→K*+μ+μ-) =[0.95±0.32±0.08]×10-6
•AFB result with full CDF Run-II data analysis will come out soon!
Control sample
• Roughly selected with pre-selection
cuts same with ones for Bsg m+m- preselection
• Control sample used for Br.
normalization and cross checks for
background modeling.
Opposite-sign mm, B decay length < 0 (OS-)
Same-sign mm,
B decay length > 0 (SS+)
Same-sign mm,
B decay length < 0 (SS-)
One m is tagged as fake,B decay length > 0
(FM+)
Prediction from
B+ g J/y K+ sample
Observed in
Bs g mm BG control sample
34
Further BG discrimination by Neural Network
• NN input variables
•
•
•
•
•
•
•
3D pointing angle
Isolation
Proper decay length
Proper decay length sig.
PT(Bs)
PT(m)
Etc...
•NeuroBayes with 14
discriminating parameters
strongly distinguish signal
and BG.
Multi-variable analysis : Neural Network
Unbiased optimization based on MC signal
and data sidebands Input
Discriminating variables
..
..
..
..
nodes
Output
(0 ~ 1)
BG shape modeling from Mmm Sideband
• Events are separated in each NN output bins
• In each bin, continuum BG is estimated from SB polinomial fitting
• < 5 GeV is rejected from the fit to avoid B g m+m- + neutrinos events)
36
Mmm distributions for Bd / Bs signal region
Completely consistent with
BG For Bd g m+m- channel.
On Bs g m+m- channel,
No evident excess observed,
while best fit result is
sizeable (SMx4.1) in highest
NN bin.
37
Bs g m+m- Branching fraction limit
SM
prediction
p-values assuming
Bs background only : 0.94%
Bs SM+background : 7.1%
Again with SM prediction
Br : (3.2  0.2) 10-9 38
bg sm+m- Branching fraction measurement
- Again Relative normalization counting • Starts with di-muon trigger datasets
with 6.8 fb-1
•Control sample events are pre-selected
by kinematical cuts
• Signal candidates are selected by
Neural Network in addition to pre-selection
for further purification
• Ratio of Br. Fraction of signal and control sample is determined by
Signal event
Yield (w/ NN cut)
NN cut
efficiency
From PDG
Control sample yield
Pre-selection
(only with pre-selection )Efficiencies ratio
From PDG
39
Reconstructed b sm+m- events and Br. fractions
Br(Λb0→Λμ+μ-)
=[1.73±0.42(stat)±0.55(syst)]×10-6
Br(Bs0→φμ+μ-) =[1.47±0.24±0.46]×10-6
Br(B+→K+μ+μ-) =[0.46±0.04±0.02]×10-6
Br(B0→K*0μ+μ-) =[1.02±0.10±0.06]×10-6
Br(B0→K0μ+μ-) =[0.32±0.10±0.02]×10-6
Br(B+→K*+μ+μ-) =[0.95±0.32±0.08]×10-6
World’s first observation !!
Other B0/+g K m+m- channels have comparably large statistics
with B factories.
40
Event-shape (AFB) analysis for b g sm+m- decay
• There are various event-shape parameters to access physics beyond the SM
(arXiv:1108.0695.)
• Most interesting parameter is FB asymmetry as a function of q2=Mmm2
compared with muon polar angle in B rest frame :
Now we have 6.8 fb-1 from CDF Run II !!
41
B→K*μ+μ-
AFB analysis with 6.8 fb-1
(simultaneous fit of K*0 and K*+
channels)
B+→K+μ+μ-
Belle result
• CDF provides b g sm+m- AFB measurement
comparable with B factories!
42
Summary & Conclusions
• B rare decays are excellent stage to search for New Physics effect.
•Using full 9.7 fb-1 of CDF Run-II data sample, we measured :
Latest CDF result stays ~2s from
• Various b g sm+m- decay channels are observed with 6.8 fb-1 of data :
Br(Λb0→Λμ+μ-) =[1.73±0.42(stat)±0.55(syst)]×10-6… world’s first!!
Br(Bs0→φμ+μ-) =[1.47±0.24±0.46]×10-6
Br(B+→K+μ+μ-) =[0.46±0.04±0.02]×10-6
Br(B0→K*0μ+μ-) =[1.02±0.10±0.06]×10-6
Br(B0→K0μ+μ-) =[0.32±0.10±0.02]×10-6
Br(B+→K*+μ+μ-) =[0.95±0.32±0.08]×10-6
• Observed AFB is comparable with B factories.
Bs0  m + m Rare decay Bs  m
0
+ -
m : FCNCs, forbidden at tree level
SM Diagrams
SM Expectation
Br : (3.2  0.2) 10-9
JHEP 1009 (2010)106
Box Diagram
Penguin Diagram
2HDM
NP Expectation
Br enhancement
Penguin Diagram
Powerful Probe to New Physics
44
The CDF Detector
•Silicon Vertex Detector
(precise position measurement)
•Tracking Chamber
(momentum measurement)
•1.4 Tesla Solenoidal magnet
• Calorimeter (red and blue)
(energy measurement)
• Muon Detector (yellow & blue)
General purpose collider detector for many physics progr
• B physics
• QCD / Top quark properties
• Electroweak physics (W, Z bosoms)
• Higgs / Exotic physics
The CDF Detector
2.0
Silicon vertex (SVX-II) detector
Central tracker drift chamber
Time-of-flight (TOF) system
EM and HAD calorimeters
3.0
Muon detector
h =1.0
TOF
Front-end, DAQ and trigger electronics
CDF Trigger
– 2 central muons
“CMU”, |h|<0.6,
– pT>1.5 GeV
– 2.7<Mmm<6.0 GeV
– pT(m)+ pT(m) >4 GeV
• “CF”:
entries / 10 MeV/c2
Data collected using dimuon trigger
• “CC”:
J/y
region
s ( M J /y )  16MeV / c2
– one central, one forward muon
“CMX”, 0.6<|h|<1.0
– pT>2 GeV
– other cuts same as above
Trigger efficiency same for muons from J/y or
Bs (for muon of a given pT) 47
analysis
region
2) Background from two-body hadronic B decays
Two-body B→ hh decays where h produces a fake
muon can contribute to the background
• fake muons dominated by p+, p-,K+, K• fake rates are determined separately using D*-tagged
D → K-p+ events
Estimate contribution to signal region by:
• take acceptance, Mhh, pT(h) from MC samples. Normalizations derived from
known branching fractions
• convolute pT(h) with pT and luminosity-dependent m-fake rates. Double fake
rate ~0.04%
48
Fake rates from D*-tagged D0 → K-p+ events
Example of D0
peaks in one bin of
pT, used to extract a
pT and luminositydependent fake rate
K-
K+
for K+ and KKaons passing muon selection:
K-
49
K+
Probing New Physics
• “Smoking gun” of some Flavor Violatin
g NP models:
– ratio BR(Bs→ m+m- BR(B0→ m+m- highly inform
ative about whether NP violates flavor significan
tly or not
– clear correlation between CP violating mixing ph
ase from Bs→ J/yf and BR(Bs→ m+m-
Altmannshofer, Buras, Gori, Paradisi, Straub,
Nucl.Phys.B830:17-94,2010
• Important complementarity with direct
searches at Tevatron and LHC
– Indirect searches can access even higher mass s
cales than LHC COM energies
New bounds on BR(B0→ m+m- and BR(Bs→ m+m- are of crucial importance, and are a
top priority at the Tevatron and LHC.
50
Probing New Physics
Plenary talk
A.Buras, Beauty 2011:
51
Determination of the p-value
Ensemble of background-only pseudo-experiments is used to determine a p-value for a
given hypothesis
• for each pseudo-experiment, we do two fits Log Likelihood Distribution
and form the log-likelihood ratio
of pseudo-experiments
2ln( Q)

with Q 
L(s + b | data)
L(b | data)
for background-only hypothesis
• in the denominator, the “signal” is fixed to
zero (I.e. we assume background only), and

in the numerator
s floats
• L(h|x) is the product of Poisson
probabilities over all NN and mass bins
• systematic uncertainties included as
nuisance parameters, modeled as Gaussian.
Result: the p-value for the
background-only hypothesis is 23.3%
52
for B0→ m+m- signal window