Search for B s m + m - and B d m + m - Decays at CDF Cheng-Ju S. Lin (Fermilab)
description
Transcript of Search for B s m + m - and B d m + m - Decays at CDF Cheng-Ju S. Lin (Fermilab)
Search for Bs+- and Bd+- Decays at CDF
Cheng-Ju S. Lin(Fermilab)
The 13th International Conference on Supersymmetry and Unification of
Fundamental Interactions
IPPP Durham 18 July 2005
Cheng-Ju Lin (Fermilab) SUSY05 2
Introduction
• In the Standard Model, the FCNC decay of B +- is heavily suppressed
910)9.05.3()( sBBR
• SM prediction is below the sensitivity of current experiments (CDF+D0): SM Expect to see 0 events at the Tevatron
(Buchalla & Buras, Misiak & Urban)
• Bd is further suppressed by CKM coupling (vtd/vts)2
SM prediction
Any signal would indicate new physics!!
3
BEYOND STANDARD MODEL
’i23 i22
b
s
R-parity violating SUSY
• In many SUSY models, the BR could be enhanced by many orders of magnitude:
• For examples: - MSSM: Br(B) is proportional to tan6BR could be as large as ~100 times the SM prediction
- Tree level diagram is allowed in R-parity violating (RPV) SUSY models. Possible to observe decay even for low value of tan
• In context of mSUGRA, Br(B) search complements direct SUSY searches: (A. Dedes et al, hep-ph/0207026) Low tan() observation of trilepton events Large tan() observation of Br(B)
4
PROBE OF NEW PHYSICS• New physics may enhance Bs and Bd differently
• Minimal-flavor-violation (MFV) assumption in SUSY yields SM relations between Bs and Bd decays
• Can observe both Bs and Bd: unique to Tevatron
• CDF has the mass resolution to distinguish two decays, M~23MeV : unique to CDF
• Either observation or null search, results will provide important clues about possible scenarios of new physics beyond SM
Monte Carlo
M(Bs)-M(Bd)~90MeV
5
TEVATRON
Main Injector
Tevatron
DØCDF
• Tevatron is the highest energy collider in the world Ecm(pp) = 1.96 TeV
• RunII physics run began in Mar 2001 (>1 fb-1 delivered so far) Expect 8fb-1 (design spec) by 2009.
• B production cross-sec is ~30b at Tevatron (1nb at PEPII)
• All B species are produced (B+ : Bd : Bs : b,)
CDF II Detector
z
y
x))2/ln(tan(
Central Drift Chamber(COT)
Silicon Vertex Detector (L00, SVXII, ISL)
Central Muon Chambers: CMU, CMP(|| < 0.6)
Central Muon Extension: CMX(0.6< || < 1.0)
Important components relevant to the analysis highlighted:
SuperconductingSolenoid (1.4T)
Cheng-Ju Lin (Fermilab) SUSY05 7
Data Sample
• Using 364pb-1 of data (Feb 02 – Aug 04) from di-muon triggers:- CMU(P) + CMU (central-central)- CMU(P) + CMX (central-extended)
• Central-central and central-extended channels treated independently in this analysis (background and efficiencies are different)
Search region
Rare B di-muon triggersrequires additional cuts to reduce background relative to inclusive J/ di-muon trigger
Cheng-Ju Lin (Fermilab) SUSY05 8
Ingredients of the Analysis
Key elements in the analysis: - Construct discriminant to select Bs signal and suppress bkg - understanding the background - accurately measure the acceptance and efficiency ratios
Analysis optimization: Figure of merit expected 90% C.L. upper limit Performed unbiased optimization
Overall picture: - Reconstructing di-muon events in the B mass window - Measure the branching ratio or set a limit Normalized to BJ/ K decays
)()()(
BRKBBR
f
f
N
NBBR
s
u
B
totalBB
totalBsBs
Bss
9
Reconstruct “normalization mode”
Selection Requires:• pT(B)>4 GeV && |y(B)|<1• pT(K+)>1GeV• good vertex fit quality• > 2
= proper decay length (B+) = ~502m]
B+K+ in rare B trigger Sample: N(CMU/P-CMU) = 1785±60
N(CMU/P-CMX) = 696±39
Count the # of B++-K+ candidates
)(3
Bp
McL D
CMU-CMU
Cheng-Ju Lin (Fermilab) SUSY05 10
For Bs+-:
Pre-selection requires:• pT(B)>4 GeV && |y(B)|<1 • > 2 • good vertex fit quality
Bs Search Sample:N(CMU-CMU) = 22459N(CMU-CMX) = 14305
(completely Bgd dominated)
“Baseline” cuts are loose cuts toreject events that are clearly background
Bs Sample Selection “baseline cuts”
CMU-CMU
Background shapes are linearfor both channels
[ (Bs) = ~502mm]
Cheng-Ju Lin (Fermilab) SUSY05 11
– Invariant +- mass, Mvtxwithin +/- 60MeV (2.5)
– Proper decay-length ():
– Isolation (Iso):
(fraction of pT from B within R=(2+2)1/2 cone of 1)
– “pointing ()”: (3D opening angle between Bs momentum and decay axis)
)(3
Bp
McL vtxD
i iiTT
T
RpBp
BpIso
)0.1()(
)(
))(( 3DLBp
We have explored various discriminating variables for selectingB events and suppress bkg. The chosen ones are:
Enhance Bs,dand Suppress Background
Cheng-Ju Lin (Fermilab) SUSY05 12
cut cut
To further reduce Bgd, we apply the additional cuts:
<0.70 rad
&&Iso > 0.50
eff(signal) ~ 92%
This leaves in data:
N(CMU-CMU) = 6242
N(CMU-CMX) = 4908
Discriminating Variables (Sig vs BKG)
~x3 down in bkg but still…
Cheng-Ju Lin (Fermilab) SUSY05 13
• We use a likelihood ratio method:
ii
i
xPxPxP
ibis
isLH
)()(
)(
Ps/b is the probability for a given sig/bkg to have a valueof x, where i runs over all discriminating variables.
• The chosen variables are: - isolation (iso) - 3D pointing (),
- proper decay length probability [P()=exp(-/Bs)]
• PDF for the individual likelihood is reconstructed from the data sideband for background and Pythia MC for signal
Likelihood Ratio Discriminant
Cheng-Ju Lin (Fermilab) SUSY05 14
Likelihood PDFs and LH Ratio
Input PDFs for Likelihood Ratio Likelihood Ratio Discriminant(CMU/P-CMU channel)
(Isolation)
Prob()
(Pointing Angle)
15
Background Estimate
• Extrapolate the number of events in the sidebands to the signal region and scale by the expected rejection of the likelihood ratio cut• Use toy-MC to estimate bkg rejection of likelihood ratio cut based on input distributions from data sidebands
LH CMU-CMU CMU-CMX cut pred obsv pred obsv
>0.50 236+/-4 235 172+/-3 168OS- >0.90 37+/-1 32 33+/-1 36 >0.99 2.8+/-0.2 2 3.6+/-0.2 3
>0.50 2.3+/-0.2 0 2.8+/-0.3 3SS+ >0.90 0.25+/-0.03 0 0.44+/-0.04 0 >0.99 <0.10 0 <0.10 0
>0.50 2.7+/-0.2 1 3.7+/-0.3 4SS- >0.90 0.35+/-0.03 0 0.63+/-0.06 0 >0.99 <0.10 0 <0.10 0
>0.50 84+/-2 84 21+/-1 19FM+ >0.90 14.2+/-0.4 10 3.9+/-0.2 3 >0.99 1.0+/-0.1 2 0.41+/-0.03 0
1.) OS- : opposite-charge dimuon < 02.) SS+ : same-charge dimuon > 03.) SS- : same-charge dimuon < 04.) FM : fake muon sample (at least one leg failed muon stub chi2 cut)
Control sample for bkgestimate cross-check:
Note: Using a wide ±100MeV window around B mass for bkg cross-check
16
Compute Acceptance and Efficiencies
• Most efficiencies are determined directly from data using inclusive J/events. The rest are taken from Pythia MC.
• (B+/Bs) = 0.297 +/- 0.008 (CMU-CMU) = 0.191 +/- 0.006 (CMU-CMX)
• LH(Bs): ranges from 70% for LH>0.9 to 40% for LH>0.99
• trig(B+/Bs) = 0.9997 +/- 0.0016 (CMU-CMU) = 0.9986 +/- 0.0014 (CMU-CMX)
s
uLHBs
KrecoBvtx
Bs
vtxB
recoBs
recoB
trigBs
trigB
Bs
B
f
f
1
• reco-(B+/Bs) = 1.00 +/- 0.03 (CMU-CMU/X)
• vtx(B+/Bs) = 0.986 +/- 0.013 (CMU-CMU/X)
• reco-K(B+) = 0.938 +/- 0.016 (CMU-CMU/X)
Red = From MC
Green = From Data
Blue = combination of MC and Data
fu/fs=3.83±0.57 (HFAG 2004)
17
We used the set of requirements which yielded the minimum a priori expected BR Limit:
0
),,()|( %90%90
obsnbgbg
CLsignalbgobs
CLsignal nNnnN
Poisson prob of observingnobs when expecting nbg
90% CL UL on Nsignal whenexpecting nbg bkgd evts
using Bayesian Method (w/ flat prior)and including uncertainties
Analysis Optimization
)/()(
)(1)(
/
%90
KJBBrf
f
N
NBBR
s
dtotBs
tot
BBsLHKJB
CLsignal
s
Quantity which vary in optimization
• For optimization, we scan over a range of LH values
• Assume 1 fb-1 of data Optimal cuts: LH>0.99
Cheng-Ju Lin (Fermilab) SUSY05 18
Looking at Results
For optimized cuts of LH >0.99 and pT(B) > 4GeV and a 60 MeV window around world avg B mass
CMU-CMU Channel CMU-CMX Channel
We observed 0 event in the signal region!
Cheng-Ju Lin (Fermilab) SUSY05 19
Bs Summary:CMU-CMU:
Single event sensitivity = (1.0±0.2) ×10-7 Expected # bkg (364pb-1) = 0.81 ± 0.12
CMU-CMX:Single event sensitivity = (1.6±0.3) ×10-7 Expected # bkg (336pb-1) = 0.66 ± 0.13
We observed 0 event which yields a combined limit of: Br(Bs) < 1.5×10-7 @ 90% CL ; 2.0×10-7 @ 95% CL
Bd Summary:We also observed 0 event which yields a combined limit of: Br(Bd) < 3.8×10-8 @ 90% CL ; 4.9×10-8 @ 95% CL
Both Bs and Bd results are a x2 better than best published results! D0 PRL 94 (2005) 042001 (240pb-1) BaBar PRL 94 (2005) 221803 (111fb-1)
B Limits
20
mSUGRA M0 vs M1/2
• We are beginning to carve into mSUGRA space
• For mh~115GeV implies 10-8<Br(Bs)<3×10-7
Solid red = excluded by theory or experimentDashed red line = light Higgs mass (mh)Dashed green line = (a)susy (in units of 10-10)Black line = Br(Bs)
Dedes, Dreiner, Nierste, PRL 87(2001) 251804
M0
[GeV
]
Excluded
Excluded
21
mSUGRA M0 vs M1/2
• We are beginning to carve into mSUGRA space
• For mh~115GeV implies 10-8<Br(Bs)<3×10-7
Solid red = excluded by theory or experimentDashed red line = light Higgs mass (mh)Dashed green line = (a)susy (in units of 10-10)Black line = Br(Bs)
Dedes, Dreiner, Nierste, PRL 87(2001) 251804
M0
[GeV
]
Excluded
Excluded
Excluded by thisnew result
22
R. Dermisek et al., JHEP 0304 (2003) 037
SO(10) Unification Model
• tan()~50 constrained by unification of Yukawa coupling
• White region is not excluded
• Unification valid for small M1/2
(~500GeV)
Red regions are excluded by either theory or experimentsGreen region is the WMAP preferred regionBlue dashed line is the Br(Bs) contourLight blue region excluded by old Bs analysis
h2>0.13
m+
<10
4GeV
mh<
111G
eV
23
SO(10) Unification Model
• New Br(Bs limit strongly disfavors this solution for mA= 500 GeV
Red regions are excluded by either theory or experimentsGreen region is the WMAP preferred regionBlue dashed line is the Br(Bs) contourLight blue region excluded by old Bs analysis
h2>0.13
m+
<10
4GeV
mh<
111G
eV
Excluded by thisnew result
R. Dermisek et al., JHEP 0304 (2003) 037
Cheng-Ju Lin (Fermilab) SUSY05 24
Summary
• Bs is a powerful probe of new physics. Could potentially provide the first hint of SUSY at the Tevatron
• Using 364pb-1 of data, CDF has obtained world best limits on Bs and Bd channels: CDF: Br(Bs) < 1.5×10-7 (2.0×10-7) @ 90% (95%) CL Br(Bd) < 3.8×10-8 (4.9×10-8) @ 90% (95%) CL
• TEVNPWG is combining CDF and D0 limits. Expect ~20% improvements over CDF limits alone • The limits are now starting to constrain interesting regions of SUSY parameter space
• We have covered an order of magnitude since Run I. Will cover at least another order of magnitude before the end of RunII
25
BACKUP SLIDES
26
Correlations Between Variables
yx
N
iii
xy
yyxx
N
1
)ˆ)(ˆ(
1
1
Correlations between discriminating variablesare negligible: straighforward to construct likelihood discriminant (see next slide)
27
Checking MC Modeling of Signal LH
For CMU-CMU:
• MC reproduces Data efficiency vs LHood cut to 10% or better
• Assign 10% (relative) systematic
• CMU-CMX MC vs Data agreement is better
Compare B+ Data and MC
28
For CMU-CMX:
• MC reproduces Data efficiency vs LHood cut to 5% or better
• Assign 5% (relative) systematic for CMU-CMX
Checking MC Modeling of Signal LH
29
Likelihood Ratio Efficiency for Bs Signal
LH(Bs) cut CMU-CMU CMU-CMX
LH>0.90 (70+/-1)% (66+/-1)%LH>0.92 (67+/-1)% (65+/-1)%LH>0.95 (61+/-1)% (60+/-1)%LH>0.98 (48+/-1)% (48+/-1)%LH>0.99 (38+/-1)% (39+/-1)%
• determined from Bs MC
• MC modeling checked by comparing LH(B+) between MC and sideband subtracted Data
(stat uncertainties only)
30
Since discriminating variables are uncorrelated, use toy MCto estimate RLH based on input distributions from data SB
Estimate LH Bkg Rejection (RLH)
KS-Prob(CMU)=11%KS-Prob(CMX)=5%
Likelihood Ratio Data vs Toy MC
RLH Cut CMU-CMU CMU-CMXLH>0.85 0.0245+/-0.0005 0.0226+/-0.0005LH>0.92 0.0130+/-0.0004 0.0120+/-0.0003LH>0.99 0.0014+/-0.0001 0.0015+/-0.0001
Likelihood Ratio Rejection from Toy MC
(Errors are stat only)
LH strongly suppresses bkg
31
RPV SUSY EXCLUSION
B. Dutta et al, PLB 538 (2002) 121
’i23 i22
b
s
R-parity violating SUSY
• Possible to exclude phase space even for small tan()
• Exclusion strongly depends on the coupling.
Excluded
32
B Sensitivity To Heavy Higgs
MFV MSSM (tan=50)(A. Dedes et al, hep-ph/0407285)
(Excluded by Bs)
• Br(Bs) is sensitive to the mass of heavy Higgs
• If the branching ratio is measured sets an upper limit on the mass of the heavy Higgs
• mA mass limit is where BR crosses the green curve (same argument hold of any tan≤ 50)
• The mass limit is fairly model independent
33
Bs Limit Projection
• Extrapolate based on the current analysis which was optimized for 1/fb
• Assume background and single-event-sensitivity scale linearly with luminosity
• Will need to re-optimize the analysis for > 3/fb