BEACH 04 J. Piedra 1

Search for FCNC Decays Bs(d) → μμ-

Motivation

Tevatron and CDF

Analysis Method

Results

Conclusion

Matthew Herndon, May 2006University of Wisconsin

Carnegie Mellon University High Energy Physics Seminar

2

The Standard ModelSuccesses of the Standard Model

Simple comprehensive theory

Explains the hundreds of common particles:

atoms, protons, neutrons, electrons

Explains the interactions between them

Basic building blocs

6 quarks: up, down...

6 leptons: electrons...

Force carrier particles: photon...

Predictions of the Standard Model have been verified to high precision

M. HerndonCMU Seminar

All common matter particles are

composites of the quarks & leptons

which interact by the exchange of

force carrier particles

3

Beyond the Standard ModelWhy look for physics beyond the Standard Model

Look for new physics that would explain these mysteries:

SUSY, Extra Dimensions ...M. Herndon

Gravity not a part of the SM

What is the very high energy

behaviour?

At the beginning of the universe?

Grand unification of forces?

Where is the Antimatter?

Why is the observed universe mostly matter?

Dark Matter?

Astronomical observations of gravitational

effects indicate that there is more matter than

we see

Standard Model fails to answer many fundamental questions

CMU Seminar

4

Searches For New PhysicsHow do you search for new physics at a collider?

Direct searches for production of new particles

Particle-antipartical annihilation

Example: the top quark

Indirect searches for evidence of new particles

Within a complex decay new particles can occur

virtually

A unique window of opportunity to find new physics before the LHC

M. Herndon

Tevatron is at the energy frontier and

a data volume frontier: 1 billion B and Charm events on tape

So much data that we can look for some very unusual decays

Where to look

Many weak decays of B and charm hadrons are very low probability

Look for contributions from other low probability processes – Non Standard Model

CMU Seminar

5

Bs → μμ: Beyond the SMLook at decays that are suppressed in the

Standard Model: Bs(d) →μμ-

Flavor changing neutral currents(FCNC) to leptons

No tree level decay in SM

Loop level transitions: suppressed

CKM , GIM and helicity(ml/mb): suppressed

SM: BF(Bs(d) →μμ-) = 3.5x10-9(1.0x10-10)G. Buchalla, A. Buras, Nucl. Phys. B398,285

New physics possibilities

Loop: MSSM: mSugra, Higgs Doublet

3 orders of magnitude enhancement

Rate tan6β/(MA)4

Babu and Kolda, Phys. Rev. Lett. 84, 228

Tree: R-Parity violating SUSY

M. Herndon

One of the best indirect search channels at the TevatronCMU Seminar

1.96TeV pp collider

Performance substantially improving

each year

Record peak luminosity in 2006:

1.8x1032sec-1cm-2

6

CDF and the Tevatron

CDF Integrated Luminosity

~1fb-1 with good run requirements

All critical systems operating

including silicon

Analysis presented here uses 780pb-1

Doubled data in 2005, predicted to

double again in 2006

-

Bs(d) →μμ- benefits from new dataCMU Seminar

Silicon Tracker

|η|<2, 90cm long, rL00 =1.3 - 1.6cm

Drift Chamber(COT)

96 layers between 44 and 132cm

7

CDF DetectorMuon coverage

|η|<1.5

Triggered to |η|<1.0

Outer chambers: high purity muons

EXCELLENT TRACKING

CMU Seminar

TRIGGERED TO 1.5 GeV/c

8

Rare B Decay Physics & TriggersLarge production rates

σ(pp → bX, |y| < 1.0, pT(B) > 6.0GeV/c) = ~30μb, ~10μb

All heavy b states produced

B0, B+, Bs, Bc, Λb, Ξb

Backgrounds: > 3 orders of magnitude higher

Inelastic cross section ~100 mb

Challenge is to pick one B decay from >103 other QCD events

Di-muon trigger

pT(μ) > ~1.5 GeV/c, |η| < ~1.0

B yields 2x Run I (lowered pT threshold, increased acceptance)

Di-muon triggers for rare decay physics

Bs(d) →μμ-, B+ →μμ-K+, B0 →μμ-K*0, Bs →μμ-φ, Λb→μμ-Λ

Trigger on di-muon masses from near 0GeV/c2 to the above Bs mass

Reduce rate by requiring outer muon chamber hits: pT(μ) > ~3.0 GeV/c or pTμ >

5.0GeV/c

TRIGGERS ARE CRITICAL

M. Herndon

-

1 Billion B and Charm Events on Tape

9

Bs → μμExperimental Challenge

M. Herndon

Primary problem is large background at hadron colliders

Analysis and trigger cuts must effectively reduce the large background around mBs

= 5.37GeV/c2 to find a possible handful of events

Key elements of the analysis are determining the efficiency and rejection

of the discriminating variables and estimating the background levelCMU Seminar

10

Analysis Method

M. Herndon

Performing this measurement requires that we

Demonstrate an understanding of the background

Optimize the cuts to reduce background

Accurately estimate α and ε: for triggers, reconstruction and selection cuts

SM predicts 0 events, this is essentially a search

Emphasis on accurately predicting the Nbg and performing an unbiased optimization

Blind ourselves to the signal region

Estimate background from sidebands and test background predictions in orthogonal

samples

€

BF(Bs → μ +μ−) =(Ncand −Nbg )

α BsεBs•αB +εB +

NB +

•fuf s

•

BR(B+ → J /ψK +) • BR(J /ψ → μ +μ−)

CMU Seminar

11

Data Sample

M. Herndon

Start the with di-muon trigger

2 CMU muons or 1 CMU and 1CMX muon with pTμ > 5.0GeV/c

CMU: pT(μ) > ~1.5 GeV/c, |η| < ~0.6, CMX: pT(μ) > ~2.0 GeV/c, 0.6 < |η| < 1.0

1 CMUP muon: pT(μ) > ~3.0 GeV/c and 1 CMU(X) muon

Apply basic quality cuts

Track, vertex and muon quality cuts

Loose preselection on analysis cuts

PT(μμ-) > 4.0 GeV/c, 3D Decay length significance > 2 …

In the mass region around the Bs: 4.669 < Mμμ < 5.969 GeV/c2

Blind region: 4σ(Mμμ), 5.169 < Mμμ < 5.469 GeV/c2

Sideband region 0.5 GeV/c2 on either side of the blinded region

Sample still background dominated

Expect < 10 Bs(d) →μμ- events to pass these cuts: based on previous limits

100M Events

23K Events

CMU Seminar

12

Normalization Data Sample

M. Herndon

Normalize to the number of

B J/K+ events

Relative normalization analysis

Some systematic uncertainties

will cancel out in the ratios of the

normalization

Example: muon trigger efficiency

the same for J/ or Bs muons for

a given pT

Apply same sample selection

criteria as for Bs(d) →μμ-

€

BF(Bs → μ +μ−) =(Ncand −Nbg )

α BsεBs•αB +εB +

NB +

•fufu

•

BR(B+ → J /ψK +) • BR(J /ψ → μ +μ−) 9.8 X 107 B+ events

Need to discriminate signal from background

Reduce background by a factor of > 1000

Signal characteristics

Final state fully reconstructed

Bs is long lived (cτ = 438 μm)

B fragmentation is hard: few additional tracks

Background contributions and characteristics

Sequential semi-leptonic decay: b → cμ-X → μ+μ-X

Double semileptonic decay: bb → μ+μ-X

Continuum μ+μ-

μ + fake, fake+fake

Partially reconstructed, short lived,

has additional tracks

13

Signal vs. Background

M. Herndon

+

-

L3D

primary vertex

di-muon vertex

PT()L3D

LT

primary vertex

di-muon vertex

+

-

PT()LT

-

Cut on mass, lifetime, how well pT points to the vertex and isolationCMU Seminar

23K Events

14

Discriminating Variables

M. Herndon

Mass M

choose 2.5σwindow: σ = 25MeV/c2

exp(λ=cτ/cτBs)

α : |φB – φvtx| in 3D

Isolation: pTB/( trk + pTB

)

exp(λ), α and Iso:

used in likelihood ratio

Optimization

Unbiased optimization

Based on simulated signal and data

sidebands

Optimize based on likely physics result

a priori expected BF limit

4 primary discriminating variables

CMU Seminar

15

CDF 3D Silicon Vertex Detector

M. Herndon

8 layer silicon detector

4 double sided layers with stereo

strips at a small angle

3 double sided layers layers with

strips at 90

3D vertexing is a powerful

discriminant

2D from previous

analysis

CMU Seminar

16

Likelihood Ratio

M. Herndon

Probability distributions used to construction the likelihood ratio

Use binned histograms to estimate the probabilities

Resulting likelihood with signal

MC and data sidebandsCMU Seminar

17

Background Estimate

M. Herndon

Estimate the background for

any given set of requirements

Typical method: apply all cuts, see

how many sideband events pass

Optimal cuts may be chosen to

reject 1-2 unusual events

Flat background distribution allows

use of wide mass window

Need to show that mass is distribution is linear

Need to investigate backgrounds that may peak in signal region: B hh

Design crosschecks to double check background estimates

Analysis dependent on an accurate background estimate

CMU Seminar

18

Background Estimate

M. Herndon

Expected number of combinatorial background in a 60 MeV/c2 signal

window

Extrapolation from sideband

0.99 expected to be near optimal cut, 1 background event typical of optimal search

Optimal point looks achievable

LH cut

Predicted Background

CMU-CMU

Predicted Background CMU-CMX

LH > 0.90 13.3±1.3 10.4±1.1

LH > 0.98 1.0±0.5 1.0±0.3

LH > 0.99 0.7±0.3 0.4±0.2 Cross-checks needed!

CMU Seminar

19

Background Cross-Checks

M. Herndon

Use independent data samples to test

background estimates

OS-: opposite sign muons, negative lifetime

(signal sample is OS+)

SS+ and SS-: same sign muons,

positive and negative lifetime

FM: OS- and OS+ fake μ enhanced,

one μ fails the muon reconstruction quality cuts

Compare predicted vs. observed # of bg. events:

3 sets of cutsLoose: LR > 0.5

Medium: LR > 0.9

Tight: LR > 0.99

+

primary vertex

Fake di-muon vertex

CMU Seminar

20

Bg. Cross-Checks Cont.

OS- Excellent control sample:

SS and fake muon could represent some bb backgroundsM. Herndon

[cm]

For >0: hist: OS pts: SS

OS+ vs. OS- OS+ vs. SS+

-CMU Seminar

21

Bg. Cross-Checks Cont.

M. Herndon

OS- and Fake μ samples

SS+,SS-: expect and find 0 events for tighter cuts

OS-Predicted Observed Probability

LR > 0.50 840±16 821 28%

LR > 0.90 118±6 136 7%

LR > 0.99 8.7±1.9 17 2%

SS- LR > 0.50 8.7±1.7 6 27%

SS+ LR > 0.50 11.0±1.8 12 40%

FMPredicted Observed Probability

LR > 0.50 221±8 196 7%

LR > 0.90 40±3 29 6%

LR > 0.99 5.1±1.1 9 12%

Using 150 MeV/c2 mass

window for cross-check

Combined CMU-CMU(X)

Think about results with tight

cuts and FM

Investigated B hh carefully

Generated inclusive MC to look for anomalous background sources - no surprises found

CMU Seminar

22

B hh Background

M. Herndon

Clearly peaks in signal region

Sideband estimates not useful

Observed/measured at CDF

ModeN Bs Mass Window

N Bd Mass Window

Bd 0.008 0.095

Bd K 0.005 0.55

Bd KK 0 0

Bs 0.035 0.006

Bs K 0.048 0.061

Bs KK 0.089 0.66

Total 0.19±0.06 1.37±0.16

Convolute known branching ratios and

acceptance with K and fake rates.

(estimated for LH > 0.99)

B hh background substantial!

CMU Seminar

23

Acceptance & Efficiencies

: efficiency - from data where possible

Using J/ψ → μ+μ- and B+ → J/ψK+ data and special unbiased J/ψ triggers

Most efficiencies relative: Exception - and LH and K

M. Herndon

From Data

From MC:

Checked with Data

From MC

Second critical part of the analysis is estimating the efficiencies

€

α • total =α • ε trig • εreco • εLH

€

reco = εSVX • εCOT • εmuon • εVtx

€

trig = εL1 • εL2 • εL 3

From Data

CMU Seminar

24

Example: SVX EfficiencyMeasurement of the efficiency of adding silicon hits to COT tracks

Use J/ψ → μ+μ- data

M. Herndon

εSVX = 74.5 ± 0.3(stat) ± 2.2(sys)%(2001-2003 data)

Average efficiency for adding silicon to two tracks: In silicon fid.

(2001-2002 data) (2001-2002 data)

Very low and went down

with pT and time!

25

New SVX EfficiencyImproved silicon pattern recognition code and detector performance

M. Herndon

old avg (88%)w/ same data(2003)

+2% and flat - improved code

+5% εSVX improved code and

new quality cuts

εSVX = 74.5% → 88.5%

2001-2003 → 2001-2005 data

CMU Seminar

26

LR Efficiency Cross-CheckEfficiencies determined using realistic MC

MC efficiencies validated by comparing with data using B+ → J/ψK+ events

Pt and isolation distributions

reweighed to match data Same

treatment for Bs

M. Herndon

LR Cut LR(data) LR(MC) Ratio

LR>0.90 0.70±0.01 0.71±0.01 0.98

LR>0.92 0.68±0.01 0.68±0.01 1.00

LR>0.96 0.60±0.01 0.61±0.01 0.98

LR>0.98 0.52±0.01 0.53±0.01 0.98

LR>0.99 0.41±0.01 0.43±0.01 0.96

Agreement good

4% systematic uncertainty assigned

5% from isolation reweighingCMU Seminar

27

Final Efficiencies & AcceptancesLR efficiencies determined using realistic MC

Pt and isolation distributions reweighed to match data using Bs J/ events

M. Herndon

LR Cut LR(CMUCMU) LR(CMUCMX)

LR>0.90 0.65±0.01 0.68±0.01

LR>0.95 0.57±0.01 0.61±0.01

LR>0.98 0.47±0.01 0.49±0.01

LR>0.99 0.37±0.01 0.41±0.01

Very good efficiency for large

background rejection

For LR > 0.99

39% efficiency, 7% sys

~Factor 2000 of rejection

Acceptance ratio:

α(B+/Bs) = 0.297 ± 0.006 (CMUCMU), 0.191 ± 0.008 (CMUCMX): 7% sys - generation model

Most efficiency ratios near 1

Two absolute efficiencies:

K+COT = 95% and K+SXV = 96%

CMU Seminar

28

Optimization ResultsTried Likelihood ratios from 0.9-0.99: LR Cut 0.99

M. Herndon

NB+: 5763 ± 101

εLR: 39%

Backgrounds:

a priori Bs limit: 1.1x10-7 95% CL

Optimized for 1fb-1

Systematic uncertainties

Bg. estimation: 28%

Bg statistical error

Sensitivity: 16%

12% fs/fu

7% acceptance

7% LR

Previous CDF result 2.0x10-7

Decay

Bhh Background

Combinatoric Background

Total Background

Bs

0.19±0.06 1.08±0.36 1.27±0.36

Bd

1.37±0.16 1.08±0.36 2.45±0.39

CMU Seminar

CDF 1 Bs result: 3.010-6

CDF Bs result: 365pb-1 2.010-7

D0 Bs result: 240pb-1 4.110-7

BaBar Bd result: 8.310-8(90%)

29

Bs(d) → μ+μ- Search Result

M. Herndon

PRL 95, 221805 2005

Results: 1(2) Bs(d) candidates observed

consistent with background expectation

Worlds Best Limits!

Decay

Total Background Observed

Bs

1.27±0.36 1

Bd

2.45±0.39 2 BF(Bs +- ) < 1.0x10-7 at 95% CL

BF(Bd +- ) < 3.0x10-8 at 95% CL

DO Note 4733-Conf

PRL 94, 221803 2005

CMU Seminar

30

Bs →μμPhysics Reach

Strongly limits specific SUSY models: SUSY SO(10) models

Allows for massive neutrino

Incorporates dark matter results

M. Herndon

BF(Bs +- ) < 1.0x10-7 at 95% CL

Excluded at 95% CL

BF(Bs +- ) = 1.0x10-7

Dark matter constraints

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

R. Dermisek et al.hep-ph/0507233R. Dermisek JHEP 0304 (2003) 037

Looks like the theorists

have a little work to do!

CMU Seminar

31

Bs →μμPhysics ReachIn addition to limiting S0(10) models – starting to impact standard MSSM scenarios: CMSSM

Blue BF(Bs → μ+μ-)

Light green: b → s

Cyan: dark matter constraints

Dashed red: Higgs mass bound

Red brown areas: excluded

Incorporates errors from fBs and mtop

M. Herndon

BF(Bs +- ) < 1.0x10-7 at 95% CL

J. Ellis Phys. Lett. B624, 47 2005

CMU Seminar

BF(Bs +- ) = 2.0(1.0)x10-7

M. Herndon 32

Conclusions

Best Bs and Bd results: well ahead of D0 and the B factories

Limit excludes part of parameter space allowed by SO(10) models

Expanding sensitivity to interesting areas of MSSM parameter space

Improving analysis to include selection NN and advanced lepton ID(from CMU: Vivek)

Improved fs results could also improve result by 10-20%(from CMU: Karen)

CDF B(s,d) →μ+μ- results

BF(Bs +- ) < 1.0x10-7 at 95% CL

BF(Bd +- ) < 3.0x10-8 at 95% CL

Worlds Best Limits!

CMU Seminar

Goal for 1fb-1 analysis: BF(Bs +- ) < 7.0x10-8 at 95% CL

33

Muon & Trigger EfficienciesL1 eff: Use J/ψ → μ+μ- trigger that only requires one muon

L3 eff: Use autoexcept trigger

M. Herndon

trig

ger

effic

ienc

y (%

)75

1

0075

10075

10

0

75

100

75

100

2 6 10PT (GeV)

L1 Muon L1(pT,η), L3 1 number

Convolute εL1 for each muon and εL3

Systematic errors L1 and L3:

Kinematic difference between

J/ψ → μ+μ- and Bs(d) →μμ-

2-Track correlations

Sample statistics

εtrig = 85 ± 3%

Offline muon reconstruction

From J/ψ → μ+μ- L1 trigger with one

muon found

Systematic from comparison to Z

εmuon = 95.9 ±1.3(stat) ± 0.6(sys)%

34

COT EfficiencyEstimated by embedding COT hits from MC muons in real data

Occupancy effects correctly accounted for

Efficiency driven by loss of hits when hit density is high

Demonstrate agreement between hit usage in data and MC

Tunable parameters can lower or raise the hit usage and the efficiency to bracket the data

M. Herndon

εCOT = 99.62 ± 0.02%

Systematic errors

Isolation dependence dominant

systematic

Pt dependence

2-track correlations

Varying the simulation tuning

Error: +0.34 -0.91

Consistent with true tracking eff.

using high pt elections identified in

the calorimeter

35

Bs →μμPhysics ReachIn addition to limiting S0(10) models – starting to impact standard MSSM scenarios: mSugra

Blue BF(Bs → μ+μ-)

Light green: b → s

Cyan: dark matter constraints

Dashed red: Higgs mass bound

Red brown areas: excluded

Incorporates errors from fBs and mt

M. Herndon

BF(Bs +- ) < 1.0x10-7 at 95% CL

J. Ellis Phys. Lett. B624, 47 2005