Evidence for Single Top Quark Production at CDF

Evidence for Single Top Quark Production at CDF

Bernd Stelzer

University of California, Los Angeles

HEP Seminar, University of Pennsylvania

September, 18th 2007

2

Outline

1. Introduction to Top Quarks

2. Motivation for Single Top Search

3. The Experimental Challenge

4. Analysis Techniques• Likelihood Function Discriminant (1.51fb-1)• Matrix Element Analysis (1.51fb-1)

• Measurement of |Vtb|

• More Tevatron Results

• Summary / Conclusions / Outlook

3

The Tevatron Collider

• Tevatron is worlds highest energy Collider (until 2008)

• Proton Anti-proton Collisions at ECM=1.96 TeV

4

Top Production at the Tevatron

Once every 10,000,000,000 inelastic collision..Once every 10,000,000,000 inelastic collision..

5

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Top Production at the Tevatron

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

•At the Tevatron, top quarks are primarily produced in pairs via the strong interaction:

•Single top quark production is also predictedby the Standard Model through theelectroweak interaction: (st ~ ½ tt)

NLO = 6.7±0.8 pb

mt=175GeV/c2

s-channelNLO = 0.88±0.07 pb

t-channelNLO = 1.98±0.21 pb

Discovered

1995!

Cross-sections at mt=175GeV/c2, B.W. Harris et al., Phys.Rev. D70 (2004) 114012, Z. Sullivan hep-ph/0408049

6

Top Quark in the Standard Model

>10 orders of magnitude!

•Top Quark is heaviest particle to date–mt=170.9 1.8 GeV/c2 March 2007–Close to the scale of electroweak symmetry breaking–Special role in the Standard Model?

•Top Quark decays within ~10-24s-No time to hadronize-We can study a ‘bare quark’

7

Why measure Single Top Production ?

Vtb

Direct measurements

Ratio from Bs oscillations

Not precisely measured

s-channelt-channel

Ceccucci, Ligeti, Sakai PDG Review 2006

Precision EW rules out “simple”fourth generation extensions,but see

J. Alwall et. al., “Is |Vtb|~1?”Eur. Phys. J. C49 791-801 (2007).

• Source of single ~100% polarized top quarks:– Short lifetime, information passed to decay products– Test V-A structure of W-t-b vertex

•Allows direct Measurement of CKM- Matrix Element Vtb:

– single top ~|Vtb|2

– indirect determinations of Vtb enforce 3x3 unitarity

8

Sensitivity to New Physics and Benchmark for WH

•Single top rate can be altered due to the presence of New Physics:- t-channel signature: Flavor changing neutral currents (t-Z/γ/g-c couplings)

- s-channel signature: Heavy W boson, charged Higgs H+, Kaluza Klein excited WKK

Tait, Yuan PRD63, 014018(2001)

Z

ct

W,H+

s (pb)

1.2

5

t (p

b)

•s-channel single top has the same final state

as WHlbb=> benchmark for WH!

(WH ~ 1/10 s-channe))

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

CMSSM Study:Buchmuller, Cavanaugh, deRoeck, S.H., Isidori, Paradisi, Ronga, Weber, G. Weiglein’07]

9

ExperimentalChallenge

10

Top Pair Production with decayInto Lepton + 4 Jets final stateare very striking signatures!

Jet4

Jet3

Event Signatures

Jet1

Jet2

Ele

ctron

MET

Single top Production with decayInto Lepton + 2 Jets final stateIs less distinct!

11

€

η =−log(tan(θ 2))

ηη = 1.0= 1.0

ηη = 2.8= 2.8

ηη = 2.0= 2.0

CDF II Detector (Cartoon)

■Silicon tracking

detectors■Central drift

chambers (COT)■Solenoid Coil■EM calorimeter■Hadronic

calorimeter■Muon scintillator

counters■Muon drift

chambers■Steel shielding

Single top analysisneeds full detector!

Thanks to great work of detector experts and shift crew!

12

CDF II Detector

Silicon detector

Central muonCentral calorimeters

Endplug calorimeters

Drift chamber tracker

13

Data Collected at CDF

This analysis uses 1.51 fbThis analysis uses 1.51 fb-1-1 (All detector components ON)(All detector components ON)

CDF is getting faster, too!6 weeks turnaround time to calibrate, validate and process raw data

Tevatron people are doing a fantastic job!3fb-1 party coming up!

Design goal

Delivered : 3.0 fb-1

Collected : 2.7 fb-1

14

Single Top Selection

Event Selection:•1 Lepton, ET >20 GeV, |ηe()|< 2.0 (1.0)

•Missing ET, (MET) > 25 GeV

•2 Jets, ET > 20 GeV, |η|< 2.8

•Veto Fake W, Z, Dileptons, Conversions, Cosmics

•At least one b-tagged jet, (displaced secondary vertex tag)

CDF W+2jet Candidate Event:CDF W+2jet Candidate Event:

Close-up View of Layer 00 Silicon DetectorClose-up View of Layer 00 Silicon Detector

Jet2

Jet1

Ele

ctron

12mm

Number of Events / 1.51 fb-1 Single Top

Background

S/B

W(l) + 2 jets 136 28300 ~1/210

W(l) + 2 jets + b-tag 61 1042 ~1/17

Run: 205964, Event: 337705Electron ET= 39.6 GeV, Missing ET = 37.1 GeVJet 1: ET = 62.8 GeV, Lxy = 2.9mmJet 2: ET = 42.7 GeV, Lxy = 3.9mm

15

B-quark Tagging and Jet Flavor Separation

•Separate tagged b-jets from charm/light jets using a Neural Network trained with tracking information

–Lxy, vertex mass, track multiplicity, impact parameter, semilepton decay information, etc...

•Used in all single top analyses

Neural Network Jet-Flavor Separator

NN Output

Charm tagging rate ~10%Mistag rate ~ 0.5%

• Exploit long lifetime of B hadrons (c ~450 m)+boost

• B hadrons travel LLxyxy~3mm ~3mm before decay with large track multiplicity

16

Mistags (W+2jets)

• Falsely tagged light quark or gluon jets

• Mistag probability parameterization obtained from inclusive jet data

Background Estimate

W+HF jets (Wbb/Wcc/Wc)

•W+jets normalization from data and

heavy flavor (HF) fraction from MC

Top/EWK (WW/WZ/Z→ττ, ttbar)

•MC normalized to theoretical cross-section

Non-W (QCD)

•Multijet events with semileptonic b-decays or mismeasured jets

•Fit low MET data and extrapolate into signal region

Wbb

WccWc

non-W

Z/D

ibMistag

s

tt

W+HF jets (Wbb/Wcc/Wc)

•W+jets normalization from data and heavy flavor (HF) fractions from ALPGEN Monte Carlo

17

Non-W Estimate

•Build non-W model from ‘anti-electron’ selection•Require at least two non-kinematic lepton ID variables to fail: EM Shower Profile 2, shower maximum matching (dX and dZ), Ehad/Eem,

•Data is superposition of non-W and W+jets contribution -> Likelihood Fit

Signal Region

Before b-tagging: After b-tagging:

Signal Region

18

W + Heavy Flavor Estimate

•Method inherited from CDF Run I (G. Unal et. al.)

•Measure fraction of W+jets events with heavy flavor (b,c) in Monte Carlo

•Normalize fractions to W+jets events found in data

Correct data for non W+jets events

€

NWbb

data = (NWbb

NW + jets

)MC ⋅K HF ⋅NW + jetsdata

€

NW + jetsdata = NCandidates

data − Nnon−W − NEWKHeavy flavor fractions

and b-tagging efficiencies from LO ALPGEN Monte

Carlo

Calibrate ALPGEN heavy flavor Fractions by comparing W + 1jet Data with ALPGEN jet Monte Carlo

Note: Similar for W+charm background

Large uncertainties from Monte Carlo estimate and heavy flavor calibration (36%)

KHF=1.4 ± 0.4

19

Signal and Background Event Yield

CDF RunII Preliminary, L=1.51 fbCDF RunII Preliminary, L=1.51 fb-1

Predicted Event Yield in W+2jetsPredicted Event Yield in W+2jets

Single top swamped by background and hidden behind background uncertainty. Makes counting experiment impossible!s-channel 23.9 ±6.1

t-channel 37.0 ±5.4

Single top 60.9 ±11.5

tt 85.3 ±17.8

Diboson 40.7 ±4.0

Z + jets 13.8 ±2.0

W + bottom 319.6 ±112.3

W + charm 324.2 ±115.8

W + light 214.6 ±27.3

Non-W 44.5 ±17.8

Total background

1042.8

±218.2

Total prediction1103.

7±230.9

Observed 1078

20

Analysis Flow Chart

Analy

sis

Event

Sele

ctio

nA

naly

sis

Event

Sele

ctio

n

CDF DataCDF Data

Monte CarloSignal/

Background

Monte CarloSignal/

Background

Apply MCCorrectionsApply MC

Corrections

Analysis TechniqueAnalysis

Technique

Result

Template Fit to Data

Template Fit to Data

DiscriminantDiscriminant

Signal

Background

Cross SectionCross Section

21

Analysis Techniques

Likelihood Discriminant

Matrix Element AnalysisMatrix Element Analysis

More Tevatron ResultsMore Tevatron Results

22

The Likelihood Function Analysis

€

pisig =

N isig

N isig + N i

bkg

Nsig

Nbkg

i, index input variable

€

LF(r x ) =

psigi (x i)i=1

nvar∏psig

i (x i)i=1

nvar∏ + pbkgi (x i)i=1

nvar∏Bkgr Signal

Unit

Are

a

tchanschan

Wbbttbar

Leading Jet ET (GeV)

Uses 7 (8) kinematic variables for t-channel (s-channel) Likelihood Functione.g. M(Wb) or kin. Solver 2, HT, QxEta, NN flavor separator, Madgraph Matrix Elements, M(jj)

Discriminant

23

Kinematic Variables

Background Signal Background Signal

Wbbttbar

Wbbttbar

tchanschan

tchanschan

HT =ET(lepton,MET,Jets)

Wbbttbar

tchanschan

24

Analysis Techniques

Likelihood DiscriminantLikelihood Discriminant

Matrix Element Discriminant

More Tevatron ResultsMore Tevatron Results

25

Matrix Element Approach

€

P(x) =dσ (pi

μ )

σ=

1

σM

2dΦ

•No single ‘golden’ kinematic variable!

•Attempt to include all available kinematic information by

using Matrix Element approach

•Start from Fermi’s Golden rule:

Cross-sections ~ |Matrix Element|2 Phase space

Calculate an event-by-event probability (based on fully differential cross-section calculation) for signal and background hypothesis

26

Matrix Element Method

€

P( plμ , p j1

μ , p j 2μ ) =

1

σdρ j1dρ j 2dpν

z | M(piμ ) |2

f (q1) f (q2)

| q1 || q2 |φ4 W jet (E jet , E part )

comb

∑∫Parton

distribution function (CTEQ5)

Leading Order matrix element (MadEvent)

W(Ejet,Epart) is the probability of measuring a jet energy Ejet when Epart was produced

Integration over part of the phase space Φ4

Event probability for signal and background hypothesis:

Input only lepton and 2 jets 4-vectors!

c

27

Transfer Functions

Eparton Ejet

Full simulation vs parton energy:

Eparton Ejet

Double Gaussian parameterization:

partoniii Ebap +=where:

E = (Eparton–Ejet)

€

W jet (E jet , E parton ) =1

2π ( p1 + p2 p5)[exp

−(δE − p1)2

2p22

+ p3 exp−(δE − p4 )2

2p52

]

Double Gaussian parameterization:

28

Event Probability Discriminant (EPD)

€

EPD =b ⋅Psin gletop

b ⋅Psin gletop + b ⋅PWbb + (1− b) ⋅PWcc + (1− b) ⋅PWcj

;b = Neural Network b-tagger output

•We compute probabilities for signal and background hypothesis per eventUse full kinematic correlation between signal and background events

•Define ratio of probabilities as event probability discriminant (EPD):

SignalBackground

29

Event Probabilty Discriminant

S/B~1/1In most sensitive bin!

•S/B~1/17 over full range•Likelihood fit will pin down background in low score

region

30

Cross-Checks

31

Cross-Checks in Data Control Samples

•Validate method in various data control samples

•W+2 jets data (veto b-jets, selection orthogonal to the candidate sample)

•Similar kinematics, with very little contribution from top (<0.5%)

px py pz E

Second Leading Jet

Leading Leading Jet

Lepton (Electron/Muon)

32

Cross-Checks in Data Control Samples

CDF Run II Preliminary

•b-tagged dilepton + 2 jets sample

•Purity: 99% ttbar•Discard lepton with lower pT

•b-tagged lepton + 4 jets sample

•Purity: 85% ttbar•Discard 2jets with lowest

pT

33

Monte Carlo Modeling Checks

34

Template Fitto the data

35

Binned Likelihood Fit

Binned Likelihood Function:

Expected mean in bin k:

All sources of systematic uncertainty included as nuisance parameters

Correlation between Shape/Normalization uncertainty considered (δi)

βj = σj/σSM parameter

single top (j=1)

W+bottom (j=2)

W+charm (j=3)

Mistags (j=4)

ttbar (j=5)

k = Bin index

i = Systematic effect

δi = Strength of effect

εji± = ±1σ norm. shifts

κjik± = ±1σ shift in bin k

36

Rate vs Shape Systematic Uncertainty

DiscriminantDiscriminant

•Rate systematics give fit templates freedom to move vertically only

•Shape systematics allow templates to ‘slide horizontally’ (bin by bin)

Shape systematics

Rate and

Systematic uncertainties can affect rate and template shape

37

Sources of Systematic Uncertainty

Source Rate Uncert. Shape Uncert.

W + bottom 36%

W + charm 36%

Mistags 15%

ttbar 21%

Non-W 40%

Jet Energy Scale 1..13%

Initial State Radiation

3.2%

Final State Radiation

5.3%

Parton Dist. Function

1.4%

Monte Carlo Modeling

1.6%

Efficiencies/b-tag SF

5%

Luminosity 6%

CDF RunII Preliminary, L=1.51fb-1

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

38

Results

39

Matrix Element Analysis

• Matrix Element analysis observes excess over background expectation• Likelihood fit result for combined search:

€

Single Top = 3.0−1.1+1.2 pb

40

ME Separate Search

s-channels=1.1 pb+1.0

−0.8

t-channelt=1.9 pb+1.0

−0.9

•Perform separate likelihood fit fors-channel and t-channel signal•Both signal templates float independently

41

Likelihood Function Discriminant

• Likelihood Function analysis also observes excess over background expectation• Observed deficit previously in 0.955 fb-1

€

Single Top = 2.7−1.1+1.3 pb

42

Likelihood Function 2D Fit

1.41.1

1.21.0

1.1 pb

1.3 pb

s

t

+−

+−

=

=

43

Signal Significance

44

Hypothesis Testing

•Calculate p-value: Faction of background-only pseudo-experiments with a test statistic value as signal like (or more) as the value observed in data

•Define Likelihood ratio test statistic:

•Systematic uncertainties included in pseudo-experiments

•Use median p-value as measure for the expected sensitivity

Median p-value = 0.13% (3.0)

€

Q =L(data | s + b)

L(data | b)

More signal like Less signal like

Observed p-value = 0.09% (3.1)

L. Read, J. Phys. G 28, 2693 (2002)T. Junk, Nucl. Instrum. Meth. A 434, 435 (1999)

3.1 Evidence

45

Hypothesis Testing

Median p-value = 0.20% (2.9)

More signal like Less signal like

Observed p-value = 0.31% (2.7)

46

Signal Features

47

Central Electron CandidateCharge: -1, Eta=-0.72 MET=41.85, MetPhi=-0.83 Jet1: Et=46.7 Eta=-0.61 b-tag=1 Jet2: Et=16.6 Eta=-2.91 b-tag=0QxEta = 2.91 (t-channel signature)EPD=0.95

Single Top Candidate Event

Jet1

Jet2

Lepton

Run: 211883, Event: 1911511

€

u,d

€

d,u

48

Single Top Signal Features

EPD>0.90

EPD>0.95

Look for signal featuresin high score output

49

QxEta Distributions in Signal Region

EPD>0.9EPD>0.9

3) 4)

EPD>0.95EPD>0.95

50

m(W,b) Distributions in Signal Region

EPD>0.9EPD>0.9 EPD>0.95EPD>0.95

51

Unconstrained Likelihood Fit

Remove all background normalization constraints and perform a five parameter likelihood fit (all template shapes float freely) Best fit for signal almost unchanged. Uncertainty increased by about 20%

52

Direct |Vtb| Measurement

•Using the Matrix Element cross Section PDF we measure |Vtb|

•Assume Standard Model V-A coupling

and |Vtb| >> |Vts|, |Vtd|

|Vtb|= 1.02 ± 0.18 (experiment) ± 0.07 (theory)

t-channel

Z. Sullivan, Phys.Rev. D70 (2004) 114012

Flat prior 0 < |Vtb|2

< 1

|Vtb|>0.55 at 95% C.L.

s-channel

53

Single TopResults from DØ

54

D0 Results

First direct limit on Vtb:0.68 <|Vtb|< 1 @ 95%CL or

|Vtb| = 1.3 ± 0.2

First direct limit on Vtb:0.68 <|Vtb|< 1 @ 95%CL or

|Vtb| = 1.3 ± 0.2

Boosted Decision Tree

PRL 98 18102 (2007)

Expected p-value = 1.9% (2.1)Observed p-value = 0.04% (3.4)

3.4 Evidence

55

Summary of Results

Expected

3.0

2.9

2.6

2.1

1.9

2.2

Observed

3.1

2.7

3.4

3.2

2.7

Summary

•CDF analyses more sensitive•D0 observes upward fluctuationIn 900 pb-1 of data

Combined:2.3

/3.6

56

CDF Single Top History

First Tevatron Run II result using 162 pb-1

single top < 17.5 pb at 95 % C.L.

2004: Simple analysis while refining Monte Carlo samples and analysis tools

2 Years

2006: Established sophisticated analysesCheck robustness in data control samples

2007: Evidence for single top quark production using 1.5 fb-1 (expected and observed!)

•Development of powerfulanalysis techniques (Matrix Element, NN, Likelihood Discriminant)•NN Jet-Flavor Separatorto purify sample•Refined background estimates and modeling•Increase acceptance (forward electrons)•10x more data

Phys. Rev. D71 012005

57

Conclusion

• Evidence for electroweak single top quark production at the Tevatron established by CDF and D0 experiment!

• First direct measures of CKM matrix element |Vtb|

• Advanced analysis tools essential to establish small signals buried underneath large backgrounds

• Entering the era of single top physics. 4-5 sigma observation possible with >3 fb-1 of data - Perhaps CDF is lucky this time..

• Separate s-channel from t-channel, measure more top properties, e.g. top polarization etc..

• Exciting times! The race for first observation is on..

• Important milestone along the way to the Higgs!

58

Search for Heavy W Boson

Limit at 95% C.L. M(W´) > 760 GeV/c2 for M(W´) > M(νR) M(W´) > 790 GeV/c2 for M(W´) < M(νR)

W•Search for heavy W boson in W + 2, 3 jets

•Assume Standard Model coupling strengths(Z. Sullivan, Phys. Rev. D 66, 075011, 2002)•Perform fit to MWjj distribution Previous Limits:

•CDF Run I: M(WR) > 566 GeV/c2 at 95% C.L.•D0 Run II: M(WR) > 630 GeV/c2 at 95% C.L.

59

LHC is the Future

Large Hadron Collider

60

LHC is the Future

•LHC will be a top quark factoryσtt ~ 800 pbσt-channel ~ 243 pb (153 pb for top and 90 pb for antitop

production)σs-channel ~ 11 pb (6.6 pb for top and 4.8 pb for antitop production)σWt ~ 50-60 pb (negligible at the Tevatron)

•First precision t-channel measurement (10%) expected after 1st year of running (10 fb-1/year)

•s-channel measurement harder because of small cross section

and large backgrounds (sounds familiar!)

•The associated Wt production is tough because of large top-pair background (W+3jets signature)

Wt- production

Additional single top process at the LHC! (negligible at the Tevatron)

61

Backup Slides

Backup