13 th Nordic LHC Physics Workshop Helsinki October 26, 2006

Measurement of the top quark properties with jet flavour and charge taggingChristophe Clément (CERN/ Stockholm University)

13th Nordic LHC Physics WorkshopHelsinkiOctober 26, 2006

B(t→Wb)

• B(t→Wb)=1 usually assumed by CDF and D analyses

• B(t→Wb) might deviate from unity:– Additional quark singlets or doublets– ”Pollution” of top sample by non-top

process!– Non-SM processes in the production– Non-SM in the decay (H+,...)

• Experimentally B(t→Wb) affects number of b-jets need to experimentally discriminate b/w t→Wb and t→Wqlight

Vtb unconstrained without 3x3 unitarity constraint

October 16, 2006 C.Clément CERN EP seminar 9/39October 16, 2006 C.Clément CERN EP seminar 9/39

Vtb

• Select a top-enriched sample– e+jets and +jets channels– larges statistics, good S/B

• # events with 0, 1 and ≥2 b-jets– B(t→Wb)– b-tagging efficiency– Jet identification efficiency– Probability to tag background

Deriving B(t→Wb) experimentally....

October 16, 2006 C.Clément CERN EP seminar 10/39October 16, 2006 C.Clément CERN EP seminar 10/39

t

t

W

Wb-jets?light jets?

Missing transverse

energyOne high pT

isolated lepton

light, c-jets

• .

• W+jets

• Z+jets

• WW, WZ, ZZ

• single top

• multijet

Lepton + jets sample composition

October 16, 2006 C.Clément CERN EP seminar 11/39October 16, 2006 C.Clément CERN EP seminar 11/39

t t

True isolated lepton processes

Ntrue

fake isolated lepton processes

Nfake

Fake isolated electron Jets with leading /πo , convertions, γwith random tracks,...

Fake isolated inside jets from heavy flavor or in flight decaysDetermine Ntrue , Nfake on a

statistical basis- from data by deriving the probability for jets to fake isolated leptons

• .

• W+jets

• Z+jets

• WW, WZ, ZZ

• single top

• multijet

Lepton + jets sample composition

October 16, 2006 C.Clément CERN EP seminar 12/39October 16, 2006 C.Clément CERN EP seminar 12/39

t t

Ntrue

Nfake

•Small•derived from MC•Lepton, jet efficiencies calibrated on data• σfrom data or NLO

Nother

Nbefore tag = Ntt + NWj + Nfake +

Nother

Nn-tags = Pntt (B(t→Wb)) Nn

tt + PnWj Nn

wj + N’nfake + Pn

other Nn

other

n-tags = 0, 1, 2Fit B(t→Wb), Ntt, Nwj to the Nn-tags

Tagging probability

Explicitely reconstruct displaced secondary vertices: Secondary Vertex Tagger

1. Taggable jets are: Calorimeter jets with pT>15 GeV, ||<2.5 ΔR(calo jet, track jet)=0.5, ≥ 2 tracks in ΔR=0.5, Δz<2cm ≥ 1 hit in the innermost tracking detector, pT>0.5 GeV ≥ 1 track with pT>1 GeV

Decouple b-tagging from experimental issues

2. Tagged jets: Are taggable jets Contain a SV, χ2

Lxy > n σLxy

Similar algorithm used by CDF

How to identify b-quark jets...

October 16, 2006 C.Clément CERN EP seminar 13/39October 16, 2006 C.Clément CERN EP seminar 13/39

Lxy=

Taggability,tagger independent

B-tagging efficiency

b-tagging efficiency

From di-jet data: extract b-tagging efficiency for muonic b-jets

We need the b-tagging efficiency for ”all kinds of b-jets”

PPbbtagtag(E(ETT,,) =) =

b,MCb,MC ------------------------------------ bb→→, data, data Taggability Taggability C Ctaggabilitytaggability(b) (b) bb→→,MC,MC

b→b→, data, data

bb

Transform semi-muonic b-tag efficiency into inclusive one

October 16, 2006 C.Clément CERN EP seminar 14/39October 16, 2006 C.Clément CERN EP seminar 14/39

Similar techniques described inPhys. Rev. D71, 052003 (2005)

• Fit B(t→Wb) and Ntt simultaneously to 12 bins: e/+ 3/≥4 jets, 0/1/≥2 tags

• B(t→Wb) is constrained by relative 0/1/≥2 tags populations

• Very poor S/B in 0-tag sample, Nttbar~√Nobs

• Perform measurement of Ntt in 0-tag as additional constrain.

October 16, 2006 C.Clément CERN EP seminar 16/39October 16, 2006 C.Clément CERN EP seminar 16/39

3 jets

≥4 jets ≥4 jets, 0 tag

Number of b-tagged jets

Number of b-tagged jets Topological discriminant

October 16, 2006 C.Clément CERN EP seminar 23/39October 16, 2006 C.Clément CERN EP seminar 23/39

Phys. Lett. B 639 (2006)

No 0

-tag

B(t→Wb) = 1.03+0.19-0.17

Confidence contour plots in R, Ntt

B(t→Wb) B(t→Wb)

68% CL : B(t→Wb)>0.78 |Vtb|>0.8895% CL : B(t→Wb)>0.61 |Vtb|>0.78

OROR

??

Top Quark Charge

October 16, 2006 C.Clément CERN EP seminar 25/39October 16, 2006 C.Clément CERN EP seminar 25/39

Lift ambiguity present in all top analyses!t→W+b or ”t”→W-b

Test exotic models...t

bQ1

Q4

+2/3

-1/3

-4/3mixing

M(Q4)~175GeV

Mtop~270GeV

Phys.Rev. D65 (2002) 053002

OROR ??

Q jettrack i

qi . pTi0.6

track ipTi

0.6 kinematic fit++ == Qtop

Ingredients

What is the expected shape of Qtop for SM top”top” with 4e/3 charge?

October 16, 2006 C.Clément CERN EP seminar 26/39October 16, 2006 C.Clément CERN EP seminar 26/39

Analysis strategy Discriminate between |Qtop| = 2e/3 and |Q”top”|=4e/3

Two |Qtop| per event

Use the pure sample-- lepton+4≥jets events with 2 SVT

Compute the jet charge of the 2 b-tagged jets

Associate the b-jets to correct W boson (charged lepton)

Combine the 2 jet charges and the lepton charge to derive the 2 |Qtop|

Compare the observed |Qtop| with expected SM and exotic distributions

Double tagged events

October 16, 2006 C.Clément CERN EP seminar 27/39October 16, 2006 C.Clément CERN EP seminar 27/39

t t

S/B~10

Jet Charge Algorithm

Compute jet charge only for b-tagged jets - (2 per events)

Jet charge =

Optimizaton on MC gives a=0.6Sum over tracks with

pT>0.5GeV, ΔR(track, jet) <0.5 of the jet axis

Algorithm:

i

aT

i

aTi

jet

i

i

p

pqQ

October 16, 2006 C.Clément CERN EP seminar 28/39October 16, 2006 C.Clément CERN EP seminar 28/39

Derive expected shape of Qjet from data

with minimal input from simulation

pTi,qi

Jet Charge Performance in Data

Tag and probe method in ”pure” events

In reality: Is it pure ? ? flavor excitation? g→ ? B→ B →D → → B→ light hadrons → Charge misidentification

Tight di-jet sample

>3.0

October 16, 2006 C.Clément CERN EP seminar 30/39October 16, 2006 C.Clément CERN EP seminar 30/39

_bb

Ideal case: sign of q = sign of qb

_bb

_cc

_bb

´ _Bo→ Bo

Charge flipping processes

Is the triple tag sample pure ”bb”? The fraction of c-jets in the triple tag sample is determined by pTrel fit of the

order of a few percents,

Flavor excitation/ splitting?

>3.02 b-jets back to back dominate

Phys. Rev. D 65, 094006 (2002)

-

October 16, 2006 C.Clément CERN EP seminar 32/39October 16, 2006 C.Clément CERN EP seminar 32/39

P+ (Qjet) = (1-xc) (1-xflip)Pb (Qjet)+ (1-xc)xflip Pb + xc Pc

- -

Fraction of derived from pT

rel spectrum of (1+2

-1%)

_ccFraction of charge

flipping processes30±1% from MC, Cross checked on data

Similar equation for P- (Qjet)

4 Unknown p.d.f’sPb, Pb, Pc, Pc

Tight di-jet sample

+

p.d.f of Qjet in probe jet

- -

p.d.f.’s Qb, Qb, Qc, Qc from data...--

October 16, 2006 C.Clément CERN EP seminar 33/39October 16, 2006 C.Clément CERN EP seminar 33/39

p.d.f.’s Qb, Qb, Qc, Qc from data...--

October 16, 2006 C.Clément CERN EP seminar 34/39October 16, 2006 C.Clément CERN EP seminar 34/39

P+ (Qjet) = 0.69 Pb (Qjet) + 0.30 Pb + 0.01 Pc

P- (Qjet) = 0.30 Pb (Qjet) + 0.69 Pb + 0.01 Pc

P´+(Qjet) = 0.567 Pb (Qjet) + 0.243 Pb + 0.19 Pc

P´- (Qjet) = 0.243 Pb (Qjet) + 0.243 Pb + 0.19 Pc

- -

-

-

-

-

Tight di-jet sample

loose di-jet sample

-Correct for different

t→Wb and bb kinematics

SM Top and Exotic Quark Charge Observables Use kinematic fit to assign b-jets to correct W-bosons in MC

BlSM

blSM

qqQ

qqQ

2

1

qb = qjet of the b-jet associated to the leptonic side of the event

qB = qjet of the b-jet asssociated to the hadronic side of the event

qB

qb

qB

qb

October 16, 2006 C.Clément CERN EP seminar 35/39October 16, 2006 C.Clément CERN EP seminar 35/39

t t

blEX

BlEX

qqQ

qqQ

2

1

qBqBqb

qb

Admixture of Q4 and t-quark - not excluded by - fQ4 <0.52 at 68% C.L.

σtt

Result on ~0.4 fb-1 of D data 21 l + jets double tagged events 16 events with converged kinematic fit 32 measured top charges

Perform likelihood ratio test b/w 4e/3 and 2e/3 hypothesis

Exclude 4e/3 at 92%C.L. (91 % expected)

October 16, 2006 C.Clément CERN EP seminar 37/39October 16, 2006 C.Clément CERN EP seminar 37/39

hep-ex/060844hep-ex/060844

Conclusions• World best/only measurement

• Acknowledge collaborators: J. Strandberg (Stockholm University, now at Michigan) and

P. Hansson (Kungliga Tekniska Högskolan)

• Calibration of b-tagging algorithm and jet charge algorithm on data was crucial to establish these measurements

• Calibration of efficiencies, fake rates from DATA is often at the heart of analyses at Tevatron and most tricky part.

• Probably even more so at LHC experiments, large leap forward, need to understand data before we can convince ourselves we

October 16, 2006 C.Clément CERN EP seminar 39/39October 16, 2006 C.Clément CERN EP seminar 39/39

B(t→Wb) = 1.03+0.19-0.17

Exclude 4e/3 at 92%C.L. (91 % expected)

hep-ex/060844

Phys. Lett. B 639 (2006)

Complement slides

How top quark was discovered!How top quark was discovered!

Used for most measured top Used for most measured top quark properties so farquark properties so far

tt W

W

Dileptonb

b

ll

vv

tt W

W

Lepton+jetsb

b

lq

vq’

tt W

W

All hadronic

b

b

q

q’q

q’

Top Pair Final StatesTop Pair Final States

October 16, 2006 C.Clément CERN EP seminar 8/39October 16, 2006 C.Clément CERN EP seminar 8/39

44%

ee+e+ ~5%

”golden channel”e+jets & +jets ~32%

Top Quark DecayTop Quark Decay

3 quark generations + direct measurements of Vub and Vcb predict Vtb,~1 B(t→Wb)~1

SM predicts FCNC decays are tiny, t→Wq is dominant

In SM top decays via V-A charged current

⇒ Mostly left handed b-quarks in the decay

ig

2 2b 1 5 V tb tW

October 26, 2006 C.Clément Noridc LHC Workshop 5/39October 26, 2006 C.Clément Noridc LHC Workshop 5/39

Single Tag l+jets

Double Tag l+jets

Jet multiplicity

Jet multiplicity

October 16, 2006 C.Clément CERN EP seminar 17/39October 16, 2006 C.Clément CERN EP seminar 17/39

R=1.0, σ=7pb R=0.5, σ=7pbIllustration:How data constraintsB(t→Wb) and σtt ...

Pntt versus B(t→Wb)

• Probability to see n-tags in events (Pntt )

depends on the number of b-jets

• Pn tt = R2 Pn tag(tt→bb) + 2 R(1-R)Pn tag(tt→bql) + (1-R)2Pn

tag(tt→qlql)

t t

October 16, 2006 C.Clément CERN EP seminar 15/39October 16, 2006 C.Clément CERN EP seminar 15/39

ttbar → l+jets

Take into account ”contamination” by → llt t

• Fit B(t→Wb) and Ntt simultaneously to 12 bins: e/+ 3/≥4 jets, 0/1/≥2 tags

• B(t→Wb) is constrained by relative 0/1/≥2 tags populations

• Very poor S/B in 0-tag sample, Nttbar~√Nobs

• Perform measurement of Ntt in 0-tag as additional constrain.

October 16, 2006 C.Clément CERN EP seminar 16/39October 16, 2006 C.Clément CERN EP seminar 16/39

Prediction 7pb & B(t→Wb)=1

Lower limit on B(t→Wb) and |Vtb|

• Prior π(B(t→Wb))=0 outside [0,1]

• Monte Carlo integration over 191 nuisance parameters associated to systematic errors

• Provides a 2D p.d.f. for B(t→Wb) and Ntt

• Limit on |Vtb| can be derived using |Vtb|=√B(t→Wb) (SM)

68% CL : B(t→Wb)>0.78 |Vtb|>0.8895% CL : B(t→Wb)>0.61 |Vtb|>0.78

October 16, 2006 C.Clément CERN EP seminar 24/39October 16, 2006 C.Clément CERN EP seminar 24/39

The 0-tag sample

• Ntt ~ √N(0-tag)

• Without further information on the 0-tag events low B(t→Wb) + large Ntt (σtt) can still be consistent with data

• Use topological properties of events in 0-tag sample for addititional constraint of Ntt in the 0-tag sample.

Cro

ss

Sect

ion

RPreliminary result from summer 2004 (170pb-1) No 0-tag sample used

October 16, 2006 C.Clément CERN EP seminar 18/39October 16, 2006 C.Clément CERN EP seminar 18/39

Low B(t→Wb)Large σtt

Likelihood discriminant in l+4jets 0-tag sample

Data and prediction for

7pb and R=1

Data

October 16, 2006 C.Clément CERN EP seminar 20/39October 16, 2006 C.Clément CERN EP seminar 20/39

Likelihood discriminant output

Nevents

Systematics on template shapes

Some systematic uncertainties can affect the template shapes...

Systematics on template shapes on ttbar→ l+jet

JESJetIDJet energy resolutionW-modeling (W+jets)TaggabilityTagging probabilities for b, c and light jets

October 16, 2006 C.Clément CERN EP seminar 21/39October 16, 2006 C.Clément CERN EP seminar 21/39

1. SphericityS = 3(λ2+λ3)/2 , λ’s smallest eigenvalues of

momentum tensor M (ttbar S~1)

2. K’Tmin

K’Tmin = Rminjj/EW

T with EWT = El

T + MET

3. CentralityC=HT/H , HT is scalar sum of jets ET and

H is the sum of the jet energies.

4. H’T2

H’T2 = HT2/Hz, HT2 : pT sum of all jets but leading jet,

Hz is the scalar sum of all jets |Ez| plus |Ez| of the neutrino (W-assumption)

Likelihood discriminant in 0-tag samplel + 4 jets before tagging

October 16, 2006 C.Clément CERN EP seminar 19/39October 16, 2006 C.Clément CERN EP seminar 19/39

Results (230 pb-1)

October 16, 2006 C.Clément CERN EP seminar October 16, 2006 C.Clément CERN EP seminar 22/39 22/39

3 jets ≥4 jets

≥4 jets, 0 tag

b-tagging efficiency from data

”n-sample”:1 jet with a

Not -taggedn

-taggedn

Not -tagged, SVT taggednSVT

-tagged, SVT taggedn ,SVT

”p-sample”:2 b-2-b jets 1 jet with a

Not -taggedp

-taggedp

Not -tagged, SVT taggedpSVT

-tagged, SVT taggedp,SVT

Systematic uncertainties

October 16, 2006 C.Clément CERN EP seminar 38/39October 16, 2006 C.Clément CERN EP seminar 38/39

Kinematic fit

Jet charge specific

Data Calibration Corrections

Discriminant Power

bb

bb

VV

aaD

||

Tag and probe

Method: Z→bbOctober 16, 2006 C.Clément CERN EP seminar 31/39October 16, 2006 C.Clément CERN EP seminar 31/39

Tag and probe

Method: data

MC truth on tag side

Why does it work?The charge of the quark is correlated with the charge ofthe highest pT hadron resulting of the hadronization

Simple study carried out with Pythia:Generated QCD 2→2 process, pT>15GeV

And look at hight pT b quarks produced in the process.

Usually large number of hadrons Produced, most quite modest pT

Thís is then smeared by detector effects...

The original b-quark not always In highest pT hadron

Charge of b-quark

Charge of highest p T hadron

October 16, 2006 C.Clément CERN EP seminar 29/39October 16, 2006 C.Clément CERN EP seminar 29/39