[PPT] - Single Top Physics at the Tevatron Enrique Palencia Fermilab for PowerPoint Presentation

SLIDE 1

Single Top Physics at the Tevatron

Enrique Palencia

Fermilab

for the CDF and DØ Collaborations

Rencontres de Moriond EW 2009, La Thuille (Italy), March 7-14, 2009

SLIDE 2

Fasten your seatbelt!!!

In this talk, you will see brand new single top

results using the latest amount of data available per collaboration

..... reporting the first observation of the single top

quark production!!!!!

♦ You will see a lot of 5-sigma analyses!

By CDF and DØ, independently, at the Tevatron
Major achievement of both collaborations

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 1 of 20

SLIDE 3

Why Single Top?

Single top quark is produced via electroweak

interaction but has not been observed SO FAR

♦ σSM(t-channel/tqb) = 1.98 ± 0.25 pb (mtop = 175 GeV) ♦ σSM(s-channel/tb) = 0.88 ± 0.11 pb (mtop = 175 GeV) ♦ σSM(t¯ t) = 6.7 ± 0.8 pb (via strong interaction) ♦

B.W. Harris et al., Phys. Rev. D 66, 054024 (2002)

Z. Sullivan, Phys. Rev. D70, 114012 (2004)
Test of the Standard Model

♦ Direct measurement of |Vtb| ♦ Top quark properties: polarization, spin, W helicity,... ♦ Same final state as WH

Sensitive to new physics

♦ Search for W ′, H+ (s-channel signature) ♦ Search for FCNC,...

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 2 of 20

SLIDE 4

Event Selection and Backgrounds

Top decays most of the times to Wb
W + 2 or 3 (4 in DØ) energetic jets
One high pT isolated lepton (electron
r muon) from the leptonic decay of

the W

Large missing transverse energy, E

/T, from the neutrino

At least one jet identified as b-tagged
Main backgrounds: W+Heavy Flavor,

W+mistags, t¯ t, QCD, diboson

W+HF jets (Wbb/Wcc/Wc)

W +j

et s nor m al i zat i

n

f r

m

dat a and heavy f l avor ( HF) f r act i

n

f r

m

MC

Wbb

Wcc Wc non-W Z/Dib Mistags tt

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 3 of 20

SLIDE 5

Why it took so long? Experimental Challenge!

Single top hidden under huge background ⇒ counting experiment is NOT possible
Multivariate analyses needed to discriminate single top from backgrounds

♦ No single observable to see single top ♦ Will show several multivariate analyses in next slides

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 4 of 20

SLIDE 6

Search Strategy

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 5 of 20

SLIDE 7

Likelihood Function (LF)

Combines several sensitive variables into a single one
7 (10) variables used in the 2 (3) jet bin: HT, Q×η, Mjj, cos(l, j), log (MEt−chan)...

20 40 60 80 100 120 140 160

3
2
1

1 2 3 20 40 60 80 100 120 140 160 Q*η Events/0.155

Data s-channel t-channel Wbb ttbar Wc+Wcc W+LF NonW Z+jets,Diboson

CDF Run II Preliminary, L=3.2 fb-1 Monte Carlo Scaled to Prediction

psig

i

(xi) =

Nsig

i

(xi) Nsig

i

(xi)+Nbkg

i

(xi),

L =

Πnvar

i=1 psig i

(xi) Πnvar

i=1 psig i

(xi)+Πnvar

i=1 pbkg i

(xi)

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 6 of 20

SLIDE 8

Likelihood Function: Results

tchan

L 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Events 10

2

10

3

10

tchan

L 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Events 10

2

10

3

10

1

CDF Run II Preliminary, L=3.2fb

s-channel t-channel Wbb ttbar WC mistag ZJETS Diboson nonW

1 10 10 2 10 3 10 4 10 5

100
80
60
40
20

20 40 60 80 CDF Run II Preliminary, L=3.2 fb-1

2lnQ

Pseudoexperiments

B S+B

Observed Median S+B

LF

Lum. (fb−1)
Exp. sign.
Obs. sign.

Cross Section (pb) 3.2 4.0σ 2.4σ 1.6+0.8

−0.7

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 7 of 20

SLIDE 9

Neural Networks: Results

NN Output

1
0.5

0.5 1 Candidate Events 100 200 300 NN Output

1
0.5

0.5 1 Candidate Events 100 200 300

1

CDF II Preliminary 3.2 fb MC normalized to SM prediction All Channels

0.6 0.7 0.8 0.9 1 10 20 30 40 50 60 70 80 0.6 0.7 0.8 0.9 1 10 20 30 40 50 60 70 80

single top t t c +Wc b Wb Wc q Wq Diboson Z+jets QCD data

NN

Lum. (fb−1)
Exp. sign.
Obs. sign.

Cross Section (pb) 3.2 5.2σ 3.5σ 1.8 ± 0.6 2.3 4.1σ 5.2σ 4.7+1.2

−0.9

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 8 of 20

SLIDE 10

Matrix Elements (ME)

Compute, for each event, the probability for signal and background hypotheses
Use full event kinematic

information

Calculate probabilities for

signal and backgrounds

Build a discriminant

EPD =

b·Psig( x) b·Psig( x)+b·Pb−bkg( x)+(1−b)·Pnonb−bkg( x)

D( x) =

Psig( x) Psig( x)+Pbkg( x), (separate for s and t channels)

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 9 of 20

SLIDE 11

Matrix Elements: Results

Event Probability Discriminant

0.2 0.4 0.6 0.8 1

Candidate Events

100 200 300 400 500

Event Probability Discriminant

0.2 0.4 0.6 0.8 1

Candidate Events

100 200 300 400 500

schan tchan wbb wcc mistag ww wz zz zjets nonW ttbarlj ttbardil

Normalized to Prediction

1

CDF Run II Preliminary, L=3.2 fb

Event Probability Discriminant

0.7 0.75 0.8 0.85 0.9 0.95 1

Candidate Events

20 40 60 80

Event Probability Discriminant

0.7 0.75 0.8 0.85 0.9 0.95 1

Candidate Events

20 40 60 80

ME

Lum. (fb−1)
Exp. sign.
Obs. sign.

Cross Section (pb) 3.2 4.9σ 4.3σ 2.5+0.7

−0.6

2.3 4.1σ 4.9σ 4.3+1.0

−1.2

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 10 of 20

SLIDE 12

Boosted Decision Trees (BDT)

Sequence of binary splits using the

discriminating variable which gives best sig-bkg separation

Leaf nodes are classified as sig-like or

bkg-like depending on majority of events ending up in the respective leaf

Use large number of input variables

♦ Non-discriminating variables are automatically

ignored, but do not degrade the performance

Boosting

algorithm improves the discrimination power and statistical stability

♦ Events misclassified during a DT training are given a higher weight in the next DT training

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 11 of 20

SLIDE 13

Boosted Decision Trees: Results

BDT Output (2 jets, 1 tag)

1
0.5

0.5 1

Candidate Events

20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

1
0.5

0.5 1 20 40 60 80 100 120

BDT Output (2 jets, 1 tag)

1
0.5

0.5 1

Candidate Events

20 40 60 80 100 120

BDT Output (2 jets, 1 tag)

1
0.5

0.5 1

Candidate Events

20 40 60 80 100 120

s-channel t-channel W+light W+charm W+bottom Non-W Z+jets Diboson tt Data

1

CDF Run II Preliminary, L=3.2 fb

Normalized to Prediction

BDT

0.5 0.6 0.7 0.8 0.9 1

Events

5 10 15 20 25

BDT

0.5 0.6 0.7 0.8 0.9 1

Events

5 10 15 20 25

BDT

Lum. (fb−1)
Exp. sign.
Obs. sign.

Cross Section (pb) 3.2 5.2σ 3.5σ 2.1+0.7

−0.6

2.3 4.3σ 4.6σ 3.7+1.0

−0.8

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 12 of 20

SLIDE 14

High Score Region (BDT>0.6)

[GeV]

T

H

150 200 250 300 350 400

Candidate Events

5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30 150 200 250 300 350 400 5 10 15 20 25 30

[GeV]

T

H

150 200 250 300 350 400

Candidate Events

5 10 15 20 25 30

[GeV]

T

H

150 200 250 300 350 400

Candidate Events

5 10 15 20 25 30

s-channel t-channel W+light W+charm W+bottom Non-W Z+jets Diboson tt Data

1

CDF Run II Preliminary, L=3.2 fb

Normalized to Prediction

KIT flavor sep.

0.5

0.5 1

Candidate Events

5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

0.5

0.5 1 5 10 15 20 25 30

KIT flavor sep.

0.5

0.5 1

Candidate Events

5 10 15 20 25 30

KIT flavor sep.

0.5

0.5 1

Candidate Events

5 10 15 20 25 30

s-channel t-channel W+light W+charm W+bottom Non-W Z+jets Diboson tt Data

1

CDF Run II Preliminary, L=3.2 fb

Normalized to Prediction

η × Q

2
1

1 2

Candidate Events

5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

2
1

1 2 5 10 15 20 25 30

η × Q

2
1

1 2

Candidate Events

5 10 15 20 25 30

η × Q

2
1

1 2

Candidate Events

5 10 15 20 25 30

s-channel t-channel W+light W+charm W+bottom Non-W Z+jets Diboson tt Data

1

CDF Run II Preliminary, L=3.2 fb

Normalized to Prediction

lnub

m

120 140 160 180 200 220

Candidate Events

5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25 120 140 160 180 200 220 5 10 15 20 25

lnub

m

120 140 160 180 200 220

Candidate Events

5 10 15 20 25

lnub

m

120 140 160 180 200 220

Candidate Events

5 10 15 20 25

s-channel t-channel W+light W+charm W+bottom Non-W Z+jets Diboson tt Data

1

CDF Run II Preliminary, L=3.2 fb

Normalized to Prediction

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 13 of 20

SLIDE 15

Other Analyses

Single top search in the E

/T+jets sample (new in the Tevatron!!!)

♦ Orthogonal to lepton+jets analysis ♦ Increase acceptance by ∼30%: events where the lepton was not reconstructed and hadronic taus ♦ Challenging! Huge instrumental QCD background ♦ σ = 4.9+2.5

−2.2 pb

Separate t and s channel search
Lum. (fb−1)

σs (pb) σt (pb) CDF (NN) 3.2 2.0+0.7

−0.6

0.7 ± 0.5 CDF (LF) 3.2 1.4 1.0 DØ (BDT) 0.9 0.9 3.8

Separate s-channel search (CDF

, 3.2 fb−1): use double b-tagged events and LF as multivariate technique

♦ σs = 1.5 +0.9

−0.8 pb, (σSM s

= 0.88 ± 0.11 pb)

Final NN Discriminant Output

1
0.8
0.6
0.4
0.2

0.2 0.4 0.6 0.8 1 Events/0.25 50 100 150 200 250 300 350 Final NN Discriminant Output

1
0.8
0.6
0.4
0.2

0.2 0.4 0.6 0.8 1 Events/0.25 50 100 150 200 250 300 350

DATA Top T Top S Multijet Diboson Z + h.f. W + h.f. t t

MC normalized to SM prediction

1

CDF Run II Preliminary, 2.1 fb

5 10 15 20 25 30

ln(L) ∆

1

CDF II Preliminary 3.2 fb

t-channel cross section [pb]

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

s-channel cross section [pb]

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

data 68.3% C.L. 95.5% C.L. 99.7% C.L. SM Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 14 of 20

SLIDE 16

Combination

Each experiment combines the individual results
Use the NN, BDT and ME discriminant outputs to create a second layer

combination NN discriminant ♦ Crosscheck with BLUE (Best Linear Unbiased Estimator)

NEAT: Neuro Evolution of Augmenting Topologies

♦ Superdiscriminant that uses output discriminant of individual analysis (LF , NN, ME, BDT, LF-schan) as input ♦ Candidate networks compete against each other ♦ Network topology, weights, output histogram binning, includes systematic errors in optimization procedure (using genetic algorithms) ♦ The final network is chosen based on the expected p-value

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 15 of 20

SLIDE 17

Combination: Results

Neural Network Output

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Events

1

10 1 10

2

10

3

10

4

10

Neural Network Output

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Events

1

10 1 10

2

10

3

10

4

10

S-Channel T-Channel (Dilepton) t t (Non-Dilepton) t t b W+b c W+c Mistags Z+jets WW WZ ZZ Non-W Data

1

CDF Run II Preliminary, L = 3.2 fb

Normalized to Prediction

All channels

Combination Output

0.2 0.4 0.6 0.8 1

Event Yield

100 200 300 400 500

D 2.3 fb

1

Data tb + tqb W+jets tt Multijets

0.6 0.7 0.8 0.9 1 50 75 25

Comb

Lum. (fb−1)
Exp. sign.
Obs. sign.

Cross Section (pb)

∗

3.2 5.9σ 5.0σ 2.3+0.6

−0.5

2.3 4.5σ 5.0σ 3.9 ± 0.9

∗ SD combined with E /T +jets analysis Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 16 of 20

SLIDE 18

Vtb Measurement

|Vtb|2 is proportional to the cross section

♦ Vtb is extracted from the measured cross section ♦ CDF: |Vtb| = 0.91 ± 0.11(exp) ± 0.07(th) ♦ DØ: |Vtb| = 1.07 ± 0.12 ♦ Measurement does not assume 3 generations

r unitarity
Assume SM production

♦ Flat prior in |Vtb|2 and 0 < |Vtb|2 ≤ 1 ♦ |Vtd|2 + |Vts|2 <<< |Vtb|2

Lower limits in |Vtb|

♦ CDF: |Vtb| > 0.71 (95% C.L.) ♦ DØ: |Vtb| > 0.78 (95% C.L.)

2

|

tb

|V

0.2 0.4 0.6 0.8 1

Posterior Probability Density | > 0.71 (95% C.L.)

tb

|V

68% 95%

1

CDF Run II Preliminary, L = 3.2 fb

2

|

tb

|V

0.2 0.4 0.6 0.8 1

Posterior Probability Density

2

|

tb

|V 0.2 0.4 0.6 0.8 1 Posterior Density 0.5 1 1.5 2 2.5 3 3.5 4 4.5

1

DØ Run II, 2.3 fb Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 17 of 20

SLIDE 19

Crosscheck

a) W+jets enriched sample, b) t¯

t enriched sample

♦ Nice agreement between data and MC predictions

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 18 of 20

SLIDE 20

Summary

Tevatron is in great shape!
Single top searches are challenging!
Multivariate analysis techniques have

been developed

Both

collaborations have reached the observation of single top quark production (5σ significance) Significance CDF (3.2 fb−1) DØ (2.3 fb−1) Exp. Obs. Exp. Obs. ME 4.9 4.3 4.1 4.9 BDT 5.2 3.5 4.3 4.6 NN 5.2 3.5 4.1 5.2 LF 4.0 2.4 — — LF s-chan 2.0 1.1 — — E /T+jets 1.4 2.1 — — Comb. 5.9 5.0 4.5 5.0

Single Top Production Cross Section (pb)

5

5 11

SM Prediction NLO NNNLO

Combination (All Channels) 0.5 0.6

± 2.3

)

1

(3.2 fb

MET+Jets 2.2 2.6

± 4.9

)

1

(2.1 fb

Combination (Lepton+Jets) 0.5 0.6

± 2.1

)

1

(3.2 fb

Boosted Decision Tree 0.6 0.7

± 2.1

)

1

(3.2 fb

Likelihood Function 0.7 0.8

± 1.6

)

1

(3.2 fb

Matrix Element 0.6 0.7

± 2.5

)

1

(3.2 fb

Neural Network

0.6 ± 1.8

)

1

(3.2 fb

Likelihood Function S-Channel 0.8 0.9

± 1.5

)

1

(3.2 fb

CDF Preliminary Single Top Summary

2

= 175 GeV/c

top

For M

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 19 of 20

SLIDE 21

Conclusions

First Observation of Single Top Production!!!!!
... by CDF and DØ, independently, at the Tevatron
Submitted to PRL

last Wednesday

CDF: arXiv:0903.0885 [hep-ex]

♦ www-cdf.fnal.gov/physics/new/top/public singletop.html

DØ: arXiv:0903.0850 [hep-ex]

♦ www-d0.fnal.gov/Run2Physics/top/top public web pages/top

Thanks to the hard work of many people

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 20 of 20

SLIDE 22

BACK-UP SLIDES

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 21 of 20

SLIDE 23

Tevatron Performance

p¯

p collisions at √s = 1.96 TeV

Only

place where top quarks are produced

Tevatron is performing better than ever

♦ Peak luminosity ∼ 360×1030cm−2s−1 ♦ Expected 7-8 fb−1 by end of 2009 ♦ Possibly run in 2010-11 (9-11 fb−1)

Tevatron delivered ∼ 6.2 fb−1

♦ Collected ∼ 5.2 fb−1

Analyses shown here use 2-3 fb−1

250 500 750 1000 1250 1500 1750 50 100 150 200 250 300 350 Day Delivered Luminosity (pb-1)

FY01 FY02FY03 FY04 FY05 FY06 FY07 FY08 FY09

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 22 of 20

SLIDE 24

CDF and DØ Detectors

Two multipurpose detectors at Tevatron collecting data efficiently
Large acceptance and good ID for leptons

♦ Tracking and EM calorimeter ♦ Muon systems

Good calorimetry for jet energy resolution
Silicon detectors for b-jet tagging
Data taking efficiency ∼85-90%

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 23 of 20

SLIDE 25

The CDF Detector

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 24 of 20

SLIDE 26

The DØ Detector

Calorimeter Shielding Toroid Muon Chambers Muon Scintillators η = 0 η = 1 η = 2

[m]

η = 3

–10 –5 5 10 –5 5

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 25 of 20

SLIDE 27

About the SM Theoretical Cross Sections

s-channel (tb)

♦ σNLO = 0.88 ± 0.11 pb (mtop = 175 GeV)∗ ♦ σNNLO = 1.12 ± 0.05 pb (mtop = 170 GeV)∗∗

t-channel (tqb)

♦ σNLO = 1.98 ± 0.25 pb (mtop = 175 GeV)∗ ♦ σNNLO = 2.34 ± 0.13 pb (mtop = 170 GeV)∗∗

∗ B.W. Harris et al., Phys. Rev. D 66, 054024 (2002), Z. Sullivan, Phys. Rev. D70, 114012 (2004) ∗∗ N. Kidonakis., Phys. Rev. D 74, 114012 (2006) Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 26 of 20

SLIDE 28

Extended Muon Coverage (CDF)

Increase acceptance by taking muons from complementary triggers (E

/T+ 2 jets)

New muons fill the “cracks” left by triggered muons (∼30% gain in muon

acceptance)

η

2
1.5
1
0.5

0.5 1 1.5 2 [deg] φ

180
90

90 180 triggers µ CMUP CMX η

2
1.5
1
0.5

0.5 1 1.5 2 [deg] φ

180
90

90 180 triggers µ MET+Jets CMUP CMX CMU CMP BMU CMIO SCMIO CMXNT

Improve overall single top sensitivity by ∼7%

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 27 of 20

SLIDE 29

NN for Single Top search

About 50% of the background in the W+2 jets sample do not contain b quarks

(even though a secondary vertex is required)

NN to separate jets from b-quark

from c and light-quark jets

Jet

and track variables (vertex mass, decay length, track multiplicity...)

Replace yes-no by a continous

variable

Entries 644

NN output

1
0.5

0.5 1

Events per 0.125 units

50 100 150

Entries 644 Entries 644

NN output

1
0.5

0.5 1

Events per 0.125 units

50 100 150

Entries 644

CDF II data Fit Sum (with stat. error) W + beauty W + charm W + light

W + 2 jets

1

CDF II Preliminary 955 pb

Improves single top sensitivity by 10-20%!!!

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 28 of 20

SLIDE 30

Background Estimate (CDF)

Mistags (W+2jets)

Falsely tagged light quark or gluon jets
Mistag probability parameterization
btained from inclusive jet data

W+HF jets (Wbb/Wcc/Wc)

W +j

et s nor m al i zat i

n

f r

m

dat a and heavy f l avor ( HF) f r act i

n

f r

m

MC

Top/EWK (WW/WZ/Z, ttbar, single t)

MC normalized to theoretical cross-section

Non-W (QCD)

Multijet events with

semileptonic b-decays or mismeasured jets

Fit low missing ET data and

extrapolate into signal region

Wbb Wcc Wc non-W Z/Dib Mistags tt+s t

W+HF jets (Wbb/Wcc/Wc)

W+jets normalization from data and

heavy flavor (HF) fractions from ALPGEN Monte Carlo, calibrated in generic multijet data

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 29 of 20

SLIDE 31

Event Yields

Process Number of Events in 3.2 fb−1 W + 2 jets W + 3 jets s-channel 58.1 ± 8.4 19.2 ± 2.8 t-channel 87.6 ± 13.0 26.2 ± 3.9 Wb¯ b 656.9 ± 198.0 201.3 ± 60.8 Wc¯ c 292.2 ± 90.1 98.1 ± 30.2 Wcj 250.4 ± 77.2 52.1 ± 16.0 Mistags 501.3 ± 69.6 151.9 ± 21.4 non-W 89.6 ± 35.8 35.1 ± 14.0 WW 58.5 ± 6.6 21.2 ± 2.4 WZ 28.9 ± 2.4 8.5 ± 0.7 ZZ 0.9 ± 0.1 0.4 ± 0.0 Z + jets 36.5 ± 5.6 15.6 ± 2.4 t¯ t dilepton 69.2 ± 10.0 60.2 ± 8.7 t¯ t non-dilepton 134.9 ± 19.6 421.8 ± 61.1 Total signal 145.7 ± 21.4 45.4 ± 6.7 Total prediction 2265.0 ± 375.4 1111.5 ± 129.5 Observed in data 2229 1086

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 30 of 20

SLIDE 32

Event Yields

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 31 of 20

SLIDE 33

Neural Networks (NN) Out???

Combine many variables into one more

powerful discriminant

Bayesian Neural Network (BNN)

♦ Weighted average over many networks trained interactively (average

ut

statistical fluctuations) ♦ Use 18-25 variables: Mj1j2, pT j1j2, MW

T ...

♦ 24 different networks: e, µ × 2, 3, 4 jets × 1, 2 b-tags

NeuroBayes Program

♦ Use 11-18 variables: Mlνb, Ql × η, KIT flavor separator (used by all CDF analysis) ♦ 4 different networks: 2, 3 jets × 1, 2 b- tags

KIT Flavor Sep. Output

1
0.5

0.5 1 Event Fraction 0.1 0.2

single top t t c +Wc b Wb Wc q Wq Diboson Z+jets QCD

TLC 2Jets 1Tag

1

CDF II Preliminary 3.2 fb normalized to unit area

KIT Flavor Sep. Output

1
0.5

0.5 1 Event Fraction 0.1 0.2

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 32 of 20

SLIDE 34

Extracting the Cross Section

Both CDF and DØ use Bayesian posteriors

integrated

ver

nuisance parameters (systematic uncertainties)

Flat prior taken in σs + σt
Binned likelihood fit is used (to calculate

the probability that the data comes from the distributions predicted for different amounts of signal and to build the posterior)

Single Top Cross Section [pb]

2 4 6 8 10

Posterior Probability Density Single Top Cross Section [pb]

2 4 6 8 10

Posterior Probability Density 68%

pb

0.5

+0.6

= 2.3

Single Top

σ

Single Top Cross Section [pb]

2 4 6 8 10

Posterior Probability Density

1

CDF Run II Preliminary, L = 3.2 fb

Peak and sigma are taken as central value and uncertainty, respectively

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 33 of 20

SLIDE 35

Extracting Significance

CDF performs pseudo experiments (PE) with and without SM single top

♦ Fluctuate all systematic rate and shape uncertainties in PE ♦ Test Statistic: -2 ln(Q), Q=L(sig+bkg) / L(bkg only) ♦ Obs. p-value: fraction of bkg. only PE with a value smaller than the measured one ♦ Exp. p-value: fraction of bkg. only PE with a value smaller than the median of the S+B hyp.

DØ uses ensemble tests without signal contribution

♦ Significance is the fraction of the background-only pseudo-datasets with a cross section equal

r higher than the measured cross section

Test Statistic [-2ln(Q)]

300
200
100

100 Pseudo-Experiments 1 10

2

10

3

10

4

10

5

10

6

10

7

10

8

10

S+B B Obs 68%

1

CDF Run II Preliminary, L = 3.2 fb

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 34 of 20

SLIDE 36

Systematics

Systematic Rate Shape Jet energy scale 0...16% ✔ Initial state radiation 0...11% ✔ Final state radiation 0...15% ✔ Parton distribution functions 2...3% ✔ Monte Carlo generator 1...5% — Event detection efficiency 0...9% — Luminosity 6% — NN fl avor separator — ✔ Mistag model — ✔ Non-W model — ✔ ALPGEN Q2 — ✔ MC Modeling (∆R, η(j2)) — ✔ W b¯ b+W c¯ c normalization 30% — W c normalization 30% — Mistag normalization 17...29% — Top Mass - top-pair normalization 23% ✔

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 35 of 20

SLIDE 37

Systematic Shape Uncertainties

Total of 370 shape uncertainties are evaluated (most are small)
Left: jet energy scale (JES) for the single top-quark template.
Right: factorization and renormalization scale (Q2) for Wb¯

b events.

NN Output

1
0.5

0.5 1 Event Fraction 0.05 0.1

1

CDF II Preliminary 2.7 fb normalized to unit area

single top σ JES - σ JES +

NN Output

1
0.5

0.5 1

0.5

0.5

W+2Jets 1Tag

NN Output

1
0.5

0.5 1 Event Fraction 0.02 0.04 0.06

1

CDF II Preliminary 2.7 fb normalized to unit area

b Wb = 0.5

2

Q = 2.0

2

Q

NN Output

1
0.5

0.5 1

0.5

0.5

W+2Jets 1Tag

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 36 of 20

SLIDE 38

Non-SM Searches in Single Top Sample

Search for W’-like resonances: W’→tb

♦ No evidence for resonance found M(W’)

Lum. (fb−1)

M(W’R)>M(νR) M(W’R)<M(νR) CDF 1.9 >800 GeV >825 GeV DØ 0.9 >731 GeV >739 GeV

Search for direct production of charged Higgs:

H+ →tb (DØ, 0.9 fb−1)

♦ No evidence for resonance found ♦ Type I 2HDM excluded for M(H+) = 180-185 GeV and tan(β) = 20-70 approx and limits on Types II and III are set

FCNC: non-SM single top-quark production in the

channel u(c)+g → t (CDF , 2.2 fb−1)

♦ NN used as discriminant ♦ No excess is observed ♦ σanom <1.8 pb

M(jet1,jet2,W) [GeV] 100 200 300 400 500 600 700 Events/20 GeV 20 40 60 80 100 100 200 300 400 500 600 700 20 40 60 80 100

1

DØ 0.9 fb tb + tqb t t W + jets Multijets 50 × (180 GeV)

+

H 50 × (240 GeV)

+

H 50 × (300 GeV)

+

H , 2 jets, 1+2 b-tags µ e+

Enrique Palencia, Fermilab Rencontres de Moriond EW 2009, La Thuille (Italy), March 10, 2009 37 of 20