single top search

Single Top search- Significance study using S+J B-tagging

K. NakamuraUniv. of Tsukuba

2

Motivation

|Vtb|

|Vtb|

t-channel

s-channel

Produced electroweakly couplingDirect access to W-t-b vertex

singletop~|Vtb|2Top-polarization study and V-A structure of EW top interaction

Background study for SM Higgs searchExactly the same final state as WH->WbbBackground of all channels to use W+njet background

Probe b-quark PDF (t-channel)Non-SM phenomena

Heavy W’ bosonAnomalous W-t-b coupling ……

But…Cross section is small (~40% of tt)Larger background cause of W+2jets bin

Not established a significant evidence yet …

3

Current Study and CDF NOTE

Signal Acceptance and Monte Carlo based backgrounds (CDF 7863)Validation of single-top signal samples (CDF 7701)

Event selection and MC validation

Anlysis with 0.7 fb-1 (CDF 8112)Non-W and W+njets (n=1,2,3) background estimationNeural network b-tagger (CDF 7816)Kinematics fitter for top quark reconstruction (CDF 749)Construct discriminates with Neural network with Likelihood ratio

Unit : pbLimit : 95%CL

theory Likelihood Ratio

Neural Network

t-channel 1.98±0.21 < 3.1

s-channel 0.88±0.07 < 3.2

t+s comb. 2.86±0.28 < 5.8 < 3.4

Result

Phys. Rev. D 71, 012005 (2005)

Projections for Single Top Sensitivity

(CDF 8286)NEW

4

Backgrounds

MC based background

Data based backgroundNon-W MistagsWbb Wcc Wc

5

Event Selection

Standard tight lepton selection of lepton+jet group.Detail is in appendix

Lepton ID

Et>20 GeVIsolated Conversion Veto

ElectronPt>20 GeV/cIsolated Cosmic Veto

Muon

Jet Energy CorrectionLevel4 Jet Energy Correction – not to apply abs correction

Et>15 GeV, Eta < 2.8

Et>20 GeVEt is corrected with lepton and Level4 jets

MET

6

Signal/Background Expectation

@1fb-1

7

The 2 jets bin networks have 14 input variables. The 14 input variables for 2 jet bin networks are shown in backup slide

NN discriminants

8

NN b-tagger

This NN b-tagger is applied to jets that are already tagged by SecVtx tagger.

9

Likelihood Function Analysis

t-channel

s-channel

Likelihood function uses 7 variables

Likelihood function uses 6 variables

10

|Vtb|

s-channel

|Vtb|

t-channel

Motivation of our analysis

CDF NOTE 8112

Large amount of Background for this channel.About half of the background comes from Mistags and Wc(c).

Double tag is the powerful method to rejectthese background for s-channel !

But, statistics of the signal is limited by low efficiency of double tag.

We plan to apply “tight” + “loose” b-tagging S+JSecVtx JetProb

11

Signal Sample

Generator : Mad EventEvent : single top s-channel LO Top Mass : Mt=175 GeV Name : ttop0oNew

Background Sample

Generator : Alpgen + pythiaEvent : W(->e)+nP W(->mu)+nP (n = 0,1,2,3,4)Name : ltop0n - ltop4n ltop0m – ltop4m

12

SecVtx

Use Jprob<0.05

30% events in 2-jet bin can be tagged as s+s or s+j. >>> which corresponds to 46% of at least 1 secvtx tagged events in 2-jet bin.

Estimated b-tag performance of each method for s-channel

13

W+nP Background with MC sampleMistag Matrix

12

3

4

12

3

4

12

3

4

12

3

4

Over estimate SS+SJ

2 S

Under estimate SS+SJ

i

Siε

ji

Sj

Si εε

At least 1 S

ji

Sj

Si

Sj

Ji

Jj

Si εεεεεε

ji

Sj

Ji

Jj

Si εεεε

Null-corr 100%-corr

14

W(->e)+nP Background n=0,1,2,3,4over

under

15

overunder

W+nP Normalized to 0.7 fb-1

~1/40

16

0.022BS

0.59B

S

4.8 (2.5%)

s-channel significance

7.3 (46%) s-channel26.0 (46%)

0.74 (25%)

26.7

99.7 (46%)

2.2 (2.5%)

11.4 (14%)

113.3

0.56(2.5%)0.050B

S

0.61B

S

17

Summary Double tag including s+j tagging has a potential to improve the s/n ratio and significance.

30% events in 2-jet bin can be tagged as s+s or s+j tag The fraction of s+s or s+j tagged events is 46% of at least 1 tag events According to very rough estimation,

Detail of signal selection will be confirmed. We will examine some kinematics variables for background rejection. We will estimate the sensitivity for single-top production as a function of integrated luminosity……

Plan

0.022BS

0.59B

S

0.050BS

0.61B

S improveCDFNOTE Double tag

Backup

19

CEM ele ID PHX ele ID

20

21

ID

22

MC based background

23

Non-W background

24

Mistag background

25

W+HF background

26

NN input

27

28

t-channel Likelihood valuables

29

Lepton IDStandard tight lepton selection of lepton+jet group.

Detail is in appendix

Et>20 GeVIsolated Conversion Veto

Electron

Pt>20 GeV/cIsolated Cosmic Veto

Muon

30

Jet Energy CorrectionLevel4 Jet Energy Correction – not to apply abs correction

Et>15 GeV, Eta < 2.8

MET

Et>20 GeV

Et is corrected with lepton and Level4 jets

Et

Phi

Et/Pt()

L4 Jet … 0.8L5 Jet … ~1

Phi/Phi()

31

Gen info after geometrical cuts

muonelectronlepton

L5 MET correction

32

Event Selection

(96%)(90%)(25%)(24%)(24%)(24%)(23%)(21%)

Cross checkThis is t-channell result but…

33

B-taggingPerformance of SecVtx and JetProb b-tagging

Number of events of under histogram include W+njet (n>0) final state

For the statistics of the signal double SecVtx tag is applied, about 70% statistics are increased to use S+J tagging.

34

Jet multiplicity distribution of b-tag performanceSecVtx JetProb

Cross checkJet multiplicity distribution after b tag (CDF 7863)

19.2% 64.7% 13.6% 2.1% 0.3%

Good agreement !?

Use Jprob<0.05 cuts for Jet Prob.

35

overunder

W(->)+nP Background n=0,1,2,3,4