Top quark mass in dilepton and - Fermilabboj/073008-ichep-jayatilaka.pdf · 2016-11-03 · ICHEP...

ICHEP 2008, 7/30/08 B. Jayatilaka (Duke)

Top quark mass in dilepton and all jets channels

Bo JayatilakaDuke University/CDF

On Behalf of the CDF and D∅ Collaborations

34th International Conference on High Energy PhysicsJuly 30, 2008

Philadelphia, PA

1

DØnote 4725-CONF

Measurement of the Top Quark Mass in the Dilepton Channel

The DØ CollaborationURL http://www-d0.fnal.gov(Dated: February 25, 2005)

We present a preliminary measurement of the top quark mass in the dilepton channel based onabout 230 pb!1 of data collected by the DØ experiment during Run II of the Fermilab Tevatroncollider. We show that the method used obtains consistent results using ensemble tests of eventsgenerated with the DØ Monte Carlo simulation. We apply this technique to a total of 13 dileptonevents selected in the collider data to obtain mt = 155+14

!13(stat) ± 7(syst) GeV.

Preliminary Results for Winter 2005 Conferences


Why measure the top mass?• Top mass is a fundamental parameter in SM

• Important in radiative corrections

• Why so massive?

• Only fermion with mass near EW scale

• Yukawa coupling of ~1

• Constrains SM Higgs mass and SUSY models

2

80.3

80.4

80.5

150 175 200

mH [GeV]114 300 1000

mt [GeV]m

W [

GeV

]

68% CL

!"

LEP1 and SLDLEP2 and Tevatron (prel.)

March 2008

Measurement in different decay channels

• Important to check consistency between channels (is it SM top?)

• Discrepancy could indicate new physics


Dilepton decay channel• 2 leptons, 2 jets, large missing ET

• Mostly electroweak background

• Drell-Yan, Diboson

• Instrumental background (fakes)

• Kinematically underconstrained

• Two neutrinos

• Make assumptions and integrate over at least one variable

• Advantages

• Clean: little background without need for b-tagging

• Least jets of any channel (less reliant on JES, less ambiguity in jets)

Disadvantages

• Low statistics

• 2 νs escape undetected– underconstrained system

3


Alljets decay channel

4

• Look for 6 jets

• Dominated by QCD multijet background

• Require b-tags to have appreciable signal

• Reconstructable final state

• Many permutations of jet combinations

• Advantages

• High statistics

• Hadronic Ws allow for in situ jet energy scale (JES) calibration as in l+jets channel

Disadvantages

• Large background


Reconstructing the mass

5

Reconstruct mass for each event

Form “templates” for signal (varying Mtop) and background using simulated events

Perform maximum likelihood fit to extract measured Mtop

Form per-event probability using matrix element

Integrate over unmeasured quantities

Form ensemble probability and calibrate using simulated events

1. Template-based

Advantages: Less CPU bound (systematic uncertainties less likely to be limited by CPU)Disadvantages: Only single number (recon. mass) per event in final Likelihood, all events have equal weight

Advantages: More statistical power, probability curve rather than single mass per event, events weighted naturally

Disadvantages: Complex numerical integration (much CPU)

2. Matrix Element-based


Cut-based event selection: dilepton channel

• Two well-identified leptons

• Lower statistics

• Better signal purity

• “Lepton+track”

• 2nd lepton is only required to be an isolated track

• More statistics at expense of background

6

Source CDF D∅

tt (Mt=175 GeV/c2) 164.3±5.1 86.9±3.8

Drell-Yan 50.1±7.2 12.8±1.4

Fakes 81.0±15.9 7.1±1.2

WW/WZ 15.4±1.0 4.0±0.6

Total 311.8±18.7 110.8±5.0

Data 330 107

Expected Signal Fraction 53% 78%

CDF [lepton+isolated track]

- 1 lepton with pT >20 GeV/c- 1 isolated, well-measured track with pT >20 GeV/c- 2 jets with ET >20 GeV, |η|<2- ET >25 GeV

D∅ [two identified leptons]

- 2 leptons with pT >15 GeV/celectrons up to |η|<2.5

- 2 jets with pT >20 GeV/c, |η|<2.5- HT >115 GeV

DØnote 4725-CONF







Advanced event selection: ENN

• Most selections used in top mass measurements were designed for x-section measurements

• Optimized for S:B

• Devise selection optimized for uncertainty in top mass

• Evolutionary neural net

• Evaluate “fitness function” (Mt stat. error from pseudoexperiments)

• Select best networks and breed

• Best networks do not have best S:B

• 20% improvement in CDF ME analysis over cut-based selection

7

Source CDF(ME)

tt (Mt=175 GeV/c2) 121.8±7.5

Drell-Yan 128.1±14.3

Fakes 33.5±5.9

Diboson 18±3.7

Z→ττ 12.1±2.6

Total 313.3±21.2

Data 344

!"! !"# !"$ !"% !"& !"' !"( !") !"* !"+

,-./01234056-7/289,-.:967601;

$"!

$"'

%"!

%"'

&"!

&"'

'"!

'"'

("!

<=>?56?@2A7>2B0CC2D/5?460-/6E

FG0H>-7/29?I40129?6J74K2,?1?56-7/


Number of jets0 1 2 3 4 5 6 7 8 9

Events

1

10

210

Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt

CDF RunII Preliminary (2.0/fb)

Number of jets0 1 2 3 4 5 6 7 8 9

Events

1

10

210

]2

Dilepton Mass [GeV/c0 50 100 150 200 250 300 350

Events

1

10

210

Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt


]2

Dilepton Mass [GeV/c0 50 100 150 200 250 300 350

Events

1

10

210

]2

[GeV/cT

Missing E0 20 40 60 80 100 120 140

Events

1

10

210

Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt


]2

[GeV/cT

Missing E0 20 40 60 80 100 120 140

Events

1

10

210

Number of b-tags-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Events

1

10

210

310 Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt


Number of b-tags-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Events

1

10

210

310

]2

[GeV/cT

Total Vector E0 100 200 300 400 500 600

Events

1

10

210

Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt


]2

[GeV/cT

Total Vector E0 100 200 300 400 500 600

Events

1

10

210

Neural Net Output0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Events

1

10

210

Data

Bkg. Uncert.

ee!Z

!! !Z

"" !Z

fakes

Diboson

tt


Neural Net Output0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Events

1

10

210

Matrix element dilepton mass: CDF

• Matrix element probabilities for both signal and dominant backgrounds

• Use sample selected using evolving neural net

• 2.0 fb-1 (344 events)

• Measure Mt=171.2±2.7(stat.)±2.9(syst.) GeV/c2

8

)2

(GeV/ct

M155 160 165 170 175 180 185 190 195

-ln

(P)

0

2

4

6

8

10

12

14

16

18

20

)-1

CDF Run II Preliminary (2.0 fb


Matrix element dilepton mass: D∅

• Measurement performed in eμ channel only

• Reduces Drell-Yan background

• Expect 87 tt events out of 107 data events

• Uses 2.8 fb-1 of data

• Measure Mt=172.9±3.6(stat)±2.3(syst) GeV/c2

9

DØnote 4725-CONF






(GeV)topm160 165 170 175 180

-ln L

0

0.5

1

1.5

/ ndf 2! 6.353 / 4

p0 0.02861! -0.06246

p1 0.08874! 4.707

p2 0.08088! 170.2

/ ndf 2! 6.353 / 4

p0 0.02861! -0.06246

p1 0.08874! 4.707

p2 0.08088! 170.2

4.7 GeV!=170.2top

m

-1DØ RunII Preliminary, L=1.1fb

(GeV)topm170 175 180

-ln L

0

1

2

/ ndf 2! 7.744 / 4

p0 0.04575! -0.007383

p1 0.07768! 3.799

p2 0.09391! 175.6

/ ndf 2! 7.744 / 4

p0 0.04575! -0.007383

p1 0.07768! 3.799

p2 0.09391! 175.6

3.8 GeV!=175.6top

m

-1DØ RunII Preliminary, L=1.7fb


Template dilepton mass: CDF

• Template method integrating over neutrino φ

• Combined background template

• Use lepton+isolated track sample

• Use 2.8 fb-1 of CDF data

• Measure

10

Mtop = 165.1 (stat.)±3.1(syst.)GeV/c2+3.3-3.2

0

20

40

60

80

100

100 150 200 250 300 3500

20

40

60

80

100

100 150 200 250 300 350

Data (330 ev.)

CDF RunII Preliminary (2.8 fb-1

)

Constrained fit

ttbar+Backgr.Backgr.

Reconstructed mass (GeV/c2)

Even

ts/(

10 G

eV

/c2) -!

Ln

(L)

0

0.5

1

1.5

2

2.5

3

160 165 170


Template dilepton mass: D∅

• Template method integrating over neutrino rapidity

• Include looser lepton+track selection

• Measure Mt=176.0±5.3(stat.)±2.0(syst.) GeV/c2

• Combine all Run II measurements of dilepton mass at D0

• Matrix Element eμ and Template ee, μμ, and l+track

• Mt=174.4±3.2(stat.)±2.1(syst.) GeV/c2

• χ2/dof=1.3/3

11

DØnote 4725-CONF






[GeV]top

m140 150 160 170 180 190 200 210

-lo

gL

160

165

170

175

180

185

DØ Preliminary -1 1 fb


Systematics: dilepton channel

• Systematic uncertainty in dilepton channel is dominated by JES error

• Cannot do in situ calibration

12

Source CDF (ME) GeV/c2 CDF (Template) GeV/c2

D∅(ME) GeV/c2

[Run IIb only]D∅(Template)

GeV/c2

Jet Energy Scale 2.5 2.9 2.3 1.7b Jet Energy Scale 0.4 0.4 0.3 0.5

Pileup 0.2 0.2 -- --MC Statistics 0.5 -- -- 0.1

PDFs 0.6 0.3 0.2 0.3Generator 0.9 0.2 -- 0.8

Background Shape 0.2 0.5 -- 0.3QCD Radiation 0.5 0.3 0.4 0.1

Method 0.4 -- 0.4 0.8Sample Composition 0.3 0.5 0.3 0.3Lepton Resolution 0.1 0.3 0.3 0.1

Total 2.9 3.1 2.5 2.0

DØnote 4725-CONF







Event selection: all jets channel

• Initial cuts

• 6-8 jets with ET>15 GeV

• Enforce minimum distance between jets

• Veto events with well measured leptons

• Require b-tags

• Apply NN to further reduce background

• Model background using data

• Expected background

• 2409±68 for 1-tag

• 338±28 for 2-tag

• In data, 2881 1-tag events and 537 2-tag events

13

NN output0.85 0.9 0.95 1 1.05 1.1

Even

ts/0

.01

0

20

40

60

80

100

120

)-1CDF Run II Preliminary (2.1 fb

2 tags events, JES=0

Data

Expected Bkg

= 6.7pb)! (Mtop=175, t Bkg + t

NN output0.85 0.9 0.95 1 1.05 1.1

Even

ts/0

.01

0

20

40

60

80

100

120



Template all hadronic mass: CDF

• Separate signal into 1 and 2-tag bins

• Single background template

• Data-based model

• Simultaneous fit of Mt and JES

• Uses 2.1fb-1 of CDF data

• Measure Mt=176.9±3.8(stat.+JES)±1.7(syst.) GeV/c2

14

Mtop (GeV)150 155 160 165 170 175 180 185 190 195 200

JE

S

-3

-2

-1

0

1

2

3) Contours, 1+2 tags events

max-Ln(L/L

Fitted Values

)!) = 4.5 (3max

-Ln(L/L

)!) = 2.0 (2max

-Ln(L/L

)!) = 0.5 (1max

-Ln(L/L

)-1CDF Run II Preliminary, (2.1 fb

(GeV)rectm

100 150 200 250 300

Even

ts/(

5 G

eV

)

0

10

20

30

40

50

(GeV)rectm

100 150 200 250 300

Even

ts/(

5 G

eV

)

0

10

20

30

40

50 rec

t2 tags events m

Data

t Fitted t

Fitted Bkg

/Ndof = 23.3 / 272!

Prob = 0.669


(GeV)rectm

100 150 200 250 300

Even

ts/(

5 G

eV

)

0

10

20

30

40

50

(GeV)rectm

100 150 200 250 300

Even

ts/(

5 G

eV

)

0

10

20

30

40

50 rec

t2 tags events m

Data

t Fitted t

Fitted Bkg

/Ndof = 23.3 / 272!

Prob = 0.669



Ideogram alljets: CDF

• Templates for signal and background constructed using per-event likelihoods

• ME-like calculation for likelihoods

• Signal purity is a free parameter

• Use 1.9 fb-1 of CDF data

• Mt=165.2±4.4(stat.+JES)±1.9(syst.) GeV/c2

15

)2

(GeV/ct

M150 155 160 165 170 175 180

)JE

S!

JE

S (

-0.5

0

0.5

1

1.5

2

2.5

3

)-1


Measured values

)!) = 0.5 (1 max-Ln(L/L

)!) = 2.0 (2 max-Ln(L/L

)!) = 4.5 (3 max-Ln(L/L

)-1


)2

(GeV/ct

M100 150 200 250 300

2E

ve

nts

/6.8

Ge

V/c

0

5

10

15

20

25

30

35

40

)-1


Data2

=165 GeV/ct

Signal, M

Background

/nDof = 24.9/262!

KS Prob = 0.854

)2

(GeV/ct

M100 150 200 250 300

2E

ve

nts

/6.8

Ge

V/c

0

5

10

15

20

25

30

35

40

)-1



Systematics: alljets

• Alljets channel systematics lower since JES is handled in situ

16

Source CDF (Template) GeV/c2

CDF (Template) [JES]

CDF(Ideogram) GeV/c2

Residual Jet Energy Scale 0.8 N/A 0.7

b Jet Energy Scale 0.6 0.09 0.3

Pileup 0.3 0.03 0.7

MC Statistics 1.1 0.05 0.1

PDFs 0.4 0.11 0.4

Generator 0.5 0.26 0.8

Background Shape N/A N/A 0.4

ISR/FSR 0.5 0.10 1.2

Method 0.5 0.12 0.2

Sample Composition N/A N/A 0.5

Total 1.7 0.36 1.9


Impact and conclusions

• 2.8/fb combination from July 2008

• Includes dilepton: CDF(ME), D0(combined)

• Includes template alljets

• Dilepton and alljets measurements contribute ~12% weight

• Compare measurements across channels

• Agree... for now.

• Both CDF and D0 continue analyses in these channels

• Alljets result from D0 imminent

17

DØnote 4725-CONF






)2

(GeV/ctop

M

150 160 170 180 190 2000

15

CDF March’07 2.2! 1.5 !12.4

Tevatron July’08*

1.0! 0.7 !172.4 (syst.)!(stat.)

CDF-II trk*

3.0! 6.2 !175.3

CDF-II all-j*

3.3! 2.6 !176.9

CDF-I all-j 5.7!10.0 !186.0

D0-IIb l+j*

1.7! 1.3 !173.0

D0-IIa l+j*

1.5! 1.5 !171.5

CDF-II l+j*

1.0! 1.3 !172.2

D0-I l+j 3.6! 3.9 !180.1

CDF-I l+j 5.3! 5.1 !176.1

D0-II di-l*

2.1! 3.2 !174.4

CDF-II di-l*

2.9! 2.7 !171.2

D0-I di-l 3.6!12.3 !168.4

CDF-II di-l 4.9!10.3 !167.4

Mass of the Top Quark (*Preliminary)

/dof = 6.9/11.0 (81%)2!

Figure 1: A summary of the input measurements and resulting world average mass of the topquark.

9

Date post:	28-May-2020
Category:	Documents
Upload:	others
View:	7 times
Download:	0 times

Top quark mass in dilepton and - Fermilabboj/073008-ichep-jayatilaka.pdf · 2016-11-03 · ICHEP...

Documents