Measurement of colour ﬂow in events with the ATLAS detector€¦ · Introduction Analysis Results...

Measurement of colour flow in t t̄ eventswith the ATLAS detector

ATLAS-CONF-2017-069

Fabian Wilkon behalf of the ATLAS collaboration

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/CONFNOTES/ATLAS-CONF-2017-069

Introduction Analysis Results Conclusion

IntroductionÏ Colour connection of hard objects affects energy distribution within︸︷︷︸

jet substructureandjet superstructure︷︸︸︷

between jets

Rapidity y

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Azi

mut

hφ/π

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

Rapidity y

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Azi

mut

hφ/π

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

Colour-Connected

Connected to Beam

Ï Goal: Extract information about connection type and propertiesÏ Can be used for Monte-Carlo validation / tuning, . . .Ï Allow to construct truly orthogonal event selection observable

e.g. for assigning b’s in t t̄ H(→ bb̄)

New first measurement of this type at 13 TeV New

[Paper in preparation: . ATLAS-CONF-2017-069]

Fabian Wilk Measurement of colour flow in t t̄ events with the ATLAS detector 2 / 10



The Jet Pull

Ï Measure quantities derived from jet pull vector→ pT -weighted radial jet moment

~P (J1) =∑

i∈J1

piT · ∣∣~ri

∣∣p J

T

~ri

→ jet pull vector magnitude |~P |

~C (J1, J2) =~J2 −~J1

θP (J1, J2) =^(~P (J1) ,~C (J1, J2)

)→ jet pull angle θP





Analysis Strategy

1. Apply event selection for t t̄ → `+ jets signal2. Select target jets for calculation

t

t̄

b

WHadq

q̄ ′

b̄

W −

`+ν`

Ï Probe true colour connectionÏ Select two leading non-b-tagged jets: j W

1 , j W2

t

t̄

b

W +q

q̄ ′

b̄

W −

`+ν`

Ï Probe remnant colour flow across eventÏ Select two leading b-tagged jets: j b

1 , j b2

3. Calculate observables for selected jets, measure detector-level spectra→ Use jet-associated tracks, previous measurement w/ calo clusters . Phys. Lett. B750 475

θP(

j W1 , j W

2

)θP

(j W2 , j W

1

)|~P

(j W1

)| θP

(j b1 , j b

2

)4. Remove detector effects from measurement by unfolding to particle-level

→ use D’Agostini iterative bayesian technique

5. Compare normalised particle-level spectra to simulation-based predictions

New New New



https://doi.org/10.1016/j.physletb.2015.09.051



Selection

t

t̄

b

W +q

q̄ ′

b̄

W −

`+ν`

Event Selection

Q == 1 lepton (pT > 27GeV, |η| < 2.5)Q == 0 veto lepton (pT > 25GeV, |η| < 2.5)Q ≥ 4 jets (pT > 25GeV, |η| < 2.5)Q ≥ 2 b-tags (70 % eff. WP)

Q EMissT > 20GeV

Eve

nts

0

50

100

150

200

250

300

350

400

×103

Datat t̄Single-top, t t̄ + V/HV+jets, DibosonNP/Fake leptonsSyst. Unc.

ATLAS Preliminary√s = 13 TeV, 36.1 fb−1

Charged particle θP(jW1 , jW2

)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Dat

a/M

C

0.91

1.1

Eve

nts

0

100

200

300

400

500

×103



Charged particle |~P(jW1

)|

0 0.005 0.01 0.015 0.02

Dat

a/M

C

0.91

1.1

Eve

nts

0

50

100

150

200

250

300

350

400

×103



Charged particle θP(jb1 , jb2

)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Dat

a/M

C

0.91

1.1

→ estimated signal fraction ∼ 89%





Results1

σFi

d

dσFi

ddθ

P( jW 1

,jW 2

)

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6ATLAS Preliminary√

s = 13 TeV, 36.1 fb−1DataStatistical Unc.Total Unc.Powheg+Pythia8Powheg+Pythia6aMC@NLO+Pythia8Powheg+Herwig7Sherpa


)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Pre

dict

ion

Unf

olde

d

1

1.05

1σ

Fid

dσFi

ddθ

P( jW 2

,jW 1

)

0.95

1

1.05

1.1

1.15

1.2

1.25

1.3

1.35 ATLAS Preliminary√s = 13 TeV, 36.1 fb−1

DataStatistical Unc.Total Unc.Powheg+Pythia8Powheg+Pythia6aMC@NLO+Pythia8Powheg+Herwig7Sherpa


)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1P

redi

ctio

nU

nfol

ded

1

1.05

1σ

Fid

dσFi

d

d|~ P( jW 1

) |

0

50

100

150

200

250

300

350

400 ATLAS Preliminary√s = 13 TeV, 36.1 fb−1


Charged particle |~P(jW1

)|

0 0.005 0.01 0.015 0.02

Pre

dict

ion

Unf

olde

d

0.90.95

11.051.1

Ï Uncertainties dominated by signal modellingÏ Statistical uncertainty sub-dominant

Angle: Predictions suggest stronger effect than observation→ more sloped pull angle

Magnitude: Data favours larger value→ “wider” jets

Ï Mostly poor modelling from MCsÏ Powheg+Herwig7 generally does a good job





Results

Ï Uncertainties dominated by signalmodelling

Ï Statistical uncertainty sub-dominant

Ï As expected, very flat distributionÏ Agreement reversed:

Ï Powheg+Herwig7 modelsobservation poorly

Ï Other predictions offer muchbetter description

1σ

Fid

dσFi

ddθ

P( jb 1

,jb 2

)

0.98

1

1.02

1.04

1.06

1.08

1.1 ATLAS Preliminary√s = 13 TeV, 36.1 fb−1


Charged particle θP(jb1 , jb2

)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Pre

dict

ion

Unf

olde

d

0.980.99

11.011.02





Results

Ï Comparison to maximally violating exotic(“colour-flipped”) colour flow model:

replace WHad by ad-hoc colour-octet

W +t

b

q

q̄ ′ t

b

q

q̄ ′

Ï Predicted colour flow shows significantly lesssloped pull angle

1σ

Fid

dσFi

ddθ

P( jW 1

,jW 2

)

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6ATLAS Preliminary√

s = 13 TeV, 36.1 fb−1DataStatistical Unc.Total Unc.Powheg+Pythia8Powheg+Pythia8(Colour-Flipped)


)[rad]/π

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1P

redi

ctio

nU

nfol

ded

0.9

0.95

1

1.05

Ï Goodness-of-fit tests show data prefers SM-like distribution over exotic. . . but Powheg+Pythia8 does not provide a good description either





Conclusion

Ï A measurement of colour flow using jet pull propertieswith 36.1 fb−1 of 13 TeV pp data measured by theATLAS experiment was presented

Ï Detector-level spectra are unfolded to particle-level usingiterative bayesian technique

Ï Normalised fiducial spectra are compared to various predictions obtained from MC

Ï An improved precision respective to 8 TeV measurement is achievedÏ Several Monte-Carlo predictions show poor modelling of jet pull vector properties

→ We can barely exclude the exotic colour model against our SM predictionsÏ No one-fits-all prediction:

→ Powheg+Herwig7 offers best description of pull angle from W -jets but worst for b-jets

Ï Measurement uncertainty dominated by signal modelling uncertaintiesÏ Low statistical uncertainty suggests future doubly-differential measurement




Backup


Event Selection Yields

Sample Yieldt t̄ 1026000 ± 95000

t t̄V 3270 ± 250t t̄ H 1711 ± 97

Single-top 48400 ± 5500Diboson 1440 ± 220W + jets 27700 ± 4700Z + jets 8300 ± 1400

NP/Fake leptons 53000 ± 30000Total Expected 1170000 ± 100000

Observed 1153003

→ estimated signal fraction ∼ 89%


Systematic Uncertainty Breakdown

∆θP(

j W1 , j W

2

)[%]

θP(

j W1 , j W

2

)0.0 – 0.21 0.21 – 0.48 0.48 – 0.78 0.78 – 1.0

Hadronisation 0.63 0.22 0.27 0.09Generator 0.37 0.24 0.50 0.06

Colour Reconnection 0.11 0.26 0.03 0.53b-Tagging 0.35 0.12 0.20 0.31

Non-Closure 0.25 0.07 0.08 0.30ISR / FSR 0.32 0.12 0.15 0.01

Other 0.25 0.20 0.11 0.18JER 0.12 0.13 0.21 0.03JES 0.13 0.06 0.13 0.07

Tracks 0.09 0.04 0.05 0.07Syst. 0.97 0.52 0.68 0.72Stat. 0.22 0.18 0.17 0.26Total 0.99 0.55 0.71 0.76

Ï Signal modelling uncertainties (generator, hadronisation, . . . ) evaluated usingalternate MC


χ2 Goodness-of-Fit Test Results

Sample θP(

j W1 , j W

2

)θP

(j W2 , j W

1

)θP

(j b1 , j b

2

)|~P (

j W1

) |χ2/NDF p-value χ2/NDF p-value χ2/NDF p-value χ2/NDF p-value

Powheg+Pythia8 50.7 / 3 < 0.001 20.5 / 3 < 0.001 1.5 / 3 0.690 26.2 / 4 < 0.001Powheg+Pythia6 24.0 / 3 < 0.001 8.2 / 3 0.041 3.0 / 3 0.385 9.7 / 4 0.045aMC@NLO+Pythia8 6.2 / 3 0.104 4.6 / 3 0.200 1.9 / 3 0.597 18.4 / 4 0.001Powheg+Herwig7 2.5 / 3 0.478 2.1 / 3 0.543 5.7 / 3 0.128 10.0 / 4 0.041Sherpa 23.4 / 3 < 0.001 11.7 / 3 0.008 0.2 / 3 0.974 14.6 / 4 0.006Powheg+Pythia8? 21.0 / 3 < 0.001 20.5 / 3 < 0.001 0.9 / 3 0.821 13.0 / 4 0.011Flipped Powheg+Pythia8? 48.1 / 3 < 0.001 35.6 / 3 < 0.001 2.8 / 3 0.427 18.5 / 4 < 0.001

When comparing the data with the prediction for the exotic flipped colour flow model, the model itself isconsidered as an additional source of signal modelling uncertainty and thus added to the covariance matrix.Calculations which include this additional systematic are marked with ?.


Measurement of colour ﬂow in events with the ATLAS detector€¦ · Introduction Analysis Results...

Documents

Transcript of Measurement of colour ﬂow in events with the ATLAS detector€¦ · Introduction Analysis Results...