Performance of Jet Substructure Techniques and Boosted Object Identification in ATLAS

Jim Lacey (Carleton University), for the ATLAS Collaboration

ICHEP Poster Session - 4 July 2014

Introduction

[GeV]Ttp

0 100 200 300 400500 600 700 800 900

R(W

,b)

Δ

00.20.40.60.8

11.21.41.61.8

22.2

020406080100120140160180200

Wb→, t t t→Pythia Z'ATLAS Simulation

[GeV]TWp

0 100 200 300 400 500 600 700 800

)q R

(q,

Δ

0

0.5

1

1.5

2

2.5

3

3.5

4

020406080100120140160180200

Wb→, t t t→Pythia Z' ATLAS Simulation

Jet grooming techniques are designed to improve the mass resolution of hadronically decaying boosted objects. Jet Trimming divides large-R jets into N subjets, removing soft components. Jet Pruning is similar to trimming in that it removes constituents with a small relative pT, but it additionally applies a veto on wide-angle radiation.

At the LHC, √s >> electroweak scale, therefore massive particles like top, W, Z and Higgs are often produced with a significant boost with the decay products reconstructed as a single jet. Jet substructure techniques mitigate pile-up and probe inside the jet to identify the boosted objects. Important for exploring the boosted kinematic regime, extending understanding of the SM and searching for new physics.

Jet Grooming

Trimming in combination with a jet-area-based jet 4-momentum correction are expected to be performant at the high luminosity and pile-up conditions expected during the 14 TeV running of the LHC.

Boosted W Boson Identification

Boosted Boson Identification combines jet grooming techniques with jet substructure variables, improving the boosted object tagging efficiency and QCD background rejection. Data/MC agreement for tagging variables is verified using a high purity sample of W bosons from top decays. f

y0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Norm

alis

ed E

ntr

ies

0

0.05

0.1

0.15

0.2

0.25

0.3C/A LCW jets with R=1.2

BDRS

(POWHEG)ttW+jets

Single Top

Z+jets

Stat. Uncertainty

DataK.S. = 0.525

ATLAS Preliminary

=8 TeVs -1=20.1fbintL

| < 1.2TRUTH!|

< 350 GeVTRUTH

T200 < p

WindowRECO

M

Planar Flow

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

No

rma

lise

d E

ntr

ies

0

0.05

0.1

0.15

0.2

0.25

0.3 jets with R=1.0

tanti-k

Trimmed

(POWHEG)ttW+jets

Single Top

Z+jets

Stat. Uncertainty

DataK.S. = 0.992

ATLAS Preliminary

=8 TeVs -1=20.1fbintL

| < 1.2TRUTH!|

< 350 GeVTRUTH

T200 < p

WindowRECO

M

Jet Charge

Positive Charge Efficiency0 0.2 0.4 0.6 0.8 1

Neg

ativ

e C

harg

e R

ejec

tion

1

10

210

310

ATLAS Preliminary-1 L dt = 5.8 fb∫ = 8 TeV, s

Data MC Samplett

= 1.0κ

= 0.3κ

= 1.0κ

= 0.3κ

Reconstructed Jet Charge (e)

Even

ts/0

.07e

0

100

200

300

400

500

600

700 ATLAS Preliminary-1 L dt = 5.8 fb∫ = 8 TeV, s

=1.0κ

Data 2012+µ

Data 2012-µ

t MC@NLO t+µ

t MC@NLO t-µ

Background (MC)+µ

Background (MC)-µ

Jet Charge [e]-1 0 1

Data

/MC

0

1

2

Q j =1

(pT j)X

i2Tr

qi ⇥ (piT ),

ATLAS-CONF-2013-086

ATL-PHYS-PUB-2014-004

Quark/Gluon Tagging

Tracking information is used to discriminate between quark and gluon initiated jets. A likelihood-based Quark-Gluon Tagger has been implemented and validated using 4.7 fb-1 of 7 TeV data. For data, a ~45% gluon-jet efficiency is achieved for a 70% quark-jet efficiency.

corrJVF0 0.5 1

)un

groo

med

T/p

subj

T(p

10lo

g

-2

-1

0

1

Nor

mal

ized

Ent

ries

-410

-310

-210

ATLAS Simulation Preliminary qqqq)A WZAPythia8 (W'

= 1 TeVW'M LCW R=1.0 jettAnti-k

LCW R=0.3 subjetstk

h i

corrJVF =P

k ptrkkT (PV0)

Pl ptrkl

T (PV0) +P

n�1P

l ptrklT (PVn)

(k·nPUtrk )

.

arXiv:1405.6583v1

Pure Track-based Trimming places a requirement on track-based variables of reclustered subjets. One such variable is the pile-up corrected Jet Vertex Fraction (corrJVF). Simulated W′ → WZ → qqqq events are used for performance evaluation.

ATLA

S-C

ON

F-20

14-0

18

Trimmed jetInitial jet pTi/pTjet<fcut

kt, R=Rsub

Pruned jetInitial jet pTj2/pTj1+j2<zcut or ΔRj1,j2 < Rcut

kt or C/A

Quantum Jets

Jets

10

210

310

410 = 8 TeVs,

-1 0.6 fb± L dt = 20.3 !

R=0.7 LCt

W-jet Selection, anti-k

, C/A PruningT

z = 0.1, d = m/p

= 0.1" = 75, Q-jetsN

ATLAS PreliminaryDataTop

DibosonsSingle Top

Z+jetsW+jets

stat+syst#

Volatility0 0.2 0.4 0.6 0.8 1

Data

/ M

C

0.5

1.0

1.5

Jets

410

510

610

710

810 = 8 TeVs,

-1 1.0 pb± L dt = 36.3 !

R=0.7 LCt

Dijet Selection, anti-k

, C/A PruningT

z = 0.1, d = m/p

= 0.1" = 75, Q-jetsN

ATLAS PreliminaryDataPythia Dijets

stat#

Volatility0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Data

/ M

C

0.5

1.0

1.5

The Q-Jets technique interprets jets through multiple sets of possible showering histories. Jet observables are evaluated as distributions and not simply as single quantities. The volatility provides powerful discrimination between boosted particles and QCD backgrounds.

ATLA

S-C

ON

F-20

13-0

87

Top Tagging

Mass [GeV]0 50 100 150 200 250 300 350 400

Entri

es /

10 G

eV

0102030405060708090

100Data 2012ttbar (with contained top)ttbar (non-contained top)W + jetsSingle TopZ+jetsStatistical uncertainty

PreliminaryATLAS,-1 L dt = 20.3 fb∫ = 8 TeVs

R=1.0, Trimmedtanti-k|<1.2η>500 GeV, |

Tp

y-1 -0.5 0 0.5 1 1.5 2 2.5

φ

0

0.5

1

1.5

2

2.5 Preliminary Simulation ATLAS = 1.75 TeVZ’m event, tt → Z’

= 186.5 GeVWbm = 77.3 GeV, Wm

= 1.0R tkAnti-Calorimeter clusters

= 0.2RSubjets, C/A bosonW

jetbTop radiationISR

Shower Deconstruction categorizes a jet into N sub-jets, representing a possible showering history. The probability that a given history was realized is used to distinguish jets originating from boosted heavy particles from QCD backgrounds.

ATLA

S-C

ON

F-20

14-0

03

Jet Charge is a quark-charge sensitive observable which can be used for identifying the charge of hadronically-decaying heavy particles.

ATLA

S-C

ON

F-20

14-0

03