Optimal Jet Finder (OJF) in ATLASlefebvre/talks/... · D.Lelas (UVic) Optimal Jet Finder (OJF) in...

Optimal Jet Finder (OJF) in ATLASMichel Lefebvre, Rolf Seuster, Rob McPherson, Frank Michel Lefebvre, Rolf Seuster, Rob McPherson, Frank

Berghaus Damir LelasBerghaus Damir LelasBerghaus, Damir Lelas Berghaus, Damir Lelas

((University of Victoria, British Columbia, CanadaUniversity of Victoria, British Columbia, Canada))

ATLAS T i &Ph i W k J t/T /Et i T i dATLAS T i &Ph i W k J t/T /Et i T i dATLAS Trigger&Physics Week, Jet/Tau/Etmiss Trigger and ATLAS Trigger&Physics Week, Jet/Tau/Etmiss Trigger and Combined Performance MeetingCombined Performance Meeting

20 M h 200720 M h 2007

OUTLINE:OUTLINE:

20 March 200720 March 2007

•• Introduction to OJFIntroduction to OJF

•• OJF in ATLAS: experiences, findings and results OJF in ATLAS: experiences, findings and results

•• Some thoughts and remarksSome thoughts and remarksgg

•• Conclusions and OutlookConclusions and Outlook

Optimal Jet Finder (OJF)

Introduced and developed by:

F.V. Tkachov, D.Yu. Grigoriev and E. JankowskiF.V. Tkachov, D.Yu. Grigoriev and E. Jankowski

Short introduction: Phys. Rev. Lett. 91, 061801 (2003)Phys. Rev. Lett. 91, 061801 (2003)

Int. Journal of Mod. Phys. A, Vol 17, No 21 (2002) 2783Int. Journal of Mod. Phys. A, Vol 17, No 21 (2002) 2783--2884.2884.Int. Journal of Mod. Phys. A, Vol 17, No 21 (2002) 2783Int. Journal of Mod. Phys. A, Vol 17, No 21 (2002) 2783 2884.2884.

Authors webpage (with links to source code, etc.):

http://www inr ac ru/~ftkachov/projects/jets/welcome htmlhttp://www.inr.ac.ru/ ftkachov/projects/jets/welcome.html

Presented in JetRec phone meetings in January and March 2007

Results in this talk are from private OJF implementation in Athena

Official implementation can be expected soon (R. Seuster)

Discussions with experts has startedp

D.Lelas (UVic) Optimal Jet Finder (OJF) in ATLAS 2

Optimal Jet Finder (OJF)Tries to extract as much (jet) information from a HEP event as possible

(that is what authors call ‘optimal’…)

OJF starts from rather ambitious, global point of view (i.e. from event view):

HEP event: list of particles ( t h d l i t ll t l t )

parts,,2,1, napa K=(partons • hadrons • calorimeter cells • towers • preclusters)

}{recombination matrix jetsp arts}{ nnjaz ×

tn

∑=

=p arts

1

n

aaajj pzq the 4-momentum qj of the j-th jet

expressed by 4-momenta pa of the ti l

lt li t f j t 21j

particles

result: list of jets jets,,2,1, njq j K=


Optimal Jet Finder (OJF)Any allowed value of the Recombination Matrix Recombination Matrix {{zzajaj }} from previous slidejj

describes some jet configuration

‘Whole trick’ is to find desired optimal jet configuration by minimizing ‘Whole trick’ is to find desired optimal jet configuration by minimizing

some function some function Ω({Ω({zzajaj})})Definition from Optimal Jet Finder gives (for cylindrical kinematics (pp coll.):p g ( y (pp )

{ }( ) ∑∑ ∑ ⊥⊥⊥ +⎟⎟⎠

⎞⎜⎜⎝

⎛ −+

−=Ω

p artsp art22

2 2sin

2sinh4 n

aa

n njaja

ajaTotaj EzzER

Ezjets ϕϕηη

== =⎟⎠

⎜⎝ 11 1 22 aa jR

∑ ⊥p artsn

aaajEz η ‘width’ of the j-th jet energy outside jets

p artsn ⊥q∑

=

⊥

=≡p arts

1

1n

aaaj

aj

Ezη ( ) ( )2y2x

aaa ppE +≡⊥ ∑=

−≡jets

11

n

jaja zz

( ) ∑=

≡≡p arts

1,

aaajjjj pzEq q ( )jj

j

j ϕϕ sin,cos≡⊥q

q

E is the energy of the a-th particle

0, ≥aja zzEa is the energy of the a th particle

RR>0 is a parameter with a similar meaning as the cone radius>0 is a parameter with a similar meaning as the cone radius


OJF features

The authors claim it is based on an optimal jet definition that The authors claim it is based on an optimal jet definition that solves the problem of jet definition in generalsolves the problem of jet definition in general

OJF is infrared and collinear safeOJF is infrared and collinear safeOJF is infrared and collinear safeOJF is infrared and collinear safeno issues intrinsic with seedsno issues intrinsic with seedsno issues with overlapping cones: all jets are “ready to use”no issues with overlapping cones: all jets are “ready to use”

Particle energy can be shared among jetsParticle energy can be shared among jetsHadronization is always an effect of the interaction of at least two hard Hadronization is always an effect of the interaction of at least two hard

t l i i t t j t h d th t i thit l i i t t j t h d th t i thipartons evolving into two jets, so some hadrons that emerge in this partons evolving into two jets, so some hadrons that emerge in this process can belong partially to both jetsprocess can belong partially to both jets

The association between particles and jets is obtained through the The association between particles and jets is obtained through the minimization of a global functionminimization of a global function

global variables are a biglobal variables are a bi--product of this procedureproduct of this procedureglobal variables are a biglobal variables are a bi product of this procedureproduct of this procedure

Particles (or part of them) are allowed to be outside all jetsParticles (or part of them) are allowed to be outside all jetsbut with a penalty in the global function to minimizebut with a penalty in the global function to minimizebut with a penalty in the global function to minimizebut with a penalty in the global function to minimize


OJF running modesThe algorithm can be used in two modes:The algorithm can be used in two modes:The algorithm can be used in two modes:The algorithm can be used in two modes:

1) Number of jets fixed1) Number of jets fixed-- Parameter Parameter RR (related to conventional (related to conventional RRconecone) and ) and number number

of initial random of initial random ZZajaj configurationsconfigurations to be specifiedto be specified

2) Number of jets not fixed, obtained by the algorithm2) Number of jets not fixed, obtained by the algorithm2) Number of jets not fixed, obtained by the algorithm2) Number of jets not fixed, obtained by the algorithm-- Parameter Parameter RR, number of initial random , number of initial random ZZaj aj configurations configurations andand

ParameterParameter ωω to be specified:to be specified:-- Parameter Parameter ωωcutcut to be specified:to be specified:

-- related to the jet resolution related to the jet resolution yycutcut of conventional of conventional

bi i l i hbi i l i hrecombination algorithmsrecombination algorithms

-- upper limit to the fractional transverse energy not upper limit to the fractional transverse energy not

associated to any jetsassociated to any jets


Fully hadronic t-tbar sample

Sample 5204 (13000 events)Sample 5204 (13000 events)

MC@NLO fully hadronic tMC@NLO fully hadronic t--tbar, full simulation, CSC12 prod.tbar, full simulation, CSC12 prod.

AOD produced using Athena 12.0.6 and OJF private codeAOD produced using Athena 12.0.6 and OJF private code

-- final jet cutfinal jet cut EETT > 7 GeV> 7 GeVfinal jet cut final jet cut EET T > 7 GeV> 7 GeV

-- Jet collections using OJF:Jet collections using OJF:

H1 i ht il bl C H1 i htH1 i ht il bl C H1 i ht→-- no proper H1 weights available: use Cone H1 weightsno proper H1 weights available: use Cone H1 weights

-- number of jets fixed to 6, number of jets fixed to 6, RR parameter set to parameter set to 0.4 or 0.70.4 or 0.7

→

(labels for following plots: (labels for following plots: OJFN6R4_xxxOJFN6R4_xxx and and OJFN6R7_xxxOJFN6R7_xxx, respect.), respect.)

-- number of jets not fixed, number of jets not fixed, RR parameter set to parameter set to 0.70.7 and and ωωcutcutj ,j , pp cutcut

set to set to 0.40.4 (labels for following plots: (labels for following plots: OJFR7W4_xxxOJFR7W4_xxx))

jet collection from MC Truth CaloTowers and TopoClustersjet collection from MC Truth CaloTowers and TopoClusters-- jet collection from MC Truth, CaloTowers and TopoClustersjet collection from MC Truth, CaloTowers and TopoClusters


Jet multiplicity: CaloTower jetsWith N_jets > 5 cut

even

ts

c t-t

bar e

hadr

onic

With N_jets > 5 cut

fully

h

OJF jets only OJF jets only


Jet-Parton matching (ttbar events)

Matching criteriaMatching criteriaFor each parton, look for a matching jetFor each parton, look for a matching jet

-- restrict search in a region limited by restrict search in a region limited by ΔΔRRmax max < 0.2< 0.2-- keep the closest jet in this regionkeep the closest jet in this region

Demand that a jet be matched only onceDemand that a jet be matched only once-- matched 1 to 1matched 1 to 1-- one can then study matching efficiency one can then study matching efficiency

Study events with “true” jet hypothesis (for ttbar events)Study events with “true” jet hypothesis (for ttbar events)Require all 6 partons to be matched 1 to 1Require all 6 partons to be matched 1 to 1

-- this way we can study the top mass reconstruction this way we can study the top mass reconstruction f f diff t j t l ithf f diff t j t l ithperformance for different jet algorithmsperformance for different jet algorithms

For much more quantitative details on matching eff., For much more quantitative details on matching eff., please see extra slides!!please see extra slides!!


Normalized distributions: CaloTower jetsro

nic

| < 3

Jet – Parton

All 1 to 1 matched jets in events with

ully

had

ets

in |η

R < 0.2

at least 6 jets in |eta| < 3

OJF with R = 0.7 is between Cone4 and 7

rton

in fu

east

6 je Similar results for topoJets and truthJets

(extra slides)

1 je

t-par

with

at l

ed 1

to

even

ts w

mat

cht-t

bar


Normalized Normalized ηη distributions: OJF towerJetsdistributions: OJF towerJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized E distributions: towerJetsNormalized E distributions: towerJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized E distributions: topoJetsNormalized E distributions: topoJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized pNormalized pTT distributions: towerJetsdistributions: towerJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized pNormalized pTT distributions: OJF towerJetsdistributions: OJF towerJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized pNormalized pTT distributions: topoJetsdistributions: topoJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


Normalized pNormalized pTT distributions: OJF topoJetsdistributions: OJF topoJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a


W mass plots: TowerJetsW mass plots: TowerJetst 6

jets

1

%6.15=⟩⟨M

RMS %6.16=⟩⟨M

RMS

at le

ast

hed

1 to

nt

s w

ith

ns m

atch

tbar

eve

ll pa

rton

RMS RMSroni

c t-t

3w

ith a

%0.18=⟩⟨M

RMS %1.17=⟩⟨M

RMS

fully

had

n |η

| < 3


f in

W mass plots: OJF TowerJetsW mass plots: OJF TowerJetst 6

jets

1

%2.16=⟩⟨M

RMS %3.15=⟩⟨M

RMSat le

ast

hed

1 to

⟩⟨M ⟩⟨

nts

with

ns

mat

chtb

ar e

vell

parto

nro

nic

t-t3

with

afu

lly h

adn

|η| <

3 %5.15=⟩⟨M

RMS

f in


top mass plots: TopoJetstop mass plots: TopoJetst 6

jets

1

%9.12=⟩⟨M

RMS %3.12=⟩⟨M

RMSat le

ast

hed

1 to

⟩⟨ ⟩⟨

nts

with

ns

mat

chtb

ar e

vell

parto

n

RMSroni

c t-t

3w

ith a

RMS%5.14=⟩⟨M

RMS

fully

had

n |η

| < 3 %2.13=

⟩⟨MRMS

f in


top mass plots: OJF TopoJetstop mass plots: OJF TopoJetst 6

jets

1

%0.12=⟩⟨M

RMS %5.11=⟩⟨M

RMSat le

ast

hed

1 to

⟩⟨ ⟩⟨

nts

with

ns

mat

chtb

ar e

vell

parto

nro

nic

t-t3

with

afu

lly h

adn

|η| <

3 %5.13=⟩⟨M

RMS

f in


Conclusions and OutlookThere is additional jet finder on the market, Optimal Jet Finder (OJF)j p ( )

The authors claim it is based on an optimal jet definition that solves the problem

of jet definition in general OJF is based on global event propertiesof jet definition in general. OJF is based on global event properties

It is infra-red and collinear safe and there are no seed-related problems nor

overlapping jets. It could provide more event info than standard jet algorithms

Official implementation in Athena is very much in progressp y p g

First tests with OJF in the ATLAS environment are quite encouraging.

A t ith Kt d C i d b t ld lik t d b tt ☺Agreement with Kt and Cone is good, but we would like to do better ☺

More systematic tests in progress:- fine tuning of OJF parameter spaceg p p- studies with different physical samples

Years of testing for Cone and Kt, OJF deserves a bit of attention as well

Hopefully, this is just a beginning of long and interesting journey…


Extra Slides

Normalized distributions: Topo jetsni

c t-

3Jet – Parton

All 1 to 1 matched jets in events with

y ha

dron

n |η

| < 3

R < 0.2

at least 6 jets in |eta| < 3

OJF with R = 0.7 is between Cone4 and

on in

fully

t 6 je

ts in Cone7

jet-p

arto

at l

east

d 1

to 1

jnt

s w

ithm

atch

edtb

ar e

vem t

Normalized distributions: truthJetsNormalized distributions: truthJetson

ic

< 3

lly h

adro

ts in

|η| <

ton

in fu

ast 6

jet

jet-p

art

with

at l

eaed

1 to

1ev

ents

wm

atch

et-t

bar e

Event and Jet hypothesis selection

Try to compare jet matching efficiency for Try to compare jet matching efficiency for different jetdifferent jet algorithmsalgorithms

Fully hadronic tFully hadronic t--tbar, choice of sampletbar, choice of sampleRequire at least 6 jets in |Require at least 6 jets in |ηη| < 3| < 3Require at least 6 jets in |Require at least 6 jets in |ηη| 3| 3-- clearly, algorithms with more than 6 jets will have a betterclearly, algorithms with more than 6 jets will have a better

chance at 6 jetchance at 6 jet--parton matchingparton matchingonly consider the 6 highest pT jetsonly consider the 6 highest pT jets-- algos with more than 6 jets still have a better chance at algos with more than 6 jets still have a better chance at

6 jet6 jet--parton matchingparton matchingonly consider events with exactly 6 jetsonly consider events with exactly 6 jets

h f i t 6 j th f i t 6 j t t t hit t hi-- perhaps a more fair way to compare 6 jetperhaps a more fair way to compare 6 jet--parton matching parton matching for OJF with fixed number of jets (= 6)for OJF with fixed number of jets (= 6)

No jet ENo jet E cuts appliedcuts appliedNo jet ENo jet ETT cuts appliedcuts applied

Jet-Parton matching: CaloTower jets

It is very difficult to come up with totally fair matching-fair matchingefficiency comparisons. But, even without tuning, OJF ltOJF results are already comparable to Cone and Kt

Jet-Parton matching: topoJets

It is very difficult to ith t t llcome up with totally

fair matching-efficiency comparisons. But, p ,even without tuning, OJF results are already comparable to C d KtCone and Kt

Normalized pNormalized pTT distributions: truthJetsdistributions: truthJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a

Normalized pNormalized pTT distributions: OJF truthJetsdistributions: OJF truthJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a

Normalized E distributions: OJF towerJetsNormalized E distributions: OJF towerJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a

Normalized E distributions: OJF topoJetsNormalized E distributions: OJF topoJetsev

ents

n

|η| <

3c

t-tba

r 6

jets

i nha

dron

iat

leas

t fu

lly

with

Normalized E distributions: truthJetsNormalized E distributions: truthJetsev

ents

|η

| < 3

c t-t

bar e

6 je

ts in

hadr

onic

at le

ast 6

fully

hw

ith a

W mass plots: TopoJetsW mass plots: TopoJetst 6

jets

1

at le

ast

hed

1 to

nt

s w

ith

ns m

atch

tbar

eve

ll pa

rton

roni

c t-t

3w

ith a

fully

had

n |η

| < 3

f in

W mass plots: OJF TopoJetsW mass plots: OJF TopoJetst 6

jets

1

at le

ast

hed

1 to

nt

s w

ith

ns m

atch

tbar

eve

ll pa

rton

roni

c t-t

3w

ith a

fully

had

n |η

| < 3

f in

top mass plots: TowerJetstop mass plots: TowerJetst 6

jets

1

at le

ast

hed

1 to

nt

s w

ith

ns m

atch

tbar

eve

ll pa

rton

roni

c t-t

3w

ith a

fully

had

n |η

| < 3

f in

top mass plots: OJF TowerJetstop mass plots: OJF TowerJetst 6

jets

1

at le

ast

hed

1 to

nt

s w

ith

ns m

atch

tbar

eve

ll pa

rton

roni

c t-t

3w

ith a

fully

had

n |η

| < 3

f in

Optimal Jet Finder (OJF) in ATLASlefebvre/talks/... · D.Lelas (UVic) Optimal Jet Finder (OJF) in...

Documents

Transcript of Optimal Jet Finder (OJF) in ATLASlefebvre/talks/... · D.Lelas (UVic) Optimal Jet Finder (OJF) in...