lc performance for jet reconstruction
DESCRIPTION
LC performance for Jet reconstruction. Paola Giovannini MPPMU HEC group. OUTLINE. Linearity for jets respect to Calibration Hit Truth Linearity for jets respect to Particle Jet Truth Linearity for jets with only one step reconstructed ( NEW ) - PowerPoint PPT PresentationTRANSCRIPT
LC performanc
e for Jet reconstructi
onPaola Giovannini
MPPMU HEC group
P.Giovannini, LIP ATLAS WORKSHOP
2
OUTLINE
Linearity for jets respect to Calibration Hit Truth Linearity for jets respect to Particle Jet TruthLinearity for jets with only one step reconstructed (NEW) Resolution respect to Particle Jet Truth, after recovering linearity Resolution inverting the order of the calibration
for jet performance studies: 225000 di-jet events reconstructed with the new Local Hadron Calibration weights for QGSP_BERTINI (ALL STEPS!) [15.1.0]
code to produce all the plots is available in CVS as an ARA package: /offline/PhysicsAnalysis/JetPhys/JetLCPerformanceARA/JetLCPerformanceARA
06/24/09
P.Giovannini, LIP ATLAS WORKSHOP
3
Validation with Cal Hits
Particle Jet
• default reference • provided by theory
ALWAYS be final reference BUT
• does not tell us what/how ends up in the calorimeter
• no B field, no hadronic shower, no detector geometry accounted for
Calibration Hit Jet
reference for every LC step
reflects the reference used in computing the weights
on single pions tells us HOW WELL we do
what we want to do uses algorithms to attach calibration hits to a jet these algorithms need to be validated using
ParticleID to be fully reliable
06/24/09
P.Giovannini, LIP ATLAS WORKSHOP
4
Validation with Cal Hits
06/24/09
HOW TO BUILD A TRUTH JET FROM CALIBRATION HITS
for every cluster in the RECONSTRUCTED jet
Σ
Calibration Hit Cluster Moment energy
Truth Jet Energy afterhadronic compensation (w)
==
Out Of Cluster Calibration Hit Cluster Moment energy
Truth Jet Energy due toout of cluster corrections (ooc)==
Dead Material Hit Cluster Moment energy
Truth Jet Energy due todead material corrections (dm)==
Truth Jet Energy afterLocal Hadron Calibration
SEE BACK-UP FOR MORE DETAILSON MOMENTS ALGORITHMS
P.Giovannini, LIP ATLAS WORKSHOP
5
LC Jet Linearity
06/24/09
plot Reco/Truth with Calibration hit
Numerator: Jet at a defined level of correction MC Reco
Denominator: Jet at the same defined level of correction MC TruthRATIO SHOULD BE 1
Jet after weighting ONLY
Jet after weighting AND out of cluster
Jet after weighting AND out of clusterAND dead material
(Calib-Hit)
η bin 0.3
W OOCDM
AntiKt4 used as Jet Algorithm
P.Giovannini, LIP ATLAS WORKSHOP
LC Jet Linearity
6
plot Reco/Truth with Calibration hit
W OOCDM
06/24/09
η bin 2.1 η bin 3.2
W OOCDM
Few comments: weighting step seems to work for every eta bin new physics list accounted for ! out of cluster corrections are underestimated for jetEn < 300 GeV dead material corrections are underestimated for large Eta values or small jet energies
(IF CALIBRATION HITS ARE TAKEN AS TRUTH REFERENCE)
P.Giovannini, LIP ATLAS WORKSHOP
7
LC Jet Linearity
06/24/09
plot Reco ONE STEP ONLY/Truth with Calibration hit
Numerator: Jet at a defined level of correction (MC Reco for last step only + CalHit truth)
Denominator: Jet at the same defined level of correction MC Truth (CalHit)
RATIO SHOULD BE 1
η bin 0.3
W OOCDM
AntiKt4 used as Jet Algorithm
ARECO+BTRUTH+CTRUTH
ATRUTH+BTRUTH+CTRUTH
In this way the single correction performance is isolated
P.Giovannini, LIP ATLAS WORKSHOP
8
LC Jet Linearityplot Reco ONE STEP ONLY/Truth with Calibration hit
η bin 2.1 η bin 3.2
W OOCDM
W OOCDM
This approach shows that both OOC and DM correctionsmiss 10% of energy respect to Calibration Hit Moments,
that is what LC wants to correct for
06/24/09
P.Giovannini, LIP ATLAS WORKSHOP
9
Hit Moments Validation
06/24/09
linearity plot Truth (CalibHit)/Truth (Particle Jet)
et bin 15 GeV
Numerator: Jet at a defined level of correction Calib Hits
Denominator: Particle Jet
• how much and how do we gain at every step? (even if we would make LC reconstruction perfect)
W OOCDM
W OOCDM
et bin 50 GeV
up to 20% of the Jet energy is due to DM
COMMENTS::-non uniform response in η-most of the jet energy is deposited NEAR to its constituents NEED Particle ID validation to be fully reliable
P.Giovannini, LIP ATLAS WORKSHOP
10
LC Jet Resolution
06/24/09
EMW OOCDM
Jet after weighting ONLY / Particle Jet Truth
Jet after weighting AND out of cluster / Particle Jet Truth
Jet after weighting AND out of clusterAND dead material / Particle jet Truth
restoring linearityfor every step
Jet at the em scale / Particle Jet Truth
DISTANCE BTW STEPSIS REDUCED:: IT IS A FAKE
IMPROVEMENT DUE TO NON-LINEARITIES
AntiKt4η bin 0.3
P.Giovannini, LIP ATLAS WORKSHOP
11 06/24/09
LC Jet ResolutionINVERTING ORDER OF LC CORRECTIONS..
OOC applied as the last correction..
after washing out nonlinearity effects the distancebetween the steps is reducedbut the last step is better it’s adding “knowledge”
corrected for non-linear effects
AntiKt4η bin 0.3
P.Giovannini, LIP ATLAS WORKSHOP
12 06/24/09
LC Jet ResolutionINVERTING ORDER OF LC CORRECTIONS..
W applied as the last correction..
corrected for non-linear effects
after washing out nonlinearity effects the distancebetween the steps is reducedbut the last step is better it’s adding “knowledge”
AntiKt4η bin 0.3
P.Giovannini, LIP ATLAS WORKSHOP
13 06/24/09
LC Jet Resolutionusing different algorithm parameter
corrected for non-linear effects
AntiKt6η bin 0.3
having a wider jetimproves resolution…..
P.Giovannini, LIP ATLAS WORKSHOP
14
Conclusions
06/24/09
• the first set of plots show that the actual constants don’t fully compensate for the OOC and DM Hit energy included in the Moments • Calibration Hit energy accounts for most of the Jet energy, but it is not possible to say if this energy belongs to the truth Jet ParticleID() • Resolution plots taking into account non-linearity effects have been produced[ see resolution with non linear calorimeters for more details on the method !!] at every step of the LC there is a resolution improvement after washing out non-linear effects .. wider jet seems to give a better resolution for di jet events