Jets and Hadrons Calibration Jets and Hadrons Calibration Strategy in ATLAS Strategy in ATLAS
Irene Vichou Irene Vichou University of AthensUniversity of Athens
IntroductionIntroductionCalibration Strategy (from testbeam to LHC data)Calibration Strategy (from testbeam to LHC data)Jet and hadron reconstruction Jet and hadron reconstruction In situ methodsIn situ methods
Z+jets and W jj samples
ConclusionsConclusions
July 2002 Ringberg Calibration Workshop - Irene Vichou 2
ConsiderationsConsiderations
The goal is to:The goal is to: reconstruct hadronic
entities on the calorimeter but also muons to assist the Spectrometer and transfer the calibration csts be able to reproduce as
precisely as possible their 4-momenta
This implies:This implies: pass the knowledge of
calorimeter response from the tests of modules
to the assembled detector to physics events
Physics casesPhysics cases mass reconstruction # jets and total E in SUSY
searches jet veto down to 15 GeV QCD studies (… unexpected)
RequirementsRequirements Knowledge of E scale for
jets should be ~1%
July 2002 Ringberg Calibration Workshop - Irene Vichou 3
...Considerations...Considerations
Physics systematicsPhysics systematics fragmentation ISR FSR underlying and MinBias
Detector SystematicsDetector Systematics response different to charged and
neutral hadrons non linearities B-field dead material longitudinal shower leakage lateral shower size granularity electronics noise
July 2002 Ringberg Calibration Workshop - Irene Vichou 4
Layout of ATLAS calorimetersLayout of ATLAS calorimeters
Calorimeters in ATLAS cover up to ||=5.
The electromagnetic calorimeter is a liquid argon accordion.
The hadronic calorimeter is an iron-scintillator device in the central region and Lar in the endcap and forward region
The EC cryostat houses EM, HAD and FCAL.
July 2002 Ringberg Calibration Workshop - Irene Vichou 5
EM calorimeter segmentationEM calorimeter segmentation
•3 longitudinal samplings3 longitudinal samplings•very fine strips in 1st very fine strips in 1st •thick middle samplingthick middle sampling•good for good for //separationseparation
presampler in front forpresampler in front forenergy recoveryenergy recovery
July 2002 Ringberg Calibration Workshop - Irene Vichou 6
Calibration StrategyCalibration Strategy
Before LHC….Before LHC…. Calibration with single particles at testbeamCalibration with single particles at testbeam transfer of calibration constants to all modules transfer of calibration constants to all modules
so that all calorimeters are set at the EM scaleso that all calorimeters are set at the EM scale
LHC start !LHC start ! intercalibration with single particles at LHCintercalibration with single particles at LHC in situ samplesin situ samples
W jet jet Z(photon)+jet balance
for E regions not reached before? MC...for E regions not reached before? MC...
July 2002 Ringberg Calibration Workshop - Irene Vichou 7
TileCal calibration with CsTileCal calibration with Cs
AIMAIM: : Equalize cell responses and Equalize cell responses and balance optical non-uniformities.balance optical non-uniformities.
All individual optical components All individual optical components are “seen” in a shadowgram.are “seen” in a shadowgram.
The precision in the cell The precision in the cell response isresponse is 0.2% !!! 0.2% !!!
There is a good correlationThere is a good correlationwith muons ~ 3%with muons ~ 3%
July 2002 Ringberg Calibration Workshop - Irene Vichou 8
Calibration in test beam Calibration in test beam
eeee
TileCal moduleTileCal modulee/e/ response = f(e/h, response = f(e/h,00 fraction) fraction) e/h (TileCal) = 1.3 ± 0.01e/h (TileCal) = 1.3 ± 0.01check resolution check resolution
Muon response (90Muon response (90oo) is due to ionisation) is due to ionisation•1 in 8 modules will see particles1 in 8 modules will see particles•muon signals correlate with Cs, muon signals correlate with Cs, so...so...Cs constants will be used to transfer theCs constants will be used to transfer theresponses to particles responses to particles
July 2002 Ringberg Calibration Workshop - Irene Vichou 9
Tuning the Monte-CarloTuning the Monte-Carlo
Combined Test of EM and HAD calorimeters in beams Combined Test of EM and HAD calorimeters in beams (only once for final ATLAS modules!)(only once for final ATLAS modules!)
E ResolutionE Resolution
60% sampling term, 1.8% cst term60% sampling term, 1.8% cst term2 GeV noise term2 GeV noise term
LinearityLinearity e/e/ ratio ratio 1.35 to 1.371.35 to 1.37
• able to have a suitable hadronic shower Monte-Carloable to have a suitable hadronic shower Monte-Carlo• useful to establish energy reconstruction methodsuseful to establish energy reconstruction methods
July 2002 Ringberg Calibration Workshop - Irene Vichou 10
Tuning the MC (G4)Tuning the MC (G4)
Pion Test beam resultsPion Test beam resultsvs vs
G4 for hadronic showers G4 for hadronic showers for Lar HECfor Lar HEC
July 2002 Ringberg Calibration Workshop - Irene Vichou 11
E/p for isolated hadronsE/p for isolated hadrons
and and kk’s from ’s from W(Z) + jets W(Z) + jets ( (++--) + jets) + jets, with , with h h at LHC at LHC
Backgrounds:Backgrounds:• QCD jetsQCD jets• multi-particle multi-particle decaysdecays under control with under control with cuts on Ncuts on Ntracktrackand isolationand isolation
Triggerable (Triggerable ( and E and ETTmissmiss))
p(tracker) = E(calorimeter)p(tracker) = E(calorimeter)
E/p bias:E/p bias:0.3% from QCD0.3% from QCD3.5% total, mainly from 3.5% total, mainly from decaysdecays containing containing
Need to reject the Need to reject the bcg bcg by aby a factor of 5 factor of 5
July 2002 Ringberg Calibration Workshop - Irene Vichou 12
E/p for isolated hadronsE/p for isolated hadrons
Make use of the fine EM caloMake use of the fine EM calogranularity of the 1st sapling.granularity of the 1st sapling.
A rejection of 11 is achieved A rejection of 11 is achieved at at =0.3=0.3
The residual bias in E/p dueThe residual bias in E/p dueto the to the bcg is bcg is 0.4%0.4%
Nr of Nr of -strips-strips
E(em1)/E(em)E(em1)/E(em)
July 2002 Ringberg Calibration Workshop - Irene Vichou 13
sample #events
SIGNAL 530 x103
QCD bcg 30 x103
bcg 20 x103
E/p for hadronsE/p for hadrons Statistics for 10 fbStatistics for 10 fb-1-1
Global E/p bias is Global E/p bias is 0.6 %0.6 % (OK if < 1%) (OK if < 1%) The E reaches The E reaches
120 GeV in TileCal (barrel + extended) 240 GeV in the Endcap Lar
It is a good intercalibration methodIt is a good intercalibration method Gives absolute scale for hadronsGives absolute scale for hadrons Direct comparison with beam tests for h’s/mip in EM Direct comparison with beam tests for h’s/mip in EM
July 2002 Ringberg Calibration Workshop - Irene Vichou 14
… … before using the in situ… before using the in situ… (in these studies)(in these studies)
fixed-cone algorithmfixed-cone algorithm
•““Benchmark” Benchmark” E= E= 11xxEE11++22xxEE22++33xxEE33++44xxEE44
•““a la H1” a la H1” E=EE=E11++22((2,j2,j))xx2,j 2,j + + 33((3,j3,j))xx3,j 3,j ++ 44xxEE44
where where 1…41…4 are PS,EM,HAD,cryostat correction terms are PS,EM,HAD,cryostat correction terms andand j j’s refer to cells ’s refer to cells and and EE4 4
`̀ss are function parametrisations wrt to parton energyare function parametrisations wrt to parton energy
Preserves linearity (calo effects) and resolution Preserves linearity (calo effects) and resolution
Jet Energy ReconstructionJet Energy Reconstruction
Cones Cones RR=0.4 and 0.7=0.4 and 0.7Seed 2 GeV,Ecell>2Seed 2 GeV,Ecell>2 noise noiseEtower Etower (0.1x0.1)(0.1x0.1) > 0.2 GeV > 0.2 GeV
Jet ?Jet ?
Jet E ?Jet E ?
E Eem had3 1
July 2002 Ringberg Calibration Workshop - Irene Vichou 15
Jet Energy ReconstructionJet Energy Reconstruction
Sampling(% GeV1/2)
Constant()
Barrel 52.3 1.1 1.7 0.1
Endcap 64.2 2.4 3.6 0.2
Jet Energy Resolution termsJet Energy Resolution terms
And dependece on pseudorapidityAnd dependece on pseudorapidity
50%50% 3%3%
R=0.7R=0.7
samplingsampling constantconstant
July 2002 Ringberg Calibration Workshop - Irene Vichou 16
Taus calibrationTaus calibration
Will be covered in Ambreesh’s talk as it is part of testWill be covered in Ambreesh’s talk as it is part of testcarried out in the new ATHENA frameworkcarried out in the new ATHENA framework
July 2002 Ringberg Calibration Workshop - Irene Vichou 17
Extrapolation to high EExtrapolation to high E
And after calibrating…..And after calibrating…..What if wrong extrapolation to energies not reached What if wrong extrapolation to energies not reached before?before?
Example: Use of a Monte-Carlo not reproducing well the e/hExample: Use of a Monte-Carlo not reproducing well the e/h
norm at 100 GeVnorm at 100 GeV
norm at 500 GeVnorm at 500 GeV
July 2002 Ringberg Calibration Workshop - Irene Vichou 18
In situ samplesIn situ samples
NeedNeed: Absolute for verification: Absolute for verificationWishWish: Statistics should be adequate above bcg: Statistics should be adequate above bcg Trigger available Trigger available Wide Energy coverage Wide Energy coverage
ChoicesChoices: Mass reconstruction of the W: Mass reconstruction of the W• only light q jetsonly light q jets• limited E and limited E and reachreach
Balance of pBalance of pTT between Z and leading jet between Z and leading jet • wider E andwider E andreachreach• b-jet contentb-jet content• good to transfer calibration across calorimeters?good to transfer calibration across calorimeters?
July 2002 Ringberg Calibration Workshop - Irene Vichou 19
In situ samples In situ samples
Simulated using PYTHIA/JETSETSimulated using PYTHIA/JETSET
Fast simulationFast simulation• smearing with calorimeter resolutionsmearing with calorimeter resolution• jet finding from deposited energy in towersjet finding from deposited energy in towers
Full simulationFull simulation
• using GEANT 3.21using GEANT 3.21• G-Calor calorimeter response modelG-Calor calorimeter response model• addition of pile-up events and electronics noiseaddition of pile-up events and electronics noise• digital filtering for cell energy reconstruction digital filtering for cell energy reconstruction
In the following studies In the following studies
• Fixed cone algorithm used for jet reconstructionFixed cone algorithm used for jet reconstruction• Benchmark/Sampling method for jet E calculationBenchmark/Sampling method for jet E calculation and also weighting method for the Z+jetsand also weighting method for the Z+jets
July 2002 Ringberg Calibration Workshop - Irene Vichou 20
In situ W In situ W jj jj
W’s are abundant through the tt events sampleW’s are abundant through the tt events sampleAt low L 1500 lAt low L 1500 lvvjjbb final states/day !jjbb final states/day !
The jj final states of the W decay should satisfy:The jj final states of the W decay should satisfy:
•Events easily triggered from isolated lepton and 2 b-jets.Events easily triggered from isolated lepton and 2 b-jets.•No physics bcg’s.No physics bcg’s.•Combinatorial bcg under the W peak ~ 30%.Combinatorial bcg under the W peak ~ 30%.
ppTTpartonparton/p/pTT
jetjet
~50 GeV out of cone losses
>200 GeV jet overlap pbms (add cut of R>0.8)
M E E Mjj j j j j W 2 11 2 1 2( cos )
Characteristics:Characteristics:
July 2002 Ringberg Calibration Workshop - Irene Vichou 21
In situ W In situ W jj jj
After calibrating (full sim):
>70 GeV 1% is achievable
~50 GeV 3% looks fair
Further studies showed:Further studies showed:overestimation of corrected jet energy towards the high end of overestimation of corrected jet energy towards the high end of E-spectrum was due to underestimation of the angle.E-spectrum was due to underestimation of the angle.
Problem solved by taking into account E,Problem solved by taking into account E,and and in the minimisation in the minimisationprocedure and correct energies and angles.procedure and correct energies and angles.
Result: Result: E of parton and jet agree within ~ 1% over the range 50-250 GeVE of parton and jet agree within ~ 1% over the range 50-250 GeV
July 2002 Ringberg Calibration Workshop - Irene Vichou 22
In situ WIn situ Wjet jetjet jet
Statistics: Statistics: 100 K100 K W’s for W’s for 10 fb10 fb-1-1
The W The W jj decay in tt events should provide the jj decay in tt events should provide the wished 1% absolute scale precision from 50 to wished 1% absolute scale precision from 50 to few hundred GeV for few hundred GeV for light quarkslight quarks
Lower end of spectrum:Lower end of spectrum: interplay between FSR and out of cone losses for specific
reconstruction algorithms
Higher endHigher end residual effects due to overlap are significant and special
treatment has to be applied to overcome this without significant loss of statistics (e.g. by asking well-separated jets)
July 2002 Ringberg Calibration Workshop - Irene Vichou 23
In situ Z+jetIn situ Z+jet
Direct Z production, events with Z+jets where ZDirect Z production, events with Z+jets where Zee or Zee or ZMethod: pMethod: pT T balance between Z (leptonic/precisely measured) andbalance between Z (leptonic/precisely measured) and
highest phighest pTT jet jet
Large cross section, no statistics problemLarge cross section, no statistics problem
… … but balance is not perfectbut balance is not perfect
HP +UE +FSR +ISR0.0260.015
0.0050.002
0.0500.034
0.0970.071
ppTT imbalance wrt to processes imbalance wrt to processes
20<p20<pTT<60 GeV jets<60 GeV jets
60<p60<pTT<120 GeV jets<120 GeV jets
fractional imbalance between Z and jet: fractional imbalance between Z and jet:
FrIp p
pT
Z
T
j
T
Z
July 2002 Ringberg Calibration Workshop - Irene Vichou 24
In situ Z+jetIn situ Z+jet
ppT T imbalance wrt to selection cutsimbalance wrt to selection cuts1 jet pT>20 GeV <3.2
>3.06 rad >3.06 radLoose jet veto
>3.06 radTight jet veto
0.0970.071
0.0500.042
0.0490.041
0.0440.033
Selections to favour topologies with back-to-back Z and jetSelections to favour topologies with back-to-back Z and jet•azimuthal angle between Z and leading jet > 3.06 radazimuthal angle between Z and leading jet > 3.06 rad•loose jet veto: no other jet ploose jet veto: no other jet pTT>40 GeV and |>40 GeV and ||<3.2 (|<3.2 (possible at high Lpossible at high L))•tight jet veto: no other jet ptight jet veto: no other jet pTT>15 GeV and |>15 GeV and ||<4.9|<4.9
20<p20<pTT<60 GeV<60 GeV60<p60<pTT<120 GeV <120 GeV
Fr IFr I
July 2002 Ringberg Calibration Workshop - Irene Vichou 25
In situ Z+jetIn situ Z+jet
Impact of cuts assessed also with full simulationImpact of cuts assessed also with full simulation
Example is the jet veto capability with comparison betweenExample is the jet veto capability with comparison betweenfast and full simulationfast and full simulation
Fraction on events with >1 jet is reconstructed wrt to the pFraction on events with >1 jet is reconstructed wrt to the pTT threshold applied threshold applied
Cone Cone R = 0.7R = 0.7
adjustment of padjustment of pTT threshold to take into account threshold to take into account the EM scale calibration in full simthe EM scale calibration in full sim
Good agreement between fast and full sim in jet veto efficiency!Good agreement between fast and full sim in jet veto efficiency!
July 2002 Ringberg Calibration Workshop - Irene Vichou 26
In situ Z+jetIn situ Z+jet
20-60 GeV 60-120 GeV >120 GeV
Barrel(… 1.7)
340K9K
180K4K
44K1K
End-cap(… 3.2)
190K5K
90K1.5K
16K0.5K
Forward(… 4.9)
30K 10K 1K
Rates after cuts are adequate for statistical sensitivity of 1%Rates after cuts are adequate for statistical sensitivity of 1%for pfor pTT<200 GeV. <200 GeV.
Nr of b-jets assuming 50% tagging efficiency hopefullyNr of b-jets assuming 50% tagging efficiency hopefullysufficient to constrain b-jet scale to % level. sufficient to constrain b-jet scale to % level.
+jet sample offers 30 times more events but more tend to bias+jet sample offers 30 times more events but more tend to biasin selecting back-to-back eventsin selecting back-to-back events
Number of jets after cuts for 10 fbNumber of jets after cuts for 10 fb-1-1 integrated luminosity in integrated luminosity in calorimeter and pcalorimeter and pT T regionsregions
Nr of Nr of b-jetsb-jets90% purity90% purity
NB overall jet sample content:NB overall jet sample content: 28% gluon , 54% light quark, 12% c and 6% b-jets28% gluon , 54% light quark, 12% c and 6% b-jets
July 2002 Ringberg Calibration Workshop - Irene Vichou 27
In situ Z+jetIn situ Z+jet
Comparison between full and fast simulationComparison between full and fast simulation1 jet pT>20 GeV <3.2
>3.06 rad >3.06 radLoose jet veto
>3.06 radTight jet veto
0.0530.106
0.0400.091
0.0390.089
0.0330.076
0.0570.089
0.037 0.027 0.0240.050
Fast:Fast:
Full:Full:
all jetsall jetsb-jetsb-jets
Average Fractional ImbalanceAverage Fractional Imbalancein three pin three pT T ranges ranges
ppTT>40 GeV and |>40 GeV and ||<3.2|<3.2
July 2002 Ringberg Calibration Workshop - Irene Vichou 28
In situ Z+jetIn situ Z+jet
20-60 60-120 >1200.049 0.015 0.004
GeVGeV
Accuracy of the calibration in this sampleAccuracy of the calibration in this sample
Reconstructed jet pReconstructed jet pTT rescaled to balance the Z p rescaled to balance the Z pTT wrt to the original parton pwrt to the original parton pTT
Fractional imbalanceFractional imbalanceof the two pof the two pTT’s’s
Systematics and effect on fractional imbalanceSystematics and effect on fractional imbalance
Uncertainties in modelling ISR:Uncertainties in modelling ISR:1.5 factor in1.5 factor inQCD QCD ±±1.5%(0.3%) at low (high) p1.5%(0.3%) at low (high) pTT
Topology uncertainties on back-to-back:Topology uncertainties on back-to-back: cut tightening by 0.06 rad cut tightening by 0.06 rad -0.7% to -0.1% -0.7% to -0.1%
Cone enlarging to Cone enlarging to R=0.7 R=0.7 0.3% 0.3%
July 2002 Ringberg Calibration Workshop - Irene Vichou 29
Z+jet sampleZ+jet sample
Reconstructed energy with a weighting methodReconstructed energy with a weighting method
The aThe aii’s are derived using the Z as a reference’s are derived using the Z as a reference
Parametrise the aParametrise the aii’s as functions of E’s as functions of ETTjetjet
E f a ET
rec
i cell jet ( , ) / cosh( )
Non-lineraitiesNon-lineraitieswrt to the partonwrt to the parton
July 2002 Ringberg Calibration Workshop - Irene Vichou 30
Z+jet resolutionsZ+jet resolutions
bisectorbisector
ppTT jet jet
ppTT Z Z
KKTT
KKTT
Bisector methodBisector method
sensitive to ISR and detector sensitive to ISR and detector resolutionresolutionsensitive to ISRsensitive to ISR
So, resolution due to jet recSo, resolution due to jet rec D 2 2
July 2002 Ringberg Calibration Workshop - Irene Vichou 31
Z+jet resolutionsZ+jet resolutions
Difference in resolution obtained with the bisector andDifference in resolution obtained with the bisector andthe ‘classic’/true definition.the ‘classic’/true definition.
July 2002 Ringberg Calibration Workshop - Irene Vichou 32
In situ jet calibrationIn situ jet calibration
The 1% jet calibration uncertainty is difficult but The 1% jet calibration uncertainty is difficult but it seems not impossible with the current studiesit seems not impossible with the current studies
The two types of in situ samples are somewhat The two types of in situ samples are somewhat complementary complementary
W jj will provide very accurately the E-scale at intermediate
energies but systematics rise at the low and high end of spectrum
Z to jet balance can reach higher energies and || (even the forward) systematics are very high at low energies, but above 60 GeV
the 1% requirement seems possible can offer separate calibration of b-jets can transfer calibration across calorimeters can check the linearity of calorimeter (TeV values at the
endcap and forward regions)
July 2002 Ringberg Calibration Workshop - Irene Vichou 33
Finally...Finally...
Test beam/ single particles and/or electronics Test beam/ single particles and/or electronics calibration to set the cell at the EM scalecalibration to set the cell at the EM scale
transfer to modules not tested where needed
Assign a detector Monte-Carlo that describes Assign a detector Monte-Carlo that describes well the observed with particleswell the observed with particles
remember the high E end..
Work on the machinery for jet/hadrons Work on the machinery for jet/hadrons reconstruction/calibration (even if not final reconstruction/calibration (even if not final choices) and keep flexiblechoices) and keep flexible
no best algorithm or best calibration scheme NOW be ready to try any useful data
In situ calibration cross-check at LHC start (only In situ calibration cross-check at LHC start (only feasibility studies and predictions now)feasibility studies and predictions now)
in situ intercalibration also