Update On Jet Energy Corrections

1
Update On Jet Energy Corrections

• Robert M. Harris
• Fermilab
• Physics / Trigger Week
• October 26, 2007

2
Outline

• MC Jet Correction
• Factorization Plan and Status
• Developments in Factorized Corrections
• Zero Suppression Correction
• Response vs. PT for g jet events.
• Jet Correction Methodologies
• Developments in Monolithic Corrections
• Quark Jet Correction Studies
• Global Fit Investigation
• Conclusions

Discussing status and all updates shown to
JetMET And JESTF since September 15-21 CMS
Week For discussion of developments in the top
quark group please see talk by J.
DHondt yesterday and CMS-AN 2007 / 029
3
MC Jet Corrections for CMSSW_1_5_2Kostas
Kousouris (JetMET Meet Oct. 4)

• Standard corrections from MC truth released
  October 6
• For Midpoint Cone R0.5, 0.7, Iterative Cone
  R0.5, 0.7, KT D0.6
• Official support expected soon for KT D0.4,
  Seedless Cone R0.5, 0.7
• Un-official version of these corrections have
  been supplied to the experts.
• Similar correction as Spring07 in barrel and
  endcap
• Smaller correction than Spring07 in forward which
  has improved response.
• Asymmetric HF response vs. h at low PT is not
  corrected.

4
Factorization Plan

• Factorized jet corrections will replace MC Jet
  Corrections in CMSSW

Postdocs in JES Task Force Jochen Cammin
(Rochester) L5 Flavour Corrections Selda Esen
(Brown) L4 EMF Corrections Kostas Kousouris
(Fermilab) L2 h L3 PT corrections

• Breaks correction into natural pieces.
• Found in stages of time precision
• A. MC truth (now)
• B. Data-driven techniques using MC (soon)
• C. Data-driven using actual in-situ data (08)
• D. Eventually back to MC tuned on data.
• Factorized corrections are work of the Jet Energy
  Scale Task Force in JetMET
• Especially the hard working postdocs !

5
Factorization Status

• L1 (offset) 1st pass Zero Suppression correction
  done including code to use it.
• Need pile-up component of correction next.
• L2 (h) MC truth h dependence correction almost
  done.
• No correction code yet.
• 1st study of dijet balance for data-driven phase
  of correction done.
• L3 (pT) MC truth pT dependence correction almost
  done.
• No correction code yet.
• Study of g jet balance for data-driven phase of
  correction begun.
• L4 (EMF) MC truth EMF dependence factorization
  almost done
• Study showing feasibility for limited range of pT
  and h completed, and work ongoing.
• L5 (flavour) MC truth flavour dependence study
  ongoing.
• L6 (UE) No developments
• L7 (parton) No developments but we have a
  volunteer.

6
Zero Suppression Correction Olga Kodolova
(JetMET meet Oct. 4)

• Adds back jet energy lost due to calorimeter
  thresholds that suppress noise
• Significant correction 25 for 20 GeV CaloJet in
  Barrel
• Correction decreases with ET, and has complex h
  dependence ( ? hlt2, ? hgt2)
• Correction restores jets to there measured vales
  w/o thresholds.
• 1st Zero Suppression Correction completed and
  available now in CMSSW
• Special CMSSW_1_3_4 sample w/o HCAL cell level
  ZSP or Ecal Selective Readout.

Fractional Correction vs. ET and h
Jet Response Closure Test vs. ET
Jets After Corrections Jets w/o Thresholds
Jets Before Corrections
7
Jet Response vs. PT from g Jet EventsAnwar
Bhatti (JES Task Force Oct. 2)

• g Jet PT balance or MPF provide jet response
  vs. PT in barrel
• High rate calibration process, but lots of subtle
  issues need study.
• MPF response a little higher than PT bal. at most
  PT (radiation?).
• Response to GenJet a little larger than PT bal.
  (particle vs. parton ?).
• PT bal. response a little higher than dijet
  system response (q vs. g?).
• Also prior work by O. Kodolova et al and J. Cammin

g
Jet
Jet Response
Jet Response
Photon PT (GeV)
Photon PT (GeV)
8
Jet Correction MethodologiesKostas Kousouris
(JetMET Oct. 18)

• Four methodologies explored for finding
  correction vs. pT in barrel.
• Binning in GenJet or CaloJet pT, measuring
  response with ratio or difference.
• Using QCD sample w/o weights (flat spectrum)
  and with weights (QCD spectrum).
• MC Jet Method bin in GenJet pT and measure
  CaloJet pT / GenJet pT
• Answers What is response of GenJet? independent
  of input spectrum.
• CaloJet Method bin in CaloJet pT and measure
  CaloJet pT - GenJet pT
• Answers What GenJet pT gave my CaloJet pT?
  which depends on input spectrum.
• CaloJet method closes better vs corrected CaloJet
  pT.
• Similar to work of S. Esen, G. Landsberg and P.
  Schieferdecker. Discussion ongoing.

Spectrum Dependence
Correction from Methods
CaloJet Closure Test
Correction
Correction
MC Jet Method
No Weights
MC Jet Method
CaloJet Method
No Weights
CaloJet Method
Corrected CaloJet PT (GeV)
CaloJet PT (GeV)
CaloJet PT (GeV)
9
Quark Jet Correction Studies

• Quark jet corrections
• Loïc Quertenmont, Dorian Kcira, Giacomo
  Bruno (JetMET Oct. 4)
• MC Jet overcorrects W from top
• Because MC Jet for mixture of q g ?
• Because of closure issues in MC jet ?
• Quark jet correction found from Z sample
• Private monolithic correction.
• Reconstructs correct W mass.

Quark Jet Corr.
No Corr.
MC Jet Corr.

• g jet corrections applied to qqH
• Efe Yazgan and Alexandre Nikitenko (JetMET
  Oct. 4)
• Jet corrections derived from g jet sample in
  CMSSW ala ORCA method.
• Overcorrects qqH jets at high h because of large
  amounts of gluons at high h in the g jet sample.

g Jet Corr.
MC Jet Corr.
No Corr.
Maiko Takahashi
10
Global Fit for Hcal/Jet CalibrationChristian
Autermann (JetMET Oct. 18)

• Trial exploration of framework to combine all
  factors and perform global fit.
• Work of C. Autermann, C. Sander, P. Schleper, H.
  Stadie, R. Wolf at Hamburg.
• Test of the tools calibration of jets and of all
  2048 CaloTowers.
• Use g jet sample and form c2
• Between the g energy and corrected jet energy in
  110K back-to-back clean events.
• Between the track momentum and corrected tower
  energy in 1.5K gjet events
• Minimize the c2 to find the correction function
  for each CaloTower and Jet.

Jet Response
Jet Resolution
Tower K-factor per event
11
Summary Note

• MC Jet corrections available for 1_5_2
• Factorized corrections will replace it soon.
• 1st Zero Suppression correction available.
• g jet studies begun for correction vs. jet pT
  in barrel.
• Jet correction methodologies for the factorized
  correction being actively studied and discussed
• Work on monolithic corrections continues to teach
  us valuable lessons.
• We are working on completing a draft of our JES
  note for December CMS week.
• 90 done 28 pages, 24 plots.
• Needs little more work and editing.