1. Jet Calibration at CDF Sandro De Cecco Sandro De Cecco
18 mars Paris
Sandro De Cecco

2. Jet Production Measurements
Hadronic showers
EM showers
Unfold measurements to hadron level  Correct for efficiency, smearing
Correct theory (pQCD) for non-perturbative effects  Underlying Event, Fragmentation
Q ~ QCD
Q >> QCD
Well defined jet algorithm required  At calorimeter, hadron and parton levels
18 mars Paris
Sandro De Cecco

3. CDF calorimeter Wall Had Plug Central Central and Wall (||<1.2):
Scintillating tile with lead (iron) as absorber material in EM (HAD) section
Coarse granularity: ~800 towers
Non-compensating
non-linear response to hadrons
Rather thin: 4 interaction lengths
Nearly no noise
Resolutions:
EM energies: /E=13.5% / √E  1.5%
HAD energies: /E=50% / √E  3%
Plug (1.2<||<3.6):
Similar technology to central
Resolution:
EM energies: /E=16 % / √E  1%
HAD energies: /E=80 % / √E  5%
Thicker: 7 interaction lengths
Plug
Central
18 mars Paris
Sandro De Cecco

4. Cone Jet Algorithms and pQCD
Iterative cone algorithms
Staring from seeds, iteratively cluster particles in cones of radius RCONE and look for stable cones (pT-weighted centroid)
Infrared and Collinear Safety
Fixed order pQCD contains not fully cancelled infrared divergences
Inclusive jet cross section affected at NNLO
Tevatron Run II Cone Algorithm: Midpoint
Uses midpoints between pairs of proto-jets as additional seeds  Infrared and collinear safety restored
Merging/Splitting
Emulated in NLO pQCD calculation by merging 2 partons only if they are within R’ = RCONE  RSEP of each other
Arbitrary parameter RSEP: prescription RSEP = 1.3 (based on parton level approximate arguments)
below threshold
(no jets)
above threshold
(1 jet)
18 mars Paris
Sandro De Cecco

5. kT Algorithm Inclusive kT algorithm
Merging pairs of nearby particles in order of increasing relative pT
D parameter controls merging termination and characterizes size of resulting jets
pT classification inspired by pQCD gluon emissions
Infrared and Collinear safe to all orders in pQCD
No merging/splitting
No RSEP issue comparing to pQCD
jet
18 mars Paris
Sandro De Cecco

6. CDF Jet calibration overview
Calibrate EM and HAD calorimeters in situ
Reconstruct jets (JetClu cone algorithm):PTraw
Correct jets in plug calorimeter w.r.t. central “relative corrections”: frel
use di-jet data (versus )
Correct for Multiple pp Interactions : UEM
Correct measured jets back to particle level jets: tune MC simulation: fabs
Response of calorimeter to single particles
Fragmentation: Pt spectra in data
Correct for Underlying Event: UE
Correct particle jet back to parton: OC
and … Splash-out systematics
Systematic error associated with each step
18 mars Paris
Sandro De Cecco

7. Two different approaches
CDF and DØ use very different approaches
Documented in
CDF Run 2: hep-ex/ (accepted by NIM)
DØ Run 1: NIM A424: (1999)
DØ Run 2:
Main difference:
CDF uses test beam and single particles measured in-situ to understand absolute response of single particles
deduce jet response using simulation
Cross check with calibration processes like photon-jet data
DØ uses photon-jet data to measure absolute response
Extra correction for “showering” necessary
Other differences:
CDF corrects separately for underlying event, multiple interactions, out-of-cone energy
DØ includes all these effects into one correction factor
18 mars Paris
Sandro De Cecco

8. Overview: CDF and DØ CDF calibrates PT DØ calibrates Energy
PTcorr: calibrated jet PT
PTraw: raw jet PT
F: eta-dependent correction
R: absolute response
MI: multiple interactions
DØ calibrates Energy
Ecorr: calibrated jet E
Eraw: raw jet E
F: eta-dependent correction
R: absolute response
O: offset energy
includes MI, noise, UE
S: showering corrections
Systematic error associated with each step
additional corrections to get to parton energy
18 mars Paris
Sandro De Cecco

9. In Situ Calorimeter Calibration I
Minimum Ionising Particle (MIP):
J/Ψ and W muons
peak in HAD calo: ≈2 GeV
Peak in EM calo: ≈300 MeV
Check time stability and run1 versus run2
Applicable where muon coverage: <1.4
E/p of electrons:
p calibrated on J/Ψ mass
Time dependence of E/p checked
18 mars Paris
Sandro De Cecco

10. In Situ Calorimeter Calibration II
Z→ ee peak:
Set absolute EM scale in central and plug
Compare data and MC: mean and resolution
Applied in Central and Plug
MinBias events:
Occupancy above some threshold: e.g. 500 MeV
Time stability
Phi dependent calibrations: resolution
18 mars Paris
Sandro De Cecco

11. Relative Corrections - dijets
18 mars Paris
Sandro De Cecco

12. Relative Corrections Mapping out cracks and response of calorimeter
Central at ~1 by definition
D0:
Response similar in central and forward
Two rather large cracks
CDF:
Response of forward better than of central
Three smaller cracks
Difficulties:
depends on ET
Can be (most often is initially) different for data and MC
18 mars Paris
Sandro De Cecco

13. Detector to Particle Level
Do not use data since no high statistics calibration processes at high Et>100 GeV
Extracted from MC  MC needs to
Simulate accurately the response of detector to single particles (pions, protons, neutrons, etc.): CALORIMETER SIMULATION
Describe particle spectra and densities at all jet Et: FRAGMENTATION
Measure fragmentation and single particle response in data and tune MC to describe it
Use MC to determine correction function to go from observed to “true”/most likely Et:
Etrue=f ( Eobs, , conesize)
18 mars Paris
Sandro De Cecco

14. Single Particle Response
Low Pt (1-10 GeV) in situ calibration:
Select “isolated” tracks and measure energy in tower behind them
Dedicated trigger
Perform average BG subtraction
Tune GFlash to describe E/p distributions at eack p (use π/p/K average mixture in MC)
High Pt (>8 GeV) uses test beam:
Could try τ-leptons
Non-linearity: response drops by 30% between 10 and 1 GeV
18 mars Paris
Sandro De Cecco

15. Fragmentation
Due to non-linearity of CDF calorimeter big difference between e.g.
1 10 GeV pion
10 1 GeV pions
Measure number of and Pt spectra of particles in jets at different Et values as function of track Pt:
Requires understanding track efficiency inside jets
Ideally done for each particle type (π, p, K)
E.g. difference in fragmentation between
Herwig and Pythia may result in different response
18 mars Paris
Sandro De Cecco

16. Absolute Correction from MC
Wanted:
Most likely true Et value for given measured Et value
BUT cannot be obtained universally for all analyses since it depends on Et spectrum:
E.g. most likely value in falling spectrum dominated by smearing from lower Et bins
Different for flat Et spectrum (e.g. top or new resonance)
CDF:
Provide standard “generic” jet corrections using flat Pt spectrum
Individual analyses determine their “specific” residual corrections themselves from their MC
18 mars Paris
Sandro De Cecco

17. Flat vs. QCD Spectra For both spectra With a Flat spectrum.
Avg
For both spectra
There is an average PT shift of hadron jets to calorimeter jets.
With a Flat spectrum.
After accounting for the average shift there are roughly as many low PT as high PT jets “smearing” into the calorimeter PT bin.
With a QCD spectrum
After accounting for the average shift, there are significantly more low PT jets than high PT jets “smearing” into the calorimeter PT bins.
The QCD spectrum correction is therefore significantly lower.
Hadron Jet PT
Calorimeter Jet PT
Avg
Hadron Jet PT
Calorimeter Jet PT
18 mars Paris
Sandro De Cecco

18. Absolute Corrections Use MC with “flat” Et distribution
Separately for each cone size: 0.4, 0.7 and 1.0
Correction factor decreases with Et due to non-linearity of calorimeter, e.g. cone 0.4:
50 GeV: ≈25%
500 GeV: ≈15%
Systematic errors due to
Test beam precision
γ-Jet and Z-jet balancing agreement between data and simulation after correction 
18 mars Paris
Sandro De Cecco

19. Photon-Jet PT balance CDF pTjet/pT-1
Agreement within 3% but differences in distributions
Data, Pythia and Herwig all a little different
These are physics effects!
18 mars Paris
Sandro De Cecco

20. Z-jet PT balance CDF pTjet/pTZ-1 pTjet/pTZ-1
Better agreement of data and MC than in photon-jet data
In progress of understanding this better together with Herwig and Pythia authors – different L scale
18 mars Paris
Sandro De Cecco

21. Multiple pp Interactions
Extra pp interactions will increase the jet Et values of primary hard interaction
subtract off average energy in cone per interaction:
Number of interactions = Number of observed vertices
Random cone in MinBias data: Et versus Nvtx
E.g. ≈0.8 GeV per vertex for cone 0.7
18 mars Paris
Sandro De Cecco

22. Corrections from Particle Jet to Parton
Underlying event (UE) and Out-of-cone (OOC) energy
Only used if parton energy is wanted
Requires MC modeling of UE and OOC
Differences are taken as systematic uncertainty
18 mars Paris
Sandro De Cecco

23. Underlying Event Underlying event definition:
“beam-beam remnants”: energy from interaction of spectator partons
“Initial state radiation”: energy radiated off hard process before main interaction
Not wanted when e.g. measuring the top quark mass
Can be estimated using Monte Carlo
Measurements led to tuning of MC generators: PYTHIA, Herwig+Jimmy
18 mars Paris
Sandro De Cecco

24. Measuring the Underlying Event
Leading Jet Direction
“Transverse” region very sensitive to the “underlying event”!
Many studies exist about underlying event:
Checkout talks by Rick Field/U. of Florida
At LHC we will need to measure it:
Expect it to be much harder than at Tevatron
18 mars Paris
Sandro De Cecco

25. Out of Cone Energy (OOC)
Original parton energy that escapes the cone
E.g. due to gluon radiation
Jet shape in MC must describe data:
measure energy flow in annuli around jet
Differences between data and MC
Lead to rather large systematic uncertainty
Herwig
Pythia
0.06
0.04
Data
0.02
(OOCData-OOCMC)/pT
0.0
0.02
0.04
Herwig
Pythia
0.06 40 80
120 PT (GeV)
18 mars Paris
Sandro De Cecco

26. Energy Flow Inside Jets
Jet shapes governed by multi-gluon emission from primary parton
Test of parton shower models
Sensitive to underlying event structure
Sensitive to quark and gluon mixture in the final state
Phys. Rev. D 71, (2005)
(1-)
37 < pT < 380 GeV/c
18 mars Paris
Sandro De Cecco

27. Underlying Event & Hadronization correction
jet
Calorimeter level
1.2
Hadron level
Parton level
18 mars Paris
Sandro De Cecco

28. Jet Energy Scale Uncertainties
CDF and DØ achieve similar uncertainties after following very different paths before
18 mars Paris
Sandro De Cecco

29. Other calibration samples
18 mars Paris
Sandro De Cecco

30. In-situ Measurement of JES
Additionally, use Wjj mass resonance (Mjj) to measure the jet energy scale (JES) uncertainty
2D fit of the invariant mass of the non-b-jets
and the top mass:
JES M(jj) GeV/c2
Mjj
Measurement of JES scales directly with data statistics
18 mars Paris
Sandro De Cecco

31. Wjj Calibration in Top Events
CDF (1 fb-1): JES = 0.99 ± 0.02
DØ (0.3 fb-1): JES = 0.99 ± 0.03
Fit for ratio of JES in data to JES in MC
Constrain JES to 2% using 166 events
At LHC will have 45,000 top events/month!
18 mars Paris
Sandro De Cecco
(jet)

32. Z->bb Zbb decay mode:
Suppresses QCD background more than signal
Difficult to trigger
CDF uses secondary vertex trigger
D0 uses semi-leptonic decays collected by muon trigger
Use this to measure difference between data and MC JES, e.g. DØ:
Data:
=81.0 +/- 2.2
=10.7 +/- 2.1
MC:
=83.3
=13.0
18 mars Paris
Sandro De Cecco

33. Using jets in analysis
18 mars Paris
Sandro De Cecco

34. W/Z bosons + Jets (HF) production
The study of W / Z / g + Jets production is very relevant:
to QCD:
High q2 ( ~Mboson) interactions, perturbative theory
Less leading diagrams ( wrt pure Jets production)
Test Monte Carlo generators (ALPGEN…)
Probe for protons PDF (in particular W/Z + H.F. )
for many high pT analysis where W/Z+HF is a main background:
Top quark cross section and mass
Single Top cross section
Search for low mass Higgs boson
several SUSY searches
18 mars Paris
Sandro De Cecco

35. W+jet(s) Production ..+ PS Background top, Higgs, SUSY…
Stringent test of pQCD predictions
Test Ground for ME+PS techniques
(Special matching  MLM, CKKW to avoid
double counting on ME+PS interface)
..+ PS
18 mars Paris
Sandro De Cecco
Affected by cutoff and soft radiation

36. Soft radiation in Z+jet(s)
Implementation of proper modeling
of UE still needed in W/Z+Jet(s)
Monte Carlo….very important
LHC will use "extra jets" veto in
Higgs analyses to reduce QCD bckg.
18 mars Paris
Sandro De Cecco

37. Diphoton Production g g g g g g q q q g q g g LO (Pythia) off
Phys. Rev. Lett. 95, (2005)
Diphoton Production
relevant for future H-> gg
searches at the LHC
(K-factors applied to LO pQCD MCs )
LO (Pythia) off g q q g g g q g q g g
NLO (DIPHOX) g g
18 mars Paris
Sandro De Cecco

38. Electrons and Jets Hadronic Calorimeter Energy
Electromagnetic Calorimeter Energy
Jets can look like electrons, e.g.:
photon conversions from 0’s: ~13% of photons convert (in CDF)
early showering charged pions
And there are lots of jets!!!
18 mars Paris
Sandro De Cecco

39. The Isolation Cut  candidates Non-isolated isolated Isolation cut:
Draw cone of size 0.4 around object
Sum up PT of objects inside cone
Use calorimeter or tracks
Typical cuts:
<10% x ET
<2-4 GeV
Isolation is very powerful for isolated leptons
E.g. from W’s, Z’s
Rejects background from leptons inside jets due to:
b-decays
photon conversions
pions/kaons that punch through or decay in flight
pions that shower only in EM calorimeter
This is a physics cut!
Efficiency depends on physics process
The more jet activity the less efficient
Depends on luminosity
Extra interactions due to pileup
 candidates
Non-isolated isolated
18 mars Paris
Sandro De Cecco

40. Jets faking Electrons Jets faking “loose” electrons Fake Rate (%)
Jets can pass electron ID cuts,
Mostly due to
early showering charged pions
Conversions:0ee+X
Semileptonic b-decays
Difficult to model in MC
Hard fragmentation
Detailed simulation of calorimeter and tracking volume
Measured in inclusive jet data at various ET thresholds
Prompt electron content negligible:
Njet~10 billion at 50 GeV!
Fake rate per jet:
Loose cuts: 5/10000
Tight cuts: 1/10000
Typical uncertainties 50%
Jets faking “loose” electrons
Fake Rate (%)
18 mars Paris
Sandro De Cecco

41. Feynman Talk at Coral Gables (December 1976)
1st transparency
Last transparency
“Feynman-Field
Jet Model”
18 mars Paris
Sandro De Cecco

42. Backup
18 mars Paris
Sandro De Cecco

43. Underlying Event Everything but the hard scattering process
Initial state soft radiations
Beam-beam remnants
Multiple Parton Interactions (MPI)
Studied in the transverse region
Leading jet sample
Back-to-back sample
18 mars Paris
Sandro De Cecco

44. New Underlying Event StudiesMB
ET(jet#2)/ET(jet#1) > 0.8
Back-to-Back
Df12 > 150o
Suppresses contribution
from additional hard radiation
Pythia Tune A describes the data
Herwig underestimates UE activity
Extended to 250 GeV jets
18 mars Paris
Sandro De Cecco

45. New Underlying Event Studies
MAX (MIN) for the largest (smallest) charged particle density in transverse region
MAX-MIN sensitive to remaining hard contribution
MIN specially sensitive to the remnant-remnant contribution
18 mars Paris
Sandro De Cecco

46. Studies on Jet Fragmentation
Jet shape dictated by multi-gluon emission form primary parton
Test of parton shower models and their implementations
Sensitive to quark/gluon final state mixture and run of strong coupling
Sensitive to underlying event structure in the final state
18 mars Paris
Sandro De Cecco

47. Jet shapes PYTHIA Tune A describes the data PRD. 71, 112002 (2005)
(enhanced ISR + MPI tuning)
PYTHIA default too narrow
MPI are important at low Pt
HERWIG too narrow at low Pt
We know how to model the UE at
2 TeV for QCD jet processes
18 mars Paris
Sandro De Cecco

48. Studies on Δf between jets
PRL 94, (2005) 
Studies on Δf between jets
LO in Df
NLO in Df
LO limited at hard (Mercedes Star)
and soft limits for third emission
NLO closer to data…however
soft gluon contributions are needed
PYTHIA (enhanced ISR) & Herwig
provide best description across the
different regions…
18 mars Paris
Sandro De Cecco