Jet efficiencies experience at D

1
Jet efficiencies experience at DØ

• Jet ID
• Choose energy to be clustered into jets
• Choose clustering algorithms and parameters
• Choose sets of selection criteria
• Prepare optimize jet ID tools
• Choose a basic set of cuts
• Fancier optional cuts (e.g. jet-vertex matching)
• Measure reconstruction and selection
  efficiencies
• For this talk Ill concentrate on measuring
  reconstruction and selection efficiencies,
• Known in DØ as jet certification (4)

• Amnon HarelUniversity of Rochester

2
Jet certification timeline
Reviewingbasicassump-tions
First fullcertification
1st use of conclusions
Several early measurements
2nd fullcertification
First (preliminary) results with jets
This data published with methodologiesdescribed
in this talk
3
The theme

• Concentrating aspects relevant to LHC.
• The recurring theme
• Limited resolution makes significant bias
  possible
• large (exponential) slope dont help either

4
An example

• The natural way to measure jet ID efficiency is
  to take your data sample, and bin isGood in the
  jet pT. But theres a bias layer fractions (EMF
  and CHF) are correlated to the energy scale and
  resolution.

These are the same di-jet events, using a fully
efficient trigger. Red EMF of the jet with
130(unbiased).
The bias increases the high EMF inefficiency by
100-400
5
Jet Reconstruction Eff.

• The point of a reconstruction eff. measurement is
  that the probe jet may not exist!
• Need a tag probe procedure to identify events
  where were sure the jet should exist, even if
  the detector says no.
• Need a tag track to find the ? of the jet
• Must trust MC to model how this criteria effects
  jet efficiencies ?

• Possible samples
• Zjets good sE - fairly pure sideband
  subtractions possible low statistics
• ?jets good sE impure high statistics
• backgrounds have unusual JES (neither ? nor
  generic jets!)
• di jets bad sE very pure high statistics
• We will see that having a good tag sE is an
  illusion
• ? Di jets are by far the best sample for jet
  reconstruction efficiency measurements

6
Hemisphere imbalance

• Contrary to popular belieftwo object production
  is NOT balanced in pT
• That is only true on average
• The spread is 4-5 GeV for any axis (at the
  Tevatron)
• update of hep-ph/9808467 with hep-ex/0412050 by
  M.Begel
• Theory due to the pT carried by the partons
  (neglected in naïve PDF picture)
• Experiment beam remnants carry significant pT
  down the beam pipe
• Hence even with infinite precision on the tag
  energy, the resolution of the predicted probe pT
  will be no better than 4 GeV, which is enough for
  an interesting (?) bias
• ?Good sE on the probe is an illusion
• Is this of any numerical importance?

7
Jet reco. Naïve tag probe

• The natural way to measure jet reco. efficiency
  is to take your tag probe sample, and bin
  events in the tag jet pT. Then

In terms of passing or failing jet reco.
In terms of of tags and of probes
Only one problem This gives completely the wrong
results Do you see the mistake?
8
The 1st red flag
MC
MC
No jet
With jet
Data vs. MCdiscrepancies ?Irrelevant
No jet
With jet
Data
Data
Naïve method assumesthese pTs are the same!
9
Toy MC
10
Toy MC - II
11
Jet ID Eff.
Tag probe jet exists Probe jet ID

• Jet ID is easier can tag dijet events and probe
  jet ID
• No need for a track
• Why bother with tag probe?
• To get pT dependence of jet ID eff. despite the
  jet ID E bias shown on slide 4

Tag jet
Possible samples Same discussion as for jet
reco. eff. di jets are best
12
Conclusions

• Limited resolution makes significant bias
  possible
• Imposing pT bins ? bias
• Requiring that a jet is or is not reconstructed ?
  different biases
• Requiring that a jet pass jet ID ? bias
• Contrary to popular belief two object
  production is NOT balanced in pT
• That is only true on average
• The RMS is 5-6 GeV at the LHC (4-5GeV at the
  Tevatron)
• It is not straightforward to measure pT
  dependent jet reco. eff.
• Can not correct the results of the naïve
  approach
• Difficult to fit from observable spectra
• Might be possible by unbiasing (and smoothing)
  the pT dependence of the pass and fail samples
  before combining them into an efficiency
• Probably dominated by energy resolution anyways ?
  just get that right ?
• Jet ID eff. are fairly easy to measure
• Tag probe needed to avoid jet ID ? E bias

13

• Back Up Slides

14
?
b