How to clean jets

We describe here the recommended procedure how to clean jets from detector effects, beam background and cosmics. We classify jets in three categories: The bad, the ugly and the good. Our experience is that unless one triggers on high-pt jets and high MET events, looser bad cleaning is enough.

Discriminating variables

The bad jet definition is based on the following variables:

  • emf (aka emfrac): electromagnetic fraction
  • fmax (aka fracSamplingMax): maximum energy fraction in one calorimeter layer
  • HECf: energy fraction in the HEC
  • LArQ (aka LArQuality): jet quality is the fraction of energy corresponding to LAr cells with a cell Q-factor greater than 4000. The cell Q-factor measures the difference between the measured pulse shape ($a_{i}^{meas}$) and the predicted pulse shape ($a_{i}^{pred}$) that is used to reconstruct the cell energy. It is computed as $\sum\limits_{samples}^{} {(a_{i}^{meas}-a_{i}^{pred})^{2}} $ and it is stored as 16-bit integer.
  • HECq: same as the LArQuality except calculated only with the HEC
  • NegativeE: negative energy in the jet
  • t (aka Timing): jet time computed as the energy squared cells mean time
  • AverageLArQF : jet quality computed as the energy squared cells mean quality
  • LArQmean: normalized jet quality computed as the energy squared cells mean quality (qmean=AverageLArQF/65335)
  • n90: minimum number of cells containing at least 90% of the jet energy
  • eta: is the eta at the emscale (aka emscale_eta)
  • chf: is the jet charged fraction. It is the ratio of the sum pt of tracks (sumPtTrk) associated to the jets divided by the CALIBRATED jet pt. In rel15 data, sumPtTrk can be computed from AOD using the JetVertex AssociationTool. In rel 16 data, this variable is stored as a jet moment called sumPtTrk.
How to access the cleaning variables in Athena from AOD

Bad jets


Bad jets are jets not associated to real energy deposits in the calorimeters. They arise from various sources, ranging from hardware problems (HEC spike, EM coherent,...), LHC beam conditions, and cosmic-ray showers. Several sets of cuts are provided to identify such jets:

  • Looser: new set of cuts with a reduced inefficiency. Should be the default for most Atlas analysis.
  • Loose: unchanged wrt to release 16 2010 recommendation (Should not be used anymore unless you encounter issues with the looser criteria. In this case, please contact also the jet/etmiss group)
  • Medium: unchanged wrt to release 16 2010 recommendation
  • Tight: New criteria with improved rejection (should only be used for performance or systematics studies)


More information can be found in this talk

Technical implementation:

Will be available soon


Will be available soon


  • Cleaning should be applied to both data and MC. This is new wrt previous recommendation
  • Please, note that the jet cleaning cuts can reject electrons, photons or muons. Make sure that in your analysis lepton-ID takes priority over jet-ID.

"Ugly" jets:


Good jets

Good jets are jets which are not bad neither ugly.

How to clean MET

Remove events with at least one bad jet with pt_calibrated>=20GeV (no eta cut). Until further notice calibrated=EM+JES. The cleaning should be applied to both data and MC. For most Atlas physics analysis, the looser criteria is recommended. The per event inefficiency should be of the order of few per mill. If it is larger, please contact the jet/etmiss group.

LAr Hole

Since the hole in the LAr calorimeter affects the jet and etmiss reconstruction, several strategies to deal with this problem are proposed. More information can be found here.

Do nothing

Jet and MET performance are degraded but it is reproduced by mc11. Analysis weakly affected by this problem would probably prefer to not apply any veto.

Simple veto

A simple and robust strategy (but not optimal in term of signal acceptance) is to veto events if a calorimeter jets fall in the vicinity of LAR hole: This treatment should only be applied for data in period E to H and also to the corresponding MC samples with the LAr hole emulated

The veto should be applied if the jet satisfy -0.1<eta<1.5 -0.1 and -0.9<phi<-0.5 (EMJES-corrected eta and phi should be used) since it was demonstrated that only these jets are strongly affected by the LAr hole. The veto should be applied to jets with a EMJES and BCH_CORR_JET corrected pt greather than a pt threshold:

* jet_emjes_pt*(1-BCH_CORR_CELL)/(1-BCH_CORR_JET) > threshold

A snippet to apply the simple veto can be found in LArHole_v2.00.hpp More information can be found in this talk

The exact implementation can be adjusted upon need and in particular: * The threshold can be adjusted upon needs by balancing the signal acceptance loss and the increase of background for a given analysis * Some analysis would prefer to apply a jet veto instead of the veto

Smart veto

Remove events only if the estimated energy missed in the LAr hole contributes significantly to MET using the variable: isBad2.png

Events are rejected if Ep>E_min and Ep>frac*MET where E_min and frac are two parameters and MET is the transverse missing energy.

The default values (as used in the SuSy group, see fake estimator twiki) are E_min=10GeV and frac=0.1. The exact implementation can be adjusted upon need and in particular the parameter values.

A snippet to apply the smart veto can be found in LArHole_v2.00.hpp More information can be found in this talk

BCH_CORR_XXX variables:

Two tools exist to estimate the energy in calorimeter dead cells:
  • Cell level correction using neighboring cells assuming the same energy density
    • More details can be found here
    • BCH_CORR_CELL= fraction of the jet energy coming from this correction
    • Apply at the cell level before jet and met reconstruction
    • Use for dead tile modules
  • Jet level correction using jet shape from MC
    • More details can be found here
    • BCH_CORR_JET= fraction of the jet energy coming from this correction
    • Can only be computed after jet reconstruction
    • Computed for all dead cells: dead Tile module, LAr hole,...

%RESPONSIBLE% MakovecNikola
%REVIEW% Never reviewed

Topic attachments
I Attachment History Action Size Date Who Comment
JPEGjpg Ep_formula.jpg r1 manage 2.7 K 2011-12-14 - 18:32 MakovecNikola  
PNGpng Ep_formula.png r1 manage 3.5 K 2011-12-14 - 18:42 MakovecNikola  
PNGpng isBad.png r1 manage 18.5 K 2010-09-29 - 21:33 MakovecNikola  
PNGpng isBad_rel17.png r2 r1 manage 31.8 K 2011-12-14 - 19:11 MakovecNikola  
Edit | Attach | Watch | Print version | History: r19 < r18 < r17 < r16 < r15 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r19 - 2011-12-17 - TancrediCarli
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback