Official Isolation Working Points

Introduction

Official isolation working points are provided by the isolation forum for electrons, photons and muons. They can be accessed through the IsolationSelectionTool inside the IsolationSelection package.

The scale factors are provided by egamma and MCP groups, as documented here.

Current official working points for which SFs are provided

• The isolation at trigger level is (almost) as tight as in FCTight ISO WP, so any analysis using single electron trigger for signals between 27 and 60GeV where it applies with any other isolation WP are biased in this respect! You can see the isolation applied in this twiki
• Some working points make use of different variables/cone sizes (shown in red in the table above), and are therefore incompatible with lepton triggers that include isolation (see here).
• For the exact track or calo isolation cuts used in the efficiency-targeted working points can be obtained following these instructions.
• The efficiency-targeted working points (those not having "FC" in their name) have a bad background rejection at low pT (pT<20 GeV). For low pT leptons analyses, you should probably chose one of the FC (FixCut) working points.

Other lepton isolation WPs proposed to PA groups

• %W% These WPs might replace the "nominal" ISO WPs in future, depending on the feedback received from PA groups
• %W% SFs available for some ISO WPs, find descriptions below each table

• Preliminary SFs for these muon ISO Pflow WPs already obtained! if you want to use them, please contact IFF group.

• Helps to reduce a lot the fake lepton background (as the relative isolation cut on the isolation variable with a dependent pt cone size is really inefficient to reject the background)
• Such isolation WPs can reduce a lot the prompt lepton isolation efficiency in the boosted regions (when the leptons and the (b-)jets are close-by)!
• %W% Preliminary SFs for these muon ISO experimental WPs already obtained! if you want to use them, please contact IFF group. To get the SFs, the ISO WP name corresponds to HighMu ISO WPs; this will change in future.
• Do we really need cone dependent isolation variable for the low pt region?
• These WPs might replace the current ones, but more feedback from the PA groups (especially from the analyses targeting boosted scenarious) is needed before doing this swap.
• Try also some variations of the nominal and experimental ISO WPs with a cone 30 for electrons instead of cone 20 (better background rejection in the low pt region. Maybe use cone20 in the high pt regions as the prompt electron isolation efficiency for a cone20 ISO WPS is expected to be lower in the high pt region, in (signal) events with many jets).

• A version of the Rel20.7 FixedCutTrackCone40 ISO WPs will be studied in the HWW sub-group (targeting a publication in 2020); This WP might have a better performance than FixedCutTight _new (higher prompt electron efficiency for a better background rejection) => however from the studies performed by the IFF group, this affirmation seems be true only in the pt < 50 GeV region, therefore studies/feedback on the performance above 50 GeV are needed before removing FixedCutTight _new ISO WP. In addition, a cone 40 is expected to perform very bad in the context of a search with busy environments (and/or boosted topologies)
• FCTightTrackOnly _new is currently requested by few sub-groups targeting boosted topologies from EXOTICS group. A similar performance can be obtain with FCLoose WP in the intermediate pt region, and no isolation cut above ~100GeV. To recover the background rejection in the high pt region, the computation of the track isolation variable must be improved. Preliminary studies are shown here (we need manpower to fully cover this topic!)

Photons

For photons, the leakage corrections must be used together with this version of IsolationSelectionTool.

• FixedCutLoose should not be used for low pT photons (pT<25 GeV): since there is no offset in the cut, it becomes too tight at low pT

Procedure for requesting a new isolation working point

New isolation working points can be requested by analysis teams if the currently supported ones are not fulfilling their needs. A presentation in the isolation forum with the motivation of the additional working point is needed, please include the information in the following check-list with the justification of the proposal:

• Origin of your signal/background leptons (light flavour, heavy flavour, conversions) and pT range
• Have all the currently supported WPs being tried? How much worse they are (quantitatively) in terms of final sensitivity of the analysis?
• Are there more customers for your proposed WP?
• Do you plan to use this proposed WP for triggered-matched leptons? Does that introduce any limitations (see IsolationTrigger)?
• If using other variables beyond the standard WPs:
  • Justify (quantitatively) why current variables ones are not suitable
  • If requesting ptcone, justify (quantitatively) why ptvarcone is not suitable

List of older or intermediate test working points

WPs used for Rel20.7 studies and first Rel21 studies (starting with IsolationSelection-00-06-07, until august 2018)

Analyses are welcome to provide feedback on the working points. They can document the result of their investigation on this wiki page.

Leptons

Several "FixedCut" points (i.e. not using cut maps to get a target efficiency, but using a simple cut on the isolation) have been introduced. The Tight working point was supposed to be removed since nobody listed in in the EOYE survey initially, but finally one analysis is using it so it is keep for EOYE (but avoid to use it unless strongly justified: it will probably be removed after EOYE).

For electrons and photons, the new leakage corrections must be used together with this version of IsolationSelectionTool.

For muons, the performance of these working points can be found in MuonIsolationPerformance for muons.

• For most working points, the track isolation uses variable ptvarcone20 for electron while for muon ptvarcone30 is used. topoetcone20 is used in calo isolation cut for both lepton flavors.
• Some working points make use of different variables/cone sizes (shown in red in the table above), and are therefore incompatible with lepton triggers that include isolation (see here)
• For the exact track or calo isolation cuts used in the efficiency-targeted working points can be obtained following these instructions.
• The efficiency-targeted working points (those not having "FixedCut" in their name) have a bad background rejection at low pT (pT<20 GeV). For low pT leptons analyses, you should probably chose one of the FixedCut working point.
• Values of the cuts applied to the isolation variables for each working point at η=0:

Leptons PU-robust

There are also five experimental working points that perform better in high pileup conditions. In order to use them, the necessary variables must be calculated and stored in your derivations. Each WP has the same definition for electrons and muons. If you try to use these WPs, please let us know how they perform for you. The working points labeled "Tight" are designed to have similar efficiency to the existing FixedCutTight, and the points labeled "Loose" are designed to have similar efficiency to FixedCutLoose. FixedCutHighMuTrackOnly has similar efficiency to FixedCutTightTrackOnly.

Photons

oct. 1st

Oct. 6th

Notes:

• FixedCutLoose should not be used for low pT photons (pT<25 GeV): since there is no offset in the cut, it becomes too tight at low pT

Versions of working points / cut maps

%W% The cut maps should be re-obtained using Rel21 samples, however no manpower to perform this task. Please contact the IFF group conveners if you'd like to help with this.

Analysis release _/ 2.2.31 (electron cut maps based on 25ns MC15 samples)

In IsolationSelection-00-02-00, the following modifications are present:

* Muon cut maps: 50 ns MC15 Z to ll samples for pT>15 GeV and J/psi for pT< 15 GeV (as in Analysis release 2.3.16)

• Electron cut maps: 25 ns MC15 Z to ll samples for pT>15 GeV and J/psi for pT<15 GeV
• Photon working points: optimized on 50 ns H to two photons samples
• egamma leakage corrections: rel 20_2

To use the new cut maps of electron, you must update the leakage corrections to the release 20 version in your analysis code. Follow the recipe here. Note that the muon cut maps will not be updated to 25 ns samples. Since the difference of target efficiencies is less than 2% between 25ns and 50 ns, a single cut map will be used for both types of data (50 ns for muons, 25 ns for electrons - the reason to update electrons is because the leakage was also updated).

More info

Usage and examples

Working example can be find in util/testIsolationSelectionTool.cxx

Basic use

CP::IsolationSelectionTool iso_1( "iso_1" );
  CHECK( iso_1.setProperty("MuonWP","Gradient") );
  CHECK( iso_1.setProperty("ElectronWP","Tight") );
  CHECK( iso_1.setProperty("PhotonWP","Cone40") );
  CHECK( iso_1.initialize() );

In event loop

const xAOD::PhotonContainer* photons(nullptr);
    CHECK( event.retrieve(photons,m_sgKeyPhotons) );
    for (auto x : *photons) {
       if (iso_1.accept( *x )) Info(APP_NAME," Photon passes Isolation");
    }

The tool will use the corresponding WP based on the type of the object passed to it.

Using more than one WP

CP::IsolationSelectionTool iso_3( "iso_3" );
  CHECK( iso_3.setProperty("MuonWP","Loose") );
  CHECK( iso_3.initialize() );

  /// add Tight WP
  CHECK( iso_3.addWP("Tight", xAOD::Type::Muon));

The accept method returns an TAccept object.

iso_3.accept( *x )

For ntuple processing

One can pass an C++ struct object to the tool to use the WP defined in m_objWPs. It would be useful for ntuple processing.

/// use some object rather than xAOD objects
  CP::IsolationSelectionTool iso_3( "iso_3" );
  CHECK( iso_3.setProperty("MuonWP","Loose") );
  CHECK( iso_3.initialize() );

  /// take the Loose, tight and GradientLoose from muon WP
  CHECK( iso_3.addWP("Tight", xAOD::Type::Muon));
  CHECK( iso_3.addMuonWP("Gradient"));
  for(auto x: iso_3.getMuonWPs()) {CHECK(iso_3.addWP(x, xAOD::Type::Other)); Info("adding WP", x->name().c_str());}

  strObj strMuon;
  strMuon.isolationValues.resize(xAOD::Iso::numIsolationTypes);

In event loop:

const xAOD::MuonContainer* muons(nullptr);
    CHECK( event.retrieve(muons,m_sgKeyMuons) );
    for (auto x : *muons) {
      strMuon.pt = x->pt();
      strMuon.eta = x->eta();
      /// make sure the variable needed by your WP is passed to the tool
      x->isolation(strMuon.isolationValues[xAOD::Iso::topoetcone20], xAOD::Iso::topoetcone20);
      x->isolation(strMuon.isolationValues[xAOD::Iso::ptvarcone30], xAOD::Iso::ptvarcone30);

      /// isolation selection
      if (iso_3.accept( strMuon )) Info(APP_NAME," Muon passes Isolation");
    }

Using IParticle interface

For other xAOD objects inherited from IParticle, there is also an interface provided. The IPariticle should have the isolation auxdata accessible using the standard accessors. The example code (not tested):

CP::IsolationSelectionTool iso_1( "iso_1" );
  CHECK( iso_1.setProperty("MuonWP","Gradient") );
  CHECK( iso_1.initialize() );
  CHECK( iso_1.setIParticleCutsFrom(xAOD::Type::Muon))

In event loop

const xAOD::TrackParticle* tracks(nullptr);
    CHECK( event.retrieve(tracks,m_sgKeyTracks) );
    for (auto x : *tracks) {
       if (iso_1.accept( *x )) Info(APP_NAME," Track passes Isolation");
    }

Retrieve the cut values from the WP

The script scripts/get_cutMap.py inside the package (available starting from IsolationSelection-00-03-00, link to trunk version) can be used to directly retrieve the cut map used for a WP. The cutMap is kept in /cvmfs/atlas.cern.ch/repo/sw/database/GroupData/IsolationSelection/. Alternatively, one can use the following code to retrieve the cuts used for each object:

const xAOD::MuonContainer* muons(nullptr);
    CHECK( event.retrieve(muons,m_sgKeyMuons) );
    IsolationWP *WP = iso_3.getMuonWPS()->at(0); // First, retrieve the desired working point (in this case the first one).
    WP->saveCutValues(); // Enables the storage of the cut values.

    for (auto x : *muons) {

      // Isolation selection
      if (iso_3.accept(*x)) Info(APP_NAME," Muon passes Isolation"); // The values of the cuts are stored with the call of this function.
      std::map<xAOD::IsolationType, float> cuts = WP->cutValues(); // Retrieves the cut values in a std::map. This must be called after accept and after saveCutValues(), if not this will be a null pointer.
      float cut = cuts.at(xAOD::Iso::topoetcone20); // In this example, assuming the working point included a calo based isolation (topoetcone20), the corresponding cut value can be retrieved.

    }

The IsolationWP class is defined in IsolationSelection/IsolationWP.h.

Customized cuts

If none of the working point suits your need, you can implement your own cut in the following way:

For electron and muon, one can add working points targeting at fixed efficiency (cuts will adjusted to provide the required efficiency):

// use a user configured Muon WP?
  CP::IsolationSelectionTool iso_2( "iso_2" );
  CHECK( iso_2.initialize() );

  /// use "myTestWP" WP for muon
  std::vector< std::pair<xAOD::Iso::IsolationType, std::string> > myCuts;
  myCuts.push_back(std::make_pair<xAOD::Iso::IsolationType, std::string>(xAOD::Iso::ptcone20, "0.1*x+90")); /// ptcone20 cut has efficiency min(int(0.1*pT[GeV])+90, 99) percents
  myCuts.push_back(std::make_pair<xAOD::Iso::IsolationType, std::string>(xAOD::Iso::topoetcone20, "0.2*x+80")); /// topoetcone20 cut has efficiency min(int(0.2*pT[GeV])+80, 99) percents
  CHECK( iso_2.addUserDefinedWP("myTestWP", xAOD::Type::Muon, myCuts));

Or using cuts directly (the efficiency is not fixed):

// use a user configured Muon WP?
  CP::IsolationSelectionTool iso_2( "iso_2" );
  CHECK( iso_2.initialize() );

  /// use "myTestWP" WP for muon
  std::vector< std::pair<xAOD::Iso::IsolationType, std::string> > myCuts;
  myCuts.push_back(std::make_pair<xAOD::Iso::IsolationType, std::string>(xAOD::Iso::ptcone20, "0.058*x")); /// ptcone20<=0.058*pT [MeV]
  myCuts.push_back(std::make_pair<xAOD::Iso::IsolationType, std::string>(xAOD::Iso::topoetcone20, "0.015*(x+100000)")); /// topoetcone20<=0.015*(pT+100000) [MeV]
  CHECK( iso_2.addUserDefinedWP("myTestWP", xAOD::Type::Photon, myCuts, "", CP::IsolationSelectionTool::Cut));

How to switch from ElectronIsolationSelection to IsolationSelection

Feel free to improve this.

Header

// #include "ElectronIsolationSelection/IsolationSelectionTool.h"
#include "IsolationSelection/IsolationSelectionTool.h"

Class name

// CP::IsolationSelectionTool* iso_loose;
CP::IsolationSelectionTool*  iso_tool;

Configuration

//  iso_loose = new CP::IsolationSelectionTool("iso_loose");
//  iso_loose->setProperty("WorkingPoint","Loose");

  iso_tool = new CP::IsolationSelectionTool("iso_tool");
  iso_tool->setProperty("muonWP","Loose");
  iso_tool->setProperty("electronWP","Gradient");

Calling the tool

No change.

Major updates:
SandrineLaplace - 2015-09-30

%RESPONSIBLE% MatthewKlein
%REVIEW% Never reviewed