Matching Monte Carlo generator using ΔR

Complete: 3

Introduction

The best way to match Monte Carlo generator particles to reconstructed objects (tracks, super-clusters, etc.) is to consider the detailed information of Geant4 hits, match the corresponding reconstructed hits (RecHits) and establish the match with an appropriate criterion, e.g.: a large fraction of RecHits matching to Geant4-hits coming from the same generator particle.

Some of such tools are available for tracks. Anyway, they can't be used in AOD since the matching is done at thelevel of Geant4 tracks, that are not saved in AOD, instead of generator particles.

While a G4-based match is being developed, a possible quick, but not perfect, alternative is to consider the matching done using a metric that measures the distance of reconstructed objects from generator particle. The simplest metric to use is:

  • ΔR = √(Δφ2 + Δη2).

Common MC matching tools using ΔR exists and will be used for CSA07 analysis.

Matching Distributions

The following plots show the ΔR of reconstructed tracks w.r.t. the closest Monte Carlo final state muon. The sample contains inclusive Z→μ+μ- produced with 1_3_1.

Plots are separated for:

  • global muons (muons matched with tracker track)
  • tracks (reconstructed in the tracker only)
  • stand-alone muon (muon reconstructed in the muon detector only)

Global Muons


Δφ

Δη

ΔR
σ(Δφ) = 0.00017 σ(Δη) = 0.00035 <ΔR> = 0.00047

Tracks


Δφ

Δη

ΔR
σ(Δφ) = 0.00017 σ(Δη) = 0.00035 <ΔR> = 0.00081

Stand-Alone Muons


Δφ

Δη

ΔR
σ(Δφ) = 0.0149 σ(Δη) = 0.0076 <ΔR> = 0.055

Conclusions

  • Resolutions in φ and η are different, so the metrics ΔR = √(Δφ2 + Δη2) is not the best to perform the MC truth association. A metric that is asymmetric in Δφ and Δη would have better performances. The cut usually adopted in many analysis channels is ΔR < 0.15.

  • Stand-alone muons have a much worse resolution in both φ and η w.r.t. tracker tracks. Monte Carlo matching should have a relaxed cut for an efficient match of stand-alone muons. A cut too tight would induce artificial inefficiency, while a cut too loose would increase the chance of "fake" associations.

  • The correct procedure to treat Monte Carlo association should be based on Geant4 to RecHit matching. More effort is needed to develop such association modules in order to produce results that are usable with AOD, for instance in form of association maps of reco'd objects to GenParticleCandidates.

Review Status

Editor/Reviewer and Date Comments
LucaLista - 13 Jul 2007 page content last edited

Responsible: LucaLista
Last reviewed by:

Edit | Attach | Watch | Print version | History: r10 < r9 < r8 < r7 < r6 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r10 - 2007-07-16 - JennyWilliams



 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    CMSPublic All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback