2018 Tau Reconstruction Comparison at the HLT ( CMS DP-2018/035)

The high level trigger (HLT) reconstructed hadronically decaying tau leptons (τh) with a cone-based reconstruction algorithm throughout Run I and the first portion of Run II. In 2018, the Hadron Plus Strips (HPS) τh reconstruction algorithm, which is currently used for offline analysis, was introduced to the HLT to test its performance. In early June 2018, many of the τh trigger paths switched from the cone-based to the HPS reconstruction algorithm after showing simultaneous rate reductions and efficiency gains.

The cone-based τh algorithm uses a signal cone ranging from ΔR = 0.08 to 0.12 which contains the τ decay products. Charged hadrons, photons, and neutral hadrons within the cone are included in the reconstructed τh candidate.

The HPS τh algorithm reconstructs the various decay modes of hadronically decaying tau leptons[1]. The charged hadrons and photons within the signal cone are combined together in multiple combinations and ranked based on their consistency with a genuine τh decay. The highest ranked combination is selected as the reconstructed τh candidate.
[1] “Performance of reconstruction and identification of tau leptons in their decays to hadrons and tau neutrino in LHC Run-2” CMS Collaboration, CMS-PAS-TAU-16-002

The performance of the two τh reconstruction methods, cone-based versus HPS, is assessed by comparing the rate and selection efficiency for the associated trigger paths.

The performance is presented for two types of triggers: μτh triggers and τhτh triggers. The μτh trigger requires both objects to be isolated and has a pT requirement of 20 GeV for the μ and 27 GeV for the τh. The τhτh trigger requires both τh are isolated and have pT > 35 GeV. To maintain a reasonable rate, the cone-based trigger has a relatively tight isolation requirement on both τh unless both have pT > 40 GeV in which case the isolation requirements are relaxed. In comparison, the HPS trigger is able to maintain a consistently looser isolation working point across the full pT spectrum.

Rate measurements: The rate associated with each trigger path is measured during the start of a recent LHC fill with average PU = 50.

Efficiency measurements: Trigger efficiency is measured using tag and probe in Z -> ττ -> μτh events. 2018 CMS data is used corresponding to 14.3 fb-1. An isolated muon which passes one of the single muon HLT paths is used as the tag. A τh candidate, the probe, is required to have pT > 20 GeV, |η| < 2.1, to pass offline isolation and identification criteria, and to pass anti-muon and anti-electron discriminants designed to suppress Z -> μμ/ee events The contribution of jets faking τh are suppressed in the measurement with:

  • A transverse mass(μ, MET) < 40 GeV cut to suppress the rate of W+Jets events.
  • A visible mass(μ, τh) cut to enhance the relative contribution of Z -> ττ, 40 GeV < visible mass < 80 GeV.
  • μ-τh pairs with same charge are subtracted from the μ-τh pairs with opposite charge which helps eliminate the contributions from QCD multijet processes. The final selection yields approximately 90% genuine τh purity. The τhτh trigger efficiency is measured with dedicated μτh triggers where the τh reconstruction and selection is identical to that used in the associated τhτh paths.

    μτh Trigger Performance
    The τh leg efficiency of the μ(20)τh(27) trigger is shown for both τh reconstruction algorithms. The HPS path has a sharper turn-on compared to the cone-based path which is visible at the pT = 27 GeV threshold. The associated rate at average PU= 50 for the μτh triggers are 5.2 Hz for the cone-based algorithm and 4.6 Hz for the HPS algorithm. The efficiency of both paths is relatively comparable.
    [Get pdf version]
    [Get png version with x linear scale]
    [Get pdf version with x linear scale]
    Contact: Tyler Ruggles

    τhτh Trigger Performance
    The per-leg efficiency of the τh(35)τh(35) trigger is shown for both τh reconstruction algorithms. The curves compare the selection of the highest efficiency τhτh triggers with an affordable rate. Measured at average PU = 50, the rate of the cone-based algorithm is 50 Hz and the HPS algorithm is 39 Hz, a 20% reduction for the HPS triggers. The HPS algorithm shows large benefits for the τhτh trigger path including a large rate reduction and a simultaneous efficiency increase.
    [Get pdf version]
    [Get png version with x linear scale]
    [Get pdf version with x linear scale]
    Contact: Tyler Ruggles

    -- ElisabettaGallo - 2018-07-03

  • Topic attachments
    I Attachment History Action Size Date Who Comment
    PDFpdf mu20_tau27_eff.pdf r1 manage 15.3 K 2018-07-05 - 21:23 ElisabettaGallo  
    PNGpng mu20_tau27_eff.png r1 manage 14.8 K 2018-07-05 - 21:23 ElisabettaGallo  
    PDFpdf mu20_tau27_eff_logX.pdf r1 manage 15.2 K 2018-07-05 - 21:23 ElisabettaGallo  
    PNGpng mu20_tau27_eff_logX.png r1 manage 15.8 K 2018-07-05 - 21:23 ElisabettaGallo  
    PDFpdf tau35_eff.pdf r1 manage 15.1 K 2018-07-05 - 21:23 ElisabettaGallo  
    PNGpng tau35_eff.png r1 manage 14.7 K 2018-07-05 - 21:23 ElisabettaGallo  
    PDFpdf tau35_eff_logX.pdf r1 manage 15.0 K 2018-07-05 - 21:23 ElisabettaGallo  
    PNGpng tau35_eff_logX.png r1 manage 15.7 K 2018-07-05 - 21:23 ElisabettaGallo  
    Edit | Attach | Watch | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
    Topic revision: r4 - 2018-07-08 - ElisabettaGallo
      • 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 &© 2008-2023 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