Tau-related plots from the CSC book

This page contains approved plots and results in the order as they appear in the CSC note. Only the CSC note contains all the relevant information and should thus be consulted if any of the plots is used.

Additional Information

  • A full version of the tau CSC chapter is available via CDS (ATL-COM-PHYS-2008-068)
  • If you want to cite the CSC book (full text) in your publication, please do it the following way:
    ATLAS Collaboration,
    Expected Performance of the ATLAS Experiment,
    Detector, Trigger and Physics,
    CERN-OPEN-2008-020, Geneva, 2008.
  • An .eps version of each of the plots displayed below is available by clicking on the corresponding thumbnail.


tau_et_cor tau_et_cor2_type2
Caption: The visible transverse energy of τ leptons from different physics processes: top quark decays, W/Z production, Standard Model vector boson fusion Higgs boson production for mH=120 GeV with H->ττ, for τ leptons from low energy Supersymmetry with a light stau (SU1 sample), heavy Z' bosons, and heavy Higgs bosons from bbH production in the MSSM with tanβ=20(45) for masses of 400 GeV (800 GeV).

Performance of the ATLAS detector for tau identification

Tracking and Vertexing

Reconstruction efficiency and track quality

Caption: Reconstruction efficiency for tracks from charged π's for one- and three-prong hadronic τ decays from W->τν and Z->ττ signal samples as a function of the transverse momentum of the track (left) and of the pseudorapidity for three different ranges of track pT (right).

Charge mis-identification and tracks from conversions

Caption: Percentage of one- and three prong τ lepton hadronic decays within reconstructed one-, two- and three-prong τhad candidates by the track-based algorithm, matched to true τ decays. Tracks in a cone of Δ R=0.2 around the leading good quality track are considered. A transverse momentum of pT>9 GeV is required for the leading track. An estimate for electron contamination and charge misidentification is given in addition. Separately specified are results for a subsample where no hadronic secondary interaction of primary charged π was recorded inside the inner detector volume. Events from Z->ττ and W->τν samples were used. (eps version).

Seeds for track-based τhad candidates Reconstructed as single-prong Reconstructed as three-prong Reconstructed as two-prong
Electron contamination (from conversion) 1.5% 5.7% 2.9%
τ->π+/-0 ν 96.1% 3.8% 23.8%
τ-> 3π+/-0 ν 3.9% 96.2% 76.2%
Charge misid. 1.7% 3.6%
(no had. interact.) 0.4% 2.1%

Impact parameter

Caption: Transverse (left) and longitudinal (right) impact parameter resolution as a function of |η| from a one-prong Z->ττ sample. The open (full) circles are from τ->π(π0)ν (τ->μνν) events.

Caption: Significances of the impact parameters d0 (left) and z0sin(θ) (right) for 1-prong τhad candidates reconstructed by the track-based algorithm. Distributions are shown for τhad candidates reconstructed from τ decays and for fake candidates which do not originate from the decays of b- or c-hadrons.

Secondary vertex reconstruction and transverse flight path

Caption: Residuals of the secondary vertex position parallel and perpendicular to the direction of flight of the τhad candidate using the adaptive vertex fitter. Fully (solid) and partially (open) matched three-prong τhad candidates reconstructed with the track-based algorithm from Z->ττ and W->τν processes are used.

Caption: Resolution and mean of the distribution of residuals of the secondary vertex position in the directions parallel and transverse to that of the reconstructed momentum vector of the τhad candidate as obtained from the adaptive vertex fitter. Candidates with up to three associated tracks reconstructed by the track-based algorithm were used. The resolution quoted is the σ of the core Gaussian of a double Gaussian fit in the range [-4 mm, 4mm] in the parallel direction and [-50 μm, 50μm] in the transverse direction. The 68.3% and 95% coverages are also quoted. (eps version).
Resolution Mean 68.3% 95%
Fully matched 3-prong 0.593 +/- 0.008 mm 0.006 +/- 0.006 mm 1.27mm 5.33mm
Partially matched 0.703 +/- 0.030 mm -0.035 +/- 0.020 mm 3.83 mm > 15 mm
Combined 0.613 +/- 0.008 mm 0.004 +/- 0.006 mm 1.89 mm 11.37 mm
Fully matched 3-prong 10.1 +/- 0.2 μm 0.2 +/- 0.1 μm 14.4 μm 36.9 μm
Partially matched 11.3 +/- 0.5 μm -0.1 +/- 0.2 μm 20.9 μm 72.2 μm
Combined 10.5 +/- 0.2 μm 0.1 +/- 0.1 μm 16.4 μm 48.1 μm

Caption: Resolution on the transverse flight path reconstructed with the adaptive vertex fitter for fully matched three-prong τhad candidates as a function of the transverse momentum (left) and the pseudorapidity (right). Standard deviations of Gaussians fitted to central intervals covering 80% of the residual distributions are shown (black points). In addition the 68.3% and 95% coverages of the distributions of residuals of the secondary vertex position are shown (dashed and dot-dashed lines).}

Caption: Significance of the transverse flight path for fully matched and partially matched three-prong and for fake candidates with and without hadrons containing b or c quarks (the contribution from semileptonic decays of b/c jets into τ leptons was not subtracted).

Reconstruction of π0 subclusters

Caption: Single prong candidates: fractions with zero, one and two or more reconstructed π0 subclusters. (eps version).

decay mode no π0 subclusters 1 π0 subcluster >= 2 π0 subclusters
all τhadν 32% 35% 33%
τ->πν 65% 20% 15%
τ->ρν 15% 50% 35%
τ->a1 (-> 2π0π)ν 9% 34% 57%

Caption: The energy response obtained for the visible energy from τ->ρν events using candidates with one π0 subcluster (left). The invariant mass of the visible decay products for hadronic single-prong τ->ρν, τ->a1 (-> 2 π0 π) ν, and τ->πν decays using candidates from W->τν events with at least one π0 subcluster reconstructed (right).

Combined veto on electron tracks

Caption: The efficiency of the electron veto algorithm for W->τν (rectangles) and W->eν (triangles) events as a function of |η| and pT of the leading track.

Caption: Veto for ele and muon tracks Efficiency for hadronically decaying τ leptons and true electrons from W->τν for passing the electron veto algorithm. The numbers given are normalized to true electrons with pT >9 GeV and |η|<2.5 (vs. true e) and to reconstructed one-prong or three-prong candidates with the leading track being matched to a π from W->τν events (vs. reconstructed τhad). The probability that an electron from W->eν events with pT >9 GeV and |η|<2.5 is reconstructed as one-prong (three-prong) candidate is ~70% (~0.7%). In addition the performance of the standard algorithm for electron reconstruction is shown. The statistical uncertainty on the numbers presented here is at the level of (0.1-0.5)%. (eps version).

Candidates Reconstructed as single-prong Reconstructed as three-prong Overall
Electron-veto algorithm
τ from W->τν (vs reconstructed τhad) 94.1% 96.2% 94.9%
Electron from W->eν (vs true e) 1.5% <0.1% 1.6%
Standard algorithm (tight selection)
τ from W->τν (vs reconstructed τhad) 99.9% 99.9% 99.9%
Electron from W->eν (vs true e) 15.6% 0.4% 16.4%
Standard algorithm (medium selection)
τ from W->τν (vs reconstructed τhad) 90.6% 95.1% 92.1%
Electron from W->eν (vs true e) 4.2% 0.2% 4.6%

Offline algorithms for tau reconstruction

The calorimeter-based algorithm

Caption: The ratio of the reconstructed ET and the true (ETτ-vis) transverse energy of the hadronic τ decay products is shown as a function of the visible true transverse energy ETτ-vis (left), calculated in |η|<2.5 and |η| (right) for taus from Z->ττ (triangles) and A->ττ with mA=800 GeV (squares) decays. The ordinate value is the mean and the error bars correspond to the sigma of the Gaussian fit performed in the range 0.8<ET/ETτ-vis. The results are obtained after applying the loose likelihood selection, see below.

Caption: The distributions of a few discriminating variables (electromagnetic radius, energy isolation, transverse energy width in the η strip layer and ET over pT, 1 of the leading track) used in the calorimeter-based tau identification for true tau decays and jets with visible transverse cluster energies ET in the range from 40 to 60 GeV and track multiplicities between 1 and 3.

Caption: Left: The log likelihood (LLH) distribution for τ leptons (solid) and jets from QCD production (dashed). The likelihood is applied after a preselection on the number of associated tracks, i.e. requiring 1<= Ntr<=3. (Candidates with LLH<-10 had variables outside the boundaries of histograms used when obtaining the PDFs for the likelihood calculation). Right: Efficiency for τ leptons and rejection against jets for different ET ranges, achieved with the likelihood selection.

The track-based algorithm

The energy-flow approach

Caption: The fractional energy response for single-prong (left) and three-prong (right) true τhad candidates reconstructed with the track-based algorithm. Events from a W->τν sample are shown.

Identification with calorimetric and tracking variables

Caption: The distributions for signal and backgrounds for the visible mass mviseflow and ratio of the transverse energy in the isolation and core region ETisol/ETcore for single-prong candidates, and variance Wtracksτ and invariant mass of the track system mtrk3p for three-prong candidates. Distributions are shown for the candidates in the transverse energy range ET=20-40 GeV.

Overall efficiency and rejection

Caption: Efficiencies and rejection rates for different discrimination techniques for the track-based algorithm for fixed efficiencies. The efficiencies are normalized to all hadronic τ decays. The rejection rates are calculated with respect to jets reconstructed from true particles in the Monte Carlo. Events from Z->ττ signal samples and QCD dijets were used. The errors given are statistical only. (eps version).
Selection Efficiency Rejection Rejection Rejection Rejection
cuts TMVA cuts NN PDRS
ET = 10-30 GeV :
one-prong 0.33 225 +/- 10 435 +/- 30 510 +/- 40 460 +/- 40
three-prong 0.28 360 +/- 25 470 +/- 40 740 +/- 70 670 +/- 60
ET = 30-60 GeV :
one-prong 0.42 140 +/- 10 170 +/- 10 440 +/- 40 320 +/- 30
three-prong 0.45 60 +/- 2 9 0 +/- 10 160 +/- 10 130 +/- 10

Comparison of the two algorithms

Caption: Expected performance for the track-based algorithm with a neural-network selection (left) and the calorimeter-based algorithm with the likelihood selection (right). The rejection rates against jets from Monte-Calo particles as a function of the efficiency for hadronic τ decays for various ranges of the visible transverse energy are shown. For signal events Z->ττ and bbH, H->ττ with mH=800 GeV were used, for the background QCD dijet samples were used.

Caption: Rejection against jets from Monte Carlo true particles for a 30% efficiency and separately for the one-prong (1p) and three-prong (3p) candidates. The efficiencies are normalized to true hadronic τ decays. For the signal Z->ττ events and events from bbH, H->ττ with mH=800 GeV were used; for the background QCD dijet-samples were used. The errors given are statistical only. (eps version).
Algorithm ET = 10-30 GeV ET = 30-60 GeV ET = 60-100 GeV ET > 100 GeV
Track-based 1p: 740 +/- 70 1p: 1030 +/- 160
(neural network) 3p: 590 +/- 50 3p: 590 +/- 70
Calo-based 1p: 1130 +/- 50 1p: 2240 +/- 140 1p: 4370 +/- 280
(likelihood) 3p: 187 +/- 3 3p: 310 +/- 7 3p: 423 +/- 8

Caption: Track multiplicity distributions obtained for hadronic τ decays with a visible transverse energy above 20 GeV and below 60 GeV using the track-based τ identification algorithm. The distributions are shown after reconstruction, after cut-based identification and finally after applying the neural network (NN) discrimination technique for an efficiency of 30% for the signal (left) and the background (right).

Caption: Track multiplicity distributions obtained for hadronic τhad-decays with visible transverse energy above 20 GeV and below 60 GeV using the calorimeter-based τ identification. The distributions are shown after reconstruction and after applying the likelihood discrimination technique (medium selection) for the signal (left) and the background (right).

Fake-rates from QCD di-jet samples

Caption: Example of selections on a MC dijet sample, generated with 70<=pT<= 140 GeV. The two jets have to fulfill Δφ >= (π -0.3) in order to be back to back in φ (left) and have similar pT values (right).

Caption: The τhad fake rate from QCD jets and its statistical uncertainty for the available Monte Carlo statistics and for expected 100 pb-1 of data in bins of pT for both τhad reconstruction algorithms. (eps version).
Calorimeter-based algorithm Track-based algorithm
pT range MC stat. Expected stat. error MC stat. Expected stat. error
(GeV) (%) for 100 pb-1 (%) (%) for 100 pb-1 (%)
15-40 2.3 +/- 0.3 +/- 0.02 2.5+/-0.5 +/- 0.02
40-80 5.2 +/- 2.2 +/- 0.01 6.7+/-2.2 +/- 0.01
80-120 0.5 +/- 0.2 +/- 0.001 1.8+/-0.6 +/- 0.002
120-160 0.2 +/- 0.2 +/- 0.002 1.4+/-0.6 +/- 0.004

Tau leptons in Standard Model processes

W->τν inclusive production

Caption: The track multiplicity spectrum of accepted τhad candidates after selection as described in the text with thresholds respectively ETmiss > 50 GeV (left) and ETmiss > 60 GeV (right). The expected event numbers are given for an integrated luminosity of 100 pb-1.

Caption: Expected number of events in 100 pb-1 of data for signal and background after subsequent steps of the selection. The track-based algorithm has been used for τhad reconstruction. The QCD background has been estimated combining fast and full simulation. Given are the expected number of events of track multiplicity one to three, i.e. contributing to signal region only. (eps version).
Selection W->τν W->eν W->μν QCD dijet ttbar, Z->ee, Z->ττ
Trigger tau20i+EFxE30 8.8 104 6.1 105 3.2 104 4.8 108 3.0 105
Identified τ + ETmiss > 30 GeV 2.0 104 2600 200 3.0 106 1600
ETmiss > 50 GeV 4200 530 90 5.0 104 550
Veto fake ETmiss topology 3600 500 80 1.8 104 150
Require jet pT > 15 GeV 3240 450 60 3200 80
Increase to ETmiss > 60 GeV 1550 150 25 500 30

The Z->ττ inclusive production

Caption: Left: The reconstructed visible mass of the (l τhad) pair for Z->ττ decays (solid line) and QCD, W->lν, Z->ll backgrounds (dashed line). Right: The reconstructed visible mass of the (l τhad) pair from Z->ττ decays as a function of the τhad energy scale (right). The dashed lines correspond to +/- 1 σ and +/- 3 σ with respect to the reconstructed peak position. The results were obtained with the calorimeter-based algorithm.

Caption: Expected number of events in 100 pb-1 of data for signal and background after reconstruction of the τ candidate with the calorimeter-based algorithm and after application of the selection cuts for the Z->ττ channel. The QCD background has been estimated combining fast and full simulation. (eps version).
Selection Z->ττ W->lν QCD dijet ttbar Z->ll
Isolated lepton 1.5 104 16.7 105 1.1 107 2.6 104 2.2 105
ETmiss > 20 GeV 4750 14.3 105 3.2 105 2.4 104 1.0 104
mT(l, ETmiss) > 30 GeV 3200 2.6 104 1.8 105 3650 3200
ΣET < 400 GeV 3000 2.4 104 1.7 105 1280 2800
b-jet veto 2780 2.4 104 2.7 104 135 2600
τhad -id + Δφ(l, τhad) cuts 630+/-30 210 +/-10 74+/-11 10+/-2 30+/-5
OS events, ml, τhad = 37-75 GeV 520+/-30 45 +/-5 29+/-5 < 5 10 +/-5

The τ leptons from ttbar production

Caption: Combined b-tagging weights using impact parameter and secondary vertex information for the first two leading ET jets, both in ttbar->W(eνe, μνμ)W(τhadντ)bbbar and W+3-jets background. The eτ (μτ) channel is shown on the left (right). The cut value of 7 on the b-tagging weight is indicated with the arrows. An integrated luminosity of 100 pb-1 of data is assumed.

Caption: Expected number of events in 100 pb-1 of data for ttbar-> W(lν)W(τhad, ντ) bbbar signal and background after subsequent steps in the selection. The track-based algorithm has been used for τhad reconstruction. (eps version).
Selection ttbar (l, τhad) W->lν + 3 jets single t Z->ll + 2 jets
Isolated lepton pT > 20 GeV 1300 3.9 105 4300 630
Identified τhad pT > 15 GeV 190 22000 210 120
1st jet ET > 50 GeV, 2nd jet ET > 30 GeV 170 4000 170 35
ETmiss > 25 GeV 150 3400 150 15
ΣET > 250 GeV 150 1750 130 10
Opposite-sign events 130 850 54 <10
1 b-jet tag 67 28 20

Major updates:
-- WolfgangMader - 24 Oct 2008 -- ChristophAnders - 22 Jul 2008 -- ChristophAnders - 19 Jul 2008

Responsible: MichelJanus and ChristophAnders
Last reviewed by: Never reviewed

Edit | Attach | Watch | Print version | History: r26 < r25 < r24 < r23 < r22 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r26 - 2011-01-26 - PatrickJussel
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Atlas 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