Trigger efficiencies for E+Jets comparing tag and probe electron pairs for which the probe pass HLT cuts with all tag and probe pairs.


The triggers of interest are:

Triggers L1 Seed Last Filter
HLT_Ele10_LW_L1R L1_SingleEG5 hltL1NonIsoHLTNonIsoSingleElectronEt15PixelMatchFilter
HLT_Ele15_SW_L1R L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt15PixelMatchFilter
HLT_Ele15_SW_CaloEleID_L1R L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt15CaloEleIdPixelMatchFilter
HLT_Ele17_SW_CaloEleID_L1R L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt17CaloEleIdPixelMatchFilter
HLT_Ele17_SW_TightEleID_L1R L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt17TightEleIdDphiFilter
HLT_Ele22_SW_TighterEleID_L1R_v2 L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt22TighterEleIdDphiFilter
HLT_Ele22_SW_TighterEleID_L1R_v3 L1_SingleEG8 hltL1NonIsoHLTNonIsoSingleElectronEt22TighterEleIdDphiFilter


Data Samples Run Numbers
Run2010A-Sep17ReReco_v2/RECO 135821 - 144114
Run2010B-PromptReco-v2/AOD 146240 149442

Trigger Object Matching

To determine the trigger efficiencies, each PAT electron in the MC sample should be matched to an L1 seed or trigger event. This is done in the following way:

  • L1 Seed matching criteria:
    • $\Delta\phi$ < 0.522
    • for $\eta$ < 1.4791, $\Delta\eta$ < 0.261. For $\eta$ > 1.4791, $\Delta\eta$ < 0.5

  • HLT matching criteria:
    • $\Delta R$ < 0.2

The L1 seed matching criteria are derived from Sam Harper's L1 seed matching code, which can be found here. A simple $\Delta R$ matching for the L1 seed runs into trouble in the endcap, where the x-y geometry means that the crystals are large in eta. A more sophisticated, but also more relaxed cut is required here.

ttbar Electron Cuts and Plots

The cuts used are:


  • Tight Cut (Analysis cuts taken from here, excluding $p_{T}$ cut, and with eidRobustTight)
    • $p_{T}$ > 1.0 GeV
    • $|\eta|$ < 1.4442 or 1.5660 < $|\eta|$ < 2.4
    • $d_{0}$ < 0.200 mm
    • ComRelIso < 0.1
    • SimpleEleId70Iso


    • $p_{T}$ > 1.0 GeV
    • $|\eta|$ < 2.5
    • simpleEleId70Iso

Cuts on Pairs: *Exactly two electrons *Oppositely charged *Invariant mass between 76 and 106 GeV/c^2

Tag/tag pairs are counted as two tag and probe pairs with the tag switched.

The $p_{T}$ plot overflow has been transferred to the last bin in the histogram.

All efficiency plots were generated by defining a TGraphAsymmErrors in ROOT, and calling the member function BayesDivide(). This function takes two histograms (one the numerator and the other the denominator) and computes the efficiencies, while assigning suitable asymmetric errors using Bayesian statistics. Information about the TGraphAsymmErrors class can be found here, and a paper explaining the computation of the errors (which goes beyond normal binomial distribution errors) can be found here.

Each efficiency plot is fitted with the following function:

$y=\frac{A}{2}(1 + erf(\frac{x - x_{0}}{s\sqrt{x}}))$

In the plots below, p0 = A, p1 = $x_{0}$ and p2 = s. p0 corresponds to the high $p_{T}$ limit of the efficiency, $x_{0}$ the midpoint of the slope, and s the gradient of the slope at $x=x_{0}$.


new plots for this trigger here...

TABLE 2 Barrel Endcap
All electrons
L1 Seed to Electrons Efficiency
HLT to Electrons Efficiency
HLT to L1 Seed Efficiency

The L1 seed threshold should be at about 8 GeV, with the fit report a threshold of about 9 GeV.

The fit for $\epsilon^{HLT}_{L1}$ has been performed only on data points above 12, ignoring the "wiggle" at low $p_{T}$. Reading off the value of p0 from the fits, we have at high transverse momentum

TABLE 3 Efficiencies from fit (Barrel) Efficiencies from fit (Endcap)
$\epsilon^{L1}_{ele}$ 0.9998 0.9998
$\epsilon^{HLT}_{L1}$ 0.9943 0.9918
$\epsilon^{HLT}_{ele}$ 0.9941 0.9918

We should expect that $\epsilon^{HLT}_{ele}=\epsilon^{L1}_{ele}\cdot\epsilon^{HLT}_{L1}$. From Table 2, we obtain $\epsilon^{L1}_{ele}\cdot\epsilon^{HLT}_{L1}=0.9941$ for the barrel, and $\epsilon^{L1}_{ele}\cdot\epsilon^{HLT}_{L1}=0.9916$ for the endcap.

Comparing the plots shown above, we see that the slopes of the trigger efficiency for the L1 seed and the HLT begin at different transverse momenta: around 5 GeV for the former and 12 GeV for the latter. This accounts for the wiggle that we see in the plot of $\epsilon^{HLT}_{L1}$: at low transverse momentum, both the L1 seed and the HLT have similar efficiencies with respect to all PAT electrons. But between 5 and 12 GeV, the L1 seed is increasingly likely to pass electrons, while the HLT still has the same efficiency as before, leading to a dip in $\epsilon^{HLT}_{L1}$.

TABLE 4Sorted ascending Barrel Endcap
All electrons
HLT to Electrons Efficiency
HLT to L1 Seed Efficiency
L1 Seed to Electrons Efficiency

TABLE 5 Barrel Endcap
All electrons
L1 Seed to Electrons Efficiency
HLT to Electrons Efficiency
HLT to L1 Seed Efficiency

TABLE 5 Barrel Endcap
All electrons
L1 Seed to Electrons Efficiency
HLT to Electrons Efficiency
HLT to L1 Seed Efficiency

TABLE 6 Barrel Endcap
PAT pt - Gen. pt
(PAT pt - Gen. pt)/Gen. pt
PAT pt - HLT pt
(PAT pt - HLT pt)/HLT pt

TABLE 7 Barrel Endcap
PAT pt - Gen. pt vs. Gen. pt
(PAT pt - Gen. pt)/Gen. pt vs. Gen. pt
PAT pt - HLT pt vs. HLT pt
(PAT pt - HLT pt)/HLT pt vs HLT pt

TABLE 12 Barrel Endcap
All electrons
L1 Seed to Electrons Efficiency
HLT to Electrons Efficiency
HLT to L1 Seed Efficiency

For the L1 seed to electrons efficiency of the barrel electrons, the fit is very poor, despite using several different sets of initial parameters for the fit.

-- HongwanLiu - 16-Jun-2010

Edit | Attach | Watch | Print version | History: r12 < r11 < r10 < r9 < r8 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r12 - 2010-11-29 - unknown
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 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