Energy scale corrections to be applied to data are determined in 50η bins for central electrons using J/ψ → ee and Z → ee events as well as E/p studies with isolated electrons fromW [112, 57]. In addition, corrections in 8 η bins in the forward region are derived using Z →ee events. In case of the Z analysis, the scale corrections are
W →eν Z →ee Trigger + 0.987±0.001(stat) 1.000±0.000(stat) data/MC SF 0.995±0.004(stat+syst) 1.000±0.000(stat+syst)
Table 6.10: L1 EM14 and EF e15 medium trigger efficiencies and data/MC scale factors determined with the tag-and-probe method on Z →ee events.
η
Figure 6.8: Electron energy scale correction factors vs η, obtained with Z → ee events, to be applied to electrons in data [57].
estimated by constraining theMee distribution to follow the Z line shape in MC.
The energy is then scaled in data depending on the η of the electron, by applying the formula Ecorr = Emeas/(1 +α), where α is the correction factor. α varies from approximately 1% in the barrel to 5−6% in the FCAL, which is presented in Figure 6.8.
The systematic uncertainty of the energy scaling is provided as a function of electron ETin 6 ηbins. The most dominant sources of uncertainty originate from the limited knowledge of material in front of the calorimeter as well as the presampler energy scale. Overall, the uncertainty is of the level of about 0.5% for the energy range most relevant for W and Z measurements.
The simulated data have been found not to reproduce theZ →eemass resolution in data after the appropriate energy scale correction has been applied. The MC resolution must therefore be smeared to better fit the data. Studies on J/ψ events have shown that the energy resolution for lower energy electrons, where the sampling term dominates, is in good agreement with MC. Only the constant term is thus
adjusted.
The smearing function applied on MC, depends on the energy and pseudorapidity of the electron. The constant term is scaled up/down according to its relative un-certainty in order to gauge the systematic errors due to the energy resolution on the resulting CW/Z factors.
The ETmiss needs to be refined both for the case of correcting the energy in data and smearing the MC resolution. Due to the complexity of the ETmiss calculations, only the vectorial change of the identified electron from theW due to the calibration or smearing is propagated to the ETmiss vector. Comparisons have been made with a more rigorous study, involving correcting the energy of all the topological clusters in the event and re-summing them to obtain the calibrated ETmiss vector. The more advanced study shows an excellent agreement with the approximate method used in this analysis.
6.3.6 Background estimation
QCD background
The QCD background is estimated with a similar method to what was performed for the 315 nb−1 cross section results (see section 6.2.2); the ETmiss distribution is fitted with the template method to obtain the QCD background contribution to the observedW →eν candidates [101, 103]. A signal template, including the electroweak and t¯t background processes, is obtained with simulated data. The QCD template is acquired from electron candidates passing loose and track quality cuts but failing the remainder of the medium and tight requirements. To further reject signal events in the control sample, the isolated electrons are excluded by requiring ETcone(0.3)/ET > 0.2.
A large statistics QCD dijet MC sample (containing 50 million events) is used to cross check the ETmiss distribution obtained from the data template.
Several parameters are varied in order to estimated the systematic uncertainty on the estimated W →eν QCD background:
• The nominal fit rangeof the ETmiss is altered from 0-100 GeV to 0-50 GeV and 10-100 GeV.
• The impact on the ETmiss is accounted for by varying the energy scale within its uncertainty (see section 6.3.5).
• TheQCD template selectionis adjusted by altering which medium and tight cuts that are required to be failed or passed. In addition, the isolation cut is modified.
• The signal template is adjusted by changing from Pythia to MC@NLO for the preliminary results and from MC@NLO to POWHEG+Pythiafor the final differential analysis.
The resulting number of background events for W is shown in Table 6.11.
The QCD background present among the observed Z → ee candidates is very small and hence difficult to estimate. A fit is thus performed on the invariant mass distribution for events with the medium-medium selection relaxed to a combination of medium, loose and loose with the additional requirement of 4 hits in the silicon detectors. For the scenario of relaxing the medium requirement for both electrons, the EF e15 medium trigger is replaced by EF e20 loose. The fit is performed with a Crystal Ball distribution convoluted with a Breit-Wigner function for the signal shape and Landau, linear and exponential distributions are considered for the background shape. In addition, different ranges of the invariant mass are considered for the fit.
The resulting background is scaled by loose-medium background rejection factors.
The resulting number of background events for Z is shown in Table 6.12.
Other backgrounds
The electroweak and t¯t backgrounds for W → eν and Z → ee are taken from simu-lated data, with the samples presented in Table 6.2. A systematic uncertainty on the inclusive cross sections of 5%, 6% and 7% for W/Z, t¯t and diboson backgrounds are concluded. The total uncertainty on the electroweak backgrounds are negligible for the final result.
The inclusive number of QCD and other background events with the corresponding uncertainties are summarized in Table 6.11 for W → eν and Table 6.12 for Z → ee. For the differential analysis with the improved W → eν selection, the QCD background is found to be larger but with smaller relative uncertainty. To obtain differential cross sections the background estimation procedure is performed in each
|ηe| (W) and |yZ|(Z) bin. The results are shown in Table C.4 forW+, Table C.5 for W− and Table C.6 forZ, in Appendix C.