Control Region definition

The special selection defining the CR used for determining electrons misidentified as photons is explained in this section. The CR is characterised by the requirement of an additional electron fulfilling all the photon requirements. The nominal t¯t selection requirements are imposed with, specifically, the following additional criteria.

• The appropriate lepton trigger must have fired, see Sec. 3.6.

• The event must contain at least four good jets (j) with pT(j) >25 GeV of which at least one must be tagged as ab-jet.

• The event must contain at least one good electron (muon) withET(e)>25 (20)GeV matched to the trigger object. A second good electron with E_T(e)>20 GeV must be present in the event. In events with two (or more) electrons withE_T(e)>25GeV, either of those electrons can be faking a photon. These events are treated twice in the selection each time identifying one as the electron and one as the electron faking a photon (labelled in this section asfake).

The fake must fulfil the following requirements:

– E_T(f)>20 GeV,1.52<|η(f)|<2.37 and |η(f)|<1.37;

– ∆R(j, f)>0.5,∆R(l, f)>0.7;

– for the electron channel the m(e, f) must be outside a 5 GeV window around mZ = 91 GeV.

Events are categorised into the ee channel and µe channel. The former corresponds to the t¯t electron channel with an additional fake, while the latter corresponds to thet¯tmuon channel with and additional fake. Totals of 325 electron and 467 muon events pass the above event selection.

Reconstructed events are reweighted according to thef.r.(e→ γ) as a function of η and E_T, shown in Fig. 6.3. The event weight (w_f.r.(e→γ)) is obtained by summing the weights of each fake candidatefi

w_f.r.(e→γ) =

Ne-fakes

X

i=1

f.r.(e→γ) [E_T(f_i), η(f_i)] (6.5) over all fake candidates (N_e-fakes), any electron with E_T(e) >25 GeV is considered to be a fake.

This procedure avoids any selection bias upon the fake identification. The number of events with an electron faking a photon, after reweighing, are found to be 29 and 42 for the electron and muon channels respectively. Table 6.1 compares the estimated e → γ background to the expectations from simulations.

Contribution ee channel [events] µe channel [events]

t¯t 17.15 ±0.19 (stat)±4.67 (sys) 31.07 ±0.27 (stat) ±6.72 (syst) t¯tγ 0.38± 0.03 (stat)±1.13 (sys) 0.69±0.04 (stat) ±1.70 (sys) Z+jets 2.14± 0.22 (stat)±1.91 (sys) 0.12 ±0.04 (stat) ±4.10 (syst) W +jets <0.06(stat⊕sys) <0.01 (stat⊕sys)

Multijets 2.52± 0.11 (stat) 0.26 ±0.01(stat)

Dibosons 0.09± 0.02 (stat)±0.03 (sys) 0.06±0.02 (stat) ±0.08 (sys) Single top 0.44± 0.05 (stat)±0.10 (sys) 0.89±0.07 (stat) ±0.21 (sys) Total Expected 22.78 ±0.32 (stat)±5.17 (sys) 33.09 ±0.28 (stat) ±8.06 (syst) Data 29.40 ±1.55 (stat)± 2.7 (sys) 41.46±1.92 (stat) ±4.20 (sys)

Table 6.1: Estimated number of events with an electron misidentified as a photon. Systematic uncertainties correspond to those detailed on table 6.2. A 10% uncertainty is assigned to data candidates corresponding to the uncertainties on the fake rates obtained in Sec. 6.1.1. The ex-pectation was obtained at reconstruction level using the same selection and reweighting as for Data.

The data-to-simulation comparison (see Fig. 6.4), after event selection and reweighting, shows a reasonable agreement when systematic uncertainties (comprising the jet, lepton and missing transverse energy modelling) are considered. A detailed description on each component can be found in Chap. 7.

On the t¯tsample uncertainties are found to be of the order of 21% and are mainly driven by the jet energy scale (14%). The knowledge ont¯t,ttγ¯ and Single top is limited by the systematic uncertainties, while the estimation forZ+jets,Z+jetsand dibosons is limited by the size of data.

Table 6.2 shows the full breakdown for each simulated sample.

e/γ backgrounds 116

Uncertainty ee channel[%]

Source W+jets Z+jets Dibosons Single top tt¯ t¯tγ

Jet energy scale <0.01 29.30 22.09 16.49 16.17 21.71

Jet energy resolution <0.01 79.64 26.26 13.92 3.96 6.19

Jet reconstruction efficiency <0.01 0.97 <0.01 3.90 0.14 0.00 Electron energy scale <0.01 7.30 <0.01 <0.01 0.30 2.95 Electron energy resolution <0.01 5.18 10.73 5.67 0.39 2.52 Cell-out and soft terms <0.01 6.045 <0.01 <0.01 0.15 2.05

Pile-up <0.01 4.49 <0.01 0.00 0.14 0.06

Total 0.03 89.41 38.22 22.56 21.40 29.62

Uncertainty µe channel [%]

Source W+jets Z+jets Dibosons Single top tt¯ t¯tγ

Jet energy scale <0.01 6.03 0.63 36.30 16.54 19.28

Jet Energy resolution <0.01 3.46 82.80 8.28 2.35 7.01

Jet reconstruction efficiency <0.01 <0.01 <0.01 <0.01 0.06 0.23 Muon momentum resoultion <0.01 <0.01 <0.01 0.80 0.14 < 0.01 Muon momentum scale <0.01 <0.01 <0.01 <0.01 0.00 < 0.01 Cell-out and soft terms <0.01 <0.01 <0.01 0.00 0.13 0.54

Cell Out Down <0.01 <0.01 <0.01 0.00 0.14 0.43

Pile-up <0.01 <0.01 <0.01 0.02 0.09 0.67

Total 0.03 34.13 137.05 23.03 21.64 24.78

Table 6.2: Summary of systematic uncertainties for the ee (µe) channel are shown on the top (bottom).

The f.r.(e → γ) is estimated with a systematic uncertainty of 10% (see Sec. 6.1.1) con-sequently the estimated event yield in the electron and muon channels is 29.4 ± 1.6 (stat) ± 2.9 (sys) and 41.5 ± 1.9 (stat) ± 4.2 (syst) events respectively. This estimation is compatible with the expectation of 22.78 ± 0.32 (stat) ± 5.17 (syst) events for the electron channel and 33.09±0.28 (stat) ±8.06 (syst) events for the muon channel. The difference in the yielded num-ber of events from simulation and data is associated with the mismodelling of the jet multiplicites by MC@NLO see Fig. 6.5. This is a known feature allready observed by ATLAS with the measure-ment of thet¯tcross section as a function of jet multiplicity [118]. In fact, as theMC@NLOsimulation underestimates the number of jets in the event, the probability of an electron being identified as a photon (due to mis-matched jet tracks being associated to calorimeter clusters) decreases. Thus, the overall simulation-based determination of electrons faking photons is underestimated. This further motivates the choice of a background estimation derived from data.

) / GeV fakee(N

Figure 6.4: Transverse energy of the electron faking a photon candidate. The plot on the left (right) shows event candidates in the ee (µe) channel. The band includes the simulation-based statistical uncertainty from all samples, as well as detector uncertainties (see main text) applied on thet¯t sample. The agreement is significantly better when detector uncertainties on the other simulation samples are added. The last bin contains any overflow.

Events

Figure 6.5: The jet multiplicity distribution is shown. The plot on the left (right) shows event candidates in theee (µe)channel. The band includes the simulation-based statistical uncertainty from all samples, as well as detector uncertainties (see main text) applied on thet¯tsample. Both distributions illustrate the underestimation of the MC@NLO simulation program (labelled as t¯t) with increasing jet multiplicities in the event.