• Keine Ergebnisse gefunden

Multijet background and non-prompt leptons

6. Signal and background modelling 53

6.3. Multijet background and non-prompt leptons

Events where jets or photons are reconstructed as electrons can pass the analysis selection.

Thesefake electrons can occur in multijet events and thus this background is sometimes called multijet background. Estimation of the fake leptons is difficult as the probability of a singular jet or photon faking an electron is very small but the cross-section of multijet background is very high and thus a very large number of simulated events would be required. Furthermore, electrons and muons that originate from semileptonic decays of hadrons can pass the charged lepton identification and isolation criteria for prompt leptons. Simulation of these contributions

6. Signal and background modelling

Sample Generator ME PDF Shower Normalisation Cross section [pb]

t¯t Powheg+EvtGen NNPDF3.0 Pythia8 NNLO+NNLL 831.76 Single top (W t) Powheg+EvtGen NNPDF3.0 Pythia8 (app.)NLO 71.7 Single top (t) Powheg+EvtGen NNPDF3.0 Pythia8 (app.)NLO 70.43 Single top (s) Powheg+EvtGen NNPDF3.0 Pythia8 (app.)NLO 3.35

W+jets Sherpa2.2.1 NNPDF3.0 NNLO 20080.0

Z+jets Sherpa2.2.1 NNPDF3.0 NNLO 2107.0

Diboson Sherpa2.2.2 NNPDF3.0 NLO 176.0

t¯tZ MadGraph5aMC@NLO NNPDF3.0 Pythia8 NLO 0.88 t¯tW MadGraph5aMC@NLO NNPDF3.0 Pythia8 NLO 0.60 t¯tH MadGraph5aMC@NLO NNPDF3.0 Pythia8 NLO 0.51

Table 6.1.: A summary of basic MC generator settings used to simulate various SM processes.

is difficult, and thus if the contribution of these processes is expected to be large the background has to be estimated using a data-driven technique.

The Matrix Method [219] is used to estimate the contribution of the multijet background in the lepton+jets channel. Two separate selections on charged leptons are applied: tight andloose selection, leading to two different datasets for both electrons and muons. The difference between the selection is defined by applying gradient isolation criteria for tight leptons, while for loose leptons, no requirement on the isolation is imposed, see Sections5.1and 5.2. The total number of events passing theloose (Nloose) andtight (Ntight) selection reads

Nloose =Nrealloose+Nfakeloose,

Ntight =Nrealtight+Nfaketight, (6.1) whereNloose(tight)

real denotes the total number of events passing theloose(tight) criteria that consist of real prompt leptons, and similarly forNloose(tight)

fake . Since every event passing thetightselection must also pass theloose selection, selection efficiencies for real and fake leptons can be defined as

real = Nrealtight Nrealloose, fake = Nfaketight

Nfakeloose. (6.2)

Thus the interesting value, the number of fake leptons passing the tight selection, can be calcu-lated combining Equation (6.1) and (6.2)

Nfaketight = fake

realfake

realNloose−Ntight

. (6.3)

Technically, the estimation is implemented by applying per-event weights,w, to the data events passing theloose selection

w= fake

realfake (real−f), (6.4)

6.3. Multijet background and non-prompt leptons wheref is equal to 1 if the (loose) lepton passes thetight criteria and 0 otherwise.

The real lepton efficiencies are estimated in measurements described in Section 5.1 for elec-trons, and in Section 5.2 for muons. The fake lepton efficiencies are measured in fake enriched regions of exactly one lepton and at least one jet. To further increase the probability of the lepton to be a fake lepton, the selected events are required to have ETmiss < 20 GeV and low transverse mass of the W boson,mWT <60 GeV, where

mWT ≡ q

ETmisspT(1−cos (∆φ)), (6.5)

to decrease the chance that a neutrino is present in the event. The real and fake efficiencies are parametrised as a function of differential distributions of various parameters. This analysis utilises the following parametrisation of the efficiencies: the distribution of the leading (highest in pT) jet pT and the distribution of ∆R between the lepton and the closest reconstructed jet are used for muons while for electrons distributions of leptonpTand ∆φbetween the lepton and the missing transverse momentum are exploited.

The multijet background in the dilepton channel is not estimated via the Matrix Method as the contribution in this channel is expected to be small and conservative uncertainties on the estimation are applied as will be discussed in Section 9.3. The processes with the highest contribution to non-prompt lepton in the dilepton channel areW+jets,t¯tsingle lepton channel and single top processes with only one genuine charged lepton in the final state. In the dilepton channel, the multijet background contribution is estimated from the MC simulation by splitting events into two categories: events with real prompt leptons and events with fake or non-prompt leptons using the MC truth information. All the SM processes where the real lepton contribution is expected are processed in this way, furthermore, processes where no real lepton can pass the selection but other kinematic properties are similar, are also considered.

CHAPTER 7

Event selection

The measurement of the top-quark decay width uses data recorded by the ATLAS detector at a centre-of-mass energy of 13 TeV in proton-proton collisions in years 2015, 2016, 2017 and 2018.

Section 7.1 describes the used dataset and the corresponding data taking conditions. In the following, the signature of thet¯tdecay in events with at least one electron or one muon selection oft¯tevents is split into two orthogonal channels: lepton+jets and dilepton channels. Section7.2 summarises basic selection criteria that are identical for both the lepton+jets and the dilepton channels, optimised to select events that are rich in tt¯events and to suppress non-t¯t events.

Individual selections in the lepton+jets and the dilepton channels are presented in Section7.3 and Section7.4, respectively.

7.1. Dataset

The dataset is split into four separate subsets, each corresponding to a different year of data taking. Each subset consists of a number of individual runs, representing time periods of stable beams of the LHC. Each run contains many luminosity blocks (LBs), where each LB corresponds to one minute of data taking. Not all events recorded by the ATLAS detector are used in the analysis, only events passing certain quality criteria are selected. Data quality criteria require all detector subsystems to be in the fully operational status. The “good” LBs are stored in a good run list (GRL) that is available for each year separately.

The LHC conditions that affect the instantaneous luminosity and the pile-up values varied between and within each year of the data collection period are illustrated in Figure 7.1 and Figure 7.2. The pile-up conditions change significantly between years 2015, where the average number of interactions, hµi, peaked at around 13.4, and 2017 hµi was 37.8 with some runs reaching up to hµi = 80. The total luminosity corresponding to the GRL in each year is summarised in Table7.1. The luminosity is measured with techniques described in Section4.2.7.

7.2. Preselection

Events passing either of the two considered channels, lepton+jets or dilepton channel, need to pass the basic selection criteria summarised in this chapter. Firstly, events need to fire one of