• Keine Ergebnisse gefunden

parton shower and UE, any unstable hadrons are decayed according to the branching ra-tios given, for example, by the Particle Data Group [3]. This aspect cannot be neglected, particularly for cases of subsequent or multiple decays as with hadrons containing charm or bottom quarks.

4.5 Simulation of the ATLAS detector and pile-up

The simulation of proton-proton collisions from the matrix element to the final state objects is only the first step when performing an analysis based on actual physics data.

The MC generators are able to predict the outcomes of proton-proton collisions in terms of physical particles, but in order to compare the generated events to experimental data, it is essential to simulate how the particles interact with the detector material.

The Atlas experiment employs the Geant 4 software package [43] in order to simu-late the response of the Atlasdetector to the proton-proton collisions produced at the Lhc [44]. This involves simulating how particles pass through the various parts of the detector and create hits (or not in case of a neutrino). The simulated hits are digitised to simulate the signals produced by the electronic readout system. The reconstruction algorithms turn these signals into tracks in the ID and MS as well as energy depositions in the calorimeter cells. These tracks and energy depositions are used to reconstruct electrons, muons, photons and hadrons i.e. jets.

Before the simulation of the detector response, however, one particular and very crucial aspect is taken into account within the MC simulated samples to reproduce the Lhc conditions. This aspect is called pile-up and it refers to the concept that the mean num-ber of proton-proton interactions per crossing of the proton bunches,hµi, is significantly larger than one, with an average of hµi ≈ 32 interactions during the years 2015, 2016 and 2017 [45]. This is due to the high luminosity configuration at the Lhc. The different pile-up profiles of the data recorded by Atlasin the three years are shown in Figure4.5.

The large pile-up adds an additional challenge to physics analyses, because the neigh-bouring proton-proton interactions tend to produce many soft particles that are usually not of interest, but may still interfere with the reconstruction of the hard process. It is therefore essential to have a detector with high spatial resolution as well as good momentum resolution in order to identify the particles that originate from the primary vertex and distinguish them from those particles produced by neighbouring collisions.

In simulation, pile-up is taken into account by overlaying the generated events with so-called minimum bias events assuming a value ofhµi= 20 [46]. Minimum bias events are inelastic collisions without any specific selections such as requiring the presence of a high pT lepton or jet, hence the tendency to contain soft particles. In Atlas, these events are generated using the Pythia 8 generator with the MSTW2008LO PDF set and the A2 tune [46–48]. After the full simulation of the detector response and reconstruction of objects, this simulated pile-up profile is corrected to the pile-up profile measured in data

Mean Number of Interactions per Crossing

0 10 20 30 40 50 60 70 80

/0.1]-1Recorded Luminosity [pb

0 50 100 150 200 250 300

350 ATLAS Online, 13 TeV

Ldt=86.3 fb-1

> = 13.4 µ 2015: <

> = 25.1 µ 2016: <

> = 37.8 µ 2017: <

> = 31.9 µ Total: <

2/18 calibration

Figure 4.5: The pile-up profiles of the data recorded with the Atlasexperiment during the years 2015, 2016 and 2017 of Run 2 [45].

via certain weight factors which are applied to the generated events [49]. In addition to the pile-up modelling, the object reconstruction algorithms employed in Atlas are designed in a sophisticated way to reject most particles that do not originate from the hard process, for example the jet vertex tagger algorithm outlined in Section5.3.

The following chapter describes how or under which criteria the different relevant objects, namely leptons and jets, are reconstructed in the analyses presented in Chapters 6and 7.

CHAPTER 5

Analysis objects

In the previous Section 4.5, the point was raised that analysers cannot simply compare an MC simulated sample of particles that are created during a proton-proton collision to actual physical events that are recorded with the Atlas detector. First, the MC generator is required to model how the particles will interact with the detector, what signatures they produce and thus, which electric signals the readout system is ultimately able to record. Based on this information, the Atlas collaboration applies algorithms that read in those particle signatures in the detector during a collision and reconstruct the particle candidate that produced the signature, along with its kinematic properties.

Then, sophisticated trigger algorithms may fire a signal to store the event if certain criteria are met such as the presence of a highpT lepton, as was discussed in Section3.2.

This chapter describes how the individual physics objects, namely leptons and jets, are reconstructed based on the signatures that have been left in the detector. As depicted in Figure3.6, each object produces a different signature which means that all the infor-mation from the detector has to be taken into account to maximise the reconstruction efficiency. Nonetheless, different particles may produce a similar signature inside the detector. Since the detector and readout systems are not perfect, the corresponding algorithms might reconstruct an object as a different one. This has to be considered especially for electrons, photons, jets and hadronically decayingτ-leptons.

5.1 Electrons

Electrons are the lightest charged fermions of the SM and thus do not decay [3,7]. They do, however, ionise the material of the ID and thus leave a track in it. Furthermore, they are expected to be stopped in the ECAL, predominantly by producing an electromag-netic shower through Bremsstrahlung and subsequent production of electron positron

pairs until all their energy is deposited in the ECAL [28].

The electron reconstruction algorithm [50–52] tries to find tracks in the ID with pT >

0.5 GeV that can be associated to energy depositions, called clusters, in the ECAL that fulfilEcluster/cosh (ηtrack)≡ET >2.5 GeV. The energy of the cluster comprises different components besides the actual energy inside the ECAL, namely the energy deposited in front of the ECAL and the energy leakage outside of the cluster as well as after the ECAL. To reconstruct the four-momentum of the electron, the information from the ECAL and ID are combined. The energy of the electron is the cluster energy whereas its direction of movement is based on its track. The detector does not operate with a perfect efficiency, but instead suffers from a limited precision on the energy scale, energy resolution and other properties which are detailed further in Section 6.3.1 and Chap-ter8. In order to match the reconstruction efficiency of simulated electrons (and other objects) to that of electrons in data, the above effects must be taken into account which is done via scale factors (SFs). These SFs are defined as data/MC, where represents the corresponding efficiency. The efficiency in data is measured experimentally and the SFs can be applied to simulated events in all analyses considering electrons.

As the ID is limited inη, only those candidates with|η|<2.47 are considered. However, the region 1.37<|η|<1.52 is excluded as well, because the reconstruction efficiency in this region suffers significantly from the large amount of passive material in front of the ECAL [28].

In addition to the reconstruction of the electron, the algorithms must consider multi-ple other mechanisms that could lead to another object being mis-reconstructed as an electron, thus called fake electron [28]. This involves primarily photons which undergo the same procedure in the ECAL, but leave no track in the ID unless they split into an electron positron pair at an early stage. Another example would be a photon that deposits energy into the ECAL and these clusters are matched wrongly to a track in the ID. On the other hand, jets also leave tracks in the ID and may deposit energy in the ECAL. If the jets are stopped already in the ECAL or do not deposit a lot of energy into the HCAL, they might be mis-identified as electrons due to the similar signature.

These cases are discriminated by a cut-based classification which uses information from the ID as well as the calorimeters, which is described in more detail in Section6.3.1.