The reconstruction of electrons makes use of information from both the ID and ECAL sub-detectors, hence electron candidates are recon- structed only in the central region of the detector, up to |η| < 2.47 [97].
Electron reconstruction starts by looking for energy deposits (clus- ters) in the EM calorimeter with size 3 × 5 in units of ∆η × ∆φ as lon-
gitudinal towers3with a total transverse energy above 2.5 GeV, used as
seed for a sliding window algorithm.
Track seeds from the silicon detector are then extended to the calori- meter. First, the assumption that they are pions is used to account for the energy loss due to interactions with the detector material. If the track seed cannot be successfully extended to a full track under the pion hy- pothesis and it falls into an energy cluster in the EM, it is refitted under the hypothesis that it comes from an electron, which allows for larger
energy losses. The final tracks are matched to EM clusters using the∆R
metric.
Clusters are then rebuilt using 3 × 7 (5 × 5) longitudinal towers of cells in the barrel (end-caps) of the EM calorimeter, with the energy of the clusters calibrated using studies based on simulated MC sam- ples [98]. If no tracks are associated with an ECAL cluster, it is classi- fied as a photon.
The final electron energy is given by the final calibrated cluster, while the η and φ directions are taken from the corresponding track param- eters. Furthermore, electron measurements are performed by requiring the track to be compatible with the primary vertex, in order to reduce the background from conversions and decays of secondary particles.
Lastly, for most of the analyses in ATLAS, including the ones pre- sented in this thesis, electrons within the transition region between the barrel and end-cap of the calorimeter, 1.37 < |η| < 1.52, are vetoed, as this region has a poor reconstruction and energy resolution perfor- mance.
Algorithms for electron identification (ID) are used in order to deter- mine whether the reconstructed electron is indeed an electron or another object faking it, such as converted photons or jets. These algorithms use quantities related to the shape of the electromagnetic cluster shower, track properties, as well as track-cluster matching variables.
The baseline identification algorithm is a likelihood-based method, that allows for a simultaneous evaluation of several properties via the means of signal and background probability density functions (PDFs)
of the discriminating variables. This allows for better background re- jection for a given signal efficiency than a “cut-based” algorithm. Typi- cally, three identification operating points are provided for electron ID, referred to as Loose, Medium and Tight, in order of increasing back- ground rejection. Figure 3.4 shows the combined electron reconstruc-
tion and ID efficiency as a function of ET and η.
Reco + ID efficiency 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 ATLAS Preliminary -1 = 13 TeV, 3.2 fb s <2.47 η -2.47<
Data: full, MC: open
Loose Medium Tight [GeV] T E 10 20 30 40 50 60 70 80 Data / MC 0.8 0.9 1 (a) Reco + ID efficiency 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 ATLAS Preliminary -1 = 13 TeV, 3.2 fb s >15 GeV T E
Data: full, MC: open
Loose Medium Tight η 2.5 − −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 Data / MC 0.8 0.9 1 (b)
Figure 3.4: Combined electron reconstruction and identification efficiencies
in Z → ee events as a function of the transverse energy ET (a) and as a
function of pseudorapidity η (b). These plots are taken from Ref. [97]. Another requirement imposed on reconstructed electrons is that they must be isolated, i.e. the detector activity around the electron must be minimal, in order to disentangle prompt electrons, such as those from Z → ee events, from other sources, like electrons originating from pho- ton conversions, electrons from heavy flavour hadron decays and light hadrons (mostly pions) misidentified as electrons.
Two discriminating variables are used for this purpose:
• a calorimeteric variable, ETcone0.2, defined as the scalar sum of
transverse energy clusters within a cone of ∆R = 0.2 around the
electron candidate, excluding the energy deposits associated with the candidate itself;
• a track-based variable, pvarcone0.2T , defined as the scalar sum of the transverse momenta of all tracks within a cone around the candi-
date electron track of size∆R = min(0.2,10 GeV/ET)and stem-
ming from the PV, excluding the electron associated tracks. A variety of selection requirements on these quantities are defined, resulting in a set of eight isolation working points. More details on the characteristics of the various working points can be found in Section 5 of Ref. [97].
The accuracy with which the electron efficiency is modelled by de- tector simulation plays an important role in a variety of analyses, such as cross-section measurements and various searches for new physics. The efficiency to find and select an electron is not measured as a single quantity, but is divided into different components, namely reconstruc- tion, identification, isolation, and trigger efficiencies, so that the total efficiency ε, is the product of the individual efficiencies, each one mea- sured with respect to the previous step.
In order to achieve reliable physics results, the MC simulated samples need to be corrected in order to reproduce the measured efficiencies in data, therefore, a calibration of the MC detector response is needed. The calibration is provided in terms of multiplicative scale factors (SF)
as a function of pT and η of the electron, derived as the ratio of the
efficiencies measured in data and the corresponding ones in simulation. The electron efficiencies are measured by using a tag-and-probe tech-
nique4using Z → ee events and J/ψ → ee events for the low p
Trange.
These data-to-MC correction factors are usually close to unity; devi- ations arise from the mismodelling of tracking properties or shower shapes in the calorimeter. Figure 3.5 shows the electron isolation ef-
ficiency as a function of the electron ET and η for the representative
FixedCutLoose isolation working point.
4 The tag-and-probe method uses events containing resonances whose decay into particles is
easy to identify, in this case Z → ee and J/ψ → ee. A strict selection on one of the electron candidates (called “tag”) together with the requirements on the di-electron invariant mass, and on the lifetime information for the case of J/ψ, allows for a loose pre-identification of the other electron candidate (“probe”). The probe electron is then used for the measurement of the reconstruction, identification, isolation and trigger efficiencies.
[MeV]
T
E 20 40 60 80 100 120 140
Electron isolation efficiency
0.8 0.85 0.9 0.95 1 ATLAS Preliminary -1 = 13 TeV, 3.2 fb s < 0.6 η 0.1 < FixedCutLoose Data ee MC → Z [GeV] T E 20 40 60 80 100 120 140 Data / MC 0.98 1
1.02 Stat ⊕ Syst Stat only
(a)
η
2
− −1.5 −1 −0.5 0 0.5 1 1.5 2
Electron isolation efficiency
0.98 0.985 0.99 0.995 1 ATLAS Preliminary -1 = 13 TeV, 3.2 fb s < 40 GeV T 35 GeV < E FixedCutLoose Data ee MC → Z η 2 − −1.5 −1 −0.5 0 0.5 1 1.5 2 Data / MC 0.995 1
1.005 Stat ⊕ Syst Stat only
(b)
Figure 3.5: Efficiency of the representative FixedCutLoose isolation require-
ment as a function of the transverse energy ET (a) and as a function of η
(b). Electrons are required to fulfil the Tight identification. For efficiencies
as a function of ET (η) the corresponding η (ET) range used for the probe
electron is shown on the plot. These plots are taken from Ref. [97].