Photon and electron identification with the ATLAS detector 38th International Conference on High Energy Physics - Chicago, 3-10 August 2016
Leonor Cerdá Alberich, Instituto de Física Corpuscular (UV-CSIC), for the ATLAS Collaboration
Xh
h
Photon and electron identifications are a crucial input to many ATLAS physics analysis, including H γγ, H ll, SUSY and exotic searches.
For the 13 TeV running of the LHC, a number of improvements were made to the identification algorithms to improve the efficiency at high pileup conditions.
Electron and photon efficiency measurements performed with pp collisions data at 13 TeV in 2015 corresponding to an integrated luminosity of 3.2 fb-1 are presented in this document, as well as a first look at 2016 data.
Abstract
Object Definitions and Event Selection
The ATLAS detector
Electron and Photon Identification
Electron ID efficiency measurements Photon ID efficiencies compared with the simulation
Conclusions
Mail: [email protected]
References
Photon ID efficiencies measured in data
ICHEP 2016, Aug. 3 – 10, Chicago, IL, USA
The ATLAS detector [1] is a multi-purpose apparatus with a forward-backward symmetric cylindrical geometry and nearly 4π solid angle coverage. It consists of the following detectors, with the inner detector (pixel, SCT and TRT) being inside a 2T magnetic field:
¤ Pixel and silicon microstrip (SCT) trackers covering |η|<2.5. The innermost pixel layer, the insertable B-layer (IBL) [2] that was added between Run-I and Run-II, improves the impact parameter resolutions due to its smaller radius and its reduced pixel size.
¤ Transition Radiation Tracker (TRT) covering |η|<2.0; also provides discrimination between electrons and charged hadrons based on transition radiation.
¤ Lead-liquid argon electromagnetic (EM) calorimeter covering |η|<3.2; finely segmented in ϕ and η, and with three layers in depth for |η|<2.5.
¤ Hadronic and forward calorimeters, followed by the muon system. i. Three independent analyses have been pursued to measure the photon ID efficiency:
¤ Using photons from radiative Z llγ decays
¤ Extrapolating to photons the shower shapes observed for electrons from Z ee decays
¤ Directly measuring the efficiency on samples of reconstructed photons, after determining and subtracting the hadronic background with a technique based on track isolation.
z The SF obtained for photons are closer to unity than the ones obtained for electrons because of the correction of the shower shape variables in simulation applied on photons to account for the average data-MC difference on these variables.
ii. Identification efficiencies have been determined for prompt electrons produced in Z ee and in J/ψ ee events in data and in simulated samples.
¤ The precision of the SF reaches a few percent at low ET and is below 1% at high ET
¤ Discrepancies of around 5% between data and simulation are due to known mismodelling of the shower shapes and some of the track properties by the simulation. The efficiencies are found to be robust with respect to the number of primary vertices, in the range probed by the available data.
¤ The ID algorithms based on a multivariate likelihood discriminator have been optimised such that only a small dependence on the pileup condition remains and will be investigated in the future.
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 0.6d |)0
aconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 1.37d |)0.6
aconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 1.81d |)1.52 aconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 2.37d |)1.81
aconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 0.6d |)0
aunconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 1.37d |)0.6
aunconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 1.81d |)1.52
aunconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
[GeV]TE30 100 200 1000
com
bine
d SF
0.94
0.96
0.98
1
1.02
1.04
1.06| < 2.37d |)1.81
aunconverted
-1 = 13 TeV, 3.2 fbs
ATLAS Preliminary
¤ The difference between the simulation and the data-driven measurements is taken into account by computing data-to-MC efficiency ratios, also referred to as scale factors (SF).
¤ The SF are computed separately for each method and then combined.
¤ Most SF values are close to unity. This confirms that the simulation, with the applied corrections, provides a good description of the photon shower shapes in the collision data
[1] ATLAS Collaboration, The ATLAS Experiment at the CERN Large Hadron Collider, JINST 3 (2008) S08003. [2] ATLAS Collaboration, ATLAS Insertable B-Layer Technical Design Report, CERN-LHCC-2010-013, 2010, url: https://cds.cern.ch/record/1291633. [3] ATLAS Collaboration, Electron efficiency measurements with the ATLAS detector using the 2015 LHC proton-proton collision data, ATLAS-CONF-2016-024, 2016. [4] ATLAS Collaboration, Photon identification in 2015 ATLAS data, ATL-PHYS-PUB-2016-014, 2016.
The measurement of the efficiencies of the electron identification and isolation cuts are performed with the data using tag and probe techniques with large statistics sample of Z ee and J/ψ ee decays. ¤ This method consists on passing a strict selection on one of
the electron candidates (“tag”) together with the requirements of the di-electron invariant mass, which allows for a loose pre-identification of the other electron candidate (“probe”).
¤ The probe is used for the measurement of the identification efficiencies, after accounting for the residual background contamination.
¤ The low ET range (from 7 to 20 GeV) is covered by J/ψ ee and suffers from a significant background contamination, while Z ee events are used for measurements above 15 GeV.
For Run-II, several changes to the input variables used for electron identification (ID) have been introduced [3]. ¤ Taking advantage of the IBL, the number of hits in this innermost pixel layer is used for
discriminating between electrons and converted photons. ¤ The change in the TRT gas led to modifications in the detector response and prompted the
introduction of a new discriminating variable in the electron ID algorithms.
Electron ID efficiencies in Z ee events as a function of transverse energy ET, pseudo-rapidity η and the number of reconstructed primary vertices.
Data-to-MC ratio measured in J/ψ ee events with the τ-fit and τ-cut methods (left), using Z ee events with the Zmass and Ziso methods (right) and the combined result in the
overlapping region, 15 GeV < ET < 20 GeV (middle).
K.#Brendlinger WZ#Overlap#Removal#Studies 4
G
)
�
EHDP�D[LV�
EHDP�VSRW
SL[HOV6&7
757�����OD\HUV�
ILUVW�OD\HU��VWULSV�
VHFRQG�OD\HU
WKLUG�OD\HU
ij
Ș
KDGURQLF�FDORULPHWHU
HOHFWURPDJQHWLF�FDORULPHWHU
5
5
LQVHUWDEOH�E�OD\HU
+7Ș
ij
SUHVDPSOHU
electromagnetic calorimeter
second layer
first layer (strips)
presampler
third layer hadronic calorimeter
Rφ
Rη
TRT (72 layers)
SCTpixels
insertable b-layer
beam spot
beam axis
d0
eProbabilityHT
η
φ
∆η×∆φ = 0.0031×0.098
∆η×∆φ = 0.025×0.0245
FHT
TRT pe/(pe+pπ)TRT pe/(pe+pπ)
TRT PID
η
2− 1.5− 1− 0.5− 0 0.5 1 1.5 2
scal
e-fa
ctor
0.75
0.8
0.85
0.9
0.95
1
1.05 ATLAS Preliminary < 30 GeVT25 GeV < E
Tight syst)⊕Ziso (stat
syst)⊕Zmass (stat syst)⊕Zmass/Ziso (stat
η
0 0.5 1 1.5 2 2.5
scal
e-fa
ctor
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4 ATLAS Preliminary < 10 GeVT7 GeV < E
syst)⊕-fit (stat τ ψJ/ syst)⊕-cut (stat τ ψJ/ syst)⊕ cut/fit (stat ψJ/
Tight
η
0 0.5 1 1.5 2
scal
e-fa
ctor
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4 ATLAS Preliminary < 20 GeVT15 GeV < E syst)⊕ee (stat →Z
syst)⊕ and Z (stat ψJ/ syst)⊕ee (stat →ψJ/
Tight
i. Zmass : uses the invariant mass of the tag-probe pair as the discriminating varaible for background subtraction ii. Ziso : uses the calorimetric isolation distribution of the probe electron.
i. τ-cut: applies an explicit cut on the τ distribution to enrich the selected sample with prompt J/τ. ii. τ-fit: uses the full sample and extracts the non-prompt fraction by fitting the pseudo-proper time distribution before and after applying the ID cuts.
Z ee methods J/ψ ee methods
The identification (ID) of prompt photons and the rejection of background coming mostly from photons from hadron decays relies on the high granularity of the ATLAS calorimeter [4]. There are two levels of ID: ¤ The loose ID exploits the discriminating variables (DV) only in the hadronic calorimeter and in
the electromagnetic calorimeter second sampling layer, providing a highly efficient selection with fair background rejection, typically used for the trigger and background studies.
¤ The tight ID level exploits the full granularity of the electromagnetic calorimeter and applies tighter requirements also on the DVs used by the loose ID.
In all cases, the requirements on the DVs are tuned separately for converted and unconverted photons.
In Run-II, the electron ID is based on a likelihood discrimination combining calorimeter and track
variables to separate isolated electron candidates from candidates originating from photon
conversions, hadron misidentification and heavy flavour decays. In addition, isolation variables are
used as further handles to separate signal and background.
[GeV]TE20 30 40 100 200 1000
(tig
ht)
ID¡
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
| < 0.6d |)0
aunconverted
-1 = 13 TeV, 3.2 fbsATLAS Preliminary
Electron Extrapolation
Matrix Method
a llAZ
The photon ID efficiency increases from 53-64%
(47-61%) for unconverted (converted) photons at
ET ~ 10 GeV to 88-92% (96-98%) for ET ≥ 100 GeV.
Large fluctuations of the radiative Z decays are
expected due to their large statistical uncertainties
[GeV]TE20 30 40 100 200 1000
(tig
ht)
ID¡
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
| < 0.6d |)0
aconverted
-1 = 13 TeV, 3.2 fbsATLAS Preliminary
Electron Extrapolation
Matrix Method
a llAZ
Comparison of the data-driven measurements of the identification efficiency for unconverted photons (left) and
converted photons (right)
Several methods are used to measure with data the efficiency of the photon identification requirements, to cover a broad energy spectrum: ¤ At low energy, radiative photons from Z l+l-γ decays are selected by placing kinematics
requirements on the dilepton pair, on the invariant mass of the three particles in the final state and on the quality of the two leptons
¤ In the medium energy range, similarities between electrons and photon showers are exploited using Z ee decays and photon ID efficiencies are obtained using a “tag-and-probe” method
¤ At high energy, inclusive photon samples are used by applying the matrix method, which classifies the photons between prompt and background photon candidates, passing or not tight ID criteria
[GeV]TE20 30 40 50 60 70 80
Effic
ienc
y
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1Preliminary ATLAS
(2016)-1 = 13 TeV, 8.8 fbsLooseMediumTight
Data: full, MC: open
η2− 1.5− 1− 0.5− 0 0.5 1 1.5 2
Effic
ienc
y
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1Preliminary ATLAS
(2016)-1 = 13 TeV, 8.8 fbs > 15 GeVTELooseMediumTight
Data: full, MC: open
Number of primary vertices5 10 15 20 25 30
Effic
ienc
y
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1Preliminary ATLAS
(2016)-1 = 13 TeV, 8.8 fbs > 15 GeVTELooseMediumTight
Data: full, MC: open
[GeV]TE10 20 30 40 50 60 70 80 90 100
(tig
ht)
IDε
0.5
0.6
0.7
0.8
0.9
1
1.1 PreliminaryATLAS-1 = 11.6 fbint = 13 TeV, Ls
|<2.37η|<1.37 or 1.52<|η|γUnconverted
Data 2016 MCγll→Corrected Z
[GeV]TE10 20 30 40 50 60 70 80 90 100
SF
0.9
1
1.1
[GeV]TE10 20 30 40 50 60 70 80 90 100
(tig
ht)
IDε
0.5
0.6
0.7
0.8
0.9
1
1.1 PreliminaryATLAS-1 = 11.6 fbint = 13 TeV, Ls
|<2.37η|<1.37 or 1.52<|η|γConverted
Data 2016 MCγll→Corrected Z
[GeV]TE10 20 30 40 50 60 70 80 90 100
SF
0.9
1
1.1Comparison of the radiative Z boson data-driven efficiency measurements of (un)converted photons to the Z llγ simulation
as a function of the ET, in the region 15 ≤ ET ≤ 100 GeV