Electron and Photon Reconstruction and Performance in CMS

Home , Electron

L. Finco

University of Nebraska – Lincoln

LPC Physics Forum 18th of February 2021 Outline 2

Calorimeters at Colliders

ECAL Detector at CMS

Electrons and Photons in Run 2

… and Beyond 3

Calorimeters at Colliders Definition of Calorimeter 4

• In particle physics a calorimeter is a detector measuring the energy carried by an incoming particle • Instrumented blocks of matter in which the particle interacts and deposits all its energy in the form of a cascade of particles

• The particle energy is measured in eV (MeV-GeV-TeV 106, 109, 1012 eV) • 1 eV = energy acquired by one electron accelerated by 1 V • The temperature effect of a 100 GeV particle in 1 litre of water -12 (at 20 °C) is ΔT = 3.8×10 K https://www.mpp.mpg.de/~menke/elss/home.shtml General Features of Caloremeters 5

• Calorimetry is a “destructive” method • The energy is measured by total absorption of the particle • A calorimeter converts the energy of an incident particle to a detectable signal, proportional to the incoming energy • Calorimeters can measure the energy of electromagnetic radiation (electrons, photons) or hadronic particles (charged pions, protons, neutrons) • The charged component of the particles' cascade deposits energy in the active part of the calorimeter and it is detected in the form of charge or light • This talk is mainly focused on electromagnetic calorimeters

https://www.mpp.mpg.de/~menke/elss/home.shtml Particles’ Life through a Detector 6

https://cds.cern.ch/record/2120661 Particle-Matter Interactions 7

• In matter electrons and photons loose energy interacting with nuclei and atomic electrons • Main photon interactions with matter: • Photoelectric effect • Compton scattering • Pair production • Main electron interactions with matter: • Ionization PDG 2018 • Bremsstrahlung • Čerenkov radiation • Multiple scattering

PDG 2018 Particle-Matter Interactions 8

• In matter electrons and photons loose energy interacting with nuclei and atomic electrons

• Main photon interactions with matter: Pair production • Photoelectric effect • Compton scattering • Pair production • Main electron interactions with matter: • Ionization PDG 2018 • Bremsstrahlung • Čerenkov radiation • Multiple scattering

Bremsstrahlung At colliders interesting electrons and photons have usually E > 1 GeV

PDG 2018 Particle-Matter Interactions 9

• In matter electrons and photons loose energy Pair production interacting with nuclei and atomic electrons • Main photon interactions with matter: • Photoelectric effect • Compton scattering • Pair production • Main electron interactions with matter: • Ionization Courtesy of S. Harper • Bremsstrahlung • Čerenkov radiation Bremsstrahlung • Multiple scattering

These processes are the basis of the electromagnetic shower

Courtesy of S. Harper Particle-Matter Interactions 10

Pair production

Courtesy of S. Harper

Bremsstrahlung

These processes are the basis of the electromagnetic shower

Courtesy of S. Harper Electromagnetic Shower 11

• Electrons and photons create a cascade (shower) of particles • Number of particles N • Is proportional to the energy of the incoming particle

• Increases until the energy of the electron component falls below Ec (critical energy ~ 5-10 MeV)

• With E < Ec a slow decrease in number of particles occurs as electrons are stopped and photons absorbed

• Longitudinal extension of the shower is proportional to ln(E/ Ec) Electromagnetic Shower 12

Determined by radiation length Determined byMoliere radius

• The evolution of an electromagnetic shower through a material is dictated by:

• Radiation length (X0) : • Mean distance in which an electron will lose all but 1/e of its energy • 0.78 of mean free path of a photon (photon showers will start later)

• Moliere radius (MR): • 90% of shower’s energy is contained within a cylinder of radius = 1 MR • 2 MR contain 95% of its energy Energy Resolution of EM Calorimeters 13

Intrinsic energy resolution: 퐸 ∝ 푁 휎 푁 ∝ 푁

휎(퐸) 1 ∝ 퐸 퐸

Energy resolution of real detectors

휎(퐸) 푆 푁 = ⨁ ⨁퐶 퐸 퐸 퐸

S: stochastic term from Poisson-like fluctuations

N: noise term from electronics and Example using real data pile-up C: constant term Calorimeter’s Design Characteristics 14

• Size of the detector: the longitudinal depth of the detector is dictated by

X0 and is driven by the required resolution • Small X0 allows to reduce calorimeter size (compactness) • Cell size: typically chosen such that 70-80% of energy of a centrally incoming particle is deposited in the cell, while having energy in the neighbouring cells large enough to measure the centre of gravity coordinates

• Related to MR • Granularity: size of detector elements in the transverse and longitudinal direction, which determines the ability to resolve two showers induced by nearby particles as distinct. It depends on the shower size, distance from the interaction point and transverse size of detector elements/cells 15

ECAL Detector at CMS Large Hadron Collider 16 Experimental Conditions at the LHC 17

6.5 TeV beam energy 1011 protons/bunch 1034 cm-2 s-1 machine luminosity 40.000.000 bunch crossing/second More than 30 events/crossing

CMSPublic/LumiPublicResults LHC Requirements for Calorimeters 18

• Fast response (25 ns or faster) and high granularity, to reduce pile-up induced noise • Radiation-hard detectors and electronics • Hermetic and cover the full azimuthal angle and rapidity range, to tag very forward jets and well measure the missing energy • Excellent electromagnetic energy resolution • To detect the two photon decay of an intermediate mass Higgs (golden channel together with H→ZZ →4l)

• 푚훾훾 = 2퐸1퐸2(1 − cos 휃) • Uncertainty on 푚훾훾 determined by uncertainty on photon energy and direction Courtesy of C. Mariotti LHC Requirements for Calorimeters 19

• 푚훾훾 = 2퐸1퐸2(1 − cos 휃) • Uncertainty on 푚훾훾 determined by uncertainty on photon energy and direction

JHEP 11 (2018) 185 Compact Muon Detector 20

https://cms-docdb.cern.ch/cgi-bin/PublicDocDB/RetrieveFile?docid=13631&filename=cms_160312_06.png CMS Electromagnetic Calorimeter 21 휃 휂 ≡ ln(tan ) 2

• Lead Tungstate (PbWO4) homogenous crystal calorimeter • Barrel (EB): • 36 super-modules, each 1700 crystals Wikipedia • |η|< 1.48 • Endcaps (EE): • 2 endcap sides, each 7324 crystals • 1.48<|η|<3.0 • Preshower (ES): Barrel • Sampling calorimeter Endcaps (lead, silicon strips) Preshower • 1.65<|η|<2.6

CERN/LHCC 97–33CMS TDR 4 CMS Electromagnetic Calorimeter 22

Barrel Endcaps Preshower

CERN/LHCC 97–33CMS TDR 4 CMS ECAL Crystals 23

• Incident electron/photon generates EM shower in the heavy PbWO4 material • Charged particles in the shower produce scintillation light proportional to incident particle energy • Scintillation light detected by photodetectors with internal amplification

• ECAL is made of 75848 PbWO4 crystals

Parameter Value

Radiation length 0.89 cm 2.9 cm 2.9 Moliere radius 2.2 cm Rad. hardness excellent Refractive index 2.3 Peak emission 440 nm % of light in 25 ns 80% Light yield ~10 p.e./MeV ECAL Energy Resolution 24

ECAL “standalone” energy resolution measured at the test beam (3x3 arrays of barrel crystals) • No magnetic field • No material in front of the ECAL • Negligible inter-calibration contribution in the constant term

http://cds.cern.ch/record/1000388 25

Electrons and Photons Performance in Run 2 Electrons and Photons in CMS 26

• Electrons and photons are grouped together because they both create electromagnetic showers • The energy of e/γ object is measured in the electromagnetic calorimeter

CMS-PHO-EVENTS-2021-002 ECAL Energy Reconstruction 27

• Energy in the ECAL observed as a pulse as the photons from the shower arrive over time • This pulse shape is converted to an energy (E ∝ A, signal channel amplitude) • In case of out-of-time pile-up events, multiple pulses from different bunch 10.1088/1748-0221/15/10/P10002 crossings are generated • They lead to an apparent increased amplitude measurement • The amplitudes of different pulses are resolved by fitting the multiple pulse shapes simultaneously

10.1088/1748-0221/15/10/P10002 Electron and Photon Reconstruction 28

• First step is to construct individual particles from the reconstructed energy deposits (clustering) • The clustering procedure looks for local maxima above a given threshold (1 GeV) • Energies between overlapping adjacent clusters are shared Clustering of Crystals Energy 29 Unclustered energy Energy shared by two clusters

Clusters corresponding to electrons/photons • Found 5 clusters in this region of the ECAL • Each one represents the energy deposits of a particle (electron/photon) • In practice not this clean: often clusters have a very large number of crystals with small energy fraction Superclusters 30 Supercluster Incompatible with the supercluster

• An electron or photon when it arrives to ECAL, may be accompanied by multiple secondary particles created through interactions with matter on the way to ECAL • Superclustering (clustering on clusters) aims to combine the individual electrons and photons into a single object, corresponding to the original electron/photon • It starts by taking the highest energy cluster (in this case the light blue one) and looks for compatible clusters in 휙, 휂 Electron Track Reconstruction 31

• Electrons leave a signal in both tracker and calorimeter • Electron tracks have changing curvature because of radiative energy loss due to bremsstrahlung • A dedicated tracking algorithm known as GSF tracking is used to take into account these changes in the curvature • Used to associate additional bremsstrahlung radiation and photon conversion tracks to the supercluster

Bremsstrahlung Pair production

Courtesy of S. Harper Courtesy of S. Harper Refined Superclusters 32 Supercluster

Recover of a very soft bremsstrahlung photon into the refined supercluster • Refined superclusters use the information from the tracker, to be able to link bremsstrahlung emissions to missed ECAL deposits and reject some clusters which are highly incompatible with its matched tracks Energy Corrections 33

• Energy deposited by electrons and photons in the ECAL and collected in superclusters is subject to losses • Energy lost in gaps • Large amount of material upstream of the calorimeter • Calorimeter measurements more sensitive to pileup than tracking.

• To calibrate the reconstructed energy back to generated energy a correction procedure is needed • Based on simulation EGM-17-001 • Uses machine learning techniques • Applied on data and MC Reconstruction Performance 34

Electron reconstruction efficiency is higher than 95% for ET> 20 GeV and is compatible between data and simulation within 2%

EGM-17-001 Data/MC Residual Corrections 35

• Reﬁned supercluster calibration is MC-based • Residual data/MC discrepancies corrected using the Z mass and width, by comparing Z → ee events in data and MC • Simultaneously adjust energy scale (data) and resolution (MC)

Nee MC Nee MC

data data

mee mee Data/MC Residual Corrections 36

EGM-17-001 EGM-17-001

After energy corrections = refined supercluster calibration + scale corrections Energy Resolution 37

• Excellent data/MC agreement, after application of residual scales to data and smearings to simulated events

• Overall the energy resolution through Run 2 (2016-2018) between 1% and 3.4%

• Energy resolution measured in 2017 significantly better, thanks to the refined “Legacy” calibration

EGM-17-001 Electron and Photon Identification 38

• Several variables are developed to separate electrons/photons from background (jets, photon conversions, particles from secondary vertices) • They exploit that electrons/photons are single objects which are almost fully contained in the ECAL • Many different types: • Shower-shape variables • Track matching variables • Conversion ID variables • Isolation variables Electron and Photon Identification 39

• Several variables are developed to separate electrons/photons from background (jets, photon conversion, particles from secondary vertices) • They exploit that electrons/photons are single objects which are almost fully contained in the ECAL Are the energy deposits in • Many different types: the calorimeters compatible • Shower-shape variables with coming from a single • Track matching variables electron/photon? • Conversion ID variables Does the ECAL deposit • Isolation variables have a compatible track?

Is there a large amount Are the tracks compatible with coming of other particles from the collision point? Or do they nearby the appear later on in the tracker? electron/photon? 40 Shower Shape Variables: σiηiη

• σiηiη is one of the most important electron/photon ID variables in CMS • It measures the spread of an electromagnetic shower along 흶 direction • A 5x5 array of crystals is the area where an electron/photon is almost fully contained Most energetic crystal of 5x5 array

CMSEgammaRun3IDvariable

CHEP2015: Electron ID at the HLT 푤푖 non zero if Ei > 0.9% of E5x5 Shower Shape Variables: H/E 41

• H/E is the ratio of the hadronic energy to the electromagnetic energy • Excellent ID variable used in electron and photon identification • Very well modelled in simulation

ECAL HCAL

e/γ

jets

Courtesy of S. Mukherjee EGM-17-001 42 Conversion ID Variables: R9 • 5x5 matrix contains 96.5% (97.5%) of unconverted photon energy in EB (EE)

• R9 is the energy sum of the 3×3 crystals centred on the most energetic crystal in the supercluster divided by the energy of the supercluster

• R9 helps in conversions identiﬁcation and to distinguish real photons from π0

10.1088/1748-0221/10/08/P08010 Performance of ID Variables 43

Photons

• Several selections are derived based on these variables • MVA and cut-based approaches • Several working points • Different selections barrel/endcaps

EGM-17-001 Performance of ID Variables 44

• Data/MC correction factors derived for each selection using tag-and-probe • Data/MC agreement improved from ~88% to 95% when using Legacy calibration in 2017 • Correction factors stable within 3% over the full Run 2

EGM-17-001 EGM-17-001 45

Electrons and Photons Performance beyond Run 2 Run 3 Challenges 46

• Run 3 is going to start in mid 2022 (~2 years) and collect about 200 fb-1 • It is going to bring a harsher environment for electromagnetic object reconstruction and identification compared to Run 2: • Pile-up interactions increase (slightly) • Noise in the ECAL increases because radiation damage, especially in the endcaps by a factor 1.8, 3 and 4 at |η|=1.5, 2.5 and 3 • Discrimination power of ID variables affected by high noise • Mitigate the effect by noise cleaning

CMSEgammaRun3IDvariable 47 Re-definition of ID Variables: σiηiη

• Redefinition of ID σiηiη variable sensitive to ECAL noise

• Absolute cut on Ei above noise threshold to remove spurious crystals

• Applied to both electrons and photons 푤푖 ∝ Ei • More effective in endcap

CMSEgammaRun3IDvariable CMSEgammaRun3IDvariable 48 Re-definition of ID Variables: H/E, R9

• New design of HCAL detector with replacement of readout electronics allows depth segmentation • Combine information of energy deposited in different HCAL layers • Better shower-development measurement and background rejection for electrons and photons

2016 2018

CERN-CMS-DP-2018-019 CERN-CMS-DP-2018-019

• New definitions of R9 under study, to stabilize it over the years and optimize its separation power between unconverted and converted photons Summary and Conclusions 49

• The performance of electron and photon reconstruction and identiﬁcation in CMS during LHC Run 2 data-taking period has been presented • Electrons and photons create electromagnetic showers and deposit all their energy in the electromagnetic calorimeter • Reconstruction and identification algorithms have been described and show very good performance • Excellent energy resolution ([1.5-4]%) remains stable throughout the full Run 2 and is a key ingredient of the CMS physics program • Prospects for the upcoming Run 3 have also been shown, focusing in particular on the re-optimization of the ID variables, necessitated by the harsher environment with respect to Run 2 References 50

Large part of the material taken from tutorials at CERN given by S. Harper and other E/gamma Conveners at CMS. Big thank to them!

Main references: • Calorimetry for Particle Physics, C.W. Fabjan and F. Gianotti, Rev. Mod. Phys., Vol. 75, N0. 4, October 2003 • Calorimetry, Energy Measurement in Particle Physics, R. Wigmans, Oxford University Press, 2000 • CMS Collaboration, The Electromagnetic Calorimeter Project, Technical Design Report, CERN/LHCC 97-23 (15 December 1997) 364pp • Calorimetry in particle physics experiments, Lectures by R. Arcidiacono • The CMS Collaboration, Electron and photon reconstruction and identiﬁcation with the CMS experiment at the CERN LHC, PAS-EGM-17-001 • The CMS Collaboration, Redefinition of Electron and Photon identification variables for Run-3 in the CMS Experiment, CMS-DP-2020-034 ; CERN-CMS- DP-2020-034 51

Backup The CMS Collaboration 52 CMS has over 4000 physicists, engineers, computer scientists, technicians and students from around 200 institutes and universities from more than 40 countries Trigger Selection and Performance 53

• Single and double electromagnetic objects at L1 (L1 seeds) • Information coming only from calorimeter detectors • No distinction between electrons and photons • Single and double electron/photon HLT selections • Correspond to the ﬁrst selection step of most ofﬂine analyses using electrons/photons • Must ensure a large acceptance for physics signals, while keeping the CPU time and output rate under control

• Can be very complex

Pierini Courtesy of M. M. of Courtesy Trigger Selection and Performance 54

• Trigger efficiency measured using Z → ee • Full Run 2 results stable vs pile-up and detector aging

• Efficiency of single electron trigger as a function of pT of offline electron • Double electron trigger efficiency, stable within 5% as a function of pile-up

EGM-17-001 EgammaFullRun2Data