Electron reconstruction and identification in CMS at LHC on behalf of CMS collaboration 1
Outline The CMS detector @ LHC Electron reconstruction components Performance Electron identification Measurements from data 2
CMS Detector SUPERCONDUCTING COIL CALORIMETERS HCAL Plastic scintillator/brass sandwich ECAL Scintillating PbWO4 crystals 75848 Xtals TRACKER 4 dees 36 supermodules PIXEL 3 layers (barrel) 2x2 disks (fwd) SST >8 hits, depending on η η <2.5 MUON BARREL Drift Tube Resistive Plate Cathode Strip Chambers Total weight : 12,500 t Overall diameter : 15 m Overall length : 21.6 m Magnetic field : 3.8 Tesla η <2.6 Modules 2x5 Xtal 3
Why and How Physics motivation: electrons to be measured in challenging kinematics and/or background conditions Higgs searches (H ZZ* 4e) BSM (TeV resonances, SUSY) SM processes (top, Z 2e, W eν) Main problem: bremsstrahlung radiation the tracker material budget in front of ECAL 3.8 T magnetic field 4
Electron reconstruction strategy Combination of the infos from ECAL and tracker Pre-selection Super Cluster Electron Electron Electron Electron Trajectory Seed Seed TrackCandidate Gsf track candidate Seeding: - Track seeds - SC driven pixel match filter - E T, H/E Trajectory building: - CTF builder - Electron loss modeling - No χ 2 cut - Reduced #candidates/layer Gsf track fit: - Electron loss modeling - Mode of the gaussian mixture used for p ele - Brem fraction 5
Energy clustering in ECAL Bremsstrahlung recovery search for the highest E T crystal narrow η - larger φ window around the seed superclusters are built by collecting clusters of crystals in the road Energy estimation sum of the energies of the crystals in the supercluster Position estimation energy weighted mean position of the crystals in the supercluster 6
Electron seeding Start from energy weighted mean position of super-cluster External object used to filter tracker seeds Propagate back through the B field to the two corresponding layers implicit E T /p T match Both charge hypotheses considered Same algorithm for the online (HLT) and offline selection CMS pixel detector Reduced pixel detector (discarded) HLT 2.5 HLT 2.5 7
Electron track reconstruction Starts electron tracking with seeds from SC driven match filter The Bethe-Heitler parametrization is used for the energy loss Loose χ 2, a minimum of 5 hits is required The Gaussian Sum Filter is used for the forward and backward fit The track parameters are available for each position, use most probable value of the components pdf Hits collected up to the end Transverse momentum at vertex for p T = 10 GeV/c electrons 8
Electron pre-selection After the GSF track fit, track-supercluster associations from the pixel match are preselected to build candidate electrons Aim is to be as efficient as possible with fake rate compatible with affordable data size The pre-selection is suited to any physics analysis involving primary electrons a minimum transverse energy: E T > 4 GeV/c 2 an η, φ geometrical matching: Δη < 0.02, Δφ < 0.1 a cut on hadronic energy behind cluster: H/E < 0.2 9
Performance Meaningful estimation of the track parameters at the outermost state Additional possibilities for ECAL/tracker matching Define the bremsstralung fraction f brem = (p last - p @vertex )/ p @vertex Δp correlated with brem ECAL and Tracker estimates are combined to obtain the final electron momentum magnitude the weighted mean with weights defined as the normalized inverse of the variance of each measurement when both measurements are in agreement the corrected ECAL energy or the momentum estimate for the other cases Barrel 10
Electron identification Electron identification for startup achieved using cut based identification based on simple and well understood variables Inital selection performed on 4 variables mainly insensitive to tracker misalignment: H/E, σ ηη, Δη in, Δφ in Two flavours: loose and tight Same variables also suitable for high energy electrons H/E σ ηη 11
Category based eid Introduce a classification to achieve a good separation between real electrons and fakes Divide in region the f Brem and E/p plane and define 3 classes of electrons (barrel/endcap) low-brem electrons bremming electrons bad track H/E, σ ηη, Δη in, Δφ in, E/p used to discriminate Two flavours: loose and tight in E/p 5 4.5 4 3.5 Z! ee (endcap) bremming low-brem bad track in E/p 5 4.5 4 3.5 di-jet (endcap) bremming low_brem bad track 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 0.5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 fbrem 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 fbrem 12
Efficiencies from data strategy Efficiency of trigger, reconstruction and selection steps can be measured from data using the mass constraint on the Z peak One leg with tight criteria to tag the event The other leg required to pass loose criteria and used to determine efficiencies Electron selection: HLT single electron Fiducial region cut: η < 1.44, 1.56 < η < 2.5 Electron ID: «robust» selection Track isolation: No track p T > 1.5 within 0.02 < ΔR < 0.2 E T > 20 GeV/c 2 Invariant mass: 80-100 GeV/c 2 13
Efficiencies from data results Trigger efficiency versus supercluster η Electron pre-selection efficiency versus supercluster E T 14
Summary Electron reconstruction in CMS is well established Supercluster reconstruction Electron seed filtering Track propagation Simple and well understood discriminating variables for selection of primary electron for physics Established methods to measure the electron efficiencies from data 15
Material from data Location from X-ray of the detector using conversions Amount from variables sensitive to material E/p distribution Use brem fraction from GSF e-tracks <X/X0> ~ -ln(1-f brem ) ~2% precision on X/X0 16