Jet Energy and Resolution at the Tevatron Andrew Mehta YETI meeting, 7/1/2008
7/1/2008 Andrew Mehta, YETI 08 2 Outline Introduction CDF +D0 experiments and calorimeters Jets CDF Jet Energy Scale method D0 Jet Energy Scale method Cross check of jet energy scale
7/1/2008 Andrew Mehta, YETI 08 3 Motivation (1) Knowledge of Jet Energy Scale (JES) is fundamental for hadron colliders All physics processes involve jets that span a wide E T range [0,√s/2] Important for SM measurements … Jet Energy Scale uncertainties are dominant for high P T jets Inclusive jet cross section
7/1/2008 Andrew Mehta, YETI 08 4 Motivation (2) … also most of Non-Standard Model signatures (i.e. squark-gluino production) involve jets and Missing Transverse Energy (MET) MET must be corrected for jet energy measurements. Missing E T Multiple jets Correction ~ 12% at low MET
7/1/2008 Andrew Mehta, YETI 08 5 Peak luminosity above 2.0 *10 32 cm -2 s -1 Integrated luminosity/week about 25 pb -1 Highest-energy accelerator currently operational pp collisions at √s=2 TeV ~ 3.0 fb -1 on tape Tevatron RunII
7/1/2008 Andrew Mehta, YETI 08 6 Time A jet is a composite object: complex underlying physics depends on jet definitions Use different kind of Jet algorithms: - Cone algorithms (JETCLU and MIDPOINT) - K T algorithm Corrections on Jet Energy Scale (JES) for different effects: Instrumental effects: - response to hadrons - poorly instrumented regions - Multiple p-pbar interactions Physics effects: - Underlying event - Hadronization Jet reconstruction
7/1/2008 Andrew Mehta, YETI 08 7 CDF Calorimeter Central and Wall ( | |<1.2 ): Granularity: = 15° × 0.1 (~ 800 towers) Non compensating non-linear response to hadrons Rather thin: 4 interaction lengths Small amount of noise Resolutions: - EM energies ( ,e): /E T = 13.5%/√E T +1.5% - HAD energies( ± ): /E T = 50%/√E T +3% Plug (1.2<| |<3.6): Similar technology to the central Resolutions: - EM energies ( e /E = 16%/√E+1% - HAD energies ( ± ): /E = 80%/√E+5% Thicker than central: 7 interaction lengths
7/1/2008 Andrew Mehta, YETI 08 8 D0 Calorimeter LAr sampling U absorber: Compensating linear response to hadrons 7 interaction lengths Same structure for barrel and plug Resolutions: - EM energies ( ,e): /E T = 15%/√E T +0.3% - HAD energies( ± ): /E T = 45%/√E T +5%
7/1/2008 Andrew Mehta, YETI 08 9 Calorimeter calibration: EM energy For EM energy response use: MIP peak when possible (at about 300 MeV) Z e + e - mass peak stability - Set absolute EM scale in central and plug Check calorimeter response: Use test beam (from 1980s!) and single particles measured in-situ to understand absolute response Check time dependence
7/1/2008 Andrew Mehta, YETI Calorimeter calibration: Hadronic Energy For hadron energy response use Minimum Ionizing Particles (MIP): - J/ and W muons - Peak HAD calorimeter: ~ 2 GeV Also Minimum bias events: - E.g. N towers (E T >500 MeV) Syst. Uncertainty related to Calorimeter Calibration ~ 0.5%
7/1/2008 Andrew Mehta, YETI CDF Jet Energy Scale Method Different correction factors: (f abs ) Absolute Corrections Calorimeter non-linear and non-compensating (f rel ) Relative Corrections Make response uniform in all corrections are then referred to the central region (MPI) Multiple Particle Interactions Energy from different ppbar interaction P T jet particle (R) = [ P T jet raw (R) f rel (R) – MPI(R)] f abs (R) P T parton (R) = P T jet particle (R) - UE(R) + OOC Systematic uncertainties are associated with each step Additional corrections to get to parton energy: (UE) Underlying Event Energy associated with spectator partons in a hard collision Hadron-to-Parton correction (historically defined as Out-Of-Cone)
7/1/2008 Andrew Mehta, YETI CDF Absolute Corrections Use MC simulation to determine Jet Corrections MC is adjusted by comparison with data to: simulate accurately detector response to single particle (E/p). describe jet fragmentation
7/1/2008 Andrew Mehta, YETI CDF single particle response simulation Jet composition: ~ 70 % charged particles - 10% protons - 90% pions 30 % neutral pions ( ) - EM response Remaining difference data/simulation taken as syst. uncertainty hadrons
7/1/2008 Andrew Mehta, YETI CDF fragmentation MC simulation needs to reproduce well data: Due to non-linearity of the calorimeter, non trivial correlation N particles and P T track spectra: - one 10 GeV pion: ~ 8 GeV - ten 1 GeV pions: ~ GeV Very important a good understand of track efficiency Measurement of jet shape is fundamental Integrated jet shape Data/MC different = Systematic uncertainty ~ 1%
7/1/2008 Andrew Mehta, YETI CDF Absolute Correction Almost independent on jet cone size. Depends on transverse momentum: calorimeter response is ~ 70% for 25 GeV/c jets, ~ 90% for 400 GeV/c jets. Absolute correction factor
7/1/2008 Andrew Mehta, YETI CDF Relative Corrections cracks Use dijet events. Jet corrections relative to the central calorimeter: Central (0.2<| |<0.6 jets) ~1 by definition (reference) Difference Data/MC mainly in the forward region Depends on E T jets considered
7/1/2008 Andrew Mehta, YETI CDF Multiple Interactions Overlapping interactions can overlap the jet Number of extra interactions depends on luminosity Energy offset depends on number of interactions
7/1/2008 Andrew Mehta, YETI CDF Multiple Interaction corrections Linear correlation between number of interactions and number of vertices Define random cones in the central region (0.2<| |<0.6) and determine average transverse energy associated to a cone Cone-based method: should improve to make it more general (K T ?) For cone R = 0.7, = 1.06 GeV
7/1/2008 Andrew Mehta, YETI CDF Model-dependent corrections Underlying event (UE) and Hadron-to-Parton (Out-of- cone, OOC) energy corrections used only if need parton energy Modelling is required, difference MCs as systematic uncertainties. Parton transverse momentum: P T parton (R) = P T jet particle (R) - UE(R) + OOC
7/1/2008 Andrew Mehta, YETI CDF Underlying event Particle jet could have contributions note related to hard interaction: Beam-beam renmants Initial state radiation MC tuned on Data (as Pythia Tune A) Use di-jet events
7/1/2008 Andrew Mehta, YETI CDF Out-of-Cone Correction OOC energy: energy escaping the cone radius Gluon radiation (FSR) Obtained from Pythia di-jet samples: Ratio P T parton / P T jet particle Similar performance Pythia and Herwig Systematic uncertainties from photon+jet events: Assume P T = P T jet corr. Difference Data/MC of energy inside annuli around jet axis taken as systematic uncertainty
7/1/2008 Andrew Mehta, YETI D0 Jet Energy Scale Method Different correction factors: (f abs ) Absolute Corrections Calorimeter non-linear and non-compensating (f rel ) Relative Corrections Make response uniform in all corrections are then referred to the central region (O) Offset correction For MPI, underlying event and detector noise (S) Showering correction For detector effect of energy leaking inside or outside of jet cone E T jet particle = [ E T jet raw -O] / (f rel f absl S) Note D0 correct to a particle level with corrections for underlying event, but not for out of cone corrections (different from CDF).
7/1/2008 Andrew Mehta, YETI D0 Offset Energy Corrects for all energy not associated to the hard scatter: MPI, underlying event and electronic noise Worked out from minimum bias events
7/1/2008 Andrew Mehta, YETI D0 Relative Corrections Use dijet and photon-jet events. Jet corrections relative to the central calorimeter | |<0.6 : Depends on E T jets considered due to crack
7/1/2008 Andrew Mehta, YETI D0 Absolute Correction Performed with photon-jet events Similar corrections for different η →shows relative corrections ok Absolute correction factor
7/1/2008 Andrew Mehta, YETI D0 Showering Correction Use MC to estimate energy smeared in or out due to detector effects (this is absorbed in the absolute corrections at CDF) Checks with data to evaluate the systematic error Does not account for true energy from the parton distributed outside the jet radius (OOC corrections at CDF)
7/1/2008 Andrew Mehta, YETI JES Systematic uncertainties Total systematic uncertainties for JES between 2 and 3% as a function of corrected transverse jet momentum Similar between CDF and D0 apart from out of cone correction, which is very large at low Pt for CDF CDF
7/1/2008 Andrew Mehta, YETI CDF + jet p T balance E T leading jet > 25 GeV E T (second jet) < 3 GeV (Jet- ) > 3 p T balance: Agreement Data/MC within 3% Sensitive to radiation effects when allow second jet: Herwig farther away from jet cone Data Pythia Herwig Used to test procedure – not used in calibration
7/1/2008 Andrew Mehta, YETI CDF check of scale with Look at dijet mass resonances to check b jet energy scale Trigger on two displaced tracks+ two 10 GeV jets DisplacedVertex tag, SecondryVertex Mass to select b-jets, kinematic cuts to improve S/B Fit signal and background ( direct QCD production ) templates, for varying JES Jet energy scale: ± ( stat.) ± (sys.) (agreement with 1 sigma of nominal scale factor) DiJet Invariant mass GeV
7/1/2008 Andrew Mehta, YETI CDF Z-jet p T balance Similar Herwig behaviour for Z+jet w.r.t. +jet but less visible Selection two e( ) with E T >18 GeV (p T >20 GeV) 76 < M ee( ) < 106 GeV E T leading jet > 25 GeV E T (second jet) < 3 GeV (Jet- ) > 3 These events allow us to reach lower PT than photon+jet and also cross check photon+jets results.
7/1/2008 Andrew Mehta, YETI CDF check of calibration from W Very difficult to see inclusive decays of W and Z in jets Best possibilities: - W from top decays
7/1/2008 Andrew Mehta, YETI Summary and Conclusions Hadron colliders are a very challenging environment to measure the jet energy scale Lack of simple clean processes, gluon radiation, multiple interactions, underlying event etc. make it tough. 2 very different methods adopted by CDF and D0 Gives about a 3% error on the jet energy scale Checks of various signals give faith in this scale and error
Back-up
7/1/2008 Andrew Mehta, YETI CDF uncertainties on calorimeter simulation Improvement possible with higher statistical samples Sensitive to 0.9x0.9 = 81% inner part of the tower. For tower boundaries: additional 10% uncertainty Total uncertainties:
7/1/2008 Andrew Mehta, YETI CDF single particle response simulation Single particle response Test beam In situ: Select ‘isolated’ tracks Measure energy in tower behind them Dedicated trigger Bgk subtraction Tune simulation to describe E/p distribution at each p
7/1/2008 Andrew Mehta, YETI CDF jet resolution (H1 Algorithm) Apply relative corrections to make response flat in η. Use tracks (0.5<Pt<15 GeV, Pt ordered), extrapolate to face of calorimeter Select towers within Δη=0.1 and Δφ=0.2. (Central towers are 0.1x0.26.) Take the nearest tower one if none within these limits. Order selected towers in distance from the track. Remove towers such that corresponding removed energy is always less or equal to the energy of the track. Energy already removed by a previous track is not considered by subsequent tracks. Jet is sum of all quality-selected tracks and remaining towers in the jet. Scale the final jet energy There is improvement (10-15%) but need much more work for optimization.
7/1/2008 Andrew Mehta, YETI Jet Algorithms
7/1/2008 Andrew Mehta, YETI Clusters using different Jet algorithms
7/1/2008 Andrew Mehta, YETI Lateral profile
7/1/2008 Andrew Mehta, YETI Lateral profiles scan
7/1/2008 Andrew Mehta, YETI Calorimeter simulation Use MinBias or isolated track trigger Select good tracks within central 81% of tower. No extra track within 7x7 towers, no ShowerMax cluster. Measure E/p in data Tune Gflash parameters Difference in data and simulation is taken as uncertainty. E(EM)/p E(HAD)/p E(Total)/p After BG subtraction More statistics!
7/1/2008 Andrew Mehta, YETI Photon+jet balancing P T balance between photon and jet is about 3% different among data and MC. Δφ >3, second Jet Pt<3 GeV Δφ > 3, No 2nd jet cut Herwig Pythia Data Herwig Pythia Data
7/1/2008 Andrew Mehta, YETI Photon/Z – jet balance
7/1/2008 Andrew Mehta, YETI Calorimeter simulation improvements Tower-phi boundaries improved with new Simulation from 10% uncertainty to less than 5% Old simulation New Simulation
7/1/2008 Andrew Mehta, YETI Lateral profile Measure E/p signal in 5 towers adjacent in signal defined as 1×3 strip in φ Plot E/p vs. relative eta for 5 towers In Gflash, use formula for lateral profile EM and HAD calorimeter probe different parts of the hadronic shower excluding 90° crack E/p vs η rel (Central)
7/1/2008 Andrew Mehta, YETI Jet corrections to particle level (absolute) Depends on MC simulation and how well data are reproduced, and on fragmentation Main uncertainties due to calorimeter simulation Monte Carlo simulation used to compare measured (calorimeter) jets and particle (hadron) jets.
7/1/2008 Andrew Mehta, YETI Possible improvements Absolute energy scale: Better simulation translate in lower systematic uncertainties: Old simulation (response as function of track momentum): [0-12] GeV 2.5%, [12-20] GeV 3%, [20,+] 4% New simulation (under study): 2% expected in the whole p range This would reduce absolute JES uncertainty from % to 1.4% Specific b-jet correction Using Z bbbar or photon+b Jet resolution for higgs analysis H1 algorithm: use tracking information for energy determination of charged hadrons
7/1/2008 Andrew Mehta, YETI Cross-check using prompt photons Photons are well measured in EM calorimeter Complications: number of events at high E T very low From D0 measurement, 40 evt. with L=1 fb -1 and E T > 300 GeV Background due to 0 Purity % for [20-100] GeV photon transverse energy range In CDF: use photon+jets (but also Z+jets) for cross check and to evaluate OOC corrections and JES systematic uncertainty due to Data/MC differences.