Jet Calibration in ATLAS Mark Hodgkinson ATLAS UK Meeting, IPPP January 2008

Contents Introduction to detectors used in Jet reconstruction Hadronic calibrations for jets Validation of hadronic energy scale In Situ Calibrations Other contributions from UK Conclusions Far more work goes on than I am showing in this talk - as far as I know all UK jet performance work is in this talk… So this talk is biased to topics the UK is involved in…

Introduction Calorimeters used by default for jet finding Trackers also being used for wider variety of tasks in jet group nowadays Charged component seen in tracker Charged and neutral components seen in calorimeter Jet Finders ATLAS uses Cone 0.4 and 0.7 Also use Kt D = 0.4 and D = 0.6 Jets We measure jets kinematics via a spray of hadrons after the hadronisation process Ideally associate all hadrons to the correct parton uniquely

Calorimeter input to Jet Finding CaloTowers: Towers are fixed grid of 0.1 x 0.1 in eta and phi CaloTopoClusters: Define any CaloCells with |E| > as seed cells Add any neighbouring cells (3D) with |E| > to seed Repeat with new neighbours, until no neighbours pass Noise suppression built in Search for local maxima to decide if cluster needs splitting

Why We Need a Calibration Hadronic Showers complex: -Visible electromagnetic energy (electrons, photons,  0 decays) ~50% -Visible energy from ionisation ~25% -Invisible energy from nuclear interactions (excitation, break up) ~25% -Escaped energy (e.g. neutrinos) ~2% Also account for visible energy in dead material Jet response varies over the detector - e.g. in crack regions many particles cannot be detected Charged particles with low p T bent out of cone in calorimeter A Gupta (Chicago) M Hodgkinson (Sheffield) Hadronic Scale Particle Jet Scale Parton Scale Energy not included in reconstructed jet, that does come from the hadronisation Energy from underlying event included in reconstructed jet

Jet Calibrations  2 minimisation : Where Truth reference is generator level truth jet H1 Style (BNL) Local Hadron Calibration (Oxford,Stockholm,MPI Munich) Reference energy is true single pion energy (weights also tested on test beam data!) Takes us to hadronic scale, not particle jet scale Need additional corrections (magnetic field etc) on top - not done yet Classify clusters as EM or Hadronic Hadronic clusters get weights applied (12.0.6), dead material corrections (12.0.6) and out of cluster (not in ) corrections 0.4 <  < 0.5 Plots by C Issever, K Lohwasser (Oxford) and E Bergeaas (Stockholm)

Energy Flow with eflowRec  = 0 single  ± TDR: Tracking:  p T /p T  0.036%p T  1.3% Calo:  E/E  50%/  E  Charged particle response from single pion Monte Carlo In eflowRec remove energy due to charged pion showers on cell by cell basis via ordering principle Remaining clusters undergo local hadron calibration Local Hadron TopoCluster Energy Flow H1 TopoCluster M Hodgkinson, R Duxfield, D Tovey (Sheffield) Pt = GeV Cone 0.4, J1-J3 dijets, |  | < 1.8 Improves linearity, comparable width Improved width of jet energy resolution by discarding tracks matched to showers split into multiple clusters At 6 GeV 60% of pions have < 80% energy in leading cluster - work in progress to deal with this Also plan to derive H1 style weights Hadronic W mass resolution improved due to better linearity and angular resolution Used sample 5200 (semileptonic top) 4 Highest pt jets (H1 topocluster) p T > 50 GeV Refined missing ET > 40 GeV  /E = 3.54 ± 0.49 EFlow Local Hadron Calo Only  /E = 5.02 ± 0.61

Check of Hadronic Scale Use tracking to check if hadronic scale is correct in calorimeter Minimum Bias events cover 400 MeV -> 10 GeV Tau decays (W,Z) cover 10 GeV -> 140 GeV Overlapping showers a problem just like in energy flow for jets Have to suppress backgrounds (QCD di-jets and tau decays to neutral particles) Need isolated charged hadrons - isolation criteria means taus cannot be used -> Try for QCD M Hodgkinson QCD Di-jets Can select 7000 tracks from QCD jets in first year (i.e. use shower shape cuts, but not QCD rejection) Not enough CSC (full sim) data to investigate… Cannot use Atlfast due to calorimeter cuts used to select isolated charged hadrons J. Lu, D. Gingrich (Alberta): QCD J1 QCD Minimum Bias Use loose cuts and subtract background Background E/P measured from control sample N Davidson E Barberio (Melbourne) p T =10 GeV

In Situ Calibration Jet energy scale is not uniform in h Di-jet (Sheffield, Heidelberg and Argonne) samples are used Paul Hodgson (Sheffield) QCD di-jets Di-Jet Balancing Cut on  to ensure jets back to back (suppress ISR/FSR effects) pt balance = pt_probe/pt_ref Relative Jet Energy Scale can be determined to < 1% with 3 pb -1 data (assuming nominal jet prescales, J70 prescale is 5, do not change) 0.4 <  < 0.6 is reference region Paul Hodgson Plan to try technique in FDR P Hodgson  + Jet, Z + Jet (M Hurwitz - Chicago) W mass template method - J Schwindling (Saclay) Other Methods

In Situ Calibration at High Pt Di-jet,  + jet cross-sections are high pt is low Balance high pt jet with a number of calibrated low pt jets Glasgow, Tokyo, Heidelberg, Alberta 1% error on JES, if low pt jets well calibrated Systematic bias found by D.Clements, C. Buttar Partly due to unclustered soft activity - mitigated with larger cone sizes, lower jet seed p T Jet Balancing F Ruehr (Heidelberg) Cone >1120 GeV Pt D Clements, C Buttar Pythia standalone Truth Reco

UK Software K Lohwasser (Oxford), V Giangiobbe (INFN Pisa) Should be used to make standard jet performance plots JetPerformance Y-Splitter J Butterworth, A Davison, P Sherwood (UCL) Y scale is splitting scale into subjets Useful for finding e.g. hjgh p T W->qq Mid-Point Jet Algorithm S Thompson and C Cheplakov (Glasgow) Provides infrared safety Performance as good as standard algorithms Implemented in ATLAS software

Conclusions ATLAS pursuing a range of approaches to calibrating jet energy: - Initial calibration (H1 style, Local Hadron style + jet corrections, energy flow) - Validation of hadronic calibrations with E/P methods - Jet balancing (di-jet) to get uniform  response -  + jet (Z + jet), W->qq also useful - Then we can bootstrap calibration up to the very high pt jets - jet balancing favoured UK is providing useful software for use in Jets work (JetPerformance, Y Splitter, Mid Point algorithm, eflowRec)

Documents Local Hadron Calibration Performance - E Bergeaas, C Issever, K Jon-And, K Lohwasser, B King, D Milstead: ATL-COM-LARG Energy Flow for Jets - M Hodgkinson, D Tovey, R Duxfield: ATL-COM-PHYS Jet Fragmentation and E/P work - M Hodgkinson, J Lu, D Gingrich, N Davidson, E Barberio: DiJets - various talks in JetRec phone meetings Gamma + Jet - High Pt JES - again see talks in JetRec phone meetings JetPerformance - talk in 28th Nov JetRec phone conference Ysplitter - J Butterworth, A Davison, E Ozcan, P Sherwood: ATL-PHYS-INT Mid Point Algorithm in ATLAS - A Cheplakov,S Thompson: ATL-PHYS-PUB General very detailed introduction to all these issues by P Loch and M Lefevre at:

BACKUP Extra Information

Minimum Bias N.Davidson, E.Barberio (Melbourne) Apply loose cuts and subtract background Use control sample of late showering charged hadrons (MinBias) Can define EM outercone which only contains energy from background Solve set of linear equations p T =10 GeV Recover signal distribution Also check mean is consistent within errors - yes, but need more statistics to get to 1% precision Still needs more work to reduce signal contamination in control region (3%->1%) However this idea looks promising - if it works can be used potentially for higher energy tracks in QCD jets EMHad ±± Outer

In Situ Calibration Methods Photon is well measured (EM scale) - - use balance of  + Jet to get to jet energy scale Hadronic W Decays Template Method Generate template histograms of m jj using different sigma and mean for jet resolution Find which one fits best to data Read of Jet Energy Scale from template Stable wrt event selection, combinatoric background shifts scale 1% J Schwindling (Saclay)