1. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 In situ jet energy calibration General considerations The different tools –E/p ratio on charged energetic isolated hadrons coming from  decays –Invariant mass constraint on the pair of light quark jet coming from W decays in events –p T balance between jet and boson in Z 0 +jet or  +jet events Outlook

2. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Jet energy reconstruction Physics effects –Fragmentation –ISR and FSR –Underlying and MinBias events Detector effects –Magnetic field –Dead material –Non compensation –Electronic noise –Longitudinal shower leakage Jet reconstruction –Jet definition –Energy reconstruction method Motivations –Mass reconstruction –SUSY searches –QCD studies Goal –Absolute jet energy scale known at ~ 1 % Transport of beam calibrations to the ATLAS environment –Jet energy scale at ~ 5-10 % In situ calibration is needed –E/p on single tracks –W mass –p T balance

3. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 E/p calibration Based on charged energetic isolated hadrons (   or K  ) –p precisely measured in tracker, E measured in calorimeters E/p = 1 –Comparison with test-beam results –Inter-calibration of sub-detectors –  coverage limited by the Inner Detector :  2.5 High rate signals – Backgrounds –QCD events –Events from   decays itself leptonic modes multi-prong modes modes with  0

4. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 E/p calibration : event selection  QCD filter Trigger-like selection –One  jet (  20 GeV/c) + missing E T (  35 GeV) or one isolated electron or muon (  20 GeV/c) Pre-selection –N (jet)  1 Jet candidate :  (jet)  2.5, p T  20 GeV/c –N (jet + isolated lepton)  2 –N (isolated photon) = 0 Selection –Matching cut :  R = 0.15 cone centred on jet direction N (matching tracks)  1, p T (hardest matching track)  30 GeV/c –Tracker isolation :  R = 0.4 cone centred on the hardest matching track direction N (extra-tracks)  1, p T (extra-track)  1 GeV/c

5. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 QCD reduction –Huge statistics –Fast simulation studies Number of events for 10 fb -1 –Signal 530 k –QCD background 30 k No E/p bias –  decays backgrounds360 k Almost only additional  0 ’s Global E/p bias = + 3.7 % Conclusion –QCD Under Control –Need further reduction of  0 ’s E/p calibration : results after QCD filter

6. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 E/p calibration :  0 rejection  0 rejection –Use calorimeters granularity –Need showers development  Full simulation Single Particles simulated –signal  + –background  +   +  0 3  cases after QCD filter –MIP energy in ECAL  Accepted –Major part of energy in ECAL  Rejected –Intermediate situation  Additional Cuts using ECAL first sampling :  -strips designed for  /  0 separation Relative efficiencies –for  + ~ 60 % – for  + ~ 6 %  320 k signal events for 10 fb -1 Global E/p bias < 0.6 %

7. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Calibration of jet energy using W  jj Invariant mass constraint – –Light quark jet energy calibration Use of top pair production –High rate :  = 833 pb –Use modes with one W decaying into leptons and the other into light quarks Easy to trigger and to select Almost only combinatorial background Less ambiguities than in fully hadronic decays Method developed for top mass reconstruction –Reduction of systematic errors from light quark jet energy scale Results obtained with fast simulation

8. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : event selection Selection –cone size :  R = 0.4 –N(jet)  4 |  |  2.5 p T  40 GeV/c –Two b-jets –One lepton :  or e |  |  2.5 p T  20 GeV/c –Missing E T sd Results –90 k selected W  jj for 10 fb -1 –Physical Backgrounds negligible Calibration –|M jj - |  20 GeV/c 2 –Combinatorial background ~ 35 %

9. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : first calibration try Usual jet pairing : identification of the two jet coming from the W –Minimising over all light quark jet pairs First calibration try : E CORR0 –Find event by event the corrected jet energies minimising  E taken from the Monte-Carlo Do not change jet directions Event by event analysis : no a priori knowledge of function to be used – per bins of E RAW (same statistic) – E CORR0 /E RAW = f (E RAW )

10. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : first calibration try results E CORR0 /E RAW = P1 /E RAW + P2 (ATLFAST)

11. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : first calibration try conclusion Comparison with parton –Corrected energy is systematically higher > 2 % at high energy Jet directions –  j1j2 systematically too low  of 2 to 10 % on (1-cos  j1j2 )  increasing with jets overlap –Effect of -1 to -5 % on M jj –Up to 5 % overestimation of jet energies for big overlaps Conclusion –One should correct jet energies but also jet directions

12. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : final calibration Final calibration : E CORR –Find event by event the corrected jet energies and directions minimising Change also jet directions   and   taken from the Monte-Carlo –As before : E CORR = f (E RAW ) Comparison with parton –This time the  1 % level is obtained –Good light quark jet energy calibration from 40 to 250 GeV

13. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 W  jj : final calibration results

14. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Calibration of jet energy using p T balance Improve resolution and linearity using H1 method –on Z 0 +jet or  +jet events (Z 0  e + e - or  +  - ) Easy to trig and to select p T (boson) very well measured –with p T balance calibration constraint : p T (jet) = p T (boson) High statistics with p T (boson)  40 GeV/c –  (Z 0 +jet) = 380 pb –  (  +jet) = 12 nb Allows –Large  and energy coverage –b-jet energy scale Studies based on –Full simulated Z 0 +jet events at low luminosity to test the method –Fast simulated events to estimate systematic errors

15. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 p T balance : selection of fully simulated Z 0 +jet events Generated events –Z 0 +jet with Z 0   +  - Select topologies –Only one jet ( E T th = 15 GeV ) –E T (jet)  20 GeV –|  (jet) -  (Z 0 ) -  |  0.15 –| M Z - 91.187 GeV/c 2 |  10 GeV/c 2 Residual unbalance is still important Results limited here to |  jet |  1.2

16. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 p T balance : calibration procedure in full simulation Reconstructed transverse energy : –E cell coming from electromagnetic scale calibration –a l ’s obtained by minimising for each interval in ( p T (Z 0 ), |  jet | ) plane Reconstruction of transverse energy without any a priori knowledge : –Linear interpolation of a l ’s as functions of jet transverse energy from electromagnetic scale calibration : E T (jet) Corrections to take into account residual unbalance (Initial State Radiation) –Based on Monte-Carlo Simulation

17. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 p T balance : full simulation results Parameterisation of the reconstructed energy Resolution and linearity obtained –Almost no differences between before and after interpolation –Deterioration of about 10% with respect to the optimal resolution

18. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 p T balance : full simulation calibration parameters

19. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 p T balance : full simulation Monte-Carlo corrections To reduce systematic errors, use : Correction terms –Systematic studies limited by statistics Fast simulation study required –Important resolution corrections Use bisector method to estimate them on data

20. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Bisector method Determine resolution and linearity –Z 0 +jet or  +jet events (Z 0  e + e - or  +  - ) Use azimuthal angular bisector –  -axis  –  -axis  Transverse momenta projections : K T  and K T  –  (K T  ) =  D   UNB –  (K T  ) =  UNB Intrinsic resolution given by –  D 2 =  2 (K T  ) -  2 (K T  ) Results obtained with fast simulation –Selection as for full simulation except no collinearity cut in transverse plane –Asymmetric distributions Different algorithms used to estimated peak and spread values

21. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Bisector method : example of K T  and K T  distributions

22. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Bisector method : K T  and K T  resolutions

23. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Bisector method : estimations of intrinsic resolution

24. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Bisector method : conclusions Bissector method conclusions –Good estimation of resolution using only data With the best estimators : systematics lower than 1 GeV/c –Correction on linearity required Up to about 100 GeV/c Use  +  events –with at least one  of p T  40 GeV/c  (  +  ) = 23 pb –Good estimation of linearity With the best estimators : systematics lower than 1 %

25. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Use of  +  events to correct non-linearity

26. Régis Lefèvre (LPC Clermont-Ferrand - France)ATLAS Physics Workshop - Lund - September 2001 Outlook E/p and W  jj –Use full simulation p T balance –Try to go to smaller p T –Increase the statistic of fully simulated events For the three methods –Study effects of high luminosity pile-up