In situ calibration: status of γ+jet and Z+jet Outline Introduction γ+jet status pT balance method –selection cuts comparative study at parton level, particle level, detector level of different jet algorithms Underlying Event & radial jet shape & noise suppression Effect of dijet background Z+jet Rome Workshop results Scan across eta comparison pT balance and pTmiss projection comparison Z+jet and γ+jet Conclusions and future work S. Jorgensen IFAE Hadronic Calibration Workshop Munich 3-5 May 2006
Introduction Motivations and Z0 are well calibrated objects at EM scale balancing the recoiling hadronic system potentially large statistics available: L=1033cm-2s-1 pT range from 20 GeV to ~60 GeV: Z(ll)+jet (~2Hz) γ+jet (~ 0.1 Hz) reserving 1Hz for downscaled trigger pT range > 60 GeV: (expected threshold for single ) γ+jet (~2Hz) Z+jet (~ 0.1 Hz) Issues to be understood Detector effects: response, showering Physics effects: fragmentation, gluon radiation (multijets) Compare different methods of analyses and the two data samples
γ+jet 1: pT balance Rome data. Athena 10.0.1 select gamma Electronic Noise, no Minimum Bias select gamma select highest pT jet (not dependent on jet algorithm apply phi back-to-back cut pT balance = (pT jet – pT photon)/ pT photon Fit peak region iterating a gaussian fit between ±σ around the most probable value When selecting a band in the pT of the photon, the events on the left side of diagonal are more numerous because they are produced at lower CM pT, hence with higher cross section. This induces a negative bias to the pT balance: 4 % at 35 GeV (1 % at 100 GeV) We checked that selecting events using the pT of the reconstructed pT gamma and of the reconstructed jet (calibrated) instead of the thruth level gamma and the parton did not significantly change the measured pT balance Selecting events using reconstructed gamma and jet (calibrated) instead of truth gamma and parton, do not significantly change the measured pT balance pT balance = (pT parton – pT photon)/ pT photon http://agenda.cern.ch/fullAgenda.php?ida=a057453
γ+jet 2: pT balance Biases on pT balance MOP for the different jet Parton level Particle level Kt Reconstruction level Kt Particle level Cone 0.7 Reconstruction level Cone 0.7 Particle level Cone 0.4 Reconstruction level Cone 0.4 Too close to the generation cut pT balance pT balance (pTγ+pTparton)/2 (GeV) (pTγ+pTparton)/2 (GeV) Biases on pT balance MOP for the different jet algorithms: pT balance Algorithms Cone 0.7 Cone 0.4 Kt Parton level -1 - 0% Particle level 1 - 0% -7 - -3% 7 - 1% Recon level -2 - 0% -15 - -7% 10 - 2% Black points: balance at parton level Blue points: balance for Et of jets reconstructed at particle level Red points: balance for Et of jets reconstructed from detector signals -------------------------------------- Cone 0.7 : parton level, particle level jets & detector jets balance are similar Out of cone loss compensated by Underlying Event? Cone 0.4 : particle level out of cone losses 7% (35 GeV) to 4% (100 GeV) Reconstructed level 10-5% lower (calibration not optimum) kT: particle level & reconstructed level 5% positive bias above parton To understand differences Cone7 and kT, study underlying event, effect of noise, etc (pTγ+pTparton)/2 (GeV) Standard H1 weighting: calibrated for the C7
γ+jet 3: Underlying Event Transverse plane Try to measure the mean ET of UE from the event sample Select the “transverse region” of the event: avoiding 60 degrees in Phi arround both photon and the jet (suggested by the SM group) γ 60o Transverse region jet Protojet recon lvl Protojet particle lvl Tower Et (MeV) Et (MeV) Et (MeV) Difficulties: average UE level ~10% RMS of el.noise (very sensitive to noise suppression) calorimeter non-compensation is large at low energies Mean transverse energy per ŋ x φ = 0.1 x 0.1 Average UE level ~10% RMS of el.noise (very sensitive to noise suppression) Tower (RMS of el.noise ~140 MeV) 15.1 ± 0.2 MeV Recon protojet (protojets after applying “TowerNoiseTool”) 15.8 ± 0.2 MeV Particle protojet (Σ particles per tower) 19.1 ± 0.4 MeV 3 GeV in cone 0.7
Jet constituent Et versus deltaR (R – jet R) (one entry per tower) γ+jet 4: Jet size Particle Level Jets Jet constituent Et versus deltaR (R – jet R) (one entry per tower) Kt Cone 0.7 Et (GeV) Et (GeV) Cone algorithm stops at 0.7 kT algorithm extends lo larger value of R as expected from “R parameter” set to 1 deltaR deltaR Kt extends to larger R 0.7
Fully reconstructed jets γ+jet 5: Jet size Fully reconstructed jets Jet constituents (Towers) Et versus deltaR (one entry per tower) Kt Cone 0.7 Et (GeV) Et (GeV) C4 PowerNoiseTool algorithm: starts from negative ET towers ordered (most negative first) loop over ordered positive towers (most positive first) in a window 7x7 in units of 0.1x0.1 keep summing until ET becomes positive Resulting “protojet” momentum vector is obtained by vector sum of momenta (negative energy tower has momentum vector pointing in opposite direction). That protojet can be used for subsequent negative tower compensation in the vicinity (progressive drift of the location of towers). This “drift” is one the reason why towers so far away from the jet axis can be constituents of the jet. The other eeason is that a vector sum of the momenta is done for negative energy towers: 2 neighbouring towers, one positive and one negative with similar absolute value of the energy, correspond to almost back –to-back vector. Their vector sum points at 90 deg w.r.t to the real postion of the tower in the calorimeter. This is what the protojet gice and this is what is used by the jet clustering algorithm remaining uncompensated negative towers are eliminated, as well as “unphysical” protojets (PZ>E). The amount of ET corresponding to eah category is ... GeV and ... GeV respectively deltaR deltaR Why is the size so big? This is due to features of the TowerNoiseTool
γ+jet 6: Jet ET Radial Density Profile Radial profile: Jet constituent Et density versus deltaR (integrated Et in a 0.1 ring and divided by the area) Recon level Et (GeV) Et (GeV) C7 Kt deltaR deltaR Jet constituent Et versus deltaR (integrated Et in a 0.1 ring) Et (GeV) Et (GeV) C7 Kt deltaR deltaR
γ+jet 6: Dijet background Default CBNT cuts: S/B~10% Optimised cuts: S/B~30% Efficiency γ ~ 90% Efficiency γ ~ 15% low pT sample <ET>~30 GeV Data sample Athena 7.2.0 DC1 data Mean (-0.6, 0.6) window Cone 0.4 Cone 0.7 kT Signal -13 ± 0.8% 2 ± 0.9% 1 ± 0.9% Background -15 ± 2% 1 ± 2% -1 ± 2% Tight photon selection cuts, at the cost of some efficiency, allow to select jets that have fragmented to π0 ; they also are valid “well calibrated EM scale reference” from the point of view of calibration This plot is for eT(protons)~30GeV: improves at larger ET remaining jet background ≈ π0 statistical error C. Deluca. Rome Workshop
γ+jet 7: future work Continue UE studies - Use different datasets with different level of underlying events to help understand the difference of Kt and C7 algorithms Study effect of higher jet multiplicity on the pT balance method Study gamma+jet background for pT > 30 GeV
Z+jet 1 Z->m+m- Z->e+e- (PtJet-PtZ) PtZ PtZ (GeV) Goal is to establish the systematic uncertainty Use Gaussian Fits to Compare Response vs PtZ and TRUTH vs RECONSTRUCTED Rome data Z->e+e- Z->m+m- PtZ (GeV) TRUTH RECONSTRUCTED (PtJet-PtZ) PtZ Why? Why This Too? Selection: Z+Exactly 1 Jet Implies No Additional Jet with Et>10GeV df (JetZ)>2.88radians |h| (Jet)<2.6 J. Proudfoot, Tileweek, oct 12, 2005
Z+jet 2 Study Effect of second jet Select df (LeadJet-Z)<0.26 radians; 40< PtZ<75GeV => Suggests that the imbalance at Low PtZ is related to the jet selection No 2nd Jet 10<Et<15 15<Et<20 m= -0.054 0.005 m= -0.084 0.010 m= -0.179 0.025 TRUTH JETS (PtJet-PtZ) PtZ J. Proudfoot, Tileweek, oct 12, 2005
Z+jet 3 We interpret this result as implying that there is ALWAYS a second jet in the event. You ado better in momentum balance just above the Jet Pt cut by adding in the Pt of the 2nd jet. 2nd Jet:10<Et<14 GeV 2nd Jet:14<Et<18 GeV ptZ>40GeV e.g. TRUTH JETS Lead-Jet DiJet
Z+jet: Summary Observed a variation of 10% in the Pt balance for Truth jets as a function of PtZ. Truth jets and Reco jets follow the same general trend, but are generally offset by between 5 and 10%. We do not understand the source of this offset. The Pt balance is statistically consistent for the two decay modes but the precision is only of order 1-2%. If we study two-jet events, just above the minimum jet Et cut of 10 GeV, we realize momentum balance only if we sum the Pt of the two jets. Things to be studied: Pt balance as a function of the fraction of jet energy in the electromagnetic calorimeter (this is not possible from the data stored on the AOD) Much higher statistics analysis of the multi-jet processes (including the analogous survey of balance as a function of Jet Pt, h, angle between jets, etc. Different event generators (ALPGEN, Sherpa, MadGraph, MCFM) The angular distribution between the two jets, as function of the Pt of the leading jet Different cone sizes as well as for the Kt jet algorithm Different event selection approach (e.g. on (PtZ+EtJet)/2) Analysis using the bisector method to further reduce the effect of ISR See note ATL-COM-PHYS-2005-067
Conclusions and future work γ+jet, Z+jet events are useful for comparitive jet algorithm studies to do the relative energy calibration through the detector ultimately should also provide useful information on absolute energy calibration Future work: Continue studies of both gamma+jet and Z+jet events and check their consistency