1 / 35

Jet Calibration in CMS: experience… lot of questions… and few answers… …a work in progress…

Jet Calibration in CMS: experience… lot of questions… and few answers… …a work in progress…. Attilio Santocchia INFN Perugia Frascati – 2nd Workshop sui Monte Carlo, la Fisica e le Simulazioni a LHC 22.05.2006. Motivations.

neal
Télécharger la présentation

Jet Calibration in CMS: experience… lot of questions… and few answers… …a work in progress…

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Jet Calibration in CMS: experience… lot of questions… and few answers……a work in progress… Attilio Santocchia INFN Perugia Frascati – 2nd Workshop sui Monte Carlo, la Fisica e le Simulazioni a LHC 22.05.2006

  2. Motivations • ttH channel very challenging: we have to optimize all the tools we need to use… • Which is the best Jet Algorithm to use for such a complex multi-jet final state rich with b-jets? • I used to study the fully hadronic decay  8 jets (4 light + 4 b jets) • Honestly? It’s a mess! And jets are the most important object I had to understand… Attilio Santocchia

  3. Chosen Algorithm & Data Sample • 5 Different Iterative Cone Algorithm +3 KT inclusive • ICA DeltaR=0.30, 0.35, 0.40, 0.45 and 0.50 • KT  r=0.35, 0.40 and 0.45 • COMPHEP+PYTHIA 6.215+CTEQ4L • ttH120 (200K) • ttjj (1000K) 1.6M events • ttbb (400K) • ALPGEN2+PYTHIA 6.325+CTEQ5L • tt1j exclusive (1000K) • tt2j exclusive (560K) • tt3j exclusive (68K) • tt4j inclusive (97K) • 1.725M events Attilio Santocchia

  4. Calibration • To do physics we need to go back to parton energy • 2 different aspect to be considered: • Detector Effects  PARTICLElevel correction • Physics Model Effects  PARTONlevel correction • I tried to factorize the 2 effects: • 2 different set of Calibration functions are calculate for correction to JetMC Energy and to Parton Energy • These functions are then applied to the raw jet energy in cascade to recover the initial Parton Energy ET(raw)  ET(MCjet)  ET(parton) Attilio Santocchia

  5. Raw Jet • Standard Jets from CMS simul+reco software • No Calibration • Calorimeter Noise Cut are: ET > 0.5 GeV and E > 0.8 GeV • Raw Jet ET > 5 GeV • All Jets are considered massless Attilio Santocchia

  6. MC Jets • Built from stable generator particles and ET > 0.5 (1) GeV • Muons and Neutrinos are included in the calculation • Muons and Neutrinos are excluded in the calculation • If you keep all particles  Jet is the same (difference in ET below 5%) but CPU time needed for ICA increases of a factor 2-3 • Jet is kept if ET(jet) > 10 (20) GeV • We need to b-tag MC jets to build flavored dependant calibration: • Each particle belonging to a jet is classified as daughter of a b-flavored unstable particle or not • Define b-ratio as sum of energy from particles from b divided jet energy • If b-ratio > 0.20 then jet is tagged as a MC b-jet Attilio Santocchia

  7. Jet Costituents • After ICA application, I can list all the particles used to form the Jet (Jet Components) • Each Particle in the Jet can be associated either to the partons from the hard scattering or nothing • Most of the time 100% of the particles within a jet comes from the same original parton… • But sometimes there is a mixing… Attilio Santocchia

  8. Particles  String  Partons Attilio Santocchia

  9. Jet Classification - Example JET= 4 jet= 4 EtJ=33.731 EtaJ=-1.211 PhiJ=-0.899 EJ=61.810 n= 0 Npar=192 Strg=191 Part= 4 EtP= 6.883 EtaP=-1.234 PhiP=-0.864 EP=12.832 dR= 0.042 n= 1 Npar=193 Strg=191 Part= 4 EtP=13.563 EtaP=-1.249 PhiP=-0.921 EP=25.595 dR= 0.044 n= 2 Npar=371 Strg=191 Part= 4 EtP= 1.528 EtaP=-1.243 PhiP=-0.918 EP= 2.873 dR= 0.038 n= 3 Npar=373 Strg=191 Part= 4 EtP= 1.604 EtaP=-0.925 PhiP=-0.825 EP= 2.391 dR= 0.295 n= 4 Npar=375 Strg=191 Part= 4 EtP= 4.728 EtaP=-1.204 PhiP=-0.892 EP= 8.606 dR= 0.009 n= 5 Npar=376 Strg=191 Part= 4 EtP= 1.334 EtaP=-1.417 PhiP=-1.046 EP= 2.916 dR= 0.254 n= 6 Npar=507 Strg=191 Part= 4 EtP= 2.061 EtaP=-1.067 PhiP=-0.838 EP= 3.351 dR= 0.155 n= 7 Npar=508 Strg=191 Part= 4 EtP= 2.061 EtaP=-1.027 PhiP=-0.890 EP= 3.246 dR= 0.184 JET= 5 jet= 5 EtJ=25.726 EtaJ= 1.639 PhiJ= 2.254 ThetaJ= 0.384 EJ=68.962 n= 0 Npar=238 Strg= 91 Part= 0 EtP= 3.739 EtaP= 1.902 PhiP= 2.357 EP=12.810 dR= 0.283 n= 1 Npar=258 Strg=110 Part= 3 EtP= 3.686 EtaP= 1.487 PhiP= 2.340 EP= 8.585 dR= 0.174 n= 2 Npar=261 Strg=110 Part= 3 EtP= 5.652 EtaP= 1.518 PhiP= 2.170 EP=13.564 dR= 0.147 n= 3 Npar=262 Strg=110 Part= 3 EtP= 1.627 EtaP= 1.466 PhiP= 2.034 EP= 3.716 dR= 0.279 n= 4 Npar=422 Strg= 91 Part= 0 EtP= 1.370 EtaP= 1.977 PhiP= 2.286 EP= 5.042 dR= 0.340 n= 5 Npar=438 Strg=110 Part= 3 EtP= 3.864 EtaP= 1.560 PhiP= 2.246 EP= 9.615 dR= 0.079 n= 6 Npar=439 Strg=110 Part= 3 EtP= 4.426 EtaP= 1.622 PhiP= 2.279 EP=11.641 dR= 0.031 n= 7 Npar=440 Strg=110 Part= 3 EtP= 1.457 EtaP= 1.665 PhiP= 2.249 EP= 3.990 dR= 0.027 • First Jet is PURE: all the particles comes from the same string ; the associated parton code is 4 (a W+) • Second Jet is a mixing of t_bar (code 3) and something else (code 0) • How can I treat this kind of situation? Attilio Santocchia

  10. Jet Classification – An Event • ttH fully hadronic  CompHEP • Iterative Cone Algo (Cone Size 0.4) • ET(particles isthep=1)>1GeV Index 0 Jet 3 Et=164.511 Eta=-2.067 Phi=-2.742 RatioE=0.800 Quark Higgs Index 1 Jet 0 Et=147.849 Eta=-2.614 Phi= 0.030 RatioE=1.000 Quark top Index 2 Jet 2 Et=105.950 Eta=-2.131 Phi=-0.131 RatioE=0.742 Quark W+ Index 3 Jet 1 Et=102.842 Eta=-0.176 Phi= 2.456 RatioE=0.981 Quark top_bar Index 4 Jet 4 Et= 33.731 Eta=-1.211 Phi=-0.899 RatioE=1.000 Quark W+ Index 5 Jet 5 Et= 25.726 Eta= 1.639 Phi= 2.254 RatioE=0.741 Quark top_bar Index 6 Jet 7 Et= 11.329 Eta=-1.117 Phi= 1.326 RatioE=1.000 Quark Higgs Index 7 Jet 8 Et= 8.020 Eta=-2.585 Phi=-2.695 RatioE=1.000 Quark Higgs Index 8 Jet 9 Et= 6.469 Eta= 0.227 Phi= 2.937 RatioE=1.000 Quark W- Index 9 Jet 10 Et= 4.530 Eta=-0.108 Phi=-3.023 RatioE=1.000 Quark W- Index 10 Jet 6 Et= 4.008 Eta=-2.951 Phi= 0.251 RatioE=1.000 Quark top Index 11 Jet 12 Et= 3.490 Eta=-1.442 Phi= 0.667 RatioE=1.000 Quark Higgs Index 12 Jet 11 Et= 2.566 Eta= 0.574 Phi=-2.994 RatioE=1.000 Quark W- Index 13 Jet 13 Et= 1.797 Eta=-2.101 Phi= 3.061 RatioE=1.000 Quark Higgs Index 14 Jet 14 Et= 1.566 Eta=-3.743 Phi=-0.466 RatioE=1.000 Quark top Index 15 Jet 15 Et= 1.548 Eta=-2.106 Phi=-2.194 RatioE=1.000 Quark Higgs Index 16 Jet 16 Et= 1.494 Eta=-0.598 Phi= 1.332 RatioE=1.000 Quark Higgs Index 17 Jet 17 Et= 1.325 Eta=-1.567 Phi=-0.483 RatioE=1.000 Quark W+ Index 18 Jet 18 Et= 1.320 Eta= 1.166 Phi= 2.320 RatioE=1.000 Quark top_bar Index 19 Jet 19 Et= 1.297 Eta= 2.246 Phi= 2.674 RatioE=1.000 Quark no part Index 20 Jet 20 Et= 1.263 Eta=-2.534 Phi=-0.745 RatioE=1.000 Quark top Index 21 Jet 21 Et= 1.003 Eta=-0.934 Phi=-2.936 RatioE=1.000 Quark no part • In real life, low ET Jets are objects difficult to detect • Only jets with ET>10(20)GeV will be used in the next slides Attilio Santocchia

  11. How many MC jets? tt1j – All Particles – ET>10GeV • <Njet> distribution for different jet algo • Black is ICA • Red is KT • In the table <Njet>: • red is maximum <Njet> • 1st number is AllParticle – 2nd is noMuNu Iterative Cone • If <Njet> increase when DR increase  I get more jets because of ET(jet)>10GeV • If <Njet> decrease when DR increase  overlapping KT • If <Njet> increase when r increase  Again depends on ET(jet)>10GeV • If <Njet> decrease when r increase  ??? Attilio Santocchia

  12. Jet Formation  Overlapping tt1j – All Particles – ET>10GeV Black is ICA – Red is KT Fractio of jets with jetRatio > 0.80 Fractio of jets with jetRatio > 0.90 97.3% ÷ 97.9% Attilio Santocchia

  13. bJets and cJets Classification • Look for each particle belonging to the jet… • Define b-Ratio and c-Ratio • ratioParticle = E(particle)/E(jet) • If(decayFromBquark) ratioForBjet+=ratioParticle • elseIf(decayFromCquark) ratioForCjet+=ratioParticle • A jet is a bJet if: • (ratioForBjet>ratioForCjet) && (ratioForBjet>CutBjetRatio) • A jet is a cJet if: • (ratioForCjet>ratioForBjet) && (ratioForCjet>CutCjetRatio) • What are CutBjetRatio And CutBjetRatio? • See next slides… Attilio Santocchia

  14. b-ratio Distribution b-ratio>0.2 b-ratio>0.2 tt1j Sample tt2j: ICA 0.300.50 – KT r=0.350.45 ICA 0.40: sample tt1j… tt4j Percentuale di bJet taggabili! Sono meno di quelli che mi aspetto… (in percentuale!) I bJet troppo soffici li perdo… E il gluon splitting è trascurabile Attilio Santocchia

  15. c-ratio Distribution c-ratio>0.2 c-ratio>0.2 • Same criteria  a jet is classified as a cJet when more than 20% of its energy comes from c haddrons (and there are no b in the decay chain…) • tt1j ~ 11.2% are cJets • tt2j ~ 10.4% are cJets • tt3j ~ 9.8% are cJets • tt4j ~ 8.8% are cJets Attilio Santocchia

  16. Minimum Distance (b and c partons) Fractio of jets nearer than 0.2 from a b(c) parton in the hard scattering Jet Algo is ICA and DR=0.4 Attilio Santocchia

  17. b Minimum Distance VS bRatio • Example for tt2j and ICA cone 0.4 • Here we can evaluate the gluon splitting • bJets with bRatio>0.5 and minDistB>0.5 are not coming from a b parton in the hard scattering  6.2% • bJets with bRatio>0.2 and minDistB>0.8 are not coming from a b parton in the hard scattering  4.0% Attilio Santocchia

  18. Calibration Raw Jet MC Jet • Build jets from Full Reco (FR) • Build jets from Generator (MC) particles list • Match FR-MC jets minimizing SDRFR-MC; keep jets where DRFR-MC< 0.3 • Fill 50h x 200ET histos with ET(FR)/ET(MC) for b-jets and not-b-jets in the rangeabs(eta)<5 and ET<600 GeV • Gaussian Fit if Nent>30 • For each Eta Value, fit the ET Ratio as a function of ET(raw) using the function Attilio Santocchia

  19. Calibration MC Jet Parton • Build jets from Generator (MC) particles list • Match jets-Parton minimizing SDRMC-Parton; keep jets where DRMC-Parton< 0.15 • Fill 50h x 200ET histos with ET(MC)/ET(Parton) for b-jets and not-b-jets in the rangeabs(eta)<5 and ET<600 GeV • Gaussian Fit if Nent>30 • For each Eta Value, fit the ET Ratio as a function of ET(MC) using the function Attilio Santocchia

  20. Single ET Ratio Distribution • Example for 1 of the 10000 bins in which the eta-ET plane has been divided • Red is b-jets - Black is not-b-jets • Fit done in 2 steps: • First in the whole histo range [0,2]  Get Mean and Sigma • Second in the range [mean-2.5*sigma,mean+2.5*sigma] Raw-MC jet Ratio Distribution MC jet-Parton Ratio Distribution Attilio Santocchia

  21. Ratio vs ET Distribution MC jet-Parton Ratio Distribution Raw-MC jet Ratio Distribution • Eta Ring 0.5<h<0.6 • Red is b-jets - Black is not-b-jets • No needs to distinguish b-jets for MC-Parton • Error Bar are defined as s/sqrt(N) where s is the width of the second fit and N is number of Entry (see slide #11) ET(GeV) ET(GeV) Attilio Santocchia

  22. Fitted Parameter Raw Jet MC Jet Parameter a Parameter b Parameter c • Fitted parameters as a function of eta • Red is b-jets - Black is not-b-jets h h h Here is the tracker Barrel-EndCaps Border! Attilio Santocchia

  23. Fitted Parameter MC Jet Parton (DR=0.5) Parameter a Parameter b Parameter c h h h • Fitted parameters as a function of eta • b-jets and not-b-jets are together • Each parameter fitted with a streight line Attilio Santocchia

  24. Parton Calibration - Fitted Para – Different Cone – All Particles Light Jets ● ICA 0.30 ● ICA 0.35 ● ICA 0.40 ● ICA 0.45 ● ICA 0.50 h h h bJets h h h Attilio Santocchia

  25. Ratio vs ET DistributionInput Particles is NoMuNu Eta=0.5-0.6 ET=51-54GeV Attilio Santocchia

  26. Ratio vs Eta Distribution (ET=45-48 GeV)Input Particles is NoMuNu and All Particles h h • Here AllParticles and NoMuNu show differences (for low eta) • But this is not a homogeneous functions… • Matching 0.15 is not enough? See also the Single distributions and the left tail… Attilio Santocchia

  27. MCJet to ttH Comparison (cone 0.5) • 3 different eta bin  No major difference above 40 GeV for not-b-jets and 60 GeV for b-jets • Difference due to different Calibration: • MCJet is only particle level • ttH is particle level and parton level • Parton level correction important for low ET jets Attilio Santocchia

  28. MC Jet Definition… what I learned… • Option Jet input list: • All Particles • NoMuNu • No major differences  the difference in bJets/lightJets is minor (negligible?) • Particle Calibration keeps the high difference btw bJets/lightJets… • Depends on the fragmentation used… • bJets charged spectrum different from lightJets • To recover correctly the jet energy is mandatory to tag the jets and use 2 different corrections for bJets and lightJets • Parton Calibration could be the same for all experiments (ATLAS/CMS)… • Providing the definition is the same for both od us… • All Particles keeps simple the definition for parton calibration Attilio Santocchia

  29. Invariant Masses ttH fully Hadronic • To cross-check the quality of the calibration functions, invariant masses for the W,t and Higgs particles are used • The 8 most energetic jets in the tracker are paired to the 8 partons in the final state using DR. • All the events where alle the 8 jets are paired with DR<0.3 have been selected • Invariant mass are built using the calibrated jets for each algorithm-calibration Attilio Santocchia

  30. Invariant Masses ttH fully Hadronic from MC Jets • 1000 events used for this exercise… • No DeltaR matching  Look for configuration 3+3+2 and correct JetRatio associations (see slide #9-#10) • Cone 0.5 is the more ermetic… OK but how many events survive the request 3+3+2? • Cone 0.4 is better… less ermetic but sigma is better, and Nevents is a lot better! Attilio Santocchia

  31. Invariant Masses ttH fully Hadronic • Invariant Masses for (from left to right) W,t and Higgs • Upper row is Standard CMS MC-Jet calibration and DeltaR=0.5 • Lower row is ttH-calibration and DeltaR=0.5 Minv(GeV) Minv(GeV) Minv(GeV) Minv(GeV) Minv(GeV) Minv(GeV) Attilio Santocchia

  32. Invariant Masses full Results • Resolution is defined as s/M • Numer of selected events and Resolution give a hint on the best algorithm to use • ICA DR=0.4 and inclusive KT seems good choice Attilio Santocchia

  33. Analysis from the CMS P-TDR ttH fully Hadronic • Analysys based on c2mass for jet pairing • 8 most energetic jets in |h|<2.7 • Centrality Cuts (All and Higgs) • The c2mass for 2W and 2tops within 3 sigma from expected values • Different cut for Btag and ET jets • Significance S/sqrt(N) and S/N used as benchMark Attilio Santocchia

  34. Same Analisys  Different Parton Calibration Attilio Santocchia

  35. ConclusionsNon ci sono ancora! ma… • La parton calibration è necessaria per noi sperimentali • La definizione di MCjet potrebbe essere uguale per tutti (ATLAS/CMS)… • In questo modo si potrebbe chiedere ai teorici/sperimentali di definire un unico oggetto che può essere usato da noi sperimentali per la particle calibration… • E dai teorici per la parton calibration… che è uguale per tutti… • In ogni caso la particle calibration non può essere unica per tutti i jets… • Ma una distinzione tra bJets e lightJets è necessaria… • Non ho ancora guardato il KT… mea culpa… Attilio Santocchia

More Related