1. Simon Fraser M. Vetterli –TOP 2013 – #1 Jets, Missing E T, and Pileup Systematic Uncertainties M.C. Vetterli Simon Fraser University and TRIUMF TOP 2013 September 15 th, 2013 On behalf of the ATLAS, CDF, CMS, DO Collaborations

2. Simon Fraser M. Vetterli –TOP 2013 – #2 Context-1 You need to understand the whole detector to do top-quark physics: JES & EtMiss Busy events with a variety of particles in the final state b-jets Light- quark jets Charged leptons Missing E T t t

3. Simon Fraser M. Vetterli –TOP 2013 – #3 CDF combined channel X-section (CDF note 10926) Largest uncertainty Context-2

4. Simon Fraser M. Vetterli –TOP 2013 – #4 Context-3 2 vertices 7 vertices 20 vertices Things are getting messy!

5. Simon Fraser M. Vetterli –TOP 2013 – #5 The anti-k t algorithm with R=0.4 (0.5) is used for top physics at ATLAS (CMS ) [several other values of R also used]  produces cone-like jets that are infrared and co-linear safe. Various objects are used as input (see following) CDF and D0 use an iterative cone algorithm (k t as well) Jet Reconstruction

6. Simon Fraser M. Vetterli –TOP 2013 – #6 Jet Reco Strategy - 1 CDF: calo towers D0: noise-suppressed towers ATLAS: topo calo clusters e.g. energy density, shower depth Pileup & noise suppression ATLAS

7. Simon Fraser M. Vetterli –TOP 2013 – #7 Jet Reco Strategy - 1 CDF: calo towers D0: noise-suppressed towers ATLAS: topo calo clusters e.g. energy density, shower depth E nergy density Shower Depth EM Probability Pileup & noise suppression ATLAS Important for resolution

8. Simon Fraser M. Vetterli –TOP 2013 – #8 Jet Reco Strategy - 2 Tracker ECAL HCAL

9. Simon Fraser M. Vetterli –TOP 2013 – #9 Jet Calibration Strategies 2011 data 1) Correct for pileup 2) Correct for vertex position (ATLAS; not needed for PF jets ) 3) Apply Monte-Carlo calibration factors 4) Apply residual calibration from in-situ measurements

10. Simon Fraser M. Vetterli –TOP 2013 – #10 Jet Calibration Strategies E O : offset (pileup, noise); S jet : showering correction R jet : Calorimeter response; from data (MPF) C η : η uniformity ; C MI : pileup ; C abs : calo response (MC-based)

11. Simon Fraser M. Vetterli –TOP 2013 – #11 Pileup Corrections N PV = # primary vertices (in-time pileup) μ = ave. # of interactions/bunch crossing (out-of-time pileup)

12. Simon Fraser M. Vetterli –TOP 2013 – #12 Pileup Corrections And/or use a jet-area-based correction: ρ = ave. E density ; A j = jet area Corrected Response

13. Simon Fraser M. Vetterli –TOP 2013 – #13 Response Corrections (p t, η) - Correct measured energy back to particle scale - Based on Monte Carlo simulation: Corr = p T part /p T meas - Monte Carlo simulation is validated by test-beam and single-hadron response data CMS: Apply to particle-flow jets ATLAS: Apply to jets at EM-scale (EM+JES) & jets at local hadronic calibration scale (LCW+JES)

14. Simon Fraser M. Vetterli –TOP 2013 – #14 in-situ Calibration The Monte Carlo is not perfect. Correct calibration using in-situ techniques: Balance jet transverse momentum against that of a well-measured object (Z, γ ) Two techniques: + pt-balance: balance the jet (but need OOC corrections) + Missing Projection Fraction (MPF): balance the whole hadronic recoil (no intrinsic EtMiss) MPF method does not depend on the jet algorithm to 1 st order. It is also much less sensitive to (ISR, FSR, UE). However, it does not test how well the MC models the out-of-cone correction.

15. Simon Fraser M. Vetterli –TOP 2013 – #15 in-situ Calibration – γ +jet The γ+jet uncertainty is dominated by photon purity at low pt, and the photon energy scale at high pt The MC-based calibration is off by 1-2% at ATLAS in 2011

16. Simon Fraser M. Vetterli –TOP 2013 – #16 η-dependence ATLAS: η-intercalibration CMS: ‘relative’ correction in-situ calibration using dijet events - Use jets in the central region that have been calibrated by Z+jet and/or γ+jet as a reference

17. Simon Fraser M. Vetterli –TOP 2013 – #17 Calibration at high-pt; multijet events Use calibrated jets at low pt to propagate the JES to larger pt. You can bootstrap your way up to high pt. Direct balance is used: MJB in data is compared to MJB in Monte Carlo.

18. Simon Fraser M. Vetterli –TOP 2013 – #18 Calibration at high-pt; multijet events Use calibrated jets at low pt to propagate the JES to larger pt. You can bootstrap your way up to high pt. Direct balance is used: MJB in data is compared to MJB in Monte Carlo. Flavour Dependence of JES - In-situ calibrations use Z/γ+jet events  dominated by quark-induced jets - Dijet events on the other hand are dominated by gluon-induced jet (more/softer particles) - Uncertainties are determined by varying MC (e.g. PYTHIA vs HERWIG) CMS

19. Simon Fraser M. Vetterli –TOP 2013 – #19 Heavy Flavour jets - b-jets can contain muons & neutrinos  yet another response - Use b-tagged jets in ttbar events to test the Monte Carlo: - compare track jets to calo jets - take the double-ratio data/MC - compare b-jets & light jets Extra uncertainty for b-jets (≈ 1.5%)

20. Simon Fraser M. Vetterli –TOP 2013 – #20 Heavy Flavour jets - b-jets can contain muons & neutrinos  yet another response - Use b-tagged jets in ttbar events to test the Monte Carlo: - compare track jets to calo jets - take the double-ratio data/MC - compare b-jets & light jets Extra uncertainty for b-jets (≈ 1.5%) No specific bJES uncertainty needed

21. Simon Fraser M. Vetterli –TOP 2013 – #21 Uncertainties on JES - Tevatron DO: < 2% ; ≈ 1% (20-200 GeV) CDF: ≈ 2.5% ; < 2% at low pt

22. Simon Fraser M. Vetterli –TOP 2013 – #22 Uncertainties on JES - LHC Pileup dominates at low pt

23. Simon Fraser M. Vetterli –TOP 2013 – #23 Jet Resolution Measured using dijet events PF and LCW improve jet resolution significantly, PF by using the better resolution of the tracker and the ECAL, LCW by providing “software compensation” for the calorimeter

24. Simon Fraser M. Vetterli –TOP 2013 – #24 EtMiss reconstruction The “soft” term corresponds to clustered energy in the calori- meter, but that is not part of a jet. Missing transverse momentum (E Miss, E T ) can indicate the presence of neutrinos or other (new?) non-interacting particles. It is calculated as the negative of the vectorial sum of all of the objects in the events CMS: Three kinds of E T :1) PF E T : calculated from particle flow objects  2) Calo E T : calculated from calorimeter clusters (noise threshold) 3) TC E T : Calo E T corrected for tracks [−p T (π) + p T (track)] ATLAS: Many variants of E Miss : calo-based or MET_RefFinal (uses reconstructed objects) T T

25. Simon Fraser M. Vetterli –TOP 2013 – #25 EtMiss - Pileup - The jet and soft terms are the most affected by pileup: large-area objects that are dominated by hadronic energy deposits - Corrections can be made in various ways: - as a function of N PV and μ - as a function of object area - using MVA algorithms (CMS-2012) - As a general rule, pileup corrections improve the E Miss resolution, but worsen the scale (by over-correcting the soft terms) With pileup Pileup suppressed Z μμ no real E Miss (except in bkgnd processes) T

26. Simon Fraser M. Vetterli –TOP 2013 – #26 Projections in Z+jets Events - u T is the transverse momentum of the recoil - u || should balance the transverse momentum of the Z - u is a measure of the underlying event (≈0)

27. Simon Fraser M. Vetterli –TOP 2013 – #27 Projections in Z+jets Events E Miss projected into the direction of the Z T Underestimate of the hadronic recoil - u T is the transverse momentum of the recoil - u || should balance the transverse momentum of the Z - u is a measure of the underlying event (≈0) Worse for STVF (Are soft clusters from the hard PV?)

28. Simon Fraser M. Vetterli –TOP 2013 – #28 EtMiss resolution Particle Flow jets with two different pileup suppression techniques. Pileup suppression improves EtMiss resolution

29. Simon Fraser M. Vetterli –TOP 2013 – #29 Summary Jets and EtMiss are crucial to most (all?) physics analyses; top in particular Pileup is a significant effect at the LHC Several approaches have been developed for jet reconstruction & calibration Residual corrections from in-situ techniques All four experiments have JES uncertainties at the 1-2% level (absolute calibration) Missing E T requires an understanding of the whole detector; well modeled Work continues to improve the situation even further What I didn’t cover:- Large-R jets; boosted topologies - ATLAS & CMS combined uncertainties (correlations)

30. Simon Fraser M. Vetterli –TOP 2013 – #30 Acknowledgements / Further Info To be done: Add papers & CONF/PAS notes Thank people from whom I took slides/figures

31. Simon Fraser M. Vetterli –TOP 2013 – #31 Backup Slides

32. Simon Fraser M. Vetterli –TOP 2013 – #32 Combining CMS & ATLAS Results - Correlations A working group has been formed. See (Kirschenmann, Doglioni, Malaescu) : https://indico.cern.ch/getFile.py/access?contribId=7&sessionId=1&resId=0&materialId=slides&confId=245769 https://indico.cern.ch/getFile.py/access?contribId=7&sessionId=1&resId=0&materialId=slides&confId=245769 The following areas were identified for further study of correlations between the experiments: 1) in-situ Z+jets: radiation suppression, out-of-cone bias, extrapolation to Δφ=π 2) in-situ γ+jets: same, but add photon purity 3) Flavour response: JES variation with jet composition 4) bJES: JES variation with jet composition 5) High-pt: Homogenize the treatment of high-pt uncertainties

33. Simon Fraser M. Vetterli –TOP 2013 – #33 1)Context & Motivation 2)Jet Strategies: I will use CMS & ATLAS data to illustrate 3)Jet Calibration (including in-situ techniques) 4)Missing E T 5)Summary Outline

34. Simon Fraser M. Vetterli –TOP 2013 – #34 CMS: lepton + jets PLB 720 (2013) 83 One of the largest experimental uncertainties Context-3

35. Simon Fraser M. Vetterli –TOP 2013 – #35 in-situ Calibration-1 The Monte Carlo is not perfect. Validate calibration using in-situ techniques: Balance jet transverse momentum against that of a well-measured object (Z, γ ) Two techniques: + pt-balance: balance the jet (but need OOC corrections) + Missing Projection Fraction (MPF): balance the whole hadronic recoil (no intrinsic EtMiss) MPF method does not depend on the jet algorithm to 1 st order. It is also much less sensitive to (ISR, FSR, UE). However, it does not test how well the MC models the out-of-cone correction. Z+jet good at low pt, where γ+jet has low purity γ+jet good at mid-pt where Z+jet runs out of events

36. Simon Fraser M. Vetterli –TOP 2013 – #36 Calibration at high-pt; multijet events Use calibrated jets at low pt to propagate the JES to larger pt. You can bootstrap your way up to high pt. Direct balance is used: MJB in data is compared to MJB in Monte Carlo.

37. Simon Fraser M. Vetterli –TOP 2013 – #37 Combination of in-situ techniques - Do a statistical combination of the in-situ methods as a function of p T - Use a weighted average for the final result - A different method dominates in different regions of p T Absolute Energy Scale

38. Simon Fraser M. Vetterli –TOP 2013 – #38 Uncertainties on JES - LHC Absolute Scale

39. Simon Fraser M. Vetterli –TOP 2013 – #39 Uncertainties on JES - LHC Absolute Scale

40. Simon Fraser M. Vetterli –TOP 2013 – #40 EtMiss reconstruction - 3 T Σ E T is much more affected by pileup than E Miss. A nice demonstration that pileup is φ symmetric to first order. T Missing E T must be studied for many different event topologies, from those with little real E Miss, where the soft terms dominate, to those with large real E Miss, such as SUSY. T T

41. Simon Fraser M. Vetterli –TOP 2013 – #41 EtMiss Studies The missing E T is studied in Z/γ+jet events, which have no real missing E T, and in events with real missing E T such as W+jets (W → l ν) 1) Real E Miss from bkgd evts 2) Pileup suppression reduces E Miss 3) Monte Carlo does pretty well T T Mis-modelling of the number of jets in POWHEG+PYTHIA8 In addition, strong pileup suppression in the soft term

42. Simon Fraser M. Vetterli –TOP 2013 – #42 EtMiss Studies The missing E T is studied in Z/γ+jet events, which have no real missing E T, and in events with real missing E T such as W+jets (W → l ν) 1) Real E Miss from bkgd evts 2) Pileup suppression reduces E Miss 3) Monte Carlo does pretty well T T Mis-modelling of the number of jets in POWHEG+PYTHIA8 Alpgen prediction is much better

43. Simon Fraser M. Vetterli –TOP 2013 – #43 Projections in Z+jets Events Fractional mismatch between the recoil and the Z Pileup suppression makes it worse

44. Simon Fraser M. Vetterli –TOP 2013 – #44 Projections in Z+jets Events Fractional mismatch between the recoil and the Z