Jets and jet substructure 4: substructure

Transcription

1 Jets and jet substructure 4: substructure Gavin Salam (CERN) with extensive use of material by Matteo Cacciari and Gregory Soyez TASI June 213 1

2 Two things that make special The large hierarchy of scales s MEW The huge pileup npileup ~ 2 4 [These involve two opposite extremes: low pt and high pt, which nevertheless talk to each other] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 2

3 e.g. ttbar resonances RS KK resonances t t, from Frederix & Maltoni, NB: QCD dijet spectrum is 1 3 times t t Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 3

4 Boosted EW scale objects Normal analyses: two quarks from X q q reconstructed as two jets High-p t regime: EW object X is boosted, decay is collimated, q q both in same jet X at rest X jet 1 boosted X z single fat jet jet 2 (1 z) Happens for p t 2m/R p t 32 GeV for m = m W, R =.5 Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 4

5 Very active research field Jet Declustering Seymour93 Some of the tools developed for boosted W/Z/H/top reconstruction Matrix Element YSplitter Mass Drop+Filter ATLASTopTagger Jet Shapes JHTopTagger TW Planar Flow Templates CMSTopTagger Pruning N jettiness HEPTopTagger (+ dipolarity) Trimming Twist CoM N subjettiness (Kim) N subjettiness (TvT) ACF Shower Deconstruction Multi variate tagger Qjets apologies for omitted taggers, arguable links, etc. Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 5

6 Papers on jet substructure papers / month Number of papers containing the words jet substructure Papers containing "jet substructure" + pioneering papers by Mike Seymour in 1991 and 1994 (Source: INSPIRE) Mike Seymour Butterworth, Cox, Forshaw More than 1 papers since 28 (+ some background noise) Pioneered by M. Seymour in the early 9s Exploded around year Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 6

7 Extensive experimental work ATLAS Public Results Large-R, groomed jets with pile-up Large-R jets with substructure Quark/gluon jets (see also this link) Jet substructure at LHC7 Jet properties for boosted searches Resonance searches Boosted top (hadronic) Boosted top (semileptonic) Three-jet resonance (gluino RPV) Two-jet resonance (sgluon) CMS Public Results Jet substructure in CMS Subjet multiplicity Jet mass and grooming Resonance searches: Boosted top (hadronic) Boosted top (semileptonic) Boosted W/Z From a list compiled for a recent workshop at Perimeter Institute Many more analyses in the pipeline Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 7

8 Jet masses 1/N dn/dm jet [GeV -1 ] Wj events p t,jets > 7 GeV anti-k t, R =.7 pp 7 TeV, Pythia 6.4, noue Look at jet mass distribution for 2 leading jets in pp W+jet events m jet [GeV] Pythia, underlying event switched off Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 8

9 Problem #1a: QCD jets have masses too 1/N dn/dm jet [GeV -1 ] qq qq events p t,jets > 7 GeV anti-k t, R =.7 pp 7 TeV, Pythia 6.4, noue Look at jet mass distribution for 2 leading jets in pp W+jet events qq qq events m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 9

10 Problem #1b: there are lots of QCD jets 1/N dn/dm jet [GeV -1 ] qq qq + Wj mixture p t,jets > 7 GeV anti-k t, R = pp 7 TeV, Pythia 6.4, noue Look at jet mass distribution for 2 leading jets in pp W+jet events qq qq events mixture of the two, in rough proportion m jet [GeV] Jet mass gives clear sign of massive particles inside the jet; but QCD jets are massive too must learn to reject them Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 1

11 Problem #2: jet mass v. sensitive to PU m/n dn/dm jet p t,jets > 7 GeV anti-k t, R =.7 Wj events + 1 pileup pp 7 TeV, Pythia 6.4, noue Jet mass is extremely sensitive to pileup loss of mass resolution.5 loss of ability to find signals m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

12 Tagging & Grooming Two widely used terms though there s not a consensus about what they mean Tagging reduces the background, leaves much of signal Grooming improves signal mass resolution (removing pileup, etc.), without significantly changing background & signal event numbers Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

13 One core idea for tagging Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

14 Inside the jet mass QCD Jet Mass distribution Pythia 6.4, qq qq, no UE anti-k t, R=.7 QCD jet mass distribution has the approximate 1/N dn/dlog(m jet ) LHC, 7 TeV p t,jets > 7 GeV dn d ln m α s ln p tr m Sudakov Work from 8s and 9s + Almeida et al m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

15 z 1/N dn/dlog(m jet ) [1 jet 2partons] [1 jet 2partons] 1/N dn/dlog(m jet ) QCD Jet Mass distribution.8.9 QCD Jet Mass distribution Pythia 6.4, qq qq, no UE.7 Pythia 6.4, anti-k qq qq, t, R=.7 no UE.7.8 QCD Jet Mass distribution LHC, anti-k 7 t TeV, R=.7 Pythia 6.4, qq qq, no UE.6.7 LHC, anti-k 7 TeV.6 t, R=.7 LHC, 7 TeV p t,jets > 7 GeV.4.5 p t,jets > 7 GeV.4 p t,jets > 7 GeV m jet m[gev] 1 jet [GeV] m 1 jet [GeV] 1.1 1/N dn/dlog(m jet ) z.1 Inside the jet mass Inside the jet mass Inside the jet mass QCD jet mass distribution has the approximate QCD jet mass distribution has the approximate dn d ln m α s ln p tr dn Sudakov d ln m α s ln p tr m Sudakov Work from 8s and 9s Work from 8s and 9s Almeida et al 8 + Almeida et al 8 The logarithm comes from integral The logarithm comes from integral over over soft soft divergence divergence of of QCD: QCD: p t 2 m 2 p t 2R2 R2 dz dz z 2-body 2-body.1.1 phasespace m jet [GeV] m jet [GeV] A hard cut on z reduces QCD background & simplifies its shape Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

16 z 1/N dn/dlog(m jet ) [1 jet 2partons] [1 jet 2partons] [1 jet 2partons] 1/N dn/dlog(m jet ) QCD Jet Mass distribution QCD Jet Mass distribution Pythia 6.4, qq qq, no UE.7.8 Pythia 6.4, anti-k qq qq, t, R=.7 no UE.7.8 QCD QCD Jet Jet Mass Mass distribution distribution Pythia 6.4, LHC, anti-k qq qq, 7 t TeV, R=.7 no UE.6.7 Pythia 6.4, qq qq, no UE LHC, anti-k.7 7 TeV t, R=.7 anti-k.6 t, R=.7 LHC, 7 TeV.5.6 LHC, 7 TeV.6.5 p t,jets > 7 GeV p t,jets > 7 GeV.4 p t,jets > 7 GeV t,jets 7 GeV after cut on z > after cut on z > (a la BERS) (a la BERS) m jet m[gev] 1 jet m[gev] jet [GeV] 1 jet [GeV] 1 keep keep.1 1/N 1/N dn/dlog(m jet jet ) ) z z reject.1.1 cut on cut z on z reject Inside the jet mass Inside the jet mass Insidethe thejet mass QCD jet mass distribution has the approximate QCD jet jet mass mass distribution has has the the approximate dn d ln m α s ln p tr dn dn Sudakov ln s ln tr d ln m α s ln p tr m Sudakov Work from 8s and 9s Work Workfrom 8s 8sand 9s 9s et al 8 + Almeidaet etal al 8 8 The The logarithm comes from from integral over oversoft soft divergenceof of of QCD: p t 2 m 2 2 p p t 2R2 t 2 R2 R2 dz dz dz z 2-body 2-body 2-body phasespace.1 phasespace phasespace m m jet [GeV] jet [GeV] jet [GeV] A hard cut cut on on z reduces QCD background & simplifies its its shape Gavin Salam (CERN) Jets lecture 3 (Gavin Salam) Jets and CERN jet substructure Academic Training (4) March/April TASI, June / 2915

17 Inside the jet mass /N dn/dlog(m jet ) QCD Jet Mass distribution Pythia 6.4, qq qq, no UE anti-k t, R=.7 LHC, 7 TeV p t,jets > 7 GeV 1/N dn/dlog(m jet ) W+jet Jet Mass distribution Pythia 6.4, pp Wj, no UE anti-k t, R=.7 LHC, 7 TeV p t,jets > 7 GeV.1 after cut on z >.25.5 (a la BERS) m jet [GeV] m jet [GeV] 1 1 keep z.1 reject cut on z z body phasespace.1 2-body phasespace 1 1 m jet [GeV] 1 1 m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

18 Inside the jet mass /N dn/dlog(m jet ) QCD Jet Mass distribution after cut on z >.25 Pythia 6.4, qq qq, no UE anti-k t, R=.7 LHC, 7 TeV p t,jets > 7 GeV 1/N dn/dlog(m jet ) after cut on z >.25 W+jet Jet Mass distribution Pythia 6.4, pp Wj, no UE anti-k t, R=.7 LHC, 7 TeV p t,jets > 7 GeV (a la BERS) m jet [GeV] m jet [GeV] 1 1 keep keep z.1 reject cut on z z.1 reject cut on z.1 2-body phasespace.1 2-body phasespace 1 1 m jet [GeV] 1 1 m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

19 .25 qq qq + Wj mixture.2 m/n dn/dm jet.15.1 p t,jets > 7 GeV anti-k t, R =.7 Signal + bkgd after cut on z.5 + CA subjet (z >.25) m jet [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

20 One core idea for grooming [see blackboard] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

21 Grooming Groomed jet masses [Boost 21 writeup] Plain jet mass (anti-kt) Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 2

22 How do the tools work in practice? Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

23 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

24 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm dmin is dij = e 5 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

25 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

26 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

27 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

28 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

29 Identifying jet substructure: try out anti-k t p t /GeV 5 4 anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

30 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm dmin is dij =.8513 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

31 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

32 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

33 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

34 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

35 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

36 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm dmin is dib = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

37 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

38 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm dmin is dib = 1.96 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

39 Identifying jet substructure: try out anti-k t p t /GeV anti-k t algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). Anti-k t gradually makes its way through the secondary blob no clear identification of substructure associated with 2nd parton y

40 p t /GeV 5 4 k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

41 Identifying jet substructure: try out k t p t /GeV 5 4 k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

42 p t /GeV 5 4 k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

43 Identifying jet substructure: try out k t p t /GeV 5 4 k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

44 p t /GeV 5 4 k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z) y

45 Identifying jet substructure: try out k t p t /GeV k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

46 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

47 Identifying jet substructure: try out k t p t /GeV k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

48 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

49 Identifying jet substructure: try out k t p t /GeV k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

50 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

51 Identifying jet substructure: try out k t p t /GeV k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

52 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y

53 Identifying jet substructure: try out k t p t /GeV k t algorithm dmin is dij = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering Its last step is to merge two hard pieces. Easily undone to identify underlying kinematics y

54 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering Its last step is to merge two hard pieces. Easily undone to identify underlying kinematics y

55 p t /GeV k t algorithm dmin is dib = Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering Its last step is to merge two hard pieces. Easily undone to identify underlying kinematics y

56 p t /GeV k t algorithm Identifying jet substructure: try out k t How well can an algorithm identify the blobs of energy inside a jet that come from different partons? This is crucial for identifying the kinematic variables of the partons in the jet (e.g. z). k t clusters soft junk early on in the clustering y Its last step is to merge two hard pieces. Easily undone to identify underlying kinematics This meant it was the first algorithm to be used for jet substructure. Seymour 93 Butterworth, Cox & Forshaw 2

57 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

58 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

59 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

60 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

61 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

62 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

63 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

64 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? y

65 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV Cambridge/Aachen algorithm 5 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination y

66 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them y

67 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them y

68 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm DeltaR_{ij} = How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y

69 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y

70 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm DeltaR_{ij} > 2 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y

71 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y

72 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm DeltaR_{ij} > 2 How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y

73 Identifying jet substructure: Cam/Aachen p t /GeV 5 4 Cambridge/Aachen algorithm How well can an algorithm identify the blobs of energy inside a jet that come from different partons? C/A identifies two hard blobs with limited soft contamination, joins them, and then adds in remaining soft junk y The interesting substructure is buried inside the clustering sequence it s less contamined by soft junk, but needs to be pulled out with special techniques Butterworth, Davison, Rubin & GPS 8 Kaplan, Schwartz, Reherman & Tweedie 8 Butterworth, Ellis, Rubin & GPS 9 Ellis, Vermilion & Walsh 9 G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, / 1

74 anti-k t algorithm k t algorithm Cambridge/Aachen p t /GeV p t /GeV p t /GeV y y y Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

75 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL Zbb BACKGROUND Cluster event, C/A, R=1.2 Butterworth, Davison, Rubin & GPS 8 arbitrary norm.

76 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL Zbb BACKGROUND Fill it in, show jets more clearly Butterworth, Davison, Rubin & GPS 8 arbitrary norm.

77 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV.6.4 Consider hardest jet, m =15GeV Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

78 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV split: m =15GeV, max(m 1,m2) m =.92 repeat Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

79 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV split: m =139GeV, max(m 1,m2) m =.37 mass drop Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

80 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV check: y 12 p t2 p t1.7 OK + 2 b-tags (anti-qcd) Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

81 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV.6.4 R filt =.3 Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

82 pp ZH ν νb m H =115GeV Herwig Jimmy FastJet 2.3 SIGNAL 2 < p tz < 25 GeV m H [GeV] Zbb BACKGROUND.8 2 < p tz < 25 GeV.6.4 R filt =.3: take 3 hardest, m = 117 GeV Butterworth, Davison, Rubin & GPS m H [GeV] arbitrary norm.

83 pp ZH ννbb - Boosted Higgs analysis Cluster with a large R Undo the clustering into subjets, until a large mass drop is observed Re-cluster with smaller R, and keep only 3 hardest jets Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

84 Mass-Drop Tagger + Filtering in FastJet #include fastjet/tools/massdroptagger.hh #include fastjet/tools/filter.hh JetDefinition jet_def(cambridge_algorithm, 1.2); ClusterSequence cs(input_particles, jet_def); jets = sorted_by_pt(cs.inclusive_jets()); // define the tagger and use it double mu =.667, ycut =.9; MassDropTagger md_tagger(mu, ycut); PseudoJet tagged = md_tagger(jets[]); // check it was tagged OK by verifying (tagged!= ) // define the filter and use it Filter filter(.3,selectornhardest(3)); PseudoJet filtered = filter(tagged); // this is the Higgs!! The real analysis is slightly more refined (b-tagging, dynamical filter radius, etc) but the main features are already present here Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

85 Ellis, Vermilion and Walsh 9 Pruning Pruning [7, 8] takes an initial jet, and from its mass deduces apruningradiusr prune = R fact 2m p t,wherer fact is a parameter of the tagger. It then reclusters the jet and for every clustering step, involving objects a and b, it checks whether ab > R prune and min(p ta,p tb ) <z cut p t,(a+b),wherez cut is a second parameter of the tagger. If so, then the softer of the a and b is discarded. Otherwise a and b are recombined as usual. Clustering then proceeds with the remaining objects, applying the pruning check at each stage. #include fastjet/tools/pruner.hh // define pruner double zcut =.1, Rfact =.5; Pruner pruner(cambridge_algorithm, zcut, Rfact); PseudoJet pruned_jet = pruner(jet); Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

86 Krohn, Thaler & Wang 9 Trimming Different condition for retaining jets (pt-cut rather than nfilt hardest) with respect to filtering, but otherwise identical #include fastjet/tools/filter.hh // define trimmer Filter trimmer(.3,selectorptfractionmin(.3)); Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

87 different (2-body) substructure tools Detailed relative positions depend on physics context (and are possibly contentious!) Grooming filtering trimming pruning MDT jet mass N-sub jettiness Tagging Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

88 different (2-body) substructure tools Detailed relative positions depend on physics context (and are possibly contentious!) Grooming filtering trimming combined pruning MDT jet mass N-sub jettiness Tagging Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

89 different (2-body) substructure tools Detailed relative positions depend on physics context (and are possibly contentious!) Grooming filtering trimming combined pruning multivariate taggers MDT jet mass N-sub jettiness Tagging Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

90 different (2-body) substructure tools Detailed relative positions depend on physics context (and are possibly contentious!) Grooming filtering trimming template tagger combined pruning multivariate taggers Q-jets jet deconstruction MDT jet mass N-sub jettiness Tagging Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

91 3rd party tools for FastJet Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

92 Comparing top taggers Herwig, 5 < p t < 6 GeV Herwig++, 2 < p t < 8 GeV From the extensive Boost 211 report, which reviewed taggers discussed software, determined performance on MC, etc. Bottom line: some taggers clearly better than others. But many taggers behave similarly & details depend on analysis (+ MC choice) Gavin Gavin Salam Salam (CERN) (CERN/Princeton/CNRS) Jets Theory and jet of substructure Fat Jets (4) Higgs Hunting TASI, June / 28 88

93 Seeing W s and tops in a single jet s = 7 TeV W s in a single jet tops in a single jet 2-1 L dt = 4.7 fb Leading Fat Jet Mass [GeV] (a) Events / 2 GeV 12 1 CMS single-jet 8 W mass peak in events with a lepton and 6 separate b-tagged jet. 4 Uses pruning (+ mass-drop condition on split jet) 2 ATLAS -1 L dt = 4.7 fb s = 7 TeV Data 211 Z' (1 TeV) σ = 1.3 pb tt Multijet th Pruning + Mass Drop requirement NB: combined in IR unsafe way Leading Top-Quark Candidate Mass [GeV] with HEPTopTagger Figure 7. Signal region distributions of (a) the mass of the leading p T fat jet an Gavin Salam (CERN) Jets and of jet the substructure leading p T (4) top-quark candidate. Also shown are TASI, thejune prediction 213 for SM89t t (b)

94 Searches with substructure tools A range of techniques being used for varied BSM scenarios Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 9

95

96 EXTRA MATERIAL Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

97 What do we know currently? Boost 21 proceedings: The [Monte Carlo] findings discussed above indicate that while [pruning, trimming and filtering] have qualitatively similar effects, there are important differences. For our choice of parameters, pruning acts most aggressively on the signal and background followed by trimming and filtering. At the time: No clear picture of why the taggers might be similar or different No clear picture of how the parameter choices affect the taggers Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

98 The right MC study can already be instructive (testing on background [quark] jets) m/ d / dm m [GeV], for p t = 4 TeV plain jet mass Pythia 6 DW, parton-shower level, no UE, pp 14 TeV, p t,gen > 4 TeV, qq qq, R = 1 m/p t Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

99 The right MC study can already be instructive (testing on background [quark] jets) m/ d / dm m [GeV], for p t = 4 TeV plain jet mass.5 Mass-drop tagger (y cut =.9, µ=.67) Pruner (z cut =.1).4 Trimmer (z cut =.1, R trim =.2) m/p t Pythia 6 DW, parton-shower level, no UE, pp 14 TeV, p t,gen > 4 TeV, qq qq, R = 1 Different taggers are apparently quite similar Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

100 The right MC study can already be instructive (testing on background [quark] jets) m/ d / dm m [GeV], for p t = 4 TeV plain jet mass.5 Mass-drop tagger (y cut =.9, µ=.67) Pruner (z cut =.1).4 Trimmer (z cut =.1, R trim =.2) m/p t Pythia 6 DW, parton-shower level, no UE, pp 14 TeV, p t,gen > 4 TeV, qq qq, R = 1 But only for a limited range of masses Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

101 Pileup Jet non-perturbative phase (hadronisation) perturbative radiation background radiation (underlying event) Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

102 Pileup Jet non-perturbative phase (hadronisation) perturbative radiation background radiation (underlying event) background radiation (pileup) Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

103 Pileup Jet non-perturbative phase (hadronisation) perturbative radiation background radiation (underlying event) Out of time pileup (especially ATLAS) background radiation (pileup) Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

104 Effect of pileup on 2 TeV Z qq, M = 2 GeV no pileup qq, M = 2 GeV ~25 pileup 1/N dn/dbin / anti-k t, R=.7 Q w f=.12 = 26.1 GeV arxiv: /N dn/dbin / anti-k t, R=.7 Q w f=.12 = 51.9 GeV PU =.25 mb -1 arxiv: dijet mass [GeV] dijet mass [GeV] Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

105 Pileup for real ~ 2 m ATLAS a few cm Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

106 What goes into the jets? ATLAS Calorimeter towers, after pre-clustering them into topoclusters CMS Particle flow objects ~ 1) charged tracks Matthew Low s ideal detector 2) neutrals: calorimeter towers not associated with charged tracks (or leftover bits of calo if Ecalo Etrack Ecalo ) A CMS particle-flow expert would shudder at this description Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 1

107 What goes into the jets? ATLAS Calorimeter towers, after pre-clustering them into topoclusters CMS Particle flow objects ~ 1) charged tracks Matthew Low s ideal detector 2) neutrals: calorimeter towers not associated with charged tracks (or leftover bits of calo if Ecalo Etrack Ecalo ) A CMS particle-flow expert would shudder at this description Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 213 1

108 How do you remove pileup (PU)? 1. Offset method [Tevatron / old ATLAS] Count # of pileup vertices (npileup): p subtracted t,jet = p t,jet c n pileup c ~.5 GeV 2. CMS Throw out charged tracks not from the main primary vertex Use area/median(fastjet) method for neutral PU 3. Area/median (FastJet) method Determine density of pileup pt / unit area ρ p subtracted t,jet = p t,jet A jet Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

109 pt Add ghosts, infinitesimally soft particles, to track area of jet in y φ plane φ y Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

110 pt Add ghosts, infinitesimally soft particles, to track area of jet in y φ plane φ y Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

111 pt p tj / A j [GeV] Most jets are from pileup; all have similar pt / area 25 k t algorithm, R= y j Estimate ρ for a given event as = median jets pt,jet A jet Add ghosts, infinitesimally soft particles, to track area of jet in y φ plane φ y Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

112 pileup subtraction performance.15 k t, R=.7 no pileup 1/N dn/dm [GeV -1 ].1.5 LHC, high lumi Z at 2 TeV no pileup, sub pileup pileup, sub m [GeV] Used in CMS for neutral part of PU, and (I think) for recent ATLAS results Gavin Salam (CERN) Jets and jet substructure (4) TASI, June

113 Some observables particularly sensitive to PU Arbitrary units ATLAS Preliminary Data 211, Ldt = 4.7 fb anti-k t with R=1. LCW, No jet grooming jet 6 p < 8 GeV, η <.8 T 1 N PV 4 5 N PV 7 8 N PV 11 N PV 12-1 E.g. jet mass ATLAS Preliminary -1 Data 211, Ldt = 4.7 fb anti-k t with R=1. LCW, f =.5, R =.3 cut sub jet 6 p < 8 GeV, η <.8 T 1 N PV 4 5 N PV 7 8 N PV 11 N PV 12 For these, plain pileup subtraction is only OKish Leading jet mass, m (a) Data: anti-k t, R = 1.: Ungroomed jet 1 But why are we interested in the jet mass? (b) Data: anti-k t, R = 1.: Trimm its Gavin Salam (CERN).18 ATLAS Preliminary - Simulation Jets and jet substructure (4) ATLAS Preliminary TASI, June 213- Simula 14 its