Jets: where from, where to?

Transcription

1 Jets: where from, where to? Gavin P. Salam LPTHE, UPMC Paris 6 & CNRS London workshop on standard-model discoveries with early LHC data UCL, London, 31 March 29 Based on work with M. Cacciari, M. Dasgupta, L. Magnea, J. Rojo & G. Soyez

2

3 Jets, G. Salam, LPTHE (p. 3) Introduction Jets as projections p π π φ K LO partons NLO partons parton shower hadron level Jet Def n Jet Def n Jet Def n Jet Def n jet 1 jet 2 jet 1 jet 2 jet 1 jet 2 jet 1 jet 2 Projection to jets provides universal view of event

4 Jets, G. Salam, LPTHE (p. 4) Introduction QCD jets flowchart Jet (definitions) provide central link between expt., theory and theory And jets are an input to almost all analyses

5 Jets, G. Salam, LPTHE (p. 4) Introduction QCD jets flowchart Jet (definitions) provide central link between expt., theory and theory And jets are an input to almost all analyses

6 Jets, G. Salam, LPTHE (p. 5) Where from? What tools do we have?

7 Jets, G. Salam, LPTHE (p. 6) Where from? The cone algorithm CDF JetClu CDF MidPoint CDF MidPoint with searchcones D/ Run II Cone (midpoint) ATLAS Iterative Cone CMS Iterative Cone PxCone PyCell/CellJet/GetJet

8 Jets, G. Salam, LPTHE (p. 6) Where from? The cone algorithm Each cone involves CDF JetClu different code CDF MidPoint different physics CDF MidPoint with searchcones D/ Run II Cone (midpoint) ATLAS Iterative Cone CMS Iterative Cone PxCone PyCell/CellJet/GetJet

9 Jets, G. Salam, LPTHE (p. 6) Where from? The cone algorithm Each cone involves CDF JetClu different code CDF MidPoint different physics CDF MidPoint with searchcones Each cone is essentially D/ Run II Cone (midpoint) infrared unsafe ATLAS Iterative Cone collinear unsafe CMS Iterative Cone or some detector-influenced mixture of the PxCone two PyCell/CellJet/GetJet

10 Jets, G. Salam, LPTHE (p. 6) Where from? The cone algorithm Each cone involves CDF JetClu different code CDF MidPoint different physics CDF MidPoint with searchcones Each cone is essentially D/ Run II Cone (midpoint) infrared unsafe ATLAS Iterative Cone collinear unsafe CMS Iterative Cone or some detector-influenced mixture of the PxCone two PyCell/CellJet/GetJet Maybe half the cones are incompletely documented

11 Jets, G. Salam, LPTHE (p. 7) Where from? This is complex It s theoretically unsatisfactory (IR unsafety is inconsistent with perturbative calculations, even at LO) [and NLO calculations have seen $5 million investment] Last meaningful order JetClu, ATLAS MidPoint CMS it. cone Known at cone [IC-SM] [IC mp-sm] [IC-PR] Inclusive jets LO NLO NLO NLO W/Z + 1 jet LO NLO NLO NLO 3 jets none LO LO NLO [nlojet++] W/Z + 2 jets none LO LO NLO [MCFM] m jet in 2j + X none none none LO NLO

12 Jets, G. Salam, LPTHE (p. 7) Where from? This is complex It s theoretically unsatisfactory (IR unsafety is inconsistent with perturbative calculations, even at LO) [and NLO calculations have seen $5 million investment] But change has tended to be slow and hard-going E.g.: midpoint cone was proposed for Tevatron Run II: it was (only) a patch for earlier algorithm s IR safety issues its adoption was only partial at Tevatron Most of LHC s physics studies ignored it

13 Jets, G. Salam, LPTHE (p. 8) Where from? xc-sm CDF JetClu CDF MidPoint D/ Run II Cone (midpoint) ATLAS Iterative Cone Cones: what to use? SISCone find all stable cones run split merge on overlaps [GPS & Soyez 7] xc-pr CMS Iterative Cone PyCell/CellJet/GetJet anti-k t cluster min d ij = min(k 2 ti, k 2 tj ) R 2 ij if d ib = k 2 smallest, i jet ti [Cacciari, GPS & Soyez 8]

14 Jets, G. Salam, LPTHE (p. 8) Where from? xc-sm CDF JetClu CDF MidPoint D/ Run II Cone (midpoint) ATLAS Iterative Cone Cones: what to use? SISCone find all stable cones run split merge on overlaps [GPS & Soyez 7] xc-pr CMS Iterative Cone PyCell/CellJet/GetJet anti-k t cluster min d ij = min(k 2 ti, k 2 tj ) R 2 ij if d ib = k 2 smallest, i jet ti [Cacciari, GPS & Soyez 8]

15 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

16 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

17 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

18 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

19 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

20 Jets, G. Salam, LPTHE (p. 9) Where from? Build up of jets with anti-k t ( ) 1 d ij = min, 1 k 2 ti k 2 tj Recombine smallest R 2 ij

21 ICPR has circular jets But collinear unsafe So does anti-k t safe from theory point of view Cones with split-merge (SISCone) shrink to remove soft junk

22 ICPR has circular jets But collinear unsafe So does anti-k t safe from theory point of view Cones with split-merge (SISCone) shrink to remove soft junk

23 ICPR has circular jets But collinear unsafe So does anti-k t safe from theory point of view Cones with split-merge (SISCone) shrink to remove soft junk

24

25 Jets, G. Salam, LPTHE (p. 11) Where from? Sequential recombination algs The k t algorithm Find smallest d ij = min(k 2 ti, k2 tj ) R2 ij /R2 and recombine; If d ib = k 2 ti is smallest, call i a jet. The Cambridge/Aachen algorithm Repeatedly recombine objects with smallest R 2 ij, until all R ij > R Both involve a tradeoff: useful information from clustering hierarchy irregularity of the jets My favourite: Cam/Aachen (it s more easily twisted to fit your needs)

26 Jets, G. Salam, LPTHE (p. 11) Where from? Sequential recombination algs The k t algorithm Find smallest d ij = min(k 2 ti, k2 tj ) R2 ij /R2 and recombine; If d ib = k 2 ti is smallest, call i a jet. The Cambridge/Aachen algorithm Repeatedly recombine objects with smallest R 2 ij, until all R ij > R Both involve a tradeoff: useful information from clustering hierarchy irregularity of the jets My favourite: Cam/Aachen (it s more easily twisted to fit your needs)

27 Jets, G. Salam, LPTHE (p. 12) Where from? A full set of IRC-safe jet algorithms Generalise inclusive-type sequential recombination with d ij = min(k 2p ti,k 2p tj ) R 2 ij/r 2 d ib = k 2p ti Alg. name Comment time p = 1 k t Hierarchical in rel. k t CDOSTW 91-93; ES 93 N ln N exp. p = Cambridge/Aachen Hierarchical in angle Dok, Leder, Moretti, Webber 97 Scan multiple R at once N ln N Wengler, Wobisch 98 QCD angular ordering p = 1 anti-k t Cacciari, GPS, Soyez 8 Hierarchy meaningless, jets reverse-k t Delsart like CMS cone (IC-PR) N 3/2 SC-SM SISCone Replaces JetClu, ATLAS GPS Soyez 7 + Tevatron run II MidPoint (xc-sm) cones N 2 ln N exp. All these algorithms coded in (efficient) C++ at (Cacciari, GPS & Soyez 5-8)

28 Jets, G. Salam, LPTHE (p. 13) Where from? FastJet 2.3.x also contains CDF JetClu (legacy) CDF MidPoint cone (legacy) PxCone (legacy) FastJet 2.4 will add (in next few weeks) D/ Run II cone (legacy) ATLAS Iterative cone (legacy) CMS Iterative cone (legacy) Trackjet (legacy) A whole range of e + e algorithms Tools to help you build your own seq. rec. algorithms [NB: many algs available also in SpartyJet]

29 Jets, G. Salam, LPTHE (p. 13) Where from? FastJet 2.3.x also contains CDF JetClu (legacy) CDF MidPoint cone (legacy) PxCone (legacy) FastJet s FastJet inclusion 2.4 will of add many (inlegacy next few cones weeks) is not an endorsement of them. D/ Run II cone (legacy) They are toatlas be deprecated Iterative cone for any (legacy) new physics CMS Iterative analysis. cone (legacy) Trackjet (legacy) A whole range of e + e algorithms Tools to help you build your own seq. rec. algorithms [NB: many algs available also in SpartyJet]

30 Jets, G. Salam, LPTHE (p. 14) Understanding Can we understand our tools?

31 Jets, G. Salam, LPTHE (p. 15) Understanding Jet def n differences Jet definitions differ mainly in: 1. How close two particles must be to end up in same jet [discussed in the 9s, e.g. Ellis & Soper] 2. How much perturbative radiation is lost from a jet [indirectly discussed in the 9s (analytic NLO for inclusive jets)] 3. How much non-perturbative contamination (hadronisation, UE, pileup) a jet receives [partially discussed in 9s Korchemsky & Sterman 95, Seymour 97]

32 Jets, G. Salam, LPTHE (p. 16) Understanding Reach the reach of jet algorithms p t2 / p t1 p t1 R p t2 p t1 R p t2 1 jet? 2 jets? R 12 / R SISCone (xc-sm) reaches further for hard radiation than other algs

33 Jets, G. Salam, LPTHE (p. 16) Understanding Reach the reach of jet algorithms p t2 / p t1 p t1 R p t2 p t1 R p t2 1 jet? 2 jets? x = p t,2 /p t,1 1.5 T Cam/Aachen R CA = 1; R T =.4 Cam/ Aachen parton level R 12 / R R / R T SISCone (xc-sm) reaches further for hard radiation than other algs

34 Jets, G. Salam, LPTHE (p. 16) Understanding Reach the reach of jet algorithms p t2 / p t1 x = p t,2 /p t,1 1.5 p t1 R 1 jet? p t2 T Cam/Aachen R CA = 1; R T =.4 Cam/ Aachen parton level p t1 R 12 / R R 2 jets? p t R / R T x = p t,2 /p t,1 1.5 T anti-k t R CA = 1; R T =.4 anti-k t xcpr parton level R / R T SISCone (xc-sm) reaches further for hard radiation than other algs

35 Jets, G. Salam, LPTHE (p. 16) Understanding Reach the reach of jet algorithms p t2 / p t1 x = p t,2 /p t,1 1.5 p t1 R 1 jet? p t2 T Cam/Aachen R CA = 1; R T =.4 Cam/ Aachen parton level p t1 R 12 / R R 2 jets? p t R / R T x = p t,2 /p t,1 x = p t,2 /p t, T anti-k t R CA = 1; R T =.4 anti-k t xcpr parton level R / R T T SISCone (f=.75) R CA = 1; R T =.4 SISCone parton level R / R T SISCone (xc-sm) reaches further for hard radiation than other algs

36 Jets, G. Salam, LPTHE (p. 16) Understanding Reach the reach of jet algorithms p t2 / p t1 x = p t,2 /p t,1 1.5 p t1 R 1 jet? p t2 T Cam/Aachen R CA = 1; R T =.4 Cam/ Aachen parton level p t1 R 12 / R R 2 jets? p t R / R T x = p t,2 /p t,1 x = p t,2 /p t, T anti-k t R CA = 1; R T =.4 anti-k t xcpr parton level R / R T T SISCone (f=.75) R CA = 1; R T =.4 SISCone parton level R / R T SISCone (xc-sm) reaches further for hard radiation than other algs

37 Jets, G. Salam, LPTHE (p. 17) Understanding Perturbative p t Jet p t v. parton p t : perturbatively? The question s dangerous: a parton is an ambiguous concept Three limits can help you: Threshold limit e.g. de Florian & Vogelsang 7 Parton from color-neutral object decay (Z ) Small-R (radius) limit for jet One simple result p t,jet p t,parton p t = α { s π lnr 1.1CF quarks.94c A +.7n f gluons + O (α s ) only O (α s ) depends on algorithm & process cf. Dasgupta, Magnea & GPS 7

38 Jets, G. Salam, LPTHE (p. 18) Understanding Non-perturbative p t Jet p t v. parton p t : hadronisation? Hadronisation: the parton-shower hadrons transition Method: infrared finite α s à la Dokshitzer & Webber 95 prediction based on e + e event shape data could have been deduced from old work Korchemsky & Sterman 95 Seymour 97 Main result.4 GeV p t,jet p t,parton shower R { CF quarks gluons C A cf. Dasgupta, Magnea & GPS 7 coefficient holds for anti-k t ; see Mrinal s talk for k t alg.

39 Jets, G. Salam, LPTHE (p. 19) Understanding Non-perturbative p t Underlying Event (UE) Naive prediction (UE colour dipole between pp): { p t.4 GeV R2 2 CF q q dipole gluon dipole C A DWT Pythia tune or ATLAS Jimmy tune tell you: p t 1 15 GeV R2 2 This big coefficient motivates special effort to understand interplay between jet algorithm and UE: jet areas How does coefficient depend on algorithm? How does it depend on jet p t? How does it fluctuate? cf. Cacciari, GPS & Soyez 8

40 Jets, G. Salam, LPTHE (p. 2) Understanding Non-perturbative p t E.g. SISCone jet area 1. One hard particle, many soft Jet area = Measure of jet s susceptibility to uniform soft radiation Depends on details of an algorithm s clustering dynamics. SISCone, any R, f.391

41 Jets, G. Salam, LPTHE (p. 2) Understanding Non-perturbative p t E.g. SISCone jet area 2. One hard stable cone Jet area = Measure of jet s susceptibility to uniform soft radiation Depends on details of an algorithm s clustering dynamics. SISCone, any R, f.391

42 Jets, G. Salam, LPTHE (p. 2) Understanding Non-perturbative p t E.g. SISCone jet area 3. Overlapping soft stable cones Jet area = Measure of jet s susceptibility to uniform soft radiation Depends on details of an algorithm s clustering dynamics. SISCone, any R, f.391

43 Jets, G. Salam, LPTHE (p. 2) Understanding Non-perturbative p t E.g. SISCone jet area 4. Split the overlapping parts Jet area = Measure of jet s susceptibility to uniform soft radiation Depends on details of an algorithm s clustering dynamics. SISCone, any R, f.391

44 Jets, G. Salam, LPTHE (p. 2) Understanding Non-perturbative p t E.g. SISCone jet area 5. Final hard jet (reduced area) Jet area = Measure of jet s susceptibility to uniform soft radiation Depends on details of an algorithm s clustering dynamics. SISCone s area (1 hard particle) = 1 4 πr2 SISCone, any R, f.391

45 Jets, G. Salam, LPTHE (p. 21) Understanding Summary Jet algorithm properties: summary k t Cam/Aachen anti-k t SISCone reach R R R (1 + p t2 p t2 )R p t,pt αsc i π lnr lnr lnr ln1.35r p t,hadr.4 GeVC i R.7? 1? area = πr 2.81 ± ± πr 2 C i πb ln αs(q ) α s(rp t).52 ±.41.8 ± ±.7 In words: k t : area fluctuates a lot, depends on p t (bad for UE) Cam/Aachen: area fluctuates somewhat, depends less on p t anti-k t : area is constant (circular jets) SISCone: reaches far for hard radiation (good for resolution, bad for multijets), area is smaller (good for UE)

46 Jets, G. Salam, LPTHE (p. 22) Where to? Using our understanding (concentrate on R-dependence)

47 Jets, G. Salam, LPTHE (p. 23) Where to? Small v. large jet radius (R) HSBC Small jet radius Large jet radius single LO: jet radius irrelevant

48 Jets, G. Salam, LPTHE (p. 23) Where to? Small v. large jet radius (R) HSBC Small jet radius Large jet radius θ θ perturbative fragmentation: large jet radius better (it captures more)

49 Jets, G. Salam, LPTHE (p. 23) Where to? Small v. large jet radius (R) HSBC Small jet radius Large jet radius π + K L π K + π π + K L π K + π non perturbative hadronisation non perturbative hadronisation θ θ non-perturbative fragmentation: large jet radius better (it captures more)

50 Jets, G. Salam, LPTHE (p. 23) Where to? Small v. large jet radius (R) HSBC Small jet radius Large jet radius π + K L π K + π π + K L π K + π non perturbative hadronisation non perturbative hadronisation θ θ UE UE underlying ev. & pileup noise : small jet radius better (it captures less)

51 Jets, G. Salam, LPTHE (p. 23) Where to? Small v. large jet radius (R) HSBC Small jet radius Large jet radius multi-hard-parton events: small jet radius better (it resolves partons more effectively)

52 Jets, G. Salam, LPTHE (p. 24) Where to? What R is best for an isolated jet? PT radiation: q : p t α sc F π p t lnr Hadronisation: q : p t C F R.4 GeV Underlying event: q,g : p t R GeV 2 Minimise fluctuations in p t Use crude approximation: p 2 t p t 2 in small-r limit (?!) cf. Dasgupta, Magnea & GPS 7

53 Jets, G. Salam, LPTHE (p. 24) Where to? What R is best for an isolated jet? PT radiation: q : p t α sc F π p t lnr Hadronisation: q : p t C F R Underlying event:.4 GeV q,g : p t R GeV 2 Minimise fluctuations in p t Use crude approximation: p 2 t p t 2 δp t 2 pert + δp t 2 h + δp t 2 UE [GeV2 ] GeV quark jet LHC quark jets p t = 5 GeV δp t 2 h δp t 2 UE δp t 2 pert R in small-r limit (?!) cf. Dasgupta, Magnea & GPS 7

54 Jets, G. Salam, LPTHE (p. 24) Where to? What R is best for an isolated jet? PT radiation: q : p t α sc F π p t lnr Hadronisation: q : p t C F R Underlying event:.4 GeV q,g : p t R GeV 2 Minimise fluctuations in p t Use crude approximation: p 2 t p t 2 δp t 2 pert + δp t 2 h + δp t 2 UE [GeV2 ] LHC quark jets p t = 1 TeV 1 TeV quark jet δp t 2 pert δp t 2 UE R in small-r limit (?!) cf. Dasgupta, Magnea & GPS 7

55 Jets, G. Salam, LPTHE (p. 24) Where to? What R is best for an isolated jet? PT radiation: q : p t α sc F π p t lnr 4 At low p t, small R limits relative impact of UE Hadronisation: q : p t C 3 At high Fp t, perturbative effects dominate over.4 GeV R non-perturbative R best 1. Underlying event: q,g : p t R GeV 2 Minimise fluctuations in p t Use crude approximation: p 2 t p t 2 δp t 2 pert + δp t 2 h + δp t 2 UE [GeV2 ] LHC quark jets p t = 1 TeV 1 TeV quark jet δp t 2 pert δp t 2 UE R in small-r limit (?!) cf. Dasgupta, Magnea & GPS 7

56 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.3 qq, M = 1 GeV SISCone, R=.3, f=.75 Q w f=.24 = 24. GeV arxiv: p Resonance X dijets q q X q p dijet mass [GeV] q

57 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.3 qq, M = 1 GeV SISCone, R=.3, f=.75 Q w f=.24 = 24. GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

58 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.4 qq, M = 1 GeV SISCone, R=.4, f=.75 Q w f=.24 = 22.5 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

59 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.5 qq, M = 1 GeV SISCone, R=.5, f=.75 Q w f=.24 = 22.6 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

60 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.6 qq, M = 1 GeV SISCone, R=.6, f=.75 Q w f=.24 = 23.8 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

61 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.7 qq, M = 1 GeV SISCone, R=.7, f=.75 Q w f=.24 = 25.1 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

62 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.8 qq, M = 1 GeV SISCone, R=.8, f=.75 Q w f=.24 = 26.8 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

63 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R =.9 qq, M = 1 GeV SISCone, R=.9, f=.75 Q w f=.24 = 28.8 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

64 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R = 1. qq, M = 1 GeV SISCone, R=1., f=.75 Q w f=.24 = 31.9 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

65 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R = 1.1 qq, M = 1 GeV SISCone, R=1.1, f=.75 Q w f=.24 = 34.7 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

66 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R = 1.2 qq, M = 1 GeV SISCone, R=1.2, f=.75 Q w f=.24 = 37.9 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

67 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R = 1.3 qq, M = 1 GeV SISCone, R=1.3, f=.75 Q w f=.24 = 42.3 GeV arxiv: p Resonance X dijets jet q q X q p dijet mass [GeV] jet q

68 Jets, G. Salam, LPTHE (p. 25) Where to? Dijet mass: scan over R [Pythia 6.4] 1/N dn/dbin / R = 1.3 qq, M = 1 GeV SISCone, R=1.3, f=.75 Q w f=.24 = 42.3 GeV arxiv: ρ L from Q w f= qq, M = 1 GeV SISCone, f=.75 arxiv: dijet mass [GeV] R After scanning, summarise quality v. R. Minimum BEST picture not so different from crude analytical estimate

69 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 1 GeV qq, M = 1 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

70 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 15 GeV qq, M = 15 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

71 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 2 GeV qq, M = 2 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

72 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 3 GeV qq, M = 3 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

73 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 5 GeV qq, M = 5 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

74 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 7 GeV qq, M = 7 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

75 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 1 GeV qq, M = 1 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

76 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 2 GeV qq, M = 2 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

77 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 4 GeV qq, M = 4 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

78 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 4 GeV qq, M = 4 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

79 Jets, G. Salam, LPTHE (p. 26) Where to? Scan through q q mass values ρ L from Q w f= m qq = 4 GeV qq, M = 4 GeV SISCone, f=.75 arxiv: Best R is at minimum of curve Best R depends strongly on mass of system Increases with mass, just like crude analytical prediction NB: current analytics too crude BUT: so far, LHC s plans involve running with fixed smallish R values e.g. CMS arxiv: R NB: 1, plots for various jet algorithms, narrow qq and gg resonances from Cacciari, Rojo, GPS & Soyez 8

80 Jets, G. Salam, LPTHE (p. 27) Where to? quality: 5 algorithms, 3 processes

81 Jets, G. Salam, LPTHE (p. 27) Where to? quality: 5 algorithms, 3 processes

82 Jets, G. Salam, LPTHE (p. 27) Where to? quality: 5 algorithms, 3 processes

83 Jets, G. Salam, LPTHE (p. 28) Where to?

84 Jets, G. Salam, LPTHE (p. 29) Conclusions These studies show that: Choice of jet definition matters (it s worth a factor of in lumi) There is no single best jet definition LHC will span two orders of magnitude in p t (experiments should build in flexibility) There is logic to the pattern we see (it fits in with crude analytical calculations)

85 Jets, G. Salam, LPTHE (p. 3) Conclusions Towards jetography These studies motivate a more systematic approach: More realistic analytical calculations e.g. using known differences between algorithms Consideration of backgrounds Consideration of multi-jet signals (relation to boosted W/Z/H/top (subjet) ID methods) Design of more optimal jet algorithms (R alone may not give enough freedom cf. filtering )

86 Jets, G. Salam, LPTHE (p. 3) Conclusions Towards jetography These studies motivate a more systematic approach: More realistic analytical calculations e.g. using known differences between algorithms Consideration of backgrounds Consideration of multi-jet signals (relation to boosted W/Z/H/top (subjet) ID methods) Design of more optimal jet algorithms (R alone may not give enough freedom cf. filtering ) Jetography: auto-focus for jets

87 Jets, G. Salam, LPTHE (p. 31) Conclusions To conclude Past experience (CDF/JetClu) suggests that if an IRC unsafe legacy algorithm remains available within an experiment, the majority of analyses will use it. Maybe not the pure QCD analyses But all the others There are no longer any good reasons to prefer IRC unsafe algorithms. As a community, let us try and make sure LHC does the right job. So we get full value form perturbative QCD And so that we can move on to more useful questions

88 Jets, G. Salam, LPTHE (p. 32) Extras EXTRAS

89 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)]

90 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)]

91 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)]

92 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)]

93 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)]

94 Jets, G. Salam, LPTHE (p. 33) Extras ICSM Cone basics I: IC-SM Many cone algs have two main steps: Find some/all stable cones cone pointing in same direction as the momentum of its contents Resolve cases of overlapping stable cones By running a split merge procedure [Blazey et al. (Run II jet physics)] Qu: How do you find the stable cones? Until recently used iterative methods: use each particle as a starting direction for cone; use sum of contents as new starting direction; repeat. Iterative Cone with Split Merge (IC-SM) e.g. Tevatron cones (JetClu, midpoint) ATLAS cone

95 Jets, G. Salam, LPTHE (p. 34) Extras ICSM Seeded IC-SM: infrared issue Use of seeds is dangerous p T (GeV/c) stable cones from seeds 1 1 Extra soft particle adds new seed changes final jet configuration. This is IR unsafe. Kilgore & Giele 97 Partial fix: add extra seeds at midpoints of all pairs, triplets,...of stable cones. Adopted for Tevatron Run II But only postpones the problem by one order... Analogy: if you rely on Minuit to find minima of a function, in complex cases, results depend crucially on starting points

96 Jets, G. Salam, LPTHE (p. 34) Extras ICSM Seeded IC-SM: infrared issue Use of seeds is dangerous p T (GeV/c) add soft particle 1 1 Extra soft particle adds new seed changes final jet configuration. This is IR unsafe. Kilgore & Giele 97 Partial fix: add extra seeds at midpoints of all pairs, triplets,...of stable cones. Adopted for Tevatron Run II But only postpones the problem by one order... Analogy: if you rely on Minuit to find minima of a function, in complex cases, results depend crucially on starting points

97 Jets, G. Salam, LPTHE (p. 34) Extras ICSM Seeded IC-SM: infrared issue Use of seeds is dangerous p T (GeV/c) resolve overlaps 1 1 Extra soft particle adds new seed changes final jet configuration. This is IR unsafe. Kilgore & Giele 97 Partial fix: add extra seeds at midpoints of all pairs, triplets,...of stable cones. Adopted for Tevatron Run II But only postpones the problem by one order... Analogy: if you rely on Minuit to find minima of a function, in complex cases, results depend crucially on starting points

98 Jets, G. Salam, LPTHE (p. 34) Extras ICSM Seeded IC-SM: infrared issue Use of seeds is dangerous p T (GeV/c) resolve overlaps 1 1 Extra soft particle adds new seed changes final jet configuration. This is IR unsafe. Kilgore & Giele 97 Partial fix: add extra seeds at midpoints of all pairs, triplets,...of stable cones. Adopted for Tevatron Run II But only postpones the problem by one order... Analogy: if you rely on Minuit to find minima of a function, in complex cases, results depend crucially on starting points

99 Jets, G. Salam, LPTHE (p. 35) Extras ICSM pt/gev pt/gev Midpoint IR problem GeV y Stable cones with midpoint: {1,2} & {3} {1,2} & {2,3} & {3} Jets with midpoint (f =.5) {1,2} & {3} {1,2,3} 3 y Midpoint cone alg. misses some stable cones; extra soft particle extra starting point extra stable cone found MIDPOINT IS INFRARED UNSAFE Or collinear unsafe with seed threshold

100 Jets, G. Salam, LPTHE (p. 35) Extras ICSM pt/gev pt/gev Midpoint IR problem GeV y Stable cones with midpoint: {1,2} & {3} {1,2} & {2,3} & {3} Jets with midpoint (f =.5) {1,2} & {3} {1,2,3} 3 y Midpoint cone alg. misses some stable cones; extra soft particle extra starting point extra stable cone found MIDPOINT IS INFRARED UNSAFE Or collinear unsafe with seed threshold

101 Jets, G. Salam, LPTHE (p. 35) Extras ICSM pt/gev pt/gev Midpoint IR problem GeV y Stable cones with midpoint: {1,2} & {3} {1,2} & {2,3} & {3} Jets with midpoint (f =.5) {1,2} & {3} {1,2,3} 3 y Midpoint cone alg. misses some stable cones; extra soft particle extra starting point extra stable cone found MIDPOINT IS INFRARED UNSAFE Or collinear unsafe with seed threshold

102 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

103 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

104 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

105 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

106 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

107 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

108 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

109 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

110 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

111 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

112 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

113 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat

114 Jets, G. Salam, LPTHE (p. 36) Extras ICPR Iterative Cone [with progressive removal] Procedure: Find one stable cone By iterating from hardest seed particle Call it a jet; remove its particles from the event; repeat Iterative Cone with Progressive Removal (IC-PR) e.g. CMS it. cone, [Pythia Cone, GetJet],... NB: not same type of algorithm as Atlas Cone, MidPoint, SISCone

115 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

116 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

117 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

118 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

119 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

120 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

121 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

122 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

123 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

124 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

125 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

126 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

127 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

128 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

129 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

130 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

131 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity jet 2 Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

132 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity jet 2 Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

133 Jets, G. Salam, LPTHE (p. 37) Extras ICPR ICPR iteration issue 5 cone iteration cone axis cone p T (GeV/c) jet 1 rapidity jet 2 Collinear splitting can modify the hard jets: ICPR algorithms are collinear unsafe = perturbative calculations give

134 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

135 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

136 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

137 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

138 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

139 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

140 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

141 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

142 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

143 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

144 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

145 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

146 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

147 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

148 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

149 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

150 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

151 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Next cone edge on particle Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

152 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

153 Jets, G. Salam, LPTHE (p. 38) Extras SCSM Seedless [Infrared Safe] cones (SC-SM) p t /GeV Aim to identify all stable cones, independently of any seeds Procedure in 1 dimension (y): find all distinct enclosures of radius R by repeatedly sliding a cone sideways until edge touches a particle check each for stability then run usual split merge In 2 dimensions (y,φ) can design analogous procedure SISCone GPS & Soyez y This gives an IRC safe cone alg.

154 Jets, G. Salam, LPTHE (p. 39) Extras IRC safety Magnitude of IR safety impact (midpoint) dσ midpoint(1) /dp t / dσ SISCone /dp t (a) Pythia 6.4 pp s = 1.96 TeV hadron-level (with UE) hadron-level (no UE) parton-level R=.7, f=.5, y < p t [GeV] Impact depends on context: from a few % (inclusive) to 5% (exclusive) rel. diff. for dσ/dm 2 rel. diff. for dσ/dm NLOJet R=.7, f=.5 Mass spectrum of jet 2 midpoint() -- SISCone SISCone M (GeV) NLOJet R=.7, f=.5 R 23 < 1.4 Mass spectrum of jet 2 midpoint() -- SISCone SISCone M (GeV)

155 Jets, G. Salam, LPTHE (p. 4) Extras IRC safety Does lack of IRC safety matter? I do searches, not QCD. Why should I care about IRC safety? If you re looking for an invariant mass peak, it s not 1% crucial IRC unsafety R is ill-defined A huge mass peak will stick out regardless Well, actually my signal s a little more complex than that... If you re looking for an excess over background you need confidence in backgrounds E.g. some SUSY signals Check W+1 jet, W+2-jets data against NLO in control region Check W+n jets data against LO in control region Extrapolate into measured region IRC unsafety means NLO senseless for simple topologies, LO senseless for complex topologies Breaks consistency of whole Wastes 5,,$/ /CHF/e But I like my cone algorithm, it s fast, has good resolution, etc. Not an irrelevant point has motivated significant work

156 Jets, G. Salam, LPTHE (p. 4) Extras IRC safety Does lack of IRC safety matter? I do searches, not QCD. Why should I care about IRC safety? If you re looking for an invariant mass peak, it s not 1% crucial IRC unsafety R is ill-defined A huge mass peak will stick out regardless Well, actually my signal s a little more complex than that... If you re looking for an excess over background you need confidence in backgrounds E.g. some SUSY signals Check W+1 jet, W+2-jets data against NLO in control region Check W+n jets data against LO in control region Extrapolate into measured region IRC unsafety means NLO senseless for simple topologies, LO senseless for complex topologies Breaks consistency of whole Wastes 5,,$/ /CHF/e But I like my cone algorithm, it s fast, has good resolution, etc. Not an irrelevant point has motivated significant work

157 Jets, G. Salam, LPTHE (p. 4) Extras IRC safety Does lack of IRC safety matter? I do searches, not QCD. Why should I care about IRC safety? If you re looking for an invariant mass peak, it s not 1% crucial IRC unsafety R is ill-defined A huge mass peak will stick out regardless Well, actually my signal s a little more complex than that... If you re looking for an excess over background you need confidence in backgrounds E.g. some SUSY signals Check W+1 jet, W+2-jets data against NLO in control region Check W+n jets data against LO in control region Extrapolate into measured region IRC unsafety means NLO senseless for simple topologies, LO senseless for complex topologies Breaks consistency of whole Wastes 5,,$/ /CHF/e But I like my cone algorithm, it s fast, has good resolution, etc. Not an irrelevant point has motivated significant work

158 Jets, G. Salam, LPTHE (p. 4) Extras IRC safety Does lack of IRC safety matter? I do searches, not QCD. Why should I care about IRC safety? If you re looking for an invariant mass peak, it s not 1% crucial IRC unsafety R is ill-defined A huge mass peak will stick out regardless Well, actually my signal s a little more complex than that... If you re looking for an excess over background you need confidence in backgrounds E.g. some SUSY signals Check W+1 jet, W+2-jets data against NLO in control region Check W+n jets data against LO in control region Extrapolate into measured region IRC unsafety means NLO senseless for simple topologies, LO senseless for complex topologies Breaks consistency of whole Wastes 5,,$/ /CHF/e But I like my cone algorithm, it s fast, has good resolution, etc. Not an irrelevant point has motivated significant work

159 Jets, G. Salam, LPTHE (p. 4) Extras IRC safety Does lack of IRC safety matter? I do searches, not QCD. Why should I care about IRC safety? If you re looking for an invariant mass peak, it s not 1% crucial IRC unsafety R is ill-defined A huge mass peak will stick out regardless Well, actually my signal s a little more complex than that... If you re looking for an excess over background you need confidence in backgrounds E.g. some SUSY signals Check W+1 jet, W+2-jets data against NLO in control region Check W+n jets data against LO in control region Extrapolate into measured region IRC unsafety means NLO senseless for simple topologies, LO senseless for complex topologies Breaks consistency of whole Wastes 5,,$/ /CHF/e But I like my cone algorithm, it s fast, has good resolution, etc. Not an irrelevant point has motivated significant work