Jets SubstructureIntroductionReminder: Jet algorithmsSubstructure ObservablesPileup and its removalResultsCoffeeEhud Duchovni, Weizmann Institute of Science, Physics in Collision 2012, Slovakia

IntroductionJet algorithms collect groups of particles based on a common property usually the direction of motion or small invariant mass w.r.t. a given direction.At high boosts the decay products of a heavy object, e.g. Z, W, H or t may be contained in a single jet.One would like to tell such jets from ordinary high pT QCD ones.Looking for jet substructure: searching for differences between particles that have been classified as similar. --> sort of contradiction Result is prone to be jet-algorithm dependent Observables must be defined with some care (mass, planarity,... Correct for detector effectsSubstructure is studies for highly boosted massive jets (E/m>2)

Effect of BoostThe higher the jets pT the narrower the jet is.pTJet width vs. its momentumE/M>2

Implication: Creation of Massive JetsHighly boosted hadronic Top quark may be fully contained by a single jetLow pTHigh pT

Boosted Top (Simulation)Top signal is largely enhancedNot all is lost! Boosted top jets appear as massive jets

MotivationNew physics must be characterized by high mass/energyBoosted jets are a very likely outcome of BSM physicsDecayed objects like top, W, Z and Higgs are likely to be contained by a single massive jetChallenge: Tell the left configuration from the right one

Reminder:Jet AlgorithmsWealth of jet algorithms Cone algorithmSequential re-combinationInfrared and collinear safetybadGoodKtAnti KtCambridge/Aachen10-1p

Observables of Jets Substructure (incomplete list)Energy spreadWidthMassAngularityEccentricitySub-clusteringPlanar flowSub-jettinessMany observables, partially overlappingGuy Gur-Ari, Michele Papucci, and Gilad Perez. arXiv: 1101.2905 [hep-ph]

Integrated and Differential Jet Shape

LightHeavyJets Width (Girth)

Separating Quark Jets from Gluon JetsDue to their color charge gluon jets are wider than quark jets and have higher (charged) particle multiplicityLinear radial moments (Girth)Girth(width) may help enriching jet sample with gluon/quark jets.Jason Gallicchio, Matthew D. SchwartzarXiv:1106.3076v2CA=3; CF=4/3Simulation

AngularityAngularity is defined by:Usually the a=-2 case is used namely:Angularity manifests itself by having a lower and upper limits. The lower corresponds to symmetric decay while the higher to a very asymmetric one.minmax

Angularity as a Separator of QCD/Signal(s)Angularity can tell QCD from boosted VAlmeida et al., hep-ph:0807.0234 using MadGraph Z=mJet/pTAngularity is suggested as a separator between QCD and heavy object jets

EccentricityE~0Eccentricity of jets is defined by 1 vmax/vmin, where vmax & vmin are the maximal & minimal valuesof variances of jet constituents along the principle & minor axes.E>0Eccentricity is weakly correlated with mass and width but strongly correlated with the planar flow

Planar FlowPlanar flow is helpful in separating two-body decays from multi-body decaysTwo-body decays, e.g. boosted H, W or Z should give very low planar flow while 3-body decays, e.g. top quarks should give relatively high value of planar flowchoose a plane that minimizes the pT outside, and measures the sum of this pT.Planar flow can distinguish between two and three body decays

N-SubjettinessJesse Thaler and Ken Van Tilburg arXiv 1011.2268 [hep-ph]

CorrelationsThe various shape/substructure measures are correlated with each other using MC. The exact level of correlation depends on the MC used, but general picture can be obtained by studying any of the MC. Here, Pythia 6.4 is used:AngularityPlanar flowEccentricityWidthMasspTAngularityPlanar flowEccentricityWidthpTMass

Correlations for Massive JetsWhile inspecting high mass (M>100GeV) jets, the correlations between substructure observables are changed.AngularityPlanar flowEccentricityWidthMasspTAngularityPlanar flowEccentricityWidthpTMass

Variation of Correlation as a Function of the MassThe move from all jets to heavy ones induces changes in the correlations between various observables. The effect is quite small.CijAll-CijMassive

AngularityPlanar flowEccentricitypTMassAngularity0.130.230.090.08-0.11Planar flow-0.19-0.09-0.140.10Eccentricity0.070.080.01pT0.280.35Mass0.04

Typical eventThe Pileup Problem

Jets ContaminationJets would include significant amount of incoherent energy from MI and UENeed to find a method to remove the excessive contribution

GroomingJet grooming: Family of algorithms that seek to get rid of the softer components in and around a jet from UE or pileup and keep the constituents of the hard scatter for further analysis. Pruning Filtering Trimmingbrush, clean, coach, comb, curry, dress, drill,educate, lick into shape, make attractive, make presentable, nurture, preen, prep, pretty up,prim, prime, primp, put through grind, put through mill, ready, refine, refresh, rub down,shape up, sleek, slick up, smarten up, spiff up,spruce up, tend, tidy, train, turn out Ambiguous procedures neighborhood-dependent

Mass Drop/Filtering(J. Butterworth, A. Davidson, M. Rubin, G. Salam; http://arxiv.org/abs/0802.2470) BDRSUndo last stage of C/AJ1(m1,pt1)J2(m2,pt2)m1>m2But there is another option of split:J(mj,ptj)Goal: remove incoherent energy and get the cleaned heavy jetmj1>mj2split to 2 partonic constituentsLump the two partons in one jet(blue) and the incoherent b.g. into the other

Mass Drop Criteria (BDRS)=m1/mjYcut=max(PT2(j1),pT2(j2)________________min(PT2(j1),pT2(j2)SmallSmallLarge0.092/3SmallLargeStopLargedrop j2 and use j1 as new j=m1/mjYcutstopIteratively removes incoherent energy depositions till system is consistent with the decay of a heavy object into two light ones.

FilteringThe jet is split into sub-jets in an attempt to unveil its inner structure (2, 3 or more sub-jets)three parameters,YCut, RFilt

Trimming

D. Krohn, J. Thaler, L. Wang, http://arxiv.org/abs/0912.1342Use a jet algorithm (kt or C/A) to create small subjets of size Rsub from the constituents of the large-R jet: any subjets failing pTi / pT < fcut are removedRe-cluster with RsubSelect pT,i/pT>fcutKeep only selected subjetsRsubfCutTwo free parameters, RSub & fCut

PruningS. Ellis, C. Vermilion, J. Walsh, http://arxiv.org/abs/0912.0033Apply: pT,i/(pT,i+pT,j)>Zcutor Rij>RcutRecombine jet constituents with C/A or kt while vetoing wide angle (Rcut) and softer (zcut) constituents. Does not recreate subjets but prunes at each point in jet reconstructionpT,i>alpha*pT,j ,alpha~1RCutZCutRemoveTwo parametersRCut and ZCut

Summary of Parameters StudiedJet Algorithms

AlgorithmRGroomingParametersAnti-kt1.0PruningRcut=0.1, 0.2, 0.3; Zcut=0.01, 0.1Anti-kt1.2Trimmingfcut=0.01, 0.03, 0.05; Rsub=0.2, 0.3C/AM/D+Filteringmu=0.20, 0.33, 0.67

Integrated and Differential Shapes

Sanity Check: Pileup Effect on Mass is Scaling with R & NvAs expected: the correction is larger for wider jetsmR4

Pileup Correction to the Mass MeasurementPhenomenological calculations seem to be in agreement with dataR. Alon , E. Duchovni, G. Perez, A.P. Pranko, P. Sinervo. PRD 84 (2011) 114025

Correcting for Pileup by FilteringFiltering seems to work very well

Looking at the Performance of Trimming and PruningTrimming works quite wellRecommended values are fCut=0.05 & RSub=0.3

Jets Massit works well!Pileup is partially under control

Can now look at the Data

Jets MassFiltering improves the agreement between the data and simulations

Jets Widthfor the record

Jets EccentricityEven the eccentricity is not eccentricR=0.6R=1.0

Planar Flow

Jets AngularityNice agreement with data and with upper and lower bounds on angularity

N-Sub-JettinessAlso looks nice

New 2011 Results: Angularity Borrowed from Lily Asquith Boost2012 talk

New 2011 Results: MassBorrowed from Lily Asquith Boost2012 talk

UsageX->ttWHZHothers

Tagging Top Jets at CMSBasic requests:jets with pT>250 GeV||0.05*pTJetR(1,2)>0.4-0.0004*pTIterate rejecting failed subjets and decluster again

jetABReverse the C/A one step backwardCheck if subclusters pass the requirementsReverse again the C/A one step backwardAABBCheck again if subsubclusters pass the requirementsPick up the (a,a,b,b)(A,b,b)(A,A,B)configuration for further studyRequire:3 or four subjets,minimal mij>50;subjets mass~mtop [100,250]:

Top Mass Reconstruction

Top Mistag Rate

Mass Bounds on Hypothetical tt ResonanceText

ATLAS Boosted tt Studydij>40

ConclusionNice topic for talksComplex issueNeed to correct for pileup and future looks,...Excellent testing ground for pQCD - predictions are already confronted with dataAlready used to identify highly boosted top-quark initiated jetsMay come handy in the search for Higgs via WH and ZH

Backup

Track-Based Vs Calorimeter-Based4606

Boosted TopText

tt ResonanceText

Jet ShapeText

N-Subjettiness of Inclusive QCD jetsText

Complementary Cone Technique

CMS W-TaggerUse pruned jets.The total mass of the jet is inside [60,100]Last Mass Drop is 0.42 last jets should be roughly equal in pt

Looking for Boosted TopJet Mass (leading pt=350, anti-kt R=1.0 with b-tag)Rediscovering the top in its hadronic decay mode

Infra-red and co-linear safety

integrated?

back defdefine intuitivelyref understandcheck which ones are not defined?add guy paperconclusionprepare diff plotA candidate Z boson event in the dimuon decay with 25 reconstructed vertices. This event was recorded on April 15th 2012 and demonstrates the high pileup environment in 2012 running (when the beta* was reduced to 0.6m). For this display the track pT threshold is 0.4 GeV and all tracks are required to have at least 3 Pixel and 6 SCT hits. The vertices shown are reconstructed using tracks with pT greater than 0.4 GeV, but with tighter requirements on the number of hits on the tracks than in the 2011 reconstruction.

explain vertices

picturescatteradd delta m proportional to R^4understand bettermiller DPElow range?definition1203.4606what is the limitboosted topc.s. limits of 1pb not yet sensitive to kkc.s. of sm+np must be lower than 2.6*c.s. of smwhere there other selection cuts?