Jet reconstruction in W + jets events at the LHC Ulrike Schnoor Michigan State University High Energy Physics Institutsseminar IKTP TU Dresden, Nov 4, 010
Jet reconstruction in W + jets events at the LHC 1. Jets in hadronic collisions. Jet measurements at ATLAS 3. Jet reconstruction algorithms 4. Study on W + n jets Monte Carlo samples 5. Data study 6. Conclusions
1. Jets in hadronic collisions Jets and QCD Running coupling α s QCD confinement strong coupling at low energies no colored particles measured Asymptotic freedom small coupling at high energies asymptotically free Source: Particle Data Group 3
1. Jets in hadronic collisions proton-proton collision hard scattering partons hadronization jets in the calorimeter 4
1. Jets in hadronic collisions Experiment hadrons Jet reconstruction QCD Lagrangian quarks, gluons 5
1. Jets in hadronic collisions Reconstruction algorithms: group calorimeter signals into jets Find kinematic properties requirements: Detector independent Similar on all levels Fast execution Easy calibration 6
1. Jets in hadronic collisions Levels of reconstruction Parton level Truth = hadron = particle level: parton showering, hadronization, underlying event Calorimeter level: full ATLAS detector simulation 7
W + jets processes Background to Standard Model and Beyond Standard Model processes Top production Higgs boson production Test pqcd Detector performance tests http://www.physorg.com/news15581609.html Reconstruction of leptons, missing E T, jets 8
. Jet measurement in ATLAS Large Hadron Collider at CERN pp collider http://atlas.ch/atlas_photos/selected-photos/lhc/990606_01_layout_sch.jpg Center-of-mass energy 7 TeV 9
. Jet measurement in ATLAS ATLAS detector at the LHC Hadronic calorimeter: Barrel: Tile Calorimeter Endcap: Liquid Argon 10
. Jet measurement in ATLAS Tower building η ϕ = 0.1 0.1 Recombination E > 0 Calorimeter towers ProtoJets (em scale) Calorimeter cells (em scale) (em scale) Topological Clusters (em scale) Topological clustering: nearest neighbors (topo-)cluster level 11
Topological clusters hadron jets tower jets Topo-cluster calibration instead of jet calibration reconstruction more dynamic cluster jets Similar to running on hadron level Monte Carlo 1
. Jet measurement in ATLAS Tower building η ϕ = 0.1 0.1 Recombination E > 0 Calorimeter towers ProtoJets (em scale) Calorimeter cells (em scale) (em scale) jet finding Calorimeter Jets (em scale) Topological Clusters (em scale) Topological clustering: nearest neighbors (topo-)cluster level 13
. Jet measurement in ATLAS Hadronic calibration Calibration to particle level In-situ Calibration Calorimeter Jets (em scale) Calorimeter Jets (hadronic scale) Physics Jets (hadronic scale) Refined physics jets Cell weighting, jet energy scale Algorithm effects, additional corrections Underlying event, pileup correction 14
3. Jet reconstruction algorithms (Iterative) cone algorithms Clustering jets according to proximity in space SISCone Sequential recombination algorithms Clustering jets according to proximity in momentum space k T, anti-k T 15
Iterative cone algorithms 1. Initial seed particle i (= traditional way) Seed energy above certain threshold 16
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 17
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 18
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed 19
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed 0
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed 1
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed 3
Iterative cone algorithms 1. Initial seed particle i. Particles within cone of radius R: R ij = η i η j ) + ( ϕ i ϕ j ) < ( R 3. Resulting momentum is new seed 4. Iteration until stable cone is found 4
Cone algorithms overlapping cones Progressive removal (IC-PR) hardest particle is seed remove stable cone after each iteration repeat until no particle is left Split-merge (IC-SM) find all stable cones merge two overlapping cones if shared particles contain fraction f of softer jet s energy otherwise split cones, assign shared particles to cone whose axis is closer typically f = 0.5, 0.75 5
Cone algorithms Infrared and collinear safety (IRC) Infrared safety Emission of a soft parton does not change results of jet finding a) b) Collinear safety Collinear splitting of a particle does not change results of jet finding a) b) 6
Cone algorithms Infrared and collinear safety (IRC) Infrared unsafety IC-SM IRC safe quantities in fixed-order QCD calculations: jets 1 jet Collinear unsafety IC-PR Singularities from soft emissions collinear splitting 1 jet jets cancel with divergent contributions from virtual loops 7
SISCone algorithm Seedless infrared safe cone algorithm Computational geometric approach to find relevant initial subsets of particles using split-merge procedure IRC safe: no seeds 8
Sequential recombination algorithms Determine d ij and d ib of each particle i Find d min = min{d ij ;d ib } Distance measure: d ij = min( p d ib = p T p, i p p T, i, pt, j ) R R ij d min {d ij } d min {d ib } New particle k: µ k µ i p = p + p µ j i is a jet. Remove i. with R ij = ( η i η j ) + ( ϕ i ϕ j ) STOP if no particle left 9
Sequential recombination algorithms k T algorithm p = 1 clusters soft particles with small relative momentum first Distance measure: d ij = min( p d ib = p T p, i p p T, i, pt, j ) R R ij anti-k T algorithm p = -1 clusters hardest particle with particles that have small relative momentum first with R ij = ( η i η j ) + ( ϕ i ϕ j 30 )
SpartyJet Analysis framework allowing to run different algorithms with different parameters simultaneously http://projects.hepforge.org/spartyjet/ 31
SpartyJet Modular structure: Input Maker class read 4momenumt input jet list object Ntuple Maker class Store results in ROOT ntuples initial jet list final jet list JetAlgorithm class Perform any kind of operation on initial jet list JetTool e.g. cut selection, Pdg ID selection, JetTool Jet area tool, JetTool and of course: Jet finding tool (actual jet reconstruction) 3
SpartyJet Runs algorithms on topo-clusters Algorithms: ATLAS, CMS, CDF, D0, PYTHIA Celljet, FastJet library (k T, anti-k T, plug-in for SISCone) Jet tools: kinematic cut tools, jet area tools, PDG ID selection tool, p T density tool, http://projects.hepforge.org/spartyjet/ 33
4. Study on W + n jets Monte Carlo samples ALPGEN samples for W e ν + n jets CM energy 7 TeV n = 0, 1,, 3, 4, 5 Parton showering, hadronization: HERWIG Underlying event: JIMMY Cuts Strawman A selection η(jet) >.8 E T miss > 5 GeV Cross section measurement electron η <.47 W m T > 40 GeV σ = N L p T (jet) > 30 GeV electron p T > 0 GeV electron IsEM robusttight 34
Transverse momentum distributions W + 1 jet sample, R = 0.7, Cluster level algorithm σ /nb SISCone 0.1871 anti-k T 0.176 k T 0.1755-5 % - 0. % anti-k T and k T very similar below only considering SISCone and anti-k T 35
Transverse momentum distributions W + 1 jet, Cluster level, R = 0.7 σ (SISCone) = 0.1871 nb σ (anti-k T ) = 0.1755 nb σ(anti-k T ) < σ(siscone) p p T T ( SISCone) ( anti k ) T 100 00 300 400 500 36
Jet areas W + 1 jet, Cluster level, R = 0.7 a π R a π R (SISCone) = 1.057 (anti-k T ) =1.006 Ratio of area to circle of radius R 37
Jet areas k T : largest SISCone: smallest Anti-k T : most regular Determine area by clustering ghost particles 38
Cluster level correction Correction based on area Run k T (R = 0.5) algorithm on all jets (no p T cut) Determine median p T density Determine jet area grey: signal jets colored: low p T jets Subtract underlying event p T, corr = p ρ T, uncorr p T a 39
Transverse momentum distributions SISCone Anti-k T σ/nb 0.1871 0.1755 σ corr /nb 0.1787 0.1668 W + 1 jet, Cluster level, R = 0.7 σ corr < σ 40
Transverse momentum distributions W + 1 jet, Cluster level, R = 1.0 Δσ(SISCone) 0.0538 nb - % Δσ(anti-k T ) 0.0593 nb - 6% p T, corr = p ρ T, uncorr p T a Correction is larger for anti-k T : SISCone jet has larger area and larger p T 100 00 300 400 500 41
Transverse momentum distributions W + 3 jets, Parton level, leading jet, R = 1.0 σ (SISCone) = 0.0049 nb σ (anti-k T ) = 0.0079 nb Higher multiplicities: SISCone more likely to merge jets 100 00 300 400 500 4
Level comparison W + jets, SISCone, R = 0.7, leading jet Parton level, Cluster level p p T T ( Parton) ( Cluster) 43
5. Data study L1 Calorimeter stream: electron p T > 15 GeV Luminosity of the sample: 300 nb -1 Cuts looser cuts Cross section measurement σ = N L η(jet) > 3.1 E T miss > 10 GeV electron η <.47 W m T > 0 GeV p T (jet) > 0 GeV electron p T > 10 GeV electron IsEM loose 44
5. Data Study ATLAS Data Second leading jet, R = 0.7 inclusive W + 0 jets Cluster level Monte Carlo second jet, R = 0.7 45
6. Conclusions Differences of algorithms illuminate aspects of jet physics k T and anti-k T rather similar SISCone areas larger σ(siscone) > σ(anti-k T ) at low jet multiplicities SISCone more likely to merge two jets σ(siscone) < σ(anti-k T ) at high jet multiplicities Future study will investigate R = 0.5, 0.6 Data similar to inclusive distributions 46