PaperlayerView exact revision on arXiv

Source-linked AI summary

The Catchment Area of Jets

Matteo Cacciari, Gavin P. Salam, Gregory Soyez

arXiv:0802.1188v2hep-ph

TL;DR

The paper asks how jet areas quantify contamination from underlying-event and pileup radiation, whose effects can also modify hard-particle clustering. It defines passive and active areas and studies them analytically for three jet algorithms, with Monte Carlo comparisons and pileup-subtraction foundations. Active areas can differ substantially from πR^2 even for single-particle jets, while leading-order anomalous-dimension predictions agree well with hadron-level simulations.

  • Problem

    Underlying-event and pileup radiation substantially modify jet kinematics and clustering, motivating a quantitative measure of jets’ contamination susceptibility.

    Source: §1 · p. 3

  • Method

    The paper defines passive and active ghost-based jet areas and analyzes them for k_t, Cambridge/Aachen, and SISCone, including Monte Carlo comparisons and subtraction applications.

    Source: §1 · p. 4, §6 · p. 37

  • Results

    Active area for a single-particle SISCone jet is πR^2/4 rather than πR^2, while leading-order anomalous-dimension predictions agree remarkably well with hadron-level Monte Carlo area measurements.

    Source: §6 · p. 37

  • Takeaways & Limitations

    Jet areas support jet-by-jet pileup subtraction, for which individual jet areas matter because area fluctuations are significant and back-reaction is small.

    Source: §6 · p. 39

    The single-gluon calculation cannot contain all relevant hadron-level physics, so future work should examine multi-soft-gluon radiation and its matching to the non-perturbative regime.

    Source: §2.3 · p. 9

Abstract

from arXiv · show

The area of a jet is a measure of its susceptibility to radiation, like pileup or underlying event (UE), that on average, in the jet's neighbourhood, is uniform in rapidity and azimuth. In this article we establish a theoretical grounding for the discussion of jet areas, introducing two main definitions, passive and active areas, which respectively characterise the sensitivity to pointlike or diffuse pileup and UE radiation. We investigate the properties of jet areas for three standard jet algorithms, k_t, Cambridge/Aachen and SISCone. Passive areas for single-particle jets are equal to the naive geometrical expectation \pi R^2, but acquire an anomalous dimension at higher orders in the coupling, calculated here at leading order. The more physically relevant active areas differ from \pi R^2 even for single-particle jets, substantially so in the case of the cone algorithms like SISCone with a Tevatron Run-II split--merge procedure. We compare our results with direct measures of areas in parton-shower Monte Carlo simulations and find good agreement with the main features of the analytical predictions. We furthermore justify the use of jet areas to subtract the contamination from pileup.

1 Introduction

The paper addresses how underlying-event and pileup radiation modify jet kinematics and clustering at hadron colliders. It introduces jet areas to quantify contamination susceptibility and studies passive and active definitions across three algorithms, alongside analytical, Monte Carlo, and subtraction applications.

Motivation

  • Underlying-event radiation adds 3−5 GeV per unit rapidity at the Tevatron and 10−15 GeV at the LHC, while pileup can add up to 100 GeV per unit rapidity.These effects can be comparable to perturbative corrections for jets with transverse momenta of several tens of GeV.Source: §1 · p. 3
  • UE and pileup both contaminate jets directly and alter how hard-scatter particles are clustered into jets.The paper treats the latter mechanism as back-reaction on jet structure.Source: §1 · pp. 3–4

Jet-area concept

  • Jet area measures the rapidity–azimuth region from which a jet captures uniformly distributed UE or pileup radiation.Naive particle-surface and convex-hull definitions fail because particles are pointlike and neighboring hulls can overlap ambiguously.Source: §1 · p. 3

Jet-area definitions

  • Passive area scans one infinitely soft ghost and measures sensitivity to pointlike radiation, whereas active area counts densely added ghosts and measures sensitivity to diffuse radiation.Active and passive areas differ because active ghosts can cluster among themselves as well as with event particles.Source: §1 · pp. 3–4

Scope and approach

  • The study examines k_t, Cambridge/Aachen, and SISCone, using energy ordering and, where useful, the small-radius approximation at leading single-logarithmic order.It also studies area changes from soft emissions, pure ghost jets, UE/pileup back-reaction, Monte Carlo predictions, and pileup subtraction.Source: §1 · p. 4

2 Passive Area

Passive area measures where an infinitely soft ghost is captured by a jet, quantifying susceptibility to pointlike UE/PU contamination. For the three studied algorithms, it equals πR^2 for one-particle jets but develops algorithm-dependent corrections with soft radiation and can be represented geometrically for k_t jets.

2 Passive Area

  • A passive area is the region where one infinitely soft ghost clusters with a given jet; the scalar definition is used here for simplicity.The corresponding four-vector area has additional directional and mass information for larger jets, but scalar-area results generally carry over up to powers of R.Source: §2 · pp. 4–5

2.1 Areas for 1-particle configurations

  • For a single-particle jet, the passive area is πR^2 for inclusive k_t, Cambridge/Aachen, and SISCone.Source: §2.1 · p. 5

2.2 Areas for 2-particle configurations

  • For two strongly ordered particles, the passive area depends on their separation: collinear safety preserves πR^2 at zero separation, while distinct separation regimes produce different areas.For Δ12 < R the particles form one jet with an area based on the union of two radius-R circles; for R < Δ12 < 2R they form separate jets, and for Δ12 > 2R each has area πR^2.Source: §2.2 · pp. 5–6

2.2.2 Cambridge/Aachen

  • Cambridge/Aachen differs from k_t near Δ12 ∼ R/2 because its separation-only clustering measure changes the recombination order and therefore the passive area.For Δ12 > R, Cambridge/Aachen has the same area as k_t.Source: §2.2.2 · p. 7

2.2.3 SISCone

  • SISCone matches the other algorithms for R < Δ12 < 2R but shows large differences at smaller separations and only negative passive-area corrections in these configurations.For Δ12 < R, a single stable cone gives area πR^2; for R < Δ12 < 2R, split–merge assigns the ghost to the closer cone.Source: §2.2.3 · p. 7

2.3 Area scaling violation

  • These perturbative area corrections require an infrared cutoff because passive areas are infrared unsafe and depend on non-perturbative effects such as hadronisation.A single-gluon calculation omits multi-particle hadron-level structure, although the authors note that it agrees well with Monte Carlo features elsewhere.Source: §2.3 · p. 9
  • Soft-gluon radiation gives passive areas an algorithm-dependent logarithmic dependence on jet transverse momentum, strongest for k_t and much smaller for Cambridge/Aachen and SISCone.The k_t coefficient can be non-negligible, whereas the Cambridge/Aachen and SISCone coefficients are smaller and similar, with opposite sign for SISCone.Source: §2.3 · pp. 7, 8, 10

2.4 n-particle properties and the Voronoi area

  • For general configurations, only k_t passive areas admit a direct particle-by-particle construction: each area is a Voronoi cell intersected with a radius-R circle, and jet areas sum constituent-particle areas.The Voronoi construction directly determines k_t passive areas and coincides with passive and active definitions in dense events.Source: §2.4 · pp. 10–12

3 Active Area

Active area is defined by clustering a dense ensemble of infinitesimal ghosts with the event and measuring the ghosts captured by each jet. Because ghost clustering creates fluctuating substructures and boundaries, active areas require averaging over ghost ensembles and characterize susceptibility to diffuse soft radiation.

  • A dense coverage of infinitesimal ghosts is clustered with the event, allowing the number of captured ghosts to measure each jet’s active area.The ghost density is kept high while the total ghost transverse-momentum density remains infinitesimal.Source: §3 · pp. 12–13
  • Ghosts can cluster among themselves into macroscopic subjets, producing ragged jet boundaries that depend on the initial ghost configuration.This randomness persists even as the ghost density tends to infinity.Source: §3 · p. 12
  • A unique active-area value is obtained by averaging over many ghost sets in addition to taking the infinite-density limit.The number of ghost sets needed depends on the situation; physical events often require only a few sets.Source: §3 · pp. 12–13
  • Active area approximates a jet’s average susceptibility to high-density soft radiation such as minimum-bias pileup.The standard deviation across ghost ensembles measures contamination variability and is connected to attainable momentum resolution.Source: §3 · p. 13
  • Adding ghosts can create pure ghost jets that fill otherwise empty regions and resemble jets formed from purely soft pileup radiation.For SISCone, pure-ghost areas depend strongly on microscopic ghost-distribution details because of split–merge clustering.Source: §3 · p. 13

3.1 Areas for 1-particle configurations and for ghost jets

For one-particle and pure-ghost configurations, active areas differ between sequential-recombination algorithms and SISCone. kt and Cambridge/Aachen exhibit fluctuating, broadly distributed areas, whereas SISCone permits an analytic hard-jet result but has strongly procedure-dependent pure-ghost areas.

3.1.1 kt and Cambridge/Aachen

  • The kt and Cambridge/Aachen single-particle area calculations are performed numerically by clustering dense ghost ensembles with FastJet.The setup uses roughly 100 ghosts per unit area and studies jets away from the rapidity boundaries.Source: §3.1.1 · p. 14
  • For kt and Cambridge/Aachen, single-hard-particle active-area distributions are broad because ghost-density fluctuations seed subsequent growth of the soft jet region.The algorithms’ ability to adapt to local structure explains the propagation of random ghost fluctuations into the final area.Source: §3.1.1 · p. 14
  • Pure ghost jets in kt and Cambridge/Aachen typically have smaller active areas than jets containing a hard particle.This implies that hard jets can emerge more clearly from soft junk such as minimum-bias or pileup radiation.Source: §3.1.1 · p. 15

3.1.2 SISCone

  • SISCone has a fixed active area for a single hard particle, independent of infinitely dense ghost fluctuations, and this area is analytically tractable.Its split–merge geometry reduces the hard jet to a circle of radius R/2 for commonly used f values above fmax.Source: §3.1.2 · pp. 16–17
  • SISCone pure-ghost area distributions depend strongly on ghost-distribution details and the split–merge threshold, with smaller f producing occasional large ghost jets and f ≳0.6 producing mostly very small ones.The analysis recommends f = 0.75 to reduce the risk of monster jets, while retaining the infrared-safe ˜pt ordering variable.Source: §3.1.2 · pp. 17–18
  • Monster ghost jets can also arise with ordinary Pythia pileup momenta, so the issue is not restricted to the artificial ghost setup.The observed rate with SISCone’s ˜pt choice is about 5% of events, while other comparison variables can produce more frequent occurrences.Source: §3.1.2 · p. 17

3.2 Areas for 2-particle configurations

Two-particle active areas are studied as functions of the hard–soft separation, revealing smaller-than-passive areas for kt and Cambridge/Aachen and an analytically describable but more strongly reduced SISCone area. The SISCone behavior follows from how the second particle removes pure-ghost cones during split–merge clustering.

3.2.1 kt and Cambridge/Aachen

  • Two-particle kt and Cambridge/Aachen areas and their standard deviations are obtained numerically as functions of the hard–soft separation.The analysis explicitly tracks both the mean active area and fluctuations across ghost ensembles.Source: §3.2.1 · p. 18
  • For kt and Cambridge/Aachen, two-particle active areas remain smaller than passive areas while preserving their dependence on particle separation.The active area reaches the single-particle value only somewhat beyond Δ12 = 2R, unlike the passive area.Source: §3.2.1 · p. 19, Figure 8 · p. 19

3.2.2 SISCone

  • SISCone admits an analytic two-particle active-area result whose geometry changes across Δ12 < R, R < Δ12 < 2R, and Δ12 ≥ 2R.For Δ12 > 2R, the area becomes πR2/4.Source: §3.2.2 · pp. 19–20
  • The analytic SISCone treatment is limited to strong energy ordering because comparable transverse momenta introduce merge steps whose area effects are not analytically understood.This limitation applies once two or more particles have commensurate transverse momenta.Source: §3.2.2 · p. 19
  • SISCone’s two-particle active area is qualitatively and quantitatively farther from the passive result than the kt and Cambridge/Aachen areas.The comparison is shown against the other algorithms as a function of the hard–soft separation.Source: Figure 8 · p. 19, §3.2.2 · p. 20, Figure 10 · p. 21
  • The SISCone two-particle area can exceed its one-particle area because the second particle makes some previously reducing pure-ghost cones unstable.Removing those cones makes the split–merge reduction weaker than in the one-particle case.Source: §3.2.2 · pp. 20–21

3.3 Area scaling violation

Perturbative radiation modifies both the mean active area and its fluctuations, producing an area-scaling violation analogous to the passive case. The resulting active area remains infrared unsafe, with algorithm-dependent anomalous-dimension coefficients.

  • For active areas of sequential-recombination algorithms, the two-particle area can differ from the one-particle value even for separations beyond 2R.This removes the explicit 2R upper limit used in the corresponding passive-area integral.Source: §3.3 · p. 21
  • The active area remains dependent on Q0, indicating that it is infrared unsafe like the passive area.The paper treats this as a feature expected for sensible area definitions tied to UE or pileup sensitivity.Source: §3.3 · p. 21
  • The anomalous-dimension coefficients for kt and Cambridge/Aachen are close to their passive counterparts, whereas SISCone’s coefficient has the opposite sign.The SISCone coefficient is obtained from its analytic two-particle area, while the sequential-recombination results use numerical integration and Monte Carlo evaluation.Source: §3.3 · p. 22
  • The active-area correction from perturbative radiation reflects both changes in the mean area and changes in ghost-ensemble fluctuations.The analysis separates these two mechanisms when expanding the area contribution.Source: §3.3 · p. 22
  • The active-area calculation combines perturbative-emission averages with ghost-ensemble averages and neglects terms lacking logarithmic enhancement or higher-order αs dependence.This specifies the approximation used for the scaling-violation result.Source: §3.3 · pp. 21–22

3.4 n-particle properties and large-n behaviour

For dense events, distinct sensible jet-area definitions converge because particles sharpen jet boundaries; the convergence is bounded by boundary-tile scaling and supported by measured fractal dimensions.

3.4 n-particle properties and large-n behaviour

  • For kt active areas, the area can be expressed as a sum of individual particle areas, although no simple geometrical construction is known for those individual areas.This follows because ghosts cluster among themselves and with hard particles before hard particles cluster with one another.Source: §3.4.1 · p. 22

3.4.2 Equivalence of all areas for large n

  • The ambiguity between area definitions is bounded by the total area of boundary tiles, whose scaling depends on the jet boundary’s fractal structure.Dense coverage is characterized by an interparticle spacing λ much smaller than the jet radius R.Source: §3.4.2 · p. 23
  • As long as the boundary fractal dimension satisfies ℸ_JA < 2, differences between area definitions vanish in the infinitely dense limit; the studied algorithms satisfy 1 ≤ ℸ_JA < 2.The case ℸ_JA = 2 would make the boundary occupy area comparable to the jet itself, for which the authors find no evidence.Source: §3.4.2 · p. 23
  • The expected convergence is generally faster than the boundary-tile upper bound because saturating that bound would require same-sign area differences across every tile.Uncorrelated tile-level differences instead lead to stronger cancellation.Source: §3.4.2 · p. 23
  • For kt and Cambridge/Aachen, the measured boundary fractal dimensions are approximately ℸ_kt ≃ ℸ_Cam ≃ 1.20–1.25.The measurement uses pure ghost jets because their higher multiplicity facilitates reliable extraction, and the authors suspect the result also applies to single-particle jets.Source: §3.4.2 · p. 23
  • In dense events, active, passive, and Voronoi areas converge to the same result, enabling the use of whichever area is fastest to compute.This has practical relevance for highly populated events such as heavy-ion collisions.Source: §3.4.2 · pp. 23–24

4 Back reaction

Back reaction describes how finite-soft minimum-bias radiation changes which non-MB particles enter a jet, with effects depending strongly on the jet algorithm and on whether the background is pointlike or diffuse. The analysis finds algorithm-dependent scaling and generally modest average momentum effects, while diffuse backgrounds preserve the main functional structure but alter normalizations.

4 Back reaction

  • Back reaction is the change in a jet’s non-MB particle content caused by adding finite-soft minimum-bias radiation, and its size depends significantly on the jet algorithm.It can affect particles near the MB scale and, more rarely, particles much harder than that scale.Source: §4 · p. 24

4.1 Back reaction from pointlike minimum-bias

  • The pointlike calculation integrates over loss and gain probabilities for a soft perturbative emission, using effective back-reaction areas that quantify where MB can alter jet contents.The analysis examines the limit ptm ≪ pt2 ≪ pt1 and treats loss and gain separately.Source: §4.1 · pp. 25–26
  • For pointlike MB with pt2 ∼ ptm, the effective back-reaction area is about 0.5πR^2 for SISCone versus about 0.1πR^2 for kt and Cambridge/Aachen.At high pt2, pointlike MB never induces back reaction in SISCone, whereas sequential-recombination algorithms retain a suppressed probability ∼ptm/pt2.Source: §4.1 · p. 27
  • For pointlike MB, the average back-reaction momentum change scales with ln(pt1/ptm), with coefficients relatively small and comparable in magnitude to area-contamination coefficients.Its analytical form resembles ordinary minimum-bias contamination, although the underlying mechanisms differ.Source: §4.1 · p. 27
  • For finite pt1, SISCone has a residual probability ∼ptm/pt1 of changing jet structure even when pt2 is much harder than the MB scale, giving a non-zero momentum-dispersion coefficient.This is distinct from the strict high-pt2 limit in which pointlike MB produces no SISCone back reaction.Source: §4.1 · pp. 27–28

4.2 Back reaction from diffuse MB

  • For SISCone with f > f_max ≃ 0.391, a uniform diffuse background does not alter cone stability, and split–merge behavior keeps sufficiently separated soft particles in distinct jets.Corrections are expected to scale with spatial fluctuations in the MB transverse-momentum density.Source: §4.2 · p. 28
  • Diffuse MB preserves the main functional form of pointlike back reaction but changes the normalizations, reducing them by a factor of 2 for loss and approximately 10 for gain.The numerical comparison is made for kt and Cambridge/Aachen, with results shown for R = 1 and checked also for R = 0.7.Source: §4.2 · pp. 28–29, Figure 12 · p. 29
  • For diffuse MB, the average back-reaction expressions retain nearly identical forms to the pointlike case, with overall scale ρ and a modified logarithm argument.The coefficients are slightly smaller and remain modest relative to overall minimum-bias contamination.Source: §4.2 · p. 29

5 Areas in (simulated) real life

Monte Carlo studies test analytical jet-area predictions in realistic events and establish their usefulness for pileup subtraction. The comparisons show reasonable agreement for area distributions and mean/dispersion trends, while subtraction recenters contamination and narrows fluctuations.

5.1 Jet area distributions and anomalous dimension

  • At parton-shower level, passive-area distributions preserve πR^2 peaks for SISCone and Cambridge/Aachen, whereas k_t areas broaden substantially from collinear emissions.Cambridge/Aachen also extends above πR^2, while SISCone develops a shoulder below πR^2; these features reflect the algorithms’ differing sensitivity to soft emissions.Source: §5.1 · p. 30
  • Active and passive areas differ most for sparse parton-shower events, with SISCone extending to A = πR^2/4, but become much more similar after hadronisation, UE, and pileup.With many particles, only active areas are shown because the two distributions nearly coincide; SISCone retains the smallest dispersion and k_t the largest.Source: §5.1 · pp. 30, 32, Figure 13 · p. 31

5 Areas in (simulated) real life

  • The realistic-event study is qualitative for full area distributions, so mean active areas and dispersions provide the quantitative comparison to the analytical results.Herwig and Pythia simulations are used to test whether the single-gluon picture survives parton showering and hadronisation.Source: §5 · p. 30, §5.1 · p. 32

5.1 Jet area distributions and anomalous dimension

  • Analytical predictions agree reasonably with simulated mean active areas and remarkably well with their dispersions across algorithms and jet transverse momenta.The slopes are predicted, while Q0 is fitted from the mean; the standard deviation is then predicted and agrees with the simulations.Source: §5.1 · p. 32
  • The quantitative interpretation is limited by the small-angle, first-order analytical approximation and by generator-dependent behavior, especially the factor-of-two smaller k_t slope from Pythia showers.Finite ghost and particle densities may also distort SISCone results and force an extreme fitted Q0, while back-reaction estimates based on rare tails poorly represent typical events.Source: §5.1 · p. 32, §5.2 · p. 34

5.2 Jet areas and pileup subtraction

  • Jet-area subtraction successfully centers residual pileup contamination near zero and reduces transverse-momentum fluctuations, yielding narrower distributions after subtraction.The remaining fluctuation is consistent with the area-weighted local pileup variance, and the paper notes improved narrowing of reconstructed spectra such as a top-mass peak.Source: §5.2 · p. 35, Figure 16 · p. 36

6 Conclusions

The paper defines passive and active jet areas to quantify susceptibility to pointlike and diffuse UE/pileup radiation, then characterizes their algorithm dependence, fluctuations, and subtraction relevance. It finds substantial differences among algorithms and identifies both practical applications and scope limitations.

  • Passive and active areas use infinitely soft ghosts to model pointlike and diffuse UE/pileup, respectively, and apply to kt, Cambridge/Aachen, and SISCone.Source: §6 · p. 37
  • For single-particle jets, passive areas equal πR^2, whereas SISCone’s active area is πR^2/4, invalidating a general cone-area assumption.Source: §6 · p. 37
  • Soft-gluon emission changes jet areas and yields a leading-order anomalous dimension that agrees with the p_t dependence measured in hadron-level Monte Carlo simulations.Source: §6 · p. 37
  • Mean areas follow ⟨A_SISCone,R⟩≲⟨A_Cam,R⟩≲⟨A_kt,R⟩, suggesting SISCone is least affected by UE contamination under the stated balance of pointlike and diffuse components.Source: §6 · p. 37
  • The active-area and anomalous-dimension results rely on the small-R limit, while back reaction is not reducible to one number.Source: §6 · p. 39
  • Jet areas support jet-by-jet pileup subtraction, because area fluctuations require using each jet’s individual area and highly populated events make passive and active areas identical.Source: §6 · p. 39
  • The analysis uses leading-logarithmic first-order calculations and mainly mean areas, leaving area distributions and intermediate ghost structures for further study.Source: §6 · p. 39
  • For SISCone, f≃0.5 can produce monster jets, whereas f≃0.75 eliminates that problem; anti-kt achieves πR^2 passive and active areas with zero anomalous dimension for single-particle jets.Source: §6 · p. 39

A Definitions of the three jet algorithms

The paper studies kt, Cambridge/Aachen, and SISCone. The first two sequentially recombine particles using pair and beam distances, while SISCone identifies stable cones and resolves overlaps through split–merge.

B Transition from one-particle jet to soft jet.

A trigger particle immersed in diffuse soft background transitions between hard-particle and background-dominated area behavior as its transverse momentum decreases. The background sets an effective cutoff near πR^2ρ.

C Fluctuations of the active area

The appendix derives the leading-order mean and fluctuation formulas for active areas by averaging over perturbative emissions and ghost ensembles. It relates the appendix expressions algebraically to the main-text results.

Loading 0802.1188v2…