Jets and Event-Shape Observables

Jets and event shapes are measurement functions for energy flow. A jet algorithm maps many particles to a smaller set of recombined momenta, while an event shape maps the entire final state to one or several numbers. Their perturbative usefulness rests on infrared and collinear safety, a declared recombination prescription, and a resolution scale that makes the measured question unambiguous.

Required background. Infrared and Collinear Safety supplies the unresolved-limit test. Measurement Functions and Inclusive Observables supplies the weighted phase-space definition and binning conventions.

Sequential recombination as a measurement map

For hadron-collider kinematics, a broad family of sequential algorithms uses

$$d_{ij} =\min(k_{Ti}^{2p},k_{Tj}^{2p}) \frac{\Delta R_{ij}^2}{R^2}, \qquad d_{iB}=k_{Ti}^{2p},$$

where

$$\Delta R_{ij}^2=(y_i-y_j)^2+(\phi_i-\phi_j)^2.$$

At each step, find the smallest distance. If it is $d_{ij}$, recombine $i$ and $j$; if it is $d_{iB}$, declare $i$ a jet and remove it. The parameter $R$ sets an angular resolution, while $p$ controls how soft and hard objects cluster. The four-vector recombination rule, tie-breaking convention, rapidity acceptance, and minimum jet transverse momentum are part of the observable.

The anti-$k_T$ choice $p=-1$ clusters soft radiation into nearby hard seeds and produces rigid conical boundaries in well-separated configurations. Its infrared and collinear properties and geometric behavior are established in Cacciari, Salam, and Soyez 2008, §§ 1–2, pp. 1–8. Naming an algorithm without $R$, recombination, and cuts is still an incomplete measurement definition.

The unresolved-limit test for jets

Append a soft momentum $q$ with $k_{Tq}\to0$. For an IRC-safe algorithm, it either clusters into a nearby jet or becomes a vanishing-energy jet without changing the hard-jet configuration in the limit. Replace a momentum $p$ by exactly collinear daughters $zp$ and $(1-z)p$; their mutual distance must vanish so they recombine before changing resolved jets.

This test must be applied to the complete observable after jet finding. A safe clustering algorithm followed by an unsafe operation can produce an unsafe result. Examples include labeling a jet by the flavor of its single hardest constituent or counting all jets with zero transverse-momentum threshold.

At finite resolution, near-degenerate clustering sequences may differ on boundaries. Those boundaries are harmless only when they form a measure-zero set and do not turn unresolved emissions into finite changes over a singular region.

Event shapes

In $e^+e^-$ annihilation at center-of-mass energy $Q$, thrust is

$$T=\max_{|\boldsymbol n|=1} \frac{\sum_i|\boldsymbol p_i\cdot\boldsymbol n|} {\sum_i|\boldsymbol p_i|}, \qquad \tau=1-T.$$

Two narrow back-to-back jets give $\tau\to0$, while a more isotropic event has larger $\tau$. The thrust axis is defined by the maximization; it is not an external input. Soft emissions perturb the sums continuously, and collinear daughters reproduce their parent’s contribution, so thrust is IRC safe.

Broadening-type observables weight momentum transverse to an axis,

$$B\sim\frac{1}{Q}\sum_i|\boldsymbol p_i\times\boldsymbol n|,$$

while energy correlation functions use products of energies and angular distances. Each definition has its own recoil response and singular regions. A formula that looks energy weighted is not automatically safe if the axis, constituent selection, or normalization changes discontinuously.

For any event shape $V$, the measurement function for a bin $[V_a,V_b)$ is

$$F_n^{[a,b)}(\{p\}_n) =\mathbf1_{[V_a,V_b)}(V(\{p\}_n)).$$

Writing this explicitly clarifies normalization and turns the IRC test into a statement about $V_{n+1}\to V_n$ away from bin boundaries.

Resolution scales and logarithmic regions

A jet veto $p_T^{\mathrm{veto}}\ll Q$, a small jet radius $R\ll1$, or an event-shape endpoint $V\ll1$ introduces logarithms such as

$$\alpha_s^n\ln^m\!\frac{Q}{p_T^{\mathrm{veto}}}, \qquad \alpha_s^n\ln^m R, \qquad \alpha_s^n\ln^m V.$$

IRC safety makes coefficients finite but does not keep these logarithms small. Fixed-order predictions must therefore identify the region where scale hierarchies require resummation or where nonperturbative power corrections become comparable to the measured value.

An observable that restricts radiation to only part of phase space can generate non-global logarithms. A simple independent-emission Sudakov factor then misses correlated radiation across the measured boundary. Globalness, clustering, recoil, and color structure must be checked before importing a standard resummation formula. Generic hadron-collider event shapes designed for resummation illustrate these constraints in Banfi, Salam, and Zanderighi 2010, §§ 2–3, pp. 4–22.

A reproducible definition template

For a jet or event-shape prediction, publish:

1. the input four-momenta and any particle-species exclusions;
2. the distance or shape formula, frame, and axis definition;
3. the recombination rule, radius, and stopping condition;
4. acceptance cuts, resolution thresholds, bins, and normalization;
5. the treatment of missing momentum, invisible particles, or beam remnants;
6. analytic soft and collinear limits, including cuts and labels;
7. the perturbative region and known logarithmic or power-correction boundaries.

This list is enough for an independent implementation. Detector calibration, pileup mitigation, tuning, and a current jet-substructure catalog are outside the page’s parton-level scope. No runnable observable laboratory is currently available, so the formulas and limit tests here provide the static reproducibility check.

Common pitfalls

Treating the algorithm name as a full definition. Parameters, recombination, acceptance, and cuts affect the measurement and must be stated.

Testing clustering but not the downstream label. Flavor tags, grooming steps, or leading-constituent requirements can reintroduce unresolved sensitivity.

Equating IRC safety with perturbative reliability everywhere. Endpoint, small-radius, veto, and non-global logarithms can require resummation; low scales can require nonperturbative input.

Comparing parton-level and detector-level jets as identical objects. Reconstruction and response are additional maps with their own uncertainties.

Check your understanding

Take a three-particle configuration and replace one momentum by two exactly collinear daughters. Work through the sequential distance ordering and show that the daughters recombine to the parent before the resolved jets change. Then add a soft momentum and verify the limiting hard jets are unchanged.

Where to continue

• Resummation and Fixed-Order Matching treats endpoint and veto logarithms without double counting fixed-order terms.
• Validation and Theory Uncertainties records algorithm, limit, integration, and perturbative checks.
• Hard, Jet, and Soft Factorization separates the dynamical regions probed by suitable jet observables.
• QCD Radiation, Jets, and Event Shapes develops the process-level QCD treatment.

References

• Banfi, Andrea, Gavin P. Salam, and Giulia Zanderighi. “Phenomenology of Event Shapes at Hadron Colliders.” Journal of High Energy Physics 06 (2010): 038. DOI. Open preprint.
• Cacciari, Matteo, Gavin P. Salam, and Gregory Soyez. “The Anti-$k_T$ Jet Clustering Algorithm.” Journal of High Energy Physics 04 (2008): 063. DOI. Open preprint.