Jets and infrared-safe observables

Real-emission terms can become singular when a radiated parton has vanishing energy or when two massless partons become collinear. Virtual corrections contain related singular structure. For a suitable sufficiently inclusive observable, contributions associated with experimentally indistinguishable states cancel after the terms are combined, with initial-state collinear effects absorbed into PDFs where appropriate. The observable must assign compatible values in these limits; a bare count of partons generally fails this test.

Infrared and collinear safety can be expressed through a measurement function $F$. Adding a momentum that tends to zero should recover the old observable. Replacing one momentum by two exactly collinear momenta whose sum is the original should do the same. These are limiting requirements, not a statement that any finite-energy extra particle can never change a jet. A well-defined transverse-momentum threshold can change a selected sample near a boundary without automatically violating the limiting property.

For anti-$k_T$, compute pair distances using inverse squared transverse momenta and angular separation $\Delta R^2=(\Delta y)^2+(\Delta\phi)^2$, with the azimuthal difference wrapped around its periodic range. Also compute a beam distance for each object. Find the smallest distance globally: a pair minimum triggers recombination, while a beam minimum declares that object a completed jet. Recompute distances after each recombination; a single decision does not define the whole event.

Because the inverse transverse-momentum weight favors hard objects, soft radiation is clustered around hard cores in a geometrically regular way. The radius parameter $R$ controls the angular reach in rapidity and azimuth, not a literal fixed distance in meters. The usual four-vector recombination scheme adds constituent four-momenta, allowing a jet to have a nonzero invariant mass. Different algorithms and radius choices define different observables, so a theory-data comparison must state both.

Infrared safety makes a perturbative expansion meaningful, but does not ensure that a low fixed order is numerically accurate. Small radii, vetoes, or sharply restricted event shapes can generate large logarithms that require resummation. Hadronization and activity associated with the rest of the collision can shift reconstructed jet energies. Their importance depends on the observable and scale; calling a jet infrared safe does not remove all power corrections or all modeling choices.

Detector-level jets add response, pileup mitigation, reconstruction thresholds, and calibration uncertainties. A particle-level definition specifies which stable final-state particles are included, and unfolding must account for migrations to a reconstructed selection. Quark- and gluon-jet labels can also depend on the theoretical definition rather than being perfect event-by-event truth. Preserve these distinctions when interpreting a tagger score or a simulation label as evidence for a particular short-distance process.