Taggus: Jet-Flavour Tagger for Z+Jet
- Taggus is a jet-flavour tagger defined for Z+jet production that uses a jet angularity cut to differentiate between quark-like and gluon-like jets.
- It employs an IRC-safe observable with resummed and matched perturbative calculations to ensure theoretical robustness in jet substructure analyses.
- By enhancing initial-state gluon purity, Taggus provides a novel method to probe gluon PDFs and refine our understanding of partonic processes at the LHC.
Searching arXiv for the requested topic and companion work.
Taggus is the informal name for a jet-flavour tagger developed for jet production at the LHC, in which the leading jet is classified through a simple cut on a jet angularity. Its central purpose is not merely quark/gluon discrimination in the final state, but the enhancement of initial-state gluon purity: at leading order in QCD, tagging the highest- jet as quark-initiated preferentially selects the channel and therefore events containing a gluon in the initial state. The construction is based on an infrared and collinear safe observable, admits a perturbative treatment with resummation, and is proposed as a potentially useful handle on gluonic degrees of freedom of the proton and, more speculatively, on the gluon parton distribution function (PDF) (Caletti, 2021).
1. Partonic rationale and conceptual scope
Taggus is formulated in the context of electroweak boson production in association with jets, with emphasis on the dominant leading-order subprocesses
This kinematic setup creates a direct partonic correlation between the flavour of the observed leading jet and the composition of the initial state. If the leading jet is identified as quark-initiated, then at leading order it originates from , so the event necessarily contains an initial-state gluon. In this sense, Taggus is a quark tag on the leading jet that functions operationally as an initial-state gluon tag (Caletti, 2021).
The broader context is quark/gluon discrimination through jet substructure. The study situates itself within a literature in which parton-level quark/gluon labels are generically ambiguous, but hadronic fiducial definitions and grooming procedures can provide theoretically controlled observables. Jet angularities are especially attractive because they are calculable, infrared and collinear safe for the choices used here, and obey Casimir scaling at leading logarithmic accuracy. The resulting tagged observables are proposed as complementary probes of the gluon PDF, since standard inclusive jet measurements involve mixtures of partonic channels even when gluon-initiated contributions are already numerically dominant.
A recurrent misconception is that Taggus is simply a generic quark/gluon classifier. Its intended use is narrower and more specific: in jet events, the flavour tag on the leading jet is exploited as a proxy for the initial-state flavour composition. The tagging problem is therefore embedded in a hadronic process with a well-defined channel interpretation rather than treated as an abstract jet-classification task.
2. Observable definition and cut-based construction
The tagger is built from a standard jet angularity,
where is the transverse momentum of constituent , 0 is the jet radius parameter, and
1
In the Taggus study, 2 is fixed to 3, 4 is required for IRC safety, and the main numerical focus is 5, commonly called the Width. Small 6 values correspond to energy concentrated near the jet axis and are therefore quark-like, whereas larger values are associated with broader, gluon-like jets (Caletti, 2021).
The observable is computed on the leading jet, defined as the jet with the highest transverse momentum in the event. The tagger itself is deliberately simple: choose a cut value 7, and declare the leading jet to be quark-tagged if
8
Jets with 9 are effectively treated as gluon-like. The construction is therefore a one-variable, cut-based classifier rather than a multivariate discriminator.
Its operating characteristics are defined through efficiencies for the two leading-order channels. For a given final-state flavour 0 and initial state 1, the tagging efficiency is
2
This yields 3 for quark jets from 4 and 5 for gluon jets from 6 that are mis-tagged as quark-like. Plotting 7 against 8 gives the ROC curve. The working point is fixed by imposing a target enhanced gluon purity, so 9 is not arbitrary but determined by the desired initial-state flavour enrichment.
3. IRC safety, Sudakov structure, and Casimir scaling
A defining feature of Taggus is that it is built on an IRC-safe observable. With 0 and 1, the jet angularity is stable under both soft emissions and collinear splittings: a soft constituent contributes negligibly as 2, while collinear splittings preserve the relevant momentum sum and leave the angular structure essentially unchanged. Consequently, not only 3 itself but also the cut observable “cross section for events with 4” is infrared and collinear finite and admits perturbative calculation (Caletti, 2021).
At leading logarithmic accuracy, the cumulative distributions of many IRC-safe observables satisfy Casimir scaling. For angularities, the resummed Sudakov form factor takes the schematic form
5
with
6
At this level, quark and gluon distributions differ only through the colour factor 7, so gluon jets are broader because their larger colour charge generates more radiation. This provides the basic theoretical explanation for why low-8 regions are quark-dominated and therefore useful for tagging.
The analysis does not stop at leading logarithms. It explicitly incorporates the first deviation from the typical Casimir-scaling behaviour by using more recent angularity calculations beyond LL. Schematically,
9
where the 0 and higher terms are no longer simple overall rescalings by 1. These corrections matter because precision quark/gluon discrimination depends on the detailed shape difference between quark- and gluon-initiated distributions, not merely on the leading 2 hierarchy.
4. Resummed calculation and the role of grooming
The perturbative framework underlying Taggus combines resummation with fixed-order matching. Angularity distributions in 3jet production are resummed to NLL4 accuracy in the small-5 region and matched to NLO so that the prediction reduces to fixed-order behaviour at large 6. The tagged 7 transverse-momentum distribution, obtained after imposing 8, includes resummation of 9 at NLL and matching to NLO. The proceedings article is a compact presentation of a broader program developed in companion work, including a dedicated tagging study (Caletti et al., 2021).
The treatment of non-perturbative effects is explicitly acknowledged. The tagged transverse-momentum spectrum is supplemented with a non-perturbative correction factor, described as a multiplicative hadronization or underlying-event correction extracted from Monte Carlo, with PYTHIA explicitly mentioned in the proceedings summary. The uncertainty budget is framed in the usual way, through renormalization-scale, factorization-scale, and resummation-scale variations, together with modelling of non-perturbative effects.
Grooming is incorporated through Soft Drop. The procedure reclusters the jet with Cambridge–Aachen and recursively tests whether a branching satisfies
0
If the condition fails, the softer branch is removed and the declustering continues. In this study 1, and several 2 values are considered. The purpose is to remove soft wide-angle radiation, thereby reducing hadronization and underlying-event contamination and sharpening the correlation between angularity and the initiating parton’s colour structure (Caletti, 2021).
The phenomenological message is modest but definite: grooming does not qualitatively alter the low-3 quark dominance, yet it modestly improves the quark/gluon discriminating power and stabilizes the distribution. Taggus can therefore be formulated in both groomed and ungroomed variants.
5. Purity enhancement and tagged observables
The untagged initial-state gluon purity is defined as
4
where 5 corresponds to 6 and 7 to 8. After quark-tagging the leading jet, the enhanced gluon purity becomes
9
This formula makes explicit that the gain arises from retaining quark jets from the 0 channel while suppressing, though not eliminating, gluon jets from the 1 channel that are misidentified as quark-like (Caletti, 2021).
PYTHIA studies reported in the proceedings show that the untagged 2jet sample is already gluon-rich, with 3 over a broad 4 range. This reflects the dominance of 5 at LHC energies, driven by the large gluon PDF at moderate 6. Applying the angularity cut enhances the initial-state gluon purity by about 7 percentage points across a wide 8 range, bringing the tagged sample to roughly 9 gluon purity, both with and without grooming.
Several distributions characterize this behaviour. Width distributions at 0, shown separately for quark and gluon jets using the BSZ flavour–1 algorithm, exhibit a low-2 region dominated by quarks and a larger-3 region with significant gluon support. ROC curves quantify the associated quark efficiency and gluon rejection for both groomed and ungroomed cases. Finally, the tagged 4 5 spectrum is computed at NLO+NLL6, including the non-perturbative correction factor, so the purity enhancement is not presented merely as a Monte Carlo curiosity but as a calculable collider observable.
Experimentally, the strategy is straightforward: reconstruct 7jet events, identify the leading jet, measure its angularity with or without Soft Drop grooming, apply the cut 8, and form the tagged 9 transverse-momentum spectrum. The resulting observable can also be considered in double-differential form with respect to 0 and angularity.
6. Significance, caveats, and relation to gluon-PDF studies
The principal significance of Taggus lies in its combination of simplicity and perturbative control. It uses one continuous observable and one cut, yet remains theoretically well-defined because the tagging criterion is IRC safe and embedded in an NLO+NLL1 framework. This distinguishes it from heuristic quark/gluon taggers whose interpretation can depend strongly on parton-shower definitions or non-perturbative modelling (Caletti, 2021).
Its proposed PDF relevance follows directly from flavour enrichment. Standard inclusive 2jet observables probe mixtures of partonic channels, whereas the tagged spectra retain similar kinematic reach in 3 and 4 but with improved gluon purity. The proceedings stop short of a full PDF-sensitivity analysis, but they explicitly motivate the possibility that precision measurements of tagged 5jet spectra could enter global fits as complementary constraints on the gluon distribution.
Several limitations are equally explicit. Non-perturbative effects remain important and are only mitigated, not removed, by Soft Drop grooming. Pile-up is not discussed in detail, though the summary notes that experimental implementations would require mitigation techniques, particularly for ungroomed angularities. The detailed discrimination power and non-perturbative sensitivity depend on the jet radius 6, even though IRC safety itself does not. Perturbative systematics arise from the usual scale variations and from the truncation of the resummation and matching procedure. Experimental systematics include jet calibration and the measurement of angularities.
A deeper conceptual caveat concerns flavour labelling itself. The study stresses the theoretical cleanliness of the observable but also sits within the broader understanding that quark/gluon tagging is generically ambiguous at the parton level. The hadron-level tagger is calculable and meaningful; the mapping back to an unambiguous parton-flavour ontology remains, as in all such studies, a modelling-dependent interpretation. That caveat does not invalidate Taggus, but it sharply defines what the method can claim: a controlled way of enriching samples in initial-state gluons through a leading-jet angularity cut, rather than an exact event-by-event identification of microscopic partonic history.
In that restricted but important sense, Taggus occupies a distinctive niche. It is a jet-flavour tagger formulated specifically for 7jet production, grounded in Sudakov physics and Casimir scaling, augmented by NLO+NLL8 predictions and grooming studies, and aimed at constructing observables with enhanced sensitivity to gluonic proton structure.