Angular-Ordered Parton Shower (AOPS) Overview

• Angular-Ordered Parton Shower (AOPS) is a QCD algorithm that generates successive emissions in decreasing angle to enforce soft-gluon coherence.
• It employs quasi-collinear splitting kernels, a Sudakov evolution variable, and global momentum-reconstruction to faithfully capture the infrared structure of QCD.
• Different recoil schemes, such as transverse-momentum, virtuality, and dot-product preserving, balance formal accuracy with experimental phenomenology.

Searching arXiv for recent and foundational papers on angular-ordered parton showers in Herwig and related coherence/recoil formulations. Angular Ordered Parton Shower (AOPS) denotes a class of parton-shower algorithms in which successive emissions are generated in decreasing angle relative to the emitting parton, so that soft-gluon coherence is enforced through angular ordering rather than through virtuality ordering alone. In the Herwig lineage, AOPS is the canonical coherent-branching formulation: it is built from repeated $1\to2$ splittings, quasi-collinear splitting kernels, a Sudakov evolution variable $\tilde q$, and a global momentum-reconstruction procedure. In modern discussions, AOPS is not treated merely as a historical Monte Carlo ansatz, but as one concrete implementation of the infrared structure of QCD, whose formal accuracy depends crucially on the definition of the ordering variable, the recoil prescription, and the treatment of initial-state versus final-state radiation (Nagy et al., 2017, Cormier et al., 2018, Bewick et al., 2019).

1. Coherence and the angular-ordering principle

The defining physical motivation of AOPS is color coherence. Wide-angle soft gluons do not resolve the internal structure of a narrow jet; instead, they couple to the total color charge seen at the larger angular scale. In shower language, this is implemented by generating emissions at large opening angle first and constraining later splittings to occur at smaller angles. In the terminology of the coherent-branching picture, this is the classic reason angle ordering is favored for final-state showers (Nagy et al., 2014, Forshaw et al., 2020).

A convenient formulation of the soft-coherence statement appears after azimuthal averaging of the dipole soft kernel. For emission from parton $i_n$ with color partner $j_n$,

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$

The step function is the angular-ordering constraint: radiation from $i_n$ is allowed only when the emission angle is smaller than the opening angle of the parent dipole. In this limit, soft wide-angle physics is represented by the restriction itself, and the shower becomes strongly ordered in an angular variable such as

$\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$

For two-jet $e^+e^-\to q\bar q$ kinematics, the coherent-branching approximation is especially powerful because the primary legs are back-to-back, so the angular constraint is saturated exactly for emissions from the primary partons (Forshaw et al., 2020).

The same physical point can be stated graphically. In the analysis of ordering variables, the relevant configuration is a large-angle emission followed by a much smaller-angle internal splitting, summarized by

$\bm\theta_{12}\ll \bm\theta\, .$

This is identified explicitly as the basic physics argument behind angle-ordered showers. The associated color-coherence relation implies that the sum over competing shower histories reproduces the same soft-gluon emission pattern as the earliest, widest-angle branching (Nagy et al., 2014).

Within a more abstract operator formulation, angular ordering is not a separate principle but one possible realization of the infrared evolution of QCD. A shower evolution operator $U$ obeys

$\tilde q$0

so probability preservation follows from $\tilde q$1. In that framework, angular ordering, virtuality ordering, and transverse-momentum ordering are different implementations of the same infrared-sensitive structure, provided the singular limits are reproduced consistently (Nagy et al., 2017).

2. Kinematics, evolution variables, and coherent branching

In Herwig-style AOPS, the momentum of a shower parton is written in Sudakov form,

$\tilde q$2

with $\tilde q$3 the emitter momentum, $\tilde q$4 a lightlike reference vector aligned with the color partner, and $\tilde q$5 transverse to both. For a final-state branching $\tilde q$6, the canonical evolution variable is

$\tilde q$7

where $\tilde q$8 is the light-cone momentum fraction defined by $\tilde q$9 and $i_n$0. The differential branching probability is

$i_n$1

with $i_n$2 the quasi-collinear splitting function (Cormier et al., 2018, Bewick et al., 2019).

In the coherent-branching formalism for massless quarks, the same variable is often written directly in terms of transverse momentum,

$i_n$3

Angular ordering then takes the explicit form

$i_n$4

which encodes the successive narrowing of allowed angular phase space. The NLL coherent-branching evolution equation for the jet mass distribution is written with the splitting function

$i_n$5

For quasi-collinear massive quarks, the evolution variable becomes

$i_n$6

and the splitting function generalizes to

$i_n$7

This is the quasi-collinear heavy-quark evolution underlying Herwig’s angular-ordered shower (Hoang et al., 2018).

A notable phenomenological feature of the Herwig implementation is that heavy-quark radiation is not described by a strict dead-cone veto. Instead, the shower gives a smooth suppression of soft radiation from heavy quarks for angles $i_n$8, which is one reason the algorithm is used as a mass-aware default in heavy-flavor applications (Cormier et al., 2018).

The same framework also imposes intrinsic phase-space restrictions. In benchmark studies, the angular-ordered shower is described as conservative in hard-emission regions, with an intrinsic dead zone and a veto scale imposed in practice because the shower can otherwise emit radiation with transverse momentum larger than the hard-process scale (Bellm et al., 2016).

3. Recoil prescriptions and logarithmic accuracy

A central modern result is that the formal accuracy of AOPS is not determined by angular ordering alone. Once more than one emission is present, the same ordering variable $i_n$9 admits multiple kinematic interpretations, and each interpretation implies a different recoil scheme. The three schemes analyzed explicitly are the transverse-momentum-preserving scheme, the virtuality-preserving scheme, and the dot-product-preserving scheme (Ravasio et al., 2019, Bewick et al., 2019).

In the transverse-momentum-preserving scheme, one enforces

$j_n$0

This is the original Herwig choice. It preserves the independence of well-separated soft emissions in rapidity and therefore maintains the soft-collinear logarithmic structure. Its drawback is phenomenological: recoil accumulates in the emitter virtuality and tends to overpopulate hard, non-logarithmic regions of phase space (Ravasio et al., 2019).

In the virtuality-preserving scheme, the parent virtuality is kept fixed instead. For a second emission, the first emission’s transverse momentum becomes

$j_n$1

This quantity can become negative, even when both emissions are soft. The scheme therefore distorts the first emission when the second is added, spoiling the factorized structure of widely separated soft emissions. In the final-state analysis, this leads to an explicit NLL mismatch in thrust,

$j_n$2

and the scheme is identified as not formally NLL accurate (Bewick et al., 2019).

The dot-product-preserving scheme was proposed as a compromise between logarithmic correctness and phenomenology. Its defining relation is

$j_n$3

For multiple emissions, the first gluon’s transverse momentum is modified as

$j_n$4

in the final-state discussion, while angular ordering,

$j_n$5

guarantees that $j_n$6 does not become negative. In the soft limit, later emissions do not significantly alter earlier soft transverse momenta, so the formal soft-collinear structure is preserved more faithfully than in the virtuality-preserving case (Ravasio et al., 2019).

Phenomenologically, the dot-product-preserving scheme plus a phase-space veto is reported to give the best overall agreement with LEP event-shape data among the recoil prescriptions studied. The accompanying conclusion is nuanced: the transverse-momentum-preserving scheme is conceptually sound in the soft limit, the virtuality-preserving scheme improves the hard region but compromises formal accuracy, and the dot-product-preserving scheme is the best compromise in practice (Ravasio et al., 2019).

This sequence of results underpins an important misconception correction. The problem is not angular ordering itself. The critical issue is how one reconstructs kinematics and assigns recoil once multiple emissions are present. In that sense, modern analyses treat AOPS as a family of coherent-branching algorithms rather than a unique prescription fixed solely by the choice of an angular variable (Bewick et al., 2019).

4. Initial-state radiation and consistent ISR/FSR evolution

Initial-state radiation in AOPS is formulated through backward evolution. For a single ISR branching $j_n$7, with $j_n$8 the space-like child and $j_n$9 the time-like child,

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$0

with

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$1

The one-emission ISR ordering variable is

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$2

This is the angular-ordered ISR evolution variable used in Herwig 7 (Bewick et al., 2021).

With multiple ISR emissions, the same ambiguity encountered in FSR reappears. The general relation

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$3

shows that one cannot preserve simultaneously the virtuality, the dot product, and the transverse momentum of an earlier branching. The ISR study reaches the same formal conclusion as the final-state analysis: the $$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$4-preserving scheme is discarded, while the $$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$5-preserving and dot-product-preserving schemes remain logarithmically acceptable (Bewick et al., 2021).

For two soft ISR emissions, the distinction is explicit. In the $$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$6-preserving scheme,

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$7

whereas in the dot-product-preserving scheme,

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$8

In the fully soft, strongly ordered limit, the two become equivalent at the level relevant for logarithmic accuracy. The ISR paper therefore recommends the dot-product-preserving scheme because it is consistent with the preferred final-state prescription and gives a uniform recoil philosophy for both sectors (Bewick et al., 2021).

Momentum conservation is restored through a global recoil strategy. In Drell–Yan production, both incoming partons are shower progenitors, and the final-state color singlet absorbs the net recoil. The default longitudinal reshuffling preserves the off-shell momentum of the progenitor that does not contain the hardest emission, and for a single emission it exactly reproduces the Catani–Seymour initial-initial dipole kinematics. The associated rescaling satisfies

$$\langle P_{i_n j_n}\rangle_{\phi_{n,i_n}}= \frac{\Theta(\theta_{j_n,i_n}-\theta_{n,i_n})}{1-\cos\theta_{n,i_n}}\, .$$9

so the reshuffling induces only power-suppressed changes and does not affect logarithmic accuracy (Bewick et al., 2021).

The phenomenological tuning of the ISR sector was carried out using $i_n$0-boson production at the LHC at 7 TeV. The tuned shower prefers

$i_n$1

in the CMW scheme and intrinsic $i_n$2 around $i_n$3 GeV, with the best overall fit reported for the dot-product-preserving recoil scheme combined with POWHEG(aS). This suggests that the formal recoil improvement has direct phenomenological consequences in the low-$i_n$4 Drell–Yan region (Bewick et al., 2021).

5. Matching, uncertainty estimation, and benchmark phenomenology

In contemporary use, AOPS is usually embedded in matched event generation rather than employed as an isolated LL approximation. Within Herwig 7, the Matchbox framework interfaces the angular-ordered shower to NLO hard processes and supports both subtractive MC@NLO-type matching and multiplicative Powheg-type matching. In the multiplicative case, the angular-ordered shower can be supplemented with truncated showers to restore soft, large-angle radiation not described by the hardest-emission correction alone (Rauch et al., 2016, Rauch et al., 2016).

A standard formal feature is the use of a resummation profile scale $i_n$5, which regulates the transition between the hard matching region and the shower region while preserving the shower’s resummation properties. In uncertainty studies at LO+PS, shower-level variations are organized around hard, veto, and shower scales, $i_n$6, $i_n$7, and $i_n$8, varied through a full 27-point envelope. A key outcome is that one-dimensional variations underestimate the actual perturbative uncertainty, whereas combined scale variations provide a controlled baseline. In the same benchmark program, the resummation and theta profiles are favored, while hfact and the power shower distort the Sudakov structure or jet spectra too strongly (Bellm et al., 2016).

The most detailed comparisons between AOPS and dipole evolution concern VBF $i_n$9 production and $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$0 production and decay. In the VBF process

$\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$1

the angular-ordered and dipole showers are compared under the same hard-process setup, the same MMHT2014 PDFs, and the same VBF cuts. Both showers reduce the accepted rate relative to fixed order because additional radiation induces migration across the VBF selection, especially through changes in the tagging-jet system. For electroweak observables such as the four-lepton invariant mass, AOPS and the dipole shower give broadly similar results, whereas radiation-sensitive observables such as the third-jet rapidity

$\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$2

show stronger shower dependence. Even there, both showers suppress central radiation relative to fixed-order NLO, consistent with the color-singlet nature of VBF exchange (Rauch et al., 2016, Rauch et al., 2016).

In top-quark pair production, the angular-ordered shower is used as Herwig’s default shower and contrasted with the dipole shower in matched $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$3 predictions. The two algorithms are systematically different: AOPS evolves in the angular variable $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$4, reconstructs kinematics through a more elaborate global procedure, and tends to be more restrictive in jet activity, whereas the dipole shower is closer to Catani–Seymour kinematics and usually allows more radiation. NLO matching reduces uncertainty bands for many observables, but boosted top observables such as $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$5 and $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$6 remain shower sensitive. Comparisons to LHC data show that regions where non-top-specific tunes deviate markedly from data are also regions with large shower and matching uncertainties (Cormier et al., 2018).

Across these benchmarks, the overarching conclusion is not that AOPS is universally superior to dipole evolution, but that it remains a controlled coherent-radiation algorithm when paired with sensible profile choices, matched matrix elements, and full uncertainty variation. The main residual sensitivity is concentrated in observables driven by additional hard radiation, high jet multiplicity, or cut migration (Bellm et al., 2016, Rauch et al., 2016, Cormier et al., 2018).

6. Extensions, boundaries, and alternative formulations

AOPS has been extended substantially beyond its traditional QCD final-state role. Spin correlations can be incorporated through the Collins–Knowles density-matrix algorithm, for which the angular-ordered shower is described as an ideal setting because each parton line evolves recursively and approximately independently. The characteristic correlated distribution between successive branching planes has the form

$\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$7

and Herwig 7 implements the necessary basis rotations between production and shower frames to propagate these correlations through the angular-ordered cascade (Richardson et al., 2018).

The same angularly ordered architecture has also been generalized to electroweak and beyond-the-Standard-Model radiation. The electroweak extension in Herwig 7 embeds helicity-dependent quasi-collinear $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$8, $\zeta_{n,i_n}=1-\cos\theta_{n,i_n}\, .$9, $e^+e^-\to q\bar q$0, and Higgs branchings into the existing QCD+QED shower, using Dawson’s prescription for longitudinal vector bosons to avoid spurious high-energy singular behavior. At still greater generality, a model-independent BSM shower based on UFO input and ufo2herwig derives helicity-resolved splitting kernels for scalar, fermion, and vector emissions while retaining the same angular-ordering structure and the dot-product-preserving definition

$e^+e^-\to q\bar q$1

These developments show that the ordering principle itself is not specific to QCD, provided suitable quasi-collinear branching kernels are available (Masouminia et al., 2021, Lee et al., 2023).

The infrared cutoff $e^+e^-\to q\bar q$2 introduces another nontrivial consequence. For quasi-collinear massive quarks, a finite shower cutoff implies that the generator mass is not the pole mass but a $e^+e^-\to q\bar q$3-dependent short-distance mass scheme: $e^+e^-\to q\bar q$4 The same analysis finds that the soft cutoff shift is universal for massless and boosted massive quarks, and numerical comparisons with Herwig 7 support the interpretation

$e^+e^-\to q\bar q$5

This is directly relevant to the field-theoretic interpretation of direct-reconstruction top-mass measurements (Hoang et al., 2018).

Finally, recent work has clarified the conceptual boundary of the topic. The shower introduced in “A new approach to color-coherent parton evolution” is explicitly not an angular-ordered parton shower in the traditional sense. It replaces explicit angular ordering with a positive partial-fraction decomposition of the soft eikonal, a non-trivial azimuthal-angle dependence in the splitting kernels, and a global recoil mapping proven to be NLL safe. The coherence pattern is therefore recovered without an explicit angular veto. This marks it as a generalization or alternative formulation of color-coherent evolution rather than a minor refinement of AOPS (Herren et al., 2022).

Taken together, these developments place AOPS in a precise contemporary position. It is the canonical coherent-branching shower formulation of the Herwig family; it remains viable at the level of LL and many NLL-sensitive applications when recoil is handled carefully; it can be extended to spin, electroweak, and BSM radiation; and its limits are now sufficiently well understood that alternative color-coherent formulations can be defined relative to it rather than confused with it.