Multi-Propagator Angular Integrals

• Multi-propagator angular integrals are defined as angular integrations in d=4-2ε dimensions with several linear propagator denominators, crucial for isolating angular dependencies in scattering problems.
• They employ Mellin–Barnes representations, IBP reductions, and dimensional-shift identities to achieve analytic solutions expressed in hypergeometric, polylogarithmic, and Clausen functions.
• These integrals are pivotal in multi-leg scattering and phase-space analyses, linking geometric invariants like Euclidean Gram determinants to physical observables and reduction methodologies.

Multi-propagator angular integrals are angular integrations over a light-like or unit direction in $d=4-2\varepsilon$ dimensions with several linear propagator-like denominators of the form $(v_i\!\cdot k)^{-j_i}$ or $(p_i\!\cdot q)^{-j_i}$. They occur in perturbative quantum field theory whenever radial and angular variables are separated in phase-space or loop integrals, especially in soft-collinear and multi-leg kinematics. The subject has developed from general Mellin–Barnes and multivariate $H$-function representations for $n$ denominators, through complete all-order $\varepsilon$-expansions for one- and two-denominator cases, to systematic IBP-, dimensional-shift-, and differential-equation-based solutions for three denominators and, more recently, four-denominator families (Somogyi, 2011, Lyubovitskij et al., 2021, Haug et al., 2024, Haug et al., 1 Aug 2025).

1. Definition and kinematic setting

A general $n$-denominator angular integral may be written as

$$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$

with $d=4-2\varepsilon$, complex or integer exponents $j_k$, and rotationally invariant angular measure $(v_i\!\cdot k)^{-j_i}$0 on the $(v_i\!\cdot k)^{-j_i}$1-sphere. The total solid angle is

$(v_i\!\cdot k)^{-j_i}$2

Equivalent normalized forms divide by $(v_i\!\cdot k)^{-j_i}$3, which removes Euler–Mascheroni constants from the $(v_i\!\cdot k)^{-j_i}$4-expansion (Somogyi, 2011, Haug et al., 2024).

The dependence on external data is entirely through scalar invariants. In the modern phase-space-oriented notation one introduces normalized vectors $(v_i\!\cdot k)^{-j_i}$5, with

$(v_i\!\cdot k)^{-j_i}$6

and refers to $(v_i\!\cdot k)^{-j_i}$7 as “masses”: a propagator is massless if $(v_i\!\cdot k)^{-j_i}$8, and massive otherwise. For the three-denominator family studied in detail in 2024, the master class is

$(v_i\!\cdot k)^{-j_i}$9

where $(p_i\!\cdot q)^{-j_i}$0 counts the number of non-zero masses among $(p_i\!\cdot q)^{-j_i}$1 (Haug et al., 2024).

A central structural fact is scaling invariance: overall rescalings of the external vectors only factor out as simple powers. This makes the integrals intrinsically functions of dimensionless cross-ratio-like combinations of the invariants. In practical applications, these objects appear after isolating the angular dependence of emitted radiation or loop momentum directions, so they encode the geometry of multi-leg scattering in a particularly compressed form.

2. Analytic representations and special-function classes

The first general solution for $(p_i\!\cdot q)^{-j_i}$2-denominator angular integrals in $(p_i\!\cdot q)^{-j_i}$3 dimensions expresses them through a multivariate $(p_i\!\cdot q)^{-j_i}$4-function. In Mellin–Barnes form, the result has one complex variable $(p_i\!\cdot q)^{-j_i}$5 for each non-vanishing invariant $(p_i\!\cdot q)^{-j_i}$6, and the integral takes the schematic form

$(p_i\!\cdot q)^{-j_i}$7

with $(p_i\!\cdot q)^{-j_i}$8 and $(p_i\!\cdot q)^{-j_i}$9 a multidimensional Mellin–Barnes integral whose Gamma-function structure is fixed by the exponents and by the quadratic invariant matrix (Somogyi, 2011). This representation is completely general and provides the natural starting point for analytic continuation, residue calculus, and $H$0-expansion.

For one and two denominators, the Mellin–Barnes representation reduces to familiar hypergeometric functions. The one-mass one-denominator integral admits Gauss $H$1 representations; the massless two-denominator integral is again a Gauss $H$2; the single-massive two-denominator case is an Appell $H$3; and the double-massive two-denominator case can be written as a Lauricella $H$4 function (Lyubovitskij et al., 2021, Somogyi, 2011). These closed forms are not merely formal: they support all-order $H$5-expansions in terms of Nielsen polylogarithms and their two-variable generalizations for all angular integrals with up to two denominators (Lyubovitskij et al., 2021).

A later development generalizes this philosophy by introducing an Euler integral representation similar to Lee–Pomeransky representation for the general $H$6-denominator case. In that formulation the angular integral is governed by a polynomial

$H$7

which plays the role of a one-loop-like Lee–Pomeransky polynomial and makes IBP and differential-equation machinery directly applicable to phase-space angular integrals (Haug et al., 1 Aug 2025). This suggests a strong structural parallel between loop integrals and angular phase-space integrals, even though the latter do not require reversed unitarity.

3. Reduction mechanisms: IBP, partial fractioning, and dimensional shift

The reduction theory of multi-propagator angular integrals combines several techniques. For three denominators, angular IBPs are derived directly on the sphere from

$H$8

with tangential vectors built from $H$9 and $n$0. These relations generate linear recurrences in the indices $n$1, including lowering and raising relations that reduce arbitrary integer powers to a small master basis (Haug et al., 2024).

For the three-denominator family, every $n$2 with integer indices reduces to

$n$3

up to permutations, and the only genuinely new three-denominator master is $n$4; all masters with at most two denominators are known analytically to all orders in $n$5 (Haug et al., 2024). Earlier work had already established exact recursion relations reducing all one- and two-denominator angular integrals to a short basic set, together with all-order $n$6-expansions for those building blocks (Lyubovitskij et al., 2021).

A complementary mechanism is rotationally invariant geometric partial fractioning. The two-point splitting lemma decomposes a product of two linear denominators by inserting an auxiliary vector on the line connecting two given velocities. In the two-mass case one chooses the interpolation so that the new vector is massless, thereby reducing double-massive integrals to sums of single-massive ones (Lyubovitskij et al., 2021). The 2024 three-denominator analysis extended this principle: by iterative two-point splitting, multi-mass three-denominator masters are reduced to massless and single-massive masters, and a scaled basis $n$7 makes these relations purely additive (Haug et al., 2024).

Dimensional-shift identities provide the third pillar. For general angular integrals one has a shift $n$8, or equivalently $n$9, relating an integral in $\varepsilon$0 to integrals with shifted indices in $\varepsilon$1 dimensions (Haug et al., 2024). Later work formulated general-$\varepsilon$2 dimensional recurrences in terms of Minkowski and Euclidean Gram determinants and proposed explicit branch decompositions that reduce the number of scales in master integrals from $\varepsilon$3 to $\varepsilon$4 (Haug et al., 1 Aug 2025).

4. Three denominators and the emergence of Euclidean Gram structures

The three-denominator problem marks a qualitative change in analytic structure. In the 2024 solution, a special dimensional-shift identity relates the master $\varepsilon$5 to two-denominator masters and to $\varepsilon$6 in $\varepsilon$7. Its coefficients involve the Minkowski Gram determinant

$\varepsilon$8

and the Euclidean Gram determinant

$\varepsilon$9

Geometrically, $n$0, so it is proportional to the squared Euclidean volume of the tetrahedron spanned by the spatial vectors and vanishes if and only if the vectors are coplanar (Haug et al., 2024).

A decisive consequence of this identity is that all $n$1 poles and the finite $n$2 term of $n$3 are fixed by known two-denominator integrals; the genuinely new three-denominator contribution starts at $n$4 (Haug et al., 2024). That new term is the first known instance, for angular integrals, of a contribution proportional to a Euclidean Gram determinant in the $n$5-expansion.

In the massless case the $n$6 term contains

$n$7

where $n$8 is a symmetric combination of Clausen functions. The Clausen function is

$n$9

The coefficient is expressible through angles built from $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$0 and linear combinations of the invariants, and analogous Clausen sums appear in the single-, double-, and triple-massive cases (Haug et al., 2024). The paper further notes the relation

$$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$1

which links the Clausen function to the Bloch–Wigner dilogarithm and thereby to hyperbolic geometry. This interweaves Euclidean volume data, spherical solid-angle data, and hyperbolic tetrahedral structures in a single analytic object. A plausible implication is that three-denominator angular integrals form a geometric boundary case between conventional phase-space polylogarithms and higher-dimensional Gram-determinant structures.

5. Mellin–Barnes methods and small-mass asymptotics

A separate but complementary line of development studies angular integrals by expansion by regions in the small-mass limit $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$2. The key step is to convert the angular integrals into parametric integrals over $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$3 whose integrands are polynomials raised to powers linear in $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$4, making them directly compatible with $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$5 (Smirnov et al., 2024). The method is then completed with Mellin–Barnes representations,

$$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$6

and Barnes-lemma reductions for the region contributions (Smirnov et al., 2024).

For two, three, and four denominators, the contributing regions have a strikingly simple pattern: one massless region and one massive region for each non-zero small mass. In the three-denominator case, the three massive regions produce universal factors proportional to $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$7, while the massless region reconstructs the purely massless three-denominator integral (Smirnov et al., 2024). The sum yields the full leading-power small-mass asymptotics and reproduces the known pole and finite structures.

The resulting conjecture for the general $$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$8-denominator master integral is

$$\Omega_{j_1,\dots,j_n}(p_1,\dots,p_n;\epsilon)\equiv \int d\Omega_{d-1}(q)\,\prod_{k=1}^{n}(p_k\!\cdot\! q)^{-j_k},$$9

and, from pole cancellation in the fully massive case,

$d=4-2\varepsilon$0

These formulas were checked explicitly for $d=4-2\varepsilon$1 (Smirnov et al., 2024). They also clarify the relation between region decomposition and algebraic mass-splitting identities: each massive region corresponds to an angular integral with a single small mass, while the massless region is the fully massless integral.

6. Four denominators and the extension to general $d=4-2\varepsilon$2

Four-denominator angular integrals remained the next major unresolved family after the three-denominator solution. One approach, based on general-$d=4-2\varepsilon$3 Euler representations, IBP reduction, dimensional recurrence, and differential equations, solved the previously unknown four-denominator angular integrals for any number of masses to finite order in $d=4-2\varepsilon$4 and introduced a branch-integral decomposition that reduces the number of scales in the master integrals from $d=4-2\varepsilon$5 to $d=4-2\varepsilon$6 (Haug et al., 1 Aug 2025). In that framework, the $d=4-2\varepsilon$7 four-denominator master is most naturally written in terms of Clausen functions of Gram-determinant angles, and the $d=4-2\varepsilon$8 result follows by dimensional shift.

A second approach evaluates four-denominator angular phase-space integrals directly by Mellin–Barnes methods. In the fully massless case the canonical four-denominator integral is represented by a six-fold Mellin–Barnes integral; in the single-massive case by a seven-fold one. After analytic continuation and conversion to real parametric integrals, the answers are expressed through weight-2 Goncharov polylogarithms. The fully massless finite part has a 17-letter alphabet, while the single-massive finite part has an 11-letter alphabet; multiple-massive cases are reduced to the single-massive case by partial fraction decomposition (Ahmed et al., 21 Aug 2025). The pole terms are completely symmetric: $d=4-2\varepsilon$9 symmetry in the fully massless case and $j_k$0 symmetry among the massless legs in the single-massive case (Ahmed et al., 21 Aug 2025).

The Mellin–Barnes-to-GPL program has since been pushed further. A 2026 treatment computes three-denominator massless integrals to $j_k$1, three-denominator single-massive integrals to $j_k$2, and four-denominator massless and single-massive integrals to $j_k$3, all analytically in terms of GPLs (Ahmed et al., 2 Apr 2026). The same work derives recursion relations that reduce higher propagator powers to the master integrals and emphasizes that partial fraction decomposition reduces multi-massive configurations to sums of single-massive ones (Ahmed et al., 2 Apr 2026).

These results sharpen the distinction between two viewpoints. In one viewpoint, specific low-$j_k$4 families are solved completely by differential equations or Mellin–Barnes technology. In the other, general-$j_k$5 structure is encoded in recurrence, branch, and dimensional-shift formulas. Together they indicate that multi-propagator angular integrals are not an isolated list of special cases but an organized integral class with a workable reduction theory.

7. Applications, geometric content, and methodological position

The primary arena for multi-propagator angular integrals is phase-space integration with multiple observed particles. Three-denominator angular integrals arise when partial fractioning can no longer reduce all eikonal factors to at most two denominators, and the 2024 three-denominator solution was explicitly motivated by phase-space calculations with multiple observed final-state particles (Haug et al., 2024). The practical pipeline described in that work is: reduce powers by angular IBP, reduce mass patterns by two-point splitting, use dimensional shift to obtain poles and finite parts from known lower-point integrals, solve the $j_k$6 differential equation for the genuinely new term, and insert the $j_k$7-expanded result back into the full phase-space integral (Haug et al., 2024).

The same objects also appear in loop calculations once radial and angular integrations are separated. Earlier treatments emphasize that loop and phase-space angular sectors are governed by the same rotationally invariant denominators and can be handled by closed hypergeometric representations, geometric partial fractioning, or, in older massless settings, by Gegenbauer polynomial techniques and uniqueness methods (Lyubovitskij et al., 2021, Kotikov et al., 2018). This suggests a transfer of loop-integral technology to phase space that bypasses reversed unitarity for a substantial class of observables.

A recurring misconception is that angular integrals with many denominators are always reducible to the two-denominator case. That is true in generic 4D kinematics only when the available linear dependences and partial fraction identities suffice. The 2024 three-denominator analysis shows precisely where this simplification fails and why a new master integral $j_k$8 is required (Haug et al., 2024). A second misconception is that the nontrivial analytic content must already appear in the pole terms. For three denominators, the opposite is true: the pole and finite parts are controlled by known two-denominator data, while the genuinely new structure begins only at $j_k$9 through the $(v_i\!\cdot k)^{-j_i}$00-term and its Clausen-function coefficient (Haug et al., 2024).

The broader significance of the subject lies in its convergence of methods and structures. Multi-propagator angular integrals now support Mellin–Barnes representations, Euler/Lee–Pomeransky-like representations, exact IBP systems, dimensional recurrences, and differential equations, while their analytic answers range over $(v_i\!\cdot k)^{-j_i}$01, Appell $(v_i\!\cdot k)^{-j_i}$02, Lauricella functions, Nielsen polylogarithms, Clausen functions, and Goncharov polylogarithms (Somogyi, 2011, Lyubovitskij et al., 2021, Haug et al., 1 Aug 2025). This suggests that they occupy an intermediate position between classical special-function theory and modern multi-scale Feynman-integral geometry.