How gluon leading singularities discover curves on surfaces

We study the leading singularities for pure gluon amplitudes obtained by on-shell gluing of three-particle amplitudes for an arbitrary graph in any number of dimensions. By encoding the polarization vector contractions in a graphical way, on-shell gluing “discovers” curves on surfaces, and we find that the leading singularity is determined by a simple combinatorial question: what are all ways of covering the graph with non-overlapping curves such that each edge is covered exactly once? This precisely matches the formula from the surfaceology formulation of gluons, where the leading singularities are given by maximal residues, with the combinatorial problem arising from the linearized form of the $u$ variables. At loop-level we describe how the novelties associated with spin sums (related with the need for ghosts when working off-shell using Lagrangians) can be easily encoded in this combinatorial picture. Matching the leading singularities also lets us settle an open question in the surface formulation of gluons, determining the exponents of the closed curves at any loop order.

💡 Research Summary

The paper presents a novel geometric‑combinatorial framework for computing leading singularities (LS) of pure gluon amplitudes in non‑supersymmetric Yang‑Mills theory. Starting from on‑shell gluing of three‑point gluon vertices, the authors encode every Lorentz contraction of polarization vectors as a “contraction curve” drawn on the fat‑graph representing the amplitude. They discover that each LS term corresponds to a distinct way of covering the graph with non‑overlapping curves such that every edge is traversed exactly once. Each curve C is associated with a kinematic invariant X_C, the squared momentum flowing through the curve, and a monomial ∏ X_C appears in the LS.

This combinatorial picture matches precisely the “surfaceology” formulation, where a surface S with a triangulation T is dual to a fat‑graph. The surface integral contains variables u_C for each homotopy class of curves, raised to powers X_C. By expanding the u‑variables to linear order in the positive parameters y_P, residues in y‑space give the LS. The authors show that the on‑shell contraction curves generate exactly the same linearized u‑terms, establishing an equivalence between the two approaches.

At tree level the mapping is straightforward: a curve enters the graph at edge (i,i+1), follows a sequence of left/right turns, and its momentum is obtained by adding the entering edge momentum and the momentum of each edge taken on a right turn. The square of this momentum yields X_C. Different contraction patterns may produce the same monomial, sometimes with opposite signs; the paper introduces a systematic “V‑Rule” to determine which monomials survive after cancellations.

Loop‑level extensions are more subtle. Simple metric‑contraction gluing fails to produce gauge‑invariant LS; one must include ghost‑like corrections. Graphically these corrections appear as self‑intersecting curves and closed curves that wind around punctures on the surface. Closed curves encode the dependence on spacetime dimension D through an exponent Δ_J. The authors verify the known one‑loop result Δ = 1 − D and, by matching the graphical picture to the surface integral, derive a general rule for Δ_J at arbitrary loop order, resolving a previously open problem.

The paper also sketches the inclusion of fermion loops. γ‑matrix traces are represented by additional weights on curves, and the same tiling‑of‑the‑fat‑graph logic applies, suggesting a unified treatment of gluon and quark contributions.

Overall, the work demonstrates that the seemingly intricate spin‑color algebra of gluon LS can be reduced to a purely combinatorial tiling problem on a surface. This provides a transparent geometric interpretation, simplifies higher‑loop calculations, and bridges on‑shell methods with the surface‑integral approach, opening avenues for efficient computation of non‑supersymmetric amplitudes in arbitrary dimensions.