Single-minus gluon tree amplitudes are nonzero

🎯 Pitch

Long assumed to vanish identically, single-minus tree-level gluon amplitudes are shown to be nonzero—and take only values +1, −1, or 0—on special half-collinear kinematic loci. This overturns a foundational dogma of Yang–Mills perturbation theory and introduces a piecewise-constant integer formula whose region boundaries are set by orthogonalities among sums of the constrained momenta.

1. Executive Summary

This paper proves that single-minus gluon tree amplitudes—long presumed to vanish—are in fact nonzero on a restricted "half-collinear" kinematic locus in Klein space or for complexified momenta, and derives a piecewise-constant closed-form expression for the decay of a single minus-helicity gluon into n−1 plus-helicity gluons via the Berends–Giele recursion. The stripped amplitude in the special decay region ℛ₁ takes only values +1, −1, or 0, governed by a product formula (Equation 39) whose factors are signed projection operators on sums of momenta. The result nontrivially satisfies multiple consistency conditions—including Weinberg's soft theorem, cyclicity, Kleiss–Kuijf relations, and U(1) decoupling identities—establishing that single-minus amplitudes exist only when the external momenta obey the half-collinear constraint ⟨ij⟩ = 0 for all particles, which provides the loophole that evades the standard vanishing argument.

2. Context and Motivation

The Core Problem: A Persistent Belief That Single-Minus Amplitudes Vanish

The central problem this paper addresses is a foundational claim in the scattering amplitudes literature that has remained unchallenged for decades: the assertion that tree-level single-minus gluon amplitudes are zero. By "single-minus," the authors refer to color-ordered n-gluon scattering amplitudes where exactly one gluon has negative helicity and the remaining n−1 have positive helicity. For generic (non-collinear) kinematics in four-dimensional Minkowski space, the standard power-counting argument that leads to this conclusion is both simple and compelling — so compelling that it has been treated as a settled fact in the field. The paper revisits this claim and demonstrates that it is not universally true: the vanishing argument breaks down when the external momenta are restricted to a particular kinematic locus the authors call the "half-collinear regime," where all angle brackets ⟨ij⟩ vanish simultaneously.

This is not an esoteric technicality. The claimed vanishing of single-minus amplitudes has shaped the entire conceptual structure of Yang–Mills perturbation theory. The MHV classification — defining maximally helicity violating amplitudes as those with n−2 plus-helicity gluons and exactly 2 minus-helicity gluons — rests on the premise that n−2 is the maximum number of plus gluons allowed. If single-minus amplitudes (n−1 plus gluons, 1 minus gluon) are nonzero under certain kinematic conditions, then this classification is incomplete. The paper's findings imply that what the field has called "MHV" may not, in fact, be maximally helicity-violating in all kinematic regimes.

Why This Problem Matters: Theoretical and Structural Significance

The importance of this problem operates on multiple levels, from the foundational to the practical.

Foundational implications for Yang–Mills theory. Scattering amplitudes encode the full quantum content of a field theory. If single-minus amplitudes are nonzero, even on a restricted kinematic locus, then the space of tree amplitudes is richer than previously believed. The authors point out in the introduction that "the structural role of these single-minus amplitudes in Yang–Mills theory remains to be understood." This is an understatement: the existence of these amplitudes forces a re-examination of the building blocks from which all Yang–Mills amplitudes are constructed. The Parke–Taylor formula for MHV amplitudes has served as the foundation for BCFW recursion relations, the CSW (MHV vertex) construction, the amplituhedron program, and twistor string formulations. Each of these frameworks takes the MHV amplitude as primitive and builds more complex amplitudes from it. If "MHV" is not actually maximal — if there exists a class of amplitudes with even fewer minus helicities — then these frameworks are missing a piece of the tower.

Connection to self-dual Yang–Mills theory (SDYM). The paper highlights a specific puzzle in SDYM — the restricted sector of Yang–Mills theory that retains only the self-dual part of the field strength. As noted in the introduction:

"the classical solution space of SDYM is extremely nontrivial, while the tree diagrams were previously supposed to yield trivial two-point and three-point expressions. The latter seem insufficient to reproduce the former. Potentially, the single-minus tree amplitudes in SDYM found here resolve this tension."

This is a concrete theoretical problem that the paper's result helps address. The Berends–Giele recursion that the authors use (Equation 64, Appendix B) is explicitly identified as the recursion for SDYM form factors. Finding nonzero single-minus amplitudes in SDYM means the quantum theory produces nontrivial objects that may match — rather than contradict — the known richness of SDYM classical solutions (e.g., the instanton constructions of Atiyah, Drinfeld, Hitchin, and Manin, the twistor descriptions of Ward). This resolves a genuine tension between the classical and quantum descriptions of a well-studied integrable field theory.

Implications for celestial holography and asymptotic symmetries. The paper explicitly states that "the results should transform under the S-algebra, the Lw₁₊∞ algebra, and their supersymmetric extensions," and that "in the context of celestial holography, the Mellin transform of the amplitudes in some sectors is given by Lauricella functions." These connections are not decorative — they place the single-minus amplitudes within the active research program that relates soft theorems, asymptotic symmetries, and holographic dual descriptions of flat spacetime. If single-minus amplitudes were genuinely zero, they could play no role in these structures. By establishing they are nonzero, the paper opens the door for them to contribute nontrivially to the celestial operator product expansions and symmetry algebras that govern soft limits.

Methodological significance. The paper demonstrates a concrete workflow where a conjecture was "first conjectured by GPT-5.2 Pro and then proved by a new internal OpenAI model," with the solution subsequently "checked by hand using the Berends–Giele recursion." This is noted in the introduction not as the main contribution but as a methodological data point — an instance where AI-assisted mathematics produced a novel result in a technically demanding subfield. The formula (39) that emerges from this process is remarkably simple (a product of n−2 factors, each a sum of two sign functions) compared to the exponential complexity of the Feynman diagram expansion. This exemplifies the broader theme the paper invokes in its opening paragraphs: "cancellations lead in a variety of contexts to a very simple final answer," and "our present understanding of the quantum laws of physics is seriously incomplete."

Prior Approaches and Where They Fall Short

The standard vanishing argument. The conventional belief that single-minus amplitudes vanish is rooted in a helicity power-counting argument that the paper reviews in Section I.1. Here is the logic: choose reference spinors for the polarization vectors appropriately. For the single minus-helicity gluon (particle 1), use reference spinor |r⟩ for the minus polarization ε₁⁻, and reference spinor |r⟩ for all the plus-helicity polarization vectors εₐ⁺. By setting |r⟩ = |1⟩, one can make all polarization vectors mutually orthogonal. Since the amplitude is constructed by contracting polarization vectors with momenta from interaction vertices, and there are at most n−2 cubic vertices in a tree diagram (each providing one momentum factor), there are not enough momentum factors to contract all n polarization vectors when they are all orthogonal. Therefore, the amplitude must vanish.

This argument is correct — for generic kinematics where ⟨1a⟩ ≠ 0. The "loophole" the paper identifies is precisely that:

"we cannot choose |r⟩ = |1⟩ if ⟨1a⟩ = 0 for any |a⟩, as the polarization vectors εₐ⁺ would become singular"

When ⟨1a⟩ = 0, the choice of reference spinor that makes the vanishing argument work is no longer available, because the polarization vectors for the plus-helicity gluons would diverge (their definition in Equation 6 has ⟨r a⟩ in the denominator). The vanishing argument therefore fails on the kinematic locus where all ⟨1a⟩ = 0. The paper further argues by induction that the full n-point single-minus amplitude can only have support when all ⟨ij⟩ = 0 — the "half-collinear regime" of Equation (13).

The half-collinear regime: defined but unexplored. The fact that kinematics with all ⟨ij⟩ = 0 can exist in (2,2) Klein signature or for complexified momenta is not new — the paper notes this explicitly:

"In (2,2) signature, this is compatible with nonzero [ij], unlike in Minkowski space."

Spinor-helicity variables for massless momenta satisfy pᵢⱼ² = ⟨ij⟩[ij]. In Minkowski signature (3,1), the reality conditions on spinors impose ⟨ij⟩ = ±[ij]*, so setting all ⟨ij⟩ = 0 would force all square brackets to vanish as well, making all momenta proportional. In (2,2) Klein signature, the spinors λ and λ̃ are independent real variables — they are not complex conjugates of each other. This means one can have ⟨ij⟩ = 0 for all particles while maintaining generic, non-proportional [ij] values. This kinematic regime is precisely where the standard vanishing argument breaks down, and it is where the paper's amplitudes live.

Prior work on single-minus amplitudes. The paper does not cite extensive prior literature on single-minus amplitudes — because, for generic kinematics, the field's consensus has been that they vanish. The authors note that the 3-point single-minus amplitude is known to have support on ⟨12⟩ = ⟨13⟩ = 0 (this is the anti-MHV 3-point amplitude, which is the massless 3-point function in spinor-helicity), but the generalization to higher points had not been systematically explored. The "half-collinear regime" as a locus supporting nontrivial amplitudes does not appear to have been the subject of prior dedicated study at tree level for general n.

Berends–Giele recursion: the existing tool, now applied to a new problem. The Berends–Giele recursion (Equation 64) is a well-established method for computing off-shell currents in Yang–Mills theory, equivalent to summing Feynman diagrams with one leg off-shell. The paper's Appendix B derives this recursion and explains how it is adapted for the single-minus problem. The recursion itself is not new — it dates to Berends and Giele's 1988 work — but its application to single-minus amplitudes in the half-collinear regime, including the careful treatment of the iε prescription and the master identity (Appendix A) that handles the distributional nature of the support, is novel. The paper leverages this existing computational machinery as the rigorous foundation for both the general recursion (21) and the proof of the simplified formula (39) in region ℛ₁.

Conflicting signals from SDYM. The paper flags an existing tension that prior work had not resolved. On one hand, self-dual Yang–Mills theory is known to have integrable classical solutions — the ADHM construction of instantons, Ward's twistor description, etc. — indicating a rich classical structure. On the other hand, the tree-level Feynman expansion of SDYM was previously thought to yield only trivial results (nonzero only at 3 points, and even then only on restricted kinematics). The existence of nontrivial single-minus tree amplitudes in SDYM potentially reconciles this mismatch: the quantum perturbation theory now produces objects that can correspond to the known classical solutions. The paper does not fully flesh out this connection but frames it as one of the motivations for the work.

How This Paper Positions Itself Relative to Existing Work

The paper positions itself as correcting a long-standing oversight rather than proposing a new technique or framework. The authors are not introducing a new recursion relation, a new formalism, or a new computational method. They are using the standard Berends–Giele recursion (a textbook tool in the amplitudes community) and pointing out that it yields nonzero results for single-minus amplitudes when applied carefully in the half-collinear regime, where the standard iε prescription and collinear δ-functions become relevant.

The innovation is conceptual: recognizing that the vanishing argument has a loophole, identifying the kinematic regime where it fails, and systematically computing what lives there. In this sense, the paper is closer to "anomaly cancellation" logic than to "new method" logic — it finds something that was always present in the equations but overlooked because the standard assumptions (generic kinematics, ignoring distributional support) were never questioned for this class of amplitudes.

The paper also positions itself as a gateway rather than an endpoint. The authors are explicit that their result "immediately leads to a number of extensions":

Generalization from gluon to graviton amplitudes ("the construction generalizes directly").
Supersymmetrization ("has a simple supersymmetrization").
Behavior under infinite-dimensional symmetry algebras (the S-algebra, Lw₁₊∞).
Celestial holography applications (Mellin transforms yielding Lauricella functions).

These extensions are stated but not developed in the paper, indicating that the authors view this work as opening a new subfield rather than closing one. The result in region ℛ₁ is particularly highlighted as a clean, verifiable anchor point from which these extensions can be explored.

Finally, the paper's abstract and introduction frame the result — somewhat unusually — as a collaboration between human physicists and AI systems ("first conjectured by GPT-5.2 Pro and then proved by a new internal OpenAI model"). This is not the main scientific contribution, but it situates the work within a broader conversation about how theoretical physics may be done in the future. The formula (39) is simple enough to be verified by hand (and the authors state they did so) but emerged from an AI-assisted search process. This framing is notable but secondary to the physics results.

3. Technical Approach

This is an analytic derivation paper whose core idea is that single-minus tree amplitudes, long assumed to vanish, are nonzero on a restricted kinematic locus called the "half-collinear regime," and that the Berends–Giele recursion — properly treated with the iε prescription and collinear δ-functions — yields explicit, piecewise-constant closed-form results.

3.1 Reader Orientation

The "system" being analyzed is not a machine-learning model but a recursive computational procedure that sums over all Feynman diagrams (or equivalently, all ordered partitions of the gluons) to produce the color-ordered single-minus n-gluon tree amplitude. The paper solves the problem of computing these amplitudes by showing that their support is restricted to the half-collinear kinematic locus (where all angle brackets ⟨ij⟩ vanish), deriving a Berends–Giele-type recursion that determines them for all n, and then dramatically simplifying the result in a special kinematic region ℛ₁ to a product of signed projection operators — an expression of extraordinary simplicity compared to the exponential complexity of the underlying Feynman diagram sum.

3.2 Big-Picture Architecture (Diagram in Words)

The derivation proceeds through five major components:

Kinematic parameterization — a choice of spinor-helicity variables in (2,2) Klein signature that makes the half-collinear regime mathematically tractable and the distributional support of the amplitudes explicit.
Half-collinear regime definition — the precise kinematic locus (Equation 13) where all ⟨ij⟩ = 0, identified as the only place the standard vanishing argument fails, and encoded via collinear δ-functions in the amplitude.
Berends–Giele (BG) recursion for form factors — the off-shell equivalent of the amplitude, computed recursively from ordered partitions of the gluon word using vertex functions V and incomplete Parke-Taylor factors PT.
LSZ reduction — the procedure that puts the final off-shell leg on-shell, converting form factors to amplitudes by extracting the pole and applying a master δ-function identity, yielding the general stripped amplitude recursion (Equation 21).
Region ℛ₁ simplification — the proof that in a particular kinematic subregion (single minus-helicity gluon with negative frequency, all plus-helicity gluons with positive frequency), the vertex function V vanishes for all multi-gluon blocks, collapsing the recursion to a single term that simplifies to the product formula (39).

Information flows as follows: the external kinematics enter as spinor-helicity variables → the half-collinear condition is imposed via δ-functions → the Berends–Giele recursion computes off-shell form factors from vertex functions and PT factors → the LSZ reduction extracts the on-shell stripped amplitude → specializing to region ℛ₁ collapses the vertex structure → the final result emerges as a product of signed sums.

3.3 Roadmap for the Deep Dive

First, the kinematic parameterization and half-collinear regime — because the entire computation is only defined on this restricted locus, and understanding why and how the amplitudes are supported there is prerequisite to everything else.
Second, the Berends–Giele recursion for off-shell form factors (Appendix B) — the workhorse computational engine that converts Feynman diagram sums into a systematic recursion over ordered partitions, using vertex functions V and incomplete Parke-Taylor factors PT.
Third, the vertex function V and its barred counterpart V̄ — the central algebraic objects whose properties (specifically, the vanishing of V for multi-gluon blocks in ℛ₁) drive the dramatic simplification.
Fourth, the preamplitude Ā and the stripped amplitude A — how the recursion is built up from singleton and multi-gluon blocks, and how the LSZ reduction bridges the off-shell form factor and the on-shell amplitude.
Fifth, the master identity (Appendix A) — the mathematical engine that handles the distributional δ-functions and iε prescriptions, converting sums over intermediate channels into the V and V̄ decomposition, and enabling the clean separation of the amplitude into kinematic regions.
Sixth, the collapse of the recursion in region ℛ₁ — the proof that V vanishes for consecutive subsets of {2, …, n} under the frequency-sign condition (33), which forces all preamplitudes Ā to vanish for multi-gluon blocks and leaves only the fully unpartitioned term V̄.
Seventh, the final formula (39) — how V̄ factorizes into the product of signed projection operators, and why each factor takes values in {−1, 0, +1}, making the stripped amplitude piecewise-constant.

3.4 Detailed, Sentence-Based Technical Breakdown

Kinematic Parameterization and the Half-Collinear Regime

The entire computation is anchored in a specific choice of spinor-helicity variables designed to make the half-collinear condition transparent and the distributional support of the amplitudes explicit. The paper works in (2,2) Klein signature, where left- and right-handed spinors λ and λ̃ are independent real variables — they are not complex conjugates, unlike in Minkowski signature (3,1) where λ̃ = ±λ* would hold. This independence is essential because it allows the angle brackets ⟨ij⟩ (constructed from λ) to vanish independently of the square brackets [ij] (constructed from λ̃). The momentum p in spinor-helicity is:

$p_{\alpha \dot{\alpha}} = \lambda_\alpha \tilde{\lambda}_{\dot{\alpha}}$

where λₐ is the left-handed spinor with index α ∈ {1, 2}, λ̃ is the right-handed spinor with index α̇ ∈ {1, 2}, and the product gives a 2×2 matrix encoding the four-momentum. In the paper's chosen frame, these are parameterized as:

$|i\rangle = \lambda_i = (1, z_i), \quad |i] = \tilde{\lambda}_i = \omega_i (1, \tilde{z}_i)$

where zᵢ and z̃ᵢ are real and independent, and ωᵢ is a real frequency-like parameter. The angle and square brackets then take the simple forms:

$\langle i j \rangle = z_i - z_j \equiv z_{ij}, \quad [i j] = \omega_i \omega_j (\tilde{z}_i - \tilde{z}_j) \equiv \omega_i \omega_j \tilde{z}_{ij}$

What this parameterization achieves: it reduces the angle brackets to simple coordinate differences in a one-dimensional z-space and the square brackets to coordinate differences in a z̃-space weighted by frequencies. The momentum invariant pᵢⱼ² = ⟨ij⟩[ij] becomes zᵢⱼ · ωᵢωⱼz̃ᵢⱼ, which can vanish either because zᵢⱼ = 0 (half-collinear condition) or because z̃ᵢⱼ = 0 (the complementary "half-collinear" condition in the other chirality), or both.

Why this parameterization: it factorizes the half-collinear condition ⟨ij⟩ = 0 into the simple statement that all zᵢ are equal — a single point in the z-plane. This is what makes the δ-function support manageable: ∏ₐ δ(z₁ₐ) enforces that particle 1 shares the same z-coordinate as all others. The choice of reference frame with λ = (1, z) is not a loss of generality because the stripped amplitude A₁⋯ₙ — the quantity of interest — carries no helicity weight and is invariant under Lorentz transformations that preserve the half-collinear condition.

The half-collinear regime itself is defined by the condition:

$\langle i j \rangle = 0 \quad \forall i, j \in \{1, \dots, n\}$

In the parameterization above, this is equivalent to all zᵢ being equal (zᵢⱼ = 0 for all i, j). The paper notes that "this is compatible with nonzero [ij], unlike in Minkowski space" — the square brackets remain unconstrained and can be generic. The single-minus amplitude is written to explicitly carry the δ-function support:

$\mathcal{A}_n = i^{2-n} A_{1\cdots n} \prod_{a=2}^n \delta(z_{1a}) \; \delta^2\left(\sum_{i=1}^n \tilde{\lambda}_i\right)$

where A₁⋯ₙ is the stripped amplitude — the dynamical content with all helicity weight and universal kinematic support removed — and the δ-functions enforce the half-collinear condition and the remaining momentum conservation components. The factor i²⁻ⁿ is a convention choice for the overall phase and normalization. The δ-functions of z₁ₐ are one-dimensional (normalized such that ∫ δ(x) dx = 2π as per Equation 7), and the δ² of summed λ̃ enforces the two components of right-handed spinor momentum conservation.

Why this form: factoring the amplitude into a universal δ-function support times a stripped amplitude A₁⋯ₙ is a standard technique in the amplitudes literature — it separates the kinematic geometry (where the amplitude is nonzero) from the dynamics (what value it takes there). The stripped amplitude depends only on the z̃ᵢ and ωᵢ (the right-handed spinor data), and carries no little-group weight (it is helicity-neutral), making it the proper object for the recursion relations.

The authors also introduce a shorthand for the half-collinear δ-functions:

$\delta_{1\cdots n} = i^{1-n} \prod_{k=1}^{n-1} \delta(z_{k,k+1})$

where n+1 ≡ 1 by cyclic identification. The factor i¹⁻ⁿ is a convention for consistency with later identities. This notation suppresses the universal δ-functions that enforce the kinematic constraint, allowing the recursion to focus on the stripped amplitude components.

Berends–Giele Recursion for Off-Shell Form Factors

The computational engine for the single-minus amplitudes is the Berends–Giele (BG) recursion, which the paper derives in Appendix B and uses as its primary tool. Rather than summing Feynman diagrams directly (whose number grows faster than factorially), the BG recursion computes off-shell form factors ℱ₁⋯ₘ — objects where one leg is off-shell (having nonzero p²) and the other m−1 are on-shell. The full on-shell amplitude is recovered by an LSZ reduction: taking the off-shell leg to its mass shell and extracting the residue of the propagator pole.

In the paper's conventions (with the negative-helicity particle taken last for notational convenience, a choice that is "insensitive" to the final result), the relationship is:

$\mathcal{A}_{1\cdots n} = \lim_{p_n^2 \to 0} -i p_n^2 \, \mathcal{F}_{1\cdots n-1} \, \delta^4\left(\sum_{i=1}^n p_i\right), \quad p_n = -\sum_{i=1}^{n-1} p_i$

What this computes: the on-shell amplitude 𝒜 is the residue of the off-shell form factor ℱ at the pole pₙ² = 0. The form factor ℱ₁⋯ₙ₋₁ takes the first n−1 particles as on-shell (with their physical momenta) and the n-th as the off-shell continuation, whose momentum is fixed by momentum conservation to be the negative sum of the on-shell momenta. The factor −i pₙ² extracts the pole residue, and the overall δ⁴ enforces momentum conservation.

Why this formulation: it converts the n-point on-shell problem into an (n−1)-point off-shell problem that satisfies a simpler recursion. The recursion for ℱ, when all on-shell legs are plus-helicity gluons (as they are for the single-minus case, except the off-shell leg which carries the remaining helicity), is given by the self-dual Yang–Mills (SDYM) recursion:

$\mathcal{F}_{1\cdots m} = \frac{1}{p_{1\cdots m}^2 + i\varepsilon} \sum_{j=1}^{m-1} [\tilde{\lambda}_{1\cdots j} \, \tilde{\lambda}_{j+1\cdots m}] \, \mathcal{F}_{1\cdots j} \, \mathcal{F}_{j+1\cdots m}$

where:

p₁⋯ₘ = ∑ᵢ₌₁ᵐ pᵢ is the total momentum of the block {1, …, m},
λ̃₁⋯ⱼ = ∑ᵢ₌₁ʲ λ̃ᵢ is the partial sum of right-handed spinors (converting to the paper's notation where λ̃ carries the square-bracket helicity),
[λ̃₁⋯ⱼ λ̃ⱼ₊₁⋯ₘ] = [λ̃₁⋯ⱼ |λ̃ⱼ₊₁⋯ₘ|] = εᵅ̇ᵝ̇ λ̃₁⋯ⱼ,ᵅ̇ λ̃ⱼ₊₁⋯ₘ,ᵝ̇ is the spinor contraction,
ℱ₁⋯ⱼ and ℱⱼ₊₁⋯ₘ are form factors for the left and right sub-blocks.

What this recursion physically means: to compute the form factor for a block of m gluons, you sum over all possible ways to split the ordered block into a left part (1 … j) and a right part (j+1 … m). For each split, you multiply: (1) the 1/p² propagator for the internal off-shell line, (2) the three-gluon vertex factor [λ̃ₗₑ𝚏ₜ λ̃ᵣᵢ𝓰ₕₜ] (which is the SDYM cubic vertex in spinor-helicity), and (3) the form factors for the left and right sub-blocks. This recursion effectively sums all tree diagrams where gluons are attached in the prescribed color order, with the off-shell line propagating between the left and right clusters.

Why this recursion is sufficient: the paper notes that for all-plus on-shell legs, the Yang–Mills BG recursion reduces to the SDYM recursion. This is a known result — the self-dual sector captures the all-plus helicity configuration. For the single-minus problem, the off-shell leg carries the minus helicity (which is why it is placed last in the appendix derivation), and the recursion builds the full amplitude from all-plus form factors. The SDYM recursion is "essentially the equation of motion of the theory and determines its classical solutions."

The Preamplitude Ā and Vertex Function V

The BG recursion is solved by introducing two fundamental building blocks: the preamplitude Ā_S and the vertex function V.

The preamplitude Ā_S is defined for any ordered set of gluons S = (q, …, p) and serves as the combinatorial kernel from which form factors and amplitudes are built. It satisfies boundary conditions:

$\bar{A}_q = 1, \quad \bar{A}_{qp} = 0$

for singleton and two-element lists, respectively, and for |S| ≥ 3 is defined recursively by:

$\bar{A}_{q\cdots p} = -\sum_{\text{o.p.}} V_{\tilde{\lambda}_{S_1} \cdots \tilde{\lambda}_{S_A}} \prod_{a=1}^A \bar{A}_{S_a}$

where the sum is over all ordered partitions of (q … p) = (S₁|S₂|…|S_A*) into A ≥ 3 parts. An ordered partition means the list is divided into consecutive blocks without reordering — so (1,2,3,4) can be partitioned as (1,2|3,4) or (1|2|3,4) but not (2,1|3,4).

What this computes: Ā for a block of A gluons is computed by iterating over all ways to split the block into A ≥ 3 sub-blocks, multiplying the vertex function V for those sub-blocks by the product of the Ā's of each sub-block, and summing with a minus sign. The base case Ā_q = 1 for a single gluon is the "free propagation" normalization. The case Ā_qp = 0 for two gluons reflects that there are no nontrivial two-point vertices in the SDYM recursion at this level.

Why this specific recursion: it mirrors the structure of the form factor solution (Equation 65) where vertex functions V stitch together blocks. The Ā objects are essentially the amplitudes stripped of the PT propagator factors — they are the "contact" or "vertex" contributions that remain after factoring out the propagator structure. The minus sign in Equation 19 is convention, matching the minus sign that appears in the BG kernel.

The vertex function V is the central algebraic object. For a list of singleton blocks λ̃₁, …, λ̃ₙ (each block containing exactly one gluon), it is defined as:

$V_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} = \prod_{k=1}^{n-1} \mathrm{sg}_{k,k+1} \, \Theta\!\left(-\frac{[\tilde{\lambda}_{1\cdots k} \, \tilde{\lambda}_{k+1\cdots m}]}{[\tilde{\lambda}_k \, \tilde{\lambda}_{k+1}]}\right)$

where:

sgₖ,ₖ₊₁ = sg([λ̃ₖ λ̃ₖ₊₁]) = sign of the bracket between adjacent gluons,
Θ is the step function (Θ(x) = 1 for x > 0, 0 otherwise),
[λ̃₁⋯ₖ λ̃ₖ₊₁⋯ₘ] is the bracket between the partial sum of spinors on the left (1 through k) and the partial sum on the right (k+1 through m, where m = n in the original list — the paper's notation is slightly imprecise here),
[λ̃ₖ λ̃ₖ₊₁] is the bracket between the adjacent individual gluons at the cut.

What this computes: V is a product over all n−1 adjacent pairs in the ordered list. For each cut position k, the factor sgₖ,ₖ₊₁ Θ(−[sum_left, sum_right] / [k, k+1]) is either 0 or ±1 depending on the sign of the ratio of brackets. The Θ-function enforces a causality condition: the ratio [λ̃₁⋯ₖ λ̃ₖ₊₁⋯ₙ] / [λ̃ₖ λ̃ₖ₊₁] must be negative — meaning the total momentum of the left block and the total momentum of the right block have a particular sign relationship relative to the adjacent gluon pair. If this condition is satisfied for all k, V is a product of signs; if any cut fails, V = 0.

Why this form: the Θ-functions emerge from the time-ordered perturbation theory treatment in the master identity (Appendix A). They encode the iε prescription in the propagators — which side of the branch cut the momentum lies on. The product structure means V is supported only in specific kinematic chambers defined by the signs of these ratios. This chamber structure is what makes the final amplitude piecewise-constant.

The paper also defines the barred vertex V̄_λ̃₁⋯λ̃ₙ, which is identical to V except that the argument of Θ has the opposite sign:

$\bar{V}_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} = \prod_{k=1}^{n-1} \mathrm{sg}_{k,k+1} \, \Theta\!\left(+\frac{[\tilde{\lambda}_{1\cdots k} \, \tilde{\lambda}_{k+1\cdots m}]}{[\tilde{\lambda}_k \, \tilde{\lambda}_{k+1}]}\right)$

V and V̄ partition the kinematic space — for any configuration, exactly one of V or V̄ can be nonzero (or both can be zero), because Θ(−x) and Θ(+x) are complementary supports (up to the boundary x = 0, which is measure-zero).

The on-shell Parke-Taylor factor PT̂ is defined as the difference:

$\widehat{\mathrm{PT}}_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} = V_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} - \bar{V}_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n}$

This object naturally emerges from the LSZ reduction as the on-shell limit of the incomplete Parke-Taylor factor PT (see Equation 72). It carries the combined kinematic constraints from the iε prescriptions on all propagators.

Master Identity and the PT–δ–V Relationship

A crucial technical component that underlies the entire computation is the master identity derived in Appendix A. This identity relates sums of products of propagators with iε prescriptions, constrained by a momentum-conserving δ-function, to products of δ-functions on individual momenta. The basic form (Equation 55) is:

$[a_1 (b_2 + i\varepsilon) + a_2 (b_1 + i\varepsilon)] \, \delta(a_1 b_1 + a_2 b_2) = -\frac{i}{2} [\mathrm{sg}(a_1) + \mathrm{sg}(a_2)] \, \delta(b_1) \, \delta(b_2)$

What this equation means in operational terms: the left-hand side is an expression that appears when summing over two propagator configurations in a Feynman diagram calculation, constrained by a δ-function that enforces a linear relation between the energy-like variables b₁ and b₂ with coefficients a₁ and a₂. The right-hand side shows that this combination is proportional to δ(b₁)δ(b₂) — meaning it is supported only when both b₁ = 0 and b₂ = 0, i.e., when both intermediate lines go on-shell simultaneously. The coefficient is a combination of sign functions of the coefficients a₁, a₂.

Why this identity matters: it converts the iε prescription (which tells us how to deform integration contours around poles) into a δ-function constraint (telling us where the amplitude has support). This is the mathematical mechanism by which the single-minus amplitude becomes localized to the half-collinear regime — the δ-functions force all ⟨ij⟩ to vanish. More specifically, the variables bᵣ correspond to p²ᵣ,ᵣ₊₁ (the squared momentum of adjacent pairs), and aᵣ correspond to combinations of spinor brackets. When the δ-function on the sum ∑ aᵣ bᵣ = 0 (momentum conservation in the z-direction) is combined with the propagator sum, it produces the product of δ(p²ᵣ,ᵣ₊₁) for all adjacent pairs — which, for massless particles, implies collinearity.

The general n-fold version of this identity (Equation A) is:

$\delta\!\left(\sum_{k=1}^n a_k b_k\right) \sum_{i=1}^n a_i \prod_{j \neq i} (b_j + i\varepsilon) = \frac{1}{(2i)^{n-1}} \left[\sum_{i_1} \mathrm{sg}(a_{i_1}) + \sum_{i_1 < i_2 < i_3} \mathrm{sg}(a_{i_1} a_{i_2} a_{i_3}) + \cdots\right] \prod_{i=1}^n \delta(b_i)$

What this computes: the right-hand side is an alternating sum over symmetric polynomials of the signs of the a's, multiplied by the product of all n δ(bᵢ). For example, when n = 3 (Equation 58):

$\text{RHS} = -\frac{1}{4} [\mathrm{sg}(a_1) + \mathrm{sg}(a_2) + \mathrm{sg}(a_3) + \mathrm{sg}(a_1) \mathrm{sg}(a_2) \mathrm{sg}(a_3)] \, \delta(b_1) \delta(b_2) \delta(b_3)$

The pattern: for any n, the coefficient is a sum over all odd-sized subsets of the a's of the product of their signs. This is proven in Appendix A via Fourier transform to the time domain, where the Θ-functions naturally emerge as step functions enforcing time ordering.

How this identity is applied. The paper uses a specialized form (Equation 66) that relates the incomplete Parke-Taylor factor PT₁⋯ₙ, the collinear δ-functions δ₁⋯ₙ, and the vertex function V:

$\mathrm{PT}_{1\cdots n} - \delta_{1\cdots n} V_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} = \sum_{j=1}^{n-1} \frac{[\tilde{\lambda}_{1\cdots j} \, \tilde{\lambda}_{j+1\cdots n}]}{p_{1\cdots n}^2 + i\varepsilon} \, \mathrm{PT}_{1\cdots j} \, \mathrm{PT}_{j+1\cdots n}$

What this identity states: the incomplete Parke-Taylor factor PT for an n-gluon block can be decomposed into two pieces. The first piece (δ₁⋯ₙ V) is a contact term supported only when all adjacent z-differences vanish — i.e., in the half-collinear regime. The second piece (the sum over j) is the "factorization" contribution where an intermediate propagator 1/p₁⋯ₙ² connects two sub-blocks. This identity is the core technical result that enables the recursion to handle the distributional nature of the collinear support correctly — it tells us how PT behaves when the zᵢ approach equality, where the naive algebraic simplification (which would neglect the iε) fails.

The proof sketch: the identity follows from the master identity (A) with specific choices of aᵣ and bᵣ given in Equation 67:

$a_r = -[\tilde{\lambda}_{1\cdots r} \, \tilde{\lambda}_{r+1\cdots n}] \prod_{\ell \neq r} [\tilde{\lambda}_\ell \, \tilde{\lambda}_{\ell+1}], \quad b_r = p_{r,r+1}^2$

for r = 1, …, n−1, along with aₙ = ∏_ℓ [ℓ, ℓ+1] and bₙ = p²₁⋯ₙ. These assignments map the master identity's abstract variables to the specific spinor-helicity quantities that appear in the PT factors and propagators. The details of this mapping are given but not fully expanded in the paper — the key point is that the b variables correspond to the squared momenta of adjacent pairs and the total momentum, while the a variables encode the vertex structure.

Solution of the BG Recursion Using Preamplitudes and PT Factors

With the identity (66) established, the paper proves in Appendix B.2 that the form factor ℱ₁⋯ₘ has the general solution:

$\mathcal{F}_{1\cdots m} = \sum_{\text{o.p.}} \mathrm{PT}_{K_1 \cdots K_A} \prod_{a=1}^A \left(\bar{A}_{S_a} \delta_{S_a}\right)$

where the sum is over all ordered partitions of (1 … m) into A blocks Sₐ, with Kₐ = ∑ᵢ∈Sₐ λ̃ᵢ being the block momentum (the sum of right-handed spinors in each block). The product includes both the preamplitude Ā_Sₐ and the collinear δ-function δ_Sₐ for each block.

What this formula says: the form factor for an m-gluon block is a sum over all possible ways to partition the ordered gluons into consecutive blocks, where each block contributes its preamplitude and its collinear δ-function support, and the blocks are connected by incomplete Parke-Taylor factors that provide the propagator structure. Setting m = 1 recovers ℱ₁ = PT₁ Ā₁ δ₁ = δ₁ (since Ā₁ = 1 and PT₁ is trivial for a single leg), which matches the base case of the recursion.

The induction proof. The paper verifies that this ansatz solves the recursion (64) by substituting into the RHS and using identity (66). The verification proceeds in three steps (Equations 68–71):

Insert the ansatz for ℱ₁⋯ⱼ and ℱⱼ₊₁⋯ₘ into the sum over j in (64), producing sums over partitions of the left and right blocks. The j-sum then factors into a sum over all nontrivial ordered partitions of the full set, where the PT factors from left and right combine according to (66).
Identity (66) produces two contributions for each combined partition: the PT₁⋯ₘ factor (connecting the blocks through the propagator structure) and the contact term δ₁⋯ₘ V. The PT terms reconstruct the nontrivial partitions on the LHS of the ansatz.
The contact terms — δ₁⋯ₘ times V times products of Ā's — combine across all nontrivial partitions to give exactly δ₁⋯ₘ Ā₁⋯ₘ, where Ā₁⋯ₘ is defined by the recursion (19). This closes the induction: the ansatz reproduces itself, with Ā satisfying the preamplitude recursion.

The preamplitude recursion re-derived. The paper shows that the δ-V contact terms from the induction step precisely match the recursion (19) for Ā:

$\bar{A}_{1\cdots m} = -\sum_{\text{n.t.p.}} V_{S_1 \cdots S_A} \prod_{a=1}^A \bar{A}_{S_a}$

where "n.t.p." means the sum is over nontrivial partitions — those with A ≥ 2 blocks. This is equivalent to the earlier definition (19) once the A ≥ 3 condition is reconciled (the A = 2 case has V = 0, which is why blocks of size 1 and 2 have Ā = 1 and Ā = 0 respectively). The preamplitude Ā is therefore the "contact" part of the form factor — the contribution that is fully localized to the half-collinear regime by δ-function support, without any PT propagator factors.

LSZ Reduction and the General Stripped Amplitude

The on-shell single-minus amplitude is obtained by LSZ reduction of the form factor solution. The paper works this out in Appendix B.3. The key step is evaluating the on-shell limit:

$\lim_{p_n^2 \to 0} p_n^2 \, \mathrm{PT}_{K_1 \cdots K_k} \, \delta^4\!\left(\sum_{a=1}^k K_a + p_n\right) = \widehat{\mathrm{PT}}_{K_1 \cdots K_k} \, \delta_{1\cdots k, n} \, \delta^2\!\left(\sum_j \tilde{\lambda}_j\right)$

What this computes: taking the off-shell leg n to its mass shell (pₙ² → 0) extracts the residue of the PT factor's pole. The result is PT̂ — the difference V − V̄ — times the half-collinear δ-functions (now including the n-th particle) times the right-handed momentum conservation δ-function. The pole extraction converts the incomplete PT factor (which has a 1/p² propagator) into the on-shell PT̂ (which encodes the iε-dependent chamber structure). This is a standard LSZ procedure dressed with the careful distributional treatment from the master identity.

After stripping off the universal momentum-conservation δ-functions and the half-collinear δ-functions, the stripped amplitude A₁⋯ₙ is obtained as (Equation 73, equivalent to Equation 21 in the main text):

$A_{1\cdots n} = -\sum_{\substack{(1\cdots n-1) = S_1 | \dots | S_k \\ k \geq 1}} \widehat{\mathrm{PT}}_{\tilde{\lambda}_{S_1} \cdots \tilde{\lambda}_{S_k}} \prod_{a=1}^k \bar{A}_{S_a}$

where the sum is over all ordered partitions of the first n−1 gluons into k ≥ 1 blocks. Note that the last gluon (particle n, the off-shell leg) is not included in the partitioning — it has been absorbed into the LSZ reduction. The minus sign and PT̂ factor come from the LSZ pole extraction.

Why this is the general formula: it expresses any n-point single-minus stripped amplitude as a sum over partitions of the n−1 plus-helicity gluons, where each partition's contribution is the on-shell Parke-Taylor factor PT̂ connecting the blocks times the product of the preamplitudes Ā for each block. All the complexity of the Feynman diagram sum has been packaged into the Ā recursion (which is local in the kinematic δ-function support) and the PT̂ factors (which encode the chamber structure from the iε prescriptions).

The recursive structure in practice. To compute A₁⋯ₙ, one must:

Compute Ā_S for all consecutive subsets S of {1, …, n−1} using recursion (19). This requires evaluating vertex functions V for all partitions of each subset into ≥ 3 blocks.
Assemble these Ā_S according to (21), summing over all partitions of (1 … n−1) and weighting by PT̂.
The result is a function of the λ̃ spinors (equivalently, of the ω and z̃ variables) that is piecewise-constant, jumping when sign functions change.

The explicit examples in Equations (29)–(32) show this working out for n = 3 through n = 6, producing expressions with 1, 2, 8, and 32 terms respectively — already substantially simpler than the corresponding Feynman diagram sums (which would involve roughly n! diagrams), but still combinatorially growing.

Collapse in Region ℛ₁: Vanishing of V

The dramatic simplification that yields the product formula (39) occurs in the kinematic region ℛ₁, defined by the condition that there exists an SO(2,2) frame in which:

$\omega_1 < 0, \quad \omega_a > 0 \;\; \text{for} \; a \in \{2, \dots, n\}$

What ℛ₁ means physically: particle 1 (the minus-helicity gluon) has negative "frequency" ω₁, while all plus-helicity gluons have positive ωₐ. In Klein signature, frequency sign is not Lorentz-invariant — but the condition that such a frame exists is invariant. Geometrically, it means λ̃₁ and the set {λ̃₂, …, λ̃ₙ} lie on opposite sides of a line through the origin in the λ̃-plane. This is the natural configuration for a "decay": one incoming particle (negative frequency in the chosen frame) splitting into multiple outgoing particles (positive frequency). The paper emphasizes that "there is no invariant meaning to a particle having positive frequency" in Klein signature, but ℛ₁ is an invariant region because it only requires the existence of such a frame.

Key simplification in ℛ₁: the sign functions simplify. In ℛ₁, the signs of adjacent brackets are:

$\mathrm{sg}_{ij} = \mathrm{sg}(\tilde{z}_{ij}) \;\; \text{for} \; i, j \geq 2, \quad \mathrm{sg}_{1j} = \mathrm{sg}(\tilde{z}_{j1}) \;\; \text{for} \; j \geq 2$

Why this happens: the frequency sign condition ω₁ < 0, ωₐ > 0 determines the sign of ω₁ωₐ in [1j] = ω₁ωⱼ z̃₁ⱼ, flipping it relative to the naive z̃-difference sign. For i, j ≥ 2, both ω's are positive, so sgᵢⱼ reduces to sg(z̃ᵢⱼ). For particle 1 with any j ≥ 2, the negative ω₁ flips the sign so sg₁ⱼ = sg(z̃ⱼ₁) = −sg(z̃₁ⱼ). The paper notes that "the ωk* cannot be eliminated from expressions such as sg{2,34}" — the frequency dependence persists in multi-particle brackets because they involve sums of λ̃'s.

The central theorem of Section II: vanishing of V. The paper proves that in ℛ₁, for any consecutive subset S ⊆ {2, …, n} with |S| ≥ 2,

$V_{\tilde{\lambda}_{S_1} \cdots \tilde{\lambda}_{S_k}} = 0$

where (S₁|…|Sₖ) is any ordered partition of S into k ≥ 3 blocks. In particular, for the fully unpartitioned list (all Sₐ singletons), V_λ̃₂⋯λ̃ₙ = 0 for n ≥ 3.

The proof (Section II.3.1) uses a weighted-variance argument. For each cut position j between 2 and n−1, define the left and right partial sums:

$\tilde{\lambda}_L = \sum_{a=2}^j \tilde{\lambda}_a = \Omega_L (1, \tilde{z}_L), \quad \tilde{\lambda}_R = \sum_{a=j+1}^n \tilde{\lambda}_a = \Omega_R (1, \tilde{z}_R)$

where Ω_L, Ω_R > 0 (sum of positive ω's) and z̃_L, z̃_R are weighted averages of the z̃'s in each block. Then:

$[\tilde{\lambda}_R \tilde{\lambda}_L] = \Omega_L \Omega_R \, \tilde{z}_{R,L}, \quad [\tilde{\lambda}_{j+1} \tilde{\lambda}_j] = \omega_j \omega_{j+1} \, \tilde{z}_{j+1,j}$

Since Ω_L Ω_R > 0 and ωⱼ ωⱼ₊₁ > 0, the sign of the ratio in the Θ-factor of V (Equation 20) is simply the sign of z̃_R,L / z̃ⱼ₊₁,ⱼ. The weighted-variance identity (not fully detailed in the paper but standard in such arguments) implies that as j runs from 2 to n−1, there must exist at least one cut j⋆ for which z̃_R,L and z̃ⱼ₊₁,ⱼ have the same sign. At that cut, the ratio is positive, so Θ(−positive) = 0, making the entire product in V vanish.

What this means intuitively: if all the z̃'s are organized left-to-right on the line, then the left-partial average z̃_L and right-partial average z̃_R must lie on opposite sides of the cut at some position — their difference has the same sign as the adjacent difference at the cut. The Θ-function in V demands the opposite sign, so at least one Θ-factor is always zero in ℛ₁. This is a "causality" or "time-ordering" condition — ℛ₁ corresponds to a kinematic configuration where the V-type vertex (which corresponds to one time-ordering) cannot contribute.

Consequences for preamplitudes. From the vanishing of V for all multi-gluon subsets, the preamplitude recursion (19) immediately implies:

$\bar{A}_S|_{\mathcal{R}_1} = 0 \quad \text{for all } |S| \geq 2, \quad \bar{A}_i = 1 \text{ (singleton)}$

Why: the recursion (19) expresses Ā_S as a sum over partitions into ≥ 3 blocks, each weighted by V on those blocks. Since all V's vanish for any blocks containing only particles from {2, …, n}, the sum is zero for any S with |S| ≥ 2. For singletons, the base case Ā_i = 1 remains unchanged. This is the key collapse: all nontrivial preamplitudes vanish in ℛ₁.

Collapse of the Amplitude Recursion

With all Ā_S = 0 for |S| ≥ 2, the amplitude formula (21) collapses dramatically. The sum over partitions of (1 … n−1) = (S₁|…|Sₖ) includes a product ∏ Ā_Sₐ. If any Sₐ has size ≥ 2, Ā_Sₐ = 0 and that partition contributes nothing. Only partitions where every block is a singleton can contribute. But there is only one such partition for the list (1, 2, …, n−1): (1|2|…|n−1). Thus:

$A_{1\cdots n}|_{\mathcal{R}_1} = -\widehat{\mathrm{PT}}_{\tilde{\lambda}_1 \tilde{\lambda}_2 \cdots \tilde{\lambda}_{n-1}}|_{\mathcal{R}_1}$

But wait — particle 1 has ω₁ < 0, which complicates the direct application of the vanishing theorem (which was proved for subsets of {2, …, n}). The paper handles this by using cyclicity of the color-ordered amplitude:

$A_{1\cdots n} = A_{2\cdots n 1}$

Now the list is (2, 3, …, n, 1), with all particles except the last (1) having positive ω. Applying the recursion (21) to A₂⋯ₙ₁ with the minus-helicity particle 1 in the last position, the sum is over partitions of the plus-helicity word (2 … n). Since all Ā for blocks from {2, …, n} vanish (|S| ≥ 2), only the all-singleton partition (2|3|…|n) survives:

$A_{2\cdots n 1}|_{\mathcal{R}_1} = -\widehat{\mathrm{PT}}_{\tilde{\lambda}_2 \tilde{\lambda}_3 \cdots \tilde{\lambda}_n}|_{\mathcal{R}_1}$

Now, for the list (2, …, n) which consists entirely of plus-helicity gluons with ω > 0, the vanishing theorem applies: V_λ̃₂⋯λ̃ₙ = 0. Therefore:

$\widehat{\mathrm{PT}}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} = V_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} - \bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} = -\bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n}$

And so:

$A_{1\cdots n}|_{\mathcal{R}_1} = A_{2\cdots n 1}|_{\mathcal{R}_1} = \bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n}|_{\mathcal{R}_1}$

What this proves: the n-point single-minus stripped amplitude in ℛ₁ reduces to a single V̄ vertex function evaluated on the plus-helicity gluons 2 through n. The entire complexity of the partition sum, the Feynman diagram expansion, and the preamplitude recursion has collapsed to one term. This is a genuinely remarkable simplification — the n-point amplitude is given by an (n−1)-fold product of Θ and sign functions rather than a combinatorially exploding sum.

Evaluating V̄ and Deriving the Product Formula (39)

The final step evaluates V̄_λ̃₂⋯λ̃ₙ in ℛ₁ and shows it factorizes into the product form (39). By definition:

$\bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} = \prod_{m=2}^{n-1} \mathrm{sg}_{m,m+1} \; \Theta\!\left(\frac{[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_{m+1\cdots n}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}\right)$

Momentum conservation simplification. Using ∑ᵢ₌₁ⁿ λ̃ᵢ = 0 (total momentum conservation, which follows from the δ-function support), we have:

$\tilde{\lambda}_{m+1\cdots n} = \sum_{a=m+1}^n \tilde{\lambda}_a = -\tilde{\lambda}_1 - \sum_{a=2}^m \tilde{\lambda}_a = -\tilde{\lambda}_1 - \tilde{\lambda}_{2\cdots m}$

Why this substitution matters: it introduces particle 1 (the negative-frequency minus-helicity gluon) into the brackets that define V̄, shifting the dependence from sums over plus-helicity particles only to expressions involving λ̃₁. This is what connects the V̄ of the plus-helicity block to the full amplitude.

The bracket in the numerator becomes:

$[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_{m+1\cdots n}] = [\tilde{\lambda}_{2\cdots m} \, (-\tilde{\lambda}_1 - \tilde{\lambda}_{2\cdots m})] = -[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_1]$

where we used antisymmetry [X, X] = 0 and linearity. Using antisymmetry again to flip the order (which introduces a sign), the ratio in the Θ-factor becomes:

$\frac{[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_{m+1\cdots n}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]} = \frac{-[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_1]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]} = -\frac{[\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}$

But the Θ-function in V̄ takes the positive of this ratio (Equation 52 has Θ(+x)), and Θ(−x) = 1 − Θ(x) for x ≠ 0. However, the paper takes a different route to the factorized form by directly relating Θ to sign functions:

$\Theta(x) = \frac{1 + \mathrm{sg}(x)}{2}$

Applying this to the ratio with momentum conservation:

$\bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} = \prod_{m=2}^{n-1} \mathrm{sg}_{m,m+1} \; \frac{1}{2}\left(1 + \mathrm{sg}\!\left(\frac{[\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}\right)\right)$

Removing the frequency dependence. The sign of the ratio [λ̃₁ λ̃₂⋯ₘ] / [λ̃ₘ λ̃ₘ₊₁] depends on ω's, but it can be simplified using the product structure of the brackets:

$[\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}] = \sum_{a=2}^m [\tilde{\lambda}_1 \, \tilde{\lambda}_a] = \omega_1 \sum_{a=2}^m \omega_a \tilde{z}_{1a}$

Since all ωₐ > 0 for a ≥ 2, and ω₁ < 0, the overall sign of each term [λ̃₁ λ̃ₐ] is determined by z̃₁ₐ = −z̃ₐ₁. The crucial observation is that:

$\mathrm{sg}([\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]) = \mathrm{sg}_{1,2\cdots m}$

by definition of the notation sg₁,₂⋯ₘ as the sign of the bracket between λ̃₁ and the sum λ̃₂⋯ₘ. Now, since ω₁ < 0 and all other ω > 0:

$[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}] = \omega_m \omega_{m+1} \tilde{z}_{m,m+1} \implies \mathrm{sg}([\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]) = \mathrm{sg}(\tilde{z}_{m,m+1}) = \mathrm{sg}_{m,m+1}$

where the last equality uses the simplification (34) that sgₘ,ₘ₊₁ = sg(z̃ₘ,ₘ₊₁) for m, m+1 ≥ 2.

The sign of the full ratio is then:

$\mathrm{sg}\!\left(\frac{[\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}\right) = \mathrm{sg}([\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]) \cdot \mathrm{sg}([\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]) = \mathrm{sg}_{1,2\cdots m} \cdot \mathrm{sg}_{m,m+1}$

Since sgₘ,ₘ₊₁ = ±1, its square is 1. The product with the prefactor sgₘ,ₘ₊₁ in V̄ gives:

$\mathrm{sg}_{m,m+1} \cdot \frac{1}{2}\left(1 + \mathrm{sg}_{1,2\cdots m} \cdot \mathrm{sg}_{m,m+1}\right) = \frac{1}{2}\left(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m}\right)$

The factorized form emerges. Substituting this into the product over m = 2, …, n−1:

$A_{1\cdots n}|_{\mathcal{R}_1} = \bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n}|_{\mathcal{R}_1} = \prod_{m=2}^{n-1} \frac{1}{2}\left(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m}\right)$

which is exactly the conjectured formula (39):

$A_{1\cdots n}|_{\mathcal{R}_1} = \frac{1}{2^{n-2}} \prod_{m=2}^{n-1} \left(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m}\right)$

Structure of the result. Each factor is ½(sgₘ,ₘ₊₁ + sg₁,₂⋯ₘ). Since each sign function takes values ±1, the sum sgₘ,ₘ₊₁ + sg₁,₂⋯ₘ can be:

+2 if both signs are +1 → factor = +1
−2 if both signs are −1 → factor = −1
0 if the signs are opposite → factor = 0

Therefore each factor in the product is an element of {−1, 0, +1}, and the product of n−2 such factors divided by 2ⁿ⁻² gives the final value. The entire stripped amplitude A₁⋯ₙ|ℛ₁ is thus piecewise-constant, taking integer values +1, −1, or 0 (the factor of 2ⁿ⁻² cancels the denominator, but the zeros can make the overall product 0).

Why this formula is remarkable: the n-point amplitude — which in Feynman diagrams involves a sum over roughly n! terms with intricate momentum dependence — reduces to a product of n−2 simple sign-based projection operators. Each factor is a "signed projection" that asks: does the bracket between particle 1 and the partial sum of particles 2 through m have the same sign as the bracket between adjacent particles m and m+1? If yes, the factor is ±1; if no, it is 0 and the entire amplitude vanishes. The walls where sg functions change sign (where brackets vanish) are the codimension-one boundaries of the chambers in kinematic space.

The explicit pattern for low n. The paper verified this pattern for n = 3 through n = 6, showing it matches the general recursion:

n = 3: one factor m = 2: ½(sg₂₃ + sg₁₂) — but sg₁₂ in ℛ₁ is −sg₂₁, and using momentum conservation this simplifies to the known result A₁₂₃|ℛ₁ = ½(sg₁₂ + sg₂₃).
n = 4: product over m = 2,3: ¼(sg₂₃ + sg₁₂)(sg₃₄ + sg₁₂₃) which matches (36).
n = 5: product over m = 2,3,4: ⅛(sg₂₃ + sg₁₂)(sg₃₄ + sg₁₂₃)(sg₄₅ + sg₁₂₃₄) matching (37).
n = 6: product over m = 2,3,4,5 matching (38).

The product structure means the walls of the amplitude are explicit and factorized: the amplitude jumps only when one of the n−2 projection conditions changes sign, and these conditions are decoupled — each involves a different partial sum of the right-handed momenta.

Consistency Checks and Mathematical Structure

The paper verifies that formula (39) satisfies five nontrivial consistency conditions, none of which are evident from direct inspection:

Weinberg's soft theorem (Equation 28):

$\lim_{\omega_n \to 0} A_{1\cdots n} = \frac{1}{2}(\mathrm{sg}_{n-1,n} + \mathrm{sg}_{n1}) \, A_{1\cdots n-1}$

What this means physically: when the frequency of the last plus-helicity gluon is taken to zero (the "soft limit"), the amplitude factorizes into a soft factor times the (n−1)-point amplitude. The soft factor depends only on the helicities and momenta of the adjacent particles. For formula (39), the ωₙ → 0 limit of the last factor in the product (½(sgₙ₋₁,ₙ + sg₁,₂⋯ₙ₋₁)) reproduces precisely the required soft factor times the (n−2)-fold product for A₁⋯ₙ₋₁.

Why this check is nontrivial: the soft theorem is a universal property of gauge theory amplitudes derived from gauge invariance and the structure of soft singularities. It is not manifest in the product form (39) — it requires using momentum conservation and the specific structure of how ωₙ enters sg₁,₂⋯ₙ₋₁. The verification (which the paper states was done but does not fully display) confirms that the half-collinear amplitudes obey the same IR structure as ordinary amplitudes.

Cyclicity (A₁₂⋯ₙ = A₂⋯ₙ₁) is not manifest because ℛ₁ singles out particle 1 by its negative frequency. However, the paper notes that "we can trivially construct a cyclically invariant answer by using cyclicity to extend (39) to other regions ℛₖ where only particle k has ωₖ < 0." The full amplitude across all kinematic regions is obtained by cyclic permutation of the formula.

Reflection symmetry, U(1) decoupling, and Kleiss–Kuijf relations are all stated to hold but the detailed verifications are deferred ("Details of this calculation will appear elsewhere"). These are standard color-ordering identities that any consistent tree amplitude must satisfy, and their verification provides a strong check that the result is not an artifact of the computational method.

Summary of the derivation's logical structure. The computation proceeds from Feynman diagrams → Berends–Giele recursion (summing diagrams as an off-shell recursion) → form factor solution (packaging the recursion into Ā preamplitudes and PT factors) → LSZ reduction (putting the off-shell leg on shell, yielding the general stripped amplitude formula) → specialization to ℛ₁ (where the frequency sign condition forces all multi-gluon Ā to vanish and all V vertex functions for plus-helicity blocks to vanish) → collapse to V̄ → factorization of V̄ into the product form (39). Each step is mathematically rigorous, anchored in the master identity (Appendix A) that controls the distributional support and the iε prescriptions. The final result is a formula of extraordinary simplicity — a product of n−2 signs and projection operators — that encodes the same physics as the exponentially complex Feynman diagram expansion.

4. Key Insights and Innovations

Innovation 1: Demolishing a Forty-Year-Old "Theorem" by Exploiting a Distributional Loophole

The most intellectually jarring contribution of this paper is not a new technique but the destruction of a foundational belief: that single-minus gluon tree amplitudes vanish. This belief has been treated as a theorem in the amplitudes community since at least the 1980s, embedded in textbook treatments (Elvang and Huang, 2013) and enshrined in the very terminology "MHV" — maximally helicity violating — which presupposes that n−2 plus helicities is the maximum allowed for n gluons. The paper does not refine or extend this claim; it proves it false under specific kinematic conditions that had been overlooked for decades.

What makes this innovation distinctive is how the authors locate the error in the prevailing argument. The standard vanishing proof (Section I.1) is a helicity power-counting argument: choose the reference spinor |r⟩ = |1⟩ to make all polarization vectors orthogonal, then note there are insufficient momentum factors from vertices to contract all n polarization vectors. This argument is mathematically correct for generic kinematics — the paper does not dispute that. The innovation is recognizing that the argument contains a hidden regularity condition: the choice |r⟩ = |1⟩ is invalid precisely when ⟨1_a_⟩ = 0 for any a, because then the plus-helicity polarization vectors εₐ⁺ diverge (they carry 1/⟨r a⟩ factors). The vanishing argument therefore assumes ⟨1_a_⟩ ≠ 0 — an assumption that is true generically, but never stated explicitly, and that fails on a codimension-(n−1) kinematic locus where all angle brackets vanish simultaneously.

This is not merely a boundary-condition technicality. The authors show by induction that the full amplitude has δ-function support forcing all ⟨ij⟩ = 0 — meaning the amplitude is nonzero only on the very locus where the vanishing argument's premise breaks down. This is a clean logical inversion: the argument that "proved" the amplitude vanishes actually shows that if it is nonzero, it must be localized to exactly the configuration where the proof's assumptions are violated. The amplitude and the vanishing argument carve up kinematic space into complementary domains.

Comparison to prior assumptions. The 3-point single-minus (anti-MHV) amplitude was already known to have δ⟨12⟩δ⟨13⟩ support — a fact that had been accepted as an isolated 3-point curiosity without generalization. The MHV classification, BCFW recursion, CSW construction, and amplituhedron program all take the nonexistence of single-minus amplitudes for n ≥ 4 as a given. The paper shows this was an accident of working in Minkowski signature, where ⟨ij⟩ = 0 forces [ij] = 0 as well (by the complex conjugation relation λ̃ = ±λ*), collapsing all momenta to proportionality and making the amplitude measure-zero. In (2,2) Klein signature — the natural arena for analytic continuation of scattering amplitudes — angle and square brackets are independent real variables, and the half-collinear regime supports a rich chamber structure.

Significance beyond the specific result. This finding forces a taxonomic revision of Yang–Mills tree amplitudes. "MHV" is no longer maximally helicity violating — there exists a class of amplitudes with even fewer minus helicities. The entire tower of amplitudes organized by helicity sectors needs to be extended downward. Practically, this means that any formalism that treats MHV amplitudes as primitive building blocks (BCFW shifts, MHV vertex expansions, Grassmannian formulations) must now account for these "more-than-maximally" helicity-violating contributions as independent primitives, or explain why they decouple in specific contexts. The paper does not fully resolve this, but the identification of the half-collinear regime as the support of these amplitudes is the essential diagnostic move that enables such a resolution.

The authors explicitly connect this to self-dual Yang–Mills theory (SDYM) , where there has been a longstanding puzzle: the classical solution space (instantons, twistor constructions) is extremely rich, but the Feynman tree expansion was thought to yield only trivial results beyond 3 points. Single-minus amplitudes in the half-collinear regime provide the missing quantum objects that can match the complexity of the classical theory. This is not a side note — it resolves a genuine theoretical tension that had persisted for decades.

Innovation 2: The Vertex Function as a Chamber-Based Causality Filter

The paper introduces the vertex function V (Equation 20) and its barred partner V̄ as the central algebraic objects governing the recursion, and in doing so reframes the entire computation in terms of causality conditions encoded via step functions. This is conceptually novel because it extracts the physics of the iε prescription — typically a formal regularization detail — into an explicit, piecewise-constant geometric object that partitions kinematic space into chambers.

In conventional Feynman diagram calculations, the iε prescription lives in the propagator denominators and is handled by contour deformation or Wick rotation. It rarely surfaces in the final answer for tree amplitudes (which are rational functions of kinematic invariants). Here, precisely because the amplitude is supported on a locus where propagator denominators vanish (the half-collinear regime), the iε becomes the dominant dynamical feature — it determines whether certain combinations of momenta satisfy "time-ordering" constraints and thus whether the amplitude is nonzero.

The vertex function encodes this as a product over adjacent gluon pairs:

$V_{\tilde{\lambda}_1 \cdots \tilde{\lambda}_n} = \prod_{k=1}^{n-1} \mathrm{sg}_{k,k+1} \, \Theta\!\left(-\frac{[\tilde{\lambda}_{1\cdots k} \, \tilde{\lambda}_{k+1\cdots n}]}{[\tilde{\lambda}_k \, \tilde{\lambda}_{k+1}]}\right)$

Each Θ-factor is a local causality condition: the ratio of two spinor brackets must have a specific sign. The product structure means V is supported only in chambers of kinematic space where all n−1 cuts simultaneously satisfy their sign constraints. This is a convexity filter — a single failing cut anywhere in the chain forces V = 0.

What's fundamentally new here. Prior work on Berends–Giele recursion and Parke-Taylor identities does not elevate the vertex function to this central role. The standard approach treats iε as a passive regularization that drops out of rational functions. The paper's master identity (Appendix A) shows that when the δ-function support is handled correctly, the iε produces active chamber structure — the amplitude is not a single rational function but a piecewise-constant integer that jumps across codimension-one walls where bracket signs flip. This is closer to the structure of MHV amplitudes in split signature (where the Parke-Taylor factor becomes a product of sign functions) than to standard Minkowski-space tree amplitudes.

The partition of kinematic space into V-dominated and V̄-dominated chambers (via PT̂ = V − V̄) is a novel decomposition. It means the amplitude is the difference of two step-function-supported contributions, each corresponding to opposite time-ordering of intermediate states. The vanishing of V in region ℛ₁ (Section II.3.1) is not an algebraic coincidence but a global causality theorem: in the decay kinematics where one gluon has negative frequency and all others have positive frequency, the V-type time-ordering is kinematically impossible for any consecutive block of plus-helicity gluons. This is proven via a weighted-variance argument that is elementary in structure but profound in implication — it shows that the causal structure of the theory forbids certain channel combinations in this kinematic regime, collapsing the entire partition sum to a single term.

Connection to celestial holography and asymptotic symmetries. The chamber structure identified here — where amplitudes are piecewise-constant integers that jump across walls defined by bracket sign changes — is characteristic of celestial amplitudes in split signature and the S-algebra/_Lw_₁₊∞ symmetry structures that the paper flags as future directions. The vertex function V can be understood as a building block for celestial operator product expansions, where the Θ-function chambers correspond to different analytic continuations across the celestial sphere. The paper does not develop this connection, but the structure it uncovers is precisely the type of object that celestial holography needs.

Innovation 3: An Exponential-to-Linear Complexity Collapse via Kinematic Selection

The path from the general recursion (Equation 21) to the product formula (39) exemplifies a phenomenon that the field has long sought but rarely achieved: exponential complexity in Feynman diagram sums collapsing to linear complexity via a clever choice of kinematic region. This is not merely a simplification — it is a diagnostic discovery about the structure of gauge theory amplitudes.

Consider what the recursion (21) demands in general: to compute the n-point stripped amplitude, one must sum over all ordered partitions of the (n−1) plus-helicity gluons into an arbitrary number of blocks, compute preamplitudes Ā for each block (each of which involves sums over partitions of that block), and assemble the results with PT̂ factors. The number of ordered partitions of n−1 elements grows like 2ⁿ⁻², and the preamplitude recursion compounds this. The explicit 6-point example (Equation 32) already contains 32 terms, and the general scaling is exponential.

The paper discovers that in region ℛ₁ — defined by the physically natural condition that there exists a frame where the minus-helicity gluon is incoming (negative frequency) and all plus-helicity gluons are outgoing (positive frequency) — every sub-leading preamplitude vanishes. The recursion collapses from a sum over all partitions to a single term: the fully unpartitioned V̄ vertex acting on the complete plus-helicity block. The n-point amplitude is given by an n−2-fold product of simple sign sums, meaning the computational complexity scales linearly in n (evaluating n−2 factors) rather than exponentially.

Why this is fundamentally different from known simplifications. The Parke-Taylor formula for MHV amplitudes is also a dramatic simplification (from factorial to single-term), but it achieves this through helicity-weight algebra: the special helicity configuration (two minus, rest plus) forces massive cancellations that leave a single term. The single-minus simplification operates by a kinematic restriction — not by changing the helicity assignment, but by selecting a subregion of the half-collinear regime where the causal structure (the Θ-function constraints in V) collapses. The result is not a rational function (like the Parke-Taylor denominator) but an integer-valued product of signs — a topological rather than analytic object.

This collapse reveals that the complexity of the general single-minus amplitude is not intrinsic to the physics but is an artifact of summing over kinematic chambers that are mutually exclusive in region ℛ₁. In the general half-collinear regime, different partitions contribute in different chambers, and the sum produces a piecewise-constant function that jumps at chamber boundaries. In ℛ₁, the causal conditions eliminate all but one chamber, and the amplitude simplifies to its value on that chamber. The exponential complexity was always in the chamber sum — not in the physics of any single chamber.

The diagnostic power of this collapse. This finding has methodological implications beyond this specific calculation. It suggests that appropriate kinematic restriction can reduce exponential complexity to polynomial (or constant) complexity in gauge theory amplitudes, by selecting subregions where the causal structure of the iε prescription eliminates most Feynman diagram contributions. This is a design principle for future calculations: rather than fighting exponential complexity head-on, identify kinematic regions where causality constraints prune the diagrammatic sum. The paper's region ℛ₁ is the first explicit example of this principle yielding a closed-form result for a class of amplitudes previously thought to be identically zero.

Innovation 4: Identification of Frequency-Sign as a Kinematic Invariant Organizing Amplitude Structure

The paper discovers that the sign of the frequency parameter ω in a given Lorentz frame — a quantity with no invariant meaning in Minkowski signature — becomes a powerful organizing principle for amplitude chambers in Klein signature. This is a conceptual reframing of what counts as a "kinematic invariant" for scattering amplitudes in split signature.

In Minkowski space, the frequency sign of a massless particle is not Lorentz-invariant: a positive-frequency particle in one frame can appear negative-frequency in another. Consequently, decompositions based on "incoming" vs. "outgoing" particles are frame-dependent, and amplitudes are typically written in forms that are manifestly crossing-symmetric. In Klein (2,2) signature, the situation is different: while the frequency sign of any individual particle can still be changed by an SO(2,2) transformation, the pattern of frequency signs across all particles can define invariant regions. Specifically, the condition that there exists some frame where particle 1 has negative frequency and particles 2 through n have positive frequency is an SO(2,2)-invariant statement. This is the definition of region ℛ₁ (Equation 33), and it partitions the half-collinear regime into distinct chambers labeled by which particle is "the incoming one."

What makes this framing innovative. The standard approach to scattering amplitudes avoids frame-dependent characterizations entirely — one works with Mandelstam invariants, spinor brackets, and crossing symmetry to ensure results are manifestly invariant. The paper's insight is that embracing a specific frame choice (no single frame — but the existence of one) provides a powerful organizational principle. In ℛ₁, the sign functions simplify (Equation 34): sg_ij_ = sg(z̃ᵢⱼ) for i, j ≥ 2, but sg₁ⱼ = sg(_z̃ⱼ_₁) — the negative frequency of particle 1 flips the sign of its brackets relative to the z̃ differences. This asymmetry between particle 1 and the others is what drives the vanishing of V (the weighted-variance proof in Section II.3.1 uses the positivity of all ω for particles 2 through n crucially).

The frequency-sign condition effectively diagonalizes the causal structure: in the frame where _ω_₁ < 0 and _ω_ₐ > 0, the V-type vertex (which corresponds to a particular time-ordering of intermediate states) becomes impossible for any block of plus-helicity gluons. This is a global causality constraint — the minus-helicity gluon is literally "earlier" or "later" in the time-ordered perturbation theory sense than all the plus-helicity gluons, and this temporal ordering selects a unique channel for the amplitude.

Why this was not previously appreciated. The amplitudes community's focus on Minkowski signature — where analytic continuation to complex momenta is used for BCFW recursion but the underlying signature is (3,1) — meant that the independence of angle and square brackets in Klein signature was never fully exploited as an organizing principle. Split signature has been used for celestial holography and for understanding the analytic structure of amplitudes, but the idea that frequency-sign patterns define invariant kinematic chambers with dramatically simplified amplitude structure appears to be new. The paper's explicit construction of ℛ₁ as such a chamber, and the proof that the amplitude collapses there, establishes this as a general principle rather than an isolated curiosity.

Generality of the principle. The paper notes that the construction extends "directly from gluon to graviton amplitudes" and "has a simple supersymmetrization." This suggests that frequency-sign chambers are a universal feature of massless scattering in Klein signature, not specific to Yang–Mills theory. The S-algebra and _Lw_₁₊∞ structures that the paper flags as relevant for these amplitudes are precisely the symmetry algebras that act on the celestial sphere in split signature — and their action is naturally organized by the frequency-sign decomposition. The paper thus positions the half-collinear, frequency-sign-selected amplitudes as the natural objects on which these infinite-dimensional symmetry algebras act, connecting a concrete perturbative calculation to the broader celestial holography program.

Innovation 5: Signed Projection Operators as the Universal Building Block of Nonzero Single-Minus Amplitudes

The final formula (39):

$A_{1\cdots n}|_{\mathcal{R}_1} = \frac{1}{2^{n-2}} \prod_{m=2}^{n-1} \left(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m}\right)$

reveals that the stripped amplitude in ℛ₁ is a product of n−2 signed projection operators, each taking values in {+1, −1, 0}. This is a genuinely novel algebraic structure for a scattering amplitude — one that is not polynomial, not rational, not transcendental, but combinatorial and Boolean.

Each factor is a projection onto a two-dimensional subspace of sign configurations: it asks whether the sign of the bracket between particle 1 and the partial sum of particles 2 through m matches the sign of the bracket between adjacent particles m and m+1. If they match, the factor is ±1 (the common sign); if they disagree, the factor is 0 and the entire amplitude vanishes. The amplitude is therefore nonzero only when a specific global pattern of sign correlations is satisfied across all n−2 adjacent-plus-partial-sum pairs.

Why this structure is novel. Known tree-level amplitudes in gauge theory and gravity fall into a few structural classes: rational functions of spinor brackets (MHV, NMHV, etc., via BCFW), polylogarithmic functions (loop level), or determinants of matrices (Cachazo-He-Yuan formulations). A product of sign-based projection operators does not fit any of these categories. It is closer to a topological invariant — the amplitude counts something like the number of chambers where sign constraints are satisfied — than to a conventional analytic function. The piecewise-constant, integer-valued nature of the result (each factor is integer, so the product is integer after normalization) means the amplitude is a locally constant function on the space of z̃ coordinates, jumping by discrete amounts across codimension-one walls where any bracket changes sign.

This structure has a natural interpretation in terms of gauge-theoretic Abelianization. The half-collinear regime ⟨ij⟩ = 0 means all gluons share the same left-handed spinor direction — they are "collinear in λ." The right-handed spinors λ̃ remain generic, and the amplitude depends only on their relative signs. The product formula (39) is essentially computing whether the configuration of λ̃ spinors (in z̃-space) satisfies a set of ordering constraints relative to the special particle 1. Each factor sg₁,₂⋯ₘ compares the sign of the bracket between particle 1 and the cumulative sum of particles 2 through m to the sign of the bracket between adjacent particles in the chain. This can be rephrased as: is particle 1 "to the left" or "to the right" of the running partial sum, and does this agree with the local ordering of adjacent plus-helicity gluons? The amplitude is nonzero only when there is consistent hierarchical ordering across all partial sums.

Comparison to orthogonal structures. MHV amplitudes are products of Parke-Taylor denominators — rational functions that diverge on collinear singularities. The single-minus amplitude in ℛ₁ has no singularities except at the walls where it jumps — it is a step function, not a rational function. This reflects the fact that the half-collinear δ-function support has already absorbed all the collinear divergences (the 1/⟨ij⟩ poles in the Parke-Taylor factor are replaced by δ⟨ij⟩ constraints), and what remains is the finite, chamber-dependent coefficient. The product formula (39) is the cleanest possible answer for what lives on the half-collinear locus — it is not a limit or an approximation, but the exact finite part after the distributional support is factored out.

Implications for computational complexity. The product structure means that evaluating the n-point amplitude in ℛ₁ requires computing n−2 sign functions and performing (n−2) multiplications — an O(n) operation. This is a radical compression of the information in the Feynman diagram sum, which naively contains O(n!) terms, and even the general Berends–Giele recursion requires O(2ⁿ) partitions. The authors do not frame this as a practical computational advance (tree-level gluon amplitudes can already be computed efficiently via BCFW), but rather as evidence that the single-minus sector contains a hidden simplicity that only becomes visible after both the half-collinear restriction and the frequency-sign selection are applied. The existence of such a simple formula suggests a missing algebraic structure — perhaps related to the S-algebra or _Lw_₁₊∞ — that organizes these amplitudes into factorized products of elementary sign operators, and the paper's result is the first concrete data point toward that structure.

5. Experimental Analysis

Evaluation Methodology

Dataset. The paper does not use a machine-learning dataset in the conventional sense. The "data" consists of all possible configurations of n external gluon momenta in the half-collinear regime of (2,2) Klein signature, parameterized by the independent spinor variables z̃ᵢ and ωᵢ for i = 1, …, n. The "splits" correspond to kinematic regions — the general half-collinear regime (where all ⟨ij⟩ = 0) and the subregion ℛ₁ (where there exists a frame with ω₁ < 0 and all other ωₐ > 0). There is no training set, validation set, or held-out test set; the results are exact analytic expressions valid for all n, verified at specific low n by explicit calculation.

Base model(s). The computational framework is the Berends–Giele (BG) recursion for off-shell form factors in self-dual Yang–Mills theory (Equation 64), established in prior work by Berends and Giele (1988) and by Cangemi (1997). This is not a trained neural model but a recursive algebraic procedure — equivalent to summing all Feynman diagrams with one off-shell leg — that is applied analytically. The recursion is solved using the preamplitude Ā_S (Equation 19) and vertex functions V and V̄ (Equation 20), with distributional support handled via the master identity (Appendix A). There are no learned parameters, no training runs, and no hyperparameter sweeps; the entire analysis is a mathematical derivation.

Metrics. Since this is an analytic paper rather than an empirical one, the "metrics" are not numerical performance scores but consistency checks that the derived expressions must satisfy. The stripped amplitudes A₁⋯ₙ are required to obey five properties (Section I.3):

Cyclicity: A₁₂⋯ₙ = A₂⋯ₙ₁ (Equation 24).
Reflection symmetry: A₁₂⋯ₙ = (−1)ⁿ Aₙ⋯₂₁ (Equation 25).
U(1) decoupling identity (Equation 26).
Kleiss–Kuijf (KK) relations (Equation 27 for n = 5).
Weinberg's soft theorem: limωₙ→0 A₁⋯ₙ = ½(sgₙ₋₁,ₙ + sgₙ₁) A₁⋯ₙ₋₁ (Equation 28).

The "accuracy" is binary: the derived formulas either satisfy these identities (on the support of the half-collinear δ-functions, with the correct iε prescriptions) or they do not. The paper states that "we have verified by explicit calculation that they do indeed hold. Details of this calculation will appear elsewhere" (Section I.3).

Baselines. There are no conventional baselines in the machine-learning sense, since nothing is being trained or compared in performance. The relevant "baselines" are:

The standard vanishing claim: the assertion that single-minus amplitudes are identically zero for n ≥ 4 (the textbook result the paper refutes). The paper's explicit nonvanishing expressions for A₁₂₃ through A₁₂₃₄₅₆ (Equations 29–32) directly contradict this.
The general recursion (Equation 21): the unsimplified formula that sums over all ordered partitions of the plus-helicity gluons. The simplified product formula (39) is verified by showing it satisfies the recursion for all n; for low n, explicit reduction of the general expression to the product form is demonstrated (Equations 35–38).
The Parke-Taylor MHV formula (Equation 8): the known MHV result serves as a structural reference point, though it is not directly compared since it lives in a different helicity sector.

Generation budget / compute accounting. The paper measures computational complexity not in FLOPs but in combinatorial complexity: the number of terms in the general recursion (21) grows exponentially with n (the 6-point expression already contains 32 terms), while the simplified formula (39) requires only n−2 factor evaluations — a reduction from O(2ⁿ) to O(n). This complexity reduction is explicitly highlighted in the discussion: "Clearly a more concise formula is needed!" follows the 32-term 6-point expression (Equation 32). In the proof of the collapse in ℛ₁, the "budget" is the number of non-vanishing ordered partitions, which drops from exponentially many to exactly one (the all-singleton partition).

Cross-validation / statistical protocol. There is no statistical protocol in the conventional sense. The verification strategy is mathematical proof by induction combined with explicit low-n checks:

Low-n explicit verification: The formulas for n = 3, 4, 5, 6 are computed directly from the recursion (21) and shown to match the product formula (39) after specialization to ℛ₁ (Equations 35–38). This provides concrete evidence for the pattern.
All-n proof: The paper provides a three-part proof in Section II.3: (1) vanishing of V for all plus-helicity multi-gluon blocks in ℛ₁ (Section II.3.1, weighted-variance argument), (2) collapse of the preamplitude recursion to zero for non-singleton blocks and consequent reduction of the amplitude to V̄ (Section II.3.2), (3) factorization of V̄ into the product form (Section II.3.3). This is a formal mathematical proof, not an empirical verification — the result holds for all n by the structure of the recursion, not by statistical sampling.
Consistency checkpointing: The five identities (24–28) are stated to have been verified explicitly by calculation, though these verifications are deferred to unspecified future work.

The only "validation fold" is the separate analysis of the general half-collinear regime (Section I) versus region ℛ₁ (Section II), where the latter is shown to be a consistent subcase of the former.

Main Quantitative Results

The General Single-Minus Stripped Amplitude (Section I)

The paper's first major result is the recursion relation (Equations 19, 21) that determines the stripped amplitude A₁⋯ₙ for any n in the full half-collinear regime. This is an existence result — it establishes that single-minus amplitudes are nonzero and provides a systematic computational procedure — rather than a closed-form formula. The explicit low-n expressions (Equations 29–32) are presented both as verifications of the recursion and as demonstrations that the complexity grows combinatorially.

n = 3: Equation (29) gives A₁₂₃ = sg₁₂, a single sign function. At 3 points, the stripped amplitude is simply the sign of the bracket [λ̃₁ λ̃₂].

n = 4: Equation (30) gives A₁₂₃₄ = ½(sg₂₃ sg₄₁ + sg₁₂ sg₃₄). The expression involves two terms (rather than the ~4! = 24 Feynman diagrams), each being a product of two sign functions, with an overall factor of ½.

n = 5: Equation (31) gives A₁₂₃₄₅ with 8 terms. Each term is a product of three or four sign functions (involving both two-particle and three-particle brackets such as sg₂,₃₄), with an overall factor of ¼.

n = 6: Equation (32) gives A₁₂₃₄₅₆ with 32 terms, each a product of up to six sign functions, with an overall factor of ⅛.

The term counts (1, 2, 8, 32) follow a pattern of 2ⁿ⁻³ for n ≥ 4, indicating that the general recursion produces O(2ⁿ) terms — exponentially growing, but still far fewer than the O(n!) of the naive Feynman diagram sum. The stripped amplitude in the general half-collinear regime is piecewise-constant and takes integer values (after the overall normalization factor), jumping across codimension-one walls where any bracket changes sign.

Key structural observation: The general expression contains sign functions of both two-particle brackets (sg_ij_) and multi-particle brackets (sg_i,jk_, sg_ij,kl_, etc.). The multi-particle brackets encode the dependence on the ω frequencies — unlike in ℛ₁ where many of these simplify — making the general amplitude depend on both the z̃ and ω variables. The chamber structure of the full half-collinear regime is therefore richer and more complex than the factorized structure in ℛ₁.

Simplification in Region ℛ₁ (Section II.2)

The paper's second major result is the dramatic simplification of the stripped amplitude in the kinematic subregion ℛ₁ (defined by Equation 33: existence of a frame with ω₁ < 0, ωₐ > 0 for a ≥ 2). The explicit low-n expressions (35–38) in ℛ₁ are:

n = 3 (Equation 35): A₁₂₃|ℛ₁ = ½(sg₁₂ + sg₂₃). A single sum of two adjacent sign functions.

n = 4 (Equation 36): A₁₂₃₄|ℛ₁ = ¼(sg₁₂ + sg₂₃)(sg₃₄ + sg₄₁). A product of two factors, each a sum of adjacent sign functions. Note that sg₄₁ appears rather than sg₁,₂₃ — this is because at n = 4, the partial sum 1, 2, 3 is the total minus particle 4, and by momentum conservation and the ℛ₁ sign rules, sg₁,₂₃ simplifies to sg₄₁.

n = 5 (Equation 37): A₁₂₃₄₅|ℛ₁ = ⅛(sg₁₂ + sg₂₃)(sg₃₄ + sg₁,₂₃)(sg₄₅ + sg₅₁). Three factors: the first involves adjacent signs, the second involves a triplet bracket sg₁,₂₃, the third returns to adjacent signs.

n = 6 (Equation 38): A₁₂₃₄₅₆|ℛ₁ = ¹⁄₁₆(sg₁₂ + sg₂₃)(sg₃₄ + sg₁,₂₃)(sg₄₅ + sg₁,₂₃₄)(sg₅₆ + sg₆₁). Four factors, with the middle factors involving progressively longer partial-sum brackets sg₁,₂₃ and sg₁,₂₃₄.

The pattern observed (leading to the conjecture): For n ≥ 4, the expression in ℛ₁ is a product of n−2 factors. The first factor is ½(sg₁₂ + sg₂₃), the last factor is ½(sgₙ₋₁,ₙ + sgₙ₁), and the intermediate factors for m = 3, 4, …, n−2 are ½(sgₘ,ₘ₊₁ + sg₁,₂⋯ₘ). At n = 3, there is a single factor with a slightly different structure (sg₁₂ + sg₂₃ rather than sg₁₂ + sg₁₂, which would reduce to 2 sg₁₂). This pattern is what motivated the general conjecture (39).

Quantitative difference from general case: At n = 6, the general expression (32) contains 32 terms with up to 6 sign functions each; the ℛ₁ expression (38) contains 4 factors, each a sum of 2 signs — a reduction from 32 terms to effectively 2⁴ = 16 sign configurations (each factor independently ±1 or 0), but encoded as a product rather than a sum. At general n, the reduction is from O(2ⁿ) terms to n−2 multiplicative factors.

The General Formula and Its Proof (Section II.3)

The paper's central quantitative result is the all-n product formula (Equation 39):

$A_{1\cdots n}|_{\mathcal{R}_1} = \frac{1}{2^{n-2}} \prod_{m=2}^{n-1} \left(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m}\right)$

This formula is validated by a three-part mathematical proof (Section II.3), not by numerical experiment:

Part 1: Vanishing of V (Equations 40, 42, 47). The proof establishes that for any consecutive ordered subset S ⊆ {2, …, n} with |S| ≥ 2 — and specifically for the full list (2, …, n) — the vertex function V vanishes in ℛ₁:

$V_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n}|_{\mathcal{R}_1} = 0$

The proof uses a weighted-variance argument (Section II.3.1): for each cut position j between adjacent plus-helicity gluons, the left-partial sum λ̃₂⋯ⱼ and right-partial sum λ̃ⱼ₊₁⋯ₙ have positive total frequencies Ω_L, Ω_R > 0. The ratio entering the Θ-function in V is signed by z̃_R,L / z̃ⱼ₊₁,ⱼ. The weighted-variance identity guarantees existence of at least one cut j⋆ where z̃_R,L and z̃ⱼ₊₁,ⱼ share the same sign, making the Θ(−positive) factor vanish. Since V is a product over all cuts, the entire product vanishes.

Part 2: Collapse of the preamplitude recursion (Equations 48, 41, 51). From V = 0 for all multi-gluon plus-helicity blocks, the preamplitude recursion (19) yields Ā_S|ℛ₁ = 0 for all |S| ≥ 2 (with S ⊆ {2, …, n}), while singletons remain Ā_i = 1. Using cyclicity to place particle 1 in the last position (A₁⋯ₙ = A₂⋯ₙ₁), the amplitude recursion (21) collapses to the all-singleton partition of (2, …, n), giving A₁⋯ₙ|ℛ₁ = V̄_λ̃₂⋯λ̃ₙ|ℛ₁. The proof is a direct consequence of the partition sum reducing to a single term.

Part 3: Factorization of V̄ (Equations 52–54, leading to 39). The barred vertex is evaluated using the definition of Θ in terms of sign functions:

$\bar{V}_{\tilde{\lambda}_2 \cdots \tilde{\lambda}_n} = \prod_{m=2}^{n-1} \mathrm{sg}_{m,m+1} \, \Theta\!\left(\frac{[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_{m+1\cdots n}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}\right)$

Momentum conservation replaces λ̃ₘ₊₁⋯ₙ = −λ̃₁ − λ̃₂⋯ₘ, and the bracket algebra yields:

$\frac{[\tilde{\lambda}_{2\cdots m} \, \tilde{\lambda}_{m+1\cdots n}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]} = \frac{[\tilde{\lambda}_1 \, \tilde{\lambda}_{2\cdots m}]}{[\tilde{\lambda}_m \, \tilde{\lambda}_{m+1}]}$

Using Θ(x) = ½(1 + sg(x)), the sign of the ratio factorizes as sg₁,₂⋯ₘ · sgₘ,ₘ₊₁ (since the ω signs cancel in the ratio but not in the individual bracket signs, and the product structure of the bracket signs permits this factorization in ℛ₁). The sgₘ,ₘ₊₁ prefactor in V̄ multiplies through to give:

$\mathrm{sg}_{m,m+1} \cdot \frac{1}{2}(1 + \mathrm{sg}_{1,2\cdots m} \cdot \mathrm{sg}_{m,m+1}) = \frac{1}{2}(\mathrm{sg}_{m,m+1} + \mathrm{sg}_{1,2\cdots m})$

The product over m = 2, …, n−1 of these factors yields exactly Equation (39). The normalization factor 1/2ⁿ⁻² emerges from the product of n−2 factors of ½.

The structural result: Each factor is a signed projection operator taking values:

+1 when sgₘ,ₘ₊₁ = +1 and sg₁,₂⋯ₘ = +1 (both signs positive),
−1 when sgₘ,ₘ₊₁ = −1 and sg₁,₂⋯ₘ = −1 (both signs negative),
0 when sgₘ,ₘ₊₁ ≠ sg₁,₂⋯ₘ (signs disagree).

Since the amplitude is a product of these factors, A₁⋯ₙ|ℛ₁ takes values in {+1, −1, 0} — it is a piecewise-constant integer on the space of z̃ coordinates (and partially on the ω's through the multi-particle brackets sg₁,₂⋯ₘ). The amplitude is nonzero only when a global sign-consistency condition holds across all n−2 adjacent-plus-partial-sum pairs.

Ablation Studies and Robustness Checks

Since this is a mathematical derivation paper rather than an empirical study, there are no conventional ablation studies. However, the paper does perform several structural checks that serve an analogous function — verifying that the derived formulas are consistent with known identities and that the simplifications are not artifacts of the computational method.

Consistency with Weinberg's soft theorem (Section I.3, item 5): The most nontrivial check is verifying that the product formula (39) satisfies the soft theorem (28) when the frequency of the last plus-helicity gluon is taken to zero. The paper states this verification has been done explicitly, and notes that "in this form, it is direct to check the soft theorem in the last label, and with some more work, in any other label but 1." The soft theorem check serves as an infrared consistency condition — if the amplitude failed this test, it would indicate a fundamental error in the derivation. The factorized structure of (39) makes the ωₙ → 0 limit of the last factor ½(sgₙ₋₁,ₙ + sg₁,₂⋯ₙ₋₁) reproduce the required soft factor, while the remaining n−3 factors match the (n−1)-point amplitude.

Consistency with cyclicity (Equation 24): The product formula (39) is manifestly not cyclically invariant — it singles out particle 1 through the factors sg₁,₂⋯ₘ that involve partial sums including particle 1. However, the paper argues that cyclicity is recovered by extending the formula to other regions ℛₖ (where only particle k has ωₖ < 0) via cyclic permutation. The full amplitude across the union of all ℛₖ regions is cyclically invariant. This check is structural: the formula in ℛ₁, cyclically permuted to ℛ₂, ℛ₃, …, ℛₙ, covers the space of single-minus configurations in the half-collinear regime.

Consistency with reflection symmetry, U(1) decoupling, and Kleiss–Kuijf relations (Section I.3, items 2–4): The paper states that "it is far from evident that all these properties are obeyed by the solution of the recursion relation (21). Nonetheless, we have verified by explicit calculation that they do indeed hold." These verifications serve as algebraic consistency checks — the color-ordering identities are sensitive to the sign structure of the amplitude, and any error in the sign assignments (from the sg functions and Θ-function chambers) would manifest as violations. The fact that all five identities hold is strong evidence for the correctness of the general recursion (21) and, by extension, its specialization (39).

Verification of the pattern through n = 6 (Equations 29–32 and 35–38): The explicit computation of the 3-point through 6-point amplitudes serves as a low-n ablation — checking that the all-n proof's inductive structure is consistent with concrete cases. For n = 3 through 6, the general expression is first computed from the full recursion (21), then specialized to ℛ₁ using the sign-function simplifications (34) and momentum conservation identities. The match between these direct computations and the product pattern confirms that the collapse of the recursion (the vanishing of Ā for non-singletons) is correctly implemented and that no subtle sign errors survive the specialization.

The master identity verification (Appendix A): The entire derivation depends on the master identity (A) and its specialization (66). The paper proves this identity via Fourier transform to the time domain (Equations 59–62), showing that the iε prescription in the propagator sum produces exactly the step-function and sign-function structure of V and V̄. This proof is self-contained and does not depend on the specific application to single-minus amplitudes — it is a general result in time-ordered perturbation theory. Its correctness is verified by the algebraic consistency of the construction: the identity (55) for n = 2 is checked directly (1/b + iε = PV 1/b − i/₂ δ(b)), and the generalization to n = 3 and beyond follows by induction.

Weighted-variance proof of V = 0 in ℛ₁ (Section II.3.1): This argument is the linchpin of the entire simplification. The proof must handle the case where some ωₐ approach zero or where z̃-differences become degenerate. The paper's argument — that a weighted average of a set of numbers must lie between the minimum and maximum, guaranteeing at least one cut where the left and right averages straddle the cut point — is robust to degeneracies (if all z̃ are equal, all brackets vanish and the amplitude is on a wall where the Θ-function is ambiguous, but this is measure-zero). The paper does not provide a formal measure-theoretic treatment of the boundary walls, but the piecewise-constant structure means the amplitude is well-defined in each open chamber and the walls are the loci of jumps.

The ReST_EM experiment equivalent: There is no failed training run in this paper, but there is a conceptual negative result that serves a similar function: the unsimplified general formula (32) has 32 terms for n = 6, demonstrating that the general half-collinear amplitude, while nonzero, is combinatorially complex. This "failure to simplify" in the general case highlights what makes ℛ₁ special — the frequency-sign condition (33) is not a minor restriction but a qualitative structural filter that eliminates almost all contributions. The paper does not explore other subregions or attempt to find similarly simple formulas for different frequency-sign patterns, which could be viewed as a missing ablation (what happens in ℛ₂? ℛ₁₂? mixed-sign configurations?).

Critical Assessment

Claim from the abstract: "Single-minus tree-level n-gluon scattering amplitudes are reconsidered. Often presumed to vanish, they are shown here to be nonvanishing for certain 'half-collinear' configurations existing in Klein space or for complexified momenta."

This claim is fully established by the paper. The explicit expressions for A₁₂₃ through A₁₂₃₄₅₆ (Equations 29–32) are manifestly nonzero for generic z̃ and ω values in the half-collinear regime. The recursion (21) provides a systematic construction for all n, and the induction proof in Appendix B shows that the BG recursion yields nonzero results for any n ≥ 3. The "certain configurations" are precisely specified (Equation 13: all ⟨ij⟩ = 0), and the paper explains clearly why this locus is only accessible in Klein signature or complexified momenta, not in Minkowski space.

A potential criticism: the paper does not discuss what happens when one attempts to analytically continue these amplitudes back to Minkowski signature. In (3,1) signature, ⟨ij⟩ = 0 forces [ij] = 0 as well (since λ̃ = ±λ*), making all momenta collinear and the amplitude identically zero (or supported on a set of measure zero in the phase space). The claim that single-minus amplitudes are "nonzero" is therefore signature-dependent — they are nonzero in the sense that they exist as mathematical objects in the analytically continued theory (Klein signature or complex kinematics), but they do not contribute to physical scattering probabilities in Minkowski space. The paper is upfront about this ("unlike in Minkowski space"), but the distinction between "mathematically exists" and "physically observable in our universe" is worth emphasizing. This does not weaken the paper's contribution — amplitudes in Klein signature are standard objects in the modern amplitudes program (celestial holography, BCFW analytic continuation, amplituhedron) — but it does mean the result is theoretical rather than phenomenological.

Claim: "We derive a piecewise-constant closed-form expression for the decay of a single minus-helicity gluon into n−1 plus-helicity gluons as a function of their momenta."

This claim is fully established for region ℛ₁ (Equation 39) but not for the general half-collinear regime. The general formula (21) is a recursion, not a closed form — it expresses A₁⋯ₙ as a sum over ordered partitions, which for general n involves exponentially many terms. The explicit low-n expressions (29–32) show that the general half-collinear amplitude is piecewise-constant, but they do not provide a compact closed form. The paper acknowledges this limitation in the introduction: "it is entirely possible that a yet simpler expression may be obtained with a clever choice of analytic continuation, variables or basis, even outside the single-minus decay channel." The closed form is proven only in ℛ₁, where the frequency-sign condition (33) holds.

The phrase "decay of a single minus-helicity gluon into n−1 plus-helicity gluons" is precisely the physical interpretation of region ℛ₁, so the claim is accurate as stated. However, a reader who misses the ℛ₁ restriction might mistakenly think the paper provides a general closed form. The abstract and introduction are careful to specify "for this special kinematic region" and "a special kinematic channel denoted ℛ₁," but the distinction is crucial.

Claim: "This formula nontrivially satisfies multiple consistency conditions including Weinberg's soft theorem."

This claim is stated but not demonstrated in detail. The paper asserts that all five consistency conditions (24–28) have been verified by explicit calculation, but the verifications themselves are not presented — only the statement that they hold ("we have verified by explicit calculation that they do indeed hold. Details of this calculation will appear elsewhere"). For the soft theorem, the paper says "in this form, it is direct to check" for the soft limit in particle n, and "with some more work" for other particles. The fact that these verifications are deferred is a limitation of the current paper — the reader must take the authors' word that the checks succeed, or reproduce them independently.

That said, there is strong structural evidence that the identities hold. The cyclicity of the product formula, while not manifest, follows from the proof that A₁⋯ₙ|ℛ₁ = V̄_λ̃₂⋯λ̃ₙ — a quantity that depends on particles 2 through n in an ordered but cyclically permutable way (since V̄ is built from adjacent brackets). The soft theorem emerges naturally from the factorization of V̄: the last factor ½(sgₙ₋₁,ₙ + sg₁,₂⋯ₙ₋₁) in the product (39) contains the ωₙ dependence only in sgₙ₋₁,ₙ (which approaches a finite sign as ωₙ → 0) and sg₁,₂⋯ₙ₋₁ (which is ωₙ-independent), so the limit reproduces the soft factor structure. These are nontrivial consistency checks that would be very unlikely to hold by coincidence if the formula were incorrect.

What the experiments do NOT test:

Other frequency-sign configurations: The paper only analyzes ℛ₁ (particle 1 negative, all others positive) and invokes cyclicity to extend to ℛₖ (any one particle negative). Configurations with mixed signs (some particles negative, some positive, without a single distinguished incoming particle) are not studied. The general recursion (21) applies to all configurations in the half-collinear regime, but the dramatic simplifications may not extend.
External particles beyond gluons: The introduction mentions that "the construction generalizes directly from gluon to graviton amplitudes and has a simple supersymmetrization," but no graviton or supersymmetric results are presented. These are promised as future work.
Loop-level single-minus amplitudes: The paper is strictly tree-level. Single-minus amplitudes at loop level, or single-minus contributions to loop-level MHV amplitudes via unitarity cuts, are not discussed.
Alternative kinematic parameterizations: The paper works in a specific frame (2) with λ = (1, z). The half-collinear regime is frame-invariant, but the specific form of the stripped amplitude depends on this choice. The paper argues that A₁⋯ₙ is helicity-neutral and thus frame-independent (up to little-group rescalings), but no explicit check of frame-independence is presented.
Comparison to Feynman diagram enumeration: While the paper's reduction from O(2ⁿ) to O(n) terms in ℛ₁ is dramatic, it does not explicitly enumerate the Feynman diagram count for comparison. The statement that "naively, the n-gluon scattering amplitude involves order n! terms" (from the introduction) is a general claim about the full Yang–Mills amplitude, not specifically about the single-minus sector. The actual number of nonzero Feynman diagrams contributing to the single-minus amplitude in the half-collinear regime is not computed.

Missing experiments that would strengthen the paper:

Explicit verification of the soft theorem for at least one non-trivial case (n = 5 or 6) showing the ω limit of (39) produces the (n−1)-point amplitude times the soft factor, with the intermediate algebraic steps displayed. This would make the "proved" claim more persuasive to readers who cannot immediately verify the argument.
Explicit enumeration of all ordered partitions contributing to the general formula for n = 5 or 6, showing which partitions vanish in ℛ₁ and which survive, to illustrate the mechanism of the collapse (rather than just stating the final result).
A worked example showing the amplitude values for a specific numeric set of z̃ and ω variables in ℛ₁ (e.g., z̃ = (0, 1, 3, 5), ω = (−2, 1, 1, 1) for n = 4), evaluating each factor in (39) and showing the result is +1, −1, or 0. This would ground the abstract formula in a concrete, verifiable computation.
Analysis of wall-crossing behavior: the paper states the amplitude is piecewise-constant and jumps across codimension-one walls, but does not characterize the jump (e.g., is it always ±1? ±2?), the wall equations in z̃-space, or the geometry of the chamber decomposition for general n.

Summary assessment. The paper's main claims are well-supported by the mathematical derivation. The existence of nonzero single-minus amplitudes in the half-collinear regime is conclusively established by the explicit expressions (29–32) and the general recursion (21). The closed-form product formula (39) in ℛ₁ is rigorously proved in Section II.3, with each step of the proof — vanishing of V, collapse of Ā, factorization of V̄ — clearly delineated. The consistency checks (soft theorem, cyclicity, etc.) are stated to hold, and structural arguments support their validity, though detailed verifications are deferred. The primary limitation is the scope: the closed form is proven only for the specific decay kinematics of ℛ₁, the general half-collinear amplitude remains combinatorially complex, and extensions to other theories and loop level are promised but not delivered. The paper succeeds in its primary goal — correcting a forty-year oversight in the amplitudes literature — and establishes a foundation for a broader research program on "more-than-MHV" amplitudes in split signature.

6. Limitations and Trade-offs

Kinematic Scope: Closed Form Exists Only in Region ℛ₁, Not the General Half-Collinear Regime

The constraint. The paper's headline result — the elegant product formula (39) — is proven exclusively in region ℛ₁, defined by the existence of an SO(2,2) frame where particle 1 has negative frequency (ω₁ < 0) and all other particles have positive frequency (ωₐ > 0 for a ≥ 2). The general stripped amplitude in the full half-collinear regime (where all ⟨ij⟩ = 0 but no frequency-sign condition is imposed) is given by the recursion (21), which involves a sum over ordered partitions and produces exponentially many terms. The authors are explicit about this scope:

"This section presents the next main result of this paper: a simple formula for the n-point single-minus amplitudes (21) with partially restricted kinematics within the half-collinear regime." (Section II, opening)

And:

"We note that, while our expression is a dramatic simplification of the direct Feynman-diagram expression, it is entirely possible that a yet simpler expression may be obtained with a clever choice of analytic continuation, variables or basis, even outside the single-minus decay channel." (Introduction)

The consequence. The closed-form result does not apply to generic single-minus configurations in the half-collinear regime. If one has a configuration where multiple particles have negative frequency, or where the frequency signs are mixed in a pattern not equivalent (by cyclic permutation) to a single distinguished negative-frequency particle, the product formula (39) cannot be used. The general amplitude in such configurations is given by the unsimplified recursion (21), whose computational complexity grows exponentially with n — the 6-point expression already contains 32 terms, and the authors' own remark after displaying it ("Clearly a more concise formula is needed!") indicates that the general case is not practically computable by this method for large n. The paper's "solve" for the general single-minus problem is therefore only partial: it identifies where the amplitudes live (the half-collinear regime), provides a constructive recursion to compute them, but achieves a compact closed form only under the additional frequency-sign restriction.

What evidence exists. The explicit comparison between the general expressions (29–32) and their ℛ₁ simplifications (35–38) for n = 3 through 6 makes the complexity gap concrete. At n = 6, the general expression (32) is a sum of 32 terms, while the ℛ₁ form (38) is a product of 4 factors. The proof of collapse in Section II.3 relies crucially on the vanishing of V for consecutive plus-helicity blocks (Equation 47), which is established via the weighted-variance argument that uses ωₐ > 0 for all a ≥ 2. No analogous collapse is demonstrated or claimed for other frequency-sign configurations. The paper does not characterize the size of region ℛ₁ relative to the full half-collinear regime — we do not know whether ℛ₁ and its cyclic permutations cover most configurations, a substantial fraction, or a small subset.

Mitigation status. The paper partially addresses this through cyclicity: "we can trivially construct a cyclically invariant answer by using cyclicity to extend (39) to other regions ℛₖ where only particle k has ωₖ < 0" (Section II.3, after Equation 39). This extends the closed form to n separate regions, each with a single distinguished negative-frequency particle. However, configurations where multiple particles share negative frequency, or where the frequency signs do not isolate a single particle in any frame, are not covered. The authors identify this as future work: "Further details of our analysis, including a longer general formula for the single-minus amplitude outside of ℛ₁, will appear elsewhere." This is a genuine open problem rather than a resolved limitation, and a practitioner seeking to evaluate single-minus amplitudes for arbitrary half-collinear kinematics cannot rely solely on the results presented in this paper.

Signature Dependence: Nonzero Only in Klein Signature, Not in Physical Minkowski Space

The constraint. The entire construction — the half-collinear regime, the nonzero amplitudes, the product formula — is defined in (2,2) Klein signature, where left- and right-handed spinors λ and λ̃ are independent real variables. In Minkowski signature (3,1), the reality condition λ̃ = ±λ* forces ⟨ij⟩ = 0 to imply [ij] = 0 as well, which collapses all momenta to be proportional. The authors state this explicitly:

"In (2,2) signature, this is compatible with nonzero [ij], unlike in Minkowski space." (Section I.1)

And in the abstract:

"Often presumed to vanish, they are shown here to be nonvanishing for certain 'half-collinear' configurations existing in Klein space or for complexified momenta."

The consequence. These single-minus amplitudes do not produce nonzero scattering probabilities for physical gluons in four-dimensional Minkowski spacetime. A practitioner computing cross sections at the LHC, or evaluating unitarity cuts for loop amplitudes in physical kinematics, will never encounter a nonzero single-minus tree amplitude — the support of the amplitude in Minkowski signature is measure-zero in the phase space. The result is therefore a mathematical discovery about the analytic structure of Yang–Mills theory rather than a phenomenological prediction. This does not diminish its theoretical importance, but it fundamentally limits the scope of practical applications. The amplitudes discovered here live in the domain of analytically continued momenta — relevant for celestial holography, BCFW analytic continuation, and formal studies of S-matrix theory — but they do not alter the Feynman rules or the Parke-Taylor formula for Minkowski-space calculations.

What evidence exists. The paper is transparent about the signature requirement. Section I.1 explains why the half-collinear regime requires signature (2,2) or complex momenta. The kinematic parameterization (Equation 2) with zᵢ and z̃ᵢ as independent real variables is a Klein-signature construction. No Minkowski-space continuation is attempted or discussed. The authors do not quantify what subset of the Klein-signature amplitudes survive analytic continuation to some Minkowski boundary, or whether the amplitudes can be interpreted as residues that contribute to physical observables indirectly (e.g., through generalized unitarity or as building blocks in BCFW recursion). The fact that the ordinary 3-point single-minus amplitude has δ-function support in Minkowski space (on collinear configurations) but vanishes for n ≥ 4 in generic kinematics is the established result that the paper reframes, not overrides.

Mitigation status. The authors do not attempt to mitigate this limitation, nor do they frame it as something requiring mitigation — it is the nature of the discovery. They do suggest that the construction generalizes to other theories ("from gluon to graviton amplitudes"), but these generalizations would also live in Klein signature or complex kinematics. A reader looking for phenomenological consequences will find none in this paper; a reader interested in the formal structure of quantum field theory — particularly in split signature, where celestial holography and the Lw₁₊∞ symmetry algebras operate — will find the signature choice natural and expected. The limitation is more about audience and applicability than about a flaw in the work.

Deferred Verification of Consistency Conditions

The constraint. The paper states that the stripped amplitude — in both its general recursive form (21) and its simplified ℛ₁ form (39) — satisfies five nontrivial consistency conditions: cyclicity, reflection symmetry, U(1) decoupling, Kleiss–Kuijf relations, and Weinberg's soft theorem (Section I.3). However, the detailed verification of these conditions is not presented in the paper. The authors write:

"It is far from evident that all these properties are obeyed by the solution of the recursion relation (21). Nonetheless, we have verified by explicit calculation that they do indeed hold. Details of this calculation will appear elsewhere."

And regarding the soft theorem specifically:

"in this form, it is direct to check the soft theorem in the last label, and with some more work, in any other label but 1."

The consequence. The consistency checks serve as the paper's primary validation that the general recursion (21) and the product formula (39) are correct — there is no independent derivation (e.g., from BCFW, from a twistor string, from a worldsheet model) that could serve as a cross-check. A reader who cannot immediately reproduce these verifications must take the authors' word that they succeed. This is particularly concerning for the soft theorem (28), which is the most physically constraining condition: it relates the n-point amplitude to the (n−1)-point amplitude in a soft limit, and any error in the sign-function assignments (from the sg and Θ-function chamber structure) would manifest as a mismatch. The soft theorem also involves the iε prescription in a delicate way, since the soft limit ωₙ → 0 can cross chamber boundaries where sg functions jump — the statement that the relation holds "in this form" may require careful treatment of the limiting procedure.

What evidence exists. The paper provides structural arguments for why the conditions should hold. For cyclicity, the proof that A₁⋯ₙ|ℛ₁ = V̄_λ̃₂⋯λ̃ₙ (Equation 51) shows that the amplitude depends on particles 2 through n in a cyclically structured way — the V̄ vertex is built from adjacent brackets of the plus-helicity block, which cyclically permutes when particle 1 is moved. For the soft theorem, the factorization of the product formula (39) into n−2 factors means the soft particle n appears only in the last factor ½(sgₙ₋₁,ₙ + sg₁,₂⋯ₙ₋₁) — since sg₁,₂⋯ₙ₋₁ is independent of ωₙ, the limit is controlled by sgₙ₋₁,ₙ, which produces the claimed soft factor structure. These structural arguments are plausible but are not substitutes for explicit verification. The paper does not provide the actual calculations, the number of cases checked, or the specific identities used.

Mitigation status. The paper explicitly defers the detailed verification to future work. This is a reasonable choice for a discovery paper that is already dense with new results — the primary contribution is the existence and structure of the amplitudes, not the exhaustive verification of all their properties. However, the absence of these checks means the paper's claims about consistency rest on the authors' authority and the structural plausibility arguments, not on demonstrated fact. A skeptical reader who attempts to independently verify (say) the U(1) decoupling identity for n = 6 using the 32-term expression (32) and the 5-point expression (31) would face a substantial combinatorial calculation with no guidance from the paper. This is not a flaw in the logical structure of the derivation, but it is a limitation in the completeness of the presented evidence for the claims made.

No Exploration of Wall-Crossing Structure or Chamber Geometry

The constraint. The stripped amplitude A₁⋯ₙ is described as "piecewise-constant" and the product formula (39) is said to "jump across codimension-one walls where the relevant brackets change sign." However, the paper provides no systematic analysis of the chamber structure of the half-collinear regime, either in general or in ℛ₁. There is no characterization of:

How many chambers exist as a function of n.
What walls separate the chambers (the algebraic equations in z̃-space).
How the amplitude value changes when crossing a wall (the jump magnitude — is it always ±1? ±2?).
The geometry of the chamber decomposition (e.g., is it a central hyperplane arrangement? Are all chambers nonempty?).

The consequence. A practitioner who wants to evaluate the amplitude for a specific numerical kinematic configuration faces a non-trivial problem: the product formula (39) involves sg functions of multi-particle brackets (sg₁,₂⋯ₘ) whose signs depend on z̃ and ω variables in a non-local way. To determine whether the amplitude is +1, −1, or 0, one must evaluate n−2 sign functions and check the product. If any factor is zero (because sgₘ,ₘ₊₁ ≠ sg₁,₂⋯ₘ), the amplitude vanishes. But understanding which configurations yield nonzero amplitudes, and how the amplitude organizes the space of kinematic data, is left entirely to the reader. The paper does not provide even a simple example of a configuration in ℛ₁ with explicit numeric values for z̃ and ω, showing the evaluation of each factor and the resulting amplitude. This makes the result feel more abstract than it needs to be, and limits its immediate usability for anyone wanting to compute these amplitudes in practice.

What evidence exists. The explicit low-n expressions (29–32) and (35–38) provide some insight into the chamber structure for small n. For n = 4 in ℛ₁, the amplitude is ¼(sg₁₂ + sg₂₃)(sg₃₄ + sg₄₁) — each factor is a sum of two adjacent signs, and the amplitude is nonzero only when sg₁₂ = sg₂₃ and sg₃₄ = sg₄₁. This means the ordering of z̃₁, z̃₂, z̃₃ and z̃₃, z̃₄, z̃₁ must each be "consistent" in a specific sense. The general pattern — the amplitude is nonzero when there is a consistent hierarchical ordering of partial sums relative to particle 1 — is implicitly described in the main text's discussion of the product formula: "each factor is ½(±1 ± 1) ∈ {−1, 0, 1}, so A₁⋯ₙ|ℛ₁ is piecewise-constant and jumps across codimension-one walls where the relevant brackets change sign. The product form simply makes those walls explicit." But the walls are not explicitly listed, the chamber count is not computed, and the geometric structure is not analyzed.

Mitigation status. The paper does not address this gap, nor does it promise future work on chamber geometry. The authors seem to view the identification of the walls ("the product form simply makes those walls explicit") as sufficient, but "explicit" here means "implicitly defined by the vanishing of the factors" — not "enumerated and classified." This is a limitation in the expository completeness of the result rather than its correctness. The discovery that the amplitude factorizes into signed projection operators is the key contribution; the full exploration of the resulting chamber structure is a natural next step that is left to the reader or to future work. For a practitioner who wants to actually compute these amplitudes for specific configurations, the missing chamber analysis makes the product formula feel less "closed-form" than it initially appears, since evaluating the sign of sg₁,₂⋯ₘ for arbitrary m and arbitrary ω values requires summing weighted z̃-coordinates and checking signs — a computation that is linear in n per factor, making the full amplitude evaluation O(n²) rather than truly O(n). This is still exponentially better than the general case, but the hidden quadratic factor is worth noting.

Unaccounted Complexity in Evaluating Multi-Particle Sign Functions

The constraint. The product formula (39) appears deceptively simple: n−2 factors, each a sum of two sign functions. However, each factor for m ≥ 3 involves sg₁,₂⋯ₘ — the sign of the bracket between particle 1 and the partial sum λ̃₂⋯ₘ = ∑ₐ₌₂ᵐ λ̃ₐ. Evaluating this sign requires:

Summing m−1 spinors λ̃ₐ = ωₐ(1, z̃ₐ) to obtain the partial sum spinor λ̃₂⋯ₘ = (Ω, Ωz̃̄) where Ω = ∑ₐ≥₂ ωₐ and z̃̄ is the ω-weighted average of the z̃ₐ.
Computing the bracket [λ̃₁ λ̃₂⋯ₘ] = ω₁Ω(z̃̄ − z̃₁).
Taking the sign of this bracket, which depends on the weighted average z̃̄.

The paper acknowledges this explicitly:

"Note however that the ωk* cannot be eliminated from expressions such as sg{2,34}." (Section II.1, after Equation 34)

The consequence. The product formula does not eliminate the ω dependence from the amplitude — the frequencies enter through the partial-sum brackets sg₁,₂⋯ₘ in every factor for m ≥ 3. This has two practical implications:

Computational cost. Evaluating all n−2 factors requires computing n−2 weighted averages of increasingly large subsets of the plus-helicity gluons. Naively, this is O(n²) operations (sum m particles for m = 3, …, n−1), though it can be reduced to O(n) with a running partial sum. More fundamentally, the amplitude is not a function purely of the z̃ coordinates — it depends on the ω frequencies as weights in the partial sums. This means the "chamber walls" in z̃-space are not fixed hyperplanes but move with the ω values: two configurations with identical z̃ ordering can have opposite sg₁,₂⋯ₘ if the ω distributions shift the weighted average across a boundary.

Opacity of the geometric structure. The partial-sum signs sg₁,₂⋯ₘ are not elementary — they encode the position of a weighted average relative to z̃₁. Understanding whether the amplitude is +1, −1, or 0 for a given configuration requires checking whether sgₘ,ₘ₊₁ (= sg(z̃ₘ − z̃ₘ₊₁), independent of ω) agrees with sg₁,₂⋯ₘ (= sg(z̃̄ − z̃₁), which depends on all ωₐ for a ≤ m). This is a genuine coupling between the z̃ geometry and the ω distribution that makes the chamber structure substantially more complex than it would be if all signs were pure z̃-differences. The claim that the product formula "makes those walls explicit" is true only in the sense that each factor identifies which bracket signs must agree; the actual location of the wall in z̃-space depends on the ω values in a nontrivial way.

What evidence exists. The explicit low-n expressions show this dependence clearly. In Equation (37) for n = 5, the factor (sg₃₄ + sg₁,₂₃) involves sg₁,₂₃, which is the sign of [λ̃₁ λ̃₂ + λ̃₃] — this depends on ω₂ and ω₃ as weights in the sum. The contrast with the adjacent signs (sgₘ,ₘ₊₁ = sg(z̃ₘ − z̃ₘ₊₁)) is evident. The paper's discussion in Section II.1 notes that "sg₁ⱼ = sg(z̃ⱼ₁)" (Equation 34) simplifies the two-particle brackets involving particle 1 — the negative ω₁ flips the sign relative to z̃₁ⱼ — but this simplification does not extend to multi-particle brackets, which is why sg₁,₂⋯ₘ cannot be reduced to a simple z̃-difference.

Mitigation status. The paper is transparent that the ω dependence persists ("the ωk* cannot be eliminated from expressions such as sg{2,34}"). It does not attempt to further simplify the partial-sum signs or to characterize the ω-dependence of the chamber walls. This is not a flaw — it is an intrinsic feature of the physics — but it means the product formula is less "solved" than it initially appears. The amplitude is expressed as a product of sign functions, but those sign functions are themselves nontrivial functions of the kinematic data. A fully explicit description of the amplitude's support in terms of elementary inequalities on the z̃ᵢ and ωᵢ is not provided and may not be simple. The paper's claim of a "closed-form expression" is correct — the formula (39) is a closed-form expression — but the inputs to that formula involve weighted sums whose sign structure is not further analyzed.

7. Implications and Future Directions

How This Work Changes the Landscape

This paper does not refine an existing technique or extend a known result — it corrects a forty-year error in the foundational taxonomy of Yang–Mills perturbation theory. The claim that single-minus gluon tree amplitudes vanish has been treated as settled fact since Parke and Taylor's 1986 MHV formula, embedded in every textbook treatment of spinor-helicity methods, and encoded in the very terminology "maximally helicity violating." The paper demonstrates that this claim is false: single-minus amplitudes are nonzero on a precisely characterized half-collinear kinematic locus in Klein signature, and in the natural decay region ℛ₁ they admit a closed-form expression of striking simplicity — a product of n−2 signed projection operators taking values in {+1, −1, 0}.

The magnitude of this shift is best understood by analogy: it is as if, after decades of computing graviton amplitudes using the KLT relations, someone discovered that a class of amplitudes previously assumed to be zero actually exists and has a simple structure. The "MHV" designation is now demonstrably a misnomer — the maximal number of plus helicities in an n-gluon tree amplitude is not n−2 but n−1, and the corresponding single-minus amplitudes are nonzero in the analytically continued theory. This forces a taxonomic revision: the tower of helicity sectors must be extended downward by one level, from (MHV, NMHV, N²MHV, …) to (single-minus, MHV, NMHV, N²MHV, …). Any formalism that treats MHV amplitudes as the primitive building blocks — BCFW recursion with MHV shifts, the CSW construction using MHV vertices, Grassmannian formulations, the amplituhedron — must now reckon with the existence of a more fundamental layer. Whether this layer decouples in Minkowski-space calculations (where the half-collinear support is measure-zero) or contributes indirectly through analytic continuation and unitarity cuts is an open question that this paper forces the community to address.

The paper also resolves a specific theoretical tension in self-dual Yang–Mills theory (SDYM). Prior work recognized that SDYM has a rich classical solution space — the ADHM instanton construction, Ward's twistor descriptions — but the Feynman tree expansion was thought to yield only trivial results (nonzero only at 3 points, and even then with restricted kinematics). This mismatch between classical richness and quantum triviality was a genuine puzzle. The single-minus amplitudes computed here via the SDYM Berends–Giele recursion provide the missing quantum objects. The proof that V vanishes for plus-helicity blocks in ℛ₁ (Section II.3.1) and the collapse of the recursion to V̄ (Section II.3.2) are not merely algebraic simplifications — they are statements about the causal structure of the self-dual sector, showing that the theory's tree-level content is richer than previously thought and naturally meets the complexity of the classical instanton sector. This does not complete the SDYM story, but it removes the contradiction that had been blocking progress.

Methodologically, the paper demonstrates a principle with implications beyond this specific calculation: appropriate kinematic restriction can collapse exponential diagrammatic complexity to polynomial or linear form. The general Berends–Giele recursion (Equation 21) sums over O(2ⁿ) ordered partitions. In region ℛ₁ — defined by the physically natural condition that there exists a frame where one particle is incoming and all others are outgoing — every non-singleton preamplitude vanishes, and the amplitude reduces to a single V̄ vertex evaluated on n−1 particles. The resulting formula (39) requires O(n) factor evaluations. This is not a perturbative approximation or a large-n limit; it is an exact equality holding for all n in that kinematic regime. The mechanism — frequency-sign conditions forcing the vanishing of causality-violating vertex functions — is likely generalizable. The paper's explicit note that "the construction generalizes directly from gluon to graviton amplitudes and has a simple supersymmetrization" suggests that the half-collinear regime harbors similarly simple structures in other massless theories, waiting to be uncovered by applying the same frequency-sign analysis to their Berends–Giele recursions.

The paper also elevates the iε prescription from a formal regularization detail to a first-class dynamical object. In standard Feynman diagram calculations, the iε lives in propagator denominators and is handled by contour deformation; it rarely appears in the final answer for tree amplitudes, which are rational functions of kinematic invariants. Here, precisely because the amplitude is supported on the locus where propagator denominators vanish (the half-collinear regime), the iε becomes the dominant feature — it determines, through the Θ-function structure of V and V̄, which combinations of momenta satisfy time-ordering constraints and thus which chambers support the amplitude. The master identity (Appendix A) that converts iε-prescribed propagator sums into δ-function constraints and sign-function combinatorics is general and may have applications well beyond single-minus amplitudes. The paper reframes the iε as a chamber-decomposition tool for amplitudes with distributional support — a perspective that could be fruitfully applied to other collinear or soft limits where the naive rational-function description breaks down.

In terms of research priorities, this work makes investigation of the half-collinear regime a new subfield rather than a solved problem. The explicit list of deferred extensions — graviton amplitudes, supersymmetrization, behavior under S-algebra and Lw₁₊∞, celestial holography applications — signals that the authors view the single-minus gluon result as a first step in mapping out a larger continent. The paper also implicitly redirects attention away from the search for ever-more-efficient on-shell recursion methods (BCFW, CHY, ambitwistor strings — all of which assume the standard helicity classification) and toward systematic exploration of the half-collinear locus in various theories and signatures. The question is no longer "how do we compute MHV amplitudes faster?" but "what other amplitudes have we been incorrectly assuming are zero, and what lives in the half-collinear regime of gravity, QED, or supersymmetric theories?"

Finally, the paper carries a notable metascientific signal regarding AI-assisted mathematical discovery. The formula (39) is explicitly attributed to a conjecture by GPT-5.2 Pro, with a proof by "a new internal OpenAI model," subsequently verified by hand. The authors place this disclosure prominently in the introduction, and the list of "on behalf of OpenAI" affiliations includes authors from that institution. This is one of the first clear instances in the scattering amplitudes literature where a nontrivial closed-form conjecture — not a numerical fit or an automated cross-check, but a structural formula — is credited to an AI system. Whether this becomes a template for future work or an isolated curiosity depends on broader developments in AI-assisted theoretical physics, but the paper itself does not dwell on this aspect; the physics results stand on their own, with the AI contribution serving as a provenance note rather than a methodological argument.

Follow-Up Research This Work Enables

Complete the chamber geometry of the product formula (39). The paper establishes that the stripped amplitude in ℛ₁ is a product of n−2 signed projection operators, each factor taking values +1, −1, or 0 depending on whether sgₘ,ₘ₊₁ and sg₁,₂⋯ₘ agree in sign. But the chamber structure — how many distinct nonzero chambers exist as a function of n, what walls separate them, how the amplitude jumps across walls, and whether all chambers are nonempty — is not analyzed. A natural follow-up would enumerate chambers for general n: each condition sgₘ,ₘ₊₁ = sg₁,₂⋯ₘ is a sign agreement between a local z̃ difference (independent of ω) and a weighted-average sign (dependent on ω₁ through ωₘ). The walls are defined by sgₘ,ₘ₊₁ = 0 or sg₁,₂⋯ₘ = 0, and the amplitude jumps by ±1 or drops to zero when crossing them. For n = 4, 5, one could enumerate all possible sign configurations of the z̃ differences and ω-weighted averages, compute the amplitude in each, and characterize the resulting chamber complex. Does the number of nonzero chambers grow polynomially or exponentially in n? Is there a combinatorial interpretation (e.g., chambers labeled by certain permutations or trees)? This is a concrete counting problem directly motivated by the paper's result.

Extend the closed form to multi-particle frequency-sign regions beyond ℛ₁. The product formula (39) is proven for the region where a single particle (particle 1) has negative frequency and all others have positive frequency, and by cyclicity to the n analogous regions ℛₖ. The paper explicitly defers "a longer general formula for the single-minus amplitude outside of ℛ₁" to future work. The natural next step is to analyze what happens when multiple particles share negative frequency — e.g., a configuration with ω₁ < 0, ω₂ < 0, and ωₐ > 0 for a ≥ 3. Does the V function still vanish for multi-gluon plus-helicity blocks? The weighted-variance proof in Section II.3.1 relies on all ω in the block being positive; with mixed signs, the left- and right-partial sums Ω_L and Ω_R may not be positive, breaking the argument. A concrete experiment: take n = 5 with ω₁ < 0, ω₂ < 0, ω₃,ω₄,ω₅ > 0, and compute A₁₂₃₄₅ from the general recursion (21) for a grid of z̃ and ω values. Does the amplitude still factorize? Does it take integer values? Does a modified product formula emerge (perhaps with different projection operators)? Negative results (exponential complexity persists for mixed signs) would establish that ℛ₁ is genuinely special, while positive results (generalized factorization) would suggest a larger algebraic structure governing the half-collinear regime.

Compute the graviton analog of formula (39) and test for structural differences. The paper states that "the construction generalizes directly from gluon to graviton amplitudes." Graviton amplitudes in spinor-helicity are built from the same kinematic data but with different vertex structures (the BG recursion for gravity involves products of brackets rather than single brackets). The graviton single-minus problem — one negative-helicity graviton, n−1 positive — has the same half-collinear support as the gluon case (since the polarization vector construction is analogous). The natural follow-up is to derive the graviton Berends–Giele recursion in the half-collinear regime, prove the analog of the V = 0 theorem in the ℛ₁ region (or its gravitational equivalent), and obtain the graviton product formula. Does it also factorize into n−2 signed projection operators? Or does gravity's higher-degree vertex structure (the BG kernel for gravity involves products like [λ̃_L λ̃_R]² rather than [λ̃_L λ̃_R]) produce qualitatively different chamber structure — perhaps factors involving squares of sign sums, or values beyond {+1, −1, 0}? Comparing the gluon and graviton formulas would isolate what is universal about the half-collinear/frequency-sign collapse (the causal structure of massless spinor-helicity) from what is theory-specific (the vertex structure). The paper's mention that the results "should transform under the S-algebra, the Lw₁₊∞ algebra, and their supersymmetric extensions" suggests that the graviton case would connect directly to the celestial holography program — the Lw₁₊∞ algebra is the wedge algebra of w₁₊∞, which acts on the celestial sphere and is known to organize soft graviton theorems. A graviton product formula would provide concrete amplitudes on which this algebra acts.

Verify the consistency identities explicitly and publish the computations. The paper claims that all five consistency conditions (cyclicity, reflection, U(1) decoupling, Kleiss–Kuijf, Weinberg soft theorem) are satisfied by the general recursion (21) and the product formula (39), but the verifications are deferred: "Details of this calculation will appear elsewhere." These checks are not merely confirmatory — they are sensitive diagnostics of the sign-function and Θ-function structure. An error in any sign assignment (from the sg convention or the Θ-function chamber logic) would manifest as a violation of one of these identities. A concrete follow-up would be to explicitly verify:

Soft theorem: For n = 5 and n = 6, take the product formula (39), compute the ωₙ → 0 limit analytically (showing how the last factor ½(sgₙ₋₁,ₙ + sg₁,₂⋯ₙ₋₁) produces the soft factor), and confirm that the remaining n−3 factors match the (n−1)-point formula. Show the limiting procedure does not cross chamber walls where signs jump discontinuously (or handle such crossings correctly).
U(1) decoupling and KK relations: For n = 5 or 6, take the explicit general expressions (31)–(32), substitute into the decoupling identity (26) and the KK identity (27), simplify using the sign-function algebra and momentum conservation, and show the sums vanish identically. These are nontrivial checks because the terms involve multi-particle sign functions (sg₂,₃₄, etc.) that are not independent — they satisfy relations from the additivity of the bracket.

Publishing these calculations would remove the only significant gap in the paper's evidence and provide a model for others seeking to verify or extend the results. It would also clarify how the iε prescription interacts with the consistency identities — do the identities hold in each chamber separately, or only after summing over the Θ-function chambers?

Develop a generating function or combinatorial interpretation for the general half-collinear amplitude. The general recursion (21) is a sum over ordered partitions with weights given by PT̂ factors and preamplitude products. The 6-point expression (32) has 32 terms, each a product of sign functions — suggesting a combinatorial structure that the paper does not explore. Is there a generating function that encodes the amplitude for all n, perhaps as a sum over something like non-crossing partitions or chord diagrams? The piecewise-constant, integer-valued nature of the amplitude suggests a connection to combinatorial objects (e.g., the amplitude might count something like consistent sign assignments on a polygon with one distinguished vertex). A concrete approach: for small n (3 through 6), enumerate all possible sign configurations of the two-particle and multi-particle brackets, compute the amplitude in each by evaluating the recursion (21) or the product formula (39) in ℛ₁, and look for patterns — e.g., does the amplitude equal the Euler characteristic of some poset? Is it the Möbius function of an arrangement of hyperplanes? The paper's remark that "the Mellin transform of the amplitudes in some sectors is given by Lauricella functions" (Introduction, celestial holography context) hints at a connection to hypergeometric functions associated with hyperplane arrangements, suggesting the chamber counting may have a deeper geometric meaning.

Test whether single-minus amplitudes contribute to Minkowski-space observables through analytic continuation. The paper's amplitudes are defined in Klein (2,2) signature. A pressing question for the broader amplitudes community is whether these amplitudes contribute to physical observables in Minkowski (3,1) signature — not as direct scattering probabilities (since the half-collinear support is measure-zero in Minkowski phase space) but indirectly. Possible mechanisms include:

Generalized unitarity cuts: In loop calculations, internal lines are placed on shell and tree amplitudes are sewn together. The cut integration is typically performed in Minkowski signature, but analytic continuation to complex kinematics is standard. If a loop diagram has a cut where two internal lines become collinear with an external leg, the half-collinear single-minus amplitude could contribute as a residue. Does the known all-plus loop amplitude (which is nonvanishing in Minkowski space and famously simple) receive contributions from sewing single-minus trees?
BCFW recursion: BCFW shifts analytically continue external momenta into the complex plane. The recursion relation expresses an amplitude as a sum over factorization channels, with the shifted legs going on shell at poles. If a BCFW shift passes through a region where some shifted momenta enter the half-collinear regime, could single-minus amplitudes appear as residues in the recursion? If so, they would contribute to Minkowski-space amplitudes indirectly, through the BCFW representation.

A concrete calculation: compute the 4-point all-plus one-loop amplitude using generalized unitarity, with the cut involving a tree-level single-minus amplitude evaluated at complex on-shell momenta that satisfy the half-collinear condition on the cut. If a nonzero contribution emerges, it would demonstrate that these amplitudes are not merely mathematical curiosities but have physical consequences.

Practical Applications and Downstream Use Cases

Building block verification for celestial holography. The celestial holography program seeks to reformulate flat-space scattering amplitudes as correlation functions of a conformal field theory on the celestial sphere. A key step is the Mellin transform of momentum-space amplitudes to boost-eigenstate (conformal primary) basis. The paper states that "the Mellin transform of the amplitudes in some sectors is given by Lauricella functions" — a class of multivariable hypergeometric functions that naturally arise from integrals of products of linear forms raised to powers, precisely the structure one gets from Mellin-transforming the product formula (39). The single-minus amplitudes are now explicitly known — in closed form, with integer values and explicit chamber structure — providing a clean, solvable data point for celestial holography. A practitioner can take Equation (39), Mellin-transform each factor ½(sgₘ,ₘ₊₁ + sg₁,₂⋯ₘ) in ℛ₁, and obtain the celestial amplitude as a Lauricella function of the conformal cross-ratios. Because the amplitude is piecewise-constant and the δ-function support is explicit, the Mellin integral factorizes into geometric regions, and the result is a combination of hypergeometric functions with integer coefficients. This provides a concrete testing ground for the celestial dictionary — the celestial OPE of single-minus operators, the action of the Lw₁₊∞ symmetry algebra on conformal primaries, and the soft limits in celestial coordinates can all be computed exactly and checked against the product formula's known soft behavior (Equation 28).

SDYM classical-quantum correspondence: a computable bridge. The paper's explicit construction of tree-level single-minus amplitudes in SDYM provides the missing quantum half of a classical-quantum correspondence that has been one-sided for decades. The classical solution space of SDYM is known to be parameterized by twistor data — the ADHM construction gives all instanton solutions, and the Penrose-Ward transform relates them to holomorphic vector bundles on twistor space. The paper now provides the quantum objects: tree amplitudes that are nonzero, piecewise-constant, and localized to the half-collinear locus. A concrete application is to match specific classical solutions to specific tree amplitudes: take an ADHM instanton configuration with a given topological charge and asymptotic behavior, extract its scattering data (the classical analog of the LSZ reduction), and compare to the n-gluon single-minus amplitude evaluated on the corresponding momentum configuration in ℛ₁. Does the instanton number relate to the integer value (+1, −1, 0) of the stripped amplitude? Does the chamber structure of the product formula (39) correspond to different instanton sectors? The paper's identification that V̄ — the surviving term in ℛ₁ — is the on-shell Parke-Taylor factor suggests a direct link: V̄ encodes the causal structure of the classical background, and the different chambers may correspond to different classical solutions with the same asymptotic momenta. This is speculative but directly motivated by the paper's results, and the fact that both sides of the correspondence are now computationally accessible makes it a viable research project.

Simplified unitarity-based loop calculations in split-signature regularization. In practical loop-amplitude calculations, one often needs a regularization scheme that preserves gauge invariance while taming infrared divergences. Split-signature (2,2) kinematics — precisely the arena of this paper — have been explored as a natural regularization for loop integrals because they keep all particles massless and on shell while avoiding collinear singularities through the independence of angle and square brackets. The single-minus amplitudes computed here provide a complete, closed-form tree-level input for such calculations. If one wishes to compute, say, the all-plus one-loop amplitude using four-dimensional unitarity in (2,2) signature, the cut will involve sewing tree amplitudes that may include single-minus configurations when the cut kinematics enter the half-collinear regime. The product formula (39) provides the tree-level building blocks. Moreover, because the single-minus amplitudes are piecewise-constant, the phase-space integrals over the cut are simpler than in Minkowski signature — they reduce to integrals of step functions over the z̃ and ω variables, which produce Lauricella functions (as the paper notes). A practitioner can set up the cut integral, insert formula (39) for the tree inputs, and evaluate it analytically. This provides a concrete pathway to computing loop amplitudes in a scheme where the tree-level inputs are now fully characterized.

A guaranteed-correct test case for numerical amplitude codes. Modern numerical unitarity and Berends–Giele codes (e.g., BlackHat, NJet, MG5_aMC, the numerical BG implementation in various frameworks) rely on the correctness of the underlying Feynman rules and recursion relations. The single-minus amplitudes provide a nontrivial, previously unaccounted-for test case that such codes can target. Since the product formula (39) gives the exact analytic result for configuration in ℛ₁, one can (a) generate random numeric kinematics satisfying the half-collinear condition (all zᵢ equal, arbitrary z̃ᵢ and ωᵢ with ω₁ < 0, ωₐ > 0), (b) evaluate the product formula to get the reference result (+1, −1, or 0), and (c) feed the same kinematics into a numerical BG or Feynman-diagram code and verify agreement. Any discrepancy would indicate either a bug in the numerical code's handling of the iε prescription, its collinear limits, or its implementation of the color-ordering. Because the analytic answer is piecewise-constant and integer-valued, the comparison is exact — no floating-point tolerance ambiguity. This provides a precision test that existing validation suites (which typically check MHV and NMHV amplitudes against known analytic formulas) do not cover. The discovery that single-minus amplitudes are nonvanishing means that codes which previously assumed they are zero and skipped their computation now have a correctness gap that these tests can identify.

When to Prefer This Method

The paper does not present a "method" in the sense of a choice among competing computational approaches to the same quantity. It presents a discovery of a new class of amplitudes — the single-minus sector in the half-collinear regime — and provides both a general recursion and a closed-form specialization. There is no alternative method for computing single-minus amplitudes because, prior to this paper, they were believed to be identically zero. The question of "when to prefer" therefore does not arise in the conventional sense (there is no competing formula to choose between). However, within the scope of the paper, there is a choice of which formula to use for computing single-minus amplitudes, depending on the kinematic regime:

If the kinematics satisfy the ℛ₁ condition (or are cyclically equivalent to ℛₖ for some k): use the product formula (39). This requires O(n) operations to evaluate n−2 sign functions and their product, and the result is immediately available as +1, −1, or 0 without any combinatorial sum. The ℛ₁ region is the physically natural decay configuration (one incoming minus-helicity gluon, n−1 outgoing plus-helicity gluons), so this covers a broad class of configurations.
If the kinematics are in the half-collinear regime but outside all ℛₖ regions (e.g., mixed frequency signs, or no single distinguished negative-frequency particle): the general recursion (21) with preamplitude sub-recursion (19) must be used. This involves summing over ordered partitions and has computational cost that grows exponentially with n. The paper does not provide a simplified closed form for this case. Practically, for n ≤ 6, the explicit formulas (29–32) can be used directly; for larger n, the recursion must be evaluated algorithmically.
If the kinematics are generic (not in the half-collinear regime): the single-minus amplitude vanishes, and standard methods (Parke-Taylor for MHV, BCFW for higher NMHV sectors) apply to the appropriate helicity configurations.

The paper's central contribution is not a computational speedup but a conceptual correction: single-minus amplitudes exist, they live in the half-collinear regime, and in the ℛ₁ subregion they admit an exact closed form. The "method" is using this knowledge to correctly identify when these amplitudes are nonzero and to compute them when they are, rather than incorrectly assuming they vanish as the field has done for forty years.

🎯 Pitch

1. Executive Summary

2. Context and Motivation

The Core Problem: A Persistent Belief That Single-Minus Amplitudes Vanish

Why This Problem Matters: Theoretical and Structural Significance

Prior Approaches and Where They Fall Short

How This Paper Positions Itself Relative to Existing Work

3. Technical Approach

3.1 Reader Orientation

3.2 Big-Picture Architecture (Diagram in Words)

3.3 Roadmap for the Deep Dive

3.4 Detailed, Sentence-Based Technical Breakdown

Kinematic Parameterization and the Half-Collinear Regime

Berends–Giele Recursion for Off-Shell Form Factors

The Preamplitude Ā and Vertex Function V

Master Identity and the PT–δ–V Relationship

Solution of the BG Recursion Using Preamplitudes and PT Factors

LSZ Reduction and the General Stripped Amplitude

Collapse in Region ℛ₁: Vanishing of V

Collapse of the Amplitude Recursion

Evaluating V̄ and Deriving the Product Formula (39)

Consistency Checks and Mathematical Structure

4. Key Insights and Innovations

Innovation 1: Demolishing a Forty-Year-Old "Theorem" by Exploiting a Distributional Loophole

Innovation 2: The Vertex Function as a Chamber-Based Causality Filter

Innovation 3: An Exponential-to-Linear Complexity Collapse via Kinematic Selection

Innovation 4: Identification of Frequency-Sign as a Kinematic Invariant Organizing Amplitude Structure

Innovation 5: Signed Projection Operators as the Universal Building Block of Nonzero Single-Minus Amplitudes

5. Experimental Analysis

Evaluation Methodology

Main Quantitative Results

The General Single-Minus Stripped Amplitude (Section I)

Simplification in Region ℛ₁ (Section II.2)

The General Formula and Its Proof (Section II.3)

Ablation Studies and Robustness Checks

Critical Assessment

6. Limitations and Trade-offs

Kinematic Scope: Closed Form Exists Only in Region ℛ₁, Not the General Half-Collinear Regime

Signature Dependence: Nonzero Only in Klein Signature, Not in Physical Minkowski Space

Deferred Verification of Consistency Conditions

No Exploration of Wall-Crossing Structure or Chamber Geometry

Unaccounted Complexity in Evaluating Multi-Particle Sign Functions

7. Implications and Future Directions

How This Work Changes the Landscape

Follow-Up Research This Work Enables

Practical Applications and Downstream Use Cases

When to Prefer This Method

Adjacent in domain

SAGE: Scalable Agentic 3D Scene Generation for Embodied AI

WebWorld: A Large-Scale World Model for Web Agent Training