arxiv: 2604.20377 · v1 · submitted 2026-04-22 · ✦ hep-th

Recognition: unknown

Off-shell recursion for all-loop planar integrands in Yang-Mills theory

Yi-Xiao Tao

Authors on Pith no claims yet

Pith reviewed 2026-05-10 00:22 UTC · model grok-4.3

classification ✦ hep-th

keywords Yang-Millsplanar integrandsoff-shell recursionloop amplitudesperturbinerghostsmatrix formalism

0 comments

The pith

Off-shell recursion extends to all-loop planar Yang-Mills integrands with ghosts.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper takes the off-shell recursion method and applies it to construct the planar loop integrands in Yang-Mills theory starting from the classical equation of motion. It finds that the gluon contributions alone can be expressed using a matrix formalism that makes the off-shell properties more transparent and opens the way to new relations among amplitudes. Including the ghost fields then gives the full recursion for the theory. The two-loop case serves as an illustration of the general procedure.

Core claim

Following the recursion steps, the pure gluon sector of the planar loop integrands can be written in matrix formalism. This matrix formalism not only makes the off-shell structure of the Yang-Mills planar integrands clearer, but also has potential use in finding amplitude relations at higher-loop levels. Furthermore, adding the ghost contribution allows writing down the whole recursion step of the Yang-Mills planar loop integrands with ghost contributions. Considering the 2-loop planar integrand recursion as a special case leads to a recursion strategy in this case.

What carries the argument

Off-shell recursion via the perturbiner method applied to the classical equations of motion, extended to the planar loop sector with matrix form for gluons and explicit ghost terms.

If this is right

The matrix form clarifies the off-shell structure of the integrands.
It has potential application in deriving amplitude relations at higher loop orders.
The full recursion including ghosts covers the complete Yang-Mills planar loop integrands.
A concrete recursion strategy applies to the two-loop case.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the matrix formalism holds, it may allow systematic derivation of integrand relations without explicit computation at each loop order.
Similar recursion could be tested in other planar theories or with additional fields.

Load-bearing premise

The off-shell recursion method extends without obstruction from the pure gluon case to the full Yang-Mills theory when ghosts are included for planar integrands.

What would settle it

Direct computation of a low-point two-loop planar integrand that disagrees with the result obtained from the proposed recursion would disprove the claim.

read the original abstract

In this letter, we focus on the application of the off-shell recursion method proposed in \cite{Tao:2025fch} in the Yang-Mills planar loop integrands, which starts with solving the classical equation of motion via the perturbiner method. Following the recursion steps, we point out that the pure gluon sector of the planar loop integrands can be written in matrix formalism. This matrix formalism not only makes the off-shell structure of the Yang-Mills planar integrands clearer, but also has potential use in finding amplitude relations at higher-loop levels. Furthermore, we add the ghost contribution and write down the whole recursion step of the Yang-Mills planar loop integrands with ghost contributions. Finally, we consider the 2-loop planar integrand recursion as a special case and conclude a recursion strategy in this case.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This applies the prior off-shell recursion to planar Yang-Mills, packages the gluon sector in matrices and adds ghosts, but only works out the two-loop case explicitly while asserting the all-loop version.

read the letter

This paper takes the off-shell recursion from the 2025 work and applies it to planar loop integrands in Yang-Mills. The concrete steps are writing the pure-gluon part in matrix form and including the ghost contributions in the full recursion relation. They then treat the two-loop case as a worked example to sketch the strategy going forward. The matrix packaging does make the off-shell structure easier to see at a glance, and that could help when hunting for amplitude relations at higher loops. Adding ghosts completes the setup for the full theory rather than leaving it at gluons only. The two-loop specialization gives a practical check on how the recursion runs in a known setting. The limitation is that the all-loop claim is presented as following directly once the general step is written down, yet no derivation or consistency check appears for three loops or higher. There are also no comparisons to existing two-loop integrand results or error estimates on the matrix representation. If the extension introduces no hidden obstructions, the formal step may be enough, but the paper leaves that as an assertion rather than a shown result. This is aimed at the handful of people who already work with recursions and perturbiner methods for planar gauge-theory integrands. A reader in that niche could use the matrix form to organize calculations or look for new relations. It deserves a serious referee because the technical move is specific and the potential utility is real, even if the higher-loop evidence needs more support in review.

Referee Report

2 major / 1 minor

Summary. The manuscript applies the off-shell recursion method of Tao:2025fch to planar Yang-Mills loop integrands, beginning from the perturbiner solution of the classical equations of motion. It states that the pure-gluon sector admits a matrix formalism, incorporates ghost contributions to obtain the complete recursion step, and treats the two-loop case explicitly as a special case from which a general recursion strategy for all loops is concluded.

Significance. If the all-loop extension holds, the matrix representation of the gluon sector would clarify the off-shell structure of planar integrands and could assist in deriving new amplitude relations at higher loops. The inclusion of ghosts completes the recursion for the full theory, providing a systematic framework that builds directly on the perturbiner method and may enable reproducible computations of loop integrands beyond two loops.

major comments (2)

[Abstract] Abstract: the central claim is that the off-shell recursion extends without obstruction to all-loop planar Yang-Mills integrands (gluons in matrix form plus ghosts). However, the explicit construction and closure of the recursion are supplied only for the two-loop special case; no derivation or consistency check is given for L≥3 that would confirm the matrix formalism and ghost insertions preserve planarity and off-shell structure at higher loops.
[2-loop planar integrand recursion] 2-loop planar integrand recursion section: the manuscript concludes a general recursion strategy from the two-loop case, but does not demonstrate how the matrix representation of the gluon sector combines with the ghost terms to close the recursion at arbitrary loop order, nor does it provide cross-checks against known two-loop results or explicit examples for L=3.

minor comments (1)

[Abstract] The abstract would be strengthened by a brief indication of the explicit form of the matrix representation for the gluon sector, even if only schematic.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We respond to each major comment below and indicate where revisions will be made.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim is that the off-shell recursion extends without obstruction to all-loop planar Yang-Mills integrands (gluons in matrix form plus ghosts). However, the explicit construction and closure of the recursion are supplied only for the two-loop special case; no derivation or consistency check is given for L≥3 that would confirm the matrix formalism and ghost insertions preserve planarity and off-shell structure at higher loops.

Authors: The off-shell recursion is formulated using the perturbiner solution to the classical equations of motion, which generates contributions order by order without reference to a fixed loop number. The matrix representation of the pure-gluon sector follows directly from the color-ordered structure of the recursion relations and applies uniformly at every order. Ghost contributions are inserted at each recursive step in a manner that completes the Yang-Mills content while preserving planarity. Because each higher-loop integrand is assembled from lower-order building blocks via the same relations, the two-loop closure demonstrates the absence of obstructions at arbitrary order. We will revise the abstract and add a short clarifying paragraph on this inductive structure in the revised manuscript. revision: yes
Referee: [2-loop planar integrand recursion] 2-loop planar integrand recursion section: the manuscript concludes a general recursion strategy from the two-loop case, but does not demonstrate how the matrix representation of the gluon sector combines with the ghost terms to close the recursion at arbitrary loop order, nor does it provide cross-checks against known two-loop results or explicit examples for L=3.

Authors: The two-loop section explicitly shows the combination of the gluon matrix with the ghost terms and presents the resulting recursion as the general strategy, since the same combination of matrix elements and ghost insertions appears at every order by the definition of the recursion. We acknowledge that an explicit L=3 example or direct cross-check would provide additional verification. Given the letter format, we limited the discussion to the framework and its two-loop illustration. We will add a brief paragraph explaining the general-order closure and note consistency with known two-loop integrands obtained via other methods, while leaving a full L=3 computation for future work. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation applies prior method to new sectors without self-referential reduction

full rationale

The paper's chain begins by invoking the off-shell recursion and perturbiner method from the self-cited Tao:2025fch, then states that the pure-gluon sector admits a matrix formalism, adds ghost terms, and treats the 2-loop case explicitly to conclude a recursion strategy. No equation or claim reduces a derived quantity (matrix form, ghost recursion, or all-loop strategy) to an input by construction, nor renames a fit as a prediction. The self-citation supplies the foundational recursion procedure; the present work supplies the application to Yang-Mills with ghosts and the 2-loop specialization. This is standard non-circular extension rather than tautological re-derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The construction relies on the perturbiner solution of the classical Yang-Mills equations of motion and on the off-shell recursion defined in the authors' prior work; no free parameters, new entities, or ad-hoc axioms are mentioned in the abstract.

axioms (2)

domain assumption The perturbiner method generates a solution to the classical Yang-Mills equations of motion that can be used as the starting point for off-shell recursion.
Invoked in the first sentence of the abstract as the entry point for the recursion.
domain assumption The off-shell recursion of Tao:2025fch applies directly to the planar sector of Yang-Mills theory.
The letter states it focuses on the application of that method.

pith-pipeline@v0.9.0 · 5430 in / 1456 out tokens · 41772 ms · 2026-05-10T00:22:21.776336+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

60 extracted references · 40 canonical work pages · 5 internal anchors

[1]

Fromk 2 12···mAcomb l12···m ·ϵ n +k 2 12···mAcontact lm12···m−1 ·ϵwith la gluon leg, usingϵ lµϵν →η µν/l2 1 as before
[2]

Fromk 2 12···mc bcomb lP withlabghost leg, usingc b l → 1/l2 1
[3]

In the ghost sewing procedure,b l orc l must be in the first position, likeb lP =b lBP

From−k 2 12···mb ccomb lP withlacghost leg, using b cl → −1/l 2 1. In the ghost sewing procedure,b l orc l must be in the first position, likeb lP =b lBP . We should note that for a 2-way kernel part with both ghost propagators and gluon propagators, cases 2 (or 3), as well as case 1, will lead to the same loop diagram and must be multiplied by a factor o...
[4]

there arem ℓ sets in total

Consider ordered cyclic set (b ℓ, aℓ,1,2,· · ·, m ℓ − 1, mℓ) and the cyclic permutation of (12· · ·m ℓ), i.e. there arem ℓ sets in total. Note that due to the cyclic properties, (1,2,· · ·, m ℓ −1, m ℓ, aℓ, bℓ) is equivalent to (aℓ, bℓ,1,2· · ·, m ℓ −1, m ℓ)
[5]

There is only 1 case for a givenkand a given set

Then, fork-divisions, which means we divide a set intokparts, of each set, we set: i)b ℓ itself to be one part of thek-division, ii) For a part not involving aℓ, there can only be 1 element in the part, like (bℓ|aℓ,1|2|3). There is only 1 case for a givenkand a given set
[6]

Then do the replacement:A aℓµAbℓν → ηµν/l2 ℓ,c bℓ baℓ →1/l 2 ℓ, andb bℓ caℓ → −1/l 2 ℓ together withk aℓ =−k bℓ =l ℓ

Consider ak-way bare (ℓ−1)-loop kernel, replace Aiµ,b i,c i of this kernel with the comb component Acomb ···µ ,b comb ··· ,c comb ··· of each part of ak-division re- spectively. Then do the replacement:A aℓµAbℓν → ηµν/l2 ℓ,c bℓ baℓ →1/l 2 ℓ, andb bℓ caℓ → −1/l 2 ℓ together withk aℓ =−k bℓ =l ℓ. This replacement is equiva- lent to turning legsa ℓ andb ℓ in...
[7]

Fi- nally we will obtain am ℓ-wayℓ-loop kernel from (ℓ−1)-loop kernelI kernel ℓ,mℓ

Sum over all possiblek-division and allk≥2. Fi- nally we will obtain am ℓ-wayℓ-loop kernel from (ℓ−1)-loop kernelI kernel ℓ,mℓ . After getting loop kernels, we can obtain the planar integrands just as (7). Note that all loop kernels appear- ing in the integrand construction formula must have all legs gluon to make sure that no ghost legs are in the final ...
[8]

Based on (7), write down the formula for 2-loop planar integrands and figure out all the loop kernels we need
[9]

For the 1-loop kernel, calculate directly. For the 2- loop kernel, sayI kernel 2,m , when we want to construct it from 1-loop kernels, we only need to consider the following elements: 1 2 I gluon 1,i (i∈ {2,3,· · ·, m+ 2}), (1− 1 2 δi, m 2 +2)I ghost 1,i (i∈ {2,3,· · ·,⌊ m 2 ⌋+ 2}) withI ghost 1,i =I kernel 1,i −I gluon 1,i (25) Note that there is no grap...
[10]

The pure gluon part can be written in the matrix formalism, and the entire 1-loop kernel can be got from the analysis in section IV
[11]

In the ancillary file, we have listed 2-way 2-loop kernels for both the pure gluon sector and the entire kernel with ghost contributions

Construct each term in (7) to get the 2-loop inte- grands. In the ancillary file, we have listed 2-way 2-loop kernels for both the pure gluon sector and the entire kernel with ghost contributions. VI. CONCLUSION In this letter, we discussed the off-shell recursion of the YMℓ-loop integrands. We proposed the matrix formal- ism of the gluon sector recursion...
[12]

Hence, once we construct the pure gluon sector in a very organized way, the amplitude relations will be easier to find

to transform all loops into scalar loops, the ghost loop will not contribute, just as in the 1-loop case [9]. Hence, once we construct the pure gluon sector in a very organized way, the amplitude relations will be easier to find. We left this in future. After obtaining the ghost parts, we are able to con- struct the total YM planar integrands. Although fo...
[13]

Systematic approach to $\ell$-loop planar integrands from the classical equation of motion

Y.-X. Tao, Phys. Rev. D112, 106002 (2025), arXiv:2504.15701 [hep-th]

work page internal anchor Pith review Pith/arXiv arXiv 2025
[14]

Q. Cao, J. Dong, S. He, and F. Zhu, Phys. Rev. D111, 065015 (2025), arXiv:2412.19629 [hep-th]. 6

work page arXiv 2025
[15]

On differential operators and unifying relations for $1$-loop Feynman integrands

K. Zhou, JHEP10, 150 (2021), arXiv:2108.04025 [hep- th]

work page internal anchor Pith review Pith/arXiv arXiv 2021
[16]

Transmutation operators and expansions for $1$-loop Feynman integrands

K. Zhou, Phys. Rev. D106, 025014 (2022), arXiv:2201.01552 [hep-th]

work page internal anchor Pith review Pith/arXiv arXiv 2022
[17]

Q. Cao, S. He, Y. Zhang, and F. Zhu, (2025), arXiv:2509.25129 [hep-th]

work page arXiv 2025
[18]

Du and C

Y.-J. Du and C. Xie, (2025), arXiv:2512.08354 [hep-th]

work page arXiv 2025
[19]

Du, C.-H

Y.-J. Du, C.-H. Fu, Y. Wang, and C. Xie, Phys. Rev. D 113, 045025 (2026), arXiv:2511.10963 [hep-th]

work page arXiv 2026
[20]

Xie and Y.-J

C. Xie and Y.-J. Du, (2025), arXiv:2509.25632 [hep-th]

work page arXiv 2025
[21]

Chen and Y.-X

Q. Chen and Y.-X. Tao, JHEP08, 038 (2023), arXiv:2301.08043 [hep-th]

work page arXiv 2023
[22]

Arkani-Hamed, H

N. Arkani-Hamed, H. Frost, G. Salvatori, P.-G. Pla- mondon, and H. Thomas, JHEP08, 194 (2025), arXiv:2309.15913 [hep-th]

work page arXiv 2025
[23]

Arkani-Hamed, H

N. Arkani-Hamed, H. Frost, G. Salvatori, P.-G. Pla- mondon, and H. Thomas, JHEP09, 033 (2025), arXiv:2311.09284 [hep-th]

work page arXiv 2025
[24]

Arkani-Hamed, Q

N. Arkani-Hamed, Q. Cao, J. Dong, C. Figueiredo, and S. He, (2024), arXiv:2408.11891 [hep-th]

work page arXiv 2024
[25]

The Cut Equation

N. Arkani-Hamed, H. Frost, and G. Salvatori, (2024), arXiv:2412.21027 [hep-th]

work page internal anchor Pith review Pith/arXiv arXiv 2024
[26]

Cao and F

Q. Cao and F. Zhu, (2025), arXiv:2503.15860 [hep-th]

work page arXiv 2025
[27]

The bi-adjoint scalar $\ell$-loop planar integrand recursion and graded inverse variables

Y.-X. Tao, JHEP03, 165 (2026), arXiv:2505.07077 [hep- th]

work page internal anchor Pith review Pith/arXiv arXiv 2026
[28]

Tao and Q

Y.-X. Tao and Q. Chen, JHEP02, 030 (2023), arXiv:2210.15411 [hep-th]

work page arXiv 2023
[29]

Tao and K

Y.-X. Tao and K. Wu, (2024), arXiv:2404.02997 [hep-th]

work page arXiv 2024
[30]

Wu and Y.-J

K. Wu and Y.-J. Du, JHEP01, 162 (2022), arXiv:2109.14462 [hep-th]

work page arXiv 2022
[31]

Tao, Phys

Y.-X. Tao, Phys. Rev. D108, 125020 (2023), arXiv:2309.15657 [hep-th]

work page arXiv 2023
[32]

Gomez, R

H. Gomez, R. Lipinski Jusinskas, C. Lopez-Arcos, and A. Quintero Velez, Phys. Rev. Lett.130, 081601 (2023), arXiv:2208.02831 [hep-th]

work page arXiv 2023
[33]

Chattopadhyay and K

P. Chattopadhyay and K. Krasnov, JHEP03, 191 (2022), arXiv:2110.00331 [hep-th]

work page arXiv 2022
[34]

F. A. Berends and W. T. Giele, Nucl. Phys. B306, 759 (1988)

1988
[35]

F. A. Berends and W. T. Giele, Nucl. Phys. B313, 595 (1989)

1989
[36]

C. R. Mafra, O. Schlotterer, S. Stieberger, and D. Tsimpis, Phys. Rev. D83, 126012 (2011), arXiv:1012.3981 [hep-th]

work page arXiv 2011
[37]

C. R. Mafra and O. Schlotterer, JHEP03, 097 (2016), arXiv:1510.08846 [hep-th]

work page arXiv 2016
[38]

C. R. Mafra, JHEP07, 080 (2016), arXiv:1603.09731 [hep-th]

work page arXiv 2016
[39]

Lee, JHEP05, 051 (2022), arXiv:2202.08133 [hep-th]

K. Lee, JHEP05, 051 (2022), arXiv:2202.08133 [hep-th]

work page arXiv 2022
[40]

Armstrong, H

C. Armstrong, H. Gomez, R. Lipinski Jusinskas, A. Lip- stein, and J. Mei, Phys. Rev. D106, L121701 (2022), arXiv:2209.02709 [hep-th]

work page arXiv 2022
[41]

K. Cho, K. Kim, and K. Lee, JHEP01, 186 (2022), arXiv:2109.06392 [hep-th]

work page arXiv 2022
[42]

K. Cho, K. Kim, and K. Lee, JHEP03, 112 (2023), arXiv:2209.11424 [hep-th]

work page arXiv 2023
[43]

K. Cho, K. Kim, and K. Lee, JHEP05, 050 (2024), arXiv:2311.01284 [hep-th]

work page arXiv 2024
[44]

Gomez and R

H. Gomez and R. L. Jusinskas, Phys. Rev. Lett.127, 181603 (2021), arXiv:2106.12584 [hep-th]

work page arXiv 2021
[45]

Tao, JHEP09, 193 (2023), arXiv:2307.14772 [hep- th]

Y.-X. Tao, JHEP09, 193 (2023), arXiv:2307.14772 [hep- th]

work page arXiv 2023
[46]

Chattopadhyay and Y.-X

P. Chattopadhyay and Y.-X. Tao, JHEP03, 100 (2024), arXiv:2401.02760 [hep-th]

work page arXiv 2024
[47]

K. G. Selivanov, Phys. Lett. B420, 274 (1998), arXiv:hep-th/9710197

work page arXiv 1998
[48]

K. G. Selivanov, Commun. Math. Phys.208, 671 (2000), arXiv:hep-th/9809046

work page arXiv 2000
[49]

A. A. Rosly and K. G. Selivanov, (1997), arXiv:hep- th/9710196

work page arXiv 1997
[50]

Rosly and K

A. Rosly and K. Selivanov, Phys. Lett. B426, 334 (1998), arXiv:hep-th/9801044

work page arXiv 1998
[51]

Mizera and B

S. Mizera and B. Skrzypek, JHEP10, 018 (2018), arXiv:1809.02096 [hep-th]

work page arXiv 2018
[52]

Generating Feynman Diagrams and Amplitudes with FeynArts 3

T. Hahn, Comput. Phys. Commun.140, 418 (2001), arXiv:hep-ph/0012260

work page Pith review arXiv 2001
[53]

Cheung, C.-H

C. Cheung, C.-H. Shen, and C. Wen, JHEP02, 095 (2018), arXiv:1705.03025 [hep-th]

work page arXiv 2018
[54]

Gomez, R

H. Gomez, R. Lipinski Jusinskas, C. Lopez-Arcos, and A. Quintero Velez, Phys. Rev. Lett.134, 111602 (2025), arXiv:2411.07939 [hep-th]. 7 SUPPLEMENT AR Y MA TERIAL

work page arXiv 2025
[55]

flip”, we mean “turn upside down

Graph factor and bare loop kernel In the loop kernel recursion, we need to define thegraph factorin order to avoid overcounting. For anℓ-loop Feynman diagram, a graph factor can be written asg=S× 1 ℓ−r . Here,Sis the symmetry factor coming from the symmetry of the Feynman diagram, andris defined as below. We first define the largest loop to be the loop th...
[56]

Here, we present the recursion for the bi-adjoint scalar theory and useϕ P|Q to denote the bi-adjoint scalar multi-particle currents

Loop kernel recursion for bi-adjoint scalars Theℓ-loop kernel can be obtained from the (ℓ−1)-loop kernel recursively. Here, we present the recursion for the bi-adjoint scalar theory and useϕ P|Q to denote the bi-adjoint scalar multi-particle currents. For anm ℓ-wayℓ-loop kernel, we have the following steps to construct it:
[57]

there arem ℓ sets in total

Consider ordered cyclic set (b ℓ, aℓ,1,2,· · ·, m ℓ −1, m ℓ) and the cyclic permutation of (12· · ·m ℓ), i.e. there arem ℓ sets in total. Note that due to the cyclic properties, (1,2,· · ·, m ℓ −1, m ℓ, aℓ, bℓ) is equivalent to (aℓ, bℓ,1,2· · ·, m ℓ −1, m ℓ)
[58]

There is only 1 case for a givenkand a given set

Then, fork-divisions, which means we divide a set intokparts, of each set, we set: i)b ℓ itself to be one part of thek-division, ii) For a part not involvinga ℓ, there can only be 1 element in the part, like (b ℓ|aℓ,1|2|3). There is only 1 case for a givenkand a given set
[59]

Then replaceϕ aℓ|aℓ ϕbℓ|bℓ with 1/l 2 ℓ andk aℓ =−k bℓ =l ℓ

Consider ak-way bare (ℓ−1)-loop kernel, replaceϕ i|i of this kernel with the comb componentϕ comb ···|··· of each part of ak-division. Then replaceϕ aℓ|aℓ ϕbℓ|bℓ with 1/l 2 ℓ andk aℓ =−k bℓ =l ℓ. This replacement is equivalent to turning legsa ℓ andb ℓ into an internal line. Then add the corresponding graph factorgto each case. Since we have demonstrated ...
[60]

Finally we will obtain am ℓ-wayℓ-loop kernel from (ℓ−1)-loop kernelI kernel ℓ,mℓ

Sum over all possiblek-division and allk≥2. Finally we will obtain am ℓ-wayℓ-loop kernel from (ℓ−1)-loop kernelI kernel ℓ,mℓ