{"id":"a080e529-cd91-4c2f-b713-e9fb273c6078","arxiv_id":"2504.20897","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A derivative-free elliptic one-form basis yields Feynman integrals whose differential equations have a new factorized form with pure-function solutions.","lead":"Loop calculations in particle physics are hard when they involve elliptic curves, a shape with a hole in it. This paper shows that a particular derivative-free choice of one-forms inside the integrand makes the resulting integrals satisfy a cleaner, newly observed type of differential equation, and conjectures this works for a whole class of cases.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The pure-function claim rests on the unproven statement that every one-form in A has only simple poles; Appendix B checks only a random line, so the load-bearing local pole-order property is not established.","rationale":"The reader's weakest_assumption identifies exactly the step on which the pure-function claim depends: the local simple-pole property of the one-forms in A is asserted, not proved, and the supporting Appendix B is an a posteriori check along a single random line. My reading of the paper confirms that this is the most load-bearing unproven condition. If the simple-pole property fails, the Chen iterated-integral representation and the conclusion that solutions are pure functions break down. The paper contains genuine independent support: closed-form expressions for the elliptic leading singularities in Appendix A, explicit DEs for six nontrivial examples in Table I and the supplementary material, and a clear statement that the universal statement is a conjecture. None of this, however, replaces a systematic check of pole orders on all singular divisors. The reader's CONDITIONAL verdict is therefore appropriate, and I do not propose any adjustment.","tokens_in":28,"tokens_out":7187,"duration_ms":208233,"concrete_test":"Extract A from the supplementary material for the two-mass sunrise and at least one five-point example. For each irreducible factor f in the set of denominators of A (discriminants of the quartic P, mass combinations, threshold factors), fix a generic point on f=0, set t=f as a local normal coordinate and keep all remaining kinematic variables symbolic. Laurent-expand every entry of A in t down to order t^{-2} and require the t^{-2} coefficient to vanish identically as a function of the remaining variables, not merely at the sampled point. Repeat this check with t equal to each factor and with t=(f,g) near codimension-two intersections f=g=0. If any entry has a nonzero t^{-2} or worse coefficient, the simple-pole/pure-function claim fails. A cheaper cross-check is to repeat Appendix B on N random lines that meet the same divisor at different points and compare pole orders.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that solutions are pure functions requires every entry of A in Eq. (20) to have at most simple poles locally around every DE singularity. This is asserted in Sec. III and 'validated' in Appendix B by pulling the DE back on a single random line gamma(t) and inspecting the generic shape (B2). A random-line pullback samples only finitely many points on each singular divisor. A one-form can have a double pole on a divisor whose leading coefficient vanishes at the sampled point, or a double pole only at an intersection of two divisors, and the line check would miss it. The argument that psi-denominators cannot create poles by choosing a local period combination with nonvanishing constant term is a local-cycle existence statement; it is not a proof that the fixed entries of A in Eq. (20), e.g. d(psi2/psi1) and d(pi_i2 - pi_i1 psi2/psi1), have no t^{-2} terms in every coordinate chart. Since the iterated-integral representation and the 'pure functions' conclusion are directly downstream of the simple-pole property, this is the most load-bearing unproven step. The six worked examples are evidence, but the paper itself labels the general statement a conjecture, and no machine-checked or divisor-by-divisor verification is supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an integrand-level construction of Feynman-integral bases for elliptic leading singularities. The authors choose algebraic one-forms of the second kind, match them to Feynman integrands as in Eq. (16), and then apply the elliptic-leading-singularity rotation (17) to define a canonical basis. They observe that the resulting differential equations take the factorized form dG = A G with A = (1 - n epsilon) A0 + epsilon A1, (A0)_ij (A1)_ij = 0, A0^2 = 0, and all one-forms having at most simple poles. They conjecture that this structure is universal, support it with six examples summarized in Table I, and spell out the unequal-mass sunrise example in detail in Eqs. (21)-(24). The paper also contains closed-form expressions for elliptic leading singularities in Appendix A and a discussion of why the second-kind differential is crucial.","tokens_in":14884,"tokens_out":4356,"duration_ms":50661,"significance":"If the conjectured form (20) is correct, the paper offers a genuinely new structural result for elliptic Feynman integrals: a canonical, epsilon-factorized-up-to-A0 system whose solutions are pure functions, obtained without epsilon-dependent rescalings or a posteriori DE manipulation. The main positive features are the parameter-free rotation (17), the explicit closed-form representation of the eLS in Appendix A, the absence of fitted free parameters, and one fully explicit worked example. The principal limitation is that the universality claim is a conjecture, and the pure-function conclusion rests on an unproved simple-pole property whose validation in Appendix B is not conclusive. The strength of the evidence is therefore somewhat below the strength of the claimed structural theorem, which is a central consideration for the recommendation.","major_comments":[{"comment":"The pure-function conclusion rests on the statement that every entry of A in Eq. (20) has at most simple poles locally around every DE singularity. The validation in Appendix B pulls the DE back on a single random line gamma(t) and infers from the generic form (B2) that only simple poles are possible. This line check samples only generic points of each singular divisor; a double pole whose leading coefficient vanishes at the sampled point, or a pole supported at an intersection of divisors, would not be detected. Since the iterated-integral representation of the solutions is directly downstream of this property, please either prove the simple-pole statement for the constructed entries or verify it divisor-by-divisor in each worked example.","section":"Sec. III and Appendix B"},{"comment":"The empirical support for the factorized form (20) is largely deferred: only the unequal-mass sunrise is presented explicitly in Eqs. (21)-(24), while the bases for the other five examples are stated to be provided in supplementary material and a followup paper. A central claim that a new DE form appears in several state-of-the-art examples needs to be checkable within the manuscript or its verified supplementary files; otherwise the claim should be explicitly weakened to a single detailed example plus a conjecture.","section":"Sec. IV, Table I"},{"comment":"The derivation of the factorized structure is not shown in detail. The text states that (20) follows from the eLS rotation (17) and the coupled DEs (13)-(14), but the mechanism that guarantees (A0)_ij (A1)_ij = 0 and A0^2 = 0 is not exhibited. A short proof for the sunrise case, or at least a clear explanation of how the structure of (13) and (14) enforces the factorization, would substantially strengthen the paper's central observation.","section":"Sec. III, Eq. (20)"}],"minor_comments":[{"comment":"The entries '4+2' and '1+0' in the third and fourth columns are not explained in the text; please define the notation for the number of master integrals and the number of third-kind eLS on each cut.","section":"Table I"},{"comment":"There are several typographical spacing errors, including 'd logintegrands' in the introduction and 'The periodψi' in Section II; a careful proofreading pass is needed.","section":"Sec. I and Sec. II"},{"comment":"The phrase 'random line' should be replaced by a deterministic and reproducible choice, since a random line is not a well-defined verification procedure for a journal publication.","section":"Appendix B"},{"comment":"The constant coefficients c1,c2 in (B1) and the constants c_i in (B2) use overlapping notation for different objects; renaming one of the two sets would improve clarity.","section":"Appendix B, Eqs. (B1) and (B2)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a well-written and interesting progress report, and the factorized form (20) is a noteworthy observation. My main reservations are verifiability rather than novelty: the pure-function claim depends on an unproved pole-order property, and most of the six claimed examples are not shown in the text. I would be willing to support publication once the missing DEs are made available and the simple-pole check is made genuinely divisor-by-divisor, or the claims are scaled back accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a serious referee. The genuinely new pieces are the derivative-free construction of the second-kind one-form and the observed factorized form (20): A = (1 - n epsilon) A0 + epsilon A1 with (A0)_ij (A1)_ij = 0 and A0^2 = 0. If it holds up, this is the integrand-first route to canonical-form-like DEs for elliptic cases that the field has been missing, with no epsilon input and no post-processing. That is a real advance over the d-log integrand method and earlier elliptic leading-singularity work.\n\nThe paper does what a good letter should: the algorithm is concrete, the unequal-mass sunrise is worked out in detail, the period and quasi-period expressions in Appendix A are self-contained, and the construction is not circular. The rotation (17) is fixed from leading singularities before any DE is derived; no parameters are fitted to obtain (20). I also don't see a citation problem. The authors are appropriately honest: universality is presented as a conjecture, and the full bases for five of the six examples are not in the letter.\n\nThe soft spot is the simple-pole property. Appendix B validates it by pulling the DE back on a random line and inspecting the generic form (B2). That test samples finitely many points on each singular divisor. It would miss a double pole that appears only at an intersection of divisors, or a double pole whose leading coefficient vanishes along the sampled line. The local-cycle argument shows some combination of periods has a nonvanishing constant term, but that does not prove the actual fixed entries such as d(psi2/psi1) have no t^{-2} terms in every chart. Since the pure-function statement and the iterated-integral representation are directly downstream of this property, this is load-bearing. I would not call it fatal: the examples are consistent with the claim, and the paper labels the general statement a conjecture. But the next version or followup should address it head-on, with a divisor-by-divisor check or a proof.\n\nThe unexplained integer n is intriguing, not a flaw. The paper is for people working on multi-loop elliptic integrals, especially top-pair and heavy-quark loop processes, and for anyone building algorithmic canonical-DE tools. It deserves a serious referee: the core idea is new, the examples are nontrivial, and the authors separate what they prove from what they conjecture. Send it out; require that the supplementary material be available to referees, and ask for a sharper statement about the pole-order claim.","headline":"A genuinely new integrand-level route to canonical forms for elliptic Feynman integrals, with the simple-pole property that pure functions rely on still a conjecture with only a posteriori checks.","tokens_in":15367,"tokens_out":3472,"would_cite":true,"duration_ms":35150,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q30","14H52"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a derivative-free choice of algebraic one-forms of the second kind on an elliptic curve puts the associated Feynman integrals into a previously unreported differential-equation form whose epsilon expansions are pure…","keywords":["elliptic leading singularities","canonical differential equations","pure functions","algebraic one-forms","second-kind differentials","Feynman integrals","iterated integrals","elliptic curves"],"falsifier":"Choose one of the verified families, say the unequal-mass sunrise, and compute the local expansion of every entry of $A$ around a kinematic point where two roots of $P(z)$ coincide, using the cycle combination selected in Appendix B. If any entry of $A$ develops a double pole in the local coordinate — for example a term proportional to $z^{-2}\\,dz$ in $\\mathrm{d}(\\psi_2/\\psi_1)$ — then the Chen iterated-integral representation with simple-pole kernels fails and the pure-function claim is false; otherwise the check would strengthen the conjecture.","tokens_in":14425,"feed_emoji":"⚛️","tokens_out":12968,"duration_ms":124095,"temperature":0.7,"pith_summary":"The paper extends the genus-zero method of d log integrands with integer leading singularities to elliptic (genus-one) geometry. It proposes building a basis of Feynman integrands by matching the integrand, after localization to a last integration variable, to a fixed set of algebraic one-forms on the elliptic curve — the holomorphic period form, a second-kind form chosen without taking derivatives, and third-kind forms with unit residues — and then rotating the basis by the matrix of elliptic leading singularities. The authors find that the differential equations satisfied by the resulting integrals take a special, previously unreported form, $\\mathrm{d}G = ((1-n\\,\\epsilon)A_0+\\epsilon A_1)G$, with $A_0$ nilpotent, $A_0$ and $A_1$ with disjoint entrywise support, and all one-forms in $A$ having at most simple poles locally. As a consequence, the $\\epsilon$-expansion of the integrals is given order by order by Chen iterated integrals over simple-pole kernels, i.e. by pure functions. The construction is verified on the two- and three-mass sunrise graphs and several two-loop elliptic families, and the authors conjecture that it works universally for integrand bases of this type.","feed_headline":"New canonical form makes elliptic Feynman integrals pure functions","feed_subtitle":"Constructed without epsilon rescaling, the integrals expand in iterated integrals of simple-pole one-forms.","key_machinery":"The load-bearing object is the derivative-free algebraic one-form of the second kind, $\\omega_\\phi = N_\\phi(z)\\, dz/\\sqrt{P(z)}$, whose numerator $N_\\phi$ is chosen so that its period integrals equal the quasi-period $\\phi$ of the elliptic curve; together with the holomorphic one-form $\\omega_\\psi$ and the third-kind forms $\\omega_{\\pi_a}$, it forms the cohomology basis of eqs. (10)–(11). These one-forms are integrated over the two cycles of the elliptic curve to give elliptic leading singularities (eLS), whose period matrix (12) is then used in the rotation (17) to define canonical integrands with unit eLS on a preferred cycle. The special differential-equation structure (20) — in particular the disjoint-entry condition $(A_0)_{ij}(A_1)_{ij}=0$ and $A_0^2=0$ — is what allows the solution to be written as Chen iterated integrals with simple-pole kernels.","core_discovery":"The central claim is that the integrand-level construction of Sec. II, using algebraic one-forms of the second kind without derivatives, produces a basis whose differential equations take the special form $\\mathrm{d}G = A G$ with $A=(1-n\\,\\epsilon)A_0+\\epsilon A_1$, $(A_0)_{ij}(A_1)_{ij}=0$, $A_0^2=0$, and all one-forms with at most simple poles, so solutions are pure functions. The matrix $A_0$ is built from the exact differentials of the period-matrix entries $\\mathrm{d}(\\psi_2/\\psi_1)$ and $\\mathrm{d}(\\pi_{a,2}-\\pi_{a,1}\\psi_2/\\psi_1)$, so it encodes the elliptic leading singularities themselves; $A_0$ is strictly upper triangular and nilpotent. The integer $n$ is observed in every example but its origin is unexplained. The authors conjecture that any Feynman integral family that can be brought to the integrand form of eq. (16) and rotated as in eq. (17) satisfies this same special differential equation.","pith_inferences":["If the conjecture holds, the special form (20) could be used as a fast diagnostic: scanning integrand ansätze for the combination of unit eLS, disjoint support of $A_0$/$A_1$, and simple poles would identify pure elliptic sectors without a full integration-by-parts reduction.","The paper's observation that replacing $\\omega_\\phi$ by a derivative of $\\omega_\\psi$ destroys the structure suggests that the second-kind form should be chosen as an algebraic object, not a derived one; a natural extension is to prove (20) directly from the Gauss–Manin connection of the elliptic curve, without computing any integrals.","A testable boundary of the conjecture lies in higher-genus or higher-dimensional cases: applying the same integrand-level recipe to a three-loop banana or a K3/Calabi–Yau sector would show whether the simple-pole and disjoint-support properties persist beyond genus one.","The component-wise orthogonality $(A_0)_{ij}(A_1)_{ij}=0$ implies that in the $\\epsilon$-expansion, no iterated integral receives simultaneous contributions from both matrices at the same index, which may lead to simpler analytic-continuation formulas than generic Fuchsian systems."],"forward_implications":["For any family in the construction, the $\\epsilon$-expansion can be written directly as iterated integrals of explicit simple-pole one-forms, so no $\\epsilon$-factorization or differential-equation post-processing is needed at higher orders.","The same basis provides an $\\epsilon$-factorized form if desired: fully diagonalizing the period matrix (18) removes the homogeneous $A_0$ block, at the cost of breaking the explicit covariance under changing the preferred cycle.","The observed but unexplained integer $n$ in $(1-n\\,\\epsilon)$ becomes an invariant of each elliptic family; determining its meaning is posed as an open problem that may organize which integrals admit this form.","Because the construction is algorithmic and uses only integrand data and closed-form elliptic leading singularities, it is suited to automation and to multi-scale applications such as top-pair and diphoton production with massive loops."],"supporting_citations":[{"why":"Establishes the canonical-form differential equations in the genus-zero case that this work generalizes to elliptic geometry.","marker":"[10]"},{"why":"Supplies the d-log integrand construction and LS equations that the elliptic construction extends to nontrivial homology.","marker":"[55]"},{"why":"Provides the conventions for the algebraic one-form basis, period matrix, and Legendre identity used in Sec. II.","marker":"[62]"},{"why":"Introduces elliptic leading singularities in the sYM context and motivates the rotation toward canonical integrands.","marker":"[31]"},{"why":"Representative of derivative-based procedures for epsilon-factorized DEs beyond polylogarithms that the paper contrasts with its derivative-free construction.","marker":"[44]"},{"why":"Defines pure functions for elliptic Feynman integrals and supplies the cohomology basis reminiscent of the one chosen here.","marker":"[54]"},{"why":"Supports the argument that solutions are pure functions and that the presence of $A_0$ does not obstruct an iterated-integral representation.","marker":"[56]"},{"why":"Justifies the choice of period combinations with convergent power series and nonvanishing constant term used in Appendix B.","marker":"[52]"},{"why":"One of the verified two-loop example families, also the source of the rescaling-homogeneity observation for the one-forms.","marker":"[66]"}],"fun_headline_variants":["Elliptic integrands yield pure functions via nilpotent A0","Simple-pole one-forms give elliptic canonical differential equations","New elliptic canonical form avoids epsilon and gives pure results","Leading singularities encode elliptic canonical form directly","Nilpotent A0 yields pure functions from elliptic integrands"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The pure-function conclusion rests on the claim that every entry of $A$ has at most simple poles locally around each singular point; this is verified only by pulling the differential equation back on one generic line $\\gamma(t)$ and inspecting the generic form (B2), not by a proof covering arbitrary paths or all singular-point configurations.","fun_headline_variants_meta":{"raw":{"variants":["Elliptic integrands yield pure functions via nilpotent A0","Simple-pole one-forms give elliptic canonical differential equations","New elliptic canonical form avoids epsilon and gives pure results","Leading singularities encode elliptic canonical form directly","Nilpotent A0 yields pure functions from elliptic integrands"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00051,"raw_usage":{"total_tokens":2462,"prompt_tokens":903,"completion_tokens":1559,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":1479}},"tokens_in":519,"tokens_out":1559,"duration_ms":11770,"temperature":1.0,"reasoning_tokens":1479,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:16:43.638608+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose one of the verified families, say the unequal-mass sunrise, and compute the local expansion of every entry of $A$ around a kinematic point where two roots of $P(z)$ coincide, using the cycle combination selected in Appendix B. If any entry of $A$ develops a double pole in the local coordinate — for example a term proportional to $z^{-2}\\,dz$ in $\\mathrm{d}(\\psi_2/\\psi_1)$ — then the Chen iterated-integral representation with simple-pole kernels fails and the pure-function claim is false; otherwise the check would strengthen the conjecture.","supporting_citations":[],"review_version":1}