{"id":"36e6817a-d8a6-4943-bbce-45294b4e9bb0","arxiv_id":"2603.03183","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Multiloop sunset integrals in D=2 are expressed as convergent sums of symmetric polynomials in logarithms of mass ratios, with a dimension-raising operator that propagates the result to D=4-2ε.","lead":"This paper gives exact, convergent series formulas for multiloop 'sunset' Feynman diagrams in two spacetime dimensions, for any masses and any loop order, plus a differential rule that lifts those results toward four dimensions. These diagrams appear in high-precision particle physics calculations, where they usually require elliptic functions and other hard special functions.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Exactness of Theorem 2.1 rests on an unproved Mellin–Barnes contour-closing estimate; eq. (25) also shows a sign discrepancy that should be checked.","rationale":"The reader’s weakest-assumption analysis identifies the contour-closing exactness as the key unproven step. I agree: the theorem’s headline claim is exactness, and the available proof says only that the Γ-functions ensure exponential decay, without the arc estimates needed to exclude residual contributions. This is a genuine gap, not a stylistic preference, because the integrand contains power factors that can compete with the Γ-function decay. I also noticed a possible sign error in eq. (25) (Γ(-n+z) vs the Γ(-n-shift) that actually appears in eq. (17)); if the printed formula is used, it could produce incorrect coefficients, though the repository implementation may well be correct. Both issues are checkable. The reader’s CONDITIONAL verdict is appropriate: the paper’s central construction is plausible and independently supported by known two-loop results and an available repository, but the exactness claim is not yet rigorously established. I do not see grounds to reject outright, because the sign issue may be typographical and the arc-bound proof is likely routine. I therefore leave the verdict unchanged.","tokens_in":21353,"tokens_out":18799,"duration_ms":150215,"concrete_test":"Independently derive Stirling-type bounds for the integrand in eq. (14) along the closing arcs of §2.2 (Re ζ → −∞, Re z_i → +∞), explicitly tracking the growth of (-p^2/m_{L+1}^2)^{1+ζ} and (m_i^2/m_{L+1}^2)^{z_i}. If the arc integrals do not vanish, Theorem 2.1’s exactness claim is false. As a complementary check, for L=2 with masses (1,2,3) and p^2 = −36.01, truncate eq. (18) at total order n=20 and compare with high-precision numerical integration of the two-loop sunset; disagreement would falsify the exactness claim directly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Theorem 2.1, eq. 18) is that the series converges absolutely for |p^2| > (m_1+...+m_{L+1})^2 and equals the integral exactly, not just asymptotically. The proof in §2.2 closes the ζ and z_i contours because 'the Γ-functions ensure exponential decay' (Fig. 2), but no arc-bound estimate is supplied. This is load-bearing: the integrand in eq. (14) also contains factors (-p^2/m_{L+1}^2)^{1+ζ} and (m_i^2/m_{L+1}^2)^{z_i}, which can grow exponentially in the relevant half-planes, and the relevant poles are double poles whose residues involve derivatives. If any arc contribution survives, eq. (18) is at best an asymptotic expansion, contradicting Remark 2.1 and undermining the later all-equal-mass and dimension-shifting results. A secondary red flag: eq. (25) states lim_{z→0} d^k/dz^k Γ(1+n+z)/Γ(-n+z) = n!^2 c_k(n), but the ratio that actually appears in eq. (17) near z_i = r_i is Γ(1+n+shift)/Γ(-n-shift). The printed sign differs; for n=1,k=1 the two give c_1(1)=+1 vs -1. This may be a typo, but it must be resolved before the coefficient formula eq. (19) is trusted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims exact, convergent large-momentum expansions for the L-loop sunset integral in D=2, for arbitrary masses (Theorem 2.1, eq. (18)) and for equal masses (Theorem 4.1, eq. (50)), plus a dimension-raising relation to D+2 (Section 5). The generic-mass result is derived via a Mellin–Barnes representation and residue sums, leading to an expansion in symmetric polynomials of logarithms with coefficients c_k(n). The equal-mass result is expressed through a flat coordinate R^{(L)}(p^2), an instanton-type series, and a Frobenius basis, with the claimed boundary coefficients generated by eq. (51). The four-dimensional application combines these two-dimensional results with Tarasov-type dimension-shifting operators.","tokens_in":21743,"tokens_out":6996,"duration_ms":73709,"significance":"If the exactness and convergence claims hold, this is a substantial result: it gives the first all-order exact expansion of sunset-type Feynman integrals, connects the expansion to mirror-symmetric structures, and provides an operational route from D=2 to D=4. The paper ships SageMath/Maple code in a repository and benchmarks several low-loop cases against known differential equations and published results; these concrete checks are a genuine strength. The main theorems, however, rest on a contour-closing argument that is not fully proved and on a coefficient identity whose printed form has a sign discrepancy. Because the central claims are plausible and likely repairable, the appropriate outcome is a major revision rather than rejection.","major_comments":[{"comment":"The exactness of Theorem 2.1 depends on the vanishing of all arc contributions when the ζ- and z_i-contours are closed. The paper states that ‘the Γ-functions ensure exponential decay’, but no arc-bound estimate is given. The integrand contains factors (-p^2/m_{L+1}^2)^{-(1+ζ)} and (m_i^2/m_{L+1}^2)^{z_i}, which can grow exponentially in the relevant half-planes, and the poles are double poles whose residues require derivatives. If any arc contribution survives, eq. (18) is at best an asymptotic expansion, contradicting Remark 2.1. A rigorous estimate, or at least a precise statement of the integration contours and a dominated-convergence argument for the resulting series, is load-bearing for the paper’s central claim.","section":"§2.2, eq. (14), Figure 2"},{"comment":"The coefficient identity is printed as lim_{z→0} d^k/dz^k [Γ(1+n+z)/Γ(-n+z)] = n!^2 c_k(n). For n=1, k=1 the left-hand side evaluates to -1, while the right-hand side gives +1 if c_1(1)=1 as stated. This sign discrepancy affects eq. (19) and hence the coefficients in eq. (18). It may be a typographical sign in the denominator (e.g. Γ(-n-z) instead of Γ(-n+z)), but the error must be resolved and the corrected formula checked against the one- and two-loop cases before the theorem can be trusted.","section":"Eq. (25), proof of Theorem 2.1"},{"comment":"The proof of the all-equal-mass representation is a sketch. The step ‘changing variables from p^2 to the flat coordinate ... yields the representation in equation (50)’ is asserted without showing how the coefficients d_r^{(L)}(l) arise or why the resulting series has the stated convergence. The coefficients are supplied by code and by tables in Appendices A–B rather than by a closed formula or a recurrence; the generating function (51) is stated but not derived in enough detail to verify all orders. As written, the all-order claim of Theorem 4.1 is not fully proven. A complete derivation, or at least a precise combinatorial/algebraic construction of d_r^{(L)}(l), is needed.","section":"§4.2, proof of Theorem 4.1"},{"comment":"The dimension-raising relation is central to the paper’s four-dimensional application, but the all-order statement is again asserted rather than proved. The text says the operators are determined because their coefficients are polynomials of known degree and because differential equations up to 20 loops are known from [40]; this does not establish a theorem for arbitrary L. The uniqueness of the polynomial solution is not demonstrated, and the reduction modulo L^{D,(L)}_⊖ is described only schematically. If the paper’s claim is that this is a proof for all L, the missing step needs to be supplied; if instead it is an algorithm/conjecture supported by low-loop checks, that should be stated explicitly.","section":"§5, eqs. (66)–(78)"}],"minor_comments":[{"comment":"The abstract says ‘dimension-lowering relation’, but Section 5 derives and consistently calls it a ‘dimension-raising’ formula. Please align the terminology.","section":"Abstract vs. Section 5"},{"comment":"The one-loop relation is labelled I^{D=4,(2)}_⊖, which conflicts with the L=1 context. This appears to be a typo, but it should be corrected to avoid confusion.","section":"§5.2, eq. (84)"},{"comment":"There are several typos: ‘availaible’ in the introduction, and ‘and and’ in the acknowledgements. Please proofread.","section":"Throughout"},{"comment":"The notation in the generating function for α_r^{(L)} is compressed: x, λ, and ζ(2k+1) enter without an explicit statement of the region of convergence or how the coefficient extraction is ordered. A short explanation would improve readability.","section":"§4, eq. (51)"},{"comment":"The formula for c_k(m) is hard to read because k, n, and m are all used nearby. At minimum, define m as the total degree r_1+...+r_{L+1} explicitly in the theorem statement, and avoid reusing n inside the sum.","section":"Eq. (19)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a credible and valuable contribution if the proof gaps are filled. The sign issue in eq. (25) is a concrete red flag that should be checked against explicit low-loop integrals before publication. The reliance on the GitHub repository for the all-order coefficients is also a concern: the referee could not audit the code, so the paper should state which results are theorems and which are computationally verified up to a given loop order. I would be comfortable with acceptance after these points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a serious paper with a real new result, and it deserves refereeing. Theorem 2.1 — the all-loop, generic-mass exact expansion for the D=2 sunset — is the kind of thing people will want to use. The derivation from Mellin–Barnes is coherent, the expansion matches external benchmarks, and the companion code means the formulas can be tested immediately. That is genuine progress, not just a conjecture.\n\nWhat I'd want fixed before I trust every line:\n\n1. The exactness claim in Theorem 2.1 depends on closing the Mellin–Barnes contours with no residual arcs. The paper says 'the Γ-functions ensure exponential decay' and shows a figure, but gives no bound. The integrand in (14) also has powers (-p^2/m^2)^{1+ζ} and (m_i^2/m^2_{L+1})^{z_i}, which grow exponentially in the relevant half-planes. If an arc contribution survives, (18) is asymptotic, not exact. This is the load-bearing step and needs a real estimate, or a citation to a theorem that covers it.\n\n2. Equation (25) looks off. It states lim_{z→0} d^k/dz^k Γ(1+n+z)/Γ(-n+z) = n!^2 c_k(n), but the ratio in (17) has denominator Γ(-z_1-...-z_{L+1}), which near the pole is Γ(-n-Σ u_i), not Γ(-n+Σ u_i). For n=1,k=1 the two choices differ in sign. This may be a typo, but since the c_k(n) are the coefficients of the main expansion, it needs to be resolved before (19) is trusted.\n\n3. The equal-mass section (Theorem 4.1) is more sketch than proof. The d_r(L)(l) coefficients are supplied by code and tables rather than a closed formula, and the generating function (51) does not visibly reproduce the printed α values. That section is credible but not yet demonstrated to all orders.\n\n4. The dimension-raising operator in Section 5 is derived by an algorithm rather than proved; the displayed one-, two-, three-loop results are concrete and checkable, so this is fixable.\n\nBottom line: Theorem 2.1 is likely right and important; the rest is promising but needs strengthening. The paper should go to a good referee who knows Mellin–Barnes and Calabi–Yau methods. I'd read a revised version carefully.","headline":"All-loop sunset representation is a real step forward, but exactness rests on an unproved contour-closing step and a sign in the coefficient formula needs checking.","tokens_in":22184,"tokens_out":4231,"would_cite":true,"duration_ms":35075,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Multiloop sunset Feynman integrals in two dimensions admit exact all-order expansions built from symmetric polynomials of logarithms, with coefficients computable to every loop order.","keywords":["sunset integral","multiloop Feynman integrals","Mellin-Barnes representation","symmetric polynomials","mirror symmetry","Picard-Fuchs equation","dimension-shifting relations","two-dimensional quantum field theory"],"falsifier":"Choose a concrete unequal-mass sunset at, say, three loops with masses (1,2,3,4) and large Euclidean p², compute the partial sums of the series in Theorem 2.1 to high order, and compare numerically to a direct high-precision evaluation of the original Feynman parameter integral; any discrepancy at the 10^{-40} level would falsify the exactness claim. Alternatively, numerically estimate the Mellin-Barnes integrand of equation (14) on large arcs |ζ|=R, |z_i|=R with R→∞; a nonvanishing arc contribution would show the residue computation misses a residual term and the equality fails.","tokens_in":21269,"feed_emoji":"🌅","tokens_out":3384,"duration_ms":37572,"temperature":0.7,"pith_summary":"The paper aims to prove that the multiloop sunset Feynman integral in two spacetime dimensions, for any number of loops and any masses, equals a convergent infinite series rather than only an asymptotic approximation. Each term is an explicit rational coefficient times powers of mass ratios times symmetric polynomials in logarithmic mass-ratio factors, so the whole transcendental content is captured by logarithms and symmetric polynomials. If true, this gives a practical, high-precision way to evaluate sunset integrals without elliptic integrals or iterated polylogarithms. For equal masses the paper further derives a closed expression using a flat coordinate, a holomorphic period, and a Frobenius basis, with coefficients built from odd zeta values and integers. It also establishes a dimension-raising relation that expresses the D+2-dimensional equal-mass sunset integral from the D-dimensional one through a differential operator of order L-1, enabling reconstruction of four-dimensional results from two-dimensional ones.","feed_headline":"Exact all-loop formula for sunset Feynman integrals","feed_subtitle":"In two dimensions, the integrals reduce to sums of symmetric log polynomials with explicit coefficients — no elliptic transcendents needed.","key_machinery":"The argument is carried by a Mellin-Barnes representation of the sunset integral, followed by contour closure in (L+1) complex variables. Multiple poles are extracted by residue derivatives; Faà di Bruno's formula and complete Bell polynomials organize the derivatives of the ratio of Gamma functions, producing the symmetric polynomials P^k_{L+1} and the logarithmic factors ℓ_i(r). In the equal-mass case, the central objects are the flat coordinate R^{(L)}(p²), the holomorphic period π^{(L)}_⊖ (whose derivative gives the period), and the Frobenius basis Frob^{(L)}_r(p²); the structure of the inhomogeneous Picard-Fuchs equation with constant source term −(L+1)! makes the all-order solution tra","core_discovery":"The central claim is Theorem 2.1: for |p^2| greater than the square of the sum of the masses, the L-loop sunset integral in two dimensions has the exact convergent expansion I^{(L)}_⊖(p²,m²) = −(1/p²) ∑_{(r_i)} [(r_1+...+r_{L+1})!²/(r_1!²...r_{L+1}!²)] ∏ (−m_i²/p²)^{r_i} ∑_k c_k(r_1+...+r_{L+1}) P^{L+1−k}_{L+1}(ℓ_i(r_i)), where ℓ_i(r)=log(−m_i²/p²)−2∑_{n=1}^r 1/n and P^k_{L+1} are elementary symmetric polynomials. The coefficients c_k are explicit finite sums involving complete Bell polynomials and polygamma values. The paper states this series represents the exact value of the integral, not just an asymptotic expansion. For equal masses, Theorem 4.1 gives an equivalent expression in terms o","pith_inferences":["Editorial inference: The same Mellin-Barnes residue method may yield exact convergent series for other convergent massive Feynman integrals in two dimensions, such as multiloop irreducible graphs with more than two external legs; the sunset is only the simplest case where the contour-closing structure is fully explicit.","Editorial inference: If the exactness claim holds, the two-dimensional results provide a practical boundary condition for a wider program of dimension-shifting for general Feynman integrals, not only the sunset family, potentially reducing four-dimensional integral evaluations to two-dimensional convergent series plus derivatives.","Editorial inference: The explicit symmetric-polynomial form suggests a combinatorial interpretation of the coefficients as complete homogeneous symmetric functions of harmonic numbers; this could be tested by checking whether the same coefficient structure appears in other D=2 scalar integrals with different propagator powers.","Editorial inference: A direct numerical test near the convergence boundary |p²|=(Σm_i)² would probe whether the series can be analytically continued beyond the stated radius, which would indicate whether the exactness property extends by continuation or fails at the normal threshold."],"forward_implications":["Sunset integrals in two dimensions can be evaluated numerically to arbitrary precision by summing an explicit, convergent series, without elliptic integrals or iterated polylogarithms.","The exact all-order expansion confirms and strengthens the general asymptotic theorem for convergent Feynman integrals: for the sunset case the large-momentum expansion is not merely asymptotic but exact, with r_max equal to the loop order.","For equal masses, the all-loop expression makes the mirror-symmetry structure explicit: coefficients are integers, rational numbers, and odd zeta values, and the leading-log terms match the Gamma-class prediction of the associated Fano variety.","The dimension-raising relation gives a systematic algorithm for computing the epsilon expansion of the four-dimensional sunset integral from the two-dimensional integral, including finite parts, using only derivatives and rational/polynomial coefficients.","The explicit coefficients, implemented in code, make the results available for checking lower-loop computations and for formal analysis of higher-loop behavior in the large-momentum regime."],"fun_headline_variants":["Exact 2D multiloop sunset integrals","All-order sunset formula in two dimensions","Sunset integrals exact in 2D, no transcendents","Multiloop sunset: convergent exact series","2D sunset integrals: exact and explicit"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The exactness of the series rests on the claim that the Mellin-Barnes integrand decays sufficiently rapidly on the chosen closing arcs so that no residual contribution survives; the paper states that the Gamma functions ensure exponential decay but does not provide the arc-bound estimate, and if that estimate fails the expansion would be only asymptotic rather than exact.","fun_headline_variants_meta":{"raw":{"variants":["Exact 2D multiloop sunset integrals","All-order sunset formula in two dimensions","Sunset integrals exact in 2D, no transcendents","Multiloop sunset: convergent exact series","2D sunset integrals: exact and explicit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000213,"raw_usage":{"total_tokens":1260,"prompt_tokens":751,"completion_tokens":509,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":495,"completion_tokens_details":{"reasoning_tokens":438}},"tokens_in":495,"tokens_out":509,"duration_ms":5477,"temperature":1.0,"reasoning_tokens":438,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T19:09:14.848027+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose a concrete unequal-mass sunset at, say, three loops with masses (1,2,3,4) and large Euclidean p², compute the partial sums of the series in Theorem 2.1 to high order, and compare numerically to a direct high-precision evaluation of the original Feynman parameter integral; any discrepancy at the 10^{-40} level would falsify the exactness claim. Alternatively, numerically estimate the Mellin-Barnes integrand of equation (14) on large arcs |ζ|=R, |z_i|=R with R→∞; a nonvanishing arc contribution would show the residue computation misses a residual term and the equality fails.","supporting_citations":[],"review_version":1}