{"id":"e9525387-e5fe-4db0-af48-22a3d717eea8","arxiv_id":"2412.00793","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Two-loop master integrals for e+e−→μ+μ− that require an electron-mass regulator are computed as a Frobenius expansion in the electron mass.","lead":"This paper computes a set of two-loop Feynman integrals needed for precise predictions of electron-positron collisions producing muon pairs, with the electron mass kept small but nonzero. The results are needed to handle collinear divergences in the next-to-next-to-leading-order QED cross section.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's scope rests on an unproven classification: it is asserted, without proof or citation, that the integrals needing electron-mass dependence are exactly the set (b) families of Fig. 1, and that the two Fig. 2 families cover all of set (b).","rationale":"The computation itself is internally coherent. The Frobenius exponents in (2.12) and the two logarithmic sectors (k=1 for λ=−2ϵ,−ϵ) are consistent with the LSS expansion algorithm for a non-diagonalizable residue at m²=0; the β-integration path at fixed θ avoids all zeroes of the alphabet letters (1∓βc vanish only at β=±1/c, outside (0,1) for physical cosθ, and 1+2βc+β² has complex roots for |c|<1); and the boundary-constant strategy (β→0, β→1 at θ=π/2, plus the dedicated j41 subsystem) is a coherent way to close the system. I found no internal inconsistency sufficient to question the master-integral results. The residual risk sits at two ends: verification depth for the hardest integrals, and scope completeness. For j48–j51, the FIESTA check was performed only at d=6−2ϵ via dimensional recurrence; this is a genuine, non-circular test (errors in the d=4−2ϵ expressions would generically propagate to d=6−2ϵ and be detected), so the likelihood of an undetected error is low, though the check is indirect. For scope, the abstract (“We consider only those families for which the account of the electron mass m is necessary”) and the Conclusion both rest on the “It can be shown” assertion of Section 1, which is given without proof or citation, and the mapping from the diagram sets of Fig. 1 to the two integral families of Fig. 2 is asserted, not demonstrated. Because this claim determines whether the paper's deliverable is complete for its stated physical purpose, and because a concrete computational test can settle it, it is the most load-bearing concern. The reader's weakest_assumption identified the same point; I adopt it and sharpen it to its two components (collinear classification, and family coverage of set (b)). The verdict stays CONDITIONAL: correctness of the presented integrals is supported by method and cross-checks, but the scope claim should be backed by the classification check, and the validation statements preferably quantified with the achieved numerical precision.","tokens_in":6346,"tokens_out":25313,"duration_ms":228834,"concrete_test":"Settle the scope claim directly: enumerate all two-loop four-point topologies for e+e−→μ+μ− (qgraf or FeynArts), classify them into the gauge-invariant sets (a)–(d) of Fig. 1, IBP-reduce every topology against the known set (a) families plus the two Fig. 2 families, and compute the leading m→0 behavior of all resulting master integrals with the same Frobenius/LSS method used in the paper (or by sector power-counting). If any topology outside set (a) and the two set (b) families acquires a logarithmic or collinear-pole-regulated dependence on m, or if any diagram of set (b) fails to reduce to the two families, the completeness claim fails and the computed set is incomplete. A cheaper partial proxy: locate the collinear poles in the published massless two-loop amplitude (Ref. [8]) and verify they sit only in the (a) and (b) contributions.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central deliverable — the master integrals whose electron-mass dependence is needed for the NNLO e+e−→μ+μ− cross section — is gated by a two-part completeness claim. Section 1 states: “It can be shown that the collinear divergences appear only in the set(a) with le > 0 and in the set(b), i.e., in the sets of diagrams where at least one photon line connects points on the electron line.” No proof or citation accompanies this. Part 1 is the collinear classification: every two-loop topology with non-analytic m→0 behavior must have a photon attached to the electron line (beyond the known set (a) corrections), and every such diagram does develop such behavior. Part 2 is the coverage statement in Section 2: “The diagrams of the set(b) on Fig. 1 are expressed in terms of the integrals of two big families depicted in Fig. 2” — namely that the planar and non-planar two-photon-exchange topologies exhaust set (b). The figure-based classification (sets (a)–(d), with (d) labeled only “permutations”) does not establish either statement, no enumeration of set (b) diagrams is given, and no IBP demonstration that each such diagram reduces to the 61 masters of the two families is shown. If any topology outside the two families also develops ln(m) behavior — the paper's own method detects this via the logarithmic sectors of the Frobenius exponents — then the advertised mass-dependent NNLO cross section would miss contributions. The claim is plausibly correct by standard collinear power counting, so this is neither a consensus dispute nor an internal inconsistency; it is a load-bearing assertion that is currently unsupported. Secondary concerns (FIESTA validation of j48–j51 only via dimensional recurrence at d=6−2ϵ; deep om=oϵ=6 files “available from the author by request”) do not displace this as the primary risk.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents the calculation of 61 two-loop master integrals for the process e+e−→μ+μ−, restricted to the families for which the electron mass m must be retained. The author argues in §1 that only the diagrams of set (a) with electron-line loops and set (b) of Fig. 1 develop collinear divergences, and that the latter reduce to the two planar/non-planar families of Fig. 2, whose integrals are defined by a single nine-denominator basis (2.4)–(2.5). The method is: LiteRed IBP reduction (61 masters), differential equations in m2, s, t; transformation to a normalized Fuchsian form at m2=0; construction of the evolution operator as a Frobenius expansion with exponents {0, 1/2−2ϵ, −4ϵ, −3ϵ, −2ϵ, −ϵ}; determination of boundary constants from the m→0 asymptotics at β→0, from β→1 at θ=π/2, and from a dedicated s→4m2 calculation for the single remaining constant; reduction of the s,t system to an ϵ-form with an 8-letter d log alphabet; and final expression of the canonical basis in Goncharov polylogarithms of β. The expansion reaches O(m^13) in m and O(ϵ^6) in ϵ, with 15 linear constraints leaving 46 independent entries. Cross-checks against FIESTA are reported for m=1/2, with the deepest integrals j48–j51 checked at d=6−2ϵ via dimensional recurrence.","tokens_in":6670,"tokens_out":24888,"duration_ms":231924,"significance":"The calculation, if correct, provides a key ingredient for the mass-dependent NNLO differential cross section of e+e−→μ+μ−, turning the collinear divergences of the massless master integrals into logarithms of m. Methodologically, the paper extends the Frobenius-expansion technique of Ref. [9] to a four-scale problem and fixes boundary constants from several independent asymptotic limits (β→0, β→1 at θ=π/2, and s→4m2 for the j41 sector) rather than from a single DRA calculation; the argument that the chopped U and U−1 still yield the exact matrices M_s and M_t via m-independence of c is sound. The paper ships machine-readable substitution files and a numerical example notebook, and the series are cross-checked against the independent program FIESTA, so the central results are open to verification rather than resting on a fitted ansatz. The main risks are the unsupported completeness classification that defines the scope and the indirect numerical check of the deepest integrals j48–j51.","major_comments":[{"comment":"The scope of the paper rests on a two-part completeness claim that is neither proved nor referenced. In §1 the author states \"It can be shown that the collinear divergences appear only in the set(a) with le > 0 and in the set(b)\", and in §2 that \"The diagrams of the set(b) on Fig. 1 are expressed in terms of the integrals of two big families depicted in Fig. 2\". No power-counting argument, citation, enumeration of the set (b) diagrams, or IBP demonstration that each such diagram reduces to the 61 masters of (2.4)–(2.5) is given. Since the advertised physical application — the mass-dependent part of the NNLO e+e−→μ+μ− cross section — is complete only if these two claims hold, this is load-bearing. I request either a citation to a published proof of the collinear classification or a brief argument (standard soft-collinear power counting), together with an explicit list of the set (b) topologies and their reduction (e.g., a table of sector mappings).","section":"§1 (Introduction) and §2"},{"comment":"The numerical validation is reported only qualitatively: \"convincing agreement\" at m=1/2 for the bulk of the integrals, and a comparison at d=6−2ϵ for j48−j51 via the dimensional recurrence, because reliable FIESTA results at d=4−2ϵ \"were not obtained\". No precision (number of matching digits per integral) is stated. Given that j48−j51 are the most complicated integrals and are checked at a different dimension from the one at which the results will be used, I ask for the achieved accuracy to be quantified and for at least one additional independent check at d=4−2ϵ (for instance, FIESTA with increased sector depth/precision, or evaluation of the dimensionally-recurred relations at several values of ϵ).","section":"§3 (Cross checks)"},{"comment":"The paper advertises the Frobenius expansion up to O(m13) and the ϵ-expansion to order 6, but the attached files JtoK2.m and KtoG4.m contain only the shallow expansions om=2 and oϵ=4; the deeper JtoK6.m and KtoG6.m are \"available from the author by request\". Since the series coefficients are the deliverable of the paper, the full-order results should be shipped as ancillary files (or included in a supplementary archive) so that the central claim of §2 and §4 can be verified by the reader.","section":"§3 (Results and files)"},{"comment":"The 15 constraints C·K=0 are asserted without the vectors C or any derivation of their origin, and the statement that they leave 46 independent entries is not automatic because K is related to the 61 master integrals by invertible transformations: exact constant-coefficient constraints among the entries of K would imply linear dependence among the master integrals themselves (most plausibly discrete symmetries of the family that the IBP reduction did not quotient out). Please clarify whether the constraints are exact identities, list the vectors C or otherwise provide the reduction map from 61 to 46 entries, and specify precisely how compatibility with the differential equations and boundary constants was checked.","section":"§2 (Eq. (2.25))"}],"minor_comments":[{"comment":"Typos and typesetting issues: the title reads \"T wo-loop\"; §1 has \"Out approach\"; §3 has \"we present out results\" and \"our results foK\"; Eq. (3.2) contains the stray \"omX\"; and several occurrences of O(m13) lack the superscript. These should be corrected in a final version.","section":"Throughout"},{"comment":"The identification of the boundary constants fixed in each limit is incomplete: the reader is told that four constants were not fixed by β→0 and that three of them were found from β→1 at θ=π/2, but only j41 is named. A short table listing the unfixed constants and the limit that determines each one would make the boundary procedure reproducible.","section":"§2"},{"comment":"The path-ordered exponential Pexp is used without definition; please define it (or the equivalent evolution-matrix notation) and state that the straight-line path from (0,θ) to (β,θ) avoids the zeros of the alphabet (2.22) in the physical region 0<β<1, |c|<1.","section":"§2 (Eq. (2.24))"},{"comment":"Given that the physical application requires joining the present mass-dependent results to the massless master integrals of Refs. [5–7], a brief remark on how the new families reduce to those integrals in the appropriate sector limits (e.g., which of the 61 masters reproduce the known m=0 integrals) would help orient the reader.","section":"§1"}],"recommendation":"major_revision","confidential_remarks":"The main request beyond the written report: the completeness classification in §1 is presented as folklore (\"It can be shown\") without a citation. If there is a standard reference for the collinear-divergence classification in two-loop QED (e.g., in the literature on e+e−→μ+μ− or Bhabha scattering at NNLO), the author should be asked to cite it. Also, the gap between the advertised O(m13) order and the attached shallow files is the kind of thing that will draw criticism from data-focused readers; the editor may wish to request the deep expansions as part of the revision. No concerns about citation practice: the self-citations are to the author's own widely used public software (LiteRed, Libra) and to his prior work on the same technique."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it computes the electron-mass-dependent expansions of the two-loop master integrals for set (b) of e+e- to mu+mu-, the one place where massless results produce collinear logarithms that cannot be treated otherwise. The result is new; the massless versions in Refs. [5-7] do not contain these. The method is standard but executed carefully: Frobenius expansion around m^2=0, differential equations in the coefficients, boundary conditions fixed from several kinematic limits instead of the heavier DRA approach, and a final canonical d-log form with GPLs. The effort to find 15 rational constraints and reduce 61 masters to 46 independent entries is a nice touch. Cross-checks against FIESTA are convincing, and the deep m-expansion means agreement even at m=1/2. This is honest, technically demanding work.\n\nThe soft spots are real but not fatal. The biggest is the scope claim, asserted in Section 1 without proof or citation: collinear divergences appear only in set (a) with le>0 and in set (b), and the two families of Fig. 2 exhaust set (b). If any other two-loop topology develops ln(m) behavior, the advertised NNLO cross section would miss it. The claim is plausibly correct by standard power counting, and the author has a long track record in these classifications, but for a paper whose entire value proposition is \"these are the integrals you need,\" the reader deserves a derivation or at least a diagram enumeration plus an IBP statement that all set (b) diagrams reduce to the two families. As written, it is a load-bearing assertion.\n\nThe other concerns are minor. The deepest expansions (om=oe=6) are kept off the arXiv and available \"by request\"; that is inconvenient for reproducibility, though the attached shallow files are enough for the planned application. The FIESTA check for j48-j51 was done at d=6-2epsilon via dimensional recurrence rather than at d=4-2epsilon; that weakens the check slightly but is a practical and reasonable compromise for the hardest integrals.\n\nCitation pattern is clean: heavy self-citation points to established software and the author's own previous method paper, none of which is problematic. If I am forced to pick the main risk, it is the completeness claim, not the algebra.\n\nMy recommendation: send it to a serious referee. The referee should ask the author to prove or carefully justify the collinear classification and the two-family coverage. If that holds up, this is a publishable and useful calculation for the precision QED community.","headline":"A solid, genuinely new two-loop massive calculation whose only real weakness is an unproven claim about which diagrams need the electron mass.","tokens_in":763,"tokens_out":740,"would_cite":true,"duration_ms":21060,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper computes the two-loop master integrals for $e^+e^-\\to\\mu^+\\mu^-$ with the electron mass retained, as a Frobenius series in $m$ up to $O(m^{13})$.","keywords":["two-loop master integrals","electron mass","Frobenius expansion","Goncharov polylogarithms","muon pair production","NNLO QED","collinear divergences","differential equations"],"falsifier":"Evaluate the most delicate master integrals (the j48–j51 family and the boundary constant of j41) with an independent numerical integration at, say, $m=0.1$ in $d=4-2\\epsilon$ and compare with the claimed $O(m^{13})$ Frobenius series. A disagreement larger than the truncation error, especially in the leading logarithm coefficients, would show that the epsilon-form transformation or the boundary constants are incorrect.","tokens_in":6152,"feed_emoji":"⚛️","tokens_out":8177,"duration_ms":72774,"temperature":0.7,"pith_summary":"The paper targets the two-loop QED corrections to $e^+e^-\\to\\mu^+\\mu^-$ in the kinematic regime where the electron mass cannot be set to zero. Its central claim is that the master integrals of the two diagram families whose massless limit produces collinear divergences are now available as a Frobenius series in $m$ up to $O(m^{13})$, with all coefficients expressed through Goncharov's polylogarithms. This matters because those collinear logarithms must be kept for physical observables at next-to-next-to-leading order (NNLO), while the other diagram families can safely be evaluated at $m=0$. The results are supplied as Mathematica substitution rules connecting the original integrals to a canonical basis.","feed_headline":"Electron mass included in two-loop muon-pair QED integrals","feed_subtitle":"Coefficients emerge as Goncharov polylogarithms, bringing mass logarithms into the NNLO cross section.","key_machinery":"The central object is the set of 61 master integrals and the rational transformations that take them to a canonical basis $\\mathbf{K}$. The calculation proceeds in four steps: reduction by integration-by-parts identities to a closed system of differential equations in $m^2$, $s$, and $t$; normalization of the $m^2$ system to Fuchsian form so that the solution is a Frobenius evolution operator $U(m^2,0)$; reduction of the boundary-constant system to $\\epsilon$-form (d log form) using the muon velocity $\\beta$ and scattering angle $\\cos\\theta$ as variables, with an eight-letter alphabet $\\{1-c,1+c,1-\\beta,\\beta,1+\\beta,1-\\beta c,1+\\beta c,1+2\\beta c+\\beta^2\\}$; and expression of the solution through Goncharov polylogarithms with alphabet $\\{0,\\pm1,\\pm(\\cos\\theta)^{-1},-e^{\\pm i\\theta}\\}$. The boundary conditions are the load-bearing input; they come from separate asymptotic limits rather than from a single limit.","core_discovery":"On the paper's own terms, the discovery is that the 61 master integrals of the two families in set (b) admit a small-mass Frobenius expansion, $\\mathbf{J}=U L T \\mathbf{K}$, where the evolution operator $U$ contains powers $(m^2)^{\\lambda_i}$ multiplied by logarithms, with exponents $\\lambda_i\\in\\{0,\\tfrac12-2\\epsilon,-4\\epsilon,-3\\epsilon,-2\\epsilon,-\\epsilon\\}$, and the canonical basis $\\mathbf{K}$ is expanded in $\\epsilon$ with coefficients built from Goncharov polylogarithms $G(a|\\beta)$. The author obtains the expansion up to $O(m^{13})$ and $O(\\epsilon^6)$ and fixes the boundary constants using the limits $\\beta\\to 0$, $\\beta\\to 1$ at $\\theta=\\pi/2$, and a dedicated 11-integral vertex subsystem at $s\\to 4m^2$. The claim is that these integrals are exactly what is needed for the electron-mass effects in the NNLO cross section.","pith_inferences":["A natural extension the author does not spell out: because the alphabet includes both $\\beta$ and $\\cos\\theta$, the results supply the full angular dependence of the cross section, not just the total rate, which is what collider experiments actually measure.","The boundary-fixing strategy suggests a general recipe for four-scale two-loop problems: extract as many constants as possible from one regular limit, cover the rest with a second limit, and isolate any leftover constant in a smaller subsystem that can be solved exactly.","If the fifteen constraints are truly identities of the integrals rather than artifacts of the truncation order, the independent integral count is 46; a future all-order proof of these relations would be a clean test of the present calculation.","A testable extension would be to compare the $O(m^{13})$ predictions against an exact-in-$m$ solution of the differential equations; agreement would confirm that no branch-cut information was lost in the boundary procedure."],"forward_implications":["The NNLO differential cross section for $e^+e^-\\to\\mu^+\\mu^-$ can be assembled with the electron-mass logarithms included, removing the need to treat collinear divergences by massless regularization.","The attached substitution rules give numerical values for the master integrals in the physical region for small $m$, and deeper expansions up to $O(m^{13})$ and $O(\\epsilon^6)$ are available on request.","Fifteen rational linear constraints reduce the 61 master integrals to 46 independent entries of the canonical basis, which simplifies the eventual insertion of these integrals into the amplitude.","The cross-checks against sector decomposition at $m=1/2$ indicate the Frobenius series remains accurate even at a moderately large mass, well beyond the nominal small-$m$ regime."],"supporting_citations":[{"why":"Supplies the integration-by-parts reduction machinery that reveals the 61 master integrals.","marker":"[10, 11]"},{"why":"Provides the transformations that put the differential systems into normalized Fuchsian and epsilon forms and extracts leading asymptotic coefficients.","marker":"[12, 13]"},{"why":"Gives the algorithm for constructing the Frobenius evolution operator as an expansion near singular points.","marker":"[14]"},{"why":"Sets out the approach used here for converting the evolution-operator equation into differential equations for the boundary coefficients.","marker":"[9]"},{"why":"Establishes the two-loop four-fermion amplitude whose collinear divergences make the electron-mass terms necessary.","marker":"[8]"},{"why":"Supplies the independent numerical evaluations used to cross-check the expanded master integrals.","marker":"[16]"}],"fun_headline_variants":["Frobenius series reveals electron mass in two-loop muon-pair QED","Goncharov polylogs express electron mass corrections in two-loop e+e- → μ+μ-","Up to m^13: electron mass enters two-loop muon-pair QED","61 master integrals cover electron mass for two-loop muon pair"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the only two-loop integrals that require a nonzero electron mass are the two families in set (b) of Fig. 1, together with the already known form-factor and photon self-energy corrections; the paper states that this can be shown but does not provide the proof.","fun_headline_variants_meta":{"raw":{"variants":["Frobenius series reveals electron mass in two-loop muon-pair QED","Goncharov polylogs express electron mass corrections in two-loop e+e- → μ+μ-","Up to m^13: electron mass enters two-loop muon-pair QED","61 master integrals cover electron mass for two-loop muon pair"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000568,"raw_usage":{"total_tokens":2628,"prompt_tokens":823,"completion_tokens":1805,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":439,"completion_tokens_details":{"reasoning_tokens":1714}},"tokens_in":439,"tokens_out":1805,"duration_ms":16586,"temperature":1.0,"reasoning_tokens":1714,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T04:59:49.280154+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the most delicate master integrals (the j48–j51 family and the boundary constant of j41) with an independent numerical integration at, say, $m=0.1$ in $d=4-2\\epsilon$ and compare with the claimed $O(m^{13})$ Frobenius series. A disagreement larger than the truncation error, especially in the leading logarithm coefficients, would show that the epsilon-form transformation or the boundary constants are incorrect.","supporting_citations":[{"cited_title":"Bonciani, A","cited_arxiv_id":null,"evidence_quote":"Establishes the two-loop four-fermion amplitude whose collinear divergences make the electron-mass terms necessary."}],"review_version":1}