Two-loop master integrals for $e^{+}e^{-}\rightarrow\mu^+\mu^-$ process with account of electron mass

Roman N. Lee

REVIEW 4 major objections 4 minor 16 references

Two-loop master integrals for $e^{+}e^{-}\rightarrow\mu^+\mu^-$ process with account of electron mass

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read 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 $O(m^{13})$.

desk verdict A solid, genuinely new two-loop massive calculation whose only real weakness is an unproven claim about which diagrams need the electron mass. read the letter →

arxiv 2412.00793 v1 pith:4QWEJTJA submitted 2024-12-01 hep-ph

Roman N. Lee This is my paper · ORCID

classification hep-ph

keywords two-loopmasterintegralselectronmassFrobeniusexpansionGoncharovpolylogarithmsmuonpairproductionNNLOQEDcollineardivergencesdifferentialequations

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

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.

What carries the argument

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.

What would settle it

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.

Watch

Extended reading notes

Core claim

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.

Load-bearing premise

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.

Editorial extensions

If this is right

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.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

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.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the 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.

The 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.

The 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.

Citation 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.

My 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.

Referee Report

4 major / 4 minor

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.

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 (4)

[§1 (Introduction) and §2] 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).
[§3 (Cross checks)] 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).
[§3 (Results and files)] 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.
[§2 (Eq. (2.25))] 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.

minor comments (4)

[Throughout] 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.
[§2] 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.
[§2 (Eq. (2.24))] 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.
[§1] 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.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity detected; derivation is self-contained, though the collinear-completeness classification is asserted without proof.

full rationale

The master integrals are obtained by IBP reduction and differential equations; the boundary constants are fixed from independent kinematic asymptotics (β→0, β→1 at θ=π/2, and s→4m² for the j41 subsystem), not by fitting to the target integrals. The numerical comparison with Fiesta is external to the derivation. Citations to LiteRed2, Libra, and the author's prior papers are tool/method citations and are not load-bearing for the announced result. The only unsupported statement is the classification in Sec. 1 that collinear divergences appear only in set(a) with l_e>0 and set(b), and the coverage claim in Sec. 2 that set(b) is expressed by the two Fig. 2 families; this is an unproven completeness assumption and a correctness risk, but it is not a circular reduction because the later calculation does not assume the values of the master integrals being derived.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The calculation is self-contained in the sense that no parameters are fitted to the final integrals. The main inputs are the identification of the relevant integral families and the choice of boundary conditions; both are physics assumptions rather than free parameters. No invented entities are introduced.

assumptions (4)

domain assumption The two families in Fig. 2 comprise exactly the set of two-loop integrals where the electron mass must be kept to regulate collinear divergences in e+e−→μ+μ−.
Invoked in Section 1: 'It can be shown that the collinear divergences appear only in set(a) with le > 0 and in set(b)'. No proof or citation is given, yet this justifies restricting the calculation to two families.
domain assumption Boundary conditions for the differential equations can be fixed from the asymptotics β→0, β→1 at θ=π/2, and a special treatment via s→4m² for one constant.
Section 2: 'We fix the boundary conditions by evaluating the asymptotics of c in different limits.' The uniqueness of the solution depends on these constants being fully determined.
standard math The straight-line path from (0,θ) to (β,θ) in the d log differential system avoids singularities, so the ordered exponential gives the solution.
Section 2, Eq. (2.23)-(2.24); standard for ϵ-form systems but requires the path to avoid the alphabet zeros p_i=0.
domain assumption The dimensional recurrence relation provides correct values at d=6−2ϵ for the integrals j48-j51, so a cross-check there validates the d=4−2ϵ results.
Section 3, Cross checks: FIESTA unreliable for j48-j51 at d=4−2ϵ, so comparison is made at d=6−2ϵ.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Two-loop master integrals for $e^{+}e^{-}\rightarrow\mu^+\mu^-$ process with account of electron mass." pith.science (2026). https://pith.science/paper/4QWEJTJA

@misc{pith2026241200793,
  author       = {Pith},
  title        = {Pith review of: Two-loop master integrals for $e^+e^-\rightarrow\mu^+\mu^-$ process with account of electron mass},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4QWEJTJA}},
  note         = {Machine review of arXiv:2412.00793}
}

abstract

We calculate a subset of two-loop master integrals relevant for the differential cross section of $e^+e^-\to \mu^+\mu^-$ process. We consider only those families for which the account of the electron mass $m$ is necessary. Our results have the form of the Frobenius series in $m$ with coefficients expressed via Goncharov's polylogarithms.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 11 canonical work pages

[9]

R. N. Lee and V. A. Stotsky,Master integrals fore+e− → 2γ process at large energies and angles, 2410.03336

arXiv
[1]

V. B. Berestetskii and I. I. Pomeranchuk,Formation of aµ - meson pair in positron annihilation, JETP 2 (1956) 580

work page 1956
[2]

F. A. Berends, R. Kleiss, S. Jadach and Z. Was,Qed radiative corrections to electron-positron annihilation into heavy fermions, Acta Phys. Pol., Series B;(Poland)14 (1983)

work page 1983
[3]

F. A. Berends, K. Gaemers and R. Gastmans,α3-contribution to the angular asymmetry in e+ e-→ µ+ µ-, Nuclear Physics B63 (1973) 381

work page 1973
[4]

Jadach and Z

S. Jadach and Z. Was,Qed 0 (α 3) radiative corrections to the reaction e+ e-→ τ+ τ-including spin and mass effects, Acta Physica Polonica. Series B15 (1984) 1151

work page 1984
[5]

Mastrolia, M

P. Mastrolia, M. Passera, A. Primo and U. Schubert,Master integrals for the nnlo virtual corrections toµe scattering in qed: the planar graphs, Journal of High Energy Physics2017 (2017) 1

work page 2017
[6]

Di Vita, S

S. Di Vita, S. Laporta, P. Mastrolia, A. Primo and U. Schubert,Master integrals for the nnlo virtual corrections toµe scattering in qed: the non-planar graphs, Journal of High Energy Physics 2018 (2018) 1

work page 2018
[7]

R. N. Lee and K. T. Mingulov,Master integrals for two-loopC-odd contribution to e+e− → ℓ+ℓ− process, 1901.04441

work page Pith review arXiv 1901

Show all 16 references

[8]

Bonciani, A

R. Bonciani, A. Broggio, S. Di Vita, A. Ferroglia, M. K. Mandal, P. Mastrolia et al.,Two-loop four-fermion scattering amplitude in qed, Physical Review Letters128 (2022) 022002

2022
[10]

R. N. Lee,LiteRed 1.4: a powerful tool for reduction of multiloop integrals, vol. 523, p. 012059, 2014,1310.1145, DOI

2014 arXiv
[11]

LiteRed2, essential update ofLiteRed package

R. N. Lee, “LiteRed2, essential update ofLiteRed package.” https://github.com/rnlg/LiteRed2
[12]

R. N. Lee,Libra: A package for transformation of differential systems for multiloop integrals, Comput. Phys. Commun.267 (2021) 108058 [2012.00279]

2021 arXiv
[13]

Libra, package for transforming first-order linear differential systems

R. N. Lee, “Libra, package for transforming first-order linear differential systems.” https://github.com/rnlg/Libra
[14]

R. N. Lee, A. V. Smirnov and V. A. Smirnov,Solving differential equations for Feynman integrals by expansions near singular points, JHEP 03 (2018) 008 [1709.07525]. – 8 –

2018 arXiv
[15]

Computer Algebra System Fermat

R. Lewis, “Computer Algebra System Fermat.”http://home.bway.net/lewis/
[16]

A. V. Smirnov, N. D. Shapurov and L. I. Vysotsky,FIESTA5: Numerical high-performance Feynman integral evaluation, Comput. Phys. Commun.277 (2022) 108386 [2110.11660]. – 9 –

2022 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [9]

R. N. Lee and V. A. Stotsky,Master integrals fore+e− → 2γ process at large energies and angles, 2410.03336

arXiv

[2] [1]

V. B. Berestetskii and I. I. Pomeranchuk,Formation of aµ - meson pair in positron annihilation, JETP 2 (1956) 580

work page 1956

[3] [2]

F. A. Berends, R. Kleiss, S. Jadach and Z. Was,Qed radiative corrections to electron-positron annihilation into heavy fermions, Acta Phys. Pol., Series B;(Poland)14 (1983)

work page 1983

[4] [3]

F. A. Berends, K. Gaemers and R. Gastmans,α3-contribution to the angular asymmetry in e+ e-→ µ+ µ-, Nuclear Physics B63 (1973) 381

work page 1973

[5] [4]

Jadach and Z

S. Jadach and Z. Was,Qed 0 (α 3) radiative corrections to the reaction e+ e-→ τ+ τ-including spin and mass effects, Acta Physica Polonica. Series B15 (1984) 1151

work page 1984

[6] [5]

Mastrolia, M

P. Mastrolia, M. Passera, A. Primo and U. Schubert,Master integrals for the nnlo virtual corrections toµe scattering in qed: the planar graphs, Journal of High Energy Physics2017 (2017) 1

work page 2017

[7] [6]

Di Vita, S

S. Di Vita, S. Laporta, P. Mastrolia, A. Primo and U. Schubert,Master integrals for the nnlo virtual corrections toµe scattering in qed: the non-planar graphs, Journal of High Energy Physics 2018 (2018) 1

work page 2018

[8] [7]

R. N. Lee and K. T. Mingulov,Master integrals for two-loopC-odd contribution to e+e− → ℓ+ℓ− process, 1901.04441

work page Pith review arXiv 1901