{"id":"528c374c-d416-4a2d-99da-ce8b5957351c","arxiv_id":"2502.03276","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A Mathematica package computes high-precision epsilon-expansions of Lauricella functions using one-dimensional Frobenius series and interpolation.","lead":"This paper presents a computer method that calculates Lauricella functions, a family of multi-variable special functions used in particle physics, to high precision, including their expansion in a small parameter epsilon. The method uses one-dimensional power series and a publicly available Mathematica package, making such calculations faster and easier to parallelize than earlier multi-dimensional approaches.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The analytic-continuation step in §3.2 is heuristic and internally inconsistent; without independent sheet-level checks at complex arguments, the claim of covering all of C^n is not established.","rationale":"The reader's weakest_assumption identifies the heuristic graph-based continuation and the unproven Pfaffian systems as the load-bearing risk. My stress-test pass converges on the same point, with the continuation part being the sharper threat because it directly supports the 'arbitrary arguments in C^n' claim and because the paper's own description contains internal inconsistencies (kappa_i >= 0 vs. imaginary-direction steps; 'all singularities real' vs. rotated cuts for complex kappa_i). The Frobenius construction itself is standard, and the worked F1 example at a real point plus the timing tables give some confidence that the pipeline runs and produces plausible numbers, but they do not exercise the complex-argument regime where the continuation heuristic matters most. The error-estimate formula Eq. (3.51) is also suspicious, but it is a secondary accuracy-reporting issue: even if the formula is optimistic, a user could choose a smaller lattice step; the continuation issue can produce wrong values without any warning. The proposed concrete test is deliberately small: F1 admits an independent MPL representation, so a few complex-point comparisons would settle whether the sheet-handling is correct. The verdict remains CONDITIONAL: accept only after such verification, not before. No change to the reader's verdict is needed.","tokens_in":22701,"tokens_out":5740,"duration_ms":57201,"concrete_test":"Evaluate F1(1/2; 1, eps; 3/2; x, y) with PrecisionLauricella at precision 30 and expansion order eps^3 for two points that force the four-step continuation with a nontrivial complex final segment, e.g. (x,y) = (1/2 + i, 1/3 - 2i) and (x,y) = (-1/2 + i, 2 - i). Independently evaluate the right-hand side of Eq. (3.53) at the same points using HandyG or GiNaC at matching precision. If any of the four epsilon coefficients disagrees beyond the error estimate from Eq. (3.51), the continuation has selected the wrong Riemann sheet; if all agree, the sheet-handling concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is high-precision evaluation of Lauricella functions at arbitrary points of C^n on the correct Riemann sheet. That claim rests on the graph-based continuation of Section 3.2: circular convergence regions are connected when the minimal distance between regions is less than 3/4 of a radius, the grid is refined until a path is found, and the +i delta cut convention selects the sheet. None of these steps is proven, and the description contains internal tensions: the text says 'in all steps, the corresponding paths will have kappa_i >= 0', yet the positive- and negative-imaginary-direction steps necessarily use complex kappa_i, and it asserts that 'all the singular points will lie on the real axis' even though footnote 8 states that complex kappa_i rotate the branch cuts in the t-plane. The single displayed example has a real target point (4/3, 7/4) and real projected singularities, so it never exercises the complex-point machinery. If the graph path crosses a cut or lands on the wrong sheet, the package will silently return plausible but wrong values, because the Frobenius recurrence itself is standard and gives no warning. The independent MPL check advertised around Eq. (3.53) is invoked but the actual numerical comparison is not shown, so the correctness of the continuation at complex arguments is currently unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a numerical method, implemented in a Mathematica package called PrecisionLauricella, for evaluating Lauricella functions whose indices depend linearly on a small parameter ε. The method computes one-dimensional Frobenius generalized power series solutions to Pfaffian differential-equation systems, performs analytic continuation in the argument space via a graph-based chaining of convergence regions, and reconstructs the Laurent expansion in ε by Lagrange interpolation over a small set of numerical ε values rather than by a symbolic expansion. The paper includes one fully worked numerical example for the Appell function F1 and timing tables for several Lauricella and Appell functions.","tokens_in":22963,"tokens_out":8158,"duration_ms":70667,"significance":"If the method is correct and robust, it provides a useful alternative to Mellin–Barnes and multi-dimensional series approaches for high-precision ε-expansions of Lauricella functions, with the practical advantages of one-dimensional series and straightforward parallelization. The ε-interpolation idea is simple and effective, and the availability of the package is a concrete asset. However, the paper does not supply the independent numerical validation that would establish the central claim of high-precision evaluation at arbitrary complex arguments; all displayed examples use real arguments, and the advertised check against multiple polylogarithms is not shown numerically. The analytic-continuation algorithm of Section 3.2 is heuristic, with internal inconsistencies that need to be resolved. The underlying Frobenius machinery is standard, but the paper's contributions—the continuation strategy and the ε-interpolation error estimate—need firmer support.","major_comments":[{"comment":"The paper claims an independent check of the F1 example using multiple polylogarithms and tools like HandyG or GiNaC, but no comparison is shown. The only numerical output is Eq. (3.59) for F1 at a real point, and Tables 1–5 report timings only, not accuracy. Since the central claim is high-precision evaluation, the absence of a validation table (including at least one genuinely complex target point) leaves the correctness of the package unestablished. Please add explicit comparisons against independent high-precision values for every function and parameter point presented.","section":"§3.4, Eq. (3.53), and §4"},{"comment":"The analytic-continuation description is internally inconsistent and heuristic. The text states that 'In all steps, the corresponding paths will have κ_i ≥ 0' and that 'all the singular points will lie on the real axis', yet the described positive- and negative-imaginary-direction steps require complex κ_i, and footnote 8 correctly notes that complex κ_i rotate the branch cuts after projection onto the t-plane. The 3/4-overlap rule and the grid-refinement-until-a-path-is-found strategy are not proved to cover all of Cn on the principal sheet. Because all displayed target points have real arguments, the complex-sheet behavior is not tested. The continuation path in Figure 4 does use complex expansion centers, which partially exercises the machinery, but a complex target point with an independent verification is still missing.","section":"§3.2"},{"comment":"The error estimate in Eq. (3.51) appears inconsistent with the four-point interpolation example. With 2n = 4 lattice points and k = 4 expansion terms (ε^0 through ε^3), the exponent 2n − 2⌊k/2⌋ = 0, giving a vacuous bound; the explicit expansions in Eq. (3.50) give O(h^4) for the constant and linear terms and O(h^2) for the quadratic and cubic terms. The general formula needs to be corrected and derived, since it is used to choose the lattice step and to claim a target precision.","section":"§3.3, Eq. (3.51)"},{"comment":"The recurrence relations for the Frobenius coefficients are stated without derivation, and the solution of the underdetermined homogeneous system (3.33) is described only vaguely: the free parameters 'may either be matched with the matrix t^{A0} or generated randomly'. For a paper whose contribution is a numerical method, this is not a complete algorithmic specification. Please either derive the recurrences and the normalization procedure in detail or provide a precise reference where they are fully documented.","section":"§3.1, Eqs. (3.32)–(3.33)"}],"minor_comments":[{"comment":"There is a typo: 'staring' should be 'starting' in the description of the third step of the continuation procedure.","section":"§3.2"},{"comment":"The package name appears as 'PrecisionLauriecella' in one paragraph; it should be 'PrecisionLauricella'.","section":"§3.4"},{"comment":"The definition '|k| = Pn j=1' is missing the summation index and upper limit; it should read '|k| = sum_{j=1}^n k_j'.","section":"§2, Eq. (2.16)"},{"comment":"The caption lists the points {0, 9/7 − i, 39/14 − i, 53/14 − i/2} as the centers of convergence regions, but it also says 'Red points denote singularities'. Please clarify which points are singularities and which are expansion centers.","section":"Figure 4 caption"},{"comment":"In Eq. (3.48) and surrounding text, the notation for the expansion variable switches between 'ϵ' and 'ε'; please use a single symbol consistently.","section":"§3.3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is better suited to a computer-physics or numerical-special-functions venue than to a hep-th journal, but that is an editorial decision. The main technical gaps—missing validation tables, the unproven and internally inconsistent continuation heuristic, and the unclear error estimate—should be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's the short version. The genuinely new thing in 2502.03276 is packaging one-dimensional Frobenius series for Lauricella functions with ε-interpolation. That combination is not in the cited literature, and the ε-lattice interpolation is a real plus: it parallelizes trivially and avoids the blow-up of symbolic ε recurrences. The core math—Frobenius solutions of Pfaffian systems—is standard, the F1 system is displayed in full, and the worked example is internally consistent. The code is public, and the timing tables give a useful sense of cost.\n\nThe soft spots are real, though not fatal. First, the claim that the method reaches all of C^n is not supported. Section 3.2 says every continuation path has κ_i ≥ 0, yet the described steps include running in negative real and imaginary directions. It also says all singularities lie on the real axis, which their own footnote 8 contradicts. The single example targets a real point; no complex target is exercised. Second, the promised independent check against multiple polylogarithms is never shown. They display the MPL expression, but no table of numerical values comparing PrecisionLauricella against GiNaC or HandyG. That missing verification matters, since the continuation heuristic is what needs checking. Third, Eq. (3.51) is printed inconsistently: with four lattice points and four ε terms it gives h^0, while the interpolation expansions in (3.50) give h^2 for the top coefficients. Probably a typo, but it should be fixed.\n\nWho is this for? People who need high-precision numerical values of F_A, F_B, F_D (n≤3) with ε-dependent indices, e.g., Feynman integral calculations. They will get a practical tool from this paper. I'd send it to a serious referee, but with explicit requests: add the verification table, correct the error formula, and rewrite the continuation section so the cut and sheet conventions are unambiguous. I would not cite it in my own work until those are addressed.","headline":"Useful new tool for Lauricella functions, but the analytic-continuation claims need verification and the missing comparison table holds it back.","tokens_in":23474,"tokens_out":6085,"would_cite":false,"duration_ms":53890,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["33C65","33C70","65D20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that Lauricella functions with indices linear in a parameter ε can be evaluated to high precision at arbitrary complex arguments by chaining one-dimensional Frobenius series through a Pfaffian system, with the…","keywords":["Lauricella functions","Frobenius method","analytic continuation","epsilon expansion","Pfaffian systems","high-precision numerical evaluation","Appell functions","Feynman integrals"],"falsifier":"Take a Lauricella function that reduces to multiple polylogarithms, say $F_1(\\tfrac12; 1, \\varepsilon; \\tfrac32; x, y)$, evaluate it at a point with negative real parts and a tiny imaginary part near a branch cut using the package, and compare with an independent high-precision evaluation from the multiple-polylogarithm representation at the same $\\varepsilon$ values; a mismatch in the reported digits would show that the continuation graph or the $+i\\delta$ cut convention picked the wrong branch.","tokens_in":22492,"feed_emoji":"🧮","tokens_out":9916,"duration_ms":77698,"temperature":0.7,"pith_summary":"This paper presents a method—implemented in the Mathematica package PrecisionLauricella—for high-precision numerical evaluation of the Lauricella functions $F_A^{(n)}$, $F_B^{(n)}$, and $F_D^{(n)}$ with $n \\leq 3$, when the function indices depend linearly on a small parameter $\\varepsilon$. The output is a Laurent series in $\\varepsilon$, the kind of object one meets when Feynman integrals are expanded in dimensional regularization. The method replaces multi-dimensional hypergeometric sums with one-dimensional Frobenius generalized power series solving the function's Pfaffian differential system along a path, and then patches these series together by analytic continuation through overlapping convergence regions. A second idea is to evaluate the whole chain at several numerical values of $\\varepsilon$ and reconstruct the Laurent coefficients by Lagrange interpolation, which makes the computation easy to parallelize. A sympathetic reader would take the central claim to be that this combination yields fast, accurate $\\varepsilon$-expansions at arbitrary complex arguments.","feed_headline":"One-dimensional series give high-precision Lauricella expansions","feed_subtitle":"Expands ε-dependent Lauricella functions into Laurent series at any complex point, avoiding slow multi-dimensional sums.","key_machinery":"The engine is the Frobenius generalized power series solution of the one-dimensional Pfaffian system $dJ/dt = M_t J$ obtained along the line $x_i = \\kappa_i t$. Near a singular point the fundamental solution matrix is expanded as $U = \\sum_{\\lambda\\in S} t^\\lambda \\sum_{n} \\sum_{k=0}^{m_\\lambda} c^{(\\lambda)}_{n,k} t^n \\log^k t$, with recurrence relations for the coefficients $c^{(\\lambda)}_{n,k}$; analytic continuation is performed by chaining such expansions through an intersection graph of convergence disks, using a three-quarter-radius overlap criterion and fixed branch cuts, and the $\\varepsilon$ dependence is reconstructed by Lagrange interpolation over lattice values of $\\varepsilon$ rather than by symbolic expansion of the recurrences.","core_discovery":"The central claim is that an $\\varepsilon$-expansion of a Lauricella function with indices linear in $\\varepsilon$ can be computed at any point of $\\mathbb{C}^n$ by four steps: restrict the Pfaffian system to a line $x_i = \\kappa_i t$; solve the resulting one-dimensional ordinary differential system by generalized Frobenius series that may contain logarithms; continue the solution from the origin by gluing local series across overlapping disks whose centers are found by a graph search, with branch cuts fixed by a $+i\\delta$ convention; and finally evaluate at several small lattice values of $\\varepsilon$, reconstructing the Laurent series by interpolation. The authors demonstrate the pipeline on the Appell function $F_1(\\tfrac12; 1, \\varepsilon; \\tfrac32; \\tfrac43, \\tfrac74)$, reporting 30 digits through order $\\varepsilon^3$, and they cross-check such values against multiple-polylogarithm representations.","pith_inferences":["The three-quarter-overlap and grid-density choices are heuristic; a cheap robustness test would be to re-run the same point with a different grid density and use the spread of results as an empirical error bar.","Because the method fixes a Riemann sheet by cut conventions, looping the continuation path around a singular locus would let the same machinery read off monodromy transformations numerically, something the paper lists as future work rather than a delivered result.","The lattice-step $h$ can be tuned against the Frobenius truncation error through the stated estimate, so an automatic step-selection rule would make the package safer for users who do not know the function's analytic structure."],"forward_implications":["The package produces Laurent expansions in $\\varepsilon$ for $F_A$, $F_B$, and $F_D$ with $n \\leq 3$ at arbitrary complex arguments, to a requested number of $\\varepsilon$ terms and decimal places.","Because the series are one-dimensional, the method avoids the nested multi-dimensional sums of Mellin–Barnes or re-expansion approaches, which should allow higher precision with comparable or less effort.","The interpolation step makes the $\\varepsilon$ lattice evaluations independent of one another, so the computation parallelizes across cores and its runtime grows roughly linearly with the number of $\\varepsilon$ terms.","The same differential-equation-plus-Frobenius machinery extends, in the authors' view, to other hypergeometric families and to Feynman master integrals, where one-dimensional generalized power series valid over kinematic space would facilitate subsequent phase-space integration."],"supporting_citations":[{"why":"Supplies the Pfaffian differential systems, in the functional bases given in Eq. (3.19), from which every series solution in the paper starts.","marker":"[32]"},{"why":"Provides the generalized power series expansions near singular points whose recurrence structure the Frobenius step follows.","marker":"[75]"},{"why":"Companion work on expansions near singular points, used for the same one-dimensional generalized power series machinery.","marker":"[76]"},{"why":"Provides high-precision numerical evaluation of multiple polylogarithms, used as an independent check of the package's results.","marker":"[86]"},{"why":"Another independent evaluator of generalized polylogarithms, cited for verification of the numerical expansions.","marker":"[87]"},{"why":"An alternative analytic-continuation and numerical-evaluation method for Appell and Lauricella functions used for comparison.","marker":"[40]"},{"why":"A reference for evaluation of the Appell $F_2$ double hypergeometric function, used as a comparison baseline.","marker":"[54]"},{"why":"Provides the one-loop triangle with six independent scales where the demonstrated $F_1$ example arises.","marker":"[7]"}],"fun_headline_variants":["Lauricella functions to 30 digits via one-dimensional series","High-precision Lauricella: Frobenius series beat multi-d sums","Laurent series for ε-dependent Lauricella: 30-digit check","Efficient parallel Lauricella evaluation via Frobenius lines"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything depends on the assumption that the differential-equation systems the package starts from are complete and correct, and that the automatic path-finding never slips to the wrong branch of the function; if either fails, the package returns a numerical value without an error flag.","fun_headline_variants_meta":{"raw":{"variants":["Lauricella functions to 30 digits via one-dimensional series","High-precision Lauricella: Frobenius series beat multi-d sums","Laurent series for ε-dependent Lauricella: 30-digit check","Efficient parallel Lauricella evaluation via Frobenius lines"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000861,"raw_usage":{"total_tokens":3703,"prompt_tokens":881,"completion_tokens":2822,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":497,"completion_tokens_details":{"reasoning_tokens":2745}},"tokens_in":497,"tokens_out":2822,"duration_ms":19107,"temperature":1.0,"reasoning_tokens":2745,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T05:16:49.372454+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a Lauricella function that reduces to multiple polylogarithms, say $F_1(\\tfrac12; 1, \\varepsilon; \\tfrac32; x, y)$, evaluate it at a point with negative real parts and a tiny imaginary part near a branch cut using the package, and compare with an independent high-precision evaluation from the multiple-polylogarithm representation at the same $\\varepsilon$ values; a mismatch in the reported digits would show that the continuation graph or the $+i\\delta$ cut convention picked the wrong branch.","supporting_citations":[],"review_version":1}