{"id":"77cfe42e-7c66-4018-bcc7-4f730df4f1b9","arxiv_id":"2412.13876","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The two-loop finite remainders for gg to ttbar g at leading colour are evaluated numerically at one benchmark point, with elliptic functions confined to the finite part.","lead":"This paper computes the two-loop quantum corrections for producing a top quark pair plus a jet in gluon fusion, at a physical phase-space point and in the leading-colour approximation. It is a key step toward NNLO predictions for a process used to measure the top quark mass at the LHC.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite remainders hinge on numerically-determined vanishing of ε-coefficients for elliptic master integrals; no independent check of the ε^4 special-function values.","rationale":"The paper's central claim is the first numerical evaluation of the two-loop leading-colour finite remainders for gg→t¯tg. The method rests on a special-function representation of the master integrals in which the elliptic sectors contribute only at order ε^4. This is asserted on the basis of AMFlow evaluations at random points, but no proof or detailed sampling statistics are given. The finite remainder (the ε^0 coefficient) receives contributions from the ε^4 coefficients of these master integrals; if any of the lower-order coefficients were actually nonzero, the pole structure would be contaminated by elliptic functions, contradicting the universal IR/UV predictions and invalidating the results. The internal cross-checks in Sec. 5 are consistent but not independent: the momentum-twistor reduction uses the same master-integral evaluation, and the pole cancellations check only the singular terms, not the finite part. Thus the completeness of the vanishing list is the single most load-bearing assumption. An independent computation of the master integrals at the benchmark point directly tests this assumption and the correctness of the special-function values. This is a concrete and decisive check, and until it is performed, the CONDITIONAL verdict is appropriate. Our concern matches the reader's weakest_assumption, so agreement is complete.","tokens_in":28082,"tokens_out":10046,"duration_ms":90315,"concrete_test":"Compute the 37 master integrals of family PBB (and the permuted family PB'B) at the benchmark point (5.1)–(5.2) using an independent high-precision method such as AMFlow or sector decomposition, and Laurent-expand in ε through order ε^4. Verify: (1) the ε^1, ε^2, ε^3 coefficients of MIs 19, 20, 35, 36, and 37 vanish to working precision; (2) the ε^4 coefficients agree with the values obtained from the published special-function representation (i.e., evaluate the 12 non-polylog functions and the polylogarithmic functions at the same point). Agreement to, say, 10 digits would confirm the completeness of the vanishing list and the correctness of the finite remainders; any mismatch would locate the error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the two-loop finite remainders are correct depends on the completeness of the list of ε-coefficients of the master integrals that are set to zero. In Sec. 4.1 the paper states that AMFlow evaluations at random phase-space points establish that the MIs {19,20,35,36,37} (and the non-logarithmic parts of MI 15) start only at order ε^4. This property is what makes the analytic cancellation of UV/IR poles possible and justifies truncating the iterated-integral solution at order ε^3 for the polylogarithmic sector. If any of these vanishings is not identically zero—for example, if it holds only on a subvariety of the physical region or is an artifact of finite numerical precision—then the special-function representation (eq. 3.16) is incomplete, the pole cancellation would fail, and the benchmark values in Table 3 would be incorrect. The paper does not report how many points were sampled, the precision used, or any argument that the vanishing is exact. Furthermore, the boundary-value conspiracy for MI 15 (eq. 4.8) is verified only numerically at the single point d0. The internal cross-checks (gauge invariance, IR pole matching, momentum-twistor reduction) do not independently validate the ε^4 special-function coefficients, because the momentum-twistor reduction shares the same master-integral evaluation chain. This is a genuine gap in the verification of the central numerical result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents the first numerical evaluation of the two-loop leading-colour QCD helicity finite remainders for the gluon-fusion process gg -> tbar t g. The computation is performed at a single physical phase-space point specified in eqs. (5.1)-(5.2), with results collected in Table 3. The technical core is a new method for expressing the master integrals of Refs. [75,76] in an over-complete basis of special functions even though the differential equations are not in canonical form and elliptic sectors are present. The authors use numerical AMFlow evaluations to identify which epsilon-coefficients of the master integrals vanish, solve the non-canonical differential equations order by order in epsilon, isolate a set of 12 non-polylogarithmic functions that appear only at order epsilon^4, and evaluate them by generalised power series with DiffExp. The helicity amplitudes are obtained through four-dimensional projectors, finite-field IBP reduction, and rational reconstruction.","tokens_in":28327,"tokens_out":4844,"duration_ms":47981,"significance":"If the result is correct, this is an important milestone: it is the first two-loop 2 -> 3 amplitude with internal masses and elliptic master integrals evaluated numerically in the physical region, and the proposed method for extracting a polylogarithmic representation from non-canonical differential equations is likely to be reused in other computations. The paper is also unusually explicit about the assumptions underpinning the method, and it ships ancillary files containing the master-integral representations, the differential equations, and a DiffExp script, together with higher-precision values for future comparisons. The main caveat is that the central numerical result rests on numerically inferred exact vanishings that are not yet independently certified, so the significance of the paper is conditional on closing that verification gap.","major_comments":[{"comment":"The claim that the master integrals in the set S\\{15} = {19,20,35,36,37} start contributing only at order epsilon^4, and that the non-logarithmic contributions to g15 vanish through order epsilon^3 via the boundary conspiracy of eq. (4.8), is established only by 'a number of numerical evaluations ... at random phase-space points'. The paper does not report how many points were sampled, what precision was used, or why the vanishing should be exact rather than very small. This vanishing pattern is load-bearing: it is what permits the analytic cancellation of UV/IR poles and justifies truncating the iterated-integral solution at order epsilon^3. The authors should either provide a quantitative account of these numerical checks (number of points, working precision, size of the supposedly vanishing coefficients) or give an analytic argument, for example from maximal cuts, that the vanishings are identically true on the physical region. Without this, the finite remainders in Table 3 are not fully certified.","section":"Sec. 4.1, bullet list after eq. (4.9)"},{"comment":"The internal cross-checks are consistent and valuable, but they do not independently validate the epsilon^4 non-polylogarithmic coefficients. The three evaluation strategies in Sec. 4.3 share the same differential equations, the same boundary values at d0, and the same algebraic transformation from master integrals to special functions, while the momentum-twistor reduction shares the same master-integral evaluation chain. An independent numerical check of the central result, such as a direct high-precision AMFlow evaluation of the finite remainder at the benchmark point or at a second phase-space point, would materially strengthen the claim that the elliptic-sector contributions are correctly captured.","section":"Sec. 5, cross-check paragraph and Table 3"}],"minor_comments":[{"comment":"The expression for F_5^(1) is missing a separator: it reads 'F (1)_5 = log(2d45) - i pi F (1)_6 = log(-2d15)', which should be two separate equations.","section":"Eq. (4.10)"},{"comment":"The paper does not state how many prime fields were used for the rational reconstruction of the rational coefficients, nor how the phase-space point was rationalised. Adding this information would improve reproducibility.","section":"Sec. 3, rational reconstruction paragraph"},{"comment":"The conclusion says the finite remainders were computed 'at benchmark physical phase-space points', but the manuscript presents results at a single point. Either add a second benchmark point or rephrase to the singular.","section":"Sec. 6, first paragraph"},{"comment":"The timing comparison is given per segment; it would be helpful to state the typical number of segments per phase-space point and the total wall-clock time per point for the current implementation, so that the practical cost of the method is clearer.","section":"Sec. 4.3, Table 2"},{"comment":"The statement that 'elliptic functions appear solely in the finite remainder' is stronger than what is demonstrated: the non-polylogarithmic sector contains 12 functions of mixed elliptic and nested-square-root origin, and the property is shown for the chosen master-integral basis and at sampled points, not as a basis-independent statement.","section":"Abstract and Sec. 4.1"}],"recommendation":"major_revision","confidential_remarks":"I agree with the conditional assessment supplied with the manuscript. The central claim is novel and likely correct, but the verification of the numerically inferred vanishing pattern in Sec. 4.1 is a genuine gap because the entire construction of the finite remainder depends on it. A major revision that adds quantitative evidence for the vanishings or an independent numerical check of Table 3 would be sufficient; I do not see a reason to reject the paper on the evidence available."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Simon, this is the first numerical evaluation of the two-loop leading-colour finite remainders for gg → ttbar g, and it delivers what the abstract promises. The genuinely new piece is the method: an over-complete basis of special functions that lets them cancel the UV/IR poles analytically even though the master-integral DEs are non-canonical and elliptic. They show the non-polylogarithmic functions sit entirely in the finite part, and that a minimal system of DEs for just those twelve functions evaluates about twenty times faster per segment than the full MI system. That is a reusable strategy for other heavy-particle amplitudes.\n\nThe paper is careful. There are multiple internal cross-checks: gauge invariance, IR pole matching, two independent reduction paths for the benchmark table (projector vs momentum-twistor), and three different DiffExp-based evaluation strategies for the master integrals and special functions. Ancillary files with the DEs, boundary values, and solving scripts are provided. That is the right level of reproducibility for a benchmark.\n\nThe soft spot is the one the stress-test flags: the list of epsilon-coefficients that vanish, and the boundary-value conspiracy for MI 15, are established by numerical AMFlow sampling at random points, not by proof. If a vanishing were point-dependent or an artifact, the pole cancellation would fail for orders below epsilon^4, but an error at exactly epsilon^4 would pollute the finite part without breaking the poles. The internal cross-checks share the same master-integral chain, so they would not catch a systematic error in the input MIs.\n\nI don't think that concern sinks the paper, but it is a real limitation and the authors are upfront about it. The vanishings are structurally expected (the poles are polylogarithmic), the random-point sampling makes accidental vanishings unlikely, and the three evaluation strategies do cross-check the epsilon^4 special-function values against direct MI evaluation. The remaining risk sits in the published MIs of Ref. [76], which is an external but peer-reviewed dependency.\n\nThe benchmark is a single phase-space point, and the fast pentagon-function evaluation of the polylogs is still future work, so this is a proof of principle, not a phenomenological tool yet. That is fine: the point is that it is the first numerical result, and it opens the door to the full NNLO amplitude.\n\nSend it to review. A good referee should ask how many AMFlow points and what precision went into the vanishing list, and should push for a second benchmark point or an independent confirmation of the finite remainders. But this is a solid, significant step and deserves serious referee time.","headline":"First numerical two-loop finite remainders for gg → ttbar g at leading colour, with a new special-function method for elliptic integrals; the main caveat is the numerically-determined vanishing pattern, but the paper is honest and the result is solid.","tokens_in":28884,"tokens_out":5611,"would_cite":true,"duration_ms":49682,"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":"This paper presents the first benchmark numerical evaluation of the two-loop helicity finite remainders for the gluon-fusion production of a top-antitop pair plus a gluon, $gg \\to t\\bar{t}g$, in the leading-colour approximation, and shows…","keywords":["two-loop amplitudes","helicity amplitudes","top-quark pair plus jet","leading colour","elliptic Feynman integrals","special function basis","differential equations","NNLO QCD"],"falsifier":"Evaluate the master integrals in the set $S=\\{15,19,20,35,36,37\\}$ at the benchmark point of eqs. (5.1)–(5.2), or at a second independent physical point, using an independent numerical method that does not assume the special-function ansatz, to order $\\epsilon^4$: if any of these integrals develops a non-polylogarithmic contribution at order $\\epsilon^0$, $\\epsilon^1$, $\\epsilon^2$, or $\\epsilon^3$, the claimed pole cancellation fails. A second concrete check is to recompute the two-loop finite remainders of table 3 at a different phase-space point and compare against a direct integration-by-parts plus numerical-integration evaluation; agreement at all digits would confirm the construction, while disagreement would falsify it.","tokens_in":27880,"feed_emoji":"⚛️","tokens_out":10480,"duration_ms":88426,"temperature":0.7,"pith_summary":"Top-quark pair production in association with a jet is one of the processes that will limit the precision of LHC measurements unless theoretical predictions reach next-to-next-to-leading order in QCD, and the missing piece has been the two-loop double-virtual amplitudes. This paper delivers the first benchmark numerical values for those two-loop helicity amplitudes in the gluon-fusion channel, at leading colour, at physical phase-space points. The obstacle was that the relevant Feynman integrals include elliptic functions, which resist the standard polylogarithmic function bases and make pole cancellation difficult. The authors solve the non-canonical differential equations by expanding the master integrals in an over-complete basis of special functions, and find that the non-polylogarithmic pieces enter only at order $\\epsilon^4$, hence only in the finite remainder. This removes the main technical obstruction to NNLO $t\\bar{t}+$jet predictions.","feed_headline":"First two-loop gluon-fusion amplitudes for top pair + jet","feed_subtitle":"Elliptic integrals stay in the finite part, clearing the way to NNLO predictions at the LHC","key_machinery":"The engine of the computation is the (over-complete) special-function basis for the master integrals. Starting from non-canonical differential equations whose connection matrices are degree-two polynomials in $\\epsilon$ and contain non-logarithmic one-forms, the authors expand the master integrals order by order in $\\epsilon$, use numerical evaluations to identify which $\\epsilon$-coefficients vanish, and thereby extract an iterated-integral (polylogarithmic) part that can be handled with the standard pentagon-function machinery. The few remaining coefficients—twelve non-polylogarithmic functions originating from master integrals with nested square roots or elliptic curves—are treated as independent generators; they satisfy a closed system of 84 differential equations, which is solved by generalised power series. This construction is what allows the UV and IR poles to be cancelled analytically while keeping the elliptic content isolated in the finite remainder. Rational coefficients of the special-function monomials are evaluated with finite-field arithmetic, and helicity amplitudes are obtained through four-dimensional projectors combined with the massive spinor-helicity formalism.","core_discovery":"The central claim is that the two-loop finite remainders for $gg \\to t\\bar{t}g$ at leading colour can be evaluated numerically in the physical phase space, despite the presence of elliptic Feynman integrals. The paper achieves this by Laurent-expanding all master integrals to order $\\epsilon^4$ and expressing the coefficients in a (potentially over-complete) basis of 237 special functions: 225 polylogarithmic ones built with the pentagon-function algorithm, plus 12 non-polylogarithmic functions labelled $F_i^{(4*)}$ that come from the master integrals in the set $\\{15,19,20,35,36,37\\}$ of the non-canonical family. A decisive structural fact is that the non-polylogarithmic master integrals vanish below order $\\epsilon^4$, so the ultraviolet and infrared poles contain only polylogarithmic functions and can be subtracted analytically; the elliptic and nested-square-root functions appear only in the finite remainder and are evaluated numerically through generalised power-series solutions of a minimal 84-function system of differential equations. Benchmark values of the finite remainders for the three independent gluon helicity configurations are given at the phase-space point of eqs. (5.1)–(5.2), normalised by the tree-level amplitude, together with cross-checks based on gauge invariance and on the predicted IR/UV pole structure.","pith_inferences":["Beyond the paper: the completeness of the vanishing-coefficient list is verified at random points but not proved; checking the set $S=\\{15,19,20,35,36,37\\}$ at a dense grid of physical points, or with an independent evaluation method, would convert the construction from numerical evidence into a robust result.","Beyond the paper: the paper leaves the fast numerical evaluation of the 225 polylogarithmic functions to future work; a concrete next test is to implement those representations and compare total per-point evaluation time against the master-integral route at many phase-space points.","Beyond the paper: the 'mysterious' master integral 15, whose derivatives contain non-logarithmic one-forms that cancel only through a conspiracy of boundary values, may admit a basis transformation that removes the obstruction; if found, it could make the entire amplitude representation canonical and possibly analytic.","Beyond the paper: the special-function construction should transfer to other leading-colour massive two-loop processes whose differential equations are polynomial in $\\epsilon$ but non-canonical, such as $t\\bar{t}H$ or $W b\\bar{b}$ production, as long as the non-polylogarithmic sectors start at the order needed for the finite part."],"forward_implications":["The benchmark finite remainders provide the first numerical anchor for the double-virtual contribution to $t\\bar{t}+$jet at NNLO in QCD, allowing future work to target fast evaluation across phase space and eventual phenomenological predictions.","Because the elliptic master integrals contribute only at order $\\epsilon^4$, the universal pole structure of the amplitude is preserved and the UV/IR subtraction can be carried out analytically with polylogarithmic functions alone.","Expressing the master integrals in the special-function basis reduces the complexity of the rational coefficients—their maximum polynomial degree drops by roughly 30% relative to a master-integral representation—and five weight-4 special functions drop out of the finite remainders entirely.","The minimal 84-function system for the non-polylogarithmic part evaluates in about 16 seconds per path segment, substantially faster than solving all two-loop master integrals, indicating a realistic route to an efficient numerical library for this amplitude.","The method extends the pentagon-function approach to integrals whose differential equations are not in canonical form, so it applies to other two-loop $2 \\to 3$ processes with internal masses where canonical forms are currently out of reach."],"supporting_citations":[{"why":"Supplies the two-loop master integrals, their non-canonical differential equations, the boundary values at $\\vec{d}_0=(2,1,-1,5,-2,1)$, and the identification of which master integrals sit in elliptic or nested-square-root sectors; the whole special-function construction rests on these results.","marker":"[76]"},{"why":"Provides the earlier form of the two-loop master integrals for the planar topology that [76] completes; the differential equations for the permuted families used here are obtained by permuting these results.","marker":"[75]"},{"why":"Describes the pentagon-function algorithm—selecting algebraically independent $\\epsilon$-coefficients through Gaussian elimination, shuffle relations, and symbol-level analysis—that this paper extends to non-canonical differential equations.","marker":"[53]"},{"why":"Supplies numerical master-integral evaluations at random phase-space points, used to determine which $\\epsilon$-coefficients vanish and to validate boundary-value relations.","marker":"[110,111]"},{"why":"Provides the generalised power-series method and its implementation used to integrate the minimal system of differential equations for the non-polylogarithmic functions.","marker":"[79,80]"},{"why":"Gives the one-loop $gg \\to t\\bar{t}g$ helicity amplitudes to $\\mathcal{O}(\\epsilon^2)$ and the momentum-twistor parameterisation of the kinematics that the projector-based helicity construction uses.","marker":"[74]"},{"why":"Supplies the massive spinor-helicity formalism, the decomposition of the amplitude, and the integral representation of the mass counterterms used to define the mass-renormalised amplitudes.","marker":"[84]"},{"why":"Four-dimensional projection technique that converts the contracted amplitude into helicity amplitudes by inverting a basis of tensor structures.","marker":"[81]"}],"fun_headline_variants":["Two-loop top pair + jet amplitudes now computable numerically","Elliptic two-loop integrals tamed for top pair + jet","First numerical two-loop gg→t tbar g helicities","Numerical two-loop amplitudes: top pair plus jet, elliptic intact","Two-loop gluon-fusion top pair + jet: elliptic solved numerically"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the master integrals and boundary values obtained from the differential equations are correct, and that the numerical finding that the non-polylogarithmic master integrals $\\{15,19,20,35,36,37\\}$ vanish below order $\\epsilon^4$ is complete rather than accidental; if any of those vanishings fails at some physical point, the analytic cancellation of the UV/IR poles and the construction of the special-function basis would break down.","fun_headline_variants_meta":{"raw":{"variants":["Two-loop top pair + jet amplitudes now computable numerically","Elliptic two-loop integrals tamed for top pair + jet","First numerical two-loop gg→t tbar g helicities","Numerical two-loop amplitudes: top pair plus jet, elliptic intact","Two-loop gluon-fusion top pair + jet: elliptic solved numerically"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000575,"raw_usage":{"total_tokens":2733,"prompt_tokens":985,"completion_tokens":1748,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":601,"completion_tokens_details":{"reasoning_tokens":1660}},"tokens_in":601,"tokens_out":1748,"duration_ms":11690,"temperature":1.0,"reasoning_tokens":1660,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:41:12.198532+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the master integrals in the set $S=\\{15,19,20,35,36,37\\}$ at the benchmark point of eqs. (5.1)–(5.2), or at a second independent physical point, using an independent numerical method that does not assume the special-function ansatz, to order $\\epsilon^4$: if any of these integrals develops a non-polylogarithmic contribution at order $\\epsilon^0$, $\\epsilon^1$, $\\epsilon^2$, or $\\epsilon^3$, the claimed pole cancellation fails. A second concrete check is to recompute the two-loop finite remainders of table 3 at a different phase-space point and compare against a direct integration-by-parts plus numerical-integration evaluation; agreement at all digits would confirm the construction, while disagreement would falsify it.","supporting_citations":[],"review_version":1}