{"id":"c93507d6-e325-4265-90fc-f9ffe406ac8e","arxiv_id":"1908.04061","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"New Padé approximants built from large-mass and threshold expansions give the two-loop top-quark mass dependence of the gg to ZZ box form factors relevant for off-shell Higgs interference.","lead":"This paper computes top-quark mass effects in the two-loop quantum corrections to gluon fusion into two Z bosons, the process used to constrain the Higgs boson width. The authors reconstruct the relevant form factors using expansions around two kinematic limits and Padé approximants, producing a new prediction for the off-shell Higgs interference.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The NLO reconstruction rests on an internal Padé error estimate with no external two-loop benchmark; if the dropped analytic threshold terms bias the axial-vector form factor, the interference prediction inherits that bias.","rationale":"The paper's method is sensible, and the one-loop validation is a genuine positive check. The concern is not that the method is internally inconsistent, but that the two-loop uncertainty estimate is self-referential: the Padé family is selected and scored using the same input data that define it, so the spread of acceptable approximants does not measure the error from the dropped analytic terms. The authors themselves flag the vector form factor as unreliable above about 500 GeV, demonstrating that such misses do occur; the paper then relies on the numerical dominance of the axial-vector form factor to keep the interference prediction trustworthy at larger MZZ. The single most load-bearing condition is therefore the accuracy of F_AA in precisely the region where the interference claim must hold, and that condition is not yet established by an external check in this manuscript. The reader's CONDITIONAL verdict already identifies this gap and asks for the correct condition, namely comparison with a full numerical two-loop calculation, so no verdict change is needed.","tokens_in":21252,"tokens_out":10083,"duration_ms":119133,"concrete_test":"Evaluate the Padé form factors at representative phase-space points, e.g. x-tilde = 0.09 and 0.25 with MZZ = 400, 600, 800 and 1000 GeV, and compare the real and imaginary parts of F_AA and of the interference combination v_f^2 F_VV + a_f^2 F_AA against the full numerical two-loop results of refs. [51,52] (Agarwal and von Manteuffel). If the numerical points fall outside the quoted uncertainty band for F_AA or for the interference combination, the reconstruction is systematically biased; if they agree within the band, the central claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the two-loop top-quark form factors entering the Higgs-interference term are reconstructed with small, quantified uncertainty. This requires the Padé ansatz of eqs. (9)-(11), constrained only by the known LME up to 1/m_t^12 and the non-analytic threshold coefficients of Appendix A, to represent correctly the unknown analytic-in-zbar terms dropped in eq. (5). At one loop the same procedure is validated against the exact result, but at two loops no independent numerical value is available in this paper; the error bars in Figs. 3-5 are the spread of a selected family of pole-filtered Padé approximants, not an estimate of the omitted analytic terms. The vector form factor is explicitly distrusted above about 500 GeV, and the interference claim is rescued by the numerical dominance of F_AA shown in Fig. 5. The load-bearing assumption is therefore that F_AA is reconstructed accurately in the invariant-mass region where the interference signal matters. If the dropped analytic terms (or the assumed z^{-1} form of the small-mass rescaling) bias F_AA, the central conclusion inherits that bias, and no internal convergence check can reveal it because all approximants share the same input information.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the two-loop top-quark mass dependence of the continuum gg->ZZ box form factors that enter the interference with off-shell Higgs production. The authors compute new non-analytic threshold-expansion coefficients for the one- and two-loop vector and axial-vector form factors (Appendix A), and combine those with the known large-mass expansion through 1/m_t^12 in a conformal-mapping/Padé reconstruction with the small-mass rescaling of eq. (9). At one loop, they check the reconstruction against the exact analytic form factors and find good agreement over the whole plotted invariant-mass range (Fig. 2). At two loops, no independent numerical result is available; the validation is limited to the spread of a family of Padé approximants (eq. (14) and the discussion in Section IV) and to comparisons among approximants built from threshold expansions truncated at O(zbar^2), O(zbar^3) and O(zbar^4) (Fig. 4). The reconstruction of the vector form factor is explicitly reliable only up to about 500 GeV; the paper argues that the interference combination is nevertheless trustworthy to larger MZZ because the axial-vector form factor dominates the combination (Fig. 5). The paper concludes that it provides a new NLO prediction with small uncertainties at small and moderate MZZ.","tokens_in":21457,"tokens_out":12039,"duration_ms":129277,"significance":"If the two-loop reconstruction is accurate, this is a useful step for the LHC off-shell Higgs-width program: it provides a practical way to retain top-quark mass effects in gg->ZZ above the top threshold, where the LME alone fails. The concrete new threshold coefficients in Appendix A, the successful one-loop benchmark, the prior validation of the method against exact two-loop results for gg->HH [28], and the availability of a numerical implementation are clear strengths. The main limitation is the absence of an independent two-loop cross-check for the gg->ZZ form factors. The paper's central claim therefore rests on an internal consistency argument, and the quoted error bands do not directly cover the unknown analytic terms dropped in eq. (5). This is a serious but addressable concern: an external benchmark or a more conservative and clearly characterized uncertainty would settle it.","major_comments":[{"comment":"The quoted uncertainty is obtained as the mean and standard deviation of a family of Padé approximants whose input data are the same LME coefficients and the same threshold coefficients of Appendix A. This measures sensitivity to the free parameters a_R and to the polynomial degrees, but not the systematic error from the analytic-in-zbar terms that are dropped in eq. (5) and never computed. Because every member of the family shares the same missing information, a common bias in those terms would not show up in the scatter. The one-loop benchmark in Fig. 2 validates the procedure in a setting where the exact answer is known; it does not validate the two-loop reconstruction. Since the abstract and Section V claim a two-loop prediction with very small uncertainties, this gap is load-bearing. I recommend either providing an independent two-loop benchmark (e.g., comparing with an exact or numeric full-mass calculation, which the paper itself notes may be feasible in refs. [51,52]) or enlarging the uncertainty estimate to cover the dropped analytic terms and restating the NLO claim accordingly.","section":"Section III, eqs. (9)-(11); Section IV, eq. (14)"},{"comment":"The convergence study of the vector form factor shows that the O(zbar^2) approximant does not overlap with the O(zbar^3) and O(zbar^4) approximants on a significant part of the phase space, and the text agrees that F_VV should be distrusted above about 500 GeV. The argument that the interference combination is still trustworthy to arbitrarily large MZZ because F_VV is numerically tiny (Fig. 5) is only valid if the axial-vector form factor is reliable in that region. However, the axial-vector form-factor error bands in Fig. 3 also increase with MZZ, and no quantitative uncertainty is quoted for the interference combination at large MZZ. The sentence in Section IV stating that the interference prediction is trustworthy up to MZZ->infinity therefore goes beyond the error analysis shown. Please quote uncertainty estimates for the combination at representative high-MZZ points, or restrict the claim to the invariant-mass range where the error is under control.","section":"Section IV, Fig. 4 and Fig. 5"}],"minor_comments":[{"comment":"In the displayed combination, the second term should be a_t^2 |F_AA>; as printed both terms are labelled |F_VV>. The same typo appears in the text sentence before Fig. 5.","section":"Fig. 5 caption and the paragraph introducing Fig. 5"},{"comment":"The pole criterion removes approximants with poles in a region of the complex omega plane; please state how many of the 100 Padé variants are discarded by this filter in the representative phase-space points, and whether the quoted error bands are computed before or after this selection.","section":"Section IV, eq. (13)"},{"comment":"The figures normalize F_VV by z and F_AA by r_Z^2 without explanation in the text. Please state these normalizations explicitly in the captions or in the text, as the reader must otherwise infer them from the plot labels.","section":"Figs. 2 and 3"},{"comment":"The sentence 'the coefficients with m=0 and even n do not contribute to the imaginary part and are therefore not listed here' is clear, but it would help if the text also stated explicitly which coefficients are non-zero and which vanish, e.g., that b_i,ln^(n,1) and b_i,ln^(2n,m) are zero, to avoid ambiguity in the ancillary file.","section":"Appendix A, after eq. (A1)"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest about the vector-form-factor limitation, but the abstract and conclusions phrase the NLO result more strongly than the internal error analysis supports. My recommendation of major revision is intended to resolve the mismatch between the Padé-family spread and the claimed uncertainty. I would be willing to reconsider once the authors either supply an external check or reformulate the uncertainty and the MZZ->infinity statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe bottom line: this is a solid, honest methods paper that produces genuinely new two-loop threshold-expansion coefficients for the gg->ZZ box form factors and uses them, together with the known large-mass expansion, in a conformal-map/Padé reconstruction of the top-quark mass dependence. The LO check is strong: the same construction reproduces the exact one-loop result across the full MZZ range. That validates the machinery in a directly relevant case, not just in the gg->HH analog from the group's earlier paper.\n\nThe new physics message is conditional, though. At NLO there is no independent numerical benchmark in the paper; the uncertainty bands in Figs. 3-5 are the spread of a family of pole-filtered Padé approximants. That is a real estimate of the ambiguity within the ansatz, but it does not capture the effect of the unknown analytic terms dropped in eq. (5). The authors are explicit about this. They also state plainly that the vector form factor should not be trusted above about 500 GeV, and they rescue the interference prediction by showing numerically that the axial-vector form factor dominates the Higgs-interference combination. I think that rescue is reasonable: the coupling factor and the reconstructed size of F_AA make F_VV a small correction, so the interference claim is probably robust even if F_VV is off.\n\nThe soft spots are the usual ones for Padé reconstruction: the small-mass rescaling in eq. (9) is an ansatz with free parameters, and the pole criterion in eq. (13) is a filter, not a guarantee. The convergence of the vector form factor is genuinely slow. None of this is fatal to the stated conclusion, because the authors flag it. The main missing piece is the external two-loop check that they say is in preparation (refs. [51,52]); until that lands, the NLO numbers should be treated as conditional, which is exactly how the reader scored it.\n\nWho should read this: anyone working on off-shell H->ZZ and indirect Higgs-width constraints. It deserves a serious referee. I would send it to review, with the request that the authors, at revision, either add the numerical comparison or keep the conditional wording front and center. The paper earns its place; it just does not close the loop at NLO by itself.\n\nRecommendation: engage with it, cite it, send it to peer review.","headline":"New two-loop threshold coefficients for gg->ZZ with a credible but not yet externally validated Padé reconstruction; the axial-vector dominance rescue of the interference prediction is reasonable.","tokens_in":22034,"tokens_out":2259,"would_cite":true,"duration_ms":24289,"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":"The paper claims that conformal-mapped Padé approximants built from large-mass and threshold expansions reconstruct the two-loop top-quark mass dependence of the $gg\\to ZZ$ box form factors that enter off-shell Higgs interference, giving…","keywords":["top quark mass effects","gg to ZZ production","off-shell Higgs interference","two-loop amplitudes","Padé approximants","conformal mapping","threshold expansion","large-mass expansion"],"falsifier":"A direct numerical evaluation of the two-loop $gg\\to ZZ$ box amplitude with full top-quark mass dependence at representative phase-space points above the top threshold (for example $M_{ZZ}=400$–$600$ GeV at small and large transverse momentum) would settle the claim: if the real or imaginary part of either reconstructed form factor deviates from the Padé uncertainty band by more than the quoted error, the reconstruction is not reliable in that region.","tokens_in":20983,"feed_emoji":"⚛️","tokens_out":9149,"duration_ms":78808,"temperature":0.7,"pith_summary":"Measuring the Higgs-boson width at the LHC relies on the off-shell interference between the Higgs-mediated and continuum gluon-fusion production of Z-boson pairs. The continuum amplitude is known exactly only at leading order, and the two-loop top-quark mass dependence relevant for the interference was previously missing. This paper claims to supply it: from the known large-mass expansion through $1/m_t^{12}$ and a newly computed expansion around the top-pair threshold, conformal mapping and Padé approximants reconstruct the vector and axial-vector box form factors that enter the interference. At one loop the reconstruction reproduces the exact result, and at two loops it yields a first NLO prediction for the top-quark contribution with small uncertainties at small and moderate $M_{ZZ}$; the vector form factor alone is presented as reliable only below about 500 GeV, while the interference prediction, dominated by the axial-vector form factor, is argued to remain trustworthy to larger $M_{ZZ}$.","feed_headline":"Two-loop top-mass effects in gg→ZZ reconstructed via Padé","feed_subtitle":"New NLO prediction for the off-shell Higgs interference keeps small uncertainties up to moderate MZZ.","key_machinery":"The central object is the conformal-map Padé reconstruction: the variable $z=M_{ZZ}^2/(4m_t^2)+i0$ is mapped by $z=4\\omega/(1+\\omega)^2$ onto the unit disc, where the amplitude, after subtraction of threshold logarithms and separation into a constant part and a part proportional to $\\ln(-4z)$, is analytic and can be approximated by rational functions $[n/m](\\omega)$. The Padé coefficients are fixed by matching the known large-mass expansion through $1/m_t^{12}$ and the new threshold coefficients through $\\bar z^4$ (order $\\bar z^5$ for the massless-cut logarithm), and a rescaling factor $(1+a_{R,i}z)$ imposes the small-quark-mass asymptotic behavior required by chirality conservation. The uncertainty estimate comes from varying the rescaling parameters and the polynomial degrees and taking the mean and standard deviation of the approximant variants.","core_discovery":"The central claim is that the full top-quark mass dependence of the one- and two-loop box form factors in $gg\\to ZZ$ can be recovered from two local expansions: the large-mass expansion already known to $1/m_t^{12}$ and a new expansion around the top-pair threshold in $\\bar z=1-z$ with $z=M_{ZZ}^2/(4m_t^2)+i0$, computed through $\\bar z^4$ (and one order higher for the logarithm from massless cuts). The two expansions are combined with the conformal map $z=4\\omega/(1+\\omega)^2$ into Padé approximants $[n/m](\\omega)$, with subtraction functions that remove the threshold logarithms and a small-mass rescaling that enforces the correct $z\\to\\infty$ behavior. The authors show that at one loop the approximants agree with the full analytic amplitude over the whole $M_{ZZ}$ range, and at two loops they give a new prediction for the NLO interference form factor with small uncertainties at small and moderate $M_{ZZ}$. Because the axial-vector form factor dominates the interference, they argue the interference prediction remains trustworthy up to large $M_{ZZ}$ even though the vector form factor alone is reliable only below about 500 GeV.","pith_inferences":["Extending beyond the paper: if the reconstruction is as reliable as the internal convergence suggests, the same threshold-coefficient-plus-Padé strategy should transfer directly to other two-loop amplitudes (for example $gg\\to ZH$ or off-shell $gg\\to ZZ$) once the relevant threshold expansions are computed, because no new conceptual machinery is needed.","Extending beyond the paper: the paper's large-$M_{ZZ}$ trustworthiness argument rests on the numerical dominance of the axial-vector form factor; a dedicated check of the reconstructed vector form factor against the full numerical two-loop amplitude would be the most direct stress test of that argument.","Extending beyond the paper: a robust NLO interference with genuine top-mass dependence could shift indirect Higgs-width determinations by more than the current parametric uncertainty, since the off-shell region above the top threshold is where the width constraint is most sensitive."],"forward_implications":["At NLO, the top-quark contribution to the off-shell Higgs interference form factor can now be evaluated with realistic top-mass dependence instead of relying on a large-mass expansion that breaks down above the top threshold.","The new form factors combine directly with the known massless-loop virtual corrections and the one-loop real corrections to produce a complete NLO prediction for the interference in $gg\\to ZZ$.","The convergence checks show the reconstruction improves systematically as higher orders in the threshold expansion are included, so the large-$M_{ZZ}$ region should be refinable by adding more expansion terms rather than by a full analytic two-loop calculation.","The same large-mass-plus-threshold Padé construction can be applied to the remaining $gg\\to ZZ$ form factors that do not interfere with the Higgs signal, and to off-shell Z-boson production where the large-mass expansion is already known."],"supporting_citations":[{"why":"Establishes the conformal-mapping plus Padé method with threshold-expansion input that this paper adapts from gg to HH to gg to ZZ.","marker":"[28]"},{"why":"Supplies the known large-mass expansion coefficients through 1/m_t^12, the analytic double-triangle form factors, and the real NLO top-quark corrections.","marker":"[27]"},{"why":"Provides the massless two-loop amplitude and the large-mass expansion information for the continuum gg to ZZ amplitude that the reconstruction builds on.","marker":"[24]"},{"why":"Gives the exact one-loop gg to ZZ amplitude used to validate the Padé reconstruction at leading order.","marker":"[20]"},{"why":"Provides the expansion-by-regions technique used to compute the new threshold coefficients in the top-pair threshold expansion.","marker":"[42, 43]"},{"why":"Supplies the conformal transformation z = 4 omega/(1+omega)^2 that maps the z plane onto the unit disc and underlies the Padé ansatz.","marker":"[50]"}],"fun_headline_variants":["Padé approximants recover two-loop top-mass effects in gg→ZZ","Two-loop gg→ZZ: top-mass effects from expansions and Padé","Off-shell Higgs interference: two-loop top-mass effects via Padé","Conformal map plus Padé: two-loop top-mass in gg→ZZ","New NLO prediction for off-shell Higgs interference at two loops"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reconstruction's load-bearing premise is that the unknown analytic-in-$\\bar z$ terms dropped from the threshold expansion, together with the known large-mass coefficients through $1/m_t^{12}$, are sufficient for the conformal-map Padé ansatz to determine the full two-loop amplitude; this is checked only against the exact one-loop result and by agreement between approximants of different order.","fun_headline_variants_meta":{"raw":{"variants":["Padé approximants recover two-loop top-mass effects in gg→ZZ","Two-loop gg→ZZ: top-mass effects from expansions and Padé","Off-shell Higgs interference: two-loop top-mass effects via Padé","Conformal map plus Padé: two-loop top-mass in gg→ZZ","New NLO prediction for off-shell Higgs interference at two loops"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000428,"raw_usage":{"total_tokens":2176,"prompt_tokens":915,"completion_tokens":1261,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":531,"completion_tokens_details":{"reasoning_tokens":1167}},"tokens_in":531,"tokens_out":1261,"duration_ms":12242,"temperature":1.0,"reasoning_tokens":1167,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:53:28.589441+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct numerical evaluation of the two-loop $gg\\to ZZ$ box amplitude with full top-quark mass dependence at representative phase-space points above the top threshold (for example $M_{ZZ}=400$–$600$ GeV at small and large transverse momentum) would settle the claim: if the real or imaginary part of either reconstructed form factor deviates from the Padé uncertainty band by more than the quoted error, the reconstruction is not reliable in that region.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the exact one-loop gg to ZZ amplitude used to validate the Padé reconstruction at leading order."}],"review_version":1}