{"id":"17ec5d37-5438-4cd5-bd82-5fedb8ffcfa1","arxiv_id":"2507.11410","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The first NNLO QCD prediction for off-shell W+W-bbbar production with massive bottom quarks at the LHC, using a double-pole approximation for the two-loop virtual and an on-shell matching for non-factorisable corrections.","lead":"The paper computes the next-to-next-to-leading order QCD prediction for top-antitop pair production with leptonic decays and massive bottom quarks at the LHC, the first such off-shell NNLO result. The result shifts the NLO cross section by about 11%, and the method points toward a general way to handle processes whose two-loop amplitudes are too hard to compute exactly.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"On-shell matching sets the two-loop non-factorisable constant, so the physical-width NNLO result inherits any Gamma_t-model contamination; the fit ansatz should be tested against a model with quadratic power corrections.","rationale":"The reader correctly identifies the missing two-loop non-factorisable corrections and the on-shell matching assumption as the weakest point. My stress-test sharpens this: the constant C(2)_nf, which dominates the extracted correction at the physical width, is fixed by the Gamma_t -> 0 matching rather than by a direct two-loop calculation. Consequently the physical-width prediction inherits any error in the assumed functional form of power corrections. The paper's own NLO validation is strong, and the two fit variants (with and without a linear power correction) already bracket some of this uncertainty, which supports a conditional rather than outright rejection. The proposed test is a minimal extension of the existing fit that would settle whether the 152 fb systematic is complete. No verdict change is needed: the CONDITIONAL verdict already reflects this risk.","tokens_in":64265,"tokens_out":8279,"duration_ms":109548,"concrete_test":"Using the 7+2 width points shown in Fig. 14/15 for the gg and q qbar channels, redo the replica-based Gamma_t -> 0 fit with a model that adds a quadratic power-correction term E(2)(Gamma_t/mt)^2 to Eq. (76), and compute the resulting sigma_NNLO at the physical width. If the central value shifts by more than about 152 fb relative to Eq. (88), the quoted numerical uncertainty omits a genuine model dependence of the on-shell matching procedure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is not the factorisable two-loop construction but the inference of Delta_sigma_NNLO,H|non-fact. In Sec. 5.3.3 the coefficient C(2)_nf is not extracted from data at the physical width: it is effectively defined through Eq. (76) as the constant that makes the Gamma_t -> 0 extrapolation of the off-shell NNLOfact result coincide with the on-shell NNLO cross section. At Gamma_t = Gamma_phys this constant dominates the quoted 1107 fb non-factorisable contribution. The fitted quantity is the full off-shell minus on-shell difference, which also includes single-top/non-resonant interference, 4FS/5FS and bottom-mass effects, and truncation subtleties; the ansatz assumes all of this is described by A(2)_nf log^2 + B(2)_nf log + C(2)_nf + D(2)(Gamma_t/mt). Any Gamma_t-dependent contamination not of this form is absorbed into C(2)_nf and shifts sigma_NNLO at the physical width by an amount not covered by the +/-152 fb fit error. The NLO comparison validates the procedure to about 0.3% at one loop, but the NNLO non-factorisable term is O(1) of the NNLO correction, so a one-loop cross-check cannot bound it. The off-diagonal NNLO channels are exact but do not probe the diagonal-channel matching assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an NLO QCD calculation of pp -> W+W- b bbar production with leptonic decays and massive bottom quarks in the 4FS, implemented in the qT-subtraction framework and validated against an independent dipole-subtraction calculation. It then constructs the double-pole approximation (DPA) for the virtual corrections at NLO and shows that the DPA reproduces exact NLO results at the per-mille-to-percent level. The main new result is an approximate NNLO prediction for the inclusive cross section: the exact double-real and real-virtual contributions and the one-loop squared terms are combined with factorisable two-loop corrections built from polarised on-shell t-tbar amplitudes and massified top-decay form factors, while the non-factorisable two-loop contribution is inferred from the Gamma_t -> 0 limit through an on-shell matching procedure. The quoted NNLO cross section is 10623(55) +- 152 fb, corresponding to an NNLO/NLO K-factor of 1.108.","tokens_in":64589,"tokens_out":6442,"duration_ms":80474,"significance":"If the matching procedure is reliable, this is the first NNLO-quality prediction for off-shell top-quark pair production with decays in the 4FS with massive bottom quarks, and it would be a significant step towards a complete NNLO calculation for this process. The paper contains several genuine strengths: the exact NLO calculation is cross-checked against an independent implementation at the 0.04% (CKMP) and 0.2% (CMP) level; the NLO DPA is validated against exact NLO in fiducial cross sections and differential distributions; the rcut extrapolation is studied with a replica method and with alternative fit models; and the small-width extrapolation is validated at LO, at NLO, and in the off-diagonal NNLO channels that are computed exactly. The paper is also unusually transparent about the limitations of the non-factorisable two-loop inference. The main caveat is that the constant term of the non-factorisable two-loop correction is not computed but is fixed by the on-shell matching, so the final NNLO number inherits the model assumptions of that matching.","major_comments":[{"comment":"The central number is controlled by the constant C^(2)_nf of the non-factorisable two-loop corrections, but this constant is not extracted from data at the physical width. It is defined through Eq. (77) as the value that makes the Gamma_t -> 0 extrapolation of the off-shell NNLO_fact result coincide with the on-shell NNLO cross section. The fitted quantity is therefore the full off-shell-minus-on-shell difference, which also includes single-top/non-resonant contributions, 4FS/5FS effects, bottom-mass effects, and truncation subtleties. The ansatz in Eq. (73) assumes that all of these are described by A^(2)_nf log^2 + B^(2)_nf log + C^(2)_nf + D^(2)(Gamma_t/mt) plus power-suppressed terms. Any Gamma_t-dependent contamination not of this form is absorbed into C^(2)_nf and shifts sigma_NNLO at the physical width by an amount that is not covered by the quoted +-152 fb fit error. The NLO validation in this section checks the procedure at one loop, where the non-factorisable correction is relatively small, but at NNLO the non-factorisable contribution is O(1) of the NNLO correction, so the one-loop cross-check cannot bound the two-loop model error. I request a concrete sensitivity test: repeat the fit with additional power-correction terms, e.g. E(2)(Gamma_t/mt)^2 or (Gamma_t/mt) log(Gamma_t/mt), and report the resulting shift in C^(2)_nf and in sigma_NNLO. Without such a test, the statement that the numerical uncertainty is below 2% is not fully supported.","section":"Sec. 5.3.3, Eq. (76)"},{"comment":"The main result does not follow from its stated components. In Table 3 the all-channel values are sigma_NNLO_qT,DPA' = 10623(55) +- 152 fb and sigma_NNLO_fact = 9546(55) fb, while the last row gives Delta_sigma_NNLO,H|non-fact = 1107 +- 152 fb. The text in Sec. 5.4 says that the total NNLO cross section is obtained as the sum of sigma_NNLO_fact and Delta_sigma_NNLO,H|non-fact, but 9546 + 1107 = 10653, not 10623. This is a 30 fb (about 0.3%) discrepancy in the headline cross section. Please correct the table or the definition of the sum, and explain where the difference comes from.","section":"Sec. 5.4, Table 3"},{"comment":"The modified on-shell projection relaxes conservation of the invariant mass Q of the event, rescales the top-quark energies, and adjusts the initial-state momenta, and it is used over the entire phase space rather than only near threshold. This projection is part of the construction of the factorisable two-loop corrections and of the reweighting in Eq. (63), so any bias introduced by the projection can propagate into the fitted non-factorisable constant through the matching in Sec. 5.1. The NLO DPA validation is reassuring, but it does not directly bound the effect on the two-loop factorisable contribution, which is much larger in absolute terms. I ask for an explicit estimate of the projection dependence, for example by comparing the modified projection with the standard Q-conserving projection in the phase-space region where both are defined, and by reporting the change in Delta_sigma_NNLO,H|non-fact when the projection choice is varied.","section":"Appendix A, Sec. 3.1"}],"minor_comments":[{"comment":"The word 'perfomed' appears in the consistency-check paragraph; it should be 'performed'.","section":"Sec. 2.2.1"},{"comment":"The phrase 'off-shell skaddones momenta' appears to contain a stray word; it should presumably read 'off-shell momenta'.","section":"Appendix A"},{"comment":"The formula for the replica standard deviation appears garbled in the printed version (the sum over sqrt terms is not a standard deviation). Please clarify the exact expression used.","section":"Sec. 5.3.3, Eq. (85)"},{"comment":"The fitted coefficients A^(2)_nf, B^(2)_nf and D^(2) are shown only through correlation plots; reporting their central values and uncertainties in a table would make the model-dependence discussion more quantitative.","section":"Sec. 5.3.3, Fig. 16"}],"recommendation":"major_revision","confidential_remarks":"This is a 'towards NNLO' paper rather than a complete NNLO calculation, and the authors are suitably careful about that. The main issue is that the non-factorisable two-loop constant is fixed by construction through the on-shell matching, and the quoted 1.5% fit uncertainty does not include model error. If the authors provide a convincing model-variation test and fix the numerical inconsistency in Table 3, the paper could be acceptable for publication; without that, the central uncertainty claim is not fully supported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is the first calculation that puts an NNLO number on off-shell W+W-bbbar production with massive bottom quarks at the LHC. That alone makes it worth a referee's time. Second thing: the NNLO cross section is not derived from scratch; it is assembled from exact real and one-loop pieces, a factorisable two-loop virtual built from known on-shell ingredients, and a fitted non-factorisable contribution. The fit is the soft spot.\n\nWhat the paper does well is substantial. The exact NLO computation in qT subtraction is checked against dipole subtraction at the 0.04% and 0.2% level in two fiducial setups, and the NLO double-pole approximation is validated in detail—better than 1% in the bulk and a few percent in tails, with a clear discussion of where it fails. The massification of the two-loop top-decay form factors and the use of the polarised two-loop ttbar amplitudes are concrete, documented steps. The technical work on the rcut and small-width extrapolations is also serious: replica-based error estimates, explicit fit models, and power-correction terms. The paper is honest about being a first step rather than a finished NNLO prediction.\n\nThe real issue is the non-factorisable two-loop contribution. It is not computed. The authors assume the functional form A log^2(Gamma_t/mt) + B log(Gamma_t/mt) + C, and fix C by matching the Gamma_t -> 0 limit to the known on-shell ttbar NNLO cross section. At the physical width this fitted constant is roughly 1100 fb, a large part of the NNLO correction. Any Gamma_t-dependent contamination—single-top interference, 4FS/5FS differences, power corrections not captured by the ansatz—gets absorbed into C and shifts the central value by an amount not covered by the quoted +-152 fb fit error. The NLO validation tests the procedure at one loop, and the off-diagonal NNLO channels are exact, but neither probes the diagonal-channel two-loop non-factorisable term. The authors do include a linear power-correction term and compare fit variants, which is good, but it does not close that gap. The scale variation band of about +-5% is a more honest estimate of the total theory uncertainty than the sub-2% numerical error.\n\nBottom line: this is a carefully built proof of concept, and the NLO blocks are solid. The paper deserves peer review and would be useful to top-quark phenomenologists and NNLO method developers. A serious referee should push the authors to test the fit ansatz against a model with additional power corrections, or otherwise bound the contamination into C. I would not take the central NNLO number as a final prediction, but I would cite it and would bring it to a reading group.","headline":"First NNLO number for off-shell W+W-bbbar with massive bottom quarks, but the fitted non-factorisable two-loop piece is load-bearing and only indirectly validated.","tokens_in":65175,"tokens_out":2122,"would_cite":true,"duration_ms":29837,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["12.38.Bx","14.65.Ha"],"model":"deepseek-v4-flash","headline":"This paper delivers the first NNLO QCD prediction for off-shell top-quark pair production with leptonic decays, estimating the missing two-loop virtual corrections through the double-pole approximation anchored to the on-shell cross…","keywords":["NNLO QCD","top-quark pair production","off-shell effects","double-pole approximation","qT subtraction","non-factorisable corrections","W+W- b bbar production","LHC physics"],"falsifier":"A direct calculation of the genuine two-loop non-factorisable contributions — soft-gluon exchanges linking the $t\\bar t$ production stage to the top or anti-top decays, i.e., two-loop six-point integrals with internal masses — in the gluon-fusion channel would settle the claim: if the resulting constant differs from the fitted $C^{(2)}_{\\rm nf}$ by more than the quoted ~1.5% systematic, the on-shell matching has fixed the wrong value. A cheaper immediate check is to rerun the $\\Gamma_t \\to 0$ fit restricted to the smallest widths ($\\Gamma_t/\\Gamma_t^{\\rm phys} \\le 0.1$), where the logarithmic ansatz should dominate; a significant shift of the extracted constant would signal contamination by power corrections beyond the fitted linear term.","tokens_in":64047,"feed_emoji":"⚛️","tokens_out":14082,"duration_ms":152323,"temperature":0.7,"pith_summary":"This paper is trying to establish that NNLO QCD accuracy is attainable for the full off-shell process $pp \\to e^+\\nu_e b\\,\\mu^-\\bar\\nu_\\mu\\bar b$ — the clean \"golden channel\" of top-quark pair production — even though the exact two-loop amplitudes for such a $2\\to 6$ process are beyond current techniques. The strategy is to compute every other ingredient exactly and to supply only the genuine two-loop virtual piece through the double-pole approximation: factorisable corrections built from polarised two-loop on-shell $t\\bar t$ production amplitudes and two-loop top-decay form factors, plus non-factorisable corrections fixed by matching the off-shell result to the known on-shell $t\\bar t$ NNLO cross section as the top-quark width goes to zero. The paper claims an inclusive NNLO cross section of $\\sigma = 10623(55)\\pm152$ fb at 13 TeV, an 11% increase over NLO, with numerical uncertainty below 2%, smaller than the roughly 5% residual perturbative uncertainty. A sympathetic reader would care because this is the first NNLO-level prediction for the dilepton $t\\bar t$ final state that keeps top-quark off-shell effects, $tW$ interference, and finite-width effects consistently, and it lays out a path for approximating two-loop amplitudes in similarly complex processes.","feed_headline":"Off-shell top-pair production computed at NNLO for the first time","feed_subtitle":"Two-loop corrections estimated by the double-pole method add 11% to the NLO rate, with numerical error below 2%.","key_machinery":"The load-bearing device is the double-pole approximation (DPA) for the two-loop virtual amplitude: the off-shell amplitude is expanded keeping only its double-resonant $t\\bar t$ topology, with the residue evaluated on projected on-shell momenta, so production and decay factorise. The factorisable part multiplies polarised tree-level production and decay amplitudes by their two-loop corrections; the non-factorisable part is assumed to have the one-log-per-loop form $\\Delta\\sigma_{\\rm NNLO,H}^{\\rm non-fact} = A^{(2)}_{\\rm nf}\\log^2(\\Gamma_t/Q_h) + B^{(2)}_{\\rm nf}\\log(\\Gamma_t/Q_h) + C^{(2)}_{\\rm nf} + O(\\Gamma_t/Q_h)$, whose unknown coefficients are extracted by computing the cross section at several artificially small top-quark widths, fitting the residual width dependence, and matching the $\\Gamma_t\\to 0$ limit to the known on-shell $t\\bar t$ NNLO cross section. Two supporting mechanisms carry the accuracy: massification, which restores bottom-quark mass logarithms in the two-loop production and decay amplitudes obtained in the massless limit, and $q_T$-subtraction with an $r_{\\rm cut}\\to 0$ extrapolation, which handles the infrared singularities of the $2\\to 6$, $2\\to 7$ and $2\\to 8$ real and real-virtual contributions.","core_discovery":"The paper's central claim is that the inclusive cross section for $W^+W^-b\\bar b$ production with leptonic decays and massive bottom quarks can be computed at NNLO accuracy by treating only the genuine two-loop virtual contribution in the double-pole approximation. The factorisable two-loop corrections are constructed from available polarised two-loop on-shell $t\\bar t$ production amplitudes combined with two-loop top-quark decay amplitudes obtained from heavy-to-light form factors, with bottom-quark mass effects restored by massification. The non-factorisable corrections, which are not directly computable, are fixed by requiring that the off-shell cross section reproduces the on-shell $t\\bar t$ NNLO cross section times branching ratios in the $\\Gamma_t \\to 0$ limit, using the known functional form $A^{(2)}_{\\rm nf}\\log^2(\\Gamma_t/Q_h) + B^{(2)}_{\\rm nf}\\log(\\Gamma_t/Q_h) + C^{(2)}_{\\rm nf}$ for their width dependence. With this construction the paper obtains $\\sigma_{\\rm NNLO} = 10623(55)\\pm 152$ fb at 13 TeV, an 11% upward shift of the NLO result, a numerical uncertainty below 2%, and residual perturbative uncertainties of about $+3.2\\%/-4.6\\%$. Along the way it establishes that the double-pole approximation reproduces the exact NLO result at the per-mille to few-percent level across fiducial and differential observables, and that the non-factorisable two-loop corrections, though roughly 20% of the factorisable ones, shift the NNLO correction at order one because other contributions largely cancel.","pith_inferences":["My inference: the same construction — exact real and real-virtual pieces, DPA for the genuine two-loop virtual, on-shell matching in the zero-width limit — is portable to other unstable-particle processes (such as $t\\bar t H$ or single-top channels) for which exact two-loop amplitudes are unavailable but an on-shell NNLO anchor exists.","My inference: the headline number is only as strong as the fitted constant $C^{(2)}_{\\rm nf}$; a future direct computation of the non-factorisable two-loop diagrams that shifts that constant would move the predicted cross section by about the quoted 1.5% systematic — the method's conclusion, not its architecture, is the vulnerable piece.","My inference: because the fit leaves the single-log coefficient $B^{(2)}_{\\rm nf}$ weakly constrained, an analytic computation of the double- and single-log coefficients from soft-gluon exponentiation on the resonant propagators would tighten the prediction more efficiently than additional Monte Carlo statistics.","My inference: extending the $\\Gamma_t \\to 0$ fit bin-by-bin would test whether the non-factorisable two-loop corrections change sign or grow in kinematic tails, as the one-loop study in the bottom-jet-tagged setup suggests; differential NNLO predictions for this process would then be the natural next target."],"forward_implications":["The inclusive $e^+\\nu_e \\mu^-\\bar\\nu_\\mu b\\bar b$ cross section at 13 TeV is predicted at $\\sigma_{\\rm NNLO} = 10623(55)\\pm 152$ fb, an 11% upward correction over NLO with numerical uncertainty below 2%.","The double-pole approximation, validated at NLO to per-mille accuracy in fiducial cross sections and a few percent in distribution tails, supplies a stand-in for two-loop amplitudes in processes where exact multi-loop results are out of reach.","Non-factorisable two-loop corrections matter at order one for the NNLO shift because other contributions largely cancel, so a DPA-based NNLO prediction that omits them would be qualitatively incomplete.","After removing spurious finite-width terms, the matched prediction gives $\\sigma^{\\Delta\\rm trunc}_{\\rm NNLO} = 10278(55)\\pm152$ fb, allowing consistent future comparisons with narrow-width treatments of $t\\bar t$ plus $tW$ production.","The exact NLO computation of the process with a resolved jet in the 4-flavour scheme with massive bottom quarks is the first of its kind and provides a new handle for jet-based background modelling."],"supporting_citations":[{"why":"Supplies the polarised two-loop on-shell $t\\bar t$ production amplitudes that seed the factorisable two-loop corrections of the off-shell calculation.","marker":"[60]"},{"why":"Provides the heavy-to-light two-loop quark form factors from which the top-quark and anti-top-quark decay amplitudes are built via massification.","marker":"[61]"},{"why":"Gives the on-shell $t\\bar t$ NNLO cross section that anchors the $\\Gamma_t\\to 0$ matching fixing the non-factorisable constant.","marker":"[2]"},{"why":"Extends the mass factorisation formula to several heavy flavours, restoring bottom-quark mass effects in the two-loop production and decay amplitudes.","marker":"[69]"},{"why":"Defines the $q_T$-subtraction formalism used to regularise the infrared singularities of the off-shell $2\\to 6$ computation.","marker":"[49]"},{"why":"Provides the extended soft-gluon approximation used for the non-factorisable one-loop corrections whose width-dependent structure the two-loop fit assumes.","marker":"[77]"},{"why":"Supplies the fiducial CMP setup and the fully differential narrow-width NNLO results used for comparisons.","marker":"[25]"},{"why":"Earlier NLO computation in the 4-flavour scheme with massive bottom quarks whose fiducial setup and cross section serve as the NLO validation target.","marker":"[41]"}],"fun_headline_variants":["First NNLO off-shell top-pair production: 11% shift, sub-2% error","Off-shell top-pair production at NNLO: double-pole trick yields 11%","NNLO top-pair with off-shell effects: exact NLO, DPA for two-loop","Off-shell top-pair rate goes NNLO: 11% up, <2% numerical error","Full NNLO for top-pair decays: factorisable DPA, non-factorisable fixed"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result rests on the assumption that the missing two-loop non-factorisable corrections have exactly the assumed dependence on the top-quark width — a squared logarithm, a single logarithm, and a constant term — and that the constant fixed by matching to the known on-shell $t\\bar t$ cross section in the zero-width limit is the correct one, a check performed so far only at lower orders and in the subdominant channels.","fun_headline_variants_meta":{"raw":{"variants":["First NNLO off-shell top-pair production: 11% shift, sub-2% error","Off-shell top-pair production at NNLO: double-pole trick yields 11%","NNLO top-pair with off-shell effects: exact NLO, DPA for two-loop","Off-shell top-pair rate goes NNLO: 11% up, <2% numerical error","Full NNLO for top-pair decays: factorisable DPA, non-factorisable fixed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000217,"raw_usage":{"total_tokens":1569,"prompt_tokens":1212,"completion_tokens":357,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":828,"completion_tokens_details":{"reasoning_tokens":236}},"tokens_in":828,"tokens_out":357,"duration_ms":4180,"temperature":1.0,"reasoning_tokens":236,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:09:19.125517+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct calculation of the genuine two-loop non-factorisable contributions — soft-gluon exchanges linking the $t\\bar t$ production stage to the top or anti-top decays, i.e., two-loop six-point integrals with internal masses — in the gluon-fusion channel would settle the claim: if the resulting constant differs from the fitted $C^{(2)}_{\\rm nf}$ by more than the quoted ~1.5% systematic, the on-shell matching has fixed the wrong value. A cheaper immediate check is to rerun the $\\Gamma_t \\to 0$ fit restricted to the smallest widths ($\\Gamma_t/\\Gamma_t^{\\rm phys} \\le 0.1$), where the logarithmic ansatz should dominate; a significant shift of the extracted constant would signal contamination by power corrections beyond the fitted linear term.","supporting_citations":[{"cited_title":"Polarized double-virtual amplitudes for heavy-quark pair production","cited_arxiv_id":"1712.08075","evidence_quote":"Supplies the polarised two-loop on-shell $t\\bar t$ production amplitudes that seed the factorisable two-loop corrections of the off-shell calculation."}],"review_version":1}