{"id":"2e50f84c-c887-407c-80f2-0629d02d530a","arxiv_id":"2412.20578","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The two-loop Reggeon-gluon-Reggeon (Lipatov) vertex in QCD is determined in dimensional regularization through finite terms and expressed in single-valued polylogarithms.","lead":"This paper computes the two-loop QCD Lipatov vertex, the effective coupling describing the emission of a gluon from a Reggeon in high-energy scattering. The result is a key ingredient for next-to-next-to-leading-logarithmic (NNLL) predictions in QCD, and it is validated across three scattering channels and against the supersymmetric Yang-Mills limit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pole/cut criterion (8.3) is the load-bearing assumption: for 2→3 it is conjectural, and if planar MR mixes into the cut at NNLL, eqs (8.36)-(8.42) are a scheme-dependent matching coefficient rather than the factorized Regge-pole vertex.","rationale":"I agree with the reader's identification of the weakest assumption: eq (8.3), the pole/cut criterion, tested for 2→2 through four loops but conjectural for 2→3. The reading of the extraction confirms that this is the single load-bearing point. The SR/MR scheme vertex is a matching coefficient by construction, and its extraction is guarded by strong internal checks: the three-channel agreement tests the channel-dependent non-planar MR parts (F^{qq}_{nonfact}, F^{qg}_{nonfact}), and through the gg channel's (36/72)F_fact it also tests the universal planar function; the maximal-weight agreement with the independently computed sYM vertex [21] and the soft-limit finiteness are further non-trivial checks. What is not tested at two loops is the identification of the cut as entirely non-planar: for qq and qg the planar part cancels between v^{SR(2)} and the conversion (8.13), so any planar component reassigned between pole and cut shifts the vertex universally without breaking any two-loop check. The authors themselves flag this gap (§8.1, conclusions) and provide the decisive test, eq (8.19), a prediction for the three-loop planar MR whose two independent rapidity-logarithm structures are fixed by data already determined at lower orders. This concern does not change the reader's verdict: the computation appears sound, the SR/MR scheme result is robust, and the pole/cut scheme interpretation of eqs (8.36)-(8.42) is conditional on (8.3), to be settled by the three-loop computation. Verdict remains CONDITIONAL.","tokens_in":84604,"tokens_out":33476,"duration_ms":308175,"concrete_test":"Perform the three-loop test proposed by the authors, eq (8.19): compute the planar (leading-colour) part of the three-loop odd-odd octet-octet 2→3 amplitude in MRK using the shock-wave formalism of §6-7 (transitions RgR3, R3gR, R3gR3 at three loops with one-loop corrections to the Reggeon states and emission vertices), and compare with the prediction M^{planar MR}|_{3-loops} = (αs/π)³(π²N_c³/18){[η1 αg^(1)(t1)+η2 αg^(1)(t2)]F_fact/4 − (η1+η2)(rΓ)³(S_A−S_B)}M^{tree}, with F_fact from (7.97) and S_A, S_B from (8.8). Agreement through the finite terms in ε confirms (8.3) for 2→3 and validates eqs (8.36)-(8.42) as the Regge-pole vertex; any discrepancy shows planar MR contaminates the cut and the extracted vertex mixes pole and cut.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result is the two-loop Lipatov vertex in the pole/cut scheme, eqs (8.36)-(8.42). The extraction has two steps, and only the second is fragile. Step one determines v^{SR(2)} as a matching coefficient by subtracting the multi-Reggeon amplitude from the full 2→3 amplitude (eq 8.9); this is well defined, and the three-channel agreement, the maximal-weight match to the sYM vertex, and the soft-limit finiteness of the function basis give strong checks of the MR computation and of the amplitudes. Step two converts to the pole/cut scheme via eq (8.13) and requires the criterion (8.3): the Regge pole is the single-Reggeon contribution plus the planar part of the multi-Reggeon contributions, while the cut is entirely non-planar. For 2→2 this was tested through four loops in refs [23,24]; for 2→3 it is a conjecture, as the authors state in §8.1 and the conclusions. No two-loop datum can test it: in the qq and qg channels F_fact cancels identically between v^{SR(2)} and the conversion (8.13), so reassigning any planar component between pole and cut produces a universal shift in v^{(2)} that all two-loop checks still pass. The gg channel's non-planar piece is (36/72)F_fact (eq 7.96), so it does test F_fact against qq/qg, but that only checks internal consistency of the MR computation, not the pole/cut identification. Hence the result is guaranteed to be the pole/cut matching coefficient; it is the Regge-pole Lipatov vertex only if (8.3) holds for 2→3.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper determines the two-loop Reggeon-gluon-Reggeon (Lipatov) vertex of QCD by matching 2→3 scattering amplitudes in multi-Regge kinematics (MRK) to a multi-Reggeon effective-theory computation. Three building blocks are combined. First, the non-planar two-loop QCD amplitudes for gg→ggg, gq→ggq and qq→qgq are expanded around the MRK limit using the differential-equation method for pentagon functions together with p-adic reconstruction of the rational coefficients. Second, the shock-wave based multi-Reggeon framework is extended to 2→3 processes; the relevant two-loop transitions in the odd-odd signature sector (R3gR3, RgR3 and R3gR) are computed explicitly, alongside all one-loop transitions. Third, the Regge-pole factorization formula, with impact factors and Regge trajectory taken from earlier 2→2 determinations, is used to extract the vertex. The matching first gives the vertex in the single-Reggeon (SR/MR) scheme and then, through eq. (8.13), in the pole/cut scheme defined by criterion (8.3) of refs. [23,24]. The final result, eqs. (8.36)–(8.42), is expressed in a compact basis of single-valued polylogarithms with manifest target-projectile symmetry and soft-limit finiteness; the one-loop vertex is given through O(ε⁴). Extensive checks are reported: numerical verification of the MRK expansions at x=10⁻¹⁰, agreement of the one-loop vertex with the known soft limit and with ref.","tokens_in":84847,"tokens_out":33429,"duration_ms":312384,"significance":"If correct, this is the first complete two-loop QCD Lipatov vertex, closing the last missing Regge-pole ingredient (beyond impact factors and trajectory) needed for NNLL predictions from MRK factorization in 2→3 scattering, and it constitutes a key building block for the NNLO BFKL kernel. The technical achievement is substantial: the MRK expansion of two-loop five-point amplitudes, the extension of the multi-Reggeon effective theory to real-gluon emission, and the construction of a manifestly finite, symmetry-adapted function basis. The paper ships analytic expressions in ancillary files and, notably, makes a concrete falsifiable three-loop prediction, eq. (8.19). The main caveat — correctly identified by the authors — is that the identification of the extracted coefficient with the factorized Regge-pole vertex relies on the conjectural pole/cut criterion (8.3) for 2→3; as discussed in the major comments, the weight-4 component of the result is robust against this caveat, while the lower-weight components are scheme-dependent within a one-parameter family. Even under this caveat, the result is a well-defined and important matching coefficient.","major_comments":[{"comment":"The pole/cut criterion (8.3) is conjectural for 2→3, and no quantity determined in this paper tests it. The extraction of v^{SR(2)} in eq. (8.9) is well defined, but the conversion in eq. (8.13) encodes the assumption that all planar multi-Reggeon (MR) contributions belong to the Regge pole. The planar MR contribution is N_c² F_fact, eq. (7.100). In the qq and qg channels its coefficient in the MR amplitude, eqs. (7.94)–(7.95), cancels exactly against the conversion term in eq. (8.13); in the gg channel the residual coefficient is 36, eq. (7.96), so that channel only verifies the internal consistency of the MR computation against the qq/qg channels. Reassigning any fraction of the planar MR between pole and cut shifts v^{(2)} by a universal function — a multiple of the bracket in eq. (8.13) — which preserves all of the stated checks: the three-channel agreement (the shift is process-independent), the maximal-weight sYM match (the shift has weight at most 2 in the finite part), and soft-limit finiteness (the shift is built from the same finite function basis). Hence eqs. (8.36)–(8.42) are guaranteed to be the pole/cut matching coefficient; they are the Regge-pole Lipatov vertex only if (8.3) holds for 2→3. The manuscript should state this residual ambiguity explicitly and quantify it: a reassignment of the planar MR shifts v^{(2)} by a universal function of transcendental weight ≤ 2 (plus the associated 1/ε² and 1/ε terms), so that the weight-4 part is scheme-robust while the lower-weight QCD-specific parts, including all n_f-dependent terms, are conditional on (8.3).","section":"§8.1, eqs. (8.3), (8.9)–(8.13), (7.94)–(7.100)"},{"comment":"The framing of the evidence for criterion (8.3) overstates its logical force. Section 8.1 presents the universality of the planar MR contributions, eq. (7.100), as a “test” of the pole/cut criterion; universality is necessary for the planar MR to be absorbed into a universal pole, but it does not exclude a universal cut component, so it is not a test of the pole/cut identification. The first genuinely discriminating test is the three-loop planar-MR prediction, eq. (8.19), which the authors themselves propose; this is a strength and should be presented as such, but the distinction between a necessary-condition check and a falsifiable test should be drawn explicitly. The abstract's unqualified claim to “determine the QCD Lipatov vertex” should be qualified in the abstract and introduction: the result determines the vertex in the pole/cut scheme defined by (8.3), whose conjectural status for 2→3 (in contrast with the four-loop tests available for 2→2) should be stated up front. The conclusions should also note that the advertised application, the NNLO BFKL kernel, inherits this conditionality for all components except the weight-4 one.","section":"Abstract; §8.1 after eq. (7.100); §9"}],"minor_comments":[{"comment":"The one-loop extraction uses only the gg→ggg channel, and no MR subtraction is performed. This is valid because the one-loop R2gR2 transition, eq. (7.18), is carried by T(−−), whose action on the gg tree-level colour structure c[8a,8a] has no [8a,8a] component, eq. (7.26), in contrast with the qq case, eq. (7.23). Spell this out where eq. (5.4) is introduced, since a reader comparing with eqs. (7.94)–(7.96) might otherwise wonder why the two-loop extraction needs MR subtraction while the one-loop one does not.","section":"§5, eq. (5.4)"},{"comment":"The statement that odd-signature factorization holds through next-to-leading logarithms should be qualified with respect to the SR/MR scheme introduced in §6: in that scheme the one-loop R2gR2 transition already contributes to the qq odd-odd octet channel, eqs. (7.18) and (7.23), so the factorization statement holds for the physical amplitude, with the SR/MR split understood as a bookkeeping convention.","section":"§4.1, eq. (4.10)"},{"comment":"The rational prefactors in the vertex contain denominators vanishing at q ≡ z − z̄ = 0 as well as at p ≡ 1 − z − z̄ = 0. The paper establishes finiteness in the soft limit p → 0 explicitly, but the regularity at q = 0 (the tr5 → 0 line of Fig. 2) should be stated and, ideally, shown to follow from the same kind of cancellation between the φ_i and μ_j terms.","section":"§8.2.3, eqs. (8.36a)–(8.42)"},{"comment":"The comparison with the literature for the one-loop vertex is quoted through O(ε²), ref. [62], while the paper presents v^{(1)} through O(ε⁴); state explicitly which orders are new and why the literature comparison stops at O(ε²).","section":"§9 and §5"},{"comment":"Please remove the “bracehtip” artifacts in eqs. (8.3) and (B.14), and repair the garbled colour indices in eqs. (7.4)–(7.5) and (7.58); these appear to be conversion artifacts and currently obstruct the reading of those equations.","section":"General typesetting"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically remarkable and the computation appears extremely well cross-checked, including agreement with an independent group's recent results [26] for the MR transitions and the finite remainders. The conditional element is structural and is acknowledged by the authors: the pole/cut criterion (8.3) is conjectural for 2→3, and no quantity computed here tests it, because the planar MR contribution cancels out of the final vertex up to a universal lower-weight shift; the three-loop prediction (8.19) is the first discriminating test. I recommend major revision, primarily to require the manuscript to frame the central claim precisely — “the vertex in the pole/cut scheme defined by (8.3)” — and to quantify the residual scheme ambiguity (universal, weight ≤ 2 shift; weight-4 component robust). I would not want the revision to dilute the presentation of the very valuable technical results. It would be useful if the final vertex expressions could be compared directly, term by term, with the independent computation of ref. [26], beyond the already-reported agreement of the remainders and MR transitions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nQuick take: the two-loop Lipatov vertex extraction is real and probably correct as an SR/MR matching coefficient, but the headline claim that this is the Regge-pole Lipatov vertex rests on the pole/cut criterion (8.3), which the authors themselves flag as a conjecture for 2→3. Two-loop data cannot test it. Still, this is a serious, careful computation that deserves a serious referee.\n\nWhat's actually new: the two-loop vertex through finite terms (eqs 8.36–8.42), the one-loop vertex through O(ε^4), analytic MRK limits of the non-planar two-loop qq, qg, gg amplitudes, and the new multi-Reggeon transitions (RgR3, R3gR, R3gR3). They also build a basis of single-valued polylogs that makes target-projectile symmetry and reality properties manifest, and they derive a three-loop planar MR prediction (8.19) as a future test.\n\nThe paper does several things well. The cross-checks are genuine: three-channel agreement, maximal-weight match to the N=4 sYM vertex, soft-limit behaviour to all orders in ε at one loop, agreement of the one-loop vertex with known results through O(ε^2), and numerical validation of the MRK amplitude expansions at x=10^-10. The vertex is extracted by matching exact amplitudes, not by fitting a target; the trajectory and impact-factor inputs come from earlier 2→2 work, so there is no circularity. The reported agreement with the concurrent independent computation [26] is an additional plus.\n\nThe soft spot is the pole/cut criterion. For 2→2 it was tested through four loops; for 2→3 it is a conjecture. The stress-test point holds: in qq and qg, F_fact cancels between v^{SR(2)} and the conversion (8.13), so any reassignment of planar MR between pole and cut leaves the extracted two-loop vertex unchanged. The gg channel fixes the coefficient of F_fact relative to qq/qg, but that is internal consistency of the MR computation, not evidence for the pole/cut identification. So the guaranteed result is the SR/MR matching coefficient; the pole/cut vertex is physical only if (8.3) holds. The authors say this themselves in §8.1 and the conclusions, which is to their credit, but it should be stated plainly to readers.\n\nThis paper is for the high-energy QCD community: people working on Regge factorization, BFKL, and multi-loop amplitudes in MRK. It moves the frontier a real step forward. I would send it to a serious referee, with specific attention to §8.1 and the conversion (8.13).","headline":"Serious two-loop computation: the Lipatov vertex is almost certainly right as an SR/MR matching coefficient, but the pole/cut-scheme interpretation rests on a conjecture that two-loop data cannot test.","tokens_in":85553,"tokens_out":3888,"would_cite":true,"duration_ms":38760,"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 determines the two-loop QCD Reggeon-gluon-Reggeon (Lipatov) vertex in dimensional regularization through finite terms, extracted from the odd-odd colour component of 2→3 amplitudes in multi-Regge kinematics after subtracting…","keywords":["Lipatov vertex","multi-Regge kinematics","Regge pole and cut","two-loop QCD amplitudes","shock-wave formalism","single-valued polylogarithms","BFKL","N=4 super Yang-Mills"],"falsifier":"Compute the three-loop planar multi-Reggeon contribution to the octet-octet component of a 2→3 amplitude at NNLL accuracy and compare it with the prediction in eq. (8.19); any deviation would show that planar multi-Reggeon exchanges mix into the Regge cut, invalidating the pole/cut extraction of the two-loop vertex.","tokens_in":84251,"feed_emoji":"⚛️","tokens_out":6047,"duration_ms":63512,"temperature":0.7,"pith_summary":"The paper determines the two-loop Reggeon-gluon-Reggeon (Lipatov) vertex of QCD, the effective vertex for emitting a real gluon from a Reggeized gluon, in dimensional regularization through finite terms. It does so by computing the two-loop gg→ggg, gq→ggq, and qq→qgq amplitudes in multi-Regge kinematics from known general-kinematics amplitudes, and matching them against a multi-Reggeon effective-theory computation. The matching requires separating the factorizing Regge-pole part from the non-planar Regge cut; with that separation, the vertex follows from any of the three channels and its maximal-transcendental-weight part agrees with the N=4 super Yang-Mills vertex. The one-loop vertex is also obtained through O($ε^{4}$). These results complete the ingredients for next-to-next-to-leading logarithmic Regge-factorization predictions in 2→3 scattering.","feed_headline":"Two-loop QCD Lipatov vertex determined","feed_subtitle":"Three partonic channels in multi-Regge kinematics now fix the vertex through finite terms.","key_machinery":"The load-bearing object is the Reggeon field W, obtained by expanding infinite lightlike Wilson lines in the shock-wave formalism: a single W carries octet colour and odd signature and represents one Reggeon, while products of W's represent multi-Reggeon states. Rapidity evolution is generated by the Balitsky-JIMWLK Hamiltonian, and real gluon emission is described by vertices W→W+g, W→WW+g, WW→WW+g, W→WWW+g, and so on. The pole/cut criterion, eq. (8.3), assigns the single-Reggeon transition plus the planar part of multi-Reggeon transitions to the Regge pole and the non-planar remainder to the Regge cut; the extraction uses the universality of the planar part across the three partonic channels to make the definition of the vertex unambiguous.","core_discovery":"The central result is the two-loop Lipatov vertex v(2)(t1,t2,|p4|2) in the pole/cut scheme, given in eqs. (8.36)–(8.42) for its dispersive and absorptive parts. It is extracted by matching the odd-odd, octet-octet components of the gg→ggg, gq→ggq, and qq→qgq amplitudes at two loops in multi-Regge kinematics to the factorization formula after subtracting the multi-Reggeon (cut) contribution, which is computed from the shock-wave formalism. The maximal-weight contribution matches the N=4 super Yang-Mills vertex, eqs. (8.22)–(8.23); the lower-weight pieces involving Nc and nf are the genuinely new QCD terms. The paper also provides the one-loop vertex through O($ε^{4}$).","pith_inferences":["The proposed three-loop planar multi-Reggeon prediction, eq. (8.19), is a sharp test of the pole/cut conjecture: an explicit computation there would settle whether the extracted two-loop vertex is genuinely the factorized Lipatov vertex or an admixture of pole and cut.","The same effective-theory setup should extend to 2→4 amplitudes in multi-Regge kinematics, where planar Regge cuts first appear; in the planar limit this could connect the Reggeon description to the remainder functions of super Yang-Mills theory.","The manifestly finite soft limit and the cancellation of spurious rational poles are structural properties any higher-loop extension of the vertex will have to reproduce, so they can serve as a practical constraint in future computations."],"forward_implications":["The two-loop Lipatov vertex completes the Regge-pole sector for 2→3 scattering at NNLL accuracy: together with the three-loop gluon Regge trajectory and two-loop impact factors, the factorizing part of the amplitude is now fully determined.","The result provides a key ingredient for an NNLO BFKL kernel, since that kernel requires the interference of two-loop and one-loop Lipatov vertices together with multi-particle central emission.","Agreement among gg→ggg, gq→ggq, and qq→qgq channels is an internal consistency check of the extraction, and agreement of the maximal-weight part with the N=4 super Yang-Mills vertex checks the transcendental structure.","The one-loop vertex through O(ε^4), with its spurious soft poles cancelled by transcendental functions, is now available for future three-loop 2→3 computations."],"supporting_citations":[{"why":"Supplies the pole/cut criterion separating Regge pole from Regge cut, and the two-loop impact factors and three-loop trajectory used in the extraction.","marker":"[23,24]"},{"why":"Establishes the shock-wave W-field Reggeon formalism and the real-emission vertices on which the multi-Reggeon computation is based.","marker":"[17]"},{"why":"Provides the multi-Regge kinematics expansion of five-point amplitudes in super Yang-Mills and the single-valued polylogarithm basis used throughout.","marker":"[21]"},{"why":"Provides the two-loop five-point QCD amplitudes in general kinematics from which the multi-Regge limit is taken.","marker":"[63–69]"},{"why":"Defines the one-loop Lipatov vertex and the factorization structure that the two-loop extraction extends.","marker":"[55]"},{"why":"Reports the final results of the multi-Reggeon computation whose full derivation is presented in this paper.","marker":"[25]"},{"why":"Gives an independent extraction of the Lipatov vertex used for comparison and cross-checks.","marker":"[26]"}],"fun_headline_variants":["Two-loop QCD Lipatov vertex pinned down","Multi-Regge QCD yields two-loop Lipatov vertex","All three channels fix two-loop Lipatov vertex","QCD Lipatov vertex at two loops, with checks","Two-loop vertex from gg, gq, and qq amplitudes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The extraction assumes that the Regge pole is exactly the sum of the single-Reggeon exchange and the planar part of multi-Reggeon exchanges, with the Regge cut entirely non-planar; this pole/cut rule is tested for 2→2 scattering through four loops but is only conjectured for 2→3 scattering at NNLL accuracy.","fun_headline_variants_meta":{"raw":{"variants":["Two-loop QCD Lipatov vertex pinned down","Multi-Regge QCD yields two-loop Lipatov vertex","All three channels fix two-loop Lipatov vertex","QCD Lipatov vertex at two loops, with checks","Two-loop vertex from gg, gq, and qq amplitudes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000651,"raw_usage":{"total_tokens":3006,"prompt_tokens":988,"completion_tokens":2018,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":604,"completion_tokens_details":{"reasoning_tokens":1935}},"tokens_in":604,"tokens_out":2018,"duration_ms":14337,"temperature":1.0,"reasoning_tokens":1935,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:18:30.598848+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the three-loop planar multi-Reggeon contribution to the octet-octet component of a 2→3 amplitude at NNLL accuracy and compare it with the prediction in eq. (8.19); any deviation would show that planar multi-Reggeon exchanges mix into the Regge cut, invalidating the pole/cut extraction of the two-loop vertex.","supporting_citations":[],"review_version":1}