{"id":"e2150d40-bc8c-48ee-8786-3ee423d9dfd8","arxiv_id":"2505.02910","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A general theorem fixes the sign of one-loop renormalization of even-dimension EFT couplings induced by equal-even-dimension operator pairs, with applications to SMEFT dimension-eight running and new non-renormalization results.","lead":"This paper proves that certain effective field theory couplings can only run in one direction as the energy scale changes, when the running comes from pairs of operators with the same even dimension. The result gives a new infrared-only tool for deciding when positivity constraints from unitarity survive quantum corrections in theories such as the Standard Model effective field theory.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The t-channel vanishing proof in Eq. (6) is the load-bearing soft spot; a nonzero t-channel cut for higher-spin internal states would break the sign-definite claim.","rationale":"The paper's strongest claim is a theorem, so the burden is on the proof of the sign. The s/u channel positivity is standard and supported by explicit examples; the t-channel vanishing is the point where the proof is least secure. I found no internal inconsistency or counterexample, and the explicit chiPT and SMEFT checks are genuine supporting evidence. The reader's weakest assumption identified the same point, including the F^3 scope issue. My concern does not change the verdict: the preprint should remain CONDITIONAL, with the t-channel calculation above as the condition for acceptance. If the check finds a nonzero t-channel cut, the central theorem and the F^3 application would need to be revised.","tokens_in":30799,"tokens_out":19540,"duration_ms":242746,"concrete_test":"Evaluate the t-channel phase-space integral in Eq. (6) for the simplest higher-spin case: two insertions of a dimension-6 four-fermion operator, e.g. (psi-bar gamma^mu psi)(psi-bar gamma_mu psi), feeding a forward-limit dimension-8 four-fermion coefficient. Do the integral with the full Lorentz tensor structure, without imposing the replacement of loop momenta by p1-p4, and check whether the coefficient of t^0 vanishes in the forward limit. If it does not, Eq. (7) fails for arbitrary spins. If it does, repeat the same check on the F^3 cut behind Eq. (10) to test the cubic-vertex extension.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (7) hinges on the claim that Disc_t A^(1) vanishes in the forward limit for any higher-dimensional operator insertion. The proof around Eq. (6) is a sketch: it tracks only the internal-momentum dependence and then says Lorentz invariance replaces the integrated internal vectors by p1-p4. That replacement is the load-bearing step. A t-channel bubble of a four-point function is a Lorentz tensor that can contain both q^mu q^nu and g^munu q^2 with q=p1-p4; the letter does not show that the trace term and contractions with external momenta vanish when the external spinor factors of dimension-(2n+2) operators are included. If the cut total momentum is not q but one of the nonvanishing sums p1+p2 or p1+p3, the statement is even less secure. The s/u channels are only semidefinite, so any nonvanishing t-channel cut of either sign would invalidate the sign-definite theorem. The SMEFT F^3 application, Eq. (10), lies outside the stated 'massless particles without cubic interactions' setup and is justified by the same bubble-counting argument, so it inherits this gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proves, or claims to prove, an 'EFT a-theorem': in the forward limit, the one-loop running of a dimension-4n coupling c_{4n} induced by double insertions of two operators of equal even mass dimension is sign-definite, dc_{4n}/d ln mu <= 0 (Eq. (7)). The proof starts from the on-shell unitarity-cut formula (4), combines the s and u cuts using positivity of the forward elastic integrand and the relation u = -s, and argues that the t cut vanishes for any insertion of a higher-dimensional operator via the Lorentz-covariance estimate in Eq. (6). Applications are given in chiral perturbation theory (Eq. (8)), in the SMEFT dimension-8 H4D4 sector including new F^3 contributions (Eq. (10)), and at dimension six for mixing of the Weinberg operator (Eq. (12)); the supplemental material adds an R^2 phi^2 gravity example, explicit RG expressions, a toy-model and an extended-Higgs-sector demonstration of RG-induced violation of tree-level positivity, and a dispersion-relation check. A corollary gives non-renormalization of forward-probed operators when only the t channel exists.","tokens_in":30982,"tokens_out":16268,"duration_ms":206749,"significance":"The claimed theorem, if fully established, would be a genuinely useful structural result: it extracts a sign-definite sector of the one-loop RG purely from unitarity, analyticity, and Lorentz invariance, without dispersion relations and without assumptions about the UV completion. It also provides a practical criterion (11) for when loop corrections can preserve or violate tree-level positivity bounds, and it supplies new explicit SMEFT RG results, the F^3 contributions to H4D4 mixing, that are of independent value. The consistency checks in the manuscript and supplement (chiPT, Weinberg-operator mixing, the R^2 phi^2 example, and the dispersion-relation consistency of the positivity-violating toy models) are credit to the paper and make the sign claim very plausible. The main limitation is that two load-bearing steps of the general proof are sketched rather than proved, so the paper is currently stronger as a collection of verified examples plus a promising general argument than as a theorem with the stated level of generality.","major_comments":[{"comment":"The vanishing of Disc_t A^(1) in the forward limit is the decisive step that turns the s/u-channel positivity into the sign-definite claim (7), and the argument in Eq. (6) is not a complete proof. A bubble integral over internal momenta ell_1, ell_2 produces a Lorentz tensor containing both q^mu q^nu and g^{mu nu} q^2 terms with q = p1 - p4; only the former is captured by the replacement ell_i -> q. The text explicitly says that it 'does not assume any details on index contractions,' but those contractions are exactly what must be checked: contributions in which the g^{mu nu} terms contract with external spinor or Lorentz structures of a dimension-(2n+2) operator could survive even at q^2 = 0, and for spinning internal lines spinor products such as <ell_1 ell_2> and [ell_1 ell_2] are not represented by (|ell>[ell|)^{2|h|+n}. Please supply a proof that every non-scalar numerator factor is proportional to q^mu (or q^2) after the phase-space integration, or state and prove the equivalent angular-momentum selection rule. As written, a nonzero t-channel cut of either sign would invalidate Eq. (7).","section":"Proof of the EFT a-theorem, Eq. (6)"},{"comment":"The non-negativity of the integrated two-insertion coefficient d_{i,m} is imposed rather than derived: the text says 'we impose ... such that each integrand is always non-negative ... for a non-negative coefficient d_{i,m}.' Unitarity gives positivity of the forward imaginary part, but the step from that to a non-negative coefficient of the single power s_i^m after summing over intermediate helicities and after the s/u combination is nontrivial. Since this is the entire content of the sign in Eq. (7), it should be formulated as a lemma with a proof, for example from the optical theorem together with the power-counting of a single insertion, rather than as an input condition.","section":"Proof of the EFT a-theorem, around Eq. (5)"},{"comment":"The stated scope of the theorem is 'massless particles without cubic interactions' (Proof section), but the SMEFT application includes F^3 insertions and the supplement includes the gravitational R^2 phi^2 example, both of which involve theories with cubic interactions. The text argues that for F^3 only bubble diagrams contribute and therefore the single-scale t-channel counting of Eq. (6) applies; this is a new claim, because the derivation of Eq. (4) and of the t-channel vanishing used the no-cubic-interaction and no-IR-divergence assumptions. The explicit RG computation in the supplement is a welcome check, and it does verify the sign for the F^3 sector, but the paper should either extend the proof to cover these sectors or present them as explicit examples rather than as corollaries of the general theorem.","section":"Phenomenological Implications, Eq. (10); Supplemental Sec. I.C"}],"minor_comments":[{"comment":"The notation in the proof is confusing: the text speaks of operators of 'mass dimension (2m+4)' and a coefficient d_{i,m} s_i^m, while the theorem is stated in terms of dimension-(2n+2) insertions and c_{4n}; please define m unambiguously and align it with n.","section":"Proof of the EFT a-theorem, after Eq. (5)"},{"comment":"The sentence that 'This singles out phi^4-theory as the only case where Disc_t A^(1) != 0 in the forward limit' is too quick: even if the counting is correct, one should state which four-point operator is renormalized in that case, since phi^4 alone has no higher-dimensional operator in the game.","section":"Proof of the EFT a-theorem, after Eq. (6)"},{"comment":"The display after Eq. (10) contains a corrupted inline figure ('xit>') and should be cleaned up.","section":"Phenomenological Implications, Eq. (10)"},{"comment":"The estimate that electromagnetic corrections to the chiPT running are O(alpha_em/4 pi) x O(c_i) <= 10^-5 would benefit from a one-line derivation or a reference, since it carries the phenomenological claim about the direction of the flow.","section":"Phenomenological Implications, after Eq. (8)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal, and the authors are transparent about the relation to Refs. [83] and [33]. The main risk is that the proof of the t-channel vanishing is not yet at the standard of a published theorem; if the authors can supply the missing steps, or explicitly demote the unproved general statements to verified examples, I would support publication. No concerns about citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper gives a general sign rule for one-loop RG in EFTs: when a dimension-4n coupling is renormalized by two insertions of even-dimension operators, the forward-limit running has a fixed sign, with the IR coupling larger than the UV one. If true, that is genuinely useful. It generalizes the explicit examples in scalar EFTs and the SMEFT dimension-8 sector to arbitrary spins and mass dimensions, and it does so using only unitarity, analyticity, and Lorentz invariance, with no UV assumptions. It also gets a few corollaries: new non-renormalization results for operators that survive the forward limit when only the t-channel contributes, and sign-definite dimension-six mixing of the Weinberg operator. The explicit checks in chi-PT, the SMEFT, and the gravitational EFT all line up with the claimed signs, and the supplemental material shows the loop-level positivity violation example is consistent with dispersion relations. The citation practice is honest: the authors credit earlier work and are clear about what is new.\n\nThe soft spot, as the stress-test note says, is the t-channel vanishing. The argument in Eq. (6) tracks only the internal-momentum spinor-helicity factors and then says Lorentz invariance lets you replace the integrated internal momenta by p1-p4. But in the t-channel cut the total momentum through the bubble is p1+p4, not p1-p4, and the possibility of g^{mu nu} pieces with P^2 factors is not addressed. Since the s/u channels are only semidefinite, a nonvanishing t-channel cut of either sign breaks the sign-definite claim. This is the load-bearing step, and it is currently asserted rather than proved. The same sketch is used to extend the theorem to F^3-type insertions in the SMEFT, which are outside the paper's stated scope of massless particles without cubic interactions. The non-negativity of the integrated coefficient d_{i,m} is also asserted rather than proved, though that one is likely fine.\n\nThese are addressable gaps, not fatal flaws. The central idea is probably right, and the explicit examples give real support. The paper deserves a serious referee. My recommendation: send it to review, and direct the referee to focus on Eq. (6). A complete proof of the t-channel statement, or a narrower theorem that avoids the need for it, would put the paper on solid ground.","headline":"A likely true general sign rule for one-loop EFT running, but the t-channel vanishing proof is a sketch and must be the referee's focus.","tokens_in":31533,"tokens_out":11234,"would_cite":true,"duration_ms":125538,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["11.10.Hi","11.55.-m","12.15.-y"],"model":"deepseek-v4-flash","headline":"One-loop EFT running has a fixed sign in the forward limit.","keywords":["renormalization group","effective field theory","positivity bounds","forward limit","unitarity cuts","on-shell methods","SMEFT","chiral perturbation theory"],"falsifier":"Compute $\\mathrm{Disc}_t A^{(1)}$ directly in the forward limit for a one-loop four-point amplitude built from two equal-dimension higher-dimensional operators in a theory with a massless scalar plus a cubic or higher-spin interaction; the counting in (6) predicts exactly zero, so any nonzero result, or any positive contribution to $dc_{4n}/d\\ln\\mu$ from the t channel, would disprove inequality (7).","tokens_in":1734,"feed_emoji":"📈","tokens_out":1643,"duration_ms":110981,"temperature":0.7,"pith_summary":"The paper establishes an EFT a-theorem: in the forward limit, the one-loop running of a dimension-4n coupling induced by two insertions of operators of the same even mass dimension has a fixed sign, with the coupling larger in the infrared than at the matching scale. The proof is meant to follow from unitarity, analyticity, and Lorentz invariance alone, with no assumption about the ultraviolet completion, so the theorem applies across chiral perturbation theory, the Standard Model EFT, and gravitational EFTs. The authors use it to determine which tree-level positivity bounds survive renormalization effects and which can be violated, and they derive new non-renormalization theorems as corollaries. A sympathetic reader should care because it extracts a general monotonicity statement out of a notoriously sign-dependent calculation.","feed_headline":"One-loop EFT running has a fixed sign in the forward limit","feed_subtitle":"Unitarity and analyticity force dimension-8 coefficients higher at low energy than at the matching scale.","key_machinery":"The on-shell one-loop RG formula (4), obtained from $dA_{\\mathrm{full}}/d\\ln\\mu=0$ and the optical theorem, is the engine: it writes $dA^{(0)}/d\\ln\\mu$ as $-1/\\pi$ times the sum of phase-space integrals of tree amplitudes over s, u, and t discontinuities. Its sign structure is analyzed in the forward limit, where the s and u integrands are non-negative by unitarity and $s=-u$ makes the combination sign-definite for even equal dimensions. The t channel is eliminated by the counting in (6): in a bubble diagram, Lorentz invariance forces the loop-momentum dependence into powers of $(p_1-p_4)$, so only massless scalars without extra derivatives survive, and those terms vanish in the forward limit for any higher-dimensional operator. Together these pieces yield the inequality (7).","core_discovery":"The central claim is inequality (7): restricted to the contribution from a pair of dimension-(2n+2) operators, the one-loop $\\beta$ function of a dimension-4n forward-limit coupling satisfies $dc_{4n}/d\\ln\\mu \\le 0$, so integrating from the cutoff to the infrared gives $c_{4n,\\mathrm{IR}} - c_{4n,\\mathrm{UV}} \\ge 0$. The argument begins from the on-shell renormalization-group formula, which equates the scale derivative of the tree amplitude with the sum of the s-, u-, and t-channel unitarity cuts of the one-loop amplitude. In the forward limit the s and u channels contribute positive-definite integrands, and because $u=-s$ the sum is sign-definite exactly when both inserted operators have equal even mass dimension; the t channel is shown to vanish for any higher-dimensional operator by helicity and derivative counting on a single-scale bubble. For $n\\ge 2$ the theorem therefore fixes the sign of the square of dimension-$(2n+2)$ insertions into dimension-$4n$ couplings, regardless of spin and ultraviolet completion.","pith_inferences":["Editorial inference: the sign-definite sector could be used as a model-selection prior in global SMEFT fits, with data showing dimension-8 coefficients growing toward the IR in the fixed combinations favoring strongly coupled ultraviolet completions and the opposite pattern favoring weakly coupled completions.","Editorial inference: if the t-channel vanishing argument survives the addition of cubic interactions, it would expose a monotone structure at the S-matrix level that may connect to quantum-information formulations of RG irreversibility; the authors mention this as a future direction but do not prove it.","Editorial inference: the theorem's mechanism suggests the sign-definite sector is special to the one-loop forward limit, so probing the same coefficients at nonzero momentum transfer or at two loops could reveal sign cancellations the present proof does not control."],"forward_implications":["In the SMEFT, the dimension-8 H4D4 couplings receive a negative beta-function contribution from double insertions of dimension-6 operators; combinations that can be probed in the forward limit, such as $c^{(1)}_{H^4D^4}+c^{(2)}_{H^4D^4}$, run to larger values in the IR, as shown explicitly in the supplemental material.","In chiral perturbation theory, the $O(p^4)$ couplings $c_1+c_2$ and $c_2$ satisfy the predicted negative running, matching the explicit one-loop beta functions.","Tree-level positivity bounds $c_{4n}\\ge 0$ are preserved by this sector of the RG when the dimension-$(2n+2)$ pair dominates, and apparent infrared violation of the bounds points to weakly coupled ultraviolet completions.","Operators that survive in the forward limit are not renormalized by operators that vanish there when only the t-channel cut exists, yielding non-renormalization of $H^2F^2D^2$ from $H^4D^2$ and $H^2F^2$ insertions and explaining zero entries in the dimension-6 SMEFT anomalous-dimension matrix.","At dimension six, sign-definite RG still follows when only one channel is present, as in the mixing of two Weinberg operators into $H^4D^2$, $LLH^2D$, and $LLLL$ operators."],"supporting_citations":[{"why":"Supplies the on-shell RG formula (4) that the proof starts from.","marker":"[49]"},{"why":"Provides the first dimension-eight SMEFT examples of fixed-sign running and the benchmark case of apparent positivity violation.","marker":"[83]"},{"why":"Supplies known dimension-eight bosonic RG entries that the theorem explains and that the H4D4 results extend.","marker":"[33]"},{"why":"Identifies which dimension-eight operators can be probed in the forward limit, fixing the SMEFT combinations the theorem constrains.","marker":"[20]"},{"why":"Establishes the tree-level positivity bounds whose RG fate the theorem classifies.","marker":"[61]"},{"why":"Provides the partial-wave and on-shell RG perspective used for the t-channel selection rule and for the gravitational example.","marker":"[56]"},{"why":"Documents loop corrections to positivity bounds, the phenomenon the paper's criterion addresses.","marker":"[76]"}],"fun_headline_variants":["EFT coefficients only rise towards the infrared","One-loop sign fixed by unitarity and analyticity","Forward-limit running sign is not arbitrary","Unitarity dictates EFT renormalization direction","Infrared principles control one-loop EFT running"],"cache_read_input_tokens":33664,"weakest_assumption_plain":"The theorem assumes that the sideways (t-channel) cut of the one-loop amplitude contributes nothing in the forward limit for any higher-dimensional operator; if any such cut survives, the fixed-sign conclusion can fail.","fun_headline_variants_meta":{"raw":{"variants":["EFT coefficients only rise towards the infrared","One-loop sign fixed by unitarity and analyticity","Forward-limit running sign is not arbitrary","Unitarity dictates EFT renormalization direction","Infrared principles control one-loop EFT running"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000552,"raw_usage":{"total_tokens":2630,"prompt_tokens":940,"completion_tokens":1690,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":556,"completion_tokens_details":{"reasoning_tokens":1620}},"tokens_in":556,"tokens_out":1690,"duration_ms":21272,"temperature":1.0,"reasoning_tokens":1620,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:40:18.479481+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $\\mathrm{Disc}_t A^{(1)}$ directly in the forward limit for a one-loop four-point amplitude built from two equal-dimension higher-dimensional operators in a theory with a massless scalar plus a cubic or higher-spin interaction; the counting in (6) predicts exactly zero, so any nonzero result, or any positive contribution to $dc_{4n}/d\\ln\\mu$ from the t channel, would disprove inequality (7).","supporting_citations":[{"cited_title":"Distler, B","cited_arxiv_id":null,"evidence_quote":"Documents loop corrections to positivity bounds, the phenomenon the paper's criterion addresses."}],"review_version":1}