{"id":"936a1c42-daa4-4108-b809-304d06e3d688","arxiv_id":"2502.07546","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"Exact UV limits of connected Schwinger functions (n≠2) for bounded interactions in d≥2 equal tree-level 1PI functions of the erf(φ/√2) theory.","lead":"Bounded measurable interaction functions in d≥2 Euclidean scalar QFT are shown to have exact UV limits after a suitable field renormalization. The connected Schwinger functions for n≠2 exist and coincide with tree-level 1PI functions of an erf interaction, with coupling set by the interaction's limits at infinity or at zero.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; Theorem 1.1(d) is supported by the proof under the stated assumption (A2).","rationale":"The reader identified assumption (A2) as the weakest point, and I agree that the proof of Theorem 1.1(d) relies essentially on it. However, since (A2) is an explicit hypothesis, the theorem is sound under that condition. I checked the dominated convergence step in (2.30)-(2.34): for η=1, the argument of V is CΛ(0)w, which diverges for any w≠0, so the half-line limits are correctly applied; the remaining dependence on CΛJ converges to CJ because J is Schwartz. The factorization bound in Proposition 2.1 is carefully proven and the estimates hold uniformly, so the remainder vanishes. The paper does not explicitly prove that functional derivatives of the cutoff generating functionals converge, but it defines the limiting n-point functions as derivatives of the limit Σ^c(J), which is a well-defined smooth functional. This interpretation is consistent with the theorem and the discussion after it. I therefore found no load-bearing concern that would change the reader's ACCEPT verdict.","tokens_in":14180,"tokens_out":38236,"duration_ms":320952,"concrete_test":"To close the only minor gap, verify that for fixed n the n-th functional derivative of the remainder R_Λ defined in (2.27), evaluated at J=0, is O(1/√log Λ). This follows by differentiating the estimates (2.25)-(2.26) and using that ⟨δ_x,CΛf⟩ is uniformly bounded; if confirmed, the cutoff connected functions converge to (1.10) as the abstract suggests.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim, Theorem 1.1(d), is conditional on (A2), which is explicitly assumed. The dominated convergence step in the proof of part (d) is valid: for each fixed x, V(CΛ(0)w) converges to the half-line limits for w≠0, the Gaussian factor is dominated, and the J-dependent shift passes to the limit because CΛJ converges uniformly to CJ on B for J∈S(R^d). The remainder R_Λ in (2.27) is bounded pointwise by O(1/√log Λ) via Proposition 2.1, and the limit functional is smooth, so the derivatives defining S^c_n in (1.10) are well-defined from the limit. The only technical point not spelled out is the convergence of functional derivatives of the cutoff objects themselves, but the paper defines the limiting Schwinger functions as derivatives of the limit, which is a consistent interpretation. I find no internal inconsistency or unstated assumption that would invalidate the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies scalar Euclidean QFTs with a bounded measurable interaction V and a UV cutoff Λ. With the field renormalization Z_Λ=C_Λ(0)^η, the modified generating functional Σ^c_Λ(J)=log(S_Λ(J)/S_{0,Λ}(J)) is shown to have an explicit UV limit for various η, under assumptions (A1) (limits of V at 0) or (A2) (limits of V at ±∞). The main result is Theorem 1.1(d): for η=1 and (A2), the connected Schwinger functions S^c_n for n≠2 exist and equal the tree-level one-particle irreducible Schwinger functions of the interaction erf(ϕ/√2) with coupling λ(V^+−V^-)/2. Section 3 extends the construction to Λ-dependent interactions V(z_Λ ϕ), with the coupling constant determined by the discontinuity of V at zero in one case. Section 4 discusses the divergence of the two-point function and possible outlooks.","tokens_in":14357,"tokens_out":22317,"duration_ms":185004,"significance":"The result is significant: it provides a large class of non-polynomial interactions for which non-Gaussian UV limits are computed exactly in d≥2, a regime where usual constructive results are scarce. The proof is self-contained and rigorous, with explicit factorization bounds (Proposition 2.1) and no fitted parameters. The universality of the n≠2 Schwinger functions, depending on V only through the coupling (V^+−V^-)/2, is surprising and clearly stated. The two-point function is explicitly excluded from the claim, so its divergence is not a flaw. The paper is honest about the limitations for Osterwalder–Schrader reconstruction, and the connection to tree-level erf theory is an insightful interpretation.","major_comments":[],"minor_comments":[{"comment":"In Eq. (2.34), the right-hand side still contains C_Λ J in the Gaussian exponent after taking the limit in V; to complete the proof of (1.8), one should explicitly pass Λ→∞ in the Gaussian factor, using the uniform convergence ⟨δ_x,C_ΛJ⟩→⟨δ_x,CJ⟩ on B for J∈S(R^d).","section":"2, proof of Theorem 1.1(d)"},{"comment":"The statement that the n≠2 Schwinger functions 'exist in the UV limit' is stronger than what is proven: the paper defines them as derivatives of the limiting functional Σ^c(J), while convergence of the cutoff derivatives is not established. Please state this convention explicitly in Theorem 1.1 or in the abstract.","section":"1 and Theorem 1.1"},{"comment":"In the estimates following Eq. (2.23), the Gaussian exponent after the shift w→w+q is written as e^{-1/2 w^T w} instead of e^{-1/2 w^T M_α^{-1} w}; this is harmless because M_α^{-1} is uniformly bounded and positive definite, but the justification should be noted.","section":"2, estimates after Eq. (2.23)"},{"comment":"The footnote in Section 3 states that in part (e) additional regularity of V may be needed; this should be made precise, as the main theorem's statement and proof do not mention such a condition.","section":"3, footnote 1"}],"recommendation":"minor_revision","confidential_remarks":"The paper is likely to be of interest to the mathematical physics community. The central result is correct and the presentation is clear overall. The minor points above are easily addressed. I do not see any need for additional experiments or references."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The result is real and novel. For bounded measurable interactions with one-sided limits at infinity, the author shows that a field renormalization Z_Λ = C_Λ(0)^η yields exact UV limits of the modified generating functional Σ^c_Λ = log(S_Λ/S_0,Λ). The interesting case η=1 gives explicit connected Schwinger functions for n≠2, which turn out to be tree-level 1PI functions of an erf interaction with coupling (V^+−V^−)/2. I checked the proof structure: the factorization estimate (Prop. 2.1) is the load-bearing piece, and the O(1/√log Λ) bound is plausible from the estimates given. The dominated convergence steps in Theorem 1.1 are handled carefully, and the assumptions are stated precisely.\n\nWhat the paper does well: it takes a class that had no systematic treatment and gives a complete, non-perturbative solution. The proof is transparent and largely self-contained, using standard propagator bounds. The generalization in Section 3 is a nice bonus: scaling the argument of V allows a second family where non-Gaussianity is controlled by the discontinuity of V at zero. The discussion of why the sine-Gordon model evades the mechanism (Wick ordering gives λ growing like e^{βC_Λ(0)/2}) is helpful and honest.\n\nSoft spots, in proportion: the two-point function diverges in the case with nontrivial n≠2 correlations, so the result does not yield a Euclidean QFT in the Osterwalder–Schrader sense. The author says this clearly and suggests a spectator-field construction, which is speculative but reasonable. Also, because the objects are defined from the modified functional, calling them Schwinger functions is only legitimate for n≠2; the paper respects that. The universality—all bounded interactions give the same erf structure—is remarkable but also means the class is less rich than one might hope. None of these are hidden flaws; they are stated limitations.\n\nThe math is solid. I did not find circular reasoning or fitted parameters. The citation pattern looks appropriate, and the author is careful to credit prior constructive QFT work.\n\nThis paper deserves a serious referee. It makes a precise, checkable claim and proves it cleanly. I would welcome it in the literature after a routine check of the estimates. My own verdict is acceptance.","headline":"A rigorous, exactly solvable class of bounded-interaction scalar QFT limits, with the two-point function explicitly left unresolved.","tokens_in":14876,"tokens_out":1673,"would_cite":true,"duration_ms":18056,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81T08","81T16","81T10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For bounded interactions with limits at infinity, a field renormalization of the form $Z_\\Lambda=C_\\Lambda(0)^\\eta$ makes all connected Schwinger functions with $n\\neq2$ exist non-perturbatively in the ultraviolet limit and coincide with…","keywords":["Euclidean quantum field theory","Schwinger functions","ultraviolet limit","bounded interactions","field renormalization","error function interaction","non-Gaussianity","constructive QFT"],"falsifier":"Compare two bounded measurable interactions with the same limits at infinity but different local shapes, for example $V_1(w)=\\tanh(w)$ and $V_2(w)=\\tanh(w)+\\epsilon e^{-w^2}\\sin(w)$, at $\\eta=1$. Formula (1.10) predicts identical connected $n\\neq2$ Schwinger functions for both, so evaluating the one-point expression (2.30) and taking $\\Lambda\\to\\infty$ should show the difference tending to zero; a nonzero limiting difference would falsify the universality claim.","tokens_in":13965,"feed_emoji":"⚛️","tokens_out":12445,"duration_ms":105369,"temperature":0.7,"pith_summary":"This paper claims that a large class of non-polynomial scalar quantum field theories in spacetime dimension $d \\ge 2$ — specifically any interaction given by a bounded measurable function $V$ with finite limits at $\\pm\\infty$ — has an exactly computable ultraviolet limit after a suitable field renormalization. The limit is encoded in the modified generating functional $\\Sigma^c_\\Lambda(J)$, the logarithm of the interacting generating functional divided by the free one. For the most interesting renormalization, $\\eta=1$, all connected Schwinger functions with $n\\neq2$ exist non-perturbatively and coincide with the tree-level one-particle irreducible Schwinger functions of the $\\mathrm{erf}(\\phi/\\sqrt2)$ interaction, with the coupling constant rescaled by $(V^+-V^-)/2$. If true, this means the short-distance behaviour of every bounded interaction is universal up to a single coupling constant, and it gives a rare non-Gaussian, exactly solvable example in $d\\ge2$. The paper also shows how to make the theory probe the discontinuity of $V$ at zero instead of its jump at infinity.","feed_headline":"Bounded interactions get exact Schwinger functions in the UV limit","feed_subtitle":"All n≠2 connected functions become exactly computable and match a single tree-level erf theory.","key_machinery":"The load-bearing object is the modified generating functional $\\Sigma^c_\\Lambda(J)$, which removes the free-field Gaussian factor and, by standard combinatorics, generates exactly the connected Schwinger functions with $n\\neq2$. The proof is carried by a factorization mechanism: for bounded $V$, the normalized expectation of $\\prod_{i=1}^\\ell V(\\phi(x_i))$ differs from the product of the individual one-point expectations by at most $c\\,\\ell^6\\|V\\|_\\infty^\\ell|B|^\\ell/\\sqrt{\\log\\Lambda}$. This happens because the ratios $C_\\Lambda(x_i-x_j)/C_\\Lambda(0)$ vanish uniformly outside a tiny diagonal neighbourhood, while the diagonal neighbourhood itself has small volume. The remaining computation is the single one-point Gaussian integral in (2.30), whose limit is controlled by the two combinations $Z_\\Lambda C_\\Lambda(0)$ and $Z_\\Lambda/C_\\Lambda(0)$; setting $Z_\\Lambda=C_\\Lambda(0)^\\eta$ selects which asymptotic information about $V$ survives.","core_discovery":"The central discovery is that the modified generating functional $\\Sigma^c_\\Lambda(J)=\\log(S_\\Lambda(J)/S_{0,\\Lambda}(J))$ has a finite ultraviolet limit for the field renormalization $Z_\\Lambda=C_\\Lambda(0)^\\eta$ under the stated assumptions. In the most interesting case $\\eta=1$ with $V^\\pm=\\lim_{w\\to\\pm\\infty}V(w)$, the limit is $$\\Sigma^c(J)=-\\$\\lambda$|B|\\frac{V^++V^-}{2}-\\$\\lambda$\\frac{V^+-V^-}{2}\\frac{1}{(2\\pi)^{1/2}}\\int_B dx\\int dw\\,\\operatorname{sgn}(w)$e^{{-\\frac12(w-\\langle\\delta_x,CJ\\rangle)^2}}$.$$ Differentiating at $J=0$ gives, for $n\\neq2$, $$S^c_n(x_1,\\dots,x_n)=-\\$\\lambda$\\frac{V^+-V^-}{2}\\,[(\\partial_w)^n\\operatorname{erf}(w/\\sqrt2)]_{w=0}\\int_B dx\\, C(x-x_1)\\cdots C(x-x_n).$$ The paper identifies these functions with the tree-level one-particle irreducible Schwinger functions of the $\\operatorname{erf}(\\phi/\\sqrt2)$ interaction with coupling constant $\\lambda(V^+-V^-)/2$. The same mechanism, with a modified interaction function $V_\\Lambda(w)=V(C_\\Lambda(0)^\\kappa w)$, produces a variant in which non-Gaussianity is controlled by the discontinuity of $V$ at zero rather than by its jump at infinity.","pith_inferences":["A testable extension the author does not pursue is to replace the constant limits $V^\\pm$ by Cesàro averages of $V$ at infinity; if the averaging limit exists, the dominated-convergence step may generalize, giving new theorems beyond the stated assumptions.","The universality result suggests an equivalence relation on bounded interactions: two functions with the same one-sided limits at infinity define the same $n\\neq2$ UV theory after renormalization, and it would be interesting to see whether this equivalence survives under local averages or mollifications of $V$.","If the factorization rate $O(1/\\sqrt{\\log\\Lambda})$ is optimal, then allowing $V$ to become rougher as $\\Lambda$ grows, while keeping it uniformly bounded, might defeat the trivial-$J$-dependence argument for $\\eta<1$ and open a route to nontrivial correlations with a finite two-point function; the author names this as an open direction rather than a proven result."],"forward_implications":["For any bounded measurable $V$ satisfying (A2), the UV limit of all connected Schwinger functions with $n\\neq2$ is finite, non-perturbative, and given by the closed form (1.10).","The limiting $n\\neq2$ correlations are universal: two different interactions with the same limits $V^\\pm$ produce identical Schwinger functions, up to the fixed coupling renormalization $\\lambda(V^+-V^-)/2$.","Non-Gaussianity is possible in every dimension $d\\ge2$; for interactions with $V^+\\neq V^-$ the four-point and higher functions are nonzero at tree level, and in the rescaled variant non-Gaussianity is governed by a discontinuity of $V$ at zero.","The two-point function is not obtained by this construction; for $\\eta=1$ it diverges, so the model is not directly a Euclidean QFT in the Osterwalder-Schrader sense, and the paper discusses a spectator-field or classical-limit route to cure this.","For $\\eta<1$, the modified generating functional has no $J$-dependence, so the only nontrivial connected functions (if any) live in the two-point sector, which is not controlled."],"supporting_citations":[{"why":"Supplies the covariance estimates used in Lemma 2.2, especially the off-diagonal decay $C_\\Lambda(x)\\le c|x|^{-(d-3/2)}$ that makes the factorization error small.","marker":"[GJ]"},{"why":"Gives the modified generating functional identity (1.4) relating $\\Sigma_\\Lambda^c$ to connected functions.","marker":"[Sa]"},{"why":"Provides the standard fact that $\\Sigma_\\Lambda^c$ generates connected Schwinger functions for $n\\neq2$.","marker":"[PT11]"},{"why":"Cited for the result that $Z_\\Lambda>1$ forces interacting scalar fields to violate canonical commutation relations, supporting the paper's reading of the divergent two-point function.","marker":"[We]"},{"why":"Cited alongside [We] for the CCR obstruction that shapes the discussion of the two-point sector.","marker":"[Ba87]"}],"fun_headline_variants":["Bounded interactions get exact UV Schwinger functions","Exact UV Schwinger functions for n≠2 in bounded QFT","Tree-level erf emerges from bounded potential UV limit","Jump in potential gives exact non-Gaussian QFT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The non-Gaussian conclusion relies on Assumption (A2) that the limits $V^\\pm$ exist; the proof replaces $V$ by constant values on the two half-lines, so a bounded function that continues to oscillate at infinity would not be covered and the explicit formula (1.10) could fail.","fun_headline_variants_meta":{"raw":{"variants":["Bounded interactions get exact UV Schwinger functions","Exact UV Schwinger functions for n≠2 in bounded QFT","Tree-level erf emerges from bounded potential UV limit","Jump in potential gives exact non-Gaussian QFT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001263,"raw_usage":{"total_tokens":5228,"prompt_tokens":1056,"completion_tokens":4172,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":672,"completion_tokens_details":{"reasoning_tokens":4107}},"tokens_in":672,"tokens_out":4172,"duration_ms":29193,"temperature":1.0,"reasoning_tokens":4107,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T12:21:56.054523+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare two bounded measurable interactions with the same limits at infinity but different local shapes, for example $V_1(w)=\\tanh(w)$ and $V_2(w)=\\tanh(w)+\\epsilon e^{-w^2}\\sin(w)$, at $\\eta=1$. Formula (1.10) predicts identical connected $n\\neq2$ Schwinger functions for both, so evaluating the one-point expression (2.30) and taking $\\Lambda\\to\\infty$ should show the difference tending to zero; a nonzero limiting difference would falsify the universality claim.","supporting_citations":[],"review_version":1}