{"id":"3c2a8b47-3bdf-4167-9928-6d193cc8e3b6","arxiv_id":"2601.06832","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Adding N corrector terms gives an O(ε) operator-norm resolvent approximation for periodic Lévy-type operators whenever α lies in (2−1/N, 2−1/(N+1)].","lead":"This mathematics paper constructs explicit correction terms that turn a slow homogenization error for Lévy-type operators into an O(ε) error. It matters because jump-process models in biology, finance, and physics need accurate approximations when fine-scale periodic structure is averaged out.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; Theorem 5.2 appears sound, but the paper contains minor statement typos and an unsupported sharpness remark.","rationale":"The central claim (Theorem 5.2) is a genuine extension of the prior result and the proof is structurally sound. The reader's weakest_assumption (uniform spectral gap) is indeed load-bearing, but it is guaranteed by the explicit assumption μ_->0, so it is not a weakness of the theorem. The reader's CONDITIONAL verdict is based on two real but minor issues: a typo in the hypotheses of Theorems 4.4, 4.6, 4.8 (1−1/N instead of 2−1/N) and an unsupported sharpness claim in the concluding remarks. Both should be corrected, but they do not break the main upper-bound result. The strongest residual risk is the dependence on the unpublished preprint [25] for the base fiber estimate; this is a verifiability concern rather than an identified error. Therefore, the reader's conditional verdict remains appropriate, and no verdict change is warranted.","tokens_in":41328,"tokens_out":43934,"duration_ms":327605,"concrete_test":"Independently re-derive Theorem 4.5 of [25] (or at least the key Lemma 2.1) from first principles of the present paper; if the ε^{2−2α} estimate holds, the main theorem's reliance on [25] is justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main estimate Theorem 5.2 is derived through a lengthy but internally consistent spectral argument. The proof's key building blocks—the uniform spectral gap (Proposition 3.1), the threshold approximation (Proposition 3.11), and the resolvent expansion (Theorems 4.3–4.8)—are coherent under the stated assumption μ_->0. I find no flaw that would invalidate the O(ε) approximation for α∈(1,2). The only concerns are non-central: Theorems 4.4, 4.6, and 4.8 state the hypothesis as '1−1/N<α<2' instead of the correct '2−1/N<α<2' (the proof and the abstract use the latter); and the concluding remark claiming order-sharpness of (5.3) lacks a proof that the first corrector K_1 is nonzero. Neither affects the upper-bound result. The main residual risk is that the proof imports Theorem 4.5 (the base ε^{2−2α} fiber estimate) from the unpublished preprint [25]; if that result were flawed, the current theorem's foundation would break.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies homogenization in L2(R^d) of a symmetric purely nonlocal Lévy-type operator A_ε with periodic coefficient μ(x/ε,y/ε) and kernel |x−y|^{-d−α}, 1<α<2. Following the authors' earlier work, the resolvent (A_ε+I)^{-1} converges in operator norm to (A0+I)^{-1} at rate ε^{2−α}. The present paper constructs a higher-order corrector expansion: for N∈N with 2−1/N<α<2, it claims the estimate (5.4), namely that (A_ε+I)^{-1}−(A0+I)^{-1}−Σ_{m=1}^N ε^{m(2−α)}K_m has operator norm O(ε) in the range 2−1/N<α≤2−1/(N+1), and O(ε^{(N+1)(2−α)}) in the complementary subrange. The correctors are explicit: K_m=(div g0∇)^m(A0+I)^{-m−1}, with g0 defined through auxiliary periodic problems. The proof combines the scaling relation (5.1)–(5.2), a Gelfand-transform decomposition, a threshold analysis of the fiber operators A(ξ), and a spectral resolvent expansion with explicit remainders.","tokens_in":41622,"tokens_out":6716,"duration_ms":66793,"significance":"If the main theorem is correct, this is a substantial contribution: it gives, for every α∈(1,2), an operator-norm approximation of the resolvent of a nonlocal periodic operator with error of order ε, by a finite sum of explicit correctors. This mirrors the Birman–Suslina corrector theory for elliptic operators and extends it to fractional-order jump processes. The proof is long but structured, with explicit constants and parameter-free corrector formulas; there are no fitted coefficients. The main residual risk is that several load-bearing estimates (notably Theorem 4.5, the base ε^{2−2α} fiber estimate) are imported from the unpublished preprint [25], so the current paper is not fully self-contained. I see no evidence of circularity: the corrector expansion is derived from a fixed spectral asymptotic expansion, not fitted to the final estimate.","major_comments":[{"comment":"The hypotheses of Theorems 4.4, 4.6, and 4.8 are stated as '1−1/N < α < 2', but all subsequent case distinctions and the proof rely on the condition '2−1/N < α < 2'. For example, equation (4.12) distinguishes cases according to 2−1/N and 2−1/(N+1), and the derivation of the remainder estimate (4.10)–(4.11) requires α>2−1/N. The stated weaker condition is inconsistent with the displayed estimate. This appears to be a systematic typo, but since it occurs in three central theorem statements, it must be corrected before publication.","section":"§4, Theorems 4.4, 4.6, 4.8"},{"comment":"The remark asserts that 'It follows from Theorem 5.2 that the precision O(ε^{2−α}) in estimate (5.3) is order-sharp.' This does not follow from the upper bound in Theorem 5.2. Sharpness would require a matching lower bound or, at minimum, a proof that the first corrector K_1 is nonzero. The manuscript provides no such argument. Either the remark should be removed or substantiated with a concrete example or a nonvanishing condition for K_1.","section":"§5.2, Concluding Remark 1"},{"comment":"The argument is heavily reliant on the unpublished preprint [25]. In particular, Theorem 4.1 (the base ε^{2−2α} fiber approximation), Theorem 4.5 (the base resolvent estimate), Theorem 4.7, and several lemmas in Sections 1–3 are quoted from [25] without proofs. Since Theorem 4.5 is a load-bearing input for the new corrector expansion, the paper would be easier to evaluate if the authors included a proof of this theorem in an appendix or cited a published version of [25]. This is a transparency concern rather than an identified error.","section":"§3–§4, dependence on [25]"}],"minor_comments":[{"comment":"The constants C1, C2, and C3 are reused with different meanings in different sections (e.g., Lemma 3.2, Lemma 3.3, Theorems 4.5–4.8). This makes cross-referencing unnecessarily difficult; a consistent numbering system for constants would help.","section":"Throughout"},{"comment":"In (1.1), the phrase 'bounded positive definite function' is imprecise; the condition given is boundedness and a two-sided positivity bound. The wording could be aligned with the displayed inequalities.","section":"§1.1"},{"comment":"The constant in the statement is denoted C2(α, μ), but the proof and Theorem 5.2 use a constant depending also on N. This is a minor notational inconsistency.","section":"§4.2, Theorem 4.8"},{"comment":"In the proof of estimate (2.28), the parameter p is eventually set to 2, but the text says 'Letting p=2 in (2.30)'. This is fine, but the role of p in the auxiliary functions could be clarified, especially because the growth constants c(p,α) blow up as p→1.","section":"§2.3, Lemma 2.5"}],"recommendation":"minor_revision","confidential_remarks":"The paper is likely correct in its main theorem, but the referee should ask the authors to fix the repeated hypothesis typo in Theorems 4.4, 4.6, and 4.8, and to either prove or withdraw the sharpness remark. The reliance on [25] is substantial; if that preprint is not yet accepted, the editor may want to verify its status, since an error there would propagate to this paper. The paper is otherwise a well-structured and significant contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real result. It upgrades the resolvent approximation for periodic Lévy-type operators from O(ε^{2−α}) to O(ε) by including explicit correctors, and the core proof holds up under scrutiny. The construction of g0 and the correctors K_m is a genuine extension of the authors' earlier leading-order work [25], and the distinction from convolution-type kernels with integrable moments is legitimate: the singular stable-like kernel requires the contour/spectral-projector machinery, not just moment expansions.\n\nWhat's new: the N-term expansion (5.4) with operator-norm remainder O(ε) for α∈(1,2) by choosing N with 2−1/N<α≤2−1/(N+1). The correctors are explicit, the error terms are tracked, and Proposition 3.11 plus Section 4 gives a coherent derivation. I checked the threshold condition in the proof; the right threshold is 2−1/N, and that is what the abstract and Theorem 5.2 use.\n\nSoft spots:\n1. Theorems 4.4, 4.6, and 4.8 state the hypothesis as 1−1/N<α<2 instead of 2−1/N<α<2. This is a typo—the constants and later estimates clearly require the stronger bound—but it appears in three theorem statements and should be fixed.\n2. Concluding remark 1 claims the O(ε^{2−α}) estimate (5.3) is order-sharp. That does not follow from Theorem 5.2 unless the first corrector K1 is nonzero. No nondegeneracy proof is given. If K1=0 or if g0=0, the approximation could be better. The claim needs a proof or should be weakened.\n3. The paper imports a substantial amount from [25], an unpublished preprint: several lemmas and Theorem 4.5 are taken from it. I did not find a flaw there, but the dependence is real and should be highlighted so referees know what is being assumed.\n4. Uniform ellipticity μ_->0 is genuinely load-bearing: it creates the spectral gap d0>0 that underpins the contour and spectral projector. That is not a flaw—it is the stated operator class—but the result does not touch degenerate coefficients.\n\nThis paper is for people working on quantitative homogenization, nonlocal operators, or spectral methods for periodic operators. The main theorem is likely correct, the typos are cosmetic, and the sharpness claim is the only substantive gap. I would send it to peer review and ask for the sharpness remark to be proved or removed.","headline":"Solid corrector expansion for Lévy-type homogenization; main theorem is right, but fix the α-range typos and don't claim sharpness without proving K1≠0.","tokens_in":42127,"tokens_out":2059,"would_cite":true,"duration_ms":21799,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B27","45K05","47A55","60J75"],"pacs":[],"model":"deepseek-v4-flash","headline":"Adding corrector terms makes resolvents of periodic Lévy operators converge to the effective resolvent with operator-norm error of order ε for every α in (1,2).","keywords":["Lévy-type operators","periodic homogenization","operator-norm resolvent estimates","correctors","spectral method","Floquet–Bloch decomposition","fractional Laplacian","stable-like jump processes"],"falsifier":"Take d = 1, α = 3/2, and a periodic coefficient with nontrivial oscillation, for instance μ(x,y) = 2 + sin(2πx) + sin(2πy). For ε small, compute the operator norm of (A_ε + I)^{-1} − (A^0 + I)^{-1} − ε^{1/2} K_1. The theorem predicts this norm decays as O(ε). If the measured decay is only O(ε^{1/2}) (i.e., subtracting the first corrector produces no improvement), the leading corrector term is wrong. A direct spectral computation on the one-dimensional cell could compare the actual constant with the explicit bound from the proof.","tokens_in":41232,"feed_emoji":"🦘","tokens_out":6447,"duration_ms":62180,"temperature":0.7,"pith_summary":"This paper proves a sharper homogenization theorem for a family of nonlocal Lévy-type operators with rapidly oscillating periodic coefficients. Previous work showed that the resolvent of the oscillating operator converges to the resolvent of a constant-coefficient effective operator with error O(ε^(2−α)), a rate that deteriorates as the jump index α approaches 2. Here the authors add explicit corrector terms—differential operators built from a matrix that solves auxiliary cell problems—and show that after subtracting the first N correctors, the operator-norm error becomes O(ε), for every α in (1,2). The significance is that even though the underlying jump processes have infinite second moments and the operators are nonlocal, quantitative homogenization works at the same linear rate familiar from elliptic equations. A byproduct is that the earlier O(ε^(2−α)) bound is order-sharp.","feed_headline":"Nonlocal jump operators homogenize at order ε with correctors","feed_subtitle":"Adding corrector terms restores linear convergence for every Lévy index α, matching elliptic rates.","key_machinery":"The proof uses the operator-theoretic spectral method: after a scaling transformation and a Gelfand transform, the operator A is decomposed into a direct integral of fiber operators A(ξ) acting on periodic functions, with discrete spectra. Homogenization is a threshold effect at the bottom of the spectrum: for small quasi-momentum ξ, the spectral projector F(ξ) and the operator A(ξ)F(ξ) are approximated by (μ_0 c_0 |ξ|^α + ⟨g^0 ξ,ξ⟩) P plus a remainder of order |ξ|^{1+α}. This threshold approximation, derived via contour integration of the resolvent and a resolvent identity for quadratic forms with a common domain, yields the corrector expansion at the operator level.","core_discovery":"The central result (Theorem 5.2) is that for any integer N with 2 − 1/N < α < 2, the resolvent of the ε-periodic Lévy operator A_ε satisfies ‖(A_ε + I)^{-1} − (A^0 + I)^{-1} − Σ_{m=1}^{N} ε^{m(2−α)} K_m‖ ≤ C ε when α ≤ 2 − 1/(N+1), and ≤ C ε^{(N+1)(2−α)} otherwise. Here A^0 is the effective operator, a multiple of the fractional Laplacian (−Δ)^{α/2}, and the correctors are K_m = (div g^0 ∇)^m (A^0 + I)^{-m−1}, with g^0 an explicit (not necessarily sign-definite) symmetric matrix defined through periodic cell problems. Thus, by choosing N so that 2 − 1/N < α ≤ 2 − 1/(N+1), one obtains an operator-norm approximation of order ε for every α in (1,2).","pith_inferences":["The same threshold-expansion strategy may extend to nonsymmetric Lévy kernels or stable-like processes with variable characteristics, where only strong-convergence rates are currently known; the spectral machinery here suggests operator-norm rates are plausible but would require a new treatment of non-self-adjoint fibers.","Because g^0 need not be sign-definite (Remark 3.12), the corrector terms are not positivity-preserving, so a probabilistic interpretation of the correction as an additional drift or diffusion is not straightforward.","The constants in the estimates diverge as α → 1 and α → 2, so the O(ε) accuracy is genuine for each fixed α but not uniform in α; tracking this divergence might yield a two-parameter asymptotic description near the endpoints.","The structure of the expansion—powers ε^{m(2−α)} with coefficients built from powers of div(g^0∇) applied to resolvents—suggests a full asymptotic series in ε^{2−α} may exist, with each coefficient determined by g^0 and the cell functions."],"forward_implications":["For any fixed α in (1,2), choosing N with 2 − 1/N < α ≤ 2 − 1/(N+1) gives an operator-norm approximation of the resolvent with error of order ε, independent of how close α is to 2.","Taking N larger than needed gives remainders of order ε^{(N+1)(2−α)}, so for each fixed α the accuracy can be made arbitrarily high by including more correctors.","The earlier O(ε^{2−α}) bound from the authors' previous work is order-sharp: no better rate is possible without correctors.","The correctors are explicit and computed from the effective operator and the matrix g^0, which is defined by solving periodic cell problems; the approximation is therefore implementable in practice.","The main estimate remains valid on any periodic lattice, with constants depending on the lattice parameters."],"fun_headline_variants":["Lévy homogenization: correctors yield O(ε) rate","Nonlocal jumps: order-ε convergence via correctors","Correctors give linear rate for Lévy-type operators","For every α, Lévy resolvents match to O(ε)","Operator estimates with correctors: Lévy case hits ε"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The coefficient μ must stay bounded below by a positive constant (uniform ellipticity); if that lower bound were zero, the spectral gap d_0 would close, and the contour Γ, the spectral projection F(ξ), and the whole corrector expansion would collapse.","fun_headline_variants_meta":{"raw":{"variants":["Lévy homogenization: correctors yield O(ε) rate","Nonlocal jumps: order-ε convergence via correctors","Correctors give linear rate for Lévy-type operators","For every α, Lévy resolvents match to O(ε)","Operator estimates with correctors: Lévy case hits ε"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000634,"raw_usage":{"total_tokens":2902,"prompt_tokens":1027,"completion_tokens":1875,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":771,"completion_tokens_details":{"reasoning_tokens":1790}},"tokens_in":771,"tokens_out":1875,"duration_ms":15541,"temperature":1.0,"reasoning_tokens":1790,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T11:16:44.018397+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take d = 1, α = 3/2, and a periodic coefficient with nontrivial oscillation, for instance μ(x,y) = 2 + sin(2πx) + sin(2πy). For ε small, compute the operator norm of (A_ε + I)^{-1} − (A^0 + I)^{-1} − ε^{1/2} K_1. The theorem predicts this norm decays as O(ε). If the measured decay is only O(ε^{1/2}) (i.e., subtracting the first corrector produces no improvement), the leading corrector term is wrong. A direct spectral computation on the one-dimensional cell could compare the actual constant with the explicit bound from the proof.","supporting_citations":[],"review_version":1}