{"id":"17ddeec3-c86a-4b0f-9048-9695207fe7a0","arxiv_id":"2504.17396","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Rapidly oscillating periodic coefficients whose period decays fast enough near the boundary yield Carleson measure estimates, and hence well-behaved elliptic measure, despite violating the DKP condition.","lead":"This paper proves that elliptic PDEs with coefficients oscillating very fast near a flat boundary can still have well-behaved harmonic measure, even though the standard DKP oscillation condition fails. It uses periodic homogenization as a new substitute for that condition, establishing a proof-of-concept for rapidly oscillating materials near a boundary.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Hypothesis (16) for \\bar A is unproved for general piecewise constant \\bar A; the unconditional result is the constant-homogenized example.","rationale":"The reader's weakest_assumption correctly identifies the same load-bearing point: Theorem 2's hypothesis (16) is not proved for general piecewise constant homogenized coefficients (14). I re-examined the proof of Theorem 2 in detail, including the gluing of local two-scale expansions across Whitney cubes, the scaling in (26) and (33), the Meyers estimate optimization, and the treatment of the interface t=1. I found no internal error: the local identities sum correctly because the cut-offs are supported in disjoint cubes, the divergence identity (9) kills the flux-corrector terms, and the weight scaling works as stated. The theorem is a genuine conditional transfer principle, and the concrete example with A_kj=A_per and A_\\infty=\\bar A_per is unconditional because \\bar A is then constant and (16) follows from classical DKP theory. The only substantive limitation is that the paper does not establish (16) for nonconstant \\bar A, and the offered sufficient condition (DKP) is generally false for such matrices. This limits the breadth of the advertised 'new class' but does not falsify the theorem or the example. Therefore I agree with the reader's identification of the weak assumption and with the ACCEPT verdict: the concern is real but is an explicit hypothesis, not an error, and the paper's central example survives it.","tokens_in":11061,"tokens_out":31606,"duration_ms":301242,"concrete_test":"Take N=1 (two-dimensional half-plane) and let \\bar A on the Whitney cubes be a checkerboard: \\bar A_{kj}=A_1 for j+k even, A_2 for j+k odd, with A_1\\neq A_2 elliptic matrices, and A_\\infty=A_1. Solve the Dirichlet problem (15) with a smooth bounded datum, e.g. f(x)=\\sin x, on a finite set of generations k=-1,\\dots,-20 (truncated at depth), using a standard finite-element solver. Monitor Q(R)=R^{-1}\\int_{T_R} t|\\nabla\\bar u|^2 for R=2^k. If Q(R) grows without bound as R\\to 0, then (16) fails for this natural piecewise constant \\bar A, confirming that the theorem's generality beyond the constant case rests on an unproved and non-obvious assumption. If Q(R) remains bounded, the concern is mitigated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is hypothesis (16) in Theorem 2. The theorem transfers the Carleson estimate from the homogenized solution \\bar u to the oscillating solution u, but it does not prove (16) for the piecewise constant \\bar A in (14). Proposition 1 supplies (16) only when \\bar A satisfies DKP, and a piecewise constant \\bar A that is not constant on Whitney cubes will generally violate DKP: near an internal face between two different constant blocks, \\alpha(Z) is of order 1 on a set of positive measure, making \\int_{T_R} \\alpha^2/\\delta diverge logarithmically as R\\to 0. Thus, unless \\bar A is constant (as in the paper's example) or satisfies some other unproved condition, the theorem's hypothesis is not verified. This does not invalidate the theorem (the hypothesis is explicit) or the example (where \\bar A is constant and (16) holds by classical theory), but it means the advertised 'new class' is, beyond the example, a conditional statement rather than an established family.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the Dirichlet problem for -div(A∇u)=0 in the upper half-space with rapidly oscillating periodic coefficients. The main result (Theorem 2) is a conditional transfer: if the solution of the homogenized problem with the piecewise constant matrix \\bar A defined in (14) satisfies the Carleson measure estimate (16), then the solution for the oscillating matrix A defined in (13) satisfies the same estimate, provided the period ε_{kj} is smaller than 2^{α(p)k} with α(p)=(3p−1)/(2(p−1)). The proof combines localized two-scale expansions, Caccioppoli inequalities, Meyers reverse Hölder estimates, and an optimization over the Whitney-cube boundary layer parameter η. A concrete example with A_{kj}=A_per and A∞=\\bar A_per gives a constant homogenized matrix, so (16) follows classically; this yields a family of coefficients that violate the DKP condition yet have well-behaved elliptic measure.","tokens_in":11282,"tokens_out":21783,"duration_ms":189811,"significance":"The paper introduces homogenization as a new mechanism for establishing Carleson measure estimates and hence absolute continuity of elliptic measure, complementing the DKP paradigm. The concrete example is a genuine infinite-dimensional family of DKP-violating coefficients for which the estimate holds, and the proof is largely self-contained, with the main estimates in Steps 1 and 2 being standard and correctly assembled. The principal limitation is that Theorem 2's hypothesis (16) is not verified for the general piecewise constant \\bar A in (14); nevertheless, the example provides a solid unconditional application and the transfer principle is of independent interest, so the contribution is significant if the framing is made precise.","major_comments":[],"minor_comments":[{"comment":"The hypothesis (16) is not verified for a general piecewise constant \\bar A in (14); Remark 1 only supplies the DKP condition, which fails when \\bar A has jumps across Whitney-cube faces, as the authors themselves note. The paper should state explicitly that the only fully verified instances of (16) are the constant (or laminate) homogenized cases, so that the generality of the theorem beyond the example is conditional on an unproved assumption.","section":"Theorem 2, hypothesis (16)"},{"comment":"The right-hand sides of (16) and of the conclusion appear as C∥f∥_{L∞(R^N)}, but the estimate is quadratic in the data; the correct term is C∥f∥^2_{L∞(R^N)}, consistent with Proposition 1 and with the proof's normalization ∥f∥_{L∞}=1.","section":"Theorem 2, statement"},{"comment":"The concrete example should specify a choice of ε_{kj} (for instance ε_{kj} ≤ 2^k via Corollary 1, or ε_{kj} ≤ 2^{α(p)k} via Theorem 2) so that the reader sees exactly how the smallness condition is satisfied in the DKP-violating family.","section":"Example after Theorem 2"},{"comment":"The localized two-scale expansion error equation (12) is central, but its derivation is only cited to [8]; since the authors provide a full derivation of the global equation (5), adding a brief sketch of the additional boundary terms for the localized case would improve self-containedness.","section":"Section 2, Step 2, equation (12)"},{"comment":"There are a few minor typographical issues: 'Meyer's inequality' should read 'Meyers' inequality', and the phrase 'the counter-example of Theorem 1' would read better as 'the counterexample of Theorem 1'.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"The paper is mathematically sound and the example is convincing. The main weakness is the conditional nature of Theorem 2, which should be clarified in the abstract and introduction so that the 'new class' is not overclaimed beyond the concrete family of examples. The missing square in the theorem statement is a minor typo. I see no load-bearing technical error, and I recommend minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read of the David–Gloria–Qi–Mayboroda note.\n\nThe genuinely new thing is using small-scale homogenization inside Whitney cubes as a substitute for the DKP condition. Earlier work by Kenig–Lin–Shen homogenizes at large scales after DKP gives small scales; here the period shrinks fast enough toward the boundary that the two-scale expansion works at every Whitney cube individually. The localized expansion with correctors and flux correctors is done carefully, and the optimization in η recovering ε^{2(p-1)/(3p-1)} is clean. That part is solid.\n\nI also think the paper is honest about its main hypothesis. Theorem 2 transfers the Carleson estimate from the homogenized solution \\bar u to the oscillating u, and it explicitly assumes (16). There is no circularity; the example then verifies (16) by taking \\bar A constant.\n\nThe soft spot is the gap between the theorem and the advertised 'new class'. For the general piecewise constant \\bar A in (14), hypothesis (16) is not proved. Proposition 1 only provides it under DKP, and the piecewise-constant \\bar A will typically fail DKP at internal faces between different constants, with α^2/δ integrating to infinity. So beyond the constant-homogenized example (and the laminate case where Corollary 1 gives an exponent), the statement 'new class of operators' is conditional on an unverified input. The example itself is enough to make the point against necessity of DKP, but the abstract slightly overstates the generality.\n\nMinor: the footnote on fractal invariance is a teaser, not a theorem; fine for a short note. The citation of Kenig–Lin–Shen is accurate and the distinction (small-scale vs large-scale homogenization) is correctly drawn.\n\nNet: this is a solid proof-of-concept, not a sweeping theory. It deserves a serious referee and I'd be happy to see it published in a good journal, perhaps after the authors soften 'new class' to 'a class of examples' or prove (16) for a broader family of \\bar A. I'd take it seriously.","headline":"Small-scale homogenization as a DKP replacement is a real idea, and the proof is honest; the main theorem is conditional on an unproved Carleson assumption for the piecewise-constant homogenized matrix, but the constant-homogenized example delivers an unconditional non-DKP case.","tokens_in":11784,"tokens_out":2960,"would_cite":true,"duration_ms":28575,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B27","35J25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Homogenization can replace the DKP small-oscillation condition for good elliptic measure.","keywords":["harmonic measure","elliptic measure","Carleson measure estimates","periodic homogenization","Dahlberg-Kenig-Pipher condition","two-scale expansion","Meyers exponent","Dirichlet problem"],"falsifier":"Take the piecewise-constant homogenized matrix $\\bar A$ from (14) with two different constant values on adjacent Whitney cubes, so it violates DKP, and compute the solution $\\bar u$ to (15) on the half-space with $f\\equiv 1$; if the integrals $R^{-1}\\int_{T_R} t|\\nabla\\bar u|^2$ along the boundary between the two regions grow without bound, then hypothesis (16) fails and Theorem 2 does not apply. Conversely, if for some $A$ of (13) with $\\varepsilon_{kj}\\le 2^{\\alpha(p)k}$ the Carleson estimate for $u$ fails while $\\bar A$ still satisfies (16), the claimed transfer would be false.","tokens_in":10885,"feed_emoji":"📐","tokens_out":10533,"duration_ms":88137,"temperature":0.7,"pith_summary":"The paper introduces homogenization theory into the study of harmonic (elliptic) measure for second-order elliptic operators in a half-space. Its goal is to show that the standard small-oscillation threshold for good boundary behavior, the Dahlberg–Kenig–Pipher (DKP) condition, can be replaced by a very different kind of closeness: resolvent closeness to a constant-coefficient problem brought about by rapid periodic oscillation. The main theorem proves that if the homogenized coefficients satisfy a Carleson estimate on gradients, and if the microscopic period in each boundary cell shrinks fast enough as the boundary is approached, then the original operator satisfies the same Carleson estimate and therefore has well-behaved elliptic measure. In the concrete case where all cells share one periodic matrix whose homogenization is constant, the coefficients violate DKP by a wide margin and the conclusion still holds. In short, violent oscillation is not an obstacle to good boundary behavior but a mechanism that can produce it.","feed_headline":"Rapidly oscillating coefficients can still give good harmonic measure","feed_subtitle":"A periodic coefficient matrix that violates DKP still satisfies the Carleson estimate that guarantees good elliptic measure.","key_machinery":"The load-bearing mechanism is a local two-scale expansion $u^{2s}=\\bar u+\\sum_{k,j}2^k\\varepsilon_{kj}\\chi_{kj}\\phi^i_{kj}(\\cdot/(2^k\\varepsilon_{kj}))\\partial_i\\bar u$, with $\\phi^i$ periodic correctors and $\\chi_{kj}$ cut-offs inside each Whitney cube. A skew-symmetric flux corrector $\\sigma^i$ makes the error $z=u-u^{2s}$ solve $-\\nabla\\cdot A\\nabla z=\\nabla\\cdot f$ with a source whose size is controlled by $\\varepsilon_{kj}$, the correctors' boundedness, and the second derivatives of $\\bar u$. Caccioppoli's inequality, De Giorgi–Nash–Moser regularity, and Meyers' reverse Hölder inequality let the authors trade the bad boundary-layer term for a small power of $\\varepsilon_{kj}$; choosing $\\eta_{kj}=\\varepsilon_{kj}^{2p/(3p-1)}$ balances the two error terms and yields $\\int_{T_R}|f|^2\\lesssim R^N$ exactly when $\\varepsilon_{kj}\\lesssim 2^{\\alpha(p)k}$.","core_discovery":"The paper's central claim is Theorem 2: let $A$ be the locally periodic matrix (13), built from periodic fields on Whitney cubes with periods $2^k\\varepsilon_{kj}$, and let $\\bar A$ be its piecewise-constant homogenization (14). If every bounded solution $\\bar u$ of $-\\nabla\\cdot \\bar A\\nabla\\bar u=0$ satisfies the Carleson estimate $R^{-N}\\int_{T_R(x)} t|\\nabla\\bar u|^2\\le C\\|f\\|^2_{L^\\infty}$, and if $\\varepsilon_{kj}\\lesssim 2^{\\alpha(p)k}$ with $\\alpha(p)=(3p-1)/(2(p-1))$ for the Meyers exponent $2p>2$, then every bounded solution $u$ of the original problem satisfies the same estimate. Taking $A_{kj}=A_{\\rm per}$ fixed and $A_\\infty=\\bar A_{\\rm per}$ makes $\\bar A$ constant, so hypothesis (16) holds by classical regularity, while $A$ itself has oscillations of order one at every scale and hence fails DKP; the elliptic measure is nevertheless well behaved. The paper presents this as the positive counterpart to known counterexamples showing that DKP is necessary within the pointwise-oscillation class.","pith_inferences":["The proof needs the homogenized solution hypothesis (16) only inside tents near the boundary; combining it with existing large-scale homogenization methods could remove the simplifying assumption that $A$ is constant for $t>1$ and yield a two-scale theory valid at all distances.","The exponent $\\alpha(p)=(3p-1)/(2(p-1))$ decreases from $+\\infty$ to $3/2$ as the Meyers exponent $p$ increases, while the improved Corollary 1 rates ($3/2$ for laminates, $1$ for constants) show the generic balance is not optimal; testing whether intermediate structure yields intermediate exponents would clarify the sharp rate.","One could seek homogenized matrices $\\bar A$ that satisfy the Carleson estimate (16) without satisfying DKP; if such matrices exist, Theorem 2 would cover a genuinely larger class of oscillating coefficients rather than only the constant-homogenization example."],"forward_implications":["Such rapidly oscillating operators can have $L^p$-solvable Dirichlet problems and absolutely continuous elliptic measure despite violating the DKP condition.","The small-scale homogenization strategy is local: the assumption that $A$ is constant for $t>1$ is used only to control large tents, so the approach can be combined with large-scale homogenization for more general far-field behavior.","If the homogenized matrix has more structure, the required period shrinkage is milder: laminates only need $\\varepsilon_{kj}\\lesssim 2^{3k/2}$, and constant matrices only $\\varepsilon_{kj}\\lesssim 2^k$ (Corollary 1).","The concrete periodic example is a positive counterpart to the known counterexample in which DKP-type failure destroys absolute continuity of elliptic measure."],"supporting_citations":[{"why":"supplies Proposition 1, the DKP-to-Carleson estimate used for the far field and as the baseline condition on homogenized solutions.","marker":"[13]"},{"why":"provides Theorem 1, the counterexample showing DKP-type failure can destroy absolute continuity, which motivates the new class.","marker":"[5]"},{"why":"furnishes the standard homogenization framework and corrector definition that underlies $\\bar A$.","marker":"[10]"},{"why":"gives the localized two-scale expansion and the error equation used to control $u-u^{2s}$.","marker":"[8]"},{"why":"supplies the reverse Hölder/Meyers estimates for gradients, used to bound the Whitney boundary-layer term.","marker":"[6]"},{"why":"provides the regularity theorem for elliptic systems behind the Meyers exponent estimates.","marker":"[18]"},{"why":"documents the lack of Lipschitz regularity at corners for piecewise-constant coefficients, explaining the need for the Meyers route.","marker":"[11]"},{"why":"gives the Lipschitz regularity for laminates used in Corollary 1 to relax the period condition.","marker":"[2]"},{"why":"represents the large-scale homogenization approach the present small-scale strategy contrasts with and can be mixed with.","marker":"[15]"}],"fun_headline_variants":["Rapid oscillations still yield good harmonic measure","Violating DKP but still well-behaved elliptic measure","Homogenization ensures good elliptic measure despite DKP failure","Periodic coefficients with fast oscillations pass Carleson estimate","Homogenization beats DKP for good harmonic measure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the homogenized problem already has the property being proved: every bounded solution of $-\\nabla\\cdot \\bar A\\nabla\\bar u=0$ must satisfy the same tent Carleson bound (16), and this is only shown when $\\bar A$ satisfies the DKP condition or is constant.","fun_headline_variants_meta":{"raw":{"variants":["Rapid oscillations still yield good harmonic measure","Violating DKP but still well-behaved elliptic measure","Homogenization ensures good elliptic measure despite DKP failure","Periodic coefficients with fast oscillations pass Carleson estimate","Homogenization beats DKP for good harmonic measure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000637,"raw_usage":{"total_tokens":2922,"prompt_tokens":921,"completion_tokens":2001,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":1923}},"tokens_in":537,"tokens_out":2001,"duration_ms":14509,"temperature":1.0,"reasoning_tokens":1923,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:43:45.861909+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the piecewise-constant homogenized matrix $\\bar A$ from (14) with two different constant values on adjacent Whitney cubes, so it violates DKP, and compute the solution $\\bar u$ to (15) on the half-space with $f\\equiv 1$; if the integrals $R^{-1}\\int_{T_R} t|\\nabla\\bar u|^2$ along the boundary between the two regions grow without bound, then hypothesis (16) fails and Theorem 2 does not apply. Conversely, if for some $A$ of (13) with $\\varepsilon_{kj}\\le 2^{\\alpha(p)k}$ the Carleson estimate for $u$ fails while $\\bar A$ still satisfies (16), the claimed transfer would be false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies Proposition 1, the DKP-to-Carleson estimate used for the far field and as the baseline condition on homogenized solutions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides Theorem 1, the counterexample showing DKP-type failure can destroy absolute continuity, which motivates the new class."},{"cited_title":"Giaquinta and L","cited_arxiv_id":null,"evidence_quote":"supplies the reverse Hölder/Meyers estimates for gradients, used to bound the Whitney boundary-layer term."},{"cited_title":"Uhlenbeck","cited_arxiv_id":null,"evidence_quote":"provides the regularity theorem for elliptic systems behind the Meyers exponent estimates."},{"cited_title":"Kenig, H","cited_arxiv_id":null,"evidence_quote":"documents the lack of Lipschitz regularity at corners for piecewise-constant coefficients, explaining the need for the Meyers route."},{"cited_title":"Chipot, D","cited_arxiv_id":null,"evidence_quote":"gives the Lipschitz regularity for laminates used in Corollary 1 to relax the period condition."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"represents the large-scale homogenization approach the present small-scale strategy contrasts with and can be mixed with."}],"review_version":1}