{"id":"5bd08807-3665-47aa-8161-3f29cf249332","arxiv_id":"1908.05892","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For 0<p<q<r, solutions of the three-scale parabolic equation converge to an elliptic homogenized equation, with 13 regimes and resonance shifted by p.","lead":"The paper proves new compactness results for multiscale and very weak multiscale convergence, then uses them to derive the homogenized limit of a parabolic equation with two rapid spatial and two rapid temporal scales. The limit is an elliptic problem in all 13 cases classified, and resonance appears when a temporal scale matches a spatial scale squared, shifted by the exponent p.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 10's uniqueness assertion for the homogenized solution is unsupported: coercivity of b is never proved and the parabolic cell problems' well-posedness is not verified.","rationale":"After checking the verification of (2)-(3) in (20)-(21), the compactness assumptions are satisfied for the solution sequence thanks to 0<p<q<r; the limit passages in Theorem 10 are consistent: negative-power terms like ε^{2+p-r}uε are handled either by the very weak convergence of ε^{-1}uε or ε^{-2}uε multiplied by a vanishing prefactor, or by setting the corresponding temporal test function to 1 once independence is established. The case split appears exhaustive and the formulas match the expected independencies. The main weakness is that Theorem 10 states uniqueness of u and a characterization of b, but the proof never shows b is coercive or that the parabolic cell problems are well-posed. This is a standard but nontrivial gap; it does not indicate the result is false, but it means the theorem as stated is not fully proven. The reader's conditional verdict is appropriate and our concern does not require changing it.","tokens_in":24506,"tokens_out":50022,"duration_ms":413365,"concrete_test":"Fix a representative parabolic case, e.g., r=4+p with q<2+p, so the cell problem is (40)-(41). For fixed s1, write (40) as ∂_{s2}u2 - ∇y2·(a∇y2u2) = ∇y2·(a(∇u + ∇y1u1)). Verify the Fredholm compatibility condition by integrating over Y2×S2 and checking that the right-hand side has zero mean. Then prove existence and uniqueness of a periodic solution u2 ∈ W2,2 by the Lax-Milgram theorem on the space of zero-mean functions, and likewise prove that the elliptic problem (41) for u1 is coercive. Finally, with these solutions, compute bξ for ξ = ∇u and verify ξ·bξ ≥ C0|ξ|^2 via the variational characterization of the cell problem. If the compatibility condition fails or b is not coercive for some admissible a, the characterization and uniqueness claim in Theorem 10 would fail.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim includes that u is the unique solution of the elliptic homogenized problem (26), with b characterized by the cell problems (27)-(57). The proof of Theorem 10 passes to the limit in (60)-(62) to derive those cell problems, but it never establishes that b is coercive or that the cell problems are well-posed. In the parabolic cases (e.g., (31), (36), (40), (42)-(43), (44), (50), (54)), existence of a periodic-in-time solution requires a compatibility condition (Fredholm alternative) which is not checked; the limiting procedure supplies one solution, but uniqueness of the cell problem is needed for b to be a well-defined characteristic of the microstructure. Without coercivity of b, the elliptic problem (26) may fail to have a unique solution, so the statement that the whole sequence converges to the unique u is not fully proven. This is a gap in the proof rather than a demonstrated contradiction, but it is load-bearing because the theorem's uniqueness and characterization assertions depend on it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes compactness results for evolution multiscale and very weak multiscale convergence for sequences bounded in L2(0,T;H1_0(Ω)) that satisfy the integral conditions (2) and (3), which replace the usual boundedness of the time derivative in L2(0,T;H^{-1}(Ω)). These results are then applied to homogenize the parabolic problem ε^p ∂t uε - ∇·(a(x/ε,x/ε^2,t/ε^q,t/ε^r)∇uε)=f with 0<p<q<r. The main homogenization theorem, Theorem 10, states that uε converges weakly in L2(0,T;H1_0(Ω)) to the unique solution u of the elliptic problem -∇·(b∇u)=f, with the homogenized coefficient b characterized by 13 families of local problems depending on the relative sizes of p,q,r. The paper highlights two phenomena: the homogenized problem is elliptic even though the original problem is parabolic, and parabolic resonance occurs when a temporal scale multiplied by ε^{-p} matches the square of a spatial scale.","tokens_in":24689,"tokens_out":22586,"duration_ms":194159,"significance":"If the gaps identified below are filled, the paper is a competent extension of the multiscale homogenization framework of Allaire-Briane and Flodén et al. The compactness theorems are of independent interest because they avoid the usual bound on the time derivative. The derivation is essentially self-contained and the nonstandard hypotheses (2)-(3) are verified directly from the equation in Section 3. The 13-case homogenization theorem gives concrete, falsifiable predictions for different scale matchings. The main weakness is that the proof of Theorem 10 does not supply the standard well-posedness and coercivity arguments on which the uniqueness and whole-sequence convergence claims rest; these arguments are routine and local, so the central claim is defensible.","major_comments":[{"comment":"The theorem asserts that u is the unique solution of the elliptic homogenized problem (26) and that b is characterized by the local problems (27)-(57). The proof passes to the limit and derives weak forms of these local problems, but it never proves that the cell problems are well-posed or that b is coercive. For the parabolic cell problems, e.g. (31), (36), (40), (42)-(43), (44), (50) and (54), existence of a periodic-in-time solution requires a Fredholm compatibility condition; this condition is indeed automatic for the displayed equations because their right-hand sides are in divergence form, but that verification is not included. Coercivity of b, needed for the uniqueness assertion in (26), follows from standard energy identities obtained by testing the local equations with u1 and u2 and integrating by parts, but these identities are not stated anywhere. Because b must be shown to be a well-defined coercive tensor for the limit u to be unique and for the whole sequence (rather than a subsequence) to converge, this is a load-bearing gap. Please add a lemma (or a paragraph in the proof of Theorem 10) proving well-posedness of each type of local problem and the estimate bξ·ξ ≥ C0|ξ|^2 for all ξ ∈ R^N.","section":"Section 3, Theorem 10 (after Eq. (62))"}],"minor_comments":[{"comment":"In the displayed weak form (58), the first occurrence of 'u(x,t)' inside the integral should be '∇u(x,t)'; as printed, the expression is not the weak form of (26) and is dimensionally inconsistent.","section":"Section 3, Eq. (58)"},{"comment":"After the divergence-free test functions are introduced, the text invokes the H^{-1}(Ω)-boundedness of ε^{-2}∇y1·v and then passes to the limit. It should say explicitly that ε^{-1}∇y1·v = ε(ε^{-2}∇y1·v) → 0 in H^{-1}(Ω); the current wording leaves this step implicit.","section":"Section 2, proof of Theorem 6"},{"comment":"The hypotheses of these theorems do not restate the joint well-separatedness of the scale lists {ε,ε^2} and {ε^q,ε^r}. This property is automatic for 0<q<r (after removing duplicates the exponents are strictly increasing), but stating it explicitly would make the theorems self-contained.","section":"Section 2, Theorem 6 and Theorem 9"},{"comment":"The notation such as 'u1 ∈ L2(Ω_T × S1; W1,2)' is nonstandard and potentially confusing; a sentence in Notation 1 clarifying that this means u1(·,s2) ∈ W1,2 for almost every (x,t,s1) would remove ambiguity.","section":"Theorem 10, cases 2, 4, 6, 7, 10, 12"},{"comment":"The relationship to the authors' earlier works [13] and [6] could be made more explicit: [13] treats one spatial and one temporal scale, and the novelty of the two-scale spatial/temporal setting and of the shifted resonance condition should be stated more concretely.","section":"Introduction"},{"comment":"Reference [5] is a preprint and [6] is 'to appear'; if published or updated versions exist, they should be cited with full bibliographic data.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a technical and incremental extension of the authors' previous work, but the core arguments are standard and likely correct once the well-posedness/coercivity gap in Theorem 10 is filled. The editor may wish to ask the authors to state clearly what is new relative to [13] and [6], since the two headline phenomena already appeared in those works in the one-scale setting. The reference list is somewhat insular; a broader comparison with the Allaire-Piatnitski and Nguetseng-Woukeng lines of research would strengthen the introduction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the compactness part is genuinely good, and the homogenization part is a plausible extension that gets ahead of its proof in exactly one load-bearing spot.\n\nThe genuinely new pieces are Theorems 6 and 9, which give (2,3)- and (3,3)-scale compactness under the integral conditions (2)-(3) instead of the usual boundedness of the time derivative in L2(H^{-1}). The proof of Theorem 6 is careful: they show the multiscale limit is independent of the fast space and time variables by testing with oscillatory test functions, then identify the gradient limit via the standard divergence-free Helmholtz decomposition from Allaire–Briane. The verification that solutions of (18) actually satisfy (2)-(3) is also clean and uses the equation in a straightforward way, exploiting the strict ordering 0<p<q<r. That part deserves credit.\n\nThe soft spot is Theorem 10. After extracting the two-scale limits, the authors derive the 13 local problems by passing to the limit in (60)-(62). Those passages are plausible but abbreviated; several cases are asserted rather than fully derived. More importantly, the paper never proves that the local problems are well-posed. In the parabolic cases (2, 4, 6, 7, 8, 10, 12), the operator ∂s - div(a∇) acting on periodic-in-time functions has a nontrivial kernel, so existence and uniqueness require a compatibility condition and a normalization over the time cell. The two-scale limit does supply one solution, but without uniqueness the coefficient b is not pinned down. And coercivity of b is never shown, so the uniqueness of the solution u to (26) does not follow. The stress-test note is on target: the gap is real, though it looks repairable and is not an internal contradiction. A careful referee should ask for a well-posedness analysis of the cell problems and a proof that b is coercive, or at least an explicit statement of the missing conditions.\n\nWho is this for? People working on periodic homogenization with several spatial and temporal scales. The compactness theorems may be the more durable part; the 13-case catalog is a useful reference once the proof is tightened. I'd engage with it: send it out for peer review, with the expectation of major revision rather than desk rejection.","headline":"A legitimate extension of the multiscale toolbox to three spatial and three temporal scales, with solid compactness theorems and a plausible 13-case homogenization classification, but the proof of Theorem 10 skips the cell-problem well-posedness and coercivity checks that the uniqueness statement needs.","tokens_in":25234,"tokens_out":3790,"would_cite":false,"duration_ms":41971,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B27","35K20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a parabolic equation with two fast spatial and two fast temporal scales, the paper proves a homogenized limit that is elliptic, with resonance occurring only under a p-shifted matching of scales.","keywords":["homogenization","multiscale convergence","very weak multiscale convergence","parabolic equations","elliptic homogenized problem","resonance","periodic coefficients","cell problems"],"falsifier":"Set $N=1$, $p=1$, $q=3$ (so $q=2+p$), $r=4$, and choose the coefficient $a(y_2,s_2)=2+\\sin(2\\pi s_2)$. Solve the two local problems of Theorem 10, case 4, to get the predicted homogenized coefficient $b$, then compute the actual effective coefficient from a two-scale asymptotic expansion of (18); agreement would confirm the $p$-shifted resonance, while a mismatch would indicate the matching condition is wrong.","tokens_in":24296,"feed_emoji":"","tokens_out":14073,"duration_ms":131500,"temperature":0.7,"pith_summary":"This paper establishes a homogenization limit for a linear parabolic equation with two rapidly oscillating spatial scales and two rapidly oscillating temporal scales (three spatial and three temporal scales in total, counting the macroscopic ones): $\\varepsilon^{p}\\partial_{t}u_{\\varepsilon}-\\nabla\\cdot(a(x/\\varepsilon,x/\\varepsilon^{2},t/\\varepsilon^{q},t/\\varepsilon^{r})\\nabla u_{\\varepsilon})=f$, with $0<p<q<r$. The paper proves that the solutions converge weakly in $L^{2}(0,T;H_{0}^{1}(\\Omega))$ to the unique solution of an elliptic homogenized problem $-\\nabla\\cdot(b\\nabla u)=f$, so the time derivative disappears from the effective equation even though the original problem is parabolic. The effective coefficient $b$ is characterized by local cell problems, listed in thirteen cases according to how $p$, $q$, and $r$ sit relative to $2+p$ and $4+p$. The paper also proves the compactness results behind this limit: bounded sequences satisfying two integral conditions admit multiscale and very weak multiscale limits without the usual bound on the time derivative. The second headline phenomenon is that cell problems become parabolic only when a temporal scale matches a spatial square after a shift by $p$, namely when $q-p=2$, $r-p=2$, $q-p=4$, or $r-p=4$.","feed_headline":"Three fast scales turn a parabolic PDE into an elliptic limit","feed_subtitle":"The effective equation drops the time derivative entirely, and cell problems go parabolic only under a p-shifted resonance.","key_machinery":"The machinery has three parts. First, evolution multiscale convergence and very weak multiscale convergence (Definitions 2 and 7) replace classical two-scale convergence for sequences with two spatial and two temporal microscopic scales. Second, Theorem 6 characterizes the $(2,3)$- and $(3,3)$-scale limits of $\\nabla u_{\\varepsilon}$ under the integral conditions (2)–(3), conditions that stand in for boundedness of $\\partial_{t}u_{\\varepsilon}$ in $L^{2}(0,T;H^{-1})$, and Theorem 9 identifies the very weak limits of $\\varepsilon^{-1}u_{\\varepsilon}$ and $\\varepsilon^{-2}u_{\\varepsilon}$ with the same correctors $u_1$ and $u_2$. Third, in the homogenization proof, two families of test functions, (59) and (61), with adjustable powers $k$ of $\\varepsilon$, are inserted into the weak formulation; choosing $k=r-p-2$, $q-p-2$, $r-p-1$, or $q-p-1$ isolates each local time variable and produces the thirteen systems of local problems that define the homogenized coefficient $b$.","core_discovery":"The central claim is Theorem 10: for every fixed $\\varepsilon$, the parabolic problem (18) has a unique solution, and as $\\varepsilon\\to 0$ the solutions converge weakly in $L^{2}(0,T;H_{0}^{1}(\\Omega))$ to the unique solution of the elliptic homogenized problem $-\\nabla\\cdot(b\\nabla u)=f$ with $u=0$ on the boundary. The gradient has the three-scale decomposition $\\nabla u_{\\varepsilon}\\rightharpoonup \\nabla u+\\nabla_{y_1}u_1+\\nabla_{y_2}u_2$, where the correctors $u_1$ and $u_2$ solve local problems that are elliptic in most of the thirteen parameter regimes and parabolic only in the resonant ones. The two phenomena emphasized by the paper are therefore: the homogenized problem is elliptic, and the resonance condition is shifted by $p$ — a temporal scale $\\varepsilon^{q}$ or $\\varepsilon^{r}$ acts like the square of a spatial scale when $q-p=2$ or $r-p=2$, with the analogue $q-p=4$ or $r-p=4$ for the second spatial scale, rather than when $q=2$ or $r=2$ as in the $p=0$ case. Theorems 6 and 9 provide the underlying compactness: under the integral conditions (2)–(3), the $(3,3)$-scale limit of the gradient is characterized by $u,u_1,u_2$, and the unbounded sequences $\\varepsilon^{-1}u_{\\varepsilon}$ and $\\varepsilon^{-2}u_{\\varepsilon}$ converge in the very weak multiscale sense to the same correctors.","pith_inferences":["The $p$-shift suggests a general rule: whenever the time derivative carries a vanishing factor $\\varepsilon^{p}$, resonance between a spatial scale $\\varepsilon^{k}$ and a temporal scale $\\varepsilon^{s}$ should occur at $s-p=2k$; this could be tested with a more general vanishing factor $\\varphi(\\varepsilon)$ instead of $\\varepsilon^{p}$.","Because the homogenized equation is elliptic, one expects an initial layer and boundary layers to carry the lost time information; a natural extension would construct correctors that capture these layers in stronger norms, which the paper does not do.","The very weak multiscale convergence of $\\varepsilon^{-1}u_{\\varepsilon}$ and $\\varepsilon^{-2}u_{\\varepsilon}$ shows that the correctors are uniquely determined even for unbounded sequences, suggesting the same condition-based compactness could handle stronger singular scalings if the test-function powers are adjusted accordingly."],"forward_implications":["If the paper is right, the macroscopic behavior of this multiscale parabolic medium is instantaneous: the effective equation is elliptic, so the initial data do not appear in the leading-order limit problem.","The effective diffusion matrix $b$ is computed from cell problems that are elliptic except when $r=2+p$, $q=2+p$, $r=4+p$, or $q=4+p$, in which case one of the cell problems becomes parabolic.","The convergence statements give rigorous meaning to the correctors: $\\varepsilon^{-1}u_{\\varepsilon}$ and $\\varepsilon^{-2}u_{\\varepsilon}$ converge very weakly to the same $u_1,u_2$ that appear in the gradient decomposition.","The compactness results do not require a uniform bound on $\\partial_t u_{\\varepsilon}$ in $L^{2}(0,T;H^{-1})$; the integral conditions (2)–(3), verified directly from the equation, are enough."],"supporting_citations":[{"why":"Previous homogenization result with a matching between one spatial and one temporal scale that this paper generalizes to the (2,3)- and (3,3)-scale settings.","marker":"[13]"},{"why":"Origin of the condition replacing boundedness of the time derivative, used to verify (20)–(21) from the equation.","marker":"[14]"},{"why":"Supplies the multiscale convergence framework and the density and orthogonal-decomposition results used in the proof of Theorem 6.","marker":"[2]"},{"why":"Provides the evolution multiscale compactness theorem (Theorem 4) and the general very weak multiscale framework used throughout.","marker":"[11]"},{"why":"Introduced the very weak convergence concept that Theorem 9 adapts to the present scales.","marker":"[12]"},{"why":"Improved the very weak convergence concept and supplies part of the definitional basis for Definition 7.","marker":"[18]"},{"why":"Describes the standard matching (temporal scale equal to the square of a spatial scale) against which the $p$-shift is measured.","marker":"[19]"},{"why":"Provides the a priori estimate used to obtain the uniform bound (19) on the solutions.","marker":"[5]"},{"why":"Gives existence and uniqueness of solutions for each fixed $\\varepsilon$ via linear monotone operator theory.","marker":"[22]"}],"fun_headline_variants":["Parabolic PDEs homogenize to elliptic limits via scale matching","Three-scale matching kills the time derivative in the limit","p-shifted resonance turns parabolic homogenization elliptic","Time scales vanish in homogenized limit of parabolic PDEs","Elliptic limit emerges from three-scale parabolic homogenization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result rests on the strict ordering $0<p<q<r$ together with the two integral conditions (2)–(3), which stand in for the usual bound on the time derivative; if those fail, the limit could keep oscillating in the fast time variables and the elliptic homogenized equation would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Parabolic PDEs homogenize to elliptic limits via scale matching","Three-scale matching kills the time derivative in the limit","p-shifted resonance turns parabolic homogenization elliptic","Time scales vanish in homogenized limit of parabolic PDEs","Elliptic limit emerges from three-scale parabolic homogenization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1424,"prompt_tokens":1068,"completion_tokens":356,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":684,"completion_tokens_details":{"reasoning_tokens":278}},"tokens_in":684,"tokens_out":356,"duration_ms":3861,"temperature":1.0,"reasoning_tokens":278,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:03:33.695562+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Set $N=1$, $p=1$, $q=3$ (so $q=2+p$), $r=4$, and choose the coefficient $a(y_2,s_2)=2+\\sin(2\\pi s_2)$. Solve the two local problems of Theorem 10, case 4, to get the predicted homogenized coefficient $b$, then compute the actual effective coefficient from a two-scale asymptotic expansion of (18); agreement would confirm the $p$-shifted resonance, while a mismatch would indicate the matching condition is wrong.","supporting_citations":[{"cited_title":"Johnsen, T","cited_arxiv_id":null,"evidence_quote":"Previous homogenization result with a matching between one spatial and one temporal scale that this paper generalizes to the (2,3)- and (3,3)-scale settings."},{"cited_title":"Lobkova : Homogenization of linear parabolic equations with a certain resonant matching between rapid spatial and temporal oscillations in peri- odically perforated domains","cited_arxiv_id":null,"evidence_quote":"Origin of the condition replacing boundedness of the time derivative, used to verify (20)–(21) from the equation."},{"cited_title":"Allaire, M","cited_arxiv_id":null,"evidence_quote":"Supplies the multiscale convergence framework and the density and orthogonal-decomposition results used in the proof of Theorem 6."},{"cited_title":"Flod´ en, A","cited_arxiv_id":null,"evidence_quote":"Provides the evolution multiscale compactness theorem (Theorem 4) and the general very weak multiscale framework used throughout."},{"cited_title":"Holmbom : Homogenization of parabolic equations: an alternative ap- proach and some corrector-type results","cited_arxiv_id":null,"evidence_quote":"Introduced the very weak convergence concept that Theorem 9 adapts to the present scales."},{"cited_title":"Nguetseng, J","cited_arxiv_id":null,"evidence_quote":"Improved the very weak convergence concept and supplies part of the definitional basis for Definition 7."},{"cited_title":"Persson : Selected Topics in Homogenization","cited_arxiv_id":null,"evidence_quote":"Describes the standard matching (temporal scale equal to the square of a spatial scale) against which the $p$-shift is measured."},{"cited_title":"Danielsson, P","cited_arxiv_id":null,"evidence_quote":"Provides the a priori estimate used to obtain the uniform bound (19) on the solutions."},{"cited_title":"Zeidler : Nonlinear functional analysis and its applications II/A: linear monotone operators","cited_arxiv_id":null,"evidence_quote":"Gives existence and uniqueness of solutions for each fixed $\\varepsilon$ via linear monotone operator theory."}],"review_version":1}