{"id":"080361fb-963f-4372-bba8-6b360870b56e","arxiv_id":"2506.15869","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Polygonal curves driven by crystalline elastic energy globally converge to stationary polygons, with a complete classification for square anisotropy.","lead":"This paper proves that polygonal curves evolving by a crystalline elastic energy exist globally and converge to stationary shapes. It also classifies all stationary curves for the square anisotropy.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 7.5 is applied to degenerate stationary curves, but its proof uses Lemma 2.5 which requires positive segment lengths; the irregular-case convergence in Theorem 7.3 is not fully supported.","rationale":"The reader's accepted verdict identifies the restriction to non-breaking facets as the main weakness. I agree that this restriction is explicit and standard, and I do not object to it. However, a more internal concern arises in the proof of the irregular-case convergence theorem. Theorem 7.3 is the strongest long-time result in the paper, and its proof depends on Proposition 7.5, the Lojasiewicz-Simon inequality for generalized stationary curves with possibly degenerate zero-curvature segments. The proof of Proposition 7.5 uses Lemma 2.5 to construct an open neighbourhood U of height vectors. Lemma 2.5 is only proved for polygonal curves whose segments all have positive length. When the reference curve has degenerate segments, as is allowed in Proposition 7.5 and needed for Theorem 7.3, the lemma does not apply directly. Additionally, the admissible height coordinates are not independent for all indices, so U is not open in the full R^n. This is not necessarily a fatal flaw: the Lojasiewicz inequality can often be formulated on subanalytic sets, and the argument may be repairable by working with the nondegenerate skeleton and treating degenerate zero-curvature segments as limiting points. But as written the proof is incomplete at a load-bearing point. The regular case, Theorem 7.1, is not affected by this concern, and the rest of the paper appears sound. Therefore I would not reject the paper, but I would ask the authors to supply the missing perturbation and subanalyticity argument for Proposition 7.5 before the irregular-case convergence theorem is taken as fully established.","tokens_in":46720,"tokens_out":18242,"duration_ms":186317,"concrete_test":"Specialize to the square anisotropy W=[-1,1]^2 and take a generalized stationary curve containing a degenerate zero-curvature segment, for instance a closed right-angle chain with one vertical segment of length zero. Write down the admissible height vectors h explicitly and check whether U is open in R^n or only a lower-dimensional subanalytic set. Then re-derive inequality (7.17) by applying the Lojasiewicz gradient inequality directly on this subanalytic set, or produce a counterexample. If the inequality cannot be derived, Theorem 7.3 lacks a proof in exactly the case it is designed to handle.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Theorem 7.3 covers the case where some segment lengths tend to zero only at infinity, and its proof relies on the second Lojasiewicz-Simon inequality, Proposition 7.5. The proof of Proposition 7.5 defines a set U of admissible height vectors and claims U is open by 'applying Lemma 2.5' to an associated curve eΓ. However, Lemma 2.5 requires every segment of eΓ to have positive length, since it bounds the perturbing heights by min_i H1(S_i). In the setting of Proposition 7.5, the generalized stationary curve Γ may contain degenerate zero-curvature segments, and the associated curves eΓ in U may also have zero-length segments; for such curves Lemma 2.5 does not apply. Moreover, for indices i not in J, i.e. with c_i=0 and |c_{i-1}|=|c_{i+1}|=1, the height h_i is determined by the neighbouring heights through the admissibility and closure conditions, so U is not an open subset of R^n but rather a lower-dimensional subanalytic set. The subsequent application of the Lojasiewicz inequality on O=g(U) therefore needs a separate justification, which is not supplied. Since Proposition 7.5 is the key ingredient proving that the trajectory is fundamental in the irregular case, Theorem 7.3's convergence conclusion is not completely supported as written.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the gradient flow of the crystalline elastic energy F_alpha(Gamma)=∫_Gamma(1+alpha(kappa^phi)^2)phi^o(nu)dH^1, restricted to phi-admissible polygonal curves whose segments translate in the normal direction and for which no facet-breaking, edge creation, or segment bending occurs. The flow is reduced to the finite ODE system (3.2) for signed heights. The authors prove short-time existence and uniqueness (Theorem 4.1), show that at a finite maximal time only zero-curvature segments can vanish, and construct a unique globally defined flow with finitely many restarts (Theorem 4.2). Under a uniform lower bound on segment lengths they prove Kuratowski convergence to a stationary curve via a crystalline Lojasiewicz-Simon inequality (Theorem 7.1); dropping that bound, they prove convergence to a possibly degenerate generalized stationary curve (Theorem 7.3). The final section gives a complete classification of stationary curves and a partial classification of translating curves for the square anisotropy.","tokens_in":46984,"tokens_out":14080,"duration_ms":137608,"significance":"If the proofs are completed, this is a solid contribution to the crystalline analogue of the elastic flow. The paper is largely self-contained: it proves a crystalline Lojasiewicz-Simon-type inequality from scratch, carefully derives the first-variation formula and the energy-dissipation identity, and gives explicit examples including self-similar Wulff-shape evolution and translating solutions. The classification results for the square anisotropy are concrete and checkable. The main theorems are honestly stated under the explicit parallel-translation and no-facet-breaking assumptions, so the restriction to polygonal curves with fixed combinatorial type is a limitation of the model rather than a hidden circularity. The central reservation is the proof of the second Lojasiewicz-Simon inequality, Proposition 7.5, which is the key ingredient for the irregular-case convergence in Theorem 7.3.","major_comments":[{"comment":"The proof asserts that for any ĥ in U, an open neighbourhood of ĥ is obtained by applying Lemma 2.5 to the associated curve Γ̂. Lemma 2.5 requires every segment of Γ̂ to have positive length, because the bound in (2.4) uses min_i H1(S_i). In the setting of Proposition 7.5 the reference curve Γ is a generalized stationary curve and may contain degenerate segments with H1(S_i)=0; Theorem 7.3 explicitly allows exactly this situation. Moreover, for an index i with c_i=0 and |c_{i-1}|=|c_{i+1}|=1, the length formula (2.3) shows that H1(S_i) is independent of h_i and is determined by h_{i-1} and h_{i+1}; the requirement H1(S_i)≥0 is then a one-sided constraint, so the admissible heights form a set with boundary rather than an open subset of R^n. Consequently U is not proved to be open, and the subsequent application of the Lojasiewicz inequality to O=g(U) is not justified as written. The proof needs either a degenerate-segment version of Lemma 2.5 or a direct subanalytic Lojasiewicz argument on a possibly lower-dimensional or boundary-contained set U. This is load-bearing: Proposition 7.5 is the key input in Step 4 of the proof of Theorem 7.3.","section":"§7.2, Proposition 7.5, proof of openness of U"},{"comment":"The paper uses the notions of 'parallel', 'admissible', and 'signed height' for polygonal curves with zero-length segments, but Definition 2.3 explicitly requires all segments to be nondegenerate, and the admissibility and index definitions in Section 2 are stated for ordinary polygonal curves. Theorem 7.3 and Proposition 7.5 need a rigorous extension of these notions to generalized polygonal curves with H1(S_i)=0, including a definition of the associated straight line for a degenerate segment and a precise statement of what it means for a nondegenerate curve to be parallel to such a generalized curve. The sentence in Step 1 that Γ∞ is 'not necessarily parallel to Γ∞' also appears to be a typo, presumably for 'parallel to Γ0'. Without these definitions, the transitivity of parallelness used in Step 4 and the identities (7.14) and (7.17) for degenerate configurations are not fully verifiable.","section":"§7.2, Theorem 7.3 and Definitions 2.3–2.5"},{"comment":"In the proof of Theorem 7.3, the sets I_k are defined using only the indices i in J, while the indices outside J are controlled later through (7.25). The argument for i outside J requires that the neighbouring segments with nonzero transition numbers have lengths bounded below uniformly in k and in the interval [t_k,r]; this follows from (3.10), but the constants in (7.25) and in the subsequent estimate (7.26) should be shown to be independent of k and of the exit time r. As written the reader must assemble these uniformity claims from the preceding estimates; a short explicit statement would make the proof complete.","section":"§7.2, Step 4 of Theorem 7.3"}],"minor_comments":[{"comment":"There is a typo: 'vector vields' should be 'vector fields'.","section":"§2.5"},{"comment":"The displayed condition 'c_{i-1}=c_{i+1}+0' should read 'c_{i-1}=c_{i+1}=0'.","section":"§4.1, proof of Theorem 4.1(c), Case 2.1"},{"comment":"The sentence 'The proof runs along the same lines of Proposition 7.4' should refer to Proposition 6.3, not to Proposition 7.4, which is the proposition being proved.","section":"§7.2, Step 2"},{"comment":"The phrase 'Theferfore it has a unique fixed point' contains a typo; it should be 'Therefore'.","section":"§4.1, proof of Theorem 4.1"},{"comment":"It would help to state explicitly that Lemma 2.5 is applied in later sections only to curves all of whose segments have positive length, since the bound (2.4) is vacuous when min_i H1(S_i)=0. This would clarify the gap identified in the first major comment.","section":"§2.8, Lemma 2.5"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and appears to be a genuine contribution, but the proof of Proposition 7.5, which is load-bearing for Theorem 7.3, needs a substantial repair. I would be willing to review a revised version. No concerns about attribution or overlap beyond the normal citation of the authors' own prior work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a serious paper: Bellettini–Kholmatov–Novaga build the first rigorous theory of crystalline elastic flow for polygonal curves, with short-time existence, uniqueness, restart, global existence, and long-time convergence, plus a full classification of stationary solutions for the square anisotropy. The regular-case convergence (Theorem 7.1) is well argued, the Lojasiewicz–Simon inequality in that setting (Prop. 6.6) is proven carefully, and the classification in Section 8 is a nice piece of work. The paper is clearly written and self-contained; the assumptions about no facet-breaking are stated openly and are standard in this field.\n\nThe soft spot is the irregular case, Theorem 7.3. The proof leans on Proposition 7.5, a Lojasiewicz–Simon inequality for generalized stationary curves that may contain degenerate (zero-length) segments. The proof of Prop. 7.5 claims that the set U of admissible height vectors is open by \"applying Lemma 2.5\" to an associated curve eΓ. Lemma 2.5 requires every segment of eΓ to have positive length, but eΓ may have zero-length segments, so the lemma does not apply. Moreover, for indices i outside the set J (i.e., c_i=0 with |c_{i-1}|=|c_{i+1}|=1), the height h_i is forced by closure and admissibility, so U is not open in R^n but rather a lower-dimensional subanalytic set. The later use of the Lojasiewicz inequality on O=g(U) therefore needs a separate justification, which is not supplied. Since Prop. 7.5 is the key step showing the trajectory is fundamental in the irregular case, Theorem 7.3 as written is not fully proved.\n\nIs this fixable? Likely yes—subanalytic methods can handle non-open sets, and one can probably formulate U as a subanalytic set and prove a suitable Lojasiewicz inequality on it. But the current text does not do that, and the gap is load-bearing for the irregular-case conclusion. The regular-case results and the classification do not depend on it.\n\nWho is this for? Researchers in anisotropic geometric flows, crystalline curvature, and elastic flows of curves. It deserves a serious referee, but the editors should ask for a major revision, not accept as is. I would send it out for review, with a clear request to fix Prop. 7.5 or, failing that, to restrict Theorem 7.3 to situations where no degenerate segments occur.","headline":"Serious paper with a real gap in the irregular-case convergence proof (Prop. 7.5); the regular case and the classification are solid, but Theorem 7.3 needs repair.","tokens_in":47438,"tokens_out":5318,"would_cite":true,"duration_ms":45873,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53E10","53E40","46N20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The crystalline elastic flow of polygonal curves always converges to a stationary curve.","keywords":["crystalline curvature","elastic flow","polygonal curves","Lojasiewicz-Simon inequality","stationary solutions","translating solutions","anisotropic energy","square anisotropy"],"falsifier":"Numerically integrate (3.2) for a closed admissible polygon with square Wulff shape, choosing initial data for which a zero-curvature segment vanishes at a finite restart time; if the restarted heights do not converge to values satisfying (7.11) for all nondegenerate segments, for instance if they oscillate periodically instead of settling, the long-time convergence claim fails. Alternatively, an explicit initial polygonal curve that develops a new segment direction under the crystalline elastic evolution would falsify the admissibility restriction itself.","tokens_in":46537,"feed_emoji":"📐","tokens_out":3322,"duration_ms":34165,"temperature":0.7,"pith_summary":"The paper studies the gradient flow of a crystalline elastic energy on polygonal curves, an energy that adds a squared crystalline-curvature term to the anisotropic length. It establishes that, under the standard assumption that segments translate in parallel, a unique flow exists, that it can be restarted when zero-curvature segments vanish, and that for closed curves it converges as time goes to infinity to a stationary, or generalized stationary, polygonal curve. This matters because it extends long-time convergence results known for smooth elastic flows to the crystalline, nonsmooth setting, and because it gives a complete classification of stationary solutions for the square anisotropy.","feed_headline":"Crystalline elastic flow converges to stationary curves","feed_subtitle":"Unique global flow for polygonal curves, restarts after vanishing segments, and a full classification for square crystals.","key_machinery":"The carrier object is the signed-height ODE system (3.2): each segment $S_i$ of an admissible polygon translates along its normal with speed set by the first variation of $F_\\alpha$, so the fourth-order geometric evolution reduces to $n$ ordinary differential equations in the signed heights $h_i$. The load-bearing identities are (3.1), which turns the length and curvature contributions into the transition number $c_i$ times the facet length of the Wulff shape, and the crystalline Lojasiewicz-Simon inequalities (Propositions 6.6 and 7.5), which give quantitative energy decay near stationary curves and force full convergence rather than mere subconvergence. The restart mechanism at the maximal time removes only zero-curvature segments, preserves admissibility, and keeps the index of the curve unchanged.","core_discovery":"For a closed admissible polygonal curve, the crystalline elastic flow is globally defined through finitely many restarts and converges in the Kuratowski sense to a closed admissible polygonal curve that is a generalized stationary solution: its nondegenerate segments satisfy the stationarity system (7.11), while any degenerate segment has zero crystalline curvature. If all segment lengths stay bounded away from zero, the limit is a true stationary curve parallel to the initial one, obtained through a crystalline Lojasiewicz-Simon inequality. For the square Wulff shape the paper classifies all stationary solutions as staircases, right-angle chains, double-right-angle chains, and the square Wulff shape of sidelength $\\sqrt{4\\alpha}$, and it partially classifies translating solutions.","pith_inferences":["If facet-breaking or spontaneous creation of new segments occurred, the ODE formulation would miss it; the convergence statements therefore indicate stability inside the admissible class, not for arbitrary polygonal evolutions.","The same Lojasiewicz-Simon machinery may extend to other crystalline fourth-order flows, such as crystalline surface diffusion, but only under a comparable admissible-class restriction, since facet-breaking is known to occur there.","A testable extension suggested by the paper's own conjectures is that unbounded curves with nonparallel half-lines diverge to infinity, while parallel co-directed half-lines should converge to a translating solution after rescaling."],"forward_implications":["Every closed admissible polygonal curve has a unique global crystalline elastic evolution, and the index of the curve is preserved through all restarts.","If the initial curve is convex, the flow exists globally without restarts and converges to a Wulff shape of radius $\\sqrt{\\alpha}$, possibly covered multiple times according to the index.","In the square-anisotropy case, the only stationary curve with nonzero index is the square Wulff shape of sidelength $\\sqrt{4\\alpha}$; all other stationary curves have index zero.","Bounded polygonal curves cannot be translating solutions, while the one-segment and one-rectangle unbounded examples give explicit grim-reaper-type translating curves.","The flow preserves convexity, in contrast with the Euclidean elastic flow, under the admissibility and no-facet-breaking assumptions."],"supporting_citations":[{"why":"Supplies the Euclidean elastic-flow long-time convergence result and the Lojasiewicz-Simon strategy that the crystalline version imitates.","marker":"[29]"},{"why":"Provides the original formulation of crystalline curvature and the facet-translation evolution law behind the ODE system.","marker":"[32]"},{"why":"Supplies the signed-height and parallel-polygon machinery used to represent the flow as a finite system of ODEs.","marker":"[19]"},{"why":"Provides the semianalytic and subanalytic Lojasiewicz inequality used to prove Propositions 6.6 and 7.5.","marker":"[26]"},{"why":"Supplies the Kuratowski compactness and curve-convergence facts used to extract limits in the long-time arguments.","marker":"[16]"}],"fun_headline_variants":["Crystalline elastic flow: restarts lead to global convergence","Polygonal elastic flow restarts finitely then converges","Square anisotropy: all stationary solutions classified","Crystalline elastic flow: finite restarts, full convergence","Global convergence via finite restarts in crystalline elastic flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire development assumes segments only translate in parallel and that no new edges appear, no segment bends or breaks, and no facet-breaking occurs, so the fourth-order flow is exactly a finite ODE system; if that geometric restriction fails, the flow as defined does not apply.","fun_headline_variants_meta":{"raw":{"variants":["Crystalline elastic flow: restarts lead to global convergence","Polygonal elastic flow restarts finitely then converges","Square anisotropy: all stationary solutions classified","Crystalline elastic flow: finite restarts, full convergence","Global convergence via finite restarts in crystalline elastic flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000393,"raw_usage":{"total_tokens":2012,"prompt_tokens":839,"completion_tokens":1173,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":1099}},"tokens_in":455,"tokens_out":1173,"duration_ms":9022,"temperature":1.0,"reasoning_tokens":1099,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:29:35.753570+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically integrate (3.2) for a closed admissible polygon with square Wulff shape, choosing initial data for which a zero-curvature segment vanishes at a finite restart time; if the restarted heights do not converge to values satisfying (7.11) for all nondegenerate segments, for instance if they oscillate periodically instead of settling, the long-time convergence claim fails. Alternatively, an explicit initial polygonal curve that develops a new segment direction under the crystalline elastic evolution would falsify the admissibility restriction itself.","supporting_citations":[{"cited_title":"Mantegazza, M","cited_arxiv_id":null,"evidence_quote":"Supplies the Euclidean elastic-flow long-time convergence result and the Lojasiewicz-Simon strategy that the crystalline version imitates."},{"cited_title":"Taylor: Overview no","cited_arxiv_id":null,"evidence_quote":"Provides the original formulation of crystalline curvature and the facet-translation evolution law behind the ODE system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the signed-height and parallel-polygon machinery used to represent the flow as a finite system of ODEs."},{"cited_title":"Lojasiewicz: On semi-analytic and subanalytic geometry","cited_arxiv_id":null,"evidence_quote":"Provides the semianalytic and subanalytic Lojasiewicz inequality used to prove Propositions 6.6 and 7.5."},{"cited_title":"Falconer: The Geometry of Fractal Sets","cited_arxiv_id":null,"evidence_quote":"Supplies the Kuratowski compactness and curve-convergence facts used to extract limits in the long-time arguments."}],"review_version":1}