{"id":"596e6726-7db6-4766-8935-51f62f85ce69","arxiv_id":"2507.00581","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Closed polygonal curves, three-branch line unions, and smooth convex curves with transverse self-intersections do not admit tight frames of exponentials for their arc-length measures.","lead":"This mathematics paper proves that the arc-length measure on the boundary of any closed polygon, and on many smooth closed curves, cannot be used as a basis of complex exponential waves. It settles a natural open question about the boundary of a square and gives a general method for proving such spectrality failures.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 6 omits singular contributions from non-adjacent parallel edges; the neighborhood U must be shrunk to exclude shifted lines, and this step is missing but repairable.","rationale":"The central framework of Theorems 3 and 4 is coherent, and the restriction argument removing support on the W_i is sound once the projection-density step is read with a general frame bound A rather than the printed constant N. The genuinely fragile point is Lemma 6, exactly as the reader identified. I do not find a fatal flaw: the non-adjacent parallel edge contributions are finite in number and cannot accumulate at the origin unless their supporting line passes through 0, in which case it is one of the W_i and can be absorbed into the singular part. Thus Lemma 6 is repaired by choosing U sufficiently small. Other apparent slips, such as the omitted frame bound A in the restricted tiling equation and the use of Re F where |F| would be safer, are cosmetic and do not affect the main argument. The conditional verdict therefore stands, and no verdict change is needed.","tokens_in":19108,"tokens_out":29274,"duration_ms":331222,"concrete_test":"For a general closed polygon, compute d = min dist(0, supp(μ_i*μ̃_j)) over all pairs (i,j) for which v_j is parallel to v_i and the support line is not one of the W_k; verify that d > 0 by finiteness and compactness, then choose U inside both the Lemma 5 covering neighborhood and B(0,d) and recompute Lemma 6 on this U to confirm that no shifted-line singularity remains. As a sanity check, do this explicitly for the square: the opposite-side terms (1,3) and (3,1) have supports on y = ±1 and the terms (2,4) and (4,2) have supports on x = ±1, so any U ⊂ (−1,1)² avoids them entirely.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Lemma 6 the decomposition μ*μ̃ = F + ν_1 + ... + ν_M near 0 is established by considering self-convolutions μ_i*μ̃_i (singular, supported on W_i) and adjacent nonparallel pairs i,i+1 (absolutely continuous, supported on Q_{−v_i,−v_{i+1}}). Non-adjacent parallel edges, such as opposite sides of a rectangle, are not discussed. By Lemma 4(2), each such pair contributes a singular measure supported on a line segment parallel to v_i. That line is generally shifted: for the square, the pair (1,3) is supported on the segment [−1,1]×{−1}, which is not contained in any W_i and does not contain 0. The proof chooses U from Lemma 5 but never shrinks it to avoid these finitely many shifted lines, so as written the claimed representation of μ*μ̃ on all of U can fail if such lines meet U. This is a genuine gap in the proof of Theorem 1(1). It is repairable: for every non-adjacent parallel pair, the support either coincides with some W_i (then it can be absorbed into ν_i) or has positive distance from the origin; choosing U smaller than the minimum of those distances restores the lemma. The authors should state this shrinking argument explicitly.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies whether compactly supported probability measures supported on piecewise smooth curves can admit tight frames of exponentials. The main results are: (1) the arc-length measure on any finite union of line segments forming a closed curve is not tight-frame spectral; (2) a finite union of line segments containing three segments emanating from a common point in distinct directions is not tight-frame spectral, which covers the \"plus space\"; and (3) the arc-length measure on a smooth closed planar curve of positive curvature with finitely many transverse self-intersections is not tight-frame spectral. The proof strategy combines a new support theorem (Theorem 3: if μ*μ̃ has a smooth positive density on an open set U, then the distribution δΛ vanishes on U away from 0) with a zero-spectral-gap result for singular measures (Theorem 4, attributed to Lev--Olevskii). For piecewise linear measures the authors analyze μ*μ̃ near the origin and reduce to distributions supported on finitely many lines; for smooth curves they prove absolute continuity of μ*μ̃ in a punctured neighborhood of 0. The paper closes with several open questions on closed curves, line spectra, Riesz bases, and fractal measures.","tokens_in":19332,"tokens_out":35722,"duration_ms":380342,"significance":"The results are of clear interest to the Fourier-analytic geometry community: they settle the boundary-of-a-square question raised in Iosevich--Lai--Liu--Wyman, give an independent proof of non-spectrality of the plus-space complementing Lu's recent work, and extend non-spectrality to self-intersecting smooth curves. The framework of Theorems 3 and 4 is a clean general mechanism likely to be reused. The paper is honest about its reliance on prior results (Lemma 2, Theorem 5, and Theorem 4 from LO15), and no parameters are fitted. The main concern is a gap in the proof of Lemma 6 concerning non-adjacent pairs of edges; this gap is local and repairable rather than a fundamental flaw.","major_comments":[{"comment":"The proof of Lemma 6 does not, as written, establish the claimed decomposition μ*μ̃ = F + ν_1 + ... + ν_M in a fixed neighborhood U of 0. After treating the self-convolutions μ_i*μ̃_i and the adjacent non-parallel pairs (i,i+1), the proof moves directly to Lemma 5 and concludes the decomposition, but all non-adjacent pairs are omitted. By Lemma 4(2), a non-adjacent parallel pair (e.g., the two horizontal sides of a square) contributes a singular measure supported on a line parallel to v_i that is shifted away from the origin; for the square this support is the segment [-1,1]×{-1}, which is not contained in any W_i and does not contain 0. The proof chooses U from Lemma 5 but never shrinks it to avoid these finitely many shifted lines, so the representation with singular parts only on the W_i can fail on that U. Non-adjacent non-parallel pairs are also not accounted for: Lemma 4(1) gives absolutely continuous parts on shifted parallelograms x_{i-1}-x_{j-1}+Q_{v_i,-v_j}, not on the origin-centered Q_{-v_i,-v_j} used in Lemma 5, and when the corresponding edges intersect these parts can have support containing 0 with boundaries not among the half-lines L_i. This is load-bearing for Theorem 1(1) and for the Π-shape remark. The gap is repairable: for each non-adjacent pair, either its contribution has positive distance from 0, or (if the edges meet) the contribution is locally constant on a sufficiently small neighborhood of 0, with possible boundary only along a subspace W_i; one can then shrink U below all the positive distances and absorb the local absolutely continuous terms into F. The authors should state this shrinking argument explicitly. They should also note that the piecewise constant function F is strictly positive on U\\(∪_i W_i), since each sector between consecutive half-lines is covered by the corresponding adjacent parallelogram and Theorem 3 requires strict positivity.","section":"Section 4.2, Lemma 6 and proof of Theorem 1(1)"}],"minor_comments":[{"comment":"Proposition 3 is stated for arbitrary measures, but its proof splits the argument only into the singular and the absolutely continuous cases; a measure with both an absolutely continuous and a singular part is not treated. This can be fixed by applying the same argument to the restriction of μ to each part, since each restriction inherits the same tight frame on its L²-space.","section":"Section 2.2, Proposition 3"},{"comment":"The identity for the projected spectrum writes the right-hand side as N and states that Λ1 has density N, but for a general tight frame the constant should be AN (or a corresponding normalized value). Since only boundedness of Λ1 is needed, this is harmless, but the formulas should be corrected.","section":"Section 4.2, proof of Theorem 1(1)"},{"comment":"The definition of δ as the minimum of δ0 and all δ_{s0,t0} assumes there is at least one self-intersection; when γ is simple the minimum over an empty set should be handled explicitly, for example by setting δ = δ0.","section":"Section 5, Theorem 7"},{"comment":"The proof that Γ is injective is one sentence; equality of two chord vectors gives the same direction and length, but the argument that this forces the same pair (s,t) should be expanded or replaced by a reference to the strict convexity of the arc.","section":"Section 5, Lemma 8"},{"comment":"To conclude δΛ(φ̂) = Σ φ̂(λ) ≥ 0, the test function φ should be chosen with nonnegative Fourier transform, for example φ = ψ*ψ̃ with ψ compactly supported; nonnegativity of φ alone does not imply φ̂ ≥ 0.","section":"Section 3, proof of Theorem 4"},{"comment":"In the proof of Lemma 4(2), 'ma' should read 'm_a' in the sentence defining the measure on [0, as].","section":"Section 2.4, Lemma 4(2)"}],"recommendation":"major_revision","confidential_remarks":"The gap in Lemma 6 is concentrated and, in my view, fixable within the paper's framework; the results are likely correct. I would encourage the editor to send the paper back for a revision rather than reject. No concerns about attribution: Theorem 4 and Lemma 2 are properly cited."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main thing you should know: this answers the square-boundary question and more. It proves no closed polygonal line carries a tight frame of exponentials, so a fortiori no spectrum; the plus space is non-spectral by an independent route; and positively curved closed curves with transverse self-intersections are out too. That is genuinely new and is a real step past ILLW22, which only had frames for polygon boundaries and a non-frame result for the circle.\n\nWhat I like: the engine is clean. Theorems 3 and 4 reduce non-spectrality to showing that μ*μ̃ is locally smooth or constant away from a few lines, then using the zero spectral gap for singular measures. Theorem 7 for smooth curves is a nice use of the area formula and a partition of unity. No fitted parameters, no circularity; the citations to prior work check out.\n\nThe soft spots are real but repairable. The stress-test note about Lemma 6 is correct. The decomposition of μ*μ̃ near 0 accounts for self-convolutions and adjacent nonparallel pairs, but not for non-adjacent parallel edges. Opposite sides of a rectangle give singular measures on shifted lines that do not sit in any W_i and are not discussed. As written, the claimed representation can fail on the chosen U. It is fixable: for each such pair, either the support lies in some W_i, or it has positive distance from 0, so shrink U below that distance. The same sentence should also dispose of non-adjacent nonparallel absolutely continuous pieces, whose shifted supports may need to be avoided or absorbed. This is a missing detail in the main proof, not a broken strategy.\n\nMinor: in the projection claim in Theorem 1(1), the tight-frame bound A is dropped; the equation should have AN on the right, and the conclusion is that the projection has finite density, not density N specifically. In Theorem 4, the bump φ needs to be chosen so its Fourier transform is nonnegative; otherwise the inequality Σ φ̂(λ) ≥ φ̂(0) is not justified. Both are one-line fixes.\n\nBottom line: central theorems are almost certainly correct and worth a serious referee. Send it out, conditional on the Lemma 6 fix.","headline":"A solid, important paper on non-spectrality of curves; the main theorems look right, but Lemma 6 needs a shrinking argument that is missing.","tokens_in":19883,"tokens_out":9904,"would_cite":true,"duration_ms":107373,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42C15","42C30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that arc-length measure on any closed polygonal line, the plus space, and smooth closed curves of positive curvature with transverse self-intersections are never tight-frame spectral.","keywords":["spectral measures","piecewise smooth curves","tight frames of exponentials","polygonal lines","arc-length measure","tempered distributions","tiling equation","plus space"],"falsifier":"Compute $\\mu \\ast \\widetilde{\\mu}$ explicitly for a rectangle and check whether, in every neighborhood of the origin, a singular component supported on a line not through the origin appears; if it does, Lemma 6 fails. More decisively, find a closed polygonal line and a countable set $\\Lambda$ satisfying $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = A$ with $A>0$; that would refute Theorem 1(1).","tokens_in":18883,"feed_emoji":"📐","tokens_out":8660,"duration_ms":92516,"temperature":0.7,"pith_summary":"A measure is spectral when complex exponentials indexed by a countable set form an orthonormal basis of its $L^2$ space; a tight frame is a slightly looser system that still gives a perfect reconstruction formula. This paper proves that such exponential systems never exist for arc-length measure on any closed polygonal line, whether or not it crosses itself, and in particular the boundary of a square is non-spectral. The same method shows that two crossing line segments, the plus space, and smooth closed curves of positive curvature with finitely many transverse self-intersections are also non-spectral. The engine is a Fourier-side analysis of the tiling equation $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = A$: singular measures force the dual object to have zero spectral gap, while these curves force a positive one, a contradiction.","feed_headline":"No closed polygonal line can be spectral","feed_subtitle":"Arc-length measures on polygon boundaries, plus-shapes, and self-crossing smooth ovals fail to admit exponential tight frames.","key_machinery":"The load-bearing object is $\\mu \\ast \\widetilde{\\mu}$, the convolution of the measure with its reflection, together with the tempered distribution $\\widehat{\\delta_\\Lambda}$. The identity $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = A$ is the tight-frame analogue of the classical spectral tiling equation, and its Fourier transform transfers information about the support of $\\mu \\ast \\widetilde{\\mu}$ to the support of $\\widehat{\\delta_\\Lambda}$. For polygonal lines, Lemma 6 decomposes $\\mu \\ast \\widetilde{\\mu}$ near the origin into a piecewise constant function plus singular measures supported on the lines through the origin in the side directions; Lemma 3 shows that a bounded function whose Fourier transform is supported on a line cannot vary along that line, removing those singular components. For smooth curves, Theorem 7 uses a partition of unity and the implicit function theorem to show that $\\mu \\ast \\widetilde{\\mu}$ has a smooth density in a punctured neighborhood of the origin.","core_discovery":"The central discovery is that spectrality of a curve can be read off from the behavior near the origin of the self-convolution $\\mu \\ast \\widetilde{\\mu}$, where $\\widetilde{\\mu}$ is the reflected measure. For a tight-frame spectrum $\\Lambda$, the tiling equation $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = A$ holds, and taking Fourier transforms ties the support of $\\widehat{\\delta_\\Lambda}$ to the complement of the support of $\\mu \\ast \\widetilde{\\mu}$. The paper establishes that where $\\mu \\ast \\widetilde{\\mu}$ is smooth and strictly positive, $\\widehat{\\delta_\\Lambda}$ can meet that region only at the origin, while any singular measure forces $\\widehat{\\delta_\\Lambda}$ to have zero spectral gap. For closed polygonal lines and for the smooth curves considered here, the self-convolution covers a neighborhood of the origin with controlled singular parts, so a spectral gap is forced. This contradiction yields Theorems 1 and 2.","pith_inferences":["The paper leaves open whether a semicircle's arc-length measure is spectral; its self-convolution misses the origin, so the current method cannot decide it, but it is the cleanest test case for whether curve spectra must be line spectra.","If the authors' Conjecture 2 holds, closedness alone, not corners or curvature signs, is the obstruction to spectrality; the two theorems here are the polygonal and everywhere-positively-curved extreme cases of that conjecture.","Conditional on Conjecture 1, the same mechanism would imply that every singular measure whose support has positive Lebesgue measure is non-spectral, turning case-by-case classifications of such measures into instances of one general principle."],"forward_implications":["The boundary of every polygon, self-intersecting or not, has no tight frame of exponentials; in particular, the unit square boundary is non-spectral.","The plus space formed by two equal-length crossing segments is non-spectral, providing an independent proof of a recent result.","Arc-length measure on any smooth closed curve of positive curvature with finitely many transverse self-intersections is non-spectral, including curves that cross themselves.","By Corollary 1, any finite union of smooth curves that contains one of these configurations is also non-tight-frame-spectral.","The method gives a general criterion: a singular measure whose self-convolution is smooth and strictly positive around the origin cannot be tight-frame spectral."],"supporting_citations":[{"why":"Establishes the equivalence between spectrality and the tiling equation $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = 1$, the starting identity for tight frames.","marker":"[JP98]"},{"why":"Supplies the rigorous support theorem for the Fourier transform of $\\delta_\\Lambda$ under function tilings, used to justify the support reasoning.","marker":"[KL16, Theorem 4.1]"},{"why":"Provides the Beurling density and convolution inequalities that yield $D^+(\\Lambda)=0$ for singular tight-frame spectra.","marker":"[Gab13]"},{"why":"Gives the uniform-density result for absolutely continuous tight-frame spectral measures, the contrast behind the singular-measure argument.","marker":"[DL14]"},{"why":"Contributes the density-zero result for spectra of singular measures used in Proposition 2.","marker":"[HLL13]"},{"why":"Showed polygon boundaries admit Fourier frames and raised the boundary-of-the-square spectrality question answered here.","marker":"[ILLW22]"},{"why":"Introduced the symmetric additive line-segment measures, including the spectral L-shape and the plus-space question.","marker":"[LLP21]"},{"why":"Provided a prior proof of plus-space non-spectrality that Theorem 1(2) reproves independently.","marker":"[Lu25]"}],"fun_headline_variants":["Closed polygonal lines never spectral","Arc-length measures on polygons: not spectral","Spectrality fails for all closed polygonal curves","Polygon perimeters are never spectral","Plus-shapes and polygons: no spectrality"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 6's claim that near the origin the self-convolution of arc-length measure on a closed polygonal line splits into a piecewise constant part plus singular pieces supported only on the lines through the origin in the directions of the sides; if opposite parallel sides produced singular contributions on shifted parallel lines inside every neighborhood of the origin, the decomposition would fail.","fun_headline_variants_meta":{"raw":{"variants":["Closed polygonal lines never spectral","Arc-length measures on polygons: not spectral","Spectrality fails for all closed polygonal curves","Polygon perimeters are never spectral","Plus-shapes and polygons: no spectrality"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001697,"raw_usage":{"total_tokens":6683,"prompt_tokens":866,"completion_tokens":5817,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":482,"completion_tokens_details":{"reasoning_tokens":5751}},"tokens_in":482,"tokens_out":5817,"duration_ms":45258,"temperature":1.0,"reasoning_tokens":5751,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:15:58.387686+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $\\mu \\ast \\widetilde{\\mu}$ explicitly for a rectangle and check whether, in every neighborhood of the origin, a singular component supported on a line not through the origin appears; if it does, Lemma 6 fails. More decisively, find a closed polygonal line and a countable set $\\Lambda$ satisfying $\\delta_\\Lambda \\ast |\\hat\\mu|^2 = A$ with $A>0$; that would refute Theorem 1(1).","supporting_citations":[],"review_version":1}