{"id":"bd8a7a47-8767-411a-ba02-1f02e598cf82","arxiv_id":"2505.03047","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every p-width of a convex polygon is realized by a finite union of billiard trajectories, and the low p-widths of the equilateral triangle and square are computed exactly.","lead":"This paper proves that every p-width of a convex polygon in the plane is achieved by a finite union of billiard trajectories, and it computes the first four p-widths of an equilateral triangle and the first three of a square. It is the first explicit p-width computation for domains with corners, extending results previously known only for smooth domains and round surfaces.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing h0 factor: with W=(1+cos πt)/π^2, h0=8/π^2≠1, so the proof yields Σ length=h0ω_p, not ω_p; Theorem 1.3 does not follow as stated.","rationale":"The reader's conditional verdict identified several technical caveats, and the normalization constant h0 was mentioned among them. I agree with that inclusion but elevate it to the primary obstruction: it is not a clarification but a quantitative factor error in the central argument. In CM23-style Allen-Cahn min-max, the unnormalized energy measure converges to h0 times the geometric limit varifold. The paper fixes the sine-Gordon potential W(t)=(1+cos πt)/π^2, for which h0=8/π^2≠1, and never divides by h0 or rescales W. Proposition 2.6 then forces the limiting measure to have mass h0ω_p, while Proposition 3.4 and the final grouping identify that same measure with a sum of billiard lengths whose total is ω_p. The resulting mismatch is concrete: for the equilateral triangle, the paper's Theorem 1.6 gives ω1=3/2 and the claimed billiard length is 3/2, but the proof would attribute mass 12/π^2≈1.216 to the limiting energy measure. This affects not only Theorem 1.3 but also Proposition 4.2 and the p=3 lower-bound arguments in Sections 6 and 7 that rely on the billiard length set. The explicit values for ω1 and ω2 may still be correct from independent sweepouts, but the advertised billiard realization theorem is not established by the written proof. Because the flaw is a concrete internal inconsistency rather than a missing technical lemma, I recommend rejecting the current version; the needed repair, normalizing W so that h0=1 or revising the theorem's statement, changes the statement of the main result.","tokens_in":16053,"tokens_out":27152,"duration_ms":256985,"concrete_test":"Compute h0 for W(t)=(1+cos πt)/π^2 by substituting H'=√(2W): h0=∫_{-1}^{1}√(2W(s))ds=8/π^2. Then trace the total mass through Section 3: Proposition 2.6 gives lim ω_{p,ε}=h0ω_p, while Proposition 3.4 would need h0=1 to conclude Theorem 1.3. As a sharper check, repeat the p=1 equilateral triangle calculation: Theorem 1.6 says ω1=3/2, and the only billiard has length 3/2, whereas the proof's limiting energy measure has mass h0·3/2=12/π^2≈1.216; a 23% discrepancy confirms the missing normalization.","verdict_should_be":"REJECT","load_bearing_attack":"Proposition 2.6 fixes the comparison h0^{-1} lim_{ε→0} ω_{p,ε}(P) = ω_p(P). For the potential fixed in Section 3, W(t)=(1+cos πt)/π^2, the heteroclinic H satisfies |H'|^2/2 = W(H), so h0=∫(|H'|^2/2+W)dt=∫_{-1}^{1}√(2W(s))ds=8/π^2≈0.8106≠1. The measure µ in Section 3 is the weak limit of the unnormalized Allen–Cahn energy densities, hence ||µ||=lim ω_{p,ε}=h0ω_p(P). Proposition 3.4 asserts µ=Σ H^1⌊η_j, so Σ length(η_j)=||µ||=h0ω_p(P), not ω_p(P). The final grouping in Section 3 therefore proves at most ω_p(P)=h0^{-1}Σ length(η_j), and the stated equality in Theorem 1.3 does not follow. The discrepancy is quantitative: for the equilateral triangle, Theorem 1.6 gives ω1(T)=3/2 while the same argument predicts a limiting measure of mass h0·3/2=12/π^2≈1.216, not 3/2. This is an internal factor inconsistency in the central argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Gromov–Guth p-widths of compact convex polygons in R^2. The main theorem (Theorem 1.3) asserts that every p-width of such a polygon is the sum of lengths of finitely many billiard trajectories. The proof follows the Allen–Cahn route of Chodosh–Mantoulidis [CM23], combined with an even reflection argument across boundary edges to reduce boundary bounces to interior line segments on a doubled domain. The paper also proves that the first width of any convex polygon equals its geometric width (Theorem 1.6), and computes the low widths of the equilateral triangle (p=1,...,4) and the square (p=1,...,3), with explicit billiard realizations. The exposition is generally clear, but the central argument contains a missing normalization constant h0 that affects the claimed equality in Theorem 1.3, and several analytic steps at polygonal corners and under reflection are either asserted without proof or only sketched.","tokens_in":16284,"tokens_out":7606,"duration_ms":70446,"significance":"If the main theorem is correct, it is a substantial step: it gives the first general billiard-realization result for p-widths in domains with non-smooth boundary, complementing the closed-surface theorem of [CM23] and the smooth-domain conjecture. The explicit computations for low p-widths of the equilateral triangle and square are concrete and genuinely useful, and the geometric-width result for the first width is clean and self-contained. The paper is also valuable for pointing out exactly why a boundary version of Allen–Cahn curvature estimates is not needed, via reflection. However, the central proof currently has a quantitative gap involving the heteroclinic energy h0, so the main theorem is not established as stated.","major_comments":[{"comment":"The proof of Theorem 1.3 misses the h0 normalization. Proposition 2.6 states that lim_{\\epsilon\\to0} \\omega_{p,\\epsilon}(P) = h0 \\omega_p(P), where h0 is the heteroclinic energy. In Section 3 the potential is fixed to be W(t)=(1+\\cos \\pi t)/\\pi^2; for this potential h0 = \\int_{-1}^{1} \\sqrt{2W(s)}\\,ds = 8/\\pi^2 \\neq 1. The measure \\mu is defined as the weak limit of the unnormalized energy densities, so its total mass is \\lim \\omega_{p,\\epsilon}(P) = h0 \\omega_p(P). Proposition 3.4 concludes \\mu = \\sum_j H^1\\lfloor \\eta_j, and the final grouping then gives \\sum_j \\operatorname{length}(\\eta_j) = h0 \\omega_p(P), not \\omega_p(P). Thus Theorem 1.3 does not follow as stated. The discrepancy is quantitative: for the equilateral triangle, Theorem 1.8 claims \\omega_1(T)=3/2, while the argument as written would produce a limiting measure of mass h0\\cdot 3/2 = 12/\\pi^2 \\approx 1.216. This must be repaired, for example by normalizing the Allen–Cahn widths by h0^{-1} from the outset or by choosing a potential with h0=1.","section":"Section 3, after Proposition 3.4"},{"comment":"The monotonicity formula at polygonal vertices is load-bearing but is asserted rather than proved. The lemma claims \\mu_\\epsilon(B_r(p)) \\le Cr for p \\in V with no boundary correction, citing [HT00] and [LPS24]. However, [LPS24] concerns smooth boundaries, and at a corner the Neumann condition does not give a smooth reflected problem; the statement 'no boundary correction is necessary' needs a detailed proof or a precise reference covering polygonal corners. This step is essential because it is used to conclude \\mu(V)=0 and hence to discard vertex contributions when grouping segments into billiard trajectories.","section":"Section 3, Lemma 3.3"},{"comment":"The proof that the even reflection \\tilde u_{\\epsilon_j} has index at most 2p on B is too compressed to be verifiable as written. The argument invokes 'standard results' and compares Dirichlet and Neumann eigenvalue counts, but it does not spell out how the spectrum of the reflected operator on B is identified with the Dirichlet and Neumann spectra on B^+, nor how the L^2-orthogonality argument controls the full quadratic form. Since this index bound is what permits the application of [CM23, Theorem 1.2] to the reflected problem, a complete proof or a precise reference is needed.","section":"Section 3, reflected index bound"},{"comment":"Proposition 2.6 states that the comparison h0^{-1} \\lim_{\\epsilon\\to0} \\omega_{p,\\epsilon}(K) = \\omega_p(K) remains valid for polygonal regions, with only the remark that 'a careful examination of the proof shows' this. This is not a trivial extension, because the Allen–Cahn min-max theory near corners and the boundary condition require additional arguments; the rest of the paper depends on this comparison. The authors should either provide the extension in detail or cite a source that proves the comparison for Lipschitz polygonal domains with Neumann boundary conditions.","section":"Section 2.3, Proposition 2.6"}],"minor_comments":[{"comment":"In the displayed formula for the conjecture, the summand is written as \\operatorname{length}(\\gamma_{p,N(p)}) but should be \\operatorname{length}(\\gamma_{p,j}).","section":"Section 1.1, Conjecture 1"},{"comment":"The displayed definition of h0 appears to have a typo: it reads H'(t)^2 + W'(H(t)); the standard heteroclinic energy is \\int (\\tfrac12|H'|^2 + W(H))\\,dt, and as written W'(H) integrates to zero by the traveling-wave equation. This should be corrected for clarity.","section":"Section 2.3, definition of h0"},{"comment":"The definition of a T-billiard trajectory switches notation between \\ell_k and \\eta_k in conditions (1\") and (2\"), which makes the definition harder to follow. Please use one symbol consistently.","section":"Section 4, T-billiard definition"},{"comment":"The statement of Proposition 4.2 writes \\omega_p(P) but the theorem is about the equilateral triangle T; the variable P here is either a typo or needs a line defining P=T.","section":"Section 4, Proposition 4.2"},{"comment":"There is a minor typo: 'a the 2-sweepout' should be 'a 2-sweepout'.","section":"Section 7.1"}],"recommendation":"major_revision","confidential_remarks":"The missing h0 factor is a genuine and easily quantifiable gap in the main theorem, but it is fixable within the manuscript's scope (normalize the Allen–Cahn widths or restate with h0). The paper has several useful results and clear probabilistic/geometric constructions, so I recommend major revision rather than rejection. The monotonicity and index-bound issues, while important, are secondary to the h0 factor and may also be repairable with added detail."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has a genuinely nice idea: prove that p-widths of convex polygons are realized by billiard trajectories via an Allen–Cahn reflection argument, and compute some low widths explicitly. The reflection trick is cute, the geometric width theorem for ω1 is simple and convincing, and the explicit sweepouts for the triangle and square give good upper bounds. But the central proof has a real normalization problem. In Section 3 the energy density is defined as the unnormalized Allen–Cahn measure, so its limit μ has mass lim ω_{p,ε} = h0 ω_p, with h0 = 8/π^2 for the sine–Gordon potential W(t)=(1+cos πt)/π^2. Proposition 3.4 then claims μ is the sum of Hausdorff measures along line segments, but the standard convergence gives μ = h0 times that sum. The paper never divides by h0, so the argument as written yields Σ length = h0 ω_p(P), not ω_p(P). The final grouping in Section 3 therefore does not prove Theorem 1.3 as stated. This is not a minor typo; it is a factor that changes the numerical content of the main theorem. The consequence propagates: the lower-bound argument for ω3(T) uses Proposition 4.2, which relies on the same flawed normalization, so that computation is not fully supported either, although the independent upper bounds may survive. The corner regularity and the index bound for the reflected solutions are also compressed, but those look fixable and are less serious. The low-width results for ω1, ω2, ω4 of the triangle and the square do not depend on the flawed step and appear solid. The paper should not be accepted as is. The fix is likely straightforward—insert the h0 factor and adjust Proposition 3.4 and the concluding argument—so the paper is worth a serious referee. I would send it to peer review with a clear request to address the normalization before publication.","headline":"The reflection argument is clever and the explicit widths are nice, but the proof of the main billiard theorem is off by the missing h0 normalization factor.","tokens_in":16870,"tokens_out":7155,"would_cite":false,"duration_ms":60945,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C42","58E12","49Q20","35J60"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every p-width of a convex polygon is achieved by billiard trajectories.","keywords":["p-widths","billiard trajectories","convex polygons","Allen-Cahn equation","min-max theory","phase transition","equilateral triangle","square"],"falsifier":"Run a numerical phase-field min-max simulation of the Allen-Cahn energy on a generic convex pentagon and inspect the limiting interface: if for some small $\\varepsilon$ the minimizing sweepout converges to a configuration with an interior triple junction or a corner-touching segment that does not reflect specularly, the billiard decomposition would be false.","tokens_in":15780,"feed_emoji":"📐","tokens_out":5520,"duration_ms":47269,"temperature":0.7,"pith_summary":"This paper proves that the p-widths of a convex polygon—a sequence of geometric invariants analogous to the Neumann Laplacian spectrum—are always realized by finite unions of billiard trajectories. For each p, the minimax length can be written exactly as the sum of lengths of finitely many straight segments that reflect off the polygon boundary. The proof routes the problem through Allen-Cahn phase transitions, reflecting critical points across the boundary to reduce the boundary case to an interior theorem. As concrete evidence, the paper computes the first four widths of the equilateral triangle inscribed in the unit circle and the first three widths of the square, matching explicit billiard configurations.","feed_headline":"Polygon p-widths are unions of billiard trajectories","feed_subtitle":"Exact widths for the equilateral triangle and square follow from a reflection argument for Allen-Cahn phase transitions.","key_machinery":"The main mechanism is Allen-Cahn phase-transition min-max: p-widths are recovered as $\\varepsilon\\to 0$ limits of critical points of the double-well energy $E_\\varepsilon[u]=\\int(\\frac{\\varepsilon}{2}|\\nabla u|^2+\\frac{1}{\\varepsilon}W(u))$, with Neumann boundary conditions. The load-bearing step is an even reflection across a boundary edge: reflecting a Neumann critical point produces a solution on the doubled domain whose index is at most $2p$, so the closed-surface billiard theorem applies and the limiting energy measure is a union of straight line segments. A monotonicity estimate at vertices, valid without boundary correction because polygon edges are straight and the gradient vanishes at convex corners, controls the mass near vertices. For the equilateral triangle, repeated reflections generate the full tessellation and force the billiards to unfold to straight lines; this yields the length set $\\{\\frac{3}{2}\\sqrt{a^2+ab+b^2}:a,b\\in\\mathbb{Z}\\}$ used in the explicit computations.","core_discovery":"For $P$ a compact convex polygon in $\\mathbb{R}^2$ and any $p=1,2,\\ldots$, there exist finitely many billiard trajectories $\\gamma_{p,1},\\ldots,\\gamma_{p,N(p)}$ in $P$, repetitions allowed, with total length equal to $\\omega_p(P)$. The trajectories are allowed to terminate either orthogonally at the boundary or at a vertex, and otherwise follow specular reflection. The first width of any convex polygon equals its geometric width $W(P)$, the width of the narrowest slab containing it. For the equilateral triangle inscribed in the unit circle the paper obtains $\\omega_1=\\omega_2=3/2$, $\\omega_3=3\\sqrt{3}/2$, and $\\omega_4=3$; for the square of side $\\sqrt{2}$ it obtains $\\omega_1=\\sqrt{2}$, $\\omega_2=2$, and $\\omega_3=2\\sqrt{2}$. In the triangle these values are realized respectively by a perpendicular bisector, the medial triangle, and two perpendicular bisectors; in the square by horizontal or vertical segments and a diagonal.","pith_inferences":["One can expect an algorithm: for a rational convex polygon, the finite graph of possible billiard trajectories with bounded length should make each p-width computable by solving a shortest-path or integer-optimization problem over reflection-unfolded straight lines; the paper's length-set lemma for the equilateral triangle is a prototype.","If the reflection-index bound survives perturbations, the same proof should give billiard decompositions for p-widths in higher-dimensional convex polytopes with totally geodesic faces, where the limiting objects would be unions of reflecting geodesic segments rather than planar billiards.","The vertex condition that billiards stop at corners suggests a spectral analogue: p-widths may encode the 'billiard graph' of a polygon, with corner hits acting as Dirichlet-like breaks; comparing $\\omega_p$ for polygons that differ only by rounding corners could test how much of the p-width is carried by vertex-terminating trajectories."],"forward_implications":["For every convex polygon and every p, the variational minimax length can be certified by an explicit finite collection of reflecting straight segments.","The first p-width of a convex polygon is always the width of its narrowest enclosing slab, so one can read $\\omega_1$ off from the polygon's geometry alone.","In an equilateral triangle, a single side or two sides cannot furnish a p-width; only full boundary-type billiards or configurations allowed by unfolding can.","The method extends, as the paper notes, to Riemannian surfaces whose boundary is piecewise totally geodesic with convex corners, replacing line segments by geodesic segments and allowing closed geodesics.","The computed values give nontrivial exact p-widths for polygons, providing benchmarks for the Weyl asymptotic and Lusternik-Schnirelman inequalities."],"supporting_citations":[{"why":"Supplies the closed-surface theorem that limits of reflected Allen-Cahn solutions are unions of straight line segments, the template for Theorem 1.3.","marker":"[CM23]"},{"why":"Provides convergence of phase interfaces and the discrepancy-function maximum principle used to control the limiting energy measure.","marker":"[HT00]"},{"why":"Gives the result that the first width of a smooth convex region is achieved by a free-boundary segment meeting the boundary orthogonally, used for $\\omega_1(P)=W(P)$.","marker":"[DM24]"},{"why":"Provides the boundary monotonicity formula context for Allen-Cahn with Neumann conditions, adapted here for polygonal vertices.","marker":"[LPS24]"},{"why":"Supplies the comparison between Almgren-Pitts and Allen-Cahn min-max widths used to identify the $\\varepsilon\\to 0$ limit with the p-width.","marker":"[Dey22]"},{"why":"Gives the finiteness of billiard bounces into a convex corner, used to group line segments into billiard trajectories.","marker":"[Lan23]"},{"why":"Provides the gradient-bound maximum principle behind the discrepancy estimate.","marker":"[Mod85]"},{"why":"Supplies integrability results for the sine-Gordon potential used in the limit analysis.","marker":"[LW22]"}],"fun_headline_variants":["Billiard trajectories realize polygon p-widths","Exact p-widths for triangle and square via billiards","Polygon p-widths as unions of billiard paths","p-widths of polygons come from billiard trajectories"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the even reflection of Allen-Cahn critical points having index at most $2p$ and on a vertex monotonicity estimate with no boundary correction; if either fails, the limiting interfaces need not be line segments.","fun_headline_variants_meta":{"raw":{"variants":["Billiard trajectories realize polygon p-widths","Exact p-widths for triangle and square via billiards","Polygon p-widths as unions of billiard paths","p-widths of polygons come from billiard trajectories"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00018,"raw_usage":{"total_tokens":1245,"prompt_tokens":826,"completion_tokens":419,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":442,"completion_tokens_details":{"reasoning_tokens":353}},"tokens_in":442,"tokens_out":419,"duration_ms":3456,"temperature":1.0,"reasoning_tokens":353,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:06:24.260471+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a numerical phase-field min-max simulation of the Allen-Cahn energy on a generic convex pentagon and inspect the limiting interface: if for some small $\\varepsilon$ the minimizing sweepout converges to a configuration with an interior triple junction or a corner-touching segment that does not reflect specularly, the billiard decomposition would be false.","supporting_citations":[],"review_version":1}