{"id":"beeda708-530b-482a-bc5d-817143fe2fd0","arxiv_id":"2411.13214","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For coin billiards, invariant curves exist for small heights or near-circular coins, vanish near the boundary for tall non-circular coins, and a complete foliation by invariant curves occurs only for circular coins.","lead":"This paper studies coin billiards, where a particle bounces between two identical curved tables glued to the top and bottom of a cylinder. It proves that such systems have stable regular orbits in some settings, but that tall non-circular coins destroy all regular orbits near the edge.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 11's Step 1 does not prove D_m≠∅: the 'flat graphs' and 'circular caustics' inference from D_m=∅ is a non sequitur; a direct repair via (2.1) is needed before the foliation result is established.","rationale":"The reader already flagged the flat-graphs argument in Theorem 11 as a gap and assigned a CONDITIONAL verdict. My stress-test agrees that this is the most load-bearing unresolved step in the central foliation claim, because Theorem C's proof as written relies on an inference that is not merely unproven but false as stated: D_m = ∅ does not make g_m constant or produce classical circular caustics. The direct expansion (2.1) supplies a short repair, so the concern is likely surmountable, but it must be written into the proof. I also note a concrete sign error in the first proof of Theorem 9 (the inequality preceding the implicit-function step), which is independently fixable and does not affect the second proof, so it does not change the overall conditional assessment. The C^5 regularity issue raised by the reader remains relevant: the direct repair and Lemma 10 both require uniform control of the O(θ^3) remainder and its derivatives, which the paper asserts but does not rigorously derive. My concrete test isolates the uniformity needed for the Theorem 11 repair; if that uniformity fails, the paper's strongest nonexistence conclusion is not supported. Since neither concern overturns the existence results (Theorem A, KAM curves) and the second proof of Theorem B appears sound modulo standard technical assumptions, the appropriate verdict remains CONDITIONAL as the reader stated; no verdict adjustment is needed.","tokens_in":24835,"tokens_out":36221,"duration_ms":332996,"concrete_test":"Replace the flat-graphs paragraph in Step 1 of Theorem 11 with the following analytic check: suppose D_m = ∅ for some m and write Π_θ T(φ,g_m(φ)) = g_m(φ). Use (2.1) to derive sup_φ |ρ'(φ)| ≤ C g_m(φ) for a constant C independent of φ and m, with g_m(φ) ≤ C'/m. Since sup|ρ'| > 0 for a noncircular Γ, this inequality must fail for all m > 2C'/sup|ρ'|, proving D_m ≠ ∅ for all sufficiently large m. The test is to verify that the O(g_m^3) remainder in (2.1) is uniform in φ under the C^5 hypothesis; if it is, the repair succeeds, and if not, Theorem C is unproven.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central nonexistence theorem for foliations (Theorem 11) depends on Step 1 showing that D_m ≠ ∅ for every sufficiently large m, where D_m = {(φ,θ) ∈ G_m : Π_θ T(φ,θ) ≠ θ}. The paper argues that if D_m = ∅ then 'each graph G_m would be flat' and the classical billiard would have circular caustics approaching ∂Γ, forcing Γ to be a disc. This is not justified: Π_θ T = θ on G_m only gives Π_θ T_1(φ,g_m(φ)) = g_m(φ); it does not imply g_m is constant, and T_1 need not send G_m to itself because its φ-coordinate changes. Thus the claimed sequence of classical circular caustics does not follow. The claim is nevertheless repairable directly from expansion (2.1): if D_m = ∅, then g_m(φ) − (2/3)ρ'(φ)g_m(φ)^2 + O(g_m(φ)^3) = g_m(φ), so |ρ'(φ)| ≤ C g_m(φ) for all φ. Since g_m(φ) → 0 uniformly as m → ∞, this forces ρ' ≡ 0 and hence Γ to be a disc. But this argument is absent from the manuscript; as written, Theorem C depends on a false/non-sequitur step. A second, independent issue is the sign error in the first proof of Theorem 9: the text asserts ℓ > −3/ρ'' implies ∂H/∂x_1 < 0, whereas substituting (14) gives ∂H/∂x_1 / ε^3 = −2/ℓ − (2/3)ρ'' > 0 under that inequality; the subsequent monotonicity of ν depends on correcting this sign.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the coin billiard map T = T2 ∘ T1 on the annulus A = T × (0, π), where T1 is a classical billiard map and T2 is the shear (φ, θ) ↦ (φ + ℓ cot θ, θ). Three main results are claimed: (A) KAM invariant curves exist in a strip near, but not accumulating on, the boundary when the coin height ℓ is small or the table is near-circular; (B) for any non-circular C^5 strictly convex table and any sufficiently large ℓ, no essential invariant curve passes through a neighbourhood of the boundary; (C) if the coin map admits a full foliation by essential invariant curves for some ℓ > 0, then the table is a disc. Theorems A and B are proved through expansions near the boundary, Moser's KAM theorem, Mather's variational method, and Herman-type Lipschitz estimates. Theorem C is proved via vertically-mapped graphs, Herman bounds, and a trichotomy argument, with a quantitative lower bound on the measure of the obstruction regions. Numerical phase portraits for elliptical coins illustrate the transition from KAM curves to chaotic behaviour.","tokens_in":25203,"tokens_out":13324,"duration_ms":133769,"significance":"If the main theorems are correct, the paper gives substantial and partly unexpected answers to Bialy's questions: in contrast with Lazutkin's classical result, tall non-circular coin billiards have no boundary KAM curves, and the circular coin is the only one whose phase space is completely foliated by essential invariant curves. The explicit formula ℓ0 = -3 / min ρ'' and the quantitative measure estimate in Theorem 11 are valuable. The paper also gives two independent proofs of Theorem B and credibly connects the results to nonintegrability, horseshoes, and positive entropy. A particular strength is that the arguments are essentially parameter-free: no constants are fitted to the conclusions, and the proofs rest on standard external results (Moser, Herman, Mather, Lazutkin). The numerical section is illustrative but not a substitute for the analytic proofs.","major_comments":[{"comment":"The assertion that D_m = ∅ would make each graph G_m flat and give circular caustics for the classical billiard is a non-sequitur. From D_m = ∅ one obtains only Π_θ T(φ, g_m(φ)) = g_m(φ), i.e. θ-invariance along G_m; it does not follow that g_m is constant, and T_1 need not map G_m to itself because its φ-coordinate changes. The conclusion can be recovered directly from expansion (2.2): if Π_θ T(φ, g_m(φ)) = g_m(φ), then (2/3)ρ'(φ)g_m(φ)^2 + O(g_m(φ)^3) = 0, so |ρ'(φ)| ≤ C g_m(φ) uniformly; since g_m → 0 uniformly as m → ∞, this forces ρ' ≡ 0 and hence Γ to be a disc. This repair is absent from the manuscript. Step 3 later contains a valid proof of the needed nonemptiness for φ⋆ in J_Γ, so the theorem is likely salvageable by reorganizing the proof, but as written the proof of point 1 depends on an unjustified step.","section":"Section 7, Theorem 11, Step 1"},{"comment":"The sign in the displayed consequence is wrong. For x1 with ρ''(x1) = min ρ'' < 0, the condition ℓ > -3/ρ''(x1) implies ∂H/∂x1 / ε^3 = -2/ℓ - (2/3)ρ''(x1) > 0, not < 0. The subsequent implicit-function conclusion requires ∂H/∂x1 > 0 in order for ν to be decreasing in both arguments, so the proof is salvageable by replacing '< 0' with '> 0' and adjusting the surrounding sentence. As written, this step is false. The second proof of Theorem 9 is independent of this error.","section":"Section 6, first proof of Theorem 9, after Eq. (14)"},{"comment":"The estimate |g'_m(φ)| ≤ (8/3)ρ''(φ)g_m(φ)^2 is not valid as stated for non-circular Γ, because ρ'' takes negative values and the right-hand side is then negative while the left-hand side is nonnegative. The proof of the estimate only controls the numerator by a constant multiple of max|ρ''|, so the correct form is an absolute-value bound such as |g'_m(φ)| ≤ C max_x |ρ''(x)| g_m(φ)^2. The use in Corollary 8 already passes to max|ρ''|, so this is a local fix, but the lemma statement and proof should be corrected.","section":"Section 5, Lemma 4(3)"}],"minor_comments":[{"comment":"Several cross-references to displayed formulas are inconsistent or ambiguous; for example, the proof of Lemma 4 refers to 'expression (5)' before the displayed derivative formula, and Section 6 repeatedly refers to '(6)' without a matching label. Please renumber and check all internal references.","section":"Throughout"},{"comment":"In Theorems 2 and 3 the positive-measure conclusion is stated for the whole displayed strip, but Moser's theorem applies to a subinterval strictly inside the coordinate range; the wording should clarify which subset of the strip carries the positive-measure family.","section":"Section 4"},{"comment":"The numerical section is descriptive; a short statement of the integration method, the number of iterates, and the resolution of the phase portraits would improve reproducibility.","section":"Section 8"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of math.DS and I found no concerns about citation or overlap. The main theorems are likely correct, but the proof of Theorem C contains a genuine gap in Step 1 and the first proof of Theorem 9 contains a sign error. Both are repairable locally, and Step 3 of Theorem 11 already contains the germ of a correct proof of the needed nonemptiness, so I recommend a major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is worth a serious referee. The coin billiard is new and the results are real: Theorem B shows that for any noncircular C^5 table and sufficiently large height, all invariant essential curves near the boundary disappear, in sharp contrast with Lazutkin's theorem for classical billiards. Theorem C is a strong foliation rigidity statement. The authors use Moser, Mather, and Herman appropriately, and the second proof of Theorem B appears independent and sound.\n\nThe soft spots are real but not fatal. The gap in Theorem 11, Step 1, is the one a referee must insist on fixing. The text claims that if D_m is empty, then each graph G_m is flat and the classical billiard would have circular caustics approaching the boundary. That inference is a non sequitur: having the coin map preserve θ on G_m only means the classical billiard map preserves θ on those points; it does not force g_m to be constant. The good news is that the intended conclusion follows directly from the near-boundary expansion (2.1): on G_m, θ − (2/3)ρ'(φ)θ² + O(θ³) = θ, so |ρ'(φ)| ≤ Cθ uniformly as m→∞, forcing ρ' ≡ 0 and hence a disc. This repair is short and should be inserted.\n\nThe other issue is a sign typo in the first proof of Theorem 9. The text asserts ∂H/∂x1 < 0 under ℓ > −3/ρ'', but substituting the displayed derivatives gives ∂H/∂x1 = ε³(−2/ℓ − (2/3)ρ'') > 0. The subsequent formulas for ∂ν/∂x0 and ∂ν/∂x2 are then still negative, so the orientation-reversing conclusion survives if you flip the claimed sign. This is a minor, easily corrected slip, not a breakdown. I also note the numerical experiments are illustrative but not reproducible from the submitted artifacts.\n\nBottom line: the central claims are likely correct, the proof strategy is sound, and the paper answers real questions. With the Step 1 repair and the sign cleanup, it would be a solid contribution to billiard and twist-map dynamics. It deserves peer review, not rejection.","headline":"New and believable results on coin billiards, but Theorem C's Step 1 has a logical gap that needs a repair before the foliation result is fully established.","tokens_in":25749,"tokens_out":5605,"would_cite":true,"duration_ms":48918,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37E40","37J40","37D50"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for noncircular coin billiards, sufficiently tall coins destroy all invariant curves near the boundary, and that the only coin billiard whose phase space is completely foliated by essential invariant curves is the…","keywords":["coin billiard","essential invariant curves","KAM theory","twist maps","billiard caustics","nonintegrability","topological entropy","Lipschitz bounds for invariant curves"],"falsifier":"Fix a strictly convex noncircular $C^5$ table (for example, an ellipse with known curvature $\\rho$), choose $\\ell > -3/\\min \\rho''$, and numerically integrate a fine grid of initial conditions in the strip $\\theta < \\delta$ for decreasing $\\delta$. If an essential invariant curve is found at arbitrarily small $\\delta$, Theorem 9 is false; equivalently, compute the Lipschitz constant of any candidate curve and check whether it violates the bound the paper derives.","tokens_in":24615,"feed_emoji":"🪙","tokens_out":9707,"duration_ms":89090,"temperature":0.7,"pith_summary":"Coin billiards are a one-parameter modification of the classical billiard: a particle bounces inside a strictly convex plane domain, then travels along a cylinder of height $\\ell$ before bouncing again, which adds a shift $\\ell\\cot\\theta$ to the collision point. This paper asks whether the resulting annulus map has invariant curves, the curves that would separate the phase space and block ergodicity. The answer has three parts. In perturbative regimes (small height, or tables close to a circle) there is a positive-measure family of invariant KAM curves near, but not accumulating on, the boundary. For any noncircular sufficiently smooth table and any height above a threshold fixed by the curvature of the boundary, all essential invariant curves disappear from a neighborhood of the boundary, in contrast to the classical billiard, where a classical theorem guarantees boundary-accumulating invariant curves. Finally, the only coin billiard whose whole phase space is foliated by essential invariant curves is the circular one. If correct, these results settle the main structural questions for the coin billiard in the large-height and foliation cases, and imply nonintegrability and positive entropy for noncircular coins that are sufficiently tall.","feed_headline":"Tall coins destroy boundary invariant curves","feed_subtitle":"For heights above a curvature threshold, noncircular coin billiards lose every invariant curve near the boundary.","key_machinery":"The load-bearing object is the coin map $T = T_2 \\circ T_1$, with the shift $T_2(\\varphi,\\theta)=(\\varphi+\\ell\\cot\\theta,\\theta)$: the unbounded shift $\\ell\\cot\\theta$ gives the map a strong twist near the boundary. Composing the standard near-boundary expansion of the billiard map with this shift yields $\\bar\\theta = \\theta - \\frac{2}{3}\\rho'(\\varphi)\\theta^2 + O(\\theta^3)$ and $\\bar\\varphi = \\varphi + \\ell/\\theta + O(\\theta)$, where $\\rho$ is the radius of curvature. The arguments then use two devices: a sequence of vertically-mapped graphs $G_m$, graphs on which the map advances the arclength coordinate by exactly $2m\\pi$, which accumulate on the boundary and force any invariant curve to lie in narrow bands; and a Lipschitz bound saying that an invariant curve near the boundary must be nearly horizontal. The contradiction in Theorem 9 is that strong twist plus curvature variation forces the map to reverse orientation on any candidate curve, while the generating-function or Lipschitz estimates rule out the required geometry.","core_discovery":"The central claim is that the coin billiard map $T = T_2 \\circ T_1$, where $T_1$ is the classical billiard map and $T_2(\\varphi,\\theta)=(\\varphi+\\ell\\cot\\theta,\\theta)$ shifts the arclength coordinate, behaves very differently from the classical billiard near the boundary. For any noncircular strictly convex $C^5$ table $\\Gamma$, there is a critical height $\\ell_0 = -3/\\min_{\\varphi}\\rho''(\\varphi)$, where $\\rho$ is the radius of curvature, such that for every $\\ell > \\ell_0$ no essential invariant curve passes through a neighborhood of the boundary of the annulus. Moreover, for every $\\ell > 0$, if the annulus is foliated by essential invariant curves then $\\Gamma$ must be a disc. The same machinery yields KAM curves in perturbative regimes (small height or near-circular tables) and, for $\\ell > \\ell_0$, drifting orbits, positive topological entropy, and nonintegrability.","pith_inferences":["The threshold $\\ell_0 = -3/\\min \\rho''$ depends only on the second derivative of the radius of curvature at its minimum; for explicit tables such as ellipses this gives a computable prediction that could be tested by direct numerical searches for boundary KAM curves.","The mechanism, unbounded twist plus a Lipschitz bound, does not use the specific form of the billiard map beyond its near-boundary expansion, so it should extend to other pensive billiards whose delay shift also diverges at the boundary.","The trichotomy in Theorem 11 suggests a route to small-height nonintegrability: an integrable noncircular coin would have to exhibit island chains with integer rotation number accumulating on the boundary, and generic splitting of their separatrices would complete the argument.","The near-boundary expansion shows that the $\\theta$-dynamics of the coin map agrees with that of the classical billiard to leading order; the destruction of invariant curves is caused by the $\\varphi$-shift, not by new $\\theta$-dynamics."],"forward_implications":["For every noncircular $C^5$ table and every $\\ell > -3/\\min \\rho''$, the coin map has no essential invariant curve in a boundary strip, and consequently has orbits that drift from the interior to the boundary.","The same non-existence forces a horseshoe and positive topological entropy for large heights.","For large heights, noncircular coin billiards are nonintegrable.","If a coin billiard at any height is completely foliated by essential invariant curves, the table is a disc; for the disc the map is integrable, so this is sharp.","In the small-height or near-circular regimes, a set of positive Lebesgue measure of Diophantine KAM curves exists in a strip away from the boundary, so those coin billiards are not ergodic."],"supporting_citations":[{"why":"Introduces the coin billiard model and the questions about invariant curves, integrability, and ergodicity that the paper answers.","marker":"[5]"},{"why":"Supplies the classical near-boundary expansion of the billiard map and the classical result that invariant curves accumulate on the boundary of an ordinary billiard.","marker":"[14]"},{"why":"Provides the generating-function technique used in the first proof of nonexistence of invariant curves near the boundary.","marker":"[17]"},{"why":"Supplies the Lipschitz bound for invariant curves of annulus twist maps used in the second proof of Theorem 9 and in the proof of Theorem 11.","marker":"[10]"},{"why":"States the analogue for ordinary billiards of the foliation-rigidity result that Theorem 11 extends to coin billiards.","marker":"[4]"}],"fun_headline_variants":["Tall noncircular coins erase boundary invariant curves","Above a height, noncircular coins lose all boundary curves","Only round coin billiards keep all invariant curves","Large coin heights force loss of near-boundary invariant curves","Coin billiards: critical height wipes out boundary curves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results hinge on the near-boundary Taylor expansion of the coin map remaining valid after one derivative in the arclength coordinate, which is what the $C^5$ regularity assumption guarantees; if the remainder terms are not controlled, the graph and Lipschitz arguments collapse.","fun_headline_variants_meta":{"raw":{"variants":["Tall noncircular coins erase boundary invariant curves","Above a height, noncircular coins lose all boundary curves","Only round coin billiards keep all invariant curves","Large coin heights force loss of near-boundary invariant curves","Coin billiards: critical height wipes out boundary curves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001131,"raw_usage":{"total_tokens":4714,"prompt_tokens":971,"completion_tokens":3743,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":587,"completion_tokens_details":{"reasoning_tokens":3663}},"tokens_in":587,"tokens_out":3743,"duration_ms":26828,"temperature":1.0,"reasoning_tokens":3663,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:45:38.251347+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix a strictly convex noncircular $C^5$ table (for example, an ellipse with known curvature $\\rho$), choose $\\ell > -3/\\min \\rho''$, and numerically integrate a fine grid of initial conditions in the strip $\\theta < \\delta$ for decreasing $\\delta$. If an essential invariant curve is found at arbitrarily small $\\delta$, Theorem 9 is false; equivalently, compute the Lipschitz constant of any candidate curve and check whether it violates the bound the paper derives.","supporting_citations":[{"cited_title":"Bialy, C","cited_arxiv_id":null,"evidence_quote":"Introduces the coin billiard model and the questions about invariant curves, integrability, and ergodicity that the paper answers."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the classical near-boundary expansion of the billiard map and the classical result that invariant curves accumulate on the boundary of an ordinary billiard."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the generating-function technique used in the first proof of nonexistence of invariant curves near the boundary."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Lipschitz bound for invariant curves of annulus twist maps used in the second proof of Theorem 9 and in the proof of Theorem 11."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the analogue for ordinary billiards of the foliation-rigidity result that Theorem 11 extends to coin billiards."}],"review_version":1}