{"id":"ad45ba48-3317-44aa-9f37-ee4c98ae0970","arxiv_id":"2506.19508","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The rotation number of a circle-map family through a rational rigid rotation is differentiable at that point under a transversality condition, with derivative equal to an explicit integral over the resonant Fourier terms.","lead":"A family of circle maps crossing a rigid rational rotation has a differentiable rotation number at that point when a Fourier-based condition holds, and the paper gives an explicit formula for the derivative. The formula is computable from the map's Fourier series, making the scaling behavior of mode-locked systems practical to analyze.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop. 4.2's Euler bound needs uniform C^1-in-x control of g_x, which Def. 1.1 does not assume; Theorem 1.3 is therefore only proved under an implicit stronger regularity hypothesis.","rationale":"The reader's weakest_assumption is the same uniform-regularity gap, and I agree with that identification. The rest of the paper, including the Fourier reduction in Proposition 2.1, the Euler passage-time argument, and the worked examples, is coherent, and formula (1.9) is plausible and numerically supported. The concern does not undermine the result in the smooth examples of Sections 3, 6, and 7; it affects the generality stated in Definition 1.1 and the rigor of Proposition 4.2 as written. The self-referential constant in (4.31) is a separate typo rather than a conceptual flaw. Hence the verdict remains conditional, with no change from the reader's recommendation.","tokens_in":14966,"tokens_out":17667,"duration_ms":193113,"concrete_test":"Take Definition 1.1 literally and try to choose mu-independent constants L, M, B, P, Q in the proof of Prop. 4.2; if this cannot be done, build G using a periodic C^1 bump of width mu^{3/2} and height mu^{1/4} on the circle, so that g->0 uniformly, sup|g_x|->infinity, mu^2 g_x->0, and the map remains a homeomorphism. Then estimate |rho(G)-T0^{-1}mu|/mu^2 numerically for mu=10^{-k}, k=3,...,7. An unbounded ratio shows Theorem 1.3 is false as stated; a bounded ratio shows the gap is only in the written proof and the theorem survives under the strengthened hypothesis g in C^1 jointly in (x,mu).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central derivative formula funnels through Proposition 4.2. There the map is written G=x+mu(a+psi)+mu^2 g and compared with Euler steps of dX/dt=a+psi(X)+mu g(X,mu). The global error bound (4.19) requires L=max|phi'| and M=max|phi phi'| to be finite on the whole mu-neighborhood, and (4.27) defines k3 via e^{LP}; likewise the lower bound m-mu0 B>0 requires a uniform bound B on |g|. But Definition 1.1 only says g is continuous in mu and C^1 in x for each mu. It does not imply sup_x|g_x(x,mu)| is bounded uniformly in mu: one can place a periodic C^1 bump of height mu^{1/4} and width mu^{3/2} on the circle, so g->0 uniformly, g_x~mu^{-5/4}, while mu^2 g_x->0, so the map remains a homeomorphism. Then L and the constant k3 used in Proposition 4.2 need not exist. The result may still be true, but the proof of the O(mu^2) error (4.21), and hence Corollary 4.4 and Theorem 1.3, is conditional on adding a uniform C^1 regularity assumption. A secondary symptom of the same under-specification is the self-referential constant k1=T0^{-2}(20J+16+2k1) in (4.31), which is not an explicit bound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies one-parameter families of circle homeomorphisms of the form F(x,μ)=x+p/q+μ(a+ψ(x))+μ^2 g(x,μ), where ψ is a mean-zero C^1 periodic function and g is continuous in μ and C^1 in x for each fixed μ. The main result, Theorem 1.3, states that if the Fourier-filtered function Ψ(x) built from the coefficients c_{nq} of ψ satisfies min(a+Ψ)>0, then the rotation number is differentiable at μ=0 with derivative (∫_0^1 dx/(a+Ψ(x)))^{-1}; if a+Ψ changes sign, the rotation number is locally locked at p/q; if max(a+Ψ)<0, the same formula holds after reversing the sign of μ. The proof reduces the rational case to q=1 by taking the qth iterate (Proposition 2.1), then compares the resulting difference equation with Euler steps of the ODE dX/dt=a+ψ(X)+μg(X,μ) (Proposition 4.2). The result is applied to Arnold maps, modified Arnold maps with higher harmonics, and piecewise linear circle maps, with numerical verification in each case.","tokens_in":15283,"tokens_out":6502,"duration_ms":63804,"significance":"If correct, Theorem 1.3 is a valuable and explicit complement to the Brunovsky-Herman results: it gives a computable derivative formula, with no fitted constants, for the rotation number at rational rigid rotations, and it explains why the local scaling is linear there rather than square-root-like at tongue boundaries. The reduction to q=1 in Proposition 2.1 is clean, the formula is testable, and the numerical comparisons are honest in the sense that they compare predicted slopes to simulations rather than tuning parameters. The result overlaps with Parkhe's theorem but offers a Fourier-series formula that is significantly easier to use. However, the proof as written has a regularity gap in the central passage-time argument and a self-referential constant in the key bound, so the main theorem is currently established only under an implicit stronger hypothesis. The paper is likely correct after a moderate revision that makes the regularity assumptions explicit and repairs the error estimates.","major_comments":[{"comment":"The proof of Proposition 4.2 requires uniform-in-μ control of g that Definition 1.1 does not provide. The Euler error bound (4.19) needs finite L = max|V'(x)| and M = max|V(x)V'(x)| on the whole μ-neighborhood, and the passage-time argument needs a uniform bound B with |g| ≤ B and the lower bound m−μ0B>0. But Definition 1.1 only assumes g is continuous in μ and C^1 in x for each fixed μ; these assumptions do not imply sup_x |g_x(x,μ)| is bounded uniformly in μ, nor even a uniform bound on |g|. For example, a periodic C^1 bump of height μ^{1/4} and width μ^{3/2} on the circle gives g→0 uniformly while g_x∼μ^{-5/4}, and μ^2g_x→0, so the map remains a small perturbation; nevertheless L and the constant k3 in (4.27) need not exist. Thus Theorem 1.3 is proved only under an implicit stronger hypothesis, such as g and g_x being bounded uniformly for |μ|<μ0. This assumption should be stated explicitly in Definition 1.1 or Proposition 4.2.","section":"§4, Proposition 4.2 and Definition 1.1"},{"comment":"Equation (4.31) defines k1 by k1 = T0^{-2}(20J+16+2k1), so k1 appears on both sides and the equation has no finite solution unless the constant term is zero. The preceding estimates give the fourth term in (4.29) as 2k2 μ^2/T0^2, so (4.31) should presumably read k1 = T0^{-2}(20J+16+2k2), which is explicit. As printed, the proof of Proposition 4.2 is incomplete at this point and needs a corrected, non-circular constant.","section":"§4, Eq. (4.31)"},{"comment":"The extension to μ<0 is deferred to \"the same argument as in [14]\". This is load-bearing because Theorem 1.3 asserts differentiability at μ=0, which requires both one-sided limits, and [14] is a preprint used as a black box here. The paper does provide an alternative proof of Theorem 7.1 for piecewise linear maps, but that proof does not by itself establish the smooth negative-μ estimate needed in Corollary 4.4. Please either prove the negative-μ case within this paper or state precisely which hypotheses from [14] are imported.","section":"§4, Corollary 4.4 (negative μ)"}],"minor_comments":[{"comment":"The inequality \"a − ψ(x) > m1 > 0\" appears to be a typo; it should presumably read \"a + ψ(x) > m1 > 0\" to imply F(x,μ)>x+p/q+m1μ+Bμ^2.","section":"§2, Lemma 2.2"},{"comment":"The constant k2 = min{m^{-2}, 2m^{-1}M^{-1}} does not follow from the displayed estimate for |T(μ)−T0|; the natural bound is of order m^{-2}, and the role of M and the choice of min rather than max are unclear. Please check and correct.","section":"§4, Eq. (4.24)"},{"comment":"The row (a,b,c)=(5,4,0.5) reports a numerical derivative 5.007 against a theoretical value 4.975; the text explains that a smaller μ-interval gives 4.977, which is plausible, but including error bars or a short convergence study would make the comparison more convincing.","section":"§6, Table 1"},{"comment":"There are several typographical issues: \"Deparment\", \"sh ow\", a stray \"S\" after (4.26), \"ec. 5\" in §7, and the notation \"10 5 iterations\" for 10^5 iterations. These should be corrected.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of math.DS and the main formula is attractive and likely correct. The two proof gaps identified above—missing uniform regularity in Proposition 4.2 and the circular constant in (4.31)—are repairable, as is the negative-μ reference to [14]. I would be happy to see a revised version. The dependence on [14] for one side of the derivative is worth editorial attention because it is a self-citation used as a black box."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Kiran — quick take on Glendinning's scaling paper. The headline result is worthwhile: for families of the form x + p/q + mu(a+psi(x)) + mu^2 g, the derivative of the rotation number at mu=0 is given by the explicit Fourier integral (integral dx/(a+Psi(x)))^(-1), with Psi the resonant Fourier sum. That is genuinely new relative to Parkhe, who left the derivative as a hard-to-evaluate integral over partial F^q/partial mu and assumed strict monotonicity. The reduction to q=1 in Proposition 2.1 is clean, and the Arnold-map corollaries — especially the q=1 sqrt(a^2-b^2) case — are attractive and match the numerical table.\n\nBut the proof of the key estimate, Proposition 4.2, has a real gap. The Euler global error bound (4.19) needs L = max|V'| and B = sup|g| to be finite uniformly in mu on the whole neighborhood. Definition 1.1 only gives g continuous in mu and C^1 in x for each mu. That does not imply uniform boundedness of g_x. The stress-test example is valid: a narrow bump of height mu^(1/4) and width mu^(3/2) has g->0 uniformly but g_x ~ mu^(-5/4), so mu g_x diverges and no uniform L exists. Under those conditions the passage-time argument does not go through. The theorem may still be true under the weaker hypothesis — I suspect it is — but the proof as written proves it under an implicit stronger regularity assumption. That should be stated, or Definition 1.1 strengthened.\n\nThere are smaller presentation issues: equation (4.31) defines k1 in terms of k1 and has no positive solution when T0^2 < 2, so it is almost certainly a typo (probably 2k2 or similar). Lemma 2.2 has a sign slip (a-psi vs a+psi). The argument that min(a+Psi)=0 forces Psi identically 0 is fine by mean zero, but the writing is terse.\n\nNet: the mathematical claim is plausible and the formula is useful, but the paper is not fully rigorous as written. It deserves a serious referee — the result is worth fixing. I would send to review and ask for a revised version that tightens the regularity assumptions and the bounds in Prop 4.2. For my own work, I would cite the formula if I need that scaling, after pointing the authors to the gap.","headline":"Explicit scaling formula for rotation number at rational rigid rotations is a real extension of Parkhe, but the proof as written has a regularity gap in Proposition 4.2 that needs patching.","tokens_in":739,"tokens_out":692,"would_cite":true,"duration_ms":48161,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37E45","39A28"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a circle map perturbed from a rational rotation, the rotation number is differentiable at the critical parameter when a transversality condition holds, with a Fourier-series formula for the slope.","keywords":["rotation number","circle maps","rational rotation","differentiability","mode locking","Fourier series","Arnold map","piecewise linear circle maps"],"falsifier":"Take the Arnold map family with $(\\alpha(\\mu),\\beta(\\mu))=(\\mu,b\\mu)$, so $q=1$, $a=1$, and $\\psi(x)=b\\sin(2\\pi x)$; Theorem 1.3 predicts $\\rho'(0)=1/\\sqrt{1-b^2}$ for $|b|<1$. Compute $\\rho(\\mu)$ numerically at $\\mu$ down to $10^{-5}$ and compare the measured slope with this value; any systematic deviation beyond the numerical error would refute the formula.","tokens_in":14756,"feed_emoji":"🌀","tokens_out":14031,"duration_ms":109525,"temperature":0.7,"pith_summary":"This paper establishes a precise condition under which the rotation number---the average speed of an orientation-preserving circle map---varies linearly with a parameter, even when the unperturbed map is a rational rotation $x \\mapsto x + p/q$. The condition is a transversality inequality: the function $a+\\Psi(x)$, built from the Fourier coefficients of the perturbation whose frequencies are multiples of $q$, must stay strictly positive. When it holds, $\\rho(\\mu)$ is differentiable at $\\mu=0$ and $\\rho'(0)=(\\int_0^1 dx/(a+\\Psi(x)))^{-1}$. If $a+\\Psi$ changes sign, the rotation number is constant on a neighborhood of $\\mu=0$, a mode-locked plateau with zero derivative. This contrasts with the square-root scaling found at the ends of Arnold tongues and extends classical differentiability results from irrational to rational rigid rotations.","feed_headline":"Rotation number scales linearly at rational rotations","feed_subtitle":"A Fourier-integral formula predicts the slope; without transversality the rotation number stays mode-locked flat.","key_machinery":"The $q$-th iterate reduction and the Euler passage-time comparison. For a perturbation of the rational rotation $p/q$, iterating the lift $q$ times yields $F^q(x,\\mu)=x+p+q\\mu(a+\\Psi(x))+\\mu^2\\hat{g}(x,\\mu)$, where $\\Psi$ keeps exactly the Fourier coefficients of $\\psi$ whose indices are multiples of $q$; translation by $p$ reduces the problem to rotation number zero. The discrete map is then read as Euler's method for the ordinary differential equation $dX/dt=a+\\Psi(X)$, and the time $T_0$ the exact solution takes to cross one full period is $T_0=\\int_0^1 dx/(a+\\Psi(x))$. Bounding the global Euler error by $O(\\mu)$ gives a rotation-number error $O(\\mu^2)$, so $\\rho(\\mu)=T_0^{-1}\\mu+O(\\mu^2)$ and hence differentiability. The transversality condition $\\min(a+\\Psi)>0$ is precisely what keeps the ODE well-posed and the passage time finite.","core_discovery":"The paper's central claim is Theorem 1.3. For a family of lifts $F(x,\\mu)=x+p/q+\\mu(a+\\psi(x))+\\mu^2 g(x,\\mu)$ with $\\psi$ a $C^1$ mean-zero periodic function and $g$ continuous in $\\mu$ and $C^1$ in $x$, define $\\Psi(x)=\\sum_{n\\in\\mathbb{Z}\\setminus\\{0\\}} c_{nq} e^{2\\pi i n q x}$ using the Fourier coefficients $c_m$ of $\\psi$. If $\\min_x(a+\\Psi(x))>0$, then $\\rho(\\mu)$ is differentiable at $\\mu=0$ and $\\rho'(0)=T_0^{-1}$ with $T_0=\\int_0^1 dx/(a+\\Psi(x))$. If $\\min(a+\\Psi)<0<\\max(a+\\Psi)$, then $\\rho(\\mu)=p/q$ for all small $\\mu$, so $\\rho'(0)=0$; if $\\max(a+\\Psi)<0$, the same formula holds after reversing the sign of $\\mu$. The mechanism is that iterating the lift $q$ times removes all non-resonant Fourier modes, so $F^q(x,\\mu)=x+p+q\\mu(a+\\Psi(x))+O(\\mu^2)$, and then the rotation number is controlled by the passage time of the differential equation $\\dot{X}=a+\\Psi(X)$.","pith_inferences":["Because only the Fourier modes whose indices are multiples of $q$ enter the slope, the linear scaling acts as a spectral filter: non-resonant harmonics affect $\\rho(\\mu)$ only at order $\\mu^2$ or higher, so the linear coefficient is a direct measurement of the resonant projection of the perturbation.","If the first nonzero resonant term appears only at order $\\mu^k$ rather than $\\mu$, the linear derivative should vanish and the leading scaling should be set by that higher-order resonant term; this is testable by adding a term $\\mu^2 h(x)$ with nonzero $c_{nq}$ to the family and measuring a presumably quadratic scaling.","The passage-time interpretation suggests that the next coefficient in the expansion of $\\rho(\\mu)$ can be obtained by carrying the Euler comparison one order further, tracking the $O(\\mu^2)$ term $g(x,\\mu)$; such an expansion would predict the curvature of the rotation-number curve as it approaches the mode-locked plateau.","This result may sharpen the practical distinction between rational and irrational frequency locking: at irrational rigid rotations all harmonics contribute to the derivative, while at rational rotations only resonant harmonics do, so the slope of the rotation number carries information about the commensurability of the perturbation."],"forward_implications":["If the theorem is correct, the rotation number near a rational rigid rotation obeys a linear scaling law whose slope is the reciprocal of the resonant passage time, so linear-response measurements can extract the resonant Fourier content of the perturbation.","In the Arnold map with $q\\ge2$, the slope is exactly $\\alpha'(0)$, independent of $\\beta'(0)$, because the sine perturbation has no Fourier component of order $nq$.","At the origin of the Arnold map ($p/q=0/1$), the slope is $\\mathrm{sgn}(\\alpha'(0))\\sqrt{|\\alpha'(0)|^2-|\\beta'(0)|^2}$ whenever $|\\alpha'(0)|>|\\beta'(0)|$.","When $a+\\Psi$ changes sign, the map is mode-locked at $p/q$ on a whole neighborhood of $\\mu=0$; differentiability still holds but with zero derivative.","The same Euler passage-time method reproduces the derivative formula for piecewise linear circle maps, giving a unified treatment of the scaling at rational rotations."],"supporting_citations":[{"why":"This reference supplies the classical result that the rotation number is differentiable at irrational rigid rotations, with derivative equal to the mean of the perturbation, and records the remark that rational rotations are not differentiable in general.","marker":"[15]"},{"why":"This reference provides the pathwise differentiability formula at irrational rotations that the paper's rational-case formula is designed to mirror.","marker":"[6]"},{"why":"This reference gives the closest prior theorem, differentiability under strict monotonicity of the rotation number, but with a less explicit derivative formula.","marker":"[24]"},{"why":"This reference states the piecewise linear analogue, Theorem 7.1, which the paper re-proves with the Euler passage-time method.","marker":"[14]"},{"why":"This reference supplies the global error bound for Euler's method used to convert passage times of the ODE into bounds on the rotation number.","marker":"[5]"},{"why":"This reference provides the standard properties of rotation numbers, including the iterate relation and the crossing lemma used to sandwich the rotation number.","marker":"[10]"},{"why":"This reference justifies the interchange of sums in the q-th iterate computation via absolute convergence of Fourier series of C^1 functions.","marker":"[18]"}],"fun_headline_variants":["Fourier formula gives rotation number's linear slope","Transversality condition yields rotation number derivative","Rational rotations: linear scaling with explicit slope","Mode-locking versus linear scaling in circle maps","Explicit derivative for rotation number at rationals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The chain of estimates needs $a+\\psi(x)+\\mu g(x,\\mu)$ to stay bounded away from zero and to have a uniform $C^1$-in-$x$ bound with finite Lipschitz constant and a uniform bound on $|g|$ over a whole neighborhood of $\\mu=0$; the paper's Definition 1.1 only assumes continuity in $\\mu$ and $C^1$-in-$x$ for each fixed $\\mu$, so without that extra uniformity the $O(\\mu^2)$ error estimates can fail.","fun_headline_variants_meta":{"raw":{"variants":["Fourier formula gives rotation number's linear slope","Transversality condition yields rotation number derivative","Rational rotations: linear scaling with explicit slope","Mode-locking versus linear scaling in circle maps","Explicit derivative for rotation number at rationals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000251,"raw_usage":{"total_tokens":1559,"prompt_tokens":952,"completion_tokens":607,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":537}},"tokens_in":568,"tokens_out":607,"duration_ms":6456,"temperature":1.0,"reasoning_tokens":537,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:33:28.815604+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the Arnold map family with $(\\alpha(\\mu),\\beta(\\mu))=(\\mu,b\\mu)$, so $q=1$, $a=1$, and $\\psi(x)=b\\sin(2\\pi x)$; Theorem 1.3 predicts $\\rho'(0)=1/\\sqrt{1-b^2}$ for $|b|<1$. Compute $\\rho(\\mu)$ numerically at $\\mu$ down to $10^{-5}$ and compare the measured slope with this value; any systematic deviation beyond the numerical error would refute the formula.","supporting_citations":[{"cited_title":"Herman, M´ esure de Lebesgue et Nombre de Rotation, in Geometry and Topology, eds","cited_arxiv_id":null,"evidence_quote":"This reference supplies the classical result that the rotation number is differentiable at irrational rigid rotations, with derivative equal to the mean of the perturbation, and records the remark that rational rotations are not differentiable in general."},{"cited_title":"Brunovsky (1974) Generic properties of the rotation numbe r of one-parameter diffeo- morphisms of the circle, Czech","cited_arxiv_id":null,"evidence_quote":"This reference provides the pathwise differentiability formula at irrational rotations that the paper's rational-case formula is designed to mirror."},{"cited_title":"Parkhe (2013) One-parameter families of circle diffeomorphis ms with strictly mono- tone rotation number, Proc","cited_arxiv_id":null,"evidence_quote":"This reference gives the closest prior theorem, differentiability under strict monotonicity of the rotation number, but with a less explicit derivative formula."},{"cited_title":"Piecewise linear circle maps and conjugation to rigid rational rotations","cited_arxiv_id":"2505.13689","evidence_quote":"This reference states the piecewise linear analogue, Theorem 7.1, which the paper re-proves with the Euler passage-time method."},{"cited_title":"Braun, Differential Equations and Their Applications: An Introduc tion to Applied Mathematics, Texts in Applied Mathematics 11, Springer, New York, 1993","cited_arxiv_id":null,"evidence_quote":"This reference supplies the global error bound for Euler's method used to convert passage times of the ODE into bounds on the rotation number."},{"cited_title":"Devaney An Introduction to Chaotic Dynamical Systems","cited_arxiv_id":null,"evidence_quote":"This reference provides the standard properties of rotation numbers, including the iterate relation and the crossing lemma used to sandwich the rotation number."},{"cited_title":"Katznelson, An introduction to harmonic analysis , 3 rd edition, CUP, Cambridge, 2004","cited_arxiv_id":null,"evidence_quote":"This reference justifies the interchange of sums in the q-th iterate computation via absolute convergence of Fourier series of C^1 functions."}],"review_version":2}