{"id":"1374f6ff-d366-4ac1-827e-6eac5bdcb0be","arxiv_id":"2507.05124","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For complex measures with Szegő coefficients of opposite signs (class T−), the paper proves a Mate-Nevai-Totik universality bound and a.e. convergence of (φ*_n φ̃_n)² along lacunary sequences, a functional version of the nonlinear Carleson problem.","lead":"This mathematics paper proves new convergence theorems for orthogonal polynomial systems tied to SU(2)-valued nonlinear Fourier series, the machinery behind quantum signal processing. It shows a universality law for the associated reproducing kernels and almost-everywhere convergence of a squared functional along lacunary subsequences, a genuine step toward but not a solution of the open nonlinear Carleson problem.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5(1) depends on the cited existence theorem [Tsa05, Lemma 3.7]/[AMT23, Theorem 11]; any hidden hypotheses there would invalidate the construction of the measure and the lacunary convergence claim.","rationale":"The reader's weakest assumption identifies precisely the cited existence theorem as the load-bearing external input. My independent reading of Sections 1–9 confirms that Theorems 1–4 are proved with detailed estimates and no obvious internal gap: the kernel universality proof (Theorem 1), the local parameter lemmas (Lemmas 8–17), the zero behavior (Theorem 3), and the lacunary argument (Theorem 4) all appear coherent. The construction in Section 9 is careful about verifying monicity, orthogonality, uniqueness, T− membership, and the integral condition (1.15). The only step left to an outside source is the existence of (F_n) for the given (a,b). This is genuinely load-bearing because without it the measure µ cannot be formed and the whole SU(2) application does not start. However, the concern is not an observed contradiction: the paper's own proof later shows strong consequences (e.g., |a*|>1/√2 in D) that likely make the cited theorem's hypotheses automatically satisfied. Therefore the verdict should remain conditional, with the concrete check being a verification of the cited theorem's exact assumptions. The arXiv text-layer corruption of a few displayed formulas (Lemmas 15, 16, 20, Appendix 10.7) is a real but secondary mechanical issue, already noted by the reader, and would not by itself change the verdict.","tokens_in":36681,"tokens_out":13624,"duration_ms":152439,"concrete_test":"Obtain the statements of [Tsa05, Lemma 3.7] and [AMT23, Theorem 11] and compare their hypotheses with the assumptions of Theorem 5(1). Specifically, check whether those results require extra conditions beyond b(0)=0, b∈H∞, ||b||∞ < 2^{-1/2}, a* outer, and |a|²+|b|²=1 on T. If they require, say, invertibility of a*−b in H∞, verify that this is already implied by the proof in Section 9 (where |a*|>1/√2 in D is shown). If a stronger condition is required that is not implied, construct a pair (a,b) satisfying Theorem 5(1)'s assumptions but violating that condition, and test numerically whether its recursively defined NLFS coefficients F_n satisfy ∑|F_n|² < ∞. If no such counterexample exists and the cited hypotheses are implied, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central application to SU(2)-valued nonlinear Fourier series, Theorem 5(1), is obtained by reducing to Theorem 4. The reduction constructs a measure µ = w d|z|/2π with w = 1/((a*−b)(a+b*)) and proves that the normalized polynomials φ_n, φ̃_n built from the NLFS are one-sided orthogonal polynomials for µ. The very first step of Section 9 is: 'By [Tsa05, Lemma 3.7] and [AMT23, Theorem 11], there exists a unique (F_n) ∈ ℓ²(N0) such that (a,b) is the nonlinear Fourier series of (F_n).' This is the only step that is not proved in the manuscript, and the paper does not restate the hypotheses of those two results. If the cited theorems require more than b(0)=0, b∈H∞(D), ||b||∞ < 2^{-1/2}, a* outer, and |a|²+|b|²=1 on T — for example an additional invertibility or spectral condition — then the class of pairs covered by Theorem 5(1) may be strictly smaller than stated. Because the existence of (F_n) is the entry point to the whole construction of µ, any mismatch would collapse the proof of (1.27). The remaining internal estimates (Theorems 1–4) appear consistent, so this is an external-dependency concern rather than an internal inconsistency.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a class T- of complex probability measures on the unit circle whose unique monic left and right orthogonal polynomials satisfy Szegő recurrences with coefficients that are negatives of each other. For this class, it proves a Mate--Nevai--Totik type universality estimate for the reproducing kernel (Theorem 1), analyzes the local parameters that describe the polynomials near a Lebesgue point (Theorems 2 and 3), and establishes almost everywhere convergence of the squared product of the normalized left and right polynomials along lacunary sequences under an L1 condition and square summability of the recurrence coefficients (Theorem 4). The final theorem applies these results to SU(2)-valued nonlinear Fourier series: for boundary data (a,b) with b(0)=0, ||b||_{H∞}<2^{-1/2}, a* outer, and |a|^2+|b|^2=1, it proves almost everywhere convergence along lacunary sequences of the functional (a_{n_k}^*+b_{n_k})^2(a_{n_k}-b_{n_k}^*)^2 to its expected limit, and it shows the constant 2^{-1/2} is sharp.","tokens_in":36746,"tokens_out":14213,"duration_ms":143830,"significance":"If the results hold, this paper establishes a genuinely new connection between a natural class of non-Hermitian orthogonal polynomials and the SU(2) nonlinear Fourier transform, resolving a lacunary version of the nonlinear Carleson problem in the small-||b|| regime. The proofs are long, detailed, and largely self-contained, with explicit constants in the estimates; the authors are also careful to state the limitations of their method and to prove sharpness of the main hypothesis. The universality theorem for the class T- and the accompanying zero/local-parameter analysis are likely to be of independent interest. The main caveat is that the final application rests on an external existence theorem that is invoked but not stated, and one step in a key lemma is unjustified as written.","major_comments":[{"comment":"The proof of Theorem 5(1) begins with the assertion 'By [Tsa05, Lemma 3.7] and [AMT23, Theorem 11], there exists a unique (F_n) ∈ ℓ²(N0) such that (a,b) is the nonlinear Fourier series of (F_n).' This is the load-bearing entry point for the entire construction of the measure µ, and the hypotheses of those cited results are not stated. Please state precisely the theorems being cited (or at least their full hypotheses) and verify explicitly that conditions (1.25)-(1.26), together with b(0)=0 and a* outer, satisfy them. If the cited theorems require additional hypotheses, such as extra regularity or an invertibility condition, then the class of pairs covered by Theorem 5(1) is narrower than stated and the application of Theorems 1-4 would fail for some of the claimed inputs.","section":"Section 9, first paragraph and Theorem 5(1)"},{"comment":"The proof of Lemma 12 applies Lemma 11 to infer |~A_n| ≤ 9|w(s)| |A_n| from the assumption |A_n| ≤ η². However, Lemma 11 is stated under the hypothesis η ≤ C, and the case at hand has |A_n| ≤ η², so the hypothesis is not satisfied. The same issue occurs for |~A_n| ≤ η² and in the corresponding arguments for the B-variables. As written, the chain leading to (4.27) and (4.28) is therefore incomplete. Since Lemma 12 is used in the proof of Theorem 2(1), and Theorem 2(1) is used in the proof of Theorem 4, this gap affects the central argument. The estimate can likely be repaired using (4.14), (4.15) and the lower bound on |B_n~B_n| that follows from |w(s)|≥1; the manuscript must supply a correct proof.","section":"Section 4, proof of Lemma 12 (after (4.26))"}],"minor_comments":[{"comment":"The displayed identity for φ*_n ~φ_n contains incorrect factors: it should state φ*_n = z^{-n}(a*_n + b_n) and ~φ_n = z^n(a_n - b*_n), so that φ*_n ~φ_n = (a*_n + b_n)(a_n - b*_n). The current formula 'z^{-n}(a_n + b*_n) z^{-n}(a_n - b*_n)' is not correct as written, although the final identity is the right one.","section":"Section 9, proof of Theorem 5"},{"comment":"Several displayed equations in Sections 5 and 6 contain garbled symbols and braces (for example, '⌟⟨rro⟪⟪⟩r⟪' appears repeatedly), making those lines unreadable. If these are not artifacts of the version we reviewed, they must be corrected in the final typeset manuscript.","section":"Equations (5.14), (5.17), (5.29), (6.7), (6.10)"},{"comment":"The phrasing 'There is a set E0(µ)⊂E(µ) of full measure such that for every s∈E0(µ), we have (1.16) implies both (1.17), (1.18)' is grammatically confusing. It would be clearer to state: 'For every s∈E0(µ), condition (1.16) implies (1.17) and (1.18).'","section":"Theorem 2, second part"},{"comment":"The title in the running header contains spacing/typing errors ('OR THOGONAL', 'FORSU', 'POL YNOMIALS'). Please ensure the final version has the correct title.","section":"Title and header"}],"recommendation":"major_revision","confidential_remarks":"The external dependency on [Tsa05, Lemma 3.7] and [AMT23, Theorem 11] is the single most delicate point in the chain from Theorems 1-4 to the nonlinear Fourier series application. It would be prudent for the editor to ask the authors to include the exact statement of the cited existence theorem and a verification of its hypotheses, since a referee without access to the Ph.D. thesis of Tsai cannot confirm that the conditions in Theorem 5 are sufficient. The gap in Lemma 12, while seemingly repairable, is exactly the kind of local technical issue that should be resolved in revision rather than left to the reader."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real contribution, not a dud. The T− class (Fn = −F̃n) appears genuinely new, and Theorem 1 gives a MNT-type reproducing kernel estimate with explicit constants and a transparent proof. The local parameter geometry (the A/B regions) is the right tool, and the transfer to SU(2) nonlinear Fourier series along lacunary sequences is a clean partial answer to a known open problem. The proofs are long but arranged honestly, with the sharpness counterexample in Theorem 5(2) providing a useful calibration. I read the full text and found no internal gap in Theorems 1–4. The Christoffel–Darboux identity, the local approximation lemmas, and the Plancherel step in Theorem 4 all check out logically. The reader's 'conditional' verdict is about the right calibration.\n\nThe genuine weak point is Section 9. Theorem 5(1) begins by invoking [Tsa05, Lemma 3.7] and [AMT23, Theorem 11] for the existence of the NLFS, and the hypotheses of those results are not restated. The stress-test note is on target: if those cited theorems require extra regularity or a spectral condition beyond outerness and the L∞ bound, then the construction of µ, and hence the whole application to NLFS, could be restricted to a smaller class than claimed. I don't think this is likely to be wrong — the same group wrote the cited preprints and the properties used are consistent with the known SU(2) NLFS theory — but the paper should clearly restate the result it relies on. That is an external dependency, not an internal contradiction.\n\nA second, more mechanical issue: several displayed identities in the arXiv text (Lemma 15 (5.14), Lemma 16 (5.29), Lemma 20 (6.21), Appendix (10.7)) are corrupted into unreadable tokens, so those steps cannot be verified as printed. The authors should fix the source text; the surrounding mathematics appears recoverable.\n\nThis paper deserves a serious referee. The class T− and the lacunary convergence result are worth knowing, and the proof skeleton is sound. I would send it to review with a request that Section 9 either prove or fully restate the external existence theorem. If the authors do that, I would take it as a solid advance.","headline":"Genuinely new class T−, a clean kernel universality theorem, and a lacunary convergence result for SU(2) NLFS; the only real risk is the unstated hypotheses of the external existence theorem behind Theorem 5.","tokens_in":37557,"tokens_out":1983,"would_cite":true,"duration_ms":23188,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42C05","42A20","30C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For complex measures in the class T−, the reproducing kernel of one-sided orthogonal polynomials obeys a Máté–Nevai–Totik type estimate, and this yields almost-everywhere convergence of the SU(2) nonlinear Fourier series functional along…","keywords":["nonlinear Fourier series","SU(2) scattering","orthogonal polynomials on the unit circle","reproducing kernels","universality","lacunary convergence","Szegő recurrence","pointwise convergence"],"falsifier":"Find one measure $\\mu\\in T^-$ with $\\sum|F_n|^2<\\infty$ and (1.15) for which, along some lacunary sequence, $(\\varphi^*_{n_k}\\tilde\\varphi_{n_k})^2$ fails to converge to $w^{-2}$ on a positive-measure set, or find one admissible pair $(a,b)$ as in Theorem 5 whose functional $(a^*_{n_k}+b_{n_k})^2(a_{n_k}-b^*_{n_k})^2$ fails on a positive-measure set.","tokens_in":36239,"feed_emoji":"🧮","tokens_out":13000,"duration_ms":111480,"temperature":0.7,"pith_summary":"The paper extends the Máté–Nevai–Totik universality theorem from classical orthogonal polynomials to one-sided left and right orthogonal polynomials attached to complex measures on the unit circle in a class the authors call $T^-$. For these measures the Szegő recurrences have coefficients $F_n$ and $-F_n$, which makes the setup match the $SU(2)$-valued nonlinear Fourier series. The main technical result is a kernel estimate: close to a Lebesgue point of the measure, the reproducing kernel $K_n$ of the one-sided polynomials differs from the Dirichlet kernel by an error controlled by a local average of the measure's deviation from its density. From this estimate the paper derives a sharp local description of the polynomials in terms of two parameters $A_{n,s}$ and $B_{n,s}$, and shows that convergence of their product is governed by which parameter becomes small. The payoff is that along lacunary subsequences the square of the product of the left and right normalized polynomials converges almost everywhere to $w(s)^{-2}$, which yields almost-everywhere convergence of the functional $(a^*_n+b_n)(a_n-b^*_n)$ for $SU(2)$ nonlinear Fourier series under the conditions $\\|b\\|_\\infty<2^{-1/2}$ and $a^*$ outer.","feed_headline":"SU(2) nonlinear Fourier series converge along lacunary subsequences","feed_subtitle":"Partial SU(2) series converge almost everywhere along lacunary indices when b is small and a* is outer.","key_machinery":"The central object is the reproducing kernel $K_n(z,\\lambda)=\\sum_{j=0}^n\\tilde\\varphi_j(z)\\varphi_j^*(\\lambda)$ built from the normalized left and right orthogonal polynomials $\\varphi_n,\\tilde\\varphi_n$ of a measure $\\mu\\in T^-$. For measures in $T^-$ the Szegő recurrences take the form $\\Phi_{n+1}-z\\Phi_n=F_{n+1}z^n\\tilde\\Phi_n^*$ and $\\tilde\\Phi_{n+1}-z\\tilde\\Phi_n=-F_{n+1}z^n\\Phi_n^*$, and the Christoffel–Darboux identity $(1-z\\lambda^{-1})K_n(z,\\lambda)=z^{n+1}\\lambda^{-n-1}\\tilde\\varphi_{n+1}^*(z)\\varphi_{n+1}(\\lambda)-\\varphi_{n+1}(z)\\tilde\\varphi_{n+1}^*(\\lambda)$ carries the argument. The local parameters $A_{n,s}=(\\varphi_n(s)-\\varphi_n(s\\gamma_n))/(2s^n)$ and $B_{n,s}=(\\varphi_n(s)+\\varphi_n(s\\gamma_n))/2$, with $\\gamma_n=e^{i\\pi/n}$, decide which of two possible limits occurs: small $A$ forces $\\varphi_n^*\\tilde\\varphi_n$ near $-w(s)^{-1}$, small $B$ forces it near $w(s)^{-1}$, and the theorems show the sequence stabilizes in the small-$B$ region.","core_discovery":"The central claim is Theorem 4: for a measure $\\mu\\in T^-$ whose coefficients $F_n$ are square summable and whose orthogonal polynomials satisfy the integral condition (1.15), along any lacunary sequence $(n_k)$ the squares of the products of normalized left and right polynomials converge almost everywhere, $(\\varphi^*_{n_k}(s)\\tilde\\varphi_{n_k}(s))^2\\to w(s)^{-2}$, where $w$ is the density of the absolutely continuous part of $\\mu$. Theorem 1 supplies the structural input: with $C\\ge2$ and $z,\\lambda$ within $C/n$ of a Lebesgue point $s$, the normalized difference $|w(s)K_n(z,\\lambda)-D_n(z,\\lambda)|/(n+1)$ is bounded by $e^{30C}L(\\mu,s,n)$, where $L(\\mu,s,n)\\to0$ at Lebesgue points. This universality-type estimate, proved through a Christoffel–Darboux formula for the one-sided polynomials, is simpler than its $SU(1,1)$ analogue. Theorem 5 converts the measure statement into a statement about $SU(2)$ nonlinear Fourier series: for $b\\in H^\\infty$ vanishing at the origin with $\\|b\\|_\\infty<2^{-1/2}$ and $a^*$ outer with $|a|^2+|b|^2=1$ almost everywhere, the functional $(a^*_{n_k}+b_{n_k})^2(a_{n_k}-b^*_{n_k})^2$ converges almost everywhere to $(a^*+b)^2(a-b^*)^2$, and the $L^\\infty$ bound on $b$ is shown to be sharp.","pith_inferences":["The lacunary restriction probably comes from the stopping-time argument in Theorem 4; controlling the local parameters on all scales, not just lacunary ones, would turn this into a full nonlinear Carleson theorem for the functional $(a^*_n+b_n)(a_n-b^*_n)$.","The sharpness example at $\\|b\\|_\\infty=2^{-1/2}$ suggests the threshold is intrinsic; a natural test is whether the conclusion survives with a BMO or weighted $L^p$ condition on $b$ in place of the $L^\\infty$ bound.","The paper's remark connecting these polynomials to Krein systems and de Branges spaces points toward a continuous analogue: the kernel estimate should translate into a universality result for Krein systems, with the local-parameter stabilization becoming a pointwise convergence statement for continuous scattering data.","Condition (1.15) is an integral average control; an editor's guess is that it can be relaxed to a logarithmic or variational condition in the lacunary theorem, since the proof only needs it to keep the small-$A$ and small-$B$ regions separated."],"forward_implications":["For every measure $\\mu\\in T^-$ with $\\sum|F_n|^2<\\infty$ and with (1.15), $\\lim_{k\\to\\infty}(\\varphi^*_{n_k}(s)\\tilde\\varphi_{n_k}(s))^2=w(s)^{-2}$ for almost every $s$ along every lacunary sequence $(n_k)$.","If additionally $|F_n|\\to0$ and $(\\varphi^*_n\\tilde\\varphi_n)^2$ converges almost everywhere to $w^{-2}$, then the full sequence $\\varphi^*_n(s)\\tilde\\varphi_n(s)$ converges almost everywhere to $w(s)^{-1}$ and $A_{n,s}B_{n,s}\\to0$.","For any admissible pair $(a,b)$ with $\\|b\\|_\\infty<2^{-1/2}$ and $a^*$ outer, the partial $SU(2)$ nonlinear Fourier series satisfies $(a^*_{n_k}+b_{n_k})^2(a_{n_k}-b^*_{n_k})^2\\to(a^*+b)^2(a-b^*)^2$ almost everywhere along lacunary sequences.","Zero spacing is tied to the same convergence: $|A_{n,s}B_{n,s}|\\ge\\varepsilon$ forces a zero of $\\varphi_n$ within distance $O(\\varepsilon^{-1}/n)$ of $s$, and conversely a zero within $\\varepsilon^{-1}/n$ forces $|A_{n,s}B_{n,s}|\\ge e^{-10/\\varepsilon}$.","The first-order expansion of the kernel estimate reproduces classical Fejér mean convergence, so the theorem contains ordinary linear Fourier analysis as a special case."],"supporting_citations":[{"why":"Supplies the SU(2) nonlinear Fourier series product representation and the existence/uniqueness of the series for square-summable sequences that identify the orthogonal polynomials with partial series data.","marker":"[Tsa05]"},{"why":"Provides the converse characterization, Theorem 11, that admissible pairs (a,b) with outer a* and $\\|b\\|_\\infty<2^{-1/2}$ arise from a unique $\\ell^2$ sequence, the external input Theorem 5 builds on.","marker":"[AMT23]"},{"why":"Gives the original Máté–Nevai–Totik universality theorem whose reproducing-kernel estimate Theorem 1 generalizes to the $T^-$ class.","marker":"[MNT91]"},{"why":"Motivates studying pointwise convergence of functionals of partial nonlinear Fourier series and supplies the SU(1,1) approach that the paper adapts to SU(2) data.","marker":"[Pol24]"},{"why":"Supplies the OPUC result connecting zero sets to pointwise asymptotics that Theorem 3 mirrors in the one-sided setting.","marker":"[BD21]"}],"fun_headline_variants":["Lacunary a.e. convergence for SU(2) nonlinear Fourier series","A.e. convergence of SU(2) Fourier functionals on lacunary subsequences","Outer a* and small b give a.e. convergence for SU(2) Fourier","Sharp bound for a.e. convergence of SU(2) Fourier functional"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is an external existence theorem: any two boundary functions $(a,b)$ with $|a|^2+|b|^2=1$ on the circle, $a^*$ having no zeros in the disk, and $b$ uniformly smaller than $2^{-1/2}$ in magnitude must be the nonlinear Fourier series of some square-summable sequence, and the paper's measure construction in Theorem 5 collapses if that theorem carries hidden extra hypotheses.","fun_headline_variants_meta":{"raw":{"variants":["Lacunary a.e. convergence for SU(2) nonlinear Fourier series","A.e. convergence of SU(2) Fourier functionals on lacunary subsequences","Outer a* and small b give a.e. convergence for SU(2) Fourier","Sharp bound for a.e. convergence of SU(2) Fourier functional"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002246,"raw_usage":{"total_tokens":8771,"prompt_tokens":1124,"completion_tokens":7647,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":740,"completion_tokens_details":{"reasoning_tokens":7558}},"tokens_in":740,"tokens_out":7647,"duration_ms":56696,"temperature":1.0,"reasoning_tokens":7558,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:32:59.062445+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find one measure $\\mu\\in T^-$ with $\\sum|F_n|^2<\\infty$ and (1.15) for which, along some lacunary sequence, $(\\varphi^*_{n_k}\\tilde\\varphi_{n_k})^2$ fails to converge to $w^{-2}$ on a positive-measure set, or find one admissible pair $(a,b)$ as in Theorem 5 whose functional $(a^*_{n_k}+b_{n_k})^2(a_{n_k}-b^*_{n_k})^2$ fails on a positive-measure set.","supporting_citations":[],"review_version":1}