{"id":"ac153e33-c33f-4c0f-8faa-97212dfd85f3","arxiv_id":"2509.04212","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Claims to confirm the L^α-Littlewood, L^1-Newman, and L^∞-Erdős conjectures via a short Clarkson-inequality argument, but the key norm-convergence step is unjustified.","lead":"A new preprint claims to prove that no polynomial with unit-modulus coefficients can be L^p-flat for any p > 0, which would settle long-standing flatness conjectures in harmonic analysis and prove finiteness of Barker sequences. The proof as written contains a central gap, so the conjectures remain open.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3 is misstated and misapplied: the printed inequality is non-homogeneous (F≡1,G≡0 already violates it), and its parameter condition r'≤s≤p forbids r=2+δ→2 for fixed α<2; the contradiction in Theorem 1 does not follow.","rationale":"After reading the proof, the central claim as written does not hold together. The Reader's weakest_assumption (L^α-flatness does not force ||f̃_n||_r→1) is a legitimate gap in the literal transition from (13) to (14), but it is not the most load-bearing: one could replace that step by the trivial bound ||f̃_n||_r≥||f̃_n||_2=1 and still run the intended contradiction. The real breakdown is before that: the Boas–Clarkson inequality quoted as Lemma 3 is false as printed and its parameter restrictions exclude the choice r=2+δ→2 for fixed α<2. Without (11), no contradiction follows. This is not an issue of current consensus; it is an internal defect.\n\nI also checked the advertised applications. Corollary 3 would require applying Theorem 1 to polynomials whose coefficients ε_n(k) depend on n, whereas the theorem assumes a single fixed sequence (a_j,c_j). No diagonal extraction or finite-block approximation is given, so the Erdős–Newman and Barker-sequence conclusions are not consequences of the stated theorem. Although this is serious for the paper's significance, the invalidity of Lemma 3 in the main proof is the most load-bearing concern. Verdict remains REJECT; no adjustment to the Reader's verdict.","tokens_in":14239,"tokens_out":18930,"duration_ms":180244,"concrete_test":"Verify Lemma 3 itself before considering the application: set X=[0,1] with Lebesgue measure, B=ℝ, F≡1, G≡0, p=α=3/2, r=2, s=α. The printed inequality (2) gives √2 ≤ 2^{1/3 - 1/2} = 2^{-1/6}, which is false; hence the lemma as quoted is not a valid theorem. Then check the parameter admissibility for the actual proof: for α=3/2 and r=2.5, r'=5/3>α, so no s can satisfy r'≤s≤α. Repeat with r=α'=3: if a corrected inequality holds, compute the resulting bound; if it does not contradict, the δ→0 argument in the paper cannot be reconstructed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"At the core of the proof of Theorem 1, equations (11)--(16), the argument applies Lemma 3 with p=α, r=2+δ, and some s. Two independent defects make this step unsound.\n\nFirst, Lemma 3 as printed cannot be true: with F≡1, G≡0 on a probability space and s=p=α, r=2, the left side is √2 while the right side is 2^{1/α' - α/(2α)} = 2^{(α-2)/(2α)} < 1. The displayed inequality has unmatched homogeneity (LHS scales as λ^{p/r}, RHS as λ^{p/s}); it is not the classical Boas–Clarkson estimate.\n\nSecond, even granting a correct inequality of that shape, the parameter condition 1<s≤p≤r, r'≤s forces r'≤α, i.e. r≥α'=α/(α-1)>2. For fixed α in (1,2), r cannot be taken arbitrarily close to 2; the final step 'letting δ→0' is therefore unavailable. Example: α=3/2 gives α'=3, whereas the proof needs r=2+δ<3.\n\nThus the inequality chain (11)-(16) does not produce the claimed contradiction. The L^α→L^r convergence gap identified by the Reader is real but is not the only obstruction; even using the valid lower bound ||f̃_n||_r≥1, the Lemma 3 step collapses. Separately, Corollary 3's n-dependent coefficients are not covered by Theorem 1, which is stated only for a fixed coefficient sequence; the advertised Erdős–Newman/Barker consequences would need a missing diagonal argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a generalization of Littlewood's L^α-flatness criterion: no sequence of normalized analytic polynomials P_n(z) = (∑_{j=1}^n a_j c_j z^j)/(∑ a_j^2)^{1/2} with ∑ a_m^2 ≤ (K/n^2) ∑ m^2 a_m^2 can be L^p-flat for any p>0. From this it derives several corollaries, including the L^α-Littlewood conjecture, the L^1-Newman and L^∞-Erdős conjectures, finiteness of Barker sequences, singularity of certain Morse cocycle spectra, and a new proof of the Beller–Newman theorem via Gauss–Fresnel polynomials. The proof of the main theorem combines Littlewood's L^α criterion with a generalized Clarkson inequality (Lemma 3) and a contradiction argument letting a parameter r approach 2.","tokens_in":14680,"tokens_out":8905,"duration_ms":77450,"significance":"If the main theorem were correct, it would indeed resolve a family of long-standing conjectures in one stroke, and the applications to ergodic theory and number theory would be notable. The paper also includes some original ideas: the reduction of the problem to real and imaginary parts and the use of generalized Clarkson inequalities is a plausible route. However, the central proof contains a serious, load-bearing gap: the deduction from L^α-flatness to convergence of higher L^r norms is not justified and is false in general. Moreover, the stated Lemma 3 is itself invalid, and the passage from the fixed-sequence Theorem 1 to n-dependent Littlewood coefficients is not covered. These issues are not local; they affect the main result and its advertised consequences.","major_comments":[{"comment":"The step from (13) to (14) assumes that ‖f̃_n‖_r → 1 for r = 2+δ > α under L^α-flatness. The hypothesis only gives ‖f̃_n‖_α → 1. L^α convergence does not control higher L^r moments; a sequence with spikes on sets of small measure illustrates the failure. Without an additional uniform bound on ‖f̃_n‖_r, the left-hand side of (11) need not converge to 2^{1/r}, and the contradiction does not follow.","section":"§1, proof of Theorem 1, eqs. (11)–(14)"},{"comment":"Lemma 3 as stated is false. Taking F≡1, G≡0 on a probability space with s=p=α and r=2 gives LHS=2^{1/2} and RHS=2^{1-1/α}; for α<2 the inequality fails. The inequality is also non-homogeneous: the LHS scales as λ^{p/r} while the RHS scales as λ^{p/s}. Moreover, the parameter condition r′≤s≤p forces r≥α′=α/(α−1)>2 for α∈(1,2), so the final 'letting δ→0' in (14)–(16) is unavailable for fixed α. Thus the chain (11)–(16) cannot produce the claimed contradiction.","section":"§1, Lemma 3"},{"comment":"Theorem 1 is stated for a fixed infinite coefficient sequence (a_j), (c_j). Corollary 3 applies it to polynomials whose coefficients ε_n(k) depend on n. This is not covered by the theorem as stated; a diagonal argument or a uniform version for n-dependent coefficients is missing. The advertised consequences (L^α-Littlewood, Newman, Erdős) concern sequences of Littlewood polynomials with signs varying with n, so this gap is load-bearing for the abstract's main claim.","section":"§1, Corollary 3"},{"comment":"The displayed identity ‖ |P_m|−1 ‖_2^2 = 2 − ∫|P_m| dz is incorrect; the right-hand side should be 2 − 2∫|P_m| dz. More substantively, Lemma 9 produces a subsequence of stretched polynomials P_{j_k}(z^{l_k}) with a product finite and nonzero a.e., and Lemma 10 gives M(µ)=∏ M(P_{j_k})^2. Neither implies that M(P_m) → 1 for the original Gauss–Fresnel polynomials. The alternative proof of the Beller–Newman theorem is therefore incomplete.","section":"§5, proof of Theorem 3"},{"comment":"The bounds involving √N do not follow from Theorem 1 or Lemma 1 for the analytic polynomial norm. Littlewood's criterion gives information about the real cosine polynomial g_N, not |P_N|. For α<2, the inequality ‖g_N‖_α ≤ (1−A)‖g_N‖_2 does not imply ‖P_N‖_α ≤ (1−C)√N, since |P_N| ≥ |g_N|. The α>2 case similarly compares to √(N/2), not √N. The statement needs a justification or reformulation.","section":"§3, Corollary 5"}],"minor_comments":[{"comment":"Numerous typos and misspellings: 'Furstermore', 'generalizatin', 'anwser', 'unimodulair', 'polynomails', 'Frenesl' for 'Fresnel', 'wearker', 'Baker sequences' for 'Barker sequences', and 'conclude the proof' repeated. The paper would benefit from a careful proofreading.","section":"Abstract/§5"},{"comment":"The line “Littlewood's argument |P_n(z) − P_n(z′)| = ±{|P_n(z)|} ± {|P_n(z′)|}” is not meaningful as written and does not explain how Littlewood's method applies.","section":"§1, proof of Corollary 3"},{"comment":"The notation ζ2 is confusing and seems to denote a constant rather than a function of ζ; the bound “|ζ|,|ζ2| < BεA” is ambiguous. Consider restating with clearer constants.","section":"§1, Lemma 4"},{"comment":"“discret” should be “discrete”, and the statement that L^α-flatness implies ultraflatness (given in the introduction) is incorrect as written: L^α-flatness only gives convergence in L^α of |Q_n|−1, not uniform convergence.","section":"§2"}],"recommendation":"reject","confidential_remarks":"The paper is very short and makes sweeping claims. The core proof has a fundamental gap: the L^α→L^r convergence step is unjustified and false in general, and Lemma 3 as stated is not a valid inequality. In addition, Corollary 3 applies the main theorem outside its hypotheses. These are not presentation issues; the main result and its advertised consequences are not established. The paper would require a substantially new proof, not just local repairs, so I cannot recommend revision in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Brief take: the paper claims to settle a cluster of famous conjectures, and the intended generalization is real, but the proof of the main theorem has a fatal gap; the corollaries do not follow from the stated theorem. I wouldn't trust any of the big conclusions as they stand.\n\nWhat is genuinely new: Littlewood's classical criterion for real trigonometric polynomials is extended to analytic polynomials with arbitrary unimodular phases. That is a meaningful step beyond the author's earlier alpha >= 4 result and Erdelyi's recent ultraflatness theorem. The idea of using Bochner-space Clarkson inequalities is plausible, and the spectral theory application (singularity of Morse cocycle spectral type) would be a nice consequence if the main theorem were true.\n\nThe soft spots are not minor. In the proof of Theorem 1, the passage from (11) to (14) replaces the L^alpha norm with an L^r norm for r = 2 + delta > alpha without argument. L^alpha-flatness does not control L^r. Even worse, Lemma 3 as printed is false: take F = 1, G = 0 on a probability space, p = alpha in (1,2), r = 2, s = alpha. The left side is sqrt(2), the right side is 2^{1/alpha'} < sqrt(2). The inequality is also non-homogeneous. And the parameter condition 1 < s <= p <= r, r' <= s rules out r -> 2 for fixed alpha < 2, so the final 'let delta -> 0' is unavailable. The contradiction in (16) does not follow.\n\nThere are also structural issues. Corollary 3 claims that Littlewood polynomials (coefficients +/-1 depending on n) are not L^alpha-flat, but Theorem 1 is stated only for a fixed coefficient sequence. A diagonal argument is missing, so the Erdos-Newman and Barker consequences are unsupported. The Liouville corollary inherits the same problem. The paper is also littered with typos and inconsistent constants, but those are the least of it.\n\nBottom line: this is a serious but deeply flawed draft. The main theorem might be salvageable if a correct Clarkson-type inequality exists and if the L^alpha -> L^r step can be justified, but the manuscript as written does not prove what it claims. I would not send this to peer review; I'd desk-reject and write a detailed explanation of where the proof breaks down.","headline":"A serious attempt at a major result, but the main proof has a fatal gap: Lemma 3 is false and the L^alpha to L^r step is unjustified.","tokens_in":15166,"tokens_out":5427,"would_cite":false,"duration_ms":48947,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11C08","42A05","42A55","37A05","37A30","42A61"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that no normalized analytic polynomial sequence in Littlewood's coefficient class is L^p-flat for any p>0, settling the L^α-Littlewood, Newman, and Erdős conjectures.","keywords":["flat polynomials","ultraflat polynomials","Erdős–Littlewood problem","unimodular polynomials","Clarkson inequalities","Barker sequences","generalized Riesz products","singular spectrum"],"falsifier":"Take any fixed unimodular sequence c_j and any fixed real sequence a_j satisfying Σ a_m² ≤ K/n² Σ m² a_m², and compute the normalized L^p deviation (∫_{S¹} ||P_n(z)|-1|^p dz)^{1/p} for increasing n. The theorem predicts the limsup of this quantity is positive for every p>0; finding one admissible sequence with this deviation tending to zero for any p would falsify the central claim.","tokens_in":14084,"feed_emoji":"🔢","tokens_out":16384,"duration_ms":151399,"temperature":0.7,"pith_summary":"This paper proves that within a broad class of analytic polynomials on the unit circle—those whose coefficients satisfy Littlewood's growth condition, where the unweighted sum of squared coefficients is bounded by K/n² times the frequency-weighted sum—no sequence of normalized polynomials can be flat in any L^p sense for p>0. The proof takes Littlewood's classical norm-gap criterion for real trigonometric polynomials and transfers it to complex polynomials by splitting each polynomial into real and imaginary parts and applying a generalized Clarkson inequality for Banach-space-valued L^p functions. If correct, this settles the L^α-Littlewood conjecture, and with it Newman's L¹ conjecture and Erdős's L^∞ conjecture. It also implies that only finitely many Barker sequences exist, that certain Morse cocycles have singular spectra, and that Gauss–Fresnel polynomials contain Mahler-flat subsequences, giving a new proof of the Beller–Newman theorem.","feed_headline":"Flat polynomials ruled out for every exponent p>0","feed_subtitle":"A coefficient bound from Littlewood kills L^p-flatness and settles Newman's and Erdős's conjectures.","key_machinery":"The key mechanism is the Littlewood coefficient condition Σ a_m² ≤ K/n² Σ m² a_m², which controls the derivative of the trigonometric polynomial and yields a uniform gap between its L^α and L² norms. The proof represents the normalized analytic polynomial as f̃_n = (g̃_n + i h̃_n)/√2, where g̃_n and h̃_n are its real and imaginary parts, each of L² norm 1, and applies the generalized Clarkson second inequality for Banach-valued L^α spaces to the pair F = g̃_n/√2 and G = i h̃_n/√2. This convexity inequality controls the L^r norm of f̃_n in terms of the L^s norms of its parts; feeding in Littlewood's norm gap and letting the exponent r tend to 2 from above yields the contradiction that establi","core_discovery":"The paper's central claim is that flatness is impossible, at every exponent p>0, for normalized analytic polynomials P_n(z) = (Σ a_j c_j z^j)/(Σ a_j²)^{1/2} whenever the real coefficients a_j obey Σ a_m² ≤ K/n² Σ m² a_m² for an absolute constant K and the c_j are unimodular. Littlewood had established norm gaps for real cosine and sine polynomials under this condition; the new step is to transfer that gap to the modulus of the complex polynomial by writing its real and imaginary parts as two real polynomials of equal norm and applying a Clarkson-type inequality to the pair. The argument derives a contradiction: if the sequence were L^α-flat, the Clarkson inequality together with Littlewood's","pith_inferences":["The theorem's scope ends where coefficients depend on the degree: Newman's Gauss-sum polynomials and Kahane-type ultraflat constructions escape the coefficient condition precisely because their coefficients are not fixed in advance, so their flatness is not contradicted by this result.","Because Littlewood's criterion is a limsup statement, the proof should also rule out L^p-flatness for coefficient sequences satisfying the condition along sparse subsequences of degrees, not only for every n.","A quantitative refinement that makes the constant A(K,α) explicit would turn the theorem into effective lower bounds on the L^α deviation of finite Littlewood polynomials; the paper notes the constant is not yet computed.","The proof template—splitting a complex function into equal-norm real and imaginary parts and applying a uniform-convexity inequality—could be tested on random coefficient models satisfying the same weighted-square condition."],"forward_implications":["Littlewood polynomials with coefficients ±1 are not L^α-flat for any α>0, confirming the L^α-Littlewood conjecture.","Newman's L¹ conjecture and Erdős's L^∞ conjecture follow, since a flat sequence at either exponent would fall inside the forbidden class.","Uniform unimodular polynomials—fixed coefficient sequences with |c_j|=1—are never ultraflat.","There are only finitely many Barker sequences, because an infinite family would produce L²-flat Littlewood polynomials.","Morse cocycles with ±1 values over an odometer have singular maximal spectral type, resolving an old ergodic-theory question through Guenais's criterion."],"supporting_citations":[{"why":"Supplies the L^α-flatness criterion and the coefficient condition; Corollary 3 applies its 'Littlewood argument' to ±1 coefficients.","marker":"[28]"},{"why":"Gives the generalized Clarkson second inequality for Banach-valued L^α spaces, the convexity estimate that carries the proof of Theorem 1.","marker":"[12]"},{"why":"Shows singularity of the Morse-cocycle spectral type when the defining polynomials are not L¹-flat; used for Corollary 4.","marker":"[21]"},{"why":"Proves the length restrictions for Barker sequences (odd lengths at most 13 or n = 4m²), needed with Theorem 1 to rule out infinitely many.","marker":"[35]"},{"why":"Attributes the result that Barker sequences yield L²-flat Littlewood polynomials, bridging flatness to finiteness of Barker sequences.","marker":"[34]"},{"why":"Provides the proof of the flatness implication for Barker-sequence Littlewood polynomials used in Theorem 2.","marker":"[13]"},{"why":"Newman's L⁴ estimate for Gauss-sum polynomials supplies the L¹-flatness used in the Mahler-flat construction.","marker":"[31]"},{"why":"The Beller–Newman result that Theorem 3 reproves via Mahler-flat Gauss–Fresnel polynomials.","marker":"[11]"},{"why":"Provides the generalized Riesz product calculus and the theorem on Mahler measures used to conclude Mahler-flatness.","marker":"[4]"}],"fun_headline_variants":["No L^p-flat polynomials for any p>0","Littlewood and Clarkson rule out flatness for all p","Flat polynomials impossible for every exponent p>0","Littlewood's condition kills flatness at every p","Settling conjectures: no flat polynomials for any p"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The proof depends on the assumption that closeness to 1 in one average sense implies closeness to 1 in a slightly stronger average sense; this is false in general because large values on very small sets can inflate the stronger average without affecting the weaker one.","fun_headline_variants_meta":{"raw":{"variants":["No L^p-flat polynomials for any p>0","Littlewood and Clarkson rule out flatness for all p","Flat polynomials impossible for every exponent p>0","Littlewood's condition kills flatness at every p","Settling conjectures: no flat polynomials for any p"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000227,"raw_usage":{"total_tokens":1340,"prompt_tokens":805,"completion_tokens":535,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":549,"completion_tokens_details":{"reasoning_tokens":454}},"tokens_in":549,"tokens_out":535,"duration_ms":5277,"temperature":1.0,"reasoning_tokens":454,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:21:24.908883+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any fixed unimodular sequence c_j and any fixed real sequence a_j satisfying Σ a_m² ≤ K/n² Σ m² a_m², and compute the normalized L^p deviation (∫_{S¹} ||P_n(z)|-1|^p dz)^{1/p} for increasing n. The theorem predicts the limsup of this quantity is positive for every p>0; finding one admissible sequence with this deviation tending to zero for any p would falsify the central claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the L^α-flatness criterion and the coefficient condition; Corollary 3 applies its 'Littlewood argument' to ±1 coefficients."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the generalized Clarkson second inequality for Banach-valued L^α spaces, the convexity estimate that carries the proof of Theorem 1."},{"cited_title":"Guenais, Morse cocycles and simple Lebesgue spectrum, Ergodic Theory Dynam","cited_arxiv_id":null,"evidence_quote":"Shows singularity of the Morse-cocycle spectral type when the defining polynomials are not L¹-flat; used for Corollary 4."},{"cited_title":"Turyn and J","cited_arxiv_id":null,"evidence_quote":"Proves the length restrictions for Barker sequences (odd lengths at most 13 or n = 4m²), needed with Theorem 1 to rule out infinitely many."},{"cited_title":"Saffari, Barker sequences and Littlewood two-sided conjectures on polyno- mials with ±1 coefficients, S´ eminaire d’Analyse Harmonique, Ann´ ee 1989/90, 139–151, Univ","cited_arxiv_id":null,"evidence_quote":"Attributes the result that Barker sequences yield L²-flat Littlewood polynomials, bridging flatness to finiteness of Barker sequences."},{"cited_title":"Borwein and M","cited_arxiv_id":null,"evidence_quote":"Provides the proof of the flatness implication for Barker-sequence Littlewood polynomials used in Theorem 2."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Newman's L⁴ estimate for Gauss-sum polynomials supplies the L¹-flatness used in the Mahler-flat construction."},{"cited_title":"Beller and D","cited_arxiv_id":null,"evidence_quote":"The Beller–Newman result that Theorem 3 reproves via Mahler-flat Gauss–Fresnel polynomials."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the generalized Riesz product calculus and the theorem on Mahler measures used to conclude Mahler-flatness."}],"review_version":1}