{"id":"30728ca9-71ae-4ce7-89f3-e33af5b92423","arxiv_id":"2501.03157","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A new Wronskian-based proof shows that if a one-dimensional Schrödinger potential has almost repetitions with rate γ > L(E), the operator has no eigenvalues.","lead":"A known theorem about one-dimensional Schrödinger operators with almost repeating potentials receives a new, shorter proof. The theorem says such operators have no bound states when the repetition is strong relative to the growth of solutions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The inference (17)→(18) is invalid for complex u: (17) does not imply the pair (u(0),u(-1)) and (-u(-1),u(0)) is a basis, so b_n need not be O(e^{-γ k_n}); a shift of origin or real-eigenfunction reduction is required.","rationale":"After reading the proof carefully, the most load-bearing step is not the unproved Lemma 2.1 but the linear-algebra extraction of b_n. Lemma 2.1 is a standard continuity estimate for transfer matrices: a telescoping proof with (8) gives the needed bound up to an extra factor |m-j|, and in the only application |m-j|=k_n, so the error remains O(e^{-(γ-L-ε')k_n}) for suitable ε'. It is not a correctness risk once proved. By contrast, the step (15)+(16)+(17)⇒(18) is logically invalid for complex-valued eigenfunctions: (17) does not control the determinant denominator. Since the theorem is stated for general bounded V with no reality assumption, the proof as written does not establish the claimed generality. However, the gap is easily repaired by shifting the origin, and the theorem itself is known; the reader's CONDITIONAL verdict already asks for a normalization fix at (17). I therefore do not change the verdict, but I would rank the origin-normalization issue above the Lemma 2.1 citation as the condition to address first.","tokens_in":5301,"tokens_out":26842,"duration_ms":229338,"concrete_test":"Write b_n = det([[u(0),u(k_n)],[u(-1),u(k_n-1)]])/(u(0)^2+u(-1)^2) and test the proof's assertion with u(0)=1,u(-1)=i: the denominator is zero, so (15) yields no O(e^{-γ k_n}) bound on b_n. To settle whether this case is reachable in the theorem's hypotheses, try to construct a bounded potential V (e.g., finitely supported or periodic) admitting an ℓ2 eigenfunction with u(0)=1,u(-1)=i and check whether the γ-repetition condition (2) can hold for some k_n with γ>L(E). Alternatively, verify that shifting the origin to a k0 with u(k0)^2+u(k0-1)^2≠0 (which exists for any nonzero ℓ2 sequence) restores (18) and (19) without changing (2) or L(E).","verdict_should_be":"UNCHANGED","load_bearing_attack":"At (16) the proof decomposes x=(u(k_n),u(k_n-1))^T as a_n v1 + b_n v2 with v1=(u(0),u(-1)) and v2=(-u(-1),u(0)). The coefficient b_n is obtained by Cramer's rule: b_n = det(v1,x)/(u(0)^2+u(-1)^2). Equation (15) bounds |det(v1,x)| by e^{-γ k_n}, so to get (18) one needs the denominator to be bounded away from zero. But (17) gives only |u(0)|^2+|u(-1)|^2>0, which is strictly weaker: for u(0)=1, u(-1)=i both hold while u(0)^2+u(-1)^2=0. In that case v1 and v2 are linearly dependent, the representation (16) is not unique, and neither b_n=O(e^{-γ k_n}) nor a_n=o(1) is justified. Since (22) uses b_n=O(e^{-γ k_n}) to discard the term b_n T_{0,k_n} v2 as O(e^{-(γ-L-ε)k_n}), the contradiction at (24) does not follow. This is not merely a citation gap; it is a false inference in the text. The fix is short: a nonzero ℓ2 solution cannot satisfy u(k)^2+u(k-1)^2=0 for all k (this forces |u(k)| constant), so one may shift the origin to a site where u(0)^2+u(-1)^2≠0; the repetition condition (2) and L(E) are invariant under such a shift. Alternatively, for real potentials one first reduces to a real eigenfunction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a proof of a sharp Gordon-type lemma for one-dimensional discrete Schrödinger operators with γ-repetition potentials, claiming that the eigen-equation Hu = Eu has no ℓ2(Z) solutions whenever γ > L(E), where L(E) is the transfer-matrix growth rate defined in equation (4). The proof combines a Wronskian estimate (equations (12)–(15)) with a transfer-matrix comparison lemma (Lemma 2.1) to derive a contradiction from an assumed ℓ2 eigenfunction. The paper then applies the theorem to Hölder-continuous quasi-periodic potentials and recovers the sharp almost Mathieu transition of Avila–You–Zhou.","tokens_in":5633,"tokens_out":5980,"duration_ms":54212,"significance":"If made fully rigorous, the paper would provide a concise new proof of a sharp Gordon lemma that is of genuine interest in the spectral theory of one-dimensional operators. The Wronskian mechanism leading to the exponential decay in (15) is elegant, and the overall structure is transparent. The corollaries for quasi-periodic and almost Mathieu operators are valuable applications. However, the current manuscript contains a load-bearing gap in the complex-eigenfunction case and an omitted proof of a key lemma, so the central claim is not yet fully supported.","major_comments":[{"comment":"The passage from (15) and (17) to (18) is not justified for complex-valued eigenfunctions. Solving (16) by Cramer's rule gives b_n = det((u(0),u(-1))^T, (u(k_n),u(k_n-1))^T) / (u(0)^2 + u(-1)^2), so the bound (15) only yields |b_n| ≤ e^{-γ k_n} / |u(0)^2 + u(-1)^2|. Condition (17) ensures the denominator is nonzero, but not that it is bounded away from zero; for instance u(0)=1, u(-1)=i satisfies (17) while u(0)^2 + u(-1)^2 = 0, in which case the two vectors in (16) are linearly dependent and the representation is not unique. Consequently, (18), and hence the estimate (22) and the final contradiction (24), do not follow as written. The fix is to shift the origin to a site with u(k)^2 + u(k-1)^2 ≠ 0: if no such site exists, then u(k)^2 + u(k-1)^2 = 0 for all k, which forces |u(k)| to be constant, contradicting u ∈ ℓ2(Z) unless u ≡ 0. Condition (2) and L(E) are invariant under such a translation. Alternatively, for real potentials one can first reduce to a real eigenfunction. This step must be included in the revised proof.","section":"§2, Lemma 2.1"},{"comment":"Lemma 2.1 is load-bearing: it is used at equation (23) to replace T_{0,k_n} by T_{-k_n,0} and thereby produce the contradiction (24). Yet the proof is not carried out; the text merely says it follows from (8) by “standard telescoping” and cites [28, Lemma 4.5]. Moreover, as stated the lemma gives a comparison with a shift by +k_n, whereas (23) uses a shift by −k_n. The argument should either state the lemma for shifts ±k_n (which the telescoping supplies using (2) for all k) or explain the reduction, and a complete proof should be included. If the uniformity in m,j or the exact constant C is not as stated, the exponential decay in (23) is unsupported.","section":"§2, Lemma 2.1"}],"minor_comments":[{"comment":"The running title on page 1 contains a typo: “EIGENV ALUES” should be “EIGENVALUES”.","section":"Title page"},{"comment":"The matrix norm in Lemma 2.1 and elsewhere should be specified explicitly as the operator norm, and the dependence of the constant C on E and V should be stated.","section":"§2, Lemma 2.1 and throughout"},{"comment":"The sentence “Notice that ∑_k |W(u, un)(k)| ≤ 2” should specify that the sum is over all k ∈ Z and indicate that the bound follows from Cauchy–Schwarz applied to the two products u(k+1)un(k) and u(k)un(k+1) separately.","section":"§2, Eq. (13)"},{"comment":"The orientation convention for the transfer matrices T_{m,j} when j > m is left implicit; stating the inverse convention explicitly would remove ambiguity in equations (20) and (21).","section":"§1, Eq. (4) and §2, Eq. (20)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is very short and the main theorem, while significant, overlaps with results in the existing literature; the value lies in the new proof. The two gaps I identified are local and fixable, and I see no indication of circularity or parameter fitting. I recommend major revision rather than rejection because the central strategy is sound. The editor may wish to require a full proof of Lemma 2.1 in the revised version rather than a citation, and to ask the author to address the complex-eigenfunction issue with the translation argument explicitly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Wencai Liu's paper is a new proof of a theorem that already exists. The theorem — no ℓ2 solutions when γ > L(E) for almost repetition potentials — was proved by Avila-You-Zhou, and the reflective analog by Jitomirskaya-Liu. The paper says so in the introduction, which is the right and honest way to position a proof paper. What is new is the method: a Wronskian estimate plus transfer-matrix comparison that yields the sharp threshold. That is a genuine contribution, and the proof is demonstrably different from the earlier ones. The Wronskian bound (12)–(14) and the coefficient bounds (18)–(19) are the core, and they are correct once one has the proper setup.\n\nThere is one real gap, and it is not merely a citation issue. Equation (17) only gives |u(0)|²+|u(-1)|²>0, but the decomposition (16) requires the denominator u(0)²+u(-1)² in Cramer's rule to be nonzero. For complex eigenfunctions, (17) does not imply that; u(0)=1, u(-1)=i is a concrete counterexample. The stress-test note is right that this is a false inference, not a missing citation. The fix is short: a nonzero ℓ² solution cannot have u(k)²+u(k−1)²=0 for all k without forcing |u(k)| constant, so a shift of origin puts the denominator away from zero. The repetition condition and L(E) are shift-invariant, so this is harmless. Alternatively, for real potentials one first reduces to a real eigenfunction. But as written, the proof has a load-bearing step that needs repair.\n\nMinor point: Lemma 2.1 is cited to [28, Lemma 4.5] rather than proved. That is standard telescoping, so acceptable in principle, but the paper should state the precise constants and the induction. It would take half a page. The citation pattern elsewhere is fine: the paper credits [1] and [19] for the known results and does not overclaim.\n\nOverall: the contribution is methodological, not a new spectral fact. It deserves a serious referee. With the basis issue fixed and Lemma 2.1 given a proper statement, this is a solid paper. I would not desk-reject it.","headline":"New Wronskian proof of a known sharp Gordon lemma; the method is real, but one linear-algebra step needs a fix before the proof is valid as written.","tokens_in":6239,"tokens_out":2226,"would_cite":true,"duration_ms":20141,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q10","47B39","39A70"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that a bounded potential with $\\gamma$-repetitions and repetition rate $\\gamma$ greater than the transfer-matrix growth rate $L(E)$ has no square-summable eigenfunctions, via a new Wronskian–transfer-matrix argument.","keywords":["discrete Schrödinger operator","almost repetition potentials","sharp repetition lemma","absence of eigenvalues","transfer matrices","Lyapunov exponent","almost Mathieu operator"],"falsifier":"For a concrete almost-repetition potential, for example an almost Mathieu-type potential with $\\beta(\\alpha) > 0$ and $|\\lambda| < e^{\\beta(\\alpha)}$, compute $\\|T_{0,k_n}(E) - T_{-k_n,0}(E)\\|$ and check whether it is bounded by $C e^{-\\gamma k_n} e^{(L(E)+\\varepsilon)|m-j|}$; exceeding that bound would break the proof's key comparison, while finding a nonzero $\\ell^2$ solution for such parameters would refute Theorem 1.1 directly.","tokens_in":5039,"feed_emoji":"⚫️","tokens_out":14754,"duration_ms":115112,"temperature":0.7,"pith_summary":"This paper supplies a new proof of the sharp repetition lemma for one-dimensional discrete Schrödinger operators: a bounded potential that almost repeats itself at distances $k_n$, with error at most $e^{-\\gamma|k_n|}$, cannot support an $\\ell^2(\\mathbb{Z})$ eigenfunction once the repetition rate $\\gamma$ exceeds the transfer-matrix growth rate $L(E)$. Earlier repetition arguments proved this only with a factor 2, and the sharp threshold was previously reached by different analytic methods; here the threshold is recovered by adapting the Wronskian strategy used for reflective repetition potentials. The result gives a sharp criterion for absence of point spectrum, and it applies directly to quasi-periodic potentials, yielding the arithmetic delocalization criterion for the almost Mathieu operator when the frequency has positive $\\beta(\\alpha)$. A reader should care because it isolates the exact mechanism by which almost-repetition kills eigenvalues: the Wronskian stays exponentially small while the shifted solution pair itself decays, forcing a contradiction.","feed_headline":"When repetition outruns matrix growth, eigenstates vanish","feed_subtitle":"A Wronskian–transfer-matrix proof shows ℓ² solutions cannot exist at rates above the sharp threshold.","key_machinery":"The central machinery is the Wronskian $W(f, g)(k) = f(k+1)g(k) - f(k)g(k+1)$ together with the multi-step transfer matrices $T_{m,j}(E)$ that evolve solution vectors $(u(k), u(k-1))$ across an interval. The argument first shows, by summing the potential differences and using the $\\ell^2$ normalization of $u$ and $u_n$, that $|W(u, u_n)(-1)| \\leq e^{-\\gamma k_n}$; then it expands $(u(k_n), u(k_n - 1)) = a_n (u(0), u(-1)) + b_n (-u(-1), u(0))$, so that $b_n$ is exponentially small and $a_n$ tends to zero. Multiplying by $T_{0,k_n}$ and replacing it by $T_{-k_n,0}$ via Lemma 2.1, the comparison $\\|T_{m,j}(E) - T_{m+k_n,j+k_n}(E)\\| \\leq C e^{-\\gamma k_n} e^{(L(E)+\\varepsilon)|m-j|}$, turns the expansion into the vector $(u(-k_n), u(-k_n - 1))$, which is $o(1)$, contradicting the nonzero left-hand side $(u(0), u(-1))$.","core_discovery":"The paper establishes Theorem 1.1: if $V$ is a bounded potential satisfying $|V(k_n + k) - V(k)| \\leq e^{-\\gamma|k_n|}$ for a sequence $k_n \\to \\pm\\infty$, and if $\\gamma > L(E)$, then the eigen-equation $Hu = Eu$ has no $\\ell^2(\\mathbb{Z})$ solutions. The proof is new: instead of the classical three-block repetition argument, it compares a candidate eigenfunction $u$ with its shifted copy $u_n(k) = u(k_n + k)$ through the Wronskian $W(u, u_n)$, uses the exponential closeness of $V$ and $V_n$ to bound $W$ by $e^{-\\gamma k_n}$, and then expands the shifted vector $(u(k_n), u(k_n - 1))$ in a two-vector basis. The expansion coefficients satisfy $b_n = O(e^{-\\gamma k_n})$ and $a_n = o(1)$, while a transfer-matrix comparison shows that applying $T_{0,k_n}$ yields $(u(-k_n), u(-k_n - 1))$, which tends to zero; this contradicts the fixed nonzero vector $(u(0), u(-1))$. Corollaries follow for $\\kappa$-Hölder continuous quasi-periodic potentials (no eigenvalues when $L(E) < \\kappa\\beta(\\alpha)$) and for the almost Mathieu operator (no eigenvalues when $|\\lambda| < e^{\\beta(\\alpha)}$ given $\\beta(\\alpha) > 0$).","pith_inferences":["The same Wronskian–expansion argument is likely to extend to block Jacobi or CMV matrices, since only the transfer-matrix comparison and the $\\ell^2$ normalization are used; verifying the analogous lemma would be a concrete test.","Because the proof tracks convergence only up to $o(1)$, a natural extension is a quantitative version that records the exponential decay rates of the coefficients $a_n$ and $b_n$, yielding explicit bounds on any hypothetical eigenfunction.","The proof's key estimate, Lemma 2.1, is cited to earlier work rather than derived in full; making that telescoping argument explicit would make the paper self-contained and would clarify the role of the constant $C$."],"forward_implications":["For every bounded $\\gamma$-repetition potential, the set of energies with $\\gamma > L(E)$ is disjoint from the point spectrum; on that set the operator has purely continuous spectrum.","For $\\kappa$-Hölder continuous quasi-periodic potentials, $H$ has no eigenvalues in the regime $\\{E : L(E) < \\kappa\\beta(\\alpha)\\}$.","For the almost Mathieu operator with $\\beta(\\alpha) > 0$, there are no eigenvalues whenever $|\\lambda| < e^{\\beta(\\alpha)}$, which is the sharp delocalization criterion.","The threshold $\\gamma > L(E)$ is optimal: the almost Mathieu example shows eigenvalues can appear when the inequality is not satisfied, so the factor 2 in the earlier repetition argument is not an artifact."],"supporting_citations":[{"why":"Supplies the telescoping transfer-matrix comparison used as Lemma 2.1, the estimate the final contradiction depends on.","marker":"[28]"},{"why":"Establishes the predecessor theorem for reflective repetition potentials and introduces the Wronskian-contraction style this proof adapts.","marker":"[22]"},{"why":"Reduces the reflective-repetition threshold to $\\gamma > L(E)$ and provides the coefficient-decomposition idea reused in this paper.","marker":"[19]"},{"why":"Proves the sharp phase transition for the almost Mathieu operator and shows the $\\gamma > L(E)$ regime is optimal.","marker":"[1]"},{"why":"Provides the Lyapunov exponent formula $L(E) = \\max\\{\\ln|\\lambda|, 0\\}$ needed to derive the almost Mathieu corollary.","marker":"[3]"}],"fun_headline_variants":["Sharp Gordon's lemma: no eigenvalues for almost repetition potentials","New proof: sharp repetition threshold means no eigenstates","Wronskian proof: no ℓ² solutions at sharp repetition rate","Almost repetition potentials have no eigenvalues, sharp proof","Transfer-matrix proof: no eigenstates at sharp rate"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on the estimate, cited to another paper rather than proved here, that shifting a long block of the potential by the almost-repetition distance changes the associated transfer matrices by at most $C e^{-\\gamma k_n} e^{(L(E)+\\varepsilon)|m-j|}$; if that comparison is not exponentially small, the final contradiction collapses.","fun_headline_variants_meta":{"raw":{"variants":["Sharp Gordon's lemma: no eigenvalues for almost repetition potentials","New proof: sharp repetition threshold means no eigenstates","Wronskian proof: no ℓ² solutions at sharp repetition rate","Almost repetition potentials have no eigenvalues, sharp proof","Transfer-matrix proof: no eigenstates at sharp rate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000979,"raw_usage":{"total_tokens":4147,"prompt_tokens":927,"completion_tokens":3220,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":543,"completion_tokens_details":{"reasoning_tokens":3138}},"tokens_in":543,"tokens_out":3220,"duration_ms":52445,"temperature":1.0,"reasoning_tokens":3138,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:54:06.965751+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a concrete almost-repetition potential, for example an almost Mathieu-type potential with $\\beta(\\alpha) > 0$ and $|\\lambda| < e^{\\beta(\\alpha)}$, compute $\\|T_{0,k_n}(E) - T_{-k_n,0}(E)\\|$ and check whether it is bounded by $C e^{-\\gamma k_n} e^{(L(E)+\\varepsilon)|m-j|}$; exceeding that bound would break the proof's key comparison, while finding a nonzero $\\ell^2$ solution for such parameters would refute Theorem 1.1 directly.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the telescoping transfer-matrix comparison used as Lemma 2.1, the estimate the final contradiction depends on."},{"cited_title":"Jitomirskaya and B","cited_arxiv_id":null,"evidence_quote":"Establishes the predecessor theorem for reflective repetition potentials and introduces the Wronskian-contraction style this proof adapts."},{"cited_title":"Jitomirskaya and W","cited_arxiv_id":null,"evidence_quote":"Reduces the reflective-repetition threshold to $\\gamma > L(E)$ and provides the coefficient-decomposition idea reused in this paper."}],"review_version":1}