{"id":"b2d66a60-49b9-4522-8483-60f7bd1e39a5","arxiv_id":"2608.11602","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every n≥3, non-trivial solutions on the simplex lattice Δ_N^{(n)} and on Z^n have at least c N^{⌈n/2⌉} (resp. c L^{⌈n/2⌉}/log L) sites where the value is not exponentially small.","lead":"This mathematics paper proves that functions on high-dimensional lattice grids that solve a discrete Schrödinger-type equation cannot be tiny almost everywhere: in every dimension n≥3, they must be non-negligible on a large set of lattice points. This settles a conjecture on unique continuation up to a logarithmic factor, although the exponents fall short of what the Anderson-Bernoulli localization proof would need.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.4 operator-norm equality is incorrect, so the stability bound (6.40) is not justified as written; the error is repairable and likely does not affect the main theorem.","rationale":"The reader identified the positivity of proper Pascal minors, cited from [Li26a], as the weakest assumption. That is indeed an external input, but it is a standard true combinatorial fact and the manuscript's own shifted-Schur machinery supports it. My stress-test found a more concrete internal error: Theorem 6.4's claimed operator norm is computed incorrectly. The integral in (6.31) evaluates to [(2-2/n)^N - 1]/(n-2), not (2-2/n)^{N-1}/(n-2), so the stated K_{n,N} is not an upper bound and the displayed inequality (6.40) is false as written. This is load-bearing because the stability proof of Theorem 2.2 uses exactly that inequality to control the perturbation P and to choose the final constant C_n. I checked that the corrected expression still has the form exp(O(N)) with base 2-2/n < 2 < e, so a larger C_n absorbs it; hence the main result is very likely correct after this fix. I found no fatal flaw in the quantitative Pascal UP, the shell estimates, the Walsh reduction, or the Voronoi pigeonholing; those parts are coherent and mutually consistent. The paper's own limitation statement about the localization application is honest and does not affect the QUC claim. The verdict should remain CONDITIONAL: the proof needs a corrected operator-norm computation and other minor fixes, but the central argument appears sound.","tokens_in":31936,"tokens_out":42727,"duration_ms":431940,"concrete_test":"Recompute the operator-norm bound in Theorem 6.4: for n=3,N=3 evaluate N/n * integral_0^1 (1+(n-2)t/n)^{N-1} dt and compare with (2-2/n)^{N-1}/(n-2); then redo (6.40) with the corrected constant K'_{n,N} = [(2-2/n)^N - 1]/(n-2) and verify that choosing C_n > log(2-2/n) + O(1) restores |h(a)| >= |g(a)|/2 and the subsequent chain of inequalities in Theorem 2.2. If the corrected bound holds, the stability argument goes through.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The stability passage from the exact-recurrence theorem to Theorem 2.2 rests on Theorem 6.4, which claims the exact operator norm of C_{n,N} is (2-2/n)^{N-1}/(n-2). This equality is false. From (6.29)-(6.31), with a=(n-2)/n, the integral is N/n * integral_0^1 (1+at)^{N-1} dt = [(1+a)^N - 1]/(n-2) = [(2-2/n)^N - 1]/(n-2), not (2-2/n)^{N-1}/(n-2). For n=N=3, the claimed value is 16/9, while the computation gives 37/27. Thus the displayed K_{n,N} is not an upper bound, and the inequality sup|p(α)| <= K_{n,N} e^{-C_n N}|g(a)| in (6.40) is not justified. Since (6.40) is used to ensure |h(a)| >= |g(a)|/2 and to pass from h to g in (6.41)-(6.43), the proof of Theorem 2.2 has a genuine gap at this point. The gap is repairable: replacing K_{n,N} by [(2-2/n)^N - 1]/(n-2) still gives an exp(O(N)) bound, which can be absorbed by taking C_n sufficiently large. The central claim is therefore very likely unaffected, but the manuscript must be corrected.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves quantitative unique continuation (QUC) on the n-dimensional simplex lattice Δ_N^{(n)} and on Z^n for every n≥3. Theorem 2.2 gives a lower bound c_n Φ_n(a) on the number of points at which a function satisfying the approximate recurrence (2.4) is not too small; Theorem 2.4 transfers this to the Schrödinger equation Δu=Vu in a cube, with decay exponent ⌈n/2⌉+1 and cardinality exponent ⌈n/2⌉ up to a logarithmic factor. The proof chain is: (i) a polynomial dictionary identifying the exact recurrence with DF=0; (ii) a quantitative Pascal uncertainty principle built from shifted Schur polynomial interpolation and proper Pascal minors; (iii) shell estimates on the simplex; (iv) a stability argument using the projection T_{A_1} onto ker D; and (v) a Walsh decomposition of ℓ1-spheres together with a Voronoi double-counting argument for Z^n. The overall architecture is coherent and the chain is checkable, but one load-bearing operator-norm computation in Step (iv) is incorrect as stated.","tokens_in":32187,"tokens_out":29101,"duration_ms":280168,"significance":"If the corrected operator-norm estimate is substituted, the paper's main theorems appear to follow, giving the first QUC on the simplex and on Z^n for arbitrary n≥3, extending [LZ22] and answering [Li26b, Conjecture 1.2] up to a (log L)^{-1} factor. The quantitative Pascal uncertainty principle and the Walsh/Voronoi reduction are useful new tools. The paper is explicit about the exponents and honestly notes that they fall short of the Anderson-Bernoulli localization threshold. The proof is essentially self-contained, apart from an imported determinant-positivity fact from [Li26a] that is used in a load-bearing way.","major_comments":[{"comment":"The claimed operator norm equality is false. With a=(n-2)/n, the integral in (6.31) evaluates to N/n ∫_0^1 (1+at)^{N-1} dt = [(1+a)^N -1]/ (n-2) = [(2-2/n)^N -1]/(n-2), not (2-2/n)^{N-1}/(n-2). The lower-bound computation in (6.36) gives the same corrected value, so the displayed K_{n,N} in (6.27) is not the operator norm. Consequently the inequality (6.40), which uses this K_{n,N}, is not justified as written, and the passage from the exact-recurrence function h to the original function g in (6.41)-(6.43) has a genuine gap. The error is repairable: replacing K_{n,N} by [(2-2/n)^N -1]/(n-2) gives the same exp(O_n(N)) behavior, and enlarging C_n in (6.40) restores the proof. The manuscript must correct this computation and all statements that depend on it.","section":"§6.2, Theorem 6.4, Eqs. (6.31) and (6.36)"},{"comment":"The proof relies on the strict positivity of proper Pascal minors, imported from [Li26a, Lemma 4.1] without statement or proof. This fact is load-bearing: it gives det P ≥ 1, which is used to define the interpolating polynomial via Cramer's rule and to control |F(t)| in (4.6), and the bounds then propagate to Theorem 4.4, Lemma 4.5, Lemma 5.1, and the main theorems. The manuscript should either prove this determinant positivity in an appendix or state it explicitly as an imported lemma, rather than referring to it inside a proof with no statement of the result.","section":"§4.1, Lemma 4.3 and Eq. (4.7)"}],"minor_comments":[{"comment":"The sentence 'by Lemma 6.3, (2.8) is equivalent to DF=0' should refer to Lemma 3.1, which is the lemma that establishes this equivalence.","section":"§5, first paragraph"},{"comment":"The line 'By Lemma , assumption (2.4) is exactly...' has a missing lemma reference; it should cite Lemma 3.1.","section":"§6.3, proof of Theorem 2.2"},{"comment":"In the n=2 remark, the range '0≤j≤2R' is undefined; it should be '0≤j≤N'.","section":"§2.1, Remark 2.1"},{"comment":"The text says 'QUC for stationary Schrödinger equation on Z^d'; the symbol should be Z^n for consistency with the rest of the paper.","section":"§7, first paragraph"},{"comment":"In the definition of the reversal matrix J_q, the range '1≤i,j≤1' should be '1≤i,j≤q'.","section":"§4.1, proof of Lemma 4.3"},{"comment":"The paper contains numerous typos, including 'equavalent', 'fucntion', 'knowlegdes', 'pespective', 'aslo', and 'descr'; a careful proofreading pass is needed.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The mathematical defect I found is concentrated in Theorem 6.4 and is clearly repairable by replacing K_{n,N} with [(2-2/n)^N -1]/(n-2). I recommend asking the authors to correct that computation and to make the dependency on [Li26a, Lemma 4.1] explicit, either by proof or by a stated imported lemma. The rest of the proof chain appears sound on my reading."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is a serious and largely correct extension of quantitative unique continuation: simplex and Z^n, all n≥3, Li's conjecture up to a log factor. Second, the stress-test note is right: Theorem 6.4's operator norm equality is wrong, but the fix is easy and the main theorem survives.\n\nWhat is actually new: the quantitative Pascal uncertainty principle built on shifted Schur interpolation, the stability theorem converting approximate recurrences to exact ones via projection onto ker D, and the Walsh decomposition of the ℓ1-sphere. These are real tools, not cosmetic repackaging. The Z^n result genuinely generalizes Li-Zhang from n=3 to all n, with the expected sharp cardinality up to log. The authors also state clearly that their exponents a=⌈n/2⌉+1, b=⌈n/2⌉ are outside what Anderson-Bernoulli localization would need; that limitation is handled honestly.\n\nThe main flaw is in Section 6. From (6.29)-(6.31), the bound should be N/n ∫_0^1 (1+(n-2)t/n)^{N-1} dt = ((2-2/n)^N - 1)/(n-2), not (2-2/n)^{N-1}/(n-2). The equality construction in (6.35)-(6.36) has the same error. So Theorem 6.4 as stated is false. The correction is still exp(O(N)), so (6.40) and the passage to Theorem 2.2 go through after enlarging C_n. This is a real gap in the written proof, but a repairable one. I would lower the reader's soundness score from 8 to 7 until it is fixed.\n\nThe other problems are mechanical. Section 6.3 has a blank \"By Lemma\" cross-reference. Section 5 cites Lemma 6.3 where Lemma 3.1 is intended. Remark 2.2 asserts the complex-valued case without proof; likely true but not demonstrated. The chain depends on determinant positivity of proper Pascal minors from [Li26a]; that is a separate combinatorial fact and I see no reason to doubt it. Citations to Li, Li-Zhang, and the authors' own related work are appropriate, not padding. No data claims to check.\n\nBottom line: send it to referees. Ask them to fix Theorem 6.4 and clean up the cross-references. The core architecture is coherent and the result is worth having.","headline":"Solid high-dimensional QUC proof with a repairable operator-norm error; deserves refereeing, not desk reject.","tokens_in":32812,"tokens_out":4592,"would_cite":true,"duration_ms":48842,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B60","35J10","39A12","05A10","05E05"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every dimension $n\\ge 3$, nonzero solutions of the discrete Schr\\\"odinger equation with bounded potential on $\\mathbb{Z}^n$ keep values within an exponential factor of their maximum on at least $c_n L^{\\lceil n/2\\rceil}/\\log L$ points…","keywords":["quantitative unique continuation","simplex lattice","discrete Schr\\\"odinger equation","Pascal uncertainty principle","Pascal matrix","shifted Schur polynomials","Walsh decomposition","lattice cube"],"falsifier":"Compute the determinant of a proper Pascal minor, for example the $3\\times 3$ minor with rows $r=(2,4,6)$ and columns $k=(0,2,6)$: if any proper minor has determinant $0$ or a non-integer less than $1$, the bound (4.6) in Lemma 4.3 fails and the proof from Lemma 4.5 onward no longer holds. Independently, on $\\mathbb{Z}^3$ take a non-constant harmonic polynomial ($V\\equiv 0$) with $u(0)\\neq 0$, evaluate it on $Q_L$ for large $L$, and count the points where $|u(x)|\\ge \\exp(-C L^3/\\log L)|u(0)|$; a count below a positive constant times $L^2/\\log L$ would refute Theorem 2.4's conclusion.","tokens_in":31645,"feed_emoji":"📐","tokens_out":12732,"duration_ms":136513,"temperature":0.7,"pith_summary":"The paper proves a quantitative unique continuation theorem for the discrete Schr\\\"odinger equation on $\\mathbb{Z}^n$, for every $n\\ge 3$: if $u$ is a solution with bounded potential and $u(0)\\neq 0$, then within the cube of side $L$ there are at least $c_n L^{\\lceil n/2\\rceil}/\\log L$ points where $|u|$ stays above $\\exp(-C_{n,K}L^{\\lceil n/2\\rceil+1}/\\log L)|u(0)|$. It proves the analogous quantitative statement on the $n$-dimensional simplex lattice, where the number of large points is measured by the weight $\\Phi_n(a)=(a_n+1)\\prod_{i=1}^{n-2}(a_i+1)/(N+1)^{\\lfloor(n-2)/2\\rfloor}$. The engine is a quantitative Pascal uncertainty principle: a polynomial with a nonzero coefficient has many normalized coefficients, in itself and in its one-step shift, that cannot be exponentially small. A Walsh decomposition of the $\\ell^1$-sphere turns the lattice problem into the simplex problem. These estimates are sharp in their cardinality up to a logarithm, and they are the stated first step toward Anderson\\--Bernoulli localization in high dimensions.","feed_headline":"Lattice Schr\\\"odinger solutions keep large values in dimension n≥3","feed_subtitle":"In a side-L cube, many points stay above an exponentially small threshold; the count is sharp up to a log factor.","key_machinery":"The load-bearing object is the quantitative Pascal uncertainty principle (Lemma 4.6 and Corollary 4.1): for a $d$-variable polynomial $p$ of total degree at most $L$ with a nonzero coefficient at exponent $\\omega$, the total number of pairs $(S,\\nu)$, $S$ a subset of coordinates and $\\nu$ an exponent, for which the normalized coefficients of $p(x+\\chi_S)$ are at least $\\delta_L^d$ times the reference coefficient is at least $\\prod_{i=1}^d(\\omega_i+2)$. The proof constructs, from proper Pascal minors and shifted Schur polynomials, a one-variable interpolation polynomial $F$ that vanishes on the large coefficients of $p$ and $p(x+1)$; Cramer's rule and the positivity of proper Pascal minors (determinant at least $1$) control $|F|$. A Walsh decomposition of the $\\ell^1$-sphere then identifies each coordinate orthant's fiber with a simplex lattice, so Theorem 2.2 transfers to $\\mathbb{Z}^n$.","core_discovery":"The central claim is that quantitative unique continuation holds, with explicit rates, on both the simplex lattice $\\Delta_N^{(n)}$ and the lattice cube $Q_L$ for arbitrary $n\\ge 3$. Specifically, Theorem 2.2 states that if $g:\\Delta_N^{(n)}\\to\\mathbb{R}$ satisfies $g(a)\\neq 0$ and $|\\sum_{i=1}^n g(\\beta+e_i)|\\le e^{-C_n N}|g(a)|$ for every $\\beta\\in\\Delta_{N-1}^{(n)}$, then at least $c_n\\Phi_n(a)$ points $\\alpha$ satisfy $|g(\\alpha)|\\ge e^{-C_n N}|g(a)|$, with $\\Phi_n(a)$ as above. Theorem 2.4 states that if $\\Delta_{\\mathbb{Z}^n}u=Vu$ on $Q_L$ with $\\|V\\|_{\\ell^\\infty}\\le K$ and $u(0)\\neq 0$, then at least $c_n L^{\\lceil n/2\\rceil}/\\log(2+L)$ points $x\\in Q_L$ satisfy $|u(x)|\\ge \\exp(-C_{n,K}L^{\\lceil n/2\\rceil+1}/\\log(2+L))|u(0)|$. The simplex decay is exponential in $N$ with no logarithmic loss; the lattice result carries an extra $\\log(2+L)$ factor in both the decay and the cardinality.","pith_inferences":["Editorial inference: the quantitative Pascal uncertainty principle is a self-contained statement about polynomials, independent of Schr\\\"odinger equations; it should apply to any constant-coefficient difference equation whose symbol gives a Pascal-type shift, so similar QUC theorems are plausible for other finite-range hopping operators.","Editorial inference: the simplex theorem's exponential-in-$N$ decay with no logarithmic loss suggests that the $(\\log L)^{-1}$ factor in the $\\mathbb{Z}^n$ theorem comes from the Walsh/Voronoi layer, not from the Pascal principle; sharpening the geometric pigeonholing would remove or reduce the log.","Editorial inference: a natural stress test is numerical: on $\\mathbb{Z}^3$ with $V=0$, count the large-value set for a generic harmonic polynomial; the proof's constants are not tracked, so the test would reveal whether the predicted power $L^2/\\log L$ appears with moderate constants or only asymptotically.","Editorial inference: the weight $\\Phi_n(a)$ singles out the smallest coordinate $a_n$; this predicts that solutions whose large initial value sits near the boundary of the simplex (small $a_n$) have fewer guaranteed large points, a direction that could be tested by constructing explicit $g$ from binomial-coefficient polynomials."],"forward_implications":["Taking $a=R\\mathbf{1}_n$ and $N=nR$ in the simplex theorem gives Theorem 2.1: a function on $\\Delta_{nR}^{(n)}$ satisfying the centered decay condition must be $\\ge e^{-C_n R}|g(R\\mathbf{1}_n)|$ on at least $c_n R^{\\lceil n/2\\rceil}$ points.","Theorem 2.4 answers the support-cardinality question for $\\mathbb{Z}^n$ up to a $(\\log L)^{-1}$ factor: the lower bound $L^{\\lceil n/2\\rceil}$ is optimal in power of $L$, as the paper notes via the cited sharpness proposition.","All the main estimates hold for complex-valued $u$ and $V$ with unchanged constants, since the proof runs line by line in the complex case (Remark 2.2).","The exponents $a=\\lceil n/2\\rceil+1$ and $b=\\lceil n/2\\rceil$ lie outside the range needed for Anderson\\--Bernoulli localization by the standard multiscale argument, so the paper does not claim localization; it supplies the quantitative unique continuation ingredient in that range."],"supporting_citations":[{"why":"supplies the qualitative Pascal uncertainty principle and the positivity of proper Pascal minors (determinant at least 1) on which Lemma 4.3 and the quantitative Pascal UP rest.","marker":"[Li26a]"},{"why":"provides the Lindstr\\\"om\\--Gessel\\--Viennot determinant theorem used to prove the proper-minor positivity in Lemma 4.3.","marker":"[Lin73,GV85]"},{"why":"provides the shifted-Schur tableau identity that yields the monotonicity of shifted Schur polynomials used in Lemma 4.1.","marker":"[OO97]"},{"why":"the three-dimensional QUC result and the shell/geometric-pigeonholing strategy that the present theorems generalize to arbitrary dimension.","marker":"[LZ22]"},{"why":"states the support-cardinality conjecture and the sharpness proposition showing the $L^{\\lceil n/2\\rceil}$ cardinality bound is optimal up to a log factor.","marker":"[Li26b]"}],"fun_headline_variants":["Quantitative unique continuation holds on simplex and Z^n lattices","Many lattice sites keep values above exponential threshold up to log","Simplex has sharp exponential decay; Z^n adds a log factor","n≥3: explicit lower bounds for lattice Schrödinger solutions","Unique continuation on Z^n cubes: count sharp up to log factor"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole argument hinges on the combinatorial fact, cited from [Li26a], that every proper minor of the Pascal matrix has determinant at least 1; if a single proper minor had determinant 0 or a non-integer value below 1, the interpolation polynomial used to control the exponential bounds would not exist with the required size, and the chain from the Pascal principle to the lattice theorem would break.","fun_headline_variants_meta":{"raw":{"variants":["Quantitative unique continuation holds on simplex and Z^n lattices","Many lattice sites keep values above exponential threshold up to log","Simplex has sharp exponential decay; Z^n adds a log factor","n≥3: explicit lower bounds for lattice Schrödinger solutions","Unique continuation on Z^n cubes: count sharp up to log factor"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000999,"raw_usage":{"total_tokens":4190,"prompt_tokens":869,"completion_tokens":3321,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":485,"completion_tokens_details":{"reasoning_tokens":3234}},"tokens_in":485,"tokens_out":3321,"duration_ms":26322,"temperature":1.0,"reasoning_tokens":3234,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:36:13.442363+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the determinant of a proper Pascal minor, for example the $3\\times 3$ minor with rows $r=(2,4,6)$ and columns $k=(0,2,6)$: if any proper minor has determinant $0$ or a non-integer less than $1$, the bound (4.6) in Lemma 4.3 fails and the proof from Lemma 4.5 onward no longer holds. Independently, on $\\mathbb{Z}^3$ take a non-constant harmonic polynomial ($V\\equiv 0$) with $u(0)\\neq 0$, evaluate it on $Q_L$ for large $L$, and count the points where $|u(x)|\\ge \\exp(-C L^3/\\log L)|u(0)|$; a count below a positive constant times $L^2/\\log L$ would refute Theorem 2.4's conclusion.","supporting_citations":[],"review_version":1}