{"id":"f08bc356-7855-4d0d-9cfd-cf0a3d1301b4","arxiv_id":"1908.00969","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Random non-Hermitian matrices with independent centered entries have the same local eigenvalue statistics at the spectral edge as the Gaussian Ginibre ensemble, with no four-moment matching needed.","lead":"This paper proves that the jittery patterns of eigenvalues near the edge of a random non-Hermitian matrix are always the same, no matter what the entries are, as long as they are centered and have enough moments. The result extends a famous universality principle, known for symmetric random matrices, to the non-symmetric case, and removes a restrictive four-moment matching condition.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 is stated under Assumption (A), but the proof and local law rely on (B); the removal via [55, Theorem 3.2] in Remark 1 is not verified and may not cover the shifted matrix X−zI.","rationale":"The reader's weakest assumption identifies exactly the load-bearing gap: the main theorem is stated under (A) only, while the paper's proof, as written, uses (B) in Proposition 1, Proposition 5, Lemma 3, and Lemma 4. The only bridge to (A) is the unproved reduction in Remark 1 via [55, Theorem 3.2]. If that theorem applies as claimed, the missing steps are routine but still worth writing out; if not, the central claim is not established for all matrices admitted by Theorem 1. I do not see a competing concern with more weight: the OU-flow comparison, the cumulant estimates in Propositions 3 and Lemma 2, and the use of the Ginibre lower-tail bound are internally coherent, and the extension in Appendix A is sketched but plausible. The paper deserves credit for the proof structure and for flagging the reduction explicitly. The verdict should remain conditional pending the verification of the external bound and its role in removing (B) from the local law.","tokens_in":28410,"tokens_out":25245,"duration_ms":261823,"concrete_test":"Check [55, Theorem 3.2] and independently re-derive (5) for H_z from it, under only Assumption (A), uniformly for |z|≤2, paying attention to the deterministic shift −zI and to whether the theorem allows the non-identically distributed entries inside the 2×2 block model. Then repeat the proofs of Proposition 1 and Lemma 4 with (5) replacing Proposition 5: in particular, verify that the integral in (4.5) over t≥l log n of P(λ1≤e^{-t}) is O(n^{-δ/3}) with the C_l supplied by [55], and that the local law at η down to n^{-1} survives without any density condition. If any step fails, the theorem should be restated under Assumption (A)+(B).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1 asserts edge universality for every i.i.d. matrix satisfying Assumption (A). The proof, however, runs through Proposition 1 (local law, (1.13)), Proposition 2, Proposition 5 (3.8), and Lemmas 3–4, all formulated under Assumption (B), a density condition. Proposition 5 is quoted from [3, Prop. 5.7], and Lemma 4 uses it to control eigenvalues |λ_i| < n^{-l} in (4.5). The mechanism advertised to remove (B) is Remark 1, where (5) is quoted from [55, Theorem 3.2]. That reduction is not executed: one must confirm (5) holds uniformly in |z|≤2 for the Hermitized matrix H_z whose off-diagonal blocks are X−zI — so the relevant 2n×2n block matrix does not have i.i.d. entries, and the diagonal shift contributes mean −z — and that the constants C_l are compatible with the choice of l in Lemma 4 and with the high-probability argument in Lemma 3. If [55, Theorem 3.2] in fact assumes a smooth-density or small-ball condition comparable to (B), or only applies to mean-zero matrices, then (5) cannot remove (B), Proposition 1 also stays hostage to (B), and Theorem 1 is proved only for ensembles satisfying (A)+(B). This is a genuine assumption gap in the stated theorem, not a disagreement with consensus.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a local edge universality theorem for n x n non-Hermitian random matrices with independent, identically distributed, centered entries. Under Assumption (A), i.e. centered entries with unit variance and all moments finite, it claims that for every fixed k and spectral parameters z_1,...,z_k on the unit circle, the k-point correlation function rescaled by n^{-1/2} around z_j converges to the corresponding Ginibre scaling limit with error O(n^{-c}). The proof strategy combines Girko's Hermitization formula, an optimal local law for the block matrix H_z, a lower-tail estimate on small singular values of X-z, and a long-time Ornstein-Uhlenbeck Green function comparison that matches the matrix X to a Ginibre matrix. Both the complex and the real Ginibre cases are covered.","tokens_in":28678,"tokens_out":13917,"duration_ms":134076,"significance":"If the main theorem is valid under Assumption (A) alone, the result is a substantial advance: it removes the four-moment matching condition from the previous non-Hermitian universality theorem of Tao and Vu and establishes the natural non-Hermitian analogue of Tracy-Widom edge universality. The paper is also valuable for its transparent proof architecture: the decomposition of the comparison into a local-law component, a small-singular-value component, and a Green function comparison is clearly laid out, and the quantitative O(n^{-c}) rate is explicitly tracked. The treatment of both the complex and the real Ginibre ensembles is an additional strength. However, the advertised theorem is not what is proved as written: the proof is conducted under Assumptions (A) and (B), and the asserted removal of (B) is delegated to a cited external theorem without a complete verification.","major_comments":[{"comment":"Theorem 1 is stated under Assumption (A) only, but the proof of Theorem 1 runs through Proposition 1 (local law), Proposition 4, Lemma 1, Lemma 3, and Lemma 4, all of which are formulated under Assumptions (A) and (B). Remark 1 asserts that Assumption (B) can be removed by the quoted bound (5) from [55, Theorem 3.2], but the removal is not executed. In particular, Proposition 1 is used throughout the paper, and no argument is given that the optimal local law for H_z holds under (A) alone. Either Theorem 1 must be restated under (A)+(B), or the authors must supply a complete proof or a precise citation that the local law in Proposition 1 holds without (B). This is a load-bearing gap between the theorem's hypothesis and the proof.","section":"Section 2, Proposition 1 and Section 4, Proposition 4"},{"comment":"The quoted statement (5) is not verified for the matrices actually needed. The relevant object is Spec(H_z) with H_z built from the shifted matrix X-zI, whose off-diagonal blocks are i.i.d. but whose diagonal blocks contain the deterministic shift -z. One must check that [55, Theorem 3.2] applies to this Hermitized form, uniformly in |z| <= 2, with constants C_l that are compatible with the choice of l in Lemma 4 and with the integration over t in (4.5). If [55, Theorem 3.2] assumes a smooth density or mean-zero structure comparable to (B), or if its constants grow too fast in l, then (5) cannot remove (B) and Theorem 1 is proved only for ensembles satisfying (A)+(B). This needs to be settled explicitly rather than by a one-line remark.","section":"Remark 1 and (5)"},{"comment":"The displayed bound in (4.5) reads E[|log lambda_1| 1(lambda_1 <= n^{-l})] = \\int_{l log n}^\\infty P(lambda_1 <= e^{-t}) dt \\lesssim n^{\\beta+1+2\\alpha/(1+\\alpha)} e^{-2\\alpha l/(1+\\alpha)}. As written, this bound does not decay in n at all when l is fixed, and the subsequent choice of l does not imply the claimed E|I_2^{(j)}| \\lesssim n^{-\\delta/3}. The intended estimate is presumably n^{\\beta+1+2\\alpha/(1+\\alpha)-2\\alpha l/(1+\\alpha)}, i.e. with n^{-2\\alpha l/(1+\\alpha)} instead of e^{-2\\alpha l/(1+\\alpha)}. Since Lemma 4 is essential for the reduction to the I_3-part in Lemma 1, this error must be corrected and the final bound re-verified.","section":"Lemma 4, Eq. (4.5)"}],"minor_comments":[{"comment":"If the local law from [4] is intended to hold under Assumption (A) alone, this should be stated explicitly and the precise theorem and hypotheses in [4] should be quoted; the current text appears to contradict itself by stating Proposition 1 under (A)+(B) and then claiming (B) is superfluous.","section":"Remark 1"},{"comment":"The notation \\sum_{abc} in (2.7) is ambiguous; it should be written as \\sum_{a,b}\\sum_c with the range of (a,b) and c specified, to avoid confusion over the summation indices.","section":"Eq. (2.7)"},{"comment":"The term '1/(a \\equiv b+n (mod 2n))' should be written as the indicator 1(a \\equiv b+n (mod 2n)) for clarity, since it is not a reciprocal.","section":"Lemma 5, Eq. (4.8)"},{"comment":"The statement 'the constant c>0 is a small constant depending on k' is vague; the proof tracks a dependence on \\delta and on the C^{2k+1} norm of F. It would help to state explicitly that c depends on k and on the choice of \\delta, and that the implicit constant in O(n^{-c}) depends on k and the norm of F.","section":"Theorem 1"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the mismatch between the theorem's hypothesis and the assumptions used in the proof. If the authors can either prove the local law under (A) alone and verify the Tao-Vu bound (5) with the required uniformity and constants, or else restate the main theorem under (A)+(B), the paper would be in much better shape. The current version, with the unexecuted Remark 1 and the apparent typo in (4.5), cannot be accepted as is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nYou should know two things up front. First, this is the paper that removes the four-moment matching condition from Tao–Vu and proves Ginibre edge statistics for general centered i.i.d. entries. The Ornstein–Uhlenbeck comparison method, which avoids Dyson Brownian motion, is genuinely new and has turned out to be influential. Second, the written proof has a real assumption gap: Theorem 1 is stated under (A), but the body of the paper proves the main comparison under (A)+(B), and the removal of (B) is delegated to a quoted theorem in [55] via Remark 1. That reduction is not carried out, and it is not obvious that [55, Thm 3.2] applies to the Hermitized block matrix H_z built from X−zI, whose blocks are shifted and structured rather than i.i.d.\n\nWhat the paper does well: the local law input is precisely quoted from Alt–Erdős–Krüger, the lower-tail control is extended from the Ginibre case via Green function comparison (Proposition 3), and the four-part decomposition through Girko's formula is careful. The proof of Lemma 2 does real work, and the technical estimates look honest. The paper also flags its own dependence on (B), both in Remark 1 and in Lemma 3, so this is not a hidden circularity; it is an unfinished reduction. The central idea — control edge statistics by running the OU flow for a long time and comparing to the explicit Ginibre benchmark — is sound, and I believe it is correct.\n\nThe soft spots, in order: (1) the (B)-to-(A) bridge, which is the main real gap; (2) the extension of the local law below η=n^{−3/4} in Appendix A is only sketched, with the Lipschitz/union-bound argument summarized rather than shown; (3) the proof leans heavily on companion papers for the local laws, so a referee needs those available. None of these refute the claim, but they are genuine technical obligations.\n\nWho this is for: probabilists and mathematical physicists working on non-Hermitian random matrices; anyone serious in the area needs to read it. It deserves a serious referee — the gap is fixable and the result is central. If I were the editor, I would send it out, with a strong request that the (B)-removal be made explicit.","headline":"Strong, important paper on non-Hermitian edge universality, but Theorem 1 as stated is not fully proved under (A) alone: the written argument uses (B), and the cited bridge from (B) to (A) needs closer checking.","tokens_in":29258,"tokens_out":2315,"would_cite":true,"duration_ms":23242,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60B20","15B52"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every centered i.i.d. non-Hermitian matrix has the same edge eigenvalue statistics as Ginibre, no Gaussian moments required.","keywords":["non-Hermitian random matrices","edge universality","Ginibre ensemble","circular law","Girko's formula","local eigenvalue statistics","Green function comparison","least singular value"],"falsifier":"Take Rademacher entries ($\\chi=\\pm1$, centred, all moments finite), let $n$ grow to about $10^4$, collect eigenvalues inside a band of width $n^{-1/2}$ around a fixed boundary point $|z|=1$, and compare the empirical two-point correlation function with the Ginibre edge kernel; if the discrepancy does not decay like a power of $n$, Theorem 1 is false.","tokens_in":28187,"feed_emoji":"🎲","tokens_out":12224,"duration_ms":102074,"temperature":0.7,"pith_summary":"The paper establishes that at the edge of the spectrum—the unit circle—the local eigenvalue statistics of any large non-Hermitian random matrix with independent, identically distributed, centred entries coincide with those of the Gaussian Ginibre ensemble. Earlier universality results required the first four moments of the entry distribution to almost match the Gaussian; this paper removes that condition. The result holds for both real and complex matrices and comes with an explicit algebraic error rate. It is the non-Hermitian analogue of Tracy-Widom edge universality.","feed_headline":"Edge statistics of all i.i.d. non-Hermitian matrices match Ginibre","feed_subtitle":"No moment matching needed: any centered i.i.d. entries give the same edge correlations as Gaussian.","key_machinery":"The machinery is Hermitization. One embeds the non-Hermitian problem into the Hermitian spectrum of the $2n\\times 2n$ block matrix $H_z$ whose off-diagonal blocks are $X-z$ and $X^*-z$, and Girko's formula expresses linear statistics of the eigenvalues of $X$ as integrals of $\\Im\\operatorname{Tr}(H_z-i\\eta)^{-1}$. The proof needs two sharp controls on $H_z$: an optimal local law for its resolvent near the cusp at zero, and a lower-tail estimate on its smallest eigenvalue, equivalently on the least singular value of $X-z$. These are obtained for general entries by running an Ornstein–Uhlenbeck flow $X_t=e^{-t/2}X+\\sqrt{1-e^{-t}}\\tilde{X}$ that interpolates between the given matrix and an independent Ginibre matrix, then comparing expectations of products of resolvent traces by a Green-function comparison. The Gaussian endpoint is controlled by an explicit lower-tail bound, and the comparison shows that the only non-negligible integrals differ between the two ensembles by $O(n^{-c})$.","core_discovery":"The central claim is Theorem 1: for every fixed integer $k\\ge 1$ and spectral parameters $z_1,\\dots,z_k$ on the unit circle, the $k$-point correlation function of the eigenvalues, rescaled by $\\sqrt{n}$, converges to the corresponding real or complex Ginibre scaling limit with an error of order $O(n^{-c})$. The only assumptions on the entries are zero mean, unit variance, and finiteness of every moment; no matching of the first four moments is required. The statement covers both the complex Ginibre ensemble, whose scaled correlations are determinantal, and the real Ginibre ensemble with its roughly $\\sqrt{n}$ real eigenvalues.","pith_inferences":["The same Ornstein–Uhlenbeck comparison should yield bulk universality once an optimal bulk local law and a bulk lower-tail bound on the least singular value of $X-z$ are available; the paper notes its method is currently restricted to the edge.","Because Assumption (B) is removed by quoting a least-singular-value bound, the theorem should hold even for discrete entry laws such as $\\pm1$ without any density condition; a fully self-contained proof would derive that quoted bound.","The explicit polynomial rate makes the prediction directly testable numerically at moderate $n$, since finite-size deviations from the Ginibre edge kernel should shrink like a power of $n$."],"forward_implications":["The four-moment matching condition of earlier non-Hermitian universality results is unnecessary at the spectral edge.","Any centred i.i.d. entry law satisfying the moment bound produces the same edge $k$-point correlations as the complex or real Gaussian ensemble.","For fixed $k$, the approach to the Ginibre limit is at a guaranteed polynomial rate $n^{-c}$, not merely a qualitative convergence.","The result gives the non-Hermitian counterpart of Tracy-Widom edge universality: edge fluctuations of non-Hermitian matrices are universal in the same sense as Hermitian edge fluctuations."],"supporting_citations":[{"why":"Supplies the optimal local law for the resolvent of $H_z$ in the edge regime, stated here as Proposition 1.","marker":"[4, Theorem 5.2]"},{"why":"Provides the local-law framework for the Hermitized matrix that the edge extension in [4] builds on.","marker":"[3, Theorem 5.2]"},{"why":"Gives the Gaussian lower-tail bound on the least singular value of the shifted Ginibre matrix, the starting point for the general-entry comparison in Proposition 2.","marker":"[18]"},{"why":"Establishes the earlier edge universality under four-moment matching; this paper removes that condition.","marker":"[54]"},{"why":"Quoted in Remark 1 to exclude very small singular values without a density assumption, making Assumption (B) superfluous.","marker":"[55, Theorem 3.2]"},{"why":"Supplies the real Ginibre scaling-limit correlation functions that are the target of the real-case statement.","marker":"[10]"},{"why":"Supplies the complex Ginibre kernel that defines the limiting correlation functions in the complex case.","marker":"[31]"}],"fun_headline_variants":["Non-Hermitian edge universality: any i.i.d. entries give Ginibre statistics","Ginibre universality for non-Hermitian matrices: no moment matching needed","Edge statistics of i.i.d. non-Hermitian matrices match Ginibre universal law","All i.i.d. non-Hermitian matrices have Ginibre edge correlations","Ginibre edge law holds for any centered i.i.d. non-Hermitian matrix"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof needs a guarantee that $X-z$ almost never has a singular value smaller than any power of $n$ while $|z|$ is within about $n^{-1/2}$ of 1; in the version of the theorem without a density condition, that guarantee is imported from a quoted bound in the literature rather than derived in this paper.","fun_headline_variants_meta":{"raw":{"variants":["Non-Hermitian edge universality: any i.i.d. entries give Ginibre statistics","Ginibre universality for non-Hermitian matrices: no moment matching needed","Edge statistics of i.i.d. non-Hermitian matrices match Ginibre universal law","All i.i.d. non-Hermitian matrices have Ginibre edge correlations","Ginibre edge law holds for any centered i.i.d. non-Hermitian matrix"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000263,"raw_usage":{"total_tokens":1499,"prompt_tokens":745,"completion_tokens":754,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":361,"completion_tokens_details":{"reasoning_tokens":644}},"tokens_in":361,"tokens_out":754,"duration_ms":7039,"temperature":1.0,"reasoning_tokens":644,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:26:58.607765+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take Rademacher entries ($\\chi=\\pm1$, centred, all moments finite), let $n$ grow to about $10^4$, collect eigenvalues inside a band of width $n^{-1/2}$ around a fixed boundary point $|z|=1$, and compare the empirical two-point correlation function with the Ginibre edge kernel; if the discrepancy does not decay like a power of $n$, Theorem 1 is false.","supporting_citations":[],"review_version":1}