{"id":"bec3db8a-d48a-4894-b524-052b0a5fbd6c","arxiv_id":"2411.18550","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For one-cut regular random matrix ensembles, the edge-scaling limit of an average with a Fisher-Hartwig singularity is universal and is expressed through solutions of the Painlevé-XXXIV equation.","lead":"This paper computes the exact large-matrix limit of a random matrix average carrying a special singular weight at the edge of the spectrum, and shows the limit is governed by Painlevé equations. The result makes edge universality precise for a broad class of one-cut potentials and for complex jump parameters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The boundary data for the central Painlevé formulas (16), (18) and the Hankel expansions (21)–(26) is imported from Lemma B.3, whose uniform β∈C\\(−∞,0) asymptotics the text admits is not precisely covered by the cited references and is not proved here.","rationale":"The reader identified the unique solvability and large-x asymptotics of the Painlevé-XXXIV model RHP B.1 as the weakest assumption. I agree. The paper is otherwise a systematic Riemann-Hilbert analysis: the transformations (39),(41),(44),(59) are standard, the parametrices are explicit, and the small-norm estimates are quantified. The Fredholm determinant proof of Theorem 1.5 uses dominated convergence plus analytic continuation in β, which is standard and acceptable. The Hankel determinant proofs are long algebraic expansions, but they are detailed enough to be checked mechanically. The genuinely external, unproved input is Lemma B.3. The text itself flags that the references do not exactly match the present model problem, so the β-uniform exponential asymptotics are asserted rather than demonstrated. Since these asymptotics determine the boundary constraints (17),(19) that are then used in (16),(18) and in the integration of the differential identities in Sections 5–6, this is the single most load-bearing concern. It does not, however, force rejection: the concern can be resolved by supplying a complete proof of Lemma B.3 or by citing a theorem that covers exactly this model problem and parameter domain. Thus the reader's CONDITIONAL verdict is unchanged.","tokens_in":61034,"tokens_out":9030,"duration_ms":80686,"concrete_test":"Perform a Deift–Zhou steepest descent analysis of RHP B.1 for general β∈C\\(−∞,0) and verify the exponentially small term in (123) and Corollary B.4, or use high-precision numerical solution of the Lax pair in Lemma B.2 for representative values (e.g. α∈{0,0.5,2}, β∈{i,1+i,−0.5+i}) and compare q(x,α,β) and σ(x,α,β) against (17),(19) for large x. Also check whether the cited results [16,27,29] actually contain the stated uniform-in-β prefactor (e^{iπα}−β); if not, Lemma B.3 needs its own proof before (16),(18),(21)–(26) can be accepted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main theorems are controlled by the large-x behavior of the model RHP B.1. Lemma B.3 (with Corollary B.4) supplies the leading and exponentially small terms of a=Q_1^12, q, and σ, and these feed directly into the boundary constraints (17), (19) and hence into the Painlevé representations (16), (18) and the Hankel determinant asymptotics (21)–(26). The paper does not prove Lemma B.3; Appendix B.2 explicitly says that references [16,27,29] do not precisely match the needs of RHP B.1 and that the results are only summarized. In particular, the exponentially small correction in (123) has a specific prefactor (e^{iπα}−β) and exponent −(2+3α)/2 in x (resp. −(1+3α)/2 for q in Corollary B.4), asserted uniformly for β∈C\\(−∞,0) on compact sets. If the cited works only cover special cases such as β=1, or use a different normalization, then the β-dependence in this boundary data is a new unproved assertion. This is load-bearing because the integrands in (16) and (18) are exactly differences of two Painlevé solutions, and the cancellation of the leading α√x terms plus the exponentially small corrections fixes the additive constants. An error in Lemma B.3 would propagate into F(s;α,β), η_α(s,β), and the full expansion (24). The surrounding RHP construction is coherent, but this external input is the least secure link.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the generating functional E_n[φ;λ,α,β;V] for Hermitian one-cut regular random matrix ensembles with a Fisher-Hartwig singularity placed at the soft edge. The main results are: (Theorem 1.5) a universal soft-edge limit of the Fredholm determinant factor F_n in terms of a kernel A_s^{αβ} built from the Painlevé-XXXIV model RHP B.1, together with Painlevé formulas (16) and (18) for the limiting hard-edge distribution F(s;α,β); (Theorem 1.7) for V=x^2, full large-n asymptotics of the Hankel determinant ratio including the n^{1/3} and log n terms, with the remainder expressed through Painlevé data; (Theorem 1.10) an extension of the Hankel ratio asymptotics to general one-cut regular V; and (Corollary 1.11) a CLT for the edge-scaled log-characteristic polynomial. The proof uses the Fokas-Its-Kitaev Riemann-Hilbert problem, nonlinear steepest descent with an outer parametrix, an Airy parametrix at the left endpoint, and a Painlevé-XXXIV parametrix at the right endpoint, followed by small-norm estimates and integration of differential identities.","tokens_in":61348,"tokens_out":4195,"duration_ms":40196,"significance":"If the results are correct, they constitute a substantial advance: they lift earlier Gaussian-only or β=1 results of Forrester-Witte and of Wu-Xu-Zhao to general one-cut regular potentials and to complex β∈C\\(−∞,0), and they give the first full expansion of E_n including the n^{1/3} and log n corrections. The Riemann-Hilbert architecture is coherent and well matched to the problem: the explicit outer and local parametrices, the small-norm ratio problem, and the Fredholm determinant convergence argument are all present. The paper also gives concrete, falsifiable predictions, e.g. the α^2/6 log n term and the reduction to the Tracy-Widom distribution when (α,β)=(0,1). The main weakness is that the controlling boundary data for the Painlevé formulas are imported from an external asymptotic statement, Lemma B.3, which the text itself describes as not precisely covered by the cited references.","major_comments":[{"comment":"The large-x asymptotics of the model RHP B.1 are load-bearing for the central claims: equation (123) and Corollary B.4 determine the boundary constraints (17) and (19), which in turn fix the additive constants in the Painlevé representations (16) and (18) and enter the Hankel determinant expansions (21)–(26). The text explicitly states that references [16,27,29] do not precisely match the needs of RHP B.1 and that the results are only summarized. In particular, the exponentially small correction in (123) contains a prefactor (e^{iπα}−β) and an exponent depending on α and β uniformly for β∈C\\(−∞,0), which is not visibly established in the cited papers. Since an error in this boundary data would propagate directly into F(s;α,β), η_α(s,β), and the full expansion (24), the authors must either prove Lemma B.3 in the needed normalization and domain, or identify a precise reference with matching hypotheses. As it stands, this is a new unproved assertion at the foundation of the main theorems.","section":"Appendix B.2, Lemma B.3 and Corollary B.4"},{"comment":"The dominated convergence argument for the Fredholm series in Theorem 1.5 is written for β≥0 using the Hadamard bound (67), which relies on positive definiteness of the kernel. The passage to β∈C\\(−∞,0) is then made by an application of Vitali's convergence theorem and the identity theorem. This requires a uniform domination of the series terms on compact subsets of β∈C\\(−∞,0), not merely pointwise convergence of the kernel. The text does not spell out the needed local-uniform L^1 bounds for complex β; such bounds presumably follow from the same exponential decay estimates used for β≥0, but the argument should be made explicit, because the Hadamard inequality used for β≥0 is not available for indefinite kernels.","section":"Section 4, proof of (14)"}],"minor_comments":[{"comment":"The arXiv header contains the typo 'MA TRICES'; the word should be 'MATRICES'.","section":"Title and abstract"},{"comment":"There are minor typos: 'Assumptoin' in the statement of Corollary 1.11 and 'lenghty' in Section 1.5. These should be corrected in the final version.","section":"Section 1.5 and Corollary 1.11"},{"comment":"The Painlevé determinant formulas are stated for β∈C\\(−∞,1), while the kernel convergence (14) is stated for β∈C\\(−∞,0). The reason for the extra restriction (invertibility of I−A_s^{αβ} on L^2(0,∞)) is explained in Section 4, but it would help the reader to state this distinction explicitly in the theorem and to note that (14) itself remains valid for β∈C\\(−∞,0).","section":"Equations (16) and (18)"},{"comment":"In the proof of Proposition 6.11 the text says 'combining the above we obtain (113)', but the displayed result of that proposition is equation (114). Later, Corollary 6.12 refers to 'combining (94), (98) and (113)', where (114) appears to be intended. Please correct the cross-references.","section":"Section 6, Proposition 6.11"},{"comment":"The proof of (21) is carried out for α∈(−1,∞)\\Z, while the theorem asserts all α>−1. The final extension by regularity is mentioned in the text before Section 3, but it would be helpful to state explicitly in the proof of Theorem 1.7 that the exceptional integer values follow by continuity in α, with the error term uniform on compact sets.","section":"Section 5, Corollary 5.10 and Theorem 1.7"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically ambitious and, on its face, the Riemann-Hilbert machinery is coherent. The central issue is the unproved external input Lemma B.3: the text admits the cited references do not cover the needed β-dependence and normalization. This is not a circularity problem — the model problem is a legitimate external input — but it is a completeness problem in a load-bearing place. If the authors can supply a proof of Lemma B.3 or a precise matching reference, the paper would be a strong candidate for acceptance. I would not reject on the basis of disagreement with consensus; the concern is purely about the validity of the imported asymptotic boundary data. I also recommend the authors clarify the dominated convergence step for complex β in the proof of (14)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nYou should know two things about arXiv:2411.18550. First, it delivers the Fredholm-determinant-level edge universality with a Fisher-Hartwig singularity for general one-cut regular V and complex β, with a CLT as a corollary. Second, the boundary data that control the Painlevé-XXXIV representations are imported from Lemma B.3, which the authors themselves say is not precisely covered by the cited references and is only summarized. That is the soft spot worth scrutinizing.\n\nWhat's genuinely new: Theorem 1.5 lifts the pointwise kernel convergence of [15] and [27] to convergence of the Fredholm determinant for all one-cut regular real-analytic V and α>-1, β∈C\\(-∞,0). The proof is a coherent RH analysis: explicit parametrices, a small-norm ratio problem, dominated convergence for the Fredholm series, and analytic continuation in β. The Painlevé formulas (16) and (18), expressing F(s;α,β) through differences of two solutions of the same Painlevé equation, are new and non-obvious. The Hankel asymptotics (21)-(26), with the n^{1/3} and log n terms, and the CLT in Corollary 1.11, go well beyond Forrester-Witte. The paper is honest about its own limitations.\n\nThe main issue is Lemma B.3. The exponentially small corrections in (123) and Corollary B.4, with their β-dependent prefactors, fix the additive constants in (16), (18) and feed into the full expansion (24). Without a proof or a reference that precisely establishes these asymptotics uniformly for β∈C\\(-∞,0) on compact sets, the controlling boundary data are unverified. The paper says [16,27,29] do not exactly match the needed setup. A referee should require the authors to either prove Lemma B.3 in the required generality or pin down a precise reference with a proof sketch. This is not fatal to the overall strategy—the RHP architecture is sound—but it is load-bearing.\n\nMinor issues: Proposition 4.3 as printed says β∈(-∞,1), which must be β∈C\\(-∞,1); the proof uses the latter. The long coefficient expansions in Section 5 and Appendix A are not checkable from the text; they appear internal and not load-bearing, but they make refereeing slow. There is also a small caveat in Proposition 5.8 excluding α∈Z, presumably removable by continuity.\n\nWho is this for? People working in random matrix theory, integrable probability, and Painlevé asymptotics. It deserves a serious referee. My recommendation: send it to peer review, and ask the authors to supply a complete proof or exact reference for Lemma B.3 before acceptance.","headline":"Genuine Fredholm-level edge universality with a Fisher-Hartwig singularity for general V and complex β, but the Painlevé boundary data rest on an imported asymptotic lemma the authors admit is not precisely proved.","tokens_in":61895,"tokens_out":3818,"would_cite":true,"duration_ms":30798,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47B35","45B05","30E25","34E05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Edge-scaled random matrix averages with a Fisher–Hartwig singularity converge to a Fredholm determinant built from a Painlevé-XXXIV Riemann–Hilbert problem, with the full Gaussian asymptotics expressed through Painlevé-II sigma functions.","keywords":["random matrix averages","Fisher-Hartwig singularity","soft edge","Painlevé-XXXIV","Fredholm determinants","Hankel determinants","Riemann-Hilbert problems","characteristic polynomial CLT"],"falsifier":"Compute numerically the left-hand side of (24) for, say, V(x)=$x^{2}$, a fixed α>−1, β≠1, and a few values of s, using Monte Carlo or high-precision quadrature for the Gaussian unitary ensemble with the Fisher–Hartwig factor, and compare the result to the right-hand side with the Painlevé-II $\\sigma$ functions evaluated via their connection formulas; a mismatch beyond the claimed o(1) error would disprove Corollary 1.9.","tokens_in":60811,"feed_emoji":"📐","tokens_out":2354,"duration_ms":23086,"temperature":0.7,"pith_summary":"The paper studies the large-n limit of invariant Hermitian random matrix ensembles when the eigenvalue average contains a Fisher–Hartwig-type singularity (a power-law modulus with a phase jump) placed exactly at the soft edge of the spectrum. It proves a universality theorem: for every one-cut regular real-analytic potential V, the edge-scaled average converges to a Fredholm determinant whose kernel is constructed from the Painlevé-XXXIV model Riemann–Hilbert problem. For the Gaussian case V(x)=$x^{2}$, it obtains the complete asymptotic expansion of this average, including the $n^{{1/3}}$ scaling and the logarithmic-in-n term, with the subleading term expressed through Painlevé-II $\\sigma$ functions. The work extends earlier Gaussian, β=1 results of Forrester and Witte, and it establishes a central limit theorem for the logarithm of the absolute value of the edge-scaled characteristic polynomial.","feed_headline":"Painelevé XXXIV governs edge-singular random matrix averages","feed_subtitle":"Universality across one-cut potentials, with full Gaussian asymptotics and a new CLT for the log-characteristic polynomial.","key_machinery":"The central object is the Painlevé-XXXIV model Riemann–Hilbert problem (RHP B.1), a piecewise-constant jump problem whose solution Q(ζ;x,α,β) satisfies a Lax pair whose compatibility yields the Painlevé-XXXIV equation for q and the Jimbo–Miwa–Okamoto σ-Painlevé-II equation for σ=1/$4x^{2}$−a. This model problem replaces the Airy parametrix near the edge and provides the boundary data (17), (19) that enters the Painlevé formulas. The argument is carried by the Deift–Zhou nonlinear steepest descent analysis of the orthogonal-polynomial RHP, combined with the Its–Izergin–Korepin–Slavnov theory of integrable integral operators for the Fredholm determinant, and differential identities (36)–(38) for the Hankel determinants.","core_discovery":"The central claim is that, for every one-cut regular real-analytic V and all α>−1, β∈C\\((−∞,0), the edge-scaled average E_n[φ_n;λ_n,α,β;V] converges to a Fredholm determinant with kernel $A_s^{{αβ}}$ built from the Painlevé-XXXIV RHP (Theorem 1.5). For the quadratic potential, the full asymptotic is ln E_n = nα/2(1−ln2) + αs $n^{{1/3}}$ + ($α^{2}$/6)ln n + Ξ_α(s,β)+o(1), with Ξ_α expressed through Painlevé-II $\\sigma$ functions (Corollary 1.9). The paper also derives ratio asymptotics for Hankel determinants with a single Fisher–Hartwig singularity for general one-cut potentials (Theorem 1.10), and a CLT for the edge-scaled log-characteristic polynomial (Corollary 1.11).","pith_inferences":["The difference-of-two-Painlevé-transcendents structure in (16) likely extends to other edge-singularity models where a single Fredholm determinant encodes both sides of a jump; one might look for analogous formulas in thinned or conditioned edge processes.","The Painlevé-XXXIV model problem is the natural 'master kernel' for edge Fisher–Hartwig singularities; one could test numerically whether the Fredholm determinant on L^2(0,∞) with kernel A_s^{αβ} reproduces known Tracy–Widom-type distributions for special parameter choices beyond (α,β)=(0,1).","The CLT (27) is stated for the log-modulus of the characteristic polynomial; a similar argument with an additional phase factor might yield a joint CLT for the real and imaginary parts, connecting to the logarithmic-correlated-field picture at the edge.","Since the ratio asymptotics (26) hold for general one-cut V, one could try to push the same RHP approach to the bulk of the spectrum, where the model problem should degenerate to a Fourier-type kernel and the Painlevé-XXXIV data should drop out, recovering known bulk Fisher–Hartwig asymptotics."],"forward_implications":["If the paper's theorems hold, the edge-scaled generating functional for these ensembles is universal: the same Painlevé-XXXIV kernel appears for every one-cut regular V, not just the Gaussian weight.","The Gaussian expansion (24) gives the full leading-order large-n behavior of the average with a Fisher–Hartwig singularity at the edge, including all terms that diverge with n and the constant Ξ_α(s,β), which is new for β≠1.","The ratio result (26) makes precise how the edge-scaling of the singularity modifies the known bulk Fisher–Hartwig ratio asymptotics, with fractional powers of n and Painlevé functions entering only in the edge-sensitive terms.","The CLT (27) shows that the logarithm of the absolute value of the edge-scaled characteristic polynomial, after subtracting a deterministic n−1/3 shift, is asymptotically normal with variance (ln n)/3.","The Fredholm determinant formula (16) expresses the limiting distribution as the exponential of an integral of a difference of two Painlevé-II sigma solutions, a structure that appears to be new for Fredholm determinants in integrable systems."],"supporting_citations":[{"why":"Supplies the Painlevé-XXXIV model RHP and identifies the kernel A_s^{α1} in a double-scaling limit, which the paper extends to general β and one-cut V.","marker":"[15]"},{"why":"Identifies A_s^{αβ} as the limit of the reproducing kernel for Gaussian Fisher–Hartwig weights with an edge-scaled singularity; the paper lifts this to Fredholm convergence.","marker":"[27]"},{"why":"The Gaussian, β=1, τ-function result (Proposition 26) that the paper generalizes to all β and to general one-cut V.","marker":"[12]"},{"why":"Provides the Deift–Zhou nonlinear steepest descent framework and the one-cut regular machinery used to analyze the orthogonal-polynomial RHP.","marker":"[7]"},{"why":"Gives the Tracy–Widom formulas and the Airy-kernel limit that the new results must reduce to in the case (α,β)=(0,1).","marker":"[24]"},{"why":"The Its–Izergin–Korepin–Slavnov theory of integrable integral operators used to derive the Painlevé expression (16) from the Fredholm determinant.","marker":"[14]"}],"fun_headline_variants":["Painlevé-XXXIV drives edge-singular universality","Edge singularity universality via Painlevé-XXXIV","Painlevé-XXXIV asymptotics for random matrix edge singularities","New Painlevé-XXXIV results for edge-singular random matrix averages","Universality at the soft edge with Painlevé-XXXIV"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole argument leans on the unique solvability and the large-x asymptotic expansions of the Painlevé-XXXIV model Riemann–Hilbert problem (summarized in Lemma B.3); if those asymptotics fail on the needed domain, the boundary conditions for the Painlevé formulas and the Hankel determinant expansions would lose their controlling input.","fun_headline_variants_meta":{"raw":{"variants":["Painlevé-XXXIV drives edge-singular universality","Edge singularity universality via Painlevé-XXXIV","Painlevé-XXXIV asymptotics for random matrix edge singularities","New Painlevé-XXXIV results for edge-singular random matrix averages","Universality at the soft edge with Painlevé-XXXIV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000995,"raw_usage":{"total_tokens":4265,"prompt_tokens":1043,"completion_tokens":3222,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":659,"completion_tokens_details":{"reasoning_tokens":3129}},"tokens_in":659,"tokens_out":3222,"duration_ms":23188,"temperature":1.0,"reasoning_tokens":3129,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:05:44.266741+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute numerically the left-hand side of (24) for, say, V(x)=$x^{2}$, a fixed α>−1, β≠1, and a few values of s, using Monte Carlo or high-precision quadrature for the Gaussian unitary ensemble with the Fisher–Hartwig factor, and compare the result to the right-hand side with the Painlevé-II $\\sigma$ functions evaluated via their connection formulas; a mismatch beyond the claimed o(1) error would disprove Corollary 1.9.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Painlevé-XXXIV model RHP and identifies the kernel A_s^{α1} in a double-scaling limit, which the paper extends to general β and one-cut V."},{"cited_title":"Wu, S.-X","cited_arxiv_id":null,"evidence_quote":"Identifies A_s^{αβ} as the limit of the reproducing kernel for Gaussian Fisher–Hartwig weights with an edge-scaled singularity; the paper lifts this to Fredholm convergence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The Gaussian, β=1, τ-function result (Proposition 26) that the paper generalizes to all β and to general one-cut V."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the Tracy–Widom formulas and the Airy-kernel limit that the new results must reduce to in the case (α,β)=(0,1)."}],"review_version":1}