{"id":"a4dfe723-a729-4e2c-85ab-fa58409e2f53","arxiv_id":"2607.16333","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Non-symmetric non-backtracking vector Dyson equations have square-root edges, cubic-root cusps, and complete stability estimates.","lead":"This paper analyzes the vector Dyson equation, a self-consistency equation used to describe spectral densities of random matrices, when the underlying matrix is allowed to be non-symmetric. It proves that for equations coming from non-backtracking tree structures, the solution has square-root edges, cubic-root cusps, and controlled stability, giving a tool for a class of problems earlier theory could not handle.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.1's Puiseux-expansion statement is unproved and the cited [8, Prop. 4.2] may not cover arbitrary nonnegative S; the finiteness of the singular set, support structure, and edge/cusp classification all depend on it.","rationale":"The reader's weakest assumption is exactly the concern I consider most load-bearing: Lemma 3.1 is used to establish the entire qualitative framework (finite singular set, finite union of intervals, Hölder regularity) that all later singularity and stability statements depend on, yet its proof is absent and the cited reference is not a standard source for this general setting. My independent reading of the manuscript confirms the paper's internal logic is otherwise coherent: the stability arguments in Sections 4–5 are detailed, the counterexample in Appendix A is explicit, and the bulk-stability proofs (Propositions 5.5, 5.7, 5.8) do not hide obvious circular steps. The secondary uniform-positivity issue is real but less severe, as one could interpret 'model parameters' as including the minimum positive entry of S; however, this should be stated explicitly. Since the reader already assigned CONDITIONAL, and my stress-test identifies the same foundational gap without finding a new fatal flaw, the appropriate verdict is unchanged. The concrete test I propose would resolve whether Lemma 3.1 is in fact valid, and if it fails, the central claim of complete stability estimates would be unsupported.","tokens_in":44856,"tokens_out":23811,"duration_ms":211825,"concrete_test":"Locate [8, Proposition 4.2] and verify that it proves the exact Puiseux statement for arbitrary nonnegative irreducible S (and not only for universal-cover Green functions). Independently, for a fixed S,a, symbolically eliminate variables in the d quadratic equations to obtain, for each i, a nonzero polynomial P_i(z,m_i)=0 satisfied by the physical branch; existence of such a polynomial implies m_i is algebraic and hence has Puiseux expansions, settling Lemma 3.1. If the resultant is identically zero for some i, or if [8] does not cover the general case, the missing proof must be supplied.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 3.1 asserts a convergent Puiseux expansion for every component m_i(z) near every real τ, citing 'Proof of Proposition 4.2 in [8]'. The lemma is never proved in the paper, and the cited reference is a paper on quantum ergodicity on graphs, not a general treatment of vector Dyson equations. This is not a cosmetic gap: the lemma is the sole support for the finiteness of the singular set A_j, the measure decomposition (2.5), the finite-union-of-intervals structure of the support S, the Hölder regularity in Theorem 2.6, and ultimately the edge/cusp classification in Theorems 2.7–2.8. If the Puiseux expansion failed at some real point—for instance, if the physical branch had an essential singularity or a non-algebraic branch point—the qualitative framework would collapse. A secondary, related under-specification is that the proofs of Theorem 2.5 and Proposition 4.4 rely on positive entries of S being uniformly bounded below (so that κ_{Q_F} ~ 1); this assumption is not stated in the hypotheses but is used to obtain uniform constants.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the vector Dyson equation -1/m(z) = z1 - a + S m(z) with entrywise nonnegative, not necessarily symmetric S. It proves existence/uniqueness and Stieltjes representation for all z in C+ (Thm 2.1), a measure decomposition with finite singular set and finite-union-of-intervals support (Thm 2.3), and Hölder regularity under local boundedness (Thm 2.6). The main results concern the stability operator I - m(z)^2 S: for symmetric matrices (Sym) and for weighted non-backtracking matrices (NB) arising from connected base graphs of minimum degree at least 3, the paper proves square-root growth at regular edges and cubic-root growth at regular cusps (Thm 2.7), and optimal bounds on the inverse of the stability operator in edge, cusp, bulk, and off-support regimes (Thm 2.8), including a |z+a|^(-1) bulk instability in a bipartite exceptional case. The proof strategy combines a Perron-Frobenius/Markov-chain normalization of the non-symmetric matrix F, a graph-theoretic resonance set for the phase matrix U, and a reduction to a scalar cubic equation. An appendix gives a cyclic-permutation counterexample showing that arbitrary nonnegative irreducible S can lose bulk stability.","tokens_in":1714,"tokens_out":1534,"duration_ms":158990,"significance":"If the results are correct, this is a meaningful extension of the Ajanki–Erdős–Krüger stability theory: it covers the non-symmetric equations arising from Green's functions on universal covers of finite graphs, and it identifies possible degeneracies through graph-theoretic resonance. The explicit Perron-Frobenius normalization (Lem. 4.2, Prop. 4.4), the resonance-set criterion (Lem. 5.4), and the concrete counterexample (App. A) are valuable concrete contributions. The paper is free of fitted parameters and does not appear circular: prior results [2,5,8,11,12,29] are used in a standard way. However, several load-bearing ingredients are delegated or implicit, and they need to be supplied before the main claims can be considered fully verified.","major_comments":[{"comment":"Lemma 3.1 asserts, without proof, a convergent Puiseux expansion for every component m_i(z) near every real tau for arbitrary nonnegative S, citing 'Proof of Proposition 4.2 in [8]'. This lemma is load-bearing: Theorem 2.3 Step 1 uses it to make A_j finite, Step 4 uses it to make the support S a finite union of intervals, and Theorem 2.6 uses it for Hölder regularity. The cited [8] is a paper on quantum ergodicity on graphs, not a source for vector Dyson equations. The manuscript should either prove Lemma 3.1 (e.g. from algebraicity of the solution) or give a precise theorem in the literature that covers exactly this setting. As written, the qualitative framework of the paper rests on an unverified assertion.","section":"§3, Lemma 3.1"},{"comment":"The proof of Theorem 2.5 says 'Since positive entries of S are uniformly bounded from below', and the proof of Proposition 4.4 uses this to conclude kappa_Q_F ~ 1. This uniform lower bound is not stated in the hypotheses of Theorem 2.5 or in Assumptions (Sym)/(NB). Without it, f_ij(z) ~ s_ij/(1+|z|)^2 does not imply that positive entries of Q_F are comparable to 1, and the constants in (2.8), (2.9), (4.14), and in Theorem 2.8 need not be uniform. Please add the lower-bound assumption explicitly, or define 'model parameters' to include min_{s_ij>0} s_ij and verify it for (NB).","section":"§3, Thm 2.5; §4, Prop. 4.4"},{"comment":"Proposition 4.8, which yields the approximate cubic equation (4.54) and error bounds (4.55), is the bridge from the stability expansion to Theorems 2.7–2.8. Its proof consists of a statement that the argument is 'essentially identical' to [2, Prop. 6.2] after replacing the orthogonal projection Q by the oblique projection P = I - r l^T. The non-symmetric setting changes the linear algebra in a non-trivial way, and the estimates (4.55a)–(4.55b) are not derived. Please include a self-contained proof, or a detailed translation that verifies every estimate in the oblique-coordinate case.","section":"§4, Prop. 4.8"},{"comment":"In the (Sym) bipartite case of Proposition 4.7, the exclusion of the alternating mode at tau = 0 uses a reflection argument with an invalid formula for the map z -> -z and then evaluates at z = i eta with '-z = z'. This step is needed for psi(tau)>0 when sigma(tau)=0, hence for the cusp part of Theorem 2.7; Proposition 5.8 also refers back to this argument to conclude m(0) is purely imaginary. Please correct the reflection map and provide a valid proof, or the bipartite exceptional cases are not supported.","section":"§4, Prop. 4.7; §5, Prop. 5.8"}],"minor_comments":[{"comment":"The displayed reflection formula contains typographical errors; it should be corrected and a proper proof of the symmetry property should be given.","section":"§4, Prop. 4.7"},{"comment":"The set of 'model parameters' is never defined. The paper repeatedly uses constants depending on them, and the implicit lower bound on positive entries of S is part of this dependence; a brief definition would remove ambiguity.","section":"§1.3"},{"comment":"In the (NB) case, the phrase 'a in R1' should be explained: a is indexed by directed edges, so the condition means a_(x,y) = a_y = a constant independent of y.","section":"Thm 2.8"}],"recommendation":"major_revision","confidential_remarks":"I think the paper is worth publishing after major revision. The most important fixes are: prove Lemma 3.1; make Proposition 4.8 self-contained; correct the reflection argument in Prop. 4.7/5.8; and state the uniform lower bound on positive entries. I did not find circularity or fitted parameters."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's your takeaway: this is a real contribution. It extends the deterministic stability theory of vector Dyson equations from symmetric S to non-symmetric, non-backtracking matrices, which is exactly the setting that arises for Green functions on universal covers and trees of finite cone type. The two new tools — the Perron-Frobenius Markov normalization and the graph-theoretic resonance set — are genuinely useful, and the bulk stability arguments in Sections 4 and 5 are detailed and internally coherent. The cyclic permutation counterexample in Appendix A is clean and shows why structural assumptions are necessary. I did not find circular reasoning or fitted parameters.\n\nThe soft spots are concentrated in one place. Lemma 3.1 claims that every component m_i(z) has a convergent Puiseux expansion at every real point, citing 'Proof of Proposition 4.2 in [8]'. But [8] is a paper on quantum ergodicity on graphs, not a general treatment of vector Dyson equations, and the lemma is never proved in the text. This is load-bearing: it underpins the finiteness of the singular set, the measure decomposition, the finite-union-of-intervals support structure, and the Hölder regularity in Theorem 2.6. Without it, the entire edge/cusp classification has no foundation. I'm not saying the claim is false — some Puiseux behavior is plausible from the algebraic structure — but it needs either a real proof or a reference that actually covers this class of equations.\n\nSecond, the proofs in Theorem 2.5 and Proposition 4.4 use that positive entries of S are uniformly bounded below, so that κ_Q ~ 1. That assumption is never stated in the hypotheses. It's an easy fix, but it should be explicit because the model-parameter constants depend on it.\n\nThird, the paper delegates the cubic-equation derivation and the final singularity classification to [2, Prop 6.2 / Thm 2.6]. That's a legitimate use of prior work if the notational translation genuinely works, and I believe it does, but a referee should check that the non-symmetric Perron coordinates don't introduce extra terms.\n\nOverall, the central argument holds up as far as I can tell, provided Lemma 3.1 is supplied. This paper deserves a serious referee, not a desk reject. I'd send it to review and ask for the Puiseux lemma to be proved or precisely cited, and for the positivity assumption to be stated.","headline":"A genuinely new stability theory for non-symmetric vector Dyson equations, but the unproved Puiseux lemma and an implicit positivity assumption need fixing before I'd trust the qualitative framework.","tokens_in":45618,"tokens_out":1917,"would_cite":true,"duration_ms":20412,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15A18","15B48","47A10","60B20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Non-symmetric vector Dyson equations with symmetric or non-backtracking S have fully controlled stability: inverse size (κ+η)^(-1/2) at edges, (κ+η)^(-2/3) at cusps, bounded in the bulk except at a bipartite point.","keywords":["vector Dyson equation","non-symmetric matrix","stability operator","non-backtracking matrix","resonance set","square-root edge","cubic-root cusp"],"falsifier":"Compute the boundary behavior of m for a small irreducible nonnegative non-symmetric S at a point where the density vanishes: if some component diverges like a power with irrational exponent, or like log η, rather than admitting a rational-power Puiseux expansion, then Lemma 3.1 fails and the decomposition theorem collapses. A more targeted check: at a regular cusp of a non-backtracking matrix on a 3-regular graph, numerically estimate the density exponent and the inverse stability norm; observing growth faster than |ω|^(1/3) or η^(-2/3) would falsify Theorems 2.7 and 2.8.","tokens_in":44727,"feed_emoji":"📈","tokens_out":10435,"duration_ms":86324,"temperature":0.7,"pith_summary":"The paper studies the vector Dyson equation -1/m(z) = z1 + a + S m(z), the nonlinear system whose solution encodes the limiting spectral density for certain random-matrix and tree-operator models. Its central claim is that, when the nonnegative matrix S is symmetric or has the non-backtracking structure coming from universal covers of finite graphs, the linearized stability operator I - m(z)^2 S is completely understood: near a regular edge its inverse has size of order (distance+η)^(-1/2), near a regular cusp at most (distance+η)^(-2/3), and in the bulk it stays bounded except in a bipartite exceptional case where it grows like |z+a|^(-1). From this stability control the authors derive that the limiting density grows like a square root at regular edges and like a cube root at regular cusps. A sympathetic reader would care because these sharp bounds on the stability operator are the deterministic inputs needed for local spectral laws and universality, and because the graph-theoretic method reveals which support-graph structures can make the operator unstable. The paper also shows that for a general nonnegative S bulk stability can genuinely fail, as in the cyclic permutation example.","feed_headline":"Square-root edges and cubic cusps for non-symmetric Dyson equations","feed_subtitle":"Random-matrix and tree-operator spectral analysis now has the deterministic input it was missing.","key_machinery":"The central object is the stability operator I - m(z)^2 S and its equivalent phase-modulus form R = U - F, where U is a unitary diagonal of phase ratios and F = |m| S |m| is a nonnegative matrix with the same support as S. In the symmetric theory F is self-adjoint and its norm is bounded by one; here that fails. The authors replace it with a Perron-Frobenius normalization of F into a Markov transition matrix Q_F, using a quantitative mixing-space estimate for I - λQ on the complement of the Perron direction. For bulk stability they introduce the resonance set R_Q of phase matrices diag(v_i) satisfying a cycle product condition, and prove that a small singular value of U - λQ forces U near R_","core_discovery":"The paper establishes that the stability operator I - m(z)^2 S, the linearization of the vector Dyson equation -1/m(z) = z1 + a + S m(z), is fully controlled for two classes of nonnegative irreducible matrices: symmetric S and non-backtracking matrices (NB). Theorem 2.8 gives two-sided bounds: at a regular edge, ||(I - m^2 S)^(-1)|| ~ max{1,(κ+η)^(-1/2)}; at a regular cusp, under (Sym) or (NB), it is at most (κ+η)^(-2/3); in the strict bulk it is bounded unless the relevant graph is bipartite and a is a constant vector, in which case the inverse grows like |z+a|^(-1). Theorem 2.7 converts these bounds into density singularities: v_i(τ±ω)=c_i ω^(1/2)+O(ω) at edges and v_i(τ+ω)=c_i |ω|^(1/3)+O","pith_inferences":["The sharp stability estimates would likely provide the deterministic core needed to extend local-law and universality proofs from symmetric to non-backtracking variance profiles, a step the paper itself does not take.","The resonance set R_S is a computable diagnostic: for a proposed non-symmetric S, checking R_S predicts where bulk instability can occur, as the cyclic permutation example already illustrates.","The minimum-degree-three condition in (NB) is used to force phase alignment on the non-backtracking graph; relaxing it to degree two may create new resonance modes and change the cusp picture, which the paper leaves open.","Because the Puiseux-expansion lemma is cited rather than proved, a direct algebraic proof for irreducible nonnegative S would make the qualitative framework self-contained."],"forward_implications":["At regular edges the limiting density has square-root growth, and at regular cusps cubic-root growth, in both the symmetric and the non-backtracking non-symmetric settings.","The inverse stability operator has two-sided bound (κ+η)^(-1/2) at regular edges and upper bound (κ+η)^(-2/3) at cusps, meaning perturbations are amplified only at the scale of the density singularity.","In the strict bulk the stability operator is bounded away from singularities, except in the bipartite case with a=a1, where z=-a is the only instability point and the inverse grows like |z+a|^(-1).","Every irreducible nonnegative S yields a density whose singular set is finite and whose support is a finite union of intervals; the measure decomposition holds without symmetry.","In the symmetric setting the primitivity assumption is removed, giving a complete characterization of bulk stability including the bipartite exceptional case."],"fun_headline_variants":["Square-root edges and cubic cusps proved for non-symmetric Dyson equations","Non-backtracking matrices yield exact stability for Dyson equations","Edge and cusp singularities in vector Dyson equations, even non-symmetric","Stability operator controlled for non-symmetric Dyson via graph structure","Cubic cusps and square-root edges: new laws for non-symmetric Dyson"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The framework rests on the unproved premise that, near every real point, each component of the solution expands in a convergent fractional-power series (a Puiseux expansion), and that the nonzero entries of S stay uniformly bounded away from zero; if either fails, the finite singular set, measure decomposition, interval support, and edge/cusp classification collapse.","fun_headline_variants_meta":{"raw":{"variants":["Square-root edges and cubic cusps proved for non-symmetric Dyson equations","Non-backtracking matrices yield exact stability for Dyson equations","Edge and cusp singularities in vector Dyson equations, even non-symmetric","Stability operator controlled for non-symmetric Dyson via graph structure","Cubic cusps and square-root edges: new laws for non-symmetric Dyson"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000946,"raw_usage":{"total_tokens":3888,"prompt_tokens":768,"completion_tokens":3120,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":512,"completion_tokens_details":{"reasoning_tokens":3029}},"tokens_in":512,"tokens_out":3120,"duration_ms":19324,"temperature":1.0,"reasoning_tokens":3029,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T04:16:09.051193+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the boundary behavior of m for a small irreducible nonnegative non-symmetric S at a point where the density vanishes: if some component diverges like a power with irrational exponent, or like log η, rather than admitting a rational-power Puiseux expansion, then Lemma 3.1 fails and the decomposition theorem collapses. A more targeted check: at a regular cusp of a non-backtracking matrix on a 3-regular graph, numerically estimate the density exponent and the inverse stability norm; observing growth faster than |ω|^(1/3) or η^(-2/3) would falsify Theorems 2.7 and 2.8.","supporting_citations":[],"review_version":2}