{"id":"5820e3ec-a372-4a60-8948-25ed1a4b9480","arxiv_id":"2412.03115","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For a symmetric subclass of 3x3-periodic hexagon tilings, the associated matrix valued orthogonal polynomials admit a full asymptotic expansion built from an explicit balayage equilibrium measure on a genus one spectral curve.","lead":"This paper proves strong asymptotic formulas for the 3x3 matrix valued orthogonal polynomials attached to a special class of doubly periodic lozenge tilings of a hexagon. The asymptotics are governed by an explicitly computed equilibrium measure on a genus one spectral curve, and they imply that zeros of the associated determinant accumulate on the unit circle.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.3 claims uniform asymptotics on all compact subsets of C\\T, but G is only defined on C\\(T∪[zmin,zmax]) and jumps across [zmin,zmax]; since zmin<1<zmax, the theorem is ill-posed as stated.","rationale":"The reader's verdict is CONDITIONAL and already includes the observation that the uniform domain in Theorem 1.3 should exclude the cut or clarify boundary values. I agree with that observation, but I see it as the single most load-bearing concern about the central claim, because it invalidates the theorem exactly as stated: the asymptotic formula is asserted on compact subsets that may contain points where the right-hand side is not defined. This is not merely a stylistic gap: the two sides of the cut give different values of G^N for odd N, while PN is entire, so no interpretation of the formula as a single-valued identity can hold on the cut. The rest of the paper gives a coherent steepest-descent argument and provides the jumps needed to repair the statement by working with the product AN G^N or with one-sided limits, so a REJECT verdict would be too strong; the appropriate disposition is a conditional acceptance requiring a restatement of Theorem 1.3 and a corresponding careful statement in Theorem 1.4(c). I do not elevate the imported Harnack-curve ordering into a separate objection: it is a published theorem, and the paper's explicit spectral curve and Lemma 2.3 supply the structural consequences needed for the construction, conditional on that theorem. The Maple-assisted identity in Lemma 2.1 is under-derived, but it is symbolically checkable and does not by itself threaten the central claim as much as the ill-posed formulation of Theorem 1.3. Therefore the reader's conditional verdict stands, and my stress test does not move it.","tokens_in":43884,"tokens_out":25381,"duration_ms":253441,"concrete_test":"Choose a compact interval K contained in (zmin,1), for instance K=[zmin+(1-zmin)/4, zmin+(1-zmin)/2], and attempt to verify (1.18) on K. The left-hand side PN(z) is entire, hence well defined for z∈K, while the right-hand side contains G(z)^N with G undefined on K by Theorem 1.3(a). Replacing K by a point x0 in (zmin,1) and using the boundary relations (1.24)-(1.25), compare the limits of the right-hand side as z→x0 from Im z>0 and Im z<0: the two limits differ by conjugation with diag(1,-1,-1) for odd N, whereas the left-hand side has a single limit PN(x0). Thus no choice of a single-valued G, AN, L can satisfy the theorem as printed. If the authors confirm that the intended uniformity domain is C\\(T∪[zmin,zmax]) (or that (1.18) is meant to hold only for the analytic product AN G^N on the cut), the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest stated claim is not well-defined in its printed form. Theorem 1.3 asserts (1.18) uniformly for z in compact subsets of C\\T, yet part (a) defines G only on C\\(T∪[zmin,zmax]), and part (d) gives the jump (1.24) across [zmin,1)∪(1,zmax]. Because Assumption 1.2(d) gives zmin<1<zmax, the cut [zmin,zmax] is contained in C\\T, so admissible compact sets may intersect it. At such points G(z)^N has no single defined value; for odd N the two boundary limits of G^N differ by conjugation with diag(1,-1,-1), while the left-hand side PN(z) is entire. The proof in Section 10.1 derives (1.18) only for z outside the lens around the unit circle; choosing the lens arbitrarily thin cannot repair this, since a compact set meeting (zmin,1) or (1,zmax) can still be separated from T but the cut point itself remains in the excluded domain of G. The intended statement must either restrict to compact subsets of C\\(T∪[zmin,zmax]) or specify one-sided boundary values on the cut, with the compensation by AN understood via (1.25). This is a literal inconsistency in the central theorem, though the surrounding RH construction appears to supply the ingredients needed for the corrected formulation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes matrix-valued orthogonal polynomials (MVOP) associated with 3x3-periodic lozenge tilings of a regular hexagon, in the special case B=C=1 under the symmetry assumptions (1.4)-(1.7). The spectral curve is reduced to the explicit genus-one curve (2.1), with branch points zmin and zmax. The authors define a probability measure mu as the balayage of delta_{P_infty} - delta_{P_1} + delta_{P_0} onto Gamma_1 union Gamma_2, prove that it is positive with real-analytic density, verify the variational equalities and the S-property, and use the associated g-functions in a Deift-Zhou steepest descent analysis of the Riemann-Hilbert problem. The main results are a strong asymptotic formula for the MVOP P_N (Theorem 1.3) and the weak convergence of the normalized zero counting measures of det P_N to the pushforward of mu onto the unit circle (Theorem 1.4). The appendix connects the MVOP to correlation kernels of random lozenge tilings.","tokens_in":44122,"tokens_out":10928,"duration_ms":98143,"significance":"If the stated results hold, this is a substantial advance: it provides the first full steepest-descent analysis for MVOP arising from doubly periodic lozenge tilings with period three, with an explicitly identified genus-one equilibrium measure and exponential error estimates. The paper is unusually concrete: the spectral curve is explicit, the positivity of the balayage measure is verified by an elementary calculation, the g-functions are defined from the equilibrium measure rather than fitted, and the global parametrix is built explicitly from Jacobi theta functions. The absence of local parametrices and the exponential smallness of the remainders are notable strengths. The main claims are falsifiable and the proofs are laid out in enough detail that the individual steps can be checked. The central issue identified below concerns the precise domain on which the strong asymptotic formula is stated, and a gap in one step of the proof of the zero-distribution theorem; both appear to be repairable without changing the overall strategy.","major_comments":[{"comment":"The statement says (1.18) holds uniformly for z in compact subsets of C\\T, but by part (a) G is defined and analytic only on C\\(T ∪ [zmin,zmax]), and part (d) gives a jump across [zmin,1)∪(1,zmax]. Since Assumption 1.2(d) gives zmin<1<zmax, the cut lies entirely in C\\T, so an admissible compact set may intersect it; at such a point G(z)^N has no single value, and for odd N the two boundary limits differ by conjugation with diag(1,-1,-1). The derivation in Section 10.1 proves (1.18) only outside the lens around T, and choosing the lens arbitrarily thin does not repair the problem, because cut points are separated from T but are not in the domain of G. The theorem should either restrict to compact subsets of C\\(T ∪ [zmin,zmax]) or specify one-sided boundary values on the cut, with the compensation by A_N understood via (1.24)-(1.25).","section":"Section 1.3, Theorem 1.3 and Eq. (1.18)"},{"comment":"The argument that det A_e and det A_o are not identically zero inside the unit circle is not fully justified. First, (1.26) is stated uniformly for compact subsets of C\\T, so it does not directly give the bound sup_{|z|≤1}|det P_N(z)| ≤ C e^{-cN} used in (10.5), because the closed unit disk meets T. Second, for odd N the function A_o is not analytic in C\\T: by Theorem 1.3(c) it is analytic in C\\(T ∪ [zmin,zmax]) and has the jump (1.25) across the cut, which splits the interior of the unit disk. The proof needs a boundary-value or maximum-principle argument on annuli or domains with boundary arcs, or a different argument, before one can conclude that the only zeros accumulating on T are those captured by the asymptotic formula.","section":"Section 10.3, proof of Theorem 1.4(c)"}],"minor_comments":[{"comment":"The sentence 'Thus there is k ∈ {2,3} such that E_{j,k}(z_j^*)' is incomplete; it should say that E_{j,k}(z_j^*) = 0.","section":"Section 8.2, paragraph after (8.8)"},{"comment":"The expression 'sup_{z∈T} det |P_N(z)|' should read 'sup_{z∈T} |det P_N(z)|'.","section":"Section 10.3"},{"comment":"The sentence 'Since η > 0 can be taken arbitrarily close to 0' should be made precise by referencing the positive distance from a given compact set to T; as written it can be read as claiming uniformity over all of C\\T, which interacts with the domain issue raised in Major Comment 1.","section":"Section 10.1, last paragraph"},{"comment":"The same symbol k1 is used in (6.8) and (6.9), while (6.10) uses k2; this is confusing because the three equations refer to three different connected components and the constants need not coincide. Renaming the constants would improve readability.","section":"Lemma 6.3(b)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong and detailed contribution that fits the journal's scope. The domain inconsistency in Theorem 1.3 is a genuine statement-level flaw but appears easily fixable, and the gap in Section 10.3 likewise seems repairable with a more careful boundary-value argument. I do not see grounds for rejection. The reliance on the Kenyon-Okounkov-Sheffield Harnack curve theorem is appropriate and clearly flagged."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear —,\n\nThe paper delivers the first strong asymptotic formula for MVOP coming from doubly periodic lozenge tilings of a hexagon in a nontrivial symmetric subclass, and the core mathematical argument holds up. What is new: an explicit equilibrium measure arising as a balayage of three point masses, a theta-function global parametrix on the genus-one spectral curve, and exponentially small errors from a Deift–Zhou steepest descent analysis. The positivity and analyticity of the equilibrium measure are proven by an explicit calculation, and the Riemann–Hilbert chain Y→X→T→S→R is laid out in detail. This is a genuine step beyond the prior two-periodic Aztec diamond results.\n\nThe soft spots are real but mostly at the level of statement. Theorem 1.3 as printed is not well-posed. It claims (1.18) uniformly for z in compact subsets of C\\T, but part (a) defines G only on C\\(T∪[zmin,zmax]) and part (d) gives a jump across [zmin,1)∪(1,zmax]. Since zmin<1<zmax, those intervals are inside C\\T. A compact set may meet them, and there G(z)^N has no single value; for odd N the two boundary limits differ by conjugation, while P_N(z) is entire. The proof derives (1.18) outside the lens around T, and arbitrarily thin lenses cannot fix this. This is a literal inconsistency in the central theorem, but it is fixable: restrict to compact subsets of C\\(T∪[zmin,zmax]) or specify one-sided boundary values with A_N compensating via (1.25). The same caveat applies to the statement of Theorem 1.4(c). I believe the RH construction supplies all ingredients for the corrected formulation, so the flaw is not load-bearing.\n\nTwo smaller notes. Lemma 2.1 uses a Maple-assisted identity for Q with no checkable derivation, which stands out in an otherwise explicit paper. And Section 10.3 ends with 'We proved part (d) of Theorem 1.4' when the proof is for part (c). Mislabel, not math.\n\nThe reliance on Kenyon–Okounkov–Sheffield for the Harnack structure is appropriate, and the paper is honest about the restricted parameter range B=C=1 and the symmetries (1.4)–(1.7). The remarks on generalization to B,C are careful.\n\nBottom line: this deserves a serious referee. I would engage with it, and I would bring it to reading group as a concrete example of RH analysis on a higher-genus curve. The authors need to correct the domain statements in Theorems 1.3 and 1.4(c) before publication, but the underlying mathematics appears sound.","headline":"First strong asymptotics for a nontrivial MVOP class with a solid core argument; the printed Theorem 1.3 has a domain-of-definition flaw that needs fixing before the results are citable as stated.","tokens_in":44706,"tokens_out":2426,"would_cite":true,"duration_ms":24638,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42C05","30E25","33E05","41A60"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a symmetric 3x3-periodic hexagon tiling, the zeros of the associated matrix-valued orthogonal polynomials converge to the unit circle with an explicitly identified density.","keywords":["matrix valued orthogonal polynomials","doubly periodic tilings","lozenge tilings of a hexagon","Riemann-Hilbert problem","steepest descent asymptotics","equilibrium measure","Harnack curves","genus one spectral curve"],"falsifier":"Take the explicit model (1.8) with $\\alpha_1=0.2$, $\\alpha_2=2$, compute the zeros of $\\det P_N$ numerically for large $N$ (say $N=100$ and $N=200$), and compare the empirical normalized counting measure with the pushforward $\\mu_*=(\\pi_z)_*\\mu$ of $\\mu=\\mathrm{Bal}(\\delta_{P_\\infty}-\\delta_{P_1}+\\delta_{P_0};\\Gamma_1\\cup\\Gamma_2)$; a mismatch, or any zeros accumulating away from the unit circle, would refute Theorems 1.3--1.4. A second check is to verify the normal-derivative identity (1.41) for the harmonic function $h=2U^\\mu+\\mathrm{Re}\\,V-\\ell$ using the $\\theta$-function formulas from Section 4.","tokens_in":43602,"feed_emoji":"🧩","tokens_out":20716,"duration_ms":175902,"temperature":0.7,"pith_summary":"Matrix-valued orthogonal polynomials (MVOP) are polynomials with matrix coefficients that are orthogonal with respect to a matrix weight on a contour; they encode the correlation functions of doubly periodic lozenge tilings of a hexagon. This paper studies the special 3x3-periodic weights satisfying the symmetry conditions (1.4)--(1.7) in the regular-hexagon case $B=C=1$, and proves a strong asymptotic formula for the degree-$N$ MVOP $P_N$: for $z$ away from the unit circle, $P_N(z)=L^N(A_N(z)+O(e^{-cN}/(1+|z|)))G(z)^N$ with an explicit matrix $G$ built from the $g$-functions of the spectral curve. The central result is that the normalized zero counting measures of $\\det P_N$ converge weakly, as $N\\to\\infty$, to a probability measure on the unit circle with a positive real-analytic density, identified explicitly as the balayage (harmonic sweep) of $\\delta_{P_\\infty}-\\delta_{P_1}+\\delta_{P_0}$ onto the first two sheets' unit circles. If the paper is right, this supplies a complete large-$N$ analysis of MVOP from hexagon tilings in a genuinely two-parameter family, and it exhibits the equilibrium measure with the S-property in the external field $\\mathrm{Re}\\,V=2\\log|z|-2\\log|\\lambda|$ as the organizing principle.","feed_headline":"Zeros of hexagon-tiling polynomials converge to an explicit density","feed_subtitle":"The density is a balayage of three special points on a genus-one spectral curve, so the tiling asymptotics is fully explicit.","key_machinery":"The machinery is organized around the genus-one spectral curve $P(z,\\lambda)=(\\lambda-z-1)^3-27(1+\\beta)\\lambda z=0$, a Harnack curve (an algebraic curve whose logarithmic amoeba map is at most two-to-one) with holomorphic symmetries $(z,\\lambda)\\mapsto(-\\lambda,-z)$ and $(z,\\lambda)\\mapsto(z^{-1},\\lambda/z)$ plus complex conjugation. The strict ordering $|\\lambda_1(z)|>|\\lambda_2(z)|>|\\lambda_3(z)|>0$ off the real axis selects three sheets and defines the eigenvector matrix $E(z)$. The Abel map sends this curve to the complex torus $\\mathbb{C}/(\\mathbb{Z}+\\tau\\mathbb{Z})$, with the three special points $P_\\infty,P_1,P_0$ at $1/6,1/2,5/6$ and the circles $\\Gamma_1\\cup\\Gamma_2$, $\\Gamma_3$ on the two halves of the $b$-cycle. The candidate limiting measure is $\\mu=\\mathrm{Bal}(\\delta_{P_\\infty}-\\delta_{P_1}+\\delta_{P_0};\\Gamma_1\\cup\\Gamma_2)$; its positivity is verified by transporting it to a vertical strip and then to the upper half-plane, where it becomes a sum of three Cauchy densities that is positive. The $g$-functions are defined from the complexified bipolar Green's kernel $G^C_{P_\\infty}$, written with Jacobi $\\theta$ functions $\\theta_1$, and satisfy the jump and normalization conditions that make the Riemann-Hilbert transformations work. The global parametrix is built row by row from meromorphic functions $\\psi_j,\\phi_j$ constructed from Jacobi $\\theta$ functions, with prescribed poles at $Q^*_j$ and zeros, arranged as Hadamard products $E\\circ\\Psi$ and $E\\circ\\Phi$, with a constant matrix $K$ fixing the behavior at infinity; parity of $N$ selects one of two such constructions. The final transformation $S\\to R$ reduces to a small-norm Riemann-Hilbert problem, giving $R=I_6+O(e^{-cN})$ and hence the exponential error in Theorem 1.3.","core_discovery":"The paper's central discovery is that, under the symmetry assumptions (1.4)--(1.7) with $B=C=1$, the monic matrix-valued orthogonal polynomial $P_N$ associated with the 3x3-periodic hexagon model has the strong asymptotic expansion $P_N(z)=L^N(A_N(z)+O(e^{-cN}/(1+|z|)))G(z)^N$ uniformly for $z$ in compact subsets of $\\mathbb{C}\\setminus\\mathbb{T}$, where $L$ is unit lower triangular, $A_N$ depends only on the parity of $N$, and $G(z)=E(z)\\operatorname{diag}(e^{g_1(z)},e^{g_2(z)},e^{g_3(z)})E(z)^{-1}$; the $g_j$ are normalized so that $g_j(z)=\\log z+O(z^{-1/3})$ as $z\\to\\infty$. From this formula the author derives that the normalized zero counting measures of $\\det P_N$ converge weakly to $\\mu_*=(\\pi_z)_*\\mu$, where $\\mu=\\mathrm{Bal}(\\delta_{P_\\infty}-\\delta_{P_1}+\\delta_{P_0};\\Gamma_1\\cup\\Gamma_2)$ is supported on the two unit circles on sheets 1 and 2 and has a positive real-analytic density. The measure $\\mu$ is characterized by the harmonic identity $\\int h\\,d\\mu=h(P_\\infty)-h(P_1)+h(P_0)$ for functions harmonic off $\\Gamma_1\\cup\\Gamma_2$, and it is shown to be the equilibrium measure of $\\Gamma_1\\cup\\Gamma_2\\cup\\Gamma_3$ in the external field $\\mathrm{Re}\\,V$ with the S-property: $2U^\\mu+\\mathrm{Re}\\,V=\\ell$ on $\\Gamma_1\\cup\\Gamma_2$ and $2U^\\mu+\\mathrm{Re}\\,V>\\ell$ on $\\Gamma_3$. The spectral curve behind all this is the explicit genus-one Harnack curve $(\\lambda-z-1)^3=27(1+\\beta)\\lambda z$ with $\\beta>0$, whose three special points $P_0,P_\\infty,P_1$ sit at $z=0$, $z=\\infty$, and $(-1,0)$ respectively.","pith_inferences":["The paper leaves general $B,C$ open; following Remark 1.8, for $B=C$ the equality in the Euler-Lagrange conditions on $\\Gamma_1\\cup\\Gamma_2$ persists with the generalized balayage (1.43), so a natural test is whether positivity, the strict inequality on $\\Gamma_3$, and the S-property hold for all $0<C\\le 1$, which would extend the strong asymptotics beyond the regular hexagon.","Because the same Riemann-Hilbert solution enters the correlation kernel (A.4), the explicit equilibrium measure should determine the rough-phase geometry and the arctic curve of the random tiling model; the paper states the derivation of these tiling consequences as a future step.","The same scheme---an explicit genus-one spectral curve, the Abel map to a torus, a theta-function global parametrix, and a full-support equilibrium measure---is the natural template for higher periodicities with analogous symmetries, with the three-point balayage replaced by the divisor supplied by Abel's theorem.","The explicit formula for $\\mu$ implies the limiting zero density is invariant under complex conjugation $z\\mapsto\\bar{z}$ (equivalently $\\theta\\mapsto-\\theta$ on the unit circle), a symmetry visible in the $N=20$ plot and directly checkable for larger $N$."],"forward_implications":["For the model (1.8), the zeros of $\\det P_N$ cluster on the unit circle with a density that can be computed from the parameter $\\beta$ alone, so the asymptotic zero distribution is completely determined by the single parameter $\\beta$ (equivalently, by $\\alpha_1,\\alpha_2$).","The strong formula (1.18) determines $\\det P_N$ up to exponentially small relative error on compact sets away from the unit circle; in particular, no zeros accumulate outside $\\mathbb{T}$ and the convergence in Theorem 1.4 is exponentially fast away from the circle.","Because $\\mu$ has full support on $\\Gamma_1\\cup\\Gamma_2$ with a positive analytic density, the lens-opening step does not require local parametrices, which is why the error term in (1.18) is exponentially small rather than merely $O(N^{-1})$.","The parity dependence of $A_N$ and the explicit jump relations (1.24)--(1.25) imply that the even- and odd-$N$ polynomials have different leading analytic factors but share the same limiting zero measure."],"supporting_citations":[{"why":"Establishes the connection between matrix-valued orthogonal polynomials and doubly periodic tilings and supplies the correlation-kernel formula used to motivate the orthogonality and the first transformation.","marker":"[32]"},{"why":"Proves unique existence of the MVOP from the tiling model and identifies the scalar orthogonality on the spectral curve behind the matrix orthogonality.","marker":"[37]"},{"why":"Supplies the Harnack-curve theorem for planar dimers that gives the strict eigenvalue ordering underlying the three-sheet structure.","marker":"[41]"},{"why":"Gives the amoeba/Harnack property and the sheet structure of the spectral curve, used throughout for the ordering $|\\lambda_1|>|\\lambda_2|>|\\lambda_3|$.","marker":"[42]"},{"why":"Provides the bipolar Green's kernel, equilibrium-measure and S-property framework on higher genus Riemann surfaces used in Theorems 1.4 and 1.7.","marker":"[10]"},{"why":"Gives the 2r x 2r Riemann-Hilbert characterization of matrix orthogonal polynomials that is the starting point of Riemann-Hilbert problem 1.10.","marker":"[38]"},{"why":"Supplies the standard small-norm Riemann-Hilbert estimates and uniqueness arguments used to conclude $R=I_6+O(e^{-cN})$.","marker":"[26]"},{"why":"Provides the lower-envelope and unicity theorems for logarithmic potentials used to pass from the $\\log|\\det P_N|$ asymptotics to weak convergence of zero counting measures.","marker":"[52]"},{"why":"Gives the explicit theta-function formula for the bipolar Green's kernel on a torus used to define the $g$-functions and to prove the variational identities.","marker":"[53]"}],"fun_headline_variants":["Hexagon tilings yield explicit zero density for matrix polynomials","Explicit limit density for zeros in periodic hexagon tilings","Zeros of hexagon tiling polynomials obey explicit measure","Explicit equilibrium measure for hexagon tiling zeros","Genus-one curve pins down zeros in hexagon tilings"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on the imported theorem, not reproved in the paper, that the doubly periodic weight matrix $W(z)$ has a Harnack spectral curve whose eigenvalues obey the strict ordering $|\\lambda_1(z)|>|\\lambda_2(z)|>|\\lambda_3(z)|>0$ for $z\\notin\\mathbb{R}$; all sheet labels, the eigenvector matrix $E(z)$, and the first Riemann-Hilbert transformation depend on that ordering.","fun_headline_variants_meta":{"raw":{"variants":["Hexagon tilings yield explicit zero density for matrix polynomials","Explicit limit density for zeros in periodic hexagon tilings","Zeros of hexagon tiling polynomials obey explicit measure","Explicit equilibrium measure for hexagon tiling zeros","Genus-one curve pins down zeros in hexagon tilings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000585,"raw_usage":{"total_tokens":2880,"prompt_tokens":1204,"completion_tokens":1676,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":820,"completion_tokens_details":{"reasoning_tokens":1595}},"tokens_in":820,"tokens_out":1676,"duration_ms":12250,"temperature":1.0,"reasoning_tokens":1595,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:46:43.685295+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the explicit model (1.8) with $\\alpha_1=0.2$, $\\alpha_2=2$, compute the zeros of $\\det P_N$ numerically for large $N$ (say $N=100$ and $N=200$), and compare the empirical normalized counting measure with the pushforward $\\mu_*=(\\pi_z)_*\\mu$ of $\\mu=\\mathrm{Bal}(\\delta_{P_\\infty}-\\delta_{P_1}+\\delta_{P_0};\\Gamma_1\\cup\\Gamma_2)$; a mismatch, or any zeros accumulating away from the unit circle, would refute Theorems 1.3--1.4. A second check is to verify the normal-derivative identity (1.41) for the harmonic function $h=2U^\\mu+\\mathrm{Re}\\,V-\\ell$ using the $\\theta$-function formulas from Section 4.","supporting_citations":[{"cited_title":"Duits and A.B.J","cited_arxiv_id":null,"evidence_quote":"Establishes the connection between matrix-valued orthogonal polynomials and doubly periodic tilings and supplies the correlation-kernel formula used to motivate the orthogonality and the first transformation."},{"cited_title":"Groot and A.B.J","cited_arxiv_id":null,"evidence_quote":"Proves unique existence of the MVOP from the tiling model and identifies the scalar orthogonality on the spectral curve behind the matrix orthogonality."},{"cited_title":"Kenyon and A","cited_arxiv_id":null,"evidence_quote":"Supplies the Harnack-curve theorem for planar dimers that gives the strict eigenvalue ordering underlying the three-sheet structure."},{"cited_title":"Kenyon, A","cited_arxiv_id":null,"evidence_quote":"Gives the amoeba/Harnack property and the sheet structure of the spectral curve, used throughout for the ordering $|\\lambda_1|>|\\lambda_2|>|\\lambda_3|$."},{"cited_title":"Bertola, A","cited_arxiv_id":null,"evidence_quote":"Provides the bipolar Green's kernel, equilibrium-measure and S-property framework on higher genus Riemann surfaces used in Theorems 1.4 and 1.7."},{"cited_title":"Gr¨ unbaum, M.D","cited_arxiv_id":null,"evidence_quote":"Gives the 2r x 2r Riemann-Hilbert characterization of matrix orthogonal polynomials that is the starting point of Riemann-Hilbert problem 1.10."},{"cited_title":"Deift, Orthogonal Polynomials and Random Matrices: a Riemann–Hilbert Approach, Courant Lecture Notes in Mathematics 3","cited_arxiv_id":null,"evidence_quote":"Supplies the standard small-norm Riemann-Hilbert estimates and uniqueness arguments used to conclude $R=I_6+O(e^{-cN})$."},{"cited_title":"Saff and V","cited_arxiv_id":null,"evidence_quote":"Provides the lower-envelope and unicity theorems for logarithmic potentials used to pass from the $\\log|\\det P_N|$ asymptotics to weak convergence of zero counting measures."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the explicit theta-function formula for the bipolar Green's kernel on a torus used to define the $g$-functions and to prove the variational identities."}],"review_version":1}