{"id":"9647eba4-9e83-43d1-977f-3581ecf098ce","arxiv_id":"2608.11043","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Constructs the multi-cut matrix Szegő factorization and proves strong asymptotics for matrix orthogonal polynomials via a vector-bundle global parametrix.","lead":"This paper constructs a matrix-valued Szegő factor for weights living on several disjoint intervals, proving the factor changes by constant unitary matrices across the gaps. It uses this to derive strong asymptotic formulas for matrix-valued orthogonal polynomials on multiple cuts, extending earlier single-interval results.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.4's pairing U(P)V(τP)^T is not holomorphic because τ is anti-holomorphic; this invalidates Theorem 3.10 and the Section 4 existence proof for RH 1.10.","rationale":"I read the paper's central claim as follows: Theorem 1.4 gives the multi-cut matrix Szegő factorization; RH Problem 1.10 is uniquely solvable via the flat unitary bundle E_K or via the Fredholm and vanishing alternative; this solvability feeds the reduction in Section 6 and produces (6.17) and (6.28). For the central claim to hold, the bundle proof must be valid or Section 5 must be fully invoked. The weakest point is Proposition 3.4. The map τ is anti-holomorphic by Definition 3.1, so V(τP) is not a holomorphic function of P; the paper's assertion to the contrary is not a benign typo. The subsequent equality U V^T = U V^* on ∂S is also false for complex frames. These two errors destroy the proof that ⟨U,U⟩_τ η is a holomorphic differential, and hence Theorem 3.10, Proposition 3.12, Lemma 4.4, and Proposition 4.5 do not follow. Since Section 6 explicitly invokes Proposition 4.5 and Proposition 4.9 for the global reduction, the asymptotic derivation has a hole unless the author either repairs the pairing or redirects the reader to Section 5 with the missing endpoint estimate. I do not think the local-parametrix transfer is the main issue: the ansatz (6.16) does cancel the gap jump e^{-inω_jσ3}, and the remaining matching is exactly [6, §3.5]; it is abbreviated but plausible. The Reader already noted the transpose and adjoint slip, but treating it as secondary undersells its impact; I therefore mark agreement as partial. Because Section 5 may provide an independent route and the mathematical statement may be true, I do not move to reject; the same CONDITIONAL verdict stands, with the repair of Proposition 3.4 as the binding condition. The concrete scalar check above settles whether the concern lands.","tokens_in":27965,"tokens_out":23398,"duration_ms":222046,"concrete_test":"Set r=1, fix a local holomorphic coordinate z on an open set of X disjoint from ∂S and from the Γ_j, and take U=V=z in the frame e_X0. Then ⟨U,V⟩_τ(P)=z z̄ = |z|^2, which is not holomorphic; equivalently ∂̄(V∘τ)≠0. Recompute the proof of Theorem 3.10 with this example: the alleged identity dω=0 fails because ω is not a holomorphic (1,0)-form. If this is confirmed, Proposition 3.4 cannot be used as it stands, and Section 4 must be repaired or replaced by Section 5 in the proof of Propositions 6.5 and 6.11.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.4 defines ⟨U,V⟩_τ(P)=U(P)V(τP)^T and says that P ↦ V(τP) is holomorphic. Since τ(z,y)=(z̄,−ȳ) is anti-holomorphic, this is false in general: in a local holomorphic coordinate z not on ∂S, V(z)=z gives V(τP)=z̄, which is anti-holomorphic. The proposition also identifies U V^T with U V^* on ∂S, which fails unless sections are real in that frame. The proof of Theorem 3.10 then builds ω=⟨U,U⟩_τ η, calls it a holomorphic 1-form, and uses dω=0; without holomorphy the Stokes argument has no basis. This is the sole argument for Theorem 3.10, hence for the isomorphism in Proposition 3.12 and for the bundle construction of Lemma 4.4 and Proposition 4.5. Section 6 explicitly solves RH Problems 6.4 and 6.10 by Propositions 4.5 and 4.9, so the final asymptotic formulas inherit the gap. Section 5 gives an independent Fredholm route, but it is not invoked in Section 6 and it would still need an estimate converting the L2 solution into the pointwise N4 endpoint bound. The local-parametrix transfer flagged by the Reader is plausible; the non-holomorphic pairing is the concrete defect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a matrix Szegő factorization for matrix-valued weights supported on a finite union of intervals, showing that the factor has constant unitary jumps across the gaps. It then builds the global parametrix for the associated Riemann–Hilbert problem using sections of flat unitary vector bundles over the hyperelliptic curve determined by the multi-cut set, and applies this machinery to obtain strong asymptotics for matrix-valued orthogonal polynomials in two settings: fixed Jacobi-type weights on multiple intervals and varying exponential weights whose equilibrium measure has multi-component support. An alternative proof of solvability via Toeplitz-index theory is also presented in Section 5.","tokens_in":28292,"tokens_out":15029,"duration_ms":130630,"significance":"If the results are correct, they would constitute a substantial extension of the one-cut asymptotics of Deaño–Kuijlaars–Román to multi-interval supports, and the vector-bundle viewpoint would offer a new route to global parametrices that avoids theta functions and special divisors. The paper is well structured and contains several useful auxiliary lemmas, and the Fredholm-index alternative in Section 5 is of independent interest. However, the main existence proof for the global parametrix currently rests on an invalid holomorphy claim in Proposition 3.4, so the central results are not yet established as they stand.","major_comments":[{"comment":"The pairing ⟨U,V⟩_τ(P) = U(P)V(τP)^T is not meromorphic on X, because τ is anti-holomorphic and therefore V(τP) is anti-holomorphic in P; for example, in a local holomorphic coordinate z away from ∂S, V(z)=z gives V(τP)=z̄, which is not holomorphic. Consequently the form ω = ⟨U,U⟩_τ η in the proof of Theorem 3.10 is not a holomorphic 1-form, and the argument dω=0 with Stokes' theorem has no basis. Moreover, on ∂S the claimed identity ⟨U,V⟩_τ = UV^* is false with the transpose; the correct identity requires the conjugate transpose. The likely fix is to define ⟨U,V⟩_τ(P) = U(P)\\overline{V(τP)}^T, which is holomorphic and satisfies the desired boundary identity. As stated, Theorem 3.10, Proposition 3.12, Lemma 4.4, and Proposition 4.5 are not proved, and the existence of the solution to RH Problem 1.10 used in Section 6 is not established. This is a load-bearing defect that must be repaired.","section":"Section 3, Proposition 3.4"},{"comment":"The Fredholm route proves invertibility of the operator T on L^2, yielding a solution of the form N = I + Cϕ with L^2 boundary values. However, it is not shown that this L^2 solution satisfies the pointwise endpoint condition (N4) of RH Problem 1.10, namely N(z) = O(|z-e|^{-1/4}) at each endpoint. Without such an estimate, Section 5 cannot replace the bundle construction in the applications of Section 6, where the endpoint behavior is needed for the matching step.","section":"Section 5, Subsections 5.2.1–5.2.3"},{"comment":"The local Bessel/Airy parametrices of [6, §3.5] and [7, §5.5] are transferred to the multi-cut setting by assertion rather than by proof. In particular, the absorption of the additional unitary jump e^{-inω_j σ3} on the gap adjacent to an internal endpoint into the prefactor E_{e,n} is described only in a few sentences. Since the local matching is essential for the asymptotic formulas (6.17) and (6.28), the author should either provide the full details of this transfer or give a precise reference where it is carried out for multi-cut problems.","section":"Section 6, Subsections 6.1.4 and 6.2.4"}],"minor_comments":[{"comment":"There is a stray period and semicolon in the sentence 'The isolated branch points are removable because the function is locally bounded there whenever U,V are holomorphic there; .'","section":"Section 3, Proposition 3.4"},{"comment":"The phrase 'has a simple zero at ξ_j' is missing the word 'at', and the final clause 'these are all the g zeros' should be rephrased for clarity.","section":"Section 3, Lemma 3.5"},{"comment":"The phrase 'Following [7, (5.24)] and 1.8' should presumably read 'and Assumption 1.8' rather than 'and 1.8'.","section":"Section 6.2.4"},{"comment":"The block-matrix computation for D^{-1}_- and D_+ is dense; an explicit display of D_± in terms of D_± and D_±^* would improve readability and help verify the jump calculations.","section":"Section 6, Proposition 6.5"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses an important problem and the vector-bundle approach is promising, but the flaw in Proposition 3.4 is serious and affects the main existence proof. The defect appears fixable by replacing the transpose with the conjugate transpose in the pairing, and the rest of the apparatus then has a good chance of working. I would encourage the editor to request a revision rather than reject, provided the author can repair the holomorphy argument and also clarify the local parametrix transfer."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real contribution with a genuine flaw in the main proof. Theorem 1.4, the matrix Szegő factorization for multiple intervals with unitary jumps on gaps, is new and, as far as I can tell, correct; the proof via Wiener–Masani and deck transformations is clean. The vector-bundle construction of the global parametrix is also a good idea and gives the right picture. The problem is Proposition 3.4. The pairing U(P)V(τP)^T is not holomorphic: τ is antiholomorphic, so V(τP) is antiholomorphic in general. The proof conflates transpose with conjugate transpose—it uses K_jK_j^T=K_jK_j^*=I and states that on ∂S the pairing equals UV^*, which only holds for U V^*. Theorem 3.10, Proposition 3.12, Lemma 4.4, and Proposition 4.5 all depend on this, so the existence proof for RH Problem 1.10 as written does not go through. Section 5 gives a separate Fredholm proof that looks correct, but Section 6 uses the bundle solution, so the final asymptotic formulas inherit the gap.\n\nThe good news is that the error looks fixable. Replace V(τP)^T with V(τP)^*. Then the pairing becomes holomorphic in the appropriate sense, the jump across Γ_j matches by unitarity, and on ∂S it becomes UU^*, so the Stokes argument in Theorem 3.10 goes through. I checked the divisor bookkeeping; it works. This is not a dead end, but it is a load-bearing correction, not a typo in a remark.\n\nThe reader's other concern is fair: the transfer of the local Bessel and Airy parametrices from the single-cut papers to internal endpoints is asserted rather than proved. The conjugation by e^{-inω_jσ3} plausibly removes the extra gap jump, but the paper should show the computation. The endpoint matching and N4 bounds need a real check, not a citation. That said, the overall strategy is standard enough that I expect it to work.\n\nThe paper is for specialists in Riemann–Hilbert steepest descent and matrix orthogonal polynomials. It deserves a serious referee—don't desk reject it—but I would not accept it without the Proposition 3.4 correction and a detailed check of the local-parametrix transfer. If those come back clean, the paper is likely correct and important.","headline":"Fresh and largely right, but Proposition 3.4's pairing is non-holomorphic and needs a correction before the asymptotics can be trusted.","tokens_in":28792,"tokens_out":14413,"would_cite":false,"duration_ms":135467,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42C05","30E25","15A23","14H60","47B35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper constructs the matrix Szegő factorisation for weights on several intervals and derives strong asymptotics for the associated matrix orthogonal polynomials.","keywords":["matrix-valued orthogonal polynomials","matrix Szegő function","multiple cuts","Riemann–Hilbert problem","steepest descent","hyperelliptic Riemann surface","vector bundles","strong asymptotics"],"falsifier":"Take $g=1$, choose a real analytic positive definite matrix weight $H$ on the two intervals (for instance a constant matrix), compute the actual monic matrix orthogonal polynomials numerically for $n=10,\\ldots,100$, and compare $e^{-ng(z)}P_n(z)$ with the right-hand side of (6.17) uniformly on a compact set away from the cuts; if the difference does not decay like $O(n^{-1})$ as $n$ grows, the claimed local-parametrix transfer and hence the advertised asymptotic formula would be falsified.","tokens_in":27780,"feed_emoji":"🧮","tokens_out":14722,"duration_ms":118125,"temperature":0.7,"pith_summary":"The paper establishes a matrix analogue of the Szegő factorisation for matrix-valued weights supported on several disjoint intervals, and uses it to derive strong asymptotics for the associated matrix orthogonal polynomials. The central theorem says that if a positive definite matrix weight $M$ on a finite union of intervals is integrable enough, then $M=D_+D_+^*=D_-D_-^*$ almost everywhere, with $D$ holomorphic and invertible in the complement of the outermost interval; across each interior gap the boundary values differ by a constant unitary matrix, $D_+=D_-U_j$. That unitary gap jump is the new feature of the multi-cut setting. The paper then shows that two natural families of matrix weights—fixed Jacobi-type weights on the intervals and varying exponential weights whose equilibrium measure fills the intervals—both reduce under steepest descent to the same model Riemann–Hilbert problem, whose unique solution gives explicit asymptotic formulas for the matrix polynomials and their recurrence coefficients.","feed_headline":"Matrix Szegő factorisation lifts to multiple intervals","feed_subtitle":"Unitary gap jumps extend steepest-descent asymptotics to matrix weights on several intervals.","key_machinery":"The load-bearing object is the matrix Szegő function $D$, defined by the factorisation $M=D_\\pm D_\\pm^*$ with a unitary jump $D_+=D_-U_j$ on each gap; its construction rests on pulling the unit-circle matrix spectral factorisation back through a universal covering map, so that deck transformations are exactly the unitary jumps. For the asymptotic half, the load-bearing mechanism is the model Riemann–Hilbert problem whose jump matrices are the constant matrix $J$ on the cuts and $\\mathrm{diag}(K_j,K_j^{-1})$ on the gaps, with $K_j=e^{-in\\omega_j}U_j$. The paper solves this problem by viewing the rows of the parametrix as sections of a rank-$r$ flat unitary vector bundle $E_K$ over the hyperelliptic surface $y^2=A_0(z)B_0(z)$, with transition matrices $K_j$ across the gap curves. A bilinear pairing $\\langle U,V\\rangle_\\tau$ built from the antiholomorphic involution $\\tau$, together with a meromorphic differential $\\eta=(A_0-B_0)/y\\,dz$, yields a vanishing theorem stating that sections with poles only at the divisor $D_-$ and a zero at infinity do not exist; Riemann–Roch then upgrades the evaluation map at infinity to an isomorphism. The concrete parametrix ansatz with $c,s$ built from $\\beta=\\prod((z-b_\\nu)/(z-a_\\nu))^{1/4}$ converts the model problem into a pair of section problems. The uniqueness alternatives are a determinant argument showing $\\det N\\equiv 1$, and a vanishing-lemma plus Toeplitz-index computation in which the index is zero because the determinant curve $\\det c^\\#(\\tau,\\mu)=(\\mu^2+(1-\\mu)^2)^r$ stays in the positive real interval $[2^{-r},1]$.","core_discovery":"On the paper's own terms, the discovery is Theorem 1.4: for a positive definite matrix weight $M$ satisfying the assumed integrability conditions, there is a holomorphic, pointwise invertible matrix function $D$ on the complement of the outer interval, normalised by $D(\\infty)>0$, with $M(x)=D_+(x)D_+(x)^*=D_-(x)D_-(x)^*$ for almost every $x$ in $E$, and with $D_+=D_-U_j$ on each gap, where $U_j$ is unitary. The factor $D$ is obtained by pulling the classical unit-circle matrix factorisation back through a universal covering map of the multi-cut domain; the deck transformations of that covering become the unitary jumps. This factorisation is then fed into a steepest descent analysis: after the $g$-function and lens transformations, both the Jacobi-type model and the varying exponential model reduce to one model problem with constant unitary jumps on the gaps, and the paper constructs its solution by encoding the unitary jump data in a flat unitary vector bundle over the hyperelliptic curve $y^2=A_0(z)B_0(z)$. Rows of the parametrix are sections of that bundle with a prescribed pole divisor $D_-$ and vanishing at one point at infinity, and an evaluation-at-infinity isomorphism provides the right number of sections. The resulting strong asymptotic formula is $e^{-ng(z)}P_n(z)=D(\\infty)c(z)F_1(z;K_1,\\ldots,K_g)D(z)^{-1}+O(n^{-1})$, uniformly on compact subsets of the complement of the outer interval.","pith_inferences":["For the open direction the paper flags—weights of the form $W_n=W(x)^n$—the same machinery would need the $g$-function analysis on the spectral curve of the eigenvalues rather than on the fixed hyperelliptic curve; formula (6.28) gives a concrete target to test there.","The unitary-jump structure of the matrix Szegő function suggests that multi-cut matrix factorisation could be phrased more generally as a flat unitary connection on the complement of $E$, with the $U_j$ as monodromy matrices; the paper does not develop this interpretation.","A numerical check of the leading term in (6.17) for $g=1$, with a constant matrix weight $H$ and $U_1$ determined by the Szegő factor, would isolate the effect of the gap jump from the local endpoint analysis and could reveal how quickly the $O(n^{-1})$ error sets in."],"forward_implications":["For Jacobi-type matrix weights on several intervals, the monic matrix orthogonal polynomials satisfy $e^{-ng(z)}P_n(z)=D(\\infty)c(z)F_1(z;K)D(z)^{-1}+O(n^{-1})$ uniformly on compact subsets of the complement of the outer interval; the same formula holds for varying exponential weights.","The gap data $K_j=e^{-in\\omega_j}U_j$ are explicit: $U_j$ comes from the matrix Szegő function and $\\omega_j$ from the equilibrium measure, so the effect of each gap on the polynomials is fully identified.","The model Riemann–Hilbert problem has exactly one solution for every choice of unitary matrices $K_j$, so the reduction from either polynomial model to the model problem is well-posed.","The vector-bundle construction also applies to scalar orthogonal polynomials: multi-cut scalar Szegő functions with modulus-one jumps across gaps can be read as meromorphic sections of flat unitary line bundles, giving a route that avoids explicit theta functions and special-divisor issues.","The asymptotic formulas imply corresponding asymptotics for the recurrence coefficients of the matrix polynomials, since those coefficients can be extracted from the expansion of the Riemann–Hilbert solution at infinity."],"supporting_citations":[{"why":"Supplies the unit-circle matrix spectral factorisation that Theorem 1.4 pulls back to the multi-cut domain through a universal covering map.","marker":"[24]"},{"why":"Provides the analytic proof of the matrix spectral factorisation theorem used together with [24].","marker":"[11]"},{"why":"Gives the boundary behaviour of universal covering maps, ensuring the pulled-back factorisations have radial boundary values on the set $E$.","marker":"[12]"},{"why":"Sets up the single-interval Jacobi-type matrix-valued orthogonal polynomial steepest descent analysis that Section 6.1 extends to several intervals.","marker":"[6]"},{"why":"Sets up the single-interval varying exponential weight analysis that Section 6.2 extends to multi-cut equilibrium measures.","marker":"[7]"},{"why":"Provides the scalar several-cut steepest descent and the equilibrium measure and $g$-function facts used for the varying exponential model.","marker":"[9]"},{"why":"Supplies the logarithmic potential theory and equilibrium measure background used in the $g$-function transformations.","marker":"[21]"},{"why":"Gives the vector bundle and Riemann–Roch background used to construct sections of $E_K$ and prove the evaluation-at-infinity isomorphism.","marker":"[15]"},{"why":"Supplies the Toeplitz operator index theorem used to prove that the model problem is Fredholm of index zero.","marker":"[3]"},{"why":"Provides the contour-integral vanishing lemma underlying the alternative existence proof in Section 5.","marker":"[25]"}],"fun_headline_variants":["Matrix Szegő factorisation with unitary jumps on multiple cuts","Unitary gap jumps in matrix Szegő factorisation for multiple intervals","Vector-bundle parametrix yields strong asymptotics for MVOPs","Multi-cut Szegő factorisation and hyper-elliptic asymptotics","Matrix Szegő factorisation on multiple cuts: unitary jumps and asymptotics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the local model solutions developed for a single interval can be transferred unchanged to each endpoint of the multi-cut problem, including internal endpoints where the extra unitary jump across the adjacent gap must be absorbed; the paper quotes this transfer rather than proving it in detail.","fun_headline_variants_meta":{"raw":{"variants":["Matrix Szegő factorisation with unitary jumps on multiple cuts","Unitary gap jumps in matrix Szegő factorisation for multiple intervals","Vector-bundle parametrix yields strong asymptotics for MVOPs","Multi-cut Szegő factorisation and hyper-elliptic asymptotics","Matrix Szegő factorisation on multiple cuts: unitary jumps and asymptotics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000711,"raw_usage":{"total_tokens":3239,"prompt_tokens":1024,"completion_tokens":2215,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":640,"completion_tokens_details":{"reasoning_tokens":2121}},"tokens_in":640,"tokens_out":2215,"duration_ms":14140,"temperature":1.0,"reasoning_tokens":2121,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:28:06.593193+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $g=1$, choose a real analytic positive definite matrix weight $H$ on the two intervals (for instance a constant matrix), compute the actual monic matrix orthogonal polynomials numerically for $n=10,\\ldots,100$, and compare $e^{-ng(z)}P_n(z)$ with the right-hand side of (6.17) uniformly on a compact set away from the cuts; if the difference does not decay like $O(n^{-1})$ as $n$ grows, the claimed local-parametrix transfer and hence the advertised asymptotic formula would be falsified.","supporting_citations":[{"cited_title":"Wiener and P","cited_arxiv_id":null,"evidence_quote":"Supplies the unit-circle matrix spectral factorisation that Theorem 1.4 pulls back to the multi-cut domain through a universal covering map."},{"cited_title":"Ephremidze, G","cited_arxiv_id":null,"evidence_quote":"Provides the analytic proof of the matrix spectral factorisation theorem used together with [24]."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the boundary behaviour of universal covering maps, ensuring the pulled-back factorisations have radial boundary values on the set $E$."},{"cited_title":"Deaño, A","cited_arxiv_id":null,"evidence_quote":"Sets up the single-interval Jacobi-type matrix-valued orthogonal polynomial steepest descent analysis that Section 6.1 extends to several intervals."},{"cited_title":"Asymptotics of matrix orthogonal polynomials on the real line","cited_arxiv_id":"2508.04908","evidence_quote":"Sets up the single-interval varying exponential weight analysis that Section 6.2 extends to multi-cut equilibrium measures."},{"cited_title":"Deift, T","cited_arxiv_id":null,"evidence_quote":"Provides the scalar several-cut steepest descent and the equilibrium measure and $g$-function facts used for the varying exponential model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the logarithmic potential theory and equilibrium measure background used in the $g$-function transformations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the vector bundle and Riemann–Roch background used to construct sections of $E_K$ and prove the evaluation-at-infinity isomorphism."},{"cited_title":"Böttcher and B","cited_arxiv_id":null,"evidence_quote":"Supplies the Toeplitz operator index theorem used to prove that the model problem is Fredholm of index zero."},{"cited_title":"Zhou,The Riemann–Hilbert problem and inverse scattering, SIAM J","cited_arxiv_id":null,"evidence_quote":"Provides the contour-integral vanishing lemma underlying the alternative existence proof in Section 5."}],"review_version":1}