{"id":"3076005e-27bc-4420-b1b7-b7d956dc540a","arxiv_id":"2509.08527","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The authors prove that, under a multiplicative genericity assumption, the OK condition on conjugacy classes suffices for solvability of the Deligne-Simpson problem, via a new relative spectral correspondence for parabolic Higgs bundles.","lead":"The paper builds a geometric bridge between parabolic Higgs bundles and sheaves on a family of surfaces, and uses it to prove that a numerical condition called OK guarantees that certain products of matrices exist. A general reader might value it as an example of geometry solving a concrete linear algebra question about matrices.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition A.3 assumes the spectral curve avoids E_j∩E_{j+1}, but no argument shows this can be ensured in the Deligne–Simpson construction, on which the residue-diagram computation depends.","rationale":"The reader's weakest assumption identifies exactly the gap that I find most load-bearing: Proposition A.3 is the local computation that connects the spectral-curve data to the residue diagrams of the Higgs field, and the Deligne–Simpson theorem depends on it. The proof as written omits the case where the integral spectral curve passes through E_j∩E_{j+1}, and no genericity argument is supplied to show that this case can be avoided in the applications. I considered other concerns: the abstract overstates the genericity assumptions of the Deligne–Simpson result, and there is an apparent inconsistency between the weighted residue condition in Section 1.2 and the unweighted condition in Definition 2.5/N(m). However, the abstract issue is a matter of precision rather than correctness, and the residue-condition inconsistency may be typographical since Section 1.2 states the weighted condition explicitly. The Proposition A.3 gap is a substantive mathematical omission in the proof of the central claim. It is not obviously fatal—the missing charts may well yield the same Jordan form, or a genericity argument may exclude the dangerous curves—but it must be resolved before the theorem is fully rigorous. The correct verdict remains CONDITIONAL, as the reader concluded, so I recommend no change to the reader's verdict.","tokens_in":53955,"tokens_out":25200,"duration_ms":571647,"concrete_test":"Take a concrete local example satisfying the exact multiplicity conditions of Lemma A.1 in which the strict transform of the spectral curve passes through E_j∩E_{j+1}, for instance ℓ=2, m=(1,1), and a curve whose tangent cone at c_2 aligns with the exceptional direction. Recompute the Jordan normal form of multiplication by y on the fiber using all charts covering the double point. If the result is still the conjugate partition of (m_1,...,m_ℓ), extend Proposition A.3 to cover the missing charts; if not, test whether a general integral member of B(m)_ξ avoids these double points and whether the construction in Lemma 5.28 can be perturbed to land in that locus.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Proposition A.3 explicitly assumes that the integral spectral curve Σℓ does not pass through the intersection points E_j∩E_{j+1} of successive exceptional divisors, writing: 'For simplicity, we assume that Σℓ does not pass through the intersection E_j∩E_{j+1} (otherwise, we will need to consider more charts)'. This assumption is used to guarantee that the local equation takes the normal form F_j = v_j^{e_j} + u_{j−1}f + c with c≠0, which is essential for the linear-independence argument that identifies the Jordan blocks of the induced Higgs field. If c=0, that argument no longer goes through, and the claimed Jordan normal form—the conjugate partition of (m_1,...,m_ℓ)—is not established. Proposition 5.26 and Step 2 of Theorem 5.29 rely directly on Proposition A.3 to compute the residue diagrams of the constructed Higgs bundle and hence the conjugacy classes of the monodromy. If a curve passes through one of these double points, the residue diagram could differ, breaking the proof that the DSP is solvable for the prescribed conjugacy classes. The paper provides no argument that a general integral member of B(m)_ξ can be chosen to avoid all such points, nor does it handle the dimension-zero case where the unique member might be forced through them. This is a genuine gap in the proof of the main theorem as stated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper generalizes the Diaconescu–Donagi–Pantev spectral correspondence to the relative setting over the base N(m) of eigenvalue data for parabolic Higgs bundles. It constructs a family of holomorphic symplectic surfaces by successive blow-ups, defines a relative moduli space of pure dimension one sheaves with support in a fixed curve class, and proves a closed embedding Q from the relative moduli space of xi-parabolic Higgs bundles into this sheaf moduli space, with an isomorphism over the integral locus. The paper also identifies the parabolic Hitchin base B(m)_xi with an affine linear system and with the zero locus of evaluation maps in the polynomial Hitchin base, and connects the OK condition to flatness and non-emptiness of these bases. As applications, it proves non-emptiness statements for moduli of stable xi-parabolic Higgs bundles and, in Theorem 5.29, claims that the OK condition is sufficient for the multiplicative Deligne–Simpson problem under a multiplicative genericity assumption, thereby addressing a conjecture of Balasubramanian–Distler–Donagi.","tokens_in":54290,"tokens_out":11172,"duration_ms":107951,"significance":"If the local analysis is completed, this is a substantial contribution. The relative spectral correspondence and the single-blow-up transformation are useful new tools, and the linear-system description of parabolic Hitchin bases is a genuinely new perspective. The paper also gives a concrete geometric route to the Deligne–Simpson problem and proves a substantial form of the BDD conjecture under multiplicative genericity. The computational control over residues via curve classes and the comparison with Simpson's criterion in Appendix B are valuable. However, the main Deligne–Simpson theorem currently rests on an unproved genericity assertion about spectral curves avoiding exceptional-divisor double points, and the abstract overstates the scope of the conjecture proved. These issues are load-bearing and require repair.","major_comments":[{"comment":"The proof assumes that the integral spectral curve Sigma_l does not pass through E_j cap E_{j+1}, writing that otherwise more charts are needed. This hypothesis is essential for the normal form F_j = v_j^{e_j} + u_{j-1}f + c with c != 0 in Eq. (40); when c = 0, the linear-independence argument for the set T in Eq. (41) collapses, and the claimed Jordan normal form (the conjugate partition of (m_1,...,m_l)) is not established. No argument is given that a general integral member of B(m)_xi avoids these double points, and the zero-dimensional-fiber case is not addressed. Since Proposition 5.26(2) and Step 2 of Theorem 5.29 rely on this residue-diagram computation, the main Deligne–Simpson theorem is currently conditional on an extra assumption. The authors should either prove the avoidance statement, including all dimension-zero cases, or carry out the additional charts and repeat the Jordan-block computation for c = 0.","section":"Appendix A, Prop. A.3"},{"comment":"To ensure that Sigma is the strict transform of C_s, the proof needs exact multiplicities at the blow-up centers, as in Remark A.2. In the case dim B(m)_xi = 0, the proof says that the extra equations are avoided for a general xi because they define a closed subset of N(m)^add. This is only valid if that subset is proper, and no dimension estimate is provided; a priori the bad locus could be all of N(m)^add. If exact multiplicity fails, Sigma need not be integral or may not belong to B(m)_xi, so Proposition 5.26 cannot be applied. This is a second load-bearing gap in the proof of Theorem 5.29.","section":"Lemma 5.28"},{"comment":"The abstract states that the paper proves that the OK condition is sufficient for solving the Deligne–Simpson problem, with no genericity qualification. However, Theorem 5.29, which is the theorem supporting this claim, assumes that the collection of eigenvalues is multiplicatively generic. The unqualified conjecture of [BDD22] is not proved; Remark 5.33 only indicates how a version without generic eigenvalues might be obtained under stronger inequalities. The abstract and introduction should state the multiplicative genericity assumption explicitly, or the conjecture should be proved in the stated generality.","section":"Abstract and Theorem 1.11"}],"minor_comments":[{"comment":"The word 'quadraple' should be 'quadruple'.","section":"Definition 2.1"},{"comment":"The phrase 'the expected dimension is always lower than the actual dimension' should read 'the actual dimension is always at least the expected dimension'; as written it is slightly confusing.","section":"Prop. 3.14"},{"comment":"The proof asserts without computation that the strengthened inequalities guarantee H^1(C, L(m)_mu tensor O(-p_i)) = 0 for mu = 2,...,r. A short Riemann–Roch verification for the g = 0 and g = 1 cases would make the argument easier to check.","section":"Prop. 5.18"}],"recommendation":"major_revision","confidential_remarks":"The paper relies on the authors' prior work [LL24] for the smooth family H(m) and for the moduli construction; this dependence is acknowledged but should be stated more prominently as the starting point. The main concern is the unproved genericity hypothesis in Appendix A, which is not a mere presentation issue: it directly affects the residue-diagram computation in Theorem 5.29. I would ask the authors to either supply a proof or explicitly reformulate the main theorems with the needed assumption added."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my read of Lee–Lee (arXiv:2509.08527).\n\nThe headline is that this is a substantial piece of algebraic geometry. The relative spectral correspondence over the full base N(m)—not just generic eigenvalues, and not just the strongly parabolic end—is a real extension of DDP18 and Su–Wang–Wen. The construction via iterated blow-ups of the family and the single-blow-up sheaf transformation Psi_p look sound and are of independent interest. The linear-system description of the parabolic Hitchin base and the non-emptiness results under the OK condition are well motivated and carefully proved. The DSP application is a genuine attempt at a geometric proof of the BDD conjecture.\n\nThat said, there are two soft spots worth flagging. First, the abstract says the OK condition is sufficient for the DSP without mentioning the multiplicatively generic eigenvalue assumption. Theorem 5.29 does include that assumption, so the abstract overclaims. Second, Appendix A, Proposition A.3, is the load-bearing local computation for the residue diagrams. The proof assumes the spectral curve does not pass through the intersection points of successive exceptional divisors and gives no argument that a general integral member of B(m)_xi avoids them. The linear systems in question impose multiplicities at the blow-up centers, so avoidance is not automatic; in the dimension-zero case the unique curve could be forced through a double point. Since Proposition 5.26 and Theorem 5.29 depend directly on the Jordan form computed there, the main theorem is conditional as written. I don't think this is fatal—presumably one can handle the extra charts or show that general curves avoid the double points—but the paper doesn't supply that argument.\n\nThe self-citation to [LL24] is fine: H(m) is imported from their prior work, but the spectral correspondence and the DSP application rest on new constructions. The citation pattern around DDP18, SWW22, BDD22 is appropriate.\n\nWho should read this: anyone working on parabolic Higgs bundles, Hitchin systems, or the Deligne–Simpson problem. The geometric perspective is fresh and the technical core is mostly solid. I would send it to a serious referee, asking them to check Proposition A.3 and the exact scope of the DSP theorem. With those fixed, this would be a strong paper.","headline":"Solid relative spectral correspondence; abstract overstates DSP theorem, and Proposition A.3 has an unhandled double-point case that makes the main proof conditional.","tokens_in":54826,"tokens_out":6644,"would_cite":true,"duration_ms":60106,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H60","14D20","14H70","14F05"],"pacs":[],"model":"deepseek-v4-flash","headline":"A relative spectral correspondence identifies parabolic Higgs bundles with sheaves on a family of blown-up surfaces, and proves that the OK condition suffices for the multiplicative Deligne-Simpson problem when the eigenvalue data is…","keywords":["parabolic Higgs bundles","spectral correspondence","Deligne-Simpson problem","OK condition","Hitchin map","holomorphic symplectic surfaces","moduli of sheaves","level functions"],"falsifier":"Take the smallest case satisfying the theorem's inequalities, for example $n=3$, $r=2$ with partitions $P_i=(1,1)$ or $(2)$, and choose a tuple $\\vec{\\xi}$ for which the integral spectral curve passes through an intersection $E_j \\cap E_{j+1}$ in the blown-up surface; then compute the Jordan normal form of the residue of the pushed-forward Higgs field in the local charts omitted in Proposition A.3. If the Jordan form differs from the conjugate partition, Theorem 5.29 fails for that curve; if it is the same, the simplifying assumption is removable and the theorem holds without it.","tokens_in":53740,"feed_emoji":"📐","tokens_out":7353,"duration_ms":59954,"temperature":0.7,"pith_summary":"This paper establishes a relative spectral correspondence: as the eigenvalues of a parabolic Higgs bundle vary over a base, the moduli space of such bundles is realized as the moduli space of pure dimension one sheaves on a family of holomorphic symplectic surfaces obtained by iterated blow-ups. This turns the image of the Hitchin map into a family of linear systems, giving a concrete linear description of the parabolic Hitchin base. Using this geometry, the paper proves that the OK condition, a vanishing of $H^1$ for certain line bundles, guarantees non-emptiness of these Hitchin bases and of the moduli spaces in broad ranges. As the main application, it proves the conjecture that the OK condition is sufficient for the multiplicative Deligne-Simpson problem under multiplicative genericity.","feed_headline":"OK condition suffices for Deligne-Simpson problem","feed_subtitle":"A family of blown-up surfaces turns parabolic Higgs bundles into sheaves, proving the long-sought sufficiency.","key_machinery":"The machinery is a family of holomorphic symplectic surfaces built by successively blowing up the ruled surface $M = \\mathbb{P}(K_C(D) \\oplus \\mathcal{O}_C)$ along the tautological sections $\\xi_{i,j}$, then removing the strict transforms of the fibers over the marked points and the infinity section. A relative curve class $\\Sigma(\\vec{m}) = r f^* C_0 - \\sum_{i,j} m_{i,j} \\Xi_{i,j}$ is chosen, and a single-blow-up transformation $\\Psi_p$ converts a parabolic sheaf on one surface into a parabolic sheaf of shorter filtration on the blow-up, encoding the parabolic structure in the exceptional divisors. Iterating $\\Psi_p$ turns a $\\vec{\\xi}$-parabolic Higgs bundle into a pure dimension one sheaf; the parabolic Hitchin base $B(\\vec{m})_{\\vec{\\xi}}$ is identified with the linear system $|\\Sigma(\\vec{m})_{\\vec{\\xi}}|$, described explicitly by vanishing of partial derivatives indexed by the level domains $G(P)$. The OK condition makes these linear systems have constant dimension, and the local Jordan normal form of the Higgs residue is read off from the intersection numbers of the exceptional divisors with the spectral curve.","core_discovery":"The central claim is Theorem 5.29: for $n \\geq 3$ conjugacy classes $C_1,\\ldots,C_n$ in $\\mathrm{GL}_r(\\mathbb{C})$ with multiplicatively generic eigenvalues, if the product of determinants is $1$ and the level-function sums satisfy $\\sum_{i=1}^n \\gamma_{P_i}(\\mu) < (n-2)\\mu + 2$ for $\\mu = 2,\\ldots,r$, then the multiplicative Deligne-Simpson problem is solvable. The same machinery yields non-emptiness of the moduli spaces of stable $\\vec{\\xi}$-parabolic Higgs bundles and a higher genus analogue of the Deligne-Simpson problem. The geometric heart is the relative spectral correspondence, which embeds the moduli of $\\vec{\\xi}$-parabolic Higgs bundles into the moduli of pure dimension one sheaves on a family of surfaces, compatibly with the Hitchin map and the Fitting support map; the embedding is an isomorphism over the locus of integral spectral curves.","pith_inferences":["The residue-diagram computation in Appendix A is the only place where the proof depends on the spectral curve avoiding the intersection points of consecutive exceptional divisors; if that assumption is removed or shown automatic for general integral curves, the Deligne-Simpson theorem would extend to all eigenvalue configurations satisfying the OK condition.","The identification of the Hitchin base with a linear system suggests the parabolic Hitchin map is flat when the OK condition holds, so one could define family versions of parabolic Hodge integrals or $P=W$-type invariants over the eigenvalue base.","The defect formulated with conjugate partitions differs from the classical defect; iterating the authors' construction along Kostov-style modifications may yield sufficiency beyond the range covered here.","The single-blow-up transformation $\\Psi_p$ may give a purely algebraic handle on the Nahm transform for parabolic sheaves, since it depends only on local filtration data."],"forward_implications":["If the OK condition holds, then $B(\\vec{m})_{\\vec{\\xi}}$ is non-empty for every eigenvalue tuple $\\vec{\\xi}$; in particular for $n \\geq 3$, $g=0$ with the stated inequalities, and for all $n \\geq 1$, $g \\geq 2$.","The strengthened OK-type inequalities imply non-emptiness of the moduli space of stable $\\vec{\\xi}$-parabolic Higgs bundles for every $\\vec{\\xi}$.","The multiplicative Deligne-Simpson problem is solvable for multiplicatively generic conjugacy classes satisfying the OK condition, confirming the conjecture that the OK condition is sufficient.","The same approach yields existence of irreducible solutions to the higher genus analogue of the Deligne-Simpson problem whenever the OK condition, or its appropriate variant, holds.","The relative spectral correspondence gives a linear description of the image of the parabolic Hitchin map, an image that in general is not defined by linear equations."],"supporting_citations":[{"why":"Supplies the absolute spectral correspondence for parabolic Higgs bundles at generic eigenvalues, which this paper relativizes over the whole family.","marker":"[DDP18]"},{"why":"Introduced the OK condition and conjectured its sufficiency for the Deligne-Simpson problem, which Theorem 5.29 proves under multiplicative genericity.","marker":"[BDD22]"},{"why":"Provides the tame non-abelian Hodge correspondence used to turn irreducible local systems with prescribed monodromy into stable parabolic Higgs bundles with prescribed residue diagrams.","marker":"[Sim90]"},{"why":"Contains the original solution of the Deligne-Simpson problem in the case of a regular conjugacy class, with which the numerical equivalence of the OK condition is checked in Appendix B.","marker":"[Sim91]"},{"why":"Classical spectral correspondence, the first step of the construction producing a pure dimension one sheaf on the ruled surface from a Higgs bundle.","marker":"[BNR89]"},{"why":"The authors' earlier construction of the relative moduli space $H(\\vec{m}) \\to N(\\vec{m})$ of $\\vec{\\xi}$-parabolic Higgs bundles, whose smoothness is used here.","marker":"[LL24]"},{"why":"Describes the image of the strongly parabolic Hitchin map, which is matched with $B(\\vec{m})_0$ in Corollary 3.13 as a consistency check.","marker":"[SWW22a]"},{"why":"Provides the known criterion for the multiplicative Deligne-Simpson problem with multiplicatively generic eigenvalues, with which the new criterion is compared in Remark 5.34.","marker":"[Kos04]"}],"fun_headline_variants":["OK condition suffices for Deligne-Simpson","Relative spectral correspondence proves Deligne-Simpson","Parabolic Higgs bundles solve Deligne-Simpson","Spectral geometry shows OK suffices for Deligne-Simpson"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the Higgs residue has the prescribed Jordan normal form assumes the integral spectral curve does not pass through the intersection points of consecutive exceptional divisors, and no argument shows that a general integral member of $B(\\vec{m})_{\\vec{\\xi}}$ avoids those points.","fun_headline_variants_meta":{"raw":{"variants":["OK condition suffices for Deligne-Simpson","Relative spectral correspondence proves Deligne-Simpson","Parabolic Higgs bundles solve Deligne-Simpson","Spectral geometry shows OK suffices for Deligne-Simpson"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000689,"raw_usage":{"total_tokens":3129,"prompt_tokens":957,"completion_tokens":2172,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":573,"completion_tokens_details":{"reasoning_tokens":2107}},"tokens_in":573,"tokens_out":2172,"duration_ms":16686,"temperature":1.0,"reasoning_tokens":2107,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:00:53.115899+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the smallest case satisfying the theorem's inequalities, for example $n=3$, $r=2$ with partitions $P_i=(1,1)$ or $(2)$, and choose a tuple $\\vec{\\xi}$ for which the integral spectral curve passes through an intersection $E_j \\cap E_{j+1}$ in the blown-up surface; then compute the Jordan normal form of the residue of the pushed-forward Higgs field in the local charts omitted in Proposition A.3. If the Jordan form differs from the conjugate partition, Theorem 5.29 fails for that curve; if it is the same, the simplifying assumption is removable and the theorem holds without it.","supporting_citations":[],"review_version":2}