{"id":"09d7d9d4-c359-4d8f-ba0b-11b54cdaad36","arxiv_id":"2608.02875","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors construct explicit non-integral Q-Hodge variations with Zariski dense monodromy, prove new S-integral point results on character varieties, and show the Hodge locus can be non-algebraic.","lead":"Non-integral variations of Q-Hodge structures with large monodromy are built explicitly from hyperbolic surface geometry. The paper shows that for such variations the Hodge locus can fail to be algebraic, so integrality is essential.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 2.5 is false as stated, and the proof of Theorem 1.2 relies on it to conclude that the constructed QVHS is non-integral.","rationale":"The reader's verdict of CONDITIONAL is appropriate: the main construction is novel and likely repairable, but the written proof has a concrete false lemma. The reader's weakest_assumption focused on Theorem 6.2 and the rationality of Kabaya's parametrization, whereas the more load-bearing gap is Proposition 2.5, which is not merely unproven but false. The reader did mention the false proposition in the rationale, so the agreement is partial. The false proposition is used directly in the proof of Theorem 1.2 to pass from trace ring containing Z[S^{-1}] to non-integrality of the QVHS; without it, the paper does not establish the central 'non-integral' adjective. The counterexample is elementary and unambiguous. However, because the fix is plausibly simple — a direct trace computation in Lemma 2.2 may prove the QVHS non-integral — the verdict should remain CONDITIONAL rather than REJECT. The authors should either correct Proposition 2.5 with suitable hypotheses or replace the citation with the direct argument. No further adjustment to the reader's verdict is needed.","tokens_in":20390,"tokens_out":13025,"duration_ms":126002,"concrete_test":"Verify the refutation of Proposition 2.5 by computing the decomposition of the C3 permutation representation: the 3x3 cyclic permutation matrix P has eigenvalues 1, ω, ω^2 over C, and the ω-eigenspace defines a complex direct factor whose trace is ω, not an integer. This settles that Proposition 2.5 is false exactly as stated. If the referee wishes to check whether the theorem survives, repeat the proof of Theorem 1.2 replacing the Proposition 2.5 citation with a direct trace computation: write N = ∏_{p∈S} p, take the element γ1 with Tr ρ(γ1) = -N - 1/N, and compute the trace of the rank-four Q-local system from Lemma 2.2 as 2·Tr ρ(γ1); for N≥3 this is not an integer, so the ambient QVHS is non-integral. If this direct computation is not supplied, the non-integrality assertion remains unproven.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 2.5 asserts that every complex direct factor of an integral representation Γ→GL_n(Z) has traces in Z. This is false: for Γ=C3, the rank-three cyclic permutation representation on Z^3 is integral, but over C it decomposes as the trivial character plus a two-dimensional sum of the two nontrivial characters, each with traces in Q(ω), ω^3=1; for the nontrivial character the trace ω is not an integer. In the proof of Theorem 1.2, the authors write 'This follows directly from the above and Proposition 2.5.' The intended inference is: since the local system ρ has trace ring containing Z[S^{-1}] (Theorem 1.6, property prPS), any integral QVHS admitting ρ as a complex direct factor would force traces of ρ to lie in Z, contradicting non-integrality. Because Proposition 2.5 is false, that inference is invalid. The C3 example shows integrality of the ambient representation does not imply integrality of a complex direct summand. The central claim of Theorem 1.2 — that the constructed QVHS is non-integral — therefore rests on a false lemma. A repair may be possible using the explicit form of the QVHS in Lemma 2.2: if the trace field is Q, the rank-four Q-local system has trace 2·Tr(ρ(γ)) for an element with trace t, and choosing t = -N - 1/N with N = ∏_{p∈S} p gives non-integral rational traces for suitable N, but this direct argument is absent from the paper. As written, the proof is incomplete.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs explicit complex local systems on curves (arbitrary genus and the four-punctured sphere) whose trace ring is strictly larger than Z, and argues that they are complex direct factors of non-integral QVHS with Zariski dense monodromy in SL2. The constructions use Maskit's and Kabaya's parametrizations of the Teichmüller component of relative character varieties, with rational parameters chosen so that the trace field is Q while the trace ring contains Z[S^{-1}]. The paper also proves finiteness/infinitude statements for OK,S-Teichmüller points, gives a new proof of Beauville's classification of signature (0,4) modular embeddings, and discusses failures of the Cattani--Deligne--Kaplan theorem and the André--Oort conjecture for non-integral QVHS.","tokens_in":20664,"tokens_out":9697,"duration_ms":92126,"significance":"If the main construction is completed, the paper would provide explicit non-integral QVHS with large monodromy, showing that integrality is essential for several structural results in Hodge theory. The use of explicit Fenchel--Nielsen-type parametrizations and exact trace-ring computations is a genuine strength, as is the concrete matching with Beauville's list. However, the key non-integrality conclusion of Theorem 1.2 currently rests on a false statement, Proposition 2.5, so the main theorem is not proven as written. The flaw appears localized and repairable by a direct trace computation, but the revision must supply that argument.","major_comments":[{"comment":"Proposition 2.5 is false as stated. The claim that every complex direct factor of an integral representation Γ → GL_n(Z) has traces in Z fails already for Γ = C3 acting on Z^3 by cyclic permutation of coordinates. Over C this representation decomposes as the trivial character plus the two nontrivial characters, whose traces lie in Q(ω), ω^3 = 1, and are not integers. Thus a complex direct summand of an integral representation need not be integral. Any argument that relies on this proposition must be replaced.","section":"§2.1, Proposition 2.5"},{"comment":"The proof of Theorem 1.2 says 'This follows directly from the above and Proposition 2.5.' Because Proposition 2.5 is false, the inference that a QVHS admitting ρ as a complex direct factor must be non-integral whenever ρ is non-integral is invalid. The intended repair is available: in the construction of Theorem 1.6 the element γ1 has eigenvalue e1 = -N with N = ∏_{p∈S} p, so Tr ρ(γ1) = -N - N^{-1}. The QVHS produced by Lemma 2.2 has trace 2(-N - N^{-1}) on the corresponding element; for instance N = 3 gives -20/3, which is not an integer. This direct argument would establish Theorem 1.2, but it must be written out in the revision.","section":"§6.3, proof of Theorem 1.2"}],"minor_comments":[{"comment":"The parameters e_i for i ≥ 2 and t_i are only specified to lie in Q. They should be chosen generically in Q ∩ (-∞,-1) and Q ∩ (0,∞) so that the excluded conditions in the definition of E(Σ,C) are avoided; for example, taking all e_i = -N and all t_i = 1 is an explicit valid choice.","section":"§6.3, proof of Theorem 1.6"},{"comment":"The formula 'e1 = -1/(∏_{p∈S} p^{-1})' is confusingly written; it should read e1 = -∏_{p∈S} p.","section":"§6.3, displayed formula for e1"},{"comment":"There are minor typographical issues: 'not S1-integral' appears to mean 'not S'-integral for a smaller set S', and 'property1 (PS)' in the statement of Theorem 1.8 should be 'property (PS)'.","section":"§1.2 and Theorem 1.8"},{"comment":"The proof of Proposition 7.3 is largely delegated to [3, Prop. 7.1.2] with only a short indication. Since [3] is published, this is acceptable, but a sentence explaining why the non-integral QVHS constructed here satisfies the same hypothesis would improve readability.","section":"§7, Proposition 7.3"}],"recommendation":"major_revision","confidential_remarks":"The false Proposition 2.5 is the only serious mathematical error I found. The repair via explicit traces is short and well within the scope of the manuscript, so I recommend requesting a major revision rather than rejection. Please also verify that the proof of Claim 1 in Theorem 4.5 does not depend in an essential way on the unpublished preprint [17]; if it does, the dependence should be made explicit and ideally eliminated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main construction is genuinely new and valuable: explicit non-integral QVHS built from Maskit and Kabaya parametrizations, plus Diophantine results on O_{K,S}-Teichmüller points, a new proof of Beauville's theorem, and a catalog of Hodge-locus pathologies for non-integral QVHS. The authors use external parametrizations and standard trace-field facts honestly, and the overall strategy is credible. This is a paper worth engaging with seriously.\n\nThe soft spot is real and load-bearing. Proposition 2.5 is false as stated: the C3 permutation representation on Z^3 is integral, but over C it splits as the trivial character plus two nontrivial characters, whose traces lie in Q(omega), not Z. The proof of Theorem 1.2 explicitly says 'This follows directly from the above and Proposition 2.5.' That inference is invalid: integrality of the ambient representation does not force integrality of a complex direct summand. Since the constructed QVHS is supposed to be non-integral, the main theorem is not proved as written.\n\nThat said, the flaw looks repairable. The ingredients are already in the paper: Lemma 2.2 gives a rank-four QVHS whose trace on an element with trace t is 2t, and the constructed local system has trace ring containing Z[S^{-1}] while its trace field is Q. Choosing a suitable element with trace -N - 1/N for N the product of the primes in S would give a non-integral rational trace directly. The repair is not long, but it needs to be written down.\n\nOther issues are minor by comparison. Section 7 is mostly a sketch and leans on the authors' own prior work, especially [3] and [17]; the claims there are plausible but not fully proved in this text. Theorem 1.6 also does not explicitly justify that the constructed curves are non-isomorphic, though that should follow from discreteness of the parameter choices. The K-rationality statement in Theorem 6.2 is compressed but likely correct.\n\nThis paper deserves a serious referee. The natural question—whether non-integral QVHS with large monodromy exist—is resolved by a construction that is likely correct, and the surrounding Diophantine and Hodge-theoretic observations are useful. My recommendation: send it to peer review with a request to fix Proposition 2.5 or replace that step with the direct trace argument, and to expand the Section 7 sketches. With those repairs, the paper would be a solid contribution.","headline":"Interesting and likely right in spirit, but the non-integrality proof rests on a false proposition, so Theorem 1.2 needs repair before the paper can be accepted.","tokens_in":21264,"tokens_out":1567,"would_cite":false,"duration_ms":16545,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D07","14H15","30F60","32G15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper constructs explicit rank-two local systems on infinitely many curves of every genus that, while not integral, are complex direct factors of variations of Q-Hodge structure with Zariski dense monodromy.","keywords":["Character varieties","S-integral points","variations of Hodge structures","non-abelian Hodge theory","Teichmüller component","Fenchel–Nielsen coordinates","Hodge locus","trace field"],"falsifier":"For a small concrete case, say $g=2$ and $S=\\{2\\}$, write down the explicit matrices produced by the parametrization with $e_1=-1/2$ and all other parameters rational, and compute the trace ring and the Zariski closure of the monodromy; if any trace lies outside $\\mathbb{Z}[1/2]$, or the representation is not discrete and faithful, or the monodromy is not Zariski dense in $\\mathrm{SL}_2$, then Theorem 1.6 would be wrong.","tokens_in":20146,"feed_emoji":"📐","tokens_out":12424,"duration_ms":103335,"temperature":0.7,"pith_summary":"By \"non-integral\" the paper means a variation of Hodge structures with rational coefficients whose monodromy trace ring is not contained in $\\mathbb{Z}$. The main construction shows that for every genus $g \\geq 2$ and every nonempty finite set $S$ of primes there exist infinitely many non-isomorphic genus $g$ curves carrying a rank-two local system whose trace field is $\\mathbb{Q}$ but whose trace ring contains $\\mathbb{Z}[S^{-1}]$, and which is a complex direct factor of an irreducible variation of $\\mathbb{Q}$-Hodge structures. The construction is explicit, using Fenchel-Nielsen-type coordinates on the Teichm\\\"uller component of the character variety, and it also gives a new proof of Beauville's classical result for the four-punctured sphere. The paper closes by exhibiting pathological Hodge loci for non-integral variations, where the Cattani-Deligne-Kaplan theorem and the Andr\\'e-Oort analogue fail.","feed_headline":"Every genus has infinitely many non-integral Hodge variations","feed_subtitle":"Using Fenchel–Nielsen coordinates, their trace field is rational while the trace ring overflows Z.","key_machinery":"The load-bearing object is Kabaya's parametrization of the $\\mathrm{PSL}_2(\\mathbb{C})$-character variety relative to a pants decomposition: it attaches to each point a tuple of eigenvalue parameters $e_i$ and twist parameters $t_i$ and builds explicit matrices, with the property that $K$-rational points of the parameter space map to $K$-points of the character variety for every subfield $K \\subset \\mathbb{C}$. Maskit's parametrization does the same for the four-punctured sphere, giving rational formulas for the three trace coordinates on the Teichm\\\"uller component. These coordinates let the authors force the trace field to be $\\mathbb{Q}$ while inserting specified denominators into the trace ring. Lemma 2.2 and Corollary 2.4 then convert trace-field $\\mathbb{Q}$ into the desired direct-factor statement about variations of $\\mathbb{Q}$-Hodge structures, and Proposition 2.5 converts trace integrality into $\\mathbb{Z}$-integrality of the local system.","core_discovery":"The central claim is Theorem 1.2: for any $g \\geq 2$ there are infinitely many non-isomorphic genus $g$ curves $X$ supporting a rank two local system $\\rho : \\pi_1(X) \\to \\mathrm{SL}_2(\\mathbb{C})$ with trivial determinant and Zariski dense monodromy in $\\mathrm{SL}_2$, such that $\\rho$ is a complex direct factor of a non-integral variation of $\\mathbb{Q}$-Hodge structures, and such that the adjoint local system $\\mathrm{ad}(\\rho)$ itself underlies a non-integral $\\mathbb{Q}$VHS. The refinement Theorem 1.6 controls the trace ring: for any nonempty finite set $S$ of primes one can arrange $e_1 = -\\prod_{p \\in S} p^{-1}$ so that $\\mathbb{Z}[S^{-1}]$ lies inside the trace ring while the trace field is still $\\mathbb{Q}$. Lemma 2.2 and Corollary 2.4 are the bridge: a Zariski dense, non-unitary local system with trace field $\\mathbb{Q}$ that underlies a complex variation of Hodge structure is automatically a complex direct factor of a $\\mathbb{Q}$VHS. The non-integrality then follows from Proposition 2.5, which says that an integral local system has traces in $\\mathbb{Z}$; here the trace ring is strictly larger than $\\mathbb{Z}$.","pith_inferences":["One extension the paper leaves implicit is whether the rationality property of the parametrization survives for other semisimple groups; if it does, analogous constructions would likely produce non-integral $\\mathbb{Q}$VHS with large monodromy in higher rank, which is explicitly left open in the paper.","The 'Murphy's law' pattern suggests that integrality is the hypothesis doing the work in the Cattani-Deligne-Kaplan theorem and in the Andr\\'e-Oort conjecture; one could test whether other finiteness theorems for Hodge loci fail as soon as $\\mathbb{Z}$-integrality is dropped.","The explicit trace formulas in the four-punctured sphere case turn the search for $S$-integral Teichm\\\"uller points into a Diophantine problem on a Markoff-type cubic; a computational scan for small $S$ could reveal whether the exact trace-ring orbits described in Proposition 5.6 are the only ones."],"forward_implications":["The non-abelian Hodge conjecture over $\\mathbb{Q}$ fails even for local systems with Zariski dense monodromy in $\\mathrm{SL}_2$.","For every genus $g \\geq 2$ and every nonempty finite set of primes $S$, there are infinitely many non-isomorphic curves whose uniformizing local system has trace ring containing $\\mathbb{Z}[S^{-1}]$, so prescribed $S$-integrality is achievable.","If the Higgs standard conjecture is true, the curves appearing in Theorem 1.2 cannot be defined over $\\mathbb{Q}$ (Conjecture 1.4).","In signature $(0,4)$, allowing no primes yields exactly the four Beauville surfaces, while any nonempty $S$ yields infinitely many mapping-class-group orbits of $S$-integral Teichm\\\"uller points (Theorem 5.1 and Proposition 5.6).","For non-integral $\\mathbb{Q}$VHS with discrete monodromy, the Hodge locus can be non-algebraic and can contain a dense set of CM points without the variation being of Shimura type; the paper introduces the algebraic Hodge locus as the replacement object (Section 7)."],"supporting_citations":[{"why":"Supplies the Fenchel-Nielsen-type parametrization of PSL(2,C)-character varieties; its K-rationality statement is the step that makes the trace field Q in Theorem 1.6.","marker":"[16]"},{"why":"Gives Maskit's explicit parameters and fundamental domain for signature (0,4), used for the trace-ring computations and the Beauville reproof.","marker":"[23]"},{"why":"Defines RVHS and the non-abelian Hodge correspondence; Lemma 2.2 builds on its Lemma 4.8 to convert trace-field Q into a direct factor of a QVHS.","marker":"[32]"},{"why":"Provides the quaternion-algebra construction of a representation from its trace field, used in Lemma 2.2.","marker":"[21]"},{"why":"Beauville's classification of stable families of elliptic curves with four singular fibers, which Theorem 4.5 reproves and matches.","marker":"[6]"},{"why":"Margulis's commensurability criterion, used in Theorem 7.2 to show the Hodge locus of a non-integral QVHS need not be algebraic.","marker":"[22]"},{"why":"The Cattani-Deligne-Kaplan theorem on algebraicity of the Hodge locus for integral variations, whose analogue is shown to fail in Section 7.","marker":"[8]"},{"why":"Hitchin's uniqueness of the VHS on the Teichmuller component, used to prove Theorem 1.11 about totally real fields.","marker":"[15]"}],"fun_headline_variants":["Murphy's law: non-integral Hodge variations for every genus","Infinite non-integral QVHS per genus: trace ring overflows Z","Cattani-Deligne-Kaplan fails: non-integral variations for all genera","Every genus: trace field Q, but trace ring not Z","Non-integral QVHS for all genera: Murphy's law in Hodge theory"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction depends on the claim that rational choices of eigenvalue and twist parameters in the parametrization produce local systems whose trace field is $\\mathbb{Q}$; if that rationality step fails, none of the constructed examples are known to be direct factors of a $\\mathbb{Q}$-Hodge variation.","fun_headline_variants_meta":{"raw":{"variants":["Murphy's law: non-integral Hodge variations for every genus","Infinite non-integral QVHS per genus: trace ring overflows Z","Cattani-Deligne-Kaplan fails: non-integral variations for all genera","Every genus: trace field Q, but trace ring not Z","Non-integral QVHS for all genera: Murphy's law in Hodge theory"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000448,"raw_usage":{"total_tokens":2285,"prompt_tokens":994,"completion_tokens":1291,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":610,"completion_tokens_details":{"reasoning_tokens":1188}},"tokens_in":610,"tokens_out":1291,"duration_ms":11096,"temperature":1.0,"reasoning_tokens":1188,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:00:50.114276+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a small concrete case, say $g=2$ and $S=\\{2\\}$, write down the explicit matrices produced by the parametrization with $e_1=-1/2$ and all other parameters rational, and compute the trace ring and the Zariski closure of the monodromy; if any trace lies outside $\\mathbb{Z}[1/2]$, or the representation is not discrete and faithful, or the monodromy is not Zariski dense in $\\mathrm{SL}_2$, then Theorem 1.6 would be wrong.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Fenchel-Nielsen-type parametrization of PSL(2,C)-character varieties; its K-rationality statement is the step that makes the trace field Q in Theorem 1.6."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Maskit's explicit parameters and fundamental domain for signature (0,4), used for the trace-ring computations and the Beauville reproof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines RVHS and the non-abelian Hodge correspondence; Lemma 2.2 builds on its Lemma 4.8 to convert trace-field Q into a direct factor of a QVHS."},{"cited_title":"Maclachlan and A","cited_arxiv_id":null,"evidence_quote":"Provides the quaternion-algebra construction of a representation from its trace field, used in Lemma 2.2."},{"cited_title":"Beauville","cited_arxiv_id":null,"evidence_quote":"Beauville's classification of stable families of elliptic curves with four singular fibers, which Theorem 4.5 reproves and matches."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Margulis's commensurability criterion, used in Theorem 7.2 to show the Hodge locus of a non-integral QVHS need not be algebraic."},{"cited_title":"Cattani, P","cited_arxiv_id":null,"evidence_quote":"The Cattani-Deligne-Kaplan theorem on algebraicity of the Hodge locus for integral variations, whose analogue is shown to fail in Section 7."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Hitchin's uniqueness of the VHS on the Teichmuller component, used to prove Theorem 1.11 about totally real fields."}],"review_version":1}