{"id":"bd337be7-7089-418b-83fa-6fe0d96a9bb2","arxiv_id":"2501.13175","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Solutions of algebraic differential equations are conjectured to be algebraic exactly when their Taylor coefficients have almost no primes in their denominators, and the conjecture is proved for Picard-Fuchs equations and for isomonodromy equations such as Painlevé VI.","lead":"This paper proposes an arithmetic test for whether a solution to a differential equation is an algebraic function: check whether only finitely many primes appear in the denominators of the coefficients of its Taylor expansion. The authors prove this test works for equations coming from geometry, including the Painlevé VI and Schlesinger systems, and show the full conjecture would imply the Grothendieck-Katz p-curvature conjecture.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop. 9.2.3's replacement proof of logarithmic conjugate spectral sequence degeneration rests on an unsupported 'splits' claim; Lemma 11.3.1 and hence the main non-linear induction depend on it.","rationale":"The paper's linear results—Theorem 3.1.1, Theorem 4.1.1, and Proposition 5.2.2—appear sound and are largely independent of the contested positive-characteristic machinery; the hypergeometric interlacing argument and the elliptic conjugate-filtration argument are checkable from the text. The central non-linear claim is Theorem 6.1.3, whose proof reduces to Theorem 10.1.3 (the Griffiths-transverse deformation of the Hodge filtration). The induction proving Theorem 10.1.3 passes through Theorem 11.2.2, and Theorem 11.2.2 uses Lemma 11.3.1; Lemma 11.3.1's proof of Hodge–de Rham degeneration for Higgs–de Rham fixed points invokes Proposition 9.2.3. Proposition 9.2.3 is precisely where the authors replace the missing logarithmic Ogus–Vologodsky theorem with a short argument. That argument asserts, without proof, that a filtration on the Higgs complex splits and therefore its spectral sequence degenerates at E2. Since the Higgs differential is not grading-preserving, an OX-module splitting does not give a filtered splitting; the degeneration is therefore not established. This is not a complaint about external consensus or a stylistic issue; it is an internal step in the central proof that needs either a new argument or a written logarithmic reference. It does not amount to a disproof of the theorem, and the authors explicitly flag the missing logarithmic theorem in the text, so the appropriate action is to retain the reader's CONDITIONAL verdict, with the condition being a complete proof of Proposition 9.2.3 or a written filtered logarithmic version of [OV07, Thm. 3.22].","tokens_in":67234,"tokens_out":13462,"duration_ms":149509,"concrete_test":"Verify the contested degeneration directly in the simplest nontrivial logarithmic case: take S an Artin k-algebra, X/S a smooth projective curve, D a nonempty relative snc divisor, and a length-three graded Higgs bundle E′ = E′_0 ⊕ E′_1 ⊕ E′_2 with non-zero θ: E′_i → E′_{i−1}⊗Ω^1_{X/S}(log D), nilpotent of order at most p−1 and length condition of Prop. 9.2.3 satisfied. Compute the E2 and E∞ pages of the spectral sequence attached to the filtration F^m on (E′, θ)_Higgs. If E2 ≠ E∞, the asserted 'splits' step is invalid; if E2 = E∞, identify the explicit splitting of filtered complexes used in the proof and check it is compatible with θ. Either outcome settles whether the proof of Prop. 9.2.3 is valid.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step for the non-linear theorem is the proof of Proposition 9.2.3. After invoking [Sch05, Cor. 5.7], the proof introduces the filtration F^m = ⊕_{i≤−m} E′_i on the Higgs complex (E′, θ)_Higgs and states that the resulting spectral sequence 'degenerates at E2 as the filtration on (E′, θ)_Higgs splits by assumption.' This is not justified as written. The grading on E′ gives a direct-sum splitting of the underlying OX-modules, but the Higgs differential θ shifts the grading by −1; it is not a morphism of filtered complexes compatible with that splitting. A filtered complex with locally free, directly summand graded pieces need not have E2 = E∞; nontrivial higher differentials can occur in length-three filtrations. The hypothesis length(F_conj) < p − dim_S(X) is not shown to produce a splitting of filtered complexes, and the paper itself notes that a filtered logarithmic version of [OV07, Thm. 3.22] has not been written down. Thus the replacement argument in Prop. 9.2.3 leaves the logarithmic E2-degeneration unproved. This matters because Lemma 11.3.1 uses Prop. 9.2.3 to degenerate the conjugate spectral sequence for Higgs–de Rham fixed points; that degeneration is used in the induction proving Theorem 11.2.2, which in turn proves Theorem 10.1.3 and hence Theorem 6.1.3. The statement may be true and repairable by a genuinely filtered logarithmic non-abelian Hodge theory, but the proof as written does not establish it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper formulates an arithmetic conjecture (Conjectures 1.1.1, 2.3.1, and 6.1.1) asserting the equivalence, for formal leaves of foliations and formal flat sections, of algebraicity, integrality, and a much weaker p-adic condition called ω(p)-integrality. It then proves the conjecture for several classes of differential equations: Picard–Fuchs bundles at cycle-class initial conditions (Theorem 3.1.1), Gauss–Manin systems of non-isotrivial families of elliptic curves and their symmetric powers (Theorem 4.1.1), hypergeometric functions near a regular singular point (Proposition 5.2.2), and the non-linear isomonodromy foliation at Picard–Fuchs initial conditions (Theorems 1.2.4 and 6.1.3), with consequences for Painlevé VI and Schlesinger systems. The paper also shows that a sufficiently general instance of the isomonodromy conjecture would imply the classical Grothendieck–Katz p-curvature conjecture, and it discusses implications for relative Fontaine–Mazur-type variational conjectures.","tokens_in":67370,"tokens_out":9890,"duration_ms":99820,"significance":"If the main results are correct, this is a substantial contribution. The linear theorems are proven by a convincing combination of crystalline Frobenius and Chebotarev arguments with analytic finite-branching arguments, and the elliptic-curve and hypergeometric results are clean and well grounded in prior work. The non-linear theorem is an ambitious and natural analogue of Katz's p-curvature theorem, and it would give the first arithmetic, denominator-based classification of algebraic solutions for non-linear isomonodromy equations such as Painlevé VI. The authors are also explicit about the conjectural framework and about which steps rely on positive-characteristic non-abelian Hodge theory; this transparency is a strength. However, the central non-linear theorem currently depends on a degeneration statement whose proof in the logarithmic setting is incomplete, so the significance of the paper can only be fully assessed after that gap is repaired.","major_comments":[{"comment":"The replacement proof of logarithmic E2-degeneration is not complete. After invoking [Sch05, Cor. 5.7], the proof considers the filtration F^m = ⊕_{i≤−m} E′_i on the Higgs complex (E′, θ)_Higgs and asserts that the resulting spectral sequence “degenerates at E2 as the filtration on (E′, θ)_Higgs splits by assumption.” The grading on E′ gives a direct-sum splitting of the underlying O_X-modules, but the Higgs differential θ shifts the grading by −1 and is not a morphism of filtered complexes compatible with that splitting; in fact θ(F^m) ⊂ F^{m+1} ⊗ Ω^1_{X/S}(log D). A filtered complex whose graded pieces are locally free direct summands need not have E2 = E∞, and the hypothesis length(F_conj) < p − dim_S(X) is not shown to produce a splitting of filtered complexes. Because the paper itself states that a filtered logarithmic version of [OV07, Thm. 3.22] has not been written down, the argument as written does not establish the logarithmic E2-degeneration claimed.","section":"§9.2, Proposition 9.2.3"},{"comment":"The gap in Proposition 9.2.3 is load-bearing for the main non-linear theorem. Lemma 11.3.1 uses Proposition 9.2.3 to degenerate the conjugate spectral sequence for a Higgs–de Rham fixed point; that degeneration is used in the induction in Theorem 11.2.2 to prove smoothness of the deformation space Def_{F_i,∇,F}, which yields Theorem 10.1.3 and hence Theorem 6.1.3 and Theorem 1.2.4. Unless Proposition 9.2.3 is supplied with a valid proof, or replaced by a written logarithmic/filtered version of the Ogus–Vologodsky theory, the central non-linear claim is not established by the text as it stands. The statement may well be true and repairable, but the present proof does not justify it.","section":"§11.2–§12, Theorem 6.1.3"}],"minor_comments":[{"comment":"The displayed isomorphism has the same bundle (E′, ∇′) on both sides; the left side should be the original bundle (E, ∇). Please correct this typo.","section":"§12.1, Lemma 12.1.1"},{"comment":"The hypothesis “the length of s is less than p − dim_A X” appears to be a typo; presumably the intended hypothesis is that the length of the filtration F (or of the conjugate filtration) is less than p − dim_A X.","section":"§11.3, Lemma 11.3.1"},{"comment":"The notation H^{2i}_dR(X_s)_s has an extra subscript and should be cleaned up; the same issue appears in the statement of Proposition 3.3.1.","section":"§3.1, Theorem 3.1.1"},{"comment":"The term “good” is imported from [EK24, §1] without definition; since the theorem is used for pairs (X, D) with a logarithmic divisor, please state the condition or give a precise pointer to the definition.","section":"§8.3, Theorem 8.3.3"},{"comment":"The definition of ω(f)-integrality uses a general function f while the conjecture and the rest of the paper use ω(p); aligning the notation would avoid confusion.","section":"§2.2, Definition 2.2.5"},{"comment":"The sentence “After replacing R with a finitely-generated localization, we may assume it is Z-smooth” has an ambiguous antecedent; please clarify whether R or Spec(A) is meant.","section":"§10.3.1, Assumption 10.3.5"}],"recommendation":"major_revision","confidential_remarks":"The paper relies heavily on literature that is not uniformly easy to verify: Schepler's thesis [Sch05], Ogus–Vologodsky [OV07], and the presentation [EG25]. For a journal at this level, the dependence should be made fully explicit and, where possible, self-contained. The unfinished proof of Proposition 9.2.3 is the main obstacle: it is a load-bearing step in the non-linear theorem, not a presentation issue. I would not recommend rejection: the linear theorems and the overall architecture are valuable, and the gap appears repairable, but it needs to be fixed before the central claim can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: read the linear half, treat the non-linear half as conditional. The conjecture is genuinely new—equating algebraicity, integrality, and ω(p)-integrality for solutions to algebraic differential equations, with the foliation version strengthening Grothendieck-Katz. The paper proves it for Picard-Fuchs equations at cycle-class initial conditions (Thm 3.1.1), for families of elliptic curves (Thm 4.1.1), and for hypergeometric functions at zero (Prop 5.2.2). I checked the interlacing argument, the conjugate-filtration argument, and the Chebotarev-plus-Fontaine-Laffaille step; they are sound. The exposition is honest about its limits—see Remarks 3.3.5 and 10.1.2, and the explicit patch in Prop 9.2.3.\n\nThe soft spot is exactly where the stress-test lands. Prop 9.2.3 claims the conjugate spectral sequence degenerates at E2 because 'the filtration on (E', θ)_Higgs splits by assumption.' That does not follow. The grading splits the underlying O_X-modules, but the Higgs field shifts the grading by -1, so it is not a map of filtered complexes compatible with the splitting. Higher differentials can occur. The hypothesis on the length of the filtration does not obviously force a filtered-splitting. This matters: Lemma 11.3.1 uses Prop 9.2.3 to degenerate the conjugate spectral sequence for Higgs–de Rham fixed points, and that degeneration feeds into the induction proving Theorem 11.2.2, then 10.1.3, then 6.1.3. So the main non-linear theorem is not established as written. The statement may well be true and repairable via a genuinely filtered logarithmic version of Ogus–Vologodsky, but that has not been written down.\n\nWho this is for: anyone working on the p-curvature conjecture, arithmetic of differential equations, or algebraic solutions of Painlevé VI. The linear results and the conjecture itself are worth taking seriously now. The non-linear part should be cited cautiously until the gap is closed.\n\nRecommendation: send to a serious referee. The paper is important enough and the linear half is strong enough to deserve referee time. The referee should focus on §9–§11 and the appendices, and should ask for a complete proof of Prop 9.2.3 or a citation to a filtered logarithmic OV theorem.","headline":"Real conjecture, solid linear results, but the non-linear main theorem currently rests on an unproved degeneration step in Prop. 9.2.3.","tokens_in":68162,"tokens_out":2527,"would_cite":true,"duration_ms":26894,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14F40","14F30","14D07","34M55","12H05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that for Picard-Fuchs equations and for non-linear isomonodromy systems such as Painlevé VI, a formal solution through an algebro-geometric initial value is algebraic if and only if the primes dividing the denominators…","keywords":["algebraic differential equations","Grothendieck-Katz p-curvature conjecture","Picard-Fuchs equations","isomonodromy foliations","Painlevé VI equation","p-integrality of Taylor coefficients","Hodge filtration","inverse Cartier transform"],"falsifier":"The cleanest single test: find a Picard-Fuchs equation and a cycle-class initial condition whose formal flat section is integral (or $\\omega(p)$-integral) but not algebraic — that would refute Theorem 3.1.1 directly; for the non-linear theorem, an isomonodromy leaf through a Picard-Fuchs point that is $\\omega(p)$-integral but not algebraic would refute Theorem 6.1.3. At the level of the missing ingredient, one concrete computation would be to exhibit a logarithmic flat bundle in characteristic $p$ for which the inverse Cartier transform fails the $p$-curvature formula of Lemma B.4.3, since that formula is the hinge of the whole reduction.","tokens_in":2737,"feed_emoji":"🧮","tokens_out":2899,"duration_ms":152146,"temperature":0.7,"pith_summary":"This paper tries to establish a single arithmetic criterion for algebraicity that spans linear and non-linear algebraic differential equations: a formal solution expanded at a non-singular point is algebraic if and only if the primes dividing the denominators of its Taylor coefficients form a finite set (integrality), and if and only if a weaker order-by-order $p$-integrality (the $\\omega(p)$-integrality condition) holds. For linear equations the paper proves the criterion for Picard-Fuchs equations — flat bundles with regular singularities arising as the relative de Rham cohomology of a smooth projective family — at initial conditions given by cycle classes, and unconditionally for families of elliptic curves. For non-linear equations it proves the criterion for isomonodromy foliations, the setting containing the Schlesinger system and the Painlevé VI equation, at initial conditions that are themselves Picard-Fuchs equations: the leaf through such a bundle is algebraic exactly when its formal leaf is integral. A sympathetic reader would care because this gives a uniform arithmetic test — check which primes divide Taylor denominators — in a non-linear setting where algebraic solutions were previously found by case-by-case classification, and because the same conjecture would imply the classical Grothendieck-Katz $p$-curvature conjecture if it held for the isomonodromy foliations over moduli spaces of curves.","feed_headline":"Denominator primes decide which solutions are algebraic","feed_subtitle":"Integral Taylor coefficients at one point now provably force algebraicity for Picard-Fuchs and Painlevé-type systems — and conversely.","key_machinery":"The central object is the denominator test itself: $\\omega(p)$-integrality, meaning that for each prime $p$ the first $\\omega(p)$ Taylor coefficients of the solution have denominators prime to $p$, for some function $\\omega(p)$ growing faster than $p$; ordinary integrality is the stricter statement that only finitely many primes ever divide denominators, which a nineteenth-century theorem already shows is necessary for algebraicity. The proof of the non-linear theorem is carried by a characteristic-$p$ comparison: the obstruction to extending the Hodge filtration Griffiths-transversally to the formal isomonodromic deformation is identified, over large primes, with the $p$-curvature of the isomonodromy foliation on the moduli stack of flat bundles (the foliation whose leaves keep the monodromy representation of the flat bundle fixed), using the inverse Cartier transform (a characteristic-$p$ correspondence between nilpotent Higgs bundles and flat bundles with nilpotent $p$-curvature) and the Higgs-de Rham flow (a periodicity procedure showing Picard-Fuchs bundles are fixed points of the transform). Integrality of the formal leaf forces that $p$-curvature to vanish; vanishing makes the Hodge filtration extend; the extension triggers a finiteness theorem for local systems underlying integral variations of Hodge structure; and finiteness of the monodromy orbit yields algebraicity of the leaf. In the linear theorems the same role is played by the crystalline Frobenius, which reads the Hodge filtration directly off the denominators of a formal flat section.","core_discovery":"The central claim, stated in Theorem 6.1.3, is that the paper's Conjecture 6.1.1 holds when $(E,\\nabla)$ is a Picard-Fuchs equation: for a flat bundle with regular singularities that arises, with its Gauss-Manin connection, as the relative de Rham cohomology of a smooth projective family, the leaf of the isomonodromy foliation through $[(E,\\nabla)]$ is algebraic if and only if the formal leaf is integral, if and only if it is $\\omega(p)$-integral. In the linear case, Theorem 3.1.1 establishes the same equivalence for the formal flat section to a Picard-Fuchs equation whose initial condition lies in the image of the cycle class map; Theorem 4.1.1 goes further for $\\mathrm{Sym}^n$ of the Gauss-Manin bundle of a non-isotrivial family of elliptic curves, proving that no non-zero formal section is integral or $\\omega(p)$-integral, so that every integral section is algebraic. The paper also verifies the conjecture's local prediction for hypergeometric functions, where integrality of the Taylor series at a non-singular point forces either algebraicity or infinite-order local monodromy.","pith_inferences":["If the denominator criterion survives in its conjectured generality, it gives a finite test against algebraicity: to certify that a candidate solution is not algebraic, one need not compute all Taylor coefficients, only find, for a growing sequence of primes $p$, one coefficient among the first $\\omega(p)$ whose denominator is divisible by $p$; the theorems prove this obstruction necessarily appea","The paper's own flagged gap — that the logarithmic version of the inverse Cartier transform needed for Proposition 9.2.3 had not been written down — is the natural stress point to examine first: if that transform is supplied, Remark 10.1.2 suggests the nilpotent-residue hypothesis would drop away and the non-linear theorem would extend to arbitrary regular-singularity bundles.","Read through the paper's meta-conjectural lens (its Conjectures 14.2.2 and 14.2.3), the non-linear theorem supplies the first arithmetic hypothesis — $\\omega(p)$-integrality of the formal leaf — under which a Griffiths-transverse Hodge filtration provably extends to the deformation; the further step, that the whole deformation underlies a true variation of Hodge structure, is what the conjectural "],"forward_implications":["The Schlesinger system and Painlevé VI now have an arithmetic classification at Picard-Fuchs initial conditions: a solution is algebraic exactly when the denominators of its power series coefficients are confined to finitely many primes, and the weaker $\\omega(p)$-integrality condition already suffices.","Non-isotrivial families of elliptic curves produce no non-zero integral formal sections in $\\mathrm{Sym}^n$ of the Gauss-Manin bundle, so the concrete recurrences controlling such Taylor coefficients must have infinitely many primes dividing denominators, as Proposition 4.0.1 states for an explicit third-order recurrence.","For hypergeometric equations, an integral Taylor series at a non-singular point that is not algebraic forces infinite-order monodromy around the singular point, confirming the local form of the conjecture in that classical case.","Verifying the conjecture for the isomonodromy foliations over the moduli stacks of curves of all sufficiently large genus would imply the full Grothendieck-Katz $p$-curvature conjecture for every flat bundle, so the new criterion is strictly stronger than the classical one.","Flat bundles with vanishing $p$-curvature modulo almost all primes automatically admit $\\omega(p)$-integral isomonodromic deformations (Lemma 13.0.1), so the non-linear conjecture specializes to a strengthening, not merely a restatement, of the $p$-curvature conjecture."],"supporting_citations":[{"why":"Supplies the classical theorem that an algebraic power series has coefficients with only finitely many prime denominators, the 'algebraic implies integral' direction of the conjecture.","marker":"[Eis52]"},{"why":"Provides the linear antecedent — the proof of the Grothendieck-Katz $p$-curvature conjecture for Gauss-Manin connections — which the paper's main theorems are announced as the non-abelian analogue of.","marker":"[Kat72]"},{"why":"Supplies the inverse Cartier transform in positive characteristic that the paper uses to relate the Hodge-filtration obstruction to $p$-curvature.","marker":"[OV07]"},{"why":"Provides the logarithmic non-abelian Hodge theory results that stand in for a version of the inverse Cartier transform the paper states has not been written down (Proposition 9.2.3).","marker":"[Sch05]"},{"why":"Supplies the Higgs-de Rham flow used to lift the characteristic-$p$ construction back to characteristic zero and to show Picard-Fuchs bundles are fixed points of the flow.","marker":"[LSZ19]"},{"why":"Provides the Artin-Rees construction and flow-functor formalism that the lifting argument in Section 10.4 follows.","marker":"[EG25]"},{"why":"Supplies the finiteness theorem equating Griffiths-transverse extension of the Hodge filtration with finiteness of the monodromy orbit, the analytic hinge of the main theorem.","marker":"[EK24]"},{"why":"Provides the monodromy and algebraicity classification of hypergeometric functions that Proposition 5.2.2 uses to prove infinite local monodromy for non-algebraic integral cases.","marker":"[BH89]"},{"why":"Supplies the integrality criterion for hypergeometric Taylor coefficients used in that same argument.","marker":"[Chr86]"},{"why":"Supplies the finiteness theorem for local systems underlying integral variations of Hodge structure that underlies the analytic step (Theorem 8.3.1).","marker":"[Del87]"}],"fun_headline_variants":["Integrality implies algebraicity for Painlevé and Picard-Fuchs","Integral power series at one point force algebraic solutions","Algebraicity proved from one integral Taylor expansion","For key differential equations, integrality implies algebraicity"],"cache_read_input_tokens":69888,"weakest_assumption_plain":"The non-linear theorem rests on a version of positive-characteristic non-abelian Hodge theory for logarithmic connections that the paper itself says has not been written down (Proposition 9.2.3); if the inverse Cartier transform or its $p$-curvature computation fails there, the comparison between the Hodge-filtration obstruction and the isomonodromy $p$-curvature breaks.","fun_headline_variants_meta":{"raw":{"variants":["Integrality implies algebraicity for Painlevé and Picard-Fuchs","Integral power series at one point force algebraic solutions","Algebraicity proved from one integral Taylor expansion","For key differential equations, integrality implies algebraicity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001307,"raw_usage":{"total_tokens":5330,"prompt_tokens":947,"completion_tokens":4383,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":4317}},"tokens_in":563,"tokens_out":4383,"duration_ms":33663,"temperature":1.0,"reasoning_tokens":4317,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:26:08.692890+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The cleanest single test: find a Picard-Fuchs equation and a cycle-class initial condition whose formal flat section is integral (or $\\omega(p)$-integral) but not algebraic — that would refute Theorem 3.1.1 directly; for the non-linear theorem, an isomonodromy leaf through a Picard-Fuchs point that is $\\omega(p)$-integral but not algebraic would refute Theorem 6.1.3. At the level of the missing ingredient, one concrete computation would be to exhibit a logarithmic flat bundle in characteristic $p$ for which the inverse Cartier transform fails the $p$-curvature formula of Lemma B.4.3, since that formula is the hinge of the whole reduction.","supporting_citations":[],"review_version":1}