REVIEW 4 major objections 5 minor 15 references
De Rham-Higgs comparison for mixed Hodge modules in positive characteristic
T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Frobenius pushforwards of twisted de Rham subcomplexes match Higgs complexes at every closed point and globally when the cotangent splits.
desk verdict Valuable new results, but Theorem 4.4's proof has a false determinant-one factorization claim, so the paper needs a serious revision before it can be trusted. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
Two mechanisms carry the argument. The first is a deformed Higgs field, a variant of the $\alpha$-transform: on the completion along the zero scheme of the coefficients $f_{i_1\cdots i_n}$, the Higgs field $\hat{\theta} - df\wedge$ is replaced by $\Theta := \hat{\theta} - \sum_{i_1,\ldots,i_n}\sum_{j\ge 0} f_{i_1\cdots i_n}^{p^j-1}\, df_{i_1\cdots i_n}\wedge$, whose Higgs complex is isomorphic to the original twisted one while its inverse Cartier transform is isomorphic to $(\hat{H}, \hat{\nabla} + df\wedge)$; the isomorphism of Cartier transforms is multiplication by the Artin--Hasse exponential $\prod \mathrm{AH}(f_{i_1\cdots i_n})$. The second is the splitting hypothesis $\Omega^1_{X/k}(\log D) = \bigoplus_{i=1}^{\beta}\Omega_i$ with $\mathrm{rank}\,\Omega_i < p - \ell$, which feeds an explicit $L$-indexed $\infty$-homotopy $Ho_\Omega$ built from higher-homotopy formulas $\varphi^{(r,s)}$ -- combinatorial sums over Frobenius liftings with coefficients $C(i,S,j)$ -- realizing the quasi-isomorphism $\varphi_\Omega: K^\bullet_{\mathrm{Hig}} \to \check{C}(U', F_\ast K^\bullet_{\mathrm{dR}})$. The same formulas decompose $F_\ast\Omega^\bullet_{X/k}(\log D)$ into Koszul complexes $K^\ast_v$ that are acyclic except at $v = 0$; compatibility of the homotopy with the divisor or the semistable map (Definition 4.2) makes the isomorphism restrict to the subcomplexes, and independence of the splitting up to homotopy yields the two-term truncation theorem.
What would settle it
Check a supersingular abelian variety of dimension $> p$ (for instance a supersingular abelian fourfold over $\mathbb{F}_2$: dimension $4 > p = 2$ and $p$-rank $0 < g-1 = 3$, so it is not quasi-$F$-split). Corollary 1.3 predicts $F_\ast\Omega^\bullet_{A/k}$ is decomposable in $D(A)$; computing its Frobenius pushforward through Hodge--Witt cohomology or the de Rham--Dieudonné theory cited in the paper and finding no direct-sum splitting would refute the central claim, while confirming a split where the quasi-$F$-split theorem does not apply would corroborate it.
Extended reading notes
Core claim
The central claim is that the paper's Question 1.1 has an affirmative answer in three regimes. Formally: for every closed point $x$ and every global function $f$, $F_\ast(K^\bullet_{\mathrm{dR}}, \nabla + df\wedge) \otimes \hat{\mathcal{O}}_{X',x'} \cong (K^\bullet_{\mathrm{Hig}}, \theta - df'\wedge) \otimes \hat{\mathcal{O}}_{X',x'}$ in the derived category, for all three kinds of subcomplexes; a corollary is that the cohomology supports of the two twisted complexes agree, and if that support is a finite set of closed points the local isomorphisms glue to a global isomorphism. Globally and without truncation: if $\Omega^1_{X/k}(\log D)$ splits as $\bigoplus_i \Omega_i$ with $\mathrm{rank}\,\Omega_i < p - \ell$, then $F_\ast\Omega^\ast(H,\nabla) \cong \Omega^\ast(E,\theta)$ in $D(X)$, and more generally $\tau_{<q}K^\bullet_{\mathrm{Hig}} \cong \tau_{<q}F_\ast K^\bullet_{\mathrm{dR}}$ for the mixed-Hodge-type subcomplexes, with $q = \infty$ under the full splitting and $q = p - \ell$ otherwise; the isomorphism is independent of the chosen splitting up to homotopy. Unconditionally: for Hodge pairs of types I--IV with nilpotent level at most $p-3$, the two-sided truncations agree, $\tau_{[a,a+1]}F_\ast K^\bullet_{\mathrm{dR}} \cong \tau_{[a,a+1]}K^\bullet_{\mathrm{Hig}}$ in $D(X')$ for every $a \ge 0$, giving isomorphisms $F_\ast H^a(K^\bullet_{\mathrm{dR}}) \cong H^a(K^\bullet_{\mathrm{Hig}})$ of cohomology sheaves.
Load-bearing premise
The load-bearing premise is that the logarithmic cotangent bundle $\Omega^1_{X/k}(\log D)$ splits into a direct sum of subbundles, each of rank strictly below $p - \ell$ and, for the subcomplex results, with the splitting compatible with the divisor or the semistable map; the global untruncated theorems have no stated proof when this splitting fails.
Editorial extensions
If this is right
- Under the splitting hypothesis the truncation disappears: $F_\ast\Omega^\ast(H,\nabla) \cong \Omega^\ast(E,\theta)$ in $D(X)$, a full untruncated de Rham--Higgs comparison that extends the Deligne--Illusie decomposition beyond the range where truncation is usually necessary.
- If $\Omega^1_{X/k}(\log D)$ splits into subbundles of rank $< p$, then $F_\ast\Omega^\bullet_{X/k}(\log D)$ is decomposable; in particular every abelian variety, including supersingular ones of dimension $> p$ with $p$-rank $< g-1$ (hence not quasi-$F$-split), has decomposable Frobenius-pushed de Rham complex.
- For Hodge pairs of types I--IV with nilpotent level at most $p-3$, one gets $\tau_{[a,a+1]}F_\ast K^\bullet_{\mathrm{dR}} \cong \tau_{[a,a+1]}K^\bullet_{\mathrm{Hig}}$ in $D(X')$ for every $a \ge 0$, and hence $F_\ast H^a(K^\bullet_{\mathrm{dR}}) \cong H^a(K^\bullet_{\mathrm{Hig}})$.
- Applications include a Künneth formula for the decomposition, $E_1$-degeneration of Hodge-to-de Rham spectral sequences for Fontaine--Faltings modules (recovering Oda's degeneration and Mumford's vanishing for supersingular abelian varieties), and mod-$p$ proofs of characteristic-0 results: dimension equalities for twisted de Rham cohomology and Kodaira--Saito vanishing.
Reading between the lines
- Because the formal-local comparison needs no hypothesis on $f$ or on the splitting, the global obstruction is purely a gluing problem for the deformed Higgs fields $\Theta$; any geometric condition that makes the local isomorphisms compatible on overlaps -- not only a cotangent splitting -- should yield the same untruncated comparison, a testable route to weaken the splitting hypothesis.
- The abelian-variety corollary suggests the splitting hypothesis is a condition on the cotangent sheaf (a direct sum of small pieces), not on $F$-splitting; one could test whether smooth varieties in characteristic $p$ whose tangent bundle splits into line bundles -- the positive-characteristic analogue of split-tangent varieties -- all satisfy the untruncated decomposition.
- The two-term truncation theorem is unconditional and is the part most likely to extend: if the argument is stable under variation of the twisting function $f$, it could yield a mod-$p$ proof of degeneration of the full irregular Hodge filtration in characteristic 0, going beyond the dimension equalities and vanishing statements derived in Section 5.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a positive-characteristic generalization of the Deligne–Illusie decomposition theorem in which the complexes under comparison are subcomplexes inspired by mixed Hodge modules (weight filtrations, intersection complexes, Kontsevich subcomplexes), twisted by a global function f. The main construction combines the inverse Cartier transform with a deformed Higgs field Θ, yielding a formal-local comparison near the zeros of certain coefficients of f (Theorem 3.1). A global comparison is then proved under the assumption that the logarithmic cotangent bundle splits as a direct sum of subbundles of rank < p−ℓ (Theorem 4.2), with results on independence of the splitting (Proposition 4.1), compatibility with Hodge pairs of types I–IV (Theorem 4.3), and two-term truncations for arbitrary a (Theorem 4.4). The final sections derive Künneth, E1-degeneration, and vanishing consequences in positive characteristic and characteristic zero.
Significance. If the main results are correct, they give a substantial extension of the Ogus–Vologodsky and Sheng–Zhang comparisons and a new route to Deligne–Illusie-type decompositions for twisted complexes, with applications to abelian varieties and semistable families. The formal-local theorem of Section 3 is a genuine contribution, with an explicit deformation via Artin–Hasse exponentials. However, the global theorems are heavily conditional on the splitting assumption (21), which is strong and fails for general smooth varieties; consequently Corollary 1.3 and the applications in Section 5 are limited to a special class of varieties. More importantly, several load-bearing steps in the proofs of Sections 4.2 and 4.4 are either omitted or based on incorrect assertions, so the global comparison theorems are not established as written.
major comments (4)
- [§4.4, proof of Theorem 4.4, chain (53)] The proof asserts that, after shrinking X, the change-of-basis matrix A defined by (dlog f, dlog m2, ..., dlog mn) = (dlog f, dlog t2, ..., dlog tn)A can be written as a product of elementary matrices of the three listed types. Since every listed transformation has determinant ±1, such a factorization would force det A = ±1. This is false in general. For example, on X = G_m × G_m over a field k of characteristic p ≠ 3, take f = xy, t2 = y, m1 = y^2/x, and m2 = x^2/y. Then m1m2 = xy = f, and dlog m2 = 2 dlog f − 3 dlog y, so A = [[1,2],[0,−3]] has determinant −3. Passing to a Zariski open subset cannot change this determinant because it is a nonzero constant in k. Hence the required chain (53) of f-compatible splittings does not exist by this argument, and Proposition 4.1 cannot be applied along such a chain. This is a load-bearing gap in the proof of Theorem 4.4.
- [§4.2, Lemma 4.2] Lemma 4.2 states that Construction 4.3 is independent of the choice of basis, but its proof is given as 'similar to that of Lemma 4.1, and we omit the details.' This lemma is the key input to Proposition 4.1, which in turn underlies the compatibility theorem and the two-term truncation theorem. The verification is not a routine detail: Construction 4.3 involves the coefficient C_o and the index set H_{Ω,o}, both of which are substantially more complicated than the corresponding data in Lemma 4.1. Without a complete proof, the homotopy between Ho_Ω and Ho_Ω′ is not established.
- [§4.2, proof of Proposition 4.1] Even if Lemma 4.2 were available, the proof of Proposition 4.1 relies on a very long coefficient comparison. The identities c^L_{S,i,j} = c^R_{S,i,j}, c^L_{l,q,i,S,i,j} = 0, and c^L_{u,q,i,S,i,j} = 0 are verified only in 'typical situations', and the remaining cases are dismissed by 'direct computation'. Since this proposition is the only mechanism for proving that the de Rham–Higgs comparison is independent of the chosen splitting, the proof as written is not complete. A full verification, or a more conceptual argument, is needed before Theorem 4.3 and Theorem 4.4 can be accepted.
- [§4.4, Theorem 4.4, proof for all types] The proof of Theorem 4.4 explicitly treats only Hodge pairs of type (IV,0) and says that the first three types 'can be checked in a similar manner.' Given the intricate compatibility and independence arguments in Sections 4.2–4.3, this is not sufficient. In particular, type (II,ℓ) involves the intersection subcomplexes whose compatibility is proved separately via [SZ2], and type (III,ℓ) involves the Kontsevich subcomplexes; the analogous chain-of-splittings argument is not supplied for these cases.
minor comments (5)
- [Introduction, Question 1.1] The phrase 'de Rhm-Higgs comparison' should read 'de Rham-Higgs comparison'.
- [§2.6, type (III)] The word 'semistble' is a typo for 'semistable'.
- [Corollary 3.2, proof] The sentence 'Since ˆO_{X,x} is fully faithful flat over O_{X,x}' should say 'faithfully flat'; the argument only requires faithful flatness to detect non-vanishing of coherent sheaves.
- [§3.2, Definition 3.2 and following] In the sentence 'V (t^α ω_I) = V (γ_i) or V (−γ_i), i ≤ i ≤ n', the index range should be '1 ≤ i ≤ n'.
- [§5.1] In the proof of Theorem 5.1, the text says 'For i = 1, 2, 3, we set H^*_i ...' but only i = 1, 2 are used; this is confusing and should be corrected.
Circularity Check
No significant circularity: the twisted de Rham–Higgs comparison is constructed from explicit deformed Higgs fields and prior published Cartier-transform inputs, not assumed as its own conclusion.
full rationale
The central comparison is not circular. In §3, the f-twisted statement is obtained by explicitly constructing a deformed Higgs field Θ (Lemma 3.1) whose inverse Cartier transform is shown, by a direct Artin–Hasse exponential calculation, to be isomorphic to (Ĥ, ∇+df), and whose Higgs complex is shown (Lemma 3.2) to be isomorphic to (Ê, θ−df). Step 2 then applies the standard Cartier decomposition to C^{-1}_F(Ê,Θ), an input theorem (Ogus–Vologodsky, Schepler), not the target statement. In §4, the untruncated global comparison is built through explicit higher-homotopy formulas (28) and coefficient identities, rather than by assuming the desired isomorphism; Proposition 4.1 verifies homotopy independence by a direct coefficient check. Citations to [SZ2] supply the prior untwisted framework and are published, externally checkable inputs; they do not assume the new twisted comparison. The proof of Theorem 4.4 does contain a serious unsupported reduction: the claim that the coordinate-change matrix A decomposes into the displayed determinant-one elementary matrices is not valid in general (e.g., on G_m×G_m with f=xy and m2=x^2/y, det A=−3), so the asserted chain (53) is not justified. This is a correctness/proof gap in one global-independence argument, not a circularity: the theorem’s content is not assumed as an input, and the formal/local comparisons stand independently. Hence circularity score 0.
Assumptions & free parameters
assumptions (6)
- domain assumption The inverse Cartier transform C^{-1} exists for nilpotent Higgs bundles of level <= p-1 on (X,D)/k and is an equivalence from HIG_ell to MIC_ell.
- domain assumption The pair (X,D)/k admits a W2(k)-lifting (X~,D~).
- domain assumption The logarithmic cotangent bundle Omega^1_{X/k}(log D) splits into subbundles of rank < p-ell in the global theorems.
- ad hoc to paper The splitting is compatible with the divisor D or the semistable family g in the sense of Definition 4.2 when dealing with subcomplexes.
- standard math The higher homotopy formalism of Sheng-Zhang [SZ1,SZ2] is correct and applies in the present generality.
- standard math Abelian varieties have trivial cotangent bundle, so Omega^1_A splits into line bundles.
Cite this review
Pith. "Pith review of De Rham-Higgs comparison for mixed Hodge modules in positive characteristic." pith.science (2026). https://pith.science/paper/UBB7XJ62
@misc{pith2026250715175,
author = {Pith},
title = {Pith review of: De Rham-Higgs comparison for mixed Hodge modules in positive characteristic},
year = {2026},
howpublished = {\url{https://pith.science/paper/UBB7XJ62}},
note = {Machine review of arXiv:2507.15175}
}
read the original abstract
Building on the nonabelian Hodge theory in positive characteristic developed by Ogus, Vologodsky, and Schepler, we propose a generalization of the decomposition theorem of Deligne and Illusie from the perspective of mixed Hodge modules. This generalization is verified in certain special cases by extending the method of Sheng and the author, leading to several interesting byproducts.
Reference graph
Works this paper leans on
-
[1]
[AS] P. Achinger, J. Suh, Some refinements of the Deligne-Illusie theorem, Algebra Number Theory17(2023), no.2, 465–496. [A WZ] P. Achinger, J. Witaszek, M. Zdanowicz, Global Frobenius liftability I, J. Eur. Math. Soc. (JEMS) 23 (2021), no. 8, 2601–2648. [Ar] D. Arapura, Kodaira-Saito vanishing theorem via Higgs bundles in positive char- acteristic, Journ...
work page 2023
-
[5]
[ESY] H. Esnault, Y. Sabbah, J. Yu, E1-degeneration of the irregular Hodge filtration. With an appendix by Morihiko Saito, J. Reine Angew. Math. 729 (2017), 171-
work page 2017
-
[9]
Katz, Nilpotent connections and the monodromy theorem: Applications of a result of Turrittin, Inst
[Kat] N. Katz, Nilpotent connections and the monodromy theorem: Applications of a result of Turrittin, Inst. Hautes ´Etudes Sci. Publ. Math.(1970), no.39, 175–232. [KKP] L. Katzarkov, M. Kontsevich, T. Pantev, Bogomolov–Tian–Todorov theorems for Landau Ginzburg models, J. Differential Geom. 105 (2017), 55–117. [LSYZ] G.-T. Lan, M. Sheng, Y.-H. Yang, K. Zu...
work page 1970
-
[11]
Oda, The first de Rham cohomology group and Dieudonn´ e modules, Ann
[Oda] T. Oda, The first de Rham cohomology group and Dieudonn´ e modules, Ann. scient. ´Ec. Norm. Sup. (4), 2, (1969), 63–135. [O94] A. Ogus, F-crystals, Griffiths transversality, and the Hodge decomposition, Ast´ erisque 221,
work page 1969
-
[1970]
[DI] P. Deligne, L. Illusie, Rel` evements modulo p2 et decomposition du complexe de de Rham, Invent. Math. 89 (1987), 247-270. [Dru] S. Druel, Vari´ et´ es alg´ ebriques dont le fibr´ e tangent est totalement d´ ecompos´ e, J. Reine Angew. Math., 522:161–171,
work page 1987
-
[1973]
An explicit infinite homotopy in nonabelian Hodge theory in positive characteristic
[SZ1] M. Sheng, Z. Zhang, An explicit infinite homotopy in nonabelian Hodge theory in positive characteristic, arXiv:2302.09231. [SZ2] M. Sheng, Z. Zhang, Intersection de Rham complexes in positive characteristic, J. Differential Geom. 127(2): 551-602 (June 2024). DOI: 10.4310/jdg/1717772421. [Shi] A. Shiho, Notes on generalizations of local Ogus-Vologods...
work page Pith review arXiv 2024
-
[1987]
50 [Ka] K. Kato, Logarithmic structures of Fontaine-Illusie, Algebraic analysis, geometry, and number theory (Baltimore, MD, 1988), 191-224, Johns Hopkins Univ. Press, Baltimore, MD,
work page 1988
-
[1989]
[FL] J.-M. Fontaine, G. Laffaille, Construction de repr´ esentation p-adiques, Ann. Sci. Ec. Norm. Sup. 15 (1982), 547-608. [J] K. Joshi, On varieties with trivial tangent bundle in characteristic p >0, Nagoya Math. J. 242 (2021), 35–51. [H] A. H¨ oring, Uniruled varieties with split tangent bundle, Math. Z. 256 (2007), no. 3, 465–479. [I90] L. Illusie, R...
work page 1982
Show all 15 references
-
[1991]
[LM] S. Li, S. Mondal, On endomorphisms of the de Rham cohomology func- tor.(English summary)Geom. Topol.28(2024), no.2, 759–802. [LZ] S. Li, D. Zhang, Exponentially twisted de Rham cohomology and rigid cohomol- ogy, arXiv:2111.05689. [Mum] D. Mumford, Abelian varieties, Tata ...
2024 arXiv
-
[1992]
Faltings, Crystalline cohomology and p-adic Galois-representations, Algebraic analysis, geometry, and number theory (Baltimore, MD, 1988), 25-80, Johns Hopkins Univ
[Fa] G. Faltings, Crystalline cohomology and p-adic Galois-representations, Algebraic analysis, geometry, and number theory (Baltimore, MD, 1988), 25-80, Johns Hopkins Univ. Press, Baltimore, MD,
1988
-
[1994]
Ogus, Higgs cohomology, p-curvature, and the Cartier isomorphism
[O04] A. Ogus, Higgs cohomology, p-curvature, and the Cartier isomorphism. Compo- sitio Math. 140 (2004), 145-164. [O18] A. Ogus, Lectures on logarithmic algebraic geometry, Cambridge Studies in Advanced Mathematics,
2004
-
[2000]
[BL] B. Bhatt J. Lurie, The prismatization of p-adic formal schemes, 2022, arXiv:2201.06124. [Bei] A. Beilinson, How to glue perverse sheaves, in K-theory, arithmetic and geometry (Moscow, 1984-1986), Lect. Notes in Math., vol. 1289, Springer, 1987, p. 42–51. [Ber] P. Berthelo...
2022 arXiv
-
[2014]
Cattani, A
[CKS] E. Cattani, A. Kaplan, W. Schmid, L2 and intersection cohomologies for a po- larizable variation of Hodge structure. Invent. Math. 87, 217-252, 1987, [Del] P. Deligne, Equations diff´ erentielles ` a points singuliers r´ eguliers, Lect. Notes in Math. 163, Springer, Berlin,
1987
-
[2015]
Simpson, Higgs bundles and local systems, Inst
[Sim] C. Simpson, Higgs bundles and local systems, Inst. Hautes ´Etudes Sci. Publ. Math. No. 75 (1992), 5-95. [SP] The Stacks Project Authors Stacks Project. https://stacks.math.columbia.edu. [Tsu] T. Tsuji, Poincar´ e duality for logarithmic crystalline cohomology, Compositio...
1992
-
[2018]
xviii+539 pp. [O24] A. Ogus, Crystalline prisms: Reflections and diffractions, present and past, arXiv:2204.06621. [OV] A. Ogus, V. Vologodsky, Nonabelian Hodge theory in characteristic p, Publ. Math. Inst. Hautes ´ etudes Sci. 106 (2007), 1-138. 51 [P1] A. Petrov. Non-decompo...
2007
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.