P=W for Lagrangian fibrations and degenerations of hyper-K\"ahler manifolds

Andrew Harder; Junliang Shen; Qizheng Yin; Zhiyuan Li

P=W for Lagrangian fibrations and degenerations of hyper-K\"ahler manifolds

T0 review · 0 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read For any Lagrangian fibration of a projective hyper-Kähler manifold, the perverse filtration coincides with the monodromy weight filtration of a type III degeneration.

desk verdict A genuinely new P=W theorem for Lagrangian fibrations, proved by cleanly assembling Shen–Yin's splitting with Soldatenkov's degenerations; the only soft spot is the terse handoff to Soldatenkov's hypotheses. read the letter →

arxiv 1908.07510 v1 pith:43I6SGJ5 submitted 2019-08-20 math.AG math.DG

Andrew Harder , Zhiyuan Li , Junliang Shen , Qizheng Yin This is my paper · ORCID

classification math.AGmath.DG MSC 14D0714D0614J42

keywords P=Wconjectureperversefiltrationmonodromyweighthyper-KählermanifoldsLagrangianfibrationtypeIIIdegenerationlimitingmixedHodgestructurehard-Lefschetzoperatoralgebra

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper establishes a $P=W$ statement for projective hyper-Kähler manifolds that admit a holomorphic Lagrangian fibration: the perverse filtration on the cohomology of the fibration equals, in even-indexed degrees, the monodromy weight filtration of a suitably chosen maximally unipotent (type III) degeneration of the same manifold. The authors construct such a degeneration for every Lagrangian fibration, not just for K3 surfaces, thereby proving the general case of a conjecture that had previously been settled only there. This matters because it moves the perverse filtration, which is defined from the geometry of the fibration, into the theory of limiting mixed Hodge structures, and it explains the known 'Perverse = Hodge' dimension equalities and the multiplicativity of the perverse filtration.

What carries the argument

The argument is carried by the Lie algebra of cohomology operators generated by all hard-Lefschetz cup-product $\mathfrak{sl}_2$-triples, together with three classes in $H^2(M,\mathbb{Q})$: a relative ample class $\eta$, the pullback $\beta$ of an ample class on the base, and a class $\rho$ of positive square orthogonal to both. The first two provide an $\mathfrak{sl}_2\times\mathfrak{sl}_2$-action whose weights split the perverse filtration; adjoining $\rho$ promotes this to an $\mathfrak{so}(5)$-action. The nilpotent operator $N_{\beta,\rho}=[L_\beta,\Lambda_\rho]$ satisfies $N^3=0$ and is identified with the logarithmic monodromy of a type III degeneration supplied by a cited construction theorem. Representation theory of $\mathfrak{so}(5)$ then shows that the weight decomposition for $N_{\beta,\rho}$ is a rotation of the perverse weight decomposition, producing the equality of filtrations.

What would settle it

Take an explicit Lagrangian fibration on a projective hyper-Kähler manifold (for instance from moduli of sheaves on a K3 surface) and check, in the Beauville–Bogomolov–Fujiki lattice, whether there is an integral class $h$ with $q(h)>0$ and $q(h,\beta)=q(h,\rho)=0$; if no such $h$ exists, the prescribed-monodromy degeneration used in the proof cannot be built, and the claimed identification would have to be verified by another construction.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3: given any holomorphic Lagrangian fibration $\pi: M \to B$ with $M$ a projective hyper-Kähler manifold, there exists a projective type III degeneration $f:\mathcal{M}\to\Delta$ (maximally unipotent logarithmic monodromy, $N^2\neq 0$, $N^3=0$) whose general fibres are deformation equivalent to $M$, together with an identification of cohomology rings $H^*(M,\mathbb{Q}) = H^*_{\lim}(\mathbb{Q})$ under which $P_k H^*(M,\mathbb{Q}) = W_{2k}H^*_{\lim}(\mathbb{Q}) = W_{2k+1}H^*_{\lim}(\mathbb{Q})$ for every $k$. In other words, the perverse filtration of the fibration is exactly the evenly indexed monodromy weight filtration of a canonical degeneration. Because the identification is an isomorphism of rings, the perverse filtration inherits multiplicativity under cup product from the monodromy weight filtration.

Load-bearing premise

The proof depends on a cited theorem that builds a type III degeneration with prescribed logarithmic monodromy, and that theorem needs an integral second-cohomology class $h$ of positive square orthogonal to the two isotropic classes; if no such class exists, the degeneration is not produced and the equality of filtrations is not established by this argument.

Editorial extensions

If this is right

The perverse filtration is multiplicative under cup product, since the monodromy weight filtration is multiplicative and the identification preserves the ring structure.
The 'Perverse = Hodge' dimension identities $\dim \mathrm{Gr}^P_i H^{i+j} = \dim \mathrm{Gr}^F_i H^{i+j}$ follow from the main equality, so perverse-filtration dimensions of Lagrangian fibrations are read off from ordinary Hodge numbers.
The conjecture for Lagrangian fibrations, previously known for K3 surfaces, is now established in all dimensions for every projective hyper-Kähler manifold carrying such a fibration.
The theorem provides a compact geometric analogue of the $P=W$ conjecture from non-abelian Hodge theory, with the monodromy weight filtration playing the role of the weight filtration on character varieties.
The identity $W_{2k}=W_{2k+1}$ forces the limiting mixed Hodge structure of the constructed degeneration to have only even weights, a structural constraint on any type III degeneration realizing a Lagrangian fibration.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same $\mathfrak{so}(5)$ bookkeeping should work for any triple of classes with the same orthogonality and positivity relations, so this version of $P=W$ is probably a property of the hard-Lefschetz operator algebra rather than of the particular fibration geometry.
The proof suggests a recipe for testing other settings: find an integral class of positive square orthogonal to two isotropic classes, build the corresponding nilpotent orbit, and compare the two weight decompositions; this is a finite lattice-and-representation-theory check in each example.
If a concrete example fails to have such an integral class, the construction used here cannot produce the degeneration, although the equality of filtrations might still hold through a degeneration with different monodromy; the paper does not address that possibility.
The methods should extend to non-projective compact hyper-Kähler manifolds or fibrations over non-projective bases as long as the period-map construction of the degeneration remains available.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper proves a real open case of P=W: for any Lagrangian fibration of a projective hyper-Kähler manifold, the perverse filtration coincides with the monodromy weight filtration of an explicit type III degeneration. Before this, that identity was known only for K3 surfaces, so the novelty is substantial. The paper is short and carefully assembled from two main ingredients: the canonical splitting of the perverse filtration from Shen–Yin [9], and Soldatenkov's construction of type III degenerations [12]. It also recovers Perverse=Hodge and multiplicativity of the perverse filtration. The authors are transparent that their proof is not logically independent of [9] because two of them are co-authors; I take that as honesty rather than a flaw.

I checked the representation-theoretic step in Section 11. The matching of the perverse decomposition with the monodromy weight decomposition via the so(5)-action is explicit and correct. The conclusion P_k = W_{2k} = W_{2k+1} follows once the degeneration with monodromy N_{β,ρ} exists.

The one soft spot is the construction of that degeneration. Section 10 delegates to [12, Theorem 4.6] and only verifies the existence of an integral class h with q(h)>0 and q(h,β)=q(h,ρ)=0. The signature argument indeed produces such h. The stress-test concern is whether [12] additionally requires h to be a polarization of type (1,1) at the relevant period point, and whether the pair (N_{β,ρ},x) is an admissible nilpotent orbit. The authors write "from the proof of [12, Theorem 4.6]" these conditions suffice, and N h = 0 follows from the orthogonality conditions, but they do not spell out the remaining verifications. If [12]'s hypotheses are exactly as stated, the proof is complete. If a polarization condition is hidden there, the two-line argument is thin. I cannot resolve this from the text alone; it is precisely the kind of thing a referee should check. This is a minor-to-moderate concern, not a load-bearing flaw.

The citation pattern is appropriate. The reference to [9] is central, and the overlap is disclosed. The reference to [12] is the right tool. No circularity: the paper does not assume the P=W equality as an input.

Who benefits: anyone working on hyper-Kähler geometry, perverse filtrations, or degenerations. It is a short note that settles a conjecture and deserves serious peer review. My recommendation: send it out with a referee who knows Soldatenkov's construction, and ask for a one-paragraph clarification of the hypotheses in [12]. I would accept contingent on that check.

Referee Report

0 major / 4 minor

Summary. The paper proves a P=W theorem for Lagrangian fibrations of projective hyper-Kähler manifolds. For any Lagrangian fibration π:M→B, it constructs a type III projective degeneration f:M→Δ whose general fiber is deformation equivalent to M and whose monodromy weight filtration satisfies P_kH^*(M,Q) = W_{2k}H^*_{lim}(Q) = W_{2k+1}H^*_{lim}(Q) under the natural identification H^*(M,Q)≅H^*_{lim}(Q). The proof combines the sl2×sl2 splitting of the perverse filtration from [9] with the Looijenga–Lunts–Verbitsky algebra, constructs a nilpotent operator N_{β,ρ} from a relative ample class β and an auxiliary positive class ρ, realizes this operator as logarithmic monodromy using Soldatenkov's degeneration theorem [12], and then matches the two weight decompositions through the representation theory of so(5,C). The paper also derives applications to the 'Perverse = Hodge' dimensional equality and to the multiplicativity of the perverse filtration.

Significance. If the cited degeneration theorem is applied correctly, this is a clean and conceptually satisfying proof of the expected P=W phenomenon for Lagrangian fibrations of compact hyper-Kähler manifolds. The argument is elegant and concise, and the explicit so(5,C) matching in Section 11 is a particularly nice part of the paper. The authors are careful to credit prior work, and the honest remark in Section 5 that the argument is not logically independent of [9] is appropriate. The theorem also gives a unified explanation of the previously known 'Perverse = Hodge' equality and the multiplicativity of the perverse filtration.

minor comments (4)

[Section 10] The proof of Theorem 3 delegates the existence of the degeneration to [12, Theorem 4.6], but the manuscript does not state the precise hypotheses of that theorem. Since this is the only step that connects the algebraically defined N_{β,ρ} to an actual degeneration, I recommend quoting the full statement of [12, Theorem 4.6] and explicitly verifying all of its hypotheses, including the role of the integral class h and any restrictions on b2(M).
[Section 10] The existence of an integral h with q(h)>0 and q(h,β)=q(h,ρ)=0 is stated in two sentences. In the case b2=4, the orthogonal complement of Vρ has signature (1,0), so a brief comment on why this positive-definite rank-one rational subspace contains an integral class of positive square would make the argument fully transparent.
[Section 11] The identification of the so(5,C) weight decomposition V^{i,j} with the decomposition P^{i,j}_C is asserted without derivation. A short explanation using H = H_η + H_β and −√−1 K_{23} = H_η − H_β would improve readability and make the matching of the two filtrations easier to follow.
[References] The reference to Soldatenkov [12] is listed as 'to appear'; if a final journal and year are available, they should be added.

Circularity Check

1 steps flagged · score 2.0 of 10

No hidden circularity: P=W is derived, not assumed, from [9]'s canonical splitting and Soldatenkov's degeneration theorem; the only flagged issue is an admitted logical dependence on the authors' own [9].

other [Section 5, after Theorem 3 and before the proof]
"As the proof of Theorem 3 uses the same ingredients as in [9], the new way of deriving these results is not logically independent."

This is the paper's own limitation note rather than a circular step. It acknowledges that Theorem 3's proof reuses the canonical splitting and LLV computations from the authors' earlier paper [9], so recovering [9, Theorem 0.2] from Theorem 3 does not give a logically independent proof of that earlier result. The note does not assert that [9] or [12] already contains the equality P_k = W_{2k}; the equality is obtained in Section 11 by matching the perverse decomposition (3) with the monodromy weight decomposition (6)-(7). The self-citation is therefore load-bearing but not question-begging.

full rationale

The central equality (1) is not an input to the proof. Section 11 computes P_k = ⊕_{i≤k} P^{i,j}_C and W_{2k} = ⊕_{d-m≤2k} W^d_m, then shows the same V^{i,j} decomposition is induced by the so(5,C)-action from the LLV algebra, so the two filtrations coincide by construction of the degeneration. The two external pillars are [9, Prop. 1.1 and Cor. 2.5] (canonical splitting of the perverse filtration via sl2 × sl2 / so(5) actions) and [12, Thm. 4.6] (existence of a type III degeneration with prescribed logarithmic monodromy N_{β,ρ}). Neither of these states the P=W equality; [9] concerns the perverse/Hodge dimension equality, while [12] is a degeneration-construction theorem by an author not on this paper. The self-citation to [9] is prominent and load-bearing, and Section 5 candidly says the resulting derivation of [9]'s results is not logically independent; however this is a dependence on prior work, not a reduction of the theorem to its own conclusion. The only other risk flagged is external-theorem applicability: the h produced by the signature argument in Section 10 is checked only for q(h)>0 and orthogonality to β and ρ, while [12, Thm. 4.6] may require h to be a (1,1) polarization and the nilpotent orbit to be admissible before global Torelli applies. That is a correctness concern about hypotheses of an external theorem, not a circularity concern. Under the stated rules, non-circular self-citation and unverified external hypotheses do not raise the circularity score; accordingly the score is kept at 2.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper relies on two substantial external theorems: the Shen-Yin splitting of the perverse filtration and Soldatenkov's degeneration construction. Both are cited as black boxes, and the paper explicitly states that its proof is not logically independent of [9]. No free parameters or invented entities are introduced.

assumptions (5)

domain assumption The Looijenga-Lunts-Verbitsky algebra g acts on H^*(M,Q) with g ≅ so(\tilde{H}^2(M,Q), \tilde{q}_M), and the sl2-triples associated with η and β split the perverse filtration as in [9, Proposition 1.1].
Invoked in Sections 7 and 11 as the backbone of the identification of the perverse filtration with a weight decomposition. It is cited from [9] (Shen-Yin) and [3], not reproven here.
domain assumption Soldatenkov's theorem [12, Theorem 4.6] constructs a type III projective degeneration of hyper-Kähler manifolds with prescribed logarithmic monodromy N_{β,ρ}, provided an integral class h with q(h)>0 and q(h,β)=q(h,ρ)=0 exists.
Used in Section 10 to produce the degeneration f: M → Δ on which the monodromy weight filtration is computed. The proof does not reproduce Soldatenkov's construction.
domain assumption The Beauville-Bogomolov-Fujiki form q_M has signature (3, b2(M)-3), with q_M(β)=0 for β a pullback of an ample class on the base of the Lagrangian fibration.
Standard structure theorem for irreducible holomorphic symplectic varieties. Used in Sections 6, 8, and 10 to choose η, ρ, and the integral class h.
standard math The representation theory of so(5,C), specifically the weight decomposition with respect to the Cartan subalgebra generated by H and -√-1 K_{23}, and the sl2-triple (L_N, H_N, Λ_N) defined in Section 11.
The matching of the indices (i,j) with the perverse and monodromy weight filtrations in Section 11 is a direct calculation in so(5,C)-representations.
domain assumption The limiting mixed Hodge structure of a type III degeneration is Hodge-Tate ([12, Theorem 3.8]).
Used in Section 5 to derive the Perverse=Hodge dimension equality from Theorem 3. Not needed for the main proof but used in the explanatory remarks.

how reviews work

0 comments

Cite this review

Pith. "Pith review of P=W for Lagrangian fibrations and degenerations of hyper-K\"ahler manifolds." pith.science (2026). https://pith.science/paper/43I6SGJ5

@misc{pith2026190807510,
  author       = {Pith},
  title        = {Pith review of: P=W for Lagrangian fibrations and degenerations of hyper-K\"ahler manifolds},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/43I6SGJ5}},
  note         = {Machine review of arXiv:1908.07510}
}

read the original abstract

We identify the perverse filtration of a Lagrangian fibration with the monodromy weight filtration of a maximally unipotent degeneration of compact hyper-K\"ahler manifolds.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 14 canonical work pages

[9]

Topology of Lagrangian fibrations and Hodge theory of hyper-K\"ahler manifolds

J. Shen and Q. Yin, Topology of Lagrangian ﬁbrations and Hodge theory of hyper- K¨ ahler manifolds,arXiv: 1812.10673. 6 Andrew Harder, Zhiyuan Li, Junliang Shen, and Qizheng Yin

work page Pith review arXiv
[12]

Soldatenkov, Limit mixed Hodge structures of hyperk¨ ahler manifolds, Mosc

A. Soldatenkov, Limit mixed Hodge structures of hyperk¨ ahler manifolds, Mosc. Math. J., to appear

work page
[1]

M. A. de Cataldo , T. Hausel , and L. Migliorini , Topology of Hitchin systems and Hodge theory of character varieties: the case A1, Ann. of Math. (2) 175 (2012), no. 3, 1329–1407

work page 2012
[2]

Fujisaw a, Polarizations on limiting mixed Hodge structures, J

T. Fujisaw a, Polarizations on limiting mixed Hodge structures, J. Singul. 8 (2014), 146–193

work page 2014
[3]

Green , Y.-J

M. Green , Y.-J. Kim , R. Laza , and C. Robles , The LL V decomposition of hyper-K¨ ahler cohomology, arXiv:1906.03432

arXiv 1906
[4]

Torus fibers and the weight filtration

A. Harder , Torus ﬁbers and the weight ﬁltration, arXiv:1908.05110

work page Pith review arXiv 1908
[5]

Koll ´ar, R

J. Koll ´ar, R. Laza , G. Sacc `a, and C. Voisin , Remarks on degenerations of hyper-K¨ ahler manifolds, Ann. Inst. Fourier (Grenoble) 68 (2018), no. 7, 2837–2882

work page 2018
[6]

Kurnosov , A

N. Kurnosov , A. Soldatenkov , and M. Verbitsky , Kuga–Satake construction and cohomology of hyperk¨ ahler manifolds,Adv. Math. 351 (2019), 275–295

work page 2019

Show all 15 references

[7]

Looijenga and V

E. Looijenga and V. A. Lunts , A Lie algebra attached to a projective variety, Invent. Math. 129 (1997), no. 2, 361–412

1997
[8]

Nagai , On monodromies of a degeneration of irreducible symplectic K¨ ahler manifolds, Math

Y. Nagai , On monodromies of a degeneration of irreducible symplectic K¨ ahler manifolds, Math. Z. 258 (2008), no. 2, 407–426

2008
[10]

Shende , The weights of the tautological classes of character variet ies, Int

V. Shende , The weights of the tautological classes of character variet ies, Int. Math. Res. Not. IMRN 2017, no. 22, 6832–6840

2017
[11]

C. T. Simpson , Higgs bundles and local systems, Inst. Hautes ´Etudes Sci. Publ. Math. No. 75 (1992), 5–95

1992
[13]

Steenbrink , Limits of Hodge structures, Invent

J. Steenbrink , Limits of Hodge structures, Invent. Math. 31 (1975/76), no. 3, 229–257

1975
[14]

M. S. Verbitski ˘ı, Action of the Lie algebra of SO(5) on the cohomology of a hyper-K¨ ahler manifold, Funct. Anal. Appl. 24 (1990), no. 3, 229–230

1990
[15]

Verbitsky, Cohomology of compact hyper-K¨ ahler manifolds and its applications, Geom

M. Verbitsky, Cohomology of compact hyper-K¨ ahler manifolds and its applications, Geom. Funct. Anal. 6 (1996), no. 4, 601–611. Lehigh University, Department of Mathematics E-mail address : anh318@lehigh.edu Fudan University, Shanghai Center for Mathematical Scienc es E-mail a...

1996

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [9]

Topology of Lagrangian fibrations and Hodge theory of hyper-K\"ahler manifolds

J. Shen and Q. Yin, Topology of Lagrangian ﬁbrations and Hodge theory of hyper- K¨ ahler manifolds,arXiv: 1812.10673. 6 Andrew Harder, Zhiyuan Li, Junliang Shen, and Qizheng Yin

work page Pith review arXiv

[2] [12]

Soldatenkov, Limit mixed Hodge structures of hyperk¨ ahler manifolds, Mosc

A. Soldatenkov, Limit mixed Hodge structures of hyperk¨ ahler manifolds, Mosc. Math. J., to appear

work page

[3] [1]

M. A. de Cataldo , T. Hausel , and L. Migliorini , Topology of Hitchin systems and Hodge theory of character varieties: the case A1, Ann. of Math. (2) 175 (2012), no. 3, 1329–1407

work page 2012

[4] [2]

Fujisaw a, Polarizations on limiting mixed Hodge structures, J

T. Fujisaw a, Polarizations on limiting mixed Hodge structures, J. Singul. 8 (2014), 146–193

work page 2014

[5] [3]

Green , Y.-J

M. Green , Y.-J. Kim , R. Laza , and C. Robles , The LL V decomposition of hyper-K¨ ahler cohomology, arXiv:1906.03432

arXiv 1906

[6] [4]

Torus fibers and the weight filtration

A. Harder , Torus ﬁbers and the weight ﬁltration, arXiv:1908.05110

work page Pith review arXiv 1908

[7] [5]

Koll ´ar, R

J. Koll ´ar, R. Laza , G. Sacc `a, and C. Voisin , Remarks on degenerations of hyper-K¨ ahler manifolds, Ann. Inst. Fourier (Grenoble) 68 (2018), no. 7, 2837–2882

work page 2018

[8] [6]

Kurnosov , A

N. Kurnosov , A. Soldatenkov , and M. Verbitsky , Kuga–Satake construction and cohomology of hyperk¨ ahler manifolds,Adv. Math. 351 (2019), 275–295

work page 2019