Constructive Proof of the Hodge Conjecture for K3 Surfaces via Nodal Degenerations

Badre Mounda

REVIEW 5 major objections 5 minor 20 references

Constructive Proof of the Hodge Conjecture for K3 Surfaces via Nodal Degenerations

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Every rational (1,1)-class on a complex K3 surface is algebraic, the paper claims, and an explicit finite-step degeneration algorithm produces the representing divisor.

desk verdict The claimed constructive proof collapses immediately because the central family (3.1) has a smooth central fiber — no nodes, no Picard–Lefschetz, no Clemens–Schmid transport — and the underlying theorem is already known via Lefschetz (1,1). read the letter →

arxiv 2507.18999 v1 pith:OT5VTBKC submitted 2025-07-25 math.AG math.CV

Badre Mounda This is my paper · ORCID

classification math.AGmath.CV MSC 14C3014J2814D0714J32

keywords HodgeconjectureK3surfacesrational(11)-classesnodaldegenerationsPicard–LefschetztheoryClemens–Schmidsequencelimitmixedstructurealgebraiccycles

topics The Hodge Conjecture

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 aims to prove the Hodge conjecture for complex K3 surfaces constructively: starting from an arbitrary rational (1,1)-class, it claims to produce an explicit algebraic divisor in finitely many steps. The route is to degenerate the surface to a quartic with at most ten ordinary double points, read the class as a rational combination of the hyperplane class and exceptional curves on the resolved central fibre, and then use the limit mixed Hodge structure to carry that combination back to the original smooth surface. If the construction is right, it converts a statement whose usual proof is an abstract existence argument into a computational recipe with a uniform bound on the number of nodes needed.

What carries the argument

The load-bearing mechanism is a one-parameter family of quartic K3 surfaces $X_t=\{f_0+t\sum_i\lambda_i g_i=0\}$ together with the resolved normal-crossings central fibre $\widetilde{X}_0$, on which the exceptional curves over the $A_1$-nodes generate, with the hyperplane class $h$, the rational Picard group. The companion tool is the Clemens–Schmid identification $\mathrm{Gr}^W_2 H^2_{\lim}\cong H^2(\widetilde{X}_0)$, which lets a class on the limit mixed Hodge structure be realized on a smooth fibre; Picard–Lefschetz theory fixes the coefficients through the equations $\langle\alpha,\gamma_i\rangle=-2m_i$. This package converts a Hodge class on the smooth surface into a rational combination of curves on the degenerate fibre, and back.

What would settle it

Put $t=0$ in the defining family $X_t=\{f_0+t\sum_i\lambda_i g_i=0\}$: the central fibre is $\{f_0=0\}$. For a smooth quartic $f_0$, no point $p_i$ can satisfy both $f_0(p_i)=0$ and $\nabla f_0(p_i)=0$, so no $A_1$-node appears at the chosen points; checking this for an explicit $f_0$, $p_i$, $g_i$, and $\lambda_i$ directly settles whether the claimed specialization of $\alpha$ exists.

Watch

Extended reading notes

Core claim

The central claim is Theorem 5.2: every rational $(1,1)$-class on a complex K3 surface is algebraic, and the proof is constructive. On a very general quartic K3 with Picard number one, the paper writes the class as $\alpha = a_0h + \sum_j a_j v_j$ in a transcendental root basis, chooses $k = \sum_j |a_j| \le 10$ points in general position, and builds a one-parameter family of quartics. Blowing up the resulting $A_1$-nodes gives a central fibre whose Picard group is generated by the hyperplane class and the exceptional $(-2)$-curves, and Picard–Lefschetz theory fixes the coefficients $m_i$ from the pairings $\langle\alpha,\gamma_i\rangle$. The Clemens–Schmid identification $\mathrm{Gr}^W_2 H^2_{\lim} \cong H^2(\widetilde{X}_0)$ then transports the algebraic combination back to a smooth fibre, giving the representative of $\alpha$. The paper also states a conjectural equivariant extension to $(2,2)$-classes on Calabi–Yau threefolds.

Load-bearing premise

The argument assumes the degenerate central fibre really develops conical double points at the chosen locations, so the given class becomes a combination of the hyperplane class and the exceptional curves; if that nodal specialization fails, the step that carries the class back to a smooth fibre has nothing to act on.

Editorial extensions

If this is right

Every rational $(1,1)$-class on any complex K3 surface would be represented by an explicit $\mathbb{Q}$-divisor, not merely shown to exist by an abstract argument.
For projective K3 surfaces, the Lefschetz $(1,1)$ part of the Hodge conjecture would follow without using the global Torelli theorem or period-map surjectivity.
The construction would be bounded: at most ten nodes are needed for quartics, and the size of the linear system to solve is governed by the $\ell^1$-norm of the class in a transcendental basis.
If the conjectural threefold statement is correct, $G$-invariant rational $(2,2)$-classes on nodal Calabi–Yau threefold degenerations would be algebraic, with exceptional surfaces replacing the $(-2)$-curves.

Reading between the lines

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

Editorial: the cleanest way to test the construction is to choose the points $p_i$ on $f_0$ and perturb $f_0$ itself so that the central fibre develops the nodes, then check numerically whether the Clemens–Schmid lift preserves the class $\alpha$; this is a finite linear-algebra computation.
Editorial: if the nodal families can be made to exist, the method effectively turns the Hodge conjecture for K3 surfaces into a search over quartics with at most ten nodes, which could be explored with computer algebra over $\mathbb{Q}$.
Editorial: the same node-surgery idea might apply to other Hodge-level settings or to families with slightly worse singularities, provided the limit mixed Hodge structure still admits a weight-graded identification with the cohomology of a resolution.

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

Two things you should know. First, the paper's headline theorem is already known: the Lefschetz (1,1) theorem gives every rational (1,1) class on a projective K3 as an algebraic cycle. The only possible contribution is the alleged algorithm. Second, the algorithm is broken from the very first step. The family X_t defined in (3.1) has central fiber {f0 = 0}, which is smooth, so there are no nodes, no vanishing cycles, and no Clemens–Schmid transport to act on. The points p_i are chosen to satisfy g_i(p_i) = 0, not f0(p_i) = 0, so they generally do not even lie on X0. The stress-test note is exactly right about this.

What the paper does well: it is clearly organized, it is honest that the underlying degeneration theory is classical (Friedman–Scattone, Persson–Pinkham), it includes worked examples and a SageMath sketch, and it cites the relevant literature. The broad strategy — specialize a given class to a combination of exceptional curves, then lift back via Clemens–Schmid — is a sensible idea if a nodal degeneration actually existed.

The soft spots are not minor. First, the central construction is internally inconsistent, as described above. Second, the bound k ≤ 10 rests on k = Σ|a_j|, where the a_j are rational coefficients in a fixed basis; for arbitrary α this sum is unbounded, so ten nodes cannot cover all classes. Third, the weight filtration claim W1 = 0 and the identification GrW_2 H^2_lim ≅ H^2(X~0) are asserted without proof and are not obviously true for a multi-node degeneration; the vanishing cycles are not shown to be orthogonal. Fourth, the "transcendental root basis" with v_j^2 = −2 is suspect — the transcendental lattice of a very general quartic generally contains no (-2)-vectors. The CY3 section is explicitly conjectural and Remark 7.9 concedes it needs more work.

In short, the load-bearing construction fails at the first step, so the paper does not contain a salvageable proof. It might serve as a cautionary example of how easy it is to confuse the central fiber of a family with the desired singular fiber, but it is not a serious contribution. My recommendation: desk reject; this does not need referee time.

Referee Report

5 major / 5 minor

Summary. The paper proposes a constructive proof of the Hodge conjecture for rational (1,1)-classes on complex K3 surfaces. The strategy is: given α ∈ H^{1,1}(X,Q), choose at most ten nodes and a one-parameter family of quartic K3s (3.1) whose central fiber acquires those nodes; after blowing up, α specializes to a rational combination of the hyperplane class and exceptional (−2)-curves; the Clemens–Schmid sequence identifies Gr^W_2 of the limit cohomology with H^2 of the resolved central fiber; and the algebraic combination is transported back to the smooth fiber. The paper also sketches a conjectural equivariant extension to (2,2)-classes on Calabi–Yau threefolds.

Significance. If the proof were correct, it would give an explicit finite-step algorithm for a statement already known for projective K3s via the Lefschetz (1,1) theorem; the genuinely new content would be the claimed constructivity and the uniform ten-node bound. The paper makes a serious effort to be algorithmic (pseudo-code, SageMath sketch), which is commendable. However, the central construction contradicts equation (3.1), the weight-filtration identifications are dimensionally impossible, and the k≤10 bound is unbounded for arbitrary α. I do not see a salvageable core: the degeneration around which everything is built does not exist as written, so the main theorem is not established.

major comments (5)

[§3, Eq. (3.1)] The family defined by X_t = {f0 + t Σ λ_i g_i = 0} has central fiber X0 = {f0 = 0}, which is smooth because f0 is a smooth quartic. The points p_i are chosen with g_i(p_i)=0 and ∇g_i(p_i)≠0, but they are not required to satisfy f0(p_i)=0; if f0(p_i)=0, then p_i is a smooth point of X0 since ∇f0(p_i)≠0. Therefore X0 does not acquire the claimed A1 nodes at the p_i, and the blow-up construction, Lemma 3.1, and the specialization formula sp(α)=m0h+Σ m_i[E_i] have no object to act on. This is a load-bearing failure at the first step of the proof.
[§5.2–5.4, Prop. 5.1] The monodromy weight filtration is misidentified. For a degeneration with k>0 nodes, N is nonzero, so any admissible weight filtration must satisfy N(W_i)⊂W_{i-2} and cannot have W1=0 while W3=H^2_lim: N would map H^2_lim into W1=0. Thus the claim Gr^W_2 H^2_lim = ker N = Im(sp) is not justified. Moreover, rank reasons contradict the identification with H^2(\tilde X_0): the local system has rank b2=22 on the smooth fiber, while the Mayer–Vietoris computation in (5.2) gives dim H^2(\tilde X_0)=22+k ≥ 23 for k≥1. No graded piece of H^2_lim has dimension 22+k, so Proposition 5.1 cannot hold as stated.
[§4, Prop. 4.1; §6.1] The bound k≤10 is incompatible with the definition k=Σ|a_j|. For an arbitrary rational (1,1)-class α, the coefficients a_j in any fixed Q-basis of T^{1,1}(X)_Q are unbounded, so k is unbounded. The statement in §6.1 that 'k≤10 is sharp and independent of α' is therefore false. Additionally, the premise that T^{1,1}(X)_Q admits a basis of mutually orthogonal rational vectors of square −2 (the 'transcendental root basis') is not established and is not generally available for a signature (2,17) quadratic space over Q.
[Theorem 5.2, §2.1, §6.4] Theorem 5.2 asserts the statement for every complex K3 surface, but Proposition 4.1 is proved only for a very general quartic K3 with ρ(X)=1, fixed once and for all in Remark 2.2. No argument is given to pass to arbitrary Picard rank or to non-quartic K3 surfaces. The claimed extension is also internally inconsistent: Example 2 in §6.4 says s=0 for a Picard-rank-20 Kummer surface and then uses v5, v13, v17, which do not exist in that case.
[§4.1, §5.5] The final lift from \tilde X_0 to the smooth fiber is not justified. The classes [E_i] are exceptional divisors that are contracted by the specialization and do not deform to curves on the smooth fiber; the specialization map from H^2(X_t) to H^2(\tilde X_0) is not surjective onto the span of the [E_i], and by the dimension count above it cannot be. The argument in §5.5 attempts to show N(ω_t)=0 using W1=0, but W1=0 is false (see the second major comment), so the claim that 'transporting back via sp^{-1} yields an algebraic representative' is a non-sequitur. Step 7 of the algorithm and the worked examples assume, rather than prove, the existence of this lift.

minor comments (5)

[§2.2] The phrase 'in cohomology one has N(v) = ⟨v,γ⟩γ^∨' mixes homology and cohomology conventions; the notation γ^∨ is not defined clearly and should be fixed.
[§4.1 and §6.2] The two pseudo-code listings are nearly identical but differ in wording for the final step ('transport α0 back to X_t' versus 'α0 → α_alg on X_t'); this is only a presentation issue.
[§6.3] The SageMath sketch contains placeholder functions and comments such as 'Dummy helpers (to be replaced by your own code)', so the claimed implementation is not actually provided.
[§6.5] The toy statistics depend on an arbitrary coefficient distribution and on the nonexistent transcendental root basis; they do not support the claimed uniform bound k≤10.
[§5.4] The comparison isomorphism is attributed to '[4, Thm. 3.13] or [5]' without a precise theorem number, and the cited statements are not verifiable from the reference list as written.

Circularity Check

1 steps flagged · score 6.0 of 10

The specialization step is circular: α0 is defined from α by solving pairings and then identified with sp(α), so the algebraic-divisor output is the input class only by construction.

self definitional [Proposition 4.1 (Section 4) and Algorithm §4.1, steps 5–7]
"Hence the linear system ⟨α,γ_i⟩ = −2m_i (1≤i≤k) admits a rational (indeed integral) solution (m_i), determining sp(α) on the central fibre. Lemma 3.1 identifies Pic(X̃_0) with ⟨h,E_1,...,E_k⟩, so sp(α) has the stated form. … Set α0 := m0h + Σ_i m_i[E_i] ∈ Pic(X̃_0) ⊗ Q."

The coefficients m_i are solved from the pairing equalities, but the assertion that sp(α) equals m0h + Σ m_i[E_i] is exactly the conclusion being proved. The pairings determine the coefficients only if sp(α) is already known to lie in the span of h and the exceptional curves with those pairings; that is the algebraicity claim. The algorithm then defines α0 as that same combination and 'lifts back' via Clemens–Schmid, so the output divisor represents α only under the unproved identification sp(α)=α0. No independent computation of sp(α) from the degeneration is supplied; the conclusion is built into the definition of α0.

full rationale

The paper's central constructive claim is that an arbitrary rational (1,1)-class α on a K3 surface can be followed through a nodal degeneration, specialized to an algebraic combination of exceptional curves, and then transported back. The circular point is in Proposition 4.1 and Algorithm §4.1: the 'specialization' sp(α) is never independently computed. Instead, the paper solves a linear system for coefficients m_i from the pairings ⟨α,γ_i⟩, declares that this 'determines sp(α)', and then defines α0 := m0h + Σ m_i[E_i], which is algebraic by construction. The later Clemens–Schmid transport of α0 back to the smooth fiber can only return α if sp(α) already equals α0, which is precisely what was to be shown. Thus the constructive procedure's output is the input class by definition rather than by a derived geometric specialization. The underlying Hodge conjecture statement for K3 surfaces is independently true via the Lefschetz (1,1) theorem, so the circularity affects the claimed construction, not the truth of the theorem. There is no problematic self-citation chain: the cited Clemens–Schmid, Steenbrink, and Picard–Lefschetz results are external classical results. Separate mathematical difficulties exist (for example, the central fiber in (3.1) is asserted to acquire prescribed nodes although {f0=0} is smooth and the chosen points p_i need not lie on it, and k=Σ|a_j| is unbounded for arbitrary α), but these are correctness issues rather than circularity and are not scored here.

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

No numeric parameters are fitted to data, so the free-parameter list is empty. The paper relies on standard Hodge-theoretic tools, Picard-Lefschetz, Clemens-Schmid, and Steenbrink, and on domain assumptions about very general quartic K3 surfaces. The decisive assumption is ad hoc: that a family of the displayed form can have a nodal central fiber with prescribed vanishing cycles. The paper's own equation makes the central fiber smooth. The paper introduces no new particles, forces, or dimensions, so the invented-entities list is empty.

assumptions (5)

standard math Picard-Lefschetz monodromy for an A1 node: N(v)=langle v,gamma rangle gamma^vee with gamma^2 = -2 and N^2 = 0.
Used in Lemma 2.3 and throughout Sections 4 and 5. This is standard for ordinary double points.
standard math Clemens-Schmid sequence and the identification Gr^W_2 H^2_lim is isomorphic to H^2(X~0) are applicable to the constructed family.
Invoked in Proposition 5.1 and Theorem 5.2. The identification is standard for semistable degenerations, but the paper does not first establish that its family is semistable with the claimed central fiber.
domain assumption A fixed transcendental root basis {v_j} of T^{1,1}(X)_Q exists with v_j^2 = -2 and with k = sum |a_j| at most 10 for every rational class alpha.
Remark 2.2 fixes such a basis and Sections 4 and 6.1 use k = sum |a_j| as the node count. For a fixed basis this sum is unbounded over all rational classes, so the bound is not guaranteed.
ad hoc to paper The family (3.1) has a central fiber that acquires exactly k A1 nodes at the chosen points p_i, and the pairings langle alpha, gamma_i rangle determine the specialization sp(alpha) = m0 h + sum m_i [E_i].
This is the load-bearing step in Proposition 4.1. It contradicts equation (3.1), where the central fiber is the smooth quartic f0 = 0, and no repair is supplied.
domain assumption The monodromy weight filtration on H^2_lim satisfies W1 = 0 and Gr^W_2 = ker N for the nodal degeneration.
Stated in Section 5.2. For nonzero vanishing cycles, Im N should contribute to W1, so this is at least unjustified and undermines the comparison in Proposition 5.1.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Constructive Proof of the Hodge Conjecture for K3 Surfaces via Nodal Degenerations." pith.science (2026). https://pith.science/paper/OT5VTBKC

@misc{pith2026250718999,
  author       = {Pith},
  title        = {Pith review of: Constructive Proof of the Hodge Conjecture for K3 Surfaces via Nodal Degenerations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OT5VTBKC}},
  note         = {Machine review of arXiv:2507.18999}
}

abstract

We give a constructive proof of the Hodge conjecture for complex $K3$ surfaces that does not rely on Torelli-type results. Starting with an arbitrary rational $(1,1)$-class $\alpha\in H^{1,1}(X,\mathbb{Q})$, we algorithmically build a one-parameter family of quartic $K3$'s acquiring at most ten $A_1$-nodes. On the central fibre $\widetilde{X}_0$, the class $\alpha$ specializes to a $\mathbb{Q}$-linear combination of the hyperplane class and the exceptional $(-2)$-curves coming from the blow-ups of the nodes. Using the Clemens--Schmid sequence together with Picard--Lefschetz theory, we identify $\Gr^W_2 H^2_{\lim}\cong H^2(\widetilde{X}_0)$ and transport this combination back to the original smooth surface as an algebraic divisor. This yields an explicit, finite-step procedure that realizes any rational $(1,1)$-class by an algebraic cycle. We also formulate an equivariant extension for $(2,2)$-classes on Calabi--Yau threefolds, indicating how the same strategy might apply in higher dimension.

Figures

Figures reproduced from arXiv: 2507.18999 by the authors.

**Figure 1.** Clemens–Schmid identification of GrW 2 H2 lim with H2 (Xe0). 5.1 Nearby and vanishing cycles Let ψ := ψtQ and ϕ := ϕtQ be the nearby and vanishing cycle complexes (Deligne). They fit into a distinguished triangle i ∗Rj∗Q −→ ψ −→ ϕ [+1] −−→ . Taking hypercohomology yields the long exact sequence · · · → H2 (X0, Q) sp −→ H2 lim N −→ H2 lim(−1) δ −→ H3 (X0, Q) → · · · , (5.1) where H2 lim := H2 (X0, ψ) is the limit MHS… view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 19 canonical work pages

[1]

Friedman, F

R. Friedman, F. Scattone,Type II degenerations ofK3 surfaces, Invent. Math.83 (1986), 1–39

work page 1986
[2]

Persson, H

U. Persson, H. Pinkham,Degeneration of surfaces with nontrivial canonical bundle, Ann. of Math. 113 (1981), 45–66

work page 1981
[3]

Schmid, Variation of Hodge structure: the singularities of the period mapping, Invent

W. Schmid, Variation of Hodge structure: the singularities of the period mapping, Invent. Math. 22 (1973), 211–319

1973
[4]

Steenbrink,Limits of Hodge structures, Invent

J. Steenbrink,Limits of Hodge structures, Invent. Math.31 (1976), 229–257

work page 1976
[5]

C. H. Clemens,Degeneration of Kähler manifolds, Duke Math. J.44 (1977), 215–290

work page 1977
[6]

Friedman, Simultaneous resolution of threefold double points, Math

R. Friedman, Simultaneous resolution of threefold double points, Math. Ann.274 (1986), 671–689

work page 1986
[7]

M. Kerr, G. Pearlstein, Polarised limiting mixed Hodge structures and their asymptotics, Compos. Math.159 (2023), 1243–1275

work page 2023
[8]

Arbeitman, C

M. Arbeitman, C. Schnell,Singularities of normal functions and extensions of variations of Hodge structure, J. Eur. Math. Soc.26 (2024), 139–178

work page 2024

Show all 20 references

[9]

Bakker, Y

B. Bakker, Y. Brunebarbe, J. Tsimerman,Tame topology of arithmetic quotients and alge- braization of period maps, Ann. of Math.195 (2022), 369–403

2022
[10]

Charles, C

F. Charles, C. Schnell,A bound on the dimension of Hodge loci, Duke Math. J.171 (2022), 1679–1701

2022
[11]

Huybrechts,Lectures onK3 Surfaces, Cambridge Univ

D. Huybrechts,Lectures onK3 Surfaces, Cambridge Univ. Press, 2016

2016
[12]

Voisin,Hodge Theory and Complex Algebraic Geometry I, Cambridge Univ

C. Voisin,Hodge Theory and Complex Algebraic Geometry I, Cambridge Univ. Press, 2002

2002
[13]

Voisin,Hodge Theory and Complex Algebraic Geometry II, Cambridge Univ

C. Voisin,Hodge Theory and Complex Algebraic Geometry II, Cambridge Univ. Press, 2003

2003
[14]

K. Kato, S. Usui,Classifying spaces of degenerating polarized Hodge structures, Annals of Math. Studies 169, Princeton Univ. Press, 2009

2009
[15]

S. Cynk, D. van Straten, Small resolutions and non-liftable Calabi–Yau threefolds , Manuscripta Math.103 (2000), 325–332

2000
[16]

Reid,Minimal models of canonical 3-folds, Adv

M. Reid,Minimal models of canonical 3-folds, Adv. Stud. Pure Math.1 (1983), 131–180

1983
[17]

Candelas, X

P. Candelas, X. C. de la Ossa, P. S. Green, L. Parkes,A pair of Calabi–Yau manifolds as an exactly soluble superconformal theory, Nucl. Phys. B359 (1991), 21–74

1991
[18]

P. S. Aspinwall, D. R. Morrison,Topological field theory and rational curves, Comm. Math. Phys. 151 (1993), 245–262

1993
[19]

Gross, P

M. Gross, P. M. H. Wilson,Large complex structure limits ofK3 surfaces, J. Differential Geom. 55 (2000), 475–546

2000
[20]

Dwork,On the zeta function of a hypersurface, Inst

B. Dwork,On the zeta function of a hypersurface, Inst. Hautes Études Sci. Publ. Math.12 (1962), 5–68. 14

1962

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Friedman, F

R. Friedman, F. Scattone,Type II degenerations ofK3 surfaces, Invent. Math.83 (1986), 1–39

work page 1986

[2] [2]

Persson, H

U. Persson, H. Pinkham,Degeneration of surfaces with nontrivial canonical bundle, Ann. of Math. 113 (1981), 45–66

work page 1981

[3] [3]

Schmid, Variation of Hodge structure: the singularities of the period mapping, Invent

W. Schmid, Variation of Hodge structure: the singularities of the period mapping, Invent. Math. 22 (1973), 211–319

1973

[4] [4]

Steenbrink,Limits of Hodge structures, Invent

J. Steenbrink,Limits of Hodge structures, Invent. Math.31 (1976), 229–257

work page 1976

[5] [5]

C. H. Clemens,Degeneration of Kähler manifolds, Duke Math. J.44 (1977), 215–290

work page 1977

[6] [6]

Friedman, Simultaneous resolution of threefold double points, Math

R. Friedman, Simultaneous resolution of threefold double points, Math. Ann.274 (1986), 671–689

work page 1986

[7] [7]

M. Kerr, G. Pearlstein, Polarised limiting mixed Hodge structures and their asymptotics, Compos. Math.159 (2023), 1243–1275

work page 2023

[8] [8]

Arbeitman, C

M. Arbeitman, C. Schnell,Singularities of normal functions and extensions of variations of Hodge structure, J. Eur. Math. Soc.26 (2024), 139–178

work page 2024