The period-index problem for hyper-K\"ahler varieties via hyperholomorphic bundles

Davesh Maulik; James Hotchkiss; Junliang Shen; Qizheng Yin; Ruxuan Zhang

REVIEW 2 major objections 3 minor 1 cited by

The period-index problem for hyper-K\"ahler varieties via hyperholomorphic bundles

T0 review · 2 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read For every hyper-Kähler variety of $K3^{[n]}$-type with a primitive polarization, the index of a Brauer class divides its period raised to the dimension; with Picard rank at least two, most classes satisfy the half-dimension bound.

desk verdict New period-index bounds for K3[n]-type varieties via Markman's hyperholomorphic bundles; the main theorems are sound and the one flagged gap in Proposition 1.3 is fillable. read the letter →

arxiv 2502.09774 v2 pith:NE36GSWD submitted 2025-02-13 math.AG

James Hotchkiss , Davesh Maulik , Junliang Shen , Qizheng Yin , Ruxuan Zhang This is my paper · ORCID

classification math.AG MSC 14F2214J2814J4216K50

keywords period-indexproblemBrauergrouphyper-KählervarietiesK3^[n]-typeprojectivelyhyperholomorphicbundlestwistedvectorBeauville–Bogomolov–FujikiformPicardrank

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 period–index problem for a hyper-Kähler variety $X$ of $K3^{[n]}$-type asks how large an exponent $e$ must be so that, for every Brauer class $\alpha$, the index $\operatorname{ind}(\alpha)$ divides the period $\operatorname{per}(\alpha)^e$; here $\operatorname{per}(\alpha)$ is the order of $\alpha$ in the Brauer group and $\operatorname{ind}(\alpha)$ is the smallest degree of a central division algebra representing it. The paper proves that $e=\dim(X)$ works for every such $X$ carrying a primitive polarization, provided $\operatorname{per}(\alpha)$ avoids the primes dividing $n!\,h\,I_X$. When the Picard rank is at least two, it proves the stronger bound $e=\dim(X)/2$ for all non-special classes whose period avoids an explicit constant $N_X$. The proof is geometric: it constructs an $\alpha$-twisted vector bundle of controlled rank on $X$, and the index of $\alpha$ must divide that rank. The significance is that these are the best bounds currently known for these varieties and they confirm, for most rank-at-least-two cases, the conjectured half-dimension bound.

What carries the argument

The engine is a projectively hyperholomorphic vector bundle $U^{[n]}$ of rank $n!\,r^n$ on a product $M^{[n]}\times S^{[n]}$ of two $K3^{[n]}$-type varieties, built from a moduli space of stable bundles on a $K3$ surface with a primitive isotropic vector of the required numerical type. Projectively hyperholomorphic means that its projectivization deforms along diagonal twistor paths in the component of the moduli space of marked varieties, so restricting the deformed bundle to a fiber gives an $\alpha$-twisted vector bundle of rank $n!\,r^n$ on any variety in that component. The lattice-theoretic part chooses $r$, $m$, and a class $L_u = 2\ell A + 2B + uD$ whose Beauville–Bogomolov–Fujiki square is divisible by $\ell$ and whose divisibility is $1$ or $2$; a lattice-theoretic classification criterion then produces the $K3$ surface and parallel transport realizing the required twisted bundle, with $r = 4\ell^2$ or $r = 4\ell$ to control the final exponent.

What would settle it

Search for a $K3^{[n]}$-type variety $X$ with a primitive polarization and a Brauer class $\alpha$ with $\operatorname{per}(\alpha)$ coprime to $n!\,h\,I_X$ for which $\operatorname{ind}(\alpha)$ does not divide $\operatorname{per}(\alpha)^{\dim X}$; Theorem 0.2 says no such pair exists. A concrete place to look is the Fano variety of a smooth cubic fourfold, where the theorem gives $\operatorname{ind}(\alpha)\mid \operatorname{per}(\alpha)^4$: a class of period $\ell$ whose index is divisible by $\ell^5$ would refute it.

Watch

Extended reading notes

Core claim

The paper's central claim is that the period–index exponent for $K3^{[n]}$-type varieties is much smaller than the general period–index conjecture predicts. Concretely, Theorem 0.2 asserts that if $X$ is of $K3^{[n]}$-type and admits a primitive polarization of degree $2h$, then for every Brauer class $\alpha$ with $\operatorname{per}(\alpha)$ coprime to $n!\,h\,I_X$ one has $\operatorname{ind}(\alpha)\mid \operatorname{per}(\alpha)^{\dim X}$. Theorem 0.5 asserts that when the Picard rank is at least two, there is an explicit integer $N_X$ such that $\operatorname{ind}(\alpha)\mid \operatorname{per}(\alpha)^{\dim X/2}$ for every non-special $\alpha$ with $\operatorname{per}(\alpha)$ coprime to $N_X$; here 'non-special' means the Beauville–Bogomolov–Fujiki norm $q(B)$ of the B-field representative is coprime to the period. The proof realizes each such class by an $\alpha$-twisted vector bundle of rank dividing $n!\,(4\,\operatorname{per}(\alpha)^2)^n$ in the first case and $n!\,(4\,\operatorname{per}(\alpha))^n$ in the second, and the elementary Lemma 1.1 converts this rank divisibility into the index bound.

Load-bearing premise

The proof depends on an earlier theorem that a certain vector bundle over Hilbert schemes of $K3$ surfaces is projectively hyperholomorphic and deforms along diagonal twistor paths; if that theorem failed on some $K3^{[n]}$-type variety, the twisted bundles used to bound the index would not exist.

Editorial extensions

If this is right

For every $K3^{[n]}$-type variety with a primitive polarization, the period–index exponent $e(X)$ is at most $\dim(X)$, which is the best bound currently known for Picard rank one.
For Picard rank at least two, the conjectured half-dimension bound holds for all non-special Brauer classes whose period avoids the explicit constant $N_X$.
For $K3^{[n]}$-type varieties admitting a Lagrangian fibration, the previous half-dimension bound is recovered with the explicit constant $N_X = C_X \cdot n!\, I_X$.
On the Fano variety of a smooth cubic fourfold, the general result gives $\operatorname{ind}(\alpha)\mid \operatorname{per}(\alpha)^4$, improving the previously known exponent $5$ for periods avoiding a uniform integer.
The constants in the theorems are effective: they are built from $I_X$, the index of the transcendental lattice in $H^2(X,\mathbb{Z})/\operatorname{Pic}(X)$, and from the Beauville–Bogomolov–Fujiki form on $\operatorname{Pic}(X)$.

Reading between the lines

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

A natural reading of the numerical obstruction in Section 2.3 is that the half-dimension bound for Picard rank one would require a genuinely different input: the same twisted-bundle method cannot handle classes for which a certain integer is a quadratic residue modulo the period.
The same mechanism should transfer to other deformation types of hyper-Kähler varieties once a projectively hyperholomorphic bundle with the same twistor-path deformation property is known; the lattice-theoretic part of the proof depends only on the Beauville–Bogomolov–Fujiki form.
A testable consequence is that the explicit constants $N_X$ in Theorem 0.5 can be computed for concrete Picard lattices and compared with direct Brauer-group calculations on Hilbert schemes of points on $K3$ surfaces.

Share X Bluesky LinkedIn Reddit HN

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 short version: this paper proves genuinely new period-index bounds for K3[n]-type hyper-Kähler varieties, and the main arguments hold up. Theorem 0.2 gives the first general dim(X) bound for all K3[n]-type, and Theorem 0.5 gives the conjectured dim(X)/2 bound for non-special classes when Picard rank is at least two. Theorem 0.4 is a clean, usable criterion. The application of Markman's projectively hyperholomorphic bundles is new in this context, and Proposition 1.3, removing the coprime and primitivity assumptions, is a real technical improvement — assuming it is correct.

The soft spot the stress-test flagged is in Proposition 1.3, Step 1: the paper asserts that the modified Mukai vector v'_0 satisfies condition (7) without proof. That is a real omission, but it is not load-bearing. The preservation of (7) follows from the standing hypotheses. Write r = ρR, m = ρμ with gcd(R, μ) = 1. The modification replaces mH by mH + rkL, so the new Mukai vector is isotropic with the same r and with s' ≡ s mod R. The divisibility of the new class in the second cohomology is still ρ, and the expression in (7) is congruent modulo R to the original expression. Since the original satisfies (7), the new one does too. The stress-test's counterexample fails because that input does not satisfy (7) to begin with, so it is not an allowed input to Proposition 1.3. The authors should add a sentence or two making this verification explicit, but the gap is easily patched.

There are minor presentation slips — Lemma 2.5 writes q(L_u, B) = 4q(B) where a direct calculation gives 2q(B) — but nothing that changes the conclusion. The lattice-theoretic steps are otherwise careful, and the numerical obstruction discussion in Section 2.3 is honest about why the method cannot prove Huybrechts' conjecture in Picard rank one.

This paper is for people working in hyper-Kähler geometry and Brauer groups. It deserves a serious referee and, after the small fixes above, publication. I would take it to reading group and would cite it if I were working on period-index problems.

Referee Report

2 major / 3 minor

Summary. The paper studies the period-index problem for hyper-Kähler varieties of K3^{[n]}-type. It constructs α-twisted vector bundles of controlled rank from Markman's projectively hyperholomorphic bundles and derives three main results: Theorem 0.2 gives ind(α) | per(α)^{dim(X)} for every Brauer class with per(α) coprime to n! h I_X when X admits a primitive polarization of degree 2h; Theorem 0.4 gives a concrete criterion under which the stronger bound ind(α) | per(α)^{dim(X)/2} holds; Theorem 0.5 applies this criterion to non-special classes when the Picard rank is at least 2, for per(α) coprime to an explicit N_X. The central mechanism is Proposition 1.3, a strengthened version of Proposition 1.2 in which the coprimality assumption on the Mukai vector is removed. The paper also explains numerical obstructions to pushing the method further, and recovers Huybrechts' Lagrangian-fibration bound for K3^{[n]}-type varieties.

Significance. If the main results are correct, the paper gives the first general dimension bound for the period-index problem on K3^{[n]}-type varieties and strong evidence for Huybrechts' half-dimension conjecture in rank at least two. The strategy is original and builds on a published construction of Markman; the statements are explicit and involve no fitted parameters. The reliance on external results is clearly identified. However, the proof of Proposition 1.3 contains a load-bearing missing verification, described below, so the results are not fully established as written.

major comments (2)

[Section 1.3, Proposition 1.3, Step 1] The assertion that the modified Mukai vector v'_0 = (r, φ(mH+rkL), s') satisfies condition (7) is made without proof. This is not a formal consequence of the other stated properties: for a K3 surface with Pic(S)=ZH and H^2=2, the vector (4,2H,1) is primitive and isotropic with gcd(r,s)=1 but fails (7), since ρ=gcd(4,2)=2 and r/ρ=2 divides (1/2)(2H/2)^2+1 = 2. Thus the preservation of (7) under the replacement M ↦ M+rkL and passage to S' is a substantive arithmetic claim. The text neither verifies it by calculation nor explains how the choices of k and L can be made to guarantee it. This matters because Proposition 1.3 is invoked in the proofs of Theorems 0.2, 0.4, and 0.5; if (7) fails for some allowed input, the twisted-bundle construction of the stated rank is not obtained.
[Section 1.3, Proposition 1.3, Step 1] After replacing S by S' and v0 by v'_0, the proof also does not explicitly verify condition (iii) of Proposition 1.2 for the new pair. This condition is needed to ensure that the resulting twisted bundle has Brauer class [B/ℓ] rather than a different class. The verification is likely straightforward using the parallel-transport relation and absorbing integral or rational Picard terms, but it is omitted; it should be included for completeness.

minor comments (3)

[Section 2.4, Lemma 2.5] In the proof of Lemma 2.5, the equality q(L_u,B)=4q(B) is incorrect: since L_u=2ℓA+2B+uD and A is orthogonal to B and D, one has q(L_u,B)=2q(B). The subsequent divisibility conclusion still goes through because a is odd, but the statement as written is false.
[Section 2.2, Proposition 2.3] The line 'Then the proposition follows from gcd(2ℓ,2n-2)=2' is terse. The reader must supply the fact that gcd(ℓ,n-1)=1 follows from the standing coprimality assumption gcd(ℓ,n!h)=1; this should be stated explicitly.
[Throughout] There are a few typographical slips, e.g. 'assumptiom' in the proof of Lemma 2.5 and the truncated phrase 'instead of 2q()' in the reader's copy; these should be corrected in revision.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the period-index bounds are derived from Markman's external hyperholomorphic-bundle theorem and standard lattice arithmetic, not from self-referential reductions.

full rationale

The derivation chain is not circular. The central constructions of alpha-twisted vector bundles are attributed to Markman's published theorem [14] (deformation along diagonal twistor paths, the Hodge isometry phi_{U^[n]}, and the bundle U^[n] itself), with the paper explicitly saying 'A key result proved in [14]' and citing [14, Section 5.6, Lemma 7.1, Equation (7.11)] for the load-bearing facts. The only self-citation is [15], by four of the five authors, and it is cited merely 'for details' of Markman's machinery; no main theorem reduces to an unverified claim from that paper. The period-index inequalities are then obtained by the elementary rank argument of Lemma 1.1 and by lattice-theoretic constructions (Lemmas 2.1-2.2, Eichler criterion, Proposition 2.3) that do not encode the target divisibility statements. The paper also recovers Huybrechts's Theorem 0.1 for K3^[n]-type, an independent external benchmark. Two flagged items do not amount to circularity: Section 2.3 explicitly concedes that the method is not expected to prove Huybrechts's conjecture in Picard rank 1, and Proposition 1.3 Step 1 asserts without proof that the modified Mukai vector v'_0 satisfies condition (7). That omitted verification is a potential proof-completeness or correctness gap in a load-bearing lemma, but it is not a fitted input, a renamed known result, or a self-referential derivation, so it leaves the circularity score at 0.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The proof rests on Markman's projectively hyperholomorphic bundles, global Torelli, and standard lattice/period map inputs, all cited. There are no fitted free parameters and no invented entities. The paper's contribution is to organize these inputs into new period-index bounds.

assumptions (6)

domain assumption Existence and deformation property of Markman's projectively hyperholomorphic bundle U^{[n]} along diagonal twistor paths.
Used in Section 1.2 to obtain α-twisted vector bundles of rank dividing n! r^n on any parallel transport of S^{[n]}; this is the engine of Proposition 1.3.
domain assumption Global Torelli theorem for K3 surfaces and K3^{[n]}-type varieties.
Used in Proposition 1.3 Step 1 and in the proof of Theorem 0.2 to find Picard-rank-1 K3 surfaces and parallel transports with prescribed class ρ(H-ℓδ)=ϵL_u.
standard math Primitive embedding results for even lattices [9, Prop. 1.8, Cor. 1.9].
Lemma 2.2 uses these to construct a class A in B^⊥∩D^⊥ of divisibility 1.
standard math Eichler criterion for unimodular lattices and Markman's surjectivity of Torelli (via [13, Theorem 9.8]).
Used in Cases A.1/A.2 and Theorem 0.4 to construct a polarized K3 surface (S,H) with prescribed H^2 and a parallel transport realizing L_u.
standard math Existence and universality of moduli spaces of stable sheaves on K3 surfaces (Yoshioka).
Provides the K3 surface M and the universal rank-r bundle U used in Section 1.2.
standard math Hensel's lemma and the existence of rational points on nonsingular quadrics over finite fields.
Proposition 3.1 and Theorem 0.5 rely on these to solve the key congruence Q(x) ≡ b w^2 mod ℓ.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The period-index problem for hyper-K\"ahler varieties via hyperholomorphic bundles." pith.science (2026). https://pith.science/paper/NE36GSWD

@misc{pith2026250209774,
  author       = {Pith},
  title        = {Pith review of: The period-index problem for hyper-K\"ahler varieties via hyperholomorphic bundles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NE36GSWD}},
  note         = {Machine review of arXiv:2502.09774}
}

abstract

We prove new bounds for the period-index problem for hyper-K\"ahler varieties of $K3^{[n]}$-type using projectively hyperholomorphic bundles constructed by Markman. We show that $\mathrm{dim}(X)$ is a bound for any $X$ of $K3^{[n]}$-type. We also show that $\frac{1}{2}\mathrm{dim}(X)$ is a bound for most Brauer classes when the Picard rank of $X$ is at least two, providing evidence for a conjecture of Huybrechts.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

The period-index conjecture is false
math.AG 2026-08 accept novelty 8.0 of 10

The period-index conjecture is disproved: for every d≥3 there is a variety with a 2-torsion Brauer class of index 2^{d-1}, exceeding the conjectural bound.

Reference graph

Works this paper leans on

16 extracted references · 12 canonical work pages · cited by 1 Pith paper

[1]

A. Auel, E. Brussel, S. Garibaldi, and U. Vishne,Open problems on central simple algebras,Transform. Groups 16 (2011), no. 1, 219–264

work page 2011
[2]

Colliot-Th´ el` ene,Exposant et indice d’alg` ebres simples centrales non ramifi´ ees,Enseign

J.-L. Colliot-Th´ el` ene,Exposant et indice d’alg` ebres simples centrales non ramifi´ ees,Enseign. Math. (2) 48 (2002), no. 1-2, 127–146

work page 2002
[3]

Colliot-Th´ el` ene,Alg` ebres simples centrales sur les corps de fonctions de deux variables (d’apr` es A

J.-L. Colliot-Th´ el` ene,Alg` ebres simples centrales sur les corps de fonctions de deux variables (d’apr` es A. J. de Jong),S´ eminaire Bourbaki, Vol. 2004/2005, Ast´ erisque No. 307 (2006), Exp. No. 949, ix, 379–413

work page 2006
[4]

Colliot-Th´ el` ene,Die Brauersche Gruppe; ihre Verallgemeinerungen und Anwendungen in der arith- metischen Geometrie,arXiv:2311.02437

J.-L. Colliot-Th´ el` ene,Die Brauersche Gruppe; ihre Verallgemeinerungen und Anwendungen in der arith- metischen Geometrie,arXiv:2311.02437

arXiv
[5]

A. J. de Jong,The period–index problem for the Brauer group of an algebraic surface,Duke Math. J. 123 (2004), no. 1, 71–94

work page 2004
[6]

A. J. de Jong and A. Perry,The period–index problem and Hodge theory,arXiv:2212.12971

arXiv
[7]

Gritsenko, K

V. Gritsenko, K. Hulek, and G. K. Sankaran,Moduli spaces of irreducible symplectic manifolds,Compos. Math. 146 (2010), no. 2, 404–434

work page 2010
[8]

Hotchkiss and A

J. Hotchkiss and A. Perry,The period–index conjecture for abelian threefolds and Donaldson–Thomas theory,arXiv:2405.03315

arXiv

Show all 16 references

[9]

Huybrechts,Lectures onK3surfaces,Cambridge Stud

D. Huybrechts,Lectures onK3surfaces,Cambridge Stud. Adv. Math., 158, Cambridge University Press, Cambridge, 2016, xi+485 pp

2016
[10]

Huybrechts,The period–index problem for hyperk¨ ahler manifolds,arXiv.2411.17604

D. Huybrechts,The period–index problem for hyperk¨ ahler manifolds,arXiv.2411.17604

arXiv
[11]

Huybrechts and D

D. Huybrechts and D. Mattei,Splitting unramified Brauer classes by abelian torsors and the period–index problem,Math. Ann., to appear
[12]

Huybrechts and S

D. Huybrechts and S. Schr¨ oer,The Brauer group of analyticK3surfaces,Int. Math. Res. Not. 2003, no. 50, 2687–2698

2003
[13]

Markman,A survey of Torelli and monodromy results for holomorphic-symplectic varieties,Complex and differential geometry, 257–322, Springer Proc

E. Markman,A survey of Torelli and monodromy results for holomorphic-symplectic varieties,Complex and differential geometry, 257–322, Springer Proc. Math., 8, Springer, Heidelberg, 2011

2011
[14]

Markman,Rational Hodge isometries of hyper-K¨ ahler varieties ofK3 [n]-type are algebraic,Compos

E. Markman,Rational Hodge isometries of hyper-K¨ ahler varieties ofK3 [n]-type are algebraic,Compos. Math. 160 (2024), no. 6, 1261–1303

2024
[15]

Maulik, J

D. Maulik, J. Shen, Q. Yin, and R. Zhang,TheD-equivalence conjecture for hyper-K¨ ahler varieties via hyperholomorphic bundles,arXiv 2408.14775v4

arXiv
[16]

Yoshioka,Stability and the Fourier–Mukai transform

K. Yoshioka,Stability and the Fourier–Mukai transform. II,Compos. Math. 145 (2009), no. 1, 112–142. 20 J. HOTCHKISS, D. MAULIK, J. SHEN, Q. YIN, AND R. ZHANG Columbia University Email address:james.hotchkiss@columbia.edu Massachusetts Institute of Technology Email address:maul...

2009

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

A. Auel, E. Brussel, S. Garibaldi, and U. Vishne,Open problems on central simple algebras,Transform. Groups 16 (2011), no. 1, 219–264

work page 2011

[2] [2]

Colliot-Th´ el` ene,Exposant et indice d’alg` ebres simples centrales non ramifi´ ees,Enseign

J.-L. Colliot-Th´ el` ene,Exposant et indice d’alg` ebres simples centrales non ramifi´ ees,Enseign. Math. (2) 48 (2002), no. 1-2, 127–146

work page 2002

[3] [3]

Colliot-Th´ el` ene,Alg` ebres simples centrales sur les corps de fonctions de deux variables (d’apr` es A

J.-L. Colliot-Th´ el` ene,Alg` ebres simples centrales sur les corps de fonctions de deux variables (d’apr` es A. J. de Jong),S´ eminaire Bourbaki, Vol. 2004/2005, Ast´ erisque No. 307 (2006), Exp. No. 949, ix, 379–413

work page 2006

[4] [4]

Colliot-Th´ el` ene,Die Brauersche Gruppe; ihre Verallgemeinerungen und Anwendungen in der arith- metischen Geometrie,arXiv:2311.02437

J.-L. Colliot-Th´ el` ene,Die Brauersche Gruppe; ihre Verallgemeinerungen und Anwendungen in der arith- metischen Geometrie,arXiv:2311.02437

arXiv

[5] [5]

A. J. de Jong,The period–index problem for the Brauer group of an algebraic surface,Duke Math. J. 123 (2004), no. 1, 71–94

work page 2004

[6] [6]

A. J. de Jong and A. Perry,The period–index problem and Hodge theory,arXiv:2212.12971

arXiv

[7] [7]

Gritsenko, K

V. Gritsenko, K. Hulek, and G. K. Sankaran,Moduli spaces of irreducible symplectic manifolds,Compos. Math. 146 (2010), no. 2, 404–434

work page 2010

[8] [8]

Hotchkiss and A

J. Hotchkiss and A. Perry,The period–index conjecture for abelian threefolds and Donaldson–Thomas theory,arXiv:2405.03315

arXiv