← All problems

Colliot-Thélène's period-index conjecture

17 years old · traced to

Let kk be a field of dimension dd, and let b1∈Br⁡(k)b1\in\operatorname{Br}(k). The index ind(α)ind(\alpha) is the square root of the rank of the unique division algebra representing α\alpha.

Period-index conjecture.

ind(α)∣(per(α))d−1ind(\alpha)\mid (per(\alpha))^{d-1}

for all α∈Br⁡(k)\alpha\in\operatorname{Br}(k).

This is the algebraic period-index conjecture attributed in the source to Colliot-Thélène. The surrounding text presents a topological period-index theorem as evidence, but gives no resolution of this conjecture.

Equivalent formulations
  1. Let XX be a smooth projective variety over an algebraically closed field, and let α∈Br⁡(X)\alpha\in\operatorname{Br}(X) be a Brauer class. Write per⁡(α)\operatorname{per}(\alpha) for its period and ind⁡(α)\operatorname{ind}(\alpha) for its index. Colliot-Thélène's conjecture. Every such class satisfies

    ind⁡(α)∣per⁡(α)dim⁡(X)−1.\operatorname{ind}(\alpha)\mid\operatorname{per}(\alpha)^{\dim(X)-1}.

    This is the general period-index conjecture. The conjecture is wide open in general, although several cases and partial bounds are known.

    source: Alessio Bottini and Daniel Huybrechts, “The period-index problem for hyperkähler varieties: Lower and upper bounds”, arXiv:2512.15131 (2025).

References
  1. Benjamin Antieau, “Cohomological obstruction theory for Brauer classes and the period-index problem”, arXiv:0909.2352 (2010).

Progress summary

Refreshed
Open · solution reported

A 2026 preprint claims the conjecture is false in three or more dimensions, but its counterexamples have not been independently verified.

Attributed to Colliot-Thélène, the conjecture predicts that every Brauer class over a field of dimension dd has index dividing its period raised to the power d−1d-1. A 2026 preprint claims explicit counterexamples beginning in dimension 33.

Known results

  • The conjecture is vacuous for d≤1d\leq 1.
  • It holds for d=2d=2 by work of de Jong, Lieblich, and de Jong–Starr.
  • A 2024 paper proves ind⁡(α)∣per⁡(α)2\operatorname{ind}(\alpha)\mid\operatorname{per}(\alpha)^2 for suitable classes on abelian threefolds, not the general case.
  • Topological and Hodge-theoretic results give evidence or conditional bounds, but do not resolve the algebraic conjecture.

August 2026 counterexample claim

The preprint The period-index conjecture is false claims, over every algebraically closed characteristic-zero field, a smooth projective threefold with per⁡(α)=2\operatorname{per}(\alpha)=2 and ind⁡(α)=8\operatorname{ind}(\alpha)=8, and higher-dimensional examples with ind⁡(αd)=2d\operatorname{ind}(\alpha_d)=2^d. It reports that ChatGPT assisted the author’s discovery process; the author wrote the paper and accepts responsibility for it.

Current status (as of September 2026): The cases d≤2d\leq 2 are settled, while the claimed counterexamples for d≥3d\geq 3 remain unverified.

Sources

Solutions 0

No solutions have been posted yet.