Skip to content

Documentary state: EDITORIALLY_ADMITTED. The manuscript has a protected editorial admission, not a theorem certification.

OAI-26 family 038: Fujita's freeness conjecture

Editorial status: SOURCE_CRITIC_COMPLETE; protected acceptance is determined by the admissions ledger. Mathematical status: NO_CERTIFICATION. This article describes the pinned manuscript claim and does not independently verify its proof.

The question in plain language

A line bundle on an algebraic variety can have global sections—functions defined locally that fit together consistently. A bundle is globally generated when these sections can provide a nonzero value at every point, so there are no unavoidable base points.

Fujita's freeness conjecture predicts a universal numerical threshold beyond which a canonical bundle made more positive by an ample line bundle is globally generated. The pinned OAI-26 overview claims the sharp threshold in every dimension for smooth complex projective varieties.

A small example

Let \(X\) be a smooth projective complex variety of dimension \(n\), let \(K_X\) denote its canonical divisor, and let \(L\) be an ample line bundle. The predicted global-generation threshold is

\[ K_X+mL\quad\text{for every integer }m\geq n+1. \]

The notation \(K_X+mL\) describes the tensor product of the canonical line bundle with the tensor power \(L^{\otimes m}\). Global generation means that for every point of \(X\) there is a global section not vanishing there.

On a smooth projective curve, the numerical threshold specializes to \(m\geq2\). The general conjecture is stronger than proving the claim for curves or surfaces individually: its bound depends only on the dimension, not on the particular variety or ample line bundle.

What is known

The source-locked Family 038 catalogue identifies one manuscript, Fujita's freeness conjecture (September 23, 2026).

The pinned primary PDF, Theorem 1.1, states that for every smooth connected projective variety over \(\mathbb{C}\) of dimension \(n\geq1\) and every ample line bundle \(L\), \(K_X+(n+1)L\) is globally generated. Corollary 6.3 extends this to every integer \(m\geq n+1\), even when \(L\) itself is not globally generated. Connectedness is not essential, and the paper separately notes that the zero-dimensional case is trivial. The bound is sharp for \(X=\mathbb{P}^n\) with \(L=\mathcal{O}(1)\): \(K_X+(n+1)L\) is trivial, whereas \(K_X+nL\) has no nonzero sections.

Global generation is a different property from very ampleness, which additionally ensures that the sections embed a variety into projective space. A proof of Fujita's freeness conjecture should not be reported as a proof of Fujita's separate very-ampleness bound without checking an additional theorem.

Likewise, the indexed claim does not cover singular varieties or varieties over arbitrary positive-characteristic fields. Extending the conclusion to those settings requires separate arguments.

GCL verification state

The documentary source is pinned to openai/math@1581d4cbdd69a0faf34b2089614fe2a814316a6a. The primary PDF was read directly at that exact source revision: Git blob c91c48b4c31f0fa53f5885d808d1a5dcc5a9d8f6, 485,593 bytes, 25 pages, SHA-256 73aefa0b0e5578dfba86cca2231608f2ba628519238cad062e49b3dc095be40b. The abstract, Theorem 1.1, sharp example, and opening proof outline were inspected.

This pass does not independently check the proof of global generation, the dimension-independent aspects of the argument, or the claimed sharpness. No Family 038-specific Lean theorem, Comparator acceptance, paper-to-formal semantic correspondence, or MATHCERT certificate has been verified here.

What remains

A bounded primary-source critic has checked Theorem 1.1, its hypotheses, the sharp projective-space example, and the extension to all permitted integers \(m\). Full proof-level review remains outstanding: the discrepancy/valuation minimization, log-canonical-center reduction, uniform restriction estimates, and Kawamata–Viehweg vanishing argument have not been independently verified.

Editorial admission still requires the exact article hash, tests, generated HTML inspection and protected merge/readback. This is documentary acceptance only; it does not verify the claimed mathematical result.

Sources and claim boundaries

Boundary: NO_CERTIFICATION. The document records a publisher-reported resolution of Fujita's freeness conjecture, not an independently established or formally certified proof.


Indexed primary manuscript paths:

  • Fujita's freeness conjecture — preprints/Fujitas-freeness-conjecture-September-23-2026/Fujitas-freeness-conjecture-September-23-2026.pdf