A Counterexample to Kaplansky's Direct-Finiteness Conjecture in Characteristic Two

OpenAI

There are a finite field K of characteristic two, a finitely presented group G containing odd-prime torsion, and scalar a,b in K[G] with ab=1 but ba≠1. The manuscript gives a terminating prescription for the witness; it does not assert K=F₂.

Sources for this record

What was already known

Group algebras over characteristic-zero fields and those of sofic groups were known directly finite. The unrestricted positive-characteristic scalar question remained open; a matrix defect alone would not settle it.

AI involvement AI-led

OpenAI attributes the manuscript and Lean artifact to an internal model. No named human checker or explanation of the finite-incidence scalar construction is documented.

The OpenAI-authored release attributes this mathematical construction to an internal model; individual-session autonomy is undocumented.

Importance Major

The scalar witness negates unrestricted Kaplansky direct finiteness and also yields a nonsurjunctive cellular automaton for the constructed group.

Limitations

The verified scope uses odd-prime torsion and an unspecified finite characteristic-two field. The separate torsion-free companion is paper-only and is outside this Formal assessment. The prescription was proved to terminate but was not executed.

Technical verification Formal (Lean 4 with Mathlib and Comparator)

The reported Lean theorem checks the characteristic-two scalar witness for a finitely presented group with odd-prime torsion. The torsion-free October companion has no cited complete check and is outside this rating.

Formal checking

Source-reported Lean proof of the finite-presentation scalar counterexample; the torsion-free strengthening is outside this theorem.

Checker
Lean 4 kernel, as reported by the OpenAI math release
Scope
Finite characteristic-two field, finitely presented odd-prime-torsion group and scalar ab=1 with ba≠1
Basis
Reported by the linked checker
Evidence source

Checking a formal statement does not establish that it matches the original problem.

Human understanding Few

No named human explanation of the finite-incidence scalar construction is documented; Few is the evidence default.

Correction and publication history

2026-10-18 · publication

The scalar one-sided inverse result is distinct from published zero-divisor, odd-characteristic surjunctivity and determinant witnesses.

Read this published revision