Rigidity and arithmetic of lattice von Neumann algebras

Claims the von Neumann algebra of a property (T) lattice remembers its group, its twist and its scale.

2 papers September 23, 2026 Explained by AI

From any group one can build an operator algebra called its group factor. A long-standing question asks how much of the group survives in that algebra. Two companion papers study groups like SL₃(ℤ), integer 3-by-3 matrices of determinant 1, and their relatives. They claim every finite-index link between such a factor and another group factor comes from actual subgroup matches. From this they claim the factor recovers the group, its twist and its scale.

First step

1980

Alain Connes, countability of the fundamental group of a property (T) group factor

Significance

Major

It gives stable recovery of the group from its factor for a broad class of lattices, answering Connes- and Popa-type rigidity questions there.

Trust

Claimed, not checked

Field

Operator algebras

2 papers

The explanation

Take the group SL₃(ℤ), all 3-by-3 integer matrices with determinant 1. Reduce each entry modulo 2. You get a map onto the 3-by-3 matrices over the two-element field with determinant 1. That finite group has (8 − 1)(8 − 2)(8 − 4) = 7 × 6 × 4 = 168 elements. The matrices that map to the identity form a subgroup of index 168. Index counts how many copies of the subgroup fill the whole group.

From a group one can build an operator algebra, its group factor. Each group element becomes a unitary operator, and the algebra is their closure. A scalar twist multiplies products by fixed complex numbers of size one. The question is how much of the group the factor remembers. Connes proved that the fundamental group of a property (T) group factor is countableSource 1. That fundamental group lists the scales t at which the factor matches its own resized copy. Property (T) is a strong rigidity condition on unitary representations. But property (T) alone does not ensure recovery. OpenAI and Zhou independently built different property (T) groups with the same group factorSource 2. The obstacle is that the comparison group can be any group, with no geometry to start from.

The papers work with a looser link called a bifinite correspondence. It is a space where both factors act, with finite size on each side. The subgroup above gives one. Match it to itself and use a one-dimensional representation. The formula then gives size n × 168 on each side, with n = 1, so 168Source 3. The first paper claims every bifinite correspondence is built from such modelsSource 4. These come from real finite-index subgroup isomorphisms and finite-dimensional projective representations. A nonzero correspondence forces the two groups to be commensurable, sharing isomorphic finite-index subgroupsSource 5.

The class is broad. It covers groups commensurable with lattices in products of simple groups over the real numbers, the complex numbers or p-adic fields. Lattices are discrete subgroups of finite covolume. SL₃(ℤ) and every finite-index subgroup belong to itSource 6. The second paper claims stable recovery. Two resized twisted factors match only when the scales agree and the groups are isomorphic, with matching twistsSource 7. Every factor in the class has trivial fundamental groupSource 8.

How sure is this? A computer proof checker called Lean has not checked any part of it. The argument builds geometry inside the correspondence over many long sections, including rank-one quaternionic and Cayley cases. No specialist review of the papers has been published.

What it doesn't show. This does not cover every property (T) group, and other papers show property (T) alone does not ensure recovery.

Where it came from

This result comes from an unreleased OpenAI model, published in the openai/math release on 6 October 2026.

Hype check

Major if it holds, for a wide class of lattices. It does not cover all property (T) groups, and it has no Lean check or published specialist review.

Words used

Group factor: The operator algebra made by turning each group element into a unitary operator and closing up; a factor has trivial centre.
ICC: Infinite conjugacy classes: every element other than the identity has infinitely many conjugates.
Property (T): A rigidity condition: any unitary representation with almost-invariant vectors has a truly invariant vector.
Commensurable: Two groups are commensurable if they have isomorphic subgroups of finite index.
Bifinite correspondence: A Hilbert space on which one factor acts from the left and another from the right, with finite size on both sides.
Scalar twist (cocycle): A rule assigning a complex number of size one to each pair of elements, used to twist multiplication.
Fundamental group of a factor: The set of scales t at which the factor is isomorphic to its own resized copy.
Projective representation: A map to matrices that respects multiplication up to fixed scalar factors.

The original papers (2)

Arithmeticity of twisted finite correspondences for lattices over local fields (PDF, opens in a new tab)
September 23, 2026
The arithmetic category and stable recovery of lattice factors (PDF, opens in a new tab)
September 23, 2026

All quoted lines (8)

Connes proved countability of the fundamental group of a property- $(T)$ group factor; his later problems ask whether nonisomorphic ICC property- $(T)$ groups have nonisomorphic factors and ask for the fundamental groups of those factors.
Arithmeticity of twisted finite correspondences for lattices over local fields Back to text
Property $(T)$ alone does not ensure group reconstruction: independent, concurrent constructions of OpenAI and Zhou give nonisomorphic ICC property- $(T)$ groups with isomorphic group factors.
Arithmeticity of twisted finite correspondences for lattices over local fields Back to text
If $n=\dim_{\mathbb C}V_\sigma$ , the elementary model has dimensions $\dim_{L_\mu(\Gamma)}E(A,B,\delta,\sigma)=n[\Lambda:B],\qquad \dim E(A,B,\delta,\sigma)_{L_\omega(\Lambda)}=n[\Gamma:A].$
Arithmeticity of twisted finite correspondences for lattices over local fields Back to text
Every bifinite $L_\mu(\Gamma)$ – $L_\omega(\Lambda)$ correspondence is unitarily isomorphic to a closed bimodule summand of a finite direct sum of the models $E(A,B,\delta,\sigma)$ in Equation (…), where $A\leq\Gamma$ and $B\leq\Lambda$ have finite index, $\delta:A\to B$ is an actual group isomorphism, and $\sigma$ is a finite-dimensional projective unitary representation satisfying Equation (…).
Arithmeticity of twisted finite correspondences for lattices over local fields Back to text
In particular, existence of a nonzero correspondence forces $\Gamma$ and $\Lambda$ to be abstractly commensurable.
Arithmeticity of twisted finite correspondences for lattices over local fields Back to text
For example, $\mathrm{SL}_3(\mathbb Z)$ and every finite-index subgroup belong to $\mathscr K$ .
The arithmetic category and stable recovery of lattice factors Back to text
Then $L_\omega(\Lambda)^t\cong L_\nu(\Pi)^s$ if and only if $t=s$ and there are an isomorphism $\delta:\Lambda\to\Pi$ and a normalized cochain $b:\Lambda\to\mathbb T$ such that $\omega(g,h)=b(g)b(h)b(gh)^{-1}\nu(\delta(g),\delta(h)).$
The arithmetic category and stable recovery of lattice factors Back to text
In particular every factor $L_\omega(\Lambda)$ in this class has trivial fundamental group.
The arithmetic category and stable recovery of lattice factors Back to text

Other sources

openai/math release notes (2026): model provenance and review status manuscripts and supporting proof artifacts produced by an internal OpenAI model

285Counterexamples to Baum–Connes and Kadison–KaplanskyOperator algebras
290Relative bicentralizers and modular spectral recoveryOperator algebras
287Isomorphism of the free group factorsOperator algebras
288Kadison's similarity conjectureOperator algebras
289Strong Kadison–Kastler stability and its spatial boundariesOperator algebras

Browse Operator algebras Open since the 1980s Claimed, not yet checked