Kadison’s orthonormal bases of unitaries problem

Does every type II1\mathrm{II}_1 factor MM, equipped with its normalized trace τ\tau, admit an orthonormal basis of L2(M,τ)L^2(M,\tau) consisting of unitaries in MM? Equivalently, does there exist a family {ui}iIM\{u_i\}_{i\in I}\subset M such that each uiu_i is unitary, τ(uiuj)=δij\tau(u_i^*u_j)=\delta_{ij} for all i,jIi,j\in I, and the linear span of {ui:iI}\{u_i:i\in I\} is dense in L2(M,τ)L^2(M,\tau)?

Progress summary

Solved

A new preprint claims to settle Kadison’s question for every type II1_1 factor, but the result has not yet been independently verified.

Kadison posed the question in 1967: whether every type II1\mathrm{II}_1 factor admits a complete orthonormal basis of trace vectors, equivalently an orthonormal basis of unitaries in the associated L2L^2-space. A new preprint claims a full solution, extending an earlier claimed solution for the separable case.

Known results

  • Group von Neumann algebras of countable discrete groups and certain group-measure-space constructions were already known cases.
  • De and Mukherjee (2023) constructed uniformly bounded self-adjoint orthonormal bases for every separable diffuse finite von Neumann algebra, without proving that the basis elements are unitaries.
  • He, Tang, and Zhang (2026) claimed the unitary-basis result for diffuse finite von Neumann algebras with separable L2L^2-space.

August 2026 full-resolution claim

The abstract of a new preprint claims to extend the separable result to arbitrary density character using a transfinite construction, thereby settling Kadison’s question for all type II1\mathrm{II}_1 factors. No peer-reviewed verification, independent exposition, identified gap, or retraction was found.

Current status (as of August 2026): The separable case and now the general case are claimed in preprints, but the all-factor resolution remains unverified.

Sources
Sources & referencesView supporting material

Solutions 0

No solutions have been posted yet.