[Paper Review] On the classification problem for nuclear C*-algebras

Andrew S. Toms|ArXiv.org|Sep 5, 2005

Advanced Operator Algebra Research8 references22 citations

TL;DR

This paper constructs a counterexample to Elliott's classification conjecture for simple, separable, nuclear C*-algebras by exhibiting two non-isomorphic algebras with identical Elliott invariants and all standard continuous homotopy invariants. The Cuntz semigroup is shown to be necessary to distinguish them, proving that existing invariants are insufficient and that classification requires stronger regularity conditions like Z-stability rather than expanded invariants.

ABSTRACT

We construct a simple, unital AH algebra which is shape equivalent to its tensor product with any infinite-dimensional UHF algebra, has the same tracial simplex as the said tensor product, and yet is not isomorphic to it. An analogous result for automorphisms is also proved.

Motivation & Objective

• To challenge the completeness of the Elliott invariant for classifying simple, separable, nuclear C*-algebras.
• To demonstrate that existing invariants—including real rank, stable rank, and continuous homotopy functors—fail to distinguish non-isomorphic algebras.
• To show that the Cuntz semigroup is the minimal invariant needed to achieve classification beyond the Elliott invariant.
• To argue that the classification program must adopt new regularity conditions (e.g., Z-stability) rather than expanding the invariant.
• To clarify the limitations of current classification methods and the necessity of stronger structural assumptions.

Proposed method

• Construct a simple, unital, separable, nuclear AH algebra A that is not isomorphic to A ⊗ U, where U is a UHF algebra.
• Show that A and A ⊗ U have identical values for all invariants in the collection F, including the Elliott invariant, real rank, stable rank, and continuous homotopy functors.
• Use the Cuntz semigroup W(·) to distinguish A and A ⊗ U, proving it is not captured by F.
• Construct an automorphism α of a C*-algebra B of period two that acts trivially on all invariants in F_R, yet is not locally inner.
• Employ pullbacks of projections via continuous maps from a product of spheres to a cube to relate Cuntz equivalence to Murray-von Neumann equivalence.
• Establish an order embedding ι: D(Y) → W(C([0,1]^N)) to link vector bundles over products of spheres to the Cuntz semigroup.

Experimental results

Research questions

• RQ1 Can two non-isomorphic simple, separable, nuclear C*-algebras have identical Elliott invariants and all standard continuous homotopy invariants?
• RQ2 Is the Cuntz semigroup necessary to distinguish such algebras, and can it be captured by existing invariants?
• RQ3 Can automorphisms that act trivially on all standard invariants fail to be locally inner, indicating failure of uniqueness?
• RQ4 What is the role of Z-stability in the classification of nuclear C*-algebras, and how does it relate to slow dimension growth?
• RQ5 Can the Cuntz semigroup be used effectively in classification theorems, or is its range problem too intractable?

Key findings

• A simple, unital, separable, nuclear AH algebra A exists such that A and A ⊗ U are not isomorphic, yet F(A) ≅ F(A ⊗ U) for all F ∈ F.
• The Cuntz semigroup distinguishes A and A ⊗ U, proving it is not captured by the Elliott invariant or any continuous homotopy invariant in F.
• An automorphism α of period two on a C*-algebra B acts as the identity on all invariants in F_R but is not locally inner, showing failure of uniqueness.
• The Cuntz semigroup is the minimal invariant needed to repair the Elliott conjecture, as it captures information beyond F.
• Existing classification results for simple unital AH algebras via the Elliott invariant are optimal and cannot be improved by adding continuous homotopy invariants.
• The results suggest that Z-stability—not expanded invariants—is the correct regularity condition for a complete classification program.