When every extended trace determines a stabilized operator algebra

← All digests

Family 301 · Operator algebras
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-25

1. The question

Can an operator algebra be reconstructed from all its ways of measuring positive elements? Usually traces alone lose too much information: projection classes, ideals and other invariants may distinguish algebras with similar ordinary trace spaces.

Family 301 claims a setting in which a suitably complete trace invariant is enough. After tensoring a separable nuclear C*-algebra with the Razak–Jacelon algebra and with compact operators, its isomorphism type is determined by the topological cone of all extended lower-semicontinuous tracial weights.

“All” is crucial. The invariant includes weights taking only zero and infinity, and these encode the ideal structure. The theorem allows arbitrary ideals, nonunital algebras, and mixtures of finite and infinite behavior. It also claims that an algebra isomorphism can realize a prescribed isomorphism of trace cones.

2. Background and essential definitions

A C*-algebra is an adjoint algebra complete in a norm satisfying $\|a^*a\|=\|a\|^2$. Its positive elements form $A_+$. Separability means a countable norm-dense subset exists; nuclearity means finite sets admit completely positive finite-dimensional approximations.

An extended tracial weight is a map $\tau:A_+\to[0,\infty]$ that is additive, positively homogeneous, vanishes at zero, and satisfies $\tau(x^*x)=\tau(xx^*)$. Lower semicontinuity means its value cannot jump upward under a norm-convergent approximation in the forbidden direction: $\tau(a)\leq\liminf\tau(a_i)$ when $a_i\to a$.

The weight need not be finite on a dense subset. For every closed two-sided ideal $I$, the cone includes

$$\tau_I(a)= \begin{cases} 0,&a\in I_+,\\ \infty,&a\in A_+\setminus I_+. \end{cases}$$

These are not normalized states. Omitting them would discard information central to the classification claim.

The cone $T(A)$ has addition, multiplication by positive real scalars, and a canonical topology. With $(a-\varepsilon)_+$ denoting the positive part obtained by cutting off the spectrum below $\varepsilon$, convergence is characterized by

$$\limsup_i\tau_i((a-\varepsilon)_+) \leq\tau(a)\leq\liminf_i\tau_i(a)$$

for every positive $a$ and every $\varepsilon>0$. A cone isomorphism must be a homeomorphism preserving addition, zero and positive scalar multiplication.

The Razak–Jacelon algebra $\mathcal W$ is simple, nuclear, stably projectionless and has a unique tracial state. Stably projectionless means no nonzero projection appears even after tensoring with compact operators. Write $\mathcal K=\mathcal K(\ell^2(\mathbb N))$. The theorem classifies $A\otimes\mathcal W\otimes\mathcal K$, not $A$ itself; both tensor factors are part of the conclusion.

3. Development history and previous partial results

Razak classified certain simple inductive limits of stably projectionless building blocks using traces and scale. Jacelon isolated $\mathcal W$, with trivial K-theory and a unique trace, as a central example. Its tensorial role suggests removing K-theoretic distinctions while retaining numerical size information.

Elliott, Robert and Santiago developed the full topological cone of extended traces and its relationship to functionals on the Cuntz semigroup. The latter records comparison classes of positive operators. Robert showed that tensoring with $\mathcal W$ corresponds to a realification of that semigroup. Thus the trace cone already determines the stabilized order invariant; the remaining question is whether the invariant can be realized by actual algebra isomorphisms.

Robert’s trace-cone classification question was recorded in 2012 and appears in the later Schafhauser–Tikuisis–White problem collection. Simple tracial cases were covered by classification results under their respective contractibility and regularity hypotheses. Other results show absorption of $\mathcal W$ for suitable simple monotracial algebras.

The traceless branch behaves differently. If every extended trace takes only zero and infinity, the cone mainly records ideals, and purely infinite classification tools address the corresponding stabilization. Neither that branch nor the simple tracial branch automatically handles an algebra with both finite and infinite regions.

Restricted nonsimple classification results treated particular extensions with additional hypotheses. M1 claims the full cone-only theorem for arbitrary ideal structure, without a universal coefficient theorem assumption. Its historical discussion emphasizes that even one proper nonzero ideal introduces difficulties absent from simple classification.

4. What the new manuscripts claim

Let $A,B$ be separable nuclear complex C*-algebras. M1 states that an isomorphism of their full topological trace cones implies

$$A\otimes\mathcal W\otimes\mathcal K \cong B\otimes\mathcal W\otimes\mathcal K.$$

All tensor products are spatial. The conclusion is an actual C*-algebra isomorphism, not just agreement of trace data, Cuntz semigroups, or asymptotic models.

Put $P=A\otimes\mathcal W\otimes\mathcal K$ and $Q=B\otimes\mathcal W\otimes\mathcal K$. After transporting the given cone map to the stabilized cones in the contravariant direction,

$$F:T(Q)\longrightarrow T(P),$$

M1 constructs an isomorphism $\Phi:P\to Q$ satisfying

$$\tau\circ\Phi=F(\tau)\qquad(\tau\in T(Q)).$$

Thus the specified invariant map is realized, not merely some map between isomorphic cones.

The algebras may be nonunital, need not be simple, and may have arbitrary primitive ideal spaces. Weights may have nondense finite domains. No extra dimension, trace-boundary or UCT condition is imposed in the main theorem.

The source also observes that these stabilizations have nuclear dimension at most one, using the separate dimension theorem in family 291 and absorption of $\mathcal Z$ by $\mathcal W$. It explicitly states that its classification argument does not use that dimension bound.

5. Why the advance matters

Importance: 5/5. If correct, this is a broad nonsimple classification theorem with a single integrated invariant. It would show that, after the stated stabilization, trace data encode not only finite numerical sizes but also the ideal geometry needed to distinguish the algebras.

The invariant is richer than the familiar simplex of tracial states. For instance,

$$I\subseteq J\quad\Longleftrightarrow\quad \tau_I+\tau_J=\tau_I.$$

The cone’s arithmetic therefore remembers inclusion between ideals. Its topology and scaling also distinguish where general weights are finite and where they vanish.

Realizing a prescribed cone map adds classification strength. It connects not just objects but the way their invariants are identified. This matters when one wants a particular trace correspondence respected by an algebra isomorphism.

6. The main idea

Recover positive-element order and ideal supports

Cuntz comparison writes $a\precsim b$ when $v_n^*bv_n\to a$ in norm. Its classes form $\mathrm{Cu}(A)$, an ordered semigroup with increasing suprema. For a positive contraction, the rank associated with a weight is

$$d_\tau(a)=\lim_{k\to\infty}\tau(a^{1/k}).$$

Robert’s realification result converts the cone isomorphism into a compatible Cuntz-semigroup isomorphism for $P,Q$. It also identifies their ideal spaces. M1 then tries to build algebra maps implementing this data.

Measure errors where a weight is finite

In a simple normalized-trace setting, an error can often be bounded uniformly over a compact trace space. Here a weight may be finite only on a proper ideal. A full positive element may even have infinite rank for every relevant nonzero weight, making a small scalar multiple of that rank useless as an error estimate.

M1 therefore fixes local controllers: spectral cuts and ideal supports on which finite quantities can be compared. Upper bounds keep errors small there, while lower support estimates prevent a map from accidentally erasing the regions where the prescribed weight is infinite.

This two-sided control is the central bookkeeping difficulty. Matching all finite numerical values would not suffice if the resulting map changed which elements had infinite weight.

Build sequence models and correct them in norm

The first maps land in a sequence algebra $Q_\infty=\ell^\infty(Q)/c_0(Q)$, where errors may vanish along coordinates. Regularized limits of weights retain lower semicontinuity through spectral cuts.

Finite-dimensional nuclear approximations and von Neumann algebra uniqueness tools produce tracial models. Small stabilizing pieces simultaneously satisfy upper trace estimates and lower ideal-support comparisons. Ideal-related absorption and stable-uniqueness methods then remove differences in operator norm, rather than leaving only agreement under traces.

The manuscript explicitly treats correction on the original module and recovery of the full trace formula, including infinite values. These are the steps turning a numerical model into an algebraic one.

Pass from models to an actual isomorphism

A suspension-and-corner construction converts point and interval models into actual maps with the required trace normalization. Approximate intertwining builds maps in both directions and adjusts successive compositions to approach the identity on increasing finite sets.

Separability permits this countable approximation scheme. The limiting maps are mutually inverse and realize the prescribed cone correspondence. The final output is therefore an isomorphism of the stabilized C*-algebras, not just an embedding into a sequence quotient.

7. Scope, limitations, and open questions

The theorem classifies after tensoring with both $\mathcal W$ and $\mathcal K$. It does not claim that equal trace cones classify the original algebras, nor that either stabilization can always be omitted. It also does not classify all C*-algebras without nuclearity or separability.

The cone must include every extended lower-semicontinuous tracial weight with the specified topology. Restricting to bounded traces, normalized states or densely finite weights would change the invariant and could lose ideal information. A set-theoretic bijection of weights or an affine identification of one trace base is not the theorem’s hypothesis.

The map-realization statement is an existence assertion, not uniqueness of the implementing algebra isomorphism. Likewise, the separate nuclear-dimension consequence is not a hidden assumption in the classification construction.

Understanding which additional hypotheses permit removing compact stabilization, or using a smaller trace invariant in particular classes, remains a separate classification question. M1’s claim is the full-cone theorem in the stated stabilized setting.

8. Sources and further reading

• Jacelon, A simple, monotracial, stably projectionless C*-algebra, for the construction and role of $\mathcal W$.
• M1’s introduction gives the precise Robert realification theorem, the earlier simple and traceless classification cases, and the formulation of the full trace-cone question.
• The source’s Sections 2–3 specify the extended-weight topology and local error ideals; Section 8 states realization of the prescribed cone map.

Source manuscripts

• M1. OpenAI, The trace cone classifies Razak–Jacelon stabilizations (September 25, 2026).

Sources consulted on 25 October 2026 at repository revision 1581d4c. Download the manuscript citations.