This is a question that is a bit outside my usual mathematical comfort zone, but I feel like an expert might know the answer.

Recall that a dimension group is an ordered abelian group $G$ with positive cone $G^+$ that arises as a limit of ordered groups of the form $\mathbb Z^d$ with the componentwise order. It is called simple if every element in $G^+\setminus\{0\}$ is an order unit.

If one considers a distinguished order unit $u\in G^+$, then one defines the state space $\mathcal{S}(G,G^+,u)$ as the set of all states (duh), meaning all order-preserving homomorphisms $s: G\to\mathbb R$ with $s(u)=1$. If $G$ is countable, then the topology of pointwise convergence will give the state space the structure of a metrizable Choquet simplex.

It is a well-known fact that every metrizable Choquet simplex $T$ can be realized as the state space of a simple dimension group $(G,G^+,u)$. However, it is also well-known that this is highly non-unique, and in fact there are many possible choices one can make in general. I am interested to know whether one can ask the following in addition:

Let $T$ be a metrizable Choquet simplex and let $f: T\to (0,1)$ be an affine continuous function. Does there exist a simple dimension group $(G,G^+,u)$ with distinguished order unit together with another element $a\in G^+$ such that $T$ is affinely homeomorphic to $\mathcal{S}(G,G^+,u)$ in such a way that $f$ corresponds to the assignment $[s\mapsto s(a)]$?

If this happens to be false in general, are there any natural sufficient conditions on the function $f$?

A note for C*-algebraists: What this question really asks is whether there exists a unital simple AF algebra $A$ having $T$ as its tracial state space and a projection $p\in A$ such that the function $f$ corresponds to the point evaluation $[\tau\mapsto\tau(p)]$.


Yes. Form Aff $(T)$; since $T$ is metrizable, Aff $(T)$ contains a countable dense set. Adjoin $f$ and the constant function $1$ to the countable dense set (creating a larger countable dense set), and let $G$ be the subgroup of Aff $(T)$ generated by this enlarged set. This is of course countable, and dense.

Therefore, equipped with the strict ordering ($G^+$ consisting of $0$ together with all the elements in $G$ that are bounded below away from zero on $T$), $G$ is a simple partially ordered group, and it is well known that dense subgroups of Aff of a Choquet simplex with respect to the strict ordering satisfy interpolation. Hence $G$ is a simple dimension group [using the result that a countable unperforated partially ordered group with interpolation is a dimension group, i.e., a limit of simpicial groups], and with respect to the order unit $u = 1$, $S(G,u)$ is naturally identified with $T$ (that is, the normalized traces of $G$ are the points of $T$).

It is easy to check that with $u = 1$ (the constant function), the map $(G,u) \to ({\rm Aff\ }S(G,u),1)$ given by $g \mapsto \hat g$ where $\hat g(s) = s(g)$ is equivalent to the inclusion $(G,1) \subset {\rm Aff } (T),1)$, and in such a way that $\hat f$ is sent to $f$.

More generally, we can do this for any countable family of $f$s, rather than just a single one.

For all of the unjustified statements here, see Goodearl's book on partially ordered abelian groups. The use of countable subgroups as here is a standard construction.

  • $\begingroup$ Thanks, David! I am happy to hear that this type of construction is known. I will sure go and check out Goodearl's book. $\endgroup$ – Gabor Szabo Nov 30 '17 at 7:33

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.