It's well known that a countable theory is uncountably categorical if and only if it is $\omega$-stable and has no Vaughtian pairs. One of the old definitions of unidimensional theory (all $\omega_1$-saturated models of the same (sufficiently large) size are isomorphic) is a weak categoricity notion, so it's natural to wonder if an analogous characterization can be made.

Hrushovski showed that unidimensional theories are supserstable. Although I'm sure it's already known, I couldn't find a proof that unidimensional theories do not have any Vaughtian pairs, so I worked out a silly proof:

Let $T$ be a countable, superstable theory with a Vaughtian pair. By Vaught's two-cardinal theorem, since $T$ has a Vaughtian pair, it has a $(\omega_1,\omega)$-model. By a result of Shelah, in a stable theory the existence of any two-cardinal model implies the existence of $(\kappa,\lambda)$-models for any $\kappa\geq\lambda\geq\omega$, so we can find a $(\kappa,\omega)$-model, $\mathfrak{M}$, of $T$, with $\kappa \gg 2^\omega$. Then if we take any non-principal ultrafilter $\mathcal{U}$ on $\omega$, $\mathfrak{M}^\omega/\mathcal{U}$ will still be a two-cardinal, $\omega_1$-saturated model of $T$, contradicting unidimensionality.

Again, I'm sure there's a better, published proof somewhere, but I couldn't find it.

I would guess it's far too much to hope that a countable theory $T$ is unidimensional if and only if it is superstable and has no Vaughtian pairs, but I also can't quite find a counterexample. So my first question is:

What is an example of a countable superstable theory with no Vaughtian pairs that fails to be unidimensional, if such a thing exists? EDIT: I believe theory of the following structure is a counterexample to this: $(2^\omega, P_i, f_i)$ where $P_i$ is a sequence of unary predicates such that $P_i(\alpha)$ is true if and only if $\alpha(i)=1$, and $f_i$ is a sequence of unary functions such that $f_i(\alpha)(j)=\alpha(j)$ if $i\neq j$ and $f_i(\alpha)(i)=1-\alpha(i)$.

Another related characterization of uncountably categorical countable theories is that a countable theory $T$ is uncountably categorical if and only if there is a definable strongly minimal set $\varphi$ such that the entire structure is layered by $\varphi$ (where layering is in the sense of Hodges' "Model Theory"). In Pillay's "Geometric Stability Theory," he shows in chapter 6 proposition 1.1 that if $T$ is a countable, superstable, unidimensional theory that fails to be $\omega$-stable, then there is a weakly minimal group definable over $\text{acl}(\varnothing)$. Since we know that $T$ doesn't have any Vaughtian pairs, by the results on layerings in Hodges we know that the entire theory is layered over this weakly minimal group.

I suspect that it's also way too much to hope that any weakly minimal group is unidimensional, so my next questions are:

What is an example of a weakly minimal group that fails to be unidimensional? Is there a good characterization of which weakly minimal groups are unidimensional?

Aside from those parts, I'm pretty confident that if a theory is layered over a unidimensional set, then it will be unidimensional, so if there is a good characterization of which weakly minimal groups are unidimensional, then I'm expecting there is a statement like: "A theory $T$ is unidimensional if and only if it is layered over a strongly minimal set or a weakly minimal group with property X." Where 'property X' is the missing ingredient to ensure that a weakly minimal group is unidimensional.


1 Answer 1


I believe the answer to the second question is that all weakly minimal groups are unidimensional.

Proof: Let $T$ be the theory of a weakly minimal group. It suffices to show that if $p$ and $q$ are non-algebraic $1$-types over a sufficiently saturated model $G$, then $p$ and $q$ are not orthogonal. So fix such $p$ and $q$. By weak minimality, $p$ and $q$ have $U$-rank $1$, and thus are generic. By stability, there is some $g\in G$ such that $q=gp$. So $p$ and $q$ are not orthogonal.

Edit: The fact that stable unidimensional theories have no Vaughtian pairs is in Lemma IX.1.10 of Shelah's book.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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