arXiv subjects
Publications and source records attributed to Danielle Ulrich.
We characterize having Borel isomorphism relation among some weakly minimal trivial theories, namely the examples of families of finite equivalence relations from recent joint work with Laskowski, and tame expansions of crosscutting-equivalence relations. We also prove a dichotomy in Borel complexity for the latter.
Malliaris and Shelah famously proved that Keisler's order $\trianglelefteq$ has infinitely many classes. In more detail, for each $2 \leq k < n < \omega$, let $T_{n, k}$ be the theory of the random $k$-ary $n$-clique free hypergraph. Malliaris and Shelah show that whenever $k+1 < k'$, then $T_{k+1, k} \not \trianglelefteq T_{k'+1, k'}$. However, their arguments do not separate $T_{k+1, k}$ from $T_{k+2, k+1}$, and the model-theoretic properties detected by their ultrafilters are difficult to evaluate in practice. We uniformize the relevant ultrafilter constructions and obtain sharper model-theoretic bounds. As a sample application, we prove the following: suppose $3 \leq k < \aleph_0$, and $T$ is a countable low theory. Suppose that every independent system $(M_s: s \subsetneq k)$ of countable models of $T$ can be independently amalgamated. Then $T_{k, k-1} \not \trianglelefteq T$. In particular, for all $k < k'$, $T_{k+1, k} \not \trianglelefteq T_{k'+1, k'}$.
We introduce a new invariant of Borel reducibility, namely the notion of thickness; this associates to every sentence $\Phi$ of $\mathcal{L}_{\omega_1 \omega}$ and to every cardinal $\lambda$, the thickness $\tau(\Phi, \lambda)$ of $\Phi$ at $\lambda$. As applications, we show that all the Friedman-Stanley jumps of torsion abelian groups are non-Borel complete. We also show that under the existence of large cardinals, if $\Phi$ is a sentence of $\mathcal{L}_{\omega_1 \omega}$ with the Schr\"{o}der-Bernstein property (that is, whenever two countable models of $\Phi$ are biembeddable, then they are isomorphic), then $\Phi$ is not Borel complete.
Building off of recent results on Keisler's order, we show that consistently, $\leq_{SP}$ has infinitely many classes. In particular, we define the property of $\leq k$-type amalgamation for simple theories, for each $2 \leq k < \omega$. If we let $T_{n, k}$ be the theory of the random $k$-ary, $n$-clique free random hyper-graph, then $T_{n, k}$ has $\leq k-1$-type amalgamation but not $\leq k$-type amalgamation. We show that consistently, if $T$ has $\leq k$-type amalgamation then $T_{k+1, k} \not \leq_{SP} T$, thus producing infinitely many $\leq_{SP}$-classes. The same construction gives a simplified proof of Shelah's theorem that consistently, the maximal $\leq_{SP}$-class is exactly the class of unsimple theories. Finally, we show that consistently, if $T$ has $<\aleph_0$-type amalgamation, then $T \leq_{SP} T_{rg}$, the theory of the random graph.