I think the submodel (elementary or arbitrary) of an atomic model may not be atomic. But I do not know an example.
Is submodel of atomic model atomic
119 Views Asked by anon https://math.techqa.club/user/anon/detail AtThere are 2 best solutions below
On
Noah had shown that (1) every elementary substructure of an atomic model is atomic, (2) not every substructure of an atomic model is atomic.
One might ask: What about a substructure of an atomic model which is elementarily equivalent to it? That is, for a complete theory $T$, can a non-atomic model of $T$ embed (non-elementarily) in an atomic model of $T$?
The answer is yes. Consider the theory $T$ of forests (simple graphs with no cycles) augmented by axioms asserting that every vertex has infinitely many neighbors. This theory is complete.
A countable model of $T$ consists of a disjoint union of copies of the complete countably branching tree (at least one copy and at most countably many copies). The atomic model $M$ is the connected one (just one copy of the tree). Now pick an arbitrary vertex $v\in M$ and remove it (and all edges out of it). The resulting graph $N$ is disconnected - each of the neighbors of $v$ is now in its own connected component, and each connected component is a copy of the complete countably branching tree.
So $N$ is a non-atomic model of $T$ which is a substructure of the atomic model $M$.
Recall for readers that an atomic model is a structure such that each type realized in the structure is generated by a single formula.
Suppose $M\preccurlyeq N$, $N$ is atomic, and $m\in M$. By elementarity we have $tp^M(m)=tp^N(m)$, and by atomicity of $N$ we have that $tp^N(m)$ is generated by a single formula $\varphi$. Putting these together we get that $\varphi$ generates the type of $m$ in $M$ as well. So elementary submodels of atomic models are atomic.
If we look merely at substructures, however, everything breaks down immediately. It's a good exercise to show that every countable graph is an induced subgraph of an atomic countable graph. (HINT: think about gluing the starting graph to a copy of $\mathbb{N}$.)