Smooth classes without AC and Robinson theories (Q4451719)

Let \(L=\{R_1,\dots,R_n\}\) be a finite relational language, and \(K\) a class of finite \(L\)-structures closed under isomorphisms and substructures, with \(\emptyset\in K\). The class \(K\) with a partial order \(\leq\) on \(K\) is called smooth if, for all \(A\), \(B\), \(C\in K\), (i) if \(A\leq B\) then \(A\subseteq B\), (ii) \(\emptyset\leq K\), (iii) if \(A\), \(B\subseteq C\) and \(A\leq C\) then \(A\cap B\leq B\). Clearly, the class of \(L\)-structures all of whose substructures are in \(K\) is first-order axiomatizable by a universal theory \(T\). If a smooth class \((K,\leq)\) has the amalgamation property (AP), the Fraïssé construction provides a unique countable model of \(T\), the \((K,\leq)\)-generic model. In the late 80's, E.~Hrushovski discovered a seminal method of constructing various smooth classes \((K,\leq)\) with stable \((K,\leq)\)-generic models. The author considers the following version of the Hrushovski construction. For a finite \(L\)-structure \(A\), let \(\delta(A)=| A| -\sum^n_{i=1}\alpha_i| R^A_i| \), where \(\alpha_1,\dots,\alpha_n\) are fixed real numbers. Let \(K\) be the class of \(A\) such that \(\delta(B)>0\) for all nonempty \(B\subseteq A\). For \(A\), \(B\in K\), let \(A\leq B\) mean that \(A\subseteq B\), and \(\delta(C)>\delta (A)\) whenever \(A\subset C\subseteq B\). Then \((K,\leq)\) turns out to be a smooth class with the AP. However, as the author shows, the class does not satisfy the algebraic closure property, a certain finiteness condition considered by Hrushovski and important in proofs of stability of generic models. Nevertheless, here some good theory of forking can be developed. For any smooth class \((K,\leq)\), \textit{E. Hrushovski} [Simplicity and the Lascar group, manuscript (1997)] considered an inductive theory \(T_{\text{nat}}\), which is a definitional extension of \(T\) such that on the finite substructures of a model \(M\) of \(T\) the relation \(\leq\) coincides with the substructure relation in the \(T_{\text{nat}}\)-expansion of \(M\); so the \(T_{\text{nat}}\)-expansion of the \((K,\leq)\)-generic model is an existentially closed model of \(T_{\text{nat}}\). From now on, let \(T_{\text{nat}}\) be the theory corresponding to the class \((K,\leq)\) defined above in terms of \(\delta\). The author shows that the class of existentially closed models of \(T_{\text{nat}}\) is not elementary (contrary to what Hrushovski expected), but \(T_{\text{nat}}\) is a Robinson theory (which means that its universal part has the AP). Hrushovski pointed out that model-theoretic methods developed for complete first-order theories should apply to the category of existentially closed models of Robinson theories, where quantifier-free types play the role of types; in particular, forking and simplicity can be defined in this context. This idea was developed by \textit{A.~Pillay} [``Forking in the category of existentially closed structures'', Quad. Mat. 6, 23--42 (2000; Zbl 0986.03031)]. Applying this approach to \(T_{\text{nat}}\) and analyzing the dependence relation on existential models of \(T_{\text{nat}}\) that comes from \(\delta\), the author shows that the class of existential models is simple. The proof is a modification and completion of Hrushovski's arguments in his paper cited above.