Pseudo real closed fields, pseudo \(p\)-adically closed fields and \(\mathrm{NTP}_{2}\) (Q331056)

The main result of this paper is a positive answer to (the first part of) the following conjecture by \textit{A. Chernikov} et al. [Proc. Am. Math. Soc. 143, No. 1, 395--406 (2015; Zbl 1386.03036)]: ``A PRC field is $\mathrm{NTP}_2$ if and only if it is bounded. Similarly, a P$p$C field is $\mathrm{NTP}_2$ if and only if it is bounded.'' Recall that the class of $\mathrm{NTP}_2$ theories (theories without tree properties of the second kind) contains both simple and NIP theories. A bounded (i.e. possessing finitely many algebraic extensions of any given degree) pseudo-real closed field $K$ has only finitely many possible orderings (Remark 3.2). If the set of these ordering is not empty, we denote them by $<_1, \dots <_n$. We are then led to study PRC fields in the language $\mathcal{L}_n := \mathcal{L} \cup \{<_1, \dots, <_n\}$, where $\mathcal{L}$ is the language of rings $\mathcal{L}_{\mathcal{R}} := \{+,-, \cdot, 0,1 \}$ together with constant symbols for elements of $K_0$, where $K_0 \prec K$ is a countable elementary substructure of $K$. The author proves his result using the approximation theorem, which essentially states that open sets in different topologies (induced by distinct orderings) on a field $F$ have non-empty intersection. An intersection of $<_i$-intervals (for $i=1, \dots, n$) is called a multi-interval. Then they proceed to deduce a ``density theorem'': a definable set is multi-dense in a finite union of multi-intervals and points. A generalization to several variables definable sets is given, in which one encounters multi-cells (and multi-boxes). Observe that if $M$ is PRC, then $M(\sqrt{-1})$ is a PAC (pseudo algebraically closed field). Also a PRC field with no orderings is a PAC field. This observation allows to transfer many results already obtained for PAC fields to the setting of PRC fields. For instance, it is shown that if $M$ is a PRC field which is not real closed neither algebraically closed then $\mathrm{Th}_{\mathcal{R}}(M)$ has the independence property (Theorem 4.10). Using the above results in addition to an amalgamation theorem for PRC fields (Theorem 3.21 and Corollary 3.22) it is then possible to show that if $M$ is a PRC field with exactly $n$ orders, then $\mathrm{Th}(M)$ is strong of burden $n$, hence $\mathrm{NTP}_2$, (Theorem 4.23), and resilient (Theorem 4.30). The burden of complete types is computed explicitly (Theorem 4.26), and it is then shown that forking equals dividing (Theorem 4.37). On the other hand, if $M$ is an unbounded PRC field then $\mathrm{Th}(M)$ is $\mathrm{TP}_2$ and is not rosy (Theorem 4.23 and Corollary 4.20). The case of a P$p$C field is handled in a similar way. A bounded pseudo $p$-adically closed field $K$ (which is not $p$-adically closed) has finitely many $p$-adic valuations (Lemma 6.1) denoted by $v_1, \dots, v_n$. Fix a bounded P$p$C field $K$, and a countable elementary substructure $K_0$ of $K$. Work with the language \[ \mathcal{L}_n := \mathcal{L} \cup \{O_i\}_{i \leq n} \cup \{P_m^i\}_{i \leq n, m \in \mathbb{N}, m >1}, \] with $\mathcal{L}$ being the language of rings together with constant symbols for the elements of $K_0$, and $O_i$ ($P^i_m$) being unary relation symbols interpreted as the valuation ring associated to the valuation $v_i$ (the set of $m$-th powers in a $p$-adic closure of $K$ with respect to $v_i$). It is shown that the $\mathcal{L}$-theory of a bounded P$p$C field which is not $p$-adically closed is not NIP (Corollary 7.4). Also, if $M$ is a bounded P$p$C field, then $\mathrm{Th}(M)$ is strong, resilient and of finite burden (Theorem 8.5) hence $\mathrm{NTP}_2$ (Corollary 8.6). The converse implication does not follow using these methods, however, since algebraic extensions of P$p$C fields are not necessarily P$p$C. Also, similarly to the case of bounded PRC fields, it is shown that in $\mathrm{Th}(M)$ forking equals dividing (Theorem 8.9).