Externally definable sets and dependent pairs (Q1955748): Difference between revisions
From MaRDI portal
Latest revision as of 12:44, 6 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Externally definable sets and dependent pairs |
scientific article |
Statements
Externally definable sets and dependent pairs (English)
0 references
18 June 2013
0 references
The paper under review studies externally definable sets in NIP theories as well as pair structures \((M,A)\), where \(M\) is a (usually NIP) structure in some language \(L\) and \(A\subseteq M\) a subset named by a new predicate \({\mathbf P}(x)\). The resulting language is denoted by \(L_{\mathbf P}\), and \(T_{\mathbf P}=\mathrm{Th}(M,A)\). A set \(X\subseteq M^k\) is called \textit{externally definable} if \(X\) is the trace on \(M\) of a definable set in some elementary extension, i.e.\ if there is a formula \(\phi(x,d)\) (\(d\in N\succ M\)) such that \(X=\phi(M,d)\). The authors call \(\phi(x,d)\) an \textit{honest definition} for \(X\) if \(X=\phi(M,d)\), and for every formula \(\psi(x,e)\) over \(M\) with \(X\subseteq\psi(M,e)\), the formula \(\psi(x,e)\) is implied by \(\phi(x,d)\). They show that honest definitions exist in any NIP theory. As an immediate corollary, one obtains \textit{S. Shelah}'s expansion theorem [Isr. J. Math. 173, 1--60 (2009; Zbl 1195.03040)] which states that the expansion \(M^{\mathrm{Sh}}\) of an NIP structure \(M\) by predicates for all externally definable subsets admits quantifier elimination, so in particular is NIP. Moreover, they show that any subset \(A\) of an NIP structure is \textit{weakly stably embedded}: if \(\phi(x)\) is an \(L(M)\)-formula, then there is an elementary extension \((M',A')\succ(M,A)\) and an \(L(A')\)-formula \(\theta(x)\) such that \(\theta(A)=\phi(A)\) and \(\theta(A')\subseteq\phi(A')\). Since the paper has appeared, it has already become clear that honest definitions are a fundamental tool in the study of NIP theories. E.g., in a sequel to the paper under review [the authors, Trans. Am. Math. Soc. 367, No. 7, 5217--5235 (2015, Zbl 1388.03035)], the authors show that in NIP theories types are uniformly definable over finite sets (UDTFS), and that UDTFS characterises NIP. Call an \(L_{\mathbf P}\)-formula \textit{bounded} if it is of the form \(Q_1 y_1\in{\mathbf P}\ldots Q_n y_n\in{\mathbf P}\phi(x,\overline{y})\) for some \(L\)-formula \(\phi\) and quantifiers \(Q_i\in\{\exists,\forall\}\). Generalising work of \textit{E. Casanovas} and \textit{M. Ziegler} [J. Symb. Log. 66, No.3, 1127-1140 (2001; Zbl 1002.03023)] and others, the authors obtain the following very general results about pairs \((M,A)\) for \(M\) an NIP structure. If the induced structure on \(A\) is NIP and every \(L_{\mathbf P}\)-formula is equivalent in \(T_{\mathbf P}\) to a bounded formula, then \(T_{\mathbf P}\) is NIP. In a final section, some examples are treated, and it is shown that numerous previous results on NIP for pair structures are subsumed in the aforementioned general result. Moreover, the authors prove a conjecture of \textit{J. Baldwin} and \textit{M. Benedikt} [Trans. Am. Math. Soc. 352, No.11, 4937--4969 (2000; Zbl 0960.03027)], stating that if \(I\) is a small indiscernible sequence with underlying dense complete linear order in an NIP structure \(M\), then the pair \((M,I)\) is NIP. It follows that any NIP theory (having infinite models) admits an NIP expansion in which an infinite definable set admits a definable linear order. The paper is overall very well written.
0 references
model theory
0 references
NIP theories
0 references
externally definable sets
0 references
honest definitions
0 references