No label defined (Q4034802)

[For the entire collection see Zbl 0745.00015.] The authors offer very interesting extensions of notions and results introduced by \textit{S. Shelah} [Classification theory and the number of nonisomorphic models (1978; Zbl 0388.03009)] and by \textit{B. Poizat} [J. Symb. Logic 48, 239-249 (1983; Zbl 0525.03023)]. They study \(P\)-stability [the first author, ``A series of stability notions'' (Russian), 10th Soviet Conf. Math. Logic, Alma-Ata, Publ. IMM AN Kazach. Sov. Rep. (1990), p. 121], a principal generalization obtained from the classical concept of stability by extending the language by a new unary predicate \(P\). This predicate is intended to characterize a proper elementary submodel. Definitions: Let \(T\) be a complete first-order theory with infinite models, \(\sigma\) its signature and \(\mathfrak C\) its monster-model. Let \(A\) be a subset of \(\mathfrak C\) and \(P\) a new unary predicate symbol. The theory \(T_ P(A)\) with the signature \(\sigma_ P(A)=\sigma_ A\cup\{P\}\) is given by the following union \[ T_ P(A)=\text{Th}({\mathfrak C},\alpha)_{\alpha\in A}\cup\bigl\{P(c_ \alpha)\mid\alpha\in A\bigr\}\cup\{\text{``}P\prec\text{''}\}, \] where \(\{\text{``}P\prec\text{''}\}\) is an infinite set of formulas expressing the fact that the interpretation of \(P\) is the elementary submodel in the signature \(\sigma\). In general, the resulting theory \(T_ P(A)\) is not complete -- let \(S_ P(A)\) denote the set of all possible completions of the theory \(T_ P(A)\). The theory \(T\) is said to be \(P\)-\(\lambda\)- stable, if there holds for any set \(A\) of cardinality \(\lambda\) that \(| S_ P(A)|\leq\lambda\). The theory \(T\) is \(P\)-stable, if it is \(P\)-\(\lambda\)-stable for some \(\lambda\). The theory \(T\) is \(P\)- superstable, if there is a cardinal \(\lambda_ 0\) such that \(T\) is \(P\)- \(\lambda\)-stable for any cardinal larger than \(\lambda_ 0\), i.e. for any \(\lambda\geq \lambda_ 0\). The theory \(T\) is \(P\)-nonstable, if there is no cardinal \(\lambda\) s.t. \(T\) is \(P\)-\(\lambda\)-stable. The paper studies the introduced notions. It gives full proofs of the corresponding theorems. The basic results obtained are summarized as follows. Theorem: Let \(T\) be a superstable theory. Then the claims a)--d) on the theory \(T\) are equivalent: a) \(T\) is not multidimensional, b) \(T\) is \(P\)- superstable, c) \(T\) is \(P\)-stable, d) \(T\) does not have the property of \(P\)-independence. The special case of stable theories with countable signature is treated at the end of the paper. The attention is shifted towards saturated models. Theorem: Let \(T\) be a \(\lambda\)-stable theory with \(\sigma\) a countable signature, \(M\precneqq N\) any elementary pair of models of \(T\) and \(\lambda\) any regular cardinal such that \(\lambda\geq| N|\). Then there exists an elementary extension \((N',M')\) of the pair \((N,M)\) in the language of the theory of pairs \(\sigma_ P=\sigma\cup \{P\}\) meeting the following conditions: 1) \(| N'|=| M'|=\lambda\), 2) models \(M'\) and \(N'\) are saturated (in the signature \(\sigma\)), 3) for any subset \(A\subset M'\), \(| A|<\lambda\), if the type \(p\in S(A)\) and \(p\cup\{\neg P\}\) is finitely satisfiable in the pair \((N',M')\), then \(p\) is true in \(N'\backslash M'\).