Model theoretic characterizations of large cardinals (Q2182021)

This outstanding paper is about a variation of compactness to characterize large cardinals. Classical results on large cardinals and inifinitary logic handle satisfaction of sets of formulas (theories). This is the case for weakly compact, strongly compact, and even for measurable, cardinals. Of course, we can consider other logics, like second-order, infinitary second-order, etc. \textit{J. A. Makowsky} [J. Symb. Log. 50, 42--48 (1985; Zbl 0623.03041)] proved that the Vopěnka principle is equivalent to the existence of a compact cardinal for every logic. Since this seems to be an upper bound for such characterizations of large cardinals, the author looks for a new kind of compactness which allows us to incorporate other large cardinals. A relevant result of \textit{M. Benda} [Ann. Math. Logic 14, 39--56 (1978; Zbl 0384.03037)] provides an idea in this direction, namely a representation for supercompact cardinals in terms of type omission in \(L_{\kappa\kappa}\)-logic. Given any \(L_{\kappa\kappa}\)-theory \(T\) and a type \(\Sigma(x,y)\) in two variables (but the result generalizes to finitely many variables) Benda [loc. cit.] seeks a model of \[ \exists x[\bigwedge\exists y\Sigma(x,y)\wedge\neg\exists y\bigwedge\Sigma(x,y)] \] which ``realizes and omits the type'', because such a model contains an element \(a\) such that \(\{\exists y\sigma(a,y):\sigma\in\Sigma\}\) is satisfiable but omits the type \(\Sigma(a,y)\) in \(y\). Since this theorem is the motivation for the paper, it is worth to formally describe all these issues. Let \(\kappa,\lambda\) be cardinals with \(\kappa\leq\lambda\). The set \(\mathcal{P}_{\kappa}(\lambda)\) consists of all subsets of \(\lambda\) of size \(<\kappa\). We have a corresponding notion of a closed set in \(\mathcal{P}_{\kappa}(\lambda)\). Benda's theorem [loc. cit.] is: \(\kappa\) is \(\lambda\)-supercompact iff for every \(L_{\kappa\kappa}\)-theory \(T\) and type \(\Sigma(x,y)=\{\varphi_\alpha(x,y):\alpha<\lambda\}\), if the set of \(\Delta\in \mathcal{P}_{\kappa}(\lambda)\) such that \[ T\cup\left\{\exists x\left(\bigwedge_{\alpha\in\Delta}\exists y\varphi_\alpha(x,y)\wedge\neg\exists y\bigwedge_{\alpha\in \Delta}\varphi_\alpha(x,y)\right)\right\} \] has a model is a club in \(\mathcal{P}_{\kappa}(\lambda)\), then there is a model of \[ T\cup\left\{\exists x\left(\bigwedge_{\alpha\in\lambda}\exists y\varphi_\alpha(x,y)\wedge\neg\exists y\bigwedge_{\alpha\in \lambda}\varphi_\alpha(x,y)\right)\right\} \] Notice that the last model omits the type \(\Sigma(a,y)\), for some element \(a\) in it. This is not a result about compactness, but about omitting type compactness, which the author calls compactness for omitting types. Benda's theorem [loc. cit.] says that if enough small parts of a type can be omitted, then the whole type can be omitted. In order to achieve new results, the author looks at other index sets than \(\mathcal{P}_{\kappa}(\lambda)\) and allows to filter the theory, that is, the theory is the union of an increasing sequence of theories. The author discusses several logics, besides infinitary, like second order, \(L^2\), infinitary second order, \(L_{\kappa\kappa}^2\), sort logic, \(L^s\) intoduced by \textit{J. Väänänen} [in: Logic colloquium '78. Proceedings of the colloquium held in Mons, August 1978. Amsterdam, New York, Oxford: North-Holland Publishing Company. 391--421 (1979; Zbl 0433.03019)], and variations of them. The new kind of index sets are called robust (\(\kappa\)-robust on \(\mathcal{P}_{\kappa}(\lambda)\)). He introduces a convenient notion of club: strong \(\kappa\)-club. His first result establishes a link between what he called \(I\)-\(\kappa\)-compactness for type omission, kind of \(\kappa\)-complete ultrafilters and elementary embeddings of the universe. With this, he provides a characterizartion for measurable cardinals in terms of the \(\mathcal{P}_{\kappa}(\lambda)\)-\(\kappa\)-compactness for type omission of \(L_{\kappa\kappa}\). Stronger results are obtained in sections devoted to second-order logic, using strong cardinals, sort logic and variations of them. There is also a section on elementary substructures of second-order logic, where a definition of elementary substructure is given. This definition includes a stronger notion of substructures in the second-order context. The authors set out a correspondig Tarski-Vaught test and generalizations for the notion of club. In the final section, there are results about extenders for strong cardinals and compactnes for type omission. This is a well-written paper easy to follow, when the reader is acquainted with infnitary logic, its model theory and large cardinals.