Existentially closed \(\mathrm{W}^*\)-probability spaces (Q2152856)

This paper lies in the area of research of continuous model theory and its applications to operator algebras, and in particular von Neumann algebras. The study of tracial von Neumann algebras from a model-theoretic point of view was formalised in [\textit{I. Farah} et al., Isr. J. Math. 201, Part A, 477--505 (2014; Zbl 1301.03037)] and [\textit{I. Farah} et al., Bull. Lond. Math. Soc. 46, No. 3, 609--628 (2014; Zbl 1303.46049)]. Since then, the theory grew exponentially in the last decade, and it was shown that many key properties of tracial von Neumann algebras can be approached via model-theoretic methods and ideas (e.g., [\textit{I. Goldbring}, Proc. Am. Math. Soc. 148, No. 11, 5007--5012 (2020; Zbl 1456.03058); \textit{I. Goldbring} and \textit{B. Hart}, Int. Math. Res. Not. 2017, No. 18, 5609--5628 (2017; Zbl 1404.03038)], or the recent volume on continuous model theory and operator algebras to appear in the series `Logic and its applications'). Tracial von Neumann algebras can be thought as the finite objects in the theory (in the theory of von Neumann algebras, not in model theoretic sense!); here, ultraproducts are tracial ones. When trying to generalise this approach to a larger class of von Neumann algebras, \textit{Y. Dabrowski} [J. Funct. Anal. 277, No. 11, Article ID 108308, 82 p. (2019; Zbl 1444.46039)] introduced a proper formalisation of the model theory for \(\sigma\)-finite von Neumann algebra together with an associated faithful normal state (see also the unpublished `Correspondences, ultraproducts, and model theory' of Goldbring, Hart, and Sinclair). These objects are called W\(^*\)-probability space, and, in the realm of factors (von Neumann algebra with trivial center) correspond to the type III case. Here the relevant ultraproduct construction is the one of Ocneanu, obtained by considering the corona algebra of those sequences on which the specified states vanish on an ultrafilter. After reviewing the proper axiomatisation of W\(^*\)-probability spaces and the construction of the Ocneanu ultraproduct, the paper focuses on the study of existentially closed (e.c. from now on) W\(^*\)-probability spaces. In model theory, e.c. structures are those with the property that inclusions preserve the value of existential formulas with parameters in the small structure. Given \(\mathcal M\subseteq\mathcal N\), we say that \(\mathcal M\) is e.c. in \(\mathcal N\) if for every existential formula \(\varphi\) with parameters in \(M\) we have that \(\varphi^\mathcal M=\varphi^\mathcal N\). A structure \(\mathcal M\) is e.c. in a class of structures \(\mathcal K\) if \(\mathcal M\) is e.c. in all superstructures coming from \(\mathcal K\). The study of e.c. structures in crucial in model theory, as it is often linked with Fraïssé theoretic constructions (and consequently quantifier elimination), model-theoretic forcing, and the existence of model companions among others. The key results of the paper are contained in \S3. Here, the authors prove many interesting results on e.c. W\(^*\)-probability, some in parallel to those obtained for e.c. II\(_1\) factors (in the tracial setting), contained in [\textit{I. Farah} et al., Fundam. Math. 233, No. 2, 173--196 (2016; Zbl 1436.03213)]. To name a few, the authors show that \begin{itemize} \item being e.c. (in the class of all W\(^*\)-probability spaces) does not depend on the choice of the faithful normal state chosen to make a \(\sigma\)-finite von Neumann algebra a W\(^*\)-probability space. \item e.c. models must be factors of type III\(_1\), and, again similarly to the II\(_1\) case, they have to tensorially absorb \emph{the} type III\(_1\) factor of Araki and Woods, \(\mathcal R_\infty\). \item e.c. model have only approximately inner automorphisms. \end{itemize} These, together with the fact that Ocneanu ultraproducts cannot tensorially absorb infinite dimensional factors, give that e.c. W\(^*\)-probability spaces are not axiomatisable, and therefore the theory does not have a model companion. Further in \S3, the focus shifts on QWEP factors and W\(^*\)-probability spaces. The QWEP of Kirchberg asks to be a quotient of an algebra with the weak expectation property, and in this setting having the QWEP can be rephrased in terms of satisfying a certain chunk of the theory of the Araki-Woods factor. Notably the authors show that \(\mathcal R_\infty\) is an e.c. QWEP W\(^*\)-probability space, and they link results on QWEP e.c. factors and W\(^*\)-probability spaces with results on enforceable (in the sense of model-theoretic forcing) objects. Leaving some open questions at the end of \S3, the paper ends with other axiomatisability results for certain classes of W\(^*\)-probability spaces, computing the quantifier depth of the theories axiomatising for example type III\(_1\) and type III\(_\lambda\) factors.