Factorial relative commutants and the generalized Jung property for \(\mathrm{II}_1\) factors (Q2071628): Difference between revisions

This paper develops and uses techniques from continuous model theory to approach questions about embeddings of II\(_1\)-factors (certain von Neumann algebras with trivial center) in their ultrapowers. Unarguably, the most important II\(_1\) factor is the hyperfinite II\(_1\) factor \(\mathcal R\), the only separable object among amenable II\(_1\) factors. A key property of the \(\mathcal R\) is that any two embeddings of \(\mathcal R\) into its ultrapower \(\mathcal R^\mathcal U\) are unitarily equivalent. Indeed, a striking result of \textit{K. Jung} [Math. Ann. 338, No. 1, 241--248 (2007; Zbl 1121.46052)] gives that \(\mathcal R\) is the only II\(_1\)-factor with this property, that is, if a II\(_1\)-factor \(N\) has the property that all embeddings of \(N\) into \(\mathcal R^\mathcal U\) are unitarily equivalent, then \(N\cong \mathcal R\). The main focus of the present paper is a generalisation of the equivalence relation considered in Jung's theorem: Instead of focusing on automorphisms modulo unitarily equivalence, one focuses on the equivalence relations of `being conjugated by an automorphisms of the range structure'. To formalise this, the authors introduce the definitions of Jung pair and generalised Jung pair. In their terminology, a pair of II\(_1\)-factors \((N,M)\) is a Jung pair if all embeddings of \(N\) into \(M^\mathcal U\) are unitarily equivalent, while the generalised notion asks for equality modulo an automorphism of \(M^\mathcal U\). In this setting, Jung's theorem asserts that if \((N,\mathcal R)\) is a Jung pair, then \(N\cong \mathcal R\). The authors prove the same for generalised Jung pairs, that is, that the only II\(_1\) factor \(N\) with the property that all embeddings of \(N\) into \(\mathcal R^\mathcal U\) are equivalent modulo an automorphism of \(\mathcal R^\mathcal U\) is \(\mathcal R\). Next, attention is paid to non-\(\mathcal R^\mathcal U\)-embeddable II\(_1\)-factors. A II\(_1\)-factor which cannot be embedded in an ultrapower of \(\mathcal R\) is said to be non-\(\mathcal R^\mathcal U\)-embeddable. Such II\(_1\)-factors exist by the recent negative answer to the Connes Embedding Problem [\textit{Z.-F. Ji} et al., ``\textsf{MIP}\(^*\)=\textsf{RE}'', Preprint (2020), \url{arxiv:2001.04383}]. The existence of a non-\(\mathcal R^\mathcal U\)-embeddable II\(_1\)-factor \(M\) such that \((M,M)\) forms a generalised Jung pair is proved in \S3. This gives examples of II\(_1\)-factors which, although being non-\(\mathcal R^\mathcal U\)-embeddable, resemble one of the key property of \(\mathcal R\) itself. Further, the authors study factorial embeddings. (An embedding of \(N\) into \(\mathcal M^{\mathcal U}\) is factorial if the relative commutant of the image of \(N\) is a factor itself.) Popa asked whether all \(\mathcal R^\mathcal U\)-embeddable factors admit a factorial embedding into \(\mathcal R^\mathcal U\). While \(\mathcal R\) is the only II\(_1\)-factor such that \emph{all} embeddings in \(\mathcal R^\mathcal U\) are factorial (by a result of \textit{N. P. Brown} [Adv. Math. 227, No. 4, 1665--1699 (2011; Zbl 1229.46041)]), there were few examples so far of II\(_1\)-factors which are \(\mathcal R^\mathcal I\)-embeddable and admit a factorial embedding into \(\mathcal R^\mathcal U\). The authors add to these, by showing that every II\(_1\)-factor which is elementary equivalent to \(\mathcal R\) has a factorial embedding into \(\mathcal R^\mathcal U\) (stronger than this, they prove that all elementary embeddings \(N\to\mathcal R^\mathcal U\) are factorial). This gives continuum many nonisomorphic II\(_1\)-factors admitting factorial embeddings into \(\mathcal R^\mathcal U\). (The study of factorial embeddings and Popa's question from a model theoretic point of view was already initiated in [\textit{I. Goldbring}, Proc. Am. Math. Soc. 148, No. 11, 5007--5012 (2020; Zbl 1456.03058)], starting from ideas of [\textit{I. Farah} et al., Fundam. Math. 233, No. 2, 173--196 (2016; Zbl 1436.03213)].) The study of (generalised) Jung pairs is done via the study of the structure of all homomorphisms of a II\(_1\)-factor \(N\) into the ultrapower of a second II\(_1\)-factor \(M\), modulo automorphism equivalence. This set, denoted \(\mathrm{Hom}_A(N,M^\mathcal U)\), is studied from a model theoretic point of view, by associating it to a certain space of types in \S4 and studying its structure as a topometric space, following [\textit{I. Ben Yaacov}, Log. Anal. 1, No. 3--4, 235--272 (2008; Zbl 1180.03040)].