Annals of Mathematics · 2010 · 259 citations · 24 references
Natural SciencesMathematicsAlgebraProbability TheoryStochastic ProcessesAbstract AlgebraErgodic TheoryGroup TheoryAmenable Group ActionsCartan SubalgebrasComplete Metric SpacesErgodic Actions Of GroupsErgodic MeasuresErgodic PropertiesProfinite GroupsVon Neumann Algebras
We prove that the normalizer of any diffuse amenable subalgebra of a free group factor L.ކ r / generates an amenable von Neumann subalgebra.Moreover, any II 1 factor of the form Q x ˝L.ކ r /, with Q an arbitrary subfactor of a tensor product of free group factors, has no Cartan subalgebras.We also prove that if a free ergodic measure-preserving action of a free group ކ r , 2 Ä r Ä 1, on a probability space .X; / is profinite then the group measure space factor L 1 .X / Ì ކ r has unique Cartan subalgebra, up to unitary conjugacy.If we take Q to be an arbitrary finite factor with ƒ cb .Q/ D 1 and let ކ r act trivially on it, then1 and the theorem implies: COROLLARY 2. If Q is a II 1 factor with the complete metric approximation property then Q x ˝L.ކ r / does not have Cartan subalgebras.Moreover, if N Q x ˝L.ކ r / is a subfactor of finite index [Jon83], then N does not have Cartan subalgebras either.This shows in particular that any factor of the form L.ކ r / x ˝R, L.ކ r 1 / x L.ކ r 2 / x ˝ , and more generally any subfactor of finite index of such a factor, has no Cartan decomposition.Besides Q D R; L.ކ r /, other examples of factors with ƒ cb . Q D 1 are the group factors L./ corresponding to ICC discrete subgroups of SO.1; n/ and SU.1; n/ ([DCH85], [Cow83]), as well as any subfactor of a tensor product of such factors.None of the factors covered by Corollary 2 were known until now not to have Cartan decomposition.Finally, if we take ކ r Õ X to be a profinite m.p. action on a probability measure space .X; /, i.e. an action with the property that L 1 .X / is a limit of an increasing sequence of ކ r -invariant finite-dimensional subalgebras Q n L 1 .X /, then M D L 1 .X / Ì ކ r is an increasing limit of the algebras Q n Ì ކ r , each one of which is an amplification of L.ކ r /.Since c.m.a.p. behaves well to amplifications and inductive limits, it follows that M has c.m.a.p., so by applying the theorem and (A.1 in [Pop06a]) we get
24