Kadison's problem for type III subfactors and the bicentralizer conjecture

Abstract

In 1967, Kadison asked "if $N$ is a subfactor of the factor $M$ for which $N^{\prime }\cap M$ consists of scalars, will some maximal abelian *-subalgebra of $N$ be a maximal abelian subalgebra of $M$?". Generalizing a theorem of Popa in the type $\mathrm{II}$ case (1981), we solve Kadison's problem for all subfactors with expectation $N\subset M$ where $N$ is either a type ${\mathrm{III}}_{\hat{I}}»$ factor with $0\le \hat{I}»<1$ or a type ${\mathrm{III}}_1$ factor that satisfies Connes's bicentralizer conjecture. Our solution is based on a new explicit formula for the bicentralizer algebras of arbitrary inclusions. This formula implies a type $\mathrm{III}$ analog of Popa's local quantization principle. We generalize Haaegrup's theorem from 1984 by connecting the relative bicentralizer conjecture to the Dixmier property. Finally, we prove this conjecture for a large class of inclusions and we prove an ergodicity theorem for the bicentralizer flow. We also give applications of our methods to ${\mathrm{II}}_1$ factors, including a new characterization of Ozawa's W*-Akemann-Ostrand property.

78 pages