We attempt presenting a notion of the Haagerup approximation property for an arbitrary von Neumann algebra by using its standard form. We also prove the expected heredity results for this property.Ruan and Q. Xu in [JRX] in terms of non-commutative L p -spaces. In particular, the non-commutative L 2 -spaces become standard forms, and hence their result is a generalization of her characterization of semidiscrete von Neumann algebras. Therefore it immediately follows that the injectivity implies the Haagerup approximation property in our sense.The Haagerup approximation property has various stabilities. Among them, we will prove the following result (Theorem 5.9): Theorem B. Let N ⊂ M be an inclusion of von Neumann algebras. Suppose that there exists a norm one projection from M onto N. If M has the Haagerup approximation property, then so does N. In [CS], M. Caspers and A. Skalski independently introduce the notion of the Haagerup approximation property. Our formulation actually coincides with theirs because in either case, the Haagerup approximation property is preserved under taking the crossed products by R-actions. (See Remark 5.8.)This paper is organized as follows: In Section 2, the basic notions are reviewed and we introduce the Haagerup approximation property for a von Neumann algebra. In Section 3, we study some permanence properties such as reduced von Neumann algebras, tensor products, the commutant and the direct sums. In Section 4, we consider the case where M is a σ-finite von Neumann algebra with a faithful normal state ϕ. We present the proof of Theorem A. We also discuss the free product of von Neumann algebras and examples. In Section 5, we study the crossed product of a von Neumann algebra by a locally compact group. We show that a von Neumann algebra has the Haagerup approximation property if and only if so does its core von Neumann algebra. The proof of Theorem B is presented.Acknowledgements. The authors are grateful to Narutaka Ozawa for various useful comments on our work. Theorem B is the answer to his question to us. The first author would like to thank Marie Choda and Yoshikazu Katayama for fruitful discussions. The authors also express their gratitude to the referees for several helpful comments and revisions.
DefinitionWe first fix notations and recall basic facts. Let M be a von Neumann algebra. We denote by M sa and M + , the set of all self-adjoint elements and all positive elements in M, respectively. We also denote by M * and M + * the space of all normal linear functionals and all positive normal linear functionals on M, respectively.Let us recall the definition of a standard form of a von Neumann algebra that is formulated by Haagerup in [Ha1].Definition 2.1. Let (M, H, J, P ) be a quadruple, where M is a von Neumann algebra, H is a Hilbert space on which M acts, J is a conjugate-linear isometry on H with J 2 = 1 H , and P ⊂ H is a closed convex cone which is self-dual, i.e., P = {ξ ∈ H | ξ, η ≥ 0 for η ∈ P }.3