arXiv subjects
Publications and source records attributed to William Arveson.
This article shows how the work of Henry Helson, especially the two papers of Helson and Lowdenslager, came to influence the development of the theory of non self adjoint operator algebras acting on Hilbert space.
A (finite or countably infinite) set G of generators of an abstract C*-algebra A is called hyperrigid if for every faithful representation of A on a Hilbert space $A\subseteq \mathcal B(H)$ and every sequence of unital completely positive linear maps $ϕ_1, ϕ_2,...$ from $\mathcal B(H)$ to itself, $$ \lim_{n\to\infty}\|ϕ_n(g)-g\|=0, \forall g\in G \implies \lim_{n\to\infty}\|ϕ_n(a)-a\|=0, \forall a\in A. $$ We show that one can determine whether a given set G of generators is hyperrigid by examining the noncommutative Choquet boundary of the operator space spanned by $G\cup G^*$. We present a variety of concrete applications and discuss prospects for further development.
Paul Halmos' work in dilation theory began with a question and its answer: Which operators on a Hilbert space can be extended to normal operators on a larger Hilbert space? The answer is interesting and subtle. The idea of representing operator-theoretic structures in terms of conceptually simpler structures acting on larger Hilbert spaces has become a central one in the development of operator theory and, more generally, noncommutative analysis. The work continues today. In this article we summarize some of these diverse results and their history.
We classify operator systems $S\subseteq \mathcal B(H)$ that act on finite dimensional Hilbert spaces by making use of the noncommutative Choquet boundary. S is said to be {\em reduced} when its boundary ideal is 0. In the category of operator systems, that property functions as semisimplicity does in the category of complex Banach algebras. We construct explicit examples of reduced operator systems using sequences of "parameterizing maps" $Γ_k: \mathbb C^r\to \mathcal B(H_k)$, $k=1,..., N$. We show that every reduced operator system is isomorphic to one of these, and that two sequences give rise to isomorphic operator systems if and only if they are "unitarily equivalent" parameterizing sequences. Finally, we construct nonreduced operator systems $S$ that have a given boundary ideal $K$ and a given reduced image in $C^*(S)/K$, and show that these constructed examples exhaust the possibilities.
Let $V$ be a norm-closed subset of the unit sphere of a Hilbert space $H$ that is stable under multiplication by scalars of absolute value 1. A {\em maximal vector} (for $V$) is a unit vector $ξ\in H$ whose distance to $V$ is maximum $d(ξ,V)=\sup_{\|η\|=1}d(η,V)$, $d(ξ,V)$ denoting the distance from $ξ$ to the set $V$. Maximal vectors generalize the {\em maximally entangled} unit vectors of quantum theory. In general, under a mild regularity hypothesis on $V$, there is a {\em norm} on $H$ whose restriction to the unit sphere achieves its minimum precisely on $V$ and its maximum precisely on the set of maximal vectors. This "entanglement-measuring norm" is unique. There is a corresponding "entanglement-measuring norm" on the predual of $\mathcal B(H)$ that faithfully detects entanglement of normal states. We apply these abstract results to the analysis of entanglement in multipartite tensor products $H=H_1\otimes ...\otimes H_N$, and we calculate both entanglement-measuring norms. In cases for which $\dim H_N$ is relatively large with respect to the others, we describe the set of maximal vectors in explicit terms and show that it does not depend on the number of factors of the Hilbert space $H_1\otimes...\otimes H_{N-1}$.
We show that states on tensor products of matrix algebras whose ranks are relatively small are {\em almost surely} entangled, but that states of maximum rank are not. More precisely, let $M=M_m(\mathbb C)$ and $N=M_n(\mathbb C)$ be full matrix algebras with $m\geq n$, fix an arbitrary state $ω$ of $N$, and let $E(ω)$ be the set of all states of $M\otimes N$ that extend $ω$. The space $E(ω)$ contains states of rank $r$ for every $r=1,2,...,m\cdot\rankω$, and it has a filtration into compact subspaces $$ E^1(ω)\subseteq E^2(ω)\subseteq ...\subseteq E^{m\cdot\rankω}=E(ω), $$ where $E^r(ω)$ is the set of all states of $E(ω)$ having rank $\leq r$. We show first that for every $r$, there is a real-analytic manifold $V^r$, homogeneous under a transitive action of a compact group $G^r$, which parameterizes $E^r(ω)$. The unique $G^r$-invariant probability measure on $V^r$ promotes to a probability measure $P^{r,ω}$ on $E^r(ω)$, and $P^{r,ω}$ assigns probability 1 to states of rank $r$. The resulting probability space $(E^r(ω),P^{r,ω})$ represents ``choosing a rank $r$ extension of $ω$ at random". Main result: For every $r=1,2,...,[\rank ω/2]$, states of $(E^r(ω),P^{r,ω})$ are almost surely entangled.
Let M and N be full matrix algebras. A unital completely positive (UCP) map ϕ:M\to N is said to preserve entanglement if its inflation ϕ\otimes \id_N : M\otimes N\to N\otimes N has the following property: for every maximally entangled pure state ρof N\otimes N, ρ\circ(ϕ\otimes \id_N) is an entangled state of M\otimes N. We show that there is a dichotomy in that every UCP map that is not entanglement breaking in the sense of Horodecki-Shor-Ruskai must preserve entanglement, and that entanglement preserving maps of every possible rank exist in abundance. We also show that with probability 1, {\em all} UCP maps of relatively small rank preserve entanglement, but that this is not so for UCP maps of maximum rank.
Normal endomorphisms of von Neumann algebras need not be extendable to automorphisms of a larger von Neumann algebra, but they always have asymptotic lifts. We describe the structure of endomorphisms and their asymptotic lifts in some detail, and apply those results to complete the identification of asymptotic lifts of unital completely positive linear maps on von Neumann algebras in terms of their minimal dilations to endomorphisms.
Let S be an operator system -- a self-adjoint linear subspace of a unital C*-algebra A such that contains 1 and A=C*(S) is generated by S. A boundary representation for S is an irreducible representation πof C*(S) on a Hilbert space with the property that $π\restriction_S$ has a unique completely positive extension to C*(S). The set $\partial_S$ of all (unitary equivalence classes of) boundary representations is the noncommutative counterpart of the Choquet boundary of a function system $S\subseteq C(X)$ that separates points of X. It is known that the closure of the Choquet boundary of a function system S is the Silov boundary of X relative to S. The corresponding noncommutative problem of whether every operator system has "sufficiently many" boundary representations was formulated in 1969, but has remained unsolved despite progress on related issues. In particular, it was unknown if $\partial_S$ is nonempty for generic S. In this paper we show that every separable operator system has sufficiently many boundary representations. Our methods use separability in an essential way.
We show that the notion of asymptotic lift generalizes naturally to normal positive maps $ϕ$ acting on von Neumann algebras M. We focus on cases in which the domain of the asymptotic lift can be embedded as an operator subsystem of M, and characterize when that subsystem is a Jordan subalgebra of M in terms of the asymptotic multiplicative properties of $ϕ$.
Starting with a unit-preserving normal completely positive map L: M --> M acting on a von Neumann algebra - or more generally a dual operator system - we show that there is a unique reversible system α: N --> N (i.e., a complete order automorphism αof a dual operator system N) that captures all of the asymptotic behavior of L, called the {\em asymptotic lift} of L. This provides a noncommutative generalization of the Frobenius theorems that describe the asymptotic behavior of the sequence of powers of a stochastic n x n matrix. In cases where M is a von Neumann algebra, the asymptotic lift is shown to be a W*-dynamical system (N,\mathbb Z), whick we identify as the tail flow of the minimal dilation of L. We are also able to identify the Poisson boundary of L as the fixed point algebra of (N,\mathbb Z). In general, we show the action of the asymptotic lift is trivial iff L is {\em slowly oscillating} in the sense that $$ \lim_{n\to\infty}\|ρ\circ L^{n+1}-ρ\circ L^n\|=0,\qquad ρ\in M_* . $$ Hence αis often a nontrivial automorphism of N.
Let X be a finite set of complex numbers and let A be a normal operator with spectrum X that acts on a separable Hilbert space H. Relative to a fixed orthonormal basis e_1,e_2, ... for H, A gives rise to a matrix whose diagonal is a sequence d=(d_1,d_2,...) with the property that each of its terms d_n belongs to the convex hull of X. Not all sequences with that property can arise as the diagonal of a normal operator with spectrum X. The case where X is a set of real numbers has received a great deal of attention over the years, and is reasonably well (though incompletely) understood. In this paper we take up the case in which X is the set of vertices of a convex polygon in the complex plane. The critical sequences d turn out to be those that accumulate rapidly in X in the sense that $$ \sum_{n=1}^\infty {\rm{dist}} (d_n,X)<\infty. $$ We show that there is an abelian group $Γ_X$ -- a quotient of $R^2$ by a countable subgroup with concrete arithmetic properties -- and a surjective mapping of such sequences $d\mapsto s(d)\inΓ_X$ with the following property: If s(d) is not 0, then d is not the diagonal of any such operator A. We also show that while this is the only obstruction when X contains two points, there are other (as yet unknown) obstructions when X contains more than two points.
Product systems are the classifying structures for semigroups of endomorphisms of B(H), in that two $E_0$-semigroups are cocycle conjugate iff their product systems are isomorphic. Thus it is important to know that every abstract product system is associated with an $E_0$-semigrouop. This was first proved more than fifteen years ago by rather indirect methods. Recently, Skeide has given a more direct proof. In this note we give yet another proof by an elementary construction.
The eigenvalues of a self-adjoint nxn matrix A can be put into a decreasing sequence $λ=(λ_1,...,λ_n)$, with repetitions according to multiplicity, and the diagonal of A is a point of $R^n$ that bears some relation to $λ$. The Schur-Horn theorem characterizes that relation in terms of a system of linear inequalities. We give a new proof of the latter result for positive trace-class operators on infinite dimensional Hilbert spaces, generalizing results of one of us on the diagonals of projections. We also establish an appropriate counterpart of the Schur inequalities that relate spectral properties of self-adjoint operators in $II_1$ factors to their images under a conditional expectation onto a maximal abelian subalgebra.
We initiate a study of Hilbert modules over the polynomial algebra A=C[z_1,...,z_d] that are obtained by completing A with respect to an inner product having certain natural properties. A standard Hilbert module is a finite multiplicity version of one of these. Standard Hilbert modules occupy a position analogous to that of free modules of finite rank in commutative algebra, and their quotients by submodules give rise to universal solutions of nonlinear relations. Essentially all of the basic Hilbert modules that have received attention over the years are standard - including the Hilbert module of the d-shift, the Hardy and Bergman modules of the unit ball, modules associated with more general domains in complex d-space, and those associated with projective algebraic varieties. We address the general problem of determining when a quotient H/M of an essentially normal standard Hilbert module H is essentially normal. This problem has been resistant. Our main result is that it can be "linearized" in that the nonlinear relations defining the submodule M can be reduced, appropriately, to linear relations through an iteration procedure, and we give a concrete description of linearized quotients.
We establish the existence and uniqueness of finite free resolutions - and their attendant Betti numbers - for graded commuting d-tuples of Hilbert space operators. Our approach is based on the notion of free cover of a (perhaps noncommutative) row contraction. Free covers provide a flexible replacement for minimal dilations that is better suited for higher-dimensional operator theory. For example, every graded d-contraction that is finitely multi-cyclic has a unique free cover of finite type - whose kernel is a Hilbert module inheriting the same properties. This contrasts sharply with what can be achieved by way of dilation theory (see Remark 2.4).
We show that for every "locally finite" unit-preserving completely positive map P acting on a C*-algebra, there is a corresponding *-automorphism αof another unital C*-algebra such that the two sequences P, P^2,P^3,... and α, α^2,α^3,... have the same {\em asymptotic} behavior. The automorphism αis uniquely determined by P up to conjugacy. Similar results hold for normal completely positive maps on von Neumann algebras, as well as for one-parameter semigroups. These results can be viewed as operator algebraic counterparts of the classical Perron-Frobenius theorem on the structure of square matrices with nonnegative entries.
We show that many invariant subspaces M for d-shifts (S_1,...,S_d) of finite rank have the property that the projection P onto M almost commutes with the S_k in the sense that the commutators PS_k - S_kP belong to the Schatten-von Neumann class L^p for every p > d. In such cases the d-tuple of operators (T_1,...,T_d) obtained by compressing (S_1,...,S_d) to the orthocomplement of M generates a *-algebra whose commutator ideal is contained in L^p, p > d. It follows that the C*-algebra generated by T_1,...,T_d is commutative modulo compact operators, the associated Dirac operator is Fredholm, and the index formula for the curvature invariant is stable under compact perturbations and homotopy for this restricted class of d-contractions. We conjecture that the latter conclusions persist under much more general circumstances.