Dilation theory in finite dimensions: the possible, the impossible and the unknown

Miscellaneous inequalities involving matrices (15A45) Dilations, extensions, compressions of linear operators (47A20) Linear operator methods in interpolation, moment and extension problems (47A57)

Abstract: This expository essay discusses a finite dimensional approach to dilation theory. How much of dilation theory can be worked out within the realm of linear algebra? It turns out that some interesting and simple results can be obtained. These results can be used to give very elementary proofs of sharpened versions of some von Neumann type inequalities, as well as some other striking consequences about polynomials and matrices. Exploring the limits of the finite dimensional approach sheds light on the difference between those techniques and phenomena in operator theory that are inherently infinite dimensional, and those that are not.

Review text

The well-known dilation theorem of Sz.-Nagy states that if \(T\) is a contraction (that is, \(\|T\|\leq 1\)) on a Hilbert space \(H\), then there exists a Hilbert space \(K \supseteq H\) and a unitary \(U\) on \(K\) such that \[ T^k = P_{H}U^kP_{H} \quad \text{for all }k\in\mathbb{N}.\tag{\(\ast\)} \] Here, \(P_{H}\) is the orthogonal projection of \(K\) onto \(H\). Sz.-Nagy's theorem has numerous applications which include simple and elegant proofs of von Neumann's inequality and an ergodic theorem for contractions. However, the theory is highly infinite dimensional in the sense that even in the case \(H\) is of finite dimension, the space \(K\) is always infinite dimensional unless \(T\) is already a unitary operator. The main goal of the paper under review is to discuss how much of dilation theory can be worked out within the realm of linear algebra. In order to do that, the authors introduce the notion of \(N\)-dilations. A unitary \(U\) is an \(N\)-dilation of \(T\) if equation (\(\ast\)) holds for all \(k=1,\dots, N\). Given a contraction \(T\), the existence of an \(N\)-dilation acting on a finite dimensional space is proven. (As the authors point out, this result was proved by \textit{E. Egerváry} [Acta Sci. Math. 15, 178--182 (1954; Zbl 0056.11004)].) A proof of von~Neumann's inequality for matrices can now be given using \(N\)-dilations. Things become more difficult and interesting when one considers dilation theory for commuting tuples of contractions. Let \(T_1,\dots, T_k\) be commuting contractions on \(H\) and let \(1\leq N\leq\infty\). A unitary \(N\)-dilation for \(T_1,\dots, T_k\) is a \(k\)-tuple of commuting unitaries \(U_1,\dots, U_k\) on \(K\supseteq H\) such that \[ T_1^{n_1}\dots T_k^{n_k} = P_{H}U_1^{n_1}\dots U_k^{n_k}P_{H} \] for all non-negative integers \(n_1,\dots, n_k\) with \(n_1+\dots+n_k\leq N\). When \(N=\infty\), we simply have the notion of unitary dilation. As is well known, Ando's theorem asserts that any pair of commuting contractions has a unitary dilation. It is shown in the paper that a unitary \(N\)-dilation exists if the contractions \(T_1,\dots, T_k\) are assumed to doubly commute. However, unitary \(N\)-dilations do not always exist in general when \(k\geq 3\). Applications of the theory are also discussed. In particular, the authors present an interesting result about polynomials in several variables. Several open questions are raised. In fact, it is an open question whether there exists a ``commutant lifting theorem in the setting of unitary \(N\)-dilations. The paper concludes with a topic about \(*\)-endomorphic dilations of completely positive maps. The authors show that this theory is truly infinite dimensional in nature: it is generally impossible to find dilations acting on finite dimensional spaces.

Recommendations

Cites work

Cited in

(27)

This page was built for publication: Dilation theory in finite dimensions: the possible, the impossible and the unknown

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2453640)