Noncommutative Analysis

Category: Research

New paper “Compressions of compact tuples”, and announcement of mistake (and correction) in old paper “Dilations, inclusions of matrix convex sets, and completely positive maps”

Ben Passer and I have recently uploaded our preprint “Compressions of compact tuples” to the arxiv. In this paper we continue to study matrix ranges, and in particular matrix ranges of compact tuples. Recall that the matrix range of a tuple A = (A_1, \ldots, A_d) \in B(H)^d is the the free set \mathcal{W}(A) = \sqcup_{n=1}^\infty \mathcal{W}_n(A), where

\mathcal{W}_n(A) = \{(\phi(A_1), \ldots, \phi(A_d)) : \phi : B(H) \to M_n is UCP \}.

A tuple A is said to be minimal if there is no proper reducing subspace G \subset H such that \mathcal{W}(P_G A\big|_G) = \mathcal{W}(A). It is said to be fully compressed if there is no proper subspace whatsoever G \subset H such that \mathcal{W}(P_G A\big|_G) = \mathcal{W}(A).

In an earlier paper (“Dilations, inclusions of matrix convex sets, and completely positive maps”) I wrote with other co-authors, we claimed that if two compact tuples A and B are minimal and have the same matrix range, then A is unitarily equivalent to B; see Section 6 there (the printed version corresponds to version 2 of the paper on arxiv). This is false, as subsequent examples by Ben Passer showed (see this paper). A couple of other statements in that section are also incorrect, most obviously the claim that every compact tuple can be compressed to a minimal compact tuple with the same matrix range. All the problems with Section 6 of that earlier paper “Dilations,…” can be quickly  fixed by throwing in a “non-singularity” assumption, and we posted a corrected version on the arxiv. (The results of Section 6 there do not affect the rest of the results in the paper, and are somewhat not in the direction of the main parts of that paper).

In the current paper, Ben and I take a closer look at the non-singularity assumption that was introduced in the corrected version of “Dilations,…”, and we give a complete characterization of non-singular tuples of compacts. This characterization involves the various kinds of extreme points of the matrix range \mathcal{W}(A). We also make a serious invetigation into fully compressed tuples defined above. We find that a matrix tuple is fully compressed if and only if it is non-singular and minimal. Consequently, we get a clean statement of the classification theorem for compacts: if two tuples A and B of compacts are fully compressed, then they are unitarily equivalent if and only if \mathcal{W}(A) = \mathcal{W}(B).

 

Advertisements

The complex matrix cube problem summer project – summary of results

In the previous post I announced the project that I was going to supervise in the Summer Projects in Mathematics week at the Technion. In this post I wish to share what we did and what we found in that week.

I had the privilege to work with two very bright students who have recently finished their undergraduate studies: Mattya Ben-Efraim (from Bar-Ilan University) and Yuval Yifrach (from the Technion). It is remarkable the amount of stuff they learned for this one week project (the basics of C*-algebras and operator spaces), and that they actually helped settle the question that I raised to them.

I learned a lot of things in this project. First, I learned that my conjecture was false! I also learned and re-learned some programming abilities, and I learned something about the subtleties and limitations of numerical experimentation (I also learned something about how to supervise an undergraduate research project, but that’s besides the point right now).

Statement of the problem

Read the rest of this entry »

Minimal and maximal matrix convex sets

The final version of the paper Minimal and maximal matrix convex sets, written by Ben Passer, Baruch Solel and myself, has recently appeared online. The publisher (Elsevier) sent us a link through which the official final version is downloadable, for anyone who clicks on the following link before May 26, 2018. Here is the link for the use of the public:

Click here to download the journal version of the paper

Of course, if you don’t click by May 26 – don’t panic! We always put our papers on the arXiv, and here is the link to that. Here is the abstract:

Abstract. For every convex body K \subseteq \mathbb{R}^d, there is a minimal matrix convex set \mathcal{W}^{min}(K), and a maximal matrix convex set \mathcal{W}^{max}(K), which have K as their ground level. We aim to find the optimal constant \theta(K) such that \mathcal{W}^{max}(K) \subseteq \theta(K) \cdot \mathcal{W}^{min}(K). For example, if \overline{\mathbb{B}}_{p,d} is the unit ball in \mathbb{R}^d with the p-norm, then we find that 

\theta(\overline{\mathbb{B}}_{p,d}) = d^{1-|1/p-1/2|} .

This constant is sharp, and it is new for all p \neq 2. Moreover, for some sets K we find a minimal set L for which \mathcal{W}^{max}(K) \subseteq \mathcal{W}^{min}(L). In particular, we obtain that a convex body K satisfies \mathcal{W}^{max}(K) = \mathcal{W}^{min}(K) only if K is a simplex.

These problems relate to dilation theory, convex geometry, operator systems, and completely positive maps. For example, our results show that every d-tuple of self-adjoint contractions, can be dilated to a commuting family of self-adjoints, each of norm at most \sqrt{d}. We also introduce new explicit constructions of these (and other) dilations.

The nightmare

In September 30 the mathematician Vladimir Voevodsky passed away. Voevodsky, a Fields medalist, is a mathematician of whom I barely heard earlier, but after bumping into an obituary I was drawn to read about him and about his career. His story is remarkable in many ways. Voevodsky comes out as brilliant, intellectually honest giant, who bravely and honestly confronted the crisis that he observed “higher dimensional mathematics” was in.

Read the rest of this entry »

Algebras of bounded noncommutative analytic functions on subvarieties of the noncommutative unit ball

Guy Salomon, Eli Shamovich and I recently uploaded to the arxiv our paper “Algebras of bounded noncommutative analytic functions on subvarieties of the noncommutative unit ball“. This paper blends in with the current growing interest in noncommutative function theory, continues and unifies several strands of my past research.

A couple of years ago, after being inspired by lectures of Agler, Ball, McCarthy and  Vinnikov on the subject, and after years of being influenced by Paul Muhly and Baruch Solel’s work, I realized that many of my different research projects (subproduct systems, the isomorphism problem, space of Dirichlet series with the complete Pick property, operator algebras associated with monomial ideals) are connected by the unifying theme of bounded analytic nc functions on subvarieties of the nc ball. “Realized” is a strong word, because many of my original ideas on this turned out to be false, and others I still don’t know how to prove. Anyway, it took me a couple of years and a lot of help, and here is this paper.

In short, we study algebras of bounded analytic functions on subvarieties of the the noncommutative (nc) unit ball :

\mathfrak{B}_d = \{(X_1, \ldots, X_d) tuples of n \times n matrices,  \sum X_i X_i < I\}

as well as bounded analytic functions that extend continuously to the “boundary”. We show that these algebras are multiplier algebras of appropriate nc reproducing kernel Hilbert spaces, and are completely isometrically isomorphic to the quotient of H^\infty(\mathfrak{B}_d) (the bounded nc analytic functions in the ball) by the ideal of nc functions vanishing on the variety. We classify these algebras in terms of the varieties, similar to classification results in the commutative case. We also identify previously studied algebras (such as multiplier algebras of complete Pick spaces and tensor algebras of subproduct systems) as algebras of bounded analytic functions on nc varieties. See the introduction for more.

We certainly plan to continue this line of research in the near future – in particular, the passage to other domains (beyond the ball), and the study of algebraic/bounded isomorphisms.

Aleman, Hartz, McCarthy and Richter characterize interpolating sequences in complete Pick spaces

The purpose of this post is to discuss the recent important contribution by Aleman, Hartz, McCarthy and Richter to the characterization of interpolating sequences (for multiplier algebras of certain Hilbert function spaces). Their recent paper “Interpolating sequences in spaces with the complete Pick property” was uploaded to the arxiv about two weeks ago; here I will just give some background and state the main result. (Even more recently these four authors released yet another paper that looks very interesting – this one.)

1. Background – interpolating sequences

We will be working with the notion of Hilbert function spaces – also called reproducing Hilbert spaces (see this post for an introduction). Suppose that H is a Hilbert function space on a set X, and k its reproducing kernel. The Pick interpolation problem is the following:

Read the rest of this entry »

Dilations, inclusions of matrix convex sets, and completely positive maps

In part to help myself to prepare for my talk in the upcoming IWOTA, and in part to help myself prepare for getting back to doing research on this subject now that the semester is over, I am going to write a little exposition on my joint paper with Davidson, Dor-On and Solel, Dilations, inclusions of matrix convex sets, and completely positive maps. Here are the slides of my talk.

The research on this paper began as part of a project on the interpolation problem for unital completely positive maps*, but while thinking on the problem we were led to other problems as well. Our work was heavily influenced by works of Helton, Klep, McCullough and Schweighofer (some which I wrote about the the second section of this previous post), but goes beyond. I will try to present our work by a narrative that is somewhat different from the way the story is told in our paper. In my upcoming talk I will concentrate on one aspect that I think is most suitable for a broad audience. One of my coauthors, Adam Dor-On, will give a complimentary talk dealing with some more “operator-algebraic” aspects of our work in the Multivariable Operator Theory special session.

[*The interpolation problem for unital completely positive maps is the problem of finding conditions for the existence of a unital completely positive (UCP) map that sends a given set of operators A_1, \ldots, A_d to another given set B_1, \ldots, B_d. See Section 3 below.]

Read the rest of this entry »

Preprint update (Stable division and essential normality…)

Shibananda Biswas and I recently uploaded to the arxiv a new version of our paper “Stable division and essential normality: the non-homogeneous and quasi-homogeneous cases“. This is the paper I announced in this previous post, but we had to make some significant changes (thanks to a very good referee) so I think I have to re announce the paper.

I’ve sometimes been part of conversations where we mathematicians share with each other stories of how some paper we wrote was wrongfully (and in some cases, ridiculously) rejected; and then I’ve also been in conversations where we share stories of how we, as referees, recently had to reject some wrong (or ridiculous) paper. But I never had the occasion to take part in a conversation in which authors discuss papers they wrote that have been rightfully rejected. Well, thanks to the fact that I sometimes work on problems related to Arveson’s essential normality conjecture (which is notorious for having caused some embarrassment to betters-than-I), and also because I have become a little too arrogant and not sufficiently careful with my papers, I have recently become the author of a rightfully rejected paper. It is a good paper on a hard problem, I am not saying it is not, and it is (now (hopefully!)) correct, but it was rejected for a good reason. I think it is a story worth telling. Before I tell the story I have to say that both the referee and my collaborator were professional and great, and this whole blunder is probably my fault.

So Shibananda Biswas and I submitted this paper Stable division and essential normality: the non-homogeneous and quasi-homogeneous cases for publication. The referee sent back a report with several good comments, two of which turned out to be serious. The two serious comments concerned what appeared as Theorem 2.4 in the first version of the paper (and it appears as the corrected Theorem 2.4 in the current version, too). The first serious  issue was that in the proof of the main theorem we mixed up between t and t+1, and this, naturally, causes trouble (well, I am simplifying. Really we mixed between two Hilbert space norms, parametrised by t and t+1). The second issue (which did not seem to be a serious one at first) was that at some point of the proof we claimed that a particular linear operator is bounded since it is known to be bounded on a finite co-dimensional subspace; the referee asked for clarifications regarding this step.

The first issue was serious, but we managed to fix the original proof, roughly by changing t+1 back to t. There was a price to pay in that the result was slightly weaker, but not in a way that affected the rest of the paper. Happily, we also found a better proof of the result we wanted to prove in the first place, and this appears as Theorem 2.3 in the new and corrected version of the paper.

The second issue did not seem like a big deal. Actually, in the referee’s report this was just one comment among many, some of which were typos and minor things like that, so we did not really give it much attention. A linear operator is bounded on a finite co-dimensional subspace, so it is bounded on the whole space, I don’t have to explain that!

We sent the revision back, and after a while the referee replied that we took care of most things, but we still did not explain the part about the operator-being-bounded-because-it-is-bounded-on-a-finite-co-dimensional-space. The referee suggested that we either remove that part (since we already had the new proof), or we explain it. The referee added, that in either case he suggests to accept the paper.

Well, we could have just removed that part indeed and had the paper accepted, but we are not in the business of getting papers accepted for publication, we are in the business of proving theorems, and we believed that our original proof was interesting in itself since it used some interesting new techniques. We did not want to give up on that proof.

My collaborator wrote a revision with a very careful, detailed and rigorous explanation of how we get boundedness in our particular case, but I was getting angry and I made the big mistake of thinking that I am smarter than the referee. I thought to myself: this is general nonsense! It always holds. So I insisted on sending back a revision in which this step is explained by referring to a general principle that says that an operator which is bounded on a finite co-dimensional subspace is bounded.

OOPS!

That’s not quite exactly precisely true. Well, it depends what you mean by “bounded on a finite co-dimensional subspace”. If you mean that it is bounded on a closed subspace which has a finite dimensional algebraic complement then it is true, but one can think of interpretations of “finite co-dimensional” that make this is wrong: for example, consider an unbounded linear functional: it is bounded on its kernel, which is finite co-dimensional in some sense, but it is not bounded.

The referee, in their third letter, pointed this out, and at this point the editor decided that three strikes and we are out. I think that was a good call. A slap in the face and a lesson learned. I only feel bad for my collaborator, since the revision he prepared originally was OK.

Anyway, in the situation studied in our paper, the linear subspace on which the operator is bounded is a finite co-dimensional ideal in the ring of polynomials. It’s closure has zero intersection with the finite dimensional complement (the proof of this is not very hard, but is indeed non-trivial and makes use of the nature of the spaces in question), and everything is all right.  Having learned our lessons, we explain everything in detail in the current version. I hope that carefully enough.

I think that what caused us most trouble was that I did not understand what the referee did not understand. I assumed (very incorrectly, and perhaps arrogantly) that they did not understand a basic principle of functional analysis; it turned out that the referee did not understand why we are in a situation where we can apply this principle, and with hindsight this was worth explaining in more detail.

One of the most outrageous open problems in operator/matrix theory is solved!

I want to report on a very exciting development in operator/matrix theory: the von Neumann inequality for 3 \times 3 matrices has been shown to hold true. I learned this from a recent paper (with the irresistible title) “The von Neumann inequality for 3 \times 3 matrices“, posted on the arxiv by Greg Knese. In this paper, Knese explains how the solution of this outstanding open problem follows from results in a paper by Lukasz Kosinski, “The three point Nevanlinna-Pick problem in the polydisc” that appeared on the arxiv about a half a year ago. Beautifully, and not surprisingly, the solution of this operator/matrix theoretic problem follows from deep new facts in complex function theory in several variables.

To recall the problem, let us denote \|A\| the operator norm of a matrix A, and for every polynomial p in d variables we denote by \|p\|_\infty the supremum norm

\|p\|_\infty = \sup_{|z_i|\leq 1} |p(z_1, \ldots, z_d)|.

A matrix A is said to be contractive if \|A\| \leq 1.

We say that d commuting contractions A_1, \ldots, A_d satisfy von Neumann’s inequality if 

(*)  \|p(A_1,\ldots, A_d)\| \leq \|p\|_\infty.

It was known since the 1960s that (*) holds when d \leq 2. Moreover, it was known that for d \geq 3, there are counter examples, consisting of d contractive 4 \times 4 matrices that do not satisfy von Neumann’s inequality. On the other hand, it was known that (*) holds for any d if the matrices A_1, \ldots, A_d are of size 2 \times 2. Thus, the only missing piece of information was whether or not von Neumann’s inequality holds or not for three or more contractive 3 \times 3 matrices. To stress the point: it was not known whether or not von Neumann’s inequality holds for three three-by-three matrices. The problem in this form has been open for 15 years  – but the problem is much older: in 1974 Kaiser and Varopoulos came up with a 5 \times 5 counter-example, and since then both the 3 \times 3  and the 4 \times 4 cases were open until Holbrook in 2001 found a 4 \times 4 counter example. You have to agree that this is outrageous, perhaps even ridiculous, I mean, three 3 \times 3 matrices, come on!

In Knese’s paper this story and the positive solution to the problem is explained very clearly and succinctly, and is recommended reading for any operator theorist. One has to take on faith the paper of Kosinski which, as Knese stresses, is where the major new technical advance has been made (though one should not over-stress this fact, because tying things together, the way Knese has done, requires a deep understanding of this problem and of the various ingredients). To understand Kosinki’s paper would require a greater investment of time, but it appears that the paper has already been accepted for publication, so I am quite confident and happy to see this problem go down.

Spaces of Dirichlet series with the complete Pick property (or: the Drury-Arveson space in a new disguise)

John McCarthy and I have recently uploaded a new version of our paper “Spaces of Dirichlet series with the complete Pick property” to the arxiv. I would like to advertise the central discovery of this paper here.

Recall that the Drury-Arveson space H^2_d is the reproducing kernel Hilbert space on the open unit ball of a d dimensional Hilbert space, with reproducing kernel

k(z,w) = \frac{1}{1 - \langle z, w \rangle}.

It has the remarkable universal property that every Hilbert function space with the complete Pick property is naturally isomorphic to the restriction of H^2_\infty to a subset of the unit ball (see Theorem 6 and its corollary in this post), and consequently, every complete Pick algebra is a quotient of the multiplier algebra \mathcal{M}_\infty = Mult(H^2_\infty). To the best of my knowledge, no other Hilbert function spaces with such a universal property have been studied.

John and I discovered another reproducing kernel Hilbert space that turns out to be “the same” as the Drury-Arveson space H^2_\infty. Since the space H^2_\infty as been so well studied, it interesting to discover a new incarnation. The really interesting part is that the space we discovered is a space of analytic functions on a half plane (that is, a space of functions in one complex variable), rather than a space of analytic functions in infinitely many variables on the unit ball of a Hilbert space.

To be precise, the spaces we consider are spaces of Dirichlet series \mathcal{H}, of the form

\mathcal{H} = \{f(s) = \sum_{n=1}^\infty \gamma_n n^{-s} : \sum |\gamma_n|^2 a_n^{-1} < \infty \}.

(Here a_n is a sequence of positive numbers). These are Hilbert function spaces on some half plane that have a kernel of the form k(s,u) = \sum a_n n^{-s-\bar u}.

We first answer the question which of these spaces \mathcal{H} have the complete Pick property. This problem has a simple solution (which has been anticipated by similar results on spaces on the disc): if we denote by g(s) = \sum a_n n^{-s} the “generating function” of the space, and if we write

\frac{1}{g(s)} = \sum c_n n^{-s},

then \mathcal{H} is a complete Pick space if and only if c_n \leq 0 for all n \geq 2.

After we know to tell when these spaces are complete Pick, it is natural to ask which complete Pick spaces arise like this? We do not give a complete answer, but our surprising discovery is that things can easily be cooked up so to obtain the Drury-Arveson space H^2_d, where d can be any cardinal number in \{1,2,\ldots, \infty\}. For example, \mathcal{H} turns out to be “the same” as H^2_\infty if the kernel k is given by

k(s,u) = \frac{P(2)}{P(2) - P(2+s+\bar u)},

where P(s) = \sum_{p} p^{-s} is the prime zeta function (the sum is taken over all primes p).

 Now, I have been a little vague about what it means that \mathcal{H} is “the same” as H^2_\infty. In fact, this is a subtle question, and we devote a part of our paper what it means for two Hilbert function spaces to be the same — something that has puzzled us for a while.

What does this appearance of Drury-Arveson space as a space of Dirichlet series mean? Can we use this connection to learn something new on multivariable operator theory, or on Dirichlet series? How did the prime zeta function smuggle itself into this discussion? This requires further thought.