arXiv subjects

Joseph A. Ball

Publications and source records attributed to Joseph A. Ball.

At least 37 records · Page 2Linked to original sources

arXiv2014

Bounded Real Lemma and structured singular value versus diagonal scaling: the free noncommutative setting

The structured singular value $μ$ was introduced independently by Doyle and Safanov as a tool for analyzing robustness of system stability and performance in the presence of structured uncertainty in the system parameters. While the structured singular value provides a necessary and sufficient criterion for robustness with respect to a structured ball of uncertainty, it is notoriously difficult to actually compute. The method of diagonal (or simply "D") scaling, on the other hand, provides an easily computable upper bound (which we call $\hat μ$) for the structured singular value, but provides an exact evaluation of $μ$ (or even a useful upper bound for $μ$) only in special cases. However it was discovered in the 1990s that a certain enhancement of the uncertainty structure (i.e., letting the uncertainty parameters be freely noncommuting linear operators on an infinite-dimensional separable Hilbert space) resulted in the $D$-scaling procedure leading to an exact evaluation of $μ_{\text{enhanced}}$ ($μ_{\text{enhanced}} = \hat μ$), at least for the tractable special cases which were analyzed in complete detail. On the one hand this enhanced uncertainty has some appeal from the physical point of view: one can allow the uncertainty in the plant parameters to be time-varying, or more generally, one can catch the uncertainty caused by the designer's decision not to model the more complex (e.g. nonlinear) dynamics of the true plant. On the other hand, the precise mathematical formulation of this enhanced uncertainty structure makes contact with developments in the growing theory of analytic functions in freely noncommuting arguments and associated formal power series in freely noncommuting indeterminates. In this article we obtain the $\widetilde μ= \hat μ$ theorem for a more satisfactory general setting.

math.OC↗

arXiv2014

Weighted Hardy spaces: shift invariant and coinvariant subspaces, linear systems and operator model theory

The Sz.-Nagy--Foias model theory for $C_{\cdot 0}$ contraction operators combined with the Beurling-Lax theorem establishes a correspondence between any two of four kinds of objects: shift-invariant subspaces, operator-valued inner functions, conservative discrete-time input/state/output linear systems, and $C_{\cdot 0}$ Hilbert-space contraction operators. We discuss an analogue of all these ideas in the context of weighted Hardy spaces over the unit disk and an associated class of hypercontraction operators.

math.CA↗

arXiv2014

de Branges-Rovnyak spaces: basics and theory

For $S$ a contractive analytic operator-valued function on the unit disk ${\mathbb D}$, de Branges and Rovnyak associate a Hilbert space of analytic functions ${\mathcal H}(S)$ and related extension space ${\mathcal D(S)}$ consisting of pairs of analytic functions on the unit disk ${\mathbb D}$. This survey describes three equivalent formulations (the original geometric de Branges-Rovnyak definition, the Toeplitz operator characterization, and the characterization as a reproducing kernel Hilbert space) of the de Branges-Rovnyak space ${\mathcal H}(S)$, as well as its role as the underlying Hilbert space for the modeling of completely non-isometric Hilbert-space contraction operators. Also examined is the extension of these ideas to handle the modeling of the more general class of completely nonunitary contraction operators, where the more general two-component de Branges-Rovnyak model space ${\mathcal D}(S)$ and associated overlapping spaces play key roles. Connections with other function theory problems and applications are also discussed. More recent applications to a variety of subsequent applications are given in a companion survey article.

math.CA↗

arXiv2014

de Branges-Rovnyak spaces and norm-constraint interpolation

For $S$ a contractive analytic operator-valued function on the unit disk ${\mathbb D}$, de Branges and Rovnyak associate a Hilbert space of analytic functions ${\mathcal H}(S)$. A companion survey provides equivalent definitions and basic properties of these spaces as well as applications to function theory and operator theory. The present survey brings to the fore more recent applications to a variety of more elaborate function theory problems, including $H^\infty$-norm constrained interpolation, connections with the Potapov method of Fundamental Matrix Inequalities, parametrization for the set of all solutions of an interpolation problem, variants of the Abstract Interpolation Problem of Katsnelson, Kheifets, and Yuditskii, boundary behavior and boundary interpolation in de Branges-Rovnyak spaces themselves, and extensions to multivariable and Kre\uın-space settings.

math.CA↗

arXiv2014

De Branges-Rovnyak realizations of operator-valued Schur functions on the complex right half-plane

We give a controllable energy-preserving and an observable co-energy-preserving de Branges-Rovnyak functional model realization of an arbitrary given operator Schur function defined on the complex right-half plane. We work the theory out fully in the right-half plane, without using results for the disk case, in order to expose the technical details of continuous-time systems theory. At the end of the article, we make explicit the connection to the corresponding classical de Branges-Rovnyak realizations for Schur functions on the complex unit disk.

math.OC↗

arXiv2013

Scattering systems with several evolutions and formal reproducing kernel Hilbert spaces

A Schur-class function in $d$ variables is defined to be an analytic contractive-operator valued function on the unit polydisk. Such a function is said to be in the Schur--Agler class if it is contractive when evaluated on any commutative $d$-tuple of strict contractions on a Hilbert space. It is known that the Schur--Agler class is a strictly proper subclass of the Schur class if the number of variables $d$ is more than two. The Schur--Agler class is also characterized as those functions arising as the transfer function of a certain type (Givone--Roesser) of conservative multidimensional linear system. Previous work of the authors identified the Schur--Agler class as those Schur-class functions which arise as the scattering matrix for a certain type of (not necessarily minimal) Lax--Phillips multievolution scattering system having some additional geometric structure. The present paper links this additional geometric scattering structure directly with a known reproducing-kernel characterization of the Schur--Agler class. We use extensively the technique of formal reproducing kernel Hilbert spaces that was previously introduced by the authors and that allows us to manipulate formal power series in several commuting variables and their inverses (e.g., Fourier series of elements of $L^2$ on a torus) in the same way as one manipulates analytic functions in the usual setting of reproducing kernel Hilbert spaces.

math.FA↗

arXiv2013

Rational Cayley inner Herglotz-Agler functions: positive-kernel decompositions and transfer-function realizations

The Bessmertny\uı class consists of rational matrix-valued functions of $d$ complex variables representable as the Schur complement of a block of a linear pencil $A(z)=z_1A_1+\cdots+z_dA_d$ whose coefficients $A_k$ are positive semidefinite matrices. We show that it coincides with the subclass of rational functions in the Herglotz-Agler class over the right poly-halfplane which are homogeneous of degree one and which are Cayley inner. The latter means that such a function is holomorphic on the right poly-halfplane and takes skew-Hermitian matrix values on $(i\mathbb{R})^d$, or equivalently, is the double Cayley transform (over the variables and over the matrix values) of an inner function on the unit polydisk. Using Agler-Knese's characterization of rational inner Schur-Agler functions on the polydisk, extended now to the matrix-valued case, and applying appropriate Cayley transformations, we obtain characterizations of matrix-valued rational Cayley inner Herglotz-Agler functions both in the setting of the polydisk and of the right poly-halfplane, in terms of transfer-function realizations and in terms of positive-kernel decompositions. In particular, we extend Bessmertny\uı's representation to rational Cayley inner Herglotz-Agler functions on the right poly-halfplane, where a linear pencil $A(z)$ is now in the form $A(z)=A_0+z_1A_1+\cdots +z_dA_d$ with $A_0$ skew-Hermitian and the other coefficients $A_k$ positive semidefinite matrices.

math.FA↗

arXiv2012

Weighted Bergman spaces: shift-invariant subspaces and input/state/output linear systems

math.CA↗

arXiv2012

Contractive multipliers from Hardy space to weighted Hardy space

It is shown how any contractive multiplier from the Hardy space to a weighted Hardy space $H^{2}_{\bbeta}$ can be factored as a fixed factor composed with the classical Schur multiplier (contractive multiplier between Hardy spaces). The result is applied to get results on interpolation for a Hardy-to-weighted-Hardy contractive multiplier class.

math.CA↗

arXiv2012

The Takagi problem on the disk and bidisk

We give a new proof on the disk that a Pick problem can be solved by a rational function that is unimodular on the unit circle and for which the number of poles inside the disk is no more than the number of non-positive eigenvalues of the Pick matrix. We use this method to find rational solutions to Pick problems on the bidisk.

math.FA↗

arXiv2011

Convexity analysis and matrix-valued Schur class over finitely connected planar domains

We identify the set of extreme points and apply Choquet theory to a normalized matrix-measure ball subject to finitely many linear side constraints. As an application we obtain integral representation formulas for the Herglotz class of matrix-valued functions on a finitely-connected planar domain and associated continuous Agler decompositions for the matrix-valued Schur class over the domain. The results give some additional insight into the negative answer to the spectral set problem over such domains recently obtained by Agler-Harland-Raphael and Dritschel-McCullough.

math.FA↗

arXiv2011

Test functions, Schur-Agler classes and transfer-function realizations: the matrix-valued setting

Given a collection of test functions, one defines the associated Schur-Agler class as the intersection of the contractive multipliers over the collection of all positive kernels for which each test function is a contractive multiplier. We indicate extensions of this framework to the case where the test functions, kernel functions, and Schur-Agler-class functions are allowed to be matrix- or operator-valued. We illustrate the general theory with two examples: (1) the matrix-valued Schur class over a finitely-connected planar domain and (2) the matrix-valued version of the constrained Hardy algebra (bounded analytic functions on the unit disk with derivative at the origin constrained to have zero value). Emphasis is on examples where the matrix-valued version is not obtained as a simple higher-multiplicity tensoring of the scalar-valued version.

math.FA↗

arXiv2011

Canonical transfer-function realization for Schur multipliers on the Drury-Arveson space and models for commuting row contractions

We develop a $d$-variable analog of the two-component de Bran-ges-Rovnyak reproducing kernel Hilbert space associated with a Schur-class function on the unit disk. In this generalization, the unit disk is replaced by the unit ball in $d$-dimensional complex Euclidean space, and the Schur class becomes the class of contractive multipliers on the Drury-Arveson space over the ball. We also develop some results on a model theory for commutative row contractions which are not necessarily completely noncoisometric (the case considered in earlier work of Bhattacharyya, Eschmeier and Sarkar)

math.FA↗

arXiv2010

Extension of the $ν$-metric: the $H^\infty$ case

An abtract $ν$-metric was introduced by Ball and Sasane, with a view towards extending the classical $ν$-metric of Vinnicombe from the case of rational transfer functions to more general nonrational transfer function classes of infinite-dimensional linear control systems. In this short note, we give an additional concrete special instance of the abstract $ν$-metric, by verifying all the assumptions demanded in the abstract set-up. This example links the abstract $ν$-metric with the one proposed by Vinnicombe as a candidate for the $ν$-metric for nonrational plants.

math.OC↗

arXiv2010

The inverse commutant lifting problem: characterization of associated Redheffer linear-fractional maps

It is known that the set of all solutions of a commutant lifting and other interpolation problems admits a Redheffer linear-fractional parametrization. The method of unitary coupling identifies solutions of the lifting problem with minimal unitary extensions of a partially defined isometry constructed explicitly from the problem data. A special role is played by a particular unitary extension, called the central or universal unitary extension. The coefficient matrix for the Redheffer linear-fractional map has a simple expression in terms of the universal unitary extension. The universal unitary extension can be seen as a unitary coupling of four unitary operators (two bilateral shift operators together with two unitary operators coming from the problem data) which has special geometric structure. We use this special geometric structure to obtain an inverse theorem (Theorem 8.4 as well as Theorem 9.3) which characterizes the coefficient matrices for a Redheffer linear-fractional map arising in this way from a lifting problem. When expressed in terms of Hellinger-space functional models (Theorem 10.3), these results lead to generalizations of classical results of Arov and to characterizations of the coefficient matrix-measures of the lifting problem in terms of the density properties of the corresponding model spaces. The main tool is the formalism of unitary scattering systems developed in [18], [45].

math.FA↗

arXiv2010

Extension of the $ν$-metric

We extend the $ν$-metric introduced by Vinnicombe in robust control theory for rational plants to the case of infinite-dimensional systems/classes of nonrational transfer functions.

math.OC↗

arXiv2007

Multivariable generalizations of the Schur class: positive kernel characterization and transfer function realization

The operator-valued Schur-class is defined to be the set of holomorphic functions $S$ mapping the unit disk into the space of contraction operators between two Hilbert spaces. There are a number of alternate characterizations: the operator of multiplication by $S$ defines a contraction operator between two Hardy Hilbert spaces, $S$ satisfies a von Neumann inequality, a certain operator-valued kernel associated with $S$ is positive-definite, and $S$ can be realized as the transfer function of a dissipative (or even conservative) discrete-time linear input/state/output linear system. Various multivariable generalizations of this class have appeared recently,one of the most encompassing being that of Muhly and Solel where the unit disk is replaced by the strict unit ball of the elements of a dual correspondence $E^σ$ associated with a $W^{*}$-correspondence $E$ over a $W^{*}$-algebra $\cA$ together with a $*$-representation $σ$ of $\cA$. The main new point which we add here is the introduction of the notion of reproducing kernel Hilbert correspondence and identification of the Muhly-Solel Hardy spaces as reproducing kernel Hilbert correspondences associated with a completely positive analogue of the classical Szegö kernel. In this way we are able to make the analogy between the Muhly-Solel Schur class and the classical Schur class more complete. We also illustrate the theory by specializing it to some well-studied special cases; in some instances there result new kinds of realization theorems.

math.CA↗

arXiv2006

Multivariable backward-shift-invariant subspaces and observability operators

It is well known that subspaces of the Hardy space over the unit disk which are invariant under the backward shift occur as the image of an observability operator associated with a discrete-time linear system with stable state-dynamics, as well as the functional-model space for a Hilbert space contraction operator. We discuss two multivariable extensions of this structure, where the classical Hardy space is replaced by (1) the Fock space of formal power series in a collection of $d$ noncommuting indeterminates with norm-square-summable vector coefficients, and (2) the reproducing kernel Hilbert space (often now called the Arveson space) over the unit ball in ${\mathbb C}^{d}$ with reproducing kernel $k(λ, ζ) = 1/(1 - < λ, ζ>)$ ($\lam, ζ\in {\mathbb C}^{d}$ with $\| λ\|, \| ζ\| < 1$). In the first case, the associated linear system is of noncommutative Fornasini-Marchesini type with evolution along a free semigroup with $d$ generators, while in the second case the linear system is a standard (commutative) Fornasini-Marchesini-type system with evolution along the integer lattice ${\mathbb Z}^{d}$. An abelianization map (or symmetrization of the Fock space) links the first case with the second. The second case has special features depending on whether the operator-tuple defining the state dynamics is commutative or not. The paper focuses on multidimensional state-output linear systems and the associated observability operators.

math.CA↗