Numerical range. The field of values of linear operators and matrices (Q2364979)

From the authors' preface: The theories of quadratic forms and their applications appear in many parts of mathematics and the sciences. All students of mathematics have the opportunity to encounter such concepts and applications in their first course in linear algebra. This subject and its extensions to infinite dimensions comprise the theory of the numerical range \(W(T)\). There are two competing names for \(W(T)\), namely, the numerical range of \(T\) and the field of values for \(T\). The former has been favored historically by the functional analysis community, the latter by the matrix analysis community. It is a toss-up to decide which is preferable, and we have finally chosen the former because it is our habit, it is a more efficient expression, and because in recent conferences dedicated to \(W(T)\), even the linear algebra community has adopted it. Also, one universally refers to the numerical radius. We know of no book dedicated to presenting the fundamentals of this subject. Our goal here is to do so. Our hope is that this interesting and useful subject will thereby become available to a wider audience. To that end, we have chosen what we call the roadmap approach in writing this book. We want it to be quickly informative as to principal cities and main routes, but without getting the reader lost in secondary byways or overly general description. For this reason, we place and keep the subject squarely in a complex Hilbert space. This setting is the heart of the numerical range theory for bounded linear operators \(T\) and naturally contains the field of values theory for finite-dimensional matrices \(T\). The outline of the book is as follows. In Chapter 1, we have selected the most fundamental properties of the numerical range \(W(T)\). These include its convexity and its inclusion of the spectrum of \(T\) within its closure. In Chapter 2, we present mapping theorems relating \(W(T)\) to the spectral properties of \(T\). The best known of these are probably the power inequality and the dilation theory, but other important mapping properties are given there as well. Chapter 3 describes an operator trigonometry in which we ourselves have played a major role. This includes sharp criteria for accretive operator products \(\text{Re }W(T_1T_2)\geq 0\), and a theory of anti-eigenvalues of an arbitrary operator \(T\). We believe the latter will eventually become a standard chapter in linear algebra and will be useful in a variety of applications. Chapter 4 investigates connections between the numerical range \(W(T)\) and numerical analysis. This includes applications to schemes used in computational fluid dynamics and an improved convergence theory for certain numerical iterative algorithms from optimization theory. In Chapter 5, we expose important properties of \(W(T)\) for matrix theory, the finite-dimensional case. We also develop the essentials of some of the variations of \(W(T)\), such as the \(C\) numerical range, which have been of recent interest. We conclude in Chapter 6 with a presentation of the properties of certain interesting classes of operators that although no longer symmetric or normal still enjoy some important properties of those operators. These operator classes are defined in terms of their numerical range \(W(T)\) properties and include the normaloid, convexoid, and spectraloid operators.