Professional Documents
Culture Documents
Operator Analysis Hilbert Space Methods in Complex Analysis Jim Agler Full Chapter PDF
Operator Analysis Hilbert Space Methods in Complex Analysis Jim Agler Full Chapter PDF
Operator Analysis Hilbert Space Methods in Complex Analysis Jim Agler Full Chapter PDF
https://ebookmass.com/product/modern-control-state-space-
analysis-and-design-methods-arie-nakhmani/
https://ebookmass.com/product/gaussian-measures-in-hilbert-space-
construction-and-properties-kukush/
https://ebookmass.com/product/integration-with-complex-numbers-a-
primer-on-complex-analysis-aisling-mccluskey/
https://ebookmass.com/product/visual-complex-analysis-25th-
anniversary-edition-tristan-needham-2/
Visual Complex Analysis: 25th Anniversary Edition
Tristan Needham
https://ebookmass.com/product/visual-complex-analysis-25th-
anniversary-edition-tristan-needham/
https://ebookmass.com/product/numerical-methods-in-environmental-
data-analysis-moses-eterigho-emetere/
https://ebookmass.com/product/genetic-analysis-of-complex-
disease-3rd-edition-william-k-scott/
https://ebookmass.com/product/introduction-to-research-methods-
and-data-analysis-in-psychology-darren-langdridge/
https://ebookmass.com/product/research-methods-design-and-
analysis-ebook-pdf-version/
C A M B R I D G E T R AC T S I N M AT H E M AT I C S
General Editors
J I M AG L E R
University of California, San Diego
J O H N E DWA R D McC A RT H Y
Washington University in St. Louis
N I C H O L A S YO U N G
Leeds University and Newcastle University
University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi 110025, India
79 Anson Road, #06–04/06, Singapore 079906
www.cambridge.org
Information on this title: www.cambridge.org/9781108485449
DOI: 10.1017/9781108751292
© Jim Agler, John Edward McCarthy, and Nicholas Young 2020
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2020
Printed in the United Kingdom by TJ International Ltd, Padstow Cornwall
A catalogue record for this publication is available from the British Library.
ISBN 978-1-108-48544-9 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To
Sarah, Suzanne, and
Contents
vii
viii Contents
2.7 The Müntz–Szász Interpolation Theorem 56
2.8 Positivity Arguments 61
2.9 Historical Notes 70
3 Further Development of Models on the Disc 71
3.1 A Model Formula for SB(H,K) (D) 71
3.2 Lurking Isometries Revisited 74
3.3 The Network Realization Formula 75
3.4 Tensor Products of Hilbert Spaces 78
3.5 Tensor Products of Operators 79
3.6 Realization of Rational Matrix Functions and
the McMillan Degree 81
3.7 Pick Interpolation Revisited 83
3.8 The Corona Problem 85
3.9 Historical Notes 91
4 Operator Analysis on D2 93
4.1 The Space of Hereditary Functions on D2 93
4.2 The Hereditary Functional Calculus on D2 95
4.3 Models on D2 99
4.4 Models on D2 via the Duality Construction 104
4.5 The Network Realization Formula for D2 106
4.6 Nevanlinna–Pick Interpolation on D2 109
4.7 Toeplitz Corona for the Bidisc 112
4.8 Operator-Valued Functions on D2 113
4.9 Models of Operator-Valued Functions on D2 116
4.10 Historical Notes 129
5 Carathéodory–Julia Theory on the Disc and the Bidisc 131
5.1 The One-Variable Results 131
5.2 The Model Approach to Regularity on D: B-points and C-points 133
5.3 A Proof of the Carathéodory–Julia Theorem on D via Models 137
5.4 Pick Interpolation on the Boundary 141
5.5 Regularity, B-points and C-points on the Bidisc 143
5.6 The Missing Link 146
5.7 Historical Notes 147
6 Herglotz and Nevanlinna Representations in Several
Variables 148
6.1 Overview 148
6.2 The Herglotz Representation on D2 149
6.3 Nevanlinna Representations on H via Operator Theory 153
Contents ix
6.4 The Nevanlinna Representations on H2 156
6.5 A Classification Scheme for Nevanlinna Representations
in Two Variables 160
6.6 The Type of a Function 165
6.7 Historical Notes 168
7 Model Theory on the Symmetrized Bidisc 169
7.1 Adding Symmetry to the Fundamental Theorem for D2 170
7.2 How to Define Models on the Symmetrized Bidisc 173
7.3 The Network Realization Formula for G 176
7.4 The Hereditary Functional Calculus on G 177
7.5 When Is G a Spectral Domain? 182
7.6 G Spectral Implies G Complete Spectral 185
7.7 The Spectral Nevanlinna–Pick Problem 185
7.8 Historical Notes 187
8 Spectral Sets: Three Case Studies 189
8.1 Von Neumann’s Inequality and the Pseudo-Hyperbolic
Metric on D 189
8.2 Spectral Domains and the Carathéodory Metric 192
8.3 Background Material 194
8.4 Lempert’s Theorem 195
8.5 The Carathéodory Distance on G 203
8.6 Von Neumann’s Inequality on Subvarieties of the Bidisc 205
8.7 Historical Notes 211
9 Calcular Norms 213
9.1 The Taylor Spectrum and Functional Calculus 213
9.2 Calcular Norms and Algebras 215
9.3 Halmos’s Conjecture and Paulsen’s Theorem 223
9.4 The Douglas–Paulsen Norm 226
9.5 The B. and F. Delyon Norm and Crouzeix’s Theorem 233
9.6 The Badea–Beckermann–Crouzeix Norm 235
9.7 The Polydisc Norm 243
9.8 The Oka Extension Theorem and Calcular Norms 245
9.9 Historical Notes 252
10 Operator Monotone Functions 254
10.1 Löwner’s Theorems 254
10.2 An Interlude on Linear Programming 262
10.3 Locally Matrix Monotone Functions in d Variables 269
10.4 The Löwner Class in d Variables 278
x Contents
10.5 Globally Monotone Rational Functions in Two Variables 279
10.6 Historical Notes 282
PART TWO NON-COMMUTATIVE THEORY 283
11 Motivation for Non-Commutative Functions 285
11.1 Non-Commutative Polynomials 285
11.2 Sums of Squares 287
11.3 Nullstellensatz 287
11.4 Linear Matrix Inequalities 288
11.5 The Implicit Function Theorem 289
11.6 Matrix Monotone Functions 289
11.7 The Functional Calculus 290
11.8 Historical Notes 290
12 Basic Properties of Non-Commutative Functions 291
12.1 Definition of an nc-Function 291
12.2 Locally Bounded nc-Functions Are Holomorphic 295
12.3 Nc Topologies 298
12.4 Historical Notes 300
13 Montel Theorems 302
13.1 Overview 302
13.2 Wandering Isometries 303
13.3 A Graded Montel Theorem 305
13.4 An nc Montel Theorem 309
13.5 Closed Cones 310
13.6 Historical Notes 312
14 Free Holomorphic Functions 313
14.1 The Range of a Free Holomorphic Function 313
14.2 Nc-Models and Free Realizations 315
14.3 Free Pick Interpolation 319
14.4 Free Realizations Redux 324
14.5 Extending off Varieties 326
14.6 The nc Oka–Weil Theorem 327
14.7 Additional Results 328
14.8 Historical Notes 329
15 The Implicit Function Theorem 330
15.1 The Fine Inverse Function Theorem 330
15.2 The Fat Inverse Function Theorem 332
15.3 The Implicit Function Theorem 337
15.4 The Range of an nc-Function 339
Contents xi
15.5 Applications of the Implicit Function Theorem in
Non-Commutative Algebraic Geometry 341
15.6 Additional Results 343
15.7 Historical Notes 343
16 Non-Commutative Functional Calculus 344
16.1 Nc Operator Functions 344
16.2 Polynomial Approximation and Power Series 349
16.3 Extending Free Functions to Operators 354
16.4 Nc Functional Calculus in Banach Algebras 356
16.5 Historical Notes 357
Notation 358
Bibliography 361
Index 372
Preface
The philosophy of this book is that Hilbert space geometry binds function
theory and operator theory together, not only allowing each to aid the other but
creating a rich structure that can be used to discover new phenomena. There is
a “three-way street” between operator theory and function theory: sometimes
one uses function theory to prove operator theorems, sometimes one uses
operator theory to prove function theorems, and sometimes the theories are so
interwoven that one cannot even state the theorem without using the language
of both operator theory and function theory.
The main thrust of the book is to discover and prove theorems about
holomorphic functions and complex geometry with the aid of Hilbert space
geometry and operator theory. The holomorphic functional calculus permits us
to substitute commuting operators for the variables of a holomorphic function,
a step that reveals hidden properties of the function—something every function
theorist should want to do.
It is remarkable how little operator theory one needs in order to prove
significant facts in function theory. There will be no call here for the detailed
and subtle theories of particular classes of operators, but we shall make heavy
use of the functional calculus for operators. The theories of operator dilations
and spectral sets also play an important role. We explain carefully what is
required from these theories. The reader may either take this material on trust
or consult the references that we give.
Part I of the book is devoted to holomorphic functions on domains in
complex Euclidean space, beginning with scalar-valued functions on the unit
disc in the complex plane, where intuition is most easily developed. Here
the central notion of a Hilbert space model of a function is introduced, as
are several types of arguments that will recur throughout the book. Gradually
we build up to various domains in higher dimensions and to operator-valued
functions, the latter being important for engineering applications. We do not
xiii
xiv Preface
aim for maximum generality, which can come at the cost of a sacrifice of
elegance and impact.
Part II concerns non-commutative functions, that is, functions of non-
commuting variables. This is a topic of relatively recent study, and one that
is currently undergoing rapid development. It transpires that the Hilbert space
methods of Part I are well suited to this new context, and we derive many
analogs of classical theorems.
Our intended audience is graduate students and mathematicians interested
in complex function theory and/or operator theory. We do not require famil-
iarity with several complex variables, but we do assume that the reader has a
basic knowledge of complex analysis and functional analysis.
The moment you buy into the functional calculus, you’re ready to roll!
Acknowledgments
The catalyst for this book was two series of lectures, given by the first and
second authors at Newcastle University in an LMS lecture series in April
2017. The authors would like to thank the London Mathematical Society for
supporting this event. In addition, the first author was partially supported by
National Science Foundation Grant DMS 1665260, the second author was
partially supported by National Science Foundation Grant DMS 1565243,
and the third author was partially supported by UK Engineering and Physical
Sciences Research Council grants EP/K50340X/1 and EP/N03242X/1, and
London Mathematical Society grants 41219 and 41730. We would like to thank
the National Science Foundation, the UK Engineering and Physical Sciences
Research Council, and the London Mathematical Society for their support.
Parts of the book were read in draft form by several people, who caught
some mistakes. We are specially grateful to Alberto Dayan, Chris Felder,
Michael Hartz, Mark Mancuso, James Pascoe, and Jeet Sampat. All remain-
ing mistakes and typographical errors are, of course, the responsibility of
the authors.
xv
P AR T ONE
Commutative Theory
1
The Origins of Operator-Theoretic Approaches
to Function Theory
In this chapter we present some of the highlights of operator theory over the
last century and briefly describe their connections with function theory. It is
intended to establish notation and provide orientation for the reader, before we
go on to a detailed and systematic development in Chapter 2 (which can be
understood independently of this scene-setting chapter). The reader should not
attempt to master all the contents of this chapter in detail before progressing–
some of them are substantial. Our choice, inevitably, is partial; we mention
some other major contributions in historical notes at the end of the chapter, but
we do not attempt a history of the subject.
1.1 Operators
We adopt the standard Bourbaki symbols R and C for the sets of real numbers
and complex numbers, respectively, and also the notations
T = {z ∈ C: |z| = 1} and D = {z ∈ C: |z| < 1}.
For any Banach space X , we shall denote by ball X the closed unit ball of X ,
that is,
ball X = {u ∈ X : u ≤ 1}.
If X , Y are Banach spaces, we say that a linear transformation T : X → Y is
bounded if there exists c ∈ [0, ∞) such that
T u ≤ cu for all u ∈ X . (1.1)
For a bounded linear transformation T : X → Y we define T , the norm of T ,
by the formula
T = sup T u, (1.2)
u∈ball X
3
4 Commutative Theory
or equivalently, by the formula
T = inf {c: T u ≤ cu for all u ∈ X }.
When X , Y are Banach spaces, we denote by B(X ) the set of bounded
linear transformations from X to X , and by B(X , Y) the set of bounded linear
transformations from X to Y. We refer to the elements of B(X ) or B(X , Y)
as operators. B(X )and B(X , Y) are Banach spaces under the norm defined by
equation (1.2). Furthermore, B(X ) (with composition as multiplication) is a
Banach algebra with identity, as it satisfies the axioms1 1 = 1 and
T1 T2 ≤ T1 T2
for all T1 , T2 ∈ B(X ).
Every element of B(X ) has a spectrum. For T ∈ B(X ) we say that T is
invertible if there exists S ∈ B(X ) such that ST = T S = 1. The spectrum of T ,
denoted by σ (T ), is defined by
σ (T ) = {z ∈ C: z − T is not invertible}.
The set σ (T ) is a nonempty compact subset of C provided that X = {0}.
then f (T ) is defined by
n
f (T ) = ak T k .
k=1
That this series converges in the operator norm of B(X ) is a simple exercise on
the spectral radius formula and Cauchy’s formula for the radius of convergence
of a power series.
Another simple modification of the polynomial calculus is the rational cal-
culus. For K, a compact subset of C, let Rat(K) denote the algebra of rational
functions with poles off K. If T ∈ B(X ) and f ∈ Rat(σ (T )), then we may write
f (z) = p(z)q(z)−1 ,
where p and q are polynomials and q(z) = 0 for all z ∈ σ (T ). By the spectral
mapping theorem, which asserts that q(σ (T )) = σ (q(T )), the spectrum of
q(T ) does not contain 0, so that q(T ) is invertible. Therefore, f (T ) can be
defined by the formula
f (T ) = p(T )q(T )−1 . (1.3)
A sweet spot in the theory of functional calculi is the Riesz–Dunford
functional calculus. Treatments of this topic, complete with proofs, are in
6 Commutative Theory
[60, VII.4; 140, section 17.2], and many other introductory texts on functional
analysis. We shall just state the important results.
Fix an operator T ∈ B(X ). For any open set U in C such that σ (T ) ⊆ U , let
be a finite collection of closed rectifiable curves in U \ σ (T ) that winds once
around each point in the spectrum of σ (T ) and no times around each point in
the complement of U . If f ∈ Hol(U ) then, by the Cauchy integral formula,
1 f (w)
f (z) = dw,
2π i w − z
whenever z ∈ σ (T ). The Riesz–Dunford functional calculus defines f (T ) for
f ∈ Hol(U ) by the substitution z = T in this formula, that is to say,
1
f (T ) = f (w)(w − T )−1 dw. (1.4)
2π i
Note that this formula makes sense, since the assumption ⊆ C\σ (T ) implies
that w − T is invertible for all w ∈ . Moreover, the defining equation (1.4)
depends neither on the choice of contour nor on the choice of U , since by
Cauchy’s theorem, for any vectors u, v in X and X ∗ ,
1
f (T )u, v = f (w) (w − T )−1 u, v dw
2π i
is independent of .
Properties of the Riesz–Dunford functional calculus
f → f (T )
f (T ) = r(T ).
(4) The calculus is consistent with the power series calculus, that is, if U is a
disc, σ (T ) ⊆ U , and g(z) is the power series expansion of f (z) valid in U ,
then f (T ) = g(T ), where g(T ) is defined by the power series calculus.
(5) The calculus is continuous in the following sense. If U is a neighborhood
of σ (T ), if {fn } is a sequence in Hol(U ), f ∈ Hol(U ), and fn → f
uniformly on compact subsets of U , then fn (T ) → f (T ) in the operator
norm of B(X ).
The Origins of Operator-Theoretic Approaches to Function Theory 7
Another way of thinking of the Riesz–Dunford functional calculus is by
way of Runge’s theorem, which implies that, for any open set U in C, every
function f ∈ Hol(U ) is the limit, uniformly on compact subsets of U , of a
sequence of rational functions rn with poles off U . As a consequence, property
(5) can be used to define f (T ) as the norm limit of rn (T ).
The holomorphic functional calculus is one of the most important tools
in operator theory and is central to this book. We shall frequently want to
apply a holomorphic function of d complex variables to a d-tuple of pairwise
commuting operators; for this we need a generalization of the Riesz–Dunford
functional calculus. Such a generalization does indeed exist; the elaboration
of this theory is a heroic chapter in the history of operator theory. It is more
intricate than the one-variable theory and requires a significant input from the
theory of several complex variables. It was developed over several decades by
many mathematicians and, while “multivariable spectral theory” remains an
active research topic, the foundations of the functional calculus now appear to
be in a definitive form.
For a holomorphic function f of d variables and a pairwise commuting
d-tuple T = (T 1 , . . . , T d ) of operators on a Banach space X , to define what
we mean by f (T ) one strategy is again to require f to be holomorphic on
the spectrum of T and to invoke an integral representation formula. To do
this one must overcome some technical difficulties, the first of which is to
find the appropriate notion of spectrum of T . For a start, the spectrum should
contain the joint eigenvalues of T . We say that a non-zero vector x ∈ X is
a joint eigenvector of T and that λ = (λ1 , . . . , λd ) is a corresponding joint
eigenvalue of T if T j x = λj x for j = 1, . . . , d. If dim X < ∞ (and X = {0})
then T has at least one joint eigenvalue. For let λ1 be an eigenvalue of T 1 and
let E1 be the corresponding eigenspace {x ∈ X : T 1 x = λ1 x}. Since the T j
commute with T 1 , the space E1 is invariant under each T j and the operators
T j |E1 commute pairwise. Let λ2 be an eigenvalue of T 2 |E1 and let E2 be the
corresponding eigenspace {x ∈ E1 : T 2 x = λ2 x}. Continuing in this way we
arrive at a point λ = (λ1 , . . . , λd ) ∈ Cd and a non-zero subspace Ed of X such
that T j x = λj x for j = 1, . . . , d and all x ∈ Ed . This point λ ∈ Cd is a joint
eigenvalue of T . However, as we already know from the case d = 1, when X
is infinite-dimensional there need not be any joint eigenvalues of T .
In 1953 Georgii Evgen’evich Shilov [187] constructed a functional calculus
for several elements of a commutative Banach algebra. If T is a commuting
d-tuple of elements of B(X ), one can choose any commutative Banach
subalgebra A of B(X ) containing the elements of T and then define the
spectrum of T with respect to A by
σA (T ) = Cd \ ρA (T ), (1.5)
8 Commutative Theory
where
ρA (T ) = λ ∈ Cd : there exist R 1 , . . . , R d ∈ A such that
(T 1 − λ1 )R 1 + · · · + (T d − λd )R d = 1 . (1.6)
(i) T ≥ 0;
(ii) T ∗ = T and
σ (T ) ⊆ {x ∈ R: x ≥ 0};
0 = Z(x + λy), x + λy
= Zx, x + λ Zy, x + λ̄ Zx, y + |λ|2 Zy, y
= λ Zy, x + λ̄ Zx, y .
For Hilbert spaces H and K, there is a useful notation that allows the
identification of elements of the algebraic tensor product H ⊗ K with operators
from K into H. If u ∈ H and v ∈ K we define u ⊗ v ∈ B(K, H) by the formula
P (x + y) = x
where the sum on the right-hand side converges in the strong operator
topology.3
Theorem 1.18. (The spectral theorem for normal operators) If H is a
Hilbert space and N is a normal operator on H, then there exists a B(H)-
valued spectral measure E defined on the σ -algebra of Borel subsets of σ (N )
such that
N= z dE(z). (1.19)
σ (N )
See [140, section 31.6, theorem 15] or [60, theorem IX.2.2] for a proof.
An immediate corollary of the spectral theorem is the following result, a
cornerstone of the theory of C ∗ -algebras. For K a compact subset of C, let
C(K) denote the algebra of continuous functions on K with supremum norm
and involution f ∗ = f¯, the complex conjugate of f . For a normal operator
N on H, let C ∗ (N ) denote the smallest norm-closed subalgebra of B(H)
containing 1 and N and satisfying X ∗ ∈ C ∗ (N ) whenever X ∈ C ∗ (N ).
3 A sequence (T ) of operators converges to T ∈ B(H) in the strong operator topology if, for
n
every x ∈ H, Tn x − T x → 0 as n → ∞.
The Origins of Operator-Theoretic Approaches to Function Theory 13
Theorem 1.20. (The continuous functional calculus for normal operators)
Let H be a Hilbert space, N be a normal operator on H and E be the spectral
measure of N as in Theorem 1.18. For ϕ ∈ C(σ (N)), define ϕ(N) ∈ ϕ ∈
C(σ (N)) by the formula
ϕ(N) = ϕ(z) dE(z). (1.21)
σ (N )
The map ρ: C(σ (N)) → B(H) given by the formula ρ(ϕ) = ϕ(N) is an
isometric ∗-isomorphism of C(σ (N)) onto C ∗ (N ), that is, ρ is an algebra-
isomorphism satisfying ρ(ϕ̄) = ρ(ϕ)∗ and ρ(ϕ) = ϕ for all ϕ ∈ B(H).
This theorem illustrates the fact that not all functional calculi are holo-
morphic functional calculi. Nevertheless, the continuous functional calculus
for normal operators agrees with the Riesz–Dunford functional calculus on
functions that are holomorphic on a neighborhood of the spectrum of a normal
operator. In particular, if N is a normal operator, f ∈ Rat(σ (N )), and f (N) is
defined by equation (1.3), then
f (N ) = f (z) dE(z). (1.22)
σ (N )
4 In [72, chapter V] he wrote “By the depth and novelty of its ideas, it is a turning point in the
history of Functional Analysis, and indeed deserves to be considered as the very first paper
published in that discipline.”
14 Commutative Theory
One could say that 2 is the “purest” Hilbert space, in that it exhibits the
geometry of Hilbert space in the presence of minimal mathematical structure.
Other Hilbert spaces typically have elements that are functions or operators
of some kind, which may involve delicate notions such as measurability or
analyticity, and such spaces often occur in applications. However, all separable
infinite-dimensional Hilbert spaces are isomorphic ([172, section 82] or [214,
theorem 4.19]), and so it is natural to try to gain insight into operators on a
general Hilbert space by studying them on the “bare” Hilbert space 2 .
One operator on 2 that has been much studied is the unilateral shift S. This
is the operator defined on 2 by the formula
This condition ensures that the series (1.24) converges locally uniformly on D
and thereby defines f as an analytic function in D. The inner product on H2 is
defined by
∞
∞
∞
an zn , bn zn = an b̄n . (1.26)
n=0 n=0 n=0
Condition (1.25) also ensures that the series on the right-hand side of equation
(1.26) converges, so that the inner product is well defined.
Thus elements of H2 can be regarded either as power series or as analytic
functions on D. It is useful to have a criterion for membership of H2 , and an
expression for the norm, in terms of function values (as distinct from Taylor
coefficients). In fact, for an analytic function f on D to belong to H2 , it is
necessary and sufficient that
2π
lim sup |f (reiθ )|2 dθ < ∞. (1.27)
r→1− 0
The Origins of Operator-Theoretic Approaches to Function Theory 15
When this inequality holds, the lim sup is in fact a limit as r → 1−, and
1
2π 2
dθ
f H2 = lim |f (re )|
iθ 2
. (1.28)
r→1− 0 2π
There are numerous ways to generalize the notion of spectral sets to several
variables. One idea is to juice up the inequality (1.35) to all functions f
holomorphic on a neighborhood of K.
Lemma 1.38. Let T ∈ B(H) and assume that K ⊆ C is compact. Then K is
a spectral set for T if and only if σ (T ) ⊆ K and, for all f holomorphic in a
neighborhood of K,
Proof. The proof of necessity is immediate from the observation that the
elements of Rat(K) are holomorphic on a neighborhood of K.
To prove the converse, assume that σ (T ) ⊆ K and inequality (1.35) holds.
Fix a neighborhood U of K and f ∈ Hol(U ). By Runge’s theorem, there exists
18 Commutative Theory
a sequence of rational functions {fn } in Rat(U − ) such that fn converges locally
uniformly to f on U . By the continuity of the functional calculus,
The linear map Df (z0 ), when there is one, is unique and is called the derivative
or Fréchet derivative of f at z0 . It is also denoted by f (z0 ). The mapping f
is holomorphic on if it is differentiable at every point of .
It is clear that if f is holomorphic then the function of one variable
λ → f (λ, z2 , z3 , . . . , zd )
is analytic in λ for any fixed values of z2 , . . . , zd , and likewise for each of the
d co-ordinates. One can summarize this statement by saying that holomorphic
maps are holomorphic in each variable separately. Remarkably enough, the
converse statement is also true: any separately holomorphic map on an open
set in Cd is holomorphic. This is a theorem of Hartogs, which caused a surprise
when it was proved in 1906. Another criterion for a function to be holomorphic
is that it have a convergent power series expansion in a neighborhood of
every point.
Lemma 1.38 suggests the following definition of spectral sets in several
variables. For a compact set K in Cd , let Hol(K) denote the set of functions
holomorphic on a neighborhood of K (the neighborhood can depend on the
function).
Definition 1.41. Let H be a Hilbert space, let T be a pairwise commuting d-
tuple of operators on H and let K be a compact subset of Cd . We say that K
is a spectral set for T if σ (T ) ⊆ K and, for all f ∈ Hol(K),
f (T ) = PH f (N)|H,
and then define SB(H,K) (D), the B(H, K)-valued Schur class on D, by
SB(H,K) (D) = ball H∞
B(H,K) (D).
Of course, when K = H, we abbreviate H∞ ∞
B(H,K) (D) to HB(H) (D). We give a
brief explanation of the idea of a network realization in Section 2.5.
Theorem 1.58. (Network realization formula) A mapping ϕ: D → B(H, K)
belongs to SB(H,K) (D) if and only if there exist a Hilbert space M and an
isometry
A B
V = : H ⊕ M → K ⊕ M,
C D
such that, for all λ ∈ D,
ϕ(λ) = A + Bλ(1 − Dλ)−1 C. (1.59)
Helton’s seminal observation created new synergies between the operator-
theoretic function theory and mathematical control theory communities that
continue to the present.
Remarks 1.60.
(i) If H = K and
1 1
A = −T , B = (1 − T T ∗ ) 2 , C = (1 − T ∗ T ) 2 , and D = T ∗ ,
The Origins of Operator-Theoretic Approaches to Function Theory 25
then equation (1.59) becomes equation (1.56), and V is not just isomet-
ric, it is unitary.
(ii) In the special case when dim H = dim K < ∞, dim M < ∞, and ϕ is
defined by equation (1.59), the condition that V be isometric implies
that ϕ is a rational inner function, that is, ϕ(τ ) is unitary for all τ ∈ T
and det ϕ is a scalar rational inner function (a finite Blaschke product).
Conversely, if ϕ ∈ H∞ B(H,K) is a rational inner function, then there exists
M with dim M < ∞ and an isometric V such that equation (1.59) holds.
(iii) In the case when H = C, we may assume that
A B a 1⊗β
V = = ,
C D γ ⊗1 D
where a ∈ C, β ∈ M, γ ∈ M, and D ∈ B(M). In this case the realization
formula (1.59) becomes
ϕ(λ) = a + λ (1 − Dλ)−1 γ , β .
M
Mϕ f = ϕf for f ∈ H2 . (1.61)
The following lemma asserts that this formula does indeed define a bounded
operator.
Lemma 1.62. If ϕ ∈ H∞ (D) and Mϕ is defined by equation (1.61), then
Mϕ f ∈ H2 whenever f ∈ H2 . Furthermore, Mϕ (as an operator acting on
H2 ) has norm less than or equal to ϕD .
Proof. By equation (1.28), if f ∈ H2 , then
dt
Mϕ f 2 = lim |ϕ(reit )f (reit )|2
r→1− 2π
dt
≤ ϕD lim
2
|f (reit )|2
r→1− 2π
= ϕ2D f 2H2 .
f ∈ H orthogonal to the
To prove the reverse inclusion, consider any function 2
left-hand side of the relation (1.74). Then f , sλj = 0 for each j , which is to
say that f vanishes at all the points λ1 , . . . , λn . It follows that the function
T = PM Mψ |M.
1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside the
United States, check the laws of your country in addition to the terms
of this agreement before downloading, copying, displaying,
performing, distributing or creating derivative works based on this
work or any other Project Gutenberg™ work. The Foundation makes
no representations concerning the copyright status of any work in
any country other than the United States.
• You pay a royalty fee of 20% of the gross profits you derive from
the use of Project Gutenberg™ works calculated using the
method you already use to calculate your applicable taxes. The
fee is owed to the owner of the Project Gutenberg™ trademark,
but he has agreed to donate royalties under this paragraph to
the Project Gutenberg Literary Archive Foundation. Royalty
payments must be paid within 60 days following each date on
which you prepare (or are legally required to prepare) your
periodic tax returns. Royalty payments should be clearly marked
as such and sent to the Project Gutenberg Literary Archive
Foundation at the address specified in Section 4, “Information
about donations to the Project Gutenberg Literary Archive
Foundation.”
• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.
1.F.
1.F.4. Except for the limited right of replacement or refund set forth in
paragraph 1.F.3, this work is provided to you ‘AS-IS’, WITH NO
OTHER WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR ANY PURPOSE.
Please check the Project Gutenberg web pages for current donation
methods and addresses. Donations are accepted in a number of
other ways including checks, online payments and credit card
donations. To donate, please visit: www.gutenberg.org/donate.
Most people start at our website which has the main PG search
facility: www.gutenberg.org.