Operator Theory Systems Theory and Scattering Theory Multidimensional Generalizations

Operator Theory: Advances and
Applications
Vol. 157
Editor:
I. Gohberg
H. G. Kaper (Argonne)
S. T. Kuroda (Tokyo)
P. Lancaster (Calgary)
L. E. Lerer (Haifa)
B. Mityagin (Columbus)
V. V. Peller (Manhattan, Kansas)
L. Rodman (Williamsburg)
J. Rovnyak (Charlottesville)
D. E. Sarason (Berkeley)
I. M. Spitkovsky (Williamsburg)
S. Treil (Providence)
H. Upmeier (Marburg)
S. M. Verduyn Lunel (Leiden)
D. Voiculescu (Berkeley)
H. Widom (Santa Cruz)
D. Xia (Nashville)
D. Yafaev (Rennes)
Honorary and Advisory
Editorial Board:
C. Foias (Bloomington)
P. R. Halmos (Santa Clara)
T. Kailath (Stanford)
P. D. Lax (New York)
M. S. Livsic (Beer Sheva)
Editorial Office:
School of Mathematical
Sciences
Tel Aviv University
Ramat Aviv, Israel
Editorial Board:
D. Alpay (Beer-Sheva)
J. Arazy (Haifa)
A. Atzmon (Tel Aviv)
J. A. Ball (Blacksburg)
A. Ben-Artzi (Tel Aviv)
H. Bercovici (Bloomington)
A. Bttcher (Chemnitz)
K. Clancey (Athens, USA)
L. A. Coburn (Buffalo)
K. R. Davidson (Waterloo, Ontario)
R. G. Douglas (College Station)
A. Dijksma (Groningen)
H. Dym (Rehovot)
P. A. Fuhrmann (Beer Sheva)
B. Gramsch (Mainz)
G. Heinig (Chemnitz)
J. A. Helton (La Jolla)
M. A. Kaashoek (Amsterdam)
Birkhuser Verlag
Basel
.
Boston
.
Berlin
Operator Theory, Systems Theory
and Scattering Theory:
Multidimensional Generalizations
Daniel Alpay
Victor Vinnikov
Editors
A CIP catalogue record for this book is available from the
Library of Congress, Washington D.C., USA
Bibliographic information published by Die Deutsche Bibliothek
Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available
in the Internet at <http://dnb.ddb.de>.
ISBN 3-7643-7212-5 Birkhuser Verlag, Basel Boston Berlin
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms
or in other ways, and storage in data banks. For any kind of use permission of the copyright owner must be obtained.
2005 Birkhuser Verlag, P.O. Box 133, CH-4010 Basel, Switzerland
Part of Springer Science+Business Media
Printed on acid-free paper produced from chlorine-free pulp. TCF
Cover design: Heinz Hiltbrunner, Basel
Printed in Germany
ISBN 10: 3-7643-7212-5
ISBN 13: 978-3-7643-7212-5
9 8 7 6 5 4 3 2 1 www.birkhauser.ch
Editors:
Daniel Alpay
Victor Vinnikov
Department of Mathematics
Ben-Gurion University of the Negev
P.O. Box 653
Beer Sheva 84105
Israel
e-mail: dany@math.bgu.ac.il
vinnikov@math.bgu.ac.il
2000 Mathematics Subject Classification 47A13, 47A40, 93B28
Contents
Editorial Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
J. Ball and V. Vinnikov
Functional Models for Representations of the Cuntz Algebra . . . . . . . . . 1
T. Banks, T. Constantinescu and J.L. Johnson
Relations on Non-commutative Variables and
Associated Orthogonal Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
M. Bessmertny
Functions of Several Variables in the Theory of Finite
Linear Structures. Part I: Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
S. Eidelman and Y. Krasnov
Operator Methods for Solutions of PDEs
Based on Their Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
D.S. Kalyuzhny-Verbovetzki
On the Bessmertny Class of Homogeneous Positive
Holomorphic Functions on a Product of Matrix Halfplanes . . . . . . . . . . . 139
V. Katsnelson and D. Volok
Rational Solutions of the Schlesinger System and
Isoprincipal Deformations of Rational Matrix Functions II . . . . . . . . . . . 165
M.E. LunaElizarrar as and M. Shapiro
Preservation of the Norms of Linear Operators Acting
on some Quaternionic Function Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
P. Muhly and B. Solel
Hardy Algebras Associated with W
-correspondences
(Point Evaluation and Schur Class Functions) . . . . . . . . . . . . . . . . . . . . . . . . 221
M. Putinar
Notes on Generalized Lemniscates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
M. Reurings and L. Rodman
One-sided Tangential Interpolation for HilbertSchmidt
Operator Functions with Symmetries on the Bidisk . . . . . . . . . . . . . . . . . . 267
F.H. Szafraniec
Favards Theorem Modulo an Ideal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Operator Theory:
Advances and Applications, Vol. 157, viixvi
c 2005 Birkhauser Verlag Basel/Switzerland
Editorial Introduction
Daniel Alpay and Victor Vinnikov
La seduction de certains problèmes vient de leur
defaut de rigueur, comme des opinions discor-
dantes quils suscitent: autant de dicultes dont
sentiche lamateur dInsoluble.
(Cioran, La tentation dexister, [29, p. 230])
This volume contains a selection of papers on various aspects of operator theory
in the multi-dimensional case. This last term includes a wide range of situations
and we review the one variable case rst.
An important player in the single variable theory is a contractive analytic func-
tion on the open unit disk. Such functions, often called Schur functions, have a
rich theory of their own, especially in connection with the classical interpolation
problems. They also have dierent facets arising from their appearance in dierent
areas, in particular as:
characteristic operator functions, in operator model theory. Pioneering works
include the works of Livsic and his collaborators [54], [55], [25], of Sz. Nagy
and Foias [61] and of de Branges and Rovnyak [23], [22].
scattering functions, in scattering theory. We mention in particular the Lax
Phillips approach (see [53]), the approach of de Branges and Rovnyak (see
[22]) and the inverse scattering problem of network theory [38]; for a solution
of the latter using reproducing kernel Hilbert space methods, see [8], [9].
transfer functions, in system theory. It follows from the BochnerChandra-
sekharan theorem that a system is linear, time-invariant, and dissipative if
and only if it has a transfer function which is a Schur function. For more
general systems (even multi-dimensional ones) one can make use of Schwartz
kernel theorem (see [76], [52]) to get the characterisation of invariance under
translation; see [83, p. 89, p. 130].
There are many quite dierent approaches to the study of Schur functions, their
various incarnations and related problems, yet it is basically true that there is only
one underlying theory.
viii D. Alpay and V. Vinnikov
One natural extension of the single variable theory is the time varying case, where
one (roughly speaking) replaces the complex numbers by diagonal operators and
the complex variable by a shift operator; see [7], [39].
The time varying case is still essentially a one variable theory, and the various
approaches of the standard one variable theory generalize together with their in-
terrelations. On the other hand, in the multi-dimensional case there is no longer
a single underlying theory, but rather dierent theories, some of them loosely
connected and some not connected at all. In fact, depending on which facet of
the one-dimensional case we want to generalize we are led to completely dierent
objects and borderlines between the various theories are sometimes vague. The
directions represented in this volume include:
Interpolation and realization theory for analytic functions on the polydisk.
This originates with the works of Agler [2], [1]. From the view point of sys-
tem theory, one is dealing here with the conservative version of the systems
known as the Roesser model or the FornasiniMarchesini model in the multi-
dimensional system theory literature; see [71], [46].
Function theory on the free semigroup and on the unit ball of C
N
. From
the view point of system theory, one considers here the realization problem
for formal power series in non-commuting variables that appeared rst in
the theory of automata, see Sch utzenberger [74], [75] and Fliess [44], [45]
(for a good survey see [17]), and more recently in robust control of linear
systems subjected to structured possibly time-varying uncertainty (see Beck,
Doyle and Glover [15] and Lu, Zhou and Doyle [59]). In operator theory, two
main parallel directions may be distinguished; the rst direction is along the
lines of the works of Drury [43], Frazho [47], [48], Bunce [26], and especially
the vast work of Popescu [65], [63], [64], [66], where various one-dimensional
models are extended to the case of several non-commuting operators. Another
direction is related to the representations of the Cuntz algebra and is along
the line of the works of Davidson and Pitts (see [36] and [37]) and Bratelli
and Jorgensen [24]. When one abelianizes the setting, one obtains results
on the theory of multipliers in the so-called Arveson space of the ball (see
[12]), which are closely related with the theory of complete NevanlinnaPick
kernels; see the works of Quiggin [70], McCullough and Trent [60] and Agler
and McCarthy [3]. We note also connections with the theory of wavelets and
with system theory on trees; see [16], [10].
Hyponormal operators, subnormal operators, and related topics. Though nom-
inally dealing with a single operator, the theory of hyponormal operators and
of certain classes of subnormal operators has many features in common with
multivariable operator theory. We have in mind, in particular, the works of
Putinar [68], Xia [81], and Yakubovich [82]. For an excellent general survey of
the theory of hyponormal operators, see [80]. Closely related is the principal
function theory of Carey and Pincus, which is a far reaching development
of the theory of Krens spectral shift function; see [62], [27], [28]. Another
Editorial Introduction ix
closely related topic is the study of multi-dimensional moment problems; of
the vast literature we mention (in addition to [68]) the works of Curto and
Fialkow [33], [34] and of Putinar and Vasilescu [69].
Hyperanalytic functions and applications. Left (resp. right) hyperanalytic
functions are quaternionic-valued functions in the kernel of the left (resp.
right) CauchyFueter operator (these are extensions to R
4
of the operator
x
+ i

y
). The theory is non-commutative and a supplementary diculty
is that the product of two (say, left) hyperanalytic functions need not be
left hyperanalytic. Setting the real part of the quaternionic variable to be
zero, one obtains a real analytic quaternionic-valued function. Conversely,
the CauchyKovalevskaya theorem allows to associate (at least locally) to
any such function a hyperanalytic function. Identifying the quaternions with
C
2
one obtains an extension of the theory of functions of one complex variable
to maps from (open subsets of) C
2
into C
2
. Rather than two variables there
are now three non-commutative non-independent hyperanalytic variables and
the counterparts of the polynomials z
n
1
1
z
n
2
2
are now non-commutative poly-
nomials (called the Fueter polynomials) in these hyperanalytic variables. The
original papers of Fueter (see, e.g., [50], [49]) are still worth a careful reading.
Holomorphic deformations of linear dierential equations. One approach to
study of non-linear dierential equations, originating in the papers of Schle-
singer [73] and Garnier [51], is to represent the non-linear equation as the
compatibility condition for some over-determined linear dierential system
and consider the corresponding families (so-called deformations) of ordinary
linear equations. From the view point of this theory, the situation when the
linear equations admit rational solutions is exceptional: the non-resonance
conditions, the importance of which can be illustrated by Bolibruchs coun-
terexample to Hilberts 21st problem (see [11]), are not met. However, anal-
ysis of this situation in terms of the system realization theory may lead to
explicit solutions and shed some light on various resonance phenomena.
The papers in the present volume can be divided along these categories as follows:
Polydisk function theory:
The volume contains a fourth part of the translation of the unpublished thesis [18]
of Bessmertny, which foreshadowed many subsequent developments and contains
a wealth of ideas still to be explored. The other parts are available in [20], [19]
and [21]. The paper of Reurings and Rodman, One-sided tangential interpolation
for HilbertSchmidt operator functions with symmetries on the bidisk, deals with
interpolation in the bidisk in the setting of H
2
rather than of H
.
Non-commutative function theory and operator theory:
The rst paper in this category in the volume is the paper of Ball and Vinnikov,
Functional models for representations of the Cuntz algebra. There, the authors
develop functional models and a certain theory of Fourier representation for a rep-
resentation of the Cuntz algebra (i.e., a row unitary operator). Next we have the
x D. Alpay and V. Vinnikov
paper of Banks, Constantinescu and Johnson, Relations on non-commutative vari-
ables and associated orthogonal polynomials, where the authors survey various
settings where analogs of classical ideas concerning orthogonal polynomials and
associated positive kernels occur. The paper serves as a useful invitation and ori-
entation for the reader to explore any particular topic more deeply. In the paper
of Kalyuzhny-Verbovetzki, On the Bessmertny class of homogeneous positive
holomorphic functions on a product of matrix halfplanes, a recent investigation
of the author on the Bessmertny class of operator-valued functions on the open
right poly-halfplane which admit a so-called long resolvent representation (i.e., a
Schur complement formula applied to a linear homogeneous pencil of operators
with positive semidenite operator coecients), is generalized to a more general
non-commutative domain, a product of matrix halfplanes. The study of the Bess-
mertny class (as well as its generalization) is motivated by the electrical networks
theory: as shown by M.F. Bessmertny [18], for the case of matrix-valued func-
tions for which nite-dimensional long resolvent representations exist, this class
is exactly the class of characteristic functions of passive electrical 2n-poles where
impedances of the elements of a circuit are considered as independent variables.
Finally, in the paper Hardy algebras associated with W
-correspondences (point
evaluation and Schur class functions), Muhly and Solel deal with an extension of
the non-commutative theory from the point of view of non-self-adjoint operator
algebras.
Hyponormal and subnormal operators and related topics:
The paper of Putinar, Notes on generalized lemniscates, is a survey of the theory
of domains bounded by a level set of the matrix resolvent localized at a cyclic
vector. The subject has its roots in the theory of hyponormal operators on the one
hand and in the theory of quadrature domains on the other. While both topics are
mentioned in the paper, the main goal is to present the theory of these domains
(that the author calls generalized lemniscates) as an independent subject matter,
with a wealth of interesting properties and applications. The paper of Szafraniec,
Orthogonality of polynomials on algebraic sets, surveys recent extensive work of
the author and his coworkers on polynomials in several variables orthogonal on an
algebraic set (or more generally with respect to a positive semidenite functional)
and three term recurrence relations. As it happens often the general approach
sheds new light also on the classical one-dimensional situation.
Hyperanalytic functions:
In the paper Operator methods for solutions of dierential equations based on
their symmetries, Eidelman and Krasnov deal with construction of explicit solu-
tions for some classes of partial dierential equations of importance in physics, such
as evolution equations, homogeneous linear equations with constant coecients,
and analytic systems of partial dierential equations. The method used involves
an explicit construction of the symmetry operators for the given partial dieren-
tial operator and the study of the corresponding algebraic relations; the solutions
Editorial Introduction xi
of the partial dierential equation are then obtained via the action of the sym-
metry operators on the simplest solution. This allows to obtain representations
of Cliord-analytic functions in terms of power series in operator indeterminates.
LunaElizarrar as and Shapiro in Preservation of the norms of linear operators act-
ing on some quaternionic function spaces consider quaternionic analogs of some
classical real spaces and in particular compare the norms of operators in the orig-
inal space and in the quaternionic extension.
Holomorphic deformations of linear dierential equations:
This direction is represented in the present volume by the paper of Katsnelson and
Volok, Rational solutions of the Schlesinger system and rational matrix functions
II, which presents an explicit construction of the multi-parametric holomorphic
families of rational matrix functions, corresponding to rational solutions of the
Schlesinger non-linear system of partial dierential equations.
There are many other directions that are not represented in this volume. Without
the pretense of even trying to be comprehensive we mention in particular:
Model theory for commuting operator tuples subject to various higher-order
contractivity assumptions; see [35], [67].
A multitude of results in spectral multivariable operator theory (many of
them related to the theory of analytic functions of several complex variables)
stemming to a large extent from the discovery by Taylor of the notions of the
joint spectrum [78] and of the analytic functional calculus [77] for commuting
operators (see [32] for a survey of some of these).
The work of Douglas and of his collaborators based on the theory of Hilbert
modules; see [42], [40], [41].
The work of Agler, Young and their collaborators on operator theory and
realization theory related to function theory on the symmetrized bidisk, with
applications to the two-by-two spectral NevanlinnaPick problem; see [5], [4],
[6].
Spectral analysis and the notion of the characteristic function for commuting
operators, related to overdetermined multi-dimensional systems. The main
notion is that of an operator vessel, due to Livsic; see [56], [57], [58]. This
turns out to be closely related to function theory on a Riemann surface; see
[79],[13].
The work of Cotlar and Sadosky on multievolution scattering systems, with
applications to interpolation problems and harmonic analysis in several vari-
ables; see [30], [31], [72].
Acknowledgments
This volume has its roots in a workshop entitled Operator theory, system theory
and scattering theory: multi-dimensional generalizations, 2003, which was held at
the Department of Mathematics of Ben-Gurion University of the Negev during the
period June 30July 3, 2003. It is a pleasure to thank all the participants for an
xii D. Alpay and V. Vinnikov
exciting scientic atmosphere and the Center of Advanced Studies in Mathemat-
ics of Ben-Gurion University of the Negev for its generosity and for making the
workshop possible.
References
[1] J. Agler. Interpolation, UCSD, preprint, 1987.
[2] J. Agler. On the representation of certain holomorphic functions dened on a poly-
disk. In volume 48 of Operator Theory: Advances and Applications, pp. 4766,
Birkhauser Verlag, Basel, 1990.
[3] J. Agler and J. McCarthy. Complete Nevanlinna-Pick kernels. J. Funct. Anal.,
175:111124, 2000.
[4] J. Agler, F.B. Yeh, and N.J. Young. Realization of functions into the symmetrised
bidisc. In Reproducing kernel spaces and applications, volume 143 of Oper. Theory
Adv. Appl., pp. 137. Birkhauser, Basel, 2003.
[5] J. Agler and N.J. Young. A commutant lifting theorem for a domain in C
2
and
spectral interpolation. J. Funct. Anal. 161 (1999), no. 2, 452477.
[6] J. Agler and N.J. Young. The two-by-two spectral Nevanlinna-Pick problem. Trans.
Amer. Math. Soc., 356(2):573585 (electronic), 2004.
[7] D. Alpay, P. Dewilde, and H. Dym. Lossless inverse scattering and reproducing ker-
nels for upper triangular operators. In Extension and interpolation of linear operators
and matrix functions, pp. 61135. Birkh auser, Basel, 1990.
[8] D. Alpay and H. Dym. Hilbert spaces of analytic functions, inverse scattering and
operator models, I. Integral Equation and Operator Theory, 7:589641, 1984.
[9] D. Alpay and H. Dym. Hilbert spaces of analytic functions, inverse scattering and
operator models, II. Integral Equation and Operator Theory, 8:145180, 1985.
[10] D. Alpay and D. Volok. Interpolation et espace de Hardy sur larbre dyadique: le cas
stationnaire. Comptes Rendus Mathematiques, 336:293298, 2003.
[11] D.V. Anosov and A.A. Bolibruch. The Riemann-Hilbert problem. Aspects of Math-
ematics, E22. Friedr. Vieweg & Sohn, Braunschweig, 1994.
[12] W. Arveson. Subalgebras of C
algebras. III. Multivariable operator theory. Acta

Mathematica, 181:159228, 1998.
[13] J.A. Ball and V. Vinnikov. Overdetermined multi-dimensional systems: state space
and frequency domain methods. In Mathematical Systems Theory in Biology, Com-
munications, Computation, and Finance (Ed. J. Rosenthal and D.S. Gilliam), IMA
Volumes in Mathematics and its Applications, Springer, BerlinNew York, 2003.
[14] C.L. Beck. On formal power series representations for uncertain systems. IEEE
Trans. Auto. Contr. 46:2 (2001), 314319.
[15] C.L. Beck, J.C. Doyle, and K. Glover. Model reduction of multi-dimensional and
uncertain systems. IEEE Trans. Auto. Contr. 41:10 (1996), 14061477.
[16] A. Benveniste, R. Nikoukhah, and A. Willsky. Multiscale system theory. IEEE Trans.
Circuits Systems I Fund. Theory Appl., 41(1):215, 1994.
[17] J. Berstel and C. Reutenauer. Rational series and their languages. EATCS Mono-
graphs on Theoretical Computer Science, Springer-Verlag, BerlinNew York, 1984.
Editorial Introduction xiii
[18] M.F. Bessmertny. Functions of several complex variables in the theory of nite linear
structures. PhD thesis, Kharkov University, Kharkov, 1982. (Russian).
[19] M.F. Bessmertny. On realizations of rational matrix functions of several com-
plex variables. In Interpolation theory, systems theory and related topics (Tel
Aviv/Rehovot, 1999), volume 134 of Oper. Theory Adv. Appl., pp. 157185. Birk-
h auser, Basel, 2002. Translated from the Russian by D. Alpay and V. Katsnelson.
[20] M.F. Bessmertny. On realizations of rational matrix functions of several complex
variables. II. In Reproducing kernel spaces and applications, volume 143 of Oper. The-
ory Adv. Appl., pp. 135146. Birkh auser, Basel, 2003. Translated from the Russian
by V. Katsnelson.
variables. III. In Current trends in operator theory and its applications (Blacksburg
VA, 2002), volume 149 of Oper. Theory Adv. Appl., pp. 133138. Birkhauser, Basel,
2004. Translated from the Russian by D.S. Kalyuzhny-Verbovetzki.
[22] L. de Branges and J. Rovnyak. Canonical models in quantum scattering theory. In
C. Wilcox, editor, Perturbation theory and its applications in quantum mechanics,
pp. 295392. Wiley, New York, 1966.
[23] L. de Branges and J. Rovnyak. Square summable power series. Holt, Rinehart and
Winston, New York, 1966.
[24] O. Bratteli and P. Jorgensen. Wavelets through a looking glass. Applied and Numer-
ical Harmonic Analysis. Birkhauser Boston Inc., Boston, MA, 2002. The world of
the spectrum.
[25] M.S. Brodski and M.S. Livsic. Spectral analysis of non-self-adjoint operators and
intermediate systems. Uspehi Mat. Nauk (N.S.), 13(1(79)):385, 1958.
[26] John W. Bunce. Models for n-tuples of noncommuting operators. J. Funct. Anal.,
57(1):2130, 1984.
[27] R.W. Carey. A unitary invariant for pairs of self-adjoint operators. J. Reine Angew.
Math., 283/284:294312, 1976.
[28] R.W. Carey and J.D. Pincus. Mosaics, principal functions, and mean motions in Von
Neumann algebras. Acta Math., 138:153218, 1977.
[29] E.M. Cioran. La tentation dexister. Collection idees. Gallimard, 1974. First pub-
lished in 1956.
[30] M. Cotlar and C. Sadosky. Generalized Bochner Theorem in algebraic scattering
systems. In Analysis at Urbana vol.II. London Math. Soc. Lecture Notes Ser. 138
(Cambridge Univ. Press, Cambridge, 1989), pp. 144169.
[31] M. Cotlar and C. Sadosky. Integral representations of bounded Hankel forms dened
in scattering systems with a multi-dimensional evolution group. Contributions to Op-
erator Theory and its Applications (Mesa, AZ, 1987) (Ed. I. Gohberg, J.W. Helton
and L. Rodman), pp. 357-375, volume 35 of Operator Theory: Advances and Appli-
cations, Birkhauser, BaselBoston, 1988.
[32] R.E. Curto. Applications of several complex variables to multiparameter spectral
theory. In Surveys of some recent results in operator theory, Vol. II, pp. 2590,
Longman Sci. Tech., Harlow, 1988.
[33] R.E. Curto and L.A. Fialkow. Solution of the truncated complex moment problem
for at data. Mem. Amer. Math. Soc. 119 (1996), no. 56.
xiv D. Alpay and V. Vinnikov
[34] R.E. Curto and L.A. Fialkow. Flat extensions of positive moment matrices: recur-
sively generated relations. Mem. Amer. Math. Soc. 136 (1998), no. 648.
[35] R.E. Curto and F.-H. Vasilescu. Standard operator models in the polydisc. II. Indi-
ana Univ. Math. J., 44(3):727746, 1995.
[36] K.R. Davidson and D.R. Pitts. The algebraic structure of non-commutative analytic
Toeplitz algebras. Math. Ann., 311(2):275303, 1998.
[37] K.R. Davidson and D.R. Pitts. Invariant subspaces and hyper-reexivity for free
semigroup algebras. Proc. London Math. Soc. (3), 78(2):401430, 1999.
[38] P. Dewilde and H. Dym. Lossless inverse scattering, digital lters, and estimation
theory. IEEE Trans. Inform. Theory, 30(4):644662, 1984.
[39] P. Dewilde and A.-J. van der Veen. Time-varying systems and computations. Kluwer
Academic Publishers, Boston, MA, 1998.
[40] R.G. Douglas. Models and resolutions for Hilbert modules. In Multivariable operator
theory (Seattle, WA, 1993), volume 185 of Contemp. Math., pp. 109131. Amer.
Math. Soc., Providence, RI, 1995.
[41] R.G. Douglas and G. Misra. Quasi-free resolutions of Hilbert modules. Integral Equa-
tions Operator Theory 47 (2003), no. 4, 435456.
[42] R.G. Douglas and V.I. Paulsen. Hilbert modules over function algebras. Longman
Sci. Tech., Harlow, 1989.
[43] S.W. Drury. A generalization of von Neumanns inequality to the complex ball. Proc.
Amer. Math. Soc., 68(3):300304, 1978.
[44] M. Fliess. Matrices de Hankel. J. Math. Pures Appl., 53:197222, 1974 & erratum
54, 1975.
[45] M. Fliess. Sur divers produits de series formelles. Bull. Soc. Math. France 102 (1974),
181191.
[46] E. Fornasini and G. Marchesini. Doubly-indexed dynamical systems: state-space
models and structural properties. Math. Systems Theory, 12(1):5972, 1978/79.
[47] A.E. Frazho. Models for noncommuting operators. J. Funct. Anal., 48(1):111, 1982.
[48] A.E. Frazho. Complements to models for noncommuting operators. J. Funct. Anal.,
59(3):445461, 1984.
[49] R. Fueter. Analytische Funktionen einer Quaternionenvariablen. Comment. Math.
Helv., 4:920, 1932.
[50] R. Fueter. Die Theorie der regularen Funktionen einer quaternionen Variablen. In
Comptes rendus du congrès international des mathematiciens, Oslo 1936, Tome I,
pp. 7591, 1937.
[51] R. Garnier. Sur une classe dequations dierentielles dont les integrales generales
ont leurs points critiques xes. Compt. Rend. de lAcademie des Sciences, Paris,
151:205208, 1910.
[52] I.M. Gelfand and N.Y. Vilenkin. Generalized functions. Volume 4: Applications of
harmonic analysis. Translated from Russian. Academic Press, New York, 19641968.
[53] P.D. Lax and R.S. Phillips. Scattering theory (revised edition), volume 26 of Pure
and Applied Mathematics. Academic Press, New York, 1989.
[54] M.S. Livsic. On the theory of isometric operators with equal deciency indices. Dok-
lady Akad. Nauk SSSR (N.S.), 58:1315, 1947.
Editorial Introduction xv
[55] M.S. Livsic. On the theory of the elementary divisors of non-Hermitian operators.
Doklady Akad. Nauk SSSR (N.S.), 60:1720, 1948.
[56] M.S. Livsic. The inverse problem for the characteristic functions of several commut-
ing operators. Integral Equations Operator Theory, 2(2):264286, 1979.
[57] M.S. Livsic. Commuting nonselfadjoint operators and mappings of vector bundles on
algebraic curves, volume 19 of Operator Theory: Advances and Applications, pp.
275279. Birkh auser Verlag, Basel, 1986.
[58] M.S. Livsic, N. Kravitski, A. Markus, and V. Vinnikov. Commuting nonselfadjoint
operators and their applications to system theory. Kluwer, 1995.
[59] W.M. Lu, K. Zhou, and J.C. Doyle. Stabilization of uncertain linear systems: An
LFT approach. IEEE Trans. Auto. Contr. 41:1 (1996), 5065.
[60] S. McCullough and T. Trent. Invariant subspaces and Nevanlinna-Pick kernels. J.
Funct. Anal., 178(1):226249, 2000.
[61] B. Sz. Nagy and C. Foias. Harmonic analysis of operators on Hilbert spaces.
Akademia Kiado, Budapest, 1966.
[62] J.D. Pincus. Some applications of operator valued analytic functions of two complex
variables. In Linear operators and approximation (Proc. Conf., Math. Res. Inst.,
Oberwolfach, 1971), pp. 6879. Internat. Ser. Numer. Math., Vol. 20. Birkhauser,
Basel, 1972.
[63] G. Popescu. Characteristic functions for innite sequences of noncommuting opera-
tors. J. Operator Theory, 22(1):5171, 1989.
[64] G. Popescu. Isometric dilations for innite sequences of noncommuting operators.
Trans. Amer. Math. Soc., 316(2):523536, 1989.
[65] G. Popescu. Models for innite sequences of noncommuting operators. Acta Sci.
Math. (Szeged), 53(3-4):355368, 1989.
[66] G. Popescu. Multi-analytic operators on Fock spaces. Math. Ann., 303(1):3146,
1995.
[67] S. Pott. Standard models under polynomial positivity conditions. J. Operator The-
ory, 41:365389, 1999.
[68] M. Putinar. Extremal solutions of the two-dimensional L-problem of moments. J.
Funct. Anal., 136(2):331364, 1996.
[69] M. Putinar and F.-H. Vasilescu. Solving moment problems by dimensional extension.
Ann. of Math. (2) 149 (1999), no. 3, 10871107.
[70] P. Quiggin. For which reproducing kernel Hilbert spaces is Picks theorem true?
Integral Equations Operator Theory, 16:244266, 1993.
[71] R. Roesser. A discrete state-space model for linear image processing. IEEE Trans.
Automatic Control, AC20:110, 1975.
[72] C. Sadosky. Liftings of kernels shift-invariant in scattering systems. In Holomorphic
Spaces (Ed. S. Axler, J.E. McCarthy and D. Sarason), Mathematical Sciences Re-
search Institute Publications Vol. 33, Cambridge University Press, 1998, pp. 303336.
[73] L. Schlesinger.

Uber die Losungen gewisser linearer Dierentialgleichungen als Funk-
tionen der singularen Punkte. Journal f ur reine und angew. Math, 129:287294, 1905.
[74] M.P. Sch utzenberger. On the denition of a family of automata. Information and
Control, 4:245270, 1961.
xvi D. Alpay and V. Vinnikov
[75] M.P. Sch utzenberger. Certain elementary families of automata. Proceedings of sym-
posium on mathematical theory of automata, Polytechnic Institute Brooklyn (1962),
139153.
[76] L. Schwartz. Theorie des distributions. Publications de lInstitut de Mathematique
de lUniversite de Strasbourg, No. IX-X. Nouvelle edition, entièrement corrigee, re-
fondue et augmentee. Hermann, Paris, 1966.
[77] J.L. Taylor. The analytic-functional calculus for several commuting operators. Acta
Math., 125:138, 1970.
[78] J.L. Taylor. A joint spectrum for several commuting operators. J. Functional Anal-
ysis, 6:172191, 1970.
[79] V. Vinnikov. Commuting operators and function theory on a Riemann surface. In
Holomorphic spaces (Berkeley, CA, 1995), pp. 445476. Cambridge Univ. Press,
Cambridge, 1998.
[80] A. L. Vol
berg, V. V. Peller, and D. V. Yakubovich. A brief excursion into the theory

of hyponormal operators. Algebra i Analiz, 2(2):138, 1990.
[81] D. Xia. On pure subnormal operators with nite rank self-commutators and related
operator tuples. Integral Equations Operator Theory, 24(1):106125, 1996.
[82] D.V. Yakubovich. Subnormal operators of nite type. II. Structure theorems. Rev.
Mat. Iberoamericana, 14(3):623681, 1998.
[83] A. Yger. Theorie et analyse du signal. Mathematiques appliquees. Ellipses,

Editions
Marketing S.A., 32 rue Bargue, Paris 15
e
, 1999.
Daniel Alpay and Victor Vinnikov
Beer-Sheva, Israel
e-mail: dany@math.bgu.ac.il
e-mail: vinnikov@math.bgu.ac.il
Operator Theory:
Advances and Applications, Vol. 157, 160
Functional Models for Representations
of the Cuntz Algebra
Joseph A. Ball and Victor Vinnikov
Abstract. We present a functional model, the elements of which are formal
power series in a pair of d-tuples of non-commuting variables, for a row-unitary
d-tuple of operators on a Hilbert space. The model is determined by a weight-
ing matrix (called a Haplitz matrix) which has both non-commutative Han-
kel and Toeplitz structure. Such positive-denite Haplitz matrices then serve
to classify representations of the Cuntz algebra O
d
with specied cyclic sub-
space up to unitary equivalence. As an illustration, we compute the weighting
matrix for the free atomic representations studied by Davidson and Pitts and
the related permutative representations studied by Bratteli and Jorgensen.
Mathematics Subject Classication (2000). Primary: 47A48; Secondary: 93C35.
1. Introduction
Let U be a unitary operator on a Hilbert space / and let c be a subspace of /.
Dene a map from / to a space of formal Fourier series f(z) =
n=
f
n
z
n
by
: k
n=
(P
L
|
n
k)z
n
where P
L
is the orthogonal projection onto the subspace c /. Note that (k) = 0
if and only if k is orthogonal to the smallest reducing subspace for | containing
the subspace c; in particular, is injective if and only if c is -cyclic for |,
i.e., the smallest subspace reducing for | and containing c is the whole space /.
Denote the range of by L; note that we do not assume that maps / into norm
square-summable series L
2
(T, c) = f(z) =
n=
f
n
z
n
:
n=
|f
n
|
2
< .
The rst author is supported by NSF grant DMS-9987636; both authors are support by a grant
from the US-Israel Binational Science Foundation.
2 J.A. Ball and V. Vinnikov
Nevertheless, we may assign a norm to elements of L so as to make a coisometry:
|k|
2
1
= |P
(ker )
k|
2
K
.
Moreover, we see that if we set k =
n=
f
n
z
n
for a k / (so f
n
= P
L
|
n
k),
then
|k =
n=
(P
L
|
n1
k)z
n
=
n=
f
n1
z
n
= z
n=
f
n1
z
n1
= z
n=
f
n
z
n
= M
z
k,
i.e., the operator | is now represented by the operator M
z
of multiplication by
the variable z on the space L.
We can make this representation more explicit as follows. The standard ad-
joint
[]
of with respect to the L
2
-inner product on the target domain is dened
at least on polynomials:
_
k,
N
j=N
p
j
z
j
_
L
2
=
_
k,
[]
_
_
N
j=N
p
j
z
j
_
_
_
K
where we have set
[]
_
_
N
j=N
p
j
z
j
_
_
=
N
j=N
|
j
p
j
.
Furthermore, the range
[]
T of
[]
acting on polynomials (where we use T
to denote the subspace of L
2
(T, c) consisting of trigonometric polynomials with
coecients in c) is dense in (ker )
, and for
[]
p an element of this dense set
(with p T), we have
[]
p,
[]
p)
1
=
[]
p,
[]
p)
K
=
[]
p, p)
L
2.
This suggests that we set W =
[]
(well dened as an operator from the space
of c-valued polynomials T to the space L(Z, c) of formal Fourier series with co-
ecients in c) and dene a Hilbert space L
W
as the closure of WT in the inner
product
Wp, Wq)
1
W
= Wp, q)
L
2 .
The Toeplitz structure of W (i.e., the fact that W
i,j
= P
L
|
ji
[
L
depends only on
the dierence i j of the indices) implies that the operator M
z
of multiplication
Functional Models 3
by z is isometric (and in fact unitary) on L
W
. Conversely, starting with a positive
semidenite Toeplitz matrix [W
i,j
] with W
i,j
=

W
ij
, we may form a space L
W
and associated unitary operator |
W
equal to the multiplication operator M
z
acting
on L
W
as a functional model for a unitary operator. While the space L
W
in
general consists only of formal Fourier series and there may be no bounded point
evaluations for the elements of the space, evaluation of any one of the Fourier
coecients is a bounded operator on the space, and gives the space at least the
structure of a formal reproducing kernel Hilbert space, an L
2
-version of the usual
reproducing kernel Hilbert spaces of analytic functions arising in many contexts;
we develop this idea of formal reproducing kernel Hilbert spaces more fully in the
separate report [4].
Note that a unitary operator can be identied with a unitary representation
of the circle group T or of the C
-algebra C(T). Given any group G or C
-algebra
A, there are two natural problems: (1) classication up to unitary equivalence of
unitary representations of Gor of A, and (2) classication up to unitary equivalence
of unitary representations which include the specication of a -cyclic subspace.
While the solution of the rst problem is the loftier goal, the second problem is
arguably also of interest. Indeed, there are problems in operator theory where a
-cyclic subspace appears naturally as part of the structure; even when this is not
the case, a solution of the second problem often can be used as a stepping stone
to a solution of the rst problem. In the case of G = T or A = C(T), the theory of
L
W
spaces solves the second problem completely: given two unitary operators | on
/ and |
on /
with common cyclic subspace c contained in both / and /
, then
there is a unitary operator U : / /
satisfying U| = |
U and U[
L
= I
L
if and
only if the associated Toeplitz matrices W
i,j
= P
L
|
ji
[
L
and W
i,j
= P
L
|
ji
[
L
are identical, and then both | and |
are unitarily equivalent to |

W
on L
W
with
canonical cyclic subspace W c L
W
. A little more work must be done to analyze
the dependence on the choice of cyclic subspace c and thereby solve the rst
classication problem. Indeed, if we next solve the trigonometric moment problem
for W and nd a measure on T (with values equal to operators on c) for which
W
n
=
_
T
z
n
d(z), then we arrive at a representation for the original operator | as
the multiplication operator M
z
on the space L
2
(). Alternatively, one can use the
theory of the Hellinger integral (see [5]) to make sense of the space of boundary
values of elements of L
W
as a certain space of vector measures (called charts in
[5]), or one can view the space L
W
as the image of the reproducing kernel Hilbert
space L() appearing prominently in work of de Branges and Rovnyak in their
approach to the spectral theory for unitary operators (see, e.g., [6]), where
(z) =
_
T
+z
z
d(z) for z in the unit disk D,
under the transformation (f(z), g(z)) f(z) + z
1
g(z
1
). In any case, the rst
(harder) classication problem (classication of unitary representations up to uni-
tary equivalence without specication of a -cyclic subspace) is solved via use of
the equivalence relation of mutual absolute continuity on spectral measures. For
this classical case, we see that the solution of the second problem serves as a
stepping stone to the solution of the rst problem, and that the transition from
the second to the rst involves some non-trivial mathematics (e.g., solution of the
trigonometric moment problem and measure theory).
The present paper concerns representations of the Cuntz algebra O
d
(see,
e.g., [8] for the denition and background), or what amounts to the same thing, a
d-tuple of operators | = (|
1
, . . . , |
d
) on a Hilbert space / which is row-unitary,
i.e.,
_
_
|
1
.
.
.
|
d
_
_
_
|
1
. . . |
d
=
_
_
I
.
.
.
I
_
_
,
_
|
1
. . . |
d
_
|
1
.
.
.
|
d
_
_
= I.
Equivalently, | = (|
1
, . . . , |
d
) is a d-tuple of isometries on / with orthogonal
ranges and with span of the ranges equal to the whole space /. It is known that
O
d
is NGCR, and hence the rst classication problem for the case of O
d
is in-
tractable in a precise sense, although particular special cases have been worked
out (see [7, 9]). The main contribution of the present paper is that there is a sat-
isfactory solution of the second classication problem (classication up to unitary
equivalence of unitary representations with specication of -cyclic subspace) for
the case of O
d
via a natural multivariable analogue of the spaces L
W
sketched
above for the single-variable case.
In detail, the functional calculus for a row-unitary d-tuple | = (|
1
, . . . , |
d
),
involves the free semigroup T
d
on a set of d generators g
1
, . . . , g
d
; elements of
the semigroup are words w of the form w = g
i
n
. . . g
i
1
with i
1
, . . . , i
n
1, . . . , d.
If w = g
i
n
. . . g
i
1
, set |
w
= |
i
n
|
i
1
. The functional model for such a row-unitary
d-tuple will consist of formal power series of the form
f(z, ) =
v,wJ
d
f
v,w
z
v
w
(1.1)
where z = (z
1
, . . . , z
d
) and = (
1
, . . . ,
d
) is a pair of d non-commuting variables.
The formalism is such that z
i
z
j
,= z
j
z
i
and
i
j
,=
j
i
for i ,= j but z
i
j
=
j
z
i
for
all i, j = 1, . . . , d. In the expression (1.1), for w = g
i
n
g
i
1
we set z
w
= z
i
n
z
i
1
and similarly for . The space L
W
of non-commuting formal power series which
serves as the functional model for the row-unitary | = (|
1
, . . . , |
d
) with cyclic
subspace c will be determined by a weighting matrix
W
v,w;,
= P
L
|
w
|
v
|
[
L
with row-index (v, w) and column index (, ) in the Cartesian product T
d

T
d
. On the space L
W
is dened a d-tuple of generalized shift operators |
W
=
(|
W,1
, . . . , |
W,d
) (see formula (2.12) below) which is row-unitary and which have
the subspace W c as a -cyclic subspace. Matrices W (with rows and columns
indexed by T
d
T
d
) arising in this way from a row-unitary | can be characterized
by a non-commutative analogue of the Toeplitz property which involves both a
non-commutative Hankel-like and non-commutative Toeplitz-like property along
Functional Models 5
with a non-degeneracy condition; we call such matrices Cuntz weights. Such
Cuntz weights serve as a complete unitary invariant for the second classication
problem for the Cuntz algebra O
d
: given two row-unitary d-tuples | = (|
1
, . . . , |
d
)
on / and |
= (|
1
, . . . , |
d
) on /
with common -cyclic subspace c contained in

both / and /
, then there is a unitary operator U : / /
such that U|
j
= |
j
U
and U|
j
= |
j
U for j = 1, . . . , d and U[
L
= I
L
if and only if the associated Cuntz
weights W
v,w;,
= P
L
|
w
|
v
|
[
L
and W
v,w;,
= P
L
|
w
|
v
|
[
L
are identical, and then both | and |
are unitarily equivalent to the model row-

unitary d-tuple |
W
= (|
W,1
, . . . , |
W,d
) acting on the model space L
W
with canon-
ical -cyclic subspace W c L
W
.
The parallel with the commutative case can be made more striking by view-
ing L
W
as a non-commutative formal reproducing kernel Hilbert space, a natural
generalization of classical reproducing kernel Hilbert spaces to the setting where
the elements of the space are formal power series in a collection of non-commuting
indeterminates; we treat this aspect in the separate report [4].
A second contribution of this paper is the application of this functional model
for row-unitary d-tuples to the free atomic representations and permutative rep-
resentations of O
d
appearing in [9] and [7] respectively. These representations are
of two types: the orbit-eventually-periodic type, indexed by a triple (x, y, ) where
x and y are words in T
d
and is a complex number of modulus 1, and the orbit-
non-periodic case, indexed by an innite word x = g
k
1
g
k
2
g
k
n
. Davidson and
Pitts [9] have identied which pairs of parameters (x, y, ) or x give rise to unitarily
equivalent representations of O
d
, which parameters correspond to irreducible rep-
resentations, and how a given representation can be decomposed as a direct sum or
direct integral of irreducible representations. The contribution here is to recover
these results (apart from the identication of irreducible representations) as an
application of the model theory of L
W
spaces and the calculus of Cuntz weights.
The approach shares the advantages and disadvantages of the de Branges-Rovnyak
model theory for single operators (see [6]). Once Cuntz weights W are calculated,
identifying unitary equivalences is relatively straightforward and obtaining decom-
positions is automatic up to the possible presence of overlapping spaces. There is
some hard work involved to verify that the overlapping space is actually trivial in
specic cases of interest. While these results are obtained in an elementary way
in [9], our results here show that a model theory calculus, a non-commutative
multivariable extension of the single-variable de Branges-Rovnyak model theory,
actually does work, and in fact is straightforward modulo overlapping spaces.
The paper is organized as follows. After the present Introduction, Section
2 lays out the functional models for row-isometries and row-unitary operator-
tuples in particular. We show there that the appropriate analogue for a bi-innite
Toeplitz matrix is what we call a Haplitz operator. Just as Toeplitz operators
W = [W
ij
]
i,j=...,1,0,1,...
have symbols

W(z) =
n=
W
n
z
n
, it is shown that
associated with any Haplitz operator W is its symbol
W(z, ), a formal power series

in two sets of non-commuting variables (z
1
, . . . , z
d
) and
1
, . . . ,
d
). These symbols
serve as the set of free parameters for the class of Haplitz operators; many questions
concerning a Haplitz operator W can be reduced to easier questions concerning
its symbol

W(z, ). In particular, positivity of the Haplitz operator W is shown to
be equivalent to a factorization property for its symbol

W(z, ) and for the Cuntz
defect D
W
(z, ) of its symbol (see Theorem 2.8). Cuntz weights are characterized
as those positive semidenite Haplitz operators with zero Cuntz defect.
Section 3 introduces the analogue of L
and H
, namely, the space of in-

tertwining maps L
W,W
T
between two row-unitary model spaces L
W
and L
W
, and
the subclass of such maps (analytic intertwining operators) which preserve the
subspaces analogous to Hardy subspaces. The contractive, analytic intertwining
operators then form an interesting non-commutative analogue of the Schur class
which has been receiving much attention of late from a number of points of view
(see, e.g., [2]). These results can be used to determine when two functional models
are unitarily equivalent, or when a given functional model decomposes as a direct
sum or direct integral of internal pieces (modulo overlapping spaces). Section 4
gives the application of the model theory and calculus of Cuntz weights to free
atomic and permutative representations of O
d
discussed by Davidson and Pitts [9]
and Bratteli and Jorgensen [7] mentioned above.
In a separate report [3] we use the machinery developed in this paper (es-
pecially the material in Section 3) to study non-commutative analogues of Lax-
Phillips scattering and unitary colligations, how they relate to each other, and
how they relate to the model theory for row-contractions developed in the work
of Popescu ([12, 13, 14, 15]).
2. Models for row-isometries and row-unitaries
Let T be the free semigroup on d generators g
1
, . . . , g
d
with identity. A generic
element of T
d
(apart from the unit element) has the form of a word w = g
i
n
g
i
1
,
i.e., a string of symbols
n

1
of nite length n with each symbol
k
belonging
to the alphabet g
1
, . . . , g
d
. We shall write [w[ for the length n of the word
w =
n

1
. If w =
n

1
and v =
m

1
are words, then the product vw
of v and w is the new word formed by the juxtaposition of v and w:
vw =
m

1
n

1
.
We dene the transpose w
of the word w = g
i
n
g
i
1
by w
= g
i
1
g
i
n
. We
denote the unit element of T
d
by (corresponding to the empty word). In partic-
ular, if g
k
is a word of unit length, we write g
k
w for g
k
n

1
if w =
n

1
.
Although T
d
is a semigroup, we will on occasion work with expressions involving
inverses of words in T
d
; the meaning is as follows: if w and v are words in T
d
, the
expression wv
1
means w
if there is a w
T
d
for which w = w
v; otherwise we
say that wv
1
is undened. An analogous interpretation applies for expressions
of the form w
1
v. This convention requires some care as associativity can fail: in
general it is not the case that (wv
1
) w
= w (v
1
w
).
Functional Models 7
For c an auxiliary Hilbert space, we denote by (T
d
, c) the set of all c-valued
functions v f(v) on T
d
. We will write
2
(T
d
, c) for the Hilbert space consisting
of all elements f in (T
d
, c) for which
|f|
2
2
(J
d
,L)
:=
vJ
d
|f(v)|
2
L
< .
Note that the space
2
(T
d
, c) amounts to a coordinate-dependent view of the
Fock space studied in [1, 9, 10, 12, 13]. It will be convenient to introduce the
non-commutative Z-transform f

f(z) on (T
d
, c) given by
f(z) =
vJ
d
f(w)z
w
where z = (z
1
, . . . , z
d
) is to be thought of as a d-tuple of non-commuting variables,
and we write
z
w
= z
i
n
z
i
1
if w = g
i
n
g
i
1
.
We denote the set of all such formal power series

f(z) also as L(T
d
, c) (or L
2
(T
d
, c)
for the Hilbert space case). The right creation operators S
R
1
, . . . , S
R
d
on
2
(T
d
, c)
are given by
S
R
j
: f f
where f
(w) = f(wg
1
j
)
with adjoint given by
S
R
j
: f f
where f
(w) = f(wg
j
).
(Here f(wg
1
j
) is interpreted to be equal to 0 if wg
1
j
is undened.) In the non-
commutative frequency domain, these right creation operators (still denoted by
S
R
1
, . . . , S
R
d
for convenience) become right multiplication operators:
S
R
j
:

f(z)

f(z) z
j
, S
R
j
:

f(z)

f(z) z
1
j
.
In the latter expression z
w
z
1
j
is taken to be 0 in case the word w is not of the
form w
g
j
for some w
T
d
. The calculus for these formal multiplication operators
is often easier to handle; hence in the sequel we will work primarily in the non-
commutative frequency-domain setting L(T
d
, c) rather than in the time-domain
setting (T
d
, c).
Let / be a Hilbert space and | = (|
1
, . . . , |
d
) a d-tuple of operators on /.
We say that | is a row-isometry if the block-operator row-matrix
_
|
1
|
d
:
d
k=1
/ /
is an isometry. Equivalently, each of |
1
, . . . , |
d
is an isometry on / and the image
spaces im|
1
, . . . , im|
d
are pairwise orthogonal. There are two extreme cases of
row-isometries |: (1) the case where | is row-unitary, i.e.,
_
|
1
. . . |
d
is uni-
tary, or equivalently, im|
1
, . . . , im|
d
span the whole space /, and (2) the case
where | is a row-shift, i.e.,
n0
spanim|
v
: [v[ = n = 0;
here we use the non-commutative multivariable operator notation
|
v
= |
i
n
. . . |
i
1
if v = g
i
n
g
i
1
.
A general row-isometry is simply the direct sum of these extreme cases by the
Wold decomposition for row-isometries due to Popescu (see [14]). It is well known
that the operators S
R
1
, . . . , S
R
d
provide a model for any row-shift, as summarized
in the following.
Proposition 2.1. The d-tuple of operators (S
R
1
, . . . , S
R
d
) on the space L
2
(T
d
, c) is
a row-shift. Moreover, if | = (|
1
, . . . , |
d
) is any row-shift on a space /, then |
is unitarily equivalent to (S
R
1
, . . . , S
R
d
) on L
2
(T
d
, c), with c = /
_
d
k=1
|
k
/
.
To obtain a similar concrete model for row-unitaries, we proceed as follows.
Denote by (T
d
T
d
, c) the space of all c-valued functions on T
d
T
d
:
f : (v, w) f(v, w).
We denote by
2
(T
d
T
d
, c) the space of all elements f (T
d
T
d
, c) for which
|f|
2
2
(J
d
J
d
,L)
:=
v,wJ
d
|f(v, w)|
2
< .
The Z-transform f

f for elements of this type is given by
f(z, ) =
v,w
f(v, w)z
v
w
.
Here z = (z
1
, . . . , z
d
) is a d-tuple of non-commuting variables as before, and =
(
1
, . . . ,
d
) is another d-tuple of non-commuting variables, but we specify that
each
i
commutes with each z
j
for i, j = 1, . . . , d. For the case d = 1, note that
2
(T
1
, c) is the standard
2
-space over the non-negative integers
2
(Z
+
, c), while
2
(T
1
T
1
, c) =
2
(Z
+
Z
+
, c)
appears to be a more complicated version of
2
(Z, c). Nevertheless, we shall see
that the weighted modications of
2
(T
d
T
d
, c) which we shall introduce below do
collapse to
2
(Z, c) for the case d = 1. Similarly, one should think of L
2
(T
d
, c) as a
non-commutative version of the Hardy space H
2
(D, c) over the unit disk D, and of
the modications of L
2
(T
d
T
d
, c) to be introduced below as a non-commutative
analogue of the Lebesgue space L
2
(T, c) of measurable norm-square-integrable
c-valued functions on the unit circle T.
In the following we shall focus on the frequency domain setting L
2
(T
d

T
d
, c) rather than the time-domain setting
2
(T
d
, T
d
, c), where it is convenient
to use non-commutative multiplication of formal power series; for this reason we
shall write simply f(z, ) for elements of the space rather than

f(z, ). Unlike the
unilateral setting L
2
(T
d
, c) discussed above, there are two types of shift operators
on L
2
(T
d
T
d
, c) of interest, namely:
S
R
j
: f(z, ) f(z, ) z
j
, (2.1)
U
R
j
: f(z, ) f(0, )
1
j
+f(z, ) z
j
(2.2)
Functional Models 9
where f(0, ) is the formal power series in = (
1
, . . . ,
d
) obtained by formally
setting z = 0 in the formal power series for f(z, ):
f(0, ) =
wJ
d
f
,w
w
if f(z, ) =
v,wJ
d
f
v,w
z
v
w
.
One can think of S
R
j
as a non-commutative version of a unilateral shift (even
in this bilateral setting), while U
R
j
is some kind of bilateral shift. We denote by
S
R[]
j
and U
R[]
j
the adjoints of S
R
j
and U
R
j
in the L
2
(T
d
T
d
, c)-inner product
(to avoid confusion with the adjoint with respect to a weighted inner product to
appear below). An easy computation shows that
S
R[]
j
: f(z, ) f(z, ) z
1
j
, (2.3)
U
R[]
j
: f(z, ) f(0, )
j
+f(z, ) z
1
j
. (2.4)
Note that
U
R[]
i
S
R
j
: f(z, ) U
R[]
i
(f(z, ) z
j
) =
i,j
f(z, )
and hence we have the useful identity
U
R[]
i
S
R
j
=
i,j
I. (2.5)
On the other hand
S
R
j
U
R[]
j
: f(z, ) S
R
j
(f(0, )
j
+f(z, )z
1
j
)
= f(0, )
j
z
j
+ [f(z, )z
1
j
]z
j
and hence
_
_
I
d
j=1
S
R
j
U
R[]
j
_
_
: f(z, ) f(z, )
d
j=1
f(0, )
j
z
j

d
j=1
[f(z, )z
1
j
]z
j
= f(0, )
d
j=1
f(0, )
j
z
j
and hence
_
_
I
d
j=1
S
R
j
U
R[]
j
_
_
: f(z, ) f(0, )
_
_
1
d
j=1
z
j
j
_
_
. (2.6)
Now suppose that | = (|
1
, . . . , |
d
) is a row-unitary d-tuple of operators on a
Hilbert space /, c is a subspace of /, and we dene a map : / L(T
d
T
d
, c) by
k =
v,wJ
d
(P
L
|
w
|
v
k)z
v
w
. (2.7)
Then
|
j
k =
v,wJ
d
(P
L
|
w
|
v
|
j
k)z
v
w
=
wJ
d
(P
L
|
wg
j
k)
w
+
v,w: v,=
(P
L
|
w
|
vg
1
j
k)z
v
w
= (U
R
j
k)(z, ) (2.8)
while
|
j
k =
v,wJ
d
(P
L
|
v
|
w
|
j
k)z
v
w
=
v,wJ
d
(P
L
|
v
|
wg
j
k)z
v
w
= (S
R[]
j
k)(z, ). (2.9)
If we let W =
[]
(where
[]
is the adjoint of with respect to the Hilbert
space inner product on / and the formal L
2
-inner product on L(T
d
T
d
, c)), then
[]
: z
e |
e
and W :=
[]
= [W
v,w;,
]
v,w,,J
d
where
W
v,w;,
= P
L
|
w
|
v
|
[
L
. (2.10)
If im is given the lifted norm | |
,
|k|
= |P
(ker )
k|
K
,
then one easily checks that W T(T
d
T
d
, c) im and
|Wp|
2
= |
[]
p|
2
K
=
[]
p,
[]
p)
K
= Wp, p)
L
2.
Thus, if we dene a space L
W
as the closure of W T(T
d
T
d
, c) in the norm
|Wp|
2
1
W
= Wp, p)
L
2, (2.11)
then L
W
= im isometrically. From the explicit form (2.10) of W
v,w;,
it is easy
to verify the intertwining relations
U
R
j
W = WS
R
j
, S
R[]
j
W = WU
R[]
j
on T(T
d
T
d
, c).
If we dene |
W
= (|
W,1
, . . . , |
W,d
) on L
W
by
|
W,j
: Wp U
R
j
Wp = WS
j
p, (2.12)
then, from the intertwining relations
|
j
= U
R
j
, |
j
= S
R[]
j
for j = 1, . . . , d
Functional Models 11
and the fact that is coisometric, we deduce that |
W
is row-unitary on L
W
with
adjoint |
J
= (|
W,1
, . . . , |
W,d
) on L
W
given by
|
W,j
: Wp S
R[]
j
Wp = WU
R[]
j
p. (2.13)
If is injective (or, equivalently, if c is -cyclic for the row-unitary d-tuple
| = (|
1
, . . . , |
d
)), then |
W
= (|
W,1
, . . . , |
W,d
) on L
W
is a functional model
row-unitary d-tuple for the abstractly given row-unitary d-tuple | = (|
1
, . . . |
d
).
Our next goal is to understand more intrinsically which weights
W = [W
v,w;,
]
can be realized in this way as (2.10) and thereby lead to functional models for
row-unitary d-tuples |. From identity (2.5) we see that (S
R
1
, . . . , S
R
d
) becomes a
row-isometry if we can change the inner product on L
2
(T
d
T
d
, c) so that the
adjoint S
R
j
of S
j
in the new inner product is U
R[]
j
. Moreover, if we in addition
arrange for the new inner product to have enough degeneracy to guarantee that
all elements of the form f(0, )(1
d
j=1
z
j
j
) have zero norm, then the d-tuple
(S
R
1
, . . . , S
R
d
) in this new inner product becomes row-unitary. These observations
suggest what additional properties we seek for a weight W so that it may be of
the form (2.10) for a row-unitary |.
Let W be a function from four copies of T
d
into bounded linear operators on
c with value at (v, w, , ) denoted by W
v,w;,
. We think of W as a matrix with
rows and columns indexed by T
d
T
d
; thus W
v,w;,
is the matrix entry for row
(v, w) and column (, ). Denote by T(T
d
, T
d
, c) the space of all polynomials in
the non-commuting variables z
1
, . . . , z
d
,
1
, . . . ,
d
:
T(T
d
, T
d
, c) =p(z, ) =
v,wJ
d
p
v,w
z
v
w
: p
v,w
c and
p
v,w
= 0 for all but nitely many v, w.
Then W can be used to dene an operator from T(T
d
T
d
, c) into L(T
d
T
d
, c)
by extending the formula
W : e z
v,wJ
d
W
v,w;,
e z
v
w
.
for monomials to all of T(T
d
, T
d
, c) by linearity. Note that computation of the
L
2
-inner product Wp, q)
L
2 involves only nite sums if p and q are polynomials,
and therefore is well dened. We say that W is positive semidenite if
Wp, p)
L
2 0 for all p T(T
d
T
d
, c).
Under the assumption that W is positive semidenite, dene an inner product on
W T(T
d
T
d
, c) by
Wp, Wq)
1
W
= Wp, q)
L
2 . (2.14)
Modding out by elements of zero norm if necessary, W T(T
d
T
d
, c) is a pre-
Hilbert space in this inner product. We dene a space
L
W
= the completion of W T(T
d
T
d
, c) in the inner product (2.14). (2.15)
Note that the (v, w)-coecient of Wp WT(T
d
T
d
, c) is given by
[Wp]
v,w
, e)
L
= Wp, W(z
v
w
e))
1
W
(2.16)
and hence the map
v,w
: f f
v,w
extends continuously to the completion L
W
of W T(T
d
T
d
, c) and L
W
can be identied as a space of formal power series
in the non-commuting variables z
1
, . . . , z
d
and
1
, . . . ,
d
, i.e., as a subspace of
L(T
d
T
d
, c). (This is the main advantage of dening the space as the completion
of W T(T
d
T
d
, c) rather than simply as the completion of T(T
d
T
d
, c) in the
inner product given by the right-hand side of (2.14).) Note that then, for f L
W
and , T
D
we have
,
f, e)
L
= f
,
, e)
L
= f, W[z
e])
1
W
from which we see that
,
: c L
W
is given by
,
: e W[z
e].
By using this fact we see that
v,w
,
e, e
)
L
=
,
e,
v,w
e
)
1
W
= W[z
e], W[z
v
w
e
])
1
W
= W[z
e], z
v
w
e
)
1
W
= W[z
e], z
v
w
e
)
L
2
= W
v,w;,
e, e
)
L
and we recover the operator matrix entries W
v,w;,
of the operator W from the
family of operators
,
(, T
d
) via the factorization
W
v,w;,
=
v,w
,
.
Conversely, one can start with any such factorization of W (through a general
Hilbert space / rather than / = L
W
as in the construction above). The following
theorem summarizes the situation.
Theorem 2.2. Assume that W = [W
v,w;,
]
v,w,,J
d
is a positive semidenite
(T
d
T
d
)(T
d
T
d
) matrix of operators on the Hilbert space c with a factorization
of the form
W
v,w;,
=
v,w
,
for operators
v,w
: / c for some intermediate Hilbert space / for all v, w, ,
T
d
. Dene an operator : / L(T
d
T
d
, c) by
: k
v,wJ
d
(
v,w
k)z
v
w
.
Let L
W
be the Hilbert space dened as the completion of WT(T
d
T
d
, c) in the
lifted inner product
Wp, Wp)
1
W
= Wp, p)
L
2.
Then is a coisometry from / onto L
W
with adjoint given densely by
: Wp
v,wJ
d
,
p
,
for p(z, ) =
,
p
,
z
a polynomial in T(T
d
T
d
, c). In particular, L
W
is
given more explicitly as L
W
= im.
Proof. Note than W (as a densely dened operator on L
2
(T
d
T
d
, c) with domain
containing at least T(T
d
T
d
, c)) factors as W =
[]
, where
[]
is the formal
L
2
-adjoint of dened at least on polynomials by
[]
: p(z, w)
,
p
,
for p(z, w) =
,
p
,
z
.
In particular,
[]
T(T
d
T
d
, c) is contained in the domain of when is con-
sidered as an operator from / into L
W
with dom = k /: k L
W
. Since
is dened in terms of matrix entries
v,w
and evaluation of Fourier coecients
p p
v,w
is continuous in L
W
, it follows that as an operator from / into L
W
with domain as above is closed. For an element k / of the form
[]
p for a
polynomial p T(T
d
T
d
, c), we have
Wp, Wp)
1
W
= Wp, p)
L
2
=
[]
k, k)
L
2
=
[]
k,
[]
k)
K
.
Hence maps
[]
T(T
d
T
d
, c) isometrically onto the dense submanifold W
T(T
d
T
d
, c) of L
W
. From the string of identities
k,
[]
p)
K
= k, p)
L
2
= k, Wp)
1
W
and the density of W T(T
d
T
d
, c) in L
W
, we see that ker = (
[]
T(T
d

T
d
, c))
. Hence is isometric from a dense subset of the orthogonal complement of

its kernel (
[]
T(T
d
T
d
, c)) onto a dense subset of L
W
(namely, WT(T
d
T
d
, c)).
Since is closed, it follows that necessarily : / L
W
is a coisometry. Finally,
notice that
k, Wp)
1
W
= k, p)
L
2
= k,
[]
p)
K
: Wp
[]
p for p T(T
d
T
d
, c)
and the formula for
follows. This completes the proof of Theorem 2.2.

We seek to identify additional properties to be satised by W so that the
operators |
W,j
dened initially only on W T(T
d
T
d
, c) by
|
W,j
: Wp WS
R
j
p
and then extended to all of L
W
by continuity become a row-isometry, or even a
row-unitary operator-tuple. From (2.5), the row-isometry property follows if it can
be shown that |
W,j
: Wp WU
R[]
j
p, or equivalently,
WS
R
j
= U
R
j
W on T(T
d
T
d
, c). (2.17)
Similarly, from (2.6) we see that the row-unitary property will follow if we show
in addition that
W
_
p(0, )
_
1
d
k=1
z
k
k
__
= 0 for all p T(T
d
T
d
, c). (2.18)
The next theorem characterizes those operator matrices W for which (2.17) and
(2.18) hold.
Theorem 2.3. Let W be a (T
d
T
d
) (T
d
T
d
) matrix with matrix entries equal
to operators on the Hilbert space c. Then:
1. W satises (2.17) if and only if
W
,w;g
j
,
= W
,wg
j
;,
, (2.19)
W
v,w;g
j
,
= W
vg
1
j
,w;,
for v ,= (2.20)
for all v, w, , T
d
, j = 1, . . . , d, where we interpret W
vg
1
j
,w;,
to be 0
in case vg
1
j
is not dened.
2. Assume that W is selfadjoint. Then W satises (2.17) and (2.18) if and only
if W satises (2.19), (2.20) and in addition
W
,w;,
=
d
j=1
W
,wg
j
;,g
j
(2.21)
for all w, T
d
.
Proof. By linearity, it suces to analyze the conditions (2.17) and (2.18) on mono-
mials f(z, ) = z
e for some , T
d
and e c. We compute
WS
R
j
(z
e) = W(z
g
j
e)
=
v,wJ
d
W
v,w;g
j
,
z
v
w
(2.22)
while, on the other hand,
U
R
j
W(z
e) = U
R
j
_
_

v,wJ
d
W
v,w;,
ez
v
w
_
_
=
wJ
d
W
,w;,
wg
1
j
+
v,wJ
d
W
v,w;,
z
vg
j
w
. (2.23)
Equality of (2.22) with (2.23) for all , and e in turn is equivalent to (2.19) and
(2.20).
Assume now that W satises (2.19) and is selfadjoint. Then we have
W
v,w;,g
j
= [W
,g
j
;v,w
]
= [W
,;vg
j
,w
]
= W
vg
j
,w;,
and hence
W
vg
j
,w;,
= W
v,w;,g
j
. (2.24)
To verify the condition for (2.18), if f(z, ) = z
e, then f(0, ) = 0 if
,= and W
_
f(0, )
_
1
d
j=1
z
j
j
__
= 0 trivially. Thus it suces to consider
the case = and f(z, ) =
e. Then f(0, ) =
e and
f(0, )
_
_
1
d
j=1
z
j
j
_
_
=
e
d
j=1
z
g
j
g
j
e.
Then
W
_
_
f(0, )
_
_
1
d
j=1
z
j
j
_
_
_
_
= W
_
_
e
d
j=1
z
j
g
j
e
_
_
=
v,wJ
d
_
_
W
v,w;,

d
j=1
W
v,w;g
j
,g
j
_
_
e z
v
w
This last quantity set equal to zero for all T
d
and e c is equivalent to
W
v,w;,
=
d
j=1
W
v,w;g
j
,g
j
. (2.25)
If v = , we may use (2.19) to see that
W
,w;g
j
,g
j
= W
,wg
j
;,g
j
and (2.25) collapses to (2.21). If v ,= , write v = v
g
k
for some k. From (2.20) we
have
W
v,w;g
j
,g
j
=
j,k
W
v
,w;,g
k
and hence
d
j=1
W
v,w;g
j
,g
j
= W
v
,w;,g
k
= W
v
g
k
,w;,
(by (2.24))
= W
v,w;,
and hence (2.25) for this case is already a consequence of (2.20) and (2.24). This
completes the proof of Theorem 2.3.
Note that (2.19) is a Hankel-like property for the operator matrix W for this
non-commutative setting, while (2.20) is a Toeplitz-like property. We shall there-
fore refer to operators W : T(T
d
T
d
, c) L(T
d
T
d
, c) for which both (2.19)
and (2.20) are valid as Haplitz operators. We shall call positive semidenite Haplitz
operators having the additional property (2.21) Cuntz weights. As an immediate
corollary of the previous result we note the following.
Corollary 2.4. Let W be a positive semidenite operator
W : T(T
d
T
d
, c) L(T
d
T
d
, c)
and consider the d-tuple |
W
= (|
W,1
, . . . , |
W,d
) of operators on L
W
dened
densely by
|
W,j
: Wp WS
R
j
p for p T(T
d
T
d
, c). (2.26)
Then:
1. W is Haplitz if and only if |
W
is a row-isometry.
2. W is a Cuntz weight if and only if |
W
is row-unitary.
In either case, |
W,j
and |
W,j
are then given on arbitrary elements f L
W
by
|
W,j
: f(z, ) f(0, )
1
j
+f(z, ) z
j
, (2.27)
|
W,j
: f(z, ) f(0, )
j
+f(z, ) z
1
j
. (2.28)
Proof. The results are immediate consequences of (2.5) and (2.6). To prove (2.27)
and (2.28), note the two expressions for |
W,j
and for |
W,j
on polynomials in case
W = W
is a Haplitz operator (as in (2.12) and (2.13)), and then note that the
rst formula necessarily extends by continuity to all of L
W
since the map f f
v,w
is continuous on any space L
W
.
Remark 2.5. If W = [W
v,w;,
]
v,w,,J
d
where W
v,w;,
= P
L
|
w
|
v
|
[
L
for a row-unitary d-tuple | as in (2.10), then it is easily checked that W is a Cuntz
weight. To see this, note
W
,w;g
j
,
= P
L
|
w
|
j
|
L
= P
L
|
wg
j
|
L
= W
,wg
j
;,
L
simply from associativity of operator composition, and (2.19) follows. Similarly,
for v ,= , we have
W
v,w;g
j
,
= P
L
|
w
|
v
|
j
|
L
= P
L
|
w
|
vg
1
j
|
L
= W
vg
1
j
,w;,
from the row-isometry property of |:
|
i
|
j
=
i,j
I.
Hence (2.20) follows and W is Haplitz. To check (2.21), we use the row-coisometry
property of | (
d
j=1
|
j
|
j
= I) to see that
d
j=1
W
,wg
j
;,g
j
=
d
j=1
P
L
|
w
|
j
|
j
|
L
= P
L
|
w
|
L
= W
,w;,
.
From the formula (2.10) we see that W has the additional normalization property
W
,;,
= I
L
.
We shall be particularly interested in []-Haplitz operators, i.e., (T
d
T
d
)
(T
d
T
d
) operator matrices W for which both W and W
[]
are Haplitz. (Of course
a particular class of []-Haplitz operators are the selfadjoint Haplitz operators
Haplitz operators W with W = W
[]
.) For these operators the structure can be
arranged in the following succinct way.
Proposition 2.6. Suppose that W = [W
v,w;,
] is a []-Haplitz operator matrix, and
dene a T
d
T
d
operator matrix

W = [
W
v,w
] by
W
v,w
= W
v,w;,
.
Then W is completely determined from

W according to the formula
W
v,w;,
=
_
W
(v
1
)
,w
if [v[ [[,
,w(v
1
)
if [v[ [[.
(2.29)
Conversely, if

W is any T
d
T
d
matrix, then formula (2.29) denes a []-
Haplitz operator matrix W.
Proof. Suppose rst that W = [W
v,w;,
] is []-Haplitz. Then we compute, for
[v[ [[,
W
v,w;,
= W
v
1
,w;,
by (2.20)
=
_
W
[]
,;v
1
,w
_
=
_
W
[]
,;(v
1
)
,w
_
by (2.19) for W
[]
= W
(v
1
)
,w;,
=

W
(v
1
)
,w
while, for [v[ [[ we have
W
v,w;,
=
_
W
[]
,;v,w
_
=
_
W
[]
v
1
,;,w
_
by (2.20) for W
[]
= W
,w;v
1
,
= W
,w(v
1
)
;,
by (2.19)
=
_
W
[]
,;,w(v
1
)
=
_
W
[]
,;
,w(v
1
)
by (2.19)
= W
,w(v
1
)
;,
=

W
,w(v
1
)
and the rst assertion follows.

Conversely, given

W = [
W
v,
], dene W = [W
v,w;,
] by (2.29). Then verify
W
,w;g
j
,
=

W
,w
versus W
,wg
j
;,
=

W
,w
and (2.19) follows for W.
Similarly, compute, for v ,= ,
W
v,w;g
j
,
=
_
W
(vg
1
j

1
)
,w
if [v[ > [[,
,w(g
j
v
1
)
if [v[ [[
versus, again for v ,= ,
W
vg
1
j
,w;,
=
_
0 if v ,= v
g
j
,
W
v
,w;,
if v = v
g
j
where
W
v
,w;,
=
_
W
v
,w
if [v
[ [[,
,w(v
1
)
if [v
[ [[
and (2.20) follows for W.
From (2.29) we see that W
[]
is given by
W
[]
v,w;,
= (W
,;v,w
)
=
_
_
_
_
W
(v
1
)w
,
_
if [[ [v[,
_
W
w
,(v
1
)
if [[ [v[.
(2.30)
Using (2.30) we see that
W
[]
,w;g
j
,
=
_
W
g
j
w
,
_
while
W
[]
,wg
j
;,
=
_
W
g
j
w
,
_
and (2.19) follows for W

[]
. Similarly, for v ,= ,
W
[]
v,w;g
j
,
=
_
_
_
_
W
g
j
v
1
)w
,
_
if [v[ [[,
_
W
w
,(vg
1
j

1
)
if [v[ > [[
versus
W
[]
vg
1
j
,w;,
=
_
0 if v ,= v
g
j
,
W
[]
v
,w;,
if v = v
g
j
where
W
[]
v
,w;,
=
_
_
_
_
W
(v
1
)w
,
_
if [v
[ < [[,
_
W
w
,(v
1
)
if [v
[ [[
and (2.20) follows for W
[]
as well. We conclude that W as dened by (2.29) is
[]-Haplitz as asserted.
The formula (2.29) motivates the introduction of the symbol

W(z, ) for the
[]-Haplitz operator W dened by
W(z, )e = (We)(z, ) =
v,wJ
d
W
v,w;,
ez
v
w
.
For any []-Haplitz W and given any , T
d
, we have
v,w
W
v,w;,
ez
v
w
= W(ez
)
= W((S
R
)
(U
R[]
)
e)
= (U
R
)
W((U
R[]
)
e)
= (U
R
)
(S
R[]
)
(We).
Hence the matrix entries W
v,w;,
for W are determined from a knowledge of
matrix entries of the special form W
v,w;,
(i.e., the Fourier coecients of the
symbol

W(z, )) via
v,wJ
d
W
v,w;,
ez
v
w
= (U
R
)
(S
R[]
)
_
_

v,wJ
d
W
v,w;,
ez
v
w
_
_
. (2.31)
This gives a method to reconstruct a []-Haplitz operator directly from its symbol
(equivalent to the reconstruction formula (2.29) from the matrix entries W
v,w;,
).
This can be made explicit as follows.
Proposition 2.7. Let W = [W
v,w;,
] be a []-Haplitz operator as above with sym-
bol

W(z, ) =
v,wJ
d
W
v,w;,
z
v
w
. Then, for a general polynomial f(z, ) =
,
f
,
z
T(T
d
T
d
, c), we have
W[f](z, ) =

W(z
, )k
per
(z, )f(z, z
1
)[
z
=0
+
W(z
, )f(z, z
1
)[
z
=z
, (2.32)
where we have set k
per
(z, ) equal to the perverse Szego kernel
k
per
(z, ) =
,=
(z
1
)
v
(
1
)
v
(2.33)
and where z
= (z
1
, . . . , z
d
) is another set of non-commuting indeterminants, each
of which commutes with z
1
, . . . , z
d
,
1
, . . . ,
d
, and where it is understood that the
evaluation at z
= 0 is to be taken before the multiplication with k

per
(z, ).
In particular, for the case where p(z) =
is an analytic polynomial
in T(T
d
, c), then
W[p](z) =

W(0, )k
per
(z, )p(z) +
W(z, )p(z). (2.34)

If in addition

W(0, ) = I, then (2.34) simplies further to
W[p](z) =

W(z, )p(z). (2.35)
Proof. We compute
W[z
e] = W(S
R
)
(U
R[]
)
e
= (U
R
)
(S
R[]
)
W(z, )e
_
= (U
R
)
W(z, )(z
1
)
_
e.
Note that [
W(z, )(z
1
)
](0, ) =
w
W
,w;,
w
. Therefore
U
R
j
:

W(z, )(z
1
)
w
W
,w;,
1
j
e +
v,w
W
v
,w;,
z
v
w
z
j
e,
U
R
k
U
R
j
:

W(z, )(z
1
)
U
R
k
_
w
W
,w;,
1
j
e +
v,w
W
v
,w;,
z
v
w
z
j
e
_
=
v
W
,w;,
1
j

1
k
e +
w
W
,w;,
1
j
z
k
e
+
v,w
W
v
,w;,
z
v
w
z
j
z
k
e
and then by induction we have
W : z
: =
,=
w
W
,w;,
w
(
1
)
e +
v,w
W
v
,w;,
z
v
w
z
e
=
_

wJ
d
W
,w;,
w
_
k
per
(z, ) z
e +
_
W(z, )(z
1
)
_
z
e
=

W(z
, )k
per
(z, )z
(z
1
)
e[
z
=0
+
W(z
, )z
(z
1
)
e[
z
=z
=

W(z
, )k
per
(z, )f(z, z
1
)[
z
=0
+
W(z
, )f(z, z
1
)[
z
=z
for the case f(z, ) = z
e, and formula (2.32) now follows by linearity. The

formulas (2.34) and (2.35) are specializations of (2.32) to the case where f = p
T(T , c) and where also

W(0, ) = I respectively.
Positive semideniteness of a selfadjoint Haplitz operator W as well as the
validity of (2.21) required to be a Cuntz weight can be characterized directly
in terms of the symbol

W(z, ) as follows. For a general formal power series
A(z) =
wJ
d
A
w
z
w
with operator coecients, we set A(z)
wJ
d
A
w
z
w
wJ
d
A
z
w
. For a formal power series K(z, ) in the two sets of non-commuting
variables z = (z
1
, . . . , z
d
) and = (
1
, . . . ,
d
), we already have the notion of K
is a positive kernel; by this we mean that the NFRKHS H(K) is well dened,
or equivalently, that K has a factorization of the form K(z, ) = Y (z)Y ()
for
some formal non-commuting power series Y (z) =
vJ
d
Y
v
z
v
. See [4]. The next
result suggests that we introduce the following related notion: we shall say that
the formal power series K(z, ) is a positive symbol if K
(z, ) = K(, z) is a pos-

itive kernel, i.e., if K has a factorization of the form K(z, ) = Y ()Y (z)
. This
terminology will be used in the sequel. In addition, for any symbol

W(z, ), we
dene the Cuntz defect D
W
(z, ) by
D
W
(z, ) =

W(z, )
d
k=1
z
1
k
W(z, )
1
k
.
Theorem 2.8. A Haplitz operator W : T
d
T
d
L(c) is positive semidenite if
and only if both its symbol

W(z, ) and the Cuntz defect of its symbol D
W
(z, )
are positive symbols, i.e., there exist formal power series Y (z) =
wJ
d
Y
w
z
w
and
(z) =
wJ
d
w
z
w
so that
W(z, ) = Y ()Y (z)
, (2.36)
W(z, )
d
j=1
z
1
j
W(z, )
1
j
= ()(z)
. (2.37)
The Haplitz operator W is a Cuntz weight if and only if its symbol

W(z, ) is a
positive symbol and its Cuntz defect D
W
(z, ) is zero:
W(z, )
d
j=1
z
1
j
W(z, )
1
j
= 0. (2.38)
Proof. Suppose that W is a positive semidenite Haplitz operator. From the theory
of reproducing kernels, this means that W has a factorization
W
v,w;,
= X
v,w
X
,
(2.39)
for some operators X
v,w
: L c. As W is selfadjoint Haplitz, we have
W
v,w
= W
v,w;,
=
_
W
,;v,w
_
=
_
W
,v
;,w
_
= W
,w;,v
= X
,w
X
,v
or
W
v,w
= Y
w
Y
(2.40)
where we have set Y
w
= X
,w
. The identity (2.40) in turn is equivalent to (2.36)
with Y (z) =
wJ
d
Y
w
z
w
.
Derivation of the necessity of the factorization (2.37) lies deeper. From Corol-
lary 2.4 we know that the operators (|
W,1
, . . . , |
W,d
) given by (2.26) form a row-
isometric d-tuple on L
W
. Furthermore, factorization (2.39) implies that the map
:
v,w,J
d
(X
v,w
)z
v
w
is a coisometry. Without loss of generality, we may assume that is actually
unitary. Hence there is a row-isometric d-tuple (1
1
, . . . , 1
d
) of operators on L
dened by
1
j
= |
W,j
for j = 1, . . . , d.
Set
L = closed span Y
c : T
d
.
We claim: 1
j
: Y
e Y
g
j
e for e c and j = 1, . . . , d. Indeed, note that
Y
e =
v,w
(X
v,w
Y
e)z
v
w
=
v,w
(X
v,w
X
,
e)z
v
,
w
=
v,w
(W
v,w;,
e)z
v
w
= [W(
e)](z, )
|
W,j
Y
e = |
W,j
W(
e)
= W(U
R[]
j
(
e))
= W(
g
j
e)
=
v,w
(W
v,w;,g
j
e)z
v
w
= Y
g
j
e
and the claim follows.
Thus

L is invariant for 1
j
for 1 j d. As (1
1
, . . . , 1
d
) is a row-isometry,
it follows that (1
1
, . . . , 1
d
)[
1
is a column contraction, or
_
_
_
_
_
_
_
n
j=1
_
_
Y
j
g
1
.
.
.
Y
j
g
d
_
_
e
j
_
_
_
_
_
_
_
2
|
n
j=1
Y
j
e
j
|
2
for all
1
, . . . ,
n
T
d
and e
1
, . . . , e
n
c. Thus
[
W
v
j
,v
i
]
i,j=1,...,n
= [Y
v
i
Y
v
j
]
i,j=1,...,n
_
Y
v
1
.
.
.
Y
v
n
_
_
_
|
1
. . . |
d
_
|
1
.
.
.
|
d
_
_
_
Y
v
1
. . . Y
v
n
=
d
=1
_
_
Y
g
v
1
.
.
.
Y
g
v
n
_
_
_
Y
v
1
g
. . . Y
v
n
g
=
d
=1
[
W
g
j
,v
i
g
]
i,j=1,...,n
(2.41)
and hence we have the inequality
_
W
v
j
,v
i

d
=1
W
g
j
,v
i
g
_
i,j=1,...,d
0. (2.42)
This is exactly the matrix form of the inequality (2.37), and hence (2.37) follows.
Suppose now that W is a Cuntz weight. Then (|
W,1
, . . . , |
W,d
) is row-unitary.
It follows that (1
1
, . . . , 1
d
) is row-unitary, and hence that (1
1
, . . . , 1
d
)[
1
is a col-
umn isometry. It then follows that equality holds in (2.42), from which we get the
equality (2.38) as asserted.
Conversely, suppose that W is a selfadjoint Haplitz for which the two factor-
izations (2.36) and (2.37) hold. Then we have
W
v,w
= Y
w
Y
where Y
v
:

L c.
We may assume that

L = closed span Y
v
c : v T
d
. From (2.37), by reversing
the order of the steps in (2.41) we see that the operators (T
1
, . . . , T
d
) dened by
T
j
Y
v
e = Y
vg
j
e
extend to dene a column contraction on

L. By induction, T
u
Y
v
= Y
vu
e or
Y
v
T
u
= Y
vu
for all u T
d
and e c, and
_
T
1
. . . T
d
:
_
L
.
.
.
L
_

L
is a contraction. Let (1
1
, . . . , 1
d
) on L

L be the row-isometric dilation of
(T
1
, . . . , T
d
) (see [14]), so
_
1
1
. . . 1
d
:
_
_
L
.
.
.
L
_
_
L is isometric and P
1
1
u
= T
u
P
1
.
Set
X
v,w
= Y
w
P
1
1
v
: L c
for v, w T
d
. Then, for [v[ [[ we have
X
v,w
X
,
= Y
w
P
1
1
w
1
= Y
w
P
1
1
v
1
Y
= Y
w
T
v
1
Y
= Y
w
Y
(v
1
)
=

W
(v
1
)
,w
= W
v,w;,
where we used (2.29) for the last step. Similarly, if [v[ [[ we have
X
v,w
X
,
= Y
w
P
1
1
w
1
= Y
w
P
1
1
(v
1
)
= Y
w
T
(v
1
)
= Y
w(v
1
)
Y
=

W
,w(v
1
)
= W
v,w;,
and the factorization W
v,w,
= X
v,w
X
,
shows that W is positive semidenite
as wanted.
If the symbol

W of the Haplitz operator W satises (2.36) and (2.38), then
we see that the d-tuple (T
1
,...,T
d
) in the above construction is a row-isometry, in
which case the row-isometric extension (1
1
,...,1
d
) of the row-coisometry (T
1
,...,T
d
)
is in fact row-unitary. The construction above applies here to give W
v,w;,
=
Y
w
P
1
1
v
1
, but now with 1 equal to a row-unitary d-tuple. Hence

d
j=1
W
,wg
j
;,g
j
=
d
j=1
Y
wg
j
P
1
Y
g
j
=
d
j=1
Y
w
P
1
1
j
1
j
Y
= Y
w
Y
= W
,w;,
and (2.21) follows, i.e., W is a Cuntz weight in case (2.37) is strengthened to (2.38).
The theorem now follows.
Remark 2.9. If W is a []-Haplitz operator, then the adjoint W
[]
of W has symbol
W
[]
(z, ) =
v,w
W
v,w;,
z
v
w
where
W
v,w,
= (W
,;v,w
)
= (W
,v
;,w
)
= W
[]
,w;,v
= W
[]
,;w
,v
= (W
w
,v
;,
)
W
[]
(z, ) =

W(, z)
. (2.43)
(where we use the convention that (z
v
)
= z
v
and (
w
)
=
w
). Thus the
selfadjointness of a []-Haplitz operator W can be expressed directly in terms of
the symbol: W = W
[]
(as a []-Haplitz operator) if and only if
W(z, ) =

W(, z)
. (2.44)
Note that (2.44) is an immediate consequence of the factorization (2.36) required
for positive semideniteness of the Haplitz operator W.
In case W is a Cuntz weight with the additional property
W
,;,
= I
L
, (2.45)
then the coecient space c can be identied isometrically as a subspace of L
W
via
the isometric map V
W
: c L
W
given by V
W
: e We. Let us say that a Cuntz
weight W with the additional property (2.45) is a normalized Cuntz weight. Then
we have the following converse of Corollary 2.4 for the case of normalized Cuntz
weights. A more precise converse to Corollary 2.4, using the notion of row-unitary
scattering system, is given in [4].
Theorem 2.10. Let | = (|
1
, . . . , |
d
) be a row-unitary d-tuple of operators on the
Hilbert space / and let c be a subspace of /. Let W be the (T
d
T
d
) (T
d
T
d
)
matrix of operators on c dened by
W
v,w;,
= P
L
|
w
|
v
|
L
. (2.46)
Then W is a normalized Cuntz weight and the map dened by
: k
v,wJ
d
(P
L
|
w
|
v
k)z
v
w
(2.47)
is a coisometry from / onto L
W
which satises the intertwining property
|
j
= |
W,j
for j = 1, . . . , d.
In particular, if the span of |
w
|
v
e: v, w T
d
and e c is dense in /, then
is unitary and the row-unitary d-tuple | is unitarily equivalent to the normalized
Cuntz-weight model row-unitary d-tuple |
W
via . Moreover, if W
is any other
normalized Cuntz weight and
any other coisometry from / onto L

W
such that
|
j
= |
W
,j
for j = 1, . . . , d and
e = W
(z
e) for all e c, then W
= W
and
= .
Proof. Apart from the uniqueness statement, Theorem 2.10 has already been de-
rived in the motivational remarks at the beginning of this section and in Remark
2.5. We therefore consider only the uniqueness assertion.
Suppose that W
is another normalized Cuntz weight on c for which there

is a coisometry
: / L
W
with
: e W
e for e c for all e c and with

|
j
= |
W
,j
for j = 1, . . . , d.
We claim that W
= W, or W
v,w;,
= W
v,w;,
for all v, w, , T
d
.
Since W
is also normalized,
in fact is isometric on c, and hence also on

span
v,wJ
d
|
w
|
v
c. Hence
W
v,w;,
e, e
)
L
= |
w
|
v
|
e, e
)
K
= |
e, |
v
|
w
)
K
=
(|
e),
(|
v
|
w
))
1
W
= |
W
|
e, |
v
W
|
w
)
1
W
= |
W
|
e, |
v
W
|
w
W
W
)
1
W
= W
(z
e), W
(z
v
w
e
))
1
W
= W
(z
e), z
v
w
e
)
L
2
= W
v,w;,
e, e
)
L
.
We conclude that W = W
as claimed. This completes the proof of Theorem

2.10.
For purposes of the following corollary, let us say that two normalized Cuntz
weights W and W
on Hilbert spaces c and c
respectively are unitarily equivalent

if there is a unitary operator : c c
such that W
v,w;,
=
v,w;,
for all
v, w, , T
d
.
Corollary 2.11. The assignment
(|
1
, . . . , |
d
; c) [W
v,w;,
] = [P
L
|
w
|
v
|
[
L
]
of an equivalence class of normalized Cuntz weights to a row-unitary d-tuple
(|
1
, . . . , |
d
) of operators on a Hilbert space / (or, equivalently, an isometric rep-
resentation of the Cuntz algebra O
d
) together with a choice of -cyclic subspace c
provides a complete unitary invariant for the category of row-unitary d-tuples to-
gether with cyclic subspace. Specically, if (|
1
, . . . , |
d
) is row-unitary on the Hilbert
space / with cyclic subspace c / and (|
1
, . . . , |
d
) is another row-unitary on
the Hilbert space /
with cyclic subspace c
, then there is a unitary transformation

: / /
with |
j
= |
j
for j = 1, . . . , d and such that c = c
if and only if
W
v,w;,
=
W
v,w;,
for all v, w, , T
d
, where
W
v,w;,
= P
L
|
w
|
v
|
[
L
,
W
v,w;,
= P
L
|
w
|
v
|
[
L
, and
the unitary : c c
is given by = [
L
.
Remark 2.12. Theorem 2.10 can be formulated directly in terms of the model
row-unitary |
W
= (|
W,1
, . . . , |
W,d
) on the model space L
W
as follows. For W a
Cuntz weight on c, the matrix entries W
v,w;,
(for v, w, , T
d
) are given by
W
v,w;,
= i
W
|
w
W
|
v
W
|
W
|
W
i
W
(2.48)
where i
W
: c L
W
is the injection operator i : e We. To see (2.48) directly,
note that
W
v,w;,
e, e
)
L
= W(z
e), z
v
w
e
)
L
2
= W(z
e), W(z
v
w
e
))
1
W
= WS
R
(U
R[]
)
e, WS
Rv
(U
R[]
)
w
)
1
W
= |
W
|
W
We, |
v
W
|
w
W
We
)
1
W
= i
W
|
w
W
|
v
W
|
W
|
W
i
W
e, e)
L
and (2.48) follows.
Remark 2.13. From the point of view of classication and model theory for rep-
resentations of the Cuntz algebra, the weakness in Corollary 2.11 is the demand
that a choice of cyclic subspace c be specied. To remove this constraint what
is required is an understanding of when two Cuntz weights W and W
are such
that the corresponding Cuntz-algebra representations |
W
and |
W
are unitarily
equivalent, i.e., when is there a unitary intertwining map S: L
W
L
W
such that
S|
W,j
= |
W
,j
S for j = 1, . . . , d.
Preliminary results on this problem appear in Section 3 where the analytic
intertwining maps S (S: H
W
H
W
) are analyzed.
Remark 2.14. An instructive exercise is to sort out the model for the case d = 1.
Then the alphabet consists of a single letter g and the semigroup T
d
can be
identied with the semigroup Z
+
of non-negative integers (with the empty word
set equal to 0 and a word w = g . . . g set equal to its length [w[ N Z
+
). Hence
a Haplitz weight W is a matrix with rows and columns indexed by Z
+
Z
+
. The
Haplitz property (2.19) means
W
0,n;k+1,
= W
0,n+1;k,
(2.49)
while (2.20) means
W
m+1,n;k+1,
= W
m,n;k,
. (2.50)
Condition (2.21) reduces to
W
0,n;0,k
= W
0,n+1;0,k+1
. (2.51)
and (2.29) becomes
W
m,n;k,
=
_
W
mk+,n
if m k,
W
,n+km
if m k
where
W
i,j
= W
i,j;0,0
= (W
0,0;i,j
)
= (W
0,i;0,j
)
= W
0,j;0,i
=
_
W
0,ji;0,0
=

W
0,ji
if j i,
W
0,0;0,ij
if j i
where, for j i,
W
0,0;0,ij
= (W
0,ij;0,0
)
= (
W
0,ij
)
or alternatively
W
0,0;0,ij
= (W
0,ij;0,0
)
= (W
0,0;ij,0
)
= W
ij,0;0,0
=

W
ij,0
.
Hence if we set
T
k
=
_
W
0,k
if k 0,
W
k,0
= (
W
0,k
)
if k 0,
then

W
i,j
= T
ij
and W
m,n;k,
collapses to
W
m,n;k,
= T
n+km
.
The space L
W
is spanned by elements of the form W(z
k
e) for k, Z
+
and
e c. However, from (2.51) we see that W(z
k
e) = W(z
k+1
+1
e) and hence
we may identify W(z
k
e) simply with W(z

k
e) (where now k in general lies
in Z). The reduced weight

W has row and columns indexed by Z and is given by
W
m,k
= T
km
. In this way, we see that a Haplitz weight for the case d = 1 reduces
to a Laurent matrix [T
km
]
m,kZ
. If we then solve the trigonometric moment
problem to produce a operator-valued measure on the unit circle such that
T
j
=
_
T
z
j
(z),
we then obtain a version of the spectral theorem for the unitary operator |.
3. Analytic intertwining operators between model spaces
Let W and W
be two Cuntz weights (or more generally positive semidenite

Haplitz operators) with block-matrix entries equal to operators on Hilbert spaces
c and c
respectively. Let us say that a bounded operator S: L

W
L
W
is an
intertwining operator if
S|
W,j
= |
W
,j
S, S|
W,j
= |
,j
S
for j = 1, . . . , d. While L
W
is the analogue of the Lebesgue space L
2
, the subspace
H
W
:= closure in L
W
of WT(T
d
, c) is the analogue of the Hardy space
H
2
. Let us say that an intertwining operator S: L
W
L
W
with the additional

property S: H
W
H
W
is an analytic intertwining operator. Note that for d = 1

with L
W
and L
W
simply taken to be L
2
(Z, c) and L
2
(Z, c
), the contractive
analytic intertwining maps are multiplication operators with multiplier T from
the Schur class o(c, c
) of contractive, analytic, operator-valued functions on the

unit disk. In analogy with this classical case, we denote by o
nc
(W, W
) (the non-
commutative Schur class associated with Cuntz weights W and W
) the class of
all contractive, analytic, intertwining operators from L
W
to L
W
. A particularly
nice situation is when there is a formal power series T(z) so that
S(We) = W
[T(z)e] for each e c. (3.1)

When we can associate a formal power series T(z) with the operator So
nc
(W,W
)
in this way, we think of T(z) as the symbol for S and write

S(z) = T(z). The
purpose of this section is to work out some of the calculus for intertwining maps
S o
nc
(W, W
) and their associated symbols T(z), and understand the converse

direction: given a power series T(z), when can the formula (3.1) be used to dene
an intertwining operator S o
nc
(W, W
)? Two particular situations where this

occurs are: (i) W
is a general positive Haplitz operator and T(z) is an analytic

polynomial (so we are guaranteed that T(z)e is in the domain of W
), and (ii) W
is a Haplitz extension of the identity and T(z)e L

2
(T
d
, c
) for each e c (so

again T(z)e is guaranteed to be in the domain of W
). Here we say that the Haplitz

operator W
is a Haplitz extension of the identity if W

;v,;,
=
v,
I
L
where
v,
is the Kronecker delta-function equal to 1 for v = and equal to 0 otherwise. The
setting (ii) plays a prominent role in the analysis of Cuntz scattering systems and
model and dilation theory for row-contractions in [3]. Our main focus here is on
the setting (i).
In general, given a formal power series T(z) =
wJ
d
T
w
z
w
in the non-
commuting variables z = (z
1
, . . . , z
d
) with coecients T
w
equal to bounded oper-
ators between two Hilbert spaces c and c
(T
w
L(c, c
) for w T
d
), we dene
M
T
: T(T
d
, c) L(T
d
, c
) by
(M
T
p)(z) = T(z) p(z) :=
wJ
d
_
_

v,v
: vv
=w
T
v
p
v
_
_
z
w
if p(z) =
wJ
d
p
w
z
w
is a polynomial with coecients in c. There are various
modes of extension of M
T
to spaces of formal power series in a double set of non-
commuting variables (z, ) (with z = (z
1
, . . . , z
d
) and = (
1
, . . . ,
d
) here again
we are assuming that zs do not commute among themselves, s do not commute
among themselves, but z
i
j
=
j
z
i
for i, j = 1, . . . , d). One such operator is the
Laurent operator L
T
dened as follows. We take the domain of L
T
to be the space
T(T
d
T
d
, c) of polynomials f(z, ) =
v,wJ
d
f
v,w
z
v
w
L
2
(T
d
T
d
, c) where
f
v,w
c vanishes for all but nitely many v, w T
d
. Dene L
T
on monomials by
L
T
(z
v
w
e) = S
Rv
U
R[]w
T(z)e
and extend by linearity to T(T
d
T
d
, c). Then L
T
[
1(J
d
],L)
= M
T
[
1(J
d
],L)
.
The dening properties of L
T
: T(T
d
T
d
, c) L(T
d
T
d
, c) are:
L
T
[
1(J
d
],L)
= M
T
[
1(J
d
],L)
, (3.2)
L
T
U
R[]
j
= U
R[]
j
L
T
for j = 1, . . . , d on T(T
d
T
d
, c) (3.3)
L
T
S
R
j
= S
R
j
L
T
for j = 1, . . . , d on T(T
d
T
d
, c). (3.4)
Viewing L(T
d
T
d
, c) as the dual of T(T
d
T
d
, c) in the L
2
-inner product, we
see that L
[]
T
is well dened as an operator of the form L
[]
T
: T(T
d
T
d
, c
)
L(T
d
T
d
, c).
The following proposition gives some useful formulas for the action of L
T
and L
[]
T
.
Proposition 3.1. Let T(z) =
vJ
d
T
v
z
v
be a formal power series.
1. Then the action of the Laurent operator L
T
on a polynomial
f(z, ) =
,
f
,
z
in T(T
d
T
d
, c) is given by
L
T
[f](z, ) = T(
1
)f(z, ) [T(
1
)f(z, )][
=0
+T(z
)f(z, z
1
)[
z
=z
(3.5)
where the z
-variables are taken to the left of the z-variables in the last term
before the evaluation z
= z.
2. The action of L
[]
T
on a general polynomial g(z, ) =
v,w
g
v,w
z
v
w
T(T
d
T
d
, c
) is given by
L
[]
T
[g](z, ) = T()
[g(z, ) +k
per
(z, )g(z, 0)] (3.6)
where we use the convention T()
vJ
d
T
and where k
per
(z, ) is
the perverse Szego kernel as in (2.33).
Remark 3.2. We see that the formulas (3.5) and (3.6) actually dene L
T
and L
[]
T
from all of L(T
d
T
d
, c) into L(T
d
T
d
, c
) and from L(T

d
T
d
, c
) into L(T
d
T
d
, c) respectively. This is analogous to the fact that the operator M
T
: f(z)
T(z)f(z) is well dened as an operator on formal power series M
T
: L(T
d
, c)
L(T
d
, c
), as only nite sums are involved in the calculation of a given coecient

of T(z) f(z) in terms of the (innitely many) coecients of f(z).
Proof. By denition, for e c and T
d
, we have
L
T
(
e) = U
R[]
L
T
e
= U
R[]
vJ
d
T
v
z
v
_
e
where
U
R[]
k
(T(z)e) = T
k
e +T(z)z
1
k
e,
U
R[]
j
U
R[]
k
L
T
e = U
R[]
j
(T
k
e +T(z)z
1
k
e)
= T
j
e +T
k
j
e +T(z)z
1
k
z
1
j
e
and then by induction
L
T
(
e) = (U
R[]
)
L
T
(e) =
: =
,=
T
e +T(z)(z
1
)
J
d
T
(
1
)
e T
e +T(z)(z
1
)
e
= T(
1
)
e [T(
1
)
e][
=0
+T(z
)(z
1
)
[
z
=z
and formula (3.5) follows for the case where f(z, ) =
.
For a general monomial f(z, ) = z
, we have
L
T
(z
e) = L
T
(S
R
)
(U
R[]
)
e
= (S
R
)
_
(U
R[]
)
L
T
e
_
= (S
R
)
_
_

J
d
T
(
1
)
e T
e +T(z)(z
1
)
_
_
e
=
J
d
T
(
1
)
e T
e +
_
T(z)(z
1
)
_
z
e
= T(
1
) z
e [T(
1
)z
e][
=0
+T(z
)z
(z
1
)
[
=z
and formula (3.5) follows for the case of a general monomial. The general case of
(3.5) now follows by linearity.
To verify (3.6), we rst compute the action of L
[]
T
on a monomial; for
u, v, , T
d
, e
and e c, compute
L
[]
T
z
v
w
e
, z
e)
L
2 = z
v
w
e
, L
T
z
e)
L
2
= z
v
w
e
, (M
)z
e T
e + (T(z)(z
1
)
)z
e)
L
2
=
,v
M
T
w
e
e)
L
2
,v
w,
T
, e)
L
+
w,
z
v
e
, (T(z)(z
1
)
)z
e)
L
2 . (3.7)
From this we read o that, for w ,= we have
L
[]
T
(z
v
w
e
) = (M
T
w
)z
v
e
= T()
z
v
w
e
By linearity we see that the formula (3.6) holds for the case when f T(T
d

(T
d
), c), i.e., for the case where f(z, 0) = 0.
For the case where w = , the second term on the right-hand side of (3.7)
(with the sign reversed) is equal to
,v
T
, e)
L
=
,v
T
z
v
e
, z
e)
L
2 =
,v
T()
z
v
e
, z
e)
L
2
and thus exactly cancels with the rst term. We are therefore left with
L
[]
T
z
v
e
, z
e)
L
2 = z
v
e
, T
(v
1
)
z
v
e)
L
2
= e
, T
(v
1
)
e)
L
= (T
(v
1
)
)
, e)
L
from which we deduce that
L
[]
T
(z
v
e
) =
,
(T
(v
1
)
)
,: v=v
(T
v
,: v=v
,: v=v
T()
(
1
)
v
= T()
(
1
)
v
(z
1
)
v
z
v
_
= T()
[z
v
+k
per
(z, )z
v
]
which is formula (3.6) for the case where f(z, ) = z
v
. By linearity (3.6) holds if
f(z, ) = f(z) =
v
f
v,
z
v
. Since any f T(T

d
T
d
, c) can be decomposed as
f(z, ) = [f(z, ) f(z, 0)] +f(z, 0)
where f(z, ) f(z, 0) T(T
d
(T
d
), c) and f(z, 0) T(T , c), we
can use linearity once again to see that (3.6) holds for all f T(T
d
T
d
, c).
There is a second type of extension of a multiplication operator M
T
dened
as follows. Suppose that W
is a positive semidenite Haplitz operator on c
, W is
a positive semidenite Haplitz operator on c, and T(z) is a formal power series in
the z-variables only such that T(z)e is in the domain of the Haplitz operator W
for each e c; as was mentioned above, two cases where this occurs are (i) when
W
is general and T(z) is a polynomial, and (ii) when W
is a Haplitz extension
of the identity and M
T
is a bounded multiplier from L
2
(T
d
, c) to L
2
(T
d
, c
). We
then dene L
W,W
T
: WT(T
d
T
d
, c) W
T(T
d
T
d
, c
) by
L
W,W
T
: Wp W
L
T
p. (3.8)
If it happens that |W
L
T
p|
1
W
M|Wp|
1
W
for some M < , or equivalently,
that
M
2
W L
[]
T
W
L
T
0,
then L
W,W
T
is well dened and has a continuous extension (also denoted by L
W,W
T
)
L
W,W
T
: L
W
L
W
.
Moreover one easily checks that
L
W,W
T
|
W,j
= |
W
,j
L
W,W
T
and L
W,W
T
|
W,j
= |
,j
L
W,W
T
.
We shall be particularly interested in the case where L
W,W
T
denes a contraction,
isometry or unitary operator between L
W
and L
W
; the following proposition

summarizes the situation. The proof is an elementary continuation of the ideas
sketched above.
Proposition 3.3. Let W and W
be positive semidenite Haplitz operators with ma-

trix entries W
v,w;,
and W
;v,w;,
equal to bounded operators on Hilbert spaces
c and c
respectively, let T(z) =
vJ
d
T
v
z
v
be a non-commutative analytic poly-
nomial with coecients T
v
equal to bounded operators from c to c
, and dene
L
W,W
T
: WT(T
d
T
d
, c) T(T
d
T
d
, c
)
by (3.8). Then:
1. L
W,W
T
extends by continuity to dene a contraction operator from L
W

L
W
if and only if the Haplitz operator WL

[]
T
W
L
T
is positive semidenite.
2. L
W,W
T
extends by continuity to dene an isometry from L
W
into L
W
if and
only if W L
[]
T
W
L
T
= 0.
3. L
W,W
T
extends by continuity to dene a unitary operator from L
W
onto L
W
if and only if WL
[]
T
W
L
T
= 0 and the subspace span
(z, )T(z)e: e
c is -cyclic for |
W
.
In each of these cases, the extended operator L
W,W
T
: L
W
L
W
has the additional

properties
L
W,W
T
|
W,k
= |
W
,k
L
W,W
T
and L
W,W
T
|
W,k
= |
,k
L
W,W
T
for k = 1, . . . d.
(3.9)
The next sequence of results will help us compute the symbol

V (z, ) for a
Haplitz operator V of the form V = L
[]
T
W
L
T
.
Proposition 3.4. Assume that W
= [W
;v,w;,
] is a selfadjoint Haplitz operator
(with matrix entries W
;v,w;,
equal to bounded operators on the Hilbert space c
)
and that T(z) =
vJ
d
T
v
z
v
is a polynomial (so T
v
= 0 for all but nitely many
v T
d
) with coecients equal to bounded operators from c into c
. Then W
L
T
,
L
[]
T
W
and L
[]
T
W
L
T
are well dened []-Haplitz operators with respective symbols
L
T
(z, ) =
(z, )T(z) +
(0, )k
per
(z, )T(z) (3.10)
L
[]
T
W
(z, ) =T()
(z, ) +T()
k
per
(z, )
(z, 0) (3.11)
L
[]
T
W
L
T
(z, ) =T()
(z, )T(z) + T()
(0, )k
per
(z, )T(z)] (3.12)
+T()
k
per
(z, )[
(z, 0)T(z) + (
(0, z
1
)
W
;,
)T(z)]
(3.13)
The latter formula can also be written in the more selfadjoint form
L
[]
T
W
L
T
(z, ) = T()
(z, )T(z)
+T()
(0, )k
per
(z, )T(z)] + [T()
k
per
(z, )
(z, 0)]T(z)
+T()
k
per
(z, )[(
(0, z
1
)
W
;,
)T(z)]
+ [T()
(
1
, 0)
W
;,
)]k
per
(z, )T(z)
+T()
W
;,
k
per
(z, )T(z). (3.14)
In case W
is a []-Haplitz extension of the identity, formulas (3.10), (3.11)

and (3.13) simplify to
L
T
(z, ) =

W
(z, )T(z), (3.15)
L
[]
T
W
(z, ) = T()
(z, ), (3.16)
L
[]
T
W
L
T
(z, ) = T()
(z, )T(z) +T()
k
per
(z, )T(z). (3.17)
Proof. The formula (3.10) for

W
L
T
(z, ) is an immediate consequence of for-
mula (2.34) in Proposition 2.7. Since L
[]
T
W
= (W
L
T
)
[]
, the formula (3.11) for
L
[]
T
W
(z, ) then follows by applying the general formula (2.43) to W
L
T
.
To verify the formula (3.13) for

L
[]
T
W
L
T
(z, ), we compute
L
[]
T
W
L
T
(z, ) = L
[]
T
_
L
T
(z, )
_
(z, )
= T()
(z, )T(z) +T()
(0, )k
per
(z, )T(z)]
+T()
k
per
(z, )
_
(z, 0)T(z) + [
(0, )k
per
(z, )T(z)][
=0
_
= T()
(z, )T(z) +T()
(0, )k
per
(z, )T(z)]
+T()
k
per
(z, )
_
(z, 0)T(z) + (
(0, z
1
)
W
;,
)T(z)
_
and (3.13) follows.
To see that (3.15), (3.16) and (3.17) follows from (3.10), (3.11) and (3.13),
observe that, for

W
(z, 0) =

W
(0, ) = I,
(0, )k
per
(z, )T(z) = 0
T()
k
per
(z, )
(z, 0) = 0
(0, z
1
)
W
;,
= 0.
To see that (3.14) is equivalent to (3.13), proceed as follows. First we note
the general identity, for g() =
w
g
w
w
a formal power series in the variables
= (
1
, . . . ,
d
) and f(z) =
v
f
v
z
v
a formal power series in the variables z =
(z
1
, . . . , z
d
),
g()k
per
(z, )f(z) =
v,w
_
_
,=
g
wv
f
v
v
_
_
z
v
w
. (3.18)
Next, let us write the right-hand side of (3.13) as k
0
(z, ) + k
1
(z, ) + k
2
(z, ) +
k
3
(z, ) where
k
0
(z, ) = T()
(z, )T(z)
k
1
(z, ) = T()
(0, )k
per
(z, )T(z)]
k
2
(z, ) = T()
k
per
(z, )[
(z, 0)T(z)]
k
3
(z, ) = T()
k
per
(z, )[(
(0, z
1
)
W
;,
)T(z)].
We use the general identity (3.18) to compute
k
2
(z, ) = T()
k
per
(z, )
_
(z, 0)T(z)
_
= T()
k
per
(z, )
_
v
_

=v
W
;
,
T
_
z
v
_
=
v,w
_
_

,v
S
2
T
(wv
W
;
,
T
_
_
z
v
w
(3.19)
where we have set
S
2
= (
, v
):
= v
v, v
,= .
Similarly,
k
1
(, z)
=
_
T()
k
per
(z, )
(z, 0)
_
T(z)
=
_
_
v,w
_
_
,=
T
(wv
W
;v
v
_
_
z
v
w
_
_
_
=
v,w
_
_

(
,v
)S
1
T
(wv
W
;v
_
_
z
v
w
where
S
1
= (
, v
):
= v, v
,= .
Note that the map
: (
, v
) (
, v
) := (v
, v
)
maps S
1
injectively into S
2
, with the remainder set S
2
S
1
be given by
S
2
S
1
= (
, v
):
= v
for some
,= ,
v.
When forming the dierence k
2
(z, ) k
1
(, z)
, the terms in k
2
(z, ) over in-
dices in (S
1
) exactly cancel with the terms of k
1
(, z)
and we are left with the

associativity defect
k
2
(z, ) k
1
(, z)
v,w
_
_

(
,v
)S
2
\S
1
T
(wv
,
T
_
_
z
v
w
. (3.20)
On the other hand, we compute, again using (3.18),
k
3
(, z)
=
_
T()
W(
1
, 0)
W
,
)
_
k
per
(z, )T(z)
=
_
_
w
_
_
,=
T
(wv
W
v
,
_
_
w
_
_
k
per
(z, )T(z)
=
v,w
_
_
,=
,=
T
(wv
W
v
,
T
v
v
_
_
z
v
w
. (3.21)
A close comparison of (3.20) and (3.21) shows that
k
2
(z, ) k
1
(, z)
k
3
(, z)
v,w
_
_

,=
T
W
;,
T
v
_
_
z
v
w
= T()
W
;,
k
per
(z, )T(z) =: k
4
(z, ). (3.22)
By using this last identity (3.22) we now get
k
0
(z, ) +k
1
(z, ) +k
2
(z, ) +k
3
(z, ) = k
0
(z, ) + [k
1
(z, ) +k
1
(, z)
]
+ [k
2
(z, ) k
1
(, z)
] +k
3
(z, )
= k
0
(z, ) + [k
1
(z, ) +k
1
(, z)
]
[k
3
(z, )
+k
4
(z, )] +k
3
(z, )
= k
0
(z, ) + [k
1
(z, ) +k
1
(, z)
]
+ [k
3
(z, ), +k
3
(, z)
] +k
4
(z, )
where the last line is exactly equal to the right-hand side of (3.14). Thus (3.13)
and (3.14) are equivalent as asserted.
Finally, suppose that W
is a Haplitz extension of the identity, i.e., W

;v,;,
=
v,
I
L
. From (2.19) we see that
W
;v,w
= W
;v,w;,
= W
;v,;w
,
=
v,w
I
L
.
In particular

W
;,w
=
,w
I
L
and W
;v,
=
v,
I
L
from which we get

W
(z, 0) =
I
L
and

W
(0, ) = I
L
. Hence we are in the situation where formula (2.35)

applies for the action of

W on an analytic polynomial. With this simplication,
formulas (3.15), (3.16) and (3.17) follow as specializations of the more general
formulas (3.10), (3.11) and (3.13).
Remark 3.5. From the identity (2.43) we know that the kernel
k(z, ) :=

L
[]
T
W
L
T
(z, )
must satisfy k(z, ) = k(, z)
(since L
[]
T
W
L
T
= (L
[]
T
W
L
T
)
[]
). While this prop-
erty is not apparent from the rst formula (3.13) for k(z, ), it is apparent from
the second formula (3.13).
4. Application: Free atomic/permutative representations
of the Cuntz algebra O
d
We illustrate the machinery developed in the preceding sections by applying it to
free atomic representations of the Cuntz algebra O
d
studied in [9], a class more gen-
eral than but closely related to the permutative representations of O
d
studied in [7].
We rst remark that a representation of the C
-algebra known as the Cuntz

algebra O
d
amounts to a specication of a Hilbert space /
together with a d-tuple

|
= |
,1
, . . . , |
,d
of operators on /
which form a row-unitary operator:

|
,i
|
,j
=
i,j
I
K
,
d
k=1
|
,k
|
,k
= I
K
(see, e.g., [8]). To simplify the terminology, we generally shall say row-unitary
rather than representation of the Cuntz algebra O
d
.
We introduce the class of free atomic representations of O
d
studied by David-
son and Pitts [9] in the framework used by Bratteli and Jorgensen [7] for the study
of the special case of permutative representations. We let / be a Hilbert space with
orthonormal basis e
i
: i 1 indexed by some index set 1. Let = (
1
, . . . ,
d
)
be a function system of order d on 1; by this we mean that each
k
: 1 1 is an
injective function from 1 to 1
k
(i) =
k
(i
) for some i, i
1 =i = i
for each k = 1, . . . , d, (4.1)

the
k
s have pairwise disjoint images
im
k
im
= for k ,= in 1, . . . , d, (4.2)
and the union of the images of the
k
s is all of 1:
d
k=1
im
k
= 1. (4.3)
In addition assume that we are given a collection of numbers
=
k,i
T: k = 1, . . . , d; i 1
each of modulus 1 indexed by 1, . . . , d 1. For convenience we shall refer to the
combined set (, ) as an iterated function system. We then dene a weighted shift
operator |
k
(for k = 1, . . . , d) on / by extending by linearity the action on the
basis vectors e
i
: i 1 given by
|
k
: e
i

k,i
e
k
(i)
. (4.4)
Then, by using that each
k,i
has modulus 1 and the properties (4.1)(4.3) of an
iterated function system, it is easy to see that | = (|
1
, . . . , |
d
) is row-unitary.
When we wish to make the dependence on and =
k,i
: k = 1, . . . , d; i 1
explicit, we write |
,
= (|
,
1
, . . . , |
,
d
) in place of | = (|
1
, . . . , |
d
).
We shall need a certain calculus associated with an iterated function system
(, ). For a given i 1 there is a unique k = k(i) 1, . . . , d such that i =
k
(i
)
for some (necessarily unique) i
1. Then we write i
=
1
k
(i). For v T
d
a word
of the form v = g
k
n
g
k
1
(with k
1
, . . . , k
d
1, . . . , d), we dene
v
as the
composition of maps
v
=
k
n

k
1
. If k
is an element of 1, . . . , d not
equal to this particular k, then we say that
1
k
(i) is undened (or empty). More
generally, given i 1 and a natural number n, there is a unique word v = g
k
n
. . . g
k
1
in T
d
of length n so that
v
(i
) = i for some (necessarily unique) i
1. When
this is the case we then write i
= (
v
)
1
(i) = (
1
)
v
(i). If v
is another word
in T
d
of length n not equal to this particular v, then we say that (
1
)
v
(i) is
undened or empty.
This calculus extends to the set of modulus-1 multipliers as follows. If g
k
is
a single letter (word of length 1) in T
d
and i 1, we dene
g
k
i
=
k,i
. Inductively,
if v = g
k
n
g
k
1
T
d
is a word of length n and i 1, we dene
v
i
=
k
n
,
k
n1
k
1
(i)

k
2
,
k
1
(i)
k
1
,i
.
Similarly, given v = g
k
n
g
k
1
T
d
and i 1, if there is some (necessarily unique)
i
1 so that
v
(i
) = i, then we dene
i
=
k
1
,(
1
)
v
(i)

k
n1
,
1
k
n
(i)
k
n
,i
;
otherwise, we set
v
i
equal to 1. Properties of this calculus then are
vv
i
=
v
(i)
i
,
v
i
= (
v
i
) if
v
(i
) = i. (4.5)
This calculus gives us a simple notation for calculating products of |
,
1
, . . . , |
,
d
on basis vectors e
i
; indeed, if we simplify the notation by writing simply | for
|
,
, we have
|
v
e
i
=
v
i
e
v
(i)
,
|
v
e
i
=
v
i
e
(
1
)
v
(i)
. (4.6)
Here it is understood that e
(
1
)
v
(i)
= 0 if (
1
)
v
(i) is undened.
The calculus is further streamlined if we introduce some more formalism.
As was already used above, we know that, for a given i 1, there is a unique
k = k(i) 1, . . . , d such that i =
k
(i
) for some (necessarily unique) i

1. Iteration of this observation enables to dene inductively a function : 1
(1,...,d1)
N
by
(i) = (k
1
, i
1
), (k
2
, i
2
), . . . , (k
n
, i
n
), . . . (4.7)
by the conditions
k
1
(i
1
) = i,
k
2
(i
2
) = i
1
, . . . ,
k
n+1
(i
n+1
) = i
n
, . . . .
On occasion we need to separate the rst and second components of as fol-
lows. Dene
1
: 1 g
1
, . . . , g
d
N
(viewed as the set of innite words in the
alphabet g
1
, . . . , g
d
having a beginning on the left but no end on the right) and
2
: 1 1
N
by
1
(i) = g
k
1
g
k
2
g
k
n
, (4.8)
2
(i) = i
1
, i
2
, . . . , i
n
, . . . (4.9)
if (i) is as in (4.7). After a cosmetic change of notation, the function
1
is exactly
the object associated with the iterated function system termed the coding func-
tion by Bratteli-Jorgensen in [7]. The function (4.7) has already appeared in an
implicit way in the (, )-calculus described in the previous paragraph. Indeed, if
v is a word of length n in T
d
, then (
1
)
v
(i) is dened exactly when

1
(i) has the
form
1
(i) = vv
for some v
g
1
, . . . , g
d
N
. Also, one can compute
v
i
directly
in terms of the original set of multipliers =
k,i
: k 1, . . . , d, i 1 via
i
=
k
1
,i
1

k
n
,i
n
where (k
1
, i
1
), (k
2
, i
2
), . . . , (k
n
, i
n
) are the rst n terms of the sequence (i).
As a consequence of (4.6) it is easy to see that: given a subset 1
1,
the smallest subspace H of / containing each of the basis vectors e
i
for i 1
and reducing for |

,
is the subspace H := closed span e
i
: i 1
where 1
v,wJ
d
w
(
1
)
v
(1
). An application of this general principle tells us that a sub-

set 1
1 has the property that the associated subspace H := closed span e

i
: i
1
/ is reducing for |
,
if and only if 1
has the property

1
=
d
_
k=1
k
(1
). (4.10)
A consequence of this latter statement is: for a given i 1, the basis vector e
i
is -cyclic for |
,
if and only if the action of on 1 is ergodic, i.e., the only
subset 1
1 having the property (4.10) is 1
= and 1
= 1. Note that this

latter statement is independent of the choice of i 1; thus e
i
is -cyclic for |
,
for some i 1 if and only if e
i
is -cyclic for each i 1. In general 1 partitions
into ergodic subsets: 1 = 1
: / where 1
= for ,=
, 1
is
invariant under both and
1
, and the restriction of to 1
is ergodic for each

/. Then the corresponding row-unitary d-tuple |
,
splits as a direct sum
|
,
=
,
|
. Hence, for purposes of studying the problem of classication

up to unitary equivalence of the row-unitary d-tuples |
,
, there is no loss of
generality in assuming that 1 is ergodic with respect to , i.e., the only subsets
1
of 1 satisfying (4.10) are and the whole set 1.

Thus throughout the rest of this section we assume that is ergodic on 1.
Let us now x an i 1, . . . , d; then e
i
is -cyclic with respect to |
,
. Our next
goal is to compute the symbol W
e
i
)
(z, ) of | = |
,
with respect to the cyclic
subspace e
i
) = span e
i
. The coecients of W
e
i
)
by denition are operators on
the one-dimensional space e
i
); we identify the element ce
i
of e
i
) with the complex
number c C and an operator ce
i
d ce
i
on e
i
) with the complex number d.
Under these conventions, we compute
W
e
i
)
(z, ) =
v,wJ
d
_
P
e
i
)
|
w
|
v
[
e
i
)
_
z
v
w
=
v,w
P
e
i
)
_
|
w
i
e
(
1
)
v
(i)
_
z
v
w
=
v,w:
w
((
1
)
v
(i))=i
w
(
1
)
v
(i)
i
z
v
w
=
(i
,v,w)S
w
i

v
i
z
v
w
(4.11)
where we have set
o := (i
, v, w): i
1, . . . , d, v, w T
d
with
v
(i
) =
w
(i
) = i. (4.12)
If (i) = (k
1
, i
1
), (k
2
, i
2
), . . . , (k
n
, i
n
), . . . , we let x be the innite word x =
1
(i) = g
k
1
g
k
2
g
k
n
(see (4.7) and(4.8)). For each m = 0, 1, 2, . . . , let x
m
be
the nite word of length m given by
x
m
= g
k
1
g
k
m
(with x
0
= ) for each m = 0, 1, 2, . . . . To complete the notation, set i
0
= i.
Suppose that i
1, . . . , d and that v and w are words in T

d
with [w[ [v[.
From the denitions we see that (i
, v, w) o if and only if there are non-negative

integers j and N so that
v = x
j
, w = x
j+N
=: x
j
y, i
j
= i
j+N
= i
. (4.13)
Note that (4.13) is trivially true if we take N = 0 and j = 0, 1, 2, . . . arbitrary;
from this observation we see that
T := (i
m
, x
m
, x
m
): m = 0, 1, 2, . . . o. (4.14)
Suppose next that (i
, v, w) = (i
, x
j
, x
j+N
) := (i
, x
j
, x
j
y) is in o for some N > 0.
Then by denition we have
i =
x
j
(i
) =
x
j
y
(i
)
from which it follows that
y
(i
) = i
. From the rule dening x =

1
(i), we see
next that
x = x
j
yyy =: x
j
y
.
In the statements below, we shall use the following notation. Given a triple of
parameters (x, y, ) with x, y T
d
and a complex number of modulus 1 ( T),
we let

W
x,y,
(z, ) denote the symbol
W
x,y,
(z, ) =
: x=
,=
=y,
,=
m, m0
m
z
(xy
m
)
xy
m
(4.15)
and W
x,y,
= [W
x,y,
v,w;,
] the associated []-Haplitz operator as in (2.29) in Proposi-
tion 2.6. In case x = , we shall abbreviate
W
,y,
(z, ) to
W
y,
(z, ), and similarly
W
,y,
to W
y,
; thus
W
y,
(z, ) =
=y,
,=
m, m0
m
z
(y
m
)
y
m
. (4.16)
Similarly, if x = x
1
x
2
x
3
. . . x
n
. . . is an innite word (with letters x
j
g
1
, . . . , g
d
for j N), we let

W
x
(z, ) denote the symbol
W
x
(z, ) =
m=0
z
x
x
m
(4.17)
with associated []-Haplitz operator W
x
. We are well on our way to proving the
following result.
Proposition 4.1. Suppose that (, ) is an iterated function system with the action
of = (
1
, . . . ,
d
) on the index set 1 ergodic as above. Fix an index i 1 and
the associated -cyclic vector e
i
for the row-unitary d-tuple |
,
and let
(i) = (k
1
, i
1
), (k
2
, i
2
), . . . , (k
n
, i
n
), . . .
with x equal to the innite word
x =
1
(i) = g
k
1
g
k
2
g
k
n

as in (4.7) and (4.8) and x
m
= g
k
1
g
k
m
T
d
as above. Then:
Case 1: The orbit-eventually-periodic case: Suppose that there are integers j 0
and N > 0 so that
(k
j
+N
, i
j
+N
) = (k
j
, i
j
) for all j
j.
Assume also that j and N are chosen to be minimal with these properties. Then
x has the form
x = x
j
y
for a word y T
d
of length N such that y has no common tail with x
j
:
x = x
v and y = y
v =v = ,
the set o dened by (4.12) is given by
o =(i
]]
, , ): x
j
=
where
,=
_
(i
]x
j
]+]]
, x
j
y
m
, x
j
y
m
): m, m = 0, 1, 2, . . . , and y =
with
,= ,
and the symbol

W
e
i
)
(z, ) for the row-unitary |
,
with respect to the cyclic sub-
space e
i
) has the form
W
e
i
)
(z, ) =

W
x
j
,y,
(z, ) (4.18)
where =
y
x
j
(i)
and

W
x,y,
(z, ) in general is as in (4.15). Hence |
,
is unitar-
ily equivalent to the model row-unitary |
W
x,y, on the space L
W
x,y, (with x = x
j
,
=
y
i
and
1
(i) = x
m
y
as above).
Case 2: The orbit-non-periodic case: Suppose that there is no integer N > 0 as in
Case 1. Then the o in (4.12) is given by
o = T := (i
m
, x
m
, x
m
): m = 0, 1, 2, . . .
and the symbol

W
e
i
)
(z, ) for the row-unitary |
,
with respect to the cyclic sub-
space e
i
) has the form
W
e
i
)
(z, ) =

W
x
(z, ). (4.19)
with

W
x
(z, ) as in (4.17). Hence in this case |
,
is unitarily equivalent to the
model row-unitary |
W
x on the space L
W
x (where x =
1
(i) as above).
Proof. The description of the set o for the two cases follows from the discussion
immediately preceding the statement of Proposition 4.1. One can check from the
denitions that the lack of a common tail between x
j
and y in the orbit-eventually-
periodic case is forced by the assumed minimality in the choice of j and N. The
formulas for W
e)
(z, ) then follow by plugging into (4.11) and using the properties
(4.5) of the -functional calculus. The statements on unitary equivalence then
follow from Theorem 2.10.
Remark 4.2. In the orbit-eventually-periodic case, we see that x =
1
(i) has the
form x = x
j
y
for some j 0 and some word y T

d
. Here the length of y
is chosen equal to the smallest possible eventual period for (i) (see (4.7)). The
smallest possible eventual period for x =
1
(i) (see (4.8)) may well be smaller: in
this case y has the form y = u
n
for some shorter word u (and then [y[ = n[u[). As
we shall see below and is obtained in [9], in this case |
,
is reducible, despite the
assumption that (, ) is ergodic. Indeed, if (, ) is ergodic, then |
,
can have
no non-trivial reducing subspaces spanned by a subset of the shift orthonormal
basis e
i
: i 1; non-trivial reducing subspaces not in the span of any subset of
e
i
: i 1 can occur.
Similarly, in the orbit-non-periodic case, the assumption is that (i) (as in
(4.7)) is never eventually periodic. This does not eliminate the possibility that
x =
1
(i) (as in (4.8)) is eventually periodic; in this case, as we shall see below
and is obtained in [9], |
,
is a direct integral of orbit-periodic representations,
and hence has many non-trivial reducing subspaces. If (, ) is ergodic, then none
of these non-trivial reducing subspaces can have an orthonormal basis consisting
of a subset of the shift basis e
i
: i 1.
A consequence of Proposition 4.1 is that, for purposes of classifying |
,
up to
unitary equivalence, we need study only the model operators |
W
x,y, (for the orbit-
eventually-periodic case) and |
W
x (for the non-periodic case). In particular, we see
that the unitary-equivalence class of |
,
depends only on the three parameters,
x, y, (where x, y T
d
and T) in the orbit-eventually-periodic case, or the
single parameter x (where x is an innite word) in the orbit-non-periodic case.
Conversely, given any parameters (x, y, ) with x, y T
d
and T, there exists
an ergodic iterated function system (, ) on an index set 1 with -cyclic vector
e
i
so that
1
(i) = xy
, =
y
x
(i)
and such that (i) is periodic (with period
[y[ once the index j satises j > [x[); indeed, one can see that the representation
u,
written down in [9] (with an appropriate choice of -cyclic vector) is of this
type, or one can use the details of the model row-operator |
W
x,y, on the model
space L
W
x,y, (see Remark 4.3 below for a brief sketch). Similarly, given an innite
word x = g
k
1
g
k
2
g
k
n
, there is an ergodic iterated function system (, 1) with
1
(i) = x and with (i) aperiodic; one can check that a representation of the
form
x
as dened in [9] has this property, or one can work with the model |
W
x
on the model space L
W
x as outlined in Remark 4.3 below. Let us denote any
row-unitary of the rst type as |
x,y,
(abbreviated to |
y,
if x = ) and of the
second type as |
x
.
Remark 4.3. If

W =

W
x,y,
(with x and y equal to nite words and T)
or if

W =

W
x
(with x equal to an innite word), then it is a direct verica-
tion that there exists a formal power series Y (z) (with coecients equal to in-
nite block-rows) in the single set of non-commuting variables z = (z
1
, . . . , z
d
) so
that

W(z, ) = Y ()Y (z)
, and that

W(z, ) has zero Cuntz defect (
W(z, ) =
d
k=1
z
1
k
W(z, )
1
k
). Hence, |
W
on L
W
denes a row-unitary d-tuple. A further
direct verication shows that |
W
on L
W
is atomic, i.e.,
|
w
W
|
v
W
(W[1]) = W(z
w
v
)
is an orthonormal basis, up to having many zero vectors and having many rep-
etitions up to a unimodular factor. Thus the index set for a shift basis for |
W
is 1 = (T
d
T
d
)/

=, where the rule

= drops elements (v, w) of T
d
T
d
for
which W[z
v
w
] = 0 and identies elements (v, w) and (v
, w
) in T
d
T
d
for which
W[z
v
w
] and W[z
v
] are the same up to a unimodular factor. The row-unitary

|
W
then induces an iterated function system (, ) on the index set 1/

=, from
which we recover
((, )) periodic with period [y[ once n > [x[,
1
((, )) = xy
, =
y
x
((,))
if W = W
x,y,
,
((, )) non-periodic,
1
(, ) = x if W = W
x
.
In this way, starting with parameters (x, y, ) or x, we produce a functional
model for the free-atomic representations of O
d
given in [7, 9].
To sort out how these model operators are related, we rst need to understand
how these various symbols are related. The following proposition sorts these issues
out.
Proposition 4.4. The following relations among symbols of the form

W
x,y,
(z, ),
W
y,
(z, ) and

W
x
(z, ) as in (4.15), (4.16) and (4.17) hold:
1. Let x, y T
d
and T given. Assume that x does not share a tail with y,
i.e.,
x = x
and y = y
for some T
d
= = .
If we set T(z) = z
x
, then
W
x,y,
(z, ) = [L
[]
T
W
y,
L
T
]
(z, ). (4.20)
2. For x
T
d
and an innite word x = x
1
x
2
x
3
. . . x
n
. . . with x
j
g
1
, . . . , g
d
for each j N given, if we set T(z) = z

x
, then
W
x
x
(z, ) = [L
[]
T
W
x
L
T
]
(z, ). (4.21)
3. For y
, y T
d
and T with y
a cyclic permutation of y (so y =
and
y
for some T
d
), if we set T(z) = z
, then
W
y,
(z, ) = [L
[]
T
W
y
,
L
T
]
(z, ). (4.22)
4. For y T
d
given of the form y = u
n
for some u T
d
, then
W
y,
(z, ) =
1
n
:
n
=
W
u,
(z, ). (4.23)
5. Given an innite word x = x
1
x
2
x
3
x
n
of the form x = uuu =: u
for some u T
d
, then
W
x
(z, ) =
_
T
W
u,
(z, ) dm() (4.24)
where m is normalized Lebesgue measure on the unit circle T.
Proof of (4.20). Consider (4.20). Note that
W
y,
(0, ) =
m0
y
m
,

W
y,
(z, 0) =
m0
m
z
(y
m
)
.
Since x does not share a tail with y, one easily checks that
W
y,
(0, )k
per
(z, )z
x
= 0 and
_
W
y,
(0, z
1
)
W
y,
,
_
z
x
= 0.
Similarly, we see that
x
k
per
(z, )[
m
z
(y
m
)
z
x
] = 0 for 0 ,= m
and hence
x
k
per
(z, )
_
_
m0
m
z
(y
m
)
z
x
_
_
=
x
k
per
(z, )z
x
: x=
,=
.
We now have all the pieces to plug into the formula (3.13) for [L
[]
T
W
y,
L
T
]
(z, );
the result is
[L
[]
T
W
y,
L
T
]
(z, ) =
x
W
y,
(z, )z
x
+
x
[
W
y,
(0, )k
per
(z, )z
x
]
+
x
k
per
(z, )[
W
y,
(z, 0)z
x
]
+
x
k
per
(z, )
__
W
y,
(0, z
1
)
W
y,
,
_
z
x
_
=
x
W
y,
(z, )z
x
+
x
k
per
(z, )[
W
y,
(z, 0)z
x
]
=
x
W
y,
(z, )z
x
+
x
k
per
(z, )
_
_
m0
m
z
(y
m
)
z
x
_
_
=
x
W
y,
(z, )z
x
: x=
,=
=

W
x,y,
(z, )
as asserted, and (4.20) follows.
Proof of (4.21). We next consider (4.21). Note that W
x
is a []-Haplitz extension
of the identity. Hence, we may apply (3.17) to compute
[L
[]
T
W
x
L
T
]
(z, ) =
x
W
x
(z, )z
x
+
x
k
per
(z, )z
x
=
x
W
x
(z, )
x
=x
,=
=

W
x
x
(z, )
and (4.21) follows as wanted.
Proof of (4.22). Now suppose that y =
, y
in T
d
and we wish to check
(4.22) as follows. By (3.13) applied to the case where W
= W
y
,
and T(z) = z
,
we know that
[L
[]
T
W
y
,
L
T
]
(z, ) = k
1
(z, ) +k
2
(z, ) +k
3
(z, ) +k
4
(z, ) (4.25)
where
k
1
(z, ) =
W
y
,
(z, )z
,
k
2
(z, ) =
W
y
,
(0, )k
per
(z, )z
_
k
3
(z, ) =
k
per
(z, )
_
W
y
,
(z, 0)z
_
k
4
(z, ) =
k
per
(z, )
__
W
y
,
(0, z
1
) W
y
,
,
_
z
_
.
Plugging in the denition (4.16) of

W
y
,
(z, ) into the formula for k
1
(z, ) then
gives
k
1
(z, ) =
_
_

: y
,=
m, m0
m
z
(y
m
)
y
m
_
_
z
: y
,=
m m0
m
z
(y
m
)
y
m
: y
,=
m, m0
m
z
(y
m
)
y
m
(4.26)
where we used the identity y
m
= y
m
,
= k
11
(z, ) +k
12
(z, ) (4.27)
where
k
11
(z, ) =
: =
u,uu
=,u
,=
m, m0
m
z
(y
m
)
y
m
u: uu
=,u
,=
m m0
m
z
(y
m+1
u)
y
m+1
u
=
u: uu
=,u
,=
m, m0
m+1
m+1
z
(y
m+1
u)
y
m+1
u
where we use that = 1,
=
(,m, m)S
11
m
z
(y
m
)
y
m
(4.28)
with
o
11
= (, m, m):
= y where
= u
for some u
,= , m, m 1,
while
k
12
(z, ) =
: v=
m m0
m
z
(y
m
)
y
m
(,m, m)S
12
m
z
(y
m
)
y
m
(4.29)
where
o
22
= (, m, m):
= y with = for some , m, m 0.

To analyze k
2
(z, ) we rst compute
W
y
,
(0, )k
per
(z, )z
=
_
_
m0
y
m
_
_
k
per
(z, )z
=
_
_
m0
y
m
_
_
=,
,=
(
1
)
=
_
_
m0
m+1
y
m
_
_
_
_

:
=,
,=
_
_
and hence
k
2
(z, ) =
_
_
m0
m+1
y
m
_
_
_
_

:
=,
,=
_
_
=
_
_
m0
m+1
y
m
_
_
_
_

:
_
_
=
_
_
m0
m+1
y
m+1
_
_
_
_

:
_
_
=
(,m, m)S
2
m
z
(y
m
)
y
m
(4.30)
where
o
2
= (, m, m):
= with
and
,= , m 1, m = 0.
Similarly, to analyze k
3
(z, ) we rst note that
W
y
,
(z, 0)z
=
_
_
m0
m
z
(y
m
)
_
_
z
m0
m
z
(y
m
)
,
and hence
k
3
(z, ) =
k
per
(z, )
_
W
y
,
(z, 0)z
_
=
_
_

u,uu
=,u
,=
u
(z
1
)
u
_
_
_
_
m0
m
z
(y
( m
)
_
_
= k
31
(z, ) +k
32
(z, ) (4.31)
where
k
31
(z, ) =
u,uu
=,u
,=
u
(z
1
)
u
u,uu
=,u
,=
u
z
u
(,m, m)S
31
m
z
(y
m
)
y
m
(4.32)
with
o
31
= (, m, m): m = m = 0,
= y with
= u
and u
,= ,
while
k
32
(z, ) =
_
_

u,uu
=,u
,=
u
(z
1
)
u
_
_
_
_
m1
m
z
(y
( m
)
_
_
=
_
_

u,uu
=,u
,=
u
(z
1
)
u
_
_
z
_
_
m0
m+1
z
(y
m
)
_
_
=
_
_

u,uu
=,u
,=
u
z
u
_
_
m+1
z
(y
m
_
=
_
_

u,uu
=,u
,=
u
z
u
_
_
_
_
m0
m+1
z
(y
m+1
)
_
_
=
(,m, m)S
32
m
z
(y
m
)
y
m
(4.33)
where
o
32
= (, m, m): m = 0, m 1,
= y with
= u
for some u
,= .
We next analyze k
4
(z, ). First note that
W
y
,
(0, z
1
) W
y
,
,
=
m1
m
(z
1
)
y
m
.
As y =
with
,= by assumption, we see that (z

1
)
y
m
z
= 0 for all m 0
k
4
(z, ) =
k
per
(z, )
_
_
_
_
m1
m
(z
1
)
y
m
_
_
z
_
_
=
k
per
(z, ) [0] = 0. (4.34)
Combining (4.27), (4.28), (4.29), (4.30), (4.31), (4.32), (4.33) and (4.34) we see
that
[L
[]
T
W
y
,
L
T
]
(z, ) =
(,m, m)S
11
S
12
S
2
S
31
S
32
m
z
(y
m
)
y
m
. (4.35)
On the other hand, by denition
W
y,
(z, ) =
(,m, m)S
m
z
(y
m
)
y
m
(4.36)
where
o = (, m, m): m 0, m 0,
= y for some
,= .
Now it is a simple matter to check that o
11
, o
12
, o
2
, o
31
and o
32
forms a parti-
tioning of o:
o = o
11
o
12
o
2
o
31
o
32
with o
11
, o
12
, o
2
, o
31
, o
32
pairwise disjoint.
This combined with (4.35) and (4.36) immediately gives (4.22) as wanted.
Proof of (4.23). To verify (4.23) we shall use the character formula for the dual
of nite cyclic group: given T,
1
n
:
n
=
r
=
_
0, r ,= r
1, r = r
(4.37)
for r, r = 0, 1, . . . , n 1. Assuming now that y = u
n
in T
d
, we compute
1
n
:
n
=
W
u,
(z, ) =
1
n
:
n
=1
=u,
,=
k,
k0
k
z
(u
k
)
u
k
.
Write k = mn+r and
k = mn+ r, where r, r = 0, 1, . . . , n1 and m, m = 0, 1, 2, . . . .

Then, continuing the computation above gives
1
n
:
n
=
W
u,
(z, )
=
1
n
:
n
=
=u,
,=
n1
r, r=0
m, m0
_
nm+r
n m+ r
z
(u
n m+ r
)
u
nm+r
_
=
=u,
,=
n1
r, r=0
m, m0
_
_
_
_
1
n
:
n
=
r
_
_

m
m
z
(y
m
u
r
)
y
m
u
r
_
_
=
=u,
,=
n1
r=0
m, m0
m
z
(y
m
u
r
)
y
m
u
r
where we used the character formula (4.37)

=
=y,
,=
m, m0
m
z
(y
m
)
y
m
where we made the change of variable = u

r
W
y,
(z, )
and (4.23) follows.
Proof of (4.24). This time we use the orthogonality relations for the circle group:
_
T
m
dm() =
_
0, m ,= m
1, m = m
for m, m = 0, 1, 2, . . . . (4.38)
We then compute
_
T
W
u,
(z, ) dm() =
_
T
_
_

:
=u,
,=
m, m0
m
z
(u
m
)
u
m
_
_
dm()
=
=u,
,=
_
_

m, m0
_
T
m
dm()
_
_
z
(u
m
)
u
m
=u,
,=
m0
z
(u
m
)
u
m
where we used (4.38)

=
w: w=u
m
,m0,
=u,
,=
z
w
w
=

W
x
(z, )
and (4.24) follows as claimed. This completes the proof of all parts of Proposi-
tion 4.4.
The following two theorems collect some results on the unitary classication
of the operators |
x,y,
and |
x
from [9]. We present alternate proofs from those in
[9] as an application of our model theory and symbol calculus.
Theorem 4.5. For x, y T
d
and T, let |
x,y,
be the associated row-unitary.
Without loss of generality (see Proposition 4.1) we assume in addition that x and
y share no common tail. Then:
1. |
x,y,
is unitarily equivalent to |
y,
.
2. If y
is a cyclic permutation of y (so y =
and y
for some ,
T
d
,
then |
y,
and |
y
,
are unitarily equivalent.
3. If y has the form y = u
n
for some u T
d
, then |
y,
is unitarily equivalent
to the direct sum row-unitary
:
n
=
|
u,
.
For the basic denitions and constructions concerning direct integral spaces
which come up in the statement of the next theorem, we refer to [11].
Theorem 4.6. For x = g
k
1
g
k
2
g
k
n
an innite word, let | = |
x
be the asso-
ciated row-unitary. Then:
1. If x
= g
k
1
g
k
2
g
k
n
is another innite word which is tail equivalent to x
in the sense that there is an innite word v so that
x = v, x
v
for some nite words ,
T
d
(not necessarily of the same length), then
|
x
is unitarily equivalent to |
x
.
2. If x is a periodic innite word, say x = uuu =: u
for some non-empty

u T
d
, then |
x
is unitarily equivalent to the direct integral
_
T
|
u,
dm(),
where m is normalized Lebesgue measure on the unit circle T, and |
u,
is
the row-unitary for the orbit-periodic case as in Theorem 4.5.
Remark 4.7. Additional results in [9] are that |
y,
is irreducible if y is primitive
(i.e., not the power of a word of shorter length), and that |
x
is irreducible if x
does not have a periodic tail. We have nothing new to say about these results, as
our calculus techniques here are tailored to exhibiting the reducibility of a given
model rather than proving that a given model is irreducible.
Proof of Theorem 4.5. As we have seen in Proposition 4.1, it suces to study
the models |
W
x,y, on the model space L
W
x,y,. For the proof of the rst two
statements, we combine the third statement of Proposition 3.3 with the formulas
(4.20) and (4.22), respectively, in Proposition 4.4. To see that L
T
is unitary rather
than merely isometric (where T(z) = z
x
and T(z) = z
respectively), it suces
to note that, for any scalar Cuntz weight W (i.e., W = [W
v,w;,
] with W
v,w;,

C for v, w, , T
d
) any vector of the form W[z
v
w
] L
W
is -cyclic for |
W
.
The validity of the third statement is suggested by the identity (4.23) in
Proposition 4.4, but there is still more work to do. The operator-theoretic content
of (4.23) is as follows: the map given by
:
:
n
=
f
(z, )
:
n
=
f
(z, ) (4.39)
is a coisometry from
:
n
=
L1
n
W
u, onto L
W
y, with initial space T equal to
the closure of the linear manifold
T
0
:=
_
_
_
:
n
=
1
n
W
u,
[p] : p T(T
d
T
d
, C)
_
_
_
(4.40)
in
:
n
=
L1
n
W
u,. To see this, note that
:
:
n
=
1
n
W
u,
[p]
:
n
=
1
n
W
u,
[p] = W
y,
[p]
with preservation of norm
_
_
_
_
_
_
:
n
=
1
n
W
u,
[p]
_
_
_
_
_
_
2
:
n
=
1
1
n
W
u,
=
:
n
=
_
1
n
W
u,
[p], p
_
L
2
=
W
y,
[p], p
_
L
2
(by (4.23))
=
_
_
W
y,
[p]
_
_
2
1
W
y,
.
This calculation shows that, indeed, is isometric from T
0
into L
W
y, with image
containing the dense subset W
y,
[p] : p T(T
d
T
d
, C) of L
W
y,. Therefore, by
approximation, the same formula (4.39) denes an isometry (still denoted ) from
T onto L
W
y,. Finally, if f =
:
n
=
f
is an element of
:
n
=
L1
n
W
u,
which is orthogonal to T
0
, then we have, for all p T(T
d
T
d
, c),
0 =
:
n
=
_
f
,
1
n
W
u,
[p]
_
1
W
u,
=
:
n
=
f
, p)
L
2
=
_

:
n
=
f
, p
_
L
2
for all polynomials p T(T
d
T
d
, C). This in turn forces
:
n
=
f
to be the
zero power series, or f =
:
n
=
f
is in the kernel of . In this way we see

that indeed denes a coisometry from
:
n
=
L1
n
W
u, onto L
W
y, with initial
space equal to T dened as the closure of T
0
in (4.40) whenever (4.24) is satised.
It is clear from the denition of and the fact that the formulas for |
W,j
and
|
W,j
is independent of W (see (2.27) and (2.28)) that we have the intertwining
relations
_
_
n
=
|1
n
W
u,
_
_
= |
W
y,
,j
_
_
n
=
|
1
n
W
u,
_
_
= |
W
y,
,j
.
Hence, if it is the case that is actually unitary rather than merely coisometric,
it will follow that
:
n
=
|1
n
W
u, and |
W
y, are unitarily equivalent (via ),
and hence also that
:
n
=
|
u,
and |
y,
are unitarily equivalent. (Note that
the spaces L1
n
W
u, and L
W
u, are the same after a trivial rescaling of the norm,
and hence |1
n
W
u, and |
W
u, are trivially unitarily equivalent.) This is the extra
ingredient required to complete the proof of Theorem 4.5.
To this point the analysis has been relatively straightforward. It now remains
only to show that the map in (4.39) is an isometry, i.e., that certain overlapping
spaces are trivial; as often happens in this model-theoretic approach, this requires
some additional work which we now show is feasible in this case. For each xed
with
n
= , let us identify L1
n
W
u, as a subspace of
n
=
L1
n
W
u, in the natural
way: f
:
n
=
,
f. To show that is an isometry, it suces to show that is
an isometry on each L1
n
W
u, ; indeed, in this case the initial space T for contains
each of the subspaces L1
n
W
, and hence, since T is linear, T then contains the
sum of the subspaces L1
n
W
u, over all with
n
= which is the whole space
:
n
=
L
W
u,. To show that is isometric on each L
W
u, , we need the following
lemma:
Lemma 4.8. Assume that y T
d
has the form y = u
n
for some u T
d
as above.
For T with
n
= and for , T
d
, the polynomial
p
,,
(z, ) =
n1
i=0
i
z
u
i
(4.41)
satises
W
u,
[z
] = W
y,
[p
,,
(z, )]. (4.42)
Assuming the validity of the lemma, we prove that is an isometry on each
L
W
u, as follows. For an arbitrary pair of monomials z
and z
, we have on
the one hand
W
u,
[z
], W
u,
[z
])
1
1
n
W
u,
= n
2
_
1
n
W
u,
[z
],
1
n
W
u,
[z
]
_
1
1
n
W
u,
= n
2
_
1
n
W
u,
[z
], z
_
L
2
= nW
u,
;,
. (4.43)
On the other hand we have
W
u,
[z
], W
u,
[z
])
1
W
y,
= W
u,
[z
], W
y,
[p
,
])
1
W
y,
= W
u,
[z
], p
,
)
L
2
= z
, W
u,
[p
,
])
L
2
=
_
z
v,wJ
d
n1
i=0
W
u,
v,w;
,u
i
i
z
v
w
_
L
2
=
n1
i=0
z
, W
u,
,;
,u
i
i
z
)
L
2
=
n1
i=0
i
W
u,
,u
i
;,
=
_
n1
i=0

i
W
u,
(
1
)
,u
i
, [
[ [[,
n1
i=0

i
W
u,
,u
i
(
1
)
, [
[ [[
(by (2.29))
=
n1
i=0
i
W
u,
;,
= nW
u,
;,
. (4.44)
By combining (4.43) and (4.44), we see that is isometric on each L1
n
W
u, (for
T with
n
= ), and the proof of Theorem 4.5 is complete, once we complete
the proof of Lemma 4.8.
Proof of Lemma 4.8. For p
,,
(z, ) =:
,J
d
p
,,
,
z
given by (4.41), we
need to verify (4.42). In terms of coecients, (4.42) can be expressed as
W
u,
v,w;,
=
,
W
y,
v,w;,
p
,,
,
. (4.45)
Assume for the moment that [v[ [[. We start with the right-hand side of (4.45)
and compute
,
W
y,
v,w;,
p
,,
,
=
n1
i=0
W
y,
v,w;,u
i
i
(by the denition of p
,,
)
=
1
n
:
n
=
n1
i=0
W
u,
v,w;,u
i
i
(by (4.23))
=
1
n
:
n
=
n1
i=0
W
u,
(v
1
)
(u
i
)
,w
i
(by (2.29))
=
1
n
:
n
=
n1
i=0
W
u,
(v
1
)
,w
i
=
:
n
=
W
u,
v,w;,
_
1
n
n1
i=0
i
_
=
:
n
=
W
u,
v,w;,
(
,
)
= W
u,
v,w;,
and (4.45) follows. This completes the proof of Lemma 4.8.
We now tackle the proof of Theorem 4.6.
Proof of Theorem 4.6. For the proof of the rst statement in Theorem 4.6, use the
third statement of Proposition 3.3 combined with formula (4.21) in Proposition
4.4. Again we use that W(z
v
w
) is -cyclic for |
W
for any scalar Cuntz weight W.
The validity of the second statement in Theorem 4.6 is suggested by the
formula (4.24) in Proposition 4.4, but, as was the case for the analogous statement
in Theorem 4.5, there remains some non-trivial work to do. We are given an innite
word x of the form x = u
for a nite word u T

d
. By a continuous analogue
of the argument in the proof of the last part of Theorem 4.5, one sees that the
operator-theoretic content of (4.24) is: the operator :
_
T
L
W
u, dm() L
W
x
given by
: f()
_
T
f() dm()
is a coisometry from the direct integral space
_
T
L
W
u, dm() onto L
W
x with
initial space

T equal to the closure in
_
T
L
W
u, dm() of the linear manifold
T
0
= f : f() = W
u,
[p] : p T(T
d
T
d
, C).
Again, from (2.27) and (2.28), the intertwining conditions
__

T
|
W
u,
,j
_
= |
W
x,
__

T
|
W
u,
,j
_
= (|
W
x)
for j = 1, . . . , d
are clear. If we can show that is actually unitary rather than merely coisomet-
ric (i.e., that is isometric), it will then follow that |
W
x is unitarily equivalent
(via
) with
_
T
|
W
u, dm(), and hence also that |
x
is unitarily equivalent to
_
T
|
u,
dm(), and the proof of Theorem 4.6 will be complete. As in the proof of
Theorem 4.5, proof that the coisometry is actually unitary (i.e., that certain over-
lapping spaces are trivial) is where the extra work enters in in this model-theoretic
approach.
Given a Borel subset B of T, we dene a linear submanifold of
_
T
L
W
u, dm() by
T
B,0
= f : f() =
B
()W
u,
[p] : p T(T
d
T
d
, C). (4.46)
A useful fact concerning the space
_
T
L
W
u, dm() is that the span of the subspaces
T
B,0
over all Borel subsets B of T is dense in
_
T
L
W
u, dm(). To see this, we
show that any f
_
T
L
W
u, dm() orthogonal to T
B,0
for each Borel subset B
of T is zero. Thus, suppose that
f() =
,
f
,
()z
_

T
L
W
u, dm()
is orthogonal to the subspace T
B,0
for B any Borel subset of T. Then, for any
words , T
d
, we have
0 =
_
B
f(), W
u,
[z
])
1
W
u,
=
_
B
f(), z
)
L
2 dm()
=
_
B
f
,
() dm()
for all Borel subsets B of T. Hence it follows that f
,
() = 0 for m-almost every
T for all , , which in turn forces f to be zero in
_
T
L
W
u, dm() as wanted.
Thus, to show that is isometric on
_
T
L
W
u, dm(), it suces to show that
is isometric on the subspace T
B,0
for each Borel subset B T. Note that, for
f() =
B
()W
u,
[p] T
B,0
(with p equal to a polynomial in z and ), we see
that : f W
B
[p] L
W
B where we have set W
B
equal to the positive Haplitz
operator
W
B
v,w;,
=
_
B
W
u,
v,w;,
dm() with symbol

W
B
(z, ) =
_
B
W
u,
(z, ) dm().
(4.47)
Furthermore, via an interchange of integration and summation we see that
|f|
2
_
T
1
u,
dm()
= |f|
2
1
W
B
if f() =
B
()W
u,
[p], p T(T
d
T
d
, C).
Thus, for a given Borel subset B T, to show that is isometric from T
B,0
into
J
x
is the same as to show that the inclusion map f f is isometric from the
dense subset W
B
T(T
d
T
d
, C) of L
W
B into L
W
x. Thus the proof of Theorem 4.6
is complete once we show: for each Borel subset B T and any choice of words
, ,
T
d
,
_
W
B
[z
], W
B
[z
]
_
1
W
B
=
_
W
B
[z
], W
B
[z
]
_
1
W
x
. (4.48)
For this purpose we need the following more elaborate continuous analogue
of Lemma 4.8. In the statement of the lemma, for 0 < r < 1 and T we dene
the Poisson kernel T
r
() by
T
r
() =
k=
r
]k]
k
.
Lemma 4.9. For 0 < r < 1, T and , T
d
, dene the polynomial p
,r
,
(z, )
T(T
d
T
d
, C) by
p
,r
,
(z, ) =
_
k=0
r
k
k
z
u
k
k=1
r
k
k
z
(u
k1
)
if
has the form = u
j
,
= u,
,= , j = 0, 1, 2, . . .
0, otherwise.
(4.49)
Then
W
u,
[p
,r
,
] = T
r
(/)W
u,
[z
]. (4.50)
Moreover, if x is an innite word of the form x = uuu =: u
and we dene
W
B
as in (4.47), then, if we dene the formal power series p
B
,
by
p
B
,
[z, ) = lim
r1
_
B
p
,r
,
(z, ) dm(), (4.51)
then p
B
,
is in the domain of W
x
and
W
B
[z
] = W
x
[p
B
,
]. (4.52)
Assuming the validity of the Lemma, we now complete the proof of Theorem
4.6 as follows. A consequence of (4.24) and (4.50) in Lemma 4.9 is
W
x
p
,r
,
, p
,r
)
L
2dm() =
_
T
W
u,
[p
,r
,
], p
,r
)
L
2 dm()
=
_
T
T
r
(/)W
u,
[z
], p
,r
)
L
2 dm()
=
_
T
T
r
(/)z
, W
u,
[p
,r
])
L
2 dm()
=
_
T
T
r
(/)z
, T
r
(
/)W
u,
[z
])
L
2 dm()
lim
r1
W
x
[p
,r
,
], p
,r
)
L
2 = T
r
(
/)z
, W
u,
[z
])
L
2 . (4.53)
Integration of both sides of (4.53) with respect to over the Borel set B then
gives
W
x
[p
B
,
], p
,r
)
L
2 =
_
B
T
r
(
/)z
, W
u,
[z
])
L
2 dm() (4.54)
where p
B
,
is as in (4.51).
With these preliminaries out of the way, we start with the right-hand side of
(4.48) and compute
_
W
B
[z
], W
B
[z
]
_
1
W
x
=
W
x
[p
B
,
], W
x
[p
B
,
]
_
1
W
x
(by (4.52))
=
W
x
[p
B
,
], p
B
_
L
2
= lim
r
1
_
B
_
W
x
[p
B
,
], p
,r
_
L
2
dm(
) (by denition of p
B
,
)
= lim
r
1
_
B
__
B
T
r
(
/)
_
z
, W
u,
[z
]
_
L
2
dm()
_
dm(
) (by (4.54))
=
_
B
_
z
, W
u,
[z
]
_
L
2
dm(
) (since T is an approximate identity)

=
_
z
, W
B
[z
]
_
L
2
=
_
W
B
[z
], W
B
[z
]
_
1
W
B
and (4.48) follows as wanted. The only remaining piece of the proof of Theorem
4.6 is now the proof of Lemma 4.9.
Proof of Lemma 4.9. We rst check the general identities
W
u,
v,w;,u
k
=
k
W
u,
v,w;,
(4.55)
W
u,
v,w;
(u
k1
)
,u
j
=
k
W
u,
v,w;,u
j
where
= u with
,= . (4.56)
To check (4.55), in case [v[ [[, by (2.29) we have
W
u,
v,w;,u
k
=

W
u,
(v
1
)
(u
k
)
,w
=
k
W
u,
(v
1
)
,w
=
k
W
u,
v,w;,
,
and similarly, if [v[ < [[, we have
W
u,
v,w;,u
k
=

W
u,
(u
k
)
,w(v
1
)
=
k
W
u,
,w(v
1
)
=
k
W
u,
v,w;,
and (4.55) follows in all cases.
To check (4.56), one must consider three special cases:
Case 1: [v[ k[u[ +[[,
Case 2: [[ [v[ < k[u[ +[[, and
Case 3: [v[ < [[.
Each of these cases is handled by using (2.29) to reduce the issue to coecients
of the symbol

W
,
and then using the special structure of

W
u,
,
apparent from
the denition (4.16). As the computations are somewhat tedious, we will not go
through all the details here, but simply accept the validity of (4.56).
We next verify that p
B
,
x
for each , T
d
and
Borel subset B T. It is not dicult to see that the non-zero coecients of P
B
,
,
although innite in number, are square summable in modulus, since they roughly
correspond to the Fourier coecients for the L
2
-function
B
on the circle T. Since
W
x
is supported on the symmetrized diagonal (v
, v): v T
d
with non-zero
diagonal entries all equal to 1, the square summability of the coecients of p
B
,
implies that p
B
,
x
as claimed.
We are now ready to verify (4.50). Consider rst the case where is not of the
form = u
j
for some j 0, 1, 2, . . . and with
= u for some
,= . Then
p
,r
,
= 0 by denition while, by (2.29) and the denition of

W
u,
(z, ) we have
W
u,
v,w;,
=

W
u,v
(v
1
)
,w
= 0 in case [v[ [[, and
W
u,
v,w;,
=

W
u,
,w(v
1
)
= 0 in case [v[ < [[.

We conclude that (4.50) holds (in the form 0 = 0) for the case where is not of
the form u
j
for some j 0, 1, 2, . . . and with
= u for some
,= .
We next consider the case where does have the special form = u
j
(where
j 0, 1, 2, . . . and
= u for some
,= ). We compute
W
u,
[p
,r
,
] = W
u,
_

k=0
r
k
k
z
u
k
k=1
r
k
k
z
(u
k1
)
_
=
k=0
r
k
k
W
u,
[z
u
k
] +
k=1
r
k
k
W
u,
[z
(u
k1
)
]
=
k=0
r
k
k
W
u,
[z
] +
k=1
r
k
k
W
u,
[z
] (by (4.55) and (4.56))

= T
r
(/)W
u,
[z
]
and (4.50) follows.
To verify (4.52), we observe
W
B
[z
]
_
B
W
x
[p
,r
,
] dm()=W
B
[z
]
_
B
__
T
W
u,
[p
,r
,
] dm()
_
dm()
=
_
B
W
u,
[z
] dm()
_
B
__
T
T
r
(/)W
u,
[z
] dm()
_
dm() (by (4.50))
=
_
B
_
W
u,
[z
]
_
T
T
r
(/)W
u,
[z
] dm()
_
dm()
where the last expression tends to zero as r 1 (coecientwise) since the Poisson
kernel T
r
() is an approximate identity. This completes the proof of (4.52) and of
Lemma 4.9.
References
[1] W. Arveson, Subalgebras of C
-algebras III: multivariable operator theory, Acta

Math. 181 (1998), 159228.
[2] J.A. Ball, Linear systems, operator model theory and scattering: multivariable gen-
eralizations, in Operator Theory and Its Applications (Winnipeg, MB, 1998) (Ed.
A.G. Ramm, P.N. Shivakumar and A.V. Strauss), Fields Institute Communications
Vol. 25, Amer. Math. Soc., Providence, 2000, pp. 151178.
[3] J.A. Ball and V. Vinnikov, Lax-Phillips scattering and conservative linear systems:
a Cuntz-algebra multidimensional setting, Memoirs of the American Mathematical
Society, to appear.
[4] J.A. Ball and V. Vinnikov, Formal reproducing kernel Hilbert spaces: the commuta-
tive and noncommutative settings, in Reproducing Kernel Spaces and Applications
(Ed. D. Alpay), pp. 77134, OT143, Birkhauser-Verlag, Basel-Boston, 2003.
[5] S.S. Boiko, V.K. Dubovoy and A.Ja. Kheifets, Measure Schur complements and
spectral functions of unitary operators with respect to dierent scales, in Operator
Theory, System Theory and Related Topics: The Moshe Livsic Anniversary Volume,
pp. 89138, OT 123, Birkhauser (Basel-Boston), 2000.
[6] L. de Branges and J. Rovnyak, Canonical models in quantum scattering theory, in
Perturbation Theory and its Applications in Quantum Mechanics (Ed. C.H. Wilcox),
pp. 295392, Wilcox, New York-London-Sidney, 1966.
[7] O. Bratteli and P.E.T. Jorgensen, Iterated function systems and permutation rep-
resentations of the Cuntz algebra, Memoirs of the American Mathematical Society
Volume 139, Number 663 (second of 5 numbers), 1999.
[8] K.R. Davidson, C
-Algebras by Example, Fields Institute Monograph 6, American

Mathematical Society, Providence, 1996.
[9] K.R. Davidson and D.R. Pitts, Invariant subspaces and hyper-reexivity for free
semigroup algebras, Proc. London Math. Soc. 78 (1999), 401430.
[10] K.R. Davidson and D.R. Pitts, The algebraic structure of non-commutative analytic
Toeplitz algebras, Math. Ann. 311 (1998), 275303.
[11] J. Dixmier, Les algèbres doperateurs dans lespace Hilbertien (Algèbres de von Neu-
mann), Gauthier-Villars, Paris, 1969.
[12] G. Popescu, Models for innite sequences of noncommuting operators, Acta Sci.
Math. 53 (1989), 355368.
[13] G. Popescu, Characteristic functions for innite sequences of noncommuting opera-
tors, J. Operator Theory 22 (1989), 5171.
[14] G. Popescu, Isometric dilations for innite sequences of noncommuting operators,
Trans. Amer. Math. Soc. 316 (1989), 523536.
[15] G. Popescu, Multi-analytic operators on Fock spaces, Math. Ann. 303 (1995), 3146.
Joseph A. Ball
Virginia Tech
Blacksburg, Virginia 24061
e-mail: ball@math.vt.edu
Victor Vinnikov
Ben Gurion University of the Negev
Beer-Sheva 84105, Israel
e-mail: vinnikov@math.bgu.ac.il
Operator Theory:
Relations on Non-commutative Variables
and Associated Orthogonal Polynomials
T. Banks, T. Constantinescu and J.L. Johnson
Abstract. This semi-expository paper surveys results concerning three classes
of orthogonal polynomials: in one non-hermitian variable, in several isometric
non-commuting variables, and in several hermitian non-commuting variables.
The emphasis is on some dilation theoretic techniques that are also described
in some details.
Mathematics Subject Classication (2000). Primary 47A20; Secondary 42C05.
Keywords. Orthogonal polynomials in several variables, recurrence relations,
asymptotic properties.
1. Introduction
In this semi-expository paper we deal with a few classes of orthogonal polynomials
associated to polynomial relations on several non-commuting variables. Our initial
interest in this subject was motivated by the need of more examples related to [18].
On the other hand there are transparent connections with interpolation problems
in several variables as well as with the modeling of various classes of non-stationary
systems (see [1] for a list of recent references), which guided our choice of topics.
Thus we do not relate to more traditional studies on orthogonal polynomials of
several variables associated to a nite reection group on an Euclidean space or
other types of special functions of several variables, for which a recent presentation
could be found in [21], instead we are more focused on results connected with
various dilation theoretic aspects and Szeg o kernels.
Our aim is to give an introduction to this point of view. We begin our presen-
tation with a familiar setting for algebras given by polynomial dening relations
and then we introduce families of orthonormal polynomials associated to some
positive functionals on these algebras. Section 3 contains a discussion of the rst
class of orthogonal polynomials considered in this paper, namely polynomials in
one variable on which there is no relation. This choice is motivated mainly by
62 T. Banks, T. Constantinescu and J.L. Johnson
the fact that we have an opportunity to introduce some of the basic dilation the-
oretic techniques that we are using. First, we discuss (in a particular case that
is sucient for our goals) the structure of positive denite kernels and their tri-
angular factorization. Then these results are used to obtain recurrence relations
for orthogonal polynomials in one variable with no additional relations, as well as
asymptotic properties of these polynomials. All of these extend well-known results
of G. Szeg o. We conclude this section with the introduction of a Szeg o type kernel
which appears to be relevant to our setting.
In Section 4 we discuss the example of orthogonal polynomials of several
isometric variables. Most of the results are just particular cases of the correspond-
ing results discussed in Section 3, but there is an interesting new point about
the Szego kernel that appears in the proof of Theorem 4.1. We also use a certain
explicit structure of the Kolmogorov decomposition of a positive denite kernel
on the set of non-negative integers in order to produce examples of families of
operators satisfying Cuntz-Toeplitz and Cuntz relations.
The nal section contains a discussion of orthogonal polynomials of several
non-commuting hermitian variables. This time, some of the techniques described
in Section 3 are not so relevant and instead we obtain three-terms recursions in
the traditional way, and we introduce families of Jacobi matrices associated to
these recursions. Many of these results can be proved by adapting the classical
proofs from the one scalar variable case. However, much of the classical function
theory is no longer available so we present some proofs illustrating how classical
techniques have to be changed or replaced. Also some results are not presented in
the most general form in the hope that the consequent simplications in notation
would make the paper more readable.
2. Orthogonal polynomials associated to polynomial relations
In this section we introduce some classes of orthogonal polynomials in several vari-
ables. We begin with the algebra T
N
of polynomials in N non-commuting variables
X
1
, . . . , X
N
with complex coecients. Let F
+
N
be the unital free semigroup on N
generators 1, . . . , N. The empty word is the identity element and the length of the
word is denoted by [[. It is convenient to use the notation X
= X
i
1
. . . X
i
k
for = i
1
. . . i
k
F
+
N
. Thus, each element P T
N
can be uniquely written in the
form P =
F
+
N
c
, with c
,= 0 for nitely many s.

We notice that T
N
is isomorphic with the tensor algebra over C
N
. Let (C
N
)
k
denote the k-fold tensor product of C
N
with itself. The tensor algebra over C
N
is
dened by the algebraic direct sum
T (C
N
) =
k0
(C
N
)
k
.
If e
1
, . . . , e
N
is the standard basis of C
N
, then the set
1 e
i
1
e
i
k
[ 1 i
1
, . . . , i
k
N, k 1
Non-commutative Variables and Orthogonal Polynomials 63
is a basis of T (C
N
). For = i
1
. . . i
k
we write e
instead of e
i
1
e
i
k
, and
the mapping X
, F
+
N
, extends to an isomorphism from T
N
to T (C
N
),
hence T
N
T (C
N
).
It is useful to introduce a natural involution on T
2N
as follows:
X
+
k
= X
N+k
, k = 1, . . . , N,
X
+
l
= X
lN
, l = N + 1, . . . , 2N;
on monomials,
(X
i
1
. . . X
i
k
)
+
= X
+
i
k
. . . X
+
i
1
,
and nally, if Q =
F
+
2N
c
, then Q
+
=
F
+
2N
c
X
+
. Thus, T
2N
is a unital,
associative, -algebra over C.
We say that / T
2N
is symmetric if P / implies cP
+
/ for some
c C 0. Then the quotient of T
2N
by the two-sided ideal generated by / is
an associative algebra 1(/). Letting =
,
: T
2N
1(/) denote the quotient
map then the formula
(P)
+
= (P
+
) (2.1)
gives a well-dened involution on 1(/). Usually the elements of / are called the
dening relations of the algebra 1(/). For instance, 1() = T
2N
,
1(X
k
X
+
k
[ k = 1, . . . , N) = T
N
;
also,
1(X
k
X
l
X
l
X
k
[ k, l = 1, . . . , 2N)
is the symmetric algebra over C
2N
and
1(X
k
X
l
+X
l
X
k
[ k, l = 1, . . . , 2N)
is the exterior algebra over C
2N
. Examples abound in the literature (for instance,
see [20], [21], [29]).
There are many well-known diculties in the study of orthogonal polynomials
in several variables. The rst one concerns the choice of an ordering of F
+
N
. In this
paper we consider only the lexicographic order , but due to the canonical grading
of F
+
N
it is possible to develop a basis free approach to orthogonal polynomials. In
the case of orthogonal polynomials on several commuting variables this is presented
in [21]. A second diculty concerns the choice of the moments. In this paper
we adopt the following terminology. A linear functional on 1(/) is called q-
positive (q comes from quarter) if ((P)
+
(P)) 0 for all P T
N
. In this case,
((P)
+
) = ((P)) for P T
N
and
[((P
1
)
+
(P
2
))[
2
((P
1
)
+
(P
1
))((P
2
)
+
(P
2
))
for P
1
, P
2
T
N
. Next we introduce
(P
1
), (P
2
))
= ((P
2
)
+
(P
1
)), P
1
, P
2
T
N
. (2.2)
By factoring out the subspace ^
= (P) [ P T
N
, (P), (P))
= 0 and
completing this quotient with respect to the norm induced by (2.2) we obtain a
Hilbert space H
.
The index set G(/) F
+
N
of / is chosen as follows: if G(/), choose the
next element in G(/) to be the least F
+
N
with the property that the elements
(X
),
_ , and (X
) are linearly independent. We will avoid the degenerate

situation in which (1) = 0; if we do so, then G(/). Dene F
= (X
) for
G(/). For instance, G() = F
+
N
, in which case F
= X
, F
+
N
. Also,
G(X
k
X
l
X
l
X
k
[ k, l = 1, . . . , 2N) = i
1
. . . i
k
F
+
N
[ i
1
i
k
, k 0,
and
G(X
k
X
l
+X
l
X
k
[ k, l = 1, . . . , 2N) = i
1
. . . i
k
F
+
N
[ i
1
< < i
k
, 0 k N
(we use the convention that for k = 0, i
1
. . . i
k
is the empty word).
We consider the moments of to be the numbers
s
,
= (F
+
) = F
, F
, , G(/). (2.3)
The kernel of moments is given by K
(, ) = s
,
, , G(/). We notice that
is a q-positive functional on 1(/) if and only if K
is a positive denite kernel

on G(/). However, K
does not determine uniquely. One typical situation when

K
determines is X
k
X
+
k
[ k = 1, . . . , N /; a more general example is
provided by the Wick polynomials,
X
k
X
+
l
a
k,l
k,l

N
m,n=1
c
m,n
k,l
X
+
m
X
n
, k, l = 1, . . . , N,
where a
k,l
, c
m,n
k,l
are complex numbers and
k,l
is the Kronecker symbol.
The moment problem is trivial in this framework since it is obvious that the
numbers s
,
, , G(/), are the moments of a q-positive functional on 1(/) if
and only if the kernel K(, ) = s
,
, , G(/), is positive denite.
We now introduce orthogonal polynomials in 1(/). Assume that is strictly
q-positive on 1(/), that is, ((P)
+
(P)) > 0 for (P) ,= 0. In this case ^
= 0
and (T
N
) can be viewed as a subspace of H
. Also, F
G(,)
is a linearly inde-
pendent family in H
and the Gram-Schmidt procedure gives a family
G(,)
of elements in (T
N
) such that
_
a
,
F
, a
,
> 0; (2.4)
=
,
, , G(/). (2.5)
The elements
, G(/), will be called the orthonormal polynomials associated

to . An explicit formula for the orthonormal polynomials can be obtained in the
same manner as in the classical, one scalar variable case. Thus, set
D
= det [s
,
]
_
> 0, G(/), (2.6)
and from now on 1 denotes the predecessor of with respect to the lexicographic
order on F
+
N
, while + 1 denotes the successor of .
We have:
= s
1/2
,
and for ,
=
1
_
D
1
D
det
_
_
[s
,
]
_
F
. . . F
_
_
, (2.7)
with an obvious interpretation of the determinant. In the following sections we
will discuss in more details orthonormal polynomials associated to some simple
dening relations.
3. No relation in one variable
This simple case allows us to illustrate some general techniques that can be used in
the study of orthonormal polynomials. We have / = and N = 1, so 1(/) = T
2
.
The index set is N
0
, the set of non-negative integers, and F
n
= X
n
1
, n N
0
.
The moment kernel of a q-positive functional on T
2
is K
(n, m) = ((X
n
1
)
+
X
m
1
),
n, m N
0
, and we notice that there is no restriction on K
other than being

positive denite. We now discuss some tools that can be used in this situation.
3.1. Positive denite kernels on N
0
We discuss a certain structure (and parametrization) of positive denite kernels on
N
0
. The nature of this structure is revealed by looking at the simplest examples.
First, we consider a strictly positive matrix
S =
_
1 a
a 1
_
, a R.
This matrix gives a new inner product on R
2
by the formula
x, y)
S
= Sx, y), x, y R
2
,
where , ) denotes the Euclidean inner product on R
2
. Let e
1
, e
2
be the standard
basis of R
2
. By renorming R
2
with , )
S
the angle between e
1
and e
2
was modied
to the new angle = (e
1
, e
2
) such that
cos (e
1
, e
2
) =
e
1
, e
2
)
S
|e
1
|
S
|e
2
|
S
= a. (3.1)
We can visualize the renormalization process by giving a map T
S
: R
2
R
2
with
the property that T
S
x, T
S
y) = x, y)
S
for x, y R
2
, and it is easily seen that we
can choose
T
S
=
_
1 cos
0 sin
_
.
We can also notice that T
S
e
1
= e
1
and T
S
e
2
= f
2
= J(cos )e
1
, where
J(cos ) is the Julia operator,
J(cos ) =
_
cos sin
sin cos
_
,
e
1
e
2
e
1
f
2
T
S
`
-
Figure 1. Renormalization in R
2
which is the composition of a reection about the x-axis followed by the counter-
clockwise rotation R
through angle . We deduce that

a = cos = e
1
, f
2
) = e
1
, J(cos )e
1
) = J(cos )e
1
, e
1
).
The discussion extends naturally to the 3 3 case. Thus let
S =
_
_
1 a b
a 1 c
b c 1
_
_
, a, b, c R,
be a strictly positive matrix. A new inner product is induced by S on R
3
,
x, y)
S
= Sx, y), x, y R
3
,
and let e
1
, e
2
, e
3
be the standard basis of R
3
. With respect to this new inner
product the vectors e
1
, e
2
, e
3
still belong to the unit sphere, but they are no longer
orthogonal. Thus,
a = cos (e
1
, e
2
) = cos
12
,
c = cos (e
2
, e
3
) = cos
23
,
and
b = cos (e
1
, e
3
) = cos
13
.
This time, the law of cosines in spherical geometry gives a relation between the
numbers a, b, and c,
b = cos
13
= cos
12
cos
23
+ sin
12
sin
23
cos , (3.2)
where is the dihedral angle formed by the planes generated by e
1
, e
2
and, re-
spectively, e
2
, e
3
(see, for instance, [26]). Thus, the number b belongs to a disk of
center cos
12
cos
23
and radius sin
12
sin
23
. Once again the renormalization can
be visualized by a map T
S
: R
3
R
3
such that T
S
x, T
S
y) = x, y)
S
. In this case
we can choose
T
S
=
_
_
1 cos
12
cos
12
cos
23
+ sin
12
sin
23
cos
0 sin
12
sin
12
cos
23
cos
12
sin
23
cos
0 0 sin
23
sin
_
_
,
and we see that
T
S
e
1
= e
1
,
T
S
e
2
= f
2
= (J(cos
1,2
) 1)e
1
,
and
T
S
e
3
= f
3
= (J(cos
12
) 1)(1 J(cos ))(J(cos
23
) 1)e
1
.
In particular,
b = cos
13
= (J(cos
1,2
) 1)(1 J(cos ))(J(cos
2,3
) 1)e
1
, e
1
), (3.3)
which can be viewed as a dilation formula.
Now both (3.2) and (3.3) extend to a strictly positive n n matrix and
provide a parametrization and therefore a structure for positive denite kernels on
N
0
(for general results and details see [13], [16]). We apply this result to a kernel
K
associated to a strictly q-positive functional and obtain that K
is uniquely
determined by a family
k,j
0k<j
of complex numbers with the property that
[
k,j
[ < 1 for all 0 k < j. Dene d
k,j
= (1 [
k,j
[
2
)
1/2
. The extension of (3.3)
mentioned above gives
s
k,j
= s
1/2
k,k
s
1/2
j,j
U
k,j
e
1
, e
1
), k < j, (3.4)
where U
k,j
is a (j k + 1) (j k + 1) unitary matrix dened recursively by:
U
k,k
= 1 for k 0; for k < j (with n = j k),
U
k,j
= (J(
k,k+1
) 1
n1
)(1 J(
k,k+2
) 1
n2
) . . . (1
n1
J(
k,j
))(U
k+1,j
1),
where J(
l,m
) is the Julia operator associated to
l,m
,
J(
l,m
) =
_

l,m
d
l,m
d
l,m

l,m
_
,
and 1
m
denotes the m m identity matrix. For instance, we deduce from (3.4)
that:
s
01
= s
1/2
00
s
1/2
11
_
1 0
_

01
d
01
d
01

01
_ _
1
0
_
;
s
02
= s
1/2
00
s
1/2
22
_
1 0 0
_
_
01
d
01
0
d
01

01
0
0 0 1
_
_
_
_
1 0 0
0
02
d
02
0 d
02

02
_
_
_
_
12
d
12
0
d
12

12
0
0 0 1
_
_
_
_
1
0
0
_
_
,
and
s
03
= s
1/2
00
s
1/2
33
_
1 0 0 0
01
d
01
0 0
d
01

01
0 0
0 0 1 0
0 0 0 1
_
_
_
_
1 0 0 0
0
02
d
02
0
0 d
02

02
0
0 0 0 1
_
_
_
_
1 0 0 0
0 1 0 0
0 0
03
d
03
0 0 d
03

03
_
12
d
12
0 0
d
12

12
0 0
0 0 1 0
0 0 0 1
_
_
_
_
1 0 0 0
0
13
d
13
0
0 d
13

13
0
0 0 0 1
_
23
d
23
0 0
d
23

23
0 0
0 0 1 0
0 0 0 1
_
_
_
_
1
0
0
0
_
_
.
In particular, we deduce that s
01
= s
1/2
00
s
1/2
11

01
. The next formula is the complex
version of (3.2), s
02
= s
1/2
00
s
1/2
22

01
12
+s
1/2
00
s
1/2
22
d
01
02
d
12
. Then,
s
03
= s
1/2
00
s
1/2
33

01
12
23
+s
1/2
00
s
1/2
33

01
d
12
13
d
23
+s
1/2
00
s
1/2
33
d
01
02
d
12
23
s
1/2
00
s
1/2
33
d
01
02
12
13
d
23
+s
1/2
00
s
1/2
33
d
01
d
02
03
d
13
d
23
.
Explicit formulae of this type can be obtained for each s
k,j
, as well as inverse
algorithms allowing to calculate
k,j
from the kernel of moments, see [16] for
details.
A natural combinatorial question would be to calculate the number N(s
k,j
)
of additive terms in the expression of s
k,j
. We give here some details since the
calculation of N(s
k,j
) involves another useful interpretation of formula (3.4). We
notice that for k 0,
N(s
01
) = N(s
k,k+1
) = 1,
N(s
02
) = N(s
k,k+2
) = 2,
N(s
03
) = N(s
k,k+3
) = 5.
The general formula is given by the following result.
Theorem 3.1. N(s
k,k+l
) is given by the Catalan number C
l
=
1
l + 1
_
2l
l
_
.
Proof. The rst step of the proof considers the realization of s
k,j
through a time
varying transmission line (or lattice). For illustration we consider the case of s
03
in Figure 2.
A B
Figure 2. Lattice representation for s

03
Each box in Figure 2 represents the action of a Julia operator. As a whole, the
diagram represents the action of the unitary operator U
0,3
. We see in this gure
that the number of additive terms in the formula of s
03
is given by the number
of paths from A to B. In it clear that to each path from A to B in Figure 2 it
corresponds a Dyck path from C to D in Figure 3, that is, a path that never steps
below the diagonal and goes only to the right or downward.
e e C
D
e
e e e e
e e e e
e e e e
e
Figure 3. A Dyck path from C to D
Thus, each box in Figure 2 corresponds to a point strictly above the diagonal
in Figure 3. Once this one-to-one correspondence is established, we can use the
well-known fact that the number of Dyck paths like the one in Figure 3 is given
exactly by the Catalan numbers.
3.2. Spectral factorization
The classical theory of orthogonal polynomials is intimately related to the so-
called spectral factorization. Its prototype would be the Fejer-Riesz factorization
of a positive trigonometric polynomial P in the form P = [Q[
2
, where Q is a
polynomial with no zeros in the unit disk. This is generalized by Szeg o to the
Szego class of those measures on the unit circle T with log
L
1
, and very
general results along this line can be found in [31] and [33].
Here we briey review the spectral factorization of positive denite kernels on
the set N
0
described in [14]. For two positive denite kernels K
1
and K
2
we write
K
1
K
2
if K
2
K
1
is a positive denite kernel. Consider a family T = T
n
n0
of at most one-dimensional vector spaces (this restriction is made only for the
purpose of this paper) and call lower triangular array a family =
k,j
k,j0
of complex numbers
k,j
with the following two properties:
k,j
= 0 for k < j
and each column c
j
() = [
k,j
]
k0
, j 0, belongs to the Hilbert space
kj
T
k
.
Denote by H
2
0
(T) the set of all lower triangular arrays as above. An element of
H
2
0
(T) is called outer if the set c
j
() [ j k is total in
jk
T
j
for each k 0.
It is easily seen that if is an outer triangular array, then the formula
K
(k, j) = c
k
()
c
j
()
gives a positive denite kernel on N
0
. The following result extends the above
mentioned Szeg o factorization and at the same time it contains the Cholesky fac-
torization of positive matrices.
Theorem 3.2. Let K be an positive denite kernel on N
0
. Then there exists a family
T = T
n
n0
of at most one-dimensional vector spaces and an outer triangular
array H
2
0
(T), referred to as a spectral factor of the kernel K, such that
(1) K
K.
(2) For any other family T
= T
n0
of at most one-dimensional vector spaces
and any outer triangular array
H
2
0
(c, T
) such that K
K, we have
K
.
(3) is uniquely determined by (a) and (b) up to a left unitary diagonal factor.
It follows from (3) above that the spectral factor can be uniquely deter-
mined by the condition that
n,n
0 for all n 0. We say that the kernel K
belongs to the Szego class if inf
n0

n,n
> 0. If
k,j
are the parameters of K
introduced in Subsection 3.1 then it follows that the kernel K belongs to the Szego
class if and only if
inf
k0
s
1/2
k,k
n>k
d
k,n
> 0. (3.5)
This implies that T
n
= C for all n 0 (for details see [14] or [16]).
3.3. Recurrence relations
Formula (2.7) is not very useful in calculations involving the orthogonal polyno-
mials. Instead there are used recurrence formulae. In our case, / = and N = 1,
we consider the moment kernel K
of a strictly q-positive functional on T

2
and
also, the parameters
k,j
of K
as in Subsection 3.1. It can be shown that the

orthonormal polynomials associated to obey the following recurrence relations
0
(X
1
, l) =
0
(X
1
, l) = s
1/2
l,l
, l N
0
, (3.6)
and for n 1, l N
0
,
n
(X
1
, l) =
1
d
l,n+l
_
X
1
n1
(X
1
, l + 1)
l,n+l
n1
(X
1
, l)
_
, (3.7)
n
(X
1
, l) =
1
d
l,n+l
_
l,n+l
X
1
n1
(X
1
, l + 1) +
n1
(X
1
, l)
_
, (3.8)
where
n
(X
1
) =
n
(X
1
, 0) and
n
(X
1
) =
n
(X
1
, 0).
Somewhat similar polynomials are considered in [19], but the form of the
recurrence relations as above is noticed in [17]. It should be mentioned that
n
(X
1
, l)
n0
is the family of orthonormal polynomials associated to a q-positive
functional on T
2
with moment kernel K
l
(, ) = s
+l,+l
, , N
0
. Also, the
above recurrence relations provide us with a tool to recover the parameters
k,j
from the orthonormal polynomials.

Theorem 3.3. Let k
l
n
be the leading coecient of
n
(X
1
, l). For l N
0
and n 1,
l,n+l
=
n
(0, l)
k
l+1
0
. . . k
l+1
n1
k
l
0
. . . k
l
n
.
Proof. We reproduce here the proof from [8] in order to illustrate these concepts
and to introduce one more property of the parameters
k,j
. First, we deduce
from (3.7) that
n
(0, l) =
l,n+l
d
l,n+l
n1
(0, l),
while formula (3.8) gives
n
(0, l) =
1
d
l,n+l
n1
(0, l) = . . . = s
1/2
l,l
n
p=1
1
d
l,p+l
,
hence
n
(0, l) = s
1/2
l,l

l,n+l
n
p=1
1
d
l,p+l
.
Now we can use another useful feature of the parameters
k,j
, namely the fact
that they give simple formulae for determinants. Let D
m,l
denote the determinant
of the matrix [s
k,j
]
lk,jm
. By Proposition 1.7 in [13],
D
l,m
=
m
k=l
s
k,k

lj<pm
d
2
j,p
. (3.9)
One simple application of this formula is that it reveals the equality behind Fisher-
Hadamard inequality. Thus, for l n n
m, we have
D
l,m
=
D
l,n
D
n,m
D
n
,n
(k,j)
d
2
k,j
,
where = (k, j) [ l k < n n
< j m. Some other applications of (3.9)

can be found in [16], Chapter 8. Returning to our proof we deduce from (3.9) that
n
p=1
d
2
l,p+l
= s
1
l,l
D
l,l+n
D
l+1,l+n
so,
l,n+l
=
n
(0, l)
D
l,l+n
D
l+1,l+n
. (3.10)
We can now relate this formula to the leading coecients k
l
n
. From (3.7) we deduce
that
k
l
n
= s
1/2
l+n,l+n
n1
p=1
1
d
l+p,l+n
, n 1,
and using once again (3.9), we deduce
k
l
n
=
D
l,l+n1
D
l,l+n
, n 1,
which concludes the proof.
3.4. Some examples
We consider some examples, especially in order to clarify the connection with
classical orthogonal polynomials. Thus, consider rst / = 1X
+
1
X
1
. In this case
the index set is still N
0
and if is a linear functional on 1(/), then the kernel of
moments is Toeplitz, K
(n+k, m+k) = K
(n, m), m, n, k N
0
. Let be a strictly
q-positive functional on 1(/) and let
k,j
be the parameters associated to K
.
We deduce that these parameters also satisfy the Toeplitz condition,
n+k,m+k
=
n,m
, n < m, k 1. Setting
n
=
k,n+k
, n 1, k 0, and d
n
= (1[
n
[
2
)
1/2
, the
recurrence relations (3.7), (3.8) collapse to the classical Szego recursions obeyed
by the orthogonal polynomials on the unit circle,
n+1
(z) =
1
d
n+1
(z
n
(z)
n+1
n
(z)),
and
n+1
(z) =
1
d
n+1
(
n+1
z
n
(z) +
n
(z)).
Therefore
n
, n 1 are the usual Szeg o coecients, [32].
Another example is given by / = X
1
X
+
1
. In this case the index set is still
N
0
and the moment kernel of a strictly q-positive functional on 1(/) will have the
Hankel property, K
(n, m+k) = K
(n +k, m), m, n, k N
0
. Orthogonal polyno-
mials associated to functionals on 1(/) correspond to orthogonal polynomials on
the real line. This time, the parameters
k,j
associated to moment kernels have
no classical analogue. Instead there are so-called canonical moments which are
used as a counterpart of the Szeg o coecients (see [27]). Also, recurrence relations
of type (3.7), (3.8) are replaced by a three term recurrence equation,
x
n
(x) = b
n
n+1
(x) +a
n
n
(x) +b
n1
n1
(x), (3.11)
with initial conditions
1
= 0,
0
= 1 ([32]). Denitely, these objects are more
useful (for instance, it appears that no simple characterization of those
k,j
corre-
sponding to Hankel kernels is known). Still, computations involving the parameters
k,j
might be of interest. For instance, we show here how to calculate the pa-
rameters for Gegenbauer polynomials. For a number >
1
2
, these are orthogonal
polynomials associated to the weight function w(x) = B(
1
2
, +
1
2
)
1
(1 x
2
)
1
2
on (1, 1) (B denotes the beta function). We use the normalization constants from
[21], thus the Gegenbauer polynomials are
P
n
(x) =
(1)
n
2
n
( +
1
2
)
n
(1 x
2
)
1
2
d
n
dx
n
(1 x
2
)
n+
1
2
,
where (x)
n
is the Pochhammer symbol, (x)
0
= 1 and (x)
n
=
n
k=1
(x +k 1) for
n 1. We have:
h
n
=
1
B(
1
2
, +
1
2
)
_
1
1
_
P
n
(x)
_
2
(1 x
2
)
1
2
dx =
n!(n + 2)
2(2 + 1)
n
(n +)
and the three term recurrence is:
P
n+1
(x) =
2(n +)
n + 2
xP
n
(x)
n
n + 2
P
n1
(x)
(see [21], Ch. 1). We now let
n
(x, 0) denote the orthonormal polynomials associ-
ated to the weight function w, hence
n
(x, 0) =
1
n
P
n
(x). From the three term
relation we deduce
n
(0, 0) = (1)
n+1
2(2 + 1)
n
(n +)
n!(n + 2)

n
k=1
k 1
k 1 + 2
,
and also, the leading coecient of
n
(x, 0) is
k
,0
n
=
(n + 2)
n
2
n
_
+
1
2
_
n
2(2 + 1)
n
(n +)
n!(n + 2)
.
In order to compute the parameters
k,j
of the weight function w we use
Theorem 3.3. Therefore we need to calculate the values
n
(0, l) and k
,l
n
, n 1,
l 0, where k
,l
n
denotes the leading coecient of
n
(0, l). The main point for these
calculations is to notice that
n
(x, l)
n0
is the family of orthonormal polynomi-
als associated to the weight function x
2l
w(x). These polynomials are also classical
objects and they can be found for instance in [21] under the name of modied
classical polynomials. A calculation of the modied Gegenbauer polynomials can
be obtained in terms of Jacobi polynomials. These are orthogonal polynomials
associated to parameters , > 1 and weight function
2
1
B( + 1, + 1)
1
(1 x)
(1 +x)
on (1, 1) by the formula

P
(,)
n
(x) =
(1)
n
2
n
n!
(1 x)
(1 +x)
d
n
dx
n
(1 x)
+n
(1 +x)
+n
.
According to [21], Sect. 1.5.2, we have
2n
(x, l) = c
2n
P
1
2
,l
1
2
n
(2x
2
1)
and
2n+1
(x, l) = c
2n+1
xP
1
2
,l+
1
2
n
(2x
2
1),
where c
n
is a constant that remains to be determined. But rst we can already
notice that the above formulae give
2n+1
(0, l) = 0, so that
l,2n+1+l
= 0.
Theorem 3.4. For n, l 1,
2n
(0, l) = (1)
n+1
( + 1)
l
_
1
2
_
l
h
,l
2n
k=1
+l +k 1
k
and
k
,l
2n
=
( +l)
2n
_
l +
1
2
_
n
n!
( + 1)
l
_
1
2
_
l
h
,l
2n
,
k
,l
2n+1
=
( +l)
2n+1
_
l +
1
2
_
n+1
n!
( + 1)
l
_
1
2
_
l
h
,l
2n+1
,
where
h
,l
2n
=
_
+
1
2
_
n
( +l)
n
( +l)
n!
_
l +
1
2
_
n
( +l + 2n)
and
h
,l
2n+1
=
_
+
1
2
_
n
( +l)
n+1
( +l)
n!
_
l +
1
2
_
n+1
( +l + 2n + 1)
.
Proof. It is more convenient to introduce the polynomials
C
(,l)
2n
(x) =
( +l)
n
_
l +
1
2
_
n
P
(
1
2
,l
1
2
)
n
(2x
2
1),
C
(,l)
2n+1
(x) =
( +l)
n+1
(l +
1
2
)
n+1
xP
(
1
2
,l+
1
2
)
n
(2x
2
1),
and again by classical results that can be found in [21], we deduce
1 =
_
1
1
x
2l
_
2n
(x, l)
_
2
w(x)dx
= c
2
2n
_
(l+
1
2
)
n
(+l)
n
_
2
B(l+
1
2
,+
1
2
)
B(
1
2
,+
1
2
)
_
1
1
x
2l
_
C
,l
2n
(x)
_
2
1
B(l +
1
2
, +
1
2
)
(1 x
2
)
1
2
dx
= c
2
2n
_
(l+
1
2
)
n
(+l)
n
_
2
B(l+
1
2
,+
1
2
)
B(
1
2
,+
1
2
)
h
,l
2n
,
where
h
,l
2n
=
( +
1
2
)
n
( +l)
n
( +l)
n!
_
l +
1
2
_
n
( +l + 2n)
.
Using that
B(l+
1
2
,+
1
2
)
B(
1
2
,+
1
2
)
=
(
1
2
)
l
(+1)
l
, we deduce
2n
(x, l) =
( + 1)
l
_
1
2
_
l
h
,l
2n
C
,l
2n
(x).
The calculation of
2n
(0, l) reduces to the calculation of C
,l
2n
(0) which can be
easily done due to the three term relation
C
,l
2n+2
(x) =
+l + 2n + 1
n + 1
xC
,l
2n+1
(x)
+l +n
n + 1
C
,l
2n
(x).
Thus we deduce
C
,l
2n+2
(0) =
+l +n
n + 1
C
,l
2n
(0),
and by iterating this relation and using that C
,l
0
(0) = 1, we get
2n
(0, l) = (1)
n+1
( + 1)
l
_
1
2
_
l
h
,l
2n
k=1
+l +k 1
k
.
The leading coecient of
2n
(x, l) can be obtained from the corresponding formula
in [21]. Thus,
k
,l
2n
=
( +l)
2n
_
l +
1
2
_
n
n!
( + 1)
l
_
1
2
_
l
h
,l
2n
and
k
,l
2n+1
=
( +l)
2n+1
_
l +
1
2
_
n+1
n!
( + 1)
l
_
1
2
_
l
h
,l
2n+1
,
where
h
,l
2n+1
=
_
+
1
2
_
n
( +l)
n+1
( +l)
n!
_
l +
1
2
_
n+1
( +l + 2n + 1)
.
Now the parameters
k,j
can be easily calculated by using Theorem 3.4.
Of course, the explicit formulae look too complicated to be recorded here.
3.5. Asymptotic properties
In the classical setting of orthogonal polynomials on the unit circle there are several
remarkable asymptotic results given by G. Szeg o, see [25]. Let be a measure in the
Szego class, and let
n
n0
be the family of orthonormal polynomials associated
to . Then, the orthonormal polynomials have the following asymptotic properties:
n
0 (3.12)
and
1
, (3.13)
where
is the spectral factor of and the convergence is uniform on the compact

subsets of the unit disk D. The second limit (3.13) is related to the so-called Szego
limit theorems concerning the asymptotic behavior of Toeplitz determinants. Thus,
det T
n
det T
n1
=
1
[
n
(0)[
2
,
where T
n
= [s
ij
]
n
i,j=0
and s
k
kZ
is the set of the Fourier coecients of . As
a consequence of the previous relation and (3.13) we deduce Szeg os rst limit
theorem,
lim
n
det T
n
det T
n1
= [
(0)[
2
= exp(
1
2
_
2
0
log
(t)dt). (3.14)
The second (strong) Szego limit theorem improves (3.14) by showing that
lim
n
det T
n
g
n+1
()
= exp
_
1
_ _
]z]1
[
(z)/
(z)[
2
d(z)
_
, (3.15)
where g() is the limit in formula (3.14) and is the planar Lebesgue measure.
These two limits (3.14) and (3.15) have an useful interpretation in terms of asymp-
totics of angles in the geometry of a stochastic process associated to (see [25])
and many important applications. We show how these results can be extended
to orthogonal polynomials on T
1
. The formulae (3.7) and (3.8) suggest that it is
more convenient to work in a larger algebra. This is related to the so-called Toeplitz
embedding, see [19], [23]. Thus, we consider the set L of lower triangular arrays
a = [a
k,j
]
k,j0
with complex entries. No boundedness assumption is made on these
arrays. The addition in L is dened by entry-wise addition and the multiplication
is the matrix multiplication: for a = [a
k,j
]
kj
, b = [b
k,j
]
k,j0
two elements of L,
(ab)
k,j
=
l0
a
k,l
b
l,j
,
which is well dened since the sum is nite. Thus, L becomes an associative, unital
algebra.
Next we associate the element
n
of L to the polynomials
n
(X
1
, l) =
n
k=0
a
l
n,k
X
k
1
, n, l 0, by the formula
(
n
)
k,j
=
_
a
j
n,kj
k j
0 k < j;
(3.16)
similarly, the element
n
of L is associated to the family of polynomials
n
(X
1
, l) =
n
k=0
b
l
n,k
X
k
1
, n, l 0, by the formula
(
n
)
k,j
=
_
b
j
n,kj
k j
0 k < j.
(3.17)
We notice that the spectral factor
of K
is an element of L and we assume

that
belongs to the Szego class. This implies that
n
is invertible in L for all
n 0. Finally, we say that a sequence a
n
L converges to a L if (a
n
)
k,j
converges to a
k,j
for all k, j 0 (and we write a
n
a).
Theorem 3.5. Let belong to the Szego class. Then
n
0 (3.18)
and
(
n
)
1

. (3.19)
We now briey discuss the geometric setting for the kernel K
. By a classical
result of Kolmogorov (see [30]), K
is the covariance kernel of a stochastic process

f
n
n0
L
2
() for some probability space (X, /, ). That is,
K
(m, n) =
_
X
f
n
f
m
d.
The operator angle between two spaces c
1
and c
2
of L
2
() is dened by
B(c
1
, c
2
) = P
L
1
P
L
2
P
L
1
,
where P
L
1
is the orthogonal projection of L
2
() onto c
1
. Also dene
(c
1
, c
2
) = I B(c
1
, c
2
).
We can assume, without loss of generality, that f
n
n0
is total in L
2
() and
we associate to the process f
n
n0
a family of subspaces H
r,q
of L
2
() such that
H
r,q
is the closure of the linear space generated by f
k
, r k q. We consider a
scale of limits:
s lim
r
(H
0,n
, H
n+1,r
) = (H
0,n
, H
n+1,
) (3.20)
for n 0, and then we let n and deduce
s lim
n
(H
0,n
, H
n+1,
) = (H
0,
,
n0
H
n,
), (3.21)
where s lim denotes the strong operatorial limit.
We then deduce analogues of the Szeg o limit theorems (3.14) and (3.15) by
expressing the above limits of angles in terms of determinants. This is possible due
to (3.9).
Theorem 3.6. Let belong to the Szego class. Then
D
r,q
D
r+1,q
= s
r,r
det (H
r,r
, H
r+1,q
) =
1
[
qr
(0, r)[
2
(3.22)
and
lim
q
D
r,q
D
r+1,q
= s
r,r
det (H
r,r
, H
r+1,
) = [
(r, r)[
2
= s
r,r
j1
d
2
r,r+j
. (3.23)
If we denote the above limit by g
r
and
L = lim
n
0k<n<j
d
2
k,j
> 0,
then
lim
n
D
0,n
n
l=0
g
l
=
1
det (H
0,
,
n0
H
n,
)
=
1
L
. (3.24)
Details of the proofs can be found in [8].
3.6. Szeg o kernels
The classical theory of orthogonal polynomials is intimately related to some classes
of analytic functions. Much of this interplay is realized by Szeg o kernels. Here we
expand this idea by providing a Szeg o type kernel for (a slight modication of) the
space H
2
0
(T) which is viewed as an analogue of the Hardy class H
2
on the unit disk.
We mention that another version of this idea was developed in [2] (and recently
applied to a setting of stochastic processes indexed by vertices of homogeneous
trees in [3]); see also [4], [6]. The dierence is that H
2
0
(T) is larger than the space
|
2
of [2] and also, the Szeg o kernel that we consider is positive denite.
We return now to the space H
2
0
(T) introduced in Subsection 3.2. Its denition
involves a family of Hilbertian conditions therefore its natural structure should
be that of a Hilbert module (we use the terminology of [28]). Assume T
n
= C
for all n 0 and consider the C
-algebra T of bounded diagonal operators on
n0
T
n
= l
2
(N
0
). For a sequence d
n
n0
we use the notation
diag (d
n
n0
) =
_
_
d
0
0 . . .
0 d
1
. . . 0
.
.
.
.
.
. d
2
0
.
.
.
_
_
,
so that diag (d
n
n0
) belongs to T if and only if sup
n0
[d
n
[ < . We now dene
the vector space
H
2
(T) = H
2
0
(T) [ diag (c
n
()
c
n
()
n0
) T
and notice that T acts linearly on H
2
(T) by D = [
k,j
d
j
]
k,j0
, therefore H
2
(T)
is a right T-module. Also, if , belong to H
2
(T), then diag (c
n
()
c
n
()
n0
)
belongs to T, which allows us to dene
, ) = diag (c
n
()
c
n
()
n0
) ,
turning H
2
(T) into a Hilbert T-module. As a Banach space with norm || =
|, )|
1/2
, the space H
2
(T) coincides with l
(N
0
, l
2
(N
0
)), the Banach space of
bounded sequences of elements in l
2
(N
0
). A similar construction is used in [11] for
a setting of orthogonalization with invertible squares.
Next we introduce a Szeg o kernel for H
2
(T). Consider the set
B
1
= z
n
n0
C [ sup
n0
[z
n
[ < 1
and for z = z
n
n0
B
1
, H
2
(T), notice that
(z) = diag
_
n,n
+
k>n
k,n
z
k1
. . . z
n
n0
_
is a well-dened element of T. Also, for z B
1
, we dene
S
z
=
_
_
1 0 0 . . .
z
0
1 0 . . .
z
0
z
1
z
1
1 . . .
z
0
z
1
z
2
z
1
z
2
z
2
. . .
.
.
.
.
.
.
.
.
.
.
.
.
_
_
,
which is an element of H
2
(T) and the Szego kernel is dened on B
1
by the formula:
S(z, w) = S
w
, S
z
), z, w B
1
. (3.25)
Theorem 3.7. S is a positive denite kernel on B
1
with the properties:
(1) (z) = , S
z
), H
2
(T), z B
1
.
(2) The set S
z
D [ z B
1
, D T is total in H
2
(T).
Proof. Take z
1
, . . . z
m
B
1
and after reshuing the matrix [S(z
j
, z
l
)]
m
j,l=1
can be
written in the form
n0
_
c
n
(S
z
j
)
c
n
(S
z
l
)
m
j,l=1
.
Since each matrix
_
c
n
(S
z
j
)
c
n
(S
z
l
)
m
j,l=1
=
_
_
c
n
(S
z
1
)
.
.
.
c
n
(S
z
m
)
_
_
c
n
(S
z
1
)
.
.
.
c
n
(S
z
m
)
_
_
is positive,
we conclude that S is a positive denite kernel on B
1
. Property (1) of S follows
directly from denitions. For (2) we use approximation theory in L
spaces in
order to reduce the proof to the following statement: for every element H
2
(T)
for which there exists A N such that c
n
() = h l
2
(N
0
) for n A and
c
n
() = 0 for n / A, and for every > 0, there exists a linear combination L of
elements S
z
D, z B
1
, D T, such that sup
n0
|c
n
( L)| < . This can be
achieved as follows. Since the set
w
(z) =
1
1zw
[ w B
1
is total in the Hardy
space H
2
on the unit disk, we deduce that there exist complex numbers c
1
, . . .,
c
m
and w
1
, . . ., w
m
, [w
k
[ < 1 for all k = 1, . . . m, such that
|h
m
k=1
c
k
_
_
1
w
k
w
2
k
.
.
.
_
_
| < .
Then dene z
k
= w
k,n
n0
for k = 1, . . . , m, where w
k,n
= w
k
for all n 0. So
z
k
B
1
. Also dene d
A
n
= 1 for n A and d
A
n
= 0 for n / A, and consider
L =
m
k=1
c
k
S
z
k
diag
_
d
A
n
_
.
We deduce that L H
2
(T), c
n
(L) = 0 for n / A, and c
n
(L) =
m
k=1
c
k
_
_
1
w
k
w
2
k
.
.
.
_
_
for n A, so that
sup
n0
|c
n
( L) = sup
n0
|c
n
() c
n
(L)
= maxsup
nA
|h c
n
(L)|, sup
n/ A
|c
n
(L)|
= |h
m
k=1
c
k
_
_
1
w
k
w
2
k
.
.
.
_
_
| < .
4. Several isometric variables

In this section we discuss an example of a dening relation in several variables.
More precisely, we consider orthogonal polynomials in several variables satisfying
the isometric relations X
+
k
X
k
= 1, k = 1, . . . , N. We set / = 1 X
+
k
X
k
[ k =
1, . . . , N and notice that the index set of / is F
+
N
. Also if is a linear functional
on 1(/) then its kernel of moments is invariant under the action of F
+
N
on itself
by juxtaposition, that is,
K
(,
) = K
(,
), , ,
F
+
N
. (4.1)
In fact, a kernel K satises (4.1) if and only if K = K
for some linear functional

on 1(/). Positive denite kernels satisfying (4.1) have been already studied, see
for instance [9] and references therein. In particular, the class of isotropic processes
on homogeneous trees give rise to positive denite kernels for which a theory of
orthogonal polynomials (Levinson recursions) was developed in [10]. Here we dis-
cuss in more details another class of kernels satisfying (4.1) which was considered,
for instance, in [24].
4.1. Cuntz-Toeplitz relations
Consider the class of positive denite kernels with property (4.1) and such that
K(, ) = 0 if there is no F
+
N
such that = or = . (4.2)
We showed in [17] that K has properties (4.1) and (4.2) if and only if K =
K
for some q-positive functional on 1(/

CT
), where /
CT
= 1 X
+
k
X
k
[ k =
1, . . . , N X
+
k
X
l
, k, l = 1, . . . , N, k ,= l. The relations in /
CT
are dening the
Cuntz-Toeplitz algebra (see [22] for details). The property (4.2) shows that K is
quite sparse, therefore it is expected to be easy to analyze such a kernel. Still,
there are some interesting aspects related to this class of kernels, some of which
we discuss here.
Let be a strictly q-positive kernel on 1(/
CT
) and let K
be the associated
kernel of moments. Since the index set of /
CT
is still F
+
N
, and this is totally ordered
by the lexicographic order, we can use the results described in Subsection 3.1 and
associate to K
a family
,
of complex numbers with [

,
[ < 1, uniquely
determining K
by relations of type (3.4). It was noticed in [17] that K
has
properties (4.1) and (4.2) if and only if
,
=
,
and
,
= 0 if there
is no F
+
N
such that = or = . The main consequence of these
relations is that K
is uniquely determined by
=
,
, F
+
N
. We
dene d
= (1 [
[
2
)
1/2
. The orthogonal polynomials associated to satisfy
the following recurrence relations which follow easily from (3.7), (3.8) (see [17] for
details):
= s
1/2
,
and for k 1, . . . , N, F
+
N
,
k
=
1
d
k
(X
k
k1
), (4.3)
k
=
1
d
k
(
k
X
k
k1
). (4.4)
The results corresponding to Theorem 3.5 and Theorem 3.6 can be also easily
obtained (see [8]), but the constructions around the Szeg o kernel are more inter-
esting in this situation. Thus, there is only one Hilbertian condition involved in
the denition of H
0
(T) in this case. In fact, it is easy to see that H
0
(T) can be
identied with the full Fock space l
2
(F
+
N
), the l
2
space over F
+
N
. Now, concerning
evaluation of elements of H
0
(T), if we are going to be consistent with the point of
view that the points for evaluation come from the unital homomorphisms of the
polynomial algebra inside H
0
(T), then we have to consider an innite-dimensional
Hilbert space c and the set
B
1
(c) = Z = (Z
1
, . . . , Z
N
) L(c)
N
[
N
k=1
Z
k
Z
k
< I.
For = i
1
. . . i
k
F
+
N
we write Z
instead of Z
i
1
. . . Z
i
k
. Then we dene for
l
2
(F
+
N
) c and Z B
1
(c),
(Z) =
F
+
N
Z
,
which is an element of the set L(c) of bounded linear operators on the Hilbert
space c. Next, for Z B
1
(c) we dene S
Z
: c l
2
(F
+
N
) c by the formula:
S
Z
f =
F
+
N
e
(Z
f, f c.
Then S
Z
L(c, l
2
(F
+
N
) c) and we can nally introduce the Szeg o kernel on
B
1
(c) by the formula:
S(Z, W) = S
Z
S
W
, Z, W B
1
(c).
Theorem 4.1. S is a positive denite kernel on B
1
(c) with the properties:
(1) (z) = S
Z
, l
2
(F
+
N
) c, Z B
1
(c).
(2) The set S
Z
f [ Z B
1
(c), f c is total in l
2
(F
+
N
) c.
Proof. The fact that S is positive denite and (1) are immediate. More interesting
is (2) and we reproduce here the proof given in [17]. Let f = f
F
+
N
be an
element of l
2
(F
+
N
) c orthogonal to the linear space generated by S
Z
f [ Z
B
1
(c), f c. Taking Z = 0, we deduce that f
= 0. Next, we claim that for each

F
+
N
there exist
Z
l
= (Z
l,1
, . . . , Z
l,N
) B
1
(c), l = 1, . . . , 2[[,
such that
range
_
Z
1,
. . . Z
2]],
_
= c,
and
Z
l,
= 0 for all ,= , [[ [[, l = 1, . . . , 2[[.
Once this claim is proved, a simple inductive argument gives f = 0. In order
to prove the claim we need the following construction. Let e
n
ij
n
i,j=1
be the matrix
units of the algebra M
n
of n n matrices. Each e
n
ij
is an n n matrix consisting
of 1 in the (i, j)th entry and zeros elsewhere. For a Hilbert space c
1
we dene
E
n
ij
= e
n
ij
I
L
1
and we notice that E
n
ij
E
n
kl
=
jk
E
n
il
and E
n
ji
= E
n
ij
. Let = i
1
. . . i
k
so that c = c
2]]
1
for some Hilbert space c
1
(here we use in an essential way the
assumption that the space c is of innite dimension). Also, for s = 1, . . . , N, we
dene J
s
= l 1, . . . , k [ i
k+1l
= s and
Z
p,s
=
1
rJ
s
E
2]]
r+p1,r+p
, s = 1, . . . , N, p = 1, . . . , [[.
We can show that for each p 1, . . . , [[,
Z
p,
=
1
2
k
E
2]]
p,k+p
(4.5)
and
Z
p,
= 0 for ,= , [[ [[. (4.6)
We deduce
N
s=1
Z
p,s
Z
p,s
=
1
2
N
s=1
rJ
s
E
2]]
r+p,r+p1
E
2]]
r+p1,r+p
=
1
2
N
s=1
rJ
s
E
2]]
r+p,r+p
=
1
2
k
r=1
E
2]]
r+p,r+p
< I,
hence Z
p
B
1
(c) for each p = 1, . . . , [[. For every word = j
1
. . . j
k
F
+
N

we deduce by induction that
Z
p,j
k
. . . Z
p,j
1
=
1
2
k
rA
E
2]]
r+p1,r+p+k1
, (4.7)
where A
=
k1
p=0
(J
j
kp
p) 1, . . . , N and J
j
kp
p = l p [ l J
i
kp
.
We show that A
= 1 and A
= for ,= . Let q A
. Therefore, for
every p 0, . . . , k 1 we must have q + p J
j
kp
or i
k+1qp
= j
kp
. For
p = k 1 we deduce j
1
= i
2q
and since 2 q 1, it follows that q 1. Also
q 1, therefore the only element that can be in A
is q = 1, in which case we must

have = . Since l J
i
k+1l
for every l = 1, . . . , k 1, we deduce that A
= 1
and A
= for ,= . Formula (4.7) implies (4.5). In a similar manner we can

construct elements Z
p
, p = [[ + 1, . . . , 2[[, such that
Z
p,
=
1
2
k
E
2]]
p+k,p
and
Z
p,
= 0 for ,= , [[ [[.
Thus, for s = 1, . . . , N, we dene K
s
= l 1, . . . , k [ i
k
= s and
Z
p,s
=
1
rK
s
E
2]]
r+pk,r+pk1
, s = 1, . . . , N, p = [[ + 1, . . . , 2[[.
Now
_
Z
1,
. . . Z
2]],
_
=
1
2
k
_
E
2]]
1,k+1
. . . E
2]]
k,2k
E
2]]
k+1,1
. . . E
2]]
2k,k
_
,
whose range is c. This concludes the proof.
It is worth noticing that property (2) of S is no longer true if c is nite-
dimensional. In fact, for dimc = 1 the set S
Z
f [ Z B
1
(c), f c is total in
the symmetric Fock space of C
N
(see [5]).
4.2. Kolmogorov decompositions and Cuntz relations
This is a short detour from orthogonal polynomials, in order to show a construction
of bounded operators satisfying the Cuntz-Toeplitz and Cuntz relations, based on
parameters
F
+
N
]
associated to a positive denite kernel with properties
(4.1) and (4.2).
First we deal with the Kolmogorov decomposition of a positive denite kernel.
This is a more abstract version of the result of Kolmogorov already alluded to in
Subsection 3.5. For a presentation of the general result and some applications,
see [22], [30]. Here we consider K : N
0
N
0
C a positive denite kernel and
let
k,j
be the family of parameters associated to K as in Subsection 3.1. In
addition we assume K(j, j) = 1 for j 0. This is not a real loss of generality and
it simplies some calculations. We also assume [
k,j
[ < 1 for all k < j. Then we
introduce for 0 k < j the operator V
k,j
on l
2
(N
0
) dened by the formula
V
k,j
= (J(
k,k+1
) 1
n1
)(1 J(
k,k+2
) 1
n2
) . . . (1
n1
J(
k,j
)) 0
and we notice that
W
k
= s lim
j
V
k,j
(4.8)
is a well-dened isometric operator on l
2
(N
0
) for every k 0. If we dene V (0) =
I/C and V (k) = W
0
W
1
. . . W
k1
/C for k 1, then we obtain the following result
from [13].
Theorem 4.2. The map V : N
0
l
2
(N
0
) is the Kolmogorov decomposition of the
kernel K in the sense that
(1) K(j, l) = V (l), V (j)), j, l N
0
.
(2) The set V (k) [ k N
0
is total in l
2
(N
0
).
It is worth noticing that we can write explicitly the matrix of W
k
:
_
k,k+1
d
k,k+1
k,k+2
d
k,k+1
d
k,k+2
k,k+3
. . .
d
k,k+1

k,k+1
k,k+2

k,k+1
d
k,k+2
k,k+3
. . .
0 d
k,k+2

k,k+2
k,k+3
. . .
.
.
. 0 d
k,k+3
.
.
. 0
.
.
.
.
.
.
_
_
.
We also see that Theorem 3.7 and Theorem 4.1 produce various kinds of Kol-
mogorov decompositions for the corresponding Szego kernels. Based on a remark
in [12], we use Theorem 4.2 in order to obtain some large families of bounded
operators satisfying Cuntz-Toeplitz and Cuntz relations. Thus, we begin with a
positive denite kernel K with properties (4.1) and (4.2). For simplicity we also
assume K(, ) = 1 and let
F
+
N
]
be the family of corresponding parame-
ters. In order to be in tune with the setting of this paper, we assume [
[ < 1 for
all F
+
N
. Motivated by the construction in Theorem 4.2, we denote by
c
(W
0
), F
+
N
, the columns of the operator W
0
. Thus,
c
1
(W
0
) =
_
1
d
1
0
.
.
.
_
_
and c
(W
0
) =
_
_
d
1
. . . d
1
1
d
2
. . . d
1
.
.
.
0
.
.
.
_
_
for 1 .
We now dene the isometry U(k) on l
2
(F
+
N
) by the formula:
U(k) = [c
k
(W
0
)]
F
+
N
, k = 1, . . . , N.
Theorem 4.3.
(a) The family U
1
, . . . , U
N
satises the Cuntz-Toeplitz relations: U
k
U
l
=
k,l
I,
k, l = 1, . . . , N.
(b) The family U
1
, . . . , U
N
satises the Cuntz relations U
k
U
l
=
k,l
I, k, l =
1, . . . , N and
N
k=1
U
k
U
k
= I, if and only if
F
+
N
]
d
= 0. (4.9)
Proof. (a) follows from the fact that W
0
is an isometry. In order to prove (b) we
need a characterization of those W
0
which are unitary. Using Proposition 1.4.5 in
[16], we deduce that W
0
is unitary if and only if (4.9) holds.
It is worth mentioning that if we dene V () = I/C and V () = U()/C,
F
+
N
, where U() = U(i
1
) . . . U(i
k
) provided that = i
1
. . . i
k
, then V is
the Kolmogorov decomposition of the kernel K. A second remark here is that the
condition (4.9) is exactly the opposite of the condition for K being in the Szeg o
class. Indeed, it is easily seen that for a positive denite kernel with properties
(4.1) and (4.2), the condition (3.5) is equivalent to
F
+
N
]
d
> 0.
5. Several hermitian variables
In this section we discuss another example of dening relations in several variables.
The theory corresponding to this case might be viewed as an analogue of the theory
of orthogonal polynomials on the real line. We set / = Y
k
Y
+
k
[ k = 1, . . . , N
and /
= / Y
k
Y
l
Y
l
Y
k
[ k, l = 1, . . . , N, and notice that 1(/) = T
N
. Also,
1(/
) is isomorphic to the symmetric algebra over C

N
. Orthogonal polynomials
associated to 1(/
), that is, orthogonal polynomials in several commuting vari-

ables were studied intensively in recent years, see [21]. In this section we analyze
the non-commutative case. The presentation follows [15].
Let be a strictly q-positive functional on T
N
(/
2
) and assume for some
simplicity that is unital, (1) = 1. The index set of / is F
+
N
and let
F
+
N
be the orthonormal polynomials associated to . We notice that for any P, Q T
N
,
X
k
P, Q)
= (Q
+
X
k
P)
= (Q
+
X
+
k
P)
= P, X
k
Q)
,
which implies that the kernel of moments satises the relation s
,
= s
,I()
for
, , F
+
N
, where I denotes the involution on F
+
N
given by I(i
1
. . . i
k
) = i
k
. . . i
1
.
This can be viewed as a Hankel type condition, and we already noticed that even
in the one-dimensional case the parameters
k.j
of the kernel of moments of a
Hankel type are more dicult to be used. Therefore, we try to deduce three-terms
relations for the orthonormal polynomials. A matrix-vector notation already used
in the commutative case, turns out to be quite useful. Thus, for n 0, we dene
P
n
= [
]
]]=n
, n 0, and P
1
= 0.
Theorem 5.1. There exist matrices A
n,k
and B
n,k
such that
X
k
P
n
= P
n+1
B
n,k
+P
n
A
n,k
+P
n1
B
n1,k
, k = 1, . . . , N, n 0.
Each matrix A
n,k
is a selfadjoint N
n
N
n
matrix, while each B
n,k
is an
N
n+1
N
n
matrix such that
B
n
=
_
B
n,1
. . . B
n,N
is an upper triangular invertible matrix with strictly positive diagonal for every
n 0. For n = 1, B
1,k
= 0, k = 1, . . . , N. The fact that B
n
is upper triangular
comes from the order that we use on F
+
N
. The invertibility of B
n
is a consequence
of the fact that is strictly q-positive and appears to be a basic translation of this
information. It turns out that there are no other restrictions on the matrices A
n,k
,
B
n,k
as shown by the following Favard type result.
Theorem 5.2. Let
_
a
,
X
, F
+
N
, be elements in T
N
such that
= 1
and a
,
> 0. Assume that there exists a family A
n,k
, B
n,k
[ n 0, k = 1, . . . , N,
of matrices such that A
n,k
= A
n,k
and B
n
=
_
B
n,1
. . . B
n,N
is an upper
triangular invertible matrix with strictly positive diagonal for every n 0. Also
assume that for k = 1, . . . , N and n 0,
X
k
[
]
]]=n
= [
]
]]=n+1
B
n,k
+ [
]
]]=n
A
n,k
+ [
]
]]=n1
B
n1,k
,
where [
]
]]=1
= 0 and B
1,k
= 0 for k = 1, . . . , N. Then there exists a unique
strictly positive functional on 1(/) such that
F
+
N
is the family of or-
thonormal polynomials associated to .
There is a family of Jacobi matrices associated to the three-term relation in
the following way. For P 1(/)(= T
N
), dene
(P)
= P
.
Since the kernel of moments has the Hankel type structure mentioned above, it
follows that each
(P) is a symmetric operator on the Hilbert space H
with
dense domain T, the linear space generated by the polynomials
, F
+
N
.
Moreover, for P, Q T
N
,
(PQ) =
(P)
(Q),
and
(P)T T, hence
is an unbounded representation of T
n
(the GNS
representation associated to ). Also, (P) =
(P)1, 1)
for P T
N
. We dis-
tinguish the operators
k
=
(Y
k
), k = 1, . . . , N, since
F
+
N
c
) =
F
+
N
c
,
, where we use the notation
,
=
i
1
. . .
i
k
for = i
1
. . . i
k
. Let
e
1
, . . . , e
N
be the standard basis of C
N
and dene the unitary operator W from
l
2
(F
+
N
) onto H
such that W(e
) =
, F
+
N
. We see that W
1
T is the linear
space T
0
generated by e
, F
+
N
, so that we can dene
J
k
= W
1
,k
W, k = 1, . . . , N,
on T
0
. Each J
k
is a symmetric operator on T
0
and by Theorem 5.1, the matrix of
(the closure of) J
k
with respect to the orthonormal basis e
F
+
N
is
J
k
=
_
_
A
0,k
B
0,k
0 . . .
B
0,k
A
1,k
B
1,k
0 B
1,k
A
2,k
.
.
.
.
.
.
.
.
.
.
.
.
_
_
.
We call (J
1
, . . . , J
N
) a Jacobi N-family on T
0
. It is somewhat unexpected that the
usual conditions on A
n,k
and B
n,k
insure a joint model of a Jacobi family in the
following sense.
Theorem 5.3. Let (J
1
, . . . , J
N
) a Jacobi N-family on T
0
such that A
n,k
= A
n,k
and
B
n
=
_
B
n,1
. . . B
n,N
is an upper triangular invertible matrix with strictly

positive diagonal for every n 0. Then there exists a unique strictly q-positive
functional on T
N
with associated orthonormal polynomials
F
+
N
such that
the map
W(e
) =
, F
+
N
,
extends to a unitary operator from l
2
(F
+
N
) onto H
and
J
k
= W
1
,k
W, k = 1, . . . , N.
Proof. First the Favard type Theorem 5.2 gives a unique strictly q-positive func-
tional on T
N
such that its orthonormal polynomials satisfy the three-term re-
lation associated to the given Jacobi family, and then the GNS construction will
produce the required W and
,k
, as explained above.
One possible application of these families of Jacobi matrices involves some
classes of random walks on F
+
N
. Figure 4 illustrates an example for N = 2 and
more details are planned to be presented in [7].
e e
e e e e
e e
e e
e e e e
/
/
/
/
.
.
.
.
.
.
.
/
/
/
/
`
`
`
`
.
.
.
.
.
.
.
/
/
/
/
/
/

\
\
/
/
`
`
/
/
\
\
>
>
>

\
\

>
>
/
/
\
\
Figure 4. Random walks associated to a Jacobi family, N = 2
We conclude our discussion of orthogonal polynomials on hermitian variables
by introducing a Szeg o kernel that should be related to orthogonal polynomials
on T
N
. Thus, we consider the Siegel upper half-space of a Hilbert space c by
H
+
(c) = (W
1
. . . W
N
) L(c)
N
[ W
1
W
1
+. . . +W
N1
W
N1
<
1
2i
(W
N
W
N
).
We can establish a connection between B
1
(c) and H
+
(c) similar to the well-known
connection between the unit disk and the upper half plane of the complex plane.
Thus, we dene the Cayley transform by the formula
C(Z) = ((I +Z
N
)
1
Z
1
, . . . , (I +Z
N
)
1
Z
N1
, i(I +Z
N
)
1
(I Z
N
)),
which is well dened for Z = (Z
1
, . . . , Z
N
) B
1
(c) since Z
k
must be a strict
contraction (|Z
k
| < 1) for every k = 1, . . . , N. In addition, C establishes a one-
to-one correspondence from B
1
(c) onto H
+
(c). The Szego kernel on B
1
(c) can
be transported on H
+
(c) by the Cayley transform. Thus, we introduce the Szeg o
kernel on H
+
(c) by the formula:
S(W, W
) = F
W
F
W
, W, W
H
+
(c),
where F
W
= 2diag((i + W
N
))S
C
1
(W)
. Much more remains to be done in this
direction. For instance, some classes of orthogonal polynomials of Jacobi type and
their generating functions are considered in [7].
Finally, we mention that there are examples of polynomial relations for which
there are no orthogonal polynomials. Thus, consider
/ = X
+
k
X
k
[ k = 1, . . . , N X
k
X
l
+X
l
X
k
[ k, l = 1, . . . , 2N,
then 1(/) (C
N
), the exterior algebra over C
N
. If is a unital q-positive
denite functional on 1(/), then (X
2
k
) = 0 for k = 1, . . . , N. This and the
q-positivity of force (X
) = 0, ,= , therefore there is only one q-positive

functional on 1(/) which is not strictly q-positive. Therefore there is no theory
of orthogonal polynomials over this /. However, the situation is dierent for / =
X
k
X
l
+ X
l
X
k
[ k, l = 1, . . . , 2N. This and other polynomial relations will be
analyzed elsewhere.
References
[1] J. Agler and J.E. McCarthy, Pick Interpolation and Hilbert Function Spaces, Grad-
uate Studies in Mathematics, Vol. 44, Amer. Math. Soc., Providence, Rhode Island,
2002.
[2] D. Alpay, P. Dewilde and H. Dym, Lossless inverse scattering and reproducing kernels
for upper triangular operators, in Oper. Theory Adv. Appl., Vol. 47, Birkhauser,
Basel, 1990, pp. 61135.
[3] D. Alpay and D. Volok, Point evaluation and Hardy space on an homogeneous tree,
lanl, OA/0309262.
[4] W. Arveson, Interpolation problems in nest algebras, J. Funct. Anal., 3(1975), 208
233.
[5] W. Arveson, Subalgebras of C
-algebras. III: Multivariable operator theory, Acta.

Math., 181(1998), 476514.
[6] J.A. Ball and I. Gohberg, A commutant lifting theorem for triangular matrices with
diverse applications, Integr. Equat. Oper. Th., 8(1985), 205267.
[7] T. Banks and T. Constantinescu, Orthogonal polynomials in several non-commuting
variables. II, in preparation.
[8] M. Barakat and T. Constantinescu, Tensor algebras and displacement structure. III.
Asymptotic properties, Zeit. f ur Anal. Anw., 23(2004), 253274.
[9] M. Basseville, A. Benveniste, K.C. Chou, S.A. Golden, R. Nikoukhah, and A.S. Will-
sky, Modeling and estimation of multiresolution stochastic processes, IEEE Trans.
Inform. Theory, 38(1992), 766784.
[10] M. Basseville, A. Benveniste, and A.S. Willsky, Multiscale autoregressive Processes,
Part I: Schur-Levinson parametrizations, IEEE Trans. Signal Processing, 40(1992),
19151934; Part II: Lattice structures for whitening and modeling, IEEE Trans.
Signal Processing, 40(1992), 19351954.
[11] A. Ben-Artzi and I. Gohberg, Orthogonal polynomials over Hilbert modules, in Oper.
Theory Adv. Appl., Vol. 73, Birkhauser, Basel, 1994, pp. 96126.
[12] T. Constantinescu, Modeling of time-variant linear systems, INCREST preprint
No.60/1985.
[13] T. Constantinescu, Schur analysis of positive block matrices, in Oper. Theory Adv.
Appl., Vol. 18, Birkhauser, Basel, 1986, pp. 191206.
[14] T. Constantinescu, Factorization of positive-denite kernels, in Oper. Theory Adv.
Appl., Vol. 48, Birkhauser, Basel, 1990, pp. 245260.
[15] T. Constantinescu, Orthogonal polynomials in several non-commuting variables. I,
in Spectral Theory and its Applications, Theta, Bucharest, 2003, pp. 245260.
[16] T. Constantinescu, Schur Parameters, Factorization and Dilation Problems, Birk-
h auser, Basel, 1996.
[17] T. Constantinescu and J.L. Johnson, Tensor algebras and displacement structure.
II. Non-commutative Szego theory, Zeit. f ur Anal. Anw., 21(2002), 611626.
[18] T. Constantinescu and A. Gheondea, Representations of Hermitian kernels by means
of Krein spaces, Publ. RIMS, 33(1997), 917951; II. Invariant kernels, Commun.
Math. Phys., 216(2001), 409430.
[19] Ph. Delsarte, Y. Genin and Y. Kamp, On the Toeplitz embedding of arbitrary ma-
trices, Linear Algebra Appl., 51(1983), 97119.
[20] V. Drensky, Free Algebras and PI-Algebras, Springer Verlag, Berlin, 1999.
[21] C.H. Dunkl and Y. Xu, Orthogonal Polynomials of Several Variables, Cambridge
Univ. Press, 2001.
[22] D.E. Evans and Y. Kawahigashi, Quantum Symmetries on Operator Algebras, Oxford
Mathematical Monographs, Clarendon Press, 1998.
[23] C. Foias, A.E. Frazho, I. Gohberg and M.A. Kaashoek, Metric Constrained Interpo-
lation, Commutant Lifting and Systems, Birkhauser, Basel-Boston, 1998.
[24] A.E. Frazho, On stochastic bilinear systems, in Modeling and Applications of Sto-
chastic Processes, Kluwer Academic, 1988, pp. 215241.
[25] U. Grenander and G. Szego, Toeplitz Forms and their Applications, Univ. of Cali-
fornia Press, California, 1958.
[26] W.W. Hart and W.L. Hart, Plane Trigonometry, Solid Geometry, and Spherical
Trigonometry, D.C. Heath and Co., 1942.
[27] P. Henrici, Applied and Computational Complex Analysis. Volume 1: Power Series,
Integration, Conformal Mapping, Location of Zeros, Wiley-Interscience, 1974.
[28] E.C. Lance, Hilbert C
-modules, Cambridge University Press, 1995.

[29] H. Li, Noncommutative Gr obner Bases and Filtered-Graded Transfer, LNM 1795,
Springer Verlag, Berlin, 2002.
[30] K.R. Parthasarathy, An Introduction to Quantum Stochastic Calculus, Birkhauser,
Basel, 1992.
[31] M. Rosenblum and J. Rovnyak, Hardy Classes and Operator Theory, Oxford Univ.
Press, 1985.
[32] G. Szego, Orthogonal Polynomials, Colloquium Publications, 23, Amer. Math. Soc.,
Providence, Rhode Island, 1939.
[33] B. Sz.-Nagy and C. Foias, Harmonic Analysis of Operators on Hilbert Space, North
Holland, 1970.
T. Banks and T. Constantinescu
University of Texas at Dallas
Richardson, TX 75083, USA
e-mail: banks@utdallas.edu
e-mail: tiberiu@utdallas.edu
J.L. Johnson
Department of Mathematics and Computer Science
Wagner College
Staten Island, NY 10301, USA
e-mail: joeljohn@wagner.edu
Operator Theory:
Functions of Several Variables in the
Theory of Finite Linear Structures
Part I: Analysis
M. Bessmertny
Abstract. The notion of a nite linear structure is introduced which gener-
alizes the notion of a linear electrical network. A nite linear structure is a
Kirchho graph to which several pairs of external vertices (terminals) are
coupled. A Kirchho graph is a nite linear graph g to whose edges p the
quantities U
g
and I
g
are related. The values of U
g
and I
g
are complex num-
bers. It is assumed that the circuit Kirchho law holds for U
g
and the nodal
Kirchho law holds for I
g
. It is also assumed that for the quantities U
g
(gener-
alized voltages) and the quantities I
g
(generalized currents) corresponding to
the edges of the Kirchho graph the generalized Ohm law U
g
= z
g
I
g
holds.
The generalized impedances z
g
are complex numbers which are considered as
free variables. The Kirchho laws and the state equations U
g
= z
g
I
g
lead to
linear relations for the values U
g
and I
g
corresponding to the external edges
of the nite linear structures, i.e., the edges incident to the terminals. These
linear relations between external voltages and currents can be expressed either
in terms of the impedance matrix Z if the considered nite linear structure
is k-port or in terms of the transmission matrix A if the considered nite
linear structure is 2k-port. The properties of the impedance and transmis-
sion matrices Z and A as functions of the complex variables z
g
are studied.
The consideration of the present paper served as a natural motivation for the
study of the class of matrix functions which was introduced in the previous
paper MR2002589 of the author.
Mathematics Subject Classication (2000). Primary: 47B50; Secondary: 30A96
94A20.
Keywords. System theory, electrical networks, impedance matrices, transfer
matrices, functions of several complex variables.
This paper is a translation, prepared by V. Katsnelson, of a part from the third chapter from
the authors Ph.D. thesis entitled Functions of Several Variables in the Theory of Finite Linear
Structures, Kharkov, 1982.
92 M. Bessmertny
1. Summary of the graph theory
1
In graph theory the terminology and symbolism is not standardized. Below we

follow mainly [SeRe] and [Zy].
A nite graph g = gX, P is a nite set X = x ,= of points (called
vertices) connected by a nite set P = p of line segments (called edges).
Every edge connects two and only two vertices of the graph g; several edges
can be connected to one vertex.
A vertex x and an edge p are said to be incident if the vertex x is an endpoint
of the edge p. The notation x p means that the vertex x and the edge p
are incident.
Let g = gX, P be a graph with the vertices set X and the edges set P, and
let P
be a subset of the set P. The subgraph of the graph g generated by the

subset P
P is the nite graph g
= g
, P
, where the set X
consists
of all those vertices x
j
X of the graph g which are incident at least to one
p
.
An orientation of an edge p is a triple x
1
, p, x
2
, where x
1
, x
2
are the ver-
tices incident to the edge p. The orientation x
2
, p, x
1
is said to be opposite
to the orientation x
1
, p, x
2
.
The edge set P of the graph gX, P is said to be oriented if a certain
orientation is assigned to every edge p P. The oriented edges are denoted
by arrows. The direction of the arrow corresponds to the orientation of the
edge.
A path in the graph g =gX,P is a nite sequence x
1
p
1
x
2
p
2
x
3
. . . x
k
p
k
x
k+1
such that each edge p
l
is incident to the vertices x
l
and x
l+1
(1 l k).
If x
k+1
= x
1
, and the edges and the vertices (except x
1
) are not repeated,
then the path is said to be a simple circuit.
The set of all simple circuits of the graph gX, P is denoted by F.
An edge p P is incident to a circuit f F if the circuit f contains the
edge p.
We distinguish the simple circuits f = x
1
p
1
x
2
p
2
x
3
. . . x
k
p
k
x
1
and f
=
x
1
p
k
x
k
. . . x
3
p
2
x
2
p
1
x
1
by the orientations of their edges p
l
k
l=1
.
A simple circuit f F is said to be an oriented simple circuit, or a contour
of the graph gX, P if one of its two possible orientations is chosen.
The orientations of a contour f and of an edge p which is incident to the
contour, p f, are
equal if the orientation of the edge p coincides with the chosen orientation
of the contour;
opposite if the orientation of the edge p is opposite to the chosen orientation
of the contour.
A graph g is said to be connected if every two vertices can be connected by
some path.
A connected component of the graph g is a connected subgraph of g which is
not contained in another connected subgraph of g.
Functions of Several Variables in Finite Linear Structures 93
An isolated vertex of the graph g (i.e., the vertex which does not connect
to another vertices by edges) is considered as a connected component of the
graph g.
A tree of a connected graph g is a connected subgraph of g which contains
all the vertices but no circuit of the graph g.
If a tree of a graph is xed, then the edge belonging to the tree is said to be
the branch of the tree, and the edge which does not belong to the tree is said
to be the chord of the tree.
The forest of the graph g is the set of all the trees of all its connected com-
ponents.
A simple cut-set of the graph g is the set P
0
of its edges such that
a) the subgraph g
0
obtained from the graph g by removing the set P
0
has
more connected components than the original graph g;
b) no proper subset of the set P
0
satises the condition a).
It is clear that every simple cut-set of the graph g is a simple cut-set of some
connected component of g. Removing the set P
0
of the edges of some simple cut-set
increases the number of connected components of the graph by one.
In what follows we need to consider oriented simple cut-sets. Let P
0
be a
simple cut-set of a connected graph g. Then the subgraph g
0
obtained from the
graph g by removing the set P
0
of the edges has two connected components. The
cut-set P
0
of the connected graph g is oriented if each edge p of the cut-set P
0
is
assigned the orientation x
1
p x
2
, where x
1
, x
2
are vertices which are incident to
the edge p, and x
1
X
1
, x
2
X
2
.
We say that the orientations of the cut-set P
0
and of the edge p which belongs
to P
0
are equal if the directions on the edge corresponding to the orientations of
the cut-set and of the edge coincide;
are opposite if the directions on the edge corresponding to the orientations
of the cut-set and of the edge are opposite.
Denition 1.1. A graph gX, P is said to be an oriented nite graph if
1. gX, P is a nite graph.
2. The edges, the simple circuits and the simple cut-sets of the graph are enu-
merated in such a manner that:
(a) p
j
(j = 1, 2, . . . , e) are all the edges of the graph gX, P;
(b) f
k
(k = 1, 2, . . . , c) are all the simple circuits of the graph gX, P.
(c) s
( = 1, 2, . . . , s) are all the simple cut-sets of the graph gX, P;

3. The edges, the simple circuits and the simple cut-sets of the graph gX, P
are oriented.
94 M. Bessmertny
For oriented graphs, the notions of the circuit matrix B
a
and the cut-set
matrix Q
a
are introduced:
b
kj
=
_
_
1, if p
j
f
k
, and their orientations are the same;
0, if p
j
, f
k
;
1, if p
j
f
k
, and their orientations are opposite;
q
j
=
_
_
1, if p
j
s
, and their orientations are the same;

0, if p
j
, s
;
1, if p
j
s
, and their orientations are opposite.

It is known that the circuit matrix is of rank ev+r, and the cut-set matrix is of the
rank v r. (Here e, v and r are the total number of edges, vertices and connected
components, respectively.) As usual, B and Q are submatrices of the exhaustive
circuit matrix B
a
and the cut-set matrix Q
a
, respectively, consisting of some ev+
r and v r linearly independent rows. One of the main properties of the matrices
B and Q is their orthogonality: Q B
= 0. The proof can be found in [SeRe].

2
Let us dene the notion of a Kirchho graph.

Denition 1.2. A graph g = gX, P is said to be a Kirchho graph if
1. gX, P is a nite oriented graph.
2. To each oriented edge , p
j
, (, are the vertices incident to the edge
p
j
) correspond some quantities U
(j)
and I
(j)
. To the opposite oriented edge

, p
j
, correspond the quantities U
(j)
= U
(j)
; I
(j)
= I
(j)
. (We do not
concretize the nature of the quantities U and V : they can be numbers or
functions, real or complex, or, more general, can belong to abelian groups,
the quantities U
(j)
to one group, the quantities I

(j)
to another.)
Moreover the following assumptions are satised:
(a) The rst Kirchho law
p
j
S
I
(j)
= 0 (1.1)
holds for every cut-o S of the graph g. (The summation runs over the
directions on the edges p
j
S which are determined by the orientation
of the cut-o S.)
(b) The second Kirchho law
p
j
f
U
(j)
= 0 (1.2)
holds for every contour f of the graph g. (The summation runs over the
directions on the edges p
j
f which are determined by the orientation
of the contour f.)
3. The multiplication U
(j)
I
(j)
of the quantities U
(j)
and I
(j)
(not commutative
in general, but distributive with respect to addition) is dened.
Remark 1.3. Since the graph gX, P is oriented, to each edge p
j
corresponds one
distinguished direction. Introducing the vector-columns I and U with entries I
(j)
and U
(j)
, one can represent the Kirchho laws in the form

Q I = 0, B U = 0,
where Q and B are the above-described submatrices of the exhaustive cut-o and
circuit matrices Q
a
and B
a
.
Remark 1.4. Usually one formulates the rst Kirchho law for special cut-os of
the graph g, namely for the so-called nodes. A node is the set of edges which are
incident to one vertex. In [SeRe] is shown that this restricted formulation of the
rst Kirchho law is equivalent to the general one.
Langevins Theorem. For every Kirchho graph gX, P the relations
I
(j)
U
(j)
= 0,
U
(j)
I
(j)
= 0 (1.3)
hold. (The summation is taken over all edges directed according to given orienta-
tion.)
Proof. Without loss of generality we may assume that the graph is connected. Let
us x some vertex
0
of the graph g and introduce the potential (
k
) of an
arbitrary vertex
k
:
(
k
) =
(
0
,
0
)
U
(j)
,
where
(
0
,
k
) =
0
p
0
1
p
1
2
. . .
k1
p
k1
k
is an arbitrary path.
The summation is taken according to the directions
j
, p
j

j+1
on the
edges p
j
belonging to the path (
0
,
k
). In view of the second Kirchho law,
(
k
) does not depend on the choice of the path leading from
0
to
k
. Then
I
j
U
(j)
,
I
(j)

_
() ()
.
However,
,
I
(j)
() =
I
(j)
_
() = 0
since the set of all edges incident to the vertex is a simple cut-o set (or the
union of simple cut-o sets), and according to the rst Kirchho law, the sum
in the square brackets is equal to zero. The same holds for the second summand
,
I
(j)
().
In textbooks on theoretical electrical engineering one formulates Langevins
theorem (in the setting of electrical networks) in the following manner: the total
active (reactive) power produced in all energy sources is equal to the total active
(reactive) power discharged in all energy collectors.
96 M. Bessmertny
In [EfPo] there was mentioned that Langevins Theorem is the immediate
consequence of Kirchhos laws. Hence this theorem holds for every system in
which Kirchhos laws are postulated.
If the quantities U
(j)
,
and I
(j)
,
are functions of some parameter t and if
Kirchhos laws hold for every t, then, according to Langevins theorem,
I
(j)
,
(t
) U
(j)
,
(t
) = 0
for every two values t
and t
of the parameter t.
In what follows we omit the subindices , in the expressions U
(j)
,
, I
(j)
,
for the sake of simplicity specifying to what directions of the edges correspond
positive values of the quantities U
j
and I
j
. In pictures we draw the arrow near
the notation U
j
and V
j
. The direction of the arrow denotes the direction to which
the positive value of the quantity corresponds.
2. Finite linear structures
1
Electrical networks will be the progenitor of the abstract objects considered in

the paper. Therefore, the terminology taken from the theory of electrical networks
is used.
Denition 2.1. A nite linear structure is a Kirchho graph with an additional
weighting structure:
1. The set of the edges of the Kirchho graph is decomposed into three subsets
P
0
, P and

P, where the subsets P and

P have the same number of edges.
2. The edges from P
0
are said to be the elementary 2-ports. (Figure 1).
I
j
U
j
1 1
U
j
I
j
1 1
Figure 1. Elementary 2-port and its graph.

The complex numbers U
j
and I
j
, which are concordant to the direction
on the edge p
j
P
0
corresponding to the orientation of p
j
, are related by
U
j
=
j
z
j
I
j
(
j
= 1),
where z
j
is a complex number.
The complex number z
j
is said to be the impedance of the elementary
2-port p
j
. In what follows these impedances z
j
are considered as independent
variables
1
.
If
j
= +1 for all elementary 2-ports, the nite linear structure is said
to be passive.
3. The edges from P
0
and

P are paired: (p
j
, p
j
). Each pair (p
j
, p
j
) forms an
ideal 2 2 transformer (see Figure 2).
The complex numbers U
j
,I
j
and

U
j
,
I
j
, concordant
2
to the directions
corresponding to the orientations of the edges p
j
, p
j
, are related by
_
U
j
I
j
_
=
_
t
j
0
0 t
1
j
_
U
j
I
j
_
,
where transmission coecients t
j
are assumed to be non-zero real numbers.
1
1
2
2
U

U
I

I
1
1
2
2
U

U
P

P
I

I
Figure 2. Ideal 2 2-transformer and its graph.
2
Let us dene the operation of opening coupling channels in a nite linear

structure.
Let g be a nite linear structure; let x
1
and x
2
be two dierent vertices of the
graph belonging to the same connected component. Let us connect these vertices
by the edge p of the ideal 2 2 transformers (see Figure 2). The pair 1 and 1
of
the vertices incident to the edge p is left free and is said to be the exterior pair of
vertices (or terminals) of the new nite linear structure g
which is derived from

the initial structure g and the ideal 2 2 transformer.
The above-described method of obtaining the pair of exterior vertices is said
to be opening coupling channels in the structure g. The considered nite linear
1
It was not ruled out that some dierent elementary 2-ports of the considered linear structure
have the same impedances, that is, some dierent edges from P
0
are weighted by the same
complex variable.
2
The reason why the current

I
j
carries a negative sign is that most transmission engineers like
to regard their output current as coming out of the output point instead of going into the port
as per standard usage.
98 M. Bessmertny
structure with the open coupling channels (interacting with other nite linear
structures by means of these coupling channels) is an open system (see [Liv]).
3
In what follows we consider nite linear structures with open coupling channels.
The structures interact with each other by means of these channels.
Denition 2.2. A nite linear structure with k open coupling channels (i.e., with
2k exterior terminals) is said to be 2k-port.
Denition 2.3. A nite linear structure with open coupling channels is said to be
2 2k-port if there are k coupling channels forming the input of the structure and
there are k coupling channels forming the output of the structure.
Let us connect each pair of the exterior vertices by the oriented exterior edge
and relate to these exterior edges the quantities
3
U
j
, I
j
: U
in
j
, I
in
j
to the input edges
and U
out
j
, I
out
j
to the output edges
4
. Thus the state of the coupling channels of
2k-port is characterized by two k-dimensional vector-columns
U =
_
U
1
, . . . , U
k
; I =
_
I
1
, . . . , I
k
;
and the state of the coupling channels of 2 2k-port is characterized by two 2k-
dimensional vector-columns
f
in
=
_
U
in
1
, . . . , U
in
k
; I
in
1
, . . . , I
in
k
;
f
out
=
_
U
out
1
, . . . , U
out
k
; I
out
1
, . . . , I
out
k
;
2k-ports and 2 2k-ports will be depicted as it is shown in Figure 3.
I
1
I
k
1
1
k
k
U
1
U
k
I
in
1
I
in
k
1
1
k
k
U
in
1
U
in
k
U
out
1
U
out
k
I
out
1
I
out
k
Figure 3
In what follows we use often the simplest 2 2k-port the so-called ideal
2 2k-transformer (see Figure 4).
3
The values of these quantities are established after the transition to the steady state.
4
See Footnote
2
.
k
(2k)
U
k

U
k
k 2k
I
k

I
k
.
.
.
.
.
.
1
(k + 1)
U
1

U
1
1 (k + 1)
I
1

I
1
t
kk
t
1k
t
k1
t
11
(2k)
k 2k
1
(k + 1)
1 (k + 1)
. . .
. . .
.
.
.
.
.
.
p
11
p
k1
p
1k
p
kk
p
1k
p
kk
p
11
p
k1
Figure 4. Ideal 2 2k transformer and its graph.
For an ideal 2 2k-transformer the relation
_
U
I
_
=
_
t
0
0 t
1
_
I
_
holds, where U, I,

U,

U are vector-columns of the dimension k composed of the
quantities U
j
, I
j
and

U
j
,

I
j
, respectively.
The matrix t = t
pq
k
p,q=1
of the transmission numbers of the transformer
is assumed to be real and non-singular. The last requirement is equivalent to the
following: choosing the appropriate values of the quantities

U
j
,

I
j
, one can obtain
arbitrary prescribed values of the quantities U
j
, I
j
.
100 M. Bessmertny
Denition 2.4. A 22k-port (2k-port) is said to be passive if for all 2-ports which
are interior elements of the 2 2k-port (2k-port) the relations
U
j
=
j
z
j
I
j
, where
j
= +1,
hold.
4
Before we consider interaction characteristics of a nite linear structure, we

carry out a simple calculation.
The number of independent equations of the rst Kirchho law (i.e., the rank
of the cut-o matrix) is equal to v r; the number of independent equations of the
second Kirchho law (i.e., the rank of the circuit matrix) is equal to e+2k v+r.
The number of the equations related the values of U
j
and I
j
(the relations on the
inner edges) is equal to e. Here v is the number of the vertices, e is the number of
the interior, and 2k is the number of the exterior edges of the 2 2k-port; r is the
number of connected components of the Kirchho graph. The total number of the
equations is equal to 2(e +k). They relate 2(e + 2k) variables U
j
and I
j
.
If these equations are compatible and the variables U
in
j
and I
in
j
, j = 1, . . . , k,
can be chosen as free variables, then, solving the obtained system with respect
to unknown U
out
j
and I
out
j
, we obtain the relation
f
out
= A(z
1
, . . . , z
n
) f
in
,
where A(z
1
, . . . , z
n
) is a matrix function of dimension 2k 2k depending on
the variables z
1
, . . . , z
n
which are the impedances of the inner components of the
structure. The matrix function A(z
1
, . . . , z
n
) is said to be the transmission matrix
of the 2 2k-port.
Analogously, if the relation
U = Z(z
1
, . . . , z
n
) I
holds for a 2k-port, where
U =
_
U
1
, . . . , U
k
, I =
_
I
1
, . . . , I
k
are vector columns characterizing the states of the exterior edges of the 2k-port,
then the matrix function Z(z
1
, . . . , z
n
) of dimension k k is said to be the
impedance matrix of the 2k port.
Analogously, the admittance matrix of an 2k-port can be dened.
Since the considered linear structures are nite, the entries of the transmis-
sion matrix, the impedance matrix and the admittance matrix are expressed from
the coecients of the appropriate linear systems rationally. Therefore, the trans-
mission matrix, the impedance matrix and the admittance matrix are rational
functions of the variables z
1
, . . . , z
n
which are the impedances of the interior
edges of the multiport. In the next sections we obtain necessary conditions which
these matrices must satisfy if they exist. Remark that real physical problems lead
to such nite linear structures for which at least one of the interaction characteris-
tics (the transmission matrix, the impedance matrix and the admittance matrix)
exists, and very often all three matrices exist.
3. Characteristic matrices of nite linear structures
The main goal of this section is to obtain a necessary condition which matrices
must satisfy for to be the transmission matrix, the impedance or the admittance
matrix of a nite linear structure.
1
Since for all Kirchho graphs considered below the values of U

j
and I
j
are
complex numbers, Langevins theorem can be presented in the form
I
j
U
j
= 0 ,
U
j
I
j
= 0.
These equalities imply that
_
U
j
I
j
+I
j
U
j
_
= 0; i
_
U
j
I
j
I
j
U
j
_
= 0 . (L)
Characteristic properties of the matrices of multiports will be obtained as
the consequences of Langevins theorem. This approach was outlined in the paper
[Ef] by A.V. Emov. There apparently rstly Langevins theorem was applied for
deriving characteristic properties of the matrices of multiports as functions of the
complex frequency . (See also [EfPo].) We apply Langevins theorem for obtaining
characteristic properties of the matrices of nite linear structures as functions of
several complex variables.
To formulate the result about the properties of the impedance matrix of a
passive 2k-port, we recall the denition of a positive matrix function of several
complex variables.
To formulate this denition, we need the following subsets of C
n
:
T
+
7
= z : z C
n
, Re z
1
> 0, . . . , Re z
n
> 0 ,
T
7
= z : z C
n
, Re z
1
< 0, . . . , Re z
n
< 0 ,
T
+
1
= z : z C
n
, Imz
1
> 0, . . . , Imz
n
> 0 ,
T
1
= z : z C
n
, Imz
1
< 0, . . . , Imz
n
< 0 .
(3.1)
Denition 3.1. A rational k k matrix function M(z) = M(z
1
, . . . , z
n
) of complex
variables z
1
, . . . , z
n
is said to be positive if the following positivity conditions hold:
M(z) +M(z)
0 for z T
+
7
,
M(z) +M(z)
0 for z T
7
,
i (M(z)
M(z)) 0 for z T
+
1
,
i (M(z)
M(z)) 0 for z T
1
.
(3.2)
The class of positive functions was introduced in the authors PhD thesis
[Be1]. See [Be2] for an English translation of the part of this thesis containing
the denition of the class of positive functions. Denition 3.1 of the present paper
appears in [Be1] as Denition 0.3.
102 M. Bessmertny
We recall
Denition 3.2. A rational matrix function M(z
1
, . . . , z
n
) of complex variables
z
1
, . . . , z
n
is said to be real if it satises the condition M(z
1
, . . . , z
n
)
M(z
1
, . . . , z
n
).
Theorem 3.3. Assume that the impedance matrix Z(z
1
, . . . , z
n
) of some 2k-port
exists. Then Z(z
1
, . . . , z
n
) it is a k k rational matrix function of the variables
z
1
, . . . , z
n
which possesses the properties:
1. Z is real (in the sense of Denition 3.2).
2. Z(z
1
, . . . , z
n
) is a homogeneous function of homogeneity degree 1, i.e.,
Z(z
1
, . . . , z
n
) = Z(z
1
, . . . , z
n
) C.
3. Z is a symmetric matrix function:
Z(z
1
, . . . , z
n
) Z
(z
1
, . . . , z
n
) .
If moreover the 2k-port is passive, then the matrix function Z(z
1
, . . . , z
n
) is pos-
itive in the sense of Denition 3.1.
Proof. Since the impedance matrix of the 2k port exists, the appropriate system of
linear equations for the values U
j
, I
j
is compatible, and we can take the variables
I
j
corresponding to the exterior edges as free variables. Solving this system of
equations, we obtain the relations
U = Z(z
1
, . . . , z
n
) I,
I
(e)
= (z
1
, . . . , z
n
) I,
where U, I and I
(e)
are vector-columns composed from the values U
j
, I
j
(j =
1, . . . , k) and I
(e)
( = 1, . . . , e) corresponding respectively to the interior and

exterior edges of the 2k-port. Since the total numbers of the linear equations is
nite, the entries of the matrix functions Z and are rational functions of the co-
ecients of the system, thus, rational matrix functions of the variables z
1
, . . . , z
n
.
If all z
1
, . . . , z
n
take real values, then all the coecients of the system are real.
Hence, the entries of the matrices Z(z
1
, . . . , z
n
), (z
1
, . . . , z
n
) take real values
for real z
1
, . . . , z
n
. Therefore, the matrix function Z(z
1
, . . . , z
n
) satises the con-
dition
Z(z
1
, . . . , z
n
) Z(z
1
, . . . , z
n
).
The same is true for the matrix function (z
1
, . . . , z
n
).
Let C. Changing variables
z
, I
, U
, Z Z
does not aect the equalities in the considered system of linear equations.
Therefore,
(z
1
, . . . , z
n
) = (z
1
, . . . , z
n
) ,
Z(z
1
, . . . , z
n
) = Z(z
1
, . . . , z
n
) .
Let us apply Langevins theorem in the form L to the Kirchho graph. Taking
into account that the values corresponding to the directions distinguished by the
orientations on exterior edges and the output edges of the ideal transformers are
respectively U
j
, I
j
(j = 1, . . . , k) and U

, I

, we can rewrite L in the form
1jk
_
U
j
I
j
I
j
U
j
_
=
1e
_
U
(e)
I
(e)
I
(e)
U
(e)
1t
__
I

U

I
_
U

I

U
_
, (3.3)
where the sums correspond to the exterior edges j, to the elementary 2-ports ()
and to the ideal transformers (). After transformation, the sums in (3.3) take the
form
1jk
_
U
j
I
j
I
j
U
j
_
= I
_
Z
(z) Z(z)
_
I, (3.4)
1e
_
U
(e)
I
(e)
I
(e)
U
(e)
_
=
1e
_
I
(z
)I
_
=
1jn
(z
j
z
j
) I
(z)
j
(z)I, (3.5)
where
j
are diagonal matrices whose diagonal entries are either 1, 1 or 0. (The
rank of
j
is equal to the total number of the components with impedance z
j
.
See Footnote
1
.)
The summands corresponding to the ideal transformer vanish. Indeed, for the
th transformer,
_
I

U

I
_
+
_
U

I

U
_
=
_
U

, I

_
J
_
_
U

I

_
= 0,
_
I

U

I
_
U

I

U
_
= i
_
U

, I

_
J
H
T
J
H
T
_
_
U

I

_
= 0,
since
J
= 0 , J
H
T
J
H
T
= 0.
Here
J
=
_
0 1
1 0
_
; J
H
=
_
0 i
i 0
_
; T
=
_
t
0
0 t
1
_
is the transmission matrix of the th transformer.
104 M. Bessmertny
Taking into account (3.4) and (3.5), the equalities (3.3) can be rewritten as
I
_
Z
(z) +Z(z)
I =
1jn
(z
j
+z
j
)I
(z)
j
(z)I;
iI
_
Z
(z) Z(z)
I =
1jn
i(z
j
z
j
)I
(z)
j
(z)I.
Since the entries of I can be arbitrary, we have the equalities
_
Z
(z) +Z(z)
1jn
(z
j
+z
j
)
(z)
j
(z); (3.6)
i
_
Z
(z) Z(z)
1jn
i(z
j
z
j
)
(z)
j
(z). (3.7)
If all z
j
are real, the right-hand side of (3.7) vanishes. Thus, Z
(x) = Z(x) for

real x. However, Z(x) = Z(x) for real x. Hence, Z
(x) = Z(x) for real x. From the

uniqueness theorem, Z
(z) = Z(z) for all complex z.

If the considered 2k-port is passive, then all the matrices
j
are non-negative,
and the inequalities (3.2) hold for the matrix function Z.
Let us turn to consider the transmission matrices.
Recall that a square rational 2k-matrix function W(z) = w(z
1
, , . . . , z
n
)
of the complex variables z
1
, , . . . , z
n
is said to be J-expanding if the following
J-expandability conditions are satised:
W
(z)J
W(z) J
0 for z T
+
7
,
W
(z)J
W(z) J
0 for z T
7
,
W
(z)J
H
W(z) J
H
0 for z T
+
1
,
W
(z)J
H
W(z) J
H
0 for z T
1
,
where
J
=
_
0 I
m
I
m
0
_
, J
H
=
_
0 iI
m
iI
m
0
_
,
and the domains T
+
7
, T
7
, T
+
1
, T
1
were dened in (3.1).
Theorem 3.4. Assume that the transmission matrix A(z
1
, . . . , z
n
) of a passive 2
2k-port exists. Then A(z
1
, . . . , z
n
) is a real J-expanding rational matrix function
of the variables z
1
, . . . , z
n
.
As before, elementary reasoning leads us to the relations
f
out
= A(z
1
, . . . , z
n
) f
in
,
I
(e)
= (z
1
, . . . , z
n
) f
in
.
Langevins theorem in the form L results in the equalities
_
I
out
U
out
I
in
U
in
_
U
out
I
out
U
in
I
in
1e
_
U
(e)
I
e
I
(e)
U
e
_
,
where the summands in the left-hand side of the equality correspond to the exterior
edges of the 2 2k-port, and the summands in the right-hand side correspond to
the elementary 2-ports. The summands corresponding to the ideal transformers
are annihilated.
Carrying out calculations and using the matrices
J
=
_
0 I
k
I
k
0
_
and J
H
=
_
0 iI
k
iI
k
0
_
,
we obtain
f
in
_
A
(z)J
A(z) J
f
in
=
1jn
(z
j
+z
j
) f
in

(z)
j
(z) f
in
, (3.8)
f
in
_
A
(z)J
H
A(z) J
H
f
in
=
1jn
i(z
j
z
j
) f
in

(z)
j
(z) f
in
, (3.9)
where
j
0 since 22k-port is passive. Since the vector-column f
in
is arbitrary,
these inequalities result in J-expandability of the matrix function A(z
1
, . . . , z
n
).
Putting z
j
= i
j
(
j
, j = 1, . . . , n, are real) in (3.8), and z
j
= x
j
(x
j
, j =
1, . . . , n, are real) in (3.9), we obtain:
A
(i)J
A(i) J
0,
A
( x )J
H
A( x) J
H
0.
Since A(z) is rational,
A
(z)J
A(z) J
0, (3.10)
A
( z )J
H
A(z) J
H
0. (3.11)
Deriving the identities (3.10), (3.11), we did not use that the 22k-port is passive,
that is the matrices
j
are non-negative.
Thus, for arbitrary (not necessary passive) 2 2k ports the following state-
ment holds:
Theorem 3.5. Assume that the transmission matrix A(z
1
, . . . , z
n
) of a 22k-port
exists. Then A(z
1
, . . . , z
n
) is a real rational matrix function of variables z
1
, . . . ,
z
n
satisfying the symmetry conditions (3.10) and (3.11).
Editorial Remark. From (3.6), (3.7) it follows that
Z(z) =
1jn
z
j
(z)
j
(z) .
The left-hand side is analytic with respect to z, the factor (z) in the right-hand
side also is analytic with respect to z, and the factor
(z) in the right-hand side

is anti-analytic with respect to z. Therefore,
Z(z) =
1jn
z
j
()
j
(z) z, . (3.12)
106 M. Bessmertny
References
[Be1] Bessmertny i, M.F.: Funkcii neskolkih kompleksnyh peremennyh v teo-
rii koneqnyh line inyh struktur. Kandidacka dissertaci, Harkovski i
universitet. Harkov, 1982. 143 ss. (in Russian).
_
Bessmertny, M.F. Func-
tions of several complex variables in the theory of nite linear structures. PhD
thesis, Kharkov University. Kharkov, 1982. 143 pp.
[Be2] Bessmertny, M.F. On realizations of rational matrix functions of several com-

plex variables. Translated from the Russian by D. Alpay and V. Katsnelson. Oper.
Theory Adv. Appl., 134, Interpolation theory, systems theory and related topics
(Tel Aviv/Rehovot, 1999), 157185, Birkhauser, Basel, 2002.
[Ef] Efimov, A.V.: Ob odnom primenenii teoremy Lanevena v teorii cepe i.
DAN Armnsko i SSR, 49:3 (1969), 118123 (in Russian).[Efimov, A.V.: On
one application of Langevins theorem in the theory of electrical networks. DAN
Armyansk. SSR, 49:3 (1969), 118123.]
[EfPo] Efimov, A.V. i V.P. Potapov: J - rastgivawie matricy-funkcii i ih
rol v analitiqesko i teorii lektriqeskih cepe i. Uspehi matem. nauk,
28:1 (1973), 65130 (in Russian). English transl.:
Efimov, A.V. and V.P. Potapov: J-expanding matrix functions and their role
in the analytical theory of electrical circuits. Russ. Math. Surveys, 28:1 (1973),
pp. 69140.
[Liv] Livxic, M.S.: Operatory, Kolebani, Volny (otkrytye sistemy).
Nauka. Moskva, 1966. 298 ss. [In Russian]. English transl.:
Livsi c, M.S.(=Livshits, M.S.): Operators, oscillations, waves (open systems).
(Translations of Mathematical Monographs, Vol. 34.) American Mathematical
Society, Providence, R.I., 1973. vi+274 pp.
[SeRe] Seshu, S. and M.B. Reed: Linear Graphs and Electrical Networks. (Addison-
Wesley series in the engineering sciences. Electrical and control systems). Addison-
Wesley, Reading, MA, 1961. 315 p. Russian transl.:
Sexu, S., i M.B. Rid: Line inye Grafy i lektriqeskie Cepi. Izda-
telstvo Vysxa Xkola. Moskva, 1971. 448 s.
[Zy] Zykov, A.A.: Teori Koneqnyh Grafov. Izdatelstvo Nauka, Sibirskoe
Otdelenie. Novosibirsk, 1969. 543 s. [In Russian: Zykov, A.A.: Theory of
Finite Graphs.]
M. Bessmertny
Svody Square, 4
Department of Mathematics, Faculty of Physics
Kharkov National University
61077, Kharkov, Ukraine
Operator Theory:
Operator Method for Solution of PDEs
Based on Their Symmetries
Samuil D. Eidelman and Yakov Krasnov
Abstract. We touch upon operator analytic function theory as the solution
of frequent classes of the partial dierential equations (PDEs).
Mathematics Subject Classication (2000). 17A35, 30G30, 30G35, 35C05.
Keywords. Linear partial dierential equations; Second-order constant coe-
cient PDEs; Cauchy problem; Explicit solutions; Symmetry operator.
Introduction
Denition 0.1. A symmetry of PDEs is a transformation that maps any solution
to another.
It is well known that solving any problem in PDE theory can be substantially
facilitated by appropriate use of the symmetries inherit in the problem. For a given
solution of the DE, knowledge of an admitted symmetry leads to the generation
of another solution. Thus the following main question arises:
Question 0.2. Given a PDE Qu = 0. When does it have a property, that if one
starts with a simple solution and supplied symmetries, one can generate all solu-
tions?
We start with several motivation examples.
Example. The explicit (dAlembert) form of the solution to two-dimensional ellip-
tic and hyperbolic equations Qu u
xx
+ u
yy
= 0. Determining the formulas of
their solutions is now straightforward:
u(x, y) = f(K
1
) +g(K
2
); K
1
:= y +
x, K
2
:= y
x, (0.1)
with (complex) f(z), g(z).
108 S.D. Eidelman and Y. Krasnov
Example. The operator form [6], [29], [33] of the heat polynomial solution to the
parabolic equation Qu u
t
u
xx
= 0. It can be written as:
u(x, t) = p
m
(K)[1] :=
m
i=0
K
i
[a
i
], p
m
(x) =
m
i=0
a
i
x
i
, K := x + 2x
x
. (0.2)
By K[] in (0.2) we denote the action K on and K
i
[] := K[K
i1
[]].
Example. The polynomial-exponential solution of the Helmholz equation in R
2
:
Qu u
xx
+u
yy
u = 0. Namely, in [29] solutions obtained as
u(x, y) = p
m
(K)[e
x+y
], K = x
y
y
x
,
2
+
2
= (0.3)
with any complex polynomial p
m
(z).
Surprisingly, although types of the PDEs in Examples 13 are essentially
dierent, their solutions are represented in unied (unconventional) form:
u(x
1
, x
2
, . . . , x
n
) =
p
i=1
f
i
(K
1
, K
2
, . . . , K
s
)[u
0i
(x
1
, x
2
, . . . , x
n
)], (0.4)
The common factor in representations of the solutions to various type PDEs
in the form (0.4) is the existence of an symmetry operators K. We start with
Denition 0.3. Given PDE Q(x,
x
)u = 0. We could say that Q have enough
symmetries if (0.4) holds. For all f(K), u(x) is a solution to PDE Qu = 0 in R
n
.
In turn, the question whether or not a given PDO Q have enough symmetries
can not be answered without explicit calculations.
Denition 0.4. Let PDO Q have enough symmetries. Recall u(x) is a regular
solution to PDE Qu = 0 in R
n
i (0.4) holds with some entire functions f
i
(z).
Example. The heat equation has enough symmetries because (0.2) hold.
1. Let K be taken from (0.2). Obviously, e
z
is an entire function in C. Hence
e
x+t
:= e
K
[1] is a regular solution.
2. The fundamental solution of the heat equation E(x, t) =
1
2
t
e
x
2
4t
is singu-
lar. The straightforward computations show that E(x, t) belongs to the kernel
of K, (KE = 0) and therefore no analytic function F(z) fullls the represen-
tation E(x, t) := F(K)[1]. Only the Dirac delta function F(x) := (x) may
be considered as the reliable representative in this special case!
3. The function f(z) := 1/z is not analytic; therefore the function v(x, t)
v(x, t) := K
1
[u(x, t)] =
1
2t
_
x
0
e
y
2
x
2
4t
u(y, t)dy (0.5)
is neither necessarily regular nor is it the solution to heat equation for u(x, t)
in (0.2). Nevertheless, v
0
(x, t) K
1
[1] is a singular solution to (0.2):
v
0
(x, t) := K
1
[1] =
1
2
e
x
2
4t
er(
x
2
t
).
Here er(z) is an imaginary error function erf(iz)/i and K[v(x, t)] = 1.
Symmetry Operator Method 109
Denition 0.5. The minimal possible set of a PDOs K
1
, . . . , K
s
in Denition 0.3
will be called an operator-valued characteristics or an operator indeterminates.
Remark 0.6. The linear ODEs do not allow non-trivial (non-constant) operator
indeterminates since their solution space is always nite-dimensional.
Claim 0.7. Given PDE Q(x,
x
)u = 0 in R
n
. Suppose Q have enough symmetries
(see Denition 0.3). Then Q admit an operator-valued characteristics.
Proof. If such PDOs K
1
, K
2
, . . . , K
s
exists and u(x) fullls (0.4) then Q[u(x)] = 0
implies Q[K
i
[u(x)]] = 0 for i = 1, . . . , s. Therefore K
1
, K
2
, . . . , K
s
in (0.4) must
be the symmetry operators to PDO Q and Qu
0i
(x) = 0. We call functions u
0i
(x)
a simplest solution (see [4], [24], [20]).
The above cited PDEs possess, for arbitrary initial data, a regular solution, for
one may use, at least formally, the theory of entire functions in operator variables
(cf. [32], [18]). Moreover, their symmetries operators form a nite generated algebra
Lie (see [20]). However, a priori not all PDEs allow to write their solution in the
form (0.4). For a more rened use of (0.4), a deeper knowledge of symmetries to
the given PDE is indispensable.
The objectives of this paper is to show, for a given PDE Qu = 0, how to
construct its solution space ool
Q
(s, p). Namely, how to
nd their p simplest solutions,
construct s operator-valued characteristics in order to obtain (0.4),
study the conditions under which u(x) in (0.4) is a genuine function.
The following question is also of great importance:
Question 0.8. What is a correct denition of the singular solution to PDEs and
how can one construct these singular solutions using their symmetries?
Mathematical background
In the 19th century literature on dierential equations, one can nd extensive
discussions of the relation between singular and general (regular) solutions.
However, no rigorous denitions of these terms emerged.
The basic idea is that a general solution is a family of solutions with the
expected properties. It stands to reason that singular solutions, conversely, dif-
fer from a general one by a specialization of its non-smoothness. For example,
Euler eectively showed that exponential-polynomial functions, i.e., functions be-
ing represented in the form p
m
(x)e
,x)
where p
m
(x) is polynomial of order m,
forms the basis for a general solution to a linear constant coecient system. We
follow the results of L. Ehrenpreis [9] and V.P. Palomodov [25] in this direction.
The natural and important problem of nding a polynomial basis in the solution
space was studied in [19], [26], [29]. The method presented in this paper establishes
(formally) regular solutions, mainly exponential-polynomial solutions (cf. [19]).
Here and throughout the sequel, C[x] = C[x
1
, . . . , x
n
] denote the space of
(complex) exponential-polynomial functions and Q stands for a linear system of
PDO acting on the functions f C[x
1
, . . . , x
n
].
Our main observation is that by knowing a symmetry operators, it is possible
to study the solution space
ool
Q
(R
n
) := f C[x
1
, . . . , x
n
] : Q[f] = 0, ool
Q
(R
n
) ker Q. (0.6)
For a more detailed introduction we refer to L.V. Ovsiannikov [24], P.J. Olver [21],
[22], G.W. Bluman and J.D. Cole [3] and to many other texts that have followed
since the fundamental insight to symmetries of PDE due to Lie appeared.
One of the most widely accepted denitions of symmetry of PDO Q is:
Denition 0.9. (cf. [20],[21]) The linear PDO L is said to be a symmetry operator
of the operator Q in C[x] if L[u] Sol
Q
for all u Sol
Q
.
The second denition, also accepted and equivalent to the one stated above,
requires the existence of an additional operator to be in convolution with a given
operator, as follows:
Proposition 0.10. (cf. [20],[21]) The linear PDO L is a symmetry operator of the
operator Q in C[x] if (i) L : C[x] C[x] and (ii) QL = L
Q for some linear

dierential operator L
.
The rather trivial case in which L is a symmetry of Q is of the form RQ for
some linear operator R. In order to avoid such trivialities we adopt the convention
that L
1
L
2
if L
1
L
2
= RQ for some R. Denote by oym(Q) the set of all
symmetries operators for Q factorized by the above equivalence. Thus, it is not
hard to prove
Proposition 0.11. oym(Q) is an algebra with composition as a basic operation.
If the symmetry operator L of PDO Q is a mth-order PDO with exponential-
polynomial coecients (possible with values in certain associative algebra A), then
we call the operator L the m-symmetry of Q.
Denition 0.12. An operator Q is said to be of nite type if its symmetry algebra
oym(Q) is nitely generated by m-symmetries operators.
In [20], [21] it was proved that the symmetry algebra of PDO Q and nitely
generated by 1-symmetry operators, may be innite-dimensional only in R
2
. We
show in Section 5 that symmetries of the second order, linear, constant coecient
PDE are (algebraically) isomorphic to the pseudo-conformal Lie algebra so(p, m),
p m. We nish this section with
Denition 0.13. so(p, m) is the N = (p+m)(p+m1)/2-dimensional Lie algebra
of all real (p +m) (p +m) matrices / such that /(
p,m
+(
p,m
/
T
= 0 where
(
p,m
=
p
i=1
c
ii
i=1
c
(p+i)(p+i)
.
Here c
ij
is the (p +m) (p +m) matrices with one in row i, column j, and zeroes
otherwise:
c
ij
=
ij
kl
p+m
k,l=1
,
ij
kl
=
_
1, if k = i, l = j,
0, otherwise.
A basis in so(p, m) provided by N matrices
ij
:
ij
=
ji
= c
ij
c
ji
, if 1 i, j p, or q + 1 i, j p +m,
and
ij
=
ji
= c
ij
+ c
ji
otherwise.
Moreover,
ij
fullls the following commutation relations
[
ij
,
rs
] =
jr
is
+
is
jr
+
ri
sj
+
sj
ri
,
[
in
,
rs
] =
is
rn
+
ir
sn
, [
in
,
jn
] =
ij
where
mn
-Kronecker delta.
Notations
We shall be working on the Euclidean space R
n
or in C
n
, in which case we will
denote the partial dierentiation
i
:=
x
i
as dierentiation in coordinates. In
some cases we shall adopt the standard convention that a, b) is a scalar prod-
uct and multi-indices are merely markers, serving to identify the dierentiation:
u
(x) :=
1
x
1

n
x
n
u(x). Let C[D] = C[
1
, . . . ,
n
, x
1
, . . . , x
n
] denote the set of all
PDO with complex exponential-polynomial coecients. Actually, C[D] is a gener-
alization of the well-known Weil algebra. Let X R
n
be any open neighborhood
of an origin 0 X. C
(X) as usual denotes the ring of functions f : X C whose

partial derivatives of all order exists on X. Let C[[x]] = C[[x
1
, . . . , x
n
]] denote the
ring of formal, i.e., not necessarily convergent, power series in (complex) variables
x
1
, . . . , x
n
. For analytic f we will often use the Taylor (Maclaurin) series of f at 0
in the indeterminates x, otherwise for f C
we will use the formal power series

as a C-algebra homomorphism C
C[[x]] :
Tf(x) :=
!
f
()
(0).
Actually, a power series representation for the PDEs have recently developed
by many authors (see [29], [9], [25]). It is applicable to ordinary equations and gives
fruitful results. In particular, it is shown that solutions in the formal power series
of constant coecient PDEs are multisummable (see [25]). In the present paper we
treat solutions of a linear partial dierential equations and we provide proofs only
for new results. We will often construct the solution in the formal power series in
pairwise commutative operator indeterminates K
i
, i = 1, . . . , n. The solution is
C
in a neighborhood of x X and
Tf(x) :=
1
!
K
1
1
K
n
n
[f
()
(0)].
The relation between formal solutions (in operator indeterminates) and gen-
uine solutions with asymptotic expansions is an important problem. For a variety
of reasons that theory is quite technical. In this paper we describe an algorithm
which gives a constructive, countable basis for the set of exponential-power series
solutions to a given system of linear PDEs if they have enough symmetries as
it is shown in [33]. Solutions with some growth estimates of an entire functions
have power series expansion in operator indeterminates. The existence of a gen-
uine solution and it representation as functions belonging to some quasi-analytic
classes is not investigated in this paper. The multisummability of formal power
series solutions are studied for evolution equation only.
The contents of the present paper after the introduction is as follows.
In Section 1, we introduce the basic techniques involved with symmetry group
calculation. We use here ideas based on the Baker-Campbel-Hausdor formula for
the successive computation of an abstract operator exponent. If the operators
represented by successive commutators are of nite type, the symmetry operators
may be considered as perturbations of ordinary dierential operators in some sense.
The essential observation is that the multisummability of formal series for solutions
of evolution equation holds for commutative operators of nite type.
In Section 2, we give a denition of an analytic system of PDEs and give some
important examples. We devote Section 3 to discussions that the understanding
of a qualitative properties of the solution to system of PDEs acquire an algebraic
aspects. We show that many results for the system of partial dierential equations
are extensively connected with the underlying algebra. We introduce also denition
of an elliptic type system derived from the original notion of quasi-division algebras
and we connect the hyperbolicity with zero divisors structure in the underlying
algebra. However, we do not carry out the full linkage of these connections and
hope to carry it out in consequent papers.
In Section 4 our attention is focused to the power series expansion of the
solution to the Dirac operator in associative unital algebra. As a culmination of
this section, we apply symmetry operator methods to quaternionic and Cliord
analysis.
Most of the results of Section 5 are concerned with the linear, constant co-
ecient, homogeneous PDO in R
n
. We eectively show that 1-symmetries of any
homogeneous of order two PDEs are isomorphic to the pseudo-orthogonal algebra
so(p, q).
1. Evolution equations
In this Section we consider the evolution PDE in the form
Qu(t, x) :=
t
u(t, x) P(x,
x
)u(t, x) = 0, (1.1)
where one of the independent variables is distinguished as time and the other
independent variables x = x
1
, . . . , x
n
X are spatial.
We consider P to be an operator on the manifold X U where dependent
variables u are elements of U. Let G be a local one parametric group of trans-
formations on X U. In turn, we consider the rst-order dierential operator of
special type as a generator of a local group G. Our next step is
Denition 1.1. A local symmetry group of an equation (1.1) is a one-parameter
group g
t
of the transformation G, acting on X U, such that if u(t, x) U is
an arbitrary smooth enough solution of (1.1) and g
t
G then g
t
[u(, x)] is also a
solution of (1.1) for all small enough t, > 0.
S. Lie developed a technique for computing local groups of symmetries. His
observations were based on the theory of jet bundles and prolongation of the vector
elds whenever u is a solution of (1.1). An explicit formula can be found in [21].
In general, one can compute symmetries of (1.1) by the Lie-B acklund (LB)
method. Everything necessary can be found using explicit computations due to the
Baker-Campbell-Hausdor formula (in sequel: the BCH formula). This formula is
equivalent to the LB method and is based on successive commutator calculations:
K
i
= e
tP(x,)
x
i
e
tP(x,)
=
m0
1
m!
t
m
[P(x,
x
), x
i
]
m
. (1.2)
Here [a, b]
m
= [a, [a, b]
m1
], [a, b]
1
= ab ba and [a, b]
0
= b.
If all K
i
C[D] in (1.2) are PDEs of nite order, then Q in (1.1) is of nite
type (see Denition 0.12). The above considerations lead to the correctness of
Claim 1.2. Question 0.2 is answered in the positive for an evolution equation (1.1)
of nite type at least if the initial value problem u(0, x) = f(x) is well posed and
if e
t
for some is the solution of (1.1).
Proof. The solution to (1.1) may be written [12] in the form u = f(K)[e
t
].
Our next considerations initiate the following result which we will use in
Section 2:
Proposition 1.3. If Q is of nite type, then symmetry operators K
i
dened in (1.2)
and completing with identity operator, form a commutative, associative, unital
subalgebra of an algebra oym(Q) of all symmetries of Q.
Proof. Based on the verication of the commutator relation [P, x
i
]
m
= 0 for some
m and associativity of operation in the Weil algebra C[D].
1.1. Heat equation
It is not dicult to see that the BCH formula (1.2) for the heat equation
t
u = u in x R
n
(1.3)
possesses the pairwise commute operators K
i
= x
i
+ 2t
x
i
(cf. (0.2)). Therefore
the heat operator (1.3) is, of course, of nite type.
Let v
(x, t) for multi-indices , [[ = m (see [33]) be the polynomial solutions

to (1.3), (heat harmonics):
v
(x, t) := K
[1] = m!
[m/2]
s=0
]]=s
t
m2s
(m2s)!
x
!
, =
1
, . . . ,
n
. (1.4)
Existence of the following expansion was proved in [33]:
Theorem 1.4. A solution u(x, t) of (1.3) in R has an expansion
u(x, t) :=
n=0
K
n
[a
n
]
n!
=
n=0
a
n
n!
v
n
(x, t), a
n
=
n
x
u(0, 0), (1.5)
valid in the strip [t[ < , < x < , if and only if it is equal to its Maclaurin
expansion in the strip [t[ < , [x[ < .
In order to satisfy the Cauchy data u(x, 0) = f(x), the following necessary
and sucient conditions on a function f are known [33]:
Proposition 1.5. The Maclaurin expansion (1.5) leads to conditions that f(x) be
entire of growth (2, 1/(4)). In another case (1.5) is a formal representation of the
solution only.
1.2. Constant coecient evolution equation
Consider the general constant coecient evolution operator
Q :=
t
P(
x
).
From the BCH formula (1.2) follows trivially that Q is of nite type. The inni-
tesimal symmetries of Q, namely:
K
i
= e
tP(
x
)
x
i
e
tP(
x
)
= x
i
+t[P(
x
), x
i
] = x
i
+tP
i
(
x
) (1.6)
are pairwise commuting operators. The symbol P
i
(x) in (1.6) as usual stands for
partial derivation
x
i
P(x). Of course the function u
0
(t) = e
tP(0)
is one of the
simplest solutions meaning that Qu
0
(t) = 0 and u
0
(0) = 1. The solution of the
constant coecient evolution equation Qu = 0 with initial data u(x, 0) = f(x),
f(x) C[x] may be represented at least locally in the form
u(x, t) = f(K)[e
tP(0)
]. (1.7)
Here K = (K
1
, . . . , K
n
) are dened in (1.6).
Remark 1.6. Representation (1.7) forms a one-to-one correspondence between co-
ordinates x
i
and operator indeterminates K
i
.
The following theorem is a straightforward generalization of Theorem 1.4:
Theorem 1.7. A solution u(x, t) of the equation Q(
x
)u(x) = 0 has an expansion
u(x, t) :=
m=0
]]=m
1
!
K
u(0, 0)], (1.8)

valid in the ber [t[ < , x R
n
and is locally analytic if and only if u(x, 0) is
an entire function of order at most m/(m 1). In this case (1.8) is equal to its
Maclaurin expansion in the cylinder [t[ < , [[x[[ < .
Calculation of the symmetry operator indeterminates for some important
constant coecient PDE leads to the following considerations:
1.2.1. 2b-parabolic equation. In particular, the regular solution to the 2b-parabolic
equation [10] in R
n
t
u = (1)
b1
b
u (1.9)
may be written in the form:
u(x, t) = f(K
1
, . . . , K
n
)[1], K
i
= x
i
+ (1)
b1
2bt
b1
x
i
. (1.10)
1.3. Diusion with inertia equation
The Kolmogorov model of diusion with inertia based on solution of equation
Qu :=
t
u
2
x
u x
y
u = 0. (1.11)
Q is not a constant coecient PDE, but nevertheless, the operator Q(x,
x
) in
(1.11) allowed the technique of successively commutators via the BCH-formula
(1.2). Due to careful examination we obtain two 1-symmetry operators of Q in the
form
K
1
:= x + 2t
x
t
2
y
, K
2
:= y +xt +t
2
1
3
t
3
y
. (1.12)
It is possible to verify that (1.11) is of nite type and all three operators Q, K
1
, K
2
pairwise commute:
[K
1
, K
2
] = 0, [K
1
, Q] = [K
2
, Q] = 0.
Therefore, the solution of the Cauchy problem
Qu(x, y, t) = 0, u(x, y, 0) = f(x, y) (1.13)
with f(x, y) C[x] may be obtained in the operator form as follows: u(x, y, t) =
f(K
1
, K
2
)[1].
1.4. Some other evolution equations
Next we consider two PDEs [8], [4]:
The Fokker-Plank equation:
t
u =
2
x
u +x
x
u +u; (1.14)
and the harmonic oscillator
t
u =
1
2
2
x
u +
1
2
(1 x
2
)u. (1.15)
Calculation by the BCH formula (1.2) for both cases suppose to use innitely
many commutators. Nevertheless, the solution of (1.14) may be written in the form
u(x, t) = f(K)[e
t
] with 1-symmetry operator K := xe
t
+e
t
x
and solution of (1.15)
may be written as u(x, t) = f(K)[e
x
2
/2
] with operator K := xe
t
e
t
x
.
In fact there is no polynomial solution of (1.14) and (1.15). Also K in both
cases is PDO with exponential-polynomial coecients.
1.5. Evolution equations with time-depending coecients
Following [14] consider the Cauchy problem for higher-order linear homogeneous
evolution equations with coecients depending on the time variable:
t
u (x, t) =
]]m
a
(t)
x
u(x, t), u(x, 0) = f(x). (1.16)
Here x R
n
, t R, and the coecients a
(t) are real-valued continuous

functions of t [0, ]. Consider the following symmetry operators of (1.16):
K
i
= exp(P(t,
x
)) x
i
exp(P(t,
x
)) where P(t,
x
) =
]]m
_
t
0
a
()d
x
.
Using the BCH-formula we obtain an explicit formula for K = K
1
, . . . , K
n
:
K
i
= x
i
+P
i
(t,
x
) where P
i
(t, x) =
x
i
P(t, x). (1.17)
Then, at least formally using (1.17), we can write the solution of (1.16) in the form
u(x, t) = f(K)[u
0
(t)], u
0
(t) = exp
_
t
0
a
0
()d. (1.18)
Following [14], the formal representation (1.18) converges to an analytic function
u(x, t) in the strip 0 < t < i f is a real-valued entire function on R
n
of growth
(, ), meaning that
[f (z)[ = O
_
exp
_
i=1
[z
i
[
__
as z . (1.19)
Equivalently, f(x) can be written in form of the power series expansion
f(x) =
!
x
and lim sup
_
e
[[
_
1
[c
[
/]]
.
Under these conditions on the coecients of f(x), the solution (1.18) may be
expressed as a convergent series in operator indeterminates K. In [14] there are
also estimates on the maximal width of the strip of convergence of these series.
1.6. KdVB and non-linear equations
In the limiting cases the non-linear evolution equation reduces to the well-known
conventional Korteweg de Vries and Burgers (KdVB) equations, respectively. It
arises in various contexts as a model equation incorporating the eects of disper-
sion, dissipation and non-linearity. The general form of such an equation is given
by
t
u(x, t) =
1
u(x, t)
x
u(x, t) +
2
2
x
u(x, t) +
3
3
x
u(x, t) (1.20)
where
1
,
2
and
3
are some constant coecients.
These equations are both exactly solvable and they each have a wide range
of symmetries. It should be pointed out that the above-described technique of
symmetry operators is applicable to the solution of the KdVB equation.
2. System of PDE
Denition 2.1. Recall a system of PDEs for x R
n
is an analytic PDE if
(i) it may be written in the form of the rst-order (linear) system
L(x, D)u := L
k
(x, D)u
n
k=1
:=
n
i,j=1
a
k
ij
(x)
i
u
j
(x) = 0, (2.1)
(ii) the coecients a
k
ij
(x) in PDOs L
k
all are entire functions.
(iii) L(x, D) is an involutive system, meaning that there exist entire functions
b
k
ij
(x) such that the commutators [L
i
(x, D), L
j
(x, D)] fulll the relation
[L
i
(x, D), L
j
(x, D)] =
n
k=1
b
k
ij
(x)L
k
(x, D). (2.2)
The principal matrix symbol (L)(x, ) of (2.1) is dened [1] by
(L)(x, )
k
j
:=
n
i=1
a
k
ij
(x)
i
, j, k = 1, 2, . . . , n. (2.3)
In the neighborhood of the point x
0
, the solution u(x) of the equation (2.1)
is locally equivalent to the solution of the system L(x
0
, D)v(x) = 0 with constant
coecients. The constant coecient systems are evidently analytic. In this section
we study some properties of the solution spaces to analytic PDEs.
The solution space to analytic PDEs may be nite-dimensional and therefore
do not allow using operator indeterminates (K = const in Denition 0.5):
Example. Consider the following system of rst-order PDEs
x
i
u
i
=
j
x
j
u
j
,
x
i
u
j
+
x
j
u
i
= 0, 1 i < j N (2.4)
for N unknown functions u
1
(x
1
, . . . , x
N
), . . . , u
N
(x
1
, . . . , x
N
) and
k
= 1.
If N = 2 and
1
=
2
then (2.4) is the well-known Cauchy-Riemann system.
Although, for N > 2 the solution space to (2.4) is nite-dimensional as is shown in
Proposition 2.2. For N > 2 there are exactly
1
2
(N +1)(N +2) linear independent
solutions of (2.4) being represented in a form
u
m
(x
1
, . . . , x
N
) =
N
i=1
(2x
m
x
i
)
i
i
x
i
+
N
i=1
im
i
x
i
+x
m
+
m
, (2.5)
where
m
,
km
,
m
for k, m = 1, . . . , N are constants,
ij
+
ji
= 0.
Proof. Cross dierentiation of (2.4) yields for pairwise indices i, j, k 1, . . . , N
not equal one to another:
x
i
x
i
u
i
+
j
x
j
x
j
u
i

x
i
(
i
x
i
u
i
x
j
u
j
) +
j
x
j
(
x
i
u
j
+
x
j
u
i
) = 0;
x
i
x
i
u
k

j
x
j
x
j
u
k

i
x
i
x
i
u
k
+
k
x
k
x
k
u
k
(
k
x
k
x
k
u
k
+
j
x
j
x
j
u
k
) = 0;
2
x
i
x
j
u
k

x
i
(
x
j
u
k
+
x
k
u
j
)
x
k
(
x
i
u
j
+
x
j
u
i
) +
x
j
(
x
k
u
i
+
x
i
u
k
) = 0.
Recall the proof that all third-order derivatives of u
i
are also equal to zero:
x
i
x
i
x
i
u
i
= 0,
x
i
x
i
x
j
u
i
= 0,
x
i
x
i
x
j
u
j
= 0 for i ,= j, i, j 1, . . . , N.
But if N > 2, then there exists pairwise dierent i, j, k and
2
x
i
x
i
x
i
u
i
(
i
x
i
x
i

k
x
k
x
k
)(
i
x
i
u
i
x
j
u
j
)
+
j
x
j
(
i
x
i
x
i
u
j

k
x
k
x
k
u
j
) +
i
x
i
(
k
x
k
x
k
u
i
+
i
x
i
x
i
u
i
) = 0;
x
i
x
i
x
j
u
i

j
x
j
(
i
x
i
x
i
u
i
+
k
x
k
x
k
u
i
)
j
x
k
(
x
k
x
j
u
i
) = 0;
x
i
x
i
x
j
u
j

x
i
x
i
(
j
x
j
u
j

i
x
i
u
i
) +
i
x
i
x
i
x
i
u
i
= 0
for all pairwise non-equal i, j, k.
Using all the above relations we nally may conclude that only the second-
order polynomials form the common solutions of (2.4), because of
x
i
x
j
x
k
u
m
= 0
for all i, j, k, m = 1, . . . , N.
Choosing explicitly the coecients in quadratic terms, we obtain (2.5).
Belove we present results of some classical examples of the system of PDEs.
2.1. Dirac and Laplace equations in R
n
It is well known [5], [18] that the Laplacian may be factorized as = D D
where D is the Dirac operator and D is conjugate to D in the Cliord algebra
Cl
(0,n)
. Thus, one can choose the solution to the Laplace equation as a real part
of the solution to the Dirac equation. In fact, the Dirac equation is a system of
ODE with constant coecient. In R
4
one can write the Dirac equation in the form
gradu +
t
v + curl v = 0,
t
u div v = 0.
Algebras of 1-symmetries of the Dirac and Laplace equation in R
n
both are
isomorphic to the pseudo-orthogonal Lie algebra so(n + 1, 1).
2.2. Lame equation
The Lame equation
v + ( +)grad = 0, =
x
v
1
+
y
v
2
+
z
v
3
(2.6)
looks like a generalization of the Dirac equation. If we denote by u = div and by
v = curl v, we can rewrite the Lame equation equivalently as:
gradu + curl v = 0, div v = 0. (2.7)
2.3. Linearizing Navier-Stoks equation
The following equations is also a system of inhomogeneous PDEs
v gradp = 0, div v = 0 (2.8)
In fact system of constant coecient PDEs may be considered as PDEs in
algebra, as it will be shown below.
3. PDEs in algebra
An algebra A for us will always be a real n-dimensional Euclidean vector space
that is nitely generated by the orthonormal system of basis vectors e
1
, e
2
, . . . , e
n
and is equipped with a bilinear map m : A A A called multiplication. The
symbol , unless there is ambiguity, stands for the abbreviation m(x, y) as x y
in A. Of course, knowing the tensor form a
m
ij
of a bilinear map in an orthonormal
basis e
1
, e
2
, . . . , e
n
one can rewrite the multiplication m(e
i
, e
j
) as follows:
e
i
e
j
=
n
m=1
a
m
ij
e
m
. (3.1)
We use x = x
1
e
1
+ +x
n
e
n
to denote vector in R
n
as well as element x A, and
i
to denote /x
i
, for i 1, . . . , n. The symbol , ) stands for the Euclidean
scalar product.
Every A-valued function f(x) will be for us always locally real analytic that
is represented as:
f(x) := e
1
u
1
(x
1
, . . . , x
n
) +e
2
u
2
(x
1
, . . . , x
n
) + +e
n
u
n
(x
1
, . . . , x
n
), (3.2)
where u
i
(x) are also real analytic functions, i = 1, 2, . . . , n.
We begin with denition of the Dirac operator in A:
D := e
1
1
+ +e
n
n
. (3.3)
Denition 3.1. A real analytic functionf(x) is called an A-analytic if f(x) is a
solution to the system of partial dierential equations
D f(x) :=
n
i,j=1
e
i
e
j
i
u
j
(x
1
, . . . , x
n
) = 0. (3.4)
Comparing Denition 3.1 with the notion of analytic PDE (2.1), one can
easily verify that the solution to analytic PDE at least locally at point x
0
is an
A
x
0
-analytic function where A
x
0
is a local algebra with the multiplication rule ,
such that
D u(x) := L(x
0
, D)u(x). (3.5)
In this section we will study the qualitative properties of the solution to analytic
PDEs in terms of algebraic properties of the local algebras bundle (cf. (3.5)).
If A in Denition 3.1 is an algebra of complex numbers C, then (3.4) coincides
with the Cauchy-Riemann equations. This gives us a good reason to denote the
space of A-analytic functions as Hol(A) in complete agreement with the denition
of holomorphic functions in complex analysis (denoted by Hol(C)).
Remark 3.2. By A-analysis we mean the systematic study of Hol(A).
3.1. Algebraic approach to function theories
Let A be a real algebra (not necessarily commutative and/or associative). The
literature on function theory over such algebras has been developed by many au-
thors (see [11], [5], [30], [32]) and contains a range of denitions for analyticity
(holomorphicity, monogenicity). Three distinct approaches in these investigations
are mentioned.
The rst one (Weierstrass approach) regards functions on A as their conver-
gence in some sense power series expansions (cf. [16]).
The second (Cauchy-Riemann) approach concentrated on the solution to the
Dirac equation in algebra A (cf [17], [18]).
The third one is based on the function-theoretic properties known for complex
analytic functions, such as Cauchys theorem, residue theory, Cauchy integral
formula etc. (cf. [26], [13] ).
All these methods look like a generalization of analytic function theory of complex
variables (cf. [30], [32]). We use the term A-analysis for such cases (cf. Cliord or
quaternionic analysis [5], [30] if the algebra A is embedded into a Cliord algebra).
Claim 3.3. Totality of a functions on (in general non-commutative and/or non-
associative) regular algebras are splitting into the non-equivalent classes. These
classes are uniquely characterized by their unital hearts. If such a heart is in
addition an associative algebra then an A-analytic function may be expanded into
the commutative operator-valued power series.
3.2. Isotopy classes
In this subsection we will be concerned with Albert isotopies [2] of algebras:
Denition 3.4. Two n-dimensional algebras A
1
and A
2
with multiplication and
are said to be isotopic (A
1
A
2
) if there exist non-singular linear operators
K, L, M such that
x y = M(Kx Ly). (3.6)
Obviously, if in addition, K L M
1
, then two algebras A
1
and A
2
are
isomorphic (A
1
A
2
).
Denition 3.5. If for given two algebras A
1
and A
2
there exist non-singular linear
operators P, Q such that for every g(x) Hol(A
2
) the function f(x) = Pg(Qx)
belongs to Hol(A
1
) and vice versa, we would say that two function theories are
equivalent and write Hol(A
1
) Hol(A
2
).
Using these results, we obtain the important
Theorem 3.6. Two function theories are equivalent i the corresponding algebras
are isotopic.
Denition 3.7. If a
k
ij
(x
0
) in (2.1) for all xed x
0
R
n
forms a set of isotopic
algebra then we will say that L is a unique dened type PDO in , otherwise L is
a mixed type PDO. An algebra A
0
with multiplication tensor a
k
ij
(x
0
) (see. (3.1))
is called a local algebra for L in x
0
.
Obviously, if the coecients a
k
ij
(x) are considered as constants, then the
operator L in (2.1) coincides with the Dirac operator D dened by (3.3) and one
can obtain the solution to the homogeneous equation Lf = 0 as a (left) A-analytic
function for the operator D.
Claim 3.8. Decomposition of algebras into the isotopy classes in turn is a powerful
classication tool for the corresponding PDO.
3.3. Classication of the rst-order PDE
Here we study many essential notions of PDE theory by treating it in algebraic
terms. We begin by examining the conditions that the Dirac operator in A (3.3)
is a well-dened system of PDE.
3.4. Under- and overdetermined system
Let P(D)u(x) = f(x) be a system of partial dierential equations, where P(D) is
a given k l matrix of dierential operators with constant coecients, the given
f(x) (respectively, unknown u(x)) being k- (l-)tuples of functions or distributions
in x R
m
. Many authors (cf. [25]) assume usually that the system is under-
(over-) determined, if the rank of P() (cf. of its transpose P
()) is less than l for

all (cf. for some) non-zero R
m
.
The algebraic formulation of the fact that PDE with the constant coecient
(3.4) is under- (over-) determined can be roughly described as follows.
Denition 3.9. A real n-dimensional algebra A is called left (right) regular if there
exists v A, such that the linear operators L
v
, R
v
: R
n
R
n
dened by x v x
(resp. x x v) are both invertible. Otherwise, A is called a left (right) singular
algebra. In other words, A is regular i A A
2
Recall elements u, v A a left (cf. right annihilator) if u x = 0, (x v = 0)
for all x A.
Theorem 3.10. The Dirac operator D in algebra A is underdetermined i A is
singular and is overdetermined i A is regular and contains an annihilator.
Proof. For a given Dirac operator D in the corresponding algebra A dene left
(right) multiplication operators L
v
, R
v
: R
n
R
n
by x v x (resp. x
x v) as in Denition 3.9. If L
, R
are both invertible for some , then D is well

determined. Conversely, let L
v
(respectively, R
v
) be k
1
l
1
(k
2
l
2
) matrices of
dierential operators. Then D is underdetermined if k
1
< l
1
and/or k
2
> l
2
and
is overdetermined if k
1
> l
1
and/or k
2
< l
2
. The only case k
1
= l
1
= k
2
= l
2
= n
stands for the regular algebras A without annihilators and therefore, for a well-
determined Dirac operator D in R
n
.
In Denition 3.7, we consider some properties of L(x, D) to be of the same
type in a given open set x . These properties were given in terms of the existence
of one common isotopy relation between the set of the corresponding algebras. If
there exists a point on the boundary x
0
and such that an algebra A
0
with
multiplication tensor (3.1) a(x
0
)
k
ij
is not isotopic to any local algebra A
1
with
multiplication tensor a(x
1
)
k
ij
for x
1
, we will say that the PDO L(x, D) is
of degenerate type in . Many authors (see [1], [10]) present several interesting
results on the relationship between solutions in a neighborhood of the boundary
and the properties of the PDO (2.1) of same xed type. The results are too
complicated to be formulated here in detail. In the next section we will deal with
concrete (for example, an elliptic and parabolic type) PDEs and will explain in
terms of corresponding algebras their exceptional algebraic meaning.
3.5. Elliptic type PDE
One of the basic concepts in PDE is ellipticity. Actually, one can reformulate
ellipticity of the Dirac operator in the regular algebra A as a property of A to be
a division algebra.
Denition 3.11. An algebra A is a division algebra i both operations of left and
right multiplications by any non-zero element are invertible.
Proposition 3.12. The well-determined Dirac operator D in the necessary regular
(by Theorem 3.10) algebra A is elliptic i A is a division algebra.
Proof. The matrix symbol (2.3) of the well-determined elliptic partial dierential
operator (D)() is invertible for all ,= 0 (cf. [11], [28]). In contrast, it follows
immediately from the denition of ellipticity that the symbol of partial dierential
operator D is invertible matrix for all ,= 0.
Example. (D. Rusin) Let Q
be constructed from an algebra of quaternions Q

leaving the multiplication table unchanged except the square of i, i
2
= 1 +j.
Two algebras Q
and Q are non-isotopic division algebras if [[ < 2.

The above example shows that in four-dimensional space there exists a non-
equivalent elliptic function theory in the sense of Denition 3.5. In what follows, the
question how many non-equivalent (well-dened elliptic) function theories exists
may be answered in terms of an existence of non-isotopic classes of the division
algebras. In particular, in R
4
there exist well-dened elliptic function theories
dierent from the quaternionic analysis. Our focus will be on the general results
about (not necessary well-determined) elliptic system and therefore we continue
with some generalizations of the condition on an algebra to be a division algebra:
Let the notion / stand for a maximal ideal in A of the left annihilators (that
turn out to be maximal). Of course, the multiplication in / is trivial. Then for a
given algebra A:
Denition 3.13. Recall that an algebra A is a quasi-division algebra i the factor
algebra A// is a division algebra.
In turn, in quasi-division algebra the equation a x = b is soluble for all a, b
except those a, being the left annihilators of A. Now we will be able to formulate
the algebraic analogue of ellipticity [1] for the under- (over-) determined system:
Proposition 3.14. The Dirac operator D dened in (3.3) is elliptic i the algebra
A corresponding to D (see (3.1)) is a quasi-division.
Proof. Immediately follows (cf. [11],[28]) from the denition of ellipticity [1] of the
symbol of the partial dierential operator D.
Of course, if there are no other left/right annihilators except 0, every quasi-
division (in the sense of Denition 3.13) algebra is the division algebra.
For regular algebras the following result of A. Albert [2] is true:
Theorem 3.15. Every regular algebra is isotopic to the algebra with unit e. Every
n 2-dimensional unital division algebra contains an imaginary unit i being a
square root of e, (i
2
= e).
Proof. The proof of the rst part is based on Kaplansky trick [26]. Namely,
let L
x
, (R
y
) be operators of left (right) multiplication on elements x, (y) or x
y L
x
y R
y
x. If A is a regular algebra, then there exist two elements a, b
such that L
a
and R
b
are invertible. Dene a new multiplication by the formula
x y = R
1
b
x L
1
a
y and take the element e = a b. Then
e y = R
1
b
(a b) L
1
a
y = a L
1
a
y = L
a
L
1
a
y = y;
x e = R
1
b
x L
1
a
(a b) = R
1
b
x b = R
b
R
1
b
x = x
for all x, y. Combining Denition 3.13 with Proposition 2 yields that the same
statement as in Theorem 3.15 is true for a quasi-division algebras if codim(A) 2.
Moreover, in this case in A there exists a 2-dimensional subalgebra /
= C.
Example. Consider two systems of PDE, the spherical and usual Dirac equation
[21] in R
4
for scalar and vector functions u(t, x), v(t, x):
divv = 0; gradu + curl v = 0; (3.7)
t
u div v = 0; gradu +
t
v + curl v = 0. (3.8)
Both systems of PDE are elliptic, the rst one is overdetermined. To see that,
let us recall that the algebra of quaternion Q stands behind (3.8), and behind (3.7)
stands an algebra Q
with multiplication rule x y = 1/2(x x) y where is

the quaternion multiplication and x is the quaternion conjugation to x.
In turn, both algebras are quasi-division, but only Q
has non-trivial ideal of

their (left) annihilators.
3.6. Parabolic and hyperbolic type PDE
Comparing the denitions of elliptic [1], hyperbolic [28] and parabolic [10] PDE
with results of Theorem 3.10, we can see that parabolic and hyperbolic type Dirac
operators correspond to an algebra with zero divisors. Moreover, the property of
the algebras to be regular and quasi-division, the number of zero divisors and the
number of annihilators are invariants with respect to isotopy relation. This gives
rise to the following:
Proposition 3.16. The Dirac operator in a regular algebra A is:
(i) parabolic [10] i A// contains exactly one (up to scalar factor) zero divisor.
In this case A// is isotopic to an algebra with one nilpotent and no other
zero divisors;
(ii) hyperbolic [28] i A// contains at least one pair or more of (left) zero divi-
sors.
Below we formulate some common results about building an analytic function
theory in R
n
.
Claim 3.17. An A-analytic function theory equivalent (in the sense of Denition
3.5) to the solution space of an evolution equation i A is regular.
Proof. If A is regular then there exists an isotopy to an algebra with unit. But the
Dirac operator in algebra with unit 1 may be written in the form D =
t
+P(
x
).
Therefore D is of evolution type and we can apply methods of Section 1.
3.7. Applications
Let t, x, y, z be independent variables. Denote by u(t, x, y, z) the scalar and by
v(t, x, y, z) vector function in R
4
. Many physically important models may be writ-
ten as the following PDE:
t
u +div v = 0,
t
v +gradu +curl v = 0. (3.9)
Assume A is behind (3.9) and D is the Dirac operator in A. In turn, if
< 0, and ,= 0 then the system of PDE (3.9) is elliptic and the algebra A is
a division algebra.
If > 0 and ,= 0, then the system of PDE (3.9) is hyperbolic and
the algebra behind Dirac operator (3.9) is an algebra with innitely many zero
divisors.
If = 0, = 0 and ,= 0, then an algebra behind (3.9) is a quasi-division
algebra embedded into quaternions Q. Actually, if A is unital algebra, then one
can set = = = 1 and therefore sign of for ,= 0 distinguish elliptic and
hyperbolic cases.
It is convenient to consider the case n = 2 separately.
3.8. Two-dimensional case
Remark 3.18. It follows from Theorem 3.15 and ([26], Proposition 1.7) that all
two-dimensional division algebras are isotopic to C.
In particular, the function theory over any division algebras in R
2
is equiv-
alent to the complex analytic function theory. Many concepts and results in the
theory of analytic functions of complex variables can be extended to functions
satisfying the system (3.4)
Theorem 3.19. (cf. [26]) There are only three non-equivalent function theories over
two-dimensional regular algebras, namely:
(i) elliptic theory over C, which corresponds to complex analytic function theory;
(ii) parabolic (degenerate), f(x
1
, x
2
) = u(x
2
)e
1
+ (v(x
2
) x
1
u
(x
2
))e
2
in N and
e
2
1
= e
1
, e
1
e
2
= e
2
e
1
= e
2
, e
2
2
= 0;
(iii) hyperbolic, if f(x
1
, x
2
) = u(x
2
)e
1
+v(x
1
)e
2
in RR and e
2
1
= e
1
, e
1
e
2
=
e
2
e
1
= 0, e
2
2
= e
2
.
Proof. By the Unital Heart Proposition (cf. [26], 1.10) there exists exactly three
isotopy classes of regular two-dimensional real algebras. (The singular algebras are
out of the realm of our interest since the function theory over these algebras are
one-dimensional and therefore trivial.)
Remark 3.20. The algebras C, N and R R listed in Theorem 3.19 are the only
unital two-dimensional algebras. They are necessary commutative and associative
and have respectively 0, 1 and 2 zero divisors.
The following denition is about the construction of an analytic function
theory using algebraic operations only.
Denition 3.21. If for every two functions f(x), g(x) Hol(A) the product f(x)
g(x) as well as their linear combination f(x) + g(x) with real coecients ,
belongs to Hol(A), then we will call the A-analytic function theory pure algebraic.
Theorem 3.22. A-analytic function theory is pure algebraic i A is a direct sum
of R, C and N dened in Theorem 3.19.
Proof. The if part is based on results in [20] concerning existence of an innite-
dimensional symmetries in Hol(A) only in two-dimensional subalgebras and/or
in their direct sum. Every regular A-analytic function theory in algebras A being
represented as a direct sum of the two-dimensional subalgebras is equivalent in
the sense of Denition 3.5 to the pure algebraic theory. The only if part follows
from Theorem 3.19.
The operator of multiplication by a xed A-analytic function in the algebras
A mentioned in Theorem 3.19 maps the space Hol(A) onto itself. This statement is
related to the symmetries of the Dirac operator. We refer to [18] for results about
symmetries in Cliord algebras.
Claim 3.23. In R
2
every A-analytic function theory is equivalent (in the sense of
Denition 3.5) to the pure algebraic theory.
4. Power series expansions
As was proven in [26], every regular algebra is isotopic to its unital heart. Assume
that e
0
, e
1
, . . . , e
n
forms a basis in the unital associative algebra A and e
0
is its
two-sided unit element. In order to construct an A-analytic function theory, the
following A-analytic variable are used:
z
m
= x
m
e
0
+x
0
e
m
, m = 1, 2, . . . , n. (4.1)
In turn, Dz
k
= 0 for all k = 1, 2, . . . , n where D =
x
0
+e
1
x
1
+ +e
n
x
n
is the Dirac operator in the algebra A
Denote a canonical spherical homogenic polynomial solution of the Dirac
equation in A by the formula:
V
0
(x) = e
0
, V
m
(x) = z
m
, m = 1, 2, . . . , n, (4.2)
V
(x) := V
m
1
,...,m
k
(x) =
1
k!
(m
1
,...,m
k
)
z
m
1
z
m
2
z
m
k
, (4.3)
where the sum runs over all distinguishable permutations of m
1
, . . . , m
k
.
Proposition 4.1. (cf. [5]) The polynomials V
(x) of order k for all multi-indices

= m
1
, . . . , m
k
for m
i
1, 2, . . . , n are both left and right A-analytic. Any
A-analytic and homogenic of order k function p
k
(x) may be written as
p
k
(x) =
m
1
,...,m
k
V
m
1
,...,m
k
(x)
x
m
1

x
m
k
p
k
(x) (4.4)
where the sum runs over all possible combinations of m
1
, . . . , m
k
of k elements out
of 1, 2, . . . , n repetition being allowed.
Proof. (cf. [5], Theorem 11.2.3,5) Clearly, for = m
1
, . . . , m
k
k!x
0
DV
(x) =
(m
1
,...,m
k
)
n
i=0
x
0
e
i
x
i
(z
m
1
(z
m
2
z
m
k
) . . .)
= x
0
n
i=1
()
(e
m
i
z
m
1
z
m
i1
z
m
i+1
z
m
k
z
m
1
z
m
i1
e
m
i
z
m
i+1
z
m
k
)
=
n
i=1
()
(z
m
i
(z
m
1
z
m
i1
(z
m
i+1
z
m
k
) . . .) z
m
1
z
m
k
) = 0.
For the polynomial p
1
(x) in (4.4) by Eulers formula, one can show that
p
1
(x) = x
0
x
0
p
1
(x) +
n
i=1
x
i
x
i
p
1
(x) =
n
i=1
z
i
x
i
p
1
(x)
To continue the proof one can use induction.
Theorem 4.2. (cf. [5]) The function f(x) that is A-analytic in an open neighbor-
hood of the origin can be expanded into a normally convergent series of spherical
homogenic polynomials
f(x) =
k=0
_

m
1
,...,m
k
V
m
1
,...,m
k
(x)
x
m
1

x
m
k
f(0)
_
Proof. The proof is similar to the method described in ([5], Theorem 11.3.4) and is
a generalization of results [5] to the general unital associative algebra if the series
by spherical harmonics are convergent.
Claim 4.3. The polynomials V
(x) introduced in (4.1)(4.3) play an analogous role

as the powers of the complex variable z = x+iy in the theory of complex variables.
4.1. Symmetries
Let A be a unital associative algebra and D is the Dirac operator in A.
Theorem 4.4. The rst-order PDO L C[x,
x
] is a symmetry operator for D
dened in (3.3) i there exists a rst-order PDO L
C[x,
x
] such that
DL = L
D. (4.5)
Proof. Suciency of (4.5) holds trivially. To prove necessity, without loss of gen-
erality, choose a 1-symmetry operator L in the form
L = a
0
(x)D +
m
k=1
a
k
(x)
x
k
+b(x).
Here a
k
(x), k = 0, . . . , m, b(x) are C
functions.
Let now L
be chosen such that the operator R

L
= DL L
D does not
depend on
x
0
, for example, in the following form
L
= Da
0
(x) +
m
k=1
a
k
(x)
x
k
+b(x).
Then, obviously
R
L
f(x) = D(Lf(x)) L
(Df(x))
=
m
k,l=1
a
kl
(x)
x
k
x
l
f(x) +
m
k=1
b
k
(x)
x
k
f(x) +c(x)f(x). (4.6)
All these functions a
kl
(x), b
k
(x), c(x) for k, l = 1, 2, . . . , m. For (4.6) to be valid
when applied to an arbitrary f(x), it is necessary and sucient that the coef-
cients of R
L
f 0 in . That means, for f(x) 1 (every constant is a trivial
monogenic function), that c(x) 0. Taking consequently x
k
x
0
e
k
, k = 1, 2, . . . , m
as a monogenic function f(x) in (4.6), we get that all Cliord-valued coe-
cients b
k
(x) 0, k = 1, 2, . . . , m. Now if f(x) is chosen equal to one of the
following functions: x
2
0
x
2
k
+ 2x
0
x
k
e
k
, k = 1, 2, . . . , m, (actually hypercom-
plex second-order powers, that are, obviously, monogenic), we get a
kk
(x) 0,
k = 1, 2, . . . , m. In order to complete the proof, we can substitute the follow-
ing monogenic functions f(x) in (4.6): x
2
1
x
2
2
2x
1
x
2
e
12
, x
2
1
x
2
3
2x
1
x
3
e
13
,
. . . , x
2
m1
x
2
m
2x
m1
x
m
e
m1,m
. Now all the remaining coecients of R
L
are
equal to 0. The theorem is proven for L
= L
. Moreover, since all coecients

a
kl
(x), b
k
(x), c(x) of R
L
(x) in (4.6) are equal to 0, this leads to the following sys-
tem of partial dierential equations:
e
l
a
k
(x) a
k
(x)e
l
+e
k
a
l
(x) a
l
(x)e
k
= 0, k, l = 1, 2, . . . , m, (4.7)
Da
k
(x) + e
k
b(x) b(x)e
k
= 0, k = 1, 2, . . . , m, (4.8)
Db(x) = 0. (4.9)
The last equation (4.9) means that b(x) is a monogenic function. The solution
of (4.7) which can be easily veried is
a
k
(x) =
[I[ = even,
k , I
I
(x)e
I
e
k
+
[I[ = odd,
k I
I
(x)e
I
e
k
. (4.10)
Here
I
(x) are real analytic functions dened for all multi-indices I and such
that they all are independent of k.
After putting (4.10) in the remaining equations (4.8) and using the following
notations:
The last equation (4.9) means that b(x) is a monogenic function. The solution
of (4.7) which can be easily veried is
a
k
(x) =
[I[ = even,
k , I
I
(x)e
I
e
k
+
[I[ = odd,
k I
I
(x)e
I
e
k
. (4.11)
Here
I
(x) are real analytic functions dened for all multi-indices I and such
that they all are independent of k.
4.2. Quaternion analysis
From the point of view of quaternion analysis (cf. [5]) the entire smooth enough
(dierentiable in neighborhood of origin) solution of the Dirac equation can be
represented as a convergent series of quaternion harmonics dened in X R
4
.
They are the only homogeneous polynomial solutions of degree m to the Dirac
equation and
Y
m
(q) =
]]=m
c
, m = 0, 1, . . . , D Y
m
(q) = 0, (4.12)
where q = x
1
+ x
2
i + x
3
j + x
4
k, is multi-index, [[ =
1
+
2
+
3
+
4
,
x
= x
1
1
x
2
2
x
3
3
x
4
4
, and c
are the quaternion-valued constants.

Theorem 4.5. [17] The quaternion harmonics fulll the relation
2m(m+ 1)Y
m
(q) =
4
i=1
K
i
[
x
i
Y
m
(q)], (4.13)
where K
1
, . . . , K
4
are the generators of the conformal group in quaternion space
K
1
= (x
2
1
x
2
2
x
2
3
x
2
4
)
x
1
+ 2x
1
x
2
x
2
+ 2x
1
x
3
x
3
+ 2x
1
x
4
x
4
+ 2x
1
+q;
K
2
= (x
2
2
x
2
1
x
2
3
x
2
4
)
x
2
+ 2x
2
x
1
x
1
+ 2x
2
x
3
x
3
+ 2x
2
x
4
x
4
+ 2x
2
iq;
K
3
= (x
2
3
x
2
2
x
2
1
x
2
4
)
x
3
+ 2x
3
x
2
x
2
+ 2x
3
x
1
x
1
+ 2x
3
x
4
x
4
+ 2x
3
jq;
K
4
= (x
2
4
x
2
2
x
2
3
x
2
1
)
x
4
+ 2x
4
x
2
x
2
+ 2x
4
x
3
x
3
+ 2x
4
x
1
x
1
+ 2x
4
kq.
Proof. Let r
2
= x
2
1
+x
2
2
+x
2
3
+x
2
4
and let operator r
r
be the generator of dilatations
in R
4
:
r
r
:= x
1
x
1
+x
2
x
2
+x
3
x
3
+x
4
x
4
.
In fact, Y
m
(q) is the eigenfunction of the operator r
r
with eigenvalue m:
r
r
Y
m
(q) = mY
m
(q). Using these denitions, we obtain
3
i=0
K
i
[
x
i
Y
m
(q)] = 2m(m1)Y
m
(q) r
2
Y
m
(q) + 4mY
m
(q) +qD Y
m
(q).
Here q quaternion conjugate to q, is the Laplace operator. If Y
m
(q) is a quater-
nion harmonics, then Y
m
(q) = 0, D Y
m
(q) = 0. The theorem is proven.
Now we will be able to prove the main result for quaternion harmonics:
Theorem 4.6. The homogeneous polynomial Y
m
(q) is a quaternion harmonics i
Y
m
(q) := P
m0
(K)[1] +P
m1
(K)[i] +P
m2
(K)[j] +P
m3
(K)[k], (4.14)
where P
mi
are homogeneous of the same order m real polynomials.
Proof. Quaternion harmonics
x
i
Y
m
is also a quaternion harmonics of order m1.
By induction, using (4.13) we can show that
Y
m
(q) =
1
2
m
(m + 1)!
]]=m
1
!
K
[c
], c
Y
m
(0). (4.15)
Then (4.14) holds for polynomials
P
mi
(x) =
1
2
m
(m+ 1)!
]]=m
1
!
c
i
x
. (4.16)
Here c
= c
0
+c
1
i +c
2
j +c
3
k.
4.3. Cliord analysis
Dene the Cliord algebra Cl
0,n
Alg(R
2
n
) as associative unital algebra freely
generated by R
n
with usual inner product x, y) modulo the relation x
2
= [[x[[
2
for all x R
n
. Equivalently, the Cliord algebra Cl
0,n
is generated by the or-
thonormal basis e
0
, e
1
, . . . , e
n
in R
n+1
, and all theirs permutations. Here e
0
is
a unit element and e
i
satises the relationships e
i
e
j
+ e
j
e
i
= 2e
i
, e
j
)e
0
for
1 j n. More details on Cliord algebras can be found in [5], [30].
Below we present some results from [18].
There are exactly four classes of 1-symmetry operators for the Dirac operator
D in Cl
0,n
, namely:
the generators of the translation group in R
n+1
x
k
, k = 0, 1, . . . , n; (4.17)
the dilatations
R
0
= x
0
x
0
+x
1
x
1
+ +x
n
x
n
+
n
2
; (4.18)
the generators of the rotation group
J
ij
= J
ji
= x
j
x
i
x
i
x
j
+
1
2
e
ij
, i, j = 1, 2, . . . , n, i ,= j
J
i0
= J
0i
= x
0
x
i
x
i
x
0
+
1
2
e
i
, i = 1, 2, . . . , n; (4.19)
and the generators of the conformal group
K
i
=
n
s=0
2x
i
x
s
x
s
xx
x
i
+ (n + 1)x
i
xe
i
, (4.20)
for i = 0, 1, . . . , n. Here x = x
0
+ x
1
e
1
+ + x
n
e
n
and x are conjugate in the
sense of Cliord-valued functions.
Using these basic 1-symmetries we can construct the Cliord-valued opera-
tor indeterminates K A in the space Hol(Cl
0,n
) as operator action similar to
multiplication on x a. Namely, let a = a
1
e
1
+ + a
n
e
n
and a be conjugate
in the sense of Cliord algebra. Dene A = A
0
+ A
1
e
1
+ + A
n
e
n
and A
i
for
i = 0, 1, . . . , n where
A
i
= 2
m
j,=i,j=0
a
j
J
ji
2a
i
m
j=0
a
j
x
j
+ 2a
i
R
0
+aa
x
i
.
Theorem 4.7. All Cl
0,n
-analytic polynomial functions f(x) can be represented in
the neighborhood of a given point a in the form
u(x) = U
0
(K
0
A
0
, . . . , K
n
A
n
)[1] + +U
i
(K
0
A
0
, . . . , K
n
A
n
)[e
i
],
where U
i
(x), i = 0, 1, 2, . . . , 2
n
, are real homogeneous polynomials being factorized
by the relation x
2
0
+x
2
1
+ +x
2
n
= 0.
Proof. The proof is analogous [18] to the proof of Theorem 4.6.
Claim 4.8. The Cliord-valued analytic functions have a unique power series ex-
pansion in pairwise commutative operator independents K = K
0
, . . . , K
n
.
4.4. Axial monogenic polynomials
A Cliord analytic function u(x) is called axial symmetric if u(x) = F(K
0
)[x].
It was shown in [18] that all axial Cliord-valued analytic functions u(x) may
be represented in the form
Proposition 4.9. Any axial symmetric Cliord analytic function has a structure
u(x) = (x
0
, r) +(x
0
, r)x,
where r is a length of radius vector, r
2
= xx and (p, q), (p, q) are generalizing
analytic functions in sense of [32], namely:
p
(p, q) +p
p
(p, q) = q
q
(p, q) + (n 1)(p, q),
q
(p, q) +p
q
(p, q) +q
p
(p, q) = 0.
Table 1. Cliord analytic function in the axial symmetric case.
(p, q) (p, q) Cliord analytic function
(n 1)p 1 K
0
[1] = (n 1)x
0
+x
(n 2)p
2
2p K
2
0
[1] = (n 2)x
2
0
r
2
+ 2x
0
x
2p1
(12p+q
2
)
n2
2p1
(12p+q
2
)
n2
e
(K
0
)
[1] =
2x
0
1x
(12p+q
2
)
n2
Example. In [18] was dened the homogeneous monogenic polynomials in x
p
l
(x) =
i+j=l
((n 1)/2)
i
i!
((n + 1)/2)
j
j!
x
i
x
j
.
One can explicitly verify that some axial monomials can be written in the
form p
l
(x) = K
l
0
[a]. Now one can dene the multiplication rule in space of these
axially symmetric monomials. Namely, by the denition
p
l
(x) p
m
(x) = K
l
0
(K
m
0
[a]) = K
l+m
0
[a] = p
l+m
(x).
Further, in [18] some commutation relations between the 1-symmetry operators
are considered.
The knowledge of symmetries gives the key to the study of the structure of
monogenic functions. It is well known that monogenic Cliord-valued functions
can be represented as convergent series of monogenic homogeneous polynomial
functions. The problems of the representation of the axial symmetric monogenic
functions therefore may be solved by the straightforward evaluation of Cliord
analytic functions in operator indeterminates. (See examples in Table 1.)
5. Polynomial solutions to homogeneous PDE
First we clarify the structure of the exponentially-polynomial solutions to constant
coecient homogeneous PDE:
Q(
x
)u(x) :=
]]=m
a
x
u(x
1
, . . . , x
n
) = 0. (5.1)
Here a
are real for all multi-indices , and

x
stands for dierentiation. Clearly,
for m 2, (5.1) is not an evolution equation and, therefore, technique (1.6) is not
straightforward applicable.
The PDE (5.1) is said to be a homogeneous PDE of order m. Obviously, a
homogeneous PDE admits an innitesimal dilatation R
0
= x
1
1
+ +x
n
n
.
Denote by d
k,n
the dimension of the space of all homogeneous polynomials
T
k
of order k in R
n
. Thus
d
k,n
= dimT
k
=
_
n +k 1
k
_
. (5.2)
Theorem 5.1. The homogeneous PDE (5.1) admit the exponential-polynomial so-
lution u = e
<,x>
p
k
(x), p
k
(x) T
k
in C
n
, n 2 i Q() = 0, C
n
.
Proof. If = 0 and the polynomial p
k
(x) is of order 0 k < m the proof is
trivial. For k m and = 0, we have no more than d
km,n
linear algebraic
conditions on coecients of p
k
(x) T
k
in order to fulll (5.1). So the space of
the homogeneous polynomial solutions to (5.1) of order k is at least of dimension
d
Q,k
d
k,n
d
km,n
.
If ,= 0, it is possible to construct the exponential-polynomial solution in
the form u = p
k
(x)e
<,x>
only if p
k
(x) and e
<,x>
both are solutions. In fact is
(necessarily) a root of Q() = 0 with multiplicity l < m and
]]=
]]=m
_

1
1
_
. . .
_

n
n
_
x
p
k
(x)
= 0,
for = l + 1, . . . , m.
Actually, the symmetries are considered as transformations on the solution
space of a DE. However, the solution space is not usually known a priori. The
straightforward computations (cf. [20]) of symmetries turn out to be equivalent to
the solvability of the overdetermined system of PDEs.
Question 5.2. Under what condition is the overdetermined system of PDEs arising
in context of Proposition 0.10 non-trivially solvable?
To proceed with homogeneous PDE, for which we shall build the solution
space explicitly, consider
5.1. Second-order PDE
Assume Q := Q(
x
) to be the second-order PDO associated with the quadratic
form Q(x) = x
T
Ax in R
n
, (n > 2):
Q(
x
)u(x) :=
n
i,i=1
a
ij
x
i
x
j
u(x) = 0. (5.3)
Suppose the matrix A = a
ij
in (5.3) is not singular. Denote by P(x) the qua-
dratic form P(x) := x
T
A
1
x.
Theorem 5.3. Let (p, n p) denote the signature of the quadratic form Q(x) in
(5.3), (i.e., p is the number of positive entries and m is the number of negative
entries in a diagonalization). Then the space of 1-symmetries operators of (5.3)
forms N = (n+1)(n+2)/2-dimensional pseudo-orthogonal Lie algebra isomorphic
to so(p +1, np +1) (cf. Denition 0.13). The basis in ool
Q
(R
n
) (apart from the
trivial identity symmetry) consists of the following N operators:
(i) n generators of translation group in R
n
D
i
=
x
i
, i = 1, . . . , n; (5.4)
(ii) the generator of dilatation
R
0
= x
1
x
1
+x
2
x
2
+ +x
n
x
n
+
n 2
2
; (5.5)
(iii) n(n 1)/2 generators of the rotation (Lorentz) group
J
ij
=
1
4
(P
i
(x)Q
j
() P
j
(x)Q
i
()) i ,= j = 1, . . . , n; (5.6)
(iv) and the n generators of the special (pseudo-) conformal group
K
i
= (n 2)x
i
1
2
_
P(x)Q
i
() x
i
n
j=1
P
j
(x)Q
j
()
_
. (5.7)
Proof. Without loss of generality, assume Q(x) is already reduced to its canonical
diagonal form and let (p, n p) be the signature of Q(x). Hence, we can split
the coordinates x
1
, . . . , x
n
onto two subsets, such that p coordinates with positive
entries come rst
Q(
x
) :=
n
i=1
2
x
i
, here
i
=
_
1 i p ,
1 i > p .
(5.8)
Choose the general 1-symmetry operator for Q(
x
) as follows:
L(x, ) :=
n
i=1
i
b
i
(x)
x
i
+c(x). (5.9)
Following Proposition 0.10, in order to construct all symmetries of Q it is
enough to nd the function R(x) such that QL LQ = R(x)Q. Thus
x
j
b
i
(x) +
x
i
b
j
(x) = 0,
i
x
i
b
i
(x) =
j
x
j
b
j
(x) = R(x), (5.10)
for all i ,= j 1, . . . , n and
Q[b
i
(x)] + 2
x
i
c(x) = 0, Q[c(x)] = 0. (5.11)
for all i.
Applying results of Proposition 2.2 to the system (5.10) yields that there
exists only a nite set of solutions to (5.10). We nd the 1-symmetry operators to
Q with (2.5). This gives:
D
i
=
x
i
, i = 1, . . . , n, (5.12)
R
0
= x
1
x
1
+ +x
n
x
n
+ (n 2)/2, (5.13)
J
ij
=
i
x
i
x
j

j
x
j
x
i
, i, j = 1, . . . , n, (5.14)
K
i
= (n 2)x
i
j=1
j
x
2
j
x
i
+x
i
n
j=1
j
x
j
x
j
, i = 1, . . . , n. (5.15)
There are some additional important relations between operators (5.12)
(5.15) in oym(Q). For example, it can be easily veried that
Q(K) Q(x)
2
Q(
x
). (5.16)
Actually, Q(K) is a trivial symmetry operator (Q(K) 0).
Using the Denition 0.13, one can verify that the same commutator relations
for both of the Lie algebras oym(Q) and so(p+1, np+1) holds after the following
correspondences:
D
i
:=
1,i+1
+
i,n+2
, R
0
:=
1,n+2
, K
i
:=
1,i+1
i,n+2
, J
ij
:=
i+1,j+1
.
Finally, after back substitution we obtain the required condition on coecients of
L in the form (5.4)(5.7).
Theorem 5.4. (cf. [17]) Let u(x) be any locally analytic solution to the homogeneous
PDE (5.3). Then there exists a one-to-one correspondence between u(x) and the
entire function f(x) such that
u(x) = f(K
1
, . . . , K
n
)[1], (5.17)
where K
i
was dened in (5.7) up to choice of basis.
5.2. Laplace and biharmonic equation
Let
x = x
1
, . . . , x
n
R
n
and r =
_
x
2
2
+ +x
2
n
.
Consider axial symmetric harmonics u = u(x
1
, r) in R
n
:
u(x
1
, r)
2
x
1
u(x
1
, r) +
2
r
u(x
1
, r) +
n 2
r

r
u(x
1
, r) = 0
By Theorem 5.4 any axial symmetric harmonic function u(x
1
, r) in R
n
may be
represented in the form u(x
1
, r) = F(K
1
)[1]. The action of the symmetry operator
K
1
from (5.7) on the function u(x
1
, r) is dened as follows:
K
1
[u(x
1
, r)] := (n 2)x
1
u(x
1
, r) + (x
2
1
r
2
)
x
1
u(x
1
, r) + 2x
1
r
r
u(x
1
, r).
Example. Let f(x) be represented as a formal power series. Then the solution of
the boundary value problem
u(x
1
, r) = 0, u(x
1
, 0) = f(x
1
), f(x) =
m=1
a
m
x
m
may be written as u(x
1
, r) = F(K
1
)[1] where F(x) is as follows
F(x) =
m=1
(n 3)!
(n 3 + m)!
a
m
x
m
.
Table 2. Some harmonic functions in R
3
and their operator form.
u(x, 0) F(x) Harmonic function representation
x
m 1
m!
x
m 1
6
K
3
1
[1] := x
3
1
3r
2
/2
1
1x
e
x
e
K
1
[1] := 1/
_
(x
1
1)
2
+ r
2
e
x
I
0
(2
x) I
0
(2
K
1
)[1] := e
x
1
J
0
(r)
We nish with the solution of the biharmonic equation
2
u = 0 as follows:
u(x
1
, . . . , x
n
) = f(K
1
, . . . , K
n
)[1] +
n
i=1
x
i
g
i
(K
1
, . . . , K
n
)[1].
Conclusions
Every rst-order PDO with constant coecient is the Dirac operator in the
corresponding algebra.
The solution to the Dirac equation in isotopic algebras forms an equivalent
function theory.
The A-analysis in the regular algebras is equivalent to the canonical function
theory on their unital hearts.
Acknowledgments
The authors gratefully acknowledge fruitful conversations with Yuli Eidelman con-
cerning the results of this article. The authors are grateful to an anonymous referee
for important remarks and suggestions.
References
[1] S. Agmon, A. Douglis, L. Nirenberg, Estimates near the boundary for solutions of
elliptic partial dierential equations satisfying general boundary conditions, I, II,
Comm. Pure Applied Math. 12, (1959), 623727, 17, (1964), 3592.
[2] A.A. Albert Nonassociative algebras. Ann. of Math. 43, 1942, pp. 685707
[3] G.W. Bluman, J.D. Cole, Similarity methods for dierential equations. Applied
Mathematical Sciences, 13. Springer-Verlag, New York, 1974, 332 pp.
[4] G.W. Bluman, S. Kumei, Symmetries and dierential equations. Applied Mathemat-
ical Sciences, 81, Springer-Verlag, New York, 1989, 412 pp.
[5] F. Brackx, R. Delanghe, F. Sommen, Cliord Analysis, Pitman Research Notes in
Math. 76, 1982, 308 pp.
[6] J.R. Cannon, The one-dimensional heat equation. Encyclopedia of Math. and its
Appl., 23. Addison-Wesley, MA, 1984, 483 pp.
[7] A. Coman, D. Legg, Y. Pan, A Taylor series condition for harmonic extensions.
Real Anal. Exchange 28 (2002/03), no. 1, pp. 235253.
[8] M.J. Craddock, A.H. Dooley, Symmetry group methods for heat kernels. J. Math.
Phys. 42 (2001), no. 1, pp. 390418.
[9] L. Ehrenpreis, A fundamental principle for systems of linear equations with constant
coecients. in Proc. Intern. Symp. Linear Spaces, Jerusalem, 1960, pp. 161174
[10] S. Eidelman, Parabolic systems, North-Holland Publishing Company, 1969, 469 pp.
[11] R.P. Gilbert, J.L. Buchanan, First Order Elliptic Systems, Mathematics in Science
and Engineering; 163, Academic Press, 1983, 281 pp.
[12] G.R. Goldstein, J.A. Goldstein, E. Obrecht, Structure of solutions to linear evolution
equations: extensions of dAlemberts formula. J. Math. Anal. Appl. 201, (1996), no.
2, pp. 461477.
[13] I.J. Good, A simple generalization of analytic function theory. Exposition. Math, 6,
no. 4, 1988, pp. 289311.
[14] G.N. Hile, A. Stanoyevitch, Expansions of solutions of higher order evolution equa-
tions in series of generalized heat polynomials, Vol. 2002, No. 64, pp. 125.
[15] L. Hormander, The analysis of linear partial dierential operators II, Springer Verlag,
Berlin, 1983.
[16] P.W. Ketchum, Analytic functions of hypercomplex variables, Trans. Amer. Mat.
Soc., 30, # 4, 1928, pp. 641667.
[17] Y. Krasnov, Symmetries of Cauchy-Riemann-Fueter equation, Complex Variables,
vol. 41, 2000, pp. 279292.
[18] Y. Krasnov, The structure of monogenic functions, Cliord Algebras and their Ap-
plications in Mathematical Physics, vol. 2, Progr. Phys., 19, Birkh auser, Boston,
2000, pp. 247272.
[19] B. Malgrange, Sur les systèmes dierentiels à coecients constants. (French) 1963
Les

Equations aux Derivees Partielles (Paris, 1962) pp. 113122
[20] W. Miller, Symmetry and Separation of Variables, Encyclopedia of Mathematics
and its Applications, Addison-Wesley, 4, 1977, 285 pp.
[21] P.J. Olver, Applications of Lie Groups to Dierential Equations. Graduate Texts in
Mathematics 107, Springer, New York, 1993.
[22] P.J. Olver, Symmetry and explicit solutions of partial dierential equations. Appl.
Numer. Math. 10 (1992), no. 34, pp. 307324.
[23] P.J. Olver, V.V. Sokolov, Integrable Evolution Equations on Associative Algebras,
Commun. Math. Phys. 193, (1998), pp. 245268
[24] L.V. Ovsiannikov, Group analysis of dierential equations. English translation. Aca-
demic Press, Inc., New York-London, 1982. 416 pp.
[25] V.P. Palamodov, A remark on the exponential representation of solutions of dier-
ential equations with constant coecients. Mat. Sb. 76 (118) 1968, pp. 417434.
[26] P.S. Pedersen, Cauchys integral theorem on a nitely generated, real, commutative,
and associative algebra. Adv. Math. 131 (1997), no. 2, pp. 344356.
[27] P.S. Pedersen, Basis for power series solutions to systems of linear, constant coe-
cient partial dierential equations. Adv. Math. 141 (1999), no. 1, pp. 155166.
[28] I.G. Petrovsky, Partial Dierential Equations, CRC Press, Boca Raton, 1996.
[29] S.P. Smith, Polynomial solutions to constant coecient dierential equations. Trans.
Amer. Math. Soc. 329, (1992), no. 2, pp. 551569.
[30] F. Sommen, N. Van Acker, Monogenic dierential operators, Results in Math. Vol.
22, 1992, pp. 781798.
[31] F. Treves, Linear partial dierential operators, 1970.
[32] I. Vekua, Generalized analytic functions. London. Pergamon, 1962.
[33] D.V. Widder Analytic solutions of the heat equation. Duke Math. J. 29, 1962, pp.
497503.
Samuil D. Eidelman
Solomonov University
Kiev, Ukraine
e-mail: seidelman@amath.kiev.ua
Yakov Krasnov
Bar-Ilan University
Ramat-Gan, 52900, Israel
e-mail: krasnov@math.biu.ac.il
Operator Theory:
On the Bessmertny Class of Homogeneous
Positive Holomorphic Functions
on a Product of Matrix Halfplanes
Dmitry S. Kalyuzhny-Verbovetzki
Abstract. We generalize our earlier results from [9] on the Bessmertny class
of operator-valued functions holomorphic in the open right poly-halfplane
which admit representation as a Schur complement of a block of a linear ho-
mogeneous operator-valued function with positive semidenite operator coef-
cients, to the case of a product of open right matrix halfplanes. Several equiv-
alent characterizations of this generalized Bessmertny class are presented. In
particular, its intimate connection with the AglerSchur class of holomorphic
contractive operator-valued functions on the product of matrix unit disks is
established.
Mathematics Subject Classication (2000). Primary 47A48; Secondary 32A10,
47A56, 47A60.
Keywords. Several complex variables, homogeneous, positive, holomorphic,
operator-valued functions, product of matrix halfplanes, long resolvent repre-
sentations, AglerSchur class.
1. Introduction
In the PhD Thesis of M.F. Bessmertrny [4] (the translation into English of some
of its parts can be found in [5, 6, 7]) the class of rational n n matrix-valued
functions of N complex variables z = (z
1
, . . . , z
N
) C
N
, representable in the form
f(z) = a(z) b(z)d(z)
1
c(z), (1.1)
where a linear (n +p) (n +p) matrix-valued function
A(z) = z
1
A
1
+ +z
N
A
N
=
_
a(z) b(z)
c(z) d(z)
_
(1.2)
The author was supported by the Center for Advanced Studies in Mathematics, Ben-Gurion
University of the Negev.
140 D.S. Kalyuzhny-Verbovetzki
has positive semidenite matrix coecients A
j
, j = 1, . . . , N, with real entries,
was considered. Such a representation (1.1)(1.2), was called in the thesis a long
resolvent representation. The motivation of its consideration comes from the elec-
trical engineering. Bessmertrny has shown that this class is precisely the class of
all characteristic functions of passive 2n-poles, where the impedances of elements
of an electrical circuit are considered as independent variables.
In [9] a more general class B
N
(|) of holomorphic functions on the open right
poly-halfplane
N
:= z C
N
: Re z
k
> 0, k = 1, . . . , N, with values in the
C
-algebra L(|) of bounded linear operators on a Hilbert space |, which admit

a representation (1.1) with a linear pencil A(z) as in (1.2), however consisting
of operators from L(| H) where H is an auxiliary Hilbert space, such that
A
j
0, j = 1, . . . , N, was introduced. Here the Hilbert spaces are supposed to
be complex. This class B
N
(|) was called the Bessmertny class. Any function
f B
N
(|) is homogeneous of degree one and takes operator values with positive
semidenite real parts. Moreover, f can be uniquely extended to a holomorphic
and homogeneous of degree one function on the domain
N
:=
_
T
()
N
C
N
, (1.3)
so that (1.1) holds true for z
N
, as well as the homogeneity relation
f(z) = f(z), C 0, z
N
, (1.4)
and the symmetry relation
f( z) = f(z)
, z
N
(1.5)
(here z = (z
1
, . . . , z
N
) and z = ( z
1
, . . . , z
N
)). In [9] several equivalent char-
acterizations of the Bessmertny class have been established: in terms of certain
positive semidenite kernels on
N

N
, in terms of functional calculus of N-
tuples of commuting bounded strictly accretive operators on a common Hilbert
space, and in terms of the double Cayley transform. Let us briey recall the
last one. The double Cayley transform (over the variables and over the values),
T = ((f), of a function f B
N
(|) is dened for w in the open unit polydisk
D
N
:= w C
N
: [w
k
[ < 1, k = 1, . . . , N as
T(w) =
_
f
_
1 + w
1
1 w
1
, . . . ,
1 +w
N
1 w
N
_
I
/
__
f
_
1 +w
1
1 w
1
, . . . ,
1 +w
N
1 w
N
_
+I
/
_
1
.
(1.6)
For any f B
N
(|), its double Cayley transform T = ((f) belongs to the Agler
Schur class /o
N
(|), i.e., T is holomorphic on D
N
and |T(T)| 1 for ev-
ery N-tuple T = (T
1
, . . . , T
N
) of commuting strict contractions on a common
Hilbert space (see details on this class in [1]). Moreover, there exist Hilbert spaces
A, A
1
, . . . , A
N
such that A =
N
k=1
A
k
, and an Agler representation
T(w) = D +CP(w)(I
AP(w))
1
B, w D
N
, (1.7)
On the Bessmertny Class 141
where P(w) =
N
k=1
w
k
P
k
, with orthogonal projections P
k
onto A
k
, and
_
A B
C D
_
=: U = U
1
= U
L(A |).
Conversely, any function T /o
N
(|) satisfying the latter condition can be rep-
resented as the double Cayley transform, T = ((f), of some function f B
N
(|).
Let us recollect that matrices A
j
, j = 1, . . . , N, in original Bessmertnys
denition had real entries, thus functions from his class took matrix values whose
all entries were real at real points z R
N
. In [9] we have considered also a real
version of the (generalized) Bessmertny class. Namely, we have dened the real
structure on a Hilbert space | by means of an anti-unitary involution (a counter-
part of the complex conjugation), i.e., an operator =
/
: | | such that
2
= I
/
, (1.8)
u
1
, u
2
) = u
2
, u
1
) , u
1
, u
2
|. (1.9)
Such an operator is anti-linear, i.e.,
(u
1
+u
2
) = u
1
+

u
2
, , C, u
1
, u
2
|.
An operator A from L(|, ), the Banach space of all bounded linear operators
from a Hilbert space | to a Hilbert space , is called (
/
,
)-real for anti-unitary

involutions
/
and
if
A = A
/
. (1.10)
Such operators A are a counterpart of matrices with real entries. Finally, a function
f on a set C
N
such that z z , which takes values from L(|, ) is
called (
/
,
)-real if
f
(z) :=
f( z)
/
= f(z), z . (1.11)
If | = and
/
=
= then such a function is called -real. We have dened

the -real Bessmertny class RB
N
(|) as the subclass of all -real functions from
B
N
(|). The latter subclass is a counterpart of the original class considered by
Bessmertny. In [9] we have obtained dierent characterizations for RB
N
(|), too.
In the present paper we introduce and investigate analogous classes of func-
tions (either for the complex and real cases) on more general domains. First,
we dene a product of matrix halfplanes as
n
1
n
1

n
N
n
N
:= Z = (Z
1
, . . . , Z
N
) :
Z
k
C
n
k
n
k
, Z
k
+Z
k
> 0, k = 1, . . . , N (1.12)
which serves as a generalization of the open right poly-halfplane
N
. Then we
dene a counterpart of the domain
N
as
n
1
,...,n
N
:=
_
T
_
n
1
n
1

n
N
n
N
_
, (1.13)
and dene the corresponding Bessmertny classes of functions on the domain
n
1
,...,n
N
. Consideration of such classes can be also motivated by problems of the
theory of electrical networks since there are situations where matrix impedances
are considered as matrix variables (see, e.g., [10]). On the other hand, mathemat-
ical tools for such an investigation have recently appeared. Since in [9] the close
relation of the Bessmertny classes B
N
(|) and RB
N
(|) to the AglerSchur class
/o
N
(|) has been established, this has made possible the use of properties of the
latter class as a tool for investigation Bessmertnys classes. In the same manner
we make use of the recent works of C.-G. Ambrozie and D. Timotin [2], J.A. Ball
and V. Bolotnikov [3] on the AglerSchur class of function on so-called polynomi-
ally dened domains for the investigation of the Bessmertnys classes of functions
on
n
1
,...,n
N
. A counterpart of the class B
N
(|) is introduced in Section 2, where
also a useful decomposition for functions from this class is obtained. In Section 3
the relationship between the Bessmertny class on
n
1
,...,n
N
and the correspond-
ing AglerSchur class on a product of matrix disks is established. This allows us
to give a characterization of the (generalized) Bessmertny class in terms of func-
tional calculus for collections of operators. In Section 4 we describe the image of
this class under the double Cayley transform. Finally, a counterpart of the real
Bessmertny class RB
N
(|) is studied in Section 5.
2. The Bessmertny class for a matrix domain
Let us dene the class B
n
1
,...,n
N
(|) of all L(|)-valued functions f holomorphic on
the domain
n
1
,...,n
N
dened in (1.13) (see also (1.12)) which are representable as
f(Z) = a(Z) b(Z)d(Z)
1
c(Z) (2.1)
for Z
n
1
,...,n
N
, where
A(Z) = G
1
(Z
1
I
,
1
)G
1
+ +G
N
(Z
N
I
,
N
)G
N
=
_
a(Z) b(Z)
c(Z) d(Z)
_
L(| H)
(2.2)
for some Hilbert spaces /
1
, . . . , /
N
, H and operators G
k
L(| H, C
n
k
/
k
), k = 1, . . . , N.
Remark 2.1. If a function f is holomorphic on
n
1
n
1

n
N
n
N
and has
a representation (2.1)(2.2) there, then f can be extended to
n
1
,...,n
N
by ho-
mogeneity of degree one, and this extension is, clearly, holomorphic and admits
a representation 2.1 in
n
1
,...,n
N
. That is why we dene the class B
n
1
,...,n
N
(|)
straight away as a class of functions on
n
1
,...,n
N
. Keeping in mind the possibility
and uniqueness of such extension, we will write sometimes f B
n
1
,...,n
N
(|) for
functions dened originally on
n
1
n
1

n
N
n
N
.
Theorem 2.2. Let f be an L(|)-valued function holomorphic on
n
1
n
1

n
N
n
N
. Then f B
n
1
,...,n
N
(|) if and only if there exist holomorphic functions
k
(Z) on
n
1
n
1

n
N
n
N
with values in L(|, C
n
k
/
k
), k = 1, . . . , N,
such that
f(Z) =
N
k=1
k
()
(Z
k
I
,
k
)
k
(Z), Z,
n
1
n
1

n
N
n
N
(2.3)
holds. In this case the functions
k
(Z) can be uniquely extended to the holomorphic
functions on
n
1
,...,n
N
(we use the same notation for the extended functions) which
are homogeneous of degree zero, i.e., for every C0,
k
(Z) =
k
(Z), Z
n
1
,...,n
N
, (2.4)
and identity (2.3) is extended to all of Z,
n
1
,...,n
N
.
Proof. Necessity. Let f B
n
1
,...,n
N
(|). Then (2.1) holds for Z
n
1
,...,n
N
, some
Hilbert spaces H, /
1
, . . . , /
N
and a linear pencil of operators (2.2). Dene
(Z) :=
_
I
/
d(Z)
1
c(Z)
_
L(|, | H), Z
n
1
,...,n
N
.
Then for all Z,
n
1
,...,n
N
one has
f(Z) = a(Z) b(Z)d(Z)
1
c(Z)
=
_
I
/
c()
d()

_
a(Z) b(Z)d(Z)
1
c(Z)
0
_
=
_
I
/
c()
d()

_
a(Z) b(Z)
c(Z) d(Z)
_ _
I
/
d(Z)
1
c(Z)
_
= ()
A(Z)(Z).
Set
k
(Z) := G
k
(Z), k = 1, . . . , N. Clearly, the functions
k
(Z), k = 1, . . . , N,
are holomorphic on
n
1
,...,n
N
and satisfy (2.4). Rewriting the equality
f(Z) = ()
A(Z)(Z), Z
n
1
,...,n
N
, (2.5)
in the form
f(Z) =
N
k=1
k
()
(Z
k
I
,
k
)
k
(Z), Z,
n
1
,...,n
N
, (2.6)
we obtain, in particular, (2.3).
Suciency. Let f be an L(|)-valued function holomorphic on
n
1
n
1

n
N
n
N
and representable there in the form (2.3) with some holomorphic functions
k
(Z) taking values in L(|, C
n
k
/
k
), k = 1, . . . , N. Set
^ :=
N
k=1
(C
n
k
/
k
), P
k
:= P
,
k
, k = 1, . . . , N,
(Z) := col
_

1
(Z) . . .
N
(Z)
_
L(|, ^),
E := (I
n
1
, . . . , I
n
N
)
n
1
n
1

n
N
n
N
,
where I
n
denotes the identity n n matrix. From (2.3) we get
f(E) =
N
k=1
k
()
k
(E),
n
1
n
1

n
N
n
N
. (2.7)
In particular,
f(E) =
N
k=1
k
(E)
k
(E). (2.8)
By subtracting (2.8) from (2.7) we get
N
k=1
[
k
()
k
(E)]
k
(E) = 0,
n
1
n
1

n
N
n
N
,
i.e., the following orthogonality relation holds:
H := clos span
n
1
n
1
n
N
n
N
[() (E)]| clos(E)| =: A.
For any
n
1
n
1

n
N
n
N
and u | one can represent now ()u as
col
_
(E) () (E)
u A H.
On the other hand, for any u |,
n
1
n
1

n
N
n
N
one has
(E)u clos span
n
1
n
1
n
N
n
N
()|,
(() (E))u clos span
n
1
n
1
n
N
n
N
()|.
Thus, clos span
n
1
n
1
n
N
n
N
()| = A H. Let : A H ^ be
the natural embedding dened by
:
_
(E)u
(() (E))u
_
()u =
_
1
()u
.
.
.
N
()u
_
_
(2.9)
and extended to the whole A H by linearity and continuity. Set
G
k
:= (I
n
k
P
k
)
_
(E) 0
0 I
1
_
L(| H, C
n
k
/
k
), k = 1, . . . , N,
() :=
_
I
/
() (E)
_
L(|, | H),
n
1
n
1

n
N
n
N
.
Then
f(Z) = ()
A(Z)(Z), Z,
n
1
n
1

n
N
n
N
,
where A(Z) is dened by (2.2). Indeed,
()
A(Z)(Z) =
_
I
/
() (E)
_
_
(E) 0
0 I
1
_
_
N
k=1
Z
k
P
k
_
_
(E) 0
0 I
1
_ _
I
/
(Z) (E)
_
=
N
k=1
k
()
(Z
k
I
,
k
)
k
(Z) = f(Z).
Now, with respect to the block partitioning of A(Z) we have
A(Z)(Z) =
_
a(Z) b(Z)
c(Z) d(Z)
_ _
I
/
(Z) (E)
_
=
_
a(Z) +b(Z)((Z) (E))
c(Z) +d(Z)((Z) (E))
_
=:
_
f
1
(Z)
f
2
(Z)
_
.
Since for Z,
n
1
n
1

n
N
n
N
one has
()
A(Z)(Z) =
_
I
/
()
(E)

_
f
1
(Z)
f
2
(Z)
_
= f(Z),
by setting := E in this equality we get
f
1
(Z) = f(Z), Z
n
1
n
1

n
N
n
N
.
Therefore, for every Z,
n
1
n
1

n
N
n
N
we get [()(E)]
f
2
(Z) = 0.
This implies that for every Z
n
1
n
1

n
N
n
N
and u | one has
f
2
(Z)u H. But f
2
(Z)u H. Therefore, f
2
(Z)u = 0, and f
2
(Z) 0, i.e.,
c(Z) +d(Z)[(Z) (E)] 0. (2.10)
Since for every Z
n
1
n
1

n
N
n
N
the operator P(Z) :=
N
k=1
Z
k
P
k
has positive denite real part, i.e., P(Z) + P(Z)

Z
I
A
> 0 for some scalar
Z
> 0, the operator d(Z) = P
1
P(Z)[H has positive denite real part, too.

Therefore, d(Z) is boundedly invertible for all Z
n
1
n
1

n
N
n
N
. From
(2.10) we get (Z) (E) = d(Z)
1
c(Z), Z
n
1
n
1

n
N
n
N
, and
f(Z) = f
1
(Z) = a(Z) b(Z)d(Z)
1
c(Z), Z
n
1
n
1

n
N
n
N
.
Taking into account Remark 2.1, we get f B
n
1
,...,n
N
(|).
Functions (Z) (E) = d(Z)
1
c(Z) and, hence, (Z) are well dened,
holomorphic and homogeneous of degree zero on
n
1
,...,n
N
, thus (2.6) holds. The
proof is complete.
3. The class B
n
1
,...,n
N
(|) and functional calculus
Let us observe now that (2.3) is equivalent to the couple of identities
f(Z) +f()
=
N
k=1
k
()
((Z
k
+
k
) I
,
k
)
k
(Z), (3.1)
f(Z) f()
=
N
k=1
k
()
((Z
k
k
) I
,
k
)
k
(Z) (3.2)
valid for all Z,
n
1
n
1

n
N
n
N
. We will show that the double Cay-
ley transform T = ((f) applied to a function f from the Bessmertny class
B
n
1
,...,n
N
(|) and dened as
T(W) =
_
f((I
n
1
+W
1
)(I
n
1
W
1
)
1
, . . . , (I
n
N
+W
N
)(I
n
N
W
N
)
1
) I
/
_
f((I
n
1
+W
1
)(I
n
1
W
1
)
1
, . . . , (I
n
N
+W
N
)(I
n
N
W
N
)
1
) +I
/
1
(3.3)
(compare with (1.6)) turns the rst of these identities into an Agler-type identity
which characterizes the AglerSchur class of holomorphic L(|)-valued functions
on the product of open matrix unit disks
D
n
1
n
1
D
n
N
n
N
:= W = (W
1
, . . . , W
N
) C
n
1
n
1
C
n
N
n
N
:
W
k
W
k
< I
n
k
, k = 1, . . . , N.
The latter is a special case of the AglerSchur class of holomorphic L(|)-valued
functions on a domain with matrix polynomial dening function, which was studied
in [2] and [3]. This allows us to obtain one more characterization of B
n
1
,...,n
N
(|).
Let P(w), w C
n
, be a polynomial p q matrix-valued function, and
T
P
:= w C
n
: |P(w)| < 1
(here and in the sequel the norm of a p q matrix means its operator norm with
respect to the standard Euclidean metrics in C
p
and C
q
). Let (
1
P
denote the set of
commutative n-tuples T = (T
1
, . . . , T
n
) of bounded linear operators on a common
Hilbert space H
T
subject to the condition |P(T)| < 1. It was shown in [2] that
the Taylor joint spectrum
T
(T) (see [15, 16] and also [8]) of any T (
1
P
is
contained in T
P
. Thus, for any function S holomorphic on T
P
and any T (
1
P
the operator S(T) is well dened by the Taylor functional calculus (see [14, 17]
and also [8]). For the domain T
P
, the AglerSchur class /o
1
P
(c, c
) consists of
all holomorphic L(c, c
)-valued functions T on T
P
such that
|T(T)| 1, T (
1
P
. (3.4)
Recall the following theorem from [3] (the case when c = c
= C can be found
in [2]), however in a slightly simplied form which will be sucient for our pur-
pose.
Theorem 3.1. Let T be an L(c, c
)-valued function holomorphic on T

P
. Then the
following statements are equivalent:
(i) T /o
1
P
(c, c
);
(ii) there exist an auxiliary Hilbert space / and an L(C
p
/, c
)-valued func-
tion H
L
holomorphic on T
P
such that
I
L
T(w)T()
= H
L
(w) ((I
p
P(w)P()
) I
,
) H
L
()
(3.5)
holds for all w, T
P
;
(iii) there exist an auxiliary Hilbert space /and an L(c, C
q
/)-valued function
H
R
holomorphic on T
P
such that
I
L
T()
T(w) = H
R
()
((I
q
P()
P(w)) I
,
) H
R
(w) (3.6)
holds for all w, T
P
;
(iv) there exist an auxiliary Hilbert space /, an L(C
p
/, c
)-valued function
H
L
and an L(c, C
q
/)-valued function H
R
, which are holomorphic on T
P
,
such that
_
I
L
T(
T(w) T(
T()
T(w
) T(w) I
L
T(w
)T()
_
=
_
H
R
(
0
0 H
L
(w
)
_
(3.7)
__
I
q
P(
P(w) P(
P()
P(w
) P(w) I
p
P(w
)P()
_
I
,
__
H
R
(w) 0
0 H
L
()
_
holds for all w, w
, ,
T
P
;
(v) there exists a Hilbert space A and a unitary operator
U =
_
A B
C D
_
L((C
p
A) c, (C
q
A) c
) (3.8)
such that
F(w) = D +C(P(w) I
) (I
C
q
A(P(w) I
))
1
B (3.9)
holds for all w T
P
.
In [3] it was shown how to obtain from (3.5) a unitary operator (3.8) which
gives the representation (3.9) for an arbitrary T /o
1
P
(c, c
). We will show now

how to get from (3.7) a special unitary operator (3.8) and representation (3.9) for
an arbitrary T /o
1
P
(c, c
). Let (3.7) hold for such T, where a Hilbert space

/and functions H
L
, H
R
are such as in statement (iv) of Theorem 3.1. Dene the
lineals
T
0
:= span
__
(P(w) I
,
)H
R
(w)
I
L
_
e,
_
H
L
()
T()
_
e
:
w, T
P
, e c, e
(C
p
/) c,
1
0
:= span
__
H
R
(w)
T(w)
_
e,
_
(P()
I
,
)H
L
()
I
L
_
e
:
w, T
P
, e c, e
(C
q
/) c
,
and the operator U
0
: T
0
1
0
which acts on the generating vectors of T
0
as
_
(P(w) I
,
)H
R
(w)
I
L
_
e
_
H
R
(w)
T(w)
_
e, w T
P
, e c,
_
H
L
()
T()
_
e

_
(P()
I
,
)H
L
()
I
L
_
e
, T
P
, e
.
This operator is correctly dened. Moreover, U
0
maps T
0
isometrically onto 1
0
.
Indeed, (3.7) can be rewritten as
_
H
R
(
T(
H
L
(w
)(P(w
) I
,
) I
L
_ _
H
R
(w) (P()
I
,
)H
L
()
T(w) I
L
_
=
_
H
R
(
(P(
I
,
) I
L
H
L
(w
) F(w
)
_ _
(P(w) I
,
)H
R
(w) H
L
()
I
L
T()
_
,
which means that for
x =
_
(P(w) I
,
)H
R
(w)
I
L
_
e +
_
H
L
()
T()
_
e
,
x
=
_
(P(w
) I
,
)H
R
(w
)
I
L
_
e
+
_
H
L
(
T(
_
e
,
one has
U
0
x, U
0
x
) = x, x
) .
Clearly, U
0
can be uniquely extended to the unitary operator

U
0
: clos(T
0
)
clos(1
0
). In the case when
dim((C
p
/) c) clos(T
0
) = dim((C
q
/) c
) clos(1
0
) (3.10)
there exists a (non-unique!) unitary operator U : (C
p
/) c (C
q
/) c
such that U[clos(T

0
) =

U
0
. In the case when (3.10) does not hold one can set
/:= //, where / is an innite-dimensional Hilbert space, then (3.10) holds

for

/ in the place of /, and there exists a unitary operator U : (C
p

/) c
(C
q

/) c
such that U[clos(T

0
) =

U
0
. Thus, without loss of generality we
may consider that (3.10) holds.
Let U have a block partitioning
U =
_
A B
C D
_
: (C
p
/) c (C
q
/) c
.
Then, in particular,
_
A B
C D
_ _
(P(w) I
,
)H
R
(w)
I
L
_
=
_
H
R
(w)
T(w)
_
, w T
P
. (3.11)
Since for w T
P
one has |P(w)| < 1, and since |A| 1, we can solve the rst
block row equation of (3.11) for H
R
(w):
H
R
(w) = (I
C
q
1
A(P(w) I
,
))
1
B, w T
P
.
Then from the second block row of (3.11) we get
T(w) = D +C(P(w) I
,
)(I
C
q
,
A(P(w) I
,
))
1
B, w T
P
,
i.e., (3.9) with A = /.
We are interested here in the case of the AglerSchur class for the domain
T
P
where the domain T
P
is D
n
1
n
1
D
n
N
n
N
, and the polynomial which
denes this domain is
P(W) = diag(W
1
, . . . , W
N
), W D
n
1
n
1
D
n
N
n
N
.
Here W may be viewed as an (n
2
1
n
2
N
)-tuple of scalar variables (W
k
)
ij
, k =
1, . . . , N, i, j = 1, . . . , n
k
. We will write in this case /o
n
1
,...,n
N
(c, c
) instead of
/o
1
P
(c, c
), and if c = c
we will write /o
n
1
,...,n
N
(c). The class (
1
P
is identied
for T
P
= D
n
1
n
1
D
n
N
n
N
with the class (
(n
1
,...,n
N
)
of N-tuples of matrices
T = (T
1
, . . . , T
N
) B
n
1
n
1
T
B
n
N
n
N
T
over a common commutative operator
algebra B
T
L(H
T
), with a Hilbert space H
T
, such that |T
k
| < 1, k = 1, . . . , N.
Denote by /
(n
1
,...,n
N
)
the class of N-tuples of matrices R = (R
1
, . . . , R
N
)
B
n
1
n
1
R
B
n
N
n
N
R
over a common commutative operator algebra B
R
L(H
R
),
with a Hilbert space H
R
, for which there exists a real constant s
R
> 0 such that
R
k
+R
k
s
R
I
C
n
k
1
R
, k = 1, . . . , N.
Theorem 3.2. For any R /
(n
1
,...,n
N
)
,
T
(R)
n
1
n
1

n
N
n
N
,
where
T
(R) denotes the Taylor joint spectrum of the collection of operators
(R
k
)
ij
, k = 1, . . . , N, i, j = 1, . . . , n
k
.
Proof. It is shown in [13] that the Taylor joint spectrum
T
(X) of an n-tuple of
commuting bounded operators X = (X
1
, . . . , X
n
) on a common Hilbert space H
X
is contained in the polynomially convex closure of
(X), the approximate point

spectrum of X. The latter is dened as the set of points = (
1
, . . . ,
n
) C
n
for
which there exists a sequence of vectors h
H
X
such that |h
| = 1, N, and
(X
j

j
I
1
X
)h
0 as for all j = 1, . . . , n. Thus it suces to show that
(R) :=
((R
k
)
ij
: k = 1, . . . , N, i, j = 1, . . . , n
k
)
n
1
n
1
s

n
N
n
N
s
whenever R /
(n
1
,...,n
N
)
and R
k
+R
k
sI
C
n
k 1
R
> 0, k = 1, . . . , N, where
nn
s
:= M C
nn
: M +M
sI
n
,
since
n
1
n
1
s

n
N
n
N
s
is convex, and hence polynomially convex, and since
n
1
n
1
s

n
N
n
N
s

n
1
n
1

n
N
n
N
for s > 0. Suppose that =
(
1
, . . . ,
N
)
(R). Then there exists a sequence of vectors h
H
R
such that
|h
| = 1, N, and for k = 1, . . . , N, i, j = 1, . . . , n
k
one has
((R
k
)
ij
(
k
)
ij
I
1
R
)h
0 as .
Therefore, for every k 1, . . . , N and u
k
= col(u
k1
, . . . , u
kn
k
) C
n
k
one has
n
k
i=1
n
k
j=1
_
((R
k
)
ij
+ (R
k
)
ji
)h
, h
_
((
k
)
ij
+ (
k
)
ji
) h
, h
)
_
u
ki
u
kj
0
as . Since h
, h
) = 1, the subtrahend does not depend on . Therefore,

s u
k
, u
k
) = s lim
u
k
h
, u
k
h
)
lim
(R
k
+R
k
) u
k
h
, u
k
h
)
= lim
n
k
i=1
n
k
j=1
((R
k
)
ij
+ (R
k
)
ji
)h
, h
_
u
ki
u
kj
=
n
k
i=1
n
k
j=1
((
k
)
ij
+ (
k
)
ji
)u
ki
u
kj
= (
k
+
k
)u
k
, u
k
) .
Thus,
k
+
k
sI
n
k
, k = 1, . . . , N, i.e.,
n
1
n
1
s

n
N
n
N
s
, as desired.

Theorem 3.2 implies that for every holomorphic function f on
n
1
n
1

n
N
n
N
and every R /
(n
1
,...,n
N
)
the operator f(R) is well dened by the Taylor
functional calculus.
The Cayley transform dened by
R
k
= (I
C
n
k1
T
+T
k
)(I
C
n
k 1
T
T
k
)
1
, k = 1, . . . , N, (3.12)
maps the class (
(n
1
,...,n
N
)
onto the class /
(n
1
,...,n
N
)
, and its inverse is given by
T
k
= (R
k
I
C
n
k1
R
)(R
k
+I
C
n
k 1
R
)
1
, k = 1, . . . , N, (3.13)
where H
R
= H
T
. Let f be an L(|)-valued function holomorphic on
n
1
n
1

n
N
n
N
. Then its double Cayley transform T = ((f) dened by (3.3) is
holomorphic on D
n
1
n
1
D
n
N
n
N
, and by the spectral mapping theorem
and uniqueness of Taylors functional calculus (see [11]) one has
T(T) = f(R),
where T (
(n
1
,...,n
N
)
and R /
(n
1
,...,n
N
)
are related by (3.12) and (3.13).
Theorem 3.3. Let f be an L(|)-valued function holomorphic on
n
1
n
1

n
N
n
N
. Then f B
n
1
,...,n
N
(|) if and only if the following conditions are satis-
ed:
(i) f(tZ) = tf(Z), t > 0, Z
n
1
n
1

n
N
n
N
;
(ii) f(R) +f(R)
0, R /
(n
1
,...,n
N
)
;
(iii) f(Z
) := f(Z
1
, . . . , Z
N
) = f(Z)
, Z
n
1
n
1

n
N
n
N
.
n
1
,...,n
N
(|). Then (i) and (iii) easily follow from the
representation (2.1) of f. Condition (ii) on f is equivalent to condition (3.4) on T
which is dened by (3.3), i.e., to T /o
n
1
,...,n
N
(|). Let us show the latter. Since
by Theorem 2.2 f satises (2.3), and hence (3.1), one can set
Z
k
= (I
n
k
+W
k
)(I
n
k
W
k
)
1
,
k
= (I
n
k
+
k
)(I
n
k

k
)
1
, k = 1, . . . , N,
in (3.1) and get
(I
/
+T(W))(I
/
T(W))
1
+ (I
/
T()
)
1
(I
/
+T()
)
=
N
k=1
k
()
__
(I
n
k
+W
k
)(I
n
k
W
k
)
1
+ (I
n
k

k
)
1
(I
n
k
+
k
)
_
I
,
k
_

k
(W), W, D
n
1
n
1
D
n
N
n
N
,
where for k = 1, . . . , N,
k
(W) =
k
_
(I
n
1
+W
1
)(I
n
1
W
1
)
1
, . . . , (I
n
N
+W
N
)(I
n
N
W
N
)
1
_
. (3.14)
We can rewrite this in the form
I
/
T()
T(W) =
N
k=1
k
()
((I
n
k

k
W
k
) I
,
k
)
k
(W), (3.15)
where for k = 1, . . . , N,
k
(W) =
_
(I
n
k
W
k
)
1
I
,
k
_
k
(W)(I
/
T(W)) L(|, C
n
k
/
k
). (3.16)
The identity (3.15) coincides with (3.6) for our case, with
H
R
(W) = col(
1
(W), . . . ,
N
(W)) L
_
|,
N
k=1
(C
n
k
/
k
)
_
,
P(W) = diag(W
1
, . . . , W
N
).
Note, that without loss of generality we may consider all of /
k
s equal to some
space /, say, / =
N
k=1
/
k
. Then H
R
(W) L(|, C
n
1
++n
N
/). By The-
orem 3.1, this means that T /o
n
1
,...,n
N
(|).
Suciency. Let f satisfy conditions (i)(iii). Since (ii) is equivalent to T
/o
n
1
,...,n
N
(|), where T is dened by (3.3), the identity (3.15) holds with some
L(|, C
n
k
/)-valued functions
k
holomorphic on D
n
1
n
1
D
n
N
n
N
, k =
1, . . . , N, with an auxiliary Hilbert space / (spaces /
k
can be chosen equal in
(3.15)). Set
W
k
= (Z
k
I
n
k
)(Z
k
+I
n
k
)
1
,
k
= (
k
I
n
k
)(
k
+I
n
k
)
1
, k = 1, . . . , N,
in (3.15), and by virtue of (3.3) get (3.1) with
k
(Z) = ((I
n
k
+Z
k
)
1
) I
,
)

k
_
(Z
1
I
n
1
)(Z
1
+I
n
1
)
1
, . . . , (Z
N
I
n
N
)(Z
N
+I
n
N
)
1
_
(I
/
+f(Z)) L(|, C
n
k
/), k = 1, . . . , N (3.17)
(in fact, passing from (3.1) to (3.15) is invertible, and (3.17) is obtained from
(3.14) and (3.16), and vice versa). The property (iii) implies f(X) = f(X)
for
every N-tuple X = (X
1
, . . . , X
N
)
n
1
n
1

n
N
n
N
of positive denite
matrices (we will denote this set by T
(n
1
,...,n
N
)
), and for any such X and t > 0 by
(3.1) one has:
f(X) +f(tX) = (1 +t)
N
k=1
k
(tX)
(X
k
I
,
)
k
(X),
f(tX) +f(X) = (1 +t)
N
k=1
k
(X)
(X
k
I
,
)
k
(tX),
1 +t
2
[f(X) +f(X)] =
1 +t
2
N
k=1
k
(X)
(2X
k
I
,
)
k
(X),
1 +t
2t
[f(tX) +f(tX)] =
1 +t
2t
N
k=1
2
k
(tX)
(2tX
k
I
,
)
k
(tX).
By (i), the left-hand sides of these equalities coincide and equal (1+t)f(X), hence
f(X) =
N
k=1
k
(tX)
(X
k
I
,
)
k
(X) =
N
k=1
k
(X)
(X
k
I
,
)
k
(tX)
=
N
k=1
k
(X)
(X
k
I
,
)
k
(X) =
N
k=1
k
(tX)
(X
k
I
,
)
k
(tX).
It follows from the latter equalities that
0
N
k=1
[
k
(tX)
k
(X)]
(X
k
I
,
)[
k
(tX)
k
(X)]
=
N
k=1
k
(tX)
(X
k
I
,
)
k
(tX)
N
k=1
k
(tX)
(X
k
I
,
)
k
(X)
k=1
k
(X)
(X
k
I
,
)
k
(tX) +
N
k=1
k
(X)
(X
k
I
,
)
k
(X) = 0.
Thus
k
(tX)
k
(X) = 0 for every X T
(n
1
,...,n
N
)
, t > 0 and k = 1, . . . , N.
For xed k 1, . . . , N and t > 0 the function h
k,t
(Z) :=
k
(tZ)
k
(Z) is
holomorphic on
n
1
n
1

n
N
n
N
and takes values in L(|, C
n
k
/). Then
for any xed k 1, . . . , N, t > 0, u | and m C
n
k
/ the scalar function
h
k,t,u,m
(Z) := h
k,t
(Z)u, m)
C
n
k,
is holomorphic on
n
1
n
1

n
N
n
N
and
vanishes on T
(n
1
,...,n
N
)
. The latter set is the uniqueness subset in
n
1
n
1

n
N
n
N
, thus by the uniqueness theorem for holomorphic functions of several
variables (see, e.g., [12]), h
k,t,u,m
(Z) 0, hence h
k,t
(Z) 0, which means:
k
(tZ) =
k
(Z), t > 0, Z
n
1
n
1

n
N
n
N
.
It follows from the latter equality that for every Z,
n
1
n
1

n
N
n
N
and t > 0 one has
f(Z) +tf()
= f(Z) +f(t)
=
N
k=1
k
(t)
((Z
k
+t
k
) I
,
)
k
(Z)
=
N
k=1
k
()
((Z
k
+t
k
) I
,
)
k
(Z)
=
N
k=1
k
()
(Z
k
I
,
)
k
(Z) +t
k
()
k
I
,
)
k
(Z),
and the comparison of the coecients of the two linear functions in t, at the
beginning and at the end of this chain of equalities, gives:
f(Z) =
N
k=1
k
()
(Z
k
I
,
)
k
(Z), Z,
n
1
n
1

n
N
n
N
,
i.e., (2.3) with /
k
= /, k = 1, . . . , N. By Theorem 2.2, f B
n
1
,...,n
N
(|). The
proof is complete.
Corollary 3.4. Let f be an L(|)-valued function holomorphic on
n
1
,...,n
N
. Then
f B
n
1
,...,n
N
(|) if and only if the following conditions are satised:
(i) f(Z) = f(Z), C 0, Z
n
1
,...,n
N
;
(ii) f(R) +f(R)
0, R /
(n
1
,...,n
N
)
;
(iii) f(Z
) = f(Z)
, Z
n
1
,...,n
N
.
Proof. If f B
n
1
,...,n
N
(|) then (i) and (iii) follow from the representation (2.1)
(2.2) of f, and (ii) follows from Theorem 3.3.
Conversely, statements (i)(iii) of the corollary imply statements (i)(iii) of
Theorem 3.3, which in turn imply that f B
n
1
,...,n
N
(|).
Remark 3.5. By Corollary 3.4, its conditions (i)(iii) on holomorphic L(|)-valued
functions on
n
1
,...,n
N
give an equivalent denition of the class B
n
1
,...,n
N
(|), which
seems to be more natural than the original denition given above in existence
terms.
4. The image of the class B
n
1
,...,n
N
(|) under
the double Cayley transform
It was shown in the proof of Theorem 3.3 that if f B
n
1
,...,n
N
(|) then the double
Cayley transform of f, T = ((f), dened by (3.3), belongs to the AglerSchur
class /o
n
1
,...,n
N
(|). In fact, we are able to proof a stronger statement.
Theorem 4.1. A holomorphic L(|)-valued function T on D
n
1
n
1
D
n
N
n
N
can be represented as T = ((f) for some f B
n
1
,...,n
N
(|) if and only if the
following conditions are fullled:
(i) There exist a Hilbert space A and an operator
U =
_
A B
C D
_
L((C
n
1
++n
N
A) |) (4.1)
such that for W = (W
1
, . . . , W
N
) D
n
1
n
1
D
n
N
n
N
one has
T(W) = D +C(P(W) I
)(I
C
n
1
++n
N
A(P(W) I
))
1
B, (4.2)
where P(W) = diag(W
1
, . . . , W
N
) and U = U
= U
1
.
(ii) 1 / (T(0)).
n
1
,...,n
N
(|). Then (3.1) and (3.2) hold. As we have
shown in Theorem 3.3, the identity (3.1) implies the identity (3.15) for T = ((f),
with holomorphic L(|, C
n
k
/
k
)-valued functions
k
, k = 1, . . . , N, on D
n
1
n
1
D
n
N
n
N
dened by (3.14) and (3.16). Analogously, the identity (3.2) implies
T(W) T()
=
N
k=1
k
()
((W
k

k
) I
,
k
)
k
(W), (4.3)
W, D
n
1
n
1
D
n
N
n
N
.
Let us rewrite (3.15) and (4.3) in a somewhat dierent way. Since by Theorem 3.3
f B
n
1
,...,n
N
(|) satises f(Z
) = f(Z)
, Z
n
1
n
1

n
N
n
N
, one has
also
T(W
) = T(W)
, W D
n
1
n
1
D
n
N
n
N
.
Therefore, (3.15) and (4.3) are equivalent to the following two identities, respec-
tively:
I
/
T(W)T()
=
N
k=1
k
(W) ((I
n
k
W
k
k
) I
,
k
)

k
()
, (4.4)
T(W) T()
=
N
k=1
k
(W) ((W
k

k
) I
,
k
)

k
()
, (4.5)
W, D
n
1
n
1
D
n
N
n
N
,
where

k
(W) =
k
(W
are holomorphic L(C

n
k
/
k
, |)-valued functions on
D
n
1
n
1
D
n
N
n
N
. We will show that the identities (4.4) and (4.5) allow
us to construct a Hilbert space A and an operator U satisfying condition (i) of
this theorem. To this end, we will apply the construction from Section 3 (next to
Theorem 3.1) to T = ((f). In this case c = c
= |. Without loss of generality we

may consider all of /
k
s equal. Say, set /:=
N
k=1
/
k
and regard
H
R
k
=
k
L(|, C
n
k
/), H
L
k
=

k
L(C
n
k
/, |), k = 1, . . . , N.
Then (3.15), (4.3), (4.4) and (4.5) imply (3.7), and
H
L
(W
) = H
R
(W)
,
where H
L
(W) = col(H
L
1
(W), . . . , H
L
N
(W)), H
R
= col(H
R
1
(W), . . . , H
R
N
(W)), W
D
n
1
n
1
D
n
N
n
N
. Thus, T
0
= 1
0
, and the operator U
0
acts on the generating
vectors of T
0
as follows:
_
(P(W) I
,
)H
R
(W)
I
/
_
u
_
H
R
(W)
T(W)
_
u,
_
H
R
(W)
T(W)
_
u
_
(P(W) I
,
)H
R
(W)
I
/
_
u,
W D
n
1
n
1
D
n
N
n
N
.
We used here the relations T(W
) = T(W)
, H
L
(W
) = H
R
(W)
, P(W
) =
diag(W
1
, . . . , W
N
) = P(W)
. Thus U
0
= U
1
0
. Therefore,

U
0
=

U
0
1
. Since p =
q = n
1
+ +n
N
, c = c
= |, (3.10) holds. Then the operator

U =

U
0
I
((C
n
1
++n
N,)/)clos(1
0
)
L((C
n
1
++n
N
/) |)
satises U = U
1
. Since we have also U
= U
1
, (i) is satised with A = /.
Statement (ii) follows in the same way as in [9, Theorem 4.2], with E =
(I
n
1
, . . . , I
n
N
) in the place of e = (1, . . . , 1).
Suciency. Let the conditions (i) and (ii) on T be satised. Then in the
same way as in [9, Theorem 4.2] one can see that 1 / (T(W)) for all W
D
n
1
n
1
D
n
N
n
N
. Thus, the function
F(W) := (I
/
+T(W))(I
/
T(W))
1
is correctly dened and holomorphic on D
n
1
n
1
D
n
N
n
N
. It is easy to see
that
F(W) +F()
= 2(I
/
T()
)
1
(I
/
T()
T(W))(I
/
T(W))
1
, (4.6)
F(W) F()
= 2(I
/
T()
)
1
(T(W) T()
)(I
/
T(W))
1
. (4.7)
As shown in [3], it follows from (4.2) that
I
/
T()
T(W) = B
(I
C
n
1
++n
N
(P()
)A
)
1
((I
C
n
1
++n
N
P()
P(W)) I
)
(I
C
n
1
++n
N
A(P(W) I
))
1
B.
Since U = U
, we get
I
/
T()
T(W) =
N
k=1
B
(I
C
n
1
++n
N
(P()
)A)
1
(P
C
n
k
I
)((I
n
k

k
W
k
) I
)(P
C
n
k
I
)
(I
C
n
1
++n
N
A(P(W) I
))
1
B.
Analogously,
T(W) T()
=
N
k=1
B
(I
C
n
1
++n
N
(P()
)A)
1
(P
C
n
k I
)((W
k

k
) I
)(P
C
n
k I
)
(I
C
n
1
++n
N
A(P(W) I
))
1
B.
Thus, from (4.6) and (4.7) we get
F(W) +F()
=
N
k=1
k
()
((I
n
k

k
W
k
) I
)
k
(W), (4.8)
F(W) F()
=
N
k=1
k
()
((W
k

k
) I
)
k
(W), (4.9)
with
k
(W) =
2(P
C
n
k I
X
)(I
C
n
1
++n
N
A(P(W) I
))
1
B(I
/
T(W))
1
,
for all W D
n
1
n
1
D
n
N
n
N
and k = 1, . . . , N.
Since for Z
k
,
k

n
k
n
k
we have
I
n
k
(
k
+I
n
k
)
1
(
k
I
n
k
)(Z
k
I
n
k
)(Z
k
+I
n
k
)
1
= 2(
k
+I
n
k
)
1
(Z
k
+
k
)(Z
k
+I
n
k
)
1
,
(Z
k
I
n
k
)(Z
k
+I
n
k
)
1
(
k
+I
n
k
)
1
(
k
I
n
k
)
= 2(
k
+I
n
k
)
1
(Z
k
k
)(Z
k
+I
n
k
)
1
,
by setting W
k
:= (Z
k
I
n
k
)(Z
k
+ I
n
k
)
1
and
k
:= (
k
I
n
k
)(
k
+ I
n
k
)
1
in
(4.8) and (4.9) we get the identities (3.1) and (3.2) for
f(Z) = F((Z
1
I
n
1
)(Z
1
+I
n
1
)
1
, . . . , (Z
N
I
n
N
)(Z
N
I
n
N
)
1
),
with
k
(Z) =
2((Z
k
+I
n
k
)
1
I
)

k
((Z
1
I
n
1
)(Z
1
+I
n
1
)
1
, . . . , (Z
N
I
n
N
)(Z
N
I
n
N
)
1
)
for k = 1, . . . , N. Thus, by Theorem 2.2 we nally get T = ((f) where f
B
n
1
,...,n
N
(|). The proof is complete.
5. The real case
In Section 1 we have mentioned the notions of an anti-unitary involution (AUI) =
/
on a Hilbert space | (a counterpart of the operator
n
of complex conjugation
on C
n
), a (
/
,
)-real operator A L(|, ) (a counterpart of matrix with real

entries), and a (
/
,
)-real operator-valued function f (a counterpart of function

which takes real scalar or matrix values at real points). Some basic properties of
AUI were described in [9, Proposition 6.1]. We will need also the following property.
Proposition 5.1. Let
/
and
1
be AUIs on Hilbert spaces | and H, respectively.
Then the operator
/1
=
/

1
on | H which is dened on elementary
tensors u h as
(
/

1
)(u h) =
/
u
1
h (5.1)
and then extended to all of | H by linearity and continuity, is dened correctly
and is an AUI on | H.
Proof. First, let us observe that
/1
is correctly dened. To this end, note that
for arbitrary x
l
=1
u
and x
m
=1
u
from | H we have
/1
x
,
/1
x
)
/1
=
l
=1
m
=1
/
u
1
h
,
/
u

1
h
_
/1
=
l
=1
m
=1
/
u
,
/
u
_
/
1
h
,
1
h
_
1
=
l
=1
m
=1
, u
_
/
, h
_
1
=
l
=1
m
=1
, u
_
/1
= x
, x
)
/1
,
i.e.,
/1
is an anti-isometry on linear combinations of elementary tensors. Thus,
it is uniquely extended to an operator on all of | H, and the property (1.9) of
the extended operator follows by continuity. Since for arbitrary x
l
=1
u
and x
m
=1
u
from | H we have
2
/1
x
, x
_
/1
=
l
=1
m
=1
2
/
u
2
1
h
, u
_
/1
=
l
=1
m
=1
, u
_
/1
= x
, x
)
/1
,
by continuity the property (1.8) of
/1
follows as well. Thus,
/1
is an AUI on
| H.
Let | be a Hilbert space, and let =
/
be an AUI on |. Denote by
RB
n
1
,...,n
N
(|) the subclass of B
n
1
,...,n
N
(|) consisting of -real functions. The
following theorem gives several equivalent characterizations of the -real-valued
Bessmertny class RB
n
1
,...,n
N
(|) which specify for this case the characterizations
obtained above for the complex-valued Bessmertny class B
n
1
,...,n
N
(|).
Theorem 5.2. Let f be a holomorphic L(|)-valued function on
n
1
,...,n
N
, and
=
/
be an AUI on a Hilbert space |. The following statements are equivalent:
(i) f RB
n
1
,...,n
N
(|);
(ii) there exist a representation (2.1) of f and AUIs
,
k
on /
k
, k = 1, . . . , N,
such that the operators G
k
in (2.2) are (
/

1
,
n
k

,
k
)-real;
(iii) there exist a representation (2.3) of f and AUIs
,
k
on /
k
such that the
holomorphic functions
k
(Z) are (
/
,
n
k

,
k
)-real, k = 1, . . . , N;
(iv) there exist a Hilbert space A and an operator U as in (4.1) such that T = ((f)
satises (4.2) and U = U
= U
1
; moreover, there exists an AUI
on A
such that the operator U is ((
n
1
++n
N

)
/
)-real.
Proof. (i) (iii) Let (i) hold. By Theorem 2.2 there exists a representation (2.3) of
f with holomorphic L(|, C
n
k
/
k
)-valued functions
k
on
n
1
n
1

n
N
n
N
.
Let
,
k
be an AUI on /
k
, and let
n
k
be a standard AUI on C
n
k
, i.e., a complex
conjugation. Set
/
k
:= /
k
/
k
and
,
k
:=
_
0
,
k
,
k
0
_
, k = 1, . . . , N.
Clearly,
,
k
is an AUI on

/
k
.
Dene the rearrangement isomorphisms V
k
: C
n
k
(/
k
/
k
) (C
n
k
/
k
) (C
n
k
/
k
) by
_
_
m
11
m
21
.
.
.
m
1n
k
m
2n
k
_
_
m
11
.
.
.
m
1n
k
m
21
.
.
.
m
2n
k
_
_
.
Then
n
k

,
k
= V
1
k
_
0
n
k

,
k
n
k

,
k
0
_
V
k
. (5.2)
Set

k
(Z) :=
1
2
V
1
k
_

k
(Z)
(
n
k

,
k
)
k
(

Z)
/
_
,
where

Z = (Z
1
, . . . , Z
N
), and (Z
k
)
ij
= (Z
k
)
ij
, k = 1, . . . , N, i, j = 1, . . . , n
k
. By
properties of AUIs,
k
(Z) is holomorphic on
n
1
n
1

n
N
n
N
. Moreover,

k
is (
/
,
n
k

,
k
)-real. Indeed, due to (5.2) we have

k
(Z) = (
n
k

,
k
)
k
(

Z)
/
=
1
2
V
1
k
_
0
n
k

,
k
n
k

,
k
0
_
_

k
(

Z)
(
n
k

,
k
)
k
(Z)
/
_
/
=
1
2
V
1
k
_
(
n
k

,
k
)
2
k
(Z)
2
/
(
n
k

,
k
)
k
(
Z)
/
_
=
1
2
V
1
k
_

k
(Z)
(
n
k

,
k
)
k
(

Z)
/
_
=
k
(Z).
Let us show that
f(Z) =
N
k=1

k
()
(Z
k
I
,
k
)
k
(Z), Z,
n
1
n
1

n
N
n
N
.
To this end, let us show rst that for k = 1, . . . , N:
_
(
n
k

,
k
)
k
()
/
_
=
/

k
()
(
n
k

,
k
),
n
1
n
1

n
N
n
N
.
(5.3)
Indeed, for any m C
n
k

/
k
, u |,
n
1
n
1

n
N
n
N
one has
__
(
n
k

,
k
)
k
()
/
_
m, u
_
/
=
_
m, (
n
k

,
k
)
k
()
/
u
_
C
n
k
,
k
_

k
()
/
u, (
n
k

,
k
)m
_
C
n
k
,
k
=
_
/
u,
k
()
(
n
k

,
k
)m
_
/
=
_
/

k
()
(
n
k

,
k
)m, u
_
/
.
Now, for any Z,
n
1
n
1

n
N
n
N
:
N
k=1

k
()
(Z
k
I
,
k
)
k
(Z) =
1
2
N
k=1
_

k
()
(
n
k

,
k
)
k
(
)
/
_
V
k
(Z
k
I
,
k
)V
1
k
_

k
(Z)
(
n
k

,
k
)
k
(

Z)
/
_
=
1
2
N
k=1
_

k
()
(
n
k

,
k
)
k
(
)
/
_
_
Z
k
I
,
k
0
0 Z
k
I
,
k
_
_

k
(Z)
(
n
k

,
k
)
k
(

Z)
/
_
=
1
2
N
k=1
k
()
(Z
k
I
,
k
)
k
(Z)
+
1
2
N
k=1
k
(
(
n
k

,
k
)(Z
k
I
,
k
)(
n
k

,
k
)
k
(

Z)
/
=
1
2
_
N
k=1
k
()
(Z
k
I
,
k
)
k
(Z) +
N
k=1
k
(
(Z
k
I
,
k
)
k
(

Z)
/
_
=
1
2
(f(Z) +
/
f(
Z)
/
) = f(Z),
where we used (5.3), unitarity of V
k
, and identity
n
k
Z
k
n
k
= Z
k
. Thus, (iii) follows
from (i).
(iii) (ii) Let (iii) hold. As in the suciency part of the proof of Theorem 2.3
we set
^ :=
N
k=1
(C
n
k
/
k
), (Z) := col(
1
(Z), . . . ,
N
(Z)) L(|, ^),
P
k
:= P
,
k
, E = (I
n
1
, . . . , I
n
N
)
n
1
n
1

n
N
n
N
,
H := clos span
n
1
n
1
n
N
n
N
(() (E))| ^,
G
k
:= (I
n
k
P
k
)
_
(E) 0
0 I
1
_
L(| H, C
n
k
/
k
),
where : A H ^ is dened by (2.9). For (Z) =
_
I
/
(Z) (E)
_
one has
(E)| = | 0, therefore the linear span of vectors of the form (Z)u, Z
n
1
n
1

n
N
n
N
, u |, is dense in | H. Set
A
:=
N
k=1
(
n
k

,
k
).
By the assumption, we have for k = 1, . . . , N:
(
n
k

,
k
)
k
(Z) =
k
(

Z)
/
, Z
n
1
n
1

n
N
n
N
.
Therefore,
A
((Z) (E))u = ((Z) (E))
/
u H, u |.
Thus
A
H H. Moreover, H =
2
A
H
A
H, therefore
A
H = H. Set
1
:=
A
[H.
Clearly,
1
is an AUI on H, and
(
/

1
)(Z) = (

Z)
/
, Z
n
1
n
1

n
N
n
N
.
Let us verify that G
k
is (
/

1
,
n
k

,
k
)-real, k = 1, . . . , N.
(
n
k

,
k
)G
k
(Z)u
= (
n
k

,
k
)(I
n
k
P
k
)
_
(E) 0
0 I
1
_ _
I
/
(Z) (E)
_
u
= (
n
k

,
k
)(I
n
k
P
k
)
_
(E)
(Z) (E)
_
u = (
n
k

,
k
)
k
(Z)u
=
k
(

Z)
/
u = (I
n
k
P
k
)
_
(E) 0
0 I
1
_ _
I
/
(

Z) (E)
_
/
u
= G
k
(

Z)
/
u = G
k
(
/

1
)(Z)u.
Since the linear span of vectors of the form(Z)u, Z
n
1
n
1

n
N
n
N
, u
|, is dense in | H, the operator G
k
is (
/

1
,
n
k

,
k
)-real, as desired.
(ii) (i) Let f satisfy (ii). Then by Theorem 2.2 f B
n
1
,...,n
N
(|). Let us
show that the operator-valued linear function A(Z) from (2.2) is
/

1
-real.
Since G
k
is (
/

1
,
n
k

,
k
)-real, one has G
k
(
/

1
) = (
n
k

,
k
)G
k
and
(
/

1
)G
k
= G
k
(
n
k

,
k
), k = 1, . . . , N. The latter equality follows from the
fact that for every h C
n
k
/
k
, x | H:
(
/

1
)G
k
h, x)
/1
= (
/

1
)x, G
k
h)
/1
= G
k
(
/

1
)x, h)
C
n
k,
k
= (
n
k

,
k
)G
k
x, h)
C
n
k,
k
= (
n
k

,
k
)h, G
k
x)
C
n
k,
k
= G
k
(
n
k

,
k
)h, x)
/1
.
Therefore,
(
/

1
)A(

Z)(
/

1
) =
N
k=1
(
/

1
)G
k
(Z
k
I
,
k
)G
k
(
/

1
)
=
N
k=1
G
k
(
n
k

,
k
)(Z
k
I
,
k
)(
n
k

,
k
)G
k
=
N
k=1
G
k
(
n
k
Z
k
n
k
I
,
k
)G
k
=
N
k=1
G
k
(Z
k
I
,
k
)G
k
= A(Z).
The latter is equivalent to the identities
/
a(
Z)
/
= a(Z),
/
b(

Z)
1
= b(Z),
1
c(
Z)
/
= c(Z),
1
d(

Z)
1
= d(Z).
Since
2
1
= I
1
and
(
1
d(

Z)
1
1
) (
1
d(

Z)
1
) = (
1
d(

Z)
1
) (
1
d(

Z)
1
1
) = I
1
,
one has
1
d(

Z)
1
1
= (
1
d(

Z)
1
)
1
= d(Z)
1
.
Therefore,
f
(Z) =
/
f(
Z)
/
=
/
(a(
Z) b(

Z)d(

Z)
1
c(
Z))
/
=
/
a(
Z)
/
(
/
b(
Z)
1
) (
1
d(

Z)
1
1
) (
1
c(

Z)
/
)
= a(Z) b(Z)d(Z)
1
c(Z) = f(Z),
and f is
/
-real. Thus, (i) follows from (ii).
(iv) (i) Let (iv) hold. Then the operator U = U
= U
1
from (4.1) is
((
n
1
++n
N

)
/
)-real, i.e.,
_

n
1
++n
N

0
0
/
_
_
A B
C D
_
_

n
1
++n
N

0
0
/
_
=
_
A B
C D
_
.
This is equivalent to the following identities:
(
n
1
++n
N

)A(
n
1
++n
N

) = A, (
n
1
++n
N

)B
/
= B,
/
C(
n
1
++n
N

) = C,
/
D
/
= D.
Moreover, for W D
n
1
n
1
D
n
N
n
N
one has
(
n
1
++n
N

)(I
C
n
1
++n
N
A(P(W) I
))(
n
1
++n
N

)
= (
n
1
++n
N

)
2
(
n
1
++n
N

)A(
n
1
++n
N

)
(
n
1
++n
N

)(P(W) I
)(
n
1
++n
N

)
= I
C
n
1
++n
N
A(P(W) I
).
Therefore,
(
n
1
++n
N

)(I
C
n
1
++n
N
A(P(W) I
))
1
(
n
1
++n
N

)
= (I
C
n
1
++n
N
A(P(W) I
))
1
(we already used an analogous argument above). Thus,
T
(W) =
/
T(W)
/
=
/
_
D +C(P(W) I
)(I
C
n
1
++n
N
A(P(W) I
))
1
B
/
= D +C(P(W) I
)(I
C
n
1
++n
N
A(P(W) I
))
1
B
= T(W), W D
n
1
n
1
D
n
N
n
N
,
i.e., T is
/
-real. Applying the inverse double Cayley transform to T, one can see
that f is
/
-real on
n
1
n
1

n
N
n
N
, and hence on
n
1
,...,n
N
. Thus, (i)
follows from (iv).
(iii) (iv) Let f satisfy (2.3) with holomorphic (
/
,
n
k

,
k
)-real
L(|, C
n
k
/
k
)-valued functions
k
on
n
1
n
1

n
N
n
N
. As in the proof
of Theorem 4.1, we get for T = ((f) consecutively: identities (3.15) and (4.3)
with holomorphic L(|, C
n
k
/
k
)-valued functions
k
on
n
1
n
1

n
N
n
N
given by (3.14), (3.16) which are, moreover, (
/
,
n
k

,
k
)-real; then identi-
ties (4.4) and (4.5) with holomorphic L(C
n
k
/
k
, |)-valued functions

k
on
n
1
n
1

n
N
n
N
which are, moreover, (
n
k
,
k
,
/
)-real. Without loss of
generality, we consider all of /
k
s equal, i.e., set /:=
N
k=1
/
k
and regard
H
R
k
=
k
L(|, C
n
k
/), H
L
k
=

k
L(C
n
k
/, |),
,
:=
N
k=1
,
k
.
Then H
R
k
is (
/
,
n
k

,
)-real, and H
L
k
is (
n
k

,
,
/
)-real, k = 1, . . . , N.
Set A := /. Let us observe that the subspace T
0
(and hence, clos(T
0
)) is
((
n
1
++n
N

)
/
)-invariant. Indeed, for
x =
_
(P(W) I
)H
R
(W)
I
/
_
u +
_
H
R
(W
)
T(W
)
_
u
T
0
,
with some u, u
| and W, W
D
n
1
n
1
D
n
N
n
N
, we have
((
n
1
++n
N

)
/
)x
=
_
(
n
1
++n
N
P(W)
)H
R
(W)
/
_
u +
_
(
n
1
++n
N

)H
R
(W
/
T(W
)
_
u
=
_
(P(W)
n
1
++n
N

)H
R
(W)
/
_
u +
_
H
R
(W
)
/
T(W
)
/
_
u
=
_
(P(W) I
)H
R
(W)
I
/
_
/
u +
_
H
R
(W
)
T(W
)
_
/
u T
0
.
Therefore, the subspace
T
0
:= ((C
n
1
++n
N
A) |) clos(T
0
)
is also ((
n
1
++n
N

)
/
)-invariant. Indeed, for any h
1
T
0
, h
2
T
0
:
((
n
1
++n
N

)
/
)h
2
, h
1
) = ((
n
1
++n
N

)
/
)h
1
, h
2
) = 0,
thus h
2
T
0
implies ((
n
1
++n
N

)
/
)h
2
T
0
.
Now, it is easy to check that U
0
and therefore

U
0
are ((
n
1
++n
N
)
/
)-
real. Since U =

U
0
I
1
0
, and T
0
is ((
n
1
++n
N
)
/
)-invariant, the operator
U = U
= U
1
is ((
n
1
++n
N

)
/
)-real, as required.
The proof is complete.
References
[1] J. Agler. On the representation of certain holomorphic functions dened on a poly-
disc. In: Topics in operator theory: Ernst D. Hellinger memorial volume, volume 48
of Oper. Theory Adv. Appl., pages 4766. Birkhauser, Basel, 1990.
[2] C.-G. Ambrozie and D. Timotin. A von Neumann type inequality for certain domains
in C
n
. Proc. Amer. Math. Soc., 131(3) (2003), 859869 (electronic).
[3] J.A. Ball and V. Bolotnikov. Realization and interpolation for Schur-Agler-class func-
tions on domains with matrix polynomial dening function in C
n
. J. Functional
Analysis, to appear.
[4] M.F. Bessmertny. Functions of several complex variables in the theory of nite linear
structures. PhD thesis, Kharkov University, Kharkov, 1982. (Russian).
[5] M.F. Bessmertny. On realizations of rational matrix functions of several com-
plex variables. In: Interpolation theory, systems theory and related topics (Tel
Aviv/Rehovot, 1999), volume 134 of Oper. Theory Adv. Appl., pages 157185.
Birkhauser, Basel, 2002. Translated from the Russian by D. Alpay and V. Kat-
snelson.
variables. II. In: Reproducing kernel spaces and applications, volume 143 of Oper.
Theory Adv. Appl., pages 135146. Birkhauser, Basel, 2003. Translated from the
Russian by V. Katsnelson.
variables. III. In: Current trends in operator theory and its applications (Blacksburg
VA, 2002), volume 149 of Oper. Theory Adv. Appl., pp. 133138. Birkhauser, Basel,
2004. Translated from the Russian by D.S. Kalyuzhny-Verbovetzki.
[8] R.E. Curto. Applications of several complex variables to multiparameter spectral the-
ory. In: Surveys of some recent results in operator theory, Vol. II, volume 192 of
Pitman Res. Notes Math. Ser., pages 2590. Longman Sci. Tech., Harlow, 1988.
[9] D.S. Kalyuzhny-Verbovetzki. On the Bessmertny class of homogeneous positive
holomorphic functions of several variables. In: Current trends in operator theory and
its applications (Blacksburg VA, 2002), volume 149 of Oper. Theory Adv. Appl., pp.
255289. Birkh auser, Basel, 2004.
[10] G. Kron. Tensor analysis of networks. Wiley, New York, 1939.
[11] M. Putinar. Uniqueness of Taylors functional calculus. Proc. Amer. Math. Soc.,
89(4) (1983), 647650.
[12] B.V. Shabat. Introduction to complex analysis. Part II, volume 110 of Translations of
Mathematical Monographs. American Mathematical Society, Providence, RI, 1992.
Functions of several variables, Translated from the third (1985) Russian edition by
J. S. Joel.
[13] Z. Slodkowski and W.

Zelazko. On joint spectra of commuting families of operators.
Studia Math., 50 (1974), 127148.
[14] J.L. Taylor. The analytic-functional calculus for several commuting operators. Acta
Math., 125 (1970), 138.
[15] J.L. Taylor. A joint spectrum for several commuting operators. J. Functional Anal-
ysis, 6 (1970), 172191.
[16] F.-H. Vasilescu. A characterization of the joint spectrum in Hilbert spaces. Rev.
Roumaine Math. Pures Appl., 22(7) (1977), 10031009.
[17] F.-H. Vasilescu. A Martinelli type formula for the analytic functional calculus. Rev.
Roumaine Math. Pures Appl., 23(10) (1978), 15871605.
Dmitry S. Kalyuzhny-Verbovetzki
P.O. Box 653
e-mail: dmitryk@math.bgu.ac.il
Operator Theory:
Rational Solutions of the Schlesinger System
and Isoprincipal Deformations of
Rational Matrix Functions II
Victor Katsnelson and Dan Volok
Abstract. In this second article of the series we study holomorphic families
of generic rational matrix functions parameterized by the pole and zero loci.
In particular, the isoprincipal deformations of generic rational matrix func-
tions are proved to be isosemiresidual. The corresponding rational solutions
of the Schlesinger system are constructed and the explicit expression for the
related tau function is given. The main tool is the theory of joint system
representations for rational matrix functions with prescribed pole and zero
structures.
Mathematics Subject Classication (2000). Primary: 47A56;
Secondary: 34M55.
Keywords. Isoprincipal, isosemiresidual, joint system representation, Fuchsian
system, Schlesinger system.
Notation
C stands for the complex plane.
C
stands for the punctured complex plane:

C
= C 0.
C stands for the extended complex plane (= the Riemann sphere):
C = C .
z stands for the complex variable.
C
n
stands for the n-dimensional complex space.
In the coordinate notation, a point t C
n
will be written as t = (t
1
, . . . , t
n
).
Victor Katsnelson thanks Ruth and Silvia Shoham for endowing the chair that supports his
respective research. Victor Katsnelson was also supported by the Minerva foundation.
166 V. Katsnelson and D. Volok
C
n
is the set of points t C

n
, whose coordinates t
1
, . . . , t
n
are pairwise
dierent:
C
n
= C
n
_
1i,jn
i,=j
t : t
i
= t
j
.
C
mn
stands for the set of all mn matrices with complex entries.
For A C
mn
, A
C
nm
is the adjoint matrix, Im(A) is the image subspace
of A in C
m
(= the linear span of the columns of A) and Nul(A) is the null
subspace of A in C
n
.
[, ] denotes the commutator: for A, B C
mm
, [A, B] = AB BA.
I stands for the identity matrix of an appropriate dimension.
10. Simple singularity of a meromorphic matrix function
(For Sections 19, see Operator Theory: Advances and Applications, Vol. 149,
pp. 291348.)
Denition 10.1. Let R(z) be a C
mm
-valued function, holomorphic in a punctured
neighborhood of a point t C. The point t is said to be a simple pole of the matrix
function R if
R(z) =
R
t
z t
+H(z),
where R
t
C
mm
is a constant matrix and the function H is holomorphic at the
point t. The matrix R
t
is said to be the residue of the function R at the point
t. Furthermore, if r = rank(R
t
) and f
t
C
mr
and g
t
C
rm
are matrices
providing the factorization R
t
= f
t
g
t
, we shall say that f
t
is the left semiresidue
of R at t and g
t
is the right semiresidue of R at t.
Remark 10.2. The left and right semiresidues f
t
, g
t
are dened up to the transfor-
mation
f
t
f
t
c, g
t
c
1
g
t
,
where c C
rr
is an invertible matrix.
mm
-valued function, holomorphic and invertible
in a punctured neighborhood of a point t C.
1. The point t is said to be regular for the function R if both the function R
and the inverse function R
1
are holomorphic functions in the entire (non-
punctured) neighborhood of the point t, i.e., if R and R
1
are holomorphic
at the point t.
2. The point t is said to be singular for the function R if at least one of the
functions R and R
1
is not holomorphic at the point t.
In particular, the point t is singular for the function R if R is holomorphic at the
point t, but its value R(t) is a degenerate matrix. In this case, the point t is said
to be a zero of the function R.
Isoprincipal Deformations of Rational Matrix Functions II 167
mm
in a punctured neighborhood of a point t C, and let t be a singular point of R.
The singular point t is said to be simple if one of the following holds:
1. The point t is a simple pole of the function R and a holomorphy point of the
inverse function R
1
.
2. The point t is a simple pole of the inverse function R
1
and a holomorphy
point of the function R itself.
Remark 10.5. Note that, according to Denition 10.4, if t is a simple singular
point of the function R then R is a single-valued meromorphic function in the
entire (non-punctured) neighborhood of t.
Our main goal is to study a matrix function in a neighborhood of its sim-
ple singular point from the point of view of linear dierential systems. Thus we
consider the left logarithmic derivative of the function R:
Q
l
R
(z)
def
= R
(z)R(z)
1
.
Remark 10.6. One can also consider the right logarithmic derivative of R:
Q
r
R
(z) = R(z)
1
R
(z).
But then Q
r
R
is the left logarithmic derivative of the inverse function R
1
:
Q
l
R
1
(z) = (R
1
(z))
R(z) = R(z)
1
R
(z)R(z)
1
R(z) = Q
r
R
(z).
Thus in this work we shall deal mainly with left logarithmic derivatives. Therefore,
we shall use the notation Q
R
instead of Q
l
R
:
Q
R
(z)
def
= R
(z)R(z)
1
,
and omit the word left when referring to the left logarithmic derivative.
Proposition 10.7. Let R(z) be a C
mm
in a punctured neighborhood of a point t C, and let t be a simple singular point
of R. Then the point t is a simple pole for the logarithmic derivative
1
Q
R
of R.
Moreover, for the residue and the constant term of the Laurent expansion
Q
R
(z) =
Q
t
z t
+C +o(1) as z t (10.1)
the following relations hold.
1. If t is a pole of R then
Q
2
t
= Q
t
, (10.2a)
Q
t
CQ
t
= CQ
t
(10.2b)
and
Im(Q
t
) = Im(R
t
), (10.3)
where R
t
is the residue of R at t.
1
See Remark 10.6.
2. If t is a zero of R then
Q
2
t
= Q
t
, (10.4a)
Q
t
CQ
t
= Q
t
C (10.4b)
and
Nul(Q
t
) = Nul(R
t
), (10.5)
where R
t
is the residue of R
1
at t.
Proof. First, let us assume that t is a pole of R and let
R(z) =
R
t
z t
+A
0
+A
1
(z t)) +A
2
(z t)
2
+ , (10.6)
R
1
(z) = B
0
+B
1
(z t) +B
2
(z t)
2
+ , (10.7)
be the Laurent expansions of the functions R and R
1
at t. Then
R
(z) =
R
t
(z t)
2
+A
1
+ 2A
2
(z t) + (10.8)
Multiplying the Laurent expansions term by term, we obtain from (10.7) and (10.8)
Q
R
(z) =
R
t
B
0
(z t)
2

R
t
B
1
z t
R
t
B
2
+A
1
B
0
+o(1). (10.9)
Substituting the expansions (10.6), (10.7) into the identity
R
1
(z)R(z) = R(z)R
1
(z) = I,
we observe that
R
t
B
0
= B
0
R
t
= 0 and R
t
B
1
+A
0
B
0
= I. (10.10)
Hence the rst term of the expansion (10.9) vanishes and we obtain the expansion
(10.1) with
Q
t
= R
t
B
1
= A
0
B
0
I, (10.11)
C = R
t
B
2
+A
1
B
0
. (10.12)
Thus
(I +Q
t
)Q
t
= (A
0
B
0
)(R
t
B
1
) = A
0
(B
0
R
t
)B
1
= 0,
i.e., (10.2a) holds. Furthermore,
(I +Q
t
)CQ
t
= (A
0
B
0
)(R
t
B
2
+A
1
B
0
)(R
t
B
1
) =
= A
0
(B
0
R
t
)B
2
R
t
B
1
A
0
B
0
A
1
(B
0
R
t
)B
1
= 0,
i.e., (10.2b) holds as well. Finally,
Q
t
R
t
= (A
0
B
0
I)R
t
= A
0
(B
0
R
t
) R
t
= R
t
,
which, together with (10.11), implies (10.3). This completes the proof in the case
when t is a pole of R. The case when t is a zero of R can be treated analogously.
Remark 10.8. Since for any p q matrix A the subspace Nul(A) is the orthogonal
complement of the subspace Im(A
) in C
q
, the relation (10.5) can be rewritten as
Im(Q
t
) = Im(R
t
).
The latter relation, together with (10.4a), means that Q
t
is a (non-orthogonal, in
general) projector onto the subspace Im(R
t
) C
m
. Hence the right semiresidue
of R
1
at its pole t is also the right semiresidue of Q
R
at t.
Analogously, the relations (10.2a) and (10.3) mean that Q
t
is a (non-
orthogonal, in general) projector onto the subspace Im(R
t
) C
m
. Hence the
left semiresidue of R at its pole t is also the left semiresidue of Q
R
at t.
Proposition 10.7 implies that a C
mm
-valued function R(z) in a punctured
neighborhood of its simple singular point t may be viewed as a fundamental solu-
tion of a linear dierential system
R
(z) = Q(z)R(z), (10.13)

for which t is a Fuchsian singularity (see the rst part of this work [KaVo] for
details and references) and whose coecients satisfy the relations (10.2) or (10.4).
The next proposition shows that (10.2) or (10.4) are the only requirements a
dierential system (10.13) with a Fuchsian singularity t has to satisfy in order for
its fundamental solution in a punctured neighborhood of t to be single-valued and
have a simple singular point at t:
Proposition 10.9. Let Q(z) be a C
mm
-valued function, holomorphic and single-
valued in a punctured neighborhood of a point t. Let the point t be a simple pole
for Q(z), let
Q(z) =
Q
t
z t
+C +o(1) as z t (10.14)
be the Laurent expansion of the function Q at the point t and let R be a fundamental
solution of the linear dierential system
R
(z) = Q(z)R(z), z . (10.15)

Assume that one of the following two cases takes place.
1. The coecients Q
t
, C of the expansion (10.14) satisfy the relations
Q
2
t
= Q
t
, (10.16a)
Q
t
CQ
t
= CQ
t
. (10.16b)
2. The coecients Q
t
, C of the expansion (10.14) satisfy the relations
Q
2
t
= Q
t
, (10.17a)
Q
t
CQ
t
= Q
t
C. (10.17b)
Then R is a single-valued function in and t is a simple singular point of R; in
the rst case t is a pole of R, in the second case t is a zero of R.
Proof. Once again, we shall prove only the rst statement. Thus we assume that
the relations (10.2a), (10.2b) hold and consider the transformation
U(z) = (I +Q
t
+ (z t)Q
t
)R(z).
Then, because of (10.16a), the inverse transformation is given by
R(z) = (I +Q
t
+ (z t)
1
Q
t
)U(z).
Substituting these formulae and the Laurent expansion of M into the linear system
(10.15), we obtain the following linear system for U:
U
(z) =
_
(I +Q
t
)CQ
t
z t
+V (z)
_
U(z),
where the function V (z) is holomorphic in the entire (non-punctured) neighbor-
hood of the point t. In view of (10.16b), the coecients of this system are holo-
morphic at the point t, hence U is holomorphic and invertible in the entire neigh-
borhood of t and R has a simple pole at t. Since
R
1
(z) = U
1
(z)(I +Q
t
+ (z t)Q
t
),
R
1
is holomorphic at t and hence has a zero at t.
An important role in the theory of Fuchsian dierential systems is played by
multiplicative decompositions of fundamental solutions (see Section 5 of [KaVo]).
In the present setting we are interested in decompositions of the following form:
mm
in a punctured neighborhood of a point t. Let R admit in the factorization
R(z) = H
t
(z)E
t
(), = z t, z , (10.18)
where the factors H
t
(z) and E
t
() possess the following properties:
1. H
t
(z) is a C
mm
-valued function, holomorphic and invertible in the entire
neighborhood t ;
2. E
t
() is a C
mm
-valued function, holomorphic and invertible in the punc-
tured plane C
= C 0.
Then the functions E
t
() and H
t
(z) are said to be, respectively, the principal and
regular factors of R at t.
Remark 10.11. The multiplicative decomposition (10.18), which appears in De-
nition 10.10, is always possible. This follows, for example, from the results due to
G.D.Birkho (see [Birk1]). The principal factor E
t
() is, in a sense, the multiplica-
tive counterpart of the principal part of the additive (Laurent) decomposition: it
contains the information about the nature of the singularity t of R. Of course,
the principal and regular factors at the point t are determined only up to the
transformation
E
t
() M()E
t
(), H
t
(z) H
t
(z)M
1
(z t), (10.19)
where M(z) is an invertible entire C
mm
-valued function. However, once the choice
of the principal factor E
t
is xed, the regular factor H
t
is uniquely determined and
vice-versa.
A possible choice of the principal factor of the function R at its simple singular
point t is described in the following
Lemma 10.12. Let R(z) be a C
mm
in a punctured neighborhood of a point t and let t be a simple singular point of R.
Then a principal factor E
t
() of R at t can be chosen as follows.
1. If t is a pole of R, choose any matrix L C
mm
, satisfying the conditions
L
2
= L, Nul(L) = Nul(R
t
), (10.20)
where R
t
is the residue of R at t, and set for C
E
t
() = I +L
1
L. (10.21)
2. If t is a zero of R, choose any matrix L C
mm
, satisfying the conditions
L
2
= L, Im(L) = Im(R
t
), (10.22)
where R
t
is the residue of R
1
at t, and set for C
E
t
() = I L +L. (10.23)
Proof. Let us assume that t is a pole of R and that the function E
t
is given
by (10.21), where the matrix L satises the conditions (10.20). Then E
t
() is
holomorphic in C
; its inverse E
1
t
() is given by
E
1
t
() = I +L L
and is holomorphic in C
, as well. Let us now show that the function

H(z)
def
= R(z)E
1
t
(z t)
is holomorphic and invertible at t.
Indeed, in a neighborhood of t the principal part of the Laurent expansion
of H equals to
R
t
(I +L)
z t
. But by (10.20) Im(L
) = Im(R
t
) and hence
Im((I +L
)R
t
) = Im((I +L
)L
) = Im((L
2
+ L)
) = 0.
Therefore, R
t
(I +L) = 0 and H is holomorphic at t.
In the same way, the principal part of the Laurent expansion of H
1
equals
to
LB
0
z t
, where B
0
= R
1
(t) is the constant term of the Laurent expansion of
R
1
at t. But R
t
B
0
= 0 (see (10.10) in the proof of Proposition 10.7), hence
Im(B
0
L
) = Im(B
0
R
t
) = 0,
LB
0
= 0 and H
1
is holomorphic at t, as well.
The proof in the case when t is a zero of R is completely analogous.
Remark 10.13. Let us note that the formulae (10.21) and (10.23) can be rewritten
in the unied form
E
t
() =
L
(= e
Llog
).
This is precisely the form of the principal factor (with

Q = 0) which appears in
Proposition 5.6 of [KaVo].
Remark 10.14. The relations (10.20) mean that L
is a projector onto Im(R
t
).
This is equivalent to L being of the form L = pg
t
, where g
t
is the right semiresidue
of the function R at its pole t and p C
mrank(R
t
)
is such that g
t
p = I. Anal-
ogously, the relations (10.22) mean that L is a projector onto Im(R
t
). This is
equivalent to L being of the form L = f
t
q, where f
t
is the left semiresidue of the
function R
1
at its pole t and q C
rank(R
t
)m
is such that qf
t
= I. For example,
one can choose the matrix L mentioned in Lemma 10.12 as follows:
L =
_
g
t
(g
t
g
t
)
1
g
t
if t is a pole of R,
f
t
(f
t
f
t
)
1
f
t
if t is a zero of R.
11. Rational matrix functions of simple structure
In this section we apply the local results obtained in Section 10 to the study of
rational matrix functions.
Denition 11.1. A C
mm
-valued rational function R(z) is said to be a rational
matrix function of simple structure if it meets the following conditions:
1. det R(z) , 0;
2. all singular points of R are simple;
3. z = is a regular point of R.
The set of all poles of the function R is said to be the pole set of the function R
and is denoted by T
R
. The set of all zeros of the function R is said to be the zero
set of the function R and is denoted by Z
R
.
Remark 11.2. Note that if R is a rational matrix function of simple structure then
the inverse function R
1
is a rational matrix function of simple structure, as well,
and Z
R
= T
R
1.
Below we formulate the global counterparts of Propositions 10.7 and 10.9
in order to characterize Fuchsian dierential systems whose fundamental solutions
are rational matrix functions of simple structure.
Theorem 11.3. Let R(z) be a rational matrix function of simple structure with
the pole set T
R
and the zero set Z
R
. Then its logarithmic derivative
2
Q
R
(z) is a
2
See Remark 10.6.
rational function with the set of poles T
R
Z
R
; all the poles of Q
R
are simple.
Furthermore, the function Q
R
admits the additive decomposition
Q
R
(z) =
t1
R
R
Q
t
z t
, (11.1)
and its residues Q
t
C
mm
satisfy the following relations:
t1
R
R
Q
t
= 0, (11.2)
Q
2
t
=
_
Q
t
if t T
R
,
Q
t
if t Z
R
,
(11.3)
Q
t
C
t
Q
t
=
_
C
t
Q
t
if t T
R
,
Q
t
C
t
if t Z
R
,
(11.4)
where
C
t
=
1
R
R
t
,=t
Q
t
t t
. (11.5)
Proof. Since both functions R and R
1
are holomorphic in C (T
R
Z
R
), the
logarithmic derivative Q
R
is holomorphic there, as well. According to Proposition
10.7, each point of the set T
R
Z
R
is a simple pole of Q
R
, hence we can write
for Q
R
the additive decomposition
Q
R
(z) = Q
R
() +
t1
R
R
Q
t
z t
,
where Q
t
are the residues of Q
R
. Since R is holomorphic at , the entries of its
derivative R
decay as o([z[
1
) when z . The rate of decay for the logarithmic
derivative Q
R
is the same, because R
1
, too, is holomorphic at . Thus we obtain
the additive decomposition (11.1) for Q
R
and the relation (11.2) for the residues
Q
t
. Now the relations (11.3), (11.4) follow immediately from Proposition 10.7,
once we observe that the matrix C
t
given by (11.5) is but the constant term of the
Laurent expansion of Q
R
at its pole t.
Theorem 11.4. Let T and Z be two nite disjoint subsets of the complex plane C
and let Q(z) be a C
mm
-valued rational function of the form
Q(z) =
t1
Q
t
z t
, (11.6)
where Q
t
C
mm
. Let the matrices Q
t
satisfy the relations
t1
Q
t
= 0, (11.7)
Q
2
t
=
_
Q
t
if t T,
Q
t
if t Z,
(11.8)
Q
t
C
t
Q
t
=
_
C
t
Q
t
if t T,
Q
t
C
t
if t Z,
(11.9)
where
C
t
=
1
t
,=t
Q
t
t t
. (11.10)
Let R(z) be a fundamental solution of the Fuchsian dierential system
R
(z) = Q(z)R(z). (11.11)

Then R is a rational matrix function of simple structure such that
T
R
= T, Z
R
= Z.
Proof. Since the condition (11.7) implies that the point is a regular point for
the Fuchsian system (11.11), we may, without loss of generality, consider the fun-
damental solution R satisfying the initial condition R() = I. Then R(z) is a
matrix function, holomorphic (a priori, multi-valued) and invertible in the (multi-
connected) set C (T ^). However, for t T ^ the function Q admits in a
neighborhood of t the Laurent expansion
Q(z) =
Q
t
z t
+C
t
+o(1)
with the constant term C
t
given by (11.10). The coecients Q
t
and C
t
satisfy the
relations (11.8), (11.9), hence by Proposition 10.9 the function R is meromorphic
at t. Since this is true for every t T^, the function R is rational (in particular,
single-valued). Proposition 10.9 also implies that every t T (respectively, t Z)
is a simple pole (respectively, a zero) of the function R and a zero (respectively, a
simple pole) of the inverse function R
1
. Therefore, R is a rational matrix function
of simple structure with the pole set T and the zero set Z.
We close this section with the following useful
Lemma 11.5. Let R be a rational matrix function of simple structure. For t
T
R
Z
R
let R
t
denote the residue of the function R at t if t T
R
, and the residue
of the inverse function R
1
at t if t Z
R
. Then
t1
R
rank(R
t
) =
t
R
rank(R
t
). (11.12)
Proof. Let us consider the logarithmic derivative Q
R
of R. Its residues Q
t
satisfy
by Theorem 11.3 the relations (11.2) and (11.3). From (11.3) it follows that
rank(Q
t
) =
_
trace(Q
t
) if t T
R
,
trace(Q
t
) if t Z
R
.
But (11.2) implies
t1
R
trace(Q
t
) +
t
R
trace(Q
t
) = 0,
hence
t1
R
rank(Q
t
) =
t
R
rank(Q
t
).
Finally, by Proposition 10.7 (see (10.3), (10.5) there),
rank(R
t
) = rank(Q
t
), t T
R
Z
R
.
Thus (11.12) holds.
12. Generic rational matrix functions
Denition 12.1. A C
mm
-valued rational function R(z) is said to be a generic
3
rational matrix function if R is a rational matrix function of simple structure and
all the residues of the functions R and R
1
have rank one.
Lemma 12.2. Let R be a generic rational matrix function. Then the cardinalities
of its pole and zero sets
4
coincide:
#T
R
= #Z
R
. (12.1)
Proof. Since all the residues of R and R
1
are of rank one, the statement follows
immediately from Lemma 11.5.
Let R be a C
mm
-valued generic rational function. In what follows, we assume
that R is normalized by
R() = I. (12.2)
Let us order somehow the pole and zero sets of R:
T
R
= t
1
, . . . , t
n
, Z
R
= t
n+1
, . . . , t
2n
, (12.3)
where n = #T
R
= #Z
R
. Then we can write for R and R
1
the additive decom-
positions
R(z) = I +
n
k=1
R
k
z t
k
, (12.4a)
R
1
(z) = I +
2n
k=n+1
R
k
z t
k
, (12.4b)
3
In [Kats2] such functions are called rational matrix functions in general position.
4
See Denition 11.1.
where for 1 k n (respectively, n + 1 k 2n) we denote by R
k
the residue
of R (respectively, R
1
) at its pole t
k
. Since each matrix R
k
is of rank one, the
representations (12.4) can be rewritten as
R(z) = I +
n
k=1
f
k
1
z t
k
g
k
, (12.5a)
R
1
(z) = I +
2n
k=n+1
f
k
1
z t
k
g
k
, (12.5b)
where for 1 k n (respectively, n +1 k 2n) f
k
C
m1
and g
k
C
1m
are
the left and right semiresidues
5
of R (respectively, R
1
) at t
k
. Furthermore, we
introduce two n n diagonal matrices:
A
1
= diag(t
1
, . . . , t
n
), A
= diag(t
n+1
, . . . , t
2n
), (12.6)
two mn matrices :
F
1
=
_
f
1
. . . f
n
_
, F
=
_
f
n+1
. . . f
2n
_
, (12.7)
and two n m matrices :
G
1
=
_
_
_
g
1
.
.
.
g
n
_
_
_
, G
=
_
_
_
g
n+1
.
.
.
g
2n
_
_
_
. (12.8)
The matrices A
1
and A
are said to be, respectively, the pole and zero matrices

of R. The matrices F
1
and G
1
are said to be, respectively, the left and right pole
semiresidual matrices of R. Analogously, the matrices F
and G
are said to be
the left and right zero semiresidual matrices of R.
Remark 12.3. It should be mentioned that for a xed ordering (12.3) of the pole
and zero sets the pole and the zero matrices A
1
and A
A
are dened uniquely, and
the semiresidual matrices F
1
, G
1
, F
, G
are dened essentially uniquely, up to

the transformation
F
1
F
1
D
1
, G
1
D
1
1
G
1
, (12.9a)
F
, G
D
1
, (12.9b)
where D
1
, D
C
nn
are arbitrary invertible diagonal matrices. Once the choice
of the left pole semiresidual matrix F
1
is xed, the right pole semiresidual matrix
G
1
is determined uniquely, etc.
In terms of the matrices A
1
, A
, F
1
, G
, F
, G
, the representations (12.5)

take the following form:
R(z) = I +F
1
(zI A
1
)
1
G
1
, (12.10a)
R
1
(z) = I +F
(zI A
)
1
G
. (12.10b)
5
See Denition 10.1.
The representations (12.10) are not quite satisfactory for the following rea-
sons. Firstly, in view of the identity RR
1
= R
1
R = I, the information contained
in the pair of representations (12.10) is redundant: each of these representations
determines the function R (and R
1
) uniquely. Secondly, if, for example, the di-
agonal matrix A
1
and the matrices F
1
, G
1
of appropriate dimensions are chosen
arbitrarily then the rational function (12.10a) need not be generic. In our investi-
gation we shall mainly use another version of the system representation of rational
matrix functions, more suitable for application to linear dierential equations. This
is the so-called joint system representation (see [Kats2] for details and references)
of the function R(z)R
1
() of two independent variables z and w. A key role in
the theory of the joint system representation is played by the Lyapunov equations.
These are matricial equations of the form
UX XV = Y, (12.11)
where the matrices U, V, Y C
nn
are given, and the matrix X C
nn
is un-
known. If the spectra of the matrices U and V are disjoint, then the Lyapunov
equation (12.11) is uniquely solvable with respect to X for arbitrary right-hand
side Y . The solution X can be expressed, for example, as the contour integral
X =
1
2i
_
(zI U)
1
Y (zI V )
1
dz, (12.12)
where is an arbitrary contour, such that the spectrum of U is inside and the
spectrum of V is outside (see, for instance, Chapter I, Section 3 of the book
[DaKr]).
With the generic rational function R we associate the following pair of Lya-
punov equations (with unknown S
1
, S
1
C
nn
):
A
S
1
S
1
A
1
= G
F
1
, (12.13a)
A
1
S
1
S
1
A
= G
1
F
. (12.13b)
Since the spectra of the pole and zero matrices A
1
and A
do not intersect (these

are the pole and zero sets of R), the Lyapunov equations (12.13a) and (12.13b)
are uniquely solvable. In fact, since the matrices A
1
and A
are diagonal, the

solutions can be given explicitly (using the notation (12.6)(12.8)):
S
1
=
_
g
n+i
f
j
t
n+i
t
j
_
n
i,j=1
, S
1
=
_
g
i
f
n+j
t
i
t
n+j
_
n
i,j=1
. (12.14)
The matrices S
1
and S
1
are said to be, respectively, the zero-pole and pole-zero
coupling matrices of R.
Proposition 12.4. Let R(z) be a generic rational matrix function normalized by
R() = I. Then
1. the coupling matrices S
1
and S
1
of R are mutually inverse:
S
1
S
1
= S
1
S
1
= I; (12.15)
2. for the semiresidual matrices of R the following relations hold:
G
= S
1
G
1
, F
1
= F
S
1
; (12.16)
3. the function R admits the joint representation
R(z)R
1
() = I + (z )F
1
(zI A
1
)
1
S
1
1
(I A
)
1
G
, (12.17)
where A
1
, A
are the pole and zero matrices of R.

Proof. The proof of Proposition 12.4 can be found in [Kats2].
Remark 12.5. Note that, since R() = I, one can recover from the joint repre-
sentation (12.17) when z or the separate representations
R(z) = I F
1
(zI A
1
)
1
S
1
1
G
, (12.18a)
R
1
() = I +F
1
S
1
1
(I A
)
1
G
, (12.18b)
which, in view of (12.16), coincide with the representations (12.10).
Remark 12.6. In view of (12.15), (12.16), one can also write the joint representa-
tion for R in terms of the matrices F
, G
1
and the solution S
1
of the Lyapunov
equation (12.13b):
R(z)R
1
() = I (z )F
S
1
1
(zI A
1
)
1
S
1
(I A
)
1
S
1
1
G
1
. (12.19)
Thus we may conclude that the pole and zero sets together with a pair of the
semiresidual matrices (either right pole and left zero or left pole and right zero)
determine the normalized generic rational function R uniquely.
Remark 12.7. The theory of system representations for rational matrix functions
with prescribed zero and pole structures rst appeared in [GKLR], and was further
developed in [BGR1], [BGR2], and [BGRa].
The joint representations (12.17), (12.19) suggest that this theory can be
applied to the investigation of families of rational functions parameterized by the
zeros and poles loci and the corresponding
6
deformations of linear dierential
systems. The version of this theory adapted for such applications was presented
in [Kats1] and [Kats2]. Also in [Kats2] one can nd some historical remarks and a
list of references.
Proposition 12.8. Let R(z) be a generic rational matrix function normalized by
R() = I. Then its logarithmic derivative
7
admits the representation:
R
(z)R
1
(z) = F
1
(zI A
1
)
1
S
1
1
(zI A
)
1
G
(12.20)
where A
1
and A
are the pole and zero matrices of R; F

1
and G
are the left pole

and right zero semiresidual matrices of R; S
1
is the zero-pole coupling matrix
of R.
6
See Propositions 11.3, 11.4.
7
See Remark 10.6.
Proof. Dierentiating (12.17) with respect to z, we obtain
R
(z)R
1
() = F
1
_
I (z )(zI A
1
)
1
_
(zI A
1
)
1
S
1
1
(I A
)
1
G
.
Now set = z to obtain (12.20).
Remark 12.9. The representation (12.20) for the logarithmic derivative Q
R
of the
normalized generic rational matrix function R can also be rewritten in terms of
the matrices F
, G
1
and the solution S
1
of the Lyapunov equation (12.13b) (see
Remark 12.6):
R
(z)R
1
(z) = F
S
1
1
(zI A
1
)
1
S
1
(zI A
)
1
S
1
1
G
1
. (12.21)
13. Generic rational matrix functions with prescribed local data
In the previous section we discussed the question, how to represent a generic
rational function R in terms of its local data (the pole and zero sets and the
residues). The main goal of this section is to construct a (normalized) generic
rational matrix R(z) function with prescribed local data. In view of Proposition
12.4, Remark 12.6 and Remark 12.3, such data should be given in the form of two
diagonal matrices of the same dimension (the pole and zero matrices) and two
semiresidual matrices of appropriate dimensions (either right pole and left zero or
right zero and left pole)
8
.
Thus we consider the following
Problem 13.1. Let two diagonal matrices
A
1
= diag(t
1
, . . . , t
n
), A
= diag(t
n+1
, . . . , t
2n
), t
i
,= t
j
unless i = j,
and two matrices: F C
mn
, G C
nm
, be given.
ZT-version: Find a generic C
mm
-valued rational function R(z) such that
1. R() = I;
2. the matrices A
1
and A
are, respectively, the pole and zero matrices of R;

3. the matrix F is the left pole semiresidual matrix F
1
of R: F = F
1
;
4. the matrix G is the right zero semiresidual matrix G
of R: G = G
.
TZ-version: Find a generic C
mm
-valued rational function R(z) such that
1. R() = I;
2. the matrices A
1
and A
are, respectively, the pole and zero matrices of R;

3. the matrix F is the left zero semiresidual matrix F
of R: F = F
;
4. the matrix G is the right pole semiresidual matrix G
1
of R: G = G
1
.
Proposition 13.2.
1. The ZT-version of Problem 13.1 is solvable if and only if the solution S of
the Lyapunov equation
A
S SA
1
= GF (13.1)
8
Here we use the terminology introduced in Section 12.
2. The TZ-version of Problem 13.1 is solvable if and only if the solution S of
the Lyapunov equation
A
1
S SA
= GF (13.2)
Proof. The proof of Proposition 13.2 can be found in [Kats2]. Here we would like
to note that the solutions of the Lyapunov equations (13.1) and (13.2) can be
written explicitly as, respectively,
S =
_
g
i
f
j
t
n+i
t
j
_
n
i,j=1
and S =
_
g
i
f
j
t
i
t
n+j
_
n
i,j=1
, (13.3)
where g
i
is the ith row of the matrix G and f
j
is the jth column of the matrix
F. Note also that the necessity of S being invertible in both cases follows from
Proposition 12.4.
In view of Proposition 13.2, we propose the following terminology:
Denition 13.3. Let A
1
, A
, F, G be the given data of Problem 13.1. Then:

1. the solution S of the Lyapunov equation
A
S SA
1
= GF (13.4)
is said to be the ZT-coupling matrix related to the data A
1
, A
, F, G;
2. the solution S of the Lyapunov equation
A
1
S SA
= GF (13.5)
is said to be the TZ-coupling matrix related to the data A
1
, A
, F, G;
3. the data A
1
, A
, F, G are said to be ZT-admissible if the ZT-coupling matrix

related to this data is invertible;
4. the data A
1
, A
, F, G are said to be TZ-admissible if the TZ-coupling matrix

related to this data is invertible.
Proposition 13.4. Let A
1
, A
, F, G be the given data of Problem 13.1.

1. If the data A
1
, A
, F, G are ZT-admissible then the ZT-version of Problem

13.1 has the unique solution R(z) given by
R(z) = I F(I A
1
)
1
S
1
G, (13.6)
where S is the ZT-coupling matrix related to the data A
1
, A
, F, G. The
logarithmic derivative of R is given by
R
(z)R
1
(z) = F(zI A
1
)
1
S
1
(zI A
)
1
G. (13.7)
2. If the data A
1
, A
, F, G are TZ-admissible then the TZ-version of Problem

13.1 has the unique solution R(z) given by
R(z) = I +FS
1
(zI A
1
)
1
G, (13.8)
where S is the TZ-coupling matrix related to the data A
1
, A
, F, G. The
logarithmic derivative of R is given by
R
(z)R
1
(z) = FS
1
(zI A
1
)
1
S(zI A
)
1
S
1
G. (13.9)
Proof. If the data A
1
, A
, F, G are ZT-admissible then, according to Proposition

13.2, the ZT-version of Problem 13.1 has a solution R. Then the ZT-coupling
matrix S related to the data A
1
, A
, F, G is the zero-pole coupling matrix of R.

According to Proposition 12.4, the function R admits the representation (13.6) and
hence
9
is determined uniquely. Now the representation (13.7) for the logarithmic
derivative follows from Proposition 12.8.
Analogous considerations hold also in the case when the data A
1
, A
, F, G
are TZ-admissible (see Remarks 12.6, 12.9).
It was already mentioned (see Remark 12.3) that in the denition of the
semiresidual matrices there is a certain freedom. Accordingly, certain equivalency
classes rather than individual matrices F, G should serve as data for Problem 13.1.
The appropriate denitions are similar to the denition of the complex projective
space P
k1
as the space of equivalency classes of the set C
k
0 (two vectors
h
, h
C
k
0 are declared to be equivalent if h
, h
are proportional, i.e.,

h
= dh
for some d C
).
Denition 13.5. 1. Let C
mn
,c
denote the set of m n matrices which have no
zero columns. Two matrices F
, F
C
mn
,c
are declared to be equivalent:
F
c
F
, if
F
= F
D
c
, (13.10)
where D
c
is a diagonal invertible matrix.
The space P
(m1)n
c
is a factor-set of the set C
mn
,c
modulo the equivalency
relation
c
.
2. Let C
nm
,r
denote the set of n m matrices which have no zero rows. Two
matrices G
, G
C
nm
,r
are declared to be equivalent: G
r
G
, if
G
= D
r
G
, (13.11)
where D
r
is a diagonal invertible matrix.
The space P
n(m1)
r
is a factor-set of the set C
nm
,r
modulo the equivalency
relation
r
.
The factor spaces P
(m1)n
c
and P
n(m1)
r
inherit topology from the spaces C
mn
,c
and C
nm
,r
, respectively. They can be provided naturally with the structure of
complex manifolds.
If F
and F
are two
c
- equivalent mn matrices, and G
and G
are two
r
- equivalent nm matrices, then the solutions S
, S
of the Lyapunov equation

9
See Remark 12.5.
(13.1) with F
, G
, G
, substituted instead of F, G, and the same A

1
, A
are
related by
S
= D
r
S
D
c
, (13.12)
where D
c
, D
r
are the invertible diagonal matrices, which appear in (13.10), (13.11).
Similar result holds also for the Lyapunov equations (13.2).
However, since diagonal matrices commute, the expressions on the right-
hand side of (13.6) will not be changed if we replace the matrices F, G, S with the
matrices FD
c
, D
r
G, D
r
SD
c
, respectively.
Thus, the following result holds:
Proposition 13.6. Given A
1
and A
, solution of Problem 13.1 depends not on the

matrices F, G themselves but on their equivalency classes in P
(m1)n
c
, P
n(m1)
r
.
Remark 13.7. In view of Remark 12.3, if R is a generic rational matrix function
then its left and right pole semiresidual matrix F
1
and G
1
can be considered
separately as elements of the sets P
(m1)n
c
and P
n(m1)
r
, respectively. However,
simultaneously the matrices F
1
and G
1
can not be considered so. The same holds
for the pair of the zero semiresidual matrices, as well.
14. Holomorphic families of generic rational matrix functions
Denition 14.1. Let D be a domain
10
in the space C
2n
and for every t =

(t
1
, . . . , t
2n
) D let R(z, t) be a generic C
mm
-valued rational function of z
with the pole and zero matrices
A
1
(t) = diag(t
1
, . . . , t
n
), A
(t) = diag(t
n+1
, . . . , t
2n
). (14.1)
Assume that for every t
0
D and for every xed z Ct
0
1
, . . . , t
0
2n
the function
R(z, t) is holomorphic with respect to t in a neighborhood of t
0
. Assume also that
R(, t) I. (14.2)
Then the family R(z, t)
tD
is said to be a normalized holomorphic family of
generic rational functions parameterized by the pole and zero loci.
Given a normalized holomorphic family R(z, t)
tD
of generic rational func-
tions parameterized by the pole and zero loci, we can write for each xed t D the
following representations for the functions R(z, t), R
1
(z, t) and the logarithmic
derivative
Q
R
(z, t)
def
=
R(z, t)
z
R
1
(z, t) (14.3)
10
One can also consider a Riemann domain over C
2n
(see Denition 5.4.4 in [Hor]).

(see (11.1), (12.4)):
R(z, t) = I +
n
k=1
R
k
(t)
z t
k
, (14.4a)
R
1
(z, t) = I +
2n
k=n+1
R
k
(t)
z t
k
, (14.4b)
Q
R
(z, t) =
2n
k=1
Q
k
(t)
z t
k
. (14.4c)
The residues R
k
(t), Q
k
(t), considered as functions of t, are dened in the whole
domain D. It is not hard to see that these functions are holomorphic in D:
Lemma 14.2. Let D be a domain in C
2n
and let R(z, t)

tD
be a normalized
holomorphic family of generic rational functions, parameterized by the pole and
zero loci. For each xed t D and 1 k n (respectively, n + 1 k 2n) let
R
k
(t) be the residue of the rational function R(, t) (respectively, R
1
(, t)) at its
pole t
k
. Likewise, for each xed t D and 1 k 2n let Q
k
(t) be the residue of
the logarithmic derivative Q
R
(, t) at its pole t
k
. Then R
k
(t), Q
k
(t) considered as
functions of t are holomorphic in D.
Proof. Let us choose an arbitrary t
0
D and n pairwise distinct points z
1
, . . . , z
n
in C t
0
1
, . . . , t
0
n
. From the expansion (14.4a) we derive the following system of
linear equations with respect to the residue matrices R
k
(t):
n
k=1
R
k
(t)
z
t
k
= R(z
, t) I, = 1, . . . , n. (14.5)
The matrices R(z
, t) I on the right-hand side of the system (14.5) are holo-

morphic with respect to t in a neighborhood of t
0
. The determinant of this linear
system
(t) = det
_
1
z
t
k
_
1,kn
is holomorphic in a neighborhood of t
0
, as well. In fact, the determinant (t)
(known as the Cauchy determinant) can be calculated explicitly (see, for example,
[[PS], part VII, Section 1, No. 3]):
(t) =
1p<qn
(z
p
z
q
)(t
p
t
q
)
n
,k=1
(z
t
k
)
.
In particular, (t
0
) ,= 0. Hence, for k = 1, . . . , n, the functions R
k
(t) are holomor-
phic in a neighborhood of t
0
. Since this is true for any t
0
D, these functions are
holomorphic in the whole domain D. The proof for R
k
(t), when n + 1 k 2n,
and for Q
k
(t) is completely analogous.
Remark 14.3. Note that, on the one hand, the functions R(z, t), R
1
(z, t) are
rational with respect to z, and hence are holomorphic with respect to z in
C t
1
, . . . , t
2n
. On the other hand, for every xed z C, these functions are
holomorphic with respect to t in D
2n
k=1
t : t
k
= z. Thus, by Hartogs the-
orem (see for example [Shab], Chapter I, sections 2.3, 2.6), the matrix functions
R(z, t), R
1
(z, t) are jointly holomorphic in the variables z, t outside the singular
set
2n
k=1
x, t : t
k
= x. In view of Lemma 14.2, the same conclusion follows from
the representations (14.4a), (14.4b).
In order to employ the joint system representation techniques described in
Sections 12 and 13 in the present setting, we have to establish the holomorphy
not only of the residues but also of the semiresidues of R(, t). In view of Remarks
10.2 and 12.3, we have a certain freedom in denition of the semiresidues and
the semiresidual matrices. Thus one should take care in choosing the semiresidual
matrices of R(, t) for each xed t in order to obtain holomorphic functions of t.
In general, it is possible to dene the holomorphic semiresidues only locally (we
refer the reader to Appendix B of the present paper where the global holomorphic
factorization of a matrix function of rank one is discussed).
Lemma 14.4. Let M(t) be a C
mm
-valued function, holomorphic in a domain T
of C
N
, and let
rankM(t) = 1 t D. (14.6)
Then there exist a nite open covering |
p
m
p=1
of D, a collection f
p
(t)
m
p=1
of C
m1
-valued functions and a collection g
p
(t)
m
p=1
of C
1m
-valued functions
satisfying the following conditions.
1. For p = 1, . . . , m the functions f
p
(t) and g
p
(t) are holomorphic in |
p
.
2. Whenever |
p
|
p
,= , there exists a (scalar) function
p
,p
(t), holomor-
phic and invertible in |
p
|
p
, such that for every t |
p
|
p
f
p
(t) = f
p
(t)
p
,p
(t), g
p
(t) =
1
p
,p
(t)g
p
(t). (14.7)
3. For p = 1, . . . , m the function M(t) admits the factorization
M(t) = f
p
(t)g
p
(t), t U
p
. (14.8)
Proof. Let f
p
(t) be the pth column of the matrix M(t) and let
|
p
= t D : f
p
(t) ,= 0, p = 1, . . . , m. (14.9)
Then, in view of (14.6) and (14.9), |
p
m
p=1
is an open covering of D. Furthermore,
from (14.6) and (14.9) it follows that for 1 p, q m there exists a unique (scalar)
function
p,q
(t), holomorphic in |
p
, such that
f
q
(t) = f
p
(t)
p,q
(t), t |
p
. (14.10)
Now dene g
p
(t) as
g
p
(t) =
_
p,1
(t) . . .
p,m
(t)
_
, t |
p
. (14.11)
Then the function g
p
(t) is holomorphic in U
p
and, according to (14.10), the factor-
ization (14.8) holds for these f
p
(t) and g
p
(t). (14.10) also implies that whenever
|
p
|
p
,= we have
,p
(t)
p
,k
(t) =
p
,k
(t), t |
p
|
p
, 1 k m.
In particular,
,p
(t)
p
,p
(t) = 1, t |
p
|
p
,
and (14.7) follows.
Theorem 14.5. Let D be a domain in C
2n
and let R(z, t)

tD
be a normalized
holomorphic family of C
mm
-valued generic rational functions parameterized by
the pole and zero loci. Then there exist a nite open covering
11
T
A
of D,
two collections F
1,
(t)
A
, F
,
(t)
A
of C
mn
-valued functions and two
collections G
1,
(t)
A
, G
,
(t)
A
of C
nm
-valued functions satisfying the
following conditions.
1. For each A the functions F
1,
(t), F
,
(t), G
1,
(t), G
,
(t) are holo-
morphic in T
.
2. Whenever T
,= , there exist diagonal matrix functions D

1,
,
(t),
D
,
,
(t), holomorphic and invertible in T
, such that for every

t T
F
1,
(t) = F
1,
(t)D
1,
,
(t), G
1,
(t) = D
1
1,
(t)G
1,
(t), (14.12a)
F
,
(t) = F
,
(t)D
,
,
(t), G
,
(t) = D
1
,
(t)G
,
(t). (14.12b)
3. For each A and t T
the matrices F
1,
(t), F
,
(t), G
1,
(t), G
,
(t)
are, respectively, the left pole, left zero, right pole, right zero semiresidual
matrices of the generic rational function R(, t), i.e., the representations
R(z, t) = I +F
1,
(t) (zI A
1
(t))
1
G
1,
(t), (14.13a)
R
1
(z, t) = I +F
,
(t) (zI A
(t))
1
G
,
(t) (14.13b)
hold true for all t T
.
Proof. Let R
k
(t), k = 1, . . . , 2n, be the holomorphic residue functions as in Lemma
14.2. Since for every xed t the rational function R(, t) is generic, each matrix
R
k
(t) is of rank one. Hence there exists a nite open covering |
k,p
m
p=1
of D, such
that in each open set |
k,p
the function R
k
(t) admits the factorization R
k
(t) =
f
k,p
(t)g
k,p
(t) as in Lemma 14.4. Now it suces to dene A as the set of 2n-tuples
(p
1
, . . . , p
2n
) such that
2n
k=1
|
k,p
k
,= , the open covering T
A
of D by
T
(p
1
,...,p
2n
)
=
2n
k=1
|
k,p
k
,
11
The index runs over a nite indexing set A.
and the collections F
1,
(t)
A
, F
,
(t)
A
, G
1,
(t)
A
, G
,
(t)
A
by
F
1,(p
1
,...,p
2n
)
(t) =
_
f
1,p
1
(t) . . . f
n,p
n
(t)
_
,
F
,(p
1
,...,p
2n
)
(t) =
_
f
n+1,p
n+1
(t) . . . f
2n,p
2n
(t)
_
,
G
1,(p
1
,...,p
2n
)
(t) =
_
_
_
g
1,p
1
(t)
.
.
.
g
n,p
n
(t)
_
_
_
, G
,(p
1
,...,p
2n
)
(t) =
_
_
_
g
n+1,p
n+1
(t)
.
.
.
g
2n,p
2n
(t)
_
_
_
.
Denition 14.6. Let D be a domain in C

2n
and let R(z, t)

tD
be a nor-
malized holomorphic family of C
mm
-valued generic rational functions param-
eterized by the pole and zero loci. Let a nite open covering T
A
of D,
collections F
1,
(t)
A
, F
,
(t)
A
of C
mn
-valued functions and collec-
tions G
1,
(t)
A
, G
,
(t)
A
of C
nm
-valued functions satisfy the condi-
tions 1.3. of Theorem 14.5. Then the collections F
1,
(t)
A
, F
,
(t)
A
,
G
1,
(t)
A
, G
,
(t)
A
are said to be the collections of, respectively, the left
pole, left zero, right pole, right zero semiresidual functions related to the family
R(z, t)
tD
.
Now we can tackle the problem of recovery of a holomorphic family of generic
matrix functions from the semiresidual data. Once again, let D be a domain in
C
2n
and let R(z, t)

tD
be a normalized holomorphic family of C
mm
-valued
generic rational functions, parameterized by the pole and zero loci. Let collections
F
1,
(t)
A
, F
,
(t)
A
, G
1,
(t)
A
, G
,
(t)
A
be the collections of
the semiresidual matrices related to the family R(z, t)
tD
. Then for each A
and a xed t T
the matrices A
1
(t), A
(t) are the pole and zero matrices of

the generic rational function R(z, t), and F
1,
(t), F
,
(t), G
1,
(t), G
,
(t) are,
respectively, the left pole, left zero, right pole, right zero semiresidual matrices
of R(, t). The appropriate coupling matrices S
1,
(t), S
1,
(t) satisfying the
Lyapunov equations
A
(t)S
1,
(t) S
1,
(t)A
1
(t) = G
,
(t)F
1,
(t), (14.14a)
A
1
(t)S
1,
(t) S
1,
(t)A
(t) = G
1,
(t)F
,
(t) (14.14b)
are given by
S
1,
(t) =
_
g
n+i,
(t)f
j,
(t)
t
n+i
t
j
_
n
i,j=1
, S
1,
(t) =
_
g
i,
(t)f
n+j,
(t)
t
i
t
n+j
_
n
i,j=1
,
(14.15)
where for 1 k n g
k,
(t) is the kth row of G
1,
(t), g
n+k,
(t) is the kth row
of G
,
(t), f
k,
(t) is the kth column of F
1,
(t), f
n+k,
(t) is the kth column
of F
,
(t) (compare with similar expressions (12.7), (12.8), (12.14)). From the
explicit expressions (14.15) it is evident that S
1,
(t), S
1,
(t), considered as
functions of t, are holomorphic in T
. According to Proposition 12.4, the functions

S
1,
(t), S
1,
(t) are mutually inverse:
S
1,
(t)S
1,
(t) = S
1,
(t)S
1,
(t) = I, t T
, (14.16)
and the following relations hold:
G
,
(t) = S
1,
(t)G
1,
(t), F
1,
(t) = F
,
(t)S
1,
(t). (14.17)
Furthermore, for t T
the function R(z, t) admits the representation

R(z, t)R
1
(, t) = I+
+ (z )F
1,
(t)(zI A
1
(t))
1
S
1
1,
(t)(I A
(t))
1
G
,
(t), (14.18)
and, in view of Proposition 12.8, its logarithmic derivative with respect to z admits
the representation
R(z, t)
z
R
1
(z, t) = F
1,
(t)(zI A
1
(t))
1
S
1
1,
(t)(zI A
(t))
1
G
,
(t).
(14.19)
The representations (14.18), (14.19) above are local: each of them holds in the
appropriate individual subset T
. Note, however, that whenever T
,= ,
by Theorem 14.5 we have
S
1,
(t) = D
1
,
(t)S
1,
(t)D
1,
,
(t), t T
,
where the functions D
,
,
(t), D
1,
,
(t) are as in (14.12). Hence the expres-
sions (14.18), (14.19) coincide in the intersections of the subsets T
(although the
individual functions F
1,
(t), S
1,
(t), G
,
(t) do not). Here it is a self-evident
fact, because these expressions represent globally dened objects. In the next sec-
tion, where we shall use such local representations to construct globally dened
objects, it will become a requirement.
15. Holomorphic families of generic rational matrix functions
with prescribed local data
This section can be considered as a t - dependent version of Section 13. Here we
consider the problem, how to construct a normalized holomorphic family of generic
rational functions
12
with prescribed local data. The nature of such data is sug-
gested by the considerations of the previous section (see, in particular, Theorem
14.5).
Let D be a domain in C
2n
and let T
A
be a nite open covering of
D. We assume that F
(t)
A
and G
(t)
A
are collections of, respectively,
C
mn
-valued and C
nm
-valued functions, satisfying the following conditions:
12
See Denition 14.1.
(t), G
(t) are holomorphic in T
.
2. Whenever T

r,
,
(t),
D
c,
,

t T
(t) = F
(t)D
c,
,
(t), G
(t) = D
r,
,
(t)G
(t). (15.1)
Remark 15.1. Conditions 1. and 2. imply that the collections F
(t)
A
and
G
(t)
A
represent holomorphic mappings from D into the spaces
13
P
(m1)n
c
and P
n(m1)
r
, respectively.
For each A and t T
we consider the Lyapunov equation

A
1
(t)S
(t) S
(t)A
(t) = G
(t)F
(t), (15.2)
where A
1
(t) and A
(t) are as in (14.1). Its solution S
(t) is the TZ-coupling

14
matrix, related to the data A
1
(t), A
(t), F
(t), G
(t). Considered as a function of

t, S
(t) is holomorphic in T
, because the right-hand side G
(t)F
(t) is holomor-
phic in T
. The collection of functions S
(t)
A
is said to be the collection of
TZ-coupling functions related to the pair of collections F
(t)
A
, G
(t)
A
.
In view of (15.1), whenever T
,= , we have
S
(t) = D
r,
,
(t)S
(t)D
c,
,
(t), t T
, (15.3)
where D
r,
,
(t), D
c,
,
(t) are diagonal, holomorphic and invertible matrix
functions. Hence either A det S
(t) 0 or A det S
(t) , 0. In the
latter case the pair of collections F
(t)
A
and G
(t)
A
is said to be TZ-
admissible, and the set
1
=
_
A
t T
: det S
(t) = 0 (15.4)
is said to be the TZ-singular set related to the pair of collections F
(t)
A
,
G
(t)
A
.
In the same way, we can consider the collection of functions S
(t)
A
,
where for each A and t T
the matrix S
(t) is the solution of the Lyapunov

equation
A
(t)S
(t) S
(t)A
1
(t) = G
(t)F
(t). (15.5)
This collection is said to be the collection of ZT-coupling functions related to the
pair of collections F
(t)
A
, G
(t)
A
. If for every A det S
(t) , 0 in T
then the pair of collections F
(t)
A
, G
(t)
A
is said to be ZT-admissible,
and the set
1
=
_
A
t T
: det S
(t) = 0 (15.6)
is said to be the ZT-singular set related to the pair of collections F
(t)
A
,
G
(t)
A
.
13
See Denition 13.5.
14
See Denition 13.3.
Remark 15.2. Note that, in view of (15.1), if for each A S
(t) satises the

Lyapunov equation (15.2) (or (15.5)) then the subsets
= t T
: det S
(t) = 0
of the appropriate singular set agree in the intersections of the sets T
(T
) =
(T
.
2n
and let T
A
be a nite open cover-
ing of D. Let F
(t)
A
and G
(t)
A
are collections of, respectively, C
mn
-
valued and C
nm
(t), G
.
2. Whenever T

r,
,
(t),
D
c,
,

t T
(t) = F
(t)D
c,
,
(t), G
(t) = D
r,
,
(t)G
(t).
3. The pair of collections F
(t)
A
, G
(t)
A
is ZT-admissible.
Let
1
denote the ZT-singular set related to the pair of collections
F
(t)
A
, G
(t)
A
. Then there exists a unique normalized holomorphic fam-
ily R(z, t)
tD\
ZP
of rational generic functions parameterized by the pole and
zero loci such that for every A and t T
1
the matrices F
(t) and G
(t)
are, respectively, the left pole and right zero semiresidual matrices of R(, t):
F
(t) = F
1,
(t), G
(t) = G
,
(t), t T

1
. (15.7)
It is locally given by
R(z, t) = I F
(t)(zI A
1
(t))
1
S
1
(t)G
(t), t T

1
, (15.8)
where S
(t)
A
is the collection of ZT-coupling functions related to the pair of
collections F
(t)
A
, G
(t)
A
. Furthermore, the logarithmic derivative of
R(z, t) with respect to z admits the local representation
R(z, t)
z
R
1
(z, t) =
= F
(t)(zI A
1
(t))
1
S
1
(t)(zI A
(t))
1
G
(t),
t T

1
. (15.9)
Proof. In view of Proposition 13.4 and condition 3, for each A and t T
1
there exists a unique generic rational function R
(, t), normalized by R
(, t) =
I, with the pole and zero matrices A
1
(t), A
(t) and the prescribed left zero and

right pole semiresidual matrices (15.7). The function R
(, t) and its logarithmic

derivative admit the representations
R
(z, t) = I F
(t)(zI A
1
(t))
1
S
1
(t)G
(t), (15.10)
R
(z, t)R
1
(z, t) = F
(t)(zI A
1
(t))
1
S
1
(t)(zI A
(t))
1
G
(t).
From the representation (15.10) and condition 1 it follows that the family
R
(z, t)
tD
\
ZP
is holomorphic. In view of condition 2 (see also (15.3)), we
have
R
(z, t) = R
(z, t), t (T
)
1
.
Hence we can dene the holomorphic family R(z, t)
tD\
ZP
by
R(z, t) = R
(z, t), t D

1
to obtain the local representations (15.8), (15.9). The uniqueness of such a family
follows from the uniqueness of each function R
(, t).
2n
and let T
A
be a nite open cover-
ing of D. Let F
(t)
A
and G
(t)
A
are collections of, respectively, C
mn
-
valued and C
nm
(t), G
.
2. Whenever T

r,
,
(t),
D
c,
,

t T
(t) = F
(t)D
c,
,
(t), G
(t) = D
r,
,
(t)G
(t).
3. The pair of collections F
(t)
A
, G
(t)
A
is TZ-admissible.
Let
1
denote the TZ-singular set related to the pair of collections
F
(t)
A
, G
(t)
A
. Then there exists a unique normalized holomorphic fam-
ily R(z, t)
tD\
PZ
of rational generic functions parameterized by the pole and
zero loci such that for every A and t T
1
the matrices F
(t) and G
(t)
are, respectively, the left zero and right pole semiresidual matrices of R(, t):
F
(t) = F
,
(t), G
(t) = G
1,
(t), t T

1
. (15.11)
It is locally given by
R(z, t) = I +F
(t)S
1
(t)(zI A
1
(t))
1
G
(t), t T

1
, (15.12)
where S
(t)
A
is the collection of TZ-coupling functions related to the pair of
collections F
(t)
A
, G
(t)
A
. Furthermore, the logarithmic derivative of
R(, t) admits the local representation
R(z, t)
z
R
1
(z, t)
= F
(t)S
1
(t)(zI A
1
(t))
1
S
(t)(zI A
(t))
1
S
1
(t)G
(t),
t T

1
. (15.13)
Proof. The proof is analogous to that of Theorem 15.3.
16. Isosemiresidual families of generic rational matrix functions
In the present section we shall consider an important special case of holomorphic
families of generic rational functions, parameterized by the pole and zero loci t.
Namely, we are interested in the case when (either left pole and right zero or right
pole and left zero) semiresidual functions of t, determining the family as explained
in Section 14, are constant.
2n
and let R(z, t)

tD
be a normalized
mm
-valued generic rational functions, parameterized by
the pole and zero loci.
1. The family R(z, t)
tD
is said to be ZT-isosemiresidual
15
if there exists a
pair of matrices F C
mn
and G C
nm
such that for every t D the
matrices F and G are, respectively, the left pole and right zero semiresidual
matrices of the generic rational function R(, t):
F = F
1
(t), G = G
(t), t D.
2. The family R(z, t)
tD
is said to be TZ-isosemiresidual if there exists a
pair of matrices F C
mn
and G C
nm
such that for every t D the
matrices F and G are, respectively, the left zero and right pole semiresidual
matrices of the generic rational function R(, t):
F = F
(t), G = G
1
(t), t D.
Let us assume that a pair of matrices F C
mn
and G C
nm
is given.
How to construct a (TZ- or ZT-) isosemiresidual normalized holomorphic family
of C
mm
-valued generic rational functions, parameterized by the pole and zero
loci, for which the constant functions
F(t) F, G(t) G (16.1)
would be the appropriate semiresidual functions? This is a special case of the
problem considered in Section 15 (see Theorem 15.3). Note, however, that in this
case the prescribed semiresidual functions (16.1) are holomorphic in the domain
C
2n
. Therefore, we may consider its open covering consisting of the single set
the domain itself.
Following the approach described in Section 15, we consider the solutions
S
1
(t), S
1
(t) of the Lyapunov equations
A
1
(t)S
1
(t) S
1
(t)A
(t) = GF, (16.2a)

A
(t)S
1
(t) S
1
(t)A
1
(t) = GF, (16.2b)
where
A
1
(t) = diag(t
1
, . . . , t
n
), A
(t) = diag(t
n+1
, . . . , t
2n
). (16.3)
15
Iso- (from
..
o - equal - in Old Greek) is a combining form.
Then the functions S
1
(t), S
1
(t) are given explicitly by
S
1
(t) =
_
g
i
f
j
t
i
t
n+j
_
1i,jn
, (16.4a)
S
1
(t) =
_
g
i
f
j
t
n+i
t
j
_
1i,jn
, (16.4b)
where g
i
and f
j
denote, respectively, the ith row of G and the jth column of F.
In particular, the functions S
1
(t), S
1
(t) are rational with respect to t and
holomorphic in C
2n
. The next step is to verify that the constant functions (16.1)

are TZ- or ZT-admissible (that is, suitable for the construction of a holomorphic
family of generic rational functions see Theorem 15.3). This means to check that
det S
1
(t) , 0 or det S
1
(t) , 0. Note that, since the functions S
1
(t), S
1
(t)
are identical up to the permutation of variables t
k
t
n+k
, 1 k n, these
conditions are equivalent.
Denition 16.2. A pair of matrices F C
mn
and G C
nm
is said to be
admissible if the C
nn
-valued rational functions S
1
(t), S
1
(t) given by (16.4)
satisfy the (equivalent) conditions
det S
1
(t) , 0, det S
1
(t) , 0.
It turns out that the admissibility of a given pair of matrices F C
mn
,
G C
nm
can be checked by means of a simple criterion, described below.
Recall that for a matrix M =
_
m
i,j
_
n
i,j=1
,
det M =
(1)
m
1,(1)
m
n,(n)
, (16.5)
where runs over all n! permutations of the set 1, . . . , n, and (1)
is equal to
either 1 or 1 depending on the parity of the permutation .
Denition 16.3. A matrix M C
nn
is said to be Frobenius-singular if for some ,
1 n, there exist indices 1
1
< <
n; 1
1
< <
n+1
n,
such that m
i
,
j
= 0 for all 1 i , 1 j n + 1.
Theorem 16.4. A matrix M C
nn
is Frobenius-singular if and only if all n!
summands (1)
m
1,(1)
m
n,(n)
of the sum (16.5) representing the determinant
det M are equal to zero.
Theorem 16.4 is due to G. Frobenius, [Fro1]. The proof of this theorem can
be also found in [Ber], Chapter 10, Theorem 9. The book [LoPl] contains some
historical remarks concerning this theorem. See the Preface of [LoPl], especially
pp. xiiixvii of the English edition (to which pp. 1418 of the Russian translation
correspond).
Proposition 16.5. A pair of matrices F C
mn
, G C
nm
is admissible if and
only if their product GF is not a Frobenius-singular matrix.
Proof. Assume rst that the matrix GF is Frobenius-singular. Then, according to
(16.4a),
det S
1
(t) =
(1)
m
1,(1)
m
n,(n)
(t
1
t
1+(1)
) ((t
n
t
n+(n)
)
, (16.6)
where runs over all n! permutations of the set 1, . . . , n and m
i,j
= g
i
f
n+j
. If
the matrix GF is Frobenius-singular then, according to Theorem 16.4, all the
numerators m
1,(1)
m
n,(n)
of the summands in (16.6) are equal to zero, and
hence det S
1
(t) 0.
Conversely, if the matrix GF is not Frobenius-singular then, according to the
same Theorem 16.4, there exists a permutation
0
such that
m
1,
0
(1)
m
n,
0
(n)
,= 0.
Let us choose and x n pairwise dierent numbers t
0
1
, . . . t
0
n
and set t
0
n+
0
(1)
=
t
0
1
, . . . , t
n+
0
(n)
= t
0
n
, where ,= 0. Then as 0 the summand
(1)
0
m
1,
0
(1)
m
n,
0
(n)
(t
1
t
n+
0
(1)
()) (t
n
t
n+
0
(n)
())
= (1)
0
(m
1,
0
(1)
m
n,
0
(n)
)
n
is the leading term of the sum on the right-hand side of (16.6): all other summands
grow at most as O(
(n1)
).
Denition 16.6. Let a pair of matrices F C
mn
, G C
nm
be such that their
product GF is not a Frobenius-singular matrix and let the C
nn
-valued rational
functions S
1
(t), S
1
(t) be given by (16.4).
1. The set
1
= t C
2n
: det S
1
(t) = 0 (16.7)
is said to be the TZ-singular set related to the pair F, G.
2. The set
1
= t C
2n
: det S
1
(t) = 0 (16.8)
is said to be the ZT-singular set related to the pair F, G.
Remark 16.7. Note that, since det S
1
(t), det S
1
(t) are polynomials in (t
i

t
j
)
1
, the singular sets
1
,
1
related to the pair F, G are complex algebraic
varieties of codimension one in C
2n
.
Combining Proposition 16.5 and Theorems 15.3, 15.4, we obtain
Theorem 16.8. Let matrices F C
mn
and G C
nm
be such that their product
GF is not a Frobenius-singular matrix, and let
1
,
1
be the related singular
sets. Then the following statements hold true.
1. There exists a unique TZ-isosemiresidual family R(z, t)
tC
2n
\
PZ
of
C
mm
-valued generic rational functions such that for every t C
2n

1
the
matrices F and G are, respectively, the left zero and right pole semiresidual
matrices of R(, t):
F = F
(t), G = G
1
(t), t C
2n

1
. (16.9)
It is given by
R(z, t) = I +FS
1
1
(t)(zI A
1
(t))
1
G, t C
2n

1
(16.10)
where the function S
1
(t) satisfying the equation (16.2a) is given by (16.4a).
Furthermore, the logarithmic derivative of R(z, t) with respect to z admits the
representation
R(z, t)
z
R
1
(z, t)
= FS
1
1
(t)(zI A
1
(t))
1
S
1
(t)(zI A
(t))
1
S
1
1
(t)G,
t C
2n

1
. (16.11)
2. There exists a unique ZT-isosemiresidual family R(z, t)
tC
2n
\
ZP
of
C
mm
-valued generic rational functions such that for every t C
2n

1
the
matrices F and G are, respectively, the left pole and right zero semiresidual
matrices of R(, t):
F = F
1
(t), G = G
(t), t C
2n

1
. (16.12)
It is given by
R(z, t) = I F(zI A
1
(t))
1
S
1
1
(t)G, t C
2n

1
, (16.13)
1
(t) satisfying the equation (16.2b) is given by (16.4b).
Furthermore, the logarithmic derivative of R(z, t) with respect to z admits the
representation
R(z, t)
z
R
1
(z, t)
= F(zI A
1
(t))
1
S
1
1
(t)(zI A
(t))
1
G,
t C
2n

1
. (16.14)
17. Isoprincipal families of generic rational matrix functions
Our interest in holomorphic families of generic rational functions is motivated by
our intent to construct rational solutions of the Schlesinger system (see Section 18
below). Indeed, given a holomorphic family R(z, t) : t D of generic rational
functions, we can consider the linear dierential system
R(z, t)
z
= Q
R
(z, t)R(z, t), (17.1)
where Q
R
(z, t) is the logarithmic derivative of R(z, t) with respect to z. Accord-
ing to Lemma 14.2 (and in view of (14.3), (14.4c)), the system (17.1) can be
rewritten as
R(z, t)
z
=
_
2n
k=1
Q
k
(t)
z t
k
_
R(z, t), (17.2)
where the functions Q
k
(t) are holomorphic in D. The system (17.2) can be viewed
as a holomorphic family (=deformation) of Fuchsian systems parameterized by the
singularities loci. It was proved in [KaVo] (Theorem 8.2) that in the case when
the deformation (17.2) is isoprincipal the functions Q
k
(t) satisfy the Schlesinger
system.
2n
and let R(z, t)

tD
be a normalized
mm
the pole and zero loci. Assume that for 1 k 2n there exist C
mm
-valued
functions E
k
(), holomorphic and invertible in C
, such that for every t D the

function E
k
is the principal factor
16
of the function R(, t) at t
k
: there exists a
C
mm
-valued function H
k
(, t), holomorphic and invertible in a neighborhood of
t
k
, such that
R(z, t) = H
k
(z, t)E
k
(z t
k
). (17.3)
Then the family R(z, t)
tD
is said to be isoprincipal.
2n
and let R(z, t)

tD
be a normalized
mm
the pole and zero loci. The family R(z, t) : t D is isoprincipal if and only if it
is TZ-isosemiresidual.
Proof. First, assume that the family R(z, t)
tD
is TZ-isosemiresidual. Then,
according to Denition 16.1, there exist F C
mn
and G C
nm
such that for
every t D the matrices F and G are, respectively, the left zero and right pole
semiresidual matrices of the generic rational function R(, t). Then, by Lemma
10.12 (see also Remark 10.14), for k = 1, . . . , 2n and independently of t the prin-
cipal factor E
k
() of R(, t) at t
k
can be chosen in the form
E
k
() =
_
I +L
k
1
L
k
, if 1 k n,
I L
k
+L
k
, if 1 +n k 2n,
(17.4)
where
L
k
=
_
g
k
(g
k
g
k
)
1
g
k
, if 1 k n,
f
k
(f
k
f
k
)
1
f
k
, if 1 +n k 2n,
(17.5)
g
k
is the kth row of G and f
k
is the (k n)th column of F. Hence, by Denition
17.1 the family R(z, t) : t D is isoprincipal.
Conversely, assume that the family R(z, t)
tD
is isoprincipal. Then, ac-
cording to Denition 17.1, for 1 k 2n there exist functions

E
k
(), holomorphic
and invertible in C
, such that for every t D the function

E
k
is the principal
factor of the function R(, t) at t
k
. Let t
0
, t D be xed and let F C
mn
and
G C
nm
be, respectively, the left zero and right pole semiresidual matrices of
the generic rational function R(, t
0
). Let us denote by g
k
the kth row of G and
16
See Denition 10.10.
by f
k
the (k n)th column of F. Then, in view of Remark 10.11, the function

E
k
is of the form
E
k
() = M
k
()E
k
(),
where M
k
() is a C
mm
-valued function, holomorphic and invertible in C and E
k
()
is given by (17.4), (17.5). Hence for z in a neighborhood of t
k
R(z, t) admits the
representation
R(z, t) = H
k
(z, t)M
k
(z t
k
)E
k
(z t
k
),
where H
k
(, t) is a C
mm
-valued function, holomorphic and invertible at t
k
. Then,
for k = 1, . . . , n, the residue R
k
(t) of R(z, t) at t
k
is given by
R
k
(t) =
_
H
k
(t
k
, t)M
k
(0)g
k
(g
k
g
k
)
1
_
g
k
.
Therefore, G is the right pole semiresidual matrix of the function R(z, t), as well.
Analogously, for k = n + 1, . . . , 2n
E
1
() = I L
k
+
1
L
k
,
where
L
k
= f
k
(f
k
f
k
)
1
f
k
,
hence the residue R
k
(t) of R
1
(z, t) at t
k
is given by
R
k
(t) = f
k
_
(f
k
f
k
)
1
f
k
M
1
k
(0)H
1
k
(t
k
, t)
_
.
Therefore, F is the left zero semiresidual matrix of the function R(z, t), as well.
This completes the proof.
Theorem 17.2 reduces the construction of an isoprincipal family to the con-
struction of an isosemiresidual family. The latter problem has already been con-
sidered in Section 16. According to Theorems 16.8 and 17.2, from any pair of
matrices F C
mn
and G C
nm
, such that the product GF is not a Frobenius-
singular matrix, we can construct an isoprincipal family of generic rational func-
tions R(z, t)
tC
2n
\
PZ
, where
1
denotes the TZ-singular set related to the
pair of matrices F, G. This family is given by
R(z, t) = I +FS
1
1
(t)(zI A
1
(t))
1
G, (17.6)
1
(t), satisfying (16.2a), is given by (16.4a). The logarithmic
derivative of R(z, t) with respect to z is given by
R(z, t)
z
R
1
(z, t)
= FS
1
1
(t)(zI A
1
(t))
1
S
1
(t)(zI A
(t))
1
S
1
1
(t)G, (17.7)
and we obtain the following expressions for its residues Q
k
(t)
Q
k
(t) = FS
1
1
(t)I
[k]
S
1
(t)(t
k
I A
(t))
1
S
1
1
(t)G,
1 k n, (17.8a)
Q
k
(t) = FS
1
1
(t)(t
k
I A
1
(t))
1
S
1
(t)I
[kn]
S
1
1
(t)G,
n + 1 k 2n. (17.8b)
Here we use the notation
I
[k]
def
= diag(
1,k
, . . . ,
n,k
),
where
i,j
is the Kronecker delta.
Remark 17.3. Note that, according to (16.4a), the function S
1
(t) is a rational
function of t. Hence also the functions Q
k
(t) are rational functions of t.
18. Rational solutions of the Schlesinger system
It can be checked that the rational functions Q
k
(t) given by (17.8a) satisfy the
Schlesinger system
_
_
Q
k
t
=
[Q
, Q
k
]
t
t
k
, k ,= ,
Q
k
t
k
=
,=k
[Q
, Q
k
]
t
k
t
.
(18.1)
It is also not very dicult to check that
V (t)
def
= FS
1
1
(t)G (18.2)
is the potential function for this solution:
Q
k
(t) =
V (t)
t
k
, k = 1, . . . , 2n. (18.3)
Furthermore, one can show that the rational function det S
1
(t) admits the
following integral representation
det S
1
(t) = det S
1
(t
0
) exp
__
1j2n,
j,=i
trace
_
V (t)
t
i
V (t)
t
j
_
t
i
t
j
dt
i
_
. (18.4)
where t
0
and t are two arbitrary points the domain C
2n

1
, and is an arbitrary
path which begins at t
0
, ends at t and is contained in C
2n

1
.
However, the explanation of these facts lies in the considerations of Sections
2 and 3 of the rst part [KaVo] of this work. The matrix functions Q
k
(t) satisfy
the Schlesinger system, and the function V (t) is a Laurent coecient at z =
of the normalized solution (17.6) of the Fuchsian system
d R(z, t)
d z
=
_
_

1k2n
Q
k
(t)
z t
k
_
_
R(z, t), (18.5)
R(z, t) = I
V (t)
z
+o([z[
1
) as z , (18.6)
while the function
(t)
def
= det S
1
(t), (18.7)
is the tau-function related to the solution Q
1
(t), . . . , Q
2n
(t) of the Schlesinger
system.
More detailed explanation of these and other related facts will be given in
the third part of this work.
Appendix
B. The global factorization of a holomorphic
matrix function of rank one
Let M(t) = |m
p,q
(t)|
1p,qm
be a C
mm
-valued function of the variable t D,
where D is a domain in C
N
. (We can even assume that D is a Riemann domain
17
of dimension N over C
N
.) In our considerations N = 2n and D C
2n
. Let the
matrix function M be holomorphic in D and let
rankM(t) = 1 t D. (B.1)
We will try to represent M in the form
M(t) = f(t)g(t), (B.2)
wheref(t) and g(t) are, respectively, a C
m1
-valued function and a C
1m
-valued
function, both of them holomorphic
18
in D.
Let us recall that, according to Lemma 14.4, there exist a nite open covering
|
p
m
p=1
of D, a collection f
p
(t)
m
p=1
of C
m1
-valued functions and a collection
g
p
(t)
m
p=1
of C
1m
-valued functions satisfying the following conditions.
1. For p = 1, . . . , m the functions f
p
(t) and g
p
(t) are holomorphic in |
p
.
2. For p = 1, . . . , m the function M(t) admits the factorization
M(t) = f
p
(t)g
p
(t), t U
p
. (B.3)
17
See Denition 5.4.4 in [Hor].
18
In general, such a global factorization is impossible even if the factors f(t) and g(t) are only
required to be continuous rather than holomorphic: one of the obstacles is of topological nature.
3. Whenever |
p
|
p
,= , there exists a (scalar) function
p
,p
(t), holomorphic
and invertible in |
p
|
p
, such that
f
p
(t) = f
p
(t)
p
,p
(t), g
p
(t) =
1
p
,p
(t)g
p
(t) t |
p
|
p
. (B.4)
In particular,
p,p
(t) = 1 t |
p
, (B.5a)
p,p
(t) =
p,p
(t)
p
,p
(t) t |
p
|
p
|
p
. (B.5b)
The equalities (B.3), p = 1, . . . , k, are nothing more that the factorizations
of the form (B.2), with holomorphic factors f
p
(t) and g
p
(t). However, the factor-
ization (B.3) is only local: for each p the equality (B.3) holds in the open subset
|
p
of the set D. For dierent p
and p
, the factors f
p
, g
p
and f
p
, g
p
may not
agree in the intersections |
p
|
p
. To glue the factorizations (B.3) for dierent
p together, we seek scalar functions
p
(t) which are holomorphic in |
p
, do not
vanish there and satisfy the condition
f
p
(t)
p
(t) = f
p
(t)
p
(t) t |
p
|
p
. (B.6)
Then, in view of (B.3),
1
p
(t)g
p
(t) =
1
p
(t)g
p
(t) t |
p
|
p
. (B.7)
Assuming that such functions
p
, 1 p k, are found, we set
f(t)
def
= f
p
(t)
p
(t) if t |
p
, (B.8a)
g(t)
def
=
1
p
(t)g
p
(t) if t |
p
. (B.8b)
The relations (B.6), (B.7) ensure that these denitions are not contradictory. Thus
the functions f(t) and g(t) are dened for every t D. Moreover, these functions
are holomorphic in D and provide the factorization (B.2).
From (B.4) it follows that the condition (B.6) is equivalent to the condition
p
(t) =
p
,p
(t)
p
(t) t |
p
|
p
, (B.9)
where
p
,p
(t) are the functions appearing in (B.4). Thus, to ensure that the
conditions (B.6), (B.7) are in force, we have to solve the so-called second Cousin
problem (see [Shab], [Oni], [Leit] and [H or]):
Problem B.1. Let D be a complex manifold and let |
of D be an open
covering of D. For each , such that |
,= let a C-valued function

,
,
holomorphic and non-vanishing in |
, be given.
Find a collection of C-valued functions
with the following properties:

1. For every the function
is holomorphic in |
and does not vanish there.

2. Whenever |
,= , the relation
=
,
(B.10)
holds in |
.
A necessary condition for the solvability of the second Cousin problem in D
with the given data |
,
,
,
is the so-called cocycle condition:
,
=
,
,
in every non-empty triple intersection |
,
= 1 in every |
.
(B.11)
In our case this condition is fullled see (B.5). However, the cocycle condition
alone is not sucient to guarantee the existence of a solution to the second Cousin
problem it depends on D itself, as well.
Proposition B.2. (J.-P.Serre, [Ser1]; see also [Shab], Section 16; [H or], sections
5.5 and 7.4; [Oni], Section 4.4.) If D is a Stein manifold
19
which satises the
condition
20
H
2
(D, Z) = 0, (B.12)
then the second Cousin problem in D with arbitrary given data U
,
,
satis-
fying the cocycle condition (B.11) is solvable.
As we have seen, the factorization problem (B.2) can be reduced to solving
the second Cousin with a certain data. Thus, the following result holds:
Theorem B.3. Let M(t) be a C
mm
-valued function, holomorphic for t D, where
D is a Riemann domain over C
N
. Assume that M satises the condition
rankM(t) = 1 t D.
If D possesses the property: the second Cousin problem in D with arbitrary given
data U
,
,
,
satisfying the cocycle condition (B.11) is solvable, then the
matrix function M(t) admits the factorization of the form
M(t) = f(t) g(t),
where the factors f(t) and g(t) are, respectively, a C
m1
-valued function and a
C
1m
-valued function, holomorphic and non-vanishing for t D. In particular,
such is the case if D is a Stein manifold satisfying the condition (B.12).
References
[BGRa] Ball, J., Gohberg, I. and M. Rakowski. Reconstruction of a rational non-
square matrix function from local data. Integral Equations Operator Theory
20:3 (1994), pp. 249305.
[BGR1] Ball, J.A., Gohberg, I. and L. Rodman. Realization and interpolation of ra-
tional matrix functions. In: Topics in Interpolation Theory of Rational Matrix-
valued Functions, (I. Gohberg, ed.), pp. 172. Operator Theory: Advances and
Applications, OT 33, Birkhauser Verlag, Basel Boston Berlin, 1988.
19
A domain D in C
N
is a Stein manifold if and only if D is pseudoconvex, or, what is equivalent,
D is a holomorphy domain.
20
H
2
(D, Z) is the second cohomology group of D with integer coecients.
[BGR2] Ball, J.A., Gohberg, I. and L. Rodman. Interpolation of Rational Matrix
Functions. Birkhauser Verlag, Basel Boston Berlin, 1990.
[Ber] Berge, C. Theorie des Graphes et ses Applications. Dunod, Paris 1958
(French). English transl.:
Berge, C. The Theory of Graphs and its Applications. Methuen &Co, London
1962. Russian transl.:
Ber, K. Teori Grafov i ee Primeneni. Izdat. Inostranno i Liter-
atury, Moskva, 1962, 319 ss.
[Birk1] Birkhoff, G.D. The generalized Riemann problem for linear dierential equa-
tions and the allied problems for linear dierence and q-dierence equations.
Proc. Amer. Acad. Arts and Sci., 49 (1913), 521568.
Reprinted in: [Birk2], 259306.
[Birk2] Birkhoff, G.D. Collected Mathematical Papers. Vol.I. American Mathemat-
ical Society, New York, 1950, ixi, 754 pp.
[DaKr] Dalecki i, .L. i M.G. Kre in. Usto iqivost Rexeni i Differenci-
alnyh Uravneni i v Banahovom Prostranstve. Nauka, Moskva, 1970, 534
ss. English transl.:
Dalecki, Ju.L. and M.G. Kren. Stability of Solutions of Dierential Equa-
tions in Banach Space. (Translations of Mathematical Monographs, 43.) Amer-
ican Mathematical Society, Providence, RI, 1974, ivi, 386 pp.
[Fro1] Frobenius, G.

Uber zerlegbare Determinanten.(German: On decomposable
determinants.) Sitzungsberichte der K oniglichen Preussischen Akademie der
Wissenschaften zu Berlin, XVIII (1917), pp. 274277. Reprinted in: [Fro2], Band
III, pp.701704.
[Fro2] Frobenius, G. Gesammelte Abhandlungen, Springer-Verlag, Berlin Heidel-
berg NewYork, 1968.
[GKLR] Gohberg, I. ,Kaashoek, M.A., Lehrer, L., and L. Rodman. Minimal di-
visors of rational matrix functions with prescribed zero and pole structures,
pp. 241275 in: Topics in Operator Theory, Systems and Networks. (Dym, H.
and I. Gohberg, ed.) Operator Theory: Advances and Applications, OT 12,
Birkhauser, Basel Boston Stuttgart, 1984.
[GS] Gohberg, I.C. i E.I. Segal. Operatornoe obobwenie teoremy o loga-
rifmiqeskom vyqete i teoremy Ruxe. Matem. sbornik, 84:4 (1971), ss.
607629. (Russian). English transl.:
Gohberg, I. and E.I. Segal. On operator generalizations of the logarithmic
residue theorem and the theorem of Rouche, Math. USSR Sbornik, 13 (1971),
pp. 603625.
[Hor] H ormander, L. An Introduction to Complex Analysis in Several Variables.
Van Nostrand, Princeton, N.J. Toronto New York London, 1966. ix, 213
pp. Russian transl.:
Hermander, L. Vvedenie v Teori Funkci i Neskolkih Kompleksnyh
Peremennyh. Mir, Moskva, 1968, 279 ss.
[Kats1] Katsnelson, V. Fuchsian dierential systems related to rational matrix func-
tions in general position and the joint system realization , pp. 117 - 143 in: Is-
rael Mathematical Conference Proceedings, Vol. 11 (1997), Proceedings of the
Ashkelon Workshop on Complex Function Theory (May 1996), Zalcman, L. -
editor.
[Kats2] Katsnelson, V. Right and left joint system representation of a rational ma-
trix function in general position (System representation theory for dummies) ,
pp. 337400 in: Operator theory, system theory and related topics. (The Moshe
Livsic anniversary volume. Proceedings of the Conference on Operator Theory
held at Ben-Gurion University of the Negev, Beer-Sheva and in Rehovot, June
29July 4, 1997), Alpay, D. and V. Vinnikov - editors. (Operator Theory: Ad-
vances and Applications, vol. 123), Birkhauser Verlag, Basel, 2001.
[KaVo] Katsnelson, V. and D. Volok. Rational Solutions of the Schlesinger System
and Isoprincipal Deformations of Rational Matrix Functions I , pp. 291348
in: Current Trends in Operator Theory and its Applications, Ball, J.A., Hel-
ton, J.W., Klaus, M. and L. Rodman- editors. (Operator Theory: Advances
and Applications, vol. 149), Birkhauser Verlag, Basel, 2004.
[KaVo1-e] Katsnelson, V. and D. Volok. Rational Solutions of the Schlesinger System
and Isoprincipal Deformations of Rational Matrix Functions I,
arXiv.org e-Print archive: http://arxiv.org, math. CA/0304108.
[Leit] La iterer, . Golomorfnye vektornye rassloeni i princip Oka-Grau-
erta. Ss. 75 - 121 v: [SCV-Ru]. English transl.:
Leiterer, J. Holomorphic vector bundles and the Oka-Grauert Principle.
pp. 63103 in: [SCV-En].
[LoPl] Lov asz, L. and M. Plummer. Matching Theory. (North Holland Mathemat-
ics Studies 121.) North-Holland, AmsterdamNewYorkOxfordTokyo, 1986, i
xxvii, 544 pp.
Russian transl.:
Lovas, L i M. Plammer. Prikladnye Zadaqi Teorii Grafov. (Teori
parosoqetani i v matematike, fizike, himii.) Mir, Moskva, 1998, 654 ss.
[Oni] Oniwik, A.L. Metody teorii puqkov i prostranstva Xte ina. Ss. 5
73 v: [SCV-Ru]. English transl.:
Onishchik, A.L. Methods in the Theory of Sheaves and Stein spaces. pp. 261
in: [SCV-En].
[PS] P` olya, G. and G. Szeg o. Problems and Theorems in Analysis, Volume II.
Springer Verlag. BerlinHeidelbergNewYork. 1972.
[Ras] Rassloennye Prostranstva i ih Priloeni. Sbornik perevodov.
Boltnski i, V.G., Dynkin, E.B. i M.M. Postnikov. red. [Fiber
spaces and their applications.] Collection of translations., Boltyanski, V.G.,
Dynkin, E.B., and M.M. Postnikov eds.. IL, Moskva, 1958.
[Sakhn] Sahnoviq, L.A. O faktorizacii peredatoqno i operator-funkcii. Dok-
lady AN SSSR, 226:4 (1976), ss. 781784. Engl. transl.: Sakhnovich, L.A.
On the factorization of an operator-valued transfer function. Soviet. Math.
Dokl. 17 (1976), pp. 203207.
[Sch1] Schlesinger, L.

Uber die Losungen gewisser linearer Dierentialgleichungen
als Funktionen der singularen Punkte. Journal f ur reine und angew. Math, 129
(1905), pp. 287294.
[Sch2] Schlesinger, L. Vorlesungen uber lineare Dierentialgleichungen. Leipzig und
Berlin, 1908.
[Sch3] Schlesinger, L.

Uber eine Klasse von Dierentialsystemen beliebiger Ordnung
mit festen kritischen Punkten. Journal f ur reine und angew. Math, 141 (1912),
pp. 96145.
[SCV-Ru] Kompleksny i Analiz Mnogie Peremennye 4, Gindikin, S.G. i G.M.
Henkin red. (Itogi Nauki i Tehniki. Sovremennye problemy matem-
atiki. Fundamentalnye napravleni. Tom 10.) VINITI, Moskva,
1986, ss. 1284. English transl.: [SCV-En]
[SCV-En] Several Complex Variables. IV. Gindikin, S.G. and G.M. Henkin eds.
(Encyclopaedia of Mathematical Sciences, Vol. 10.) Springer-Verlag, Berlin
Heidelberg NewYork, 1990, pp. 1252. Translated from Russian original:
[SCV-Ru]
[Ser1] Serre, J.-P. Quelques problèmes globaux relatifs aux varietes de Stein.
(French). Pp. 5768 in: Colloque sur les fonctions de plusieurs variables, tenu
à Bruxelles, mars 1953. Georges Thone, Liège, 1953; Masson&Cie, Paris, 1953.
Reprinted in: [Ser2], pp. 259270. Russian transl.:
Serr, . - P. Nekotorye globalnye zadaqi, svzannye s mnogoobrazimi
Xte ina. Ss. 363371 v [Ras], a take ss. 344354 v [Ser3].
[Ser2] Serre, J.-P. uvres / Collected papers . Jean-Pierre Serre. Vol. 1. Berlin,
Springer, 1986.
[Ser3] Serr, .P. Sobranie soqineni i. Tom 1. Moskva, Nezavisimy i Moskov-
ski i Universitet, Moskovski i Centr Nepreryvnogo Matematiqeskogo
Obrazovani, 2002. 464 ss.
[Shab] Xabat, B.V. Vvedenie v Kompleksny i Analiz. Qast II. Funkcii
Neskolkih Peremennyh. Trete izdanie. Nauka, Moskva, 1985. 464 ss.
English transl.:
Shabat, B.V. Introduction to complex analysis. Part II. Functions of several
variables. (Translations of Mathematical Monographs, 110.) American Mathe-
matical Society, Providence, RI, 1992. x+371 pp.
French transl.:
Chabat, B. Introduction ` a lanalyse complexe. Tome 2. Fonctions de plusieurs
variables. (Traduit du Russe: Mathèmatiques.) [Translations of Russian Works:
Mathematics], Mir, Moscow, 1990. 420 pp.
Victor Katsnelson
Weizmann Institute of Science
Rehovot 76100, Israel
e-mail: victor.katsnelson@weizmann.ac.il
Dan Volok
BenGurion University of the Negev
e-mail: volok@math.bgu.ac.il
Operator Theory:
Preservation of the Norms of Linear Operators
Acting on Some Quaternionic Function Spaces
M. Elena Luna-Elizarraras
and Michael Shapiro
Abstract. There are considered real and quaternionic versions of some classi-
cal linear spaces, such as Lebesgue spaces, the spaces of continuous functions,
etc., as well as linear operators acting on them. We prove that, given a real
linear bounded operator on one of those spaces its quaternionic extension
keeps being bounded and its norm does not change.
Mathematics Subject Classication (2000). Primary 47A30, 47B38; Secondary
46B25, 46E15.
Keywords. Quaternionic function spaces, quaternionic linear operators, norms
of quaternionic extensions of linear operators.
1. Introduction
1.1. Functional analysis and operator theory almost always use the elds R, of
real numbers, and C, of complex numbers, as the sets of scalars. At the same
time, the last decades have shown an explosive growth of research in numerous
areas of hypercomplex analysis, in particular, in quaternionic analysis which has
led to considering linear spaces and linear operators where the linearity involves
the scalars from the skew-eld H of (real) quaternions.
Certain basic as well as more advanced facts can be found, for instance, in the
books [Ad], [BDS], [DSS], [G uSp1], [G uSp2] and in the articles [AgKu], [ShVa1],
[ShVa2], [Sh], [AS], [ASV], [Shar], [SharAl1], [SharAl2] but in many other sources
as well.
Real and complex linearities has been considered mostly in parallel, i.e., the
theorems are proved for both cases simultaneously although there are some pe-
culiarities distinguishing between them. In particular, it proved to be interesting
The research was partially supported by CONACYT project and by COFAA-IPN.
The research was partially supported by CONACYT projects as well as by Instituto Politecnico
Nacional in the framework of COFAA and CGPI programs.
206 M.E. Luna-Elizarrar as and M. Shapiro
to look at what is occurring while imbedding real spaces and operators into the
wider complex spaces and acting there operators since the choice of the latter is
not unique, at least, starting from a normed real space one can construct many
complex normed spaces extending the initial real one. The study of the phenomena
arising in this setting, seems to have started in [Ri] and afterwards the problem
has been subject to periodic interest of the mathematicians who have found many
intrinsic features as well as intersections with other areas, see, e.g., [VeSe], [Ve],
[So], [Kr1], [Kr2], [Zy], [De], [FIP].
1.2. Quite similar questions arise, equally naturally, for real spaces and operators
but now compared with their quaternionic extensions, that is, the H-linear spaces
and acting between them H-linear operators generated by the original R-linear
objects. Some of them are treated in the paper, more specically, we consider the
quaternionic analogs of some classical function spaces (L
p
-functions; continuous;
of bounded variation; etc.) endowed with the structure of a left, or right, H-linear
space, and linear operators acting on or between them. As in complex case, those
quaternionic linear operators can have many norms extending the classical ones in
their real antecedents but we choose to introduce the norms by the usual formulas
with the quaternionic modulus instead of the real or complex one, thus ensuring
that the resulting spaces remain to be Banach spaces. We prove that, under such
conditions, in all the situations we are concerned in the norm of the quaternionic
extension of a real linear operator, does not increase.
1.3. We are not aware of any directly preceding work in quaternionic situation.
For L
p
-spaces, perhaps the nearest one is [GaMa] which is concerned with n-
tuples of R-linear operators, so for n = 4 the corresponding 4-tuple can be seen
as an R-linear operator being a kind of dequaternionization of the quaternionic
extension of an initial real linear operator. Besides, we use strongly and eciently
the multiplicative structure of quaternions and a very specic character of the
skew-eld H in all the reasonings. Lemma 3.1 is a good, but not unique, example
of the quaternionic avor, especially if one would like to compare it with its
predecessors.
1.4. To x the notation, we note that the quaternionic imaginary units in the
paper are e
1
, e
2
, e
3
: e
1
e
2
= e
2
e
1
= e
3
; e
2
e
3
= e
3
e
2
= e
1
; e
3
e
1
= e
1
e
3
=
e
2
; e
2
1
= e
2
2
= e
3
3
= 1. The real unit 1 is written, sometimes, as e
0
, so that given
H we write
=
3
=0
with
0
,
1
,
2
,
3
R.
The quaternionic conjugate to is
:=
0
=1
.
Preservation of the Norms of Linear Operators 207
2. Quaternionic extension of a real space
2.1. We begin this section with stating denitions and notations.
A left-H-linear space is an abelian group (M, +) together with a function : H
M M such that for all , H and for all a, b M the next conditions are
satised:
(i) (a +b) = a + b.
(ii) ( +) a = a + a.
(iii) () a = ( a).
(iv) 1 a = a.
As always we write a instead a. The above function HM M is usually
called the action of H over M, or the multiplication by quaternionic scalars on the
left. Analogously can be dened a right-H-linear space.
2.2. Given an R-linear space X, we want to get a left-H-linear space M such that
X is injected into M and the action of H over M restricted to R H, coincides
with the one dened on the R-linear space X. There are at least two equivalent
ways for doing this. The rst of them, commented here, makes use of the ideas
that help in the construction of the eld H from the eld R.
2.2.1. Take the set M := X X X X over which we dene the action of H as
follows: given =
3
i=0
i
e
i
H, and x = (x
0
, x
1
, x
2
, x
3
) M, then
x = (
0
+
1
e
1
+
2
e
2
+
3
e
3
)(x
0
, x
1
, x
2
, x
3
)
:= (
0
x
0
1
x
1
2
x
2
3
x
3
,
0
x
1
+
1
x
0
+
2
x
3
3
x
2
,
0
x
2

1
x
3
+
2
x
0
+
3
x
1
,
0
x
3
+
1
x
2

2
x
1
+
3
x
0
) .
It is easy to verify that the above denition satises the requirements to make M
a left H-space.
It is clear that the set M is an R-linear space, and the subset (r, 0, 0, 0) [ r X
is an R-subspace of M which is identied with X and so every element of the form
(r, 0, 0, 0) M is identied with the element r X.
Let us note that
e
1
(r, 0, 0, 0) = (0, r, 0, 0),
e
2
(r, 0, 0, 0) = (0, 0, r, 0),
e
3
(r, 0, 0, 0) = (0, 0, 0, r),
hence every element that belongs to M, can be expressed as
(x
0
, x
1
, x
2
, x
3
) = x
0
+e
1
(x
1
, 0, 0, 0) +e
2
(x
2
, 0, 0, 0) +e
3
(x
3
, 0, 0, 0)
= x
0
+e
1
x
1
+e
2
x
2
+e
3
x
3
,
and M has the decomposition:
M = X +e
1
X +e
2
X +e
3
X .
Let us denote M with the above multiplication by quaternionic scalars by X
H
and
it will be called the H-extension of X.
2.2.2. The language of tensor products is quite useful to give a second way of
dening the H-extension of X. Indeed, consider
X := H
R
X = (, z) [ H, z X ,
where H is seen just as R
4
. By denition,

X is an R-linear space and

X =
X X X X but it can be converted into a left-H-linear space introducing
a multiplication by quaternionic scalars by the formula: given an arbitrary H
and (, z)

X, one sets
(, z) := (, z) .
One can check up then, that we have arrived at X
H
.
2.2.3. Since the multiplication in H is non-commutative, the two above reasonings
have their right-hand side copy.
2.3. Consider now an R-linear space X, which has an additional structure being
also a normed space. Let as above X
H
be its H-extension. We wonder if there exists
a norm | |
H
on X
H
which extends the norm | | on X. Mostly the concept of a
norm is introduced for real or complex spaces, but the properties of the skew-eld
of quaternions allows us to generalize it directly.
Denition 2.1. If M is a left-H-linear space, a function
| |
H
: M R
is called a norm on M (or simply a norm) if it satises the usual conditions:
1. |x|
H
0 x M and |x|
H
= 0 if and only if x = 0.
2. |x|
H
=[ [ |x|
H
x M and H.
3. |x +y|
H
|x|
H
+|y|
H
x, y M.
Now we are ready to dene an H-extension of a norm.
Denition 2.2. Let (X, | |) be an R-linear normed space. We say that an H-norm
in X
H
is an extension of the norm on X if | |
H
restricted to X X
H
, coincides
with the norm | |.
2.4. Another notions that can be H-extended are the notions of an operator
and its norm. Let X and Y be R-linear spaces and let X
H
and Y
H
be their
corresponding H-extensions. Let T : X Y be an R-linear operator. The H-
extension of T, denoted by T
H
, is the operator T
H
: X
H
Y
H
dened by
T
H
[x
0
+e
1
x
1
+e
2
x
2
+e
3
x
3
] := T[x
0
] +e
1
T[x
1
] +e
2
T[x
2
] +e
3
T[x
3
] .
It is direct to check that T
H
is H-linear, that is
T
H
[x] = T[x] , H and x X .
2.4.1. Let X and Y be R-linear normed spaces . Their corresponding H-extensions
X
H
and Y
H
can be endowed with many norms extending the given norms. Let us
x some norms on X
H
and Y
H
. Then it is easy to show that if T : X Y
is additionally a bounded R-linear operator, then T
H
is bounded too. Moreover,
obviously |T| |T
H
|. There does not exist a general answer for the question
when the reciprocal inequality is valid. In this paper we study some cases in which
the norm of an operator and the norm of its H-extension are the same.
3. Norms of operators between the quaternionic L
p
-spaces
3.1. For K = R, C or H, and for a real number p such that 1 < p < +,
consider the K-linear normed space L
K
p
:= L
p
(, , ; K) where is a space with
a -algebra , a -nite measure , and where the norm is
|f|
p
:=
__
[ f(x) [
p
dx
_
1/p
, f L
K
p
.
Following a tradition, L
K
p
is considered as a left-K-linear space although for K = H
it can be seen as a bi-linear space.
It is easily seen that L
C
p
and L
H
p
are, respectively, the complex and the quaternionic
extensions of the R-linear space L
R
p
with the additional property that their norms
extend the norm on L
R
p
.
The next lemma provides an important tool in the proofs of the theorems.
Lemma 3.1. Given A =
3
i=0
a
i
e
i
H, it follows that
_
S
3
[A, B)[
p
dS
B
= [A[
p
C
p
, (3.1)
where S
3
is the unit sphere in R
4
, C
p
:=
_
S
3
[1, x)[
p
dS
x
and , ) is the inner
(scalar) product in R
4
: A, B) :=
3
k=0
A
k
B
k
=
1
2
(AB +BA) .
Proof. Dene

A :=
A
|A|
. It is known that all rotations of H = R
4
are of the form
(y) = y
1
, with and points of S
3
. So, taking =

A and = 1, we dene
a rotation : S
3
S
3
by (x) :=

Ax. Then we make the change of variable
B := (x) =

Ax and since the Lebesgue measure on the unit sphere S
3
is invariant
under orthogonal transformations, we have that
_
S
3
[A, B)[
p
dS
B
=
_
S
3
[ A [

A, B)
p
dS
B
=[ A [
p
_
S
3

A, B)
p
dS
B
=[ A [
p
_
S
3
[ (1), (x))[
p
dS
x
=[ A [
p
_
S
3
[ 1, x)[
p
dS
x
.
3.1.1. The above lemma has its origin in [Zy, p.181] and [VeSe] where the com-
plexication of real spaces and operators has been considered. Lemma 3.1 exploits
eciently the rich multiplicative structure of the skew-eld of quaternions; one
can compare it with Lemma 1 in [GaMa]; see also [St].
Theorem 3.2. Let T : L
R
p
:= L
p
(
1
,
1
,
1
; R) L
R
q
:= L
q
(
2
,
2
,
2
; R) be an
R-linear operator and let T
H
: L
H
p
L
H
q
be its H-extension. Then for 1 p
q < , it follows that
|T
H
|
L
H
p
L
H
q
= |T|
L
R
p
L
R
q
.
Proof. It is enough to show that |T
H
|
L
H
p
L
H
q
|T|
L
R
p
L
R
q
.
Let f
0
, f
1
, f
2
, f
3
L
p
, and for 0, 1, 2, 3 write T[f
] =: g
. Then F :=
3
=0
e
and G :=
3
=0
e
are elements of L
H
p
and L
H
q
respectively. Let B :=
3
=0
B
be
a point of S
3
, then h :=
3
=0
B
L
R
p
, and
T[h] =
3
=0
B
T[f
] =
3
=0
B
L
R
q
.
Using the denition of norm of an operator, it is obtained that
|T|
L
R
p
L
R
q
|T[h]|
L
R
q
|h|
L
R
p
=
_
_
_
_
_
3
=0
B
_
_
_
_
_
L
R
q
_
_
_
_
_
3
=0
B
_
_
_
_
_
L
R
p
,
that is,
_
_
_
_
_
3
=0
B
_
_
_
_
_
L
R
q
|T|
L
R
p
L
R
q
_
_
_
_
_
3
=0
B
_
_
_
_
_
L
R
p
.
From the denitions of the norms on the spaces L
R
p
and L
R
q
, the above inequality
can be written as:
_
_
=0
B
()
q
d
2
()
_
1/q
|T|
L
R
p
L
R
q
_
_
=0
B
()
p
d
1
()
_
1/p
.
Raising to the power q both sides of the above inequality and integrating with
respect to the variable B :=
3
=0
B
over the sphere S

3
, we get:
_
S
3
_
=0
B
()
q
d
2
() dS
B
|T|
q
L
R
p
L
R
q
_
S
3
_
_
=0
B
()
p
d
1
()
_
q/p
dS
B
.
Now Fubini-Tonelis Theorem applies to the left side, and raising to the power
1
q
both sides we get:
_
_
2
_
S
3
=0
B
()
q
dS
B
d
2
()
_
1/q
|T|
L
R
p
L
R
q
_
_
_
S
3
_
_
=0
B
()
p
d
1
()
_
q/p
dS
B
_
_
1/q
.
(3.2)
Observe that for
2
xed, g :=
3
=0
g
() e
is a quaternion. Then Lemma 3.1

applies to the left side of the above inequality, with A = g, which gives:
_
_
2
_
S
3
=0
B
()
q
dS
B
d
2
()
_
1/q
=
__
2
_
S
3
[B, g)[
q
dS
B
d
2
()
_
1/q
=
__
2
[ g [
q
C
q
d
2
()
_
1/q
= C
1/q
q
_
_
_
2
_
3
=0
[ g
() [
2
_
q/2
d
2
()
_
_
1/q
,
(3.3)
where C
q
:=
_
S
3
[ 1, x) [
q
dS
x
.
Now we apply Jessens inequality (see [Du-Sch, p. 530]) to the second member of
inequality (3.2) and use again Lemma 3.1, with A :=
3
=0
f
() e
:
|T|
L
R
p
L
R
q
_
_
_
S
3
_
_
=0
B
()
p
d
1
()
_
q/p
dS
B
_
_
1/q
|T|
L
R
p
L
R
q
_
_
_
_
1
_
_
S
3
=0
B
()
q
dS
B
_
p/q
d
1
()
_
_
_
1/p
= |T|
L
R
p
L
R
q
_
_
1
__
S
3
[ B, A) [
q
dS
B
_
p/q
d
1
()
_
1/p
= |T|
L
R
p
L
R
q
__
1
([ A [
q
C
q
)
p/q
d
1
()
_
1/p
= |T|
L
R
p
L
R
q
C
1/q
q
__
1
[ A [
p
d
1
()
_
1/p
= |T|
L
R
p
L
R
q
C
1/q
q
_
_
_
_
1
_
3
=0
[ f
() [
2
_
p/2
d
1
()
_
_
_
1/p
.
(3.4)
From (3.2), (3.3) and (3.4),
C
1/q
q
_
_
_
2
_
3
=0
[ g
() [
2
_
q/2
d
2
()
_
_
1/q
|T|
L
R
p
L
R
q
C
1/q
q
_
_
_
_
1
_
3
=0
[f
()[
2
_
p/2
d
1
()
_
_
_
1/p
,
which implies
|T
H
[F]|
L
H
q
|T|
L
R
p
L
R
q
|F|
L
H
p
.
Since F is arbitrary, this means that
|T
H
|
L
H
p
L
H
q
|T|
L
R
p
L
R
q
.
Remark 3.3. This theorem can be obtained from Lemma 1 in [GaMa]. We prefer
to have given a proof which uses strongly a specic character of our quaternionic
situation, in particular, we base the proof on Lemma 3.1.
4. Norms of operators on spaces of additive functions
4.1. Let us denote by ba(S, ; K) =: ba(K), the set of K-valued bounded additive
functions dened on the eld of subsets of a set S; that is, is a family of subsets
of S which is closed under the nite umber of operations of union, intersection,
and complement. Obviously ba(S, ; K) is a K-linear space (although for K = H
it can be seen as a bi-linear space) which is made a normed space with the norm
given by the formula
||
ba(K)
:= var() :=
sup
_
n
i=1
[ (E
i
) [ : E
i
n
i=1
, n N, is a nite collection of disjoint sets of
_
.
Here we make a mention about the notation in the case K = H. When multipli-
cation by quaternions is made on the right, we write ()(E) = (E) for any
H, ba(S, ; H) and E .
It is proved in [DuSch] that for the case K = R or C, it is a Banach space. For
the case K = H, this can be proved analogously.
It is clear that
ba(S, ; C) = (ba(S, ; R))
C
where X
C
is the complexication of a real space X, and
ba(S, ; H) = (ba(S, ; R))
H
.
Theorem 4.1. Given T L(ba(S, ; R)) then |T| = |T
H
|.
Proof. Let :=
3
=0
e
ba(S, ; H) and := T
H
[] =
3
=0
e
T[
] =
3
=0
e
,
with
:= T[
], and let y :=
3
=0
y
be a point of the unit sphere S

3
. It follows
that
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
|T|
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
.
Integrating over the sphere S
3
we get:
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
dS
y
|T|
ba
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
dS
y
. (4.1)
In general, given ba(S, ; K), it will be useful to consider var() as the limit
of a generalized sequence, and this concept allows one to generalize the notion of
the limit. For doing this, there is considered the following reasoning.
A partially ordered set (D, ) is said to be directed if every nite subset of D has
an upper bound. A map f : D X of a directed set D into a set X is called
a generalized sequence of elements in X, or simply a generalized sequence in X.
If f : D X is a generalized sequence in the topological space X, it is said to
converge to the point p X, if to every neighborhood N of p there corresponds a
d
0
D, such that d d
0
implies f(d) N. In this case, it is also said that the
limit of f exists and is equal to p, or symbolically, lim
D
f(d) = p.
With additional hypothesis, this notion of convergence gives rise to a related notion
of uniform convergence: as before, let D be a directed set, A an arbitrary set, and
X a metric space with metric . Suppose that f maps (d, a) DA into X. Then
the statement lim
D
f(d, a) = g(a) uniformly on A, or uniformly for a A, means
that for every > 0 there exists a d
0
D such that (f(d, a), g(a)) < for d > d
0
and for every a A.
Now we can see var() as the limit of a generalized sequence in the following
manner. Let F be the family of all nite collections E
i
of disjoint sets in ,
which are ordered by E
i
F
j
to mean that each E
i
is the union of some of
the sets F
j
. Then by construction, F is a directed set. Observe that if E
i
F
j
,
then
[(E
i
)[
[(F
j
)[ .
Dene the function f : F K by
f (E
i
n
i=1
) :=
n
i=1
[(E
i
)[ .
Then
var() = lim
F
f (E
i
n
i=1
) = lim
F
n
i=1
[(E
i
)[ . (4.2)
For any y =
3
=0
y
S
3
and
3
=0
e
ba(S, ; H), we have that

3
=0
y

ba(S, ; R) and
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
= var
_
3
=0
y
_
= lim
E
j
]F
n
j=1
=0
y
(E
j
)
. (4.3)
Since the norm on a Banach space is continuous, the function var
_
3
=0
y
_
is
continuous over the sphere S
3
. From denition of the order relation on F, it follows
that the real sequence
_
_
_
n
j=1
=0
y
(E
j
)
_
_
_
E
j
]
(4.4)
is not decreasing.
Recall that Dinis Lemma assures that if a monotone sequence of continuous func-
tions dened on a compact converges point-wise to a continuous function, then it
also converges uniformly. This lemma extends easily onto the case of generalized
sequences. Since S
3
is compact, and (4.2) says that the generalized sequence (4.4)
converges point-wise to the continuous function var
_
3
=0
y
_
over S
3
, applying
Dinis Lemma we conclude that it also converges uniformly. Then the integral and
the limit can be interchanged in the next reasoning:
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
dS
y
=
_
S
3
lim
E
j
]
n
j=1
=0
y
(E
j
)
dS
y
= lim
E
j
]
n
j=1
_
S
3
=0
y
(E
j
)
dS
y
= lim
E
j
]
n
j=1
_
S
3
=0
(E
j
)e
,
3
=0
y
dS
y
.
Applying Lemma 3.1 for p = 1 we have:
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
dS
y
= lim
E
j
]
n
j=1
=0
(E
j
)e
C
1
= C
1
_
_
_
_
_
3
=0
e
_
_
_
_
_
ba(H)
.
Applying this result to the left and right sides of inequality (4.1), we obtain that
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
dS
y
= C
1
_
_
_
_
_
3
=0
e
_
_
_
_
_
ba(H)
|T|
_
S
3
_
_
_
_
_
3
=0
y
_
_
_
_
_
ba(R)
dS
y
= |T|C
1
_
_
_
_
_
3
=0
e
_
_
_
_
_
ba(H)
which implies:
_
_
_
_
_
T
_
3
=0
e
__
_
_
_
_
ba(H)
=
_
_
_
_
_
3
=0
e
_
_
_
_
_
|T|
ba
_
_
_
_
_
3
=0
e
_
_
_
_
_
and since =
3
=0
e
was an arbitrary element of ba(S, ; H), we conclude that

|T
H
| |T|.
4.2. The norms of operators on the space of functions of bounded variation
Let I be an interval of R, which can be nite or innite. Recall that the total
variation of a K-valued function f on I is dened by
var(f, I) := sup
n
i=1
[ f(b
i
) f(a
i
) [
where the supremum is taken over all nite sets of points a
i
, b
i
I with a
1
b
1

a
2
b
2
a
n
b
n
. If var(f, I) < , f is said to be of bounded variation
on I.
We denote by bv(I; K) the set of K-valued functions of bounded variation dened
on I. Clearly (bv(I; R))
C
= bv(I; C) and (bv(I; R))
H
= bv(I; H).
The K-linear space bv(I; K) becomes a normed space with the norm
|f|
bv(K)
:= var(f, I) .
Theorem 4.2. Let T : bv(I; R) bv(I; R) be an R-linear operator. Then |T| =
|T
H
|.
Proof. We dene a directed set F as the family of all nite collections I
j
n
j=1
of
disjoint subintervals of I with I
j
L
i
to mean that each I
j
is the union of
some of the sets L
i
. Writing I
j
= (a
j
, b
j
), the total variation of f on I is the limit
of a generalized sequence:
var(f, I) = lim
I
j
]F
n
j=1
[ f(b
j
) f(a
j
) [ .
The rest of the proof follows the proof of Theorem 4.1.
5. The case of spaces of continuous functions
As in the above sections, K will denote the elds R, C or the skew-eld H. Given
S a compact space, the K-linear space C(S; K) consists of K-valued continuous
functions. In the case K = H, C(S; H) is a bi-linear H-space. The norm in all these
spaces is given by the formula
|f| := sup [ f(s) [ : s S .
It is well known that for the case K = C or R, C(S; K) is a Banach space and this
result extends of course to the space C(S; H).
Denote by L(C(S; R)) the R-linear space whose elements are R-linear operators
on C(S; R).
Theorem 5.1. Let S be a Hausdor compact, let T be an element of L(C(S; R))
and take its H-extension T
H
. Then |T|
L(C(S;R))
= |T
H
|
L(C(S;H))
.
Proof. Consider the eld of subsets generated by the closed sets of S. Recall
that for the case K = R or C (see [DuSch]) a K-valued additive set function
dened on is said to be regular if for each E and > 0 there is a set F
such that F E and a set G such that E G
0
and [ (C) [< for every
C with C GF. Denote by rba(S; K) = rba(K) the K-linear space of regular
bounded additive set functions. Of course all this directly extends onto the case
K = H, where rba(S; H) forms a bi-linear H-space. The norm is introduced by
||
rba(K)
:= var() .
Again for the case K = C or R, in [DuSch, IV.6] it is shown that given f C(S; K),
it is integrable with respect to every in rba(S; K). This fact allows one to dene
an isometric isomorphism between (C(S; K))
and rba(S; K), such that given

(C(S; K))
, its corresponding rba(S; K) is such that

(f) =
_
S
f(s) (ds) , f C(S; K) .
Under this identication, it follows that (C(S; R))
can be seen as the space of

measures rba(S; R).
Following closely the proof of the above result, we extend it for K = H, but we
have to note the following: given rba(S; H), if we consider the left H-linear
space of -simple functions:
f =
n
j=1
j

B
j
,
it gives rise to the notion of left integral
_
S
f(s) (ds) .
That is, we accept to multiply the above integral by quaternions on the left
side only. Then, under the identication which is made between rba(S; H) and
(C(S; H))
, the latter is a left H-linear space.

For a given T : C(S; R) C(S; R), we consider its adjoint operator
T
: (C(S; R))
(C(S; R))
given by T
(f) = f T for every f (C(S; R))
.
Then, considering T
as an element of L(rba(S; R)) L(ba(S; R)), it satises the

hypothesis of Theorem 4.1, and we have that |T
| = |(T
)
H
|.
On the other hand, it is well known that |T| = |T
| for a K-linear operator in

case of K = R or C, and the same proof works for K = H. Then applying this fact
to the operator T
H
, it follows that |T
H
| = |(T
H
)
|. It suces now to prove that

(T
H
)
= (T
)
H
.
Let us show that this is true even for a more general situation. Take V
1
, V
2
two
K-linear spaces (with K = R or C) and let , )
V
: V
1
V
2
K be a bilinear
form. The triplet (V
1
, V
2
, , )
V
) is called a dual system. Given two dual systems
(V
1
, V
2
, , )
V
) and (W
1
, W
2
, , )
W
), let A : V
1
W
1
be a linear operator. If
there exists an operator A
: W
2
V
2
such that
A[v]; w)
W
= v; A
[w])
V
, (5.1)
for any v V
1
and w W
1
, then the operator A
is called dual, or adjoint, to the

operator A with respect to the given dual systems. It is readily seen that if A
exists, it is unique.
All this remains true for K = H with the following changes. The spaces V
1
and V
2
must be respectively a left- and a right-H-linear spaces. The form , )
V
should
be bi-linear in the following sense: it is left-H-linear in its rst argument and
right-H-linear in its second one.
Let us suppose that the given dual systems and the operator A are real, i.e., all the
objects involved are R-linear. Suppose additionally that there exists the adjoint
operator A
. We are going to prove that there exists the adjoint operator (A

H
)
of the H-extension of A and that the next formula holds:

(A
H
)
= (A
)
H
.
Let h :=
3
=0
e
V
H
1
and f :=
3
k=0
f
k
e
k
W
H
2
be arbitrary, then
A
H
[h] ; f)
W
H = A
H
_
3
=0
e
_
;
3
k=0
f
k
e
k
)
W
H =
3
=0
e
A[h
] ;
3
k=0
f
k
e
k
)
W
H
=
3
=0
e
A[h
] ;
3
k=0
f
k
e
k
)
W
H =
3
k=0
_
3
=0
e
A[h
] ; f
k
)
W
_
e
k
=
3
k=0
_
3
=0
e
; A
[f
k
])
V
_
e
k
.
On the other hand, we have that
h; (A
)
H
[f])
V
H =
3
=0
e
; (A
)
H
_
3
k=0
f
k
e
k
_
)
V
H =
3
=0
e
;
3
k=0
A
[f
k
] e
k
)
V
H
=
3
k=0
_
3
=0
e
; A
[f
k
] )
V
_
e
k
.
Then one can conclude that
A
H
[h]; f)
W
H = h; (A
)
H
[f])
V
H .
Comparing now with (5.1) for K = H and together with the uniqueness of the
adjoint operator, we obtain the desired result: (A
)
H
= (A
H
)
. In particular this
is true for the operator T above, which completes the proof of the theorem.
The next result is a quaternionic version of the theorem presented in [DuSch,
V.8.11, p. 445]:
Theorem 5.2. Let (S, , ) be a positive measure space. Then there exists a compact
Hausdor space S
1
and an isometric isomorphism : L
(S, , ; H) C(S
1
; H)
such that it maps real functions (i.e., functions real -almost everywhere) into real
functions, positive functions into positive functions, and quaternionic conjugate
functions into quaternionic conjugate functions, i.e., f = f for every f
L
(S, , ; H). Moreover, is an algebraic isomorphism in the sense that if h(s) =

f(s) g(s) -almost everywhere, then h = f g.
The next statement is an immediate consequence of the last two theorems.
Corollary 5.3. Let T : L
(
1
; R) := L
(
1
,
1
,
1
; R) L
(
2
; R) :=
L
(
2
,
2
,
2
; R) be an R-linear operator and let T
H
: L
(
1
; H) L
(
2
; H)
be its H-extension. Then
|T
H
|
L
(
1
;H)L
(
2
;H)
= |T|
L
(
1
;R)L
(
2
;R)
.
In other words, in Theorem 3.2 it is possible to take p = q = .
References
[Ad] Adler. Quaternionic Quantum Mechanics and Quantum Fields. Oxford Uni-
versity Press, 586 pp., 1995.
[AgKu] S. Agrawal, S.H. Kulkarni, Dual spaces of quaternion normed linear spaces
and reexivity. J. Anal. 8 (2000), 7990.
[AS] D. Alpay, M.V. Shapiro, Reproducing Kernel Quaternionic Pontryagin Spaces.
Integral equations and operator theory 50 (4) (2004), 431476.
[ASV] D. Alpay, M. Shapiro, D. Volok, Espaces de Branges Rovnyak et fonctions de
Schur: le cas hyper-analytique. Comtes Rendus de lAcademie des Sciences
Mathematique, Ser. I, 2004, v. 338, 437442.
[BDS] F. Brackx, R. Delanghe, F. Sommen, Cliord analysis. London: Pitman Res.
Notes in Math., v. 76, 308 pp.
[De] A. Defant, Best constants for the norm of the Complexication of Opera-
tors Between L
p
-spaces. Lecture Notes in Pure and Applied Mathematics 150
(1994), 173180.
[DSS] R. Delanghe, F. Sommen, V. Soucek, Cliord algebra and spinor-valued func-
tions. Amsterdam: Kluwer Acad. Publ, 1992, 485 pp.
[DuSch] N. Dunford, J.T. Schwartz, Linear Operators, Part I. Interscience Publishers,
Inc., New York, 1957.
[FIP] T. Figiel, T. Iwaniec, A. Pelczy nski, Computing norms and critical exponents
of some operators in L
p
-spaces. Stud. Math. 79 (1984), No. 3, 227274.
[GaMa] J. Gash, L. Maligranda, On Vector-valued Inequalities of the Marcinkiewicz-
Zygmund, Herz and Krivine Type. Math. Nachr 167 (1994), 95129.
[G uSp1] K. G urlebeck, W. Sproig, Quaternionic analysis and elliptic boundary value
problems. Berlin: Akademie-Verlag, 1989.
[G uSp2] K. G urlebeck, W. Sproig, Quaternionic and Cliord Calculus for Physicists
and Engineers. Wiley & Sons, 1997.
[Kr1] J.I. Krivine, Sur la Complexication des Operateurs de L
dans L
1
. C.R.
Acad. Sci. Paris 284 (1977), 377379.
[Kr2] J.I. Krivine, Constantes de Grothendieck et Fonctions de Type Positif sur les
Sphères. Adv. Math. 31 (1979), 1630.
[Ri] M. Riesz, Sur les Maxima des Formes Bilineaires et sur les Fonctionelles
Lineaires. Acta Math. 49 (1926), 465497.
[Sh] M.V. Shapiro, Structure of the quaternionic modules and some properites of
the involutive operators. J. of Natural Geometry (London) 1, 1 (1992), 937.
[ShVa1] M.V. Shapiro, N.L. Vasilevski, Quaternionic -hyperholomorphic functions,
singular integral operators and boundary value problems. I. -hyperholomorphic
function theory. Complex Variables. Theory and Applications 27 (1995), 17
46.
[ShVa2] M.V. Shapiro, N.L. Vasilevski, Quaternionic -hyperholomorphic functions,
singular integral operators and boundary value problems. II. Algebras of singu-
lar integral operators and Riemann-type boundary value problems. Complex
Variables. Theory and Applications 27 (1995), 6796.
[Shar] C.S. Sharma, Complex structure on a real Hilbert space and symplectic struc-
ture on a complex Hilbert space. J. Math. Phys 29, 5 (1988), 10691078.
[SharAl1] C.S. Sharma, D.F. Almeida, Semilinear operators. J. Math. Phys. 29, 11
(1988), 24112420.
[SharAl2] C.S. Sharma, D.F. Almeida, Additive functionals and operators on a quater-
nionic Hilbert space. J. Math. Phys. 30, 2 (1989), 369375.
[So] J. Sokolowski, On the Norm of the Complex Extension of the Linear Operator.
Mat. Issled. 54 (1980), 152154 (in Russian).
[St] S.B. Stechkin, On the Best Lacunary System of Functions. Izv. Acad. Nauk
SSSR, Ser. Mat. 25 (1961), 357366 (in Russian).
[Ve] I.E. Verbitski, Some Relations Between the Norm of an Operator and that of
its Complex Extension. Mat. Issled. 42 (1976), 312 (in Russian).
[VeSe] I.E. Verbitski, P.P. Sereda, About the norm of the complex extension of an
operator. Mat. Issled. 37 (1975), 201206 (in Russian).
[Zy] A. Zygmund, Trigonometric Series, Volume I. Cambridge University Press,
1968.
M. Elena Luna-Elizarraras and Michael Shapiro
Departamento de Matematicas
E.S.F.M. del I.P.N.
07338 Mexico, D.F., Mexico
e-mail: eluna@esfm.ipn.mx
e-mail: shapiro@esfm.ipn.mx
Operator Theory:
Hardy Algebras Associated with
W
-Correspondences
(Point Evaluation and Schur Class Functions)
Paul S. Muhly and Baruch Solel
1. Introduction
This is primarily an exposition of our work in [35] and [37] which builds on the
theory of tensor algebras over C
-correspondences that we developed in [31]. Op-

erator tensor algebras (and their w
-analogues, which we call Hardy algebras) form

a rich class of non-selfadjoint operator algebras that contains a large variety of op-
erator algebras that have received substantial attention in the literature in recent
years.
Among these algebras are the classical disc algebra A(D), and its weak clo-
sure, H
(T); Popescus non-commutative disc algebras [45], and their weak clo-
sures, the free semigroup algebras studied by Popescu [45] and Davidson and Pitts
[19]; quiver algebras studied by us in [32] and by Kribs and Power in [25]; certain
nest algebras; analytic crossed products, studied by Peters [41] and by McAsey
and Muhly in [28]; and others. (We will describe the construction of tensor and
Hardy algebras and give many examples in Section 2.) The theory gives a common
approach to the analysis of all these algebras and has its roots deeply embedded
in the model theory of contraction operators on the one hand and in classical ring
theory on the other.
In fact, the theory of contraction operators may be viewed as the theory of
contractive representations of the disc algebra. The representation theory of the
tensor algebras is a natural generalization of this theory that preserves many of
its features. The disc algebra may be viewed as an analytic generalization of the
polynomial algebra in one variable. The interplay between function theory and
the representation theory of the polynomial algebra has been one of the guiding
The rst author was supported by grants from the U.S. National Science Foundation and from the
U.S.-Israel Binational Science Foundation. The second author was supported by the U.S.-Israel
Binational Science Foundation and by the Fund for the Promotion of Research at the Technion.
222 Paul S. Muhly and Baruch Solel
beacons in model theory for decades [17]. The tensor algebras we analyze are
operator algebraic versions of algebras that generalize polynomial algebras and
have been of key importance in ring theory since 1947 [23] and, since 1972, have
been a major focus of attention for analyzing nite-dimensional algebras (see [21]
and [22]). (See [30] for an extended discussion of the connection between operator
tensor algebras and the theory of nite-dimensional algebras.)
Recall that the disc algebra A(D) may be realized as the algebra of all an-
alytic Toeplitz operators on l
2
(N) (or on H
2
(T)). Popescu generalizes A(D) by
considering algebras of operators on the full Fock space over a Hilbert space H of
some dimension, n say. Let T(H) = C H H
2
denote this Fock space.
Then his non-commutative disc algebra of index n is the norm closed algebra gen-
erated by the (left) creation operators. That is, his algebras are generated by the
identity operator and operators of the form () := , where H and
T(H). The Fock space may also be written as l
2
(F
n
+
) where F
n
+
is the free
semigroup on n generators. In this realization, H may be identied with all the
functions supported on the words of length one and for such a function , () is
just convolution by on l
2
(F
n
+
). Observe that when n, the dimension of H, is one,
then one recovers the disc algebra A(D) represented by analytic Toeplitz operators
on l
2
(N).
To construct more general tensor algebras one replaces the Hilbert space
H by a correspondence E over some C
-algebra (or von Neumann algebra) M.

Roughly, a correspondence is a bimodule over M that is also equipped with an
M-valued inner product. (For a precise denition see Section 2). When M = C a
correspondence over M is just a Hilbert space.
The tensor algebra associated with the correspondence E, T
+
(E), is generated
by creation operators on the Fock space T(E) = M E E
2
together with
a copy of M (formed by diagonal operators of multiplication,
(a), a M).
It follows from the results of [31] that (completely contractive) representations of
T
+
(E) are given by pairs (T, ) where T : E B(H) is a completely contractive
map and : M B(H) is a C
-representation of M that satisfy T(a b) =

(a)T()(b) for a, b M and E. (Note that we shall sometimes use for the
left multiplication on E; that is, a may be written (a).) Such pairs, (T, ), are
called covariant representations of the correspondence E. Given (T, ), one may
form the Hilbert space E
H (see the discussion following Denition 2.1). For

a M, (a) I then denes a bounded operator on this space. The complete
contractivity of T is equivalent to the assertion that the linear map

T dened
initially on the balanced algebraic tensor product EH by the formula

T(h) :=
T()h extends to an operator of norm at most 1 on the completion E
H. The
bimodule property of T, then, is equivalent to the equation
T((a) I) = (a)
T, (1)
for all a M, which means that

T intertwines and the natural representation
of M on E
H the composition of with Rieels induced representation of

L(E) determined by .
-Correspondences 223
Thus we see that, once is xed, the representations of T
+
(E) are pa-
rameterized by the elements in the closed unit ball of the intertwining space
B(E
H, H) [ (() I) = and || 1. Reecting on this leads

one ineluctably to the functional analysts imperative: To understand an algebra,
view it as an algebra of functions on its space of representations. In our setting,
then, we want to think about T
+
(E) as a space of functions on this ball. For rea-
sons that will be revealed in a minute, we prefer to focus on the adjoints of the
elements in this space. Thus we let E
= B(H, E
H) [ = (() I)
and we write D((E
) for the set B(E
H, H) [
, and || < 1.
That is, D((E
) is the norm-interior of the representation space consisting of

those (T, ) that are anchored by . One of our interests, then, is to understand
the kind of functions that elements X of T
+
(E) determine on D((E
) via the
formula
X(
) =
(X),
where
is the representation of T
+
(E) that is determined by the pair (, T)
with

T =
.
In the special case when A = E = C and is the one-dimensional repre-
sentation of A on C, E
is also one-dimensional, so D((E
) is just the open

unit disc in the complex plane and, for X T
+
(E), X(
) is the ordinary value

of X at the complex number . On the other hand, if A = E = C, and if is
scalar multiplication on a Hilbert space H (the only possible representation of C
on H), then D((E
) is the space of strict contraction operators on H and for
D((E
)
||
and X T
+
(E) = A(D), X(
) is simply the value of X at
dened through the Sz.-Nagy-Foias functional calculus [39]. For another example,
if A = C, but E = C
n
, and if is scalar multiplication on a Hilbert space H, then
D((E
) is the space of row contractions on H, (T

1
, T
2
, . . . , T
n
), of norm less than
1; i.e.
i
T
i
rI
H
for some r < 1. In this case, X(
) is given by Popescus
functional calculus [46].
In addition to parametrizing certain representations of T
+
(E), E
has another
fundamental property: It is itself a C
-correspondence - over the von Neumann

algebra (A)
. Indeed, it is not dicult to see that E
becomes a bimodule over

(A)
via the formulae: a = (I

E
a) and a = a, E
, a (A)
.
Further, if and are in E
, then the product
lies in the commutant (A)
and denes a (A)
-valued inner product , ) making E
a C
-correspondence.
In fact, since E
is a weakly closed space of operators, it has certain topological

properties making it what we call a W
-correspondence [35]. It is because E
is a
W
-correspondence over (A)
that we focus on it, when studying representations

of T
+
(E), rather than on its space of adjoints. While E
plays a fundamental
role in our study of quantum Markov processes [33], its importance here besides
providing a space on which to evaluate elements of T
+
(E) lies in the fact that a
certain natural representation of E
generates the commutant of the representation

of T
+
(E) obtained by inducing up to L(T(E)). (See Theorem 2.29.)
It is primarily because of this commutant theorem that we cast our work in
this paper entirely in terms of W
-correspondences. That is, we work with von

Neumann algebras M and W
-correspondences E over them. We still form the

Fock space T(E) and the tensor algebra T
+
(E) over E, but because T(E) is a
W
-correspondence over M, the space L(T(E)) is a von Neumann algebra. We

call the w
-closure of T
+
(E) in L(T(E)) the Hardy algebra of E and denote it
by H
(E). This is our principal object of study. In the case when M = E = C,

H
(E) it the classical H
(T) (viewed as analytic Toeplitz operators).

As we will see in Lemma 2.17, given a faithful normal representation of
M on a Hilbert space H, we may also evaluate elements in H
(E) at points in
D((E
) (since the representation associated with a point in the open unit ball
extends from T
+
(E) to H
(E)). That is, elements in H
(E) may be viewed as

functions on D((E
), also. Further, when H
(E) is so represented, one may

study the value distribution theory of these functions. In this context, we es-
tablish two capstone results from function theory: The rst, [35, Theorem 5.3]
is presented as Theorem 3.2 below, generalizes the Nevanlinna-Pick interpolation
theorem. It asserts that given two k-tuples of operators in B(H) (where H is the
representation space of ), B
1
, B
2
, . . . , B
k
, and C
1
, C
2
, . . . , C
k
, and given points
1
,
2
, . . . ,
k
in D((E
), one may nd an element X in H
(E) of norm at most

one such that
B
i
X(
i
) = C
i
,
for all i if and only if a certain matrix of maps, which resembles the classical Pick
matrix, represents a completely positive operator. This result captures numerous
theorems in the literature that go under the name of generalized Nevanlinna-Pick
theorems. Our proof of the theorem (in [35]) uses a commutant lifting theorem that
we proved in [31]. In the context of model theory, it was Sarason who introduced the
use of commutant lifting to prove the interpolation theorem ([49]). More recently,
a number of authors have been studying interpolation problems in the context of
reproducing kernel Hilbert spaces. (See [1], [20], [2], [3] and [14].)
Our second capstone result is a generalization of Schwartzs lemma (see The-
orem 3.4). It follows from our Nevanlinna-Pick theorem that an element X in
H
(E) of norm at most one denes a Pick-type matrix of maps that represents
a completely positive map. In fact, the matrix is dened using the values of X on
D((E
). Given an arbitrary operator-valued function Z on D((E
), one may
dene a matrix of maps in a similar way. We say that Z is a Schur class operator
function if this matrix denes a completely positive map. (See Denition 4.2 for
a precise statement.) Theorem 3.2 then shows that the function
X(
) is a
Schur class operator function for X in the closed unit ball of H
(E). In fact, we
show in Theorem 4.3 that every Schur class operator function arises in this way
and that every such function (with values in, say, B(c)) may be represented in the
form Z(
) = A+B(I L
D)
1
L
C where A, B, C and D are the entries of a 22

block matrix representing a coisometric operator V from c H to c (E
H)
(for some auxiliary Hilbert space H) with a certain intertwining property and L
is the operator from H to E
H that maps h to h. Borrowing terminology

from the classical function theory on the unit disc D, we call such a representation
a realization of Z and we call the coisometry V a colligation. (In general, V is a
coisometry but, under a mild assumption, it may be chosen to be unitary.)
These results, together with our work on canonical models in [36], represent
a generalization of some of the essential ingredients of a program that has been
developed successfully in model theory the interaction between operator theory
and function theory on the unit disc D and has been generalized in various ways
to the polydisc and the ball in C
n
. This program sets up (essentially) bijective
correspondences connecting the theory of unitary colligations (and their unitary
systems), the Sz.-Nagy-Foias functional model theory for contraction operators
and the discrete-time Lax-Phillips scattering theory. Each theory produces a con-
tractive operator-valued function (called the transfer function of the system, the
characteristic operator function of the completely non-unitary contraction or the
scattering function for the scattering system) from which one can recover the orig-
inal object (the system or the contraction) up to unitary equivalence. For more
details, see the works of Ball ([12]), Ball and Vinnikov ([15]), Ball, Trent and
Vinnikov ([14]) and the references there.
We shall not discuss the program in detail here but we note that Theorem 4.3
below is the generalization, to our context, of Theorem 2.1 of [12] or Theorem 2.1
of [14]. Here the elements of H
(E) play the role of multipliers and the disc D in

C is replaced by the open unit ball of (E
.
We also note that the canonical models for contraction operators are re-
placed, in our setting, by canonical models for representations of H
(E). This
theory was developed in [36] for completely non-coisometric representations (gen-
eralizing results of Popescu in [44]) and it is shown there that the characteristic
operator function for such a representation has a realization associated with a
unitary colligation.
In the next section we set the stage by dening our basic constructions,
presenting examples and emphasizing the roles of duality and point evaluation in
the theory.
Section 3 deals with the Nevanlinna-Pick theorem and Section 4 with Schur
class operator functions.
2. Preliminaries: W
-correspondences and Hardy algebras

We shall follow Lance [27] for the general theory of Hilbert C
-modules that we
use. Let A be a C
-algebra and let E be a right module over A endowed with a

bi-additive map , ) : EE A (referred to as an A-valued inner product) such
that, for , E and a A, , a) = , )a, , )
= , ), and , ) 0,
with , ) = 0 only when = 0. If E is complete in the norm || := |, )|
1/2
,
the E is called a (right) Hilbert C
-module over A. We write L(E) for the space

of continuous, adjointable, A-module maps on E; that is every element of L(E) is
continuous and if X L(E), then there is an element X
L(E) that satises

X
, ) = , X). The element X
is unique and L(E) is a C
-algebra with
respect to the involution X X
and the operator norm. If M is a von Neumann

algebra and if E is a Hilbert C
-module over M, then E is said to be self-dual in

case every continuous M-module map from E to M is given by an inner product
with an element of E. If E is a self-dual Hilbert module over M, then L(E) is a
W
-algebra and coincides with all the bounded linear maps on E [40].
A C
-correspondence over a C
-algebra A is a Hilbert C
-module E over A
endowed with a structure of a left module over A via a *-homomorphism : A
L(E). When dealing with a specic C
-correspondence E over a C
-algebra A,
it will be convenient to suppress the in formulas involving the left action and
simply write a or a for (a). This should cause no confusion in context.
Having dened a left action on E, we are allowed to form balanced tensor
products. Given two correspondences E and F over the C
-algebra A one may

dene an A-valued inner product on the balanced tensor product E
A
F by the
formula
1
,
2
2
)
E
B
F
:=
1
, (
1
,
2
)
E
)
2
)
F
.
The Hausdor completion of this bimodule is again denoted by E
A
F and is
called the tensor product of E and F.
Denition 2.1. Let M be a von Neumann algebra and let E be a Hilbert C
-module
over M. Then E is called a Hilbert W
-module over M in case E is self-dual.

The module E is called a W
-correspondence over M in case E is a self-dual C
-
correspondence over M such that the -homomorphism : M L(E) giving the
left module structure on E is normal.
It is evident that the tensor product of two W
-correspondences is again a
W
-correspondence. Note also that, given a W
-correspondence E over M and a

Hilbert space H equipped with a normal representation of M, we may form the
Hilbert space E
H (by dening
1
h
1
,
2
h
2
) = h
1
, (
1
,
2
))h
2
)). Then,
given an operator X L(E) and an operator S (N)
, the map h XSh

denes a bounded operator on E
H denoted by X S. When S = I and

X = (a), a M, we get a representation of M on this space.
Observe that if E is a W
-correspondence over a von Neumann algebra M,

then each of the tensor powers of E, viewed as a C
-correspondence over M in the

usual way, is in fact a W
-correspondence over M and so, too, is the full Fock space

T(E), which is dened to be the direct sum M E E
2
, with its obvious
structure as a right Hilbert module over M and left action given by the map
,
dened by the formula
(a) := diaga, (a),

(2)
(a),
(3)
(a), . . . , where for all
n,
(n)
(a)(
1

2

n
) = ((a)
1
)
2

n
,
1

2

n
E
n
. The
tensor algebra over E, denoted T
+
(E), is dened to be the norm-closed subalgebra
of L(T(E)) generated by
(M) and the creation operators T
, E, dened
by the formula T
= , T(E). We refer the reader to [31] for the basic

facts about T
+
(E).
Denition 2.2. Given a W
-correspondence E over the von Neumann algebra M,

the ultraweak closure of the tensor algebra of E, T
+
(E), in the w
-algebra L(T(E)),
will be called the Hardy Algebra of E, and will be denoted by H
(E).
Example 2.3. If M = E = C then T(E) may be identied with H
2
(T). The tensor
algebra in this setting is isomorphic to the disc algebra A(D) and the Hardy algebra
is the classical Hardy algebra H
(T).
Example 2.4. If M = C and E = C
n
, then T(E) may be identied with the space
l
2
(F
+
n
) where F
+
n
is the free semigroup on n generators. The tensor algebra then is
what Popescu refers to as the non-commutative disc algebra /
n
and the Hardy
algebra is its w
-closure. It was studied by Popescu ([47]) and by Davidson and

Pitts who denoted it by L
n
([19]).
Example 2.5. Let M be a von Neumann algebra and let be a unital, injective,
normal

-endomorphism on M. The correspondence E associated with is equal
to M as a vector space. The right action is by multiplication, the M-valued inner
product is a, b) = a
b and the left action is given by ; i.e. (a)b = (a)b. We

write

M for E. It is easy to check that E
n
is isomorphic to

nM. The Hardy
algebra in this case is called the non-selfadjoint crossed product of M by and
will be written M
Z
+
. This algebra is also called an analytic crossed product,
at least when is an automorphism. It is related to the algebras studied in [28]
and [41]. If we write w for T
1
(where 1 is the identity of M viewed as an element
of E), then the algebra is generated by w and
(M) and every element X in the

algebra has a formal Fourier expression
X =
n=0
w
n
b
n
where b
n

(M). This Fourier expression is actually Ceasaro-summable to X

in the ultraweak topology on H
(E) [35], but we do not need these details in the

present discussion.
Example 2.6. Here we set M to be the algebra l
(Z) and let be the automorphism

dened by ((g))
i
= g
i1
. Write E for the correspondence

M as in Example 2.5.
Another, isomorphic, way to describe E is to let M be the algebra T of all diagonal
operators on l
2
(Z), let U be the shift dened by Ue
k
= e
k1
(where e
k
is the
standard basis), and set E = UT B(l
2
(Z)). The left and right actions on E are
dened simply by operator multiplications and the inner product is UD
1
, UD
2
) =
D
1
D
2
. It is easy to check that these correspondences are indeed isomorphic and
the Hardy algebra H
(E) is (completely isometrically isomorphic to) the algebra

| of all operators in B(l
2
(Z)) whose matrix (with respect to the standard basis) is
upper triangular.
Example 2.7. Suppose that is a normal, contractive, completely positive map
on a von Neumann algebra M. Then we may associate with it the correspondence
M
M obtained by dening the M-valued inner product on the algebraic tensor

product M M via the formula a b, c d) = b
(a
c)d and completing. (The

bimodule structure is by left and right multiplications.) This correspondence was
used by Popa ([43]), Mingo ([29]), Anantharam-Delarouche ([6]) and others to
study the map . If is an endomorphism this correspondence is the one described
in example 2.5.
Example 2.8. Let M be D
n
, the diagonal n n matrices and E be the set of all
n n matrices A = (a
ij
) with a
ij
= 0 unless j = i + 1 with the inner product
A, B) = A
B and the left and right actions given by matrix multiplication. Then
the Hardy algebra is isomorphic to T
n
, the nn upper triangular matrices. In fact,
a similar argument works to show that, for every nite nest of projections ^ on a
Hilbert space H, the nest algebra alg^ (i.e., the set of all operators on H leaving
the ranges of the projections in ^ invariant) may be written as H
(E) for some

W
-correspondence E.
Example 2.9. (Quiver algebras) Let Q be a directed graph on the set V of vertices.
For simplicity we assume that both V and Q are nite sets and view each Q
as an arrow from s() (in V ) to r() (in V ). Let M be C(V ) (a von Neumann
algebra) and E (or E(Q)) be C(Q). Dene the M-bimodule structure on E as
follows: for f E, M and Q,
(f)() = f()(s()),
and
(f)() = (r())f().
The M-valued inner product is given by the formula
f, g)(v) =
s()=v
f()g(),
for f, g E and v V . The algebra H
(E) in this case will be written H
(Q)
and is the -weak closure of T
+
(E(Q)). Viewing both algebras as acting on the
Fock space, one sees that they are generated by a set S
: Q of partial
isometries (here S
= T
where
is the function in C(Q) which is 1 at and

0 otherwise) and a set P
v
: v V of projections (i.e. the generators of
(M))
satisfying the following conditions.
(i) P
v
P
u
= 0 if u ,= v,
(ii) S
= 0 if ,=
(iii) S
= P
s()
and
(iv)
r()=v
S
P
v
for all v V .
These algebras were studied in [30] and [32], and also in [25] where they were
called free semigroupoid algebras.
2.1. Representations
In most respects, the representation theory of H
(E) follows the lines of the

representation theory of T
+
(E). However, there are some dierences that will be
important to discuss here. To help illuminate these, we need to review some of the
basic ideas from [31, 32, 35].
A representation of H
(E) (or of T
+
(E)) on a Hilbert space H is completely
determined by what it does to the generators. Thus, from a representation we
obtain two maps: a map T : E B(H), dened by T() = (T
), and a map
: M B(H), dened by (a) = (
(a)). Analyzing the properties of T and

one is lead to the following denition.
Denition 2.10. Let E be a W
-correspondence over a von Neumann algebra M.

Then a completely contractive covariant representation of E on a Hilbert space H
is a pair (T, ), where
1. is a normal -representation of M in B(H).
2. T is a linear, completely contractive map from E to B(H) that is continuous
in the -topology of [11] on E and the ultraweak topology on B(H).
3. T is a bimodule map in the sense that T(SR) = (S)T()(R), E, and
S, R M.
It should be noted that there is a natural way to view E as an operator space
(by viewing it as a subspace of its linking algebra) and this denes the operator
space structure of E to which the Denition 2.10 refers when it is asserted that T
is completely contractive.
As we noted in the introduction and developed in [31, Lemmas 3.43.6] and
in [35], if a completely contractive covariant representation, (T, ), of E in B(H)
is given, then it determines a contraction

T : E
H H dened by the formula
T( h) := T()h, h E
H. The operator

T satises
T(() I) = ()
T. (2)
In fact we have the following lemma from [35, Lemma 2.16].
Lemma 2.11. The map (T, )

T is a bijection between all completely contrac-
tive covariant representations (T, ) of E on the Hilbert space H and contractive
operators

T : E
H H that satisfy equation (2). Given and a contraction

T
satisfying the covariance condition (2), we get a completely contractive covariant
representation (T, ) of E on H by setting T()h :=

T( h).
The following theorem shows that every completely contractive representa-
tion of the tensor algebra T
+
(E) is given by a pair (T, ) as above or, equivalently,
by a contraction

T satisfying (2).
Theorem 2.12. ([31, Theorem 3.10]) Let E be a W
-correspondence over a von

Neumann algebra M. To every completely contractive covariant representation,
(T, ), of E there is a unique completely contractive representation of the tensor
algebra T
+
(E) that satises
(T
) = T() E and (
(a)) = (a) a M.
The map (T, ) is a bijection between the set of all completely contractive
covariant representations of E and all completely contractive (algebra) represen-
tations of T
+
(E) whose restrictions to
(M) are continuous with respect to the

ultraweak topology on L(T(E)).
Denition 2.13. If (T, ) is a completely contractive covariant representation of a
W
-correspondence E over a von Neumann algebra M, we call the representation

of T
+
(E) described in Theorem 2.12 the integrated form of (T, ) and write
= T.
Example 2.14. In the context of Example 2.4, M = C and E = C
n
. Then, a
completely contractive covariant representation of E is simply given by a completely
contractive map T : E B(H). Writing T
k
= T(e
k
), where e
k
is the standard
basis in C
n
, and identifying C
n
H with the direct sum of n copies of H, we may
write

T as a row (T
1
, T
2
, . . . , T
n
). The condition that |
T| 1 is the condition
(studied by Popescu [45] and Davidson and Pitts [19]) that
T
i
T
i
1. Hence
representations of the non-commutative disc algebras are given by row contractions.
Example 2.15. Consider the setting of Example 2.9 and let V, Q, M and E be as
dened there. A (completely contractive covariant) representation of E is given by
a representation of M = C(V ) on a Hilbert space H and by a contractive map
T : E
H H satisfying (2) above. Write

v
for the function in C(V ) which is 1
on v and 0 elsewhere. The representation is given by the projections Q
v
= (
v
)
whose sum is I. For every Q write
for the function (on E) which is

1 at and 0 elsewhere. Given

T as above, we may dene maps T() B(H)
by T()h =

T(
h) and it is easy to check that

T

T
T()T()
and
T() = Q
r()
T()Q
s()
. Thus to every (completely contractive) representation of
the quiver algebra T
+
(E(Q)) we associate a family T() : Q of maps on H
that satisfy
T()T()
I and T() = Q
r()
T()Q
s()
. Conversely, every
such family denes a representation by writing

T(f h) =
f()T()h. Thus,
representations are indexed by such families. Note that, in fact, (T)(S
) = T()
and ( T)(P
v
) = Q
v
(where S
and P
v
are as in Example 2.9).
Remark 2.16. One of the principal diculties one faces in dealing with T
+
(E) and
H
(E) is to decide when the integrated form, T, of a completely contractive

covariant representation (T, ) extends from T
+
(E) to H
(E). This problem arises

already in the simplest situation, namely when M = C = E. In this setting, T is
given by a single contraction operator T(1) on a Hilbert space, T
+
(E) is the
disc algebra and H
(E) is the space of bounded analytic functions on the disc.

The representation T extends from the disc algebra to H
(E) precisely when

there is no singular part to the spectral measure of the minimal unitary dilation of
T(1). We are not aware of a comparable result in our general context but we have
some sucient conditions. One of them is given in the following lemma. It is not
necessary in general.
Lemma 2.17. ([35]) If |
T| < 1 then T extends to a -weakly continuous

representation of H
(E).
Other sucient conditions are presented in Section 7 of [35].
2.2. Duality and point evaluation
The following denition is motivated by condition (2) above.
Denition 2.18. Let : M B(H)be a normal representation of the von Neumann
algebra M on the Hilbert space H. Then for a W
-correspondence E over M, the

-dual of E, denoted E
, is dened to be
B(H, E
H) [ (a) = ((a) I), a M.

As we note in the following proposition, the -dual carries a natural structure
of a W
-correspondence. The reason to dene the -dual using covariance condi-

tion which is the adjoint of condition (2) is to get a right W
-module (instead
of a left W
-module) over (M)
.
Proposition 2.19. With respect to the actions of (M)
and the (M)
-valued inner
product dened as follows, E
becomes a W
-correspondence over (M)
: For
a, b (M)
, and E
, ab := (Ia)b, and for , E
, , )
(M)
:=
.
Example 2.20. If M = E = C, H is arbitrary and is the representation of C on
H, then (M)
= B(H) and E
= B(H).
Example 2.21. If is a contractive, normal, completely positive map on a von
Neumann algebra M and if E = M
M (see Example 2.7 ) then, for every faith-

ful representation of M on H, the -dual is the space of all bounded operators
mapping H into the Stinespring space K (associated with as a map from M
to B(H)) that intertwine the representation (on H) and the Stinespring repre-
sentation (on K). This correspondence was proved very useful in the study of
completely positive maps. (See [33], [38] and [34]). If M = B(H) this is a Hilbert
space and was studied by Arveson ([10]). Note also that, if is an endomorphism,
then this dual correspondence is the space of all operators on H intertwining and
.
We now turn to dene point evaluation. Note that, given as above, the
operators in E
whose norm does not exceed 1 are precisely the adjoints of the
operators of the form

T for a covariant pair (T, ). In particular, every in the
open unit ball of E
(written D(E
)) gives rise to a covariant pair (T, ) (with

=

T
) such that T is a representation of H
(E). Given X H
(E) we
may apply T to it. The resulting operator in B(H) will be denoted by X(
).
That is,
X(
) = ( T)(X)
where

T =
.
In this way, we view every element in the Hardy algebra as a (B(H)-valued)
function on D((E
).
Example 2.22. Suppose M = E = C and the representation of C on some Hilbert
space H. Then H
(E) = H
(T) and (Example 2.20) E
is isomorphic to B(H).
If X H
(E) = H
(T), so that we may view X with a bounded analytic function

on the open disc in the plane, then for S E
= B(H), it is not hard to check that

X(S
), as dened above, is the same as the value provided by the Sz.-Nagy-Foias

H
-functional calculus.
Note that, for a given D(E
), the map X X(
) is a -weakly contin-
uous homomorphism on the Hardy algebra. Thus, in order to compute X(
), it
suces to know its values on the generators. This is given in the following (easy
to verify) lemma.
Lemma 2.23. Let be a faithful normal representation of M on H and for E
write L
for the map from H to E
H dened by L
h = h. Then, for E,
a M and D(E
),
(i) (T
)(
) =
, and
(ii) (
(a))(
) = (a)
(Recall that
is a map from E
H to H.)
A formula for computing X(
), without referring to the generators, will be

presented later (Proposition 2.30).
Example 2.24. In the setting of Example 2.5 we may identify the Hilbert space
E
H =

M
H with H via the unitary operator mapping ah (in

M
H)
to (a)h. Using this identication, we may identify E
with B(H) : (a) =

((a)), a M.
Applying Lemma 2.23, we obtain w(
) = T
1
(
) =
L
1
=
(viewed
now as an operator in B(H)). Thus, if X =
w
n
b
n
(as a formal series), with
b
n
=
(a
n
) and D(E
), then
X(
) =
)
n
(a
n
)
with the sum converging in the norm on B(H). (In a sense, this equation asserts
that Ceasaro summability implies Abel summability even in this abstract setting.)
Example 2.25. Let T, U and E = UT be as in Example 2.6. Let be the identity
representation of T on H = l
2
(Z). The map V (UD
h) = Dh (for D T, h H)
is a unitary operator from E
H onto H such that, for every E
, V U
T
and, conversely, for every D T, V
lies in E
. We write
D
for V
.
Recall that the Hardy algebra is | (the algebra of all upper triangular operators on
H). Given X | we shall write X
n
for the n
th
upper diagonal of X. A simple
computation shows that, for D T with |D| < 1,
X(
D
) =
n=0
U
n
(U
D)
n
X
n
.
Note here that, in [4], the authors dened point evaluations for operators X |.
In their setting one evaluates X on the open unit ball of T and the values are also
in T. Their formula (for what in [5] is called the right point evaluation) is
X
(D) =
n=0
U
n
(U
D)
n
X
n
U
n
.
(One can also dene a left point evaluation.) The apparent similarity of the two
formulas above may be deceiving. Note that both their point evaluation and ours
can be dened also for block upper triangular operators (acting on l
2
(Z, K) for
some Hilbert space K). But, in that case, the relation between the two formu-
las is no longer clear. In fact, our point evaluation is multiplicative (that is,
(XY )(
) = X(
)Y (
)) while theirs is not. On the other hand, their point

evaluation is designed to satisfy the property that, for X | and D T,
(X X
(D))(U D)
1
| ([4, Theorem 3.4]). For our point evaluation (in the
general setting), it is not even clear how to state such a property.
Example 2.26. (Quiver algebras) Let Q be a quiver as in Example 2.9 and write
E(Q) for the associated correspondence. We x a faithful representation of M =
C(V ) on H. As we note in Example 2.15, this gives a family Q
v
of projections
whose sum is I (and, as is faithful, none is 0). Write H
v
for the range of Q
v
.
Then (M)
=
v
B(H
v
) and we write elements there as functions dened on V
with (v) B(H
v
). To describe the -dual of E we can use Example 3.4 in [35].
We may also use the description of the maps

T in Example 2.15 because every
in the closed unit ball of E
is

T
for some representation (, T) of E. Using

this, we may describe an element of E
as a family of B(H)-valued operators

() : Q
1
where Q
1
is the quiver obtained from Q by reversing all arrows.
The (M)
-module structure of E
is described as follows. For E
, (M)
and Q
1
,
()() = ()(s()),
and
()() = (r())().
The (M)
-valued inner product is given by the formula

, )(v) =
s()=v
()
(),
for , E
and v V .
Recall that the quiver algebra is generated by a set of partial isometries S
and projections P
v
(see Example 2.9). If is given and
=

T lies in the open
unit ball of (E
and

T is given by a row contraction (T()) (as in Example 2.15),
then the point evaluation for the generators is dened by S
) = T() = (
1
)
and P
v
(
) = Q
v
. For a general X H
(Q), X(
) is dened by the linearity,

multiplicativity and -weak continuity of the map X X(
).
We turn now to some general results concerning the -dual. First, the term
dual that we use is justied by the following result.
Theorem 2.27. ([35, Theorem 3.6]) Let E be a W
-correspondence over M and let

be a faithful, normal representation of M on H. If we write for the identity
representation of (M)
(on H) then one may form the -dual of E
and we have
(E

= E.
The following lemma summarizes Lemmas 3.7 and 3.8 of [35] and shows that
the operation of taking duals behaves nicely with respect to direct sums and tensor
products.
Lemma 2.28. Given W
-correspondences E,E
1
and E
2
over M and a faithful rep-
resentation of M on H, we have
(i) (E
1
E
2
)

= E
1
E
2
.
(ii) (E
1
E
2
)

= E
2
E
1
.
(iii) T(E)

= T(E
).
(iv) The map h (h) induces a unitary operator from E
H onto E
H.
(v) Applying item (iv) above to T(E) in place of E, we get a unitary operator U
from T(E
) H onto T(E) H.
Although H
(E) was dened as a subalgebra of L(T(E)) it is often useful

to consider a (faithful) representation of it on a Hilbert space. Given a faithful,
normal, representation of M on H we may induce it to a representation of the
Hardy algebra. To do this, we form the Hilbert space T(E)
H and write
Ind()(X) = X I, X H
(E).
(in fact, this is well dened for every X in L(T(E)). Such representations were
studied by M. Rieel in [48]). Ind() is a faithful representation and is an home-
omorphism with respect to the -weak topologies. Similarly one denes Ind(), a
representation of H
(E
). The following theorem shows that, roughly speaking,

the algebras H
(E) and H
(E
) are the commutant of each other.

Theorem 2.29. [35, Theorem 3.9] With the operator U as in part (v) of Lemma 2.28,
we have
U
(Ind()(H
(E
)))U = (Ind()(H
(E)))
and, consequently,
(Ind()(H
(E)))
= Ind()(H
(E)).
We may now use the notation set above to present a general formula for point
evaluation. For its proof, see [35, Proposition 5.1].
Proposition 2.30. If is a faithful normal representation of M on H, let
H
denote
the imbedding of H into T(E
) H and write P
k
for the projection of T(E
) H
onto (E
)
k
H. Also, for D(E
) and k 1, note that

k
lies in (E
)
k
and that L
k
maps (E
)
k
H into H in the obvious way (and, for k = 0, this
is
H
). Then, for every X H
(E),
X(
) =
k=0
L
k
P
k
U
(X I)U
H
where U is as dened in Lemma 2.28.
3. Nevanlinna-Pick Theorem
Our goal in this section is to present a generalization of the Nevanlinna-Pick The-
orem. First, recall the classical theorem.
Theorem 3.1. . Let z
1
, . . . , z
k
C with [z
i
[ < 1 and w
1
, . . . , w
k
C. Then the
following conditions are equivalent.
(1) There is a function f H
(T) with |f| 1 such that f(z

i
) = w
i
for all i.
(2)
_
1 w
i
w
j
1 z
i
z
j
_
0.
Since we are able to view elements of H
(E) as functions on the open unit

ball of E
, it makes sense to seek necessary and sucient conditions for nding an

element X H
(E) with norm less or equal 1 whose values at some prescribed

points,
1
, . . . ,
k
, in that open unit ball are prescribed operators C
1
, . . . , C
k
in
B(H). To state our conditions we need some notation. For operators B
1
, B
2
in
B(H) we write Ad(B
1
, B
2
) for the map, from B(H) to itself, mapping S to B
1
SB
2
.
Also, for elements
1
,
2
in D(E
), we let
1
,
2
denote the map, from (M)
to
itself, that sends a to
1
, a
2
). Then our generalization of the Nevanlinna-Pick
theorem may be formulated as follows.
Theorem 3.2. Let be a faithful normal representation of M on H. Fix
1
,...,
k
with |
i
| < 1 and B
1
, . . . , B
k
, C
1
, . . . , C
k
B(H). Then the following condi-
tions are equivalent
(1) There exists an X H
(E) with |X| 1 such that B

i
X(
i
) = C
i
for all i.
(2) The map from M
k
((M)
) into M
k
(B(H)) dened by the k k matrix
_
(Ad(B
i
, B
j
) Ad(C
i
, C
j
)) (id
i
,
j
)
1
_
is completely positive.
Remark 3.3. If M = B(H) (and, then (M)
= CI), condition (2) becomes

_
B
i
B
j
C
i
C
j
1
i
,
j
)
_
0.
This follows easily from a result of M.D. Choi ([18]).
For the complete proof of Theorem 3.2 we refer the reader to [35, Theorem
5.3]. Here we just remark that in order to prove that (1) implies (2) one uses the
complete positivity condition of (2) to construct a subspace / T(E
) H that
is invariant under Ind()(H
(E
))
and a contraction R that commutes with the

restriction of Ind()(H
(E
))
to /. Then it is possible to apply the commutant

lifting theorem of [31, Theorem 4.4] to R
to get a contraction on T(E
) H that
commutes with Ind()(H
(E
)). An application of Theorem 2.29 completes the

proof.
The following is a consequence of Theorem 3.2. It may be viewed as a gener-
alization of the classical Schwartzs lemma.
Theorem 3.4. Suppose an element X of H
(E) has norm at most one and satises

the equation X(0) = 0. Then for every
D((E
) the following assertions are

valid:
1. If a is a non-negative element in (M)
, and if , a ) a, then
X(
)aX(
, a ).
2. If
k
denotes the element E
k
, then
X(
)
k
,
k
)X(

k+1
,
k+1
).
3. X(
)X(
, ).
We now illustrate how to apply Theorem 3.2 in various settings.
Example 3.5. When M = H = E = C, we obtain Theorem 3.1.
Example 3.6. If M = E = C and if H is arbitrary, then E
= B(H) and Theorem

3.2 yields the following result.
Theorem 3.7. Given T
1
, . . . , T
k
B(H), |T
i
| < 1 and B
1
, . . . , B
k
, C
1
, . . . , C
k
in
B(H). Then the following conditions are equivalent.
(1) There exists a function f H
(T) with |f| 1 and B

i
f(T
i
) = C
i
.
(2) The map dened by the matrix (
ij
) is completely positive where
ij
(A) =
k=0
(B
i
T
k
i
AT
k
j
B
j
C
i
T
k
i
AT
k
j
B
j
).
Example 3.8. Assume M = B(H) = E. Then M
= CI and E
= C and Theorem
3.2 specializes to the following.
Theorem 3.9. Given z
1
, . . . , z
k
D and B
1
, . . . , B
k
, C
1
, . . . , C
k
in B(H), then the
following conditions are equivalent.
(1) There exists G H
(T) B(H) with |G| 1 such that B

i
G(z
i
) = C
i
for
all i.
(2)
_
B
i
B
j
C
i
C
j
1 z
i
z
j
_
0.
Example 3.10. Set M = B(H) and E = C
n
(B(H)) (that is, E is a column of n
copies of B(H)). Then M
= CI, E
= C
n
and Theorem 3.2 yields the following
theorem due to Davidson and Pitts [19], Arias and Popescu [8] and Popescu [47].
Theorem 3.11. Given
1
, . . . ,
k
in the open unit ball of C
n
and C
1
, . . . , C
k

B(H), then the following conditions are equivalent.
(1) There is a Y B(H) L
n
with |Y | 1 such that (
i
id)(Y ) = C
i
for
all i.
(2)
_
I C
i
C
j
1
i
,
j
)
_
0.
Moreover, if, for all i, the C
i
all lie in some von Neumann subalgebra N
B(H), then Y can be chosen in N L
n
.
Our nal example of this section concerns interpolation for nest algebras.
The rst interpolation result for nest algebras was proved by Lance ([26]). It was
later generalized by Anoussis ([7]) and by Katsoulis, Moore and Trent ([24]). A
related result was proved by Ball and Gohberg ([13]). The results we present below
recapture the results of [24].
Theorem 3.12. Let ^ be a nest of projections in B(H) and x B, C in B(H).
Then the following conditions are equivalent.
(1) There exists an X Alg^ with |X| 1 and BX = C.
(2) For all projections N ^, CNC
BNB
.
The vector version of this theorem is the following.
Corollary 3.13. Let ^ be a nest in B(H) and x u
1
, . . . , u
k
, v
1
, . . . , v
k
in H. Then
the following conditions are equivalent.
(1) There exists X Alg^ with |X| 1 and Xu
i
= v
i
for all i.
(2) For all N ^,
_
N
v
i
, N
v
j
)
_
_
N
u
i
, N
u
j
)
_
,
where N
denotes I N.
These results are not immediate corollaries of Theorem 3.2 because, for a
general nest ^, Alg^ is not of the form H
(E). However, when ^ is nite, Alg^

is a Hardy Algebra by Example 2.8. In this case, the conclusions are fairly straight
forward computations. The case of general nests is then handled by approximation
techniques along the lines of [26] and [9]. Full details may be found in [35, Theorem
6.8 and Corollary 6.9].
4. Schur class operator functions and realization
In this section we relate the complete positivity condition of Theorem 3.2 to the
concept of a Schur class function. As mentioned in the introduction, this may be
viewed as part of a general program to nd equivalences between canonical model
theory, non-commutative systems theory and scattering theory. The results be-
low are proved in [37].
We start with the following denition.
Denition 4.1. Let S be a set, A and B be two C
-algebras and write B(A, B) for

the space of bounded linear maps from A to B. A function
K : S S B(A, B)
will be called a completely positive denite kernel (or a CPD-kernel ) if, for all
choices of s
1
, . . . , s
k
in S, the map
K
(k)
: (a
ij
) (K(s
i
, s
j
)(a
ij
))
from M
k
(A) to M
k
(B) is completely positive.
This concept of CPD-kernels was studied in [16] (see, in particular, Lemma
3.2.1 there for conditions on K that are equivalent to being a CPD-kernel).
Denition 4.2. Let c be a Hilbert space and Z : D((E
) B(c) be a B(c)-valued
function. Then Z is said to be a Schur class operator function if
K(
) = (id Ad(Z(
), Z(
)) (id
,
)
1
is a CPD-kernel on D((E
). (We use here the notation set for Theorem 3.2).

Note that, when M = E = B(c) and is the identity representation of B(c)
on c, (M)
is CI
L
, E
is isomorphic to C and D((E
) may be identied with

the open unit ball D of C. In this case the denition above recovers the classical
Schur class functions. More precisely, these functions are usually dened as analytic
functions Z from D into the closed unit ball of B(c) but it is known that this is
equivalent to the positivity of the Pick kernel k
Z
(z, w) = (I Z(z)Z(w)
)(1
z w)
1
. The argument of [35, Remark 5.4] shows that the positivity of this kernel
is equivalent, in this case, to the condition of Denition 4.2.
Note that it follows from Theorem 3.2 that every operator in the closed unit
ball of H
(E) determines (by point evaluation) a Schur class operator function.

In fact we have the following result whose proof may be found in [37].
Theorem 4.3. ([37]) Let E be a W
-correspondence over a von Neumann algebra

M and let be a faithful normal representation of M on a Hilbert space c. For a
function Z : D((E
) B(c), the following conditions are equivalent.

(1) Z is a Schur class operator function.
(2) There is an X in the closed unit ball of H
(E) such that X(
) = Z(
) for
all D(E
).
(3) (Realization) There is a Hilbert space H, a normal representation of N :=
(M)
on H and operators A, B, C and D such that

(i) A B(c),B B(H, c), C B(c, H) and D B(H, E
H).
(ii) A, B, C and D intertwine the actions of N (on the relevant spaces).
(iii) The operator
V :=
_
A B
C D
_
:
_
c
H
_
_
c
E
H
_
is a coisometry.
(iv) For every D(E
),
Z(
) = A +B(I L
D)
1
L
C
where L
: H E
H is dened by L
h = h.
Note that X in part (2) of the Theorem is not necessarily unique. (Although,
as shown in [37], it is possible to choose such that the choice of X will be unique).
One may apply the techniques developed (in [37]) for the proof of the Theo-
rem 4.3 to establish the following extension result.
Proposition 4.4. Every function dened on a subset of the open unit ball of
(E
with values in some B(c) such that the associated kernel (dened on )
is a CPD-kernel may be extended to a Schur class operator function (dened on
all of D((E
)).
References
[1] J. Agler and J. McCarthy, Pick interpolation and Hilbert function spaces. Graduate
Studies in Mathematics, vol. 44. Amer. Math. Soc., Providence (2002).
[2] D. Alpay, P. Bruinsma, A. Dijksma and H. de Snoo, Interpolation problems, exten-
sions of symmetric operators and reproducing kernel spaces II, Int. Eq. Oper. Thy.
14 (1991), 465500.
[3] D. Alpay, P. Bruinsma, A. Dijksma and H. de Snoo, Interpolation problems, exten-
sions of symmetric operators and reproducing kernel spaces II (missing section 3),
Int. Eq. Oper. Thy. 15 (1992), 378388.
[4] D. Alpay, P. Dewilde and H. Dym, Lossless inverse scattering and reproducing kernels
for upper triangular operators, Operator Theory: Adv. Appl., Birkhauser Verlag,
Basel 47 (1990), 61133.
[5] D. Alpay, A Dijksma and Y. Peretz, Nonstationary analogues of the Herglotz repre-
sentation theorem: the discrete case, J. Funct. Anal. 66 (1999), 85129.
[6] C. Anantharaman-Delaroche, On completely positive maps dened by an irreducible
correspondence, Canad. Math. Bull. 33 (1990), 434441.
[7] M. Anoussis, Interpolating operators in nest algebras, Proc. Amer. Math. Soc. 114
(1992), 707710.
[8] A. Arias and G. Popescu, Noncommutative interpolation and Poison transforms,
Israel J. Math. 115 (2000), 205234.
[9] Wm. Arveson, Interpolation problems in nest algebras, J. Funct. Anal. 3 (1975),
208233.
[10] W.B. Arveson, Continuous analogues of Fock space, Mem. Amer. Math. Soc. 80
(1989).
[11] M. Baillet, Y. Denizeau and J.-F. Havet, Indice dune esperance conditionelle, Comp.
Math. 66 (1988), 199236.
[12] J. Ball, Linear systems, operator model theory and scattering: multivariable general-
izations, in Operator Theory and its applications (Winnipeg, 1998) (Ed. A.G. Ramm,
P.N. Shivakumar and A.V. Strauss), Fields Inst. Comm. vol. 25, Amer. Math. Soc.,
Providence, 2000, 151178.
[13] J. Ball and I. Gohberg, A commutant lifting theorem for triangular matrices with
diverse applications, Integral Equat. Operator Theory 8 (1985), 205267.
[14] J. Ball, T. Trent and V. Vinnikov, Interpolation and commutant lifting for multipliers
on reproducing kernel Hilbert spaces. Preprint.
[15] J. Ball and V. Vinnikov, Functional models for representations of the Cuntz algebra.
Preprint.
[16] S.D. Barreto, B.V.R. Bhat, V. Liebscher and M. Skeide, Type I product systems of
Hilbert modules, to appear in J. Functional Anal.
[17] H. Bercovici, Operator theory and arithmetic in H
. Mathematical Surveys and

Monographs, 26. American Mathematical Society, Providence, RI, 1988.
[18] M.D. Choi, Completely positive linear maps on complex matrices, Lin. Alg. Appl. 10
(1975), 285290.
[19] K. Davidson and D. Pitts, The algebraic structure of non-commutative analytic
Toeplitz algebras, Math. Ann. 311 (1998), 275303.
[20] H. Dym, J-contractive matrix functions, reproducing kernel Hilbert spaces and inter-
polation. CBMS Regional Conference Ser. Math. 71. Amer. Math. Soc., Providence,
RI, 1989. x+147 pp.
[21] P. Gabriel, Unzerlegbare Darstellungen I, Manuscr. Math. 6 (1972), 71103.
[22] P. Gabriel and A.V. Roiter, Representations of Finite-Dimensional Algebras, Algebra
VIII, Encyclopaedia of Mathematical Sciences, Vol. 73, Springer-Verlag, 1991.
[23] G. Hochschild, On the structure of algebras with nonzero radical, Bull. Amer. Math.
Soc. 53 (1947), 369377.
[24] E.G. Katsoulis, R.L. Moore and T.T. Trent, Interpolation in nest algebras and ap-
plications to operator corona theorems, J. Operator Th. 29 (1993), 115123.
[25] D. Kribs and S. Power, Free semigroupoid algebras, Preprint.
[26] E.C. Lance, Some properties of nest algebras, Proc. London Math. Soc. (3) 19 (1969),
4568.
[27] E.C. Lance , Hilbert C
-modules, A toolkit for operator algebraists, London Math.

Soc. Lecture Notes series 210 (1995). Cambridge Univ. Press.
[28] M. McAsey and P.S. Muhly, Representations of non-self-adjoint crossed products,
Proc. London Math. Soc. 47 (1983), 128144.
[29] J. Mingo, The correspondence associated to an inner completely positive map, Math.
Ann. 284 (1989), 121135.
[30] P.S. Muhly, A nite-dimensional introduction to operator algebra in Operator alge-
bras and applications (Samos, 1996), 313354, NATO Adv. Sci. Inst. Ser. C Math.
Phys. Sci., 495, Kluwer Acad. Publ., Dordrecht, 1997.
[31] P.S. Muhly and B. Solel, Tensor algebras over C
-correspondences (Representations,
dilations and C
-envelopes), J. Funct. Anal. 158 (1998), 389457.

[32] P.S. Muhly and B. Solel , Tensor algebras, induced representations, and the Wold
decomposition, Canad. J. Math. 51 (1999), 850880.
[33] P.S. Muhly and B. Solel, Quantum Markov processes (correspondences and dilations),
Int. J. Math. 13 (2002), 863906.
[34] P.S. Muhly and B. Solel, The curvature and index of completely positive maps, Proc.
London Math. Soc. 87 (2003), 748778.
[35] P.S. Muhly and B. Solel, Hardy algebras, W
-correspondences and interpolation the-

ory, to appear in Math. Ann.
[36] P.S. Muhly and B. Solel, On canonical models for representations of Hardy algebras.
In preparation.
[37] P.S. Muhly and B. Solel, Schur class operator functions associated with a W
-
correspondence. In preparation.
[38] P.S. Muhly, M. Skeide and B. Solel, Representations of B
a
(E), commutants of von
Neumann bimodules, and product systems of Hilbert modules. In preparation.
[39] B. Sz.-Nagy and C. Foias, Analyse Harmonique des Operateurs de lespace de Hilbert,
Akademiai Kiado (1966).
[40] W. Paschke, Inner product modules over B
-algebras, Trans. Amer. Math. Soc. 182

(1973), 443468.
[41] J. Peters, Semi-crossed products of C*-algebras, J. Funct. Anal. 59 (1984), 498534.
[42] M. Pimsner, A class of C
-algebras generalizing both Cuntz-Krieger algebras and

crossed products by Z, in Free Probability Theory, D. Voiculescu, Ed., Fields Institute
Comm. 12, 189212, Amer. Math. Soc., Providence, 1997.
[43] S. Popa, Correspondences, Preprint (1986).
[44] G. Popescu, Characteristic functions for innite sequences of noncommuting opera-
tors, J. Oper. Theory 22 (1989), 5171.
[45] G. Popescu, von Neumann inequality for B(H
n
)
1
, Math. Scand. 68 (1991), 292304.
[46] G. Popescu, Functional calculus for noncommuting operators, Mich. Math. J. 42
(1995), 345356.
[47] G. Popescu, Commutant lifting, tensor algebras and functional calculus, Proc. Edin-
burg Math. Soc. 44 (2001), 389406.
[48] M.A. Rieel, Induced representations of C
-algebras, Adv. in Math. 13 (1974), 176

257.
[49] D. Sarason, Generalized interpolation in H
, Trans. Amer. Math. Soc. 127 (1967),

179203.
Paul S. Muhly
University of Iowa
Iowa City, IA 52242, USA
e-mail: muhly@math.uiowa.edu
Baruch Solel
Technion
32000 Haifa, Israel
e-mail: mabaruch@techunix.technion.ac.il
Operator Theory:
Notes on Generalized Lemniscates
Mihai Putinar
Abstract. A series of analytic and geometric features of generalized lemnis-
cates are presented from an elementary and unifying point of view. A novel
interplay between matrix theory and elementary geometry of planar alge-
braic curves is derived, with a variety of applications, ranging from a classical
Fredholm eigenvalue problem and Hardy space estimates to a root separation
algorithm.
Mathematics Subject Classication (2000). Primary 47A56, 47A48; Secondary
15A22, 14P05.
Keywords. lemniscate, rational embedding, determinantal variety, Schwarz re-
ection, quadrature domain, Hardy space.
1. Introduction
The object of study in this article is the level set, with respect to a Hilbert space
norm, of the resolvent of a matrix localized at a vector. We call these sets gen-
eralized lemniscates in analogy with the classical lemniscates, that is level sets of
the modulus of a polynomial. The latter class of domains is well known for its ap-
plications to approximation theory and potential theory, see for instance [10, 22]
and [30].
The origin of this study goes back to some determinantal functions related
to the spectral theory of hyponormal operators. More specically, if T L(H)
is a linear bounded operator acting on the Hilbert space H and the commutator
[T
, T] = is non-negative and rank-one, then the innite determinant:

det[(T
z)
1
(T z)(T
z)(T z)
1
] =
1 |(T
z)
1
|
2
was instrumental in understanding the ne structure of these operators. Its prop-
erties were rst investigated from the point of view of perturbation theory of
Paper supported by the National Science Foundation Grant DMS 0100367.
244 M. Putinar
symmetric operators and of scattering theory, see [25], [6]; for more references and
later related works see [24] and [28].
We have remarked in [27] that all rank-one self-commutator operators T as
above, subject to the additional condition
dim
k=0
T
< ,
are in a natural correspondence to Aharonov and Shapiros quadrature domains
([2]). In particular, the spectrum of these nite type operators is precisely given
by the rational equation:
1 |(A z)
1
|
2
< 0,
where A is the nite matrix obtained by compressing T
to
_
k=0
T
. This is
a generalized lemniscate, in the terminology adopted below. This very correspon-
dence between planar algebraic curves and matrices proved to be fruitful for better
understanding the nature of the dening equation of the boundary of a quadrature
domain (already investigated by Gustafsson [14]).
An eective exact reconstruction algorithm of a quadrature domains from a
part of its moments was also derived from the same observation, [12]. Although
closely related to generalized lemniscates, we will not expand here these ideas.
The class of generalized lemniscates has emerged from such concepts and
computations. We have tried below to simplify the access to these planar domains
and to make it independent of any sophisticated theory of semi-normal operators.
Most of the operator theory and approximation theory aspects as well as impor-
tant connections to extremal moment problems are left aside. They are partially
explained in the recent survey article [28], or in the more technical papers cited
there.
One of the aims of this essay is to connect in simple terms a variety of ideas
of linear algebra, realization theory of linear systems, algebraic geometry and some
classical analysis on planar domains. We are well aware that this is only a rst
step. The material below was freely borrowed and compiled into another form from
a series of recent articles published in the last ve years: [5], [18], [26], [29]. This
was done with the hope that the entire is more than its parts.
2. Realization theory
In this section we link the specic form of the algebraic equation of a generalized
lemniscate to a Hilbert space realization of it. This is in accord to the well-known
matrix realization of transfer functions in linear systems theory.
Let
Q(z, z) =
d
j,k=0
jk
z
j
z
k
,
Notes on Generalized Lemniscates 245
be a Hermitian polynomial in (z, z), that is
jk
=
kj
, 1 j, k d. We will
assume that the leading coecient is non-zero, and normalized:
dd
= 1,
and we denote
P(z) =
d
j=0
jd
z
j
.
Then
[P(z)[
2
Q(z, z) =
d
j,k=0
[
jd
dk
jk
]z
j
z
k
.
The following result, proved in [17], can be taken as a starting point for our
discussion.
Theorem 2.1. The following conditions are equivalent:
a) The matrix A() = (
jd
dk
jk
)
d1
j,k=0
is strictly positive denite;
b) There exists a linear transformation A of C
d
with a cyclic vector so that
P(A) = 0 and
Q(z, z)
[P(z)[
2
= 1 |(A z)
1
|
2
; (2.1)
c) There exist polynomials Q
k
(z) of degree k (exactly), 0 k < d, with the
property
Q(z, z) = [P(z)[
2
d1
k=0
[Q
k
(z)[
2
. (2.2)
In c), the Q
k
s are uniquely determined if the leading coecients are required
to be positive.
Proof. Below we simply sketch the main steps of the proof. More details are in-
cluded in [17].
a) b). Assume that A() is positive denite. Then there exist linearly indepen-
dent vectors v
k
C
d
, 0 k < d, satisfying:
v
j
, v
k
) =
jd
dk
jk
,
and consequently
[P(z)[
2
Q(z, z) = |V (z)|
2
, (2.3)
where V (z) =
d1
j=0
v
j
z
j
is a vector-valued polynomial.
It follows that R : P
1
P
d
, dened in terms of homogeneous coordinates in
P
d
by R(z) = (P(z) : V (z)), is a rational map of degree d such that the image of
R spans P
d
. Elementary arguments of linear algebra imply then the existence of a
matrix A L(C
d
) with minimal polynomial P(z) and cyclic vector , such that:
V (z) = P(z)(A z)
1
.
This proves assertion b).
246 M. Putinar
b) c). To achieve the decomposition (2.2) we orthonormalize the vectors
, A, . . . , A
d1
:
e
0
=

||
,
e
1
=
A A, e
0
)e
0
| . . . |
,
e
2
=
A
2
A
2
, e
1
)e
1
A
2
, e
0
)e
0
| . . . |
,
etc. Equivalently,
= ||e
0
= c
0
e
0
(c
0
> 0),
A = c
1
e
1
+A, e
0
)e
0
(c
1
> 0),
A
2
= c
2
e
2
+A
2
, e
1
)e
1
+ (c
2
> 0),
and so on. By rearranging the terms we obtain:
P(z)(Az)
1
= (P(A) P(z))(A z)
1
= T
0
(z)A
d1
+ +T
d1
(z)
= T
0
(z)(c
d1
e
d1
+A
d1
, e
d2
)e
d2
+ )
+T
1
(z)(c
d2
e
d2
+ A
d2
, e
d3
)e
d3
+ ) + +T
d1
(z)c
0
e
0
= c
d1
T
0
(z)e
d1
+ (c
d2
T
1
(z) +A
d1
, e
d2
)T
0
(z))e
d2
+
+(c
0
T
d1
(z) +A, e
0
)T
d2
(z) + )e
0
= Q
0
(z)e
d1
+Q
1
(z)e
d2
+ +Q
d1
(z)e
0
,
where
Q
k
(z) = c
d1k
T
k
(z) +O(z
k1
).
Hence Q
k
(z) is a polynomial of degree k with leading coecient c
d1k
> 0, and
(2.2) now follows by inserting the above expression for P(z)(A z)
1
into (2.1)
and using that the e
j
are orthonormal.
c) a). If assertion c) is assumed to be true, then the vector-valued polynomial
V (z) = (Q
0
(z), Q
1
(z), . . . , Q
d1
(z))
satises (2.3). Expanding V (z) along increasing powers of z gives V (z)=
d1
j=0
v
j
z
j
where the v
j
are linearly independent vectors. Then (2.3) shows that A() is a
strictly positive Gram matrix (associated to the vectors v
j
). Hence A() is strictly
positive denite, proving a).
It remains to prove the uniqueness of the decomposition (2.2). For this we
observe that there exists a simple algorithm of nding the polynomials Q
k
. Indeed,
rst observe that the coecient of z
d
in Q(z, z) is P(z). Hence the polynomial
F
d1
(z, z) = [P(z)[
2
Q(z, z) has degree d 1 in each variable. By assumption
the coecient
1
of z
d1
z
d1
in F
d1
is positive, so that:
F
d1
(z, z) =
1
1/2
z
d1
Q
d1
(z) +O(z
d2
, z
d2
).
Therefore the polynomial Q
d1
(z) is determined by F
d1
(z, z).
Proceeding by descending recurrence in k, (k < d 1) we are led to the
polynomial
F
k
(z, z) = F
k+1
(z, z) [Q
k+1
(z)[
2
which has as leading term a positive constant
k
times z
k
z
k
. Then necessarily
F
k
(z, z) =
k
1/2
z
k
Q
k
(z) +O(z
k1
, z
k1
).
Thus Q
k
(z) is determined by F
k
(z, z). And so on until we end by setting F
0
(z, z) =
0
= [Q
0
(z, z)[
2
> 0.
Denition 2.2. A generalized lemniscate is a bounded open set of the complex
plane, given by the equation:
= z C; |(Az)
1
| > 1 (A),
where A L(C
d
) is a linear transformation and is a cyclic vector of it, or
equivalently:
= z C; [P(z)[
2
d1
k=0
[Q
k
(z)[
2
< 0,
with polynomials P, Q
k
subject to the degree conditions: degP = d, degQ
k
=
k, 0 k d 1.
Throughout this article the term generalized lemniscate refers both to the
respective algebraic curve and the domain surrounded by it.
Henceforth we call the pair (A, ) the linear data of a generalized lemniscate.
At this point we can easily make the link to the theory of determinantal curves
due to Moshe Livsic, Kravitsky, Vinnikov and their school. Specically, starting
with a matrix A L(C
d
) as above and a cyclic vector of it, we can produce a
linear pencil of matrices having the determinant equal to the polynomial Q above.
Indeed,
, ) A z
A
z I
, ) (A z)(A
z) A z
0 I
=
det[, ) (A z)(A
z)] =
[det(A z)[
2
det[(A z)
1
, (A z)
1
) I] =
(1)
d
[det(A z)[
2
[1 |(Az)
1
|
2
].
Thus we can state
Proposition 2.3. A generalized lemniscate with linear data (A, ) is given by the
determinantal equation:
= z C; (1)
d
, ) A z
A
z I
> 0. (2.4)
We refer to [21], [35], [36] and the monograph [23] for the theory of deter-
minantal curves. Again, we do not expand here the predictable implications of
formula (2.4). Some of them have been considered by Alex. Shapiro [33].
248 M. Putinar
3. The rational embedding
The realization of a generalized lemniscate as the level set of the resolvent of a
matrix has immediate geometric interpretations and consequences. One of them is
the derivation of a canonical rational embedding in an ane, or projective complex
space. Full proofs and a more detailed analysis of these aspects are contained in
[17] Sections 4 and 5.
Henceforth we denote by

C the Riemann sphere (that is the compactication
of the complex plane by one point at innity). Equivalently, this is the projective
space of dimension one:

C = P
1
(C). The projective space of dimension d will be
denoted by P
d
(C) or simply P
d
.
Let d be a positive integer, d > 1, let A be a linear transformation of C
d
,
and assume that C
d
is a cyclic vector for A. Let us denote by:
R(z) = (A z)
1
, z C (A),
the resolvent of A, localized at the vector .
Lemma 3.1. The map R :

C(A) C
d
is one to one and its range is a smooth
complex curve.
A complete proof is contained in [17] Lemma 4.1. The main idea is to consider
the resolvent equation:
R(z) R(w) = (z w)(A z)
1
(A w)
1
, z, w C (A).
Thus R(z) R(w) ,= 0 for z ,= w. For the point at innity we have R() = 0 ,=
R(z), for z C (A).
Moreover, the same resolvent equation shows that:
R
(z) = (A z)
1
R(z) ,= 0,
and similarly for the point at innity we obtain:
lim
t0
d
dt
R(1/t) = lim
t0
[t
2
(A t
1
)
2
] = ,= 0.
Actually we can pass to projective spaces and complete the above curve as
follows. Let us denote by (z
0
: z
1
) the homogeneous coordinates in P
1
, and by
(u
0
: u
1
: . . . : u
d
) the homogeneous coordinates in P
d
. Let z = z
1
/z
0
in the ane
chart z
0
,= 0 and w = (u
1
/u
0
, . . . , u
d
/u
0
) in the ane chart u
0
,= 0.
Let P(z) = det(Az), so that P(z) is a common denominator in the rational
entries of the map R(z). Let us dene, as in the preceding section, the function:
q(z, A) = P(z)R(z) = (P(z) P(A))R(z),
and remark that q(z, A) is a polynomial in z and A, of the form:
q(z, A) = A
d1
+O(A
d2
).
Actually we need for later use a more precise form of the polynomial q(z, A).
We pause here to derive it by a series of elementary computations.
We have
P(w) P(z)
w z
=
d
k=0
k
w
k
z
k
w z
=
d
k=0
k
k1
j=0
z
kj1
w
j
=
d1
j=0
(
d
k=j+1
k
z
kj1
)w
j
=
T
0
(z)w
d1
+T
1
(z)w
d2
+ +T
d1
(z),
where
d
= 1 and
T
k
(z) =
d
z
k
+
d1
z
k1
+ +
dk+1
z +
dk
.
Note that T
0
(z) = 1.
Therefore we obtain, as in the previous section:
q(z, A) = T
0
(z)A
d1
+T
1
(z)A
d2
+ +T
d1
(z). (3.1)
Since is a cyclic vector for A and dim
k=0
A
k
= d, we infer that q(z, A) ,=
0 for all z C. In addition, for an eigenvalue of A (multiple or not), we have:
(A )q(, A) = P() = 0,
therefore q(, A) is a corresponding (non-trivial) eigenvector.
At this point we can dene the completion of the map R as follows:
R(z
0
: z
1
) =
_
(P(z
1
/z
0
) : q(z
1
/z
0
, A)), z
0
,= 0,
(1 : 0 : . . . : 0), z
0
= 0.
(3.2)
By putting together these computations one obtains the following result.
Lemma 3.2. The map R : P
1
P
d
is a smooth embedding, that is, R is one to
one and its image is a smooth projective curve.
Note that R(P
1
) is a smooth rational curve of degree d in P
d
and the rational
map R has degree d. According to a classical result in algebraic geometry, R(P
1
)
is projectively isomorphic to the rational normal curve of degree d in P
d
obtained
as the range of the Veronese embedding
(z
0
: z
1
) (z
0
d
: z
0
d1
z
1
: . . . : z
1
d
).
See for details [13] pg. 178.
Actually the cyclicity condition on can be dropped, because the resolvent
(A z)
1
has values in the cyclic subspace generated by . Therefore, as a con-
clusion of these computations we can state the following result.
Theorem 3.3. Let A be a linear transformation of C
d
and let be a non-zero vector
of C
d
. Then the map R(z) = (A z)
1
extends to a rational embedding:
R : P
1
P
d
.
The range of R is contained in a linear subspace E of P
d
of dimension equal to
dim
k=0
A
k
and the values R(z) span E as a linear space.
250 M. Putinar
Above, and throughout this note, by embedding we mean a (rational) map
which separates the points and the directions at every point. In particular this
implies that R(P
1
) is a smooth rational curve.
Let us focus now on the geometry of the generalized lemniscate:
= z C; |(Az)
1
| > 1 (A).
The singular points a in the boundary of the bounded domain are given by the
equation R
(a), R(a)) = 0. The proofs above show that |R
(a)| , = 0, and on the

other hand the Hessian H(a) at a of the dening equation |R(z)|
2
= 1 is:
H(a) =
_
R
(a), R
(a)) R
(a), R(a))
R(a), R
(a)) R
(a), R
(a))
_
.
In particular rankH(a) 1, which shows that a is either an isolated point or
a singular double point of .
Our next aim is to study the reection in the boundary of the domain
dened above. More precisely, for a point s P
1
(C) we consider the multi-valued
Schwarz reection in as the set of solutions z = r
1
(s), . . . , r
d
(s) of the equation:
R(s), R(z)) = 1. (3.3)
Proposition 3.4. The multi-valued reection s (r
j
(s))
j=1
d
satises:
a) All r
j
(s) , 1 j d, for s P
1
(C) ;
b) For an appropriate numbering of the r
j
s, r
1
(s) = s and r
j
(s) , 2 j d,
for s .
Proof. Indeed, |R(s)| < 1 whenever s does not belong to . Therefore |R(z)| > 1
for every solution z of the equation (3.3). For s we obtain |R(s)| = 1, hence
one solution of (16), say r
1
, satises r
1
(s) = s and all other solutions z satisfy
necessarily |R(z)| > 1.
A rigidity result of the above rational embedding in the complement of the
sphere, compatible to the reections in the boundaries, is discussed in detail in
[17] Section 5.
This is an appropriate moment to recall the denition of a quadrature domain,
in the sense of Aharonov and Shapiro [2].
Denition 3.5. A bounded planar open set is a quadrature domain if it is
bounded by a real algebraic curve and the function z z extends continuously
from z to a meromorphic function in .
This means, in line with Proposition 3.4, that one determination, say S
1
(z), of
the Schwarz reection satises S
1
(z) = z, z , and it does not have ramication
points inside . Necessarily, it will have d poles a
1
, . . . , a
d
, d 1, there. The
number d is called the order of a quadrature domain and the terminology comes
from the simple observation that, in this case there are d weights c
1
, c
2
, . . . , c
d
with
the property:
_
f(z)dArea(z) = c
1
f(a
1
) + +c
d
f(a
d
), (3.4)
for every integrable analytic function f in , see [2] and [34]. If multiple poles of
S
1
(z) occur, then higher order derivatives of f evaluated there must be consid-
ered, correspondingly. As a matter of fact, the existence of the quadrature formula
(3.4), valid for all integrable analytic functions is equivalent to the above deni-
tion of a quadrature domain. Since their discovery thirty years ago, [2] and [32],
these domains have revealed a series of remarkable features, related to phenomena
of function and potential theory, uid mechanics, moment problems and partial
dierential equations, [34].
The case d = 1 corresponds to a disk. By abuse of language we allow non-
connected sets in the above denition. Thus, a disjoint union of disks is also a
quadrature domain.
Quadrature domains are relevant for this survey because of the following
result (which as a matter of fact was the origin of the whole project).
Theorem 3.6. A quadrature domain of order d is a generalized lemniscate of de-
gree (d, d).
The original proof of this theorem was based on non-trivial results of the
theory of semi-normal operators ([27]). An elementary way to prove it was recently
described in [18].
4. Fredholm eigenvalues
In this section we use a simple geometric feature of the multi-valued Schwarz
reection in the boundary of a generalized lemniscate and prove that a classical
problem in potential theory does not have non-trivial solutions on this class of
domains. It is worth mentioning that the similar picture on classical lemniscates
is quite dierent.
Let be a bounded, simply connected domain of the complex plane and
assume that the boundary of is smooth.
Let u C() be a continuous function. The double layer potential of u,
with respect to is the harmonic function:
D(u)(z) =
1
2
_
u()d arg( z).

An elementary computation shows that:
D(u)(z) =
_
1
2i
_
u()
z
d
_
=
_
u()[
d
2i( z)
].
Whenever z belongs to , respectively to its exterior, we mark the function D
by an index D
i
(z) respectively D
e
(z). It is known that D
i
is a continuous function
on the closure and that D
e
is continuous on the Riemann sphere

C minus .
252 M. Putinar
Moreover, at each boundary point we have representations:
D
i
(u)() =
1
2
u() +
1
2
K
(u)(),
D
e
(u)() =
1
2
u() +
1
2
K
(u)().
Thus, the jumping formula:
D
i
(u)() D
e
(u)() = u(), ,
holds. Remark also that for the constant function u = 1 we have D
i
(1) = 1 and
D
e
(1) = 0, hence K
(1) = 1.
The linear continuous transformation K
: C() C() is the classical

Neumann-Poincare singular integral operator (in two real variables). In general
this operator has better smoothness properties than the Hilbert transform.
Carl Neumanns approach to the Dirichlet problem f = 0 in , f[
= u,
was essentially the following:
Solve the equation 1/2(I + K
)v = u and then set f = D

i
(v) as a double
layer potential of v.
Thus, knowing that the operator I + K
: C() C() is invertible

solves the Dirichlet problem for an arbitrary (originally convex) domain. Later
this idea was applied by Poincare, Fredholm, Carleman to more general classes of
domains, in any number of variables. For (historical) comments we refer to [20].
Particularly relevant for potential and function theory are the solutions of
the Fredholm eigenvalue problem:
K
u = 0.
They correspond to non-trivial solutions of the following matching problem: nd
analytic functions, continuous up to the boundary f(z), z , g(z), z

C
, g() = 0, such that:
f() = g(), .
Non-trivial solutions exist on the lemniscates = z C; [r(z)[ < 1,
where r is a rational function satisfying r() = . Indeed, it is clear that f = r
and g = 1/r solve the above matching problem. For more details see [9].
The following result is reproduced (with its entire proof) from [29].
Theorem 4.1. Let be a connected and simply connected generalized lemniscate
of degree d 2. Then ker K
= 0.
Proof. The proof is an adaptation of an argument, based on analytic continuation,
from [9].
Write
= z C; |(Az)
1
| > 1 (A),
as in the preceding sections, where A is a d d matrix with cyclic vector .
We denote by S(z) the d-valued Schwarz reection, dened by the equation:
(A z)
1
, (A S(z))
1
) = 1, z C. (4.1)
Let a be a non-ramication point for S. Since |(A a)
1
| = 1, then
all local branches S
j
of S satisfy
|(AS
j
(a))
1
| 1, 1 j d.
Denote S
1
(a) = a, so that |(A S
1
(a))
1
| = 1. Since every other branch has
dierent values S
j
(a) ,= S
1
(a) we infer that:
|(AS
j
(a))
1
| > 1, 2 j d.
Therefore S
j
(a) , 2 j d, just as we proved in Proposition 3.4.
Similarly, for every exterior point b of we nd that all branches satisfy
S
j
(b) , 1 j d.
If the function S(z) has no ramication points on

C , then is the com-
plement of a quadrature domain (in the terminology of Aharonov and Shapiro)
and the statement is assured by Theorem 3.19 of [9].
On the contrary, if the algebraic function S(z) has ramication points in
C , then, by repeating the main idea in the proof of the same Theorem 3.19 of
[9], there exists a Jordan arc starting at and returning to a point a , such
that a

C , having the property that the analytic continuation of S
1
(z)
along this arc returns to a at another branch, say S
2
(z), with S
2
(a) .
Assume by contradiction that the matching problem on has a non-trivial
solution: f A(), g A(
C ), g() = 0, f() = g(), . Hence

f(S
1
()) = g(), .
Let us now let the point traverse the curve , and after returning to the
point a, describe once. By analytic continuation along this path, the matching
condition continues to hold, and becomes:
f(S
2
()) = g(), .
But now S
2
(), , remains trapped into a compact subset M .
Thus, putting together the latter identities, we obtain:
max
[f[ = max
[g[ = max
M
[f[.
By the maximum principle, the function f should be a constant. Then g is a
constant, too. But g() = 0, that is f and g are identically zero, a contradiction.
Apparently there are no other examples of domains with non-zero elements

in the kernel of K
other than the level sets of moduli of rational functions.

254 M. Putinar
5. Root separation
The aim of this section is to show that the classical method of separating roots
due to Hermite and later rened by Routh, Hurwitz, Schur, Cohn, Lienard and
Chipard, and many other authors, can be combined with the specic form of the
equation of a generalized lemniscate , to obtain matricial criteria for the root
location of an arbitrary polynomial, with respect to the domain . The technical
tools we invoke are elementary: the expression of the dening equation of the do-
main will be combined with some simple Hilbert space remarks; then Hermites
separation method (with respect to the half-space), or Schurs criterion (with re-
spect to the disk) will be used. Along the same lines, more powerful methods based
on the modern theory of the Bezoutiant can be exploited, [33].
Let
R
= z C; Q(z, z) = 1 |R(z)|
2
< 0,
be a generalized lemniscate associated to a rational function
R : C P
d
, R(z) = (A z)
1
,
as considered in Section 3.
To start we remark that a point belongs to C
R
if and only if, by deni-
tion, |R()| < 1. In its turn, the latter condition is equivalent to [R(), v)[ < 1
for all unit vectors v C
d
, or at least for the vectors of the form v = R()/|R()|,
where is not a pole of at least one, or a common zero of all, entries of R. Note
that in the last formula R() depends rationally on the root . Schurs criterion
of separation with respect to the unit disk can then be applied, see for instance
[3]. As a matter of fact the proof below allows us to consider slightly more general
rational functions.
Theorem 5.1. Let R : C P
d
be a rational function satisfying lim
z
R(z) = 0,
and let C be the set of all poles and common zeroes of R.
Then a monic polynomial f has all its roots
1
, . . . ,
n
in the open set C
R
if and only if, for every C , the polynomial
F
(X) =
n
j=1
(X
1 Q(
j
, )
_
1 Q(, )
) (5.1)
has all its roots in the unit disk.
Proof. Let f be a polynomial with all roots
1
,
2
, . . . ,
n
in the set C . Then
|R(
i
)| < 1 for all i, 1 i n. Consequently, if C we obtain:
[R(
i
), R())[ < |R()|,
which is exactly condition in the statement.
Conversely, if the condition holds for all C , then by reversing the
preceding argument we nd that |R(
i
)| < 1, 1 i n.
Note that the polynomial F
(X) is a symmetric function of the roots

j
, 1
j n, hence its coecients are rational functions of c
1
, . . . , c
n
. Therefore Schurs
criterion will involve only rational combinations of the coecients c
1
, . . . , c
n
.
Specically, if F(z) is a polynomial with complex coecients of degree d, we
dene the associated polynomials:
F(z) = F(z), F
(z) = z
d
F(
1
z
);
then the inertia of the bilinear form:
G
F
(X, Y ) =
F
(X)F
(Y ) F(X)F(Y )
1 XY
,
gives full information about the root location of F with respect to the unit disk.
That is, if G
F
has d
+
positive squares and d
negative squares, then the polynomial

F has exactly d
+
roots in the unit disk, d
roots outside the closed disk, and

d d
+
d
roots lie on the unit circle.

Variations of the above result are readily available: for instance one can re-
place the rational map R(z) by a polynomial map, or instead of F
one can consider

the polynomial involving the squares of the roots of F
, and so on.
If we want to have more information about the root location of the polynomial
f(z) = (z
1
) (z
n
), then the scalar products R(
j
), v), with v a xed
unit vector, can be replaced by an expression such as R(
j
), r(
j
)), where r(z)
is a vector-valued rational function, of norm less than one in a large disk, where
the roots are rst estimated to be. Then, by counting parameters, the degree of r
can be chosen to be dependent on n, the degree of f.
In order to state such a result, we make the following notation: for r : C
C
d
a vector-valued rational map, let
F
r
(X) =
n
j=1
(X R(
j
), r(
j
))). (5.2)
Note that this polynomial in X depends rationally on the entries
j
and is
symmetrical in them. We also denote below by [A[ the cardinality of the set A and
by V (h) the zero set of the function h. We have then
Corollary 5.2. In the conditions of Theorem 2.2, let U = tD be a disk centered at
the origin, that contains all the roots of the polynomial f(z).
Let r : C C
d
be a rational map of degree less than or equal to s on each
entry, satisfying |r(z)| 1, z U, where we assume: (2s + 1)
d
> dn.
Then, with the above notations, we have:
[V (f)
R
[ = max
r
[V (F
r
) D[,
and
[V (f)
R
[ = min
r
[V (F
r
) D[.
256 M. Putinar
Proof. Let d
+
= [V (f)
R
[ and d
= [V (f)
R
[.
Since |r(
j
)| 1 for all j, 1 j n, we have r(
j
), r(
j
))) |R(
j
)|.
Therefore, the polynomial F
r
has at least d
+
zeroes in the unit disk and at most
d
zeroes outside its closure.

To see that these bounds are attained, we remark that, due to the degree
assumption, the map r(z) can be chosen to have prescribed values at every point
i
, 1 j n. Thus we can choose the values r(
j
) so that
R(
j
), r(
j
)) = |R(
j
)|.
Going into another direction, it is easy to establish sucient criteria for the
roots of the polynomial f to be all contained in the exterior of . Let us denote
the dening rational map by R(z) = (R
1
(z), . . . , R
d
(z)).
Corollary 5.3. In the conditions of Theorem 5.1, let a
i
, 1 i d, be positive
numbers satisfying a
2
1
+a
2
2
+ +a
2
d
= 1.
Dene the polynomials:
F
i
(X) =
n
j=1
(X
R
i
(
j
)
a
i
), 1 i d. (5.3)
If the roots of each F
i
, 1 i d, are contained in the unit disk, then the
roots of f are contained in C .
Proof. It is sucient to remark that, under the assumption for the roots of F
j
, for
each xed j, 1 j n, we have
|R(
j
)|
2
i=1
|R
i
(
j
)|
2
<
d
i=1
a
2
i
= 1.
For more details and a couple of examples see [26].
6. The reproducing kernel
Let R(z) = (A z)
1
be a vector-valued rational function attached as before
to a matrix A and its cyclic vector C
d
. The complement of the associated
generalized lemniscate is a subset of the Riemann sphere
G = z

C; |R(z)| < 1.
As we have seen in Section 3, the map R : G B
d
is a smooth rational embedding
of G into the unit ball B
d
of C
d
.
There are several reasons, simplicity and rationality being one of them, to
consider the positive denite kernel
K(z, w) =
1
1 R(z), R(w))
, z, w G.
The theory of such kernels is well understood but we will not invoke deep results
about them; see [1], [4] for further applications of these kernels. We conne our-
selves to reproduce from [26] and [5] an identication, as topological vector spaces,
between the Hardy space of G and the reproducing Hilbert space with kernel K.
This identication holds under a generic smoothness assumption and has some
interesting function theoretic consequences.
Lemma 6.1. Assume the domain G is simply connected and with smooth boundary.
Then there are positive constants C
1
, C
2
such that
C
1
N
j,k=1
k
1 z
j
z
k
j,k=1
k
1 R(z
j
), R(z
k
))
C
2
N
j,k=1
k
1 z
j
z
k
,
for every N N and choice of points z
j
G,
j
C, 1 j N.
The proof, based on the (Fredholm) analysis of the singular integral operator
f
_
G
f(w)dw
1 R(z), R(w))
,
is contained in [5].
By passing for instance to the unit disk via a conformal mapping one deduces
the following result.
Theorem 6.2. Let G = z

C; |R(z)| < 1 be a smooth, simply connected
domain. The Hilbert space with reproducing kernel K(z, w) coincides as a set, but
not isometrically in general, with the Hardy space of G.
This fact was used in [5] to prove that a bounded analytic function along an
analytic arc, smoothly attached to the unit sphere B
d
, admits a bounded extension
to the Schur class of B
d
. This is a slight improvement of the known results of ex-
tending bounded analytic functions, from an attached disk, to a bounded analytic
function dened on the whole ball.
7. Examples
The transition from the dening equation of a generalized lemniscate to its linear
data is not totally trivial. Few concrete examples of this sort are known in full
detail. We list below a couple of such elementary computations (appearing in [17]).
7.1. Domains corresponding to a nilpotent matrix.
A simple and basic example of a generalized lemniscate, which in general is not
a quadrature domain, can be obtained as follows. Let us consider the nilpotent
matrix A and the cyclic vector :
A =
_
_
0 1 0
0 0 1
0 0 0
_
_
, =
_
_
a
b
c
_
_
,
258 M. Putinar
where a, b, c are complex numbers, c ,= 0. A simple computation shows that:
|(A z)
1
|
2
= [
a
z
+
b
z
2
+
c
z
3
[
2
+[
b
z
+
c
z
2
[
2
+[
c
z
[
2
.
Therefore the equation of the associated domain is:
[z[
6
< [az
2
+bz +c[
2
+[bz
2
+cz[
2
+[cz
2
[
2
.
According to Proposition 3.4, the multi-valued reection in the boundary of
this domain maps the exterior completely into its interior.
The rational embedding associated to this example is:
R(1 : z) = (z
3
: az
2
+bz +c : bz
2
+cz : cz
2
).
Similarly one can compute without diculty the corresponding objects asso-
ciated to a nilpotent Jordan block and an arbitrary cyclic vector of it. For instance
the nilpotent nn-Jordan block and the vector = (0, 0, . . . , 0, 1) give precisely
the Veronese embedding:
R(1 : z) = (z
n
: 1 : z : . . . : z
n2
: z
n1
).
7.2. The Lima con
This class of curves exhaust all quadrature domains of order two, with a double
point.
Let us consider the conformal mapping z = w
2
+ bw, where [w[ < 1 and
b 2. Then it is well known that z describes a quadrature domain of order 2,
whose boundary has the equation:
Q(z, z) = [z[
4
(2 +b
2
)[z[
2
b
2
z b
2
z + 1 b
2
= 0.
The Schwarz function of has a double pole at z = 0, whence the associated
2 2-matrix A is nilpotent. Moreover, we know that:
[z[
4
|(Az)
1
|
2
= [z[
2
|(A +z)|
2
= Q(z, z).
Therefore
|(A+z)|
2
= (2 +b
2
)[z[
2
+b
2
z + b
2
z +b
2
1,
or equivalently: ||
2
= 2 +b
2
, A, ) = b
2
and |A|
2
= b
2
1.
Consequently the linear data of the quadrature domain are:
A =
_
0
b
2
1
(b
2
2)
1/2
0 0
_
, =
_
b
2
(b
2
1)
1/2
(
b
2
2
b
2
1
)
1/2
_
.
This shows in particular that the linear data (A, ) of a quadrature domain
of order two, with a double node, is subject to very rigid conditions.
The associated rational embedding can easily be computed from the deni-
tion:
R(1 : z) = (z
2
:
b
2
(b
2
1)
1/2
z +
b
2
1
(b
2
1)
1/2
: (
b
2
2
b
2
1
)
1/2
z).
7.3. Quadrature domains with two distinct nodes
a) Suppose that is a quadrature domain with the quadrature distribution:
u(f) = af(0) +bf(1),
where we choose the constants a, b to be positive numbers. Then P(z) = z(z 1)
and
z(z 1)(Az)
1
= A + z.
Therefore the equation of the boundary of is:
Q(z, z) = [z(z 1)[
2
|A +z|
2
.
Accordingly we obtain:
||
2
=
a +b
, A, ) =
b
.
Let us denote |A|
2
= c. Then the dening polynomial becomes:
Q(z, z) = [z(z 1)[
2
1
(a[z 1[
2
+b([z[
2
1)) c.
The constant c actually depends on a, b, via, for instance, the relation Area()
= a + b, or, whenever a = b, the fact that Q(1/2, 1/2) = 0. The latter are called
special points of a quadrature domain and were studied in [15].
We can choose an orthonormal basis with respect to which we have:
A =
_
0
0 1
_
, =
_

_
.
The matricial elements , , are then subject to the relations:
[[
2
+[[
2
=
1
(a +b), +[[
2
=
1
b, [[
2
[[
2
+[[
2
= c.
An inspection of the arguments shows that the above system of equations
has real solutions , , given by the formulas:
2
=
(c b)
2
(a +b)c b
2
,
2
=
a
2
(a b) +
2
c
,
2
=
(a +b)c b
2
(a b) +
2
c
.
Let us remark that, if a = b > /4, the constant c is eectively computable,
as mentioned earlier, and becomes:
c =
1
16
+
a
2
.
This again illustrates the special nature of the linear data (A, ) of a quad-
rature domain. A simple computation shows that the corresponding canonical
embedding of the domain is:
R(1 : z) = (z(z 1) : (1 z) : z).
260 M. Putinar
We remark that in both of the above examples, the matrix A and the vector
are uniquely determined, as soon as we require that A is upper triangular.
b) In complete analogy, we can treat the case of two nodes with equal weights as
follows.
Assume that the nodes are xed at 1. Hence P(z) = z
2
1. The dening
equation of the quadrature domain of order two with these nodes is:
Q(z, z) = ([z + 1[
2
r
2
)([z 1[
2
r
2
) c,
where r is a positive constant and c 0 is chosen so that either is a union of
two disjoint open disks (in which case c = 0), or Q(0, 0) = 0, as a special point
property [15]. A short computation yields:
Q(z, z) = z
2
z
2
2rzz z
2
z
2
+(r),
where
(r) =
_
(1 r
2
)
2
, r < 1
0, r 1
Exactly as in the preceding two situations, the identication
[P(z)[
2
(1 |(A z)
1
|
2
) = Q(z, z) (7.1)
leads to (for example) the following simple linear data:
=
_
2r
0
_
, A =
_
_
0
2r
1(r)
1(r)
2r
0
_
_
.
7.4. Domains with rotational symmetry
One of the best understood classes of quadrature domains, and by extension,
generalized lemniscates, is that of domains invariant under a nite group of
rotations
= ,
where is a primitive root of order n of unity.
Assume that the equation of has the form:
Q(z, z) = [P(z)[
2
d1
k=0
[Q
k
(z)[
2
,
with degP = d, degQ
k
= k. In view of the uniqueness of the above decomposition
of Q, see Theorem 2.1, we infer:
[P(z)[ = [P(z)[
and
[Q
k
(z)[ = [Q
k
(z)[, 0 k d 1.
Assuming for instance that n = d, that is the order of the domain matches
the symmetry order, we obtain:
Q(z, z) = [z
d
a
d
[
2
d1
k=0
c
k
[z
k
[
2
,
where a C, c
k
> 0, 0 k d 1.
Among these domains only those for which a = 0 are quadrature domains.
Computations along these lines around the uniqueness question, whether the quad-
rature data determine , were carried out in [14], [15] and more recently in [7].
8. Disjoint unions of disks
In view of the preceding discussion, the constructive passage from an algebraic
equation of a generalized lemniscate to its linear data is interesting and possibly
useful for applications. We treat below a simple situation where the linear data
can be recurrently computed.
A generic class of quadrature domains with positive weights in the associated
quadrature formula is oered by the disjoint unions of disks. On uid mechanical
grounds any quadrature domain with positive weights has its roots in a disjoint
union of disks. Although these sets are not connected, their equation is obviously
within reach. We show below, following [18], how to compute inductively the as-
sociated linear data.
Lemma 8.1. Let D
i
= D(a
i
, r
i
), 1 i n, be disjoint disks and let
Q(z, w) =
n
i=1
[(z a
i
)(w a
i
) r
2
i
],
be the polarized equation dening their union. Then the matrix (Q(a
i
, a
j
))
n
i,j=1
is positive semidenite.
Proof. Let =
n
i=1
D(a
i
, r
i
). Since the union is disjoint, is a quadrature domain
with nodes at a
1
, a
2
, . . . , a
n
. Let P(z) be the monic polynomial vanishing at these
points. According to Theorem 3.6, one can write:
Q(z, w) = P(z)P(w)
d1
k=0
[Q
k
(z)[
2
,
with polynomials Q
k
of exact degree k, respectively.
Hence
Q(a
i
, a
j
) =
d1
k=0
Q
k
(a
i
)Q
k
(a
j
),
and this shows that the matrix in the statement is positive semidenite.
262 M. Putinar
A more detailed analysis shows that the matrix (Q(a
i
, a
j
))
n
i,j=1
is actually
strictly positive denite, see [18].
We denote the same disjoint union of disks
n
=
n
i=1
D(a
i
, r
i
), and we
consider the addition of an external disjoint disk; let
n+1
=
n+1
i=1
D(a
i
, r
i
) be the
enlarged set. At each stage we have a nite-dimensional Hilbert space K, a cyclic
vector K and an operator A L(K) which provide the linear data of these
sets. We write accordingly, the equation of a disjoint union of k disks as:
Q
k
(z, w) = P
k
(z)P
k
(w)[1 (A
k
z)
1
k
, (A
k
w)
1
k
)],
where A
k
L(K
k
) has cyclic vector
k
, dimK
k
= k, k = n, n + 1, and the
polynomial P
k
has degree k and annihilates A
k
.
Our aim is to understand the structure of the matrix A
n+1
and its cyclic vec-
tor
n+1
as functions of the previous data (A
n
,
n
) and the new disk D(a
n+1
, r
n+1
).
Henceforth we assume that the closed disks D(a
i
, r
i
) are still disjoint. In order to
simplify notation we suppress for a while the index n+1, e.g. a = a
n+1
, r = r
n+1
etc. The following computations are based on standard realization techniques in
linear systems theory.
Due to the multiplicativity of the dening equation for disjoint domains we
nd:
[1 (A
n
z)
1
n
, (A
n
w)
1
n
)][1
r
2
(z a)(w a)
]
= 1 (A z)
1
, (A w)
1
).
Equivalently,
(A
n
z)
1
n
, (A
n
w)
1
n
) +
r
2
(z a)(w a)
=
r
z a
(A
n
z)
1
n
,
r
w a
(A
n
w)
1
n
) +(A z)
1
, (A w)
1
).
Thus, for each z avoiding the poles, the norm of the vector
f(z) =
_
(A
n
z)
1
n
r
za
_
K
n
C
equals that of the vector
g(z) =
_
r
za
(A
n
z)
1
n
(A z)
1
_
K
n
K.
And moreover, the same is true for any linear combination
|
1
f(z
1
) + +
r
f(z
r
)| = |
1
g(z
1
) + +
r
g(z
r
)|.
Because the span of f(z), z C, is the whole space K
n
C, there exists a unique
isometric linear operator V : K
n
C K
n
K mapping f(z) to g(z). We write,
corresponding to the two direct sum decompositions
V =
_
B
C
_
,
where B : K
n
K
n
, K
n
, C : K
n
K
n+1
, K. Since V f(z) = g(z)
for all z, we nd by coecient identication:
B = r(A
n
a)
1
, = (A
n
a)
1
n
.
The isometry condition V
V = I written at the level of the above 2 2

matrix yields the identities:
_
_
_
r
2
(A
n
a)
1
(A
n
a)
1
+C
C = I,
r(A
n
a)
1
(A
n
a)
1
n
+C
= 0,
|(A
n
a)
1
n
|
2
+||
2
= 1.
(8.1)
In particular we deduce that (A
n
a)
1
(A
n
a)
1
r
2
and since this
operator inequality is valid for every radius which makes the disks disjoint, we can
enlarge slightly r and still have the same inequality. Thus, the defect operator
= [I r
2
(A
n
a)
1
(A
n
a)
1
]
1/2
: K
n
K
n
(8.2)
is strictly positive.
The identity C
C =
2
shows that the polar decomposition of the matrix
C = U denes without ambiguity an isometric operator U : K
n
K. Since
dimK = dimK
n
+1 we will identify K = K
n
C, so that the map U becomes the
natural embedding of K
n
into the rst factor. Thus the second line of the isometry
V becomes
(C ) =
_
d
0
_
: K
n
C K
n
C = K,
where d K
n
, C. We still have the freedom of a rotation of the last factor,
and can assume 0. One more time, equations (8.1) imply
_
d =
1
r
(
n

1
n
),
= [1 |(A
n
a)
1
n
|
2
|d|
2
]
1/2
.
(8.3)
From relation V f(z) = g(z) we deduce:
_
d
0
__
(A
n
z)
1
n
r
za
_
= (A z)
1
.
This shows that > 0 because the operator A has the value a in its spectrum.
At this point straightforward matrix computations lead to the following exact
description of the couple (A, ) = (A
n+1
,
n+1
) (by restoring the indices):
A
n+1
=
_
A
n
1
(A
n
a
n+1
)
1
d
0 a
n+1
_
, =
_

1
n
r
n+1
_
. (8.4)
It is sucient to verify these formulas, that is:
_
(A
n
z)
1
1
(A
n
a)
1
d
0 a z
__
d
0
__
(A
n
z)
1
n
r
za
_
=
_

1
n
r
_
.
264 M. Putinar
And this is done by direct multiplication:
n
+ (A
n
z)
1
rd
z a
(A
n
a)
1
rd
z a
=
1
n
,
which is equivalent to the known relation dr =
n
n
.
Summing up, we can formulate the transition laws of the linear data of a
disjoint union of disks.
Proposition 8.2. Let D(a
i
, r
i
), 1 i n + 1, be a disjoint family of closed disks,
and let
k
=
k
i=1
D(a
i
, r
i
), 1 k n + 1.
The linear data (A
k
,
k
) of the quadrature domain
k
can be inductively ob-
tained by the formula (8.4), with the aid of the denitions (8.2), (8.3).
Remark that letting r = r
n+1
0 we obtain I and d 0, which is
consistent with the fact that
n+1
, in measure, in case such a limit domain
is given.
Acknowledgment
I would like to thank Daniel Alpay and Victor Vinnikov for sharing the joy of the
generalized lemniscate game. I am much indebted to the anonymous referee for a
careful and critical reading of the manuscript.
References
[1] Agler, J., McCarthy, J., Pick Interpolation and Hilbert Function Spaces, Amer. Math.
Soc., Providence, R.I., 2002.
[2] Aharonov, D., Shapiro, H.S., Domains on which analytic functions satisfy quadrature
identities, J. Analyse Math. 30(1976), 3973.
[3] Akhiezer, N.I., On a minimum problem in function theory and the number of roots
of an algebraic equation inside the unit disc (in Russian), Izv. Akad. Nauk SSSR
9(1930), 11691189.
[4] Alpay, D., Dym, H., On a new class of realization formulas and their applications,
Linear Alg. Appl. 241243(1996), 384.
[5] Aplay, D., Putinar, M., Vinnikov, V., A Hilbert space approach to bounded analytic
extension in the ball, Comm. Pure Appl. Analysis 2(2003), 139145.
[6] Carey, R.W. and Pincus, J.D., An exponential formula for determining functions,
Indiana Univ. Math.J. 23 (1974), 10311042.
[7] Crowdy, D., Constructing multiply-connected quadrature domains I: algebraic curves,
preprint 2002.
[8] Ph.J. Davis, The Schwarz function and its applications, Carus Math. Mono. vol. 17,
Math. Assoc. Amer., 1974.
[9] Ebenfelt, P., Khavinson, D., Shapiro, H.S., An inverse problem for the double layer
potential, Comput. Methods. Funct. Theory 1 (2001), 387401.
[10] Eremenko, A., Hayman, W., On the length of lemniscates, Paul Erdos and his math-
ematics, I (Budapest, 1999), Bolyai Soc. Math. Stud. 11, Janos Bolyai Math. Soc.,
Budapest, 2002, pp. 241242.
[11] Foias, C. and Frazho, A.E., The commutant lifting approach to interpolation prob-
lems, Birkhauser Verlag, Basel, 1990.
[12] Golub, G., Gustafsson, B., Milanfar, P., Putinar, M. and Varah, J., Shape recon-
struction from moments: theory, algorithms, and applications, Signal Processing and
Image Engineering, SPIE Proceedings vol. 4116(2000), Advanced Signal Process-
ing, Algorithms, Architecture, and Implementations X (Franklin T. Luk, ed.), pp.
406416.
[13] Griths, P., Harris, J., Principles of Algebraic Geometry, J. Wiley Sons, New York,
1994.
[14] Gustafsson, B., Quadrature identities and the Schottky double, Acta Appl. Math. 1
(1983), 209240.
[15] Gustafsson, B., Singular and special points on quadrature domains from an algebraic
point of view, J. dAnalyse Math. 51(1988), 91117.
[16] Gustafsson, B. and Putinar, M., An exponential transform and regularity of free
boundaries in two dimensions, Ann. Sc. Norm. Sup. Pisa, 26 (1998), 507543.
[17] Gustafsson, B. and Putinar, M., Linear analysis of quadrature domains. II, Israel J.
Math. 119(2000), 187216.
[18] Gustafsson, B. and Putinar, M., Linear analysis of quadrature domains. IV, Quad-
rature Domains and Applications, The Harold S. Shapiro Anniversary Volume,
(P. Ebenfeldt et al. eds.), Operator Theory: Advances Appl. vol. 156, Birkhauser,
Basel, 2004, 147168.
[19] Horn, R.A. and Johnson, C.R., Matrix Analysis, Cambridge Univ. Press, Cambridge,
1985.
[20] Kral, J., Integral Operators in Potential Theory, Lect. Notes Math. vol. 823, Springer,
Berlin, 1980.
[21] Kravitsky, N., Rational operator functions and Bezoutian operator vessels, Integral
Eq. Operator Theory 26(1996), 6080.
[22] Kuznetsova, O.S., Tkachev, V.G., Length functions of lemniscates, Manuscripta
Math. 112 (2003), 519538.
[23] Livsic, M.S., Kravitsky, N., Markus, A.S., Vinnikov, V., Theory of commuting non-
selfadjoint operators, Kluwer Acad. Publ. Group, Dordrecht, 1995.
[24] Martin, M. and Putinar, M., Lectures on Hyponormal Operators, Birkhauser, Basel,
1989.
[25] Pincus, J.D. and Rovnyak, J., A representation for determining functions, Proc.
Amer. Math. Soc. 22(1969), 498502.
[26] Putinar, G., Putinar, M., Root separation on generalized lemniscates, Hokkaido
Math. J. 30(2001), 705716.
[27] Putinar, M., Linear analysis of quadrature domains, Ark. Mat. 33 (1995), 357376.
[28] Putinar, M., A renormalized Riesz potential and applications, in vol. Advances in
Constructive Approximation: Vanderbilt 2003, (M. Neamtu and E.Sa, eds.), Nash-
boro Press, Brentwood, TN, pp. 433466.
[29] Putinar, M., Sandberg, S., A skew normal dilation on the numerical range, Math.
Ann., to appear.
266 M. Putinar
[30] Ransford, T., Potential Theory in the Complex Domain, Cambridge Univ. Press,
Cambridge, 1995.
[31] Riesz, F. and Sz.-Nagy, B., Functional analysis, Dover Publ., New York, 1990.
[32] Sakai, M., Quadrature Domains, Lect. Notes Math. 934, Springer-Verlag, Berlin-
Heidelberg 1982.
[33] Shapiro, Alex., personal communication.
[34] Shapiro, H.S., The Schwarz function and its generalization to higher dimensions,
Univ. of Arkansas Lect. Notes Math. Vol. 9, Wiley, New York, 1992.
[35] Vinnikov, V., Complete description of determinantal representations of smooth irre-
ducible curves, Linear Alg. Appl. 125 (1989), 103140.
[36] Vinnikov, V., Elementary transformations of determinantal representations of alge-
braic curves, Linear Alg. Appl. 135 (1990), 118.
Mihai Putinar
Mathematics Department
University of California
Santa Barbara, CA 93106,USA
e-mail: mputinar@math.ucsb.edu
Operator Theory:
One-Sided Tangential Interpolation
for Hilbert-Schmidt Operator Functions
with Symmetries on the Bidisk
M.C.B. Reurings and L. Rodman
Abstract. One-sided tangential interpolation problems for functions with sym-
metries having values in the set of HilbertSchmidt operators and dened on
the bidisk are studied. General solutions are described as well as solutions
with the minimal scalar and operator-valued norms. Two types of symme-
tries are considered: (a) componentwise symmetries that operate separately
on each component of a general point in the bidisk; (b) interchange symmetry
that interchanges the two components of a general point in the bidisk. Appli-
cations are made to multipoint tangential interpolation problems of special
form.
Mathematics Subject Classication (2000). 47A56, 47A57.
Keywords. Tangential interpolation, symmetries, HilbertSchmidt operators.
1. Introduction
The letters H and (, with or without subscripts, designate complex Hilbert spaces;
all Hilbert spaces in the present paper are assumed to be separable, and all oper-
ators are assumed to be bounded and linear. The inner product and the norm in
H are denoted by , )
1
and | |
1
, respectively. If X, Y are selfadjoint operators
on a Hilbert space, we write X Y or Y X to mean that X Y is positive
semidenite.
Denote by HS
1
the (separable) Hilbert space of Hilbert-Schmidt opera-
tors acting from ( into H, and let HS
1
(D
d
) be the Hardy space of HS
1
-
valued functions which are analytic in the polydisk D
d
, dened by
D
d
= (z
1
, z
2
, . . . , z
d
) C
d
: [z
i
[ < 1, i = 1, 2, . . . , d.
Research of the rst author was supported by the Netherlands Organization for Scientic Re-
search (NWO). Research of the second author was supported in part by an NSF grant.
268 M.C.B. Reurings and L. Rodman
In this paper we focus on the bidisk D
2
. The space HS
1
(D
2
) is a separable
Hilbert space with the inner product
G, H)
1S
GH
(D
2
)
=
1
(2)
2
_
2
0
_
2
0
Trace(H(e
it
1
, e
it
2
)
G(e
it
1
, e
it
2
)) dt
1
dt
2
.
The space HS
1
(D
2
) is also a Hilbert module with respect to the trace class
operator-valued inner product
[G, H]
1S
GH
(D
2
)
=
1
(2)
2
_
2
0
_
2
0
H(e
it
1
, e
it
2
)
G(e
it
1
, e
it
2
) dt
1
dt
2
.
The Hilbert module property means, in particular, that
[GA, HB]
1S
GH
(D
2
)
= B
[G, H]
1S
GH
(D
2
)
A
for any choice of operators A, B : ( (. Note that every H HS
1
(D
2
) can
be written as a power series
H(z
1
, z
2
) =
j
1
,j
2
=0
H
j
1
,j
2
z
j
1
1
z
j
2
2
, (z
1
, z
2
) D
2
, (1.1)
where H
j
1
,j
2
HS
1
are such that
|H|
1S
GH
(D
2
)
:=
_
_
j
1
,j
2
=0
|H
j
1
,j
2
|
2
1S
GH
_
_
1/2
< .
In this paper we study certain tangential interpolation problems with the
solution set restricted to some subspace of HS
1
(D
2
). The subspace is specied
by a given symmetry. The data set of these interpolation problems consists of a
quintuple of operators
= A
1
, A
2
, B
1
, B
2
, B
, (1.2)
where A
i
: H
i
H
i
(i = 1, 2), B
1
: H H
1
, B
2
: H
1
H
2
, B
: ( H
2
are
operators such that
(A
1
) (A
2
) D. (1.3)
(Here and elsewhere (A) stands for the spectrum of an operator A.) It will
be assumed throughout the present paper that the spaces H
1
and H
2
are nite-
dimensional. (Extending the results, if possible, to innite-dimensional H
1
and
H
2
seems to be a challenging problem, and is beyond the scope of this paper.)
The interpolation problems in question are formulated as follows:
Interpolation with Symmetries on Bidisk 269
Problem 1.1. Tangential Interpolation TI
Given a data set (1.2), nd all functions H HS
1
(D
2
) satisfying the interpo-
lation condition
1
(2i)
2
_
]]=1
(I
1
2
A
2
)
1
B
2

_
_
]]=1
(I
1
1
A
1
)
1
B
1
H(, ) d
_
d = B
.
(1.4)
Problem 1.2. TI with Operator Norm constraint ONorm
Given a data set (1.2) and an operator : ( (, where 0, nd all functions
H HS
1
(D
2
) satisfying (1.4) and the operator-valued norm constraint
[H, H]
1S
GH
(D
2
)
. (1.5)
Problem 1.3. TI with Scalar Norm constraint SNorm
Given a data set (1.2) and a positive number , nd all functions H HS
1
(D
2
)
satisfying the interpolation condition (1.4) and the norm constraint
|H|
1S
GH
(D
2
)
. (1.6)
Substituting (1.1) into the left-hand side of (1.4) leads to the explicit expres-
sion of this integral in terms of the Taylor coecients of the function H:
j
1
,j
2
=0
A
j
2
2
B
2
A
j
1
1
B
1
H
j
1
,j
2
= B
. (1.7)
It follows from (1.7) that a necessary condition for the problem TI to be solvable is
Ran B
spanRan A
j
2
2
B
2
A
j
1
1
B
1
; j
1
, j
2
= 0, 1, . . . . (1.8)
Here and elsewhere in the paper we denote by Ran X the range of an operator X.
It turns out that (1.8) is also sucient for solvability of TI (see [9]).
Problems of the type TI, ONorm, and SNorm are called one-sided (left) tan-
gential problems. An analogous one-sided (right) tangential problem can be formu-
lated by having in (1.4) the unknown function H on the left side of the integrand
rather than on the right side of the integrand as it stands in (1.4). The results
for the right problem are completely analogous to the results for the left problem,
and can be obtained from each other essentially by passing to adjoint operators.
For this reason, right problems will not be considered here. Two-sided tangential
interpolation problems are more involved; they include (1.4), the corresponding
condition for the right problem, and a third two sided condition (Nudelman for-
mulation [16]); see, for example, [11], where the two-sided problem is studied in
depth for rational matrix-valued functions. In the context of the present paper,
the two-sided problems seem to be intractable, and therefore are not studied here.
Tangential interpolation problems on the unit disk, with or without norm
constraints, have been studied extensively in the literature. We only mention here
the books [11], [14] and [15]. It is of special interest to study Hilbert spaces of
analytic operator functions as sets in which the interpolants are sought. This is
done in [2], with the standard Hardy space H
2
of the unit disk as the Hilbert space
of matrix functions.
In recent years, the research on tangential interpolation was extended to
functions of several variables, using a variety of techniques, and producing a wealth
of results. It would take us too far a eld to quote many works in this area. We
only mention that the study of tangential interpolation in Hardy spaces of operator
functions of several variables was started in [1] and continued in [9]. The rst paper
discusses the one-sided tangential interpolation problem for matrix-valued Hardy
functions on the bidisk, whereas the subject of the second paper is one-sided
interpolation for Hilbert-Schmidt operator-valued functions on the polydisk. The
two-sided generalization of the latter can be found in [8].
In [9] the problems TI, ONorm and SNorm are solved in case H
1
= H
2
and
B
2
= I
1
1
. In [8] those problems are solved in their general form, because they are
particular cases of the two-sided tangential interpolation problems studied in that
paper.
In the present paper we will study Problems TI, ONorm, SNorm under addi-
tional symmetry conditions on the solutions. Tangential interpolation problems for
symmetric functions of one variable has already been studied, motivated by applied
problems in which symmetry of interpolant functions is inherent in the problem (e.
g., periodic problems [6]). See [12] and [13] for rational matrix functions and [3],
[4], [5], [7] and [10] for matrix-valued H
2
-functions of one variable, and [6] for ap-
plications in structured positive extension problems. Tangential interpolation with
symmetries in one variable was applied in [7] to two point interpolation problems;
see Section 9 for more information on multipoint problems.
Two type of symmetries are considered in the present paper for Problems TI,
ONorm, SNorm: (a) componentwise symmetries that operate separately on each
component of a general point in D
2
, and when restricted to each component, the
symmetry is represented by a conformal involutive mapping of the unit disk; (b)
interchange symmetry that interchanges the two components of a general point in
D
2
. It is a challenging open problem to extend if possible the results obtained here
to more general classes of symmetries of the bidisk.
Besides the introduction, the paper consists of eight sections. We formulate
the interpolation problems which are determined by componentwise symmetries
and prove some preliminary results on these problems in the next section. In Sec-
tion 3 we recall results from [8] and [9] that will be used later on. The main result
on solution of the interpolation problems with componentwise symmetries and its
proof are given in Section 4. A more transparent formulation of the main result, as
well as the treatment of interpolation problems with componentwise symmetries
under additional norm constraints, is possible provided the componentwise sym-
metries are specialized so that one of them is the map z z. This is done in
Sections 5 and 6. The interpolation problems with the interchange symmetry, with
or without additional norm constraints, are formulated in Section 7, and their so-
lutions are described in Section 8. Finally, the last section is devoted to multipoint
applications of the main results.
2. Interpolation with componentwise symmetries
In this section we formulate the interpolation problems TI, ONorm, and SNorm
in a subspace of HS
1
(D
2
) which is dened by componentwise symmetries.
We denote by HS
1
(J
1
, J
2
, s
1
, s
2
) the class of all H HS
1
(D
2
) which
satisfy the symmetry relation
J
1
H(s
1
(z
1
), s
2
(z
2
))J
2
= H(z
1
, z
2
), (z
1
, z
2
) D
2
, (2.1)
where J
1
: H H and J
2
: ( ( are two xed signature operators, i.e.,
J
i
= J
1
i
= J
i
, i = 1, 2,
and where s
1
and s
2
are two conformal mappings from D onto D of the forms
s
j
(z) =

j
z
1 z
j
,
j
D xed (j = 1, 2). (2.2)
It is easily seen that except for the case s
j
(z) = z(j = 1, 2), formula (2.2) presents
the general form of conformal involutive mappings from D onto itself. Also,
HS
1
(J
1
, J
2
, s
1
, s
2
) is clearly a (closed) subspace of HS
1
(D
2
). We consider
the following interpolation problems in this subspace.
Problem 2.1. TI with Componentwise Symmetries CSym
Let be given a data set (1.2) satisfying (1.8), signature operators J
1
and J
2
and
two mappings s
1
and s
2
of the form (2.2). Find all functions
H HS
1
(J
1
, J
2
, s
1
, s
2
)
satisfying the interpolation condition (1.4).
Problem 2.2. CSym with Operator Norm constraint ONorm-CSym
Let be given a data set (1.2), J
1
, J
2
, s
1
and s
2
as in Problem CSym. Let also be
given an operator
: ( (, 0.
Find all H HS
1
(J
1
, J
2
, s
1
, s
2
) satisfying the interpolation condition (1.4) and
the operator-valued norm constraint (1.5).
Problem 2.3. CSym with Scalar Norm constraint SNorm-CSym
Let be given a data set (1.2), J
1
, J
2
, s
1
and s
2
as in Problem CSym. Let also
be given a positive number . Find all H HS
1
(J
1
, J
2
, s
1
, s
2
) satisfying the
interpolation condition (1.4) and the norm constraint (1.6).
As part of the solution of problems CSym, ONorm-CSym, SNorm-CSym,
criteria for existence of solutions will be given.
Note that if H HS
1
(J
1
, J
2
, s
1
, s
2
) satises (1.4), then it also satises
1
(2i)
2
_
]]=1
(I
1
2
A
2
)
1
B
2
_
_
]]=1
(I
1
1
A
1
)
1
B
1
J
1
H(s
1
(), s
2
())J
2
d
_
d = B
.
(2.3)
We will write this equation in the form of a left-sided interpolation condition. For
this purpose, we need the following easily proved lemma.
Lemma 2.4. Let A be an operator on a Hilbert space H such that (A) D. Let
T HS
1
(D) and let s be a conformal involution mapping D onto itself. Then
_
]z]=1
(zI
1
A)
1
T(s(z)) dz =
_
]z]=1
(zI
1
s(A))
1
T(z) dz.
Hence, applying this lemma twice to (2.3) and then multiplying both sides
of the equality on the right with J
2
, gives us that H also satises
1
(2i)
2
_
]]=1
(I
1
2
s
2
(A
2
))
1
B
2
_
_
]]=1
(I
1
1
s
1
(A
1
))
1
B
1
J
1
H(, ) d
_
d = B
J
2
,
(2.4)
where in accordance to (2.2),
s
j
(A
j
) = (
j
I
1
j
A
j
)(I
1
j

j
A
j
)
1
(j = 1, 2).
The interpolation conditions (1.4) and (2.4) can be written together as
1
(2i)
2
_
]]=1
(I
1
2
1
2
A
2
)
1
B
2
_
_
]]=1
(I
1
1
1
1
A
1
)
1
B
1
H(, ) d
_
d = B
,
(2.5)
where
A
1
=
_
A
1
0
0 s
1
(A
1
)
_
, A
2
=
_
A
2
0
0 s
2
(A
2
)
_
,
B
1
=
_
B
1
B
1
J
1
_
, B
2
=
_
B
2
0
0 B
2
_
, B
=
_
B
J
2
_
.
(2.6)
Analogously to Problem TI a necessary and sucient condition for the exis-
tence of solutions H HS
1
(D
2
) of (2.5) is
Ran B
spanRan A
j
2
2
B
2
A
j
1
1
B
1
; j
1
, j
2
= 0, 1, . . . . (2.7)
If a data set satises this inclusion, together with (1.3), then we call C-
admissible. Note that if a data set is C-admissible, then it also satises (1.8).
We will assume throughout the rst part of the paper that is C-admissible. It
will be proved that the C-admissibility of is also sucient for the solvability of
CSym, i.e., for the existence of solutions H HS
1
(J
1
, J
2
, s
1
, s
2
) of (2.5). As a
rst result we obtain:
Proposition 2.5. Let be a C-admissible data set. Then the following equivalences
hold true.
(a) H HS
1
(J
1
, J
2
, s
1
, s
2
) solves CSym if and only if H satises (2.5).
(b) H HS
1
(J
1
, J
2
, s
1
, s
2
) solves ONorm-CSym if and only if H satises
(2.5) and the norm constraint (1.5).
(c) H HS
1
(J
1
, J
2
, s
1
, s
2
) solves SNorm-CSym if and only if H satises
(2.5) and the norm constraint (1.6).
We solve Problems CSym, ONorm-CSym and SNorm-CSym in two steps. The
rst step is to solve the problems without the symmetry condition, which amounts
to Problems TI, ONorm and SNorm with the data = A
1
, A
2
, B
1
, B
2
, B
. This
step was already done in [8] and [9]; in the next section we will recall results from
these papers adapted to the present setting. Then we will show how to select among
all the solutions those which also satisfy the symmetry constraint (2.1), i.e., those
H which belongs to the subspace HS
1
(J
1
, J
2
, s
1
, s
2
). In view of Proposition 2.5,
solution of Problems CSym, ONorm-CSym and SNorm-CSym will be obtained.
3. Tangential interpolation without symmetries
In this section we will describe, using the results of [8] and [9], all functions
H HS
1
(D
2
) which satisfy the left-sided interpolation condition (2.5) with
or without one of the norm constraints (1.5) or (1.6). Note that we will apply the
results of [8] and [9] not to the original data set, but rather to the set
= A
1
, A
2
, B
1
, B
2
, B
,
where A
i
, B
i
and B
are dened as in (2.6). Hence, A

i
maps H
i
H
i
into H
i
H
i
(for i = 1, 2), and
B
1
: H H
1
H
1
, B
2
: H
1
H
1
H
2
H
2
and B
: ( H
2
H
2
.
Because is assumed to be C-admissible, this exactly has the properties it must
have in order to be able to apply the results in [8] and [9], namely (A
1
)(A
2
)
D and (2.7).
Before we can formulate the results of [8] and [9], we have to introduce some
notations. First, let
2
(H) be the space dened by
2
(H) =
_
_
_
h
j
j=0
: h
j
H and
j=0
|h
j
|
2
1
<
_
_
_
and let E(z) :
2
(H) H be the operator dened by
E(z) = ( I
1
zI
1
z
2
I
1
). (3.1)
Note that for every x
2
(H), we have
|E()x|
1(D)
= |x|
2
(1)
, (3.2)
where H(D) is the Hardy space of all H-valued functions f(z) =
j=0
f
j
z
j
, f
j
H,
which are analytic on the unit disk and satisfy
j=1
|f
j
|
2
1
< . The space H(D)
is equipped with the inner product
f, g)
1(D)
=
1
2
_
2
0
f(e
it
), g(e
it
))
1
dt.
Making use of (1.1) we represent the function H as
H(z
1
, z
2
) =
j=0
z
j
1
F
j
(z
2
) = E(z
1
)F(z
2
),
where
F
j
(z) =
=0
H
j
z
and F(z) =
_
_
_
F
0
(z)
F
1
(z)
.
.
.
_
_
_
=
=0
_
_
_
H
0
H
1
.
.
.
_
_
_
z
.
For a given j, the function F
j
belongs to HS
1
(D) whereas the function F clearly
belongs to HS
2
(1)
(D). The following lemma holds true, see Lemma 2.1 in [9].
Lemma 3.1. The operator F(z) E(z
1
)F(z) is a unitary operator from
HS
2
(1)
(D) onto HS
1
(D
2
),
with respect to the operator-valued inner products
[, ]
1S
G
2
(H)
(D)
and [, ]
1S
GH
(D
2
)
.
More precisely, every function H HS
1
(D
2
) admits a representation of the
form
H(z
1
, z
2
) = E(z
1
)F(z
2
), (3.3)
where F is a uniquely dened function in HS
2
(1)
(D) such that
[H, H]
1S
GH
(D
2
)
= [F, F]
1S
G
2
(H)
(D)
. (3.4)
Note that from (3.4) it follows that also
|H|
1S
GH
(D
2
)
= |F|
1S
G
2
(H)
(D)
.
Now introduce the operator
B
+
=
1
2i
B
2
_
]z]=1
(zI
1
1
1
1
A
1
)
1
B
1
E(z) dz :
2
(H) H
2
, (3.5)
and let P : H
2
H
2
H
2
H
2
be the unique positive semidenite solution of the
Stein equation
P A
2
PA
2
= B
+
B
+
, (3.6)
so P is given by
P =
k=0
A
k
2
B
+
B
+
A
k
2
(3.7)
(the convergence here is in the operator norm). We will denote by P
[1]
the Moore
Penrose pseudoinverse of P, i.e., the operator uniquely dened by the relations
P
[1]
P
RanP
= P
RanP
P
[1]
= P
[1]
, P
[1]
P = PP
[1]
= P
RanP
,
where P
RanP
is the orthogonal projection on the range of P. (The existence of P
[1]
is guaranteed by the nite-dimensionality of H
2
.) Further, let be an arbitrary
point on the unit circle and dene as the following function on D:
(z, ) = I
2
(1)
+(z )B
+
(I
1
2
1
2
zA
2
)
1
P
[1]
(I
1
2
1
2
A
2
)
1
B
+
. (3.8)
Thus, takes values in the algebra of operators acting on
2
(H).
Theorem 3.2. Let and be arbitrary points on the unit circle. Then the function
dened by (3.8) satises
I
2
(1)
(z, )(, )
= (1 z)B
+
(I
1
2
1
2
zA
2
)
1
P
[1]
(I
1
2
1
2
A
2
)
1
B
+
(3.9)
and
(z, )(, ) = (z, ). (3.10)
This theorem is well known, see for example [3], where it is proved in the
matrix case (i.e., when all Hilbert spaces involved are assumed to be nite-dimen-
sional); in the present case the proof is the same.
Now we can give a description of all solutions of (2.5), with and without the
norm constraints (1.5) and (1.6).
Theorem 3.3. Let = A
1
, A
2
, B
1
, B
2
, B
be a data set such that

(A
1
) (A
2
) D
and (2.7) is satised. Then the following statements hold true.
(a) There exists an H HS
1
(D
2
) which satises (2.5). Moreover, all such H
are described by the formula
H(z
1
, z
2
) = H
min
(z
1
, z
2
) +E(z
1
)(z
2
, )f(z
2
), (3.11)
where is a xed point on the unit circle and H
min
is the minimal norm solution,
with respect to the norm | |
1S
GH
(D
2
)
, given by
H
min
(z
1
, z
2
) = B
1
(I
1
1
1
1
z
1
A
1
)
1
B
2
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
B
, (3.12)
and f is a free parameter in HS
2
(1)
(D). The sum in (3.11) is orthogonal with
respect to the operator-valued inner product [, ]
1S
GH
(D
2
)
, as well as with respect
to the inner product , )
1S
GH
(D
2
)
. The norm of H
min
is equal to
|H
min
|
2
1S
GH
(D
2
)
= Trace(B
P
[1]
B
).
(b) There exists a solution H HS
1
(D
2
) of (2.5) satisfying (1.5) if and only if
B
P
[1]
B
0.
Moreover, all such H are described by formula (3.11), where the parameter f
HS
2
(1)
(D) satises the operator-valued norm constraint
[f, f]
1S
G
2
(H)
(D)
:=
1
2
_
2
0
f(e
it
)
f(e
it
) dt B
P
[1]
B
,
but otherwise is free.
(c) There exists a solution H HS
1
(D
2
) of (2.5) satisfying (1.6) if and only if
2
Trace(B
P
[1]
B
).
Moreover, all such H are described by the formula (3.11), where the parameter
f HS
2
(1)
(D) satises the norm constraint
|f|
2
1S
G
2
(H)
(D)
:=
1
2
_
2
0
Trace(f(e
it
)
f(e
it
) dt
2
Trace(B
P
[1]
B
),
For the proof of this theorem we refer to [9].
Since the sum in (3.11) is orthogonal and is inner (which follows from
(3.9)), any solution H of (2.5) satises
|H|
2
1S
GH
(D
2
)
= |H
min
|
2
1S
GH
(D
2
)
+|f|
2
1S
G
2
(H)
(D)
.
So H
min
is the only solution with |H
min
|
2
1S
GH
(D
2
)
= Trace(B
P
[1]
B
).
4. Solution of CSym
We will start this section with a lemma which states that H belongs to the subspace
HS
1
(J
1
, J
2
, s
1
, s
2
) if and only if the function F in the representation (3.3) of H
satises a certain symmetry condition. If X and Y are operators, and s : D D a
function, we denote by HS
1
(X, Y, s) the subspace of HS
1
(D) that consists
of all functions f HS
1
(D) satisfying Xf(s(z))Y = f(z) for every z D.
Lemma 4.1. The following are equivalent:
(1) The function H belongs to HS
1
(J
1
, J
2
, s
1
, s
2
).
(2) H admits a representation (3.3) for some function
F HS
2
(1)
(J
1
, J
2
, s
2
),
where J
1
:
2
(H)
2
(H) is the operator dened by
J
1
=
1
2
_
2
0
E(e
it
)
J
1
E(s
1
(e
it
)) dt. (4.1)
The operator J
1
satises
J
1
E(s
1
(z)) = E(z)J
1
, (4.2)
J
1
= J
1
1
and J
1
J
1
=
1
1 [
1
[
2
(I
2
(1)
1
T
)(I
2
(1)
1
T), (4.3)
where T L(
2
(H)) is the backward block shift operator represented by the
operator matrix
T =
_
_
_
_
_
0
1
I
1
0
1

0
1
I
1
0
1

0
1
I
1
0
1
.
.
.
.
.
.
.
.
.
_
_
_
_
_
. (4.4)
Proof. To show that J
1
is bounded, we take two arbitrary vectors x = x
j
j=0
and
y = y
j
j=0
in
2
(H) and compute J
1
x, y)
2
(1)
. Then, because of (4.1) we have
J
1
x, y)
2
(1)
=
1
2
_
2
0
E(e
it
)
J
1
E(s
1
(e
it
))xdt, y)
2
(1)
.
By using (3.1) we can write out the rst term in the inner product explicitly,
namely
1
2
_
2
0
E(e
it
)
J
1
E(s
1
(e
it
))xdt =
1
2
_
_
_
_
_
_
_
j=0
_
2
0
(s
1
(e
it
))
j
J
1
x
j
dt
j=0
_
2
0
e
it
(s
1
(e
it
))
j
J
1
x
j
dt
j=0
_
2
0
e
2it
(s
1
(e
it
))
j
J
1
x
j
dt
.
.
.
_
_
_
_
_
_
_
.
Hence
J
1
x, y)
2
(1)
=
1
2
k=0
_

j=0
_
2
0
e
kit
(s
1
(e
it
))
j
J
1
x
j
dt, y
k
_
1
=
1
2
k=0
_

j=0
_
2
0
(s
1
(e
it
))
j
J
1
x
j
dt, e
ikt
y
k
_
1
=
1
2
_
_
2
0
j=0
(s
1
(e
it
))
j
x
j
dt, J
k=0
e
ikt
y
k
_
1
=
1
2
__
2
0
E(s
1
(e
it
))xdt, J
1
E(e
it
)y
_
1
=
1
2
_
2
0
E(s
1
(e
it
))x, J
1
E(e
it
)y)
1
dt = E(s
1
())x, J
1
E()y)
1(D)
.
With Cauchys inequality it now follows that
J
1
x, y)
2
(1)
|E(s
1
())x|
1(D)
|J
1
E()y|
1(D)
.
Making use of equality (3.2) and taking into account that J
2
1
= I
1
, we conclude
that |J
1
E()y|
1(D)
= |y|
2
(1)
. Setting
e
i
= s
1
(e
it
), e
it
= s
1
(e
i
), dt =
1 [
1
[
2
[1 e
i
1
[
2
d
we get
|E(s
1
())x|
2
1(D)
=
1
2
_
2
0
E(s
1
(e
it
))x, E(s
1
(e
it
))x)
1
dt
=
1
2
_
2
0
E(e
i
)x, E(e
i
)x)
1
1 [
1
[
2
[1 e
i
1
[
2
d
C
2
_
2
0
E(e
i
)x, E(e
i
)x)
1
d = C|E()x|
2
1(D)
= C|x|
2
(1)
,
where
C = max
[0,2]
_
1 [
1
[
2
[1 e
i
1
[
2
_
.
So we have derived the inequality
J
1
x, y)
2
(1)
C|x|
2
(1)
|y|
2
(1)
,
which implies that |J
1
| C. Hence J
1
is bounded.
Furthermore, since
E(z)E()
=
I
1
1 z
, (4.5)
it follows by Cauchys formula that
E(z)J
1
=
1
2
_
2
0
J
1
E(s
1
(e
it
))
1 ze
it
dt =
1
2i
_
]]=1
J
1
E(s
1
())
z
d = J
1
E(s
1
(z)),
which proves (4.2). Since s
1
(s
1
(z)) = z, z D, it follows from (4.2) that
E(s
1
(z))J
1
= J
1
E(z)
and therefore, that
E(z)J
2
1
= J
1
E(s
1
(z))J
1
= J
2
1
E(z) = E(z).
Thus, for every x
2
(H),
E(z)J
2
1
x = E(z)x
which implies by Lemma 3.1, that J
2
1
= I
2
(1)
, which proves the rst equality in
(4.3).
Next, using (4.1), (4.2), (4.5) and the equality
1 s
1
(z)s
1
()
=
(1 [
1
[
2
)(1 z)
(1 z
1
)(1
1
)
,
we get
E(z)J
1
J
1
= J
1
E(s
1
(z))J
1
=
1
2
_
2
0
J
1
E(s
1
(z))E(s
1
(e
it
))
1
E(e
it
) dt
=
1
2
_
2
0
E(e
it
)
1 s
1
(z)s
1
(e
it
)
dt =
1 z
1
2(1 [
1
[
2
)
_
2
0
e
it
1
e
it
z
E(e
it
) dt
=
1 z
1
2i(1 [
1
[
2
)
_
]]=1
(
1
)
( z)
E() d.
Making use of equalities
E(z) E(0)
z
= E(z)T and zE(z) = E(z)T
,
which follow readily from (3.1) and (4.4), we obtain by the residue theorem
E(z)J
1
J
1
=
1 z
1
1 [
1
[
2
_
z
1
z
E(z) +

1
z
E(0)
_
=
1 z
1
1 [
1
[
2
_
E(z)
1
E(z) E(0)
z
_
=
1 z
1
1 [
1
[
2
E(z)(I
2
(1)
1
T)
=
1
1 [
1
[
2
E(z)(I
2
(1)
1
T
)(I
2
(1)
1
T),
which proves the second relation in (4.3).
Finally, let H be of the form (3.3). Then, on account of (4.2),
J
1
H(s
1
(z
1
), s
2
(z
2
))J
2
= J
1
E(s
1
(z
1
))F(s
2
(z
2
))J
2
= E(z
1
)J
1
F(s
2
(z
2
))J
2
and thus, H belongs to HS
1
(J
1
, J
2
, s
1
, s
2
) if and only if
E(z
1
)J
1
F(s
2
(z
2
))J
2
= E(z
1
)F(z
2
),
which is equivalent to
J
1
F(s
2
(z
2
))J
2
= F(z
2
),
by Lemma 3.1. By denition, this means that F HS
2
(1)
(J
1
, J
2
, s
2
).
We now present solution of the Problem CSym.
Theorem 4.2. Let be an C-admissible data set. Then the problem CSym admits a
solution H HS
1
(J
1
, J
2
, s
1
, s
2
). Moreover, all such H are given by the formula
H(z
1
, z
2
) =

H(z
1
, z
2
) +E(z
1
)(z
2
, )f(z
2
), (4.6)
where is an arbitrary point on the unit circle,

H is given by
H(z
1
, z
2
) =
1
2
(H
min
(z
1
, z
2
) +J
1
H
min
(s
1
(z
1
), s
2
(z
2
))J
2
)
(here H
min
is dened by (3.12)) and where f HS
2
(1)
(D) is such that
(, )f HS
2
(1)
(J
1
, J
2
, s
2
). (4.7)
Here is dened by (3.8).
Proof. Since s
j
(s
j
(z
j
)) = z
j
, we have
1
2
(H
min
(z
1
, z
2
) +J
1
H
min
(s
1
(z
1
), s
2
(z
2
))J
2
HS
1
(J
1
, J
2
, s
1
, s
2
),
and therefore the function H given by (4.6) belongs to HS
1
(J
1
, J
2
, s
1
, s
2
) if
and only if
E(z
1
)(z
2
, )f(z
2
) HS
1
(J
1
, J
2
, s
1
, s
2
).
This happens, in view of Lemma 4.1, precisely when f HS
2
(1)
(D) is such
that (4.7) holds. So H given by (4.6) is indeed in HS
1
(J
1
, J
2
, s
1
, s
2
).
On the other hand, by Theorem 3.3, the function H
min
satises (2.5), and
therefore so does the function
J
1
H
min
(s
1
(z
1
), s
2
(z
2
))J
2
.
Indeed, rst note that from the equality
s
j
(s
j
(z
j
)) = z
j
, z
j
D,
and from Lemma 2.4 it follows that condition (2.5) is equivalent to
1
(2i)
2
_
]]=1
(I
1
2
1
2
s
2
(A
2
))
1
B
2
_
_
]]=1
(I
1
1
1
1
s
1
(A
1
))
1
B
1
H(s
1
(), s
2
()) d
_
d = B
.
(4.8)
Next, let
1
and
2
be the operators dened by
1
=
_
0 I
1
1
I
1
1
0
_
,
2
=
_
0 I
1
2
I
1
2
0
_
. (4.9)
It follows from the special structure of the matrices A
1
, A
2
, B
1
and B
2
that
i
A
i
i
= s
i
(A
i
),
1
B
1
= B
1
J
1
,
2
B
2
= B
2
1
,
2
B
= B
J
2
. (4.10)
So multiplying (4.8) on the left with
2
and making use of the equalities in (4.10)
lead to
1
(2i)
2
_
]]=1
(I
1
2
1
2
A
2
)
1
B
2
_
_
]]=1
(I
1
1
1
1
A
1
)
1
B
1
J
1
H(s
1
(), s
2
()) d
_
d = B
J
2
.
Multiplying both sides of this equality on the right with J
2
gives us that also the
function J
1
H(s
1
(), s
2
())J
2
satises (2.5).
Since the set of functions satisfying (2.5) is obviously convex, we obtain that
1
2
(H
min
(z
1
, z
2
) +J
1
H
min
(s
1
(z
1
), s
2
(z
2
))J
2
) (4.11)
also satises (2.5), and by Proposition 2.5 the function (4.11) solves CSym. It
remains to prove that E(z
1
)(z
2
, )f(z
2
) is a general solution of the homogeneous
equation (2.5), i.e., in which B
= 0. But this statement follows immediately from

Theorem 3.3.
5. Solution of CSym in case
1
= 0
In this section we will give a description of all solutions of CSym under the addi-
tional assumption that
1
= 0. It follows from (4.3) that J
1
is a signature matrix
in this case. This makes it possible to replace (4.7), which is a condition on both
and f, by a condition on the free parameter f only. In addition, the assumption
1
= 0 will be used in the next section to study the CSym problem with norm
restrictions.
We assume
1
= 0 throughout this section. Several lemmas will be needed.
Lemma 5.1. Let P be given by (3.7) and let P
[1]
be its MoorePenrose pseudoin-
verse. Then
(1 [
2
[
2
)
2
P
2
= (I
1
2
1
2

2
A
2
)P(I
1
2
1
2

2
A
2
), (5.1)
and
(1 [
2
[
2
)P
[1]
= (I
1
2
1
2

2
A
2
)
2
P
[1]
2
(I
1
2
1
2

2
A
2
). (5.2)
Proof. The symmetry relation (5.1) holds in view of the special structure of the
matrices A
1
, A
2
, B
1
and B
2
. Indeed, it follows from (4.2), (4.10) and Lemma 2.4
that
2
B
+
=
1
2i
2
B
2
_
]z]=1
(zI
1
1
1
1
A
1
)
1
B
1
E(z) dz
=
1
2i
B
2
1
_
]z]=1
(zI
11
A
1
)
1
B
1
E(z) dz
=
1
2i
B
2
_
]z]=1
(zI
11
1
A
1
1
)
1
1
B
1
E(z) dz
=
1
2i
B
2
_
]z]=1
(zI
1
1
1
1
s
1
(A
1
))
1
B
1
J
1
E(z) dz
=
1
2i
B
2
_
]z]=1
(zI
1
1
1
1
A
1
)
1
B
1
J
1
E(s
1
(z)) dz
=
1
2i
B
2
_
]z]=1
(zI
1
1
1
1
A
1
)
1
B
1
E(z)J
1
dz = B
+
J
1
.
Now use [7, Lemma 4.2].
Lemma 5.2. Let be an arbitrary point on the unit circle and let P
[1]
be the
pseudoinverse of the matrix P given by (3.7). Then the operator-valued function
given in (3.8) satises
J
1
(s
2
(z), )J
1
= (z, s
2
()). (5.3)
Proof. In order to show that satises the symmetry relation (5.3), let us consider
the matrix-valued function
D(z, ) = ( z)(I
1
2
1
2
zA
2
)
1
P
[1]
(I
1
2
1
2
A
2
)
1
, (5.4)
where is an arbitrary xed point on the unit circle. Note the following three
equalities (see [7, formula(4.20)])
I
1
2
1
2
s
2
(z)s
2
(A
2
)
=
1 [
2
[
2
1 z
2
(I
1
2
1
2

2
A
2
)
1
(I
1
2
1
2
zA
2
),
I
1
2
1
2
s
2
(A
2
) = (1
2
)(s
2
()I
1
2
1
2
A
2
)(I
1
2
1
2

2
A
2
)
1
,
(1 z
2
)(s
2
(z) ) = (1
2
)(s
2
() z). (5.5)
Using (4.10), (5.2) and (5.5), we complete the proof as in the proof of [7, Lemma
4.3]. The innite-dimensional context of the present paper is not an obstacle (in
contrast with the entirely nite-dimensional context of [7]), since we assume that
H
1
and H
2
are nite-dimensional.
Now we can describe all solutions of CSym in case
1
= 0.
Theorem 5.3. Let be an C-admissible data set. Then the problem CSym admits a
solution H HS
1
(J
1
, J
2
, s
1
, s
2
). Moreover, all such H are given by the formula
(4.6), where is an arbitrary point on the unit circle, f is a free parameter in
HS
2
(1)
(
J
1
, J
2
, s
2
), and

J
1
is the matrix dened by
J
1
= (, s
2
())J
1
. (5.6)
Proof. In view of Theorem 4.2 and its proof we only have to show that (, )f
HS
2
(1)
(J
1
, J
2
, s
2
) if and only if f HS
2
(1)
(
J
1
, J
2
, s
2
). Well then, rst
assume that (, )f belongs to HS
2
(1)
(J
1
, J
2
, s
2
). Hence we have the equality
J
1
(s
2
(z
2
), )f(s
2
(z
2
))J
2
= (z
2
, )f(z
2
).
Using Lemma 5.2 and then (3.10) we can write the left-hand side of this equality as
(z
2
, s
2
())J
1
f(s
2
(z
2
))J
2
= (z
2
, )(, s
2
())J
1
f(s
2
(z
2
))J
2
,
which is equal to (z
2
, )
J
1
f(s
2
(z
2
))J
2
. So we have the equality
(z
2
, )
J
1
f(s
2
(z
2
))J
2
= (z
2
, )f(z
2
).
Since det , 0 it follows that
J
1
f(s
2
(z
2
))J
2
= f(z
2
),
i.e., that f HS
2
(1)
(
J
1
, J
2
, s
2
).
Reversing these arguments proves the other implication.
6. Solution of ONorm-CSym and SNorm-CSym in case
1
= 0
If we also take
2
equal to zero, then the representation (4.6) is orthogonal with
respect to [, ]
1S
GH
(D
2
)
and to , )
1S
GH
(D
2
)
. In this case the minimal norm
solution H
min
of (2.5) is in the class HS
1
(J
1
, J
2
, s
1
, s
2
), which means that
H
min
=

H. Hence, H
min
is also the minimal norm solution of CSym. In general
this is not the case. However, to solve ONorm-CSym and SNorm-CSym we need
the minimal norm solution of CSym. In this section we will construct the minimal
norm solution of CSym in case
1
= 0 and
2
,= 0. We will assume
1
= 0 and
2
,= 0 throughout this section.
To do so, note that according to Lemma 3.1 and Lemma 4.1 every solution
H of CSym has a representation H(z
1
, z
2
) = E(z
1
)F(z
2
), such that
F HS
2
(1)
(J
1
, J
2
, s
2
)
and
[H, H]
1S
GH
(D
2
)
= [F, F]
1S
G
2
(H)
(D)
, |H|
1S
GH
(D
2
)
= |F|
1S
G
2
(H)
(D)
.
Moreover, we know from [9] that F is a solution of the interpolation condition
1
2i
_
]z
2
]=1
(z
2
I
1
2
1
2
A
2
)
1
B
+
F(z
2
) dz
2
= B
. (6.1)
So H is the minimal norm solution of CSym if and only if F is the minimal norm
solution of the following problem.
Problem 6.1. Sym on the Unit Disk - Sym-UD
Find all F HS
2
(1)
(J
1
, J
2
, s
2
) satisfying the interpolation condition (6.1).
The minimal norm solution of this problem is constructed in [7] in case (
and H are nite-dimensional. However, this construction (together with its proof)
in [7] also holds true under the present setting.
A necessary and sucient condition for the existence of solutions of Sym-
UD is
Ran B
spanRan A
j
2
B
+
; j
1
, j
2
= 0, 1, . . . .
This inclusion holds true, because in [8] (Lemma 4.3) it is proved that the right-
hand side of this inclusion is equal to the right-hand side of (2.7), which holds true
by assumption.
We know from Theorem 3.3 in [9] that all solutions of Sym-UD without the
symmetry condition are parametrized by
F(z
2
) = F
min
(z
2
) + (z
2
, )h(z
2
), (6.2)
where
F
min
(z
2
) = B
+
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
B
and where is given by (3.8). Further, h is free parameter in HS
2
(1)
(D
2
).
Moreover, the sum in (6.2) is orthogonal with respect to both inner products
[, ]
1S
G
2
(H)
(D)
and , )
1S
G
2
(H)
(D)
.
In [7, Theorem 5.3] a description is given of all the solutions F of Sym-UD
which are in the set HS
2
(1)
(J
1
, J
2
, s
2
). Namely,
F(z
2
) =

F(z
2
) + (z
2
, )h(z
2
), (6.3)
where
F(z
2
) =
1
2
(F
min
(z
2
) +J
1
F
min
(s
2
(z
2
))J
2
,
and h is a free parameter in HS
2
(1)
(
J
1
, J
2
, s
2
). Also in this case the represen-
tation in (6.3) is not orthogonal in general and hence

F is not the minimal norm
solution of Sym-UD. However, the minimal norm solution can be constructed from
F
min
, which is done in Section 6 of [7]. The following theorem is proved there.
Theorem 6.2. Let be an C-admissible data set and let z
0
be the xed point of s
2
,
i.e.,
z
0
=
1 +
_
1 [
2
[
2
2
.
Then all functions F HS
2
(1)
(
J
1
, J
2
, s
2
) satisfying the interpolation condition
(6.1) are parametrized by the formula
F(z
2
) =

F
min
(z
2
) + (z
2
, )h(z
2
), (6.4)
where
F
min
(z
2
) =
1
2
(z
2
z
0
)
((1 z
2
2
)F
min
(z
2
) +
_
1 [
2
[
2
J
1
F
min
(s
2
(z
2
))J
2
) (6.5)
and h is a free parameter in HS
2
(1)
(
J
1
, J
2
, s
2
). The function F of the form
(6.4) satises the norm constraint (1.5), resp. (1.6), if and only if the corresponding
parameter h is subject to
[h, h]
1S
G
2
(H)
(D)
[
F
min
,

F
min
]
1S
G
2
(H)
(D)
,
resp.
|h|
2
1S
G
2
(H)
(D)

2
|
F
min
|
2
1S
G
2
(H)
(D)
.
Moreover, the decomposition (6.4) is orthogonal with respect to [, ]
1S
G
2
(H)
(D)
and , )
1S
G
2
(H)
(D)
.
Because the decomposition (6.4) is orthogonal, it follows that

F
min
is the
minimal norm solution of Sym-UD. It is possible to compute the norm of

F
min
explicitly. We will do this following the approach of Section 6 in [7].
First note that
J
1
F
min
(s
2
(z
2
))J
2
= (1 z
2
2
)B
+
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
(I
1
2
1
2

2
A
2
)
1
B
.
(6.6)
Here we used (4.10), Lemma 5.1 and the rst equality in (5.5). Substituting F
min
and (6.6) into (6.5) gives us the following expression for

F
min
:
F
min
(z
2
) =
1 z
2
2
z
2
z
0
B
+
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
(z
0
I
1
2
1
2
A
2
)(I
1
2
1
2

2
A
2
)
1
B
.
(6.7)
It is obvious that

F
min
is also a solution of Sym-UD without the symmetry condi-
tion, so it is possible to write

F
min
in the form of (6.2). This means that there is
an h HS
2
(1)
(D) such that
F
min
(z
2
) = F
min
(z
2
) + (z
2
, )h(z
2
). (6.8)
Formula (6.40) in [7] gives us that this equality holds true for
h(z
2
) =
h(z
2
) :=
_
1 [
2
[
2
2
(z
2
z
0
)
T,
where
T =
2
B
+
(I
1
2
1
2
A
2
)
1
P
[1]
(I
1
2
1
2
A
2
)(I
1
2
1
2

2
A
2
)
1
B
.
Because is inner (this follows from (3.9)) and the sum in (6.8) is orthogonal
with respect to [, ]
1S
G
2
(H)
(D)
and , )
1S
G
2
(H)
(D)
we have
[
F
min
,

F
min
]
1S
G
2
(H)
(D)
= [F
min
, F
min
]
1S
G
2
(H)
(D)
+ [
h,
h]
1S
G
2
(H)
(D)
,
|
F
min
|
2
1S
G
2
(H)
(D)
= |F
min
|
2
1S
G
2
(H)
(D)
+ |
h|
2
1S
G
2
(H)
(D)
.
We know from Lemma 3.2 in [9] that
[F
min
, F
min
]
1S
G
2
(H)
(D)
= B
P
[1]
B
,
|F
min
|
2
1S
G
2
(H)
(D)
= Trace(B
P
[1]
B
),
and from formula (6.41) in [7] we know that
[
h,
h]
1S
G
2
(H)
(D)
=
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
T
T
|
h|
2
1S
G
2
(H)
(D)
=
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
Trace(T
T).
Hence,
[
F
min
,

F
min
]
1S
G
2
(H)
(D)
= B
P
[1]
B
+
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
T
T,
|
F
min
|
2
1S
G
2
(H)
(D)
= Trace(B
P
[1]
B
) +
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
Trace(T
T).
Further,
T
T = [
2
[
2
B
(I
1
2
1
2

2
A
2
)
1
(P
[1]
A
2
P
[1]
A
2
)(I
1
2
1
2

2
A
2
)
1
B
,
see (6.16) in [7], so
[
F
min
,

F
min
]
1S
G
2
(H)
(D)
= B
(P
[1]
+
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
(I
1
2
1
2

2
A
2
)
1
(P
[1]
A
2
P
[1]
A
2
)(I
1
2
1
2

2
A
2
)
1
)B
,
|
F
min
|
2
1S
G
2
(H)
(D)
= Trace(B
(P
[1]
+
_
1 [
2
[
2
2(1 +
_
1 [
2
[
2
)
(I
1
2
1
2

2
A
2
)
1
(P
[1]
A
2
P
[1]
A
2
)(I
1
2
1
2

2
A
2
)
1
)B
).
Now we will return to the original problems CSym, ONorm-CSym, and
SNorm-CSym. We remarked before that
H(z
1
, z
2
) = E(z
1
)F(z
2
)
is the minimal norm solution of CSym if and only if F is the minimal norm solution
of Sym-UD. So we have proved that the minimal norm solution of CSym, denoted
by

H
min
, is given by
H
min
(z
1
, z
2
) =
1 z
2
2
z
2
z
0
E(z
1
)B
+
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
(z
0
I
1
2
1
2
A
2
)(I
1
2
1
2

2
A
2
)
1
B
.
According to Lemma 4.3 in [8] this is equal to
H
min
(z
1
, z
2
) =
1 z
2
2
z
2
z
0
B
1
(I
1
1
1
1
z
1
A
1
)
1
B
2
(I
1
2
1
2
z
2
A
2
)
1
P
[1]
(z
0
I
1
2
1
2
A
2
)(I
1
2
1
2

2
A
2
)
1
B
.
(6.9)
Moreover, we have that
[
H
min
,

H
min
]
1S
GH
(D
2
)
= [
F
min
,

F
min
]
1S
G
2
(H)
(D)
,
|
H
min
|
1S
GH
(D
2
)
= |
F
min
|
1S
G
2
(H)
(D)
.
All this leads to the following theorem.
1
, A
2
, B
1
, B
2
, B
be an C-admissible data set. Then the

following statements hold true.
(a) Problem CSym has a solution H HS
1
(J
1
, J
2
, s
1
, s
2
). Moreover, all so-
lutions H are described by the formula
H(z
1
, z
2
) =

H
min
(z
1
, z
2
) +E(z
1
)(z
2
, )h(z
2
), (6.10)
where is a xed point on the unit circle and

H
min
is the minimal norm
solution, with respect to the norm | |
1S
GH
(D
2
)
, given by (6.9) and h is
a free parameter in HS
2
(1)
(
J
1
, J
2
, s
2
). The sum in (6.10) is orthogonal
with respect to the operator-valued inner product [, ]
1S
GH
(D
2
)
, as well as
with respect to , )
1S
GH
(D
2
)
.
(b) Problem ONorm-CSym has a solution H HS
1
(J
1
, J
2
, s
1
, s
2
) if and only
if the inequality
[
H
min
,

H
min
]
1S
GH
(D
2
)
0
holds true. In this case, all solution H are described by formula (6.10), where
the parameter h HS
2
(1)
(
J
1
, J
2
, s
2
) satises the operator-valued norm
constraint
[h, h]
1S
G
2
(H)
(D)
[
H
min
,

H
min
]
1S
GH
(D
2
)
,
(c) Problem SNorm-CSym has a solution H HS
1
(J
1
, J
2
, s
1
, s
2
) if and only
if the inequality
2
|
H
min
|
2
1S
GH
(D
2
)
holds true. In this case, all solutions H are described by the formula (6.10),
where the parameter h HS
2
(1)
(
J
1
, J
2
, s
2
) satises the norm constraint
|h|
2
1S
G
2
(H)
(D)

2
|
H
min
|
2
1S
GH
(D
2
)
We conclude this section with a remark (pointed out to us by the referee)
that a linear fractional change of variables allows one to reduce the general case
of conformal involutive mappings to the case when
1
= 0. Indeed, let
s
j
(z) =

j
z
1 z
j
, (j = 1, 2),
and let
(z) = z and
j
(z) =
z +
1
1]
j
]
2
j
1 +z
1
1]
j
]
2
j
(j = 1, 2)
be two automorphisms of the unit disk. It is readily checked that
1
j
s
j

j
= .
Therefore, a function H(z
1
, z
2
) satises (2.1) if and only if the function G(z
1
, z
2
) :=
H(
1
(z
1
),
2
(z
2
)) satises
J
1
G(z
1
, z
2
)J
2
= G(z
1
, z
2
), (z
1
, z
2
) D
2
.
Furthermore, H satises interpolation condition (1.4) if and only if G satises
1
(2i)
2
_
]z]=1
(zI
2
(A
2
))
1
B
2
_
_
]]=1
(I
1
(A
1
))
1
B
1
G(, z)
_
dz = B
,
by Lemma 2.4. Thus, one can use formula (6.10) (applied for G) to obtain a general
formula for H HS
1
(J
1
, J
2
, s
1
, s
2
) satisfying (1.4). However, the properties of
orthogonality and minimal norm of Theorem 6.3 generally are not preserved under
the above change of variables.
7. Interpolation with interchange symmetry
In this section we formulate our basic interpolation problems IT, ONorm and
SNorm in a subspace of HS
1
(D
2
) which is dened by another type of symmetry.
Denition 7.1. We denote by HS
1
sym
(D
2
) the class of all H HS
1
(D
2
) which
satisfy the interchange symmetry relation
H(z
1
, z
2
) = H(z
2
, z
1
) (7.1)
for all z
1
, z
2
D.
It is easily seen that HS
1
sym
(D
2
) is a subspace of HS
1
(D
2
). In this sub-
space we consider the following interpolation problems.
Problem 7.2. TI with Interchange Symmetry ISym
Let be given a data set (1.2) satisfying (1.8). Find all H HS
1
sym
(D
2
) satisfying
the interpolation condition (1.4).
Problem 7.3. ISym with Operator Norm Constraint ONorm-ISym
Let be given a data set (1.2) as in Problem ISym. Let also be given an operator
: ( (, 0. Find all H HS
1
sym
(D
2
) satisfying the interpolation condition
(1.4) and the operator-valued norm constraint (1.5).
Problem 7.4. ISym with Scalar Norm Constraint SNorm-ISym
Let be given a data set (1.2) as in Problem ISym. Let also be given positive number
. Find all H HS
1
sym
(D
2
) satisfying the interpolation condition (1.4) and the
norm constraint (1.6).
As in the case of the CSym problems, we will proceed to set up a second
interpolation problem using the symmetry. To make this approach work, additional
hypotheses are needed. Namely, we assume that H
1
= H
2
and that there exist
operators

A
1
: H
1
H
1
,

A
2
: H
1
H
1
and

B
2
: H
1
H
1
such that
(
A
1
) (
A
2
) D
and the equality
(I
1
1
A
2
)
1
B
2
(I
1
1
A
1
)
1
= (I
1
1

A
2
)
1
B
2
(I
1
1

A
1
)
1
(7.2)
holds true for all and on the unit circle. (These hypotheses will be assumed
throughout this section.) In particular, (7.2) holds if B
2
= I and A
1
A
2
= A
2
A
1
(then take

B
2
= I,

A
2
= A
1
,

A
1
= A
2
.)
Then every H HS
1
sym
(D
2
) satisfying (1.4) also satises the equality
1
(2i)
2
_
]]=1
_
]]=1
(I
1
1

A
2
)
1
B
2
(I
1
1

A
1
)
1
B
1
H(, ) d d = B
.
Interchanging the order of integration and replacing by and vice versa gives
the following interpolation condition:
1
(2i)
2
_
]]=1
(I
1
1

A
2
)
1
B
2
_
_
]]=1
(I
1
1

A
1
)
1
B
1
H(, ) d
_
d = B
.
(7.3)
The interpolation conditions (1.4) and (7.3) can be written together as
1
(2i)
2
_
]]=1
(I
1
1
1
1
A
2
)
1
B
2
_
_
]]=1
(I
1
1
1
1
A
1
)
1
B
1
H(, ) d
_
d = B
,
(7.4)
where
A
1
=
_
A
1
0
0

A
1
_
, A
2
=
_
A
2
0
0

A
2
_
, B
1
=
_
B
1
B
1
_
,
B
2
=
_
B
2
0
0

B
2
_
, B
=
_
B
_
.
(7.5)
As before, a necessary condition for (7.4) to have solutions is that the oper-
ators in (7.5) satisfy (2.7). We will call a data set
= A
1
, A
2
, B
1
, B
2
, B
I-admissible if it satises (2.7) (with A

1
, A
2
, B
1
, B
2
and B
as in (7.5)) together
with (1.3). In the next section we will assume that is I-admissible. We obtain
an analogue of Proposition 2.5:
Proposition 7.5. Let be an I-admissible data set. Then the following equivalences
hold true.
(a) H HS
1
sym
(D
2
) solves ISym if and only if H satises (7.4).
(b) H HS
1
sym
(D
2
) solves ONorm-ISym if and only if H satises (7.4) and the
(c) H HS
1
sym
(D
2
) solves SNorm-ISym if and only if H satises (7.4) and the
We will solve Problems ISym, ONorm-ISym and SNorm-CSym in the next
section. These problems will be solved in a similar pattern as Problems CSym,
ONorm-CSym and SNorm-CSym. First we will describe all the solutions of ISym.
Then we will determine the minimal norm solution of ISym, which makes it possible
to solve Problems ONorm-ISym and SNorm-ISym.
8. The solution of ISym, ONorm-ISym and SNorm-CSym
Note that the special structure of the operators in (2.6) is not used in Theorem 3.3.
Hence, if we assume that is I-admissible, then the statements (a), (b) and (c)
in Theorem 3.3, after replacing (2.5) by (7.4), also hold true for the operators
A
1
, A
2
, B
1
, B
2
and B
dened by (7.5). This means that all solutions of (7.4),

without the symmetry constraint, are described by (3.11) and that the minimal
solution is given by (3.12). In the next theorem, which is an analogue of Theo-
rem 4.2, we give a parametrization of all solutions of ISym, hence all solutions of
(7.4) also satisfying (7.1).
Theorem 8.1. Let be an I-admissible data set such that (7.2) is satised. Then
the problem ISym admits a solution H HS
1
sym
(D
2
). Moreover, all solutions H
are given by the formula
H(z
1
, z
2
) =

H(z
1
, z
2
) +E(z
1
)(z
2
, )f(z
2
), (8.1)
where is an arbitrary point on the unit circle,

H is given by
H(z
1
, z
2
) =
1
2
(H
min
(z
1
, z
2
) +H
min
(z
2
, z
1
)) (8.2)
(here H
min
is dened by (3.12)) and where f HS
2
(1)
(D) is such that
E(z
1
)(z
2
, )f(z
2
) = E(z
2
)(z
1
, )f(z
1
). (8.3)
Proof. It is easy to check that

H is in HS
1
sym
(D
2
), so the function H given by
(8.1) is an element of HS
1
sym
(D
2
) if and only if the free parameter f satises
(8.3). So the only thing we have to show is that H indeed is a solution of ISym.
First consider

H. We know from Theorem 3.3 that H
min
is a solution of (7.4).
From this it follows that also the function H
min
, dened by
H
min
(z
1
, z
2
) = H
min
(z
2
, z
1
),
is a solution of (7.4). Indeed, rst note that we can write
(I
1
1
1
1
A
2
)
1
B
2
(I
1
1
1
1
A
1
)
1
in the form
_
(I
1
1
A
2
)
1
B
2
(I
1
1
A
1
)
1
0
0 (I
1
1

A
2
)
1
B
2
(I
1
1

A
1
)
1
_
.
Because of condition (7.2) this is equal to
_
(I
1
1

A
2
)
1
B
2
(I
1
1

A
1
)
1
0
0 (I
1
1
A
2
)
1
B
2
(I
1
1
A
1
)
1
_
.
This proves the equality
(I
1
1
1
1
A
2
)
1
B
2
(I
1
1
1
1
A
1
)
1
=
1
(I
1
1
1
1
A
2
)
1
B
2
(I
1
1
1
1
A
1
)
1
1
,
(8.4)
where
1
is as in (4.9). This implies that H
min
also satises
1
(2i)
2
_
]]=1
_
]]=1
1
(I
1
1
1
1
A
2
)
1
B
2
(I
1
1
1
1
A
1
)
1
1
B
1
H
min
(, ) d d = B
,
which is equivalent (by changing the order of integration and replacing by and
vice versa) to
1
(2i)
2
_
]]=1
_
]]=1
1
(I
1
1
1
1
A
2
)
1
B
2
(I
1
1
1
1
A
1
)
1
1
B
1
H
min
(, ) d d = B
.
Now we multiply this equality on the left by
1
and note that
1
B
1
= B
1
and
1
B
= B
.
Hence also H
min
satises (7.4). Because the solution set of (7.4) is convex, it follows
that

H dened by (8.2) also satises (7.4). As a consequence of Proposition 7.5
we have that

H is a solution of ISym.
It remains to show that E(z
1
)(z
2
, )f(z
2
) is a solution of (7.4) with B
= 0.
But this immediately follows from Theorem 3.3.
The representation (4.6) in Theorem 4.2 was not orthogonal, so the minimal
norm solution of CSym had to be constructed to solve ONorm-CSym and SNorm-
CSym. In the case of ISym we do not have to do this, because it turns out that
the representation (8.1) is orthogonal. This is the following lemma.
Lemma 8.2. Let be an I-admissible data set such that (7.2) is satised. Then the
sum in (8.1) is orthogonal with respect to [, ]
1S
GH
(D
2
)
and , )
1S
GH
(D
2
)
.
Proof. We have to prove that
[
H, G]
1S
GH
(D
2
)
= 0,
H, G)
1S
GH
(D
2
)
= 0 (8.5)
where G is of the form
G(z
1
, z
2
) = E(z
1
)(z
2
, )f(z
2
)
with f a free parameter in HS
2
(1)
(D) such that (8.3) is satised, i.e. such that
G(z
1
, z
2
) = G(z
2
, z
1
). We will only prove the rst equality, because the second one
is a direct consequence of it.
Note that
[
H, G]
1S
GH
(D
2
)
=
1
2
[H
min
, G]
1S
GH
(D
2
)
+
1
2
[H
min
, G]
1S
GH
(D
2
)
.
First we will show that the rst term on the right-hand side is equal to zero. In
[9] it is proved that H
min
admits the representation
H
min
(z
1
, z
2
) = E(z
1
)F
min
(z
2
),
where F
min
is given by
F
min
(z
2
) = B
+
(I
1
1
1
1
z
2
A
2
)
1
P
[1]
B
.
Hence, with Lemma 3.1 in the present paper and Lemma 3.7 in [3] we see that
[H
min
, G]
1S
GH
(D
2
)
= [F
min
, (, )f]
1S
GH
(D
2
)
= 0.
Next, remark that H
min
admits the factorization
H
min
(z
1
, z
2
) = E(z
2
)F
min
(z
1
)
and that
G(z
1
, z
2
) = E(z
2
)(z
1
, )f(z
1
).
So again with the same lemmas as above we see that also
[H
min
, G]
1S
GH
(D
2
)
= [F
min
, (, )f]
1S
GH
(D
2
)
= 0.
This concludes the proof of the lemma.
This lemma shows that

H is the minimal norm solution of ISym. It turns
out that

H is equal to H
min
, which is the minimal norm solution of (7.4). We will
prove this statement by showing that
|
H|
1S
GH
(D
2
)
= |H
min
|
1S
GH
(D
2
)
.
Lemma 8.3. Let be an I-admissible data set such that (7.2) is satised. Then the
norm of the operator

H dened by (8.2) is given by
|
H|
2
1S
GH
(D
2
)
= Trace(B
P
[1]
B
).
Proof. We can split up |
H|
2
1S
GH
(D
2
)
in four terms as follows:
|
H|
2
1S
GH
(D
2
)
=
1
4
_
H
min
, H
min
)
1S
GH
(D
2
)
+H
min
, H
min
)
1S
GH
(D
2
)
+ H
min
, H
min
)
1S
GH
(D
2
)
+H
min
, H
min
)
1S
GH
(D
2
)
_
.
(8.6)
Recall from Theorem 3.3 that
H
min
, H
min
)
1S
GH
(D
2
)
= Trace(B
P
[1]
B
).
In the proof of the previous lemma we have remarked that H
min
admits the rep-
resentation
H
min
(z
1
, z
2
) = E(z
2
)F
min
(z
1
),
so it follows from Lemma 3.1 that
H
min
, H
min
)
1S
GH
(D
2
)
= F
min
, F
min
)
1S
GH
(D
2
)
= Trace(B
P
[1]
B
).
The last equality follows from Lemma 3.2 in [9]. Since
H
min
, H
min
)
1S
GH
(D
2
)
= H
min
, H
min
)
1S
GH
(D
2
)
,
we only have to compute H
min
, H
min
)
1S
GH
(D
2
)
.
Well then, by denition we have that H
min
, H
min
)
1S
GH
(D
2
)
is equal to
1
(2)
2
_
2
0
_
2
0
Trace
_
B
P
[1]
(e
it
1
, e
it
2
)P
[1]
B
_
dt
1
dt
2
.
Here is given by
(e
it
1
, e
it
2
) = (I
1
1
1
1
e
it
1
A
2
)
1
B
2
(I
1
1
1
1
e
it
2
A
1
)
1
B
1
B
1
(I
1
1
1
1
e
it
1
A
1
)
1
B
2
(I
1
1
1
1
e
it
2
A
2
)
1
.
Using (8.4) and the equality
1
B
1
= B
1
we can write as
(e
it
1
, e
it
2
) =
1
(I
1
1
1
1
e
it
2
A
2
)
1
B
2
(I
1
1
1
1
e
it
1
A
1
)
1
B
1
B
1
(I
1
1
1
1
e
it
1
A
1
)
1
B
2
(I
1
1
1
1
e
it
2
A
2
)
1
.
It follows from Lemma 4.3 in [8] that
B
2
(I
1
1
1
1
e
it
1
A
1
)
1
B
1
= B
+
E(e
it
1
)
,
so
(e
it
1
, e
it
2
) =
1
(I
1
1
1
1
e
it
2
A
2
)
1
B
+
B
+
(I
1
1
1
1
e
it
2
A
2
)
1
.
Now making use of (3.6), we nd
(e
it
1
, e
it
2
) =
1
(I
1
1
1
1
e
it
2
A
2
)
1
(P A
2
PA
2
)(I
1
1
1
1
e
it
2
A
2
)
1
=
1
P(I
1
1
1
1
e
it
2
A
2
)
1
+
1
(e
it
2
I
1
1
1
1
A
2
)
1
A
2
P,
which implies that
H
min
, H
min
)
1S
GH
(D
2
)
=
1
2
_
2
0
Trace((B
P
[1]
1
_
P(I
1
1
1
1
e
it
2
A
2
)
1
+ (e
it
2
I
1
1
1
1
A
2
)
1
A
2
P
_
P
[1]
B
) dt
2
= Trace(B
P
[1]
1
PP
[1]
B
)
= Trace(B
P
[1]
1
P
RanP
B
).
We obtain from (3.7) the equality
Ran P = spanRan A
l
2
B
+
; l = 0, 1, 2, . . ..
Further, Lemma 4.3 in [8] gives us that
spanRan A
j
2
2
B
2
A
j
1
1
B
1
; j
1
, j
2
= 0, 1, . . . = spanRan A
l
2
B
+
; l = 0, 1, 2, . . ..
Because of the assumption that is I-admissible, this equality implies that
B
Ran P,
hence P
RanP
B
= B
. So
H
min
, H
min
)
1S
GH
(D
2
)
= Trace(B
P
[1]
1
B
) = Trace(B
P
[1]
B
).
Substituting all the values we have found into (8.6) proves the lemma.
This lemma shows that the norm of

H is equal to the norm of H
min
. It is
obvious that

H is also a solution of (7.4) without the symmetry condition, hence
from the remark after Theorem 3.3 it follows that

H = H
min
.
We now present the solution of Problems ONorm-ISym and SNorm-ISym.
1
, A
2
, B
1
, B
2
, B
be an I-admissible data set such that

(7.2) is satised. Then the following statements hold true.
(a) Problem ISym has a solution H HS
1
sym
(D
2
). Moreover, all solutions H are
described by the formula (3.11), where is a xed point on the unit circle and
H
min
is the minimal norm solution, with respect to the norm | |
1S
GH
(D
2
)
,
given by (3.12), and where f is a free parameter in HS
2
(1)
(D) such that
(8.3) holds. The sum in (3.11) is orthogonal with respect to the operator-valued
inner product [, ]
1S
GH
(D
2
)
, as well as with respect to the inner product
, )
1S
GH
(D
2
)
.
(b) Problem ONorm-ISym has a solution H HS
1
sym
(D
2
) if and only if
B
P
[1]
B
0.
In this case, all solutions H are described by formula (3.11), where the pa-
rameter f HS
2
(1)
(D) satises the operator-valued norm constraint
[f, f]
1S
G
2
(H)
(D)
B
P
[1]
B
and (8.3), but otherwise is free.

(c) Problem SNorm-ISym has a solution H HS
1
sym
(D
2
) if and only if
2
Trace(B
P
[1]
B
).
In this case, all solutions H are described by the formula (3.11), where the
parameter f HS
2
(1)
(D) satises the norm constraint
|f|
2
1S
G
2
(H)
(D)

2
Trace(B
P
[1]
B
)
and (8.3), but otherwise is free.
9. Multipoint interpolation
The basic two-sided tangential interpolation problems for matrix functions of one
variable in the Nudelman formulation, as studied in [11] for rational matrix func-
tions, may be thought of as one-point interpolation. This is because the interpo-
lation conditions for these problems can be expressed in terms of equations such
that each equation involves tangential values of the unknown interpolant function
and its rst several derivatives at one point only (however, this point may be
dierent from one equation to another); see Chapter 16 in [11]. In contrast, the
multipoint interpolation is stated in terms of equations such that each equation
may involve in an essential way linear combinations of tangential values of the
unknown interpolant function and its rst several derivatives at several points.
There are few results available concerning multipoint interpolation, see [7], [10].
There, the multipoint (actually, two point) interpolation was studied using results
on bitangential interpolation with symmetries.
In the present framework, we shall derive results on certain multipoint in-
terpolation problems for operator functions on the bidisk, by utilizing a similar
approach of using theorems on interpolation with componentwise symmetries ob-
tained in previous sections. The data set for these problems consist of an ordered
set of six operators
m
= A
1
, A
2
, B
1,1
, B
1,2
, B
2
, B
, (9.1)
where
A
i
: H
i
H
i
, for i = 1, 2,
and
B
1,j
: H H
1
, (for j = 1, 2), B
2
: H
1
H
2
, B
: ( H
2
are operators such that (A
1
) (A
2
) D. (The subscript
m
stands for multi-
point). The interpolation condition is:
1
(2i)
2
_
]]=1
(I A
2
)
1
B
2
_
_
]]=1
(I A
1
)
1
B
1,1
H(, ) d
_
d
+
1
(2i)
2
_
]]=1
(I s
2
(A
2
))
1
B
2
_
_
]]=1
(I s
1
(A
1
))
1
B
1,2
H(, ) d
_
d = B
,
(9.2)
where the mappings s
j
: D D are given by
s
j
(z) =

j
z
1 z
j
,
j
D xed (j = 1, 2). (9.3)
Problem 9.1. Multipoint Tangential Interpolation MTI
Given a data set (9.1), nd all functions H HS
1
(D
2
) satisfying the interpo-
lation condition (9.2).
Using Lemma 2.4, rewrite (9.2) in the form
1
(2i)
2
_
]]=1
(I A
2
)
1
B
2
_
_
]]=1
(I A
1
)
1
B
1,1
H(, ) d
_
d
+
1
(2i)
2
_
]]=1
(I A
2
)
1
B
2
_
_
]]=1
(I A
1
)
1
B
1,2
H(s
1
(), s
2
()) d
_
d = B
,
(9.4)
and letting
B
1
= (B
1,1
B
1,2
) , G(z
1
, z
2
) =
_
H(z
1
, z
2
)
H(s
1
(z
1
), s
2
(z
2
))
_
: ( HH,
we see that (9.4) is the interpolation condition for a CSym problem, with the
unknown interpolant function G(z
1
, z
2
), where
J
1
=
_
0 I
I 0
_
, J
2
= I.
Theorem 4.2 now gives:
Theorem 9.2. The problem MTI admits a solution H HS
1
(D
2
) if and only
if the inclusion
Ran B
spanRan A
j
2
2
B
2
A
j
1
1
B
1
; j
1
, j
2
= 0, 1, . . . (9.5)
holds true, where
A
1
=
_
A
1
0
0 s
1
(A
1
)
_
, A
2
=
_
A
2
0
0 s
2
(A
2
)
_
,
B
1
=
_
B
1,1
B
1,2
B
1,2
B
1,1
_
, B
2
=
_
B
2
0
0 B
2
_
, B
=
_
B
_
.
(9.6)
In this case, all solutions H are given by the following formula, where the unimod-
ular number is arbitrary but xed in advance:
H(z
1
, z
2
) = [I 0]
_
1
2
__
G
1
(z
1
, z
2
)
G
2
(z
1
, z
2
)
_
+
_
G
2
(s
1
(z
1
), s
2
(z
2
))
G
1
(s
1
(z
1
), s
2
(z
2
))
__
+ E(z
1
)(z
2
, )f(z
2
) ,
(9.7)
where f HS
2
(11)
is a parameter subject to the only condition that
(, )f HS
2
(11)
(J
1
, I, s
2
),
where
J
1
=
1
2
_
2
0
E(e
it
)
_
0 I
I 0
_
E(s
1
(e
it
)) dt.
The function
_
G
1
(z
1
, z
2
)
G
2
(z
1
, z
2
)
_
in (9.7) is given by the formula
_
G
1
(z
1
, z
2
)
G
2
(z
1
, z
2
)
_
= B
1
(I z
1
A
1
)
1
B
2
(I z
2
A
2
)
1
P
[1]
B
,
where P
[1]
is the MoorePenrose pseudoinverse of the unique solution P of the
equation
P A
2
PA
2
= B
+
B
+
, B
+
=
1
2i
B
2
_
]z]=1
(zI A
1
)
1
B
1
E(z) dz.
The function in (9.7) is given by
(z, ) = I
2
(11)
+ (z )B
+
(I
1
2
1
2
zA
2
)
1
P
[1]
(I
1
2
1
2
A
2
)
1
.
To study the problem MTI under additional norm constraints, we shall use
a dierent approach, similar to that of [7, Section 7].
Consider an auxiliary interpolation problem given by the next two simulta-
neous equations:
1
(2i)
2
_
]]=1
(I A
2
)
1
B
2
_
_
]]=1
(I A
1
)
1
B
1,1
H(, ) d
_
d = D, (9.8)
1
(2i)
2
_
]]=1
(I s
2
(A
2
))
1
B
2
_
_
]]=1
(I s
1
(A
1
))
1
B
1,2
H(, ) d
_
d = B
D,
(9.9)
where D is an auxiliary operator. Letting
A
1
=
_
A
1
0
0 s
1
(A
1
)
_
,

A
2
=
_
A
2
0
0 s
2
(A
2
)
_
,
B
1
=
_
B
1,1
B
1,2
_
,

B
2
=
_
B
2
0
0 B
2
_
,

B
=
_
D
B
D
_
,
(9.10)
we rewrite equations (9.8) and (9.9) in the form
1
(2i)
2
_
]]=1
(I

A
2
)
1
B
2
_
_
]]=1
(I

A
1
)
1
B
1
H(, ) d
_
d =

B
. (9.11)
Invoking a part of Theorem 3.3, we obtain:
Proposition 9.3. For a xed D, assume that
:=
A
1
,

A
2
,
B
1
,
B
2
,

B
is a C-admissible data set. Let P

[1]
be the MoorePenrose pseudoinverse of the
unique positive semidenite solution P of the Stein equation
P

A
2
P
2
=

B
+
+
,

B
+
=
1
2i
B
2
_
]z]=1
(zI

A
1
)
1
B
1
E(z) dz. (9.12)
Then:
(a) There exists a function H HS
1
(D
2
) satisfying (9.8) and (9.9) and such
that
[H, H]
1S
GH
(D
2
)
(9.13)
if and only if
P
[1]
0. (9.14)
Moreover, all such H are described by formula (3.11), with A
1
, A
2
, B
1
, B
2
,
and B
are replaced with

A
1
,

A
2
,

B
1
,

B
2
, and

B
, respectively, and where

the parameter f HS
2
(1)
(D) is subject only to the operator-valued norm
constraint
[f, f]
1S
G
2
(H)
(D)
B
P
[1]
B
.
(b) There exists a solution H HS
1
(D
2
) satisfying (9.8) and (9.9) and such
that
|H|
1S
GH
(D
2
)
(9.15)
if and only if
2
Trace(
P
[1]
). (9.16)
Moreover, all such H are described by formula (3.11), with A
1
, A
2
, B
1
, B
2
,
and B
are replaced with

A
1
,

A
2
,

B
1
,

B
2
, and

B
, respectively, and where

the parameter f HS
2
(1)
(D) is subject only to the norm constraint
|f|
2
1S
G
2
(H)
(D)

2
Trace(B
P
[1]
B
).
To guarantee that

is C-admissible regardless of D, we assume that
spanRan

A
j
2
2
B
2
A
j
1
1
B
1
; j
1
, j
2
= 0, 1, . . . = H
2
H
2
. (9.17)
As it follows from Proposition 9.3, there exists a function H HS
1
(D
2
) satis-
fying the interpolation condition (9.2) and such that (9.13) holds true if and only
if (9.14) holds true for some D.
We now analyze when (9.14) holds true for some D. Let
P =
_
I 0
I I
_
P
_
I I
0 I
_
,
where I stands for the identity operator on H
2
. Then
P
[1]
=
_
D
P
[1]
_
D
B
_
. (9.18)
Partition
P
[1]
=
_

_
, , , : H
2
H
2
.
Denoting by
[1]
the MoorePenrose pseudoinverse of the positive denite oper-
ator , we have (cf. formula (7.15) in [7])
[1]
[1]
=
[1]
,
[1]
= . (9.19)
The rst equality in (9.19) is a part of the denition of the MoorePenrose pseu-
doinverse, whereas the second equality in (9.19) is a consequence of the posi-
tive semideniteness of

P
[1]
. Indeed, it follows that Ran Ran , and since
I
[1]
is an orthogonal projection that maps the range of to zero, we have
(I
[1]
) = 0,
which is equivalent to the second equality in (9.19). Using (9.19), write
_
D
P
[1]
_
D
B
_
= (D
+B
[1]
)(D +
[1]
B
) +B
[1]
)B
.
Comparing with (9.18), we conclude that the inequality
P
[1]
(9.20)
holds true for some D if and only if
B
[1]
)B
, (9.21)
and in this case (9.20) holds true for D =
[1]
B
. We have proved the rst

part of the following theorem; the second part can be proved analogously.
Theorem 9.4. Let there be given a data set (9.1), and assume that condition (9.17)
holds true.
(a) Let : ( ( be a positive semidenite operator. Then there exists
a function H HS
1
(D
2
) satisfying the interpolation condition (9.2) and the
operator norm constraint [H, H]
1S
GH
(D
2
)
if and only if (9.21) holds true.
(b) For a given > 0, there exists an H HS
1
(D
2
) satisfying the inter-
polation condition (9.2) and the norm constraint |H|
1S
GH
(D
2
)
if and only
if the inequality
Trace
_
B
[1]
)B
_

holds true.
Using the above approach and Theorem 3.3, one can derive in a straightfor-
ward manner a description of all solutions in the class HS
1
(D
2
) of the interpo-
lation problem (9.2) with either the operator or scalar norm constraint. However,
such a description is cumbersome, because the extraneous operator D is involved,
and therefore we leave it out.
Acknowledgment. We thank V. Bolotnikov for several useful discussions concerning
the paper.
References
[1] D. Alpay and V. Bolotnikov, On the tangential interpolation problem for matrix-
valued H
2
-functions of two variables, Proceedings of the Amer. Math. Soc. 127
(1999), 76105.
[2] D. Alpay and V. Bolotnikov, Two-sided interpolation for matrix functions with en-
tries in the Hardy space, Linear Algebra Appl., 223/224 (1995), 3156.
[3] D. Alpay, V. Bolotnikov, and Ph. Loubaton, On two-sided residue interpolation for
matrix-valued H
2
-functions with symmetries, J. Math. Anal. and Appl 200 (1996),
76105.
[4] D. Alpay, V. Bolotnikov, and Ph. Loubaton, An interpolation with symmetry and
related questions, Zeitschrift f ur Analysis und ihrer Anwand., 15 (1996), 1929.
[5] D. Alpay, V. Bolotnikov, and Ph. Loubaton, On interpolation for Hardy functions
in a certain class of domains under moment type constraints, Houston J. of Math.
23 (1997), 539571.
[6] D. Alpay, V. Bolotnikov, and Ph. Loubaton, On a new positive extension problem
for block Toeplitz matrices, Linear Algebra Appl. 268 (1998), 247287.
[7] D. Alpay, V. Bolotnikov, and L. Rodman, Tangential interpolation with symmetries
and two-point interpolation problem for matrix-valued H
2
-functions, Integral Equa-
tions and Operator Theory 32 (1998), 128.
[8] D. Alpay, V. Bolotnikov, and L. Rodman, Two-sided tangential interpolation for
Hilbert-Schmidt operator functions on polydisks, Operator Theory: Advances and
Applications 124 (2001), 6387.
[9] D. Alpay, V. Bolotnikov, and L. Rodman, One-sided tangential interpolation for
operator-valued Hardy functions on polydisks, Integral Equations and Operator The-
ory 35 (1999), 253270.
[10] D. Alpay, V. Bolotnikov, and L. Rodman, Two-sided residue interpolation in matrix
H
2
spaces with symmetries: conformal conjugate involutions, Linear Algebra Appl.
351/352 (2002), 2768.
[11] J.A. Ball, I. Gohberg, and L. Rodman, Interpolation of Rational Matrix Functions,
Birkhauser Verlag, Basel, 1990.
[12] J.A. Ball, I. Gohberg, and L. Rodman, Two sided tangential interpolation of
real rational matrix functions, Operator Theory: Advances and Applications, 64,
Birkhauser Verlag, Basel, 1993, 73102.
[13] J.A. Ball and J. Kim, Bitangential interpolation problems for symmetric rational
matrix functions, Linear Algebra and Appl. 241/243 (1996), 133152.
[14] H. Dym, J Contractive Matrix Functions, Reproducing Kernel Hilbert Spaces and
Interpolation, CBMS Reg. Conf. Ser. in Math, Vol. 71, Amer. Math. Soc., 1989.
[15] C. Foias and A.E. Frazho, The Commutant Lifting Approach to Interpolation Prob-
lems, Birkhauser Verlag, Basel, 1990.
[16] A. A. Nudelman, A new problem of the type of the moment problem, Soviet Math.
Dokl. 18 (1977), 792795.
M.C.B. Reurings
College of William and Mary
Williamsburg VA 23187-8795, USA
e-mail: mcreurin@math.wm.edu
L. Rodman
College of William and Mary
Williamsburg VA 23187-8795, USA
e-mail: lxrodm@math.wm.edu
Operator Theory:
Favards Theorem Modulo an Ideal
Franciszek Hugon Szafraniec
Abstract. Recently a treatise [2] dealing with the three term recurrence re-
lation for polynomials orthogonal on algebraic sets was born. Because the
paper is pretty sizeable in volume and rich in essence a kind of assistance in
the matter as well as some invitation to it has become a need.
Mathematics Subject Classication (2000). Primary 42C05, 47B25; Secondary
47B15.
Keywords. Polynomials in several variables, orthogonal polynomials, three
term recurrence relation, Favards theorem, ideal of polynomials, algebraic
set, Zariskis topology, symmetric operator, selfadjoint operator, joint spec-
tral measure, product polynomials, Krawchouk polynomials, Charlier poly-
nomials.
The ideology
The celebrated three term recurrence relation for polynomials p
n
n=0
orthogonal
on the real line is the formula
Xp
n
=
n
p
n+1
+
n
p
n
+
n
p
n1
, n = 0, 1, . . . ,
which comes out from applying orthogonality to the fact that the sequence
p
n
n=0
forms a basis (cf. the Glossary section for this as well as for other items)
ordered by deg p
n
= n. Another form of this formula (apparently not equivalent,
in general), appearing so often in the literature,
p
n+1
= (A
n
X +B
n
)p
n
+C
n
p
n1
, n = 0, 1, . . . ,
which, in contrast to the previous one, is ready to perform the recurrence, requires
the meaning of orthogonality to be more decisive. What usually authors do on
this occasion is they assume
the measure realizing orthogonality has innite support,
This work was supported by the KBN grant 2 P03A 037 024.
302 F.H. Szafraniec
which leaves the other case out of interest. This is not a great oence provided
one can avoid any of the two following events:
1
o
demanding some nite sets of polynomials to be orthogonal,
2
o
creating a logical inconsistency in the so called Favards theorem.
The above sins appear in a number of publications dealing with the general theory
of orthogonal polynomials on the real line (and on the unit circle too). Not to men-
tion the sinners an example of a fairly acceptable approach which sees the problem
and tries not to fall into those discrepancies is [1]. Regardless of someone neglects
measures having nite support or does not, the problem becomes of essential sig-
nicance if orthogonality of polynomials in several variables is treated: one can
not just simply say supports like spheres, for instance, are of minor signicance.
A substantial part of the paper [2] deals with quasi-orthogonality, which
elucidates the appearing algebraic issues in the main and enables us to pass to
orthogonality in a rather smooth way. This is done in the second part of [2] and
here we oer excerpts from that.
A cross-section of [2]
Bases with respect to an ideal
Let V be a proper ideal in T
d
. Denote by T
d
/V the quotient algebra (i.e., T
d
/V
is the algebra of all cosets p +V , p T
d
) and by
V
: T
d
T
d
/V the quotient
mapping (i.e.,
V
(p)
def
= p+V , p T
d
). It will be convenient to extend the equality
of two (scalar) polynomials modulo the ideal V to matrix polynomials. Given
two matrix polynomials P = [p
kl
]
m
k=1
n
l=1
and Q = [q
kl
]
m
k=1
n
l=1
, we write P
V
=Q if
p
kl
q
kl
V for all k, l. A set B T
d
is said to be linearly V -independent, if
V
(B) is a linearly independent subset of T
d
/V and
V
[
B
is injective. We say
that F T
d
is a linear V -span of B, if B F and
V
(F) = lin
V
(B). Finally,
B is said to be a (linear) V -basis of F, if B is linearly V -independent and F is a
linear V -span of B. Clearly, B is a V -basis of F if and only if B F,
V
(B) is a
basis of
V
(F) and
V
[
B
is injective.
Set
d
V
(k) =
_
_
_
dim
V
_
T
0]
d
_
= 1 for k = 0,
dim
V
_
T
k]
d
_
dim
V
_
T
k1]
d
_
for k 1,
and
V
= supj 0 : d
V
(j) ,= 0 N ;
notice that
0]
= and
d
0]
(k) = card N
d
: [[ = k =
_
k +d 1
k
_
, k 0.
We say that a sequence Y
k
n
k=0
(0 n ) of column polynomials is a
column representation of a non-empty subset B of T
d
if every element of B is
Favards Theorem Modulo an Ideal 303
an entry of exactly one column Y
i
and for every k = 0, . . . , n, entries of Y
k
are
pairwise distinct elements of B. Thus, in particular, we can identify V -bases with
their column representations: a sequence Y
k
n
k=0
of column polynomials is called
a V -basis of F if Y
k
n
k=0
is a column representation of a V -basis of F.
After the above modication we say sequence P
k
V
k=0
of column polynomials
is a rigid V -basis of T
d
, if P
k
V
k=0
is a column representation of a V -basis of T
d
such that for every k 0,
V
, P
k
T
k]
d
and (P
k
) = d
V
(k). If P
k
V
k=0
is a rigid V -
basis of T
d
, then, by Propositions 6 of [2], for every k 0,
V
,
k
i=0
P
i
is a V -basis
of T
k]
d
and the degree of each member of P
k
is equal to k. Moreover, if P
k
V
k=0
and
Q
k
V
k=0
are rigid V -basis of T
d
and
V
i=0
P
i
=
V
i=0
Q
i
, then for every k 0,
V
,
the columns P
k
and Q
k
are identical up to an arrangement of entries.
Having all these purposeful denitions done we may start to build up (a
selected fragment of) the theory.
Orthogonalization
Given a linear functional L : T
d
C, we dene the set
1
L
def
=
q1
d
p T
d
; L(pq) = 0.
It is clear that 1
L
is an ideal in T
d
such that 1
L
ker L. The latter inclusion and
the denition of 1
L
imply that 1
L
is the greatest ideal contained in ker L, and
that 1
L
is a proper ideal if and only if L is non-zero. If L is a Hermitian linear
functional (that is if L(p
) = L(p) for all p T

d
), then 1
L
is a -ideal.
A sequence Q
k
n
k=0
(nite or not) of column polynomials is said to be L-
orthonormal if L(Q
i
Q
j
) = 0 for all i ,= j, and L(Q
k
Q
k
) is the identity matrix
for every k 0, n. Notice that each L-orthonormal sequence Q
k
n
k=0
is linearly
V -independent for any ideal V contained in 1
L
.
Because measures orthogonalizing polynomials in several variables are not
always available positive denite functionals serve as substitutes for them. So a
linear functional L : T
d
C is said to be positive denite if L(p
p) 0 for
every p T
d
. Applying the Cauchy-Schwarz inequality to the semi-inner product
(p, q) L(pq
) on T
d
, we get
1
L
= p T
d
; L(pp
) = 0.
The following relates bases modulo an ideal to positive deniteness of the orthog-
onalizing functional.
Proposition 1 ([2], Proposition 32). If L : T
d
C is a non-zero linear functional,
then the following conditions are equivalent
(i) L is positive denite,
(ii) 1
L
is a -ideal and there is a rigid 1
L
-basis of T
d
, which is L-orthonormal,
(iii) 1
L
is a -ideal and there is a 1
L
-basis of T
d
, which is L-orthonormal,
(iv) there is a basis B of T
d
such that L(pp
) 0, 1 and L(qr
) = 0 for all
p, q, r B such that q ,= r.
304 F.H. Szafraniec
If (iv) holds we call B semi-orthonormal with respect to L (because the map
(p, q) L(pq
) is a semi-inner product on T
d
). Notice the dierence between
orthonormality of column polynomials and semi-orthonormality of a basis; though
they merge in Proposition 1 they concern dierent notions.
The three term relation and Favards theorem
Consider the following two conditions:
(A) Q
k
n
k=0
is a rigid V -basis of T
d
, which is L-orthonormal, and V ker L;
(B) Q
k
n
k=0
is such that for every k 0, n, (Q
k
) d
V
and deg Q
k
k and
there exists a system [A
k,j
, B
k,j
]
V
k=0
d
j=1
of scalar matrices for which, with
convention A
1,j
def
= 1 and Q
1
def
= 0, the following relation holds
X
j
Q
k
V
=A
k,j
Q
k+1
+B
k,j
Q
k
+A
k1,j
Q
k1
, j = 1, . . . , d, k 0,
V
; (3tr)
in the case
V
< + take Q
V
+1
def
= 0.
Notice that if (A) holds for some linear functional L on T
d
then apparently
L[
V
= 0, L(Q
0
) = Q
0
and L(Q
k
) = 0 for all k 1,
V
(1)
and, in view of Proposition 1, L is positive denite.
Now we can restate Theorem 36 of [2] as follows
Theorem 2. Let V be a proper -ideal in T
d
and Q
k
n
k=0
be a sequence, with n
nite or not, of real column polynomials such that Q
0
= 1. Then
1
o
(A) for some L : T
d
C implies
V
= n, V = 1
L
and (B) to hold;
2
o
Conversely, (B) with
V
= n implies the matrix [A
k,1
, . . . , A
k,d
]
to be injec-
tive, for every k N in the case
V
= or for every k = 0, 1, . . . ,
V
1 oth-
erwise, and to exist L such that (A) holds, which must necessarily be unique,
positive denite and satisfy (1).
We ought to notice that the matrices [A
k,j
] and [B
k,j
] are of the size which
is suitable to perform the operations on the column polynomials Q
k
, which are
required in (3tr).
Remark. Theorem 2 is stated in a slightly dierent way than the corresponding
result of [2]. In particular, we make more explicit the existence of L as it is in
the classical Favards theorem; the existence is a matter of Step 3 in the proof of
Theorem 16 therein and it relies upon the fact that, under the circumstances of
Theorem 2,
T
d
= V lin
_
i=0
Q
i
.
This condition determines L satisfying (1) uniquely.
Remark. The three term relation (3tr) contemplated here is in its symmetric form
which is ready to apply operator methods; the not necessarily symmetric case is
consider in full detail in Sections 5 and 7 of [2].
Set ideals and the three term relation
For p T
d
we dene as usually Z
p
def
= x R
d
; p(x) = 0. Then the set ideal
induced by R
d
1()
def
= p T
d
; Z
p
, R
d
;
is a -ideal; warning 1(Z
X
2 ) ,= (X
2
) in T
1
. Moreover, 1() = 1(
z
), where
z def
=
_
Z
p
: p T
d
and Z
p
_
is the Zariski closure of .
As referring to the quotient construction we have, cf. [2], for Z
p
,=
dimT
d
/1(Z
p
) =
I(Z
p
)
k=0
d
1(
p
)
(k) = cardZ
p
. (2)
Notice that for d = 1, it follows from (2) that
1(
p
)
+ 1 = cardZ
p
.
For a positive Borel measure on R
d
, ,= 0, dene the moment functional L
as
L
(p)
def
=
_
R
d
p d, p T
d
.
Then, cf. [2, Proposition 38]
1
L
= 1(supp ) = 1(supp
z
), (3)
and, by (2),
supp nite if and only if
\
L
< , equivalently: dimT

d
/1
L
< . (4)
The following shows that set ideals are the proper to deal with in the case when
orthogonalization is in terms of measure rather than by functionals.
Proposition 3 ([2], Proposition 41). Let V be a proper -ideal in T
d
and let R
d
be non-empty.
(i) If a sequence Q
k
V
k=0
of real column polynomials (with Q
0
= 1) satises the
condition (B) of Theorem 2 with
V
= n, and L dened by (1) is a moment
functional induced by a measure , then V = 1(supp ).
(ii) If V = 1(), then there exists a rigid V -basis Q
k
V
k=0
of T
d
composed of real
column polynomials (with Q
0
= 1), orthonormalized by some measure (with
supp =

) and satisfying the condition (B) of Theorem 2 with
V
= n.
Thus as far as properties modulo a set ideal are concerned (and this happens
when we want to think of orthogonality with respect to a measure, cf. Proposition
3 below) it is enough to restrict ourselves to supports being Zariski closed.
On the existence of orthogonalizing measures; the multiplication operators
A way to get orthonormalizing measures is to use spectral theorem for the multi-
plication operators. It goes in our modulo an ideal situation much like usually;
let us describe it briey. Assume that V is a proper -ideal in T
d
and Q
k
V
k=0
is
a sequence of real column polynomials (with Q
0
= 1) satisfying the condition (B)
of Theorem 2. By this theorem, the sequence Q
k
V
k=0
and the positive denite
306 F.H. Szafraniec
linear functional L dened by (1) full the condition (A) of Theorem 2. Then the
space T
d
/V is equipped with the inner product , )
L
given by
T
d
/V T
d
/V (q +V, r +V ) q +V, r +V )
L
def
= L(qr
) C.
Dene the multiplication operators M
X
1
, . . . , M
X
d
on T
d
/V via M
X
j
(q + V ) =
X
j
q + V for q T
d
and j 1, d. It is easily seen that (M
X
1
, . . . , M
X
d
)
L
#
s
(T
d
/V )
d
is a cyclic commuting d-tuple with the cyclic vector X
0
+ V , where
T
d
/V is equipped with the inner product , )
L
. If the d-tuple (M
X
1
, . . . , M
X
d
)
has an extension to an d-tuple T = (T
1
, . . . , T
d
) of spectrally commuting selfad-
joint operators acting possibly in a larger Hilbert space, then the functional L is
induced by the measure () = E()(X
0
+ V ), X
0
+ V ), where E stands for the
joint spectral measure of T. Hence orthonormalizes the sequence Q
k
V
k=0
and,
by Theorem 2, V = 1
L
. The following proposition sheds more light on this.

Proposition 4. Let V be a proper -ideal in T
d
, L : T
d
C be a linear functional
such that V ker L, and Q
k
V
k=0
be an L-orthonormal sequence of real column
polynomials (with Q
0
= 1), which is a rigid V -basis of T
d
. Then for every p T
d
,
(p) V if and only if p(M
X
1
, . . . , M
X
d
) = 0.
Orthogonality on the real line revised
Favards theorem
The three term recurrence relation is always meant as
xp
n
(x) =
n
p
n+1
(x) +
n
p
n
(x) +
n
p
n1
(x) (5)
to hold for all x R. Critical values of one of the coecients
n
or
n
cause
dierent kind of trouble. If some
n
= 0 and no
i
= 0 as i = 0, 1, . . . , n 1,
there is no functional establishing orthogonality of p
n
n=0
, cf. (9); if L existed,
L(p
n
Xp
n1
) = 0, this would violate the fact that
n1
= 0. On the other hand,
if some
n
= 0, the same happens with L(Xp
n
p
n+1
) = 0, like in (8), also the
recurrence can not be performed anymore and even worse the relation for that
particular n is no longer an equality between involved polynomials.
The symmetric form of (5)
xp
n
(x) = a
n
p
n+1
(x) +b
n
p
n
(x) +a
n1
p
n1
(x),
which is now controlled by one sequence a
n
n=0
of parameters, has the same
weak point as before if some a
N
= 0. Nevertheless, the exceptional case of Favards
theorem is included in the following as well.
Corollary 5 (Favards Theorem, complete version). Let p be in T
1
and p
k
N
k=0

T
1
be a sequence, with N nite or not, such that deg p
n
= n and p
0
= 1. Then
1
o
if p
k
N
k=0
is a (p)-basis of T
1
and L(p
m
p
n
) =
m,n
, m, n 1, N, for some
positive functional L such that Z
p
ker L, then N = cardZ
p
1, (p) =
q T
1
; L([q[
2
) = 0 and for n 0, N
xp
n
(x) = a
n
p
n+1
(x) +b
n
p
n
(x) +a
n1
p
n1
(x), x Z
p
,
a
1
def
= 1, p
1
def
= 0 and p
N+1
= 0 if N < +;
(6)
2
o
Conversely, if a sequence p
k
N
k=0
with N = cardZ
p
1 satises the relation
(6) then a
n
,= 0 for all n if N = + and for n 0, N 1 otherwise, and
there exists a unique positive denite functional L such that p
k
N
k=0
is a
(p)-basis of T
1
and L(p
m
p
n
) =
m,n
, m, n 1, N.
It is clear that a measure which represents the functional L has nite support
if and only if p ,= 0; if this happens the measure is unique and cardinality of its
support is equal to cardZ
p
. Favards theorem, as commonly stated, leaves that
case aside.
Krawtchouk polynomials
With 0 < p < 1 and N N the Krawtchouk polynomials K
n
( ; p, N) are usually
dened from 0 up to N by
K
n
(x; p, N)
def
=
2
F
1
_
n, x
N
1
p
_
def
=
k=0
(n)
(k)
(x)
(k)
(N)
(k)
(
1
p
)
k
k!
, (7)
where (a)
(0)
def
= 1 and (a)
(k)
def
= (a)
(k1)
(a+k1) is the Pochhammer symbol. They
satisfy the three term recurrence relation (having xed the parameters p and N
set K
n
def
= K
n
( ; p, N))
XK
n
+p(N n)K
n+1
(p(N n) + (1 p)n)K
n
+ (1 p)nK
n1
= 0. (8)
Inserting in (7)
K
n
= ((N)
(n)
p
n
)
1
k
n
,
after making proper cancellation, we can dene for all n the polynomials k
n
as
k
n
def
=
n
i=0
(N +i)
(ni)
(n)
(i)
(X)
(i)
p
ni
i!
and derive the recurrence relation which now takes the form, cf. [3],
Xk
n
k
n+1
(p(N n) +n(1 p))k
n
np(1 p)(N + 1 n)k
n1
= 0. (9)
Dening the functional L by
L(p)
def
=
N
x=0
_
N
x
_
p
x
(1 p)
Nx
p(x), p T
1
we come to the relation (again for all n)
L(k
m
k
n
) = (1)
n
n!(N)
(n)
(p(1 p))
n
m,n
which says the polynomials k
n
N
n=0
are orthogonal (and so are K
n
N
n=0
).
The normalization
K
n
def
=
_
(1)
n
n!(N)
(n)
(p(1 p))
n
_
1/2
k
n
, n = 0, 1, . . . , N (10)
308 F.H. Szafraniec
turns (9) into the symmetric form (with

K
1
def
= 0)
X

K
n
+
_
p(1 p)(n + 1)(N n)

K
n+1
(p(N n) +n(1 p))

K
n
+
_
p(1 p)n(N n + 1)

K
n1
= 0. (11)
and this can be carried out for 0 n N exclusively. However (11) does not
hold
1
for n = N (compute its left-hand side at x = N +1, for instance) and both
(8) and (9) do not hold either. Nevertheless the left-hand side of (11) is in the ideal
((X)
(N+1)
) which is the right excuse to include Krawtchouk polynomials into our
general framework. Thus all of it and Theorem 2 in particular can be illustrated
in this negligible case. More precisely, referring to the notation so far we have:
L dened by (8), 1
L
= ((X)
(N+1)
),
V
= N, Q
i
=

K
i
for i = 0, 1, . . . , N + 1
(notice it follows from (10) automatically all

K
i
= 0 for i N+1) and the relation
(11) to hold modulo the ideal ((X)
(N+1)
).
Notice that deg k
n
= n and the leading coecient is
(1)
n
n!
for all n; k
n

((X)
(N+1)
) as n N + 1, in particular k
N+1
=
(X)
(N+1)
N!
. Thus the semi-
orthogonal basis B appearing in Proposition 1, (iv) can be chosen as

K
n
if n =
0, 1, . . . , N and k
n
otherwise.
An illustrative though simple case
Product polynomials
By product polynomials in this context we mean roughly orthogonal polynomials
which are tensor product of other orthogonal polynomials. Let us be more precise
and describe that in the case d = 2. So for i = 1, 2 let p
(i)
n

n=0
be a sequence
of orthonormal polynomials with respect to a Borel measure
i
, or, equivalently,
with respect to a functional L
i
and let
Xp
(i)
n
= a
(i)
n
p
(i)
n+1
+b
(i)
n
p
(i)
n
+a
(i)
n1
p
(i)
n1
be its (formal) three term recurrence relation. The measure
def
=
1
2
corresponds
to the functional L extended from L(p
(1)
p
(2)
)
def
= L
1
(p
(1)
)L
2
(p
(2)
) for p
(1)
C[X
1
]
and p
(2)
C[X
2
]. It is clear that the polynomials p
m,n
def
= p
(1)
n
p
(2)
n
are orthonormal
with respect to
1
2
or, equivalently, with respect to L. Grouping p
m,n
m,n=0
according to their degree as
Q
n
def
= [p
(1)
n
p
(2)
0
, p
(1)
n1
p
(2)
1
, . . . , p
(1)
0
p
(2)
n
]
we come to the (still formal) three term recurrence relation

X
i
Q
n
= A
n,i
Q
n+1
+B
n,i
Q
n
+A
n1,i
Q
n1
, (12)
1
This troublesome case untouched in [3], what is a contemporary lexicon on orthogonal polyno-
mials, is left generously to the reader to be discovered and to be handled somehow.
with
A
n,1
def
=
_
_
a
(1)
n
0 0
.
.
.
.
.
.
.
.
.
.
.
.
0 a
(1)
0
0
_
_
, B
n,1
def
=
_
_
b
(1)
n
0
.
.
.
.
.
.
.
.
.
0 b
(1)
0
_
_
,
A
n,2
def
=
_
_
a
(2)
0
0 0
.
.
.
.
.
.
.
.
.
.
.
.
0 a
(2)
n
0
_
_
, B
n,2
def
=
_
_
b
(2)
0
0
.
.
.
.
.
.
.
.
.
0 b
(2)
n
_
_
.
Now the question is to determine an ideal, trivial or not, with respect to which the
recurrence relation (12) holds. Due to (3), we can consider set ideals determined
by algebraic set which are Zariski closed. For d = 1 the only possibility for this is
either the set is nite or equal to R. Thus we have got four cases two of which give
us an ideal which is not trivial: both measures are nitely supported or one of the
measures is nitely supported, the other not; the latter is interesting enough to be
discussed below in some detail.
The example
To discuss the case just mentioned take two sequences of orthonormal polynomials:
the Charlier polynomials
C
n
n=0
and the Krawtchouk polynomials
K
n
n=0
.
The Charlier polynomials C
(a)
n

n=0
, a > 0, are determined by
e
az
(1 +z)
x
=
n=0
C
(a)
n
(x)
z
n
n!
.
They are orthogonal with respect to a non-negative integer supported measure
according to
x=0
C
(a)
m
(x)C
(a)
n
(x)
e
a
a
x
x!
=
mn
a
n
n!, m, n = 0, 1, . . . .
The orthonormalized Charlier polynomials C
(a)
n

n=0
are
C
(a)
n
(x) = a
n
2
(n!)
1
2
C
(a)
n
(x)
and they satisfy the recurrence relation
X

C
(a)
n
=
_
(n + 1)a

C
(a)
n+1
+ (n +a)

C
(a)
n
+
na

C
(a)
n
.
Both Charlier and Krawtchouk polynomials have probabilistic connections (Pois-
son and binomial distribution), for more look at [4]. If we build product poly-
nomials with these two we get polynomials in two variables having the following
interesting features:
the support of the orthogonality measure is a countable subset N0, 1, . . . ,
. . . , N of the integer lattice;
the support is not Zariski closed, its Zariski closure is R 0, 1, . . . , N;
1
L
= ((X
2
)
(N)
) and
V
= +;
310 F.H. Szafraniec
the three term recurrence relation (3tr) modulo the ideal 1
L
is of full (i.e.,
innite) length;
the length of Q
n
stabilizes after n = N + 1.
Let us mention that all the details as well as extensive references to the literature
can be found in [2].
Glossary
N
def
= 0, 1, . . . and N
d
d - fold Cartesian product of N by itself;
for i N i, j
def
= i, i + 1, . . . , j if i j < + and i, j
def
= i, i + 1, . . . if j = +;
R (resp. C) stands for the eld of all real (resp. complex) numbers;
lin A the linear span of a subset A of a linear space;
basis means a Hamel basis;
i,j
the Kronecker symbol;
X
def
= X
1
1
X
d
d
for = (
1
, . . . ,
d
) N
d
;
T
d
the algebra of all polynomials in d commuting indeterminates X
1
, . . . , X
d
with complex coecients;
the unique involution in the algebra T

d
such that X
i
= X
i
for all i = 1, . . . , d;
-ideal an ideal invariant under the involution

;
deg p the degree of p T
d
;
(Q) the length (that is, the number of entries) of a column polynomial Q;
T
k]
d
def
= p T
d
: deg p k, k N;
(p
1
, . . . , p
n
) the ideal generated by a set p
1
, . . . , p
n
T
d
;
L
#
s
(T) the set of all linear operators A on an inner product space T such that
Ax, y) = x, A), x, y T.
References
[1] T.S. Chihara, An introduction to orthogonal polynomials, in Mathematics and its
Applications, vol. 13, Gordon and Breach, New York, 1978.
[2] D. Cicho n, J. Stochel and F.H. Szafraniec, Three term recurrence relation modulo
an ideal and orthogonality of polynomials of several variables, J. Approx. Theory, to
appear.
[3] R. Koekoek and R.F. Swarttouw, The Askey-scheme of hypergeometric orthogonal
polynomials and its q-analogue, Delft University of Technology, Report of the Faculty
of Technical Mathematics and Informatics no. 94-05.
[4] W. Schoutens, Stochastic processes and orthogonal polynomials, in Lecture Notes in
Statistics, vol. 145, Springer-Verlag, New York Berlin Heidelberg, 2000.
Franciszek Hugon Szafraniec
Instytut Matematyki, Uniwersytet Jagiello nski
ul. Reymonta 4, PL-30059 Krakow, Poland
e-mail: fhszafra@im.uj.edu.pl

Operator Theory Systems Theory and Scattering Theory Multidimensional Generalizations

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Operator Theory Systems Theory and Scattering Theory Multidimensional Generalizations

Uploaded by

Copyright:

Available Formats

Operator Theory: Advances and

algebras. III. Multivariable operator theory. Acta

berg, V. V. Peller, and D. V. Yakubovich. A brief excursion into the theory

-algebra C(T). Given any group G or C

with common cyclic subspace c contained in both / and /

are unitarily equivalent to |

with common -cyclic subspace c contained in

, then there is a unitary operator U : / /

are unitarily equivalent to the model row-

W(z, ), a formal power series

, namely, the space of in-

. Hence is isometric from a dense subset of the orthogonal complement of

follows. This completes the proof of Theorem 2.2.

and the rst assertion follows.

and (2.19) follows for W

= 0 is to be taken before the multiplication with k

W(z, )p(z). (2.34)

e, and formula (2.32) now follows by linearity. The

(z, ) = K(, z) is a pos-

W(z, ) = Y ()Y (z)

, but now with 1 equal to a row-unitary d-tuple. Hence

from which we see that

any other coisometry from / onto L

e) for all e c, then W

is another normalized Cuntz weight on c for which there

e for e c for all e c and with

in fact is isometric on c, and hence also on

as claimed. This completes the proof of Theorem

on Hilbert spaces c and c

respectively are unitarily equivalent

with cyclic subspace c

, then there is a unitary transformation

e) simply with W(z

be two Cuntz weights (or more generally positive semidenite

respectively. Let us say that a bounded operator S: L

with the additional

is an analytic intertwining operator. Note that for d = 1

) of contractive, analytic, operator-valued functions on the

[T(z)e] for each e c. (3.1)

) and their associated symbols T(z), and understand the converse

)? Two particular situations where this

is a general positive Haplitz operator and T(z) is an analytic

is a Haplitz extension of the identity and T(z)e L

) for each e c (so

). Here we say that the Haplitz

is a Haplitz extension of the identity if W

) and from L(T

), as only nite sums are involved in the calculation of a given coecient

. Since any f T(T

is a positive semidenite Haplitz operator on c

is general and T(z) is a polynomial, and (ii) when W

; the following proposition

be positive semidenite Haplitz operators with ma-

respectively, let T(z) =

if and only if the Haplitz operator WL

has the additional

(z, )T(z) + T()

is a []-Haplitz extension of the identity, formulas (3.10), (3.11)

(z, )T(z), (3.15)

(z, )T(z) +T()

(z, ) then follows by applying the general formula (2.43) to W

(z, )T(z) +T()

(z, )T(z) +T()

and we are left with the