Download the full version of the ebook now at ebookultra.

com

Linear Systems and Control An Operator

Perspective 1st Edition Martin J. Corless &
Arthur E. Frazho

https://ebookultra.com/download/linear-systems-
and-control-an-operator-perspective-1st-edition-
martin-j-corless-arthur-e-frazho/

Explore and download more ebook at https://ebookultra.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

An operator perspective on signals and systems 1st

Edition. Edition Arthur Frazho

https://ebookultra.com/download/an-operator-perspective-on-signals-
and-systems-1st-edition-edition-arthur-frazho/

ebookultra.com

Applied Linear Optimal Control Examples and Algorithms 1st

Edition Arthur E. Bryson

https://ebookultra.com/download/applied-linear-optimal-control-
examples-and-algorithms-1st-edition-arthur-e-bryson/

ebookultra.com

Linear Multivariable Control Systems Shankar P.

Bhattacharyya

https://ebookultra.com/download/linear-multivariable-control-systems-
shankar-p-bhattacharyya/

ebookultra.com

Predictive Control for Linear and Hybrid Systems 1st

Edition Francesco Borrelli

https://ebookultra.com/download/predictive-control-for-linear-and-
hybrid-systems-1st-edition-francesco-borrelli/

ebookultra.com
Linear Control Systems with Matlab Applications 12th
Edition B.S. Manke

https://ebookultra.com/download/linear-control-systems-with-matlab-
applications-12th-edition-b-s-manke/

ebookultra.com

Reliable Control and Filtering of Linear Systems with

Adaptive Mechanisms 1st Edition Guang-Hong Yang

https://ebookultra.com/download/reliable-control-and-filtering-of-
linear-systems-with-adaptive-mechanisms-1st-edition-guang-hong-yang/

ebookultra.com

Linear Algebra Linear Systems and Eigenvalues P.M. Van

Dooren

https://ebookultra.com/download/linear-algebra-linear-systems-and-
eigenvalues-p-m-van-dooren/

ebookultra.com

Communication Networks An Optimization Control and

Stochastic Networks Perspective Srikant R.

https://ebookultra.com/download/communication-networks-an-
optimization-control-and-stochastic-networks-perspective-srikant-r/

ebookultra.com

Advanced manufacturing an ICT and systems perspective 1st

Edition Marco Taisch

https://ebookultra.com/download/advanced-manufacturing-an-ict-and-
systems-perspective-1st-edition-marco-taisch/

ebookultra.com
Linear Systems and Control An Operator Perspective 1st
Edition Martin J. Corless & Arthur E. Frazho Digital
Instant Download
Author(s): MARTIN J. CORLESS & ARTHUR E. FRAZHO
ISBN(s): 9780824707293, 082470729X
Edition: 1
File Details: PDF, 14.68 MB
Year: 2003
Language: english
LINEAR SYSTEMS AND
CONTROL
An Operator Perspective

MARTIN J. CORLESS
ARTHUR E. FRAZHO
Purdue University
West Lafayette, Indiana, U.S.A.

MARCEL

MARCEL DEKKER, INC. NEW YORK • BASEL

i
DEKKER *
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the Library of Congress.

ISBN: 0-8247-0729-X

This book is printed on acid-free paper.

Headquarters
Marcel Dekker, Inc.
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540

Eastern Hemisphere Distribution

Marcel Dekker AG
Hutgasse4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-260-6300; fax: 41-61-260-6333

World Wide Web

http://www.dekker.com

The publisher offers discounts on this book when ordered in bulk quantities. For more infor-
mation, write to Special Sales/Professional Marketing at the headquarters address above.

Copyright © 2003 by Marcel Dekker, Inc. All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, microfilming, and recording, or
by any information storage and retrieval system, without permission in writing from the
publisher.

Current printing (last digit):

10 987 6 5 4 3 2 1

PRINTED IN THE UNITED STATES OF AMERICA

 PURE AND APPLIED MATHEMATICS

A Program of Monographs, Textbooks, and Lecture Notes

EXECUTIVE EDITORS

Earl J. Taft Zuhair Nashed

Rutgers University University of Central Florida
New Brunswick, New Jersey Orlando, Florida

EDITORIAL BOARD

M. S. Baouendi AnilNerode
University of California, Cornell University
San Diego
Donald Passman
Jane Cronin University of Wisconsin,
Rutgers University Madison

Jack K. Hale Fred S. Roberts

Georgia Institute of Technology Rutgers University

S. Kobayashi David L. Russell

University of California, Virginia Polytechnic Institute
Berkeley and State University

Marvin Marcus Walter Schempp

University of California, Universitdt Siegen
Santa Barbara
Mark Teply
W. S. Massey University of Wisconsin,
Yale University Milwaukee
 MONOGRAPHS AND TEXTBOOKS IN
PURE AND APPLIED MATHEMATICS

1. K. Yano, Integral Formulas in Riemannian Geometry (1970)

2. S. Kobayashi, Hyperbolic Manifolds and Holomorphic Mappings (1970)
3. V. S. Vladimirov, Equations of Mathematical Physics (A. Jeffrey, ed.; A. Littlewood,
trans.) (1970)
4. 6. N. Pshenichnyi, Necessary Conditions for an Extremum (L. Neustadt, translation
ed.; K. Makowski, trans.) (1971)
5. L Narici et a/., Functional Analysis and Valuation Theory (1971)
6. S. S. Passman, Infinite Group Rings (1971)
7. L Domhoff, Group Representation Theory. Part A: Ordinary Representation Theory.
Part B: Modular Representation Theory (1971,1972)
8. W. Boothbyand G. L. Weiss, eds., Symmetric Spaces (1972)
9. Y. Matsushima, Differentiate Manifolds (E. T. Kobayashi, trans.) (1972)
10. L. E. Ward, Jr., Topology (1972)
11. A. Babakhanian, Cohomological Methods in Group Theory (1972)
12. R. Gilmer, Multiplicative Ideal Theory (1972)
13. J. Yeh, Stochastic Processes and the Wiener Integral (1973)
14. J. Barros-Neto, Introduction to the Theory of Distributions (1973)
15. R. Larsen, Functional Analysis (1973)
16. K. Yano and S. Ishihara, Tangent and Cotangent Bundles (1973)
17. C. Procesi, Rings with Polynomial Identities (1973)
18. R. Hermann, Geometry, Physics, and Systems (1973)
19. N. R. Wallach, Harmonic Analysis on Homogeneous Spaces (1973)
20. J. Dieudonne, Introduction to the Theory of Formal Groups (1973)
21. /. Vaisman, Cohomology and Differential Forms (1973)
22. 8.-Y. Chen, Geometry of Submanifolds (1973)
23. M. Marcus, Finite Dimensional Multilinear Algebra (in two parts) (1973,1975)
24. R. Larsen, Banach Algebras (1973)
25. R. O. Kujala and A. L. Vitter, eds., Value Distribution Theory: Part A; Part B: Deficit
and Bezout Estimates by Wilhelm Stall (1973)
26. K. B. Stolarsky, Algebraic Numbers and Diophantine Approximation (1974)
27. A. R. Magid, The Separable Galois Theory of Commutative Rings (1974)
28. 8. R. McDonald, Finite Rings with Identity (1974)
29. J. Satake, Linear Algebra (S. Koh et al., trans.) (1975)
30. J. S. Golan, Localization of Noncommutative Rings (1975)
31. G. Klambauer, Mathematical Analysis (1975)
32. M. K. Agoston, Algebraic Topology (1976)
33. K. R. Goodearl, Ring Theory (1976)
34. L. E. Mansfield, Linear Algebra with Geometric Applications (1976)
35. N. J. Pullman, Matrix Theory and Its Applications (1976)
36. 8. R. McDonald, Geometric Algebra Over Local Rings (1976)
37. C. W. Groetsch, Generalized Inverses of Linear Operators (1977)
38. J. £. Kuczkowski and J. L Gersting, Abstract Algebra (1977)
39. C. O. Christenson and W. L. Voxman, Aspects of Topology (1977)
40. M. Nagata, Field Theory (1977)
41. R. L. Long, Algebraic Number Theory (1977)
42. W. F. Pfeffer, Integrals and Measures (1977)
43. R. L. Wheeden and A. Zygmund, Measure and Integral (1977)
44. J. H. Curtiss, Introduction to Functions of a Complex Variable (1978)
45. K. Hrbacek and T. Jech, Introduction to Set Theory (1978)
46. W. S. Massey, Homology and Cohomology Theory (1978)
47. M. Marcus, Introduction to Modern Algebra (1978)
48. E. C. Young, Vector and Tensor Analysis (1978)
49. S. B. Nadler, Jr., Hyperspaces of Sets (1978)
50. S. K. Segal, Topics in Group Kings (1978)
51. A. C. M. van Rooij, Non-Archimedean Functional Analysis (1978)
52. L Corwin and R. Szczarba, Calculus in Vector Spaces (1979)
53. C. Sadosky, Interpolation of Operators and Singular Integrals (1979)
54. J. Cronin, Differential Equations (1980)
55. C. W. Groetsch, Elements of Applicable Functional Analysis (1980)
 56. /. Vaisman, Foundations of Three-Dimensional Euclidean Geometry (1980)
57. H. I. Freedan, Deterministic Mathematical Models in Population Ecology (1980)
58. S. B. Chae, Lebesgue Integration (1980)
59. C. S. Rees etal., Theory and Applications of Fourier Analysis (1981)
60. L. Nachbin, Introduction to Functional Analysis (R. M. Aron, trans.) (1981)
61. G. Orzech and M. Orzech, Plane Algebraic Curves (1981)
62. R. Johnsonbaugh and W. E. Pfaffenberger, Foundations of Mathematical Analysis
(1981)
63. W. L. Voxman and R. H. Goetschel, Advanced Calculus (1981)
64. L. J. Co/win and R. H. Szczarba, Multivariable Calculus (1982)
65. V. I. Istratescu, Introduction to Linear Operator Theory (1981)
66. R. D. Jarvinen, Finite and Infinite Dimensional Linear Spaces (1981)
67. J. K. Beem and P. E. Ehriich, Global Lorentzian Geometry (1981)
68. D. L. Armacost, The Structure of Locally Compact Abelian Groups (1981)
69. J. W. Brewer and M. K. Smith, eds., Emmy Noether: A Tribute (1981)
70. K. H. Kim, Boolean Matrix Theory and Applications (1982)
71. T. W. Wieting, The Mathematical Theory of Chromatic Plane Ornaments (1982)
72. D. B. Gauld, Differential Topology (1982)
73. R. L. Faber, Foundations of Euclidean and Non-Euclidean Geometry (1983)
74. M. Carmeli, Statistical Theory and Random Matrices (1983)
75. J. H. Camith et al., The Theory of Topological Semigroups (1983)
76. R. L Faber, Differential Geometry and Relativity Theory (1983)
77. S. Bamett, Polynomials and Linear Control Systems (1983)
78. G. Karpilovsky, Commutative Group Algebras (1983)
79. F. Van Oystaeyen and A. Verschoren, Relative Invariants of Rings (1983)
80. /. Vaisman, A First Course in Differential Geometry (1984)
81. G. W. Swan, Applications of Optimal Control Theory in Biomedicine (1984)
82. T. Pethe and J. D. Randall, Transformation Groups on Manifolds (1984)
83. K. Goebel and S. Reich, Uniform Convexity, Hyperbolic Geometry, and Nonexpansive
Mappings (1984)
84. T. Albu and C. Nastasescu, Relative Finiteness in Module Theory (1984)
85. K. Hrbacek and T. Jech, Introduction to Set Theory: Second Edition (1984)
86. F. Van Oystaeyen and A. Verschoren, Relative Invariants of Rings (1984)
87. B. R. McDonald, Linear Algebra Over Commutative Rings (1984)
88. M. Namba, Geometry of Projective Algebraic Curves (1984)
89. G. F. Webb, Theory of Nonlinear Age-Dependent Population Dynamics (1985)
90. M. R. Bremner et al., Tables of Dominant Weight Multiplicities for Representations of
Simple Lie Algebras (1985)
91. A. E. Fekete, Real Linear Algebra (1985)
92. S. 6. Chae, Holomorphy and Calculus in Normed Spaces (1985)
93. A. J. Jerri, Introduction to Integral Equations with Applications (1985)
94. G. Karpilovsky, Projective Representations of Finite Groups (1985)
95. L. Narici and E. Beckenstein, Topological Vector Spaces (1985)
96. J. Weeks, The Shape of Space (1985)
97. P. R. Gribik and K. O. Kortanek, Extremal Methods of Operations Research (1985)
98. J.-A. Chao and W. A. Woyczynski, eds., Probability Theory and Harmonic Analysis
(1986)
99. G. D. Crown et al., Abstract Algebra (1986)
100. J. H. Carruth et al., The Theory of Topological Semigroups, Volume 2 (1986)
101. R. S. Doran and V. A. Belfi, Characterizations of C*-Algebras (1986)
102. M. W. Jeter, Mathematical Programming (1986)
103. M. Altman, A Unified Theory of Nonlinear Operator and Evolution Equations with
Applications (1986)
104. A. Verschoren, Relative Invariants of Sheaves (1987)
105. R. A. Usmani, Applied Linear Algebra (1987)
106. P. Blass and J. Lang, Zariski Surfaces and Differential Equations in Characteristic p >
0(1987)
107. J. A. Reneke et al., Structured Hereditary Systems (1987)
108. H. Busemann and B. B. Phadke, Spaces with Distinguished Geodesies (1987)
109. R. Harte, Invertibility and Singularity for Bounded Linear Operators (1988)
110. G. S. Ladde et al., Oscillation Theory of Differential Equations with Deviating Argu-
ments (1987)
111. L. Dudkin et al., Iterative Aggregation Theory (1987)
112. T. Okubo, Differential Geometry (1987)
113. D. L Stand and M. L Stand, Real Analysis with Point-Set Topology (1987)
114. T. C. Gard, Introduction to Stochastic Differential Equations (1988)
115. S. S. Abhyankar, Enumerative Combinatorics of Young Tableaux (1988)
116. H. Strade and R. Famsteiner, Modular Lie Algebras and Their Representations (1988)
117. J. A. Huckaba, Commutative Rings with Zero Divisors (1988)
118. W. D. Wallis, Combinatorial Designs (1988)
119. W. Wiestaw, Topological Fields (1988)
120. G. Karpilovsky, Field Theory (1988)
121. S. Caenepeel and F. Van Oystaeyen, Brauer Groups and the Cohomology of Graded
Rings (1989)
122. W. Kozlowski, Modular Function Spaces (1988)
123. E. Lowen-Colebunders, Function Classes of Cauchy Continuous Maps (1989)
124. M. Pavel, Fundamentals of Pattern Recognition (1989)
125. V. Lakshmikantham et a/., Stability Analysis of Nonlinear Systems (1989)
126. R. Sivaramakn'shnan, The Classical Theory of Arithmetic Functions (1989)
127. N. A. Watson, Parabolic Equations on an Infinite Strip (1989)
128. K. J. Hastings, Introduction to the Mathematics of Operations Research (1989)
129. a Fine, Algebraic Theory of the Bianchi Groups (1989)
130. D. N. Dikranjan et a/., Topological Groups (1989)
131. J.C. Morgan II, Point Set Theory (1990)
132. P. BilerandA. Witkowski, Problems in Mathematical Analysis (1990)
133. H. J. Sussmann, Nonlinear Controllability and Optimal Control (1990)
134. J.-P. Florens et a/., Elements of Bayesian Statistics (1990)
135. N. Shell, Topological Fields and Near Valuations (1990)
136. B. F. Doolin and C. F. Martin, Introduction to Differential Geometry for Engineers
(1990)
137. S. S. Holland, Jr., Applied Analysis by the Hilbert Space Method (1990)
138. J. Okninski, Semigroup Algebras (1990)
139. K. Zhu, Operator Theory in Function Spaces (1990)
140. G. 8. Price, An Introduction to Multicomplex Spaces and Functions (1991)
141. R. B. Darst, Introduction to Linear Programming (1991)
142. P. L Sachdev, Nonlinear Ordinary Differential Equations and Their Applications (1991)
143. T. Husain, Orthogonal Schauder Bases (1991)
144. J. Foran, Fundamentals of Real Analysis (1991)
145. W. C. Brown, Matrices and Vector Spaces (1991)
146. M. M. Rao and Z. D. Ren, Theory of Oriicz Spaces (1991)
147. J. S. Golan and T. Head, Modules and the Structures of Rings (1991)
148. C. Small, Arithmetic of Finite Fields (1991)
149. K. Yang, Complex Algebraic Geometry (1991)
150. D. G. Hoffman et a/., Coding Theory (1991)
151. M. O. Gonzalez, Classical Complex Analysis (1992)
152. M. O. Gonzalez, Complex Analysis (1992)
153. L W. Baggett, Functional Analysis (1992)
154. M. Sniedovich, Dynamic Programming (1992)
155. R. P. Agarwal, Difference Equations and Inequalities (1992)
156. C. Brezinski, Biorthogonality and Its Applications to Numerical Analysis (1992)
157. C. Swartz, An Introduction to Functional Analysis (1992)
158. S. 6. Nadler, Jr., Continuum Theory (1992)
159. M. A. AI-Gwaiz, Theory of Distributions (1992)
160. E. Perry, Geometry: Axiomatic Developments with Problem Solving (1992)
161. E. Castillo and M. R. Ruiz-Cobo, Functional Equations and Modelling in Science and
Engineering (1992)
162. A. J. Jerri, Integral and Discrete Transforms with Applications and Error Analysis
(1992)
163. A. Charlieret a/., Tensors and the Clifford Algebra (1992)
164. P. Bilerand T. Nadzieja, Problems and Examples in Differential Equations (1992)
165. E. Hansen, Global Optimization Using Interval Analysis (1992)
166. S. Guerre-Delabriere, Classical Sequences in Banach Spaces (1992)
167. V. C. Wong, Introductory Theory of Topological Vector Spaces (1992)
168. S. H. Kulkami and B. V. Limaye, Real Function Algebras (1992)
169. W. C. Brown, Matrices Over Commutative Rings (1993)
170. J. Loustau and M. Dillon, Linear Geometry with Computer Graphics (1993)
171. W. V. Petryshyn, Approximation-Solvability of Nonlinear Functional and Differential
Equations (1993)
172. E. C. Young, Vector and Tensor Analysis: Second Edition (1993)
173. T. A. Bick, Elementary Boundary Value Problems (1993)
174. M. Pavel, Fundamentals of Pattern Recognition: Second Edition (1993)
175. S. A. Albeverio et a/., Noncommutative Distributions (1993)
176. W. Fulks, Complex Variables (1993)
177. M. M. Rao, Conditional Measures and Applications (1993)
178. A. Janicki and A. Weron, Simulation and Chaotic Behavior of a-Stable Stochastic
Processes(1994)
179. P. Neittaanmaki and D. Tiba, Optimal Control of Nonlinear Parabolic Systems (1994)
180. J. Cronin, Differential Equations: Introduction and Qualitative Theory, Second Edition
(1994)
181. S. Heikkila and V. Lakshmikantham, Monotone Iterative Techniques for Discontinuous
Nonlinear Differential Equations (1994)
182. X. Mao, Exponential Stability of Stochastic Differential Equations (1994)
183. B. S. Thomson, Symmetric Properties of Real Functions (1994)
184. J. E. Rubio, Optimization and Nonstandard Analysis (1994)
185. J. L Bueso et a/., Compatibility, Stability, and Sheaves (1995)
186. A. N. Michel and K. Wang, Qualitative Theory of Dynamical Systems (1995)
187. M. R. Darnel, Theory of Lattice-Ordered Groups (1995)
188. Z. Naniewicz and P. D. Panagiotopoulos, Mathematical Theory of Hemivariational
Inequalities and Applications (1995)
189. L. J. Corwin and R. H. Szczarba, Calculus in Vector Spaces: Second Edition (1995)
190. L. H. Erbe et a/., Oscillation Theory for Functional Differential Equations (1995)
191. S. Agaian et a/., Binary Polynomial Transforms and Nonlinear Digital Filters (1995)
192. M. I. Gil', Norm Estimations for Operation-Valued Functions and Applications (1995)
193. P. A. Grillet, Semigroups: An Introduction to the Structure Theory (1995)
194. S. Kichenassamy, Nonlinear Wave Equations (1996)
195. V. F. Krotov, Global Methods in Optimal Control Theory (1996)
196. K. I. Beidaret a/., Rings with Generalized Identities (1996)
197. V. I. Amautov et a/., Introduction to the Theory of Topological Rings and Modules
(1996)
198. G. Sierksma, Linear and Integer Programming (1996)
199. R. Lasser, Introduction to Fourier Series (1996)
200. V. Sima, Algorithms for Linear-Quadratic Optimization (1996)
201. D. Redmond, Number Theory (1996)
202. J. K. Beem et a/., Global Lorentzian Geometry: Second Edition (1996)
203. M. Fontana et a/., Prufer Domains (1997)
204. H. Tanabe, Functional Analytic Methods for Partial Differential Equations (1997)
205. C. Q. Zhang, Integer Flows and Cycle Covers of Graphs (1997)
206. E. Spiegel and C. J. O'Donnell, Incidence Algebras (1997)
207. B. Jakubczyk and W. Respondek, Geometry of Feedback and Optimal Control (1998)
208. T. W. Haynes et a/., Fundamentals of Domination in Graphs (1998)
209. T. W. Haynes et a/., eds., Domination in Graphs: Advanced Topics (1998)
210. L. A. D'Alotto et a/., A Unified Signal Algebra Approach to Two-Dimensional Parallel
Digital Signal Processing (1998)
211. F. Halter-Koch, Ideal Systems (1998)
212. N. K. Govil et a/., eds., Approximation Theory (1998)
213. R. Cross, Multivalued Linear Operators (1998)
214. A. A. Martynyuk, Stability by Liapunov's Matrix Function Method with Applications
(1998)
215. A. Favini and A. Yagi, Degenerate Differential Equations in Banach Spaces (1999)
216. A. Illanes and S. Nadler, Jr., Hyperspaces: Fundamentals and Recent Advances
(1999)
217. G. Kato and D. Struppa, Fundamentals of Algebraic Microlocal Analysis (1999)
218. G. X.-Z. Yuan, KKM Theory and Applications in Nonlinear Analysis (1999)
219. D. Motreanu and N. H. Pavel, Tangency, Flow Invariance for Differential Equations,
and Optimization Problems (1999)
220. K. Hrbacek and T. Jech, Introduction to Set Theory, Third Edition (1999)
221. G. E. Kolosov, Optimal Design of Control Systems (1999)
222. N. L. Johnson, Subplane Covered Nets (2000)
223. B. Fine and G. Rosenberger, Algebraic Generalizations of Discrete Groups (1999)
224. M. Vath, Volterra and Integral Equations of Vector Functions (2000)
225. S. S. Miller and P. T. Mocanu, Differential Subordinations (2000)
226. R. Li et a/., Generalized Difference Methods for Differential Equations: Numerical
Analysis of Finite Volume Methods (2000)
227. H. Li and F. Van Oystaeyen, A Primer of Algebraic Geometry (2000)
228. R. P. Agarwal, Difference Equations and Inequalities: Theory, Methods, and Applica-
tions, Second Edition (2000)
229. A. B. Kharazishvili, Strange Functions in Real Analysis (2000)
230. J. M. Appell et a/., Partial Integral Operators and Integra-Differential Equations (2000)
231. A. I. Phlepko et a/., Methods for Solving Inverse Problems in Mathematical Physics
(2000)
232. F. Van Oystaeyen, Algebraic Geometry for Associative Algebras (2000)
233. D. L. Jagerman, Difference Equations with Applications to Queues (2000)
234. D. R. Hankerson et a/., Coding Theory and Cryptography: The Essentials, Second
Edition, Revised and Expanded (2000)
235. S. DascaVescu et a/., Hopf Algebras: An Introduction (2001)
236. R. Hagen et a/., C*-Algebras and Numerical Analysis (2001)
237. Y. Talpaert, Differential Geometry: With Applications to Mechanics and Physics (2001)
238. R. H. Villarreal, Monomial Algebras (2001)
239. A. N. Michel et a/., Qualitative Theory of Dynamical Systems: Second Edition (2001)
240. A. A. Samarskii, The Theory of Difference Schemes (2001)
241. J. Knopfmacher and W.-B. Zhang, Number Theory Arising from Finite Fields (2001)
242. S. Leader, The Kurzweil-Henstock Integral and Its Differentials (2001)
243. M. Biliotti et a/., Foundations of Translation Planes (2001)
244. A. N. Kochubei, Pseudo-Differential Equations and Stochastics over Non-Archimedean
Fields (2001)
245. G. Sierksma, Linear and Integer Programming: Second Edition (2002)
246. A. A. Martynyuk, Qualitative Methods in Nonlinear Dynamics: Novel Approaches to
Liapunov's Matrix Functions (2002)
247. B. G. Pachpatte, Inequalities for Finite Difference Equations (2002)
248. A. N. Michel and D. Liu, Qualitative Analysis and Synthesis of Recurrent Neural
Networks (2002)
249. J. R. Weeks, The Shape of Space: Second Edition (2002)
250. M. M. Rao and 2. D. Ren, Applications of Orticz Spaces (2002)
251. V. Lakshmikantham and D. Trigiante, Theory of Difference Equations: Numerical
Methods and Applications, Second Edition (2002)
252. T. Albu, Cogalois Theory (2003)
253. A. Bezdek, Discrete Geometry (2003)
254. M. J. Confess and A. E. Frazho, Linear Systems and Control: An Operator
Perspective (2003)
255. /. Graham and G. Kohr, Geometric Function Theory in One and Higher Dimensions
(2003)

Additional Volumes in Preparation

To our parents
Preface

This monograph is intended for engineers, scientists, and mathematicians interested in linear
systems and control. It is concerned with the analysis and control of systems whose behavior
is governed by linear ordinary differential equations. Our approach is based mainly on state
space models. These models arise naturally in the description of physical systems. Many of
the classical results in linear systems, including the linear quadratic regulator problem, are
presented. An introduction to H°° analysis and control is also given. The major prerequisite
is a standard graduate-level course in linear algebra. However, on occasion we use some
elementary results from least squares optimization theory and operator theory. These results
are reviewed in the appendix. It is believed that an operator perspective provides a deep
understanding of certain aspects of linear systems, along with a set of tools which are useful
in system analysis and control design.
The monograph begins by presenting the basic state space model used in linear systems
and control. Some physical examples from mechanics are given to motivate the state space
model. Then we study both state space and input-output stability of linear systems. Lya-
punov techniques are developed to analyze state space stability. Finally, we use these results
to present some stability results for mechanical systems.
Next we study controllability and observability for state space systems. All the classical
results on controllability and observability are presented. Standard least squares optimiza-
tion techniques are used to solve a controllability optimization problem and an observability
optimization problem. This naturally leads to the controllability and observability Gramians.
The connections between stability and these Gramians are studied. The duality between con-
trollability and observability is given. Using invariant subspaces and matrix representations
of linear operators, we present the controllable and observable decomposition for a linear
system. The singular value decomposition provides an efficient algorithm for computing the
controllable and observable decomposition.
The backward shift operator plays a fundamental role in operator theory. Here we use
the backward shift operator to develop realization theory. The backward shift approach to
realization theory provides a natural geometric setting for realization theory and simplifies
some of the classical proofs. All the standard results in realization theory are given. For
example, we show that a realization is controllable and observable if and only if it is minimal.
Then we show that all minimal realizations of the same transfer function are similar. The
classical Kalman-Ho algorithm is presented.
Using standard linear algebra techniques, we present the classical solutions to the eigen-
value placement problem in state feedback control of controllable linear systems. First we
study the single input-single output case. Ackermann's formula is presented. Then we solve
the eigenvalue placement problem for multivariable systems.
State feedback results are used to develop state estimators for detectable systems. By
combining the stabilization results with the state estimators, we present the classical observer-
based output feedback controllers. Finally, we present a short chapter on the zeros of a state
space system.
The solution to the linear quadratic regulator problem plays a fundamental role in de-
signing feedback controllers. Here we present a solution to the linear quadratic regulator
problem by using standard least squares optimization techniques. For example, the adjoint
of a certain operator readily yields the corresponding two-point boundary value problem, the
feedback solution and a closed form solution for the Riccati differential equation. We also
provide a simple derivation of the solution to the tracking problem. Many of the classical
results associated with the quadratic regulator problem are also presented. For instance, we
obtain stabilizing controllers from the solution of the algebraic Riccati equation. A spectral
factorization of a certain positive operator is presented. Then this factorization is used to
develop the classical root locus interpretation of the feedback controllers obtained from the
solution of the linear quadratic regulator problem.
Motivated by the linear quadratic regulator problem, we present a separate chapter on
the Hamiltonian matrix and its role in computing the stabilizing solution to the algebraic
Riccati equation. The Hamiltonian matrix plays a fundamental role in linear system theory
and is also used in studying H°° analysis and control.
The last two chapters present an introduction to H°° analysis and control. First we
study the analysis problem. Our approach is based on an abstract optimization problem.
Then we introduce a rnax-min optimization problem to study the full state feedback H°°
control problem. This max-min optimization problem is essentially a combination of the
linear quadratic regular problem and the disturbance attenuation problem. The solutions
to these optimization problems are obtained by completing the squares of certain operators.
This yields the appropriate Riccati equation and the corresponding H°° controller. Finally,
we also present a controller which is a trade-off between the optimal H°° and H2 controller.
The authors are indebted to George Leitmarm and Bob Skelton for their encouragement
and inspiration.

Martin J. Corless
Arthur E. Frazho
Contents

Preface v

1 Systems and Control 1

1.1 Notation 1
1.2 State space description 3
1.3 Transfer functions 6
1.3.1 Proper rational functions 7
1.3.2 Power series expansions 9
1.4 State space realizations of transfer functions 13
1.5 Notes 17

2 Stability 19
2.1 State space stability 19
2.2 Mechanical systems 20
2.3 Input-output stability 23
2.4 The H°° norm 25
2.5 Notes 27

3 Lyapunov Theory 29
3.1 Basic Lyapunov theory 29
3.1.1 Lyapunov functions 33
3.2 Lyapunov functions and related bounds 34
3.2.1 Bounds on eAt 34
3.2.2 Some system bounds 36
3.3 Lyapunov functions for mechanical systems 38
3.4 Notes 39

4 Observability 41
4.1 Observability 41
4.2 Unobservable eigenvalues and the PBH test 45
4.3 An observability least squares problem 46
4.4 Stability and observability 50
4.5 Notes 53

vii
viii CONTENTS

5 Controllability 55
5.1 Controllability 55
5.2 Uncontrollable eigenvalues and the PBH test 58
5.3 A controllability least squares problem 59
5.4 Stability and controllability 61
5.5 Notes 64

6 Controllable and Observable Realizations 67

6.1 Invariant subspaces 67
6.2 The controllable and observable decomposition 69
6.2.1 A minimal realization procedure 75
6.3 The minimal polynomial and realizations 77
6.4 The inverse of a transfer function 80
6.5 Notes 82

7 More Realization Theory 83

7.1 The restricted backward shift realization 83
7.2 System Hankel operators 89
7.3 Realizations and factoring Hankel matrices 91
7.4 Partial realizations and the Kalman-Ho Algorithm 96
7.4.1 The Kalman-Ho Algorithm 102
7.5 Matrix representation of operators 102
7.6 Shift realizations for proper rational functions 106
7.7 Jordan form realizations 113
7.8 Notes 116

8 State Feedback and Stabilizability 119

8.1 State feedback and Stabilizability 119
8.2 Simple stabilizing controllers 120
8.3 Stabilizability and uncontrollable eigenvalues 122
8.4 Eigenvalue placement 123
8.4.1 An eigenvalue placement procedure 123
8.4.2 Ackermann's Formula 125
8.5 Controllable canonical form 126
8.5.1 Transformation to controllable canonical form 128
8.5.2 Eigenvalue placement by state feedback 130
8.6 Multivariable eigenvalue placement 130
8.7 Two canonical forms 135
8.8 Transfer functions and feedback 141
8.9 Notes 143

9 State Estimators and Detectability 145

9.1 Detectability 145
9.2 State estimators 148
9.3 Eigenvalue placement for estimation error 149
CONTENTS ix

9.4 Notes 151

10 Output Feedback Controllers 153

10.1 Static output feedback 153
10.1.1 Transfer function considerations 154
10.2 Dynamic output feedback 156
10.3 Observer based controllers 158
10.4 Notes 162

11 Zeros of Transfer Functions 163

11.1 Zeros 163
11.2 The system matrix 165
11.3 Notes 170

12 Linear Quadratic Regulators 171

12.1 The finite horizon problem 171
12.1.1 Problems with control weights 174
12.2 An operator approach 175
12.2.1 An operator based solution 175
12.2.2 The adjoint system 177
12.2.3 The Riccati equation 179
12.3 An operator quadratic regulator problem 180
12.4 A linear quadratic tracking problem 182
12.5 A spectral factorization 184
12.5.1 A general tracking problem 186
12.6 The infinite horizon problem 189
12.7 The algebraic Riccati equation 190
12.8 Solution to the infinite horizoh problem 193
12.8.1 Problems with control weights 195
12.9 An outer spectral factorization 196
12.10 The root locus and the quadratic regulator 198
12.10.1 Some comments on the outer spectral factor 202
12.11 Notes 204

13 The Hamiltonian Matrix and Riccati Equations 205

13.1 The Hamiltonian matrix and stabilizing solutions 205
13.1.1 Computation of the stabilizing solution 210
13.2 Characteristic polynomial of the Hamiltonian matrix 211
13.3 Some special cases 212
13.4 The linear quadratic regulator 214
13.5 H°° analysis and control 216
13.6 The Riccati differential equation 219
13.6.1 A two point boundary value problem 219
13.6.2 Some properties 221
13.7 Notes 224
x CONTENTS

14 H°° Analysis 227

14.1 A disturbance attenuation problem 227
14.2 A Riccati equation 229
14.3 An abstract optimization problem 234
14.4 An operator disturbance attenuation problem 235
14.5 The disturbance attenuation problem revisited 236
14.5.1 The adjoint system 237
14.5.2 The Riccati equation 239
14.6 A spectral factorization 241
14.6.1 A general disturbance attenuation problem 242
14.7 The infinite horizon problem 244
14.7.1 Stabilizing solutions to the algebraic Riccati equation 247
14.7.2 An outer spectral factor 249
14.8 The root locus and the H°° norm 250
14.9 Notes 254

15 H°° Control 255

15.1 A H°° control problem 255
15.1.1 Problem solution 256
15.1.2 The central controller 261
15.2 Some abstract max-min problems 263
15.3 The Riccati differential equation and norms 268
15.4 The infimal achievable gain 270
15.5 A two point boundary value problem 271
15.5.1 The Riccati differential equation 273
15.6 The infinite horizon problem 274
15.6.1 The stabilizing solution 277
15.6.2 The scalar valued case 281
15.7 The central controller 282
15.8 An operator perspective 284
15.9 A tradeoff between norms 288
15.9.1 The L2 norm of an operator 288
15.9.2 The L 2 norm of a system 289
15.9.3 The L 2 optimal cost 291
15.9.4 A tradeoff between d^ and d2 293
15.10 A tradeoff between the H2 and H°° norms 295
15.11 Notes 297

16 Appendix: Least Squares 299

16.1 The Projection Theorem 299
16.2 A general least squares optimization problem 302
16.2.1 Computation of orthogonal projections 305
16.3 The Gram matrix 309
16.4 An application to curve fitting 312
16.5 Minimum norm problems 313
CONTENTS xi

16.6 The singular value decomposition 317

16.7 Schmidt pairs 319
16.8 A control example 323
16.8.1 State space 324
16.8.2 The L2-L°° gain 325
16.9 Notes 327

Bibliography 329

Index 337
Chapter 1

Systems and Control

In this chapter we introduce the basic state space model used throughout this monograph.
Some elementary examples from mechanical systems are given to motivate this model. Then
we introduce the transfer function for a state space model. Some basic state space realization
results are presented.

1.1 Notation
In this section we introduce some notation used throughout this monograph. The reader can
choose to go directly to the next section and refer back to this section as the need arises.
The set of all complex numbers is denoted by C. Throughout, all linear spaces are complex
vector spaces unless stated otherwise. The inner product on a linear space is denoted by
( • , • ) . To be precise, the inner product on a linear space H is a complex valued function
mapping H. x "H into C with the following properties

(ii) (a

(iii) ( M ) > 0 ;
(iv) (h, h)=0 if and only if h = 0.

Here /, g and h are arbitrary vectors in H while a and j3 are arbitrary complex numbers. The
inner product (-, •) is linear in the first variable and conjugate linear in the second variable.
An inner product space is simply a linear space with an inner product. If Ti. is an inner
product space, then the norm of the vector h is the non-negative real number defined by
\\h\\ = -^/(h,h). The Cauchy-Schwartz inequality states that \(f,h)\ < \\f\\ \\h\\ for all / and
h in T-L. Moreover, we have equality [(/, ti)\ = \\f\\ \\h\\ if and only if / and h are linearly
dependent. Finally, it is noted that if H is an inner product space, then we have the triangle
inequality, namely, ||/ + h\\ < \\f\\ + \\h\\ for all / and h in H.
Let "H be an inner product space. Then we say that a sequence {/ij}S° is a Cauchy
sequence in H if each hi is in Ti, and \\hi — hj\\ approaches zero as i and j tend to infinity.
An inner product space is complete if every Cauchy sequence converges to a vector in T~C,
2 CHAPTER 1. SYSTEMS AND CONTROL

that is, if {hi}™ is any Cauchy sequence in H, then {^}o° converges to a vector hinH. A
Hilbert space is simply a complete inner product space. A Hilbert space is separable if it is
the closure of a countable set. Throughout we only consider separable Hilbert spaces. If 7i
is a (separable) Hilbert space, then there exists an orthonormal basis {fj}™ for 7i where ra
is possibly infinite. In fact, m is the dimension of 7i. In this case, every h in 7i admits a
Fourier series expansion of the form

_ (h€H). (1.1)

Moreover, we also have Parseval's relation

(h,g)=} (h, (fj)(v?j , g) (h,g£T~{). (1-2)

.7=1

In particular, \\h\\2 = ^™ \(h,(fj) 2.

If JF is a subset of a linear space "H, then V 3~ denotes the linear span of J-', that is the
space of all linear combinations of elements of T. If H is a Hilbert space, then \/ .F is the
closed linear span of JF.
Throughout Cn is the Hilbert space determined by the set of all complex n-tuples of the
form [xi, x 2 , • • • , xn]tr where xy is in C for all j ~ 1, 2, • • • , n and tr denotes the transpose.
The inner product on Cn is given by

where x = [xi,.x 2 , • • • ,xn}tr and y = [yi,y 2 , • • • ,i/n] tr -

Furthermore, L2[a, b] is the Hilbert space formed by the set of all square integrable
Lebesgue measurable functions over the interval [a, 6], that is, / is in L 2 [a,6] if and only if
/ is Lebesgue measurable on [a, b] and

The inner product on L2[a, b] is given by

Now let U be a Hilbert space. Then L 2 ( [ a , b ] , U ) denotes the Hilbert space formed by the
set of all square integrable Lebesgue measurable functions on [a, 6] with values in U. The
inner product on L 2 ( [ a , b ] , l 4 ) is given by
1.2. STATE SPACE DESCRIPTION 3

where (f(t),g(i))u is the inner product on U. For example, I/ 2 ([a, b], C n ) is the Hilbert space
formed by the set of n-tuples of the form / = [/l5 /2, • • • , fn]tr where fj is a vector in I/2[a, b]
for all j = 1,2, • • • , n. The inner product on Z/ 2 ([a, b], C") is given by

where g = [gi,g^-- • ,gn]tr is in L 2 ([a,fc],C n ).

We say that T is an operator if T is a linear map from a linear space 7i into a linear
space 1C. The kernel or null space of T is denoted by kerT, and the range of T is denoted
by ranT. Now let H. and K. be Hilbert spaces and T be an operator from 7i into /C. The
norm of T is denned by

||T|| : = sup{||T/i|| : h e U and \\h\\ < 1} .

We say that T is a bounded operator if ||T|| is finite. Throughout this monograph we deal
mainly with bounded operators. So, all of our operators acting between Hilbert spaces are
bounded unless stated otherwise. The set of all bounded operators from Ti. into K. is denoted
by L(H, /C). Notice that if T, R and S are operators acting between the appropriate Hilbert
spaces, then \\TR\\ < \\T\\ \\R\\ and \\T + S\\ < ||T + \\S\\.
As before, let T be a bounded operator mapping H into /C. Then the adjoint T* of T is
the linear operator from /C into H uniquely determined by (Th, k) — (h,T*k) for all hinH
and k in /C. It is well known that T and T* have the same norm, that is, ||T|| = T*||. If
7i is finite dimensional, then ||T||2 equals the largest eigenvalue of T*T. Finally, it is noted
that if T is a matrix from C" into Cm, then T* is the conjugate transpose of T, that is,
T* = Ttr. On several occasions we will use some elementary results from Hilbert spaces. For
some references on Hilbert spaces see Akhiezer-Glazman [2], Balakrishnan [8], Conway [30],
Gohberg-Goldberg [53], Halmos [59] and Taylor-Lay [117].

1.2 State space description

A common model for describing a dynamic system is given by the following state space
representation
x = Ax + Bu , .
( J
y = Cx + Du
where the state x(t) belongs to X while the input u(t) lives in U and the output y(t) lives
in ,y for all time t > 0. Unless otherwise noted, we always assume that the state space X,
the input space U. and the output space y are finite dimensional Hilbert spaces. The system
operator A acts on X and B maps U. into X, while C maps X into y and D maps U into
y. The state space system in (1.3) is denoted by {A,B,C, D}. In many applications A,
B, C and D are matrices acting between the appropriate Euclidean spaces. If an initial
condition XQ = x(0) is specified and the input u is given, then there is a unique solution to
the differential equation in (1.3). In this case, the output y is uniquely determined by the
initial state XQ and the input u.
4 CHAFTER 1. SYSTEMS AND CONTROL

Example 1.2.1 The forced linear oscillator. Probably one of the simplest models in study-
ing mechanical systems is the forced linear oscillator shown in Figure 1.1. It consists of a
small body of mass m > 0, a linear spring with spring constant k > 0, and a linear dashpot
with damping coefficient c > 0. At each instant of time t, the body is subject to a force u(t).
Applying Newton's second law, we arrive at the following differential equation
my + cy + ky = u (1-4)
where y(t) is the displacement of the body from its equilibrium position. Now let x\ = y
and £2 — y- Using x\ = x-i and ±2 = y, it follows that (1.4) admits the state space form
0 1 0
-k/m —c/m 1/ra

y = [i 01

In this case the displacement y = x\ is the output. Notice that in this example D = 0.
However, if we choose the acceleration y of the mass as the output, then the corresponding
D matrix is nonzero.

u(t)

Figure 1.1: The forced linear oscillator

Example 1.2.2 By following the technique in the previous example, one can convert any
linear differential equation to a state space system of the form (1.3). To see this, consider
any scalar input-output system, with input u(i) and output y ( t ) , described by an n-th order
differential equation of the form
y(n] + a^iy(n-l} + • • • + a^y + a0y = u (1.5)
where y^ denotes the j-th derivative of y and a,- is a scalar for j = 0,1, • • • , n—l. To convert
this system to state space form, let Xi = y, x2 — y, • • • ,xn = y^n~l\ Then using Xj — Xj+i
for j = 1,2, • • • , n— 1 along with xn = ?/"), we obtain
" Xi ' ' 0 1 0
X2 0 0 1

u (1.6)

X 0 0 0
x
n _ -GO -a-i -az

y = [ 1 0 0 o o
The initial state is given by x(Q) = [y(0), 1/^(0), • • • , y( ~ \Q)]tr where tr denotes the trans-
n l

pose. Therefore, one can readily obtain a state space representation for any system described
by (1.5).
1.2. STATE SPACE DESCRIPTION

Example 1.2.3 Axisymmetric spacecraft.

Figure 1.2: The axisymmetric spacecraft

We consider here the rotational motion of an axisymmetric spacecraft which is subject to

two body-fixed input torques u\ and u-i about axes which are perpendicular to the symmetry
axis. This is illustrated in Figure 1.2. If we choose a body-fixed axes system with origin at
the center of mass, with axes corresponding to u\, u^ and whose 3-axis corresponds to the
axis of symmetry, the rotational motion of the axisymmetric spacecraft can be described by

/icJi = (J2 -
I2u>2 = (Is -
I3uj3 — 0

where uijU^,^ denote the components of the spacecraft angular velocity vector with respect
to the body-fixed axes; the positive scalars /i, /2, /3 are the principal moments of inertia of
the spacecraft with respect to the body-fixed axes and I\ — /2. From the third equation
above 0^3 is constant. If we introduce the constants a := (/i — J3)a;3//i and j3 := l//i, define
state variables x\ := u\ , x% := u>2 and output variables y\ := x\ , 3/2 '•— ^2> we obtain

xi — ax? + f3u\
x-2 = —axi + /3u-2
y\ = xi

This readily yields the state space form in (1.3) where

c=
-a o ' ~ o ^ ' o i ' - Q o
Now let us return to the state space description (1.3). For a specified initial condition
x(0) = XQ the state x(t) at time t for the system in (1.3) is uniquely given by
rt
x(t) = eAtx0+ eA(t-T)Bu(r}dT. (1.7)
Jo
6 CHAPTER 1. SYSTEMS AND CONTROL

The output y ( t ) of the state space system {A, B, C, D} in (1.3) is determined by

rt
y(t] = CeAtx0 + / CeA(t-r)Bu(r)dr + Du(i) (1.8)
Jo
where the initial state x0 = x(Q) is specified. Clearly,
ft
y(t) = CeAtx0+ G(t-r)u(r}dr (1.9)
Jo
where G is the operator valued function defined by

G(t) = CeAtB + 8(t)D. (1.10)

Here S(t} is the unit Dirac delta function. So, if the initial state XQ = 0, then (1.9) defines a
linear time invariant map of the form

y(t) = I G(t-r}u(r}dr . (1.11)

Jo
In particular, if u(i) = 8(t)uo for some constant UQ in U and x0 = 0, then y(t) — G(t}uQ.
Motivated by this, we call the function G in (1.10) the impulse response for {A,B,C, D}.
Clearly, if U — C1 and UQ = 1, then G(t] is precisely the output y(t) for u(i) — 5(t). Finally,
it is worth emphasizing that, under zero initial conditions, the input-output response of
{A, B, (7, -D} is completely determined by its impulse response via (1.11).

1.3 Transfer functions

In this section we introduce the notion of a transfer function for a linear system. Recall
that the Laplace transform of a measurable function / on [0, oo) with values in a finite
dimensional space £ is defined by
oo
e-stf(t)dt (1.12)
/
for appropriate values of s in C. (The set of all complex numbers is denoted by C.) The
Laplace transform of the function / is denoted by f and has values in 6 . Notice that a
boldface f is used to represent the Laplace transform of /. If / is exponentially bounded,
that is, if ||/(<) || < Me7* for some finite scalars M and 7, then f is an analytic function in
the region of C defined by 5R(s) > 7. Clearly, the Laplace transform operator £ is linear.
Finally, by convention, C(8) = I where 8 is the unit Dirac delta function.
Recall that £(/)(s) = sf(s) — /(O). By taking the Laplace transform of the state space
equations in (1.3), we obtain

sx(s) — XQ = Ax(s) + Bu(s)

y(s) =
1.3. TRANSFER FUNCTIONS 7

where x, y, u are the Laplace transforms of x, y, u, respectively and XQ — x(0). Solving these
equations for y in terms of XQ and u yields

y (s) = C(sl - A)-lx0 + [C(sl - A)~1B + D] u(s). (1.13)

The transfer function G for system {A, B, C, D} is denned by

G(s) = C(sI-A)-1B + D. (1.14)

Recall that the impulse response of system {A, B, C, D} given by

G(t) = CeAtB + 6(t)D.

Using the following Laplace transform result:

C(eM)(s) = (si - A)-1

we have G = £(G), that is, the transfer function G is the Laplace transform of the impulse
response G. Clearly, G is analytic everywhere except possibly at the eigenvalues of A.
When z(0) = 0, the relationship between the Laplace transform of the input u and the
Laplace transform of the output y is given by

y(*) = G(s)u(s). (1.15)

In other words, when the initial conditions are zero, the transfer function is the multiplication
operator which maps u into y. Recall that convolution in the time domain corresponds to
multiplication in the s-domain. Hence,

y(t)= I G(t-r)u(r)dr.
Jo
Obviously, G and G uniquely determine each other. So, for zero initial conditions, the
input-output response of the system {-4, B, C, D} is completely determined by its transfer
function.

1.3.1 Proper rational functions

A operator valued polynomial N is a function from C into some space C(U, y} of operators
which can be expressed as

N(s) = No + s/Vi + • • • + snNn

where NQ, ..., Nn are in £(U, y). It N is nonzero, then the degree of N (which we denote
by deg N} is the largest integer m for which Nm is non-zero. We say that an operator valued
function G of a complex variable is a proper rational function if G(s) = N ( s ) / d ( s ) where
N is an operator valued polynomial, d is a scalar valued polynomial and deg TV < degd. If
deg N < deg d, then we say that G is a strictly proper rational function. Notice that G is
a proper rational function with values in £(U, y) if and only if G admits a decomposition
8 CHAPTER 1. SYSTEMS AND CONTROL

of the form G(s) = N(s)/d(s) + D where N is an operator valued polynomial, d is a scalar

valued polynomial satisfying degN < degd, and D is a constant in £(U,y). In other
words, G is a proper rational function if and only if G admits a decomposition of the form
G(s) = G 0 (s) + D where G0 is a strictly proper rational function and D is a constant.
We claim that the transfer function G of a finite dimensional system {A on X, B, C, D}
is a proper rational function. To see this, recall that (si— A)"1 can be expressed as
(sJ-A)- 1 = adj (sI-A)/det[sI-A] (1.16)
where ad] (si — A) is the algebraic adjoint (or adjugate) of si — A and det[s/ — A] is the
characteristic polynomial of A. Since adj (si — A) is an operator valued polynomial whose
degree is strictly less than the degree of det[s/ — A], it follows that (si — A)"1 is a strictly
proper rational function. Hence, C(sl — A)~1B is also a strictly proper rational function.
Recalling from (1.14) that G(s) = C(sI-A)~lB + D, it follows that the transfer function G
is a proper rational function. Hence, G is a proper rational function of the form G = N/d
where d is the characteristic polynomial of A. Therefore, the poles of G are contained in
the eigenvalues of A. We say that A is stable if all the eigenvalues of A are contained in the
open left half plane {s : 2Rs < 0}. So, if A is stable, then G is also stable, that is, all the
poles of G are contained in the open left hand plane. Finally, it is noted that G is a strictly
proper rational function if and only if D equals zero.
Suppose G = N/d is a proper rational function. Then, we define the relative degree of G
to be deg d — deg N. If G is a transfer function for (A, B, C, D}, then G has relative degree
zero if and only if D is non-zero. When D is zero, the following result provides a state space
characterization of relative degree.
Lemma 1.3.1 Suppose G is a transfer function for a system {A, B,C, 0} and
CA]B = 0 for j = 0, l , - - - , r - 2 , (1.17a)
CAr~lB ^ 0. (1.17b)
Then
CAj(sI-A)-lB = sjG(s) for j = 0, 1, • • • ,r - 1 (1.18a)
r 1 r r l
CA (sI - A)~ B = s G(s}-CA - B. (
Furthermore, G = N/d where d is the characteristic polynomial of A while N is a polynomial
of degree deg(d)—r and whose highest order coefficient is CAr~lB. In particular, G has rela-
tive degree r. Finally, ifr equals the dimension of the state space, then G(s) — CAr~lB/d(s).
PROOF. We first demonstrate (1.18a) by induction. Since,
G(s) = C(sI-A)~1B,
the equality in (1.18a) clearly holds for j = 0. Suppose now the equality (1.18a) holds for
some j — k < r — 2. Then for j = k + 1,
CAk+l(sI-AYlB = C(A-sI + sI)Ah(sI-A)-1B
= -CAkB + sCAk(sI-A)-1B
= sk+1G(s).
1.3. TRANSFER FUNCTIONS 9

Hence, (1.18a) holds for j = k + 1. So, by induction it holds for j = 0, 1, • • • ,r — 1. We also

obtain that

CAr(sI - A}~1B = -CAr~lB + sCAT~\sI - A}~1B = -CAr~lB + srG(s) ;

hence (1.18b) also holds.
Recall that G = N/d where d is the characteristic polynomial of A and N is a polynomial
of degree strictly less than the degree of d. Hence, CAT(sI — A)~1B = M(s)/d(s) where M
is a polynomial of degree strictly less that the degree of d. It now follows from (1.18b) that

d(s) d(s)
that is,
srN(s} = CAT~1Bd(s) + M(s} .
The properties of d and M imply that srN(s) is a polynomial of degree deg(cQ and whose
highest order coefficient is CAr~lB. Hence, TV" is a polynomial of degree deg(rf) — r and the
coefficient of its highest order term is CAr~1B.

1.3.2 Power series expansions

We now claim that if G is a proper rational function with values in C(U,y}, then for
sufficiently large s, the function G admits a power series expansion of the form
do

G(s) = ^s-iGi (1.19)

i=0

where each d is in £(Z//, y). To see this let G = N/d where N isa, polynomial with values in
C(U,y] and d is a scalar valued polynomial satisfying deg./V < degof. Considering s — I/A
and letting G(A) = G(1/A), we obtain

where n is the degree of d. Here N and d are the polynomials of degree at most n defined
by N(\) = A n A r (l/A) and d(\) = A n d(l/A), respectively. Since d(0) equals the coefficient of
sn in the polynomial d, it is nonzero. Hence, G is analytic at zero. So, for sufficiently small
A, the function G has a power series expansion of the form G(X) = Y^o" ^'^' wnere each Gi
is in £(U,y). Therefore, for s sufficiently large, G admits a power series expansion of the
form (1.19). Moreover, the infinite sequence {G*}o° and G uniquely determine each other.
Now assume that G is a proper rational function of the form G = N/d where d is a
scalar valued polynomial and ./V is an operator valued polynomial defined by

d(s) = d0 + sd1 + - - - + sr*-ld^l + sndn (dn^0) (1.20)

N(s) - NQ + sNl + - - - + sr*-lNT^1 + snNn. (1.21)
10 CHAPTER 1. SYSTEMS AND CONTROL

Hence,

N(s) = d(s}G(s) = (1.22)

where Gi is the coefficient of s l in the power series expansion of G. Equating the coefficients
of sj in (1.22) for j = n , n —1, • • • ,0, yields

Nn = dnG0

(1-23)
4- dnGr,

dnG
These equations can be written in the following compact form:

G0
dil d2I ••• rfn_i/ dnl 0
d2I d3I ••• dnl 0 0 G2
= (1.24)

dn_J dnl ••• 0 0 0

Nn dnl 0 ... 0 0 0 Gn
(The block matrix in the above expression is a Hankel matrix.) Since N is a polynomial, the
coefficients of s~J in ]T^ d(s)s~"lGl are zero for all integers j > 1. For j = 1, we obtain

0 = dod + d!G2 ••• + d^Gn + dnGn+l . (1.25)

In fact, for any integer k > 1, we have

Q = d0Gk + dlGk+l + ---+dn-lGk+n-i+dnGk+n (1.26)

The above considerations lead to the following result.

Lemma 1.3.2 Let G = ]T^° s~lGl be the power series expansion for a proper rational func-
tion G and suppose d is a scalar valued polynomial of the form (1.20). Then G = N/d where
N is an operator valued polynomial if and only if (1.26) holds. Moreover, in this case, N is
given by (1.21) and (1.23).
When G = N/d and TV and d are known, the above results allow us to recursively compute
the operators in the sequence {Gi}™ directly from the polynomials N and d. Specifically,
(Gj}o can be recursively computed from (1.23) or (1.24), while {Gi}^+l is recursively com-
puted from
dn_\Gk-l + dn-l
Gk = - (1.27)

The following result uses state space methods to compute the coefficients in the power
series expansion for a transfer function.
1.3. TRANSFER FUNCTIONS 11

Proposition 1.3.3 Let ]^° s~lGi be the power series expansion for a proper rational func-
tion G. Then G is the transfer function for a system {A, B,C, D} if and only if

GQ = D and d = CAi'lB (for alii > I ). (1.28)

The proof of this result depends upon the following classical result.

Lemma 1.3.4 Let T be an operator on X satisfying \\T\\ < 1. Then I — T is invertible.

Moreover, in this case,
00

(I-T)-l = ^Tk. (1.29)

Finally, W-T)-^ <!/(!-\\T\\).

PROOF. Let Rn be the operator on X defined by Rn = Y^k=o Tk• We claim that Rn converges
to a bounded operator R = X^fclo-^ ( m ^ne operator topology) as n tends to infinity. To
see this, let r = \\T\\. Obviously, ||Tfc|| < rk for all integers k > 0. If n and m are any two
positive integers satisfying n > m, then
n n oo _ m+1
Up r? II
\\fin~Hm\\-\\
II \> A T^II <-• \> ^ IIT
1 \\ <
1C
' II <^ \ > ^ ,'
\\1 \\ <
k=m+l k=m+l fc=m+l

as m tends to infinity. The last equality follows from the geometric series ]T^ rk — 1/(1 — r).
So, Rn converges to the operator R — £)£° Tk as n approaches infinity. To verify that R is
a bounded operator simply notice that
00 00

^
.
-- i-r i-imr
To show that R is the inverse of I — T, observe that

(/ - T)Rn - (/ - T)(I + T + T2 + T3 + • • • + Tn) = I - Tn+l .

Because T"+1 converges to zero, the sequence {(/ — T)^n}o° converges to /. Since
converges to R, this implies that (I — T}R = I . Because Rn and I — T commute, we also
have R(I -T) = I. Therefore, / - T is invertible and ££° Tk is the inverse of / - T. •

PROOF OF PROPOSITION 1.3.3. By definition, G is the transfer function for a system

{A, B, C, D} if and only if G(s) = C(sl - A)~1B + D. Notice that

Whenever \s\ > \\A\\, it follows that ||^4/s|| < 1. According to Lemma 1.3.4, we have
°L Ai
(SI-A)-1 = ^T— (if |5|> || A||). (1.30)
12 CHAPTER 1. SYSTEMS AND CONTROL

Therefore, G is the transfer function for a system {A, B, C, D} if and only if it has a power
series expansion of the form
oo ,-, Ai—\ D
c* ( <A — n -L \ ^
Vjf \ <j I — LJ \~ 7 - .
(~\ 31 "\
\ J- •t~J J- /
V / / j „? ^ '

By matching like coefficients of l/s\ it follows that G(s) = ^^° Gi/sl is the transfer function
for a system {A, B, C, D} if and only if (1.28) hold. •
We say that {A, B, C, D} is a state space realization of G if G is the transfer function
for {A,B,C, D}. Consider any proper rational function G with a power series expansion
of the form ^^° Gi/s\ Then the above analysis shows that {^4, £?, C, D} is a realization of
G if and only if (1.28) holds. So, with out loss of generality we say that {A, B, C, D} is a
realization of a sequence of operators {Gi}™ if (1.28) holds.
Now suppose that {A, B, C, D} is a state space realization of G and d is a scalar valued
polynomial defined by (1.20). Then, according to Lemma 1.3.2 and Proposition 1.3.3, we
can express G as G = N/d where N is an operator valued polynomial if and only if
CAk-ld(A]B = 0 (fc > 1) (1.32)
n
where d(A) = dQI + diA + • • • dn_i/ + dnA . Since A and d(A) commute, the above
condition is equivalent to
CAld(A}AjB = 0 (i,j>0). (1-33)
Moreover, in this case, N is given by (1.21) and
Nn = dnD

(1.34)
= djD + dj+lCB

N0 =
When D = 0, we obtain that Nn — 0 and
[ No JV: • • • 7Vn_! } = [CB CAB • • • CAn-lB } T (1.35)
or
N0 CB
CAB
=T (1.36)
CAn-lB
where T is the invertible Hankel matrix given by
d\I d%I • • • dn^I rfn_i/ dnl
dzl d3I ••• dn^I dnl 0
dj dj ••• dnl 0 0

dn_i/ dnl ••• 0 0 0

dnl 0 ... 0 0 0
1.4. STATE SPACE REALIZATIONS OF TRANSFER FUNCTIONS 13

Summarizing, we have the following result.

Lemma 1.3.5 Let G be a transfer function for a state space system {A,B,C,D} and d a
scalar valued polynomial defined by (1.20). Then G = N/d where N is an operator valued
polynomial if and only if (1.33) holds. Moreover, in this case, N is given by (1.21) and
(1.34). When D = 0, the coefficients of N are also given by Nn = 0 and (1.35) or (1.36).

Remark 1.3.1 (Cayley-Hamilton Theorem) We have already seen that if G is the

transfer function for {A, B, C, D}, then it can be expressed as G = N/d where d is the char-
acteristic polynomial of A. Hence, the characteristic polynomial of A must satisfy (1.33). In
particular, if we consider B = C = I and i = j = 0, we obtain that d(A) = 0. This yields
the Cayley-Hamilton Theorem, namely if A is an operator on a finite dimensional space X,
then d(A) = 0 where d is the characteristic polynomial of A.

1.4 State space realizations of transfer functions

We say that {A on X', B, C, D} is a state space realization of G if G is the transfer function
for {A, B, C, D}, that is, (1.14) holds. The dimension of the state space X is the dimension
of the realization {A, B, C, D}. The analysis in the previous section shows that all transfer
functions are proper rational functions. This is part of the following basic result in system
theory.

Theorem 1.4.1 Let G be an operator valued function of a complex variable. Then G admits
a finite-dimensional state space realization if and only if G is a proper rational function.

PROOF. We only need to prove that a proper rational function G has a finite-dimensional
state space realization. Our proof is motivated by the form of the matrix A in equation
(1.6) of Example 1.2.2. Let G be a proper rational function with values in £(U,y). Then
G admits a decomposition of the form

G = N/d + D (1.37)

where N is a polynomial with values in C(U, y) and d is a scalar valued polynomial of the
form
Ti-l n-1
N(s) = J2 NkSk and d(s) = sn + ]T aksk. (1.38)
k=0 k=0

Here D, NO, • • • , N^i are operators in £(U,y) while a o , - - - , ^ ! are scalars. Now let
X = ©"W be the linear space formed by the set of all n-tuples of vectors from U, that is,
let X be the linear space consisting of all vectors of the form [MI, 7/2, • • • , un]tr where Uj is a
vector in U for all integers j = 1,2, • • • , n. (Recall that tr denotes transpose.) Let A be the
block matrix on X and B the block matrix from U into X and C the block matrix from X
14 CHAPTER 1. SYSTEMS AND CONTROL

into y defined by
• o / 0 0 o •
0 0 / 0 0

A = TD _
(1.39)

0 0 0
—ail —a2I
0 ,^
C = [ N0 7VX N2 ••• Nn_2 AU ]
Define V as the polynomial with values in C(U, X) given by

V(s}= [ I si s2! ••• sn~ll ]tr .

Here / is the identity on U. Then,
si -I 0 ••• 0 0 I ' 0 "
0 si -I ••• 0 0 si 0

(sI-A)V(s) =

0 0 0 ••• si -/
-I Sn~2I 0
HQ! ciil a2I • • • an_2/ an-i)I . . sn~ll . . d(s)I _
= Bd(s).
Hence, whenever s is not an eigenvalue of A,
(sI-Ar1B^
This along with the definition of (7, implies that
C(sl - A)~1B = [No N! ••• V(s)/d(s) = N(s)/d(s)
The above equation and (1.37) yields G(s) = C(sl - A)-1B + D. Therefore, {A , B , C , D}
is a finite dimensional realization of G. •

Remark 1.4.1 Let G be a proper rational function with values in C(U, y} of the form (1.37)
where N and d are polynomials as defined in (1.38). Moreover, let A, B, C be the block
matrices defined in (1.39). Then the proof of the previous theorem shows that {^4, B, (7, D}
is a realization of G. In particular, any proper rational function of the form (1.37), (1.38)
admits a state space realization whose state dimension is at most n • dimU.
We now present another finite dimensional state space realization of a proper rational
function G with values in C(U,y}. Since G is proper and rational, it has a power series
expansion of the form G(s) = J^o5"1^ where each d is in £(U,y}. Also G can be
expressed as in (1.37) where N and d are polynomials as defined in (1.38). Now let X = ©"3^
1.4. STATE SPACE REALIZATIONS OF TRANSFER FUNCTIONS 15

be the linear space formed by the set of all n-tuples of vectors from y. Let A be the block
matrix on X and B the block matrix from U into X and C the block matrix from X into y
defined by
" 0 I 0 ' G,
0 0 1

A = B= (1.40)

0 0 0
. -a0I —a\I -a

0 0 0 0

It follows from (1.27) that — a0Gi — aiG^ — • • • — a^Cn — G™+I. This along with the shift
structure of A implies that AB = [G-z G3 • • • G^if7". By induction with (1.27), one can
readily show that, for any integer k > 0, we have AkB = [Gk+i Gk+z • • • Gh+n]tr- Using
the structure of C yields CAkB = G^+i for all k > 0. Recalling Proposition 1.3.3, we see
that G(s) = C(sl - A)~1B + D, that is, {A,B,C,D} is a realization of G. Thus, any
proper rational function of the form (1.37), (1.38) admits a state space realization whose
state dimension is at most n • dim y.

Remark 1.4.2 By combining Remark 1.4.1 with our previous analysis, we obtain the fol-
lowing result. Any proper rational function G of the form (1.37), (1.38) with values in
£(U, y) admits a state space realization whose state dimension is at most n • m where m is
the minimum dimension of the spaces U and y.

In many applications such as mechanical systems one encounters higher order vector
differential equations of the form
A0y = u . (1.41)
Here the input u and output y are functions with values in U and Aj are operators on U for
j = 0, 1, • • • , n — 1. By following Example 1.2.2, set Xj = y^1^ for j — 1, 2, • • • , n. Let X be
the space defined by X = ©™ZY. Then a state space realization for this system is given by

x = Ax + Bu and y = Cx (1.42)
where A on X and B from U into X and C from X into U. are given by
" 0 7 0
0 0 /
A = B= (1.43)

0 0 0
-A2
/ 0 0
16 CHAPTER 1. SYSTEMS AND CONTROL

By taking the Laplace transform of (1.41) with all the initial conditions set equal to zero,
we obtain 17(s)y(,s) = u(s) where 17 is the operator valued polynomial denned by

= A0 A2s2 snl. (1.44)

It follows from Exercise 1 below that £l(s) l = C(sl — A)~~1B. In other words, fi 1 is the
transfer function for the system {A,B,C, 0} defined in (1.43).
A matrix of the form of A in (1.43) is called a companion matrix or a block companion
matrix when dim£Y > 1. Companion matrices play a fundamental role in systems and
control. If M is any operator on a finite dimensional space, then det[M] is the determinant
of any matrix representation of M. The following result provides a complete characterization
of the eigenvalues and eigenvectors for companion matrices.

Lemma 1.4.2 Let A be the (block) companion matrix on ®™U given in (1-43) and Q, the
corresponding operator valued, polynomial in (1.44)- Then A is an eigenvalue of A if and only
z/det[Q(A)] — 0. Furthermore, all eigenvectors v corresponding to a specified eigenvalue A
are given by
w
Xw
(1.45)

where w is a nonzero vector in the kernel o/fl(A).

PROOF. Recall that A is an eigenvalue of A if and only if Av = Xv where v is a nonzero

vector in X — ®r{U. Clearly, any such vector admits a decomposition of the form

W-2

where Wj is in U for all j. By using the structure of A, it follows that Av = Xv if and only if

These relationships are equivalent to

= X2wi, • • • , wn = Xn lwi and

+ ... + \Ai + A0)wl = 0 . (1.46)

By setting w = wi, the first n—l equations show that v admits a representation of the form
(1.45). In particular, v ^ 0 is equivalent to w ^ 0. Furthermore, the last expression in (1.46)
shows that A is an eigenvalue of A if and only if Q(X)wi = 0. Therefore, A is an eigenvalue
of A if and only if det[fi(A)] = 0. From the proceeding analysis it should be clear that all
the eigenvectors corresponding to a specified eigenvalue A are given by v in (1.45) where
n(A)u; = 0 and w ^ 0. •
1.5. NOTES 17

Scalar transfer functions. The scalar valued case plays an important role in many appli-
cations. Let g be any scalar valued proper rational function. Then g admits a decomposition
of the form
d + c2s + • • • + CnS1*-1
g(s)
&v = : 1-7. (1-47)
' "-1 n

' 0 1 0 • 0 0 " '0 '

0 0 1 • 0 0 0

A = ' ' ' , B= ' (1.48)

0 0 0 • 0 1 0
. -oo -ai -a2 • • -a^-2 -On-l . 1 _
C = Cl C2 C3 • L Cn ] .

Then our previous analysis shows that {A, B , C, 7} is a realization of g. Moreover, A is an

eigenvalue for the companion matrix A in (1.48) if and only if A is a zero of the polynomial
GO 4- a\s + • • • + a^is"^1 + sn. In this case, all corresponding eigenvectors for A are scalar
multiples of [1, A, A 2 , • • • , A"^1]*7". Finally, it is noted that the poles of g are contained in the
eigenvalues of A.
Exercise 1 Suppose G = N£l~l where N and f2 are polynomials of the form
n-l n-1
]\r(c\ — \~^
JV \^/ — / T\J, ~k
-L*K^ anr i
dllU. Ofs 1
u t r l O)l —
— <s
O T
n
J 4-
( ^ V"
^ /Ls fc
-Ti^O . Cl 49^
I J..r±C/l

fe=0 fc=0

Here NQ, • • -, Nn-i are operators in £(U,y) while >lo, • • •, -<4n-i are operators in £(U,U).
Let Af be the space defined by Af = ©"ZY. Then show that a state space realization for this
system is given by
x = Ax + Bu and y = Cx (1.50)
where A on X and B from U into X and C from X into ^ are given by
0 / 0 • • • 0 0 ] TO
0 0 / ••• 0 0 0

A = B= (1.51)

0 0
-AI -A-i
s~i

1.5 Notes
All the results in this section are classical. For further results on linear systems see Brockett
[21], Chen [26], DeCarlo [33], Delchamps [34], Fuhrmann [47], Kailath [68], Polderman-
Willems [98], Rugh [110], Skelton [114] and Sontag [116]. For some references on Hilbert
18 CHAPTER 1. SYSTEMS AND CONTROL

spaces see Akhiezer-Glazman [2], Balakrishnan [8], Conway [30], Gohberg-Goldberg [53], Hal-
mos [59] and Taylor-Lay [117]. For an in-depth study of operator theory and its applications
see Gohberg-Goldberg-Kaashoek [54, 55].
Chapter 2

Stability

In this chapter we will present some classical results on state space stability and input-output
stability. We will also present some elementary stability results for mechanical systems.
Let us establish some terminology used in this chapter and throughout the monograph.
Let P be an operator on a Hilbert space H. Then we say that P is a positive operator if
(Ph, h) > 0 for all h in Ji. The notation P > 0 means that P is positive. If P is a positive
operator, then P is a self-adjoint operator. In general when we write P > Q we mean that P
and Q are two self-adjoint operators and P — Q is positive. (It is easy to construct examples
where P — Q is positive and P and Q are not self-adjoint.) If H is finite dimensional, then P
is positive if and only if P is a self-adjoint operator and all the eigenvalues of P are greater
than or equal to zero. Following Halmos [58], we say that P is a strictly positive operator
if there exists a constant 5 > 0 such that (Ph, h) > 6 \\h\\2 for all h in H. So, P is strictly
positive if and only if P is positive and invertible. If Ji is finite dimensional, then P is
strictly positive if and only if P is a self-adjoint operator and all the eigenvalues of P are
strictly greater than zero. Moreover, in the finite dimensional case, P is strictly positive
if and only if (P/i, h) > 0 for all nonzero h in H.. So, if H. is finite dimensional, then the
notation P > 0 means that P is strictly positive. Finally, it is noted that our terminology is
slightly different from the standard terminology used in linear algebra. If T~C = C", then our
strictly positive operator is usually referred to as a positive definite matrix, and our positive
operator is usually defined as a positive semi-definite matrix.

2.1 State space stability

This section is devoted to the stability of systems described by

x = Ax (2.1)

where A is an operator on a finite dimensional Hilbert space X and t > 0. We say that the
system x = Ax is stable if
lim x(t] = 0
t->00

for every solution x to (2.1). Recall that every solution x to (2.1) is given by x ( t ) — eAtxo
where x0 is a constant vector in X. Therefore, the system x = Ax is stable if and only if eAt

19
20 CHAPTER 2. STABILITY

converges to zero as t tends to infinity. The following basic result completely characterizes
the stability of x — Ax in terms of the eigenvalues of A.
Theorem 2.1.1 The system x = Ax in a finite dimensional space is stable if and only if
all the eigenvalues of A have nonzero negative real part, that is, all the eigenvalues of A are
in {s G C : 9ft(s) < 0}. In this case, there exists positive constants m and a > 0 such that
\\eAt\\ <me-atforallt>0.

PROOF. Recall the Laplace transform result: £(eAt) = (si -A)"1. If L is any operator on the
finite dimensional space X , then det[L] denotes the determinant of any matrix representation
of L. Because A is on a finite dimensional space, (si — A)~l = N(s)/d(s) where N is the
algebraic adjoint of A and d(s) = det[s/ — A] is the characteristic polynomial of A. The
roots of d are the eigenvalues of A. Moreover, N is a polynomial with values in £(X, X]
satisfying deg N < deg d. By computing the partial fraction expansion of N/d, we obtain
that

where AI , A2 . . . , \m are the distinct roots of d, or equivalently, the eigenvalues of A and

are operators on X for all j,k. Moreover, RJT] is nonzero for all j. Hence,

Therefore, eAt approaches zeros as t approaches infinity if and only if 3fJ(Aj) < 0 for all j.
Equation (2.3) implies that

for all t > 0. Now assume that A is stable. From this it readily follows that there exists
positive constants m and a > 0 such that \\eAt\\ < me~at for all t > 0. •
Notice that the previous proof provides a method to compute eAt. To be more specific,
compute the operators Rjk in the partial fraction expansion of (si — A)~l. Then eAt is given
by (2.3).
By a slight abuse of terminology we say that an operator A on a finite dimensional space
is stable if all its eigenvalues have nonzero negative real part. Finally, eAt is stable if A is
stable.

2.2 Mechanical systems

In this section we present a general class of mechanical systems and an elementary stability
result. We have already seen a one degree-of-freedom mechanical system in Chapter 1, that is,
mq + cq + kq = 0. Obviously, m, c and k are positive operators on C1. The following example
illustrates how positive operators naturally occur in a two degree-of-freedom mechanical
system.
2.2. MECHANICAL SYSTEMS 21

Example 2.2.1 Consider the simple structure illustrated in Figure 2.1. It consists of two

I m2

Figure 2.1: A simple structure

"floors" of masses mi > 0 and m2 > 0 connected by linear springs of spring constants ki > 0,
k?. > 0 and dashpots with damping coefficients c\ > 0, c2 > 0. The scalars yi, y2 are the
horizontal displacements of the floors from their equilibrium positions. An application of
Newton's Second Law to each floor yields:

+ + c2)y'i - c2y2 + + - fc2y2 = 0

' + C2y2 - + fc22 = 0.

Let q be the vector in C2 defined by q = [yi , y2]*r where tr denotes the transpose. Then this
system can be described by the following second order vector differential equation:

Mq + Cq + Kq = 0

where the M, C, K are the operators given by

mi 0 I „ I ci + c2 -c2 +
M=
0 m2 -c2

Since mi > 0 and m2 > 0, the operator M is strictly positive. Finally, in this example, C
and K are also strictly positive.
A general mechanical system with n degrees-of-freedom is described by

Mq + Cq + Kq = 0 (2.4)

where q lives in an n-dimensional vector space Q and M is a strictly positive operator on Q,

while C and K are operators on Q. The initial conditions for this differential equation are
given by g(0) and </(0). Here M is called the inertia operator, K is the stiffness operator,
and C is the damping operator. This differential equation can be used to model a large class
of linear mechanical systems; see Meirovitch [89]. We say that system (2.4) is stable if q(t)
approaches zero as t approaches infinity for every solution q.
To convert the mechanical system in (2.4) to state space form, let X = Q © Q be the
2n-dimensional linear space consisting of all vectors of the form [qi, <?2]tr where q\ and g2
22 CHAPTER 2. STABILITY

are vectors in Q. If we set x = [<?,<?] t r , then the mechanical system (2.4) admits a state
description of the form x = Ax where A is the block matrix on X denned by

A
A -
~ - -M-1C
We claim that the mechanical system (2.4) is stable if and only its state space representation
x = Ax is stable. Clearly, if x = Ax is stable, then q(t) approaches zero as t tends to infinity,
and thus, the corresponding mechanical system is stable. Now assume that the mechanical
system is stable. Let A be any nonzero eigenvalue of A. Because A is a block companion
matrix, the eigenvector corresponding to A is a vector of the form v = [w , Xw}tr where w is
a nonzero vector in Q; see Lemma 1.4.2. For the initial condition x(0) = v, the solution to
the differential equation x — Ax is given by x ( t ) = eAtv = [extw , Xextw}tr. Since q is the first
component of x, and the mechanical system is stable, we must have 9?(A) < 0. So, all the
eigenvalues of A have nonzero negative real part. Therefore, x = Ax is stable, which proves
our claim.

Theorem 2.2.1 Let Q be the second order operator valued polynomial defined by
tt(s) = s'2M + sC + K (2.6)
where M , C and K are all operators on Q with M mvertible. Then the corresponding
mechanical system in (2.4) is stable if and only if all the roots of the polynomial det[f2(s)]
have nonzero negative real part.
PROOF. We have just seen that the mechanical system in (2.4) is stable if and only if the
system x = Ax is stable where A is given by (2.5). According to Lemma 1.4.2, the eigenvalues
of A in (2.5) are given by the set of all complex numbers A such that det[O(A)] = 0. Hence,
x = Ax is stable if and only if all the roots of det[fi] have nonzero negative real part. This
completes the proof. •
We are now ready to state a basic stability result for mechanical systems.
Theorem 2.2.2 // M, C and K are all strictly positive operators on Q, then the corre-
sponding mechanical system in (2.4) is stable.
PROOF. Let fi be the operator valued polynomial defined in (2.6). From Theorem 2.2.1, the
mechanical system in (2.4) is stable if and only if all the roots of the polynomial det[f2] have
nonzero negative real part. If A is a root of det[fi], then £l(X)v = 0 for some nonzero vector
v in Q. Thus,
0 = (n(A)v, v) = A2 (Mv, v) + X (Cv, v) + (Kv, v) . (2.7)
It is well known (use the quadratic formula) that the roots of a second order polynomial
with nonzero positive real coefficients have nonzero negative real parts. Since M, (7, and K
are positive operators and v ^ 0, it follows from (2.7) that 9?(A) < 0. Therefore, all roots of
det[0] have nonzero negative real parts arid the mechanical system is stable. •
If K is positive and singular, then the mechanical system in (2.4) is not stable. To see
this, notice that ft(0) = K and thus det[Q(0)] = 0. Hence, it follows from Theorem 2.2.1
that the mechanical is not stable. However, the following result demonstrates that stability
is possible when C is positive and singular.
2.3. INPUT-OUTPUT STABILITY 23

Theorem 2.2.3 Assume that M and K are strictly positive operators on Q and C is a
positive operator on Q. Then the corresponding mechanical system in (2.4) is stable if and
only if the kernel of
[K^M\ (2.8)

is zero for all real numbers a > 0.

PROOF. If v is a nonzero vector in the kernel of [ K — aM C ] r for some a > 0, then

f2(A)t> = 0 for A = i\/a where fi(A) is defined in (2.6). According to Theorem 2.2.1, this
readily implies that the mechanical system is not stable.
On the other hand, if the mechanical system in (2.4) is unstable, then there exists a scalar
A and a nonzero vector v in Q such that 3R(A) > 0 and f2(A)u = 0. Thus (2.7) holds. We
claim that (Cv,v) = 0. Since C is positive, (Cv,v] > 0. If (Cv,v) > 0, then (0(A)u,t>) is a
quadratic polynomial with nonzero positive coefficients. Thus, 5R(A) < 0 which contradicts
that fact that 3fc(A) > 0; hence (Cv,v) = 0. So, \2(Mv,v] + (Kv,v] = 0 and hence,
A = ±i^/a where a — (Kv,v)/(Mv,v). Since (Cv,v) = 0, we have Cv = 0. (To see this,
simply notice that 0 = (Cv,v) — ||C1//2v||2 where C1/2 is the positive square root of C.
Therefore, C1/2v = 0 and Cv = Cl/2Cl/2v = 0.) We now obtain (-aM + K)v = fi(A)v = 0.
Thus v is in the kernel of [K — aM, C]tr. This completes the proof. •

Example 2.2.2 Recall the two story structure of Example 2.2.1. There we assumed that
both of the damping coefficients c\ and c<2 were nonzero and positive. However, an application
of the above theorem shows that for stability, one only needs one of the damping coefficients
to be nonzero and positive, and the other to be greater than or equal to zero.

2.3 Input-output stability

This section is concerned with input-output stability of linear systems. To begin, consider a
linear map T defined by

y(t] = (Tu)(t) = I G(t-r)u(r) dr (2.9)

Jo
where G is a Lebesgue measurable function with values in £(U, y) and the input u(t) has
values in U while the output y(t) has values in y. Throughout we always assume that both
U and y are finite dimensional Hilbert spaces. We say that the input-output map T defined
in (2.9) is stable, if T defines a bounded operator from L2 ([0, oo),lf) into L 2 ([0, oo), }>)• We
say that G is in Ll(U,y} if G is a Lebesgue measurable function on [0, oo) with values in
£(U,y) and its L1 norm
oo
\\G(t)\\dt (2.10)
/
is finite. This sets the stage for the following useful result.

Proposition 2.3.1 IfG is in Ll(U, y), then the linear map T in (2.9) is a bounded operator
fromL2([0,oo),U) into L2 ([0,oo),y). Moreover, \\T\\ < \\G\\i.
24 CHAPTER 2. STABILITY

PROOF. If y = Tu, then an application of the Cauchy-Schwartz inequality yields

x 2
dr X

\\G(t-r)\\^\\G(t-r)\nu(r)\\dr

< \\G\1 T \\G(t-r) ur2dr

Jo
The Cauchy-Schwartz inequality was used to obtain the second inequality. By integrating
the previous inequality and changing the order of integration, we have
OO fOO ft

\\y(t}\\2dt < \G\J / |G(t-r)|| u(r)\\2 drdt

/ Jo Jo
OO fOO
/ \\G(t~r)\\dt\\u(r)\\2dr
/ Jr
- \\G\

Therefore, ||y | < ||G||i||w||. This completes the proof. •

The following result yields another useful bound for the operator T.
Corollary 2.3.2 Let G be a Lebesgue measurable function with values in £(C,y) satisfying

\\G(t)\\ <me-2at (te[0,oo)) (2.11)

for some m > 0 and a > 0. Then G is in L1 (C, y). Furthermore, the linear map T in (2.9)
defines a bounded operator from L 2 [0,oo) into L2 ([0,oo),>') and

\\T\\ < ||G||i < \\eatG\yV2^ < m/2a (2.12)

where || • \% denotes the L2 ([Q,oo),y) norm.

PROOF. Using the Cauchy-Schwartz inequality, we have
/ /-co \ 2
2
= / \G(t)\\dt)
\Jo /
/ fOO
= / \\eatG(t)\\e-atdt
\Jo ;

< r \\e G(t) \ dt- f e~2atdt

at 2

Jo Jo
= \\eatG\\22/2a < \\me-at\\2/2a = m2/4a2 .

This along with Proposition 2.3.1 readily yields the inequality in (2.12).
2.4. THE H°° NORM 25

Remark 2.3.1 Recall that we have equality in the Cauchy-Schwartz inequality if and only
if the corresponding vectors are linearly dependent. From the proof of the above inequality,
it should be clear that we have the equality ||G||i = ||eQ*G||2/\/2a in (2.12) if and only if
there is a scalar k such that HG^e"*)! = ke~at, or equivalently. ||G(t)| = ke~2at .
Now consider the linear system {^4, B, C, D} defined by
x = Ax + Bu and y = Cx + Du (2.13)
where A is on a finite dimensional state space X. As before, the input u(t) lives in U while
the output y(t) lives in y. If the initial condition is zero, then the input-output map R
corresponding to (2.13) is given by
rt
y(t) = (Ru)(t) = / CeA(t-^Bu(r)dT + Du(t). (2.14)
Jo
Motivated by this we say that {A,B,C, D} is input-output stable if R in (2.14) defines a
bounded operator from L2 ([0, oo),U) into L2 ([0, oo),y). Notice that if A is stable, that is,
if all the eigenvalues of A have nonzero negative real part, then G(t) = CeAtB is in I/1^, y).
Hence, the operator T in (2.9) corresponding to this G is bounded. Thus, R = T + D is
bounded. Therefore, if A is stable, then {A, B, C, D} is input-output stable. The converse
of this fact is not necessarily true. For example, choose A — I and B — C — D = 0. Then
clearly A is unstable while R = 0 obviously bounded. Summing up our previous analysis
yields the following result.
Proposition 2.3.3 // all the eigenvalues of A have nonzero negative real part, then the
system {A, B, C, D} is input-output stable.

2.4 The H°° norm

To compute the norm of the input-output operator T in (2.9), we need the Hardy space
H°°(U,y). Throughout H°°(U,y) is the Hardy space of all functions G which are analytic
for 3?(s) > 0, take values in £(U, y) and whose H°° norm
||G|| 00 : = sup{||G(5)||:»(5)>0} (2.15)
is finite. Notice that a rational function G is in H°°(U, y) if and only if G is a proper rational
function whose poles are in the open left half plane of C, that is, {s G C : •R(s) < 0}. Now
we are ready to state the following classical result; see Chapter IX, Section I of Foias-Frazho
[39].
Theorem 2.4.1 Suppose G is a Lebesgue measurable function with values in jC(U,y) and
D is in C(U,y). Let R be the linear map defined by

y(t) = (Ru)(t) = / G(t-r)u(r) dr + Du(t] (2.16)

Jo
where u is in L2 ([0, oo),W). Then R defines a bounded operator from L2 ([0, oo),Lt) into
L2 ([0,oo), [V) if and only if the Laplace transform G of G is a function in H°°(U,y). In
this case \\R\\ = \\G + D]}^.
26 CHAPTER 2. STABILITY

The following result can be used to obtain a proof of Proposition 2.3.1.

Lemma 2.4.2 If G is in Ll(U,y), then its Laplace transform, G is in H°°(U,y). Moreover,

PROOF. Recall that the Laplace transform G of G is defined by

oo
e~stG(t] dt .
/.
Now let us show that G analytic for §R(s) > 0. Consider any complex number s with
3fc(s) > 0. Since \e~st\ < 1, it follows that e-siG(t}\\ = \e~si\\\G(t}\\ < \\G(t)\\. Moreover,
there is a finite number (3 such that \te~st\ < /3 for all t > 0; hence \\te-stG(t}\\ < P\\G(t)\ .
Since G is in Ll(U,y), it now follows that the functions on [0,oo) given by e~stG(t) and its
derivative with respect to s are both in Ll(U,y}. So we can interchange the integral and
the derivative to show that the derivative of G exists for all 3?(s) > 0. Hence, G is analytic
in 3ft(s) > 0. Moreover, we have

Therefore, ||G| oo < \\G\\i and G is a function in H°°(U,y} which proves our claim. •
Consider any G in Ll(U,y). By the previous lemma, G is in H°°(U,y) and \\G\\n <
\G i- Theorem 2.4.1 now implies that the operator T in (2.9) defines a bounded operator
with \\T\ = ||G| oo. Hence, \\T\ < \\G\\i. This readily proves Proposition 2.3.1.
The following result is an immediate consequence of Theorem 2.4.1
Theorem 2.4.3 Let G be the transfer function for a finite dimensional system {A , B , C , D}
and R be the input-output map defined by
rt
(2.17)

Then {A, B,C, D} is input-output stable if and only if G is in H°°(U,y}. Moreover, in this
case R is a bounded operator from L'2 ([0, oo),U] into I/2 ([0, oo), y} and \\R\\ = | G||oo. In
particular, if all the eigenvalues of A have nonzero negative real part, then G is in H°°(U,y)
and {A,B,C,D} is input-output stable.
If G is the transfer function for {A, B, C, D}, then we say that G is stable if all the poles
of G are in the open left half plane, or equivalently, G is in H°°(lt,y).
If P is any self-adjoint operator on X', then A max (P) denotes the largest eigenvalue of P.
Now let G' be a Lebesgue measurable function with values in C(U,y}. Then d\(G] is the
scalar defined by
di(G) = sup{||Gu||i : u e W and ||u|| < 1} .
Note that d\(G] can be infinite. However, if G is in Ll(U, y], then d\(G] is finite and satisfies
di(G) < \\G\\!. IfU = C1, then d^G) =- \\G\\i. Finally, let N be an operator from U into
y, then the adjoint of N is denoted by TV*. Recall N* is the unique operator from y into U
defined by (Nu, y) = (u, N*y) for all u in U and y in 3^- This sets the stage for the following
result which is a generalization of Corollary 2.3.2.
Other documents randomly have
different content
 Kennedy had pulled a time-table out of his pocket and was
hurriedly consulting it.
“The service is very poor at this hour,” he remarked. “It will be an
hour before we can get the next train. We’ve missed the first by a
few minutes.”
“What do you suppose she’s up to now?” I speculated.
Kennedy shrugged silently.
We had finished dressing and for the moment there seemed to be
nothing to do but wait.
“By George!” Craig exclaimed, suddenly, starting for the door.
“Just the chance! Hardly anybody is about. We can get into her room
while she is gone. Come on!”
Paquita’s room, or rather suite, was on the floor above and in a
tower at the corner. It was difficult to get into, but from a porch at the
end of the hall we found that it was possible to step on a ledge and,
at some risk, reach one window. Kennedy did not hesitate, and I
followed.
As was to be supposed, the room was topsy-turvy, showing that
she had been at some pains to get away early and quick.
We began a systematic search, pawing with unhallowed fingers
all the dainty articles of feminine finery which might conceal some bit
of evidence that might assist us.
“Pretty clever,” scowled Kennedy, as drawer after drawer, trunk
after trunk, closet after closet, yielded nothing. “She must have
destroyed everything.”
He paused by a dainty little wicker writing-desk, which was
scrupulously clean. Even the blotters were clean, as though she had
feared some one might, by taking her hand-mirror, even read what
she blotted.
The scrap-basket had a pile of waste in it, including a couple of
evening papers. However, I turned it over and examined it while
Craig watched.
 As I did so I fairly pounced on a sheet of paper crumpled into a
ball, and eagerly straightened it out flat on the table.
“Humph!” I ejaculated in disgust. “Blank! Might have known she
wouldn’t leave anything in writing around, I suppose.”
I was about to throw it back when Kennedy took it from me. He
held it up to the light. It was still just a crumpled sheet of white paper.
He looked about. On a dressing-table stood an electric curling-iron.
He heated it and passed it over the paper until it curled with the heat.
Still it was just a blank sheet of paper.
Was he pursuing a will-o’-the-wisp?
For a moment he regarded it thoughtfully. “If I were in the
laboratory,” he ruminated, “I could tell pretty quick whether—
Wait!—that’s foolish. She hasn’t any laboratory here. Walter, fill that
basin with warm water.”
In the bottom of the basin Craig laid the sheet of paper and we
bent over it.
“Nothing doing,” I remarked, disappointed.
“Why not?” he returned, eagerly, turning the wet paper. “We had it
wrong side up!”
There, before our eyes, under the water, characters of some sort
were appearing.
“You can make a perfectly good sympathetic ink from linseed oil,
liquor of ammonia, and any of several other ingredients,” he said,
watching with me. “When writing with it dries it is invisible. Only
water will bring it out. Then when it dries it is invisible again. Look.”
I did, but could not yet make out what it was, except that it
seemed to be a hodge-podge of figures:
251533331514543245434412152515354433331552

543442254533442431521521243324432323154215
“It’s a cipher!” I exclaimed with that usual acumen that made
Kennedy smile indulgently.
 “Quite right,” he agreed, studying the peculiar scrawl of the
figures. “But if we are going to get to New York at anything like the
time she does, we must get that train. I can’t stop to decipher it now.
We’ll have plenty of time later in the morning. There’s no use staying
here, with the bird flown from the cage. I wonder whether Hastings is
up yet.”
The lawyer, who was not as young as he used to be, was not
awake, and it took some pounding on his door to wake him. As he
opened it sleepily he was prepared to give some one a piece of his
mind, until he saw that it was Kennedy and myself.
“Hello!” he suppressed a surly growl. “What’s the trouble?”
Quickly Craig told him of the strange departure of Paquita.
“Up to something again,” muttered Hastings, finding some one at
least on whom he could vent his spleen, although by this time he
was fully awake. “But, man, I can’t get away for that early train!”
“Oh, that’s all right,” reassured Kennedy. “I think it will be enough
if you come down on the express. But I wanted to tell you that when
Riley called up and said he was off after Paquita, I couldn’t think of a
place in the city that was more central than your office and I took the
liberty of telling him to call me up there, without thinking how early it
would be.”
“I’ll let you have the key,” returned Hastings, taking one from a
ring. “I’ll join you as soon as I can get away.”
“Just what I wanted,” commented Kennedy, as we left the lawyer
and hurried down to the dining-room for the few remaining minutes
before the hotel ’bus left for the station. “Besides, I wanted to get
there when no one was around, so that I could have a chance to look
at that confounded detectaphone again. Whoever it was who
installed it was clever.”
“Might not that be the purpose of Paquita’s trip to New York?” I
queried.
“I was thinking of that. Between us, Riley and ourselves ought to
be able to find that out.”
 There was just time for a hasty bite of breakfast and we went into
the dining-room, where Burke was evidently looking for us, for he
came over and sat down. No one else was about and he felt free to
talk.
“If you’re going,” he decided, after telling us of Riley’s report to
him also of Paquita, “I think I had better stay. We ought not to let any
of them remain here unobserved—not after last night,” he added.
“Quite right,” agreed Kennedy. “Have you heard anything more
about the attack on Winifred?”
Burke negatived. He was still sore at Sanchez, who seemed to
have come out of the affair with credit. I fancied that if ever the
sallow-faced man ran afoul of Burke it would go hard with him.
“I didn’t get that business straight last night,” mused the detective.
“Why should any one have wanted to kidnap Winifred? It couldn’t
have been to hurt her—for there was plenty of opportunity to do that.
It must have been to hold her somewhere and force some one to do
something. What do you think of that, Kennedy?”
“Your reasoning is very logical,” agreed Craig.
“There is only one thing missing—who was it and what was it
for?”
“Pretty large questions,” agreed Burke, good-humoredly now.
“There must have been some big reason for it. Well, I hope this trip
of Paquita’s proves to be the key to something. I almost wish I had
told Riley to stay. I’d like to go with you.”
“No,” reassured Craig. “It’s better that you should be here. We
must not leave any loopholes. You’ll communicate with me if
anything happens?”
Burke nodded and glanced hastily at his watch as a hint to us to
hurry. With a parting assurance from him, we made the dash for the
train in the hotel ’bus.
The crisp morning air as we spun up to the station was a tonic to
Kennedy. He seemed to enjoy the excitement of the chase keenly
and I must admit that I, too, felt the pleasing uncertainty of our
errand.
I had found by this time that there was an entirely different crowd
that regularly took each train. None of those whom we had seen the
previous day on the express were on this train, although I felt sure
that some of them at least would take their regular trip to the city
later, especially Shelby. Whatever happened, at least we were ahead
of them, although I doubted whether we would be ahead of Paquita
unless she had some trouble on the road.
Nothing was to be gained by the study of the other passengers,
and there was not even a chair car on the accommodation. The
papers had not arrived from New York in spite of the fact that
Westport was not very far out, and the time consumed in stopping at
every station on the road seemed to hang heavy.
Kennedy, however, was never at a loss for something to do. We
had no more than settled ourselves in the smoker with its seats of
hot, dirty, worn, antiquated railroad plush, when he pulled from his
pocket a copy he had made of the figures that had appeared on the
wet paper.
In a moment he was deeply engaged in a study of them, trying all
manner of tricks, combining them, adding them, setting figures
opposite the letters of the alphabet, everything that could occur to
him on the spur of the moment, although I knew that he had worked
out a scientific manner of reading any cipher. Still, his system of
deciphering would take time, and in the brief interval of the railroad
journey it was his intention to see whether he might not save the
labor and perhaps stumble on some simple key.
Evidently the cipher was not so simple. One after another he used
up sheets from his loose-leaf note-book, tearing up the scrawls and
throwing them out of the window, but never seeming to become
discouraged or to lose his temper at each fresh failure.
“I can’t say I’m making much progress,” he admitted, finally,
closing his note-book and taking from his wallet carefully the original
crumpled sheet I had found in the scrap-basket. “There’s just one
thing I’d like to try—not to decipher it, for that will take time, I see—
but to see if there is anything else that I missed as I looked at it so
hastily up there in that room.”
At the ice-water cooler, which never had any ice in it, nor cups
about it, he held the sheet of paper for a moment under the tepid
running water. Since he had first wet it, it had dried out and the
figures were again invisible. Then he returned to our seat and soon
was deep in the study of the original this time.
“I can say one thing,” he remarked, folding the cipher carefully so
as not to tear the weakened fibers, as we rolled into the New York
terminus, “the person who wrote that thing is a crook—has the
instincts of a spy and traitor.”
“How do you know that?” I inquired.
“How?” he repeated, quietly, glancing up sharply from a final look
at the thing. “Did you ever hear of the science of graphology—the
study of character in handwriting? Much the same thing applies to
figures. It’s all there in the way those figures are made, just as plain
as the nose on your face, even if the meaning of the cipher is still
hidden. We have a crook to deal with, and a very clever one, too,
even if we don’t know yet who it is. It’s possible to hide a good deal,
but not everything—not everything.”
From the station Kennedy and I went immediately down to the
office of Hastings. It was still very early and few offices were
occupied. Kennedy opened the door and, as I anticipated, went
directly to the spot in the office where he had unearthed, or rather
unwalled, the detectaphone transmitter.
There was not a chance that any one would be listening at the
other end, yet he proceeded cautiously. The transmitter had been
placed close to the plaster which had not been disturbed in
Hastings’s office. So efficient was the little machine that even the
plaster did not prevent sound waves from affecting its sensitive
diaphragm.
“But how could it have been put in place?” I asked as Kennedy
explored the hole he had made in the wall.
 “That’s new plaster back there,” he pointed out, peering in. “Some
one must have had access on a pretense to the next office and
placed the transmitter that way, plastering up the wall again and
painting it over. You see, Hastings wouldn’t know about that.”
“Still,” I objected, “any one going in and out of the next office
would be likely to be seen. Who has the office?”
“It’s no use to look,” replied Kennedy, as I started for the hall.
“They are as ignorant as we are. See—the wire doesn’t go there. It
goes horizontally to that box or casing in the corner which carries
steam-pipes. Then it goes down. It’s not likely it goes down very
many floors. Let’s see what’s under us.”
The office beneath bore on its doors the name of a well-known
brokerage firm. There was no reason to suspect them, and Kennedy
and I walked down another floor. There, in a little office, directly
under that occupied by Hastings, gilt letters announced simply
“Public Stenographer Exchange.” Without a doubt that was the other
end of the eavesdropper.
“There’s no one in yet, sir,” informed a cleaning woman who
happened to see us trying the door.
Kennedy was ready with a story. “That’s too bad,” he hastened,
with a glance at his watch. “They want to sublet it to me and I’d like
to look at it before I decide on another office at nine o’clock.”
“I can let you see it,” hinted the woman, rattling a string of keys.
“Can you?” encouraged Kennedy, slipping a silver coin into her
hand. “Thank you. It will save me another trip.”
She opened the door and we saw at once why Kennedy’s chance
story had seemed so plausible. Whatever furniture had been there
had been moved out, except a single plain chair and a very small
table. But on the table stood a box, the receiving end of the
detectaphone. It was the eavesdropper’s station, all right.
The woman left us a moment and we made the best of the
opportunity. Not even a scrap of paper had been left. Except for what
greeted us on our first entrance, there was nothing.
 Who had rented the place? Who had listened in, had heard and
anticipated even our careful frame-up?
Could it have been that this was the objective of the hasty visit of
Paquita, that it had been she who was so eager to destroy the
evidence of the eavesdropping on Hastings?
It was galling to have to stand here in inaction at a time when we
felt that we might be learning much if we had only so much as a hint
where else to look.
“The easiest way of finding out is to watch,” concluded Kennedy.
“We can’t just stand about in the hall. That in itself will look
suspicious. You wait here a few minutes while I see if I can find the
agent of the building.”
Around a bend in the hall I waited, trying to seem interested more
in some other office down the hall. No one appeared, however,
looking for any of the offices, and it was only a few minutes before
Kennedy returned.
“I found him,” he announced. “Of course he could tell me next to
nothing. It was as I had supposed, just some one who was an
emissary of our criminal. I doubt if it would do us much good to catch
the person now, anyway. Still, it’s worth while taking a chance on. A
girl who said she was a typewriter and stenographer hired the place,
and paid for it in cash in advance. I managed to persuade the agent
to let me have the key to this vacant office opposite. We can watch
better from that.”
We let ourselves into the opposite office, which was bare, and I
could see that I was in for a tiresome wait.
No one had arrived yet in Hastings’s office and Kennedy was
eager to receive some word from Riley as well as watch the
eavesdropping plant. Accordingly, he left me to watch while he
returned to the lawyer’s office.
Every footfall in the hall raised my hopes, only to dash them again
as the new-comer entered some other office than the one I was
watching.
 XVII

THE CIPHER LETTER

Irksome though it was to be compelled to do fruitless watching in a
vacant office, there was nothing to do but stick at it. What Paquita
might be up to was a mystery, but I knew that until we heard from
Riley we could have only the most slender chance to locate her in
the big city.
It was perhaps half an hour later when, to my relief, Kennedy
returned, bringing with him a strange man. I looked at him inquiringly.
“You’re just wasting time here, Walter,” Craig explained. “I’ve got
one of the Secret Service men here in the city to relieve you of your
job. But I very much suspect that, after what happened last night,
whoever had that place across the hall is through and would rather
lose the detectaphone receiver than risk being caught.”
“Have you had any word from Riley?”
“Not a word. I’m getting anxious,” he replied, turning to the new
man and instructing him what to do.
Kennedy was eager to get back, in case there might be a hasty
call about Paquita. I could see, too, that he was convinced that we
were baffled, at least as far as discovering who had been using the
detectaphone was concerned.
We returned quickly to Hastings’s office, which was still deserted,
and there, as we waited nervously, Kennedy drew forth the cipher
and began to study it again, but this time on an entirely different line,
following his own scientific principles, which he had laid down after
investigating the work of other expert decipherers.
My hopes rose momentarily when we heard footsteps in the hall
and the door was burst open. It was, however, merely a messenger-
boy.
“Telegram for Mr. Kennedy,” he shouted, penetrating even the
sacred inner office of Hastings.
Craig tore open the yellow envelope, read the message, and
tossed it over to me. It was from Burke at Westport.
“Wireless operators at Seaville Station,” it read, “report strange
interference. May be in reference to telautomaton. Will keep you
advised if anything happens.”
The possibility of a new twist to events was very fascinating,
though I did not understand it. I was just about to question Kennedy
about the telautomaton when the door opened again. This time it
was Hastings himself.
“Has there been any word?” he asked, eagerly.
“Nothing so far,” replied Craig. “You came on the express, I
suppose?”
“Yes,” he replied, his face wearing a puzzled expression. “I don’t
quite understand what is going on.”
“What in particular?” queried Craig, seeing that there was
something on Hastings’s mind.
“Why, Shelby, of course,” he answered. “Some change has taken
place in him. He’s not like the Shelby I used to know. Yesterday he
came in to town. He was on the train again to-day. I wasn’t the only
one who noticed it. Johnson Walcott was on the train, too. He
noticed it—called my attention to it, as a matter of fact. I saw some of
the younger men, too. Shelby as a regular commuter is a joke to
them. But it’s more than a joke, I’m thinking. Shelby never came near
Wall Street—or Broad Street—before. But now they tell me he
seems to be taking an active interest in the Maddox Munitions stock
on the curb. I don’t understand it.”
“Could he be trying to put through some deal?” I inquired, hastily.
“Perhaps he’s trying to get the control his brother would have had.”
 “I don’t doubt that he has some such scheme,” agreed Hastings.
“But—well, what do you say, Kennedy? Doesn’t it look suspicious, so
soon afterward? It may be real ambition, now. He may have
changed. But—”
Hastings’s “but” meant volumes.
Just then the telephone rang and the lawyer answered it, handing
the instrument over to Kennedy.
We listened eagerly. It was the first long-delayed report of Paquita
from Riley, and as Kennedy pursued the one-sided conversation that
we heard I gathered that, far from clearing up things, the actions of
Paquita had further muddled them.
Hastings glanced at me and shook his head sagely, whispering,
“That’s a clever and a dangerous woman: When she looks most
innocent is the time to be wary.”
I tried to pay no attention to his banal remarks, but still was
unable to follow, from what I heard, the course of the report from
Riley.
Finally it seemed as if Kennedy were cut off in the middle of a
remark or that Riley had hung up suddenly. Kennedy jiggled the
hook but was unable to get any one back again, though Central tried
for some time.
“What was it?” I asked, keenly interested.
“I’m afraid she’s putting one over on us again,” commented
Kennedy as he hung up the receiver.
“How’s that?” I asked.
“Why, it’s evidently a purposeless visit to the city, as nearly as I
can make out. Riley followed her in—had no difficulty. In fact, he
thinks that she knew she was being followed before they reached the
turnpike from Westport.”
“Where did she go after she got here?” I asked, hoping that at last
there was some clue that might lead to the “gang” which Burke
suspected, but which I was almost tempted to believe was mythical.
 “Just stopped at her city apartment,” returned Craig. “There
wasn’t any telephone handy and Riley was afraid to leave her for
fear she might come out and get away before he could get back. It
was very early. When it came time for the offices to open she made a
call at her theatrical agents again. After that she came down-town.
She wasn’t far away from us here. This will interest you, Mr.
Hastings.”
Hastings needed no prompting. He was already interested.
“Riley found her talking to a clerk in a brokerage house—Dexter &
Co. You know them?”
“Slightly. I wonder what that can mean.”
“Perhaps something to do with Shelby—at least Riley thinks so. It
was while she was talking to the clerk that he got his first chance to
telephone me. What cut him short was that he could see from the
telephone booth that she was starting away. He had to go, but he did
get time to say that he had just seen Shelby Maddox enter the same
building, though Shelby didn’t see Paquita.”
“Did she see him?”
“I suppose so. That must have been why she went away so
quickly. I suppose she didn’t want to be seen.”
“What can that girl be up to now?” considered Hastings. “You may
just rest assured that it is something devilish.”
“Any word from Sanchez?” I asked, remembering my own
experience the time I had tried to trail Paquita.
“Nothing so far,” replied Kennedy. “Riley was looking for him, but
hasn’t seen a trace of him. Except for the visit down-town, Paquita
seems to be just going about as though giving Riley something to do.
He thinks it’s mighty strange she doesn’t try to throw him off. Really,
she seems to want to be shadowed.”
“How can that be?”
Kennedy shrugged. “I don’t know. Riley promised to call up the
next chance he got.”
 “Why not go over to Dexter’s?” suggested Hastings.
“She can’t be there,” returned Kennedy. “If she was, Riley would
have had a chance to make a second call. Therefore I reason that
she must have gone away after she had seen the clerk and when
Shelby appeared. I think I’ll stay here awhile, until I hear again—
especially as I have nowhere else to go,” he decided, pulling out the
cipher from his pocket again. “We may hear some more about
Shelby and his schemes.”
Kennedy had now fallen into an earnest study of the peculiar
cryptogram which we had discovered.
“I suppose you’ve noticed that there’s no figure above five in it,”
he remarked to me, looking up for an instant from several sheets of
paper which he was covering with a hopeless jam of figures and
letters.
“I had not,” I confessed. “What of it?”
“Well, I’ve tried the numbers in all sorts of combinations and
permutations. They don’t work. Let me see. Suppose we take them
in pairs.”
For several moments he continued to figure and his face became
continuously brighter.
“There are six pairs of 33’s,” he remarked, almost to himself.
“Now, it’s well known that the letter ‘e’ is the most commonly used
letter. That’s the starting point usually in working out a cipher. Wait—
there are eight ‘15’s’—that must be ‘e.’ Yes, the chances are all for it.
Now what letter is 33, if any?”
He appeared to be in a dream as he recalled from his studies of
cryptograms what were the probabilities of the occurrence of the
particular letters. Suddenly he exclaimed:
“Perhaps it’s ‘n’—let’s see.”
Hastily he wrote down some letters and numbers in the following
order—“25enne1454.”
 He looked at it for a moment, and then his face registered the
dawn of an idea.
“By George!” he exclaimed, “we don’t have to go any further! I
have it. It’s my own name—Kennedy. Let me see how that works. I
believe it’s the system we call—”
Kennedy was again interrupted by the entrance of the
messenger-boy with another telegram. He tore it open and, as I
expected, it was a second message from Burke.
“Seaville Station has reported interference to Government. Just
received orders Washington to take up investigation. Not wireless
messages that interfere. Some mystery. When can you come out?”
Kennedy read and reread the message. To neither Hastings nor
myself did it convey any idea upon which we could build. But to
Kennedy, seemingly, it suggested a thousand and one things.
It was evident that the appeal from Burke had moved Kennedy
very much. Paquita had lured us into town, but I cannot say that it
was giving us much to show for our pains.
“What do you suppose that message can mean?” I questioned.
“What does Burke mean about the telautomaton?”
“I can’t say at this distance. There must be more to it than he has
put into the telegram. But at least it is possible that the men at the
station have stumbled over some attempt to use the wireless in
testing out the little model. It’s pretty hard to tell. Really, I wish I was
out there. A clue like this interests me much more than our little
adventuress.”
Kennedy had scarcely laid down the message from Burke when
the telephone tinkled again. He seized the receiver expectantly. By
his excitement I could see that it was Riley again.
“Yes, Riley,” we heard him answer. “Where are you now?”
The conversation was rapid-fire. As Kennedy hung up his face
showed considerable interest.
 “That woman is just making sport of Riley,” exclaimed Kennedy,
hotly, facing us in perplexity.
“Why, what is she doing?”
“Seems to be aimlessly driving about the city. I’ll bet she is just
laughing at him. I wonder what the game may be?”
“Where is she now?”
“Up-town again. I suppose that we could jump up there and
probably catch Riley somewhere, by keeping in touch with this office,
if both of us kept calling up here. But what good it would do I can’t
see. I’m disappointed. This thing has degenerated into a wild-goose
chase.”
His eye fell on the telegram from Burke, and I knew that the two
things had placed Kennedy in a dilemma. If he might have been in
two places at once, he would have been satisfied. Should he drop
everything and go to Burke or should he wait for Riley?
“We’ll let the cipher decide,” concluded Kennedy, turning to the
scribbled papers before him.
“What is the cipher system?” I asked, mechanically, my head
rather in a whirl at the fast-crowding events.
“Don’t you understand?” he cried, almost gleefully, working at the
solution of the secret writing. “I’ve got it! How stupid of me before not
to think of it. Why, it’s the old checkerboard cipher again!”
Quickly he drew on paper a series of five squares horizontally and
five vertically and filled them in with the letters of the alphabet,
placing I and J in the same square, thus using twenty-five squares.
Over the top he wrote the numbers to 5 and down the side he did the
same, as follows:

1 2 3 4 5

1 A B C D E

2 F G H IJ K
 3 L M N O P

4 Q R S T U

5 V W X Y Z

“Do you see?” he cried, eagerly. “The letter ‘e’ is in the first row,
the fifth letter—15. The letter ‘n’ is the third letter in the third row—
33. Why, it’s simple!”
It might have been simple to him now, but to Hastings and myself,
as Kennedy figured the thing out, it was little short of marvelous. For
all we could have done it, I suppose the blank scrap of paper would
still have been a hidden book.
We were crowded about Kennedy, eagerly watching what his
deciphering might yield, when the office-boy announced, “Mr. Shelby
Maddox to see you, Mr. Hastings.”
Kennedy quickly covered the papers on which he was writing with
some others on the desk, just as Shelby entered.
“Is Kennedy here?” cried Shelby. “Oh—I thought maybe you might
be. They told me that you’d gone early to the city.”
Our greeting was none too cordial, but Shelby either did not
notice it or affected not to do so.
“I wanted to ask you about that kidnapping,” he explained. “You
see, I wasn’t about when they found Mito, and it wasn’t until later that
I heard of it and the attempt on Winifred. What do you suppose, Mr.
Kennedy, was the reason? Who could have wanted to carry her off?”
Kennedy shrugged. “So far I haven’t been able to give a final
explanation,” he remarked, keenly.
“Then the kidnappers got away clean?” asked Shelby.
“It was very clever,” temporized Kennedy, “but I would hardly say
that there is no clue.”
 Shelby eyed Craig keenly, as though he would have liked to read
his mind. But Kennedy’s face did not betray whether it was much or
little that he knew.
“Well,” added Shelby, “all I’ve got to say is that some one is going
to get into trouble if anything happens to that girl.”
I was listening attentively. Was this a bluff, or not? From the
expression on Hastings’s face one would have said that he was
convinced it must have been Shelby himself who kidnapped her. I
wondered whether it was wholly interest in Winifred that prompted
Shelby’s visit and inquiry.
“At any rate,” he went on, “you’ll all be watching now against a
repetition of such a thing, won’t you? I don’t need to remind you,
Kennedy, of your promise when I talked to you before?”
Craig nodded. “I’ll give you a square deal, Mr. Maddox,” returned
Craig. “Of course I can’t work for two people at once. But I shall do
nothing for any client that I am not convinced is perfectly right. You
need not fear for Miss Walcott as long as I can protect her.”
Maddox seemed to be relieved, although he had found nothing
that pointed to the origin of the attack. Or was it because of that?
He glanced at his watch uneasily. “You’ll pardon me,” he said,
rising. “I had a few minutes and I thought I’d drop in and see you. I
must keep an appointment. Thank you for what you have said about
Winifred.”
As he withdrew I shot a hasty glance at Craig. Should I follow
him? Kennedy negatived.
Apparently not even the intrusion of Shelby had got out of his
mind either the dilemma we were in or the hidden message that he
seemed on the point of reading.
“An engagement,” commented Hastings, incredulously. “Since
when has Shelby had important engagements? More than likely it is
something to do with this Paquita woman.”
There was no mistaking the opinion that Hastings had of the
youngest scion of the house of Maddox. Nor was it unjustified.
Shelby’s escapades had been notorious, though I had always
noticed that, in the aftermath of the stories, Shelby was quite as
much, if not more, sinned against than sinning. Young men of his
stamp are subject to many more temptations than some of the rest of
us. If Shelby were coming through all right, I reflected, so much the
greater credit for him.
Kennedy either shared my own feelings toward Shelby or had
decided that he was not at present worth considering to the delay of
something more important.
I looked over his shoulder, fascinated, as he fell to work again
immediately on the cipher with the same zest which he had
displayed before Shelby’s interruption.
Rapidly Kennedy translated the figures into letters and, as each
word was set down on paper, became more and more excited.
Finally he leaped up and seized his hat.
“Confound her!” he exclaimed, “that explains it all! Look!”
Hastings and I read what he had written:
“KENNEDY MUST BE KEPT IN NEW YORK UNTIL WE FINISH
HERE.”
 XVIII

THE RADIO DETECTIVE

Here, it seemed, was a new danger. Was it to be taken as a proof of
Burke’s theory that some one, perhaps a gang, was back of Paquita!
I was almost inclined to Hastings’ opinion, for the moment. What was
the reason that Shelby had been so interested in Kennedy as to
seek him out even in the office of the lawyer of his brother who hated
him?
I could evolve no answer in my own inner consciousness for the
questions. As far as I could see we were still fighting in the dark, and
fighting an unknown.
Kennedy quickly chose one horn of the dilemma that had been
presented to him. Both the wording of the cipher and Burke’s
enigmatic message regarding the wireless which came so close on
its heels quite decided him to hurry back to Westport—that is, if one
might so call traveling on midday trains that lounged along from
station to station.
We left Hastings in a high state of excitement. Some pressing
business prevented his immediate return to Westport, and Kennedy
was evidently rather pleased than otherwise, for he did not urge him
to go.
“There’s just one thing that I must stop for, and we shall have
plenty of time, if we don’t waste it,” he planned. “I must go to the
laboratory. There’s some stuff there I want to take out if, as I foresee,
we are to have to deal with wireless in some way. Besides, I may
need some expert assistance and I want to arrange with one of the
graduate students at the university, if I can.”
 In the laboratory he found what he wanted and began gathering it
into bundles, packing up some head telephone receivers, coils of
wire, and other apparatus, some of which was very cumbersome.
The last he placed in a pile by itself.
The door opened and a young man entered.
“Oh, Watkins,” Craig directed, as I recognized one of the students
who had attended his courses, “there’s a lot of apparatus I would like
you to take out to Westport for me.”
They talked briefly in a wireless jargon which I did not understand,
and the student agreed to carry the stuff out on a late train, meeting
us at the Harbor House. At the last moment Kennedy was off for the
railroad station. It was making close connections, but we succeeded.
The ride out was nerve-racking to us under the circumstances.
We had taken the bait so temptingly displayed by Paquita and had
gone to New York. Now we could not get back fast enough. We had
not been in the city long, it is true. But had it been too long? What
had happened out in the town we were anxious to learn. I felt sure
that in our absence some of the Maddoxes might well have
attempted something which our presence would have restrained.
Burke met us at the station with a car, so sure was he that
Kennedy would return immediately on receipt of his second
message, and it was evident that he felt a great sense of relief at
regaining Kennedy’s help.
As we spun along down from the station Kennedy hastened to tell
Burke what had happened, first about Paquita as Riley had reported,
then his deciphering of the cipher message, our failure to discover
anything in the scantily furnished office at the other end of the
detectaphone wire, Shelby’s visit, and the whole peculiar train of
circumstances.
Instead of going directly to the Harbor House Burke drove us
around by the hotel dock, where we saw that there was a stranger in
a power boat apparently waiting for him.
Kennedy was just finishing his recital of our unsatisfactory
experience as we approached.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookultra.com