Optimal Control of Dynamic Systems Driven by

Vector Measures: Theory and Applications 1st

Edition Ahmed install download

https://ebookmeta.com/product/optimal-control-of-dynamic-systems-
driven-by-vector-measures-theory-and-applications-1st-edition-
ahmed/

Download more ebook from https://ebookmeta.com

We believe these products will be a great fit for you. Click
the link to download now, or visit ebookmeta.com
to discover even more!

The Dynamic Morse Theory of Control Systems 1st Edition

Josiney A Souza

https://ebookmeta.com/product/the-dynamic-morse-theory-of-
control-systems-1st-edition-josiney-a-souza/

Optimal Control Theory: Applications to Management

Science and Economics, Fourth Edition Sethi

https://ebookmeta.com/product/optimal-control-theory-
applications-to-management-science-and-economics-fourth-edition-
sethi/

Handbook of Dynamic Data Driven Applications Systems

Volume 1 2nd Edition Erik P. Blasch

https://ebookmeta.com/product/handbook-of-dynamic-data-driven-
applications-systems-volume-1-2nd-edition-erik-p-blasch/

Advancing a Different Modernism 1st Edition S A

Mansbach

https://ebookmeta.com/product/advancing-a-different-
modernism-1st-edition-s-a-mansbach/
Super Simple Maths 1st Edition Dk

https://ebookmeta.com/product/super-simple-maths-1st-edition-dk/

Optimal and Robust Control: Advanced Topics with

MATLAB®, 2nd Edition Fortuna

https://ebookmeta.com/product/optimal-and-robust-control-
advanced-topics-with-matlab-2nd-edition-fortuna/

Demon Fall 1st Edition M J Haag

https://ebookmeta.com/product/demon-fall-1st-edition-m-j-haag/

The Safety and Quality of Abortion Care in the United

States 1st Edition And Medicine Engineering National
Academies Of Sciences

https://ebookmeta.com/product/the-safety-and-quality-of-abortion-
care-in-the-united-states-1st-edition-and-medicine-engineering-
national-academies-of-sciences/

The Economic Sociology of Development 2nd Edition

Andrew Schrank

https://ebookmeta.com/product/the-economic-sociology-of-
development-2nd-edition-andrew-schrank/
Transmedia Marketing From Film and TV to Games and
Digital Media 1st Edition Anne Zeiser

https://ebookmeta.com/product/transmedia-marketing-from-film-and-
tv-to-games-and-digital-media-1st-edition-anne-zeiser/
N.U. Ahmed
Shian Wang

Optimal Control
of Dynamic
Systems Driven
by Vector
Measures
Theory and Applications
Optimal Control of Dynamic Systems Driven
by Vector Measures
N. U. Ahmed • Shian Wang

Optimal Control of Dynamic

Systems Driven by Vector
Measures
Theory and Applications
N. U. Ahmed Shian Wang
University of Ottawa University of Minnesota
Ottawa, ON, Canada Minneapolis, MN, USA

ISBN 978-3-030-82138-8 ISBN 978-3-030-82139-5 (eBook)

https://doi.org/10.1007/978-3-030-82139-5

Mathematics Subject Classification: 34H05, 49J15, 49K15, 49M05, 37N35, 93C15

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
In memory of my parents, uncles, aunts,
sisters, brothers, and my wife Feroza who
gave so much.

Dedicated to my sons: Jordan and Schockley;

daughters: Pamela, Rebeka, Mona, and Lisa;
and my grandchildren: Reynah-Sofia,
Maximus, Achilles, Eliza, Pearl, Austin, Rio,
Kira, and Jazzmine.

Dedicated to my parents Guihua Xu and

Shiwu Wang, and my sister Man Wang.
Preface

There are many prominent areas of systems and control theory that include
systems governed by linear and nonlinear ordinary and functional differential
equations, stochastic differential equations, partial differential equations including
their stochastic counterparts, and, above all, systems governed by abstract dif-
ferential equations and inclusions. The remarkable advance of this field is due
to the unprecedented interest, interaction, and contribution of pure and applied
mathematics and physical and engineering sciences. We strongly believe that this
interaction will continue simply because there are many unsolved challenging prob-
lems and emerging new ones. Such problems are of great interest to mathematicians,
scientists, and engineers.
This book is concerned with the development of optimal control theory for finite
dimensional systems governed by deterministic as well as stochastic differential
equations (which may be subject to impulsive forces) controlled by vector measures.
Impulsive controls are special cases of controls determined by vector measures. The
book has two major parts. The first part deals with deterministic dynamic systems
including differential inclusions controlled by vector measures; the second part is
concerned with stochastic dynamic systems also controlled by vector measures.
For non-convex control problems, probability measure valued functions known as
relaxed controls are used. We consider the question of existence of optimal controls
and the necessary conditions of optimality whereby optimal control policies can be
determined.
In recent years, significant applications of systems and control theory have
been witnessed in areas as diverse as physical sciences, engineering, biological
sciences, social sciences, management, and financial engineering, among many
others. In particular, the most interesting applications have taken place in areas such
as aerospace (civilian, military), space structures (space station, communication
satellites), suspension bridges, artificial heart, immunology, power system, hydrody-
namics, plasma and magneto hydrodynamics, computer communication networks,
and intelligent transportation systems. The importance of applications whereby a
theory is tested and new theories are developed is clearly recognized. This book is

vii
viii Preface

devoted mainly towards the development of theory of measure-driven differential

equations and their optimal control by vector measures.
This book contains seven chapters. In Chap. 1, we present most of the basic and
important results from analysis required to read the book smoothly. The contents of
this chapter will also serve as a quick reference for readers familiar with the subject
and a valuable guide for those not so familiar. It contains many relevant results from
measure theory and abstract functional analysis used in the text. Chapters 2 and 3
are devoted respectively to linear and nonlinear dynamic systems subject to forces
determined by vector measures covering existence and uniqueness of solutions
and regularity properties thereof. Chapter 4 deals with the question of existence
of optimal controls from the class of regular controls (vector valued measurable
functions), relaxed controls (measure valued functions), and controls determined by
vector measures. We consider both fully and partially observed control problems. In
Chap. 5, we present necessary conditions of optimality for all the control problems
considered in Chap. 4. In Chap. 6, we consider systems governed by stochastic
differential equations controlled by vector measures and present numerous results
on existence of optimal controls and necessary conditions of optimality. Here, we
consider also both fully and partially observed control problems. Corresponding to
each of the necessary conditions of optimality given in Chaps. 5 and 6, we present
important convergence results guaranteeing convergence of any algorithm based
on the necessary conditions of optimality. In Chap. 7, we present some practical
examples of application with numerical results to demonstrate the applicability of
the theories developed in this book.
The authors hope that this book will inspire young mathematicians and math-
ematically oriented scientists and engineers to further advance the theory and
application of dynamic systems driven and controlled by vector measures.
Finally, we would like to appreciate Dr. Remi Lodh, the Editor of Mathematics
with Springer Verlag, for his continued support and excellent cooperation through-
out the publishing process.

Ottawa, ON, Canada N. U. Ahmed

Minneapolis, MN, USA Shian Wang
Preface ix

Table 1 List of Notations

Notation Description
C(I, R n ) Class of continuous functions defined on the interval I
and taking values in R n , p. 2
BX σ -algebra of Borel subsets of the set X, p. 4
M(, X) Class of measurable functions from  to X, p. 5
(, B, μ) A σ -finite measure space, p. 5
B∞ (I, R n ) Banach space of bounded measurable functions from I
to R n , p. 5
M(, B, μ) ≡ M Class of μ measurable real-valued functions defined on
, p. 6
MCT Monotone Convergence Theorem, p. 11
LDCT Lebesgue dominated convergence theorem, p. 13
LBCT Lebesgue bounded convergence theorem, p. 14
(, F , P ) Complete probability space, p. 15
(, F , Ft≥0 , P ) Filtered probability space, p. 16
Lp (, B, μ), 1 ≤ p ≤ ∞ Lp spaces, p. 18
Lp (, μ)∗ Dual of the Banach space Lp (, μ), p. 20
μ Total variation norm of the measure μ, p. 21
AC(I, R n ) Class of absolutely continuous functions from I to R n , p. 23
F ix(F ) Set of fixed points of the map F , p. 26
E∗ Continuous (topological) dual of any Banach space E, p. 29
coB Convex hull of any set B, p. 30
clcoB Closure of the convex hull of the set B, p. 30
Ext (K) Set of extreme points of the set K, p. 30
RNP Radon-Nikodym property, p. 31
RND Radon-Nikodym derivative, p. 116
I Sigma algebra of subsets of the set I , p. 31
Mca (I , R n ) Space of countably additive bounded vector measures
defined on I and taking values in R n , p. 31
Sp (F ) Class of Lp -selections of the multifunction F , p. 32
cc(R n ) Class of nonempty closed convex subsets of R n , p. 33
cb(E) Class of closed bounded subsets of E, p. 33
SISO Single-input single-output, p. 35
MIMO Multi-input multi-output, p. 36
TIS Time-invariant system, p. 38
TVS Time-variant system, p. 38
P W C(I, R n ) Class of piecewise continuous and bounded functions
defined on I and taking values in R n , p. 42
(continued)
x Preface

Table 1 (continued)
Notation Description
|μ|() Variation of measure μ on the set , p. 48
M(n × m) Class of n × m matrices with real entries, p. 49
Br Closed ball of radius r centered at the origin, p. 64
P W Cr (I, R n ) Class of right-continuous functions having left limits, p. 69
AI Algebra of subsets of the set I , p. 76
Ud Set of admissible controls, p. 79
Mad Set of admissible control measures, p. 80
usc Upper semi-continuity, p. 85
lsc Lower semi-continuity, p. 85
c(L1 (I, R n )) Class of nonempty closed subsets of L1 (I, R n ), p. 90
L∞ (I, R m ) Class of essentially bounded Lebesgue measurable
functions defined on I and taking values from R m , p. 93
Uad ≡ M(I, U ) Class of measurable functions with values in U , p. 93
M1 (U ) Space of probability measures on U , p. 99
∞ (I, M(U ))
Lw Class of weak-star measurable functions defined on I
and taking values from the space of Borel measures
M(U ), p. 99
rel
Uad Unit sphere in Lw ∞ (I, M(U )) chosen as the class of
relaxed controls, p. 99
Mbf a (U ×I , R m ) Space of finitely additive R m -valued vector measures
on the algebra U ×I of subsets of the set U × I , p. 105
L(R m , R n ) Vector space of bounded linear operators from R m
to R n , p. 106
∼
= Symbol of isomorphism, p. 116
Dad Class of admissible controls, p. 121
C(U ) Linear space of real-valued continuous functions on U ,
endowed with the supnorm topology, p. 130
M(U ) Space of countably additive bounded signed measures
on the Borel sigma field BU of U having bounded
total variation, p. 130
(Mbf a (I , R n ))∗ Topological dual of Mbf a (I , R n ), p. 147
A∗ (t) Adjoint of the matrix-valued function A(t), p. 148
B ∗ (dt) Adjoint of the matrix-valued set function B(dt), p. 148
(, F , P ) Complete probability space, p. 168
Ft≥0 Increasing family of complete subsigma algebras
of the sigma algebra F , p. 168
(, F , Ft≥0 , P ) Filtered probability space, p. 168
(continued)
Preface xi

Table 1 (continued)
Notation Description
EX Expected value of the random variable X, p. 168
L(B) Itô integral of B with respect to standard Brownian
motion, p. 172
L2 (F0 , R n ) Hilbert space of R n -valued F0 -measurable random
elements having finite second moments, p. 173
Hn ≡ L2 (, R n ) Hilbert space of R n -valued random variables, p. 177
a (I, H )
B∞ n Class of Ft -adapted R n -valued stochastic
processes with finite second moments, p. 177
Fad Class of admissible operators, p. 211
N (ϑ) Null set of the measure ϑ, p. 218
L∞ (, L(R m , R n )) Space of essentially bounded Borel measurable
functions with values in L(R m , R n ), p. 223
Contents

1 Mathematical Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 Vector Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.3 Normed Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 2
1.4 Banach Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
1.5 Measures and Measurable Functions . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
1.6 Modes of Convergence and Lebesgue Integral .. .. . . . . . . . . . . . . . . . . . . . 6
1.6.1 Modes of Convergence .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
1.6.2 Lebesgue Integral.. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8
1.7 Selected Results From Measure Theory .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 12
1.8 Special Hilbert and Banach Spaces . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.8.1 Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.8.2 Special Banach Spaces .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 21
1.9 Metric Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 25
1.10 Banach Fixed Point Theorems . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 30
1.11 Frequently Used Results From Analysis . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 33
1.12 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 38
2 Linear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
2.2 Representation of Solutions for TIS . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.2.1 Classical System Models.. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 44
2.2.2 Impulsive System Models.. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 46
2.3 Representation of Solutions for TVS . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
2.3.1 Classical System Models.. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 53
2.3.2 Measure Driven System Models .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 55
2.3.3 Measure Induced Structural Perturbation . . . . . . . . . . . . . . . . . . . . 58
2.3.4 Measure Driven Control Systems. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
2.4 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 64

xiii
xiv Contents

3 Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 67
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 67
3.2 Fixed Point Theorems for Multi-Valued Maps . . .. . . . . . . . . . . . . . . . . . . . 69
3.3 Regular Systems (Existence of Solutions) .. . . . . . .. . . . . . . . . . . . . . . . . . . . 71
3.4 Impulsive Systems (Existence of Solutions) . . . . .. . . . . . . . . . . . . . . . . . . . 78
3.4.1 Classical Impulsive Models .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 79
3.4.2 Systems Driven by Vector Measures . . . . .. . . . . . . . . . . . . . . . . . . . 83
3.4.3 Systems Driven by Finitely Additive Measures.. . . . . . . . . . . . . 89
3.5 Differential Inclusions .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 98
3.6 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 108
4 Optimal Control: Existence Theory . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
4.2 Regular Controls .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
4.3 Relaxed Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 116
4.4 Impulsive Controls I .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 121
4.5 Impulsive Controls II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 124
4.6 Structural Control .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 129
4.7 Differential Inclusions (Regular Controls) . . . . . . .. . . . . . . . . . . . . . . . . . . . 132
4.8 Differential Inclusions (Measure-Valued Controls) .. . . . . . . . . . . . . . . . . 136
4.9 Systems Controlled by Discrete Measures . . . . . . .. . . . . . . . . . . . . . . . . . . . 141
4.10 Existence of Optimal Controls . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 144
4.11 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 149
5 Optimal Control: Necessary Conditions of Optimality .. . . . . . . . . . . . . . . . . 151
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 151
5.2 Relaxed Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 152
5.2.1 Discrete Control Domain . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 160
5.3 Regular Controls .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 162
5.4 Transversality Conditions . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 165
5.4.1 Necessary Conditions Under State Constraints . . . . . . . . . . . . . . 168
5.5 Impulsive and Measure-Valued Controls . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 169
5.5.1 Signed Measures as Controls . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 170
5.5.2 Vector Measures as Controls .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 175
5.6 Convergence Theorem .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 177
5.7 Implementability of Necessary Conditions of Optimality . . . . . . . . . . . 179
5.7.1 Discrete Measures .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 179
5.7.2 General Measures . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 180
5.8 Structural Controls.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 183
5.9 Discrete Measures with Variable Supports as Controls .. . . . . . . . . . . . . 185
5.10 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 195
6 Stochastic Systems Controlled by Vector Measures . .. . . . . . . . . . . . . . . . . . . . 197
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 197
6.2 Conditional Expectations .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 198
Contents xv

6.3 SDE Based on Brownian Motion . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 200

6.3.1 SDE Driven by Vector Measures (Impulsive Forces) .. . . . . . . 207
6.4 SDE Based on Poisson Random Processes . . . . . . .. . . . . . . . . . . . . . . . . . . . 212
6.5 Optimal Relaxed Controls .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 216
6.5.1 Existence of Optimal Controls .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 218
6.5.2 Necessary Conditions of Optimality.. . . . .. . . . . . . . . . . . . . . . . . . . 223
6.6 Regulated (Filtered) Impulsive Controls.. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 231
6.6.1 Application to Special Cases. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 241
6.7 Unregulated Measure-Valued Controls . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 242
6.7.1 An Application . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 247
6.8 Fully Observed Optimal State Feedback Controls . . . . . . . . . . . . . . . . . . . 248
6.8.1 Existence of Optimal State Feedback Laws .. . . . . . . . . . . . . . . . . 249
6.8.2 Necessary Conditions of Optimality.. . . . .. . . . . . . . . . . . . . . . . . . . 252
6.9 Partially Observed Optimal Feedback Controls ... . . . . . . . . . . . . . . . . . . . 253
6.9.1 Existence of Optimal Feedback Laws . . . .. . . . . . . . . . . . . . . . . . . . 253
6.9.2 Necessary Conditions of Optimality.. . . . .. . . . . . . . . . . . . . . . . . . . 258
6.10 Bellman’s Principle of Optimality . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 267
6.11 Bibliographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 273
7 Applications to Physical Examples . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 275
7.1 Numerical Algorithms .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 276
7.1.1 Numerical Algorithm I .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 278
7.1.2 Numerical Algorithm II . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 279
7.2 Examples of Physical Systems . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 281
7.2.1 Cancer Immunotherapy . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 281
7.2.2 Geosynchronous Satellites . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 283
7.2.3 Prey-Predator Model . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 288
7.2.4 Stabilization of Building Maintenance Units . . . . . . . . . . . . . . . . 296
7.2.5 An Example of a Stochastic System . . . . . .. . . . . . . . . . . . . . . . . . . . 305

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 311
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 317
Chapter 1
Mathematical Preliminaries

1.1 Introduction

To fully describe the state of a process, natural or artificial, at any point of time,
one may need to quantify a multiplicity of variables which may be arranged in any
convenient and consistent order and call it a vector. This is particularly important
for correct and complete characterization of the temporal evolution of the state of
dynamic systems. This makes us interested in vectors and vector spaces. Here in
this chapter we present some relevant results from real analysis, measure theory,
and functional analysis frequently used in later chapters of this book. In the first few
sections, we present some introductory materials from real analysis and measure
and integration. This is intended to familiarize the readers with basic notations and
terminologies used throughout the book. This will also help the readers, especially
students from science and engineering, who may not have formal background on
mathematical analysis. For better readability, a list of notations frequently used in
the book is presented in Table 1.

1.2 Vector Space

We present here the most important axioms of a vector space. Let X denote an
arbitrary set and let F ≡ R/C denote the field of real or complex numbers. The set
X is said to be a vector space over the field F if it satisfies the following two basic
properties:

(i) : if x ∈ X and α ∈ F, then αx ∈ X, (1.1)

(ii) : if x, y ∈ X then x + y ∈ X. (1.2)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

N. U. Ahmed, S. Wang, Optimal Control of Dynamic Systems Driven by Vector
Measures, https://doi.org/10.1007/978-3-030-82139-5_1
2 1 Mathematical Preliminaries

For example,

  x1  
.
X ≡ R ≡ x = .. : xi ∈ R, i = 1, 2, · · · , n
n

xn

is the set of all ordered n-tuples called a real vector space or a vector space of
dimension n over the field of real numbers R. A simple example is the daily
production of a firm that produces n distinct products which may be quantified by
a vector x ∈ R+ n ⊂ R n , where R n denotes the positive orthant of R n . This may
+
vary from day to day. Letting m denote the vector of daily mean production of n
distinct goods and y the vector of actual daily production, one may consider the
vector x = y − m, which of course lies in R n . The elements of this vector can take
positive as well as negative values.

1.3 Normed Space

A normed space is a vector space X furnished with a norm N(·) ≡ · . The norm
N is a real-valued function defined on X which satisfies the following properties:

(i) :  x ≥ 0, ∀ x ∈ X, (1.3)
(ii) :  x = 0, if and only if x = 0, (1.4)
(iii) :  αx = |α|  x , ∀ x ∈ X, α ∈ F, (1.5)
(iv) :  x + y ≤ x  +  y , ∀ x, y ∈ X. (1.6)

Few examples of frequently used normed vector spaces are given below,

n
(1) R n with  x ≡ 2
i=1 xi ;
 1/p
n
(2) np , 1 ≤ p < ∞, with norm given by  x ≡ |x
i=1 i | p ; the sequence
 1/p
∞
space p , with norm,  x = i=1 |x i | p ;
(3) Let C(I, R n ) denote the class of all continuous functions f defined on the
interval I and taking values in R n . Furnished with the norm,

 f ≡ sup{ f (t) R n , t ∈ I },

it is a normed vector space;

1.5 Measures and Measurable Functions 3

 1/p

(4) Lp (I ), 1 ≤ p < ∞, with  f ≡ I |f (t)| dt
p ; and Lp () with norm

 1/p
 f ≡ |f (ξ )| dξ
p
,


where  ⊂ R n ;
(5) Weighted Lp spaces denoted by Lp (, ρ), where ρ is a nonnegative function
defined on . The norm for this space is given by
 1/p
 f ≡ |f (ξ )|p ρ(ξ )dξ . (1.7)


The most commonly used function spaces are C(I, R n ), Lp (I, R n ), p =

1, p = 2, and p = ∞.

1.4 Banach Space

Definition 1.4.1 (Cauchy Sequence) Let X be a normed space with norm denoted
by  ·  . A sequence {xn } ∈ X is said to be a Cauchy sequence if

lim  xn+p − xn = 0 for every p ≥ 1.

n→∞

Definition 1.4.2 (Banach Space) A normed space X is said to be complete if every

Cauchy sequence has a limit. A complete normed space is called a Banach space.
The vector spaces R n , np , 1 ≤ p ≤ ∞, with the norms as defined in the
previous section, are finite dimensional Banach spaces. The spaces {p , 1 ≤ p ≤
∞}, {Lp (I ), 1 ≤ p ≤ ∞} are infinite dimensional Banach spaces. The space
of continuous functions defined on any finite interval I and endowed with the
supremum norm,  f ≡ sup{|f (t)|, t ∈ I }, is an example of an infinite
dimensional Banach space. More examples will be seen later.

1.5 Measures and Measurable Functions

Lebesgue integration stands out as one of the most monumental discoveries of

mathematics of the twentieth century. There are functions which are integrable in
the sense of Lebesgue, but their Riemann integrals are not defined. Later we present
the basic spirit of Lebesgue integration. For detailed study, the reader is referred to
the celebrated books on measure and integration as indicated in the bibliographical
notes.
4 1 Mathematical Preliminaries

For any abstract set , let A denote the class of all subsets of the set  satisfying
the following conditions:

(1) : empty set ∅ ∈ A,

(2) : A ∈ A ⇒ A ∈ A where A =  \ A,
(3) : A, B ∈ A ⇒ A ∪ B ∈ A, A ∩ B ∈ A.

In other words, the class A is closed under complementation, finite union, and finite
intersections. In this case, A is called an algebra or a field of subsets of the set . It
is also called the finitely additive class. It is clear from (1) and (2) that  ∈ A.
Let B denote the class of subsets of the same set  satisfying the following
properties:

(1) : ∅ ∈ B,
(2) : A ∈ B ⇒ A ∈ B,
(3) : Am ∈ B ⇒ Am ∈ B.
m≥1

In this case, the class B is closed under countable union and intersections. It is called
the σ -algebra or the sigma-field of subsets of the set . Clearly, the class B ⊂ A. In
this book we use primarily the sigma algebra B and occasionally the algebra A. As
we shall see later, the class A is the proper domain for finitely additive measures,
while the class B is the correct domain for countably additive measures. We use
these in the study of differential equations driven by control measures. The class of
Borel sets B is the minimal completely additive class of sets containing all closed
subsets of . In fact the class of Borel sets can be constructed by repeated operation
of countable union and intersection of closed or open sets in . For an excellent
account of this, the reader is referred to Munroe [85].
The pair (, B) is called a (Borel) measurable space. Now we are prepared to
introduce (positive) measures. A set function

μ : B −→ [0, ∞]

is said to be a measure if for every ∈ B, μ( ) is defined and takes values from

[0, ∞], and for every 1 , 2 ∈ B with 1 ⊂ 2 , we have μ( 1 ) ≤ μ( 2 ) and
μ(∅) = 0. The measure μ is said to be countably additive if for any pairwise disjoint
sequence {Ai }i≥1 of B measurable sets,
 
μ Ai = μ(Ai ).
i≥1 i≥1
1.5 Measures and Measurable Functions 5

In contrast, the measure μ defined on the algebra A is said to be finitely additive if

the identity holds only for the union of finitely many disjoint sets. Loosely speaking,
if  ⊆ R n and ∈ B ≡ B , the volume (e.g., length for n = 1; area for
n = 2) of the set is its Lebesgue measure. We denote this by λ( ) = V ol( )
and call λ as the Lebesgue measure. For rigorous definition and construction of
Lebesgue measure see any of the books on measure theory, in particular, the books
by Munroe [85], Halmos [69], and Hewitt and Stromberg [71]. The standard method
of construction of any countably additive nonnegative measure involves, to start
with, (i) a nonnegative set function vanishing on the empty set, (ii) a sequential
covering class of subsets of the space  such as the class of all open subsets of
, (iii) construction of an outer measure, (iv) Caratheodory’s characterization of
measurable sets, and finally (v) restriction of the outer measure on to the class of
Caratheodory measurable sets. This gives a measure corresponding to any given
set function. In fact, in this way one can put any measure μ on the measurable
space (, B) turning this into a measure space which is written as (, B, μ). If
μ() < ∞, the measure space (, B, μ) is called a finite measure space, and if
μ() = 1, it is called a probability space. For  = R n , one can construct Lebesgue
measure following the same series of steps (i)–(v) as described above. In this case,
for the sequential covering class, one can choose the open cubes in R n and, for the
nonnegative set function, one can use the volume of cubes. In this way one obtains
the Lebesgue outer measure and finally the Lebesgue measure by restricting the
outer measure on the class of Caratheodory measurable sets.
Let X ≡ R n , and let BX denote the σ -algebra of Borel subsets of the set X. A
function f :  → X is said to be a measurable function (Borel measurable map) if
for every B ∈ BX ,

{ω ∈  : f (ω) ∈ B} ≡ f −1 (B) ∈ B .

We denote this class of functions by M(, X). Sometimes this is also denoted by
L0 (, X). One can show that this is a real linear vector space in the sense that

(1) : f ∈ M(, X) ⇒ αf ∈ M(, X) for all α ∈ R,

(2) : f1 , f2 ∈ M(, X) ⇒ f1 + f2 ∈ M(, X).

Note that the space of real-valued measurable functions, M(, R), is an algebra in
the sense that point-wise multiplication of any two real-valued measurable functions
is also a real-valued measurable function. Clearly, the class M(, X) is quite
general and includes continuous as well as discontinuous functions. Indeed, if both
 and X are metric spaces, a function f :  −→ X is said to be continuous
if the inverse image of every open set in X is an open set in . Thus, open sets
being measurable sets, continuous functions are also measurable. Since every open
or closed set and countable union and intersection of such sets are all Borel sets
by definition, continuous functions constitute a smaller subclass of the class of
measurable functions.
6 1 Mathematical Preliminaries

Now we can present some examples of infinite dimensional spaces. A measure

space (, B, μ) is called σ -finite if there exists a countable sequence of disjoint sets
{Bn } ∈ B such that μ(Bn ) < ∞ for each n ∈ N and  = ∞ n Bn , while μ() =
∞. A class of linear spaces occasionally used is denoted by Lpoc (, B, μ), 1 ≤ p <
∞, where (, B, μ) is a sigma finite measure space. It consists of B-measurable
functions which are only locally p-th power integrable. That is, for any set ⊂ 
with ∈ B and μ( ) < ∞, we have

|f |p dμ < ∞.

This is not a normed space and hence not a Banach space. But it can be given
a countable family of semi-norms with respect to which it becomes a Frechèt
space [59]. Using the semi-norms one can introduce a metric with respect to which
it becomes a complete metric space. Here we are not interested in this space. The
class of functions for which  |f (ω)|p dμ < ∞ is endowed with a norm (topology)
given by
 1/p
 f ≡ |f |p dμ


and denoted by Lp (, B, μ).

The vector spaces p , C(I, R n ), C(, R n ), Lp (, B, μ), 1 ≤ p ≤ ∞, with
norms as defined above, are infinite dimensional Banach spaces. An important
Banach space, denoted by B∞ (I, R n ), is the vector space of bounded measurable
functions defined on the interval I ⊂ R and taking values from R n . The norm for
this space is given by

 f ≡ sup{ f (t) R n , t ∈ I }.

As we shall see later in Chaps. 2, 3, 4, 5, and 6, the Banach spaces

C(I, R n ), B∞ (I, R n ), and Lp (, B, μ), 1 ≤ p ≤ ∞, are very useful in the
study of dynamic systems. More examples will be seen later.

1.6 Modes of Convergence and Lebesgue Integral

There are many different notions of convergence in measure theory and functional
analysis. Here we present only those that have been frequently used in this book.
For detailed study the reader is referred to any of the standard books on real analysis
and measure theory, such as Halmos [69], Munroe [85], Hewitt and Stromberg [71],
Berberian [39], Royden [92].
1.6 Modes of Convergence and Lebesgue Integral 7

1.6.1 Modes of Convergence

Let (, B, μ) be a measure space and M(, B, μ) ≡ M denote the linear space
of μ measurable real-valued functions defined on . Sometimes this is denoted by
L0 (, B, μ) = L0 ().
Uniform Convergence
Definition 1.6.1 A sequence {fn } ∈ M is said to converge to f ∈ M uniformly if
for every ε > 0, there exists an integer nε such that whenever n > nε ,

|fn (ω) − f (ω)| < ε ∀ ω ∈ .

Recall that for any set E ⊂  we have used E  to denote its complement given
by E  ≡  \ E.
Almost Uniform Convergence
Definition 1.6.2 A sequence {fn } ∈ M is said to converge to f ∈ M almost
uniformly, denoted by a.u, if for every ε > 0, there exists a set Gε ∈ B with
μ(Gε ) < ε, such that

fn (ω) −→ f (ω) uniformly on Gε .

Clearly, uniform convergence implies almost uniform convergence.

Almost Everywhere Convergence
Definition 1.6.3 A sequence {fn } ∈ M is said to converge to f μ-almost
everywhere, indicated by μ − a.e, or just a.e, if

μ{ω ∈  : lim fn (ω) = f (ω)} = 0.

n→∞

In other words, the set on which fn fails to converge to f is a set of μ-measure zero.
If μ is a probability measure, i.e., μ() = 1, this mode of convergence is known as
almost sure convergence.
The space of measurable functions M is closed with respect to convergence
almost everywhere. In other words, the a.e. limit of a sequence of measurable
functions is also a measurable function. This is presented in the following result.
Theorem 1.6.4 Let (, B, μ) be a measure space and M ≡ M(, B, μ) denote
the class of measurable functions. Then, if {fn } ∈ M and fn −→ f μ − a.e, we
have f ∈ M.
Proof See Munroe [85, Corollary 20.3.2]. 

In other words, the almost everywhere limit of a sequence of measurable functions
is a measurable function.
8 1 Mathematical Preliminaries

Convergence in Measure
Definition 1.6.5 A sequence {fn } ∈ M is said to converge to f ∈ M in measure,
indicated by fn −→ f meas., if for every ε > 0,

lim μ{ω ∈  : |fn (ω) − f (ω)| > ε} = 0.

n→∞

Again, if μ is a probability measure, this mode of convergence is known as

convergence in probability.
In probability theory, there is another more general mode of convergence known
as convergence in distribution. A sequence of real random variables {Xn } is said
to converge to a random variable X in distribution if the cumulative distribution
functions Fn (x) ≡ P rob.{Xn ≤ x} of {Xn } converge to the cumulative distribution
function F (x) = P rob.{X ≤ x} of the random variable X at all continuity points x
of F.
Another frequently used mode of convergence is known as convergence in the
mean of order p ≥ 1. We will discuss this after we have introduced the Lebesgue
integral.

1.6.2 Lebesgue Integral

It is known that Riemann and Riemann–Stieltjes integrals are defined for continuous
functions. This is a serious restriction. This was overcome by the discovery of
Lebesgue integral which revolutionized mathematics in the twentieth century. Here
we shall briefly introduce this notion for readers not familiar with the subject. Let
(, B, μ) denote a measure space with μ being the Lebesgue measure. Let M ≡
M(, B, μ) denote the space of extended real-valued B measurable functions
defined on . An element f ∈ M is said to be a simple function if there exists
an integer n and a set of real numbers {a1 , a2 , · · · , an } and a family of disjoint B
measurable sets {Ei , i = 1, 2, · · · , n} such that Range(f ) = {a1 , a2 , · · · , an } and

f = ai CEi

where CEi is the indicator function of the set Ei with μ(Ei ) < ∞ for each i =
1, 2 · · · , n. This class of functions is denoted by S. For f ∈ S, its integral with
respect to the measure μ is given by
n
L(f ) ≡ f dμ = ai μ(Ei ).
 i=1

It is clear that for f ∈ S and α ∈ R, we have

L(αf ) = αL(f )
1.6 Modes of Convergence and Lebesgue Integral 9

and for f1 , f2 ∈ S we have

L(f1 + f2 ) = L(f1 ) + L(f2 ).

Clearly, Lebesgue integration is a linear operation on S. For f ∈ M, let {fn } ∈ S

be a sequence such that

lim fn = f μ a.e.
n→∞

That is, f is the almost everywhere limit of a sequence of simple functions {fn }.
We have seen that almost everywhere limit of a sequence of measurable functions
is a measurable function. Thus, f is measurable. Then, the Lebesgue integral of f
is defined by

L(f ) ≡ lim L(fn )

n→∞

possibly taking values from the set [−∞, +∞) or the set (−∞, +∞]. Thus,
Lebesgue integration denoted by the operation L is linear on M. For more on
Lebesgue integration see the celebrated books of Halmos [69], Munroe [85], Hewitt
and Stromberg [71], Berberian [39], Royden [92].
Set functions defined on any sigma algebra can take positive as well as negative
values. In this case they are called signed measures. Let B denote the sigma algebra
of Borel subsets of any abstract set . A set function

ν : B −→ R

is called a signed measure. A simple example of a signed measure is

ν(D) ≡ f (s)μ(ds), D ∈ B,
D

where μ is the Lebesgue measure and f is any Lebesgue integrable function.

Similarly, a vector valued set function

ν : B −→ R n

is called a (finite dimensional) vector measure with values in R n , where R n is given

any suitable norm  ·  . In passing we note that signed measures can also take
values from the extended real number system. They can also be complex valued.
We do not use complex valued measures in this book.
For any E ∈ B we can define the variation of ν on E by

|ν|(E) = sup  ν(σ ) ,

 
10 1 Mathematical Preliminaries

where the summation is taken over  which consists of a finite number of pairwise
disjoint B measurable sets covering the set E, and the supremum is taken over all
such finite partitions. The total variation norm of ν is given by  ν ≡ |ν|().
An interesting example of a positive measure on  is

μ(dx) = λ(dx) + γi δxi (dx) = dx + γi δxi (dx), γi ≥ 0,

where δxi (dx) denotes the Dirac measure concentrated at the point xi ∈ . For any
bounded set ⊂ ,

μ(dx) = λ( ) + γi .
{i:xi ∈ }

If the above integral is finite, the measure μ has finite variation on the set . It is
clear that for the measure μ to have bounded total
variation, it is necessary that the
Dirac measures have countable support and that γi < ∞.
Next we consider Lebesgue–Stieltjes integral denoted by,

Ls (f ) ≡ f (x)dα(x),


where α is a real-valued function on  having bounded total variation. For any such
α, we can define a signed measure as follows:

γα (G) ≡ dα(x), G ∈ B.
G

This measure can be decomposed into two positive measures μ+ −

α and μα giving

γ α = μ+ −
α − μα ,  γ α  ≡ μ+ −
α + μα ,

where  γα  denotes the total variation norm. This is called the Jordan decompo-
sition of γα . Since α is of bounded variation, these are finite (positive) measures
of bounded total variation. Now we can define Lebesgue integral of f ∈ M with
respect to these measures giving

L+ (f ) = f (x)μ+ −
α (dx), L (f ) = f (x)μ−
α (dx).
 

Given that one of them is finite, the Lebesgue–Stieltjes integral is given by

Ls (f ) = L+ (f ) − L− (f ) ≡ f (x)γα (dx)


which may take values in the extended real number system.

1.6 Modes of Convergence and Lebesgue Integral 11

Convergence in the Mean-p

Definition 1.6.6 A sequence {fn } ∈ M is said to converge to f ∈ M in the mean
of order p, for any p ∈ [1, ∞), if
 1/p
lim  fn − f p ≡ lim |fn (ω) − f (ω)| μ(dω)
p
= 0.
n→∞ n→∞ 

Clearly, convergence in the mean-p applies only to the subspace Lp (, B, μ) ⊂

M, where
 
Lp (, B, μ) ≡ f ∈ M : |f (x)|p μ(dx) < ∞ .


The class of vector spaces {Lp (, B, μ), 1 ≤ p ≤ ∞} is known as the Lebesgue
spaces. These constitute a very large class of Banach spaces as seen later in this
chapter. For p = ∞, we have the class of μ-essentially bounded measurable
functions. By this one means that there exists a finite positive number β which is
exceeded by the absolute value of the function f only on a set of μ-measure zero.
That is,

μ{x ∈  : |f (x)| > β} = 0.

The norm of the function f is defined by

 f ∞ ≡ ess − sup{|f (x)|, x ∈ },

and it is given by the smallest number β for which

μ{x ∈  : |f (x)| > β} = 0.

There are many subtle relationships between all these modes of convergence. For
details the reader may consult any book on measure and integration. For example,
the most popular books are those of Halmos [69], Munroe [85].
Here we present the relationship among various modes of convergence through
the following diagram. For detailed proof see Munroe [85] or any of the references
mentioned above. In Figs. 1.1 and 1.2, solid lines with arrow indicate that the
statement at the tail of the arrow implies the statement at the leading edge of the
arrow, while dashed lines with arrow mean that the statement at the tail implies the
existence of a convergent subsequence in the sense of the statement in the front of
the arrow.
With reference to Fig. 1.1, for measure spaces (, B, μ) with μ() = ∞,
convergence in the mean of order p, 1 ≤ p < ∞, implies convergence in measure
(M) as indicated by the solid arrow from Lp to M. Convergence in measure implies
that there exists a subsequence that converges almost everywhere (AE), as indicated
by the dashed arrow running from M to AE. The mode of almost everywhere (AE)
12 1 Mathematical Preliminaries

Fig. 1.1 General measure spaces

Fig. 1.2 Finite measure spaces

convergence does not imply any other modes of convergence. Convergence in mean
of order p, 1 ≤ p < ∞, implies the existence of an almost everywhere (AE) and
almost uniform (AU) convergent subsequence.
In the case of finite measure spaces, i.e., μ() < ∞, the modes of convergence
AE and AU are equivalent. The modes of convergence {Lp , AE, AU} all imply
convergence in measure M. Convergence in measure implies the existence of a
subsequence that converges almost everywhere.

1.7 Selected Results From Measure Theory

In this section, we present some selected results from measure theory which have
been frequently used in this book. These are monotone convergence theorem,
Fatou’s Lemma and the celebrated Lebesgue dominated convergence theorem, and
Fubini’s theorem. First we state the monotone convergence theorem.
Theorem 1.7.1 (Monotone Convergence Theorem (MCT)) Let {fn } ∈
L1 (, B, μ) be a non-decreasing sequence of nonnegative functions and let f0
be a function such that

lim fn = f0 a.e.
n→∞

Then, f0 ∈ L1 (, B, μ), if and only if, limn→∞  fn dμ < ∞, and if this is the
case, then

lim fn dμ = lim fn dμ = f0 dμ.

n→∞   n→∞ 
1.7 Selected Results From Measure Theory 13

Proof Since {fn } is monotone non-decreasing

 sequence, fn (x) ≤ f0 (x) μ a.e.
Thus, if f0 ∈ L1 (, B, μ), then lim  fn dμ < ∞. So,this is necessary  for
integrability of f0 . Clearly, by monotonicity of {fn } we have  fn dμ ≤  f0 dμ
for all n ∈ N and hence

lim fn dμ ≤ f0 dμ. (D1)

n→∞  

Let S denote the class of simple functions defined on , that is, functions
assuming only a finite number of values and let S + ⊂ S denote the class of
nonnegative simple functions. For each n ∈ N, let {Sn,k } ⊂ S + , k ∈ N, be a
non-decreasing sequence of integrable simple functions such that for each n ∈ N

lim Sn,k = fn μ a.e.

k→∞

Define the sequence of functions {Rn,k } by

Rn,k ≡ sup Si,k .

1≤i≤n

Clearly, this is also a sequence of nonnegative integrable simple functions and it is

monotone non-decreasing in both the indices {n, k}. Since {fn } is a non-decreasing
sequence, we have, for each n ∈ N and each k ≥ n,

Sn,k ≤ Rn,k ≤ Rk,k ≤ sup fi = fk . (D2)

1≤i≤k

Letting k → ∞ in the above inequality we obtain,

fn ≤ lim Rk,k ≤ f0 μ a.e.

k→∞

Now letting n → ∞, it follows from the above expression that

lim Rk,k = f0 μ a.e. (D3)

k→∞

For simple functions, it follows from the definition of Lebesgue integral that

lim Rk,k dμ = lim Rk,k dμ = f0 dμ. (D4)

k→∞   k→∞ 

From (D2) it is clear that for all k ∈ N,

Rk,k dμ ≤ fk dμ. (D5)

 
14 1 Mathematical Preliminaries

and hence, letting k → ∞, it follows from (D4) and (D5) that

f0 dμ ≤ lim fk dμ. (D6)

 k→∞ 

This shows that for integrability of f0 it is sufficient that limk→∞  fk dμ < ∞.
Combining (D1) and (D6) we arrive at the following identity:

lim fk dμ = f0 dμ.
k→∞  

This completes the proof. 


This theorem is due to Lebesgue. Later, this result was extended by B.Levi lifting
the non-negativity hypothesis. We present this in the following theorem without
proof. First let us define

f − ≡ −{f ∧ 0} ≡ − inf{f, 0}.

Theorem 1.7.2 (Beppo Levi’s Theorem (GMCT)) Let {fn } be a monotone

non-decreasing
 sequence of extended real-valued measurable functions such that
−
f
 k dμ < ∞ for some k ∈ N. Then

lim fn dμ = lim fn dμ.

n→∞   n→∞

Fatou’s lemma and Lebesgue dominated convergence theorem are frequently

used in this book. On the basis of the preceding results, we can give simple proof of
these as follows.
Theorem 1.7.3 (Fatou’s Lemma) Let {fn } be a sequence of measurable nonneg-
ative μ-integrable functions on the measure space (, B, μ) satisfying lim inf fn =
f0 μ a.e. Then

lim inf fn dμ ≤ lim inf fn dμ.

 n→∞ n→∞ 

Note that the expression on the right-hand side may assume the value +∞.
Proof The proof is based on the monotone convergence Theorem 1.7.1. Define the
sequence gn ≡ infk≥n fk . Since fn is nonnegative and integrable, this is a non-
decreasing sequence of nonnegative integrable functions satisfying 0 ≤ gn ≤ fn .
Clearly,

gn dμ ≤ fn dμ
 
1.7 Selected Results From Measure Theory 15

and hence

lim inf gn dμ ≤ lim inf fn dμ.

n→∞  n→∞ 

By the monotone convergence theorem and the fact that lim gn = lim inf fn , we
have

lim gn dμ = lim gn dμ ≡ lim inf fn dμ.

n→∞    n→∞

It follows from these two expressions that

lim inf fn dμ = lim gn dμ = lim inf gn dμ ≤ lim inf fn dμ.

 n→∞ n→∞  n→∞  n→∞ 

Thus,

f0 dμ ≤ lim inf fn dμ.

 n→∞ 

This completes the proof. 



Remark 1.7.4 It is clear from the above result that if lim infn→∞  fn dμ < ∞
then f0 ∈ L1 (, B, μ).
Remark 1.7.5 Note that in the statement of the Fatou’s Lemma, the condition that
f0 = lim inf fn can be replaced by f0 ≤ lim inf fn .
Remark 1.7.6 Fatou’s Lemma also applies to any sequence of integrable functions
which are bounded from below by an integrable function. For example, consider
the sequence of integrable function fn and suppose fn ≥ g μ a.e where g ∈
L1 (, B, μ). Then one can verify that

lim inf fn dμ ≤ lim inf fn dμ.

 

Now we can present the Lebesgue dominated convergence theorem which we

denote by LDCT.
Theorem 1.7.7 (LDCT-1) Let {fn } ∈ L1 (, B, μ) and suppose

(i) : lim fn (ω) = f (ω) μ − a.e

n→∞

(ii) : ∃ a g ∈ L+
1 (, B, μ) such that |fn (ω)| ≤ g(ω) μ − a.e.
16 1 Mathematical Preliminaries

Then f ∈ L1 (, B, μ) and

lim fn dμ = lim fn dμ = f dμ.

n→∞   n→∞ 

Proof We use Fatou’s lemma to prove the theorem. By assumption, the sequence
fn is dominated by the integrable function g and hence the two sequences fn + g
and g − fn are nonnegative integrable functions. Applying Fatou’s Lemma to the
first sequence, we obtain

lim inf fn dμ ≤ lim inf fn dμ.

 

Applying it to the second sequence, we obtain

lim inf(g − fn )dμ ≤ lim inf (g − fn )dμ.

 

This means

lim sup fn dμ ≤ lim sup fn dμ.

 

Since fn → f μ-a.e, it follows from the first inequality that

f dμ = lim inf fn dμ ≤ lim inf fn dμ.

  

Similarly it follows from the second inequality that

lim sup fn dμ ≤ lim sup fn dμ = f dμ.

  

Thus, it follows from the last two inequalities that

lim fn dμ = lim fn dμ = f dμ.

n→∞   n→∞ 

This completes the proof. 


Remark 1.7.8 (LBCT) It is clear from the above theorem that if the condition (i)
holds and condition (ii) is replaced by the assumption that the sequence is absolutely
bounded by a finite positive number, the conclusion of the theorem holds trivially.
This is known as the Lebesgue bounded convergence theorem.
1.7 Selected Results From Measure Theory 17

Two Examples
(E1) Consider  = R and define the sequence of real-valued (positive) functions
given by
√ 
fn (x) ≡ n/ 2π exp{−(1/2)n2 x 2 }, x ∈ R, n ∈ N.

This is a Gaussian 2
 density function with zero mean and variance (1/n ). It is
easy to see that  R fn (x)dx = 1 for all n ∈ N, and that fn −→ 0 a.e, but
clearly limn→∞ R fn (x)dx = 0. What is lacking here is the condition (ii) of
the LDCT-1. There is no integrable function that dominates the sequence {fn }.
Another example is the sequence {fn } given by
√
fn (x) ≡ (1/ 2π) exp{−(1/2)(x − n)2 }

 is one for all n ∈ N and it converges to

which has similar behavior. Its integral
zero for each x ∈ R, but the limn→∞ fn (x) = 1 = 0. The reason is same, there
is no integrable function that can dominate the sequence.
(E2) Consider the interval J ≡ [0, π] and the sequence of functions fn (t) ≡
sin(nt), t ∈ J. Note that

lim fn (t)dt = 0.
n→∞ J

Here |fn | ≤ 1 and so dominated by an integrable function but limn→∞ fn (t) =

0 a.e. Here condition (i) is missing. However, note that the Gaussian sequence
given by the first part of example (E1) converges, in generalized sense, to
the Dirac measure with the point {0} being the support. The second sequence,
however, converges to zero in the sense of distribution.
Lebesgue dominated convergence theorem also holds for Lp (, B, μ) spaces as
stated in the following theorem.
Theorem 1.7.9 (LDCT-2) Let {fn } ∈ Lp (, B, μ) for some p ∈ [1, ∞) and
suppose

(i) : lim fn (ω) = f (ω) μ − a.e.

n→∞

(ii) : ∃ a g ∈ L+
p (, B, μ) such that |fn (ω)| ≤ g(ω) μ − a.e.

Then f ∈ Lp (, B, μ) and

lim |fn − f |p dμ = 0.
n→∞ 
18 1 Mathematical Preliminaries

Sometimes we have to deal with multiple integrals. Fubini’s theorem proves that
under certain conditions they can be evaluated as iterated integrals which is much
simpler. For example, let us consider double integrals.
First, one integrates with respect to one variable keeping the other variable fixed
and then complete the integration by integrating with respect to the remaining
variable. This process is acceptable only if the end result is the same irrespective
of the order of integration. This is the basic content of Fubini’s theorem. Let
(i , i , μi ) i = 1, 2 be two measure spaces. Let  = 1 × 2 denote the product
σ -field on the product space  ≡ 1 × 2 and μ ≡ μ1 × μ2 denote the product
measure on the σ -field .
Theorem 1.7.10 (Fubini’s Theorem) Let f be a measurable function with respect
to the product σ -field and suppose the following integral exists:

f dμ.
1 ×2

If the functions f1 and f2 given by

f1 (x) ≡ f (x, y)μ2 (dy), f2 (y) ≡ f (x, y)μ1 (dx),

2 1

exist μ1 -a.e and μ2 -a.e, respectively, then f1 and f2 are integrable with respect to
the measures μ1 and μ2 respectively and the following identity holds:

f dμ = f1 (x)μ1 (dx) = f2 (y)μ2 (dy).

 1 2

Proof See Halmos [69], Berberian [39], Munroe [85]. 


Let (, F , P ) be a complete probability space furnished with an increasing
family of subsigma algebras Ft , t ≥ 0, of the sigma algebra F with P being
the probability measure. We call (, F , Ft ≥0, P ) a filtered probability space or
a probability space furnished with a filtration. In probability theory, Borel–Cantelli
Lemma is one of the best known results on zero-one law. The lemma states that
under certain conditions an event will occur either with probability one or zero.
It has been used in the study of continuity properties of solutions of stochastic
differential equations to be seen in Chap. 6. We present this result in the following
lemma.
Lemma 1.7.11 (Borel–Cantelli Lemma) Let {Ai } be a sequence of measurable
sets and let A denote the set of points in infinitely
∞ ∪∞ A . Then, either P (A) = 0 if
 many of the sets {Ai }, that is
A
 = ∩n=1 i=n i P (Ai ) < ∞ or P (A) = 1 if
i≥1 P (A i ) = ∞ and the sets {A i } are mutually independent.
Proof See Doob [58, Theorem 1.2, p. 104]. 

1.8 Special Hilbert and Banach Spaces 19

Doob’s martingale inequality is one of the most celebrated results in the study of
stochastic processes. A random process {ξ(t), t ≥ 0} adapted to filtration Ft , t ≥ 0,
is said to be an Ft -martingale if (i): E|ξ(t)| < ∞, t ≥ 0, and (ii) E{ξ(t)|Fs } = ξ(s)
for all t ≥ s ≥ 0. The process is said to be a submartingale if E{ξ(t)|Fs } ≥ ξ(s) for
all t ≥ s ≥ 0; and super martingale if E{ξ(t)|Fs } ≤ ξ(s) ∀ t ≥ s ≥ 0.
Theorem 1.7.12 If ξ(t), t ∈ I ≡ [0, T ] is an Ft -martingale on the filtered
probability space (, F ⊃ Ft ↑ t, P ) then
(i) P {supt ∈I |ξ(t)| > r} ≤ (1/r)E|ξ(T )|, ∀ r > 0,
(ii) if for some p > 1, E|ξ(t)|p < ∞, ∀ t ∈ I, then
 
E sup |ξ(t)|p ≤ (p/p − 1)p E|ξ(T )|p .
t ∈I

Note that for square integrable martingales, the last inequality reduces to
 
E sup |ξ(t)|2 ≤ 4E|ξ(T )|2 .
t ∈I

This particular inequality is crucially important in the study of stochastic differential

equations driven by martingales as seen in Chap. 6.

1.8 Special Hilbert and Banach Spaces

Many of the problems arising in physics and engineering are concerned with energy.
Hilbert spaces are more suitable for quantifying such entities.

1.8.1 Hilbert Spaces

Definition 1.8.1 (Hilbert Space) A Hilbert space H is a Banach space with

spacial structure. It is furnished with an inner (or scalar) product (·, ·) as defined
below.
For any two elements f, g ∈ H, the scalar product (f, g) is a real or a complex
number. Let C denote the field of complex numbers. The map

{f, g} −→ (f, g)

has the following properties:

(H1) (αf, g) = α(f, g), (f, αg) = α ∗ (f, g), α ∈ C, f, g ∈ H where α ∗ is the
complex conjugate of α,
(H2) (f, g1 + g2 ) = (f, g1 ) + (f, g2 ), ∀ f, g1 , g2 ∈ H,
20 1 Mathematical Preliminaries

(H3) (f, g) = (g, f )∗ , ∀ f, g ∈ H,

(H4)  f 2 = (f, f ), ∀ f ∈ H.
Some examples of finite dimensional Hilbert spaces are R n , E n , n2 , with the
norms as defined in the previous section. The normed spaces 2 , L2 (, μ), are
infinite dimensional Hilbert spaces. For estimates of intensity of forces, energy, or
any other physically important variables, often we use inequalities. A very important
inequality, called the Schwartz inequality, holds for Hilbert spaces.
Proposition 1.8.2 (Schwarz Inequality) For every f, g ∈ H,

|Re(f, g)| ≤ f  g  . (1.8)

Proof See [14, Proposition 1.4.4, p. 7]. 



As a corollary of this result, we have the triangle inequality.

Corollary 1.8.3 (Triangle Inequality) For every f, g ∈ H

 f + g ≤ f  +  g  . (1.9)

Proof See [14, Corollary 1.4.5, p. 8]. 


Another example of a Hilbert space is given by H 1 (, μ) which consists of
functions (equivalence classes) which along with their first derivatives are square
integrable.

H 1 (, μ) ≡ {ϕ : ϕ ∈ L2 (, μ), Dϕ = (Dxi ϕ, i = 1, 2, · · · , n) ∈ L2 (, μ)}.

The scalar product in this space is given by

(ϕ, ψ) = ϕ(x)ψ(x)μ(dx) + (Dxi ϕ)(Dxi ψ)μ(dx),

 1≤i≤n 

with the corresponding norm,

 1/2
 ϕ H 1 = |ϕ(x)|2μ(dx) + |Dxi ϕ|2 μ(dx) .
 1≤i≤n 

If  = I ≡ (0, T ) and μ is the Lebesgue measure (length), H 1 (I ) is the space of

square integrable functions which are absolutely continuous with the first derivatives
being square integrable. The norm is given by
 T T 1/2
 ϕ H 1 = |ϕ|2 dt + |ϕ̇|2 dt .
0 0
1.8 Special Hilbert and Banach Spaces 21

1.8.2 Special Banach Spaces

Now we consider a very important class of function (signal) spaces which are
Banach spaces, not Hilbert spaces. For f ∈ Lp (, B, μ), 1 ≤ p < ∞, we denote
the norm of f by
 1/p
 f p ≡ |f |p μ(dx) ,


where μ is a countably additive σ -finite positive measure on . For p = 2, this is a

Hilbert space. The spaces Lp (, B, μ), 1 ≤ p ≤ ∞ are Banach spaces. There is no
scalar product in these spaces in the sense defined for Hilbert spaces. For f, g ∈ Lp ,
(f, g) is not defined. However, there is a duality pairing as described below.
For special Banach spaces, like Lp (, B, μ) with μ any finite positive measure,
we have similar results like those of Proposition 1.8.2 and Corollary 1.8.3. These are
Hölder and Minkowski inequalities. For economy of notations, we may sometimes
omit the sigma algebra and write Lp (, μ) to denote Lp (, B, μ).
Proposition 1.8.4 (Hölder Inequality) For every f ∈ Lp (, μ) and g ∈
Lq (, μ) with 1 < p, q < ∞, (1/p) + (1/q) = 1,

|(f, g)| ≤ f p  g q . (1.10)

Proof We present a simple proof. Define

|f (x)|p |g(x)|q
a≡ p , b ≡ q (1.11)
 f p  g q

and α ≡ (1/p), β ≡ (1/q). Since the function log x, x ≥ 0, is a concave function,

we have

log(αa + βb) ≥ α log a + β log b = log(a α bβ ).

Hence, the inequality (1.10) follows. 


Remark 1.8.5 Note that the above result holds also for p = 1 and q = ∞, giving

|(f, g)| ≤ |f ||g|dμ ≤ g ∞ |f |dμ = f 1  g ∞ ,

 

where  g ∞ = μ − ess − sup{|g(x)|, x ∈ }.

As a corollary of this result, we have the following triangle inequality.
22 1 Mathematical Preliminaries

Corollary 1.8.6 (Minkowski Inequality) For f, g ∈ Lp (, μ), 1 ≤ p < ∞,

 f + g p ≤ f p +  g p . (1.12)
 
Proof Write  |f +g|p dμ =  |f +g|p−1 |f +g|dμ and apply Hölder inequality.
For details see [14, Corollary 1.8.7, p. 11]. 

The following result states that all Lp (Lebesgue) spaces are complete and hence
are Banach spaces.
Theorem 1.8.7 For any positive measure space (, B, μ) with μ() < ∞, and
for each p ≥ 1, the vector space Lp (, B, μ) is a Banach space.
Proof See [14, Theorem 1.4.8, p. 12]. 

Proposition 1.8.8 Consider the Banach spaces {Lp (, μ), 1 ≤ p ≤ ∞} and
suppose μ() < ∞. Then, for 1 ≤ p1 ≤ p2 ≤ ∞, we have

 f p1 ≤ c  f p2 , (1.13)

where c = (μ())(p2−p1 )/(p2 p1 ) . Hence, Lp2 (, B, μ) ⊂ Lp1 (, B, μ).

Note that the embedding constant c is dependent on the parameters as shown. It
follows from this inequality that the embedding

Lp2 (, μ) → Lp1 (, μ)

is continuous.
Remark 1.8.9 It is clear from the expression for the constant c that if μ() = ∞,
the inclusion does not make much of a sense indicating that the inclusion may not
even hold. Here is an example justifying this statement. Let  = [0, ∞) and μ the
Lebesgue measure. Consider the function

1
f (x) ≡ , x ∈ [0, ∞), ∞ > p ≥ 1.
x 1/p (1 + | log x|)2/p

The reader can verify that f ∈

/ Lr ([0, ∞), μ) for any r = p. For r = p we have
∞ ∞ 1
|f (x)|p dx = dx
0 0 x(1 + | log x|)2
∞ 1 ∞ 1
= dy = 2 dy = 2. (1.14)
−∞ (1 + |y|)2 0 (1 + y)2

In summary, in case μ() = ∞, there exists f ∈ Lp but f ∈

/ r=p Lr .
Remark 1.8.10 Let μ(dx) = dx denote the Lebesgue measure and ν(dx) ≡ 1/(1+
x 2 )dx be another measure both defined on R. Note that ν(R) = π, so this is a
1.8 Special Hilbert and Banach Spaces 23

finite positive measure. Then, one can easily verify that Lp (R, μ) ⊂ Lp (R, ν) for
all p ≥ 1. For f ∈ Lp (R, ν), g ∈ Lq (R, ν), with (p, q) being the conjugate
 fg√ ∈ L1 (R, ν). Similar conclusions hold for the Gaussian measure
pair, we have
νg (K) ≡ K (1/ 2π) exp{−(1/2)x 2}dx.
Another interesting property of the {Lp (, μ), 1 ≤ p ≤ ∞} spaces is stated in
the following result. Let

B1 (Lp ) ≡ {h : h ∈ Lp (, μ),  h p ≤ 1}

denote the closed unit ball of the Banach space Lp . The unit sphere is denoted by
∂B1 ≡ S1 , that is, these are the elements of B1 which have norms exactly equal to
one. This is also the boundary of the unit ball.
Proposition 1.8.11 For every f ∈ Lp (, μ) (1 ≤ p < ∞), there exists a g ∈
B1 (Lq ), where q is the conjugate of p in the sense that (1/p) + (1/q) = 1, such
that

(f, g) = f p .

Further, if 1 < q < ∞, there is only one such (unique) g ∈ B1 (Lq ).

Proof Let f (= 0) ∈ Lp (, μ) be given. Define

p−1
g(x) ≡ (1/  f p )|f (x)|p−1 signf (x), x ∈ . (1.15)

The signum function of a Borel measurable function is also Borel measurable.

Product of measurable functions is a measurable function. Thus, g as defined is
a measurable function. By integration we show that this element belongs to the unit
sphere ∂B1 (Lq ) where q is the number conjugate to the number p. Integrating |g|q
and noting that (p − 1)q = p, we find that

(p−1)q
|g|q dμ = (1/  f p ) |f |(p−1)q dμ = 1.
 

Thus, g ∈ ∂B1 (Lq ). Now computing the duality product of f with the function g
given by (1.15), we find that

p−1
(f, g) = f (x)g(x)dμ = (1/  f p ) |f |p dμ = f p . (1.16)
 

This proves the first part. For the second part, we note that the ball B1 (Lq ) is strictly
convex for all q ∈ (1, ∞) in the sense that the line segment joining any two points
f1 , f2 of the ball B1 given by

{h ∈ B1 : h = αf1 + βf2 , α, β ≥ 0, α + β = 1},

24 1 Mathematical Preliminaries

cannot touch the boundary except possibly at the end points. We prove this by
contradiction. Suppose there are two points g1 , g2 ∈ ∂B1 (Lq ) such that (f, g1 ) =
(f, g2 ) = f p . Define g = (1/2)(g1 + g2 ). Then, clearly (f, g) = f p . Since
the ball is strictly convex, g must be an interior point of the ball and so  g q < 1
and this implies that |(f, g)| < f p . This is a contradiction and so g1 must equal
g2 . This completes the proof. 

It is clear from Proposition 1.8.4 that every g ∈ Lq (, μ) determines a
continuous linear functional
 on Lp (, μ) (for 1/p + 1/q = 1) through the
integration g (f ) =  f (x)g(x)μ(dx) = f, g called the duality pairing. In fact
the converse is also true. That is, every continuous linear functional on Lp (, μ),
for 1 ≤ p < ∞, is determined by an element of Lq (, μ) through the pairing
mentioned above. Let Lp (, μ)∗ denote the class of continuous linear functionals
on Lp (, μ). This is known as the continuous dual of the Banach space Lp (, μ).
It follows from the above discussion that we have proved the well-known Riesz
representation theorem for Lp spaces as stated below.
Proposition 1.8.12 The dual of the Banach space Lp (, μ) denoted by Lp (, μ)∗
is isometrically isomorphic to the linear space Lq (, μ) for all {p, q : 1 ≤ p <
∞, (1/p) + (1/q) = 1}.
Readers interested in the detailed proof may consult any standard book on
functional analysis. From now on, we will denote by E ∗ the continuous dual (or
the conjugate) of any Banach space E. For example, the dual of Lp , 1 ≤ p < ∞,
is Lq , as seen above and, particularly, the dual of L1 is L∞ and the dual of L2 is the
L2 space itself. The dual of L∞ is the space of finitely additive measures.
The dual of the space of continuous functions C(), with  a compact subset of
R n , is the space of countably additive measures of bounded total variation denoted
by Mca (B) where B is the sigma algebra of Borel subsets of the set . In fact this
is also valid for any compact metric space . 
For example,
 the discrete measure μ(dx) ≡ i γi δxi (dx) having bounded
variation, |γi | < ∞, defines a continuous linear functional on C(). Indeed,
the functional μ given by

μ (f ) = f (x)μ(dx)


is well defined. One can easily verify that

 
|μ (f )| ≤ |γi |  f = μ  f < ∞,
i

where  μ  denotes the total variation norm of the measure μ.

In addition to the norm topology, an Lp space can be furnished with the topology
of weak convergence as defined below.
1.9 Metric Space 25

Definition 1.8.13 For any p ∈ [1, ∞), a sequence {fn } ⊂ Lp is said to converge
weakly to an element f ∈ Lp if, and only if,

fn (x)g(x)μ(dx) −→ f (x)g(x)μ(dx)
 

for every g ∈ Lq where Lq (q = (p/p − 1)) is the dual of Lp .

Definition 1.8.14 A Banach space E is said to be reflexive if it coincides with its
second dual, also called the bidual, that is, E = (E ∗ )∗ ≡ E ∗∗ .
For example, all Lp spaces with 1 < p < ∞ have this property. This is clear
from the fact that the dual of Lp is Lq for q = (p/p − 1) and that the dual of Lq is
Lp . Thus, L∗∗
p = Lp . This is a very important class of Banach spaces having some
properties which are similar to those of finite dimensional spaces as discussed in the
following section.
In general the bidual E ∗∗ is a larger space. The embedding E → E ∗∗ is proper
and continuous and it is known as the canonical embedding. In case E = E ∗∗ , that is
E is reflexive, weak, and weak-star topologies on the dual space E ∗ are equivalent.

1.9 Metric Space

Loosely speaking, a set M which is furnished with a measure of distance between

any two elements of the set satisfying some natural properties is called a metric
space. More precisely, a function d : M × M −→ [0, ∞] is called a metric if it
satisfies the following properties:
(M1) d(x, y) ≥ 0, ∀ x, y ∈ M, and d(x, y) = 0 if and only if, x = y, (non-
negativity),
(M2) d(x, y) = d(y, x) ∀ x, y ∈ M, (symmetry),
(M3) d(x, z) ≤ d(x, y) + d(y, z) ∀ x, y, z ∈ M, (triangle inequality).
Definition 1.9.1 A metric space M ≡ (M, d) is said to be complete, if every
Cauchy sequence has a limit.
Example 1 Every Banach space B ≡ (B,  · ) is a complete metric space with
respect to the metric d given by d(x, y) ≡ x − y  . Thus, all Lp (, μ) spaces,
1 ≤ p ≤ ∞, are complete metric spaces.
Example 2 Let B be a Banach space with norm  · , and let be any closed subset
of B. Then ( , d) with d(x, y) ≡ x − y , is a complete metric space. But this is
not a linear metric space, though, in this particular case, it can be embedded in the
linear space B.
Example 3 Let I be a closed bounded interval and D a closed subset of R n . Then
C(I, D), the class of continuous functions defined on I and taking values in D, is
26 1 Mathematical Preliminaries

again a complete metric space with the metric inherited from the norm of the Banach
space C(I, R n ).
An example of special importance in control theory is the class of measurable
functions M, also denoted by L0 , which is endowed with the topology of
convergence in measure. This topology is metrizable with the metric
|f − g|
d(f, g) = dμ.
 1 + |f − g|
This is equivalent to the metric

d(f, g) = min{1, |f − g|} dμ.



With respect to this metric, L0 is a complete metric space, see [71].

Definition 1.9.2 A subset K of a metric space (M, d) is said to be relatively
compact (or precompact) if, for every ε > 0, K can be covered by a finite number
of balls of radius ε. That is, there exists an integer nε and a finite set of points
{xi ∈ K, i = 1, 2, · · · , nε } such that

K⊂ B(xi , ε),
1≤i≤nε

where B(x, ) ≡ {z ∈ M : d(x, z) ≤ ε} is a closed ball of radius ε centered at x.

The set K is said to be compact, if it is also closed.
Definition 1.9.3 A subset K of a metric space (M, d) is said to be sequentially
compact if every sequence from K has a convergent subsequence with the limit
being in K. A set K ⊂ M is said to be relatively compact, if its closure is compact.
Here are some examples of compact sets. A closed bounded set in R n is compact.
A bounded set in R n is relatively compact. In fact any bounded set in a finite
dimensional space is relatively compact. In infinite dimensional Banach spaces,
this is not true. However, for reflexive Banach spaces, furnished with the weak
convergence topology, we have the following result. For proof see Dunford [59].
Theorem 1.9.4 A bounded weakly closed subset K of a reflexive Banach space is
weakly compact, as well as, weakly sequentially compact.
One of the most frequently used Banach spaces is the space of continuous
functions. Let (X, d) and (Y, ρ) be any two metric spaces. A function f : X −→ Y
is said to be continuous if the inverse image of any open set in Y is an open set in
X, that is OX ≡ f −1 (OY ) is an open set in X whenever OY is an open subset of Y.
Denote by C(X, Y ) the space of continuous functions from X to Y . Furnished with
the metric topology

γ (f, g) ≡ sup ρ(f (x), g(x)),

x∈X
1.9 Metric Space 27

C(X, Y ) is a metric space. If (Y, ρ) is a complete metric space, then so also is

(C(X, Y ), γ ). We are especially interested in the following Banach space. Let D be
a compact subset of a metric space and Y a finite dimensional Banach space. Let
C(D, Y ) denote the vector space of continuous functions furnished with the norm
topology

 f ≡ sup{|f (x)|Y , x ∈ D}.

Then it is a Banach space. That is, every Cauchy sequence {fn } in C(D, Y )
converges in the norm topology defined above to an element f ∈ C(D, Y ). In
particular, let D be a closed bounded interval I of the real line R and Y = R n .
Then, C(I, R n ) is a Banach space.
Definition 1.9.5 (Absolute Continuity) A function f ∈ C(I, R n ) is said to be
absolutely continuous if, for every ε > 0, there is δ > 0 such that, for any set of
disjoint open intervals {Ii ≡ (ti , ti+1 )} ⊂ I,

|f (ti+1 ) − f (ti )|R n < ε

whenever

|ti+1 − ti | < δ.

We use AC(I, R n ) to denote the class of absolutely continuous functions from I

to R n . It is evident that AC(I, R n ) ⊂ C(I, R n ). The question of compactness of a
set G ⊂ C(I, R n ) is very crucial in the study of existence of solutions of differential
equations and many optimization problems. This is closely related with the notion
of equicontinuity.
Definition 1.9.6 (Equicontinuity) A set G ⊂ C(I, R n ) is said to be equicontinu-
ous at a point t ∈ I, if for every ε > 0 there exists a δ > 0 such that

|f (t) − f (s)|R n < ε ∀ f ∈ G

whenever |t − s| < δ. The set G is said to be equicontinuous on I if this holds for

all t ∈ I.
The following result is very important in the study of regularity properties of
solutions of ordinary differential equations. It has also applications in the study of
optimization.
Theorem 1.9.7 (Ascoli–Arzela) A set G ⊂ C(I, R n ) is relatively compact if (i) it
is bounded in the norm topology and (ii) it is equicontinuous. The set G is compact
if it is relatively compact and the t section, R(t) ≡ {f (t), f ∈ G} is closed for
every t ∈ I.
28 1 Mathematical Preliminaries

Proof Take any sequence {xn } ⊂ G. We show that it has a convergent subsequence.
Since I is a compact interval, for every δ > 0, there exists a finite set IF ≡ {ti , i =
1, 2, · · · , m = m(δ)} ⊂ I such that for every t ∈ I , there exists a tj ∈ IF such that
|t − tj | ≤ δ. Since the set G is equicontinuous, the sequence {xn } is equicontinuous.
Thus, for any ε > 0, there exists a δ > 0 such that

sup  xn (t) − xn (s) ≤ ε whenever |t − s| ≤ δ.

n∈N

Consider the sequence {xn (tj ), j = 1, 2, · · · , m}. Since G is a bounded set it is clear
that each of the elements of the above sequence is contained in a bounded subset of
R n . Bolzano-Weierstrass theorem says that every bounded sequence of vectors in
a finite dimensional space has a convergent subsequence. Thus, for every sequence
{xn (tj ), n ∈ N} there exists a convergent subsequence {xn(j ) (tj )}. Let {n(1)} be a
subsequence of the sequence {n} for which xn(1) (t1 ) is convergent. Denote by n(2)
a subsequence of n(1) for which xn(2) (t2 ) is convergent. Carrying out this diagonal
process, we obtain a subsequence {n(m)} ⊂ {n} such that along this subsequence
every element of the set {xn (tj ), j = 1, 2, · · · , m} is convergent. For any t ∈ I,
there exists a j ∈ {1, 2, · · · , m} such that |t − tj | ≤ δ. Clearly, the required tj
depends on t ∈ I. We indicate this dependence by tj (t). Thus, we have

 xk (t) − xr (t) ≤ xk (t) − xk (tj (t))  +  xk (tj (t)) − xr (tj (t)) 
+  xr (tj (t)) − xr (t) 
≤ 2ε+  xk (tj (t)) − xr (tj (t)) 

for any k, r ∈ {n(m)}. Hence,

sup  xk (t) − xr (t) ≤ 2ε + sup  xk (tj (t)) − xr (tj (t))  .

t ∈I t ∈I

Letting k, r −→ ∞ along the subsequence {n(m)}, we obtain

lim  xk − xr C(I,R n ) ≤ 2ε.

k,r→∞

Since ε(> 0) is arbitrary, it follows from this that from every sequence of G one
can extract a subsequence which is a Cauchy sequence in C(I, R n ). Since this is
a Banach space with respect to the sup norm topology, there exists an element
x ∈ C(I, R n ) to which the sequence {xk } converges. This proves that the set G
is relatively compact. Clearly, if every t section of G is closed, the limit x ∈ G.
Hence, G is compact. This completes the proof. 

Remark 1.9.8 In fact, the Ascoli–Arzela theorem, as presented above, also holds
for compact metric spaces (M, d) replacing compact interval I. For more general
result, see [92, p. 153], [108, p. 85].
1.9 Metric Space 29

The space of continuous linear functionals on C(I, R n ) has the representation

(f ) = f (t), μ(dt), f ∈ C(I, R n ),

I

where μ is an R n -valued measure defined on the Borel subsets of the set I. For
example, note that for any z ∈ R n and the Dirac measure δs (dt), with mass
concentrated at the point s, the element νz,s (dt) ≡ zδs (dt) defines a continuous
linear functional on C(I, R n ). Indeed,

z,s (f ) ≡ f (t), νz,s (dt) = (f (s), z).

I

It is obvious that

|z,s (f )| ≤ |z|R n  f  .

Hence, the functional z,s is a bounded linear (and hence continuous) functional
on C(I, R n ). Similarly,
 for any sequence J ≡ {ti } ⊂ I and Z ≡ {zi } ⊂ R n , the
measure νZ,J ≡ zi δti (dt) defines a vector measure. The linear functional

Z,J (f ) ≡ f (t), νZ,J (dt) = (f (ti ), zi ),

I ti ∈J

defines a bounded linear functional on C(I, R n ) if and only if

|zi | < ∞.
ti ∈J

Two interesting facts emerge from this example. One, if J is a continuum and
|zτ |R n > 0 for all τ ∈ J , Z,J defines an unbounded linear functional and hence is
not an n
 element of the (topological) dual of C(I, R ). Similarly, if J is a countable set
but |zi |R n = ∞, the functional Z,J fails to define a continuous linear functional.
From these elementary observations, we conclude that the (topological) dual of
the Banach space C(I, R n ) is the space of countably additive R n -valued measures
having bounded total variation on I given by

 μ ≡ sup |μ(σ )|R n ,

 σ ∈

where  is any finite partition of the set I by disjoint members from the class
of Borel sets BI of I. The supremum is taken over all such finite partitions. We
may denote this dual space by M(I, R n ) or Mca (I, R n ) to emphasize countable
additivity. To emphasize that these are set functions, it is more appropriate to use the
notation Mca (I , R n ), and to emphasize that they have bounded variation one may
30 1 Mathematical Preliminaries

use the symbol Mcabv (I , R n ). We may use both provided there is no possibility
of confusion. Furnished with the norm topology (total variation norm), as defined
above, Mca (I , R n ) is a Banach space.
Note that every g ∈ L1 (I, R n ) also induces a countably additive bounded vector
measure through the mapping, g −→ μg given by

μg (σ ) ≡ g(t)dt, σ ∈ I .
σ

Indeed, define

g (f ) ≡ f (t), μ(dt) = f (t), g(t)dt.

I I

Clearly, it follows from this expression that for any f ∈ C(I, R n ),

|g (f )| ≤ f  |g(t)|R n dt < ∞.

I

This shows that the map e : g −→ g from L1 (I, R n ) to C(I, R n )∗ ≡

Mca (I , R n ) is a continuous embedding. That is e(L1 (I, R n )) ⊂ Mca (I , R n ).
Thus, the space of vector measures is much larger than the class of Lebesgue
integrable vector valued functions. In the study of optimal controls involving vector
measures (as controls), sometimes we need to consider the question of compactness
of subsets ⊂ Mca (I , R n ). We consider this later in the sequel.

1.10 Banach Fixed Point Theorems

Many problems of mathematical sciences lead to the questions of existence,

uniqueness, and regularity properties of solutions of equations of the form x = F (x)
in appropriate metric spaces depending on the particular application. The questions
of existence and uniqueness of solutions of such problems are called fixed point
problems. Let X ≡ (X, d) be a complete metric space and suppose F : X → X.
Definition 1.10.1 A point x ∗ ∈ X is said to be a fixed point of the map F if x ∗ =
F (x ∗ ).
Let

F ix(F ) ≡ {x ∈ X : x = F (x)},

denote the set of fixed points of the map F. If F ix(F ) = ∅, the equation, x = F (x),
has no solution. If the set F ix(F ) consists of only a single point in X, then the
1.10 Banach Fixed Point Theorems 31

equation, x = F (x), has a unique solution. Otherwise the equation has multiple
solutions.
Fixed point theorems are important tools for proof of existence of solutions of
linear and nonlinear problems. We present here only those used in this book.
For motivation, we consider the following example. An input–output model often
used in engineering is given by a linear system governed by an integral operator like
t
y(t) ≡ K(t, s)u(s)ds, t ≥ 0,
0

where the input u may be provided by an expression like

u(t) = g(t, y(t)) + r(t), t ≥ 0.

That is, the input has two components one being the output feedback through a
nonlinear device, and the other is the direct input command. In this case, we obtain
an integral equation of the form
t
y(t) = v(t) + K(t, s)g(s, y(s))ds,
0

where
t
v(t) = K(t, s)r(s)ds.
0

This is a nonlinear Volterra integral equation, and it is characterized by the kernel K

and the nonlinear map g. One can formulate this as an abstract fixed point problem

y = v + G(y) ≡ Fv (y)

in appropriately chosen function spaces. This is precisely in the form we have

presented earlier. We note that choice of the function space is predominantly
determined by the nonlinear operator g and the kernel K as seen in Ahmed [1].
More general nonlinear integral equations based on infinite Volterra series can be
found in Ahmed [17, 21].
Theorem 1.10.2 (Banach Fixed Point Theorem) Let X ≡ (X, d) be a complete
metric space, and suppose F : X −→ X is a contraction in the sense that there
exists a number α ∈ (0, 1) such that

d(F (x), F (y)) ≤ α d(x, y) ∀ x, y ∈ X.

Then F has a unique fixed point in X.

32 1 Mathematical Preliminaries

Proof We start with an arbitrary element x0 ∈ X and define the sequence

xk+1 ≡ F (xk ), k ∈ N0 ≡ {0, 1, 2, · · · }. (1.17)

For any 1 ≤ p ∈ N0 , by repeated application of the triangle inequality, we find that

n+p−1  n+p−1 
d(xn+p , xn ) ≤ d(xk+1 , xk ) ≤ α k d(x1, x0 ). (1.18)
k=n k=n

Since 1 > α > 0, it is clear that

n+p−1  ∞ 

αk ≤ αn αk ≤ α n /(1 − α) .
k=n k=0

From (1.17), (1.18) and the above inequality it is clear that, for any fixed p(1 ≤ p <
∞) ∈ N0 , we have

d(xn+p , xn ) ≤ α n /(1 − α) d(x1 , x0 )

and consequently

lim d(xn+p , xn ) = 0.
n→∞

Thus, {xn } is a Cauchy sequence (and hence bounded). Since (X, d) is a complete
metric space, there exists a unique element x ∗ ∈ X such that

d
xn −→ x ∗

in the metric as indicated. Now we must show that x ∗ is the unique fixed point of
the operator F. Again by use of the triangle inequality, the reader can easily verify
that

d(x ∗ , F (x ∗ )) ≤ d(x ∗ , xn ) + d(xn , F (x ∗ ))

= d(x ∗ , xn ) + d(F (xn−1 ), F (x ∗ ))
≤ d(x ∗ , xn ) + α d(xn−1 , x ∗ ), ∀ n ≥ 1. (1.19)

Letting n −→ ∞, it follows from this inequality that

d(x ∗ , F (x ∗ )) = 0
Random documents with unrelated
content Scribd suggests to you:
shilling, or some nonsense, and that he would give Stella to nobody that
couldn’t do that. On the face of it that is folly, you know.”
“I am not so sure that it is folly. I have heard him say something of the
kind; meaning, I suppose, that any son-in-law he would accept would have
to be as wealthy as himself.”
“But that is absolute madness, Dr. Burnet! Good heavens! who that was
as rich as old Tredgold could desire to be old Tredgold’s son-in-law? It is
against all reason. A man might forgive to the girls who are so nice in
themselves that they had such a father; but what object could one as rich as
himself—— Oh! it is sheer idiocy, you know.”
“Not to him; and he, after all, is the person most concerned,” said Dr.
Burnet, with his head cast down and rather a dejected look about him
altogether. The thought was not cheerful to himself any more than to Lady
Jane, and as a matter of fact he had not realised it before.
“But it cannot be,” she cried, “it cannot be; it is out of the question. Oh,
you are a man of resource; you must find out some way to baffle this old
curmudgeon. There must, there must,” she exclaimed, “be some way out of
it, if you care to try.”
“Trying will not invent thousands of pounds, alas! nor can the man who
has the greatest fund of resource but no money do it anyhow,” said Dr.
Burnet sententiously. “There may be a dodge——”
“That is what I meant. There must be a dodge to—to get you out of it,”
she cried.
“It is possible that the man whose existence you divine might not care to
get a wife—if she would have him to begin with—by a dodge, Lady Jane.”
“Oh, rubbish!” cried the great lady, “we are not so high-minded as all
that. I am of opinion that in that way anything, everything can be done.
Charlie Somers is a fool and Stella another; but to a sensible pair with an
understanding between them and plenty of time to work—and an old sick
man,” Lady Jane laid an involuntary emphasis on the word sick—then
stopped and reddened visibly, though her countenance was rather weather-
beaten and did not easily show.
“A sick man—to be taken advantage of? No, I think that would scarcely
do,” he said. “A sensible pair with an understanding, indeed—but then the
understanding—there’s the difficulty.”
 “No,” cried Lady Jane, anxiously cordial to wipe away the stain of her
unfortunate suggestion. “Not at all—the most natural thing in the world—
where there is real feeling, Dr. Burnet, on one side, and a lonely, sensitive
girl on the other——”
“A lonely, sensitive girl,” he repeated. And then he looked up in Lady
Jane’s face with a short laugh—but made no further remark.
Notwithstanding the safeguard of her complexion, Lady Jane this time
grew very red indeed; but having nothing to say for herself, she was wise
and made no attempt to say it. And he got up, having nothing further to add
by any possibility to his prescription, and put it into her hand.
“I must make haste home,” he said, “the snow is very blinding, and the
roads by this time will be scarcely distinguishable.”
“I am sorry to have kept you so long—with my ridiculous cold, which is
really nothing. But Dr. Burnet,” she said, putting her hand on his sleeve,
“you will think of what I have said. Let justice be done to those poor
Somers. Their poverty is something tragic. They had so little expectation of
anything of the kind.”
“It is most unlikely that I can be of any use to them, Lady Jane,” he said
a little stiffly, as he accepted her outstretched hand.
Perhaps Lady Jane had more respect for him than ever before. She held
his prescription in her hand and looked at it for a moment.
“I think I’ll take it,” she said to herself as if making a heroic resolution.
She had really a little cold.
As for the doctor, he climbed up into his dog-cart and took the reins from
the benumbed hands of Jim, who was one mass of whiteness now instead of
the black form sprinkled over with flakes of white which he had appeared at
the Cliff. It was a difficult thing to drive home between the hedges, which
were no longer visible, and with the big snow-flakes melting into his eyes
and confusing the atmosphere, and he had no time to think as long as he
was still out in the open country, without even the lights of Sliplin to guide
him. It was very cold, and his hands soon became as benumbed as Jim’s,
with the reins not sensible at all through his big gloves to his chilled fingers.
“I think we should turn to the left, here?” he said to Jim, who answered
“Yessir,” with his teeth chattering, “or do you think it should perhaps be to
the right?”
 Jim said “Yessir,” again, dull to all proprieties.
If Jim had been by himself he would probably have gone to sleep, and
allowed the mare to find her own way home, which very likely she would
have done; but Dr. Burnet could not trust to such a chance. To think much
of what had been said to him was scarcely possible in these circumstances.
But when the vague and confused glimmer of the Sliplin lights through the
snow put his mind at rest, it cannot but be said that Dr. Burnet found a great
many thoughts waiting to seize hold upon him. He was not perhaps
surprised that Lady Jane should have divined his secret. He had no
particular desire to conceal it, and though he did not receive Lady Jane’s
offer with enthusiasm, he could not but feel that her friendship and
assistance would be of great use to him—in fact, if not with Katherine, at
least with other things. It would be good for him professionally, even this
one visit, and the prescription for Lady Jane, not for Mrs. Cole, which must
be made up at the chemist’s, would do him good. A man who held the
position of medical attendant at Steephill received a kind of warrant of skill
from the fact, which would bring other patients of distinction. When Dr.
Burnet got home, and got into dry and comfortable clothes, and found no
impatient messenger awaiting him, it was with a grateful sense of ease that
he gave himself up to the study of this subject by the cheerful fire. His mind
glanced over the different suggestions of Lady Jane, tabulating and
classifying them as if they had been scientific facts. There was that hint
about the old sick man, which she had herself blushed for before it was
fully uttered, and at which Dr. Burnet now grinned in mingled wrath and
ridicule. To take advantage of an old sick man—as being that old man’s
medical attendant and desirous of marrying his daughter—was a suggestion
at which Burnet could afford to laugh, though fiercely, and with an
exclamation not complimentary to the intentions of Lady Jane. But there
were other things which required more careful consideration.
Should he follow these other suggestions, he asked himself? Should he
become a party to her plan, and get her support, and accept the privileges of
a visitor at her house as she had almost offered, and meet Katherine there,
which would probably be good for Katherine in other ways as well as for
himself? There was something very tempting in this idea, and Dr. Burnet
was not mercenary in his feeling towards Katherine, nor indisposed to do
“justice to Stella” in the almost incredible case that it ever should be in his
power to dispose of Mr. Tredgold’s fortune. He could not help another short
laugh to himself at the absurdity of the idea. He to dispose of Mr.
Tredgold’s fortune! So many things were taken for granted in this ridiculous
hypothesis. Katherine’s acceptance and consent for one thing, of which he
was not at all sure. She had evidently sent the Rector about his business,
which made him glad, yet gave him a little thrill of anxiety too, for, though
he was ten years younger than the Rector, and had no family to encumber
him, yet Mr. Stanley, on the other hand, was a handsome man, universally
pleasing, and perhaps more desirable in respect to position than an ordinary
country practitioner—a man who dared not call his body, at least, whatever
might be said of his soul, his own; and who had as yet had no opportunity
of distinguishing himself. If she repulsed the one so summarily, would she
not have in all probability the same objections to the other? At twenty-three
a man of thirty-five is slightly elderly as well as one of forty-seven.
Supposing, however, that Katherine should make no objection, which
was a very strong step for a man who did not in the least believe that at the
present moment she had even thought of him in that light—there was her
father to be taken into account. He had heard Mr. Tredgold say that about
the thousand for thousand told down on the table, and he had heard it from
the two ladies of the midge; but without, perhaps, paying much attention or
putting any great faith in it. How could he table thousand for thousand
against Mr. Tredgold? The idea was ridiculous. He had the reversion of that
little, but ancient, estate in the North, of which he had been at such pains to
inform Katherine; and he had a little money from his mother; and his
practice, which was a good enough practice, but not likely to produce
thousands for some time at least to come. He had said there might be a
dodge—and, as a matter of fact, there had blown across his mind a
suggestion of a dodge, how he might perhaps persuade his uncle to “table”
the value of Bunhope on his side. But what was the value of Bunhope to the
millions of old Tredgold? He might, perhaps, say that he wanted nothing
more with Katherine than the equivalent of what he brought; but he doubted
whether the old man would accept that compromise. And certainly, if he did
so, there could be no question of doing justice to Stella out of the small
share he would have of her father’s fortune. No; he felt sure Mr. Tredgold
would exact the entire pound of flesh, and no less; that he would no more
reduce his daughter’s inheritance than her husband’s fortune, and that no
dodge would blind the eyes of the acute, businesslike old man.
 This was rather a despairing point of view, from which Dr. Burnet tried
to escape by thinking of Katherine herself, and what might happen could he
persuade her to fall in love with him. That would make everything so much
more agreeable; but would it make it easier? Alas! falling in love on Stella’s
part had done no good to Somers; and Stella, though now cast off and
banished, had possessed a far greater influence over her father than
Katherine had ever had. Dr. Burnet was by no means destitute of sentiment
in respect to her. Indeed, it is very probable that had Katherine had no
fortune at all he would still have wished, and taken earlier more decisive
steps, to make her aware that he wished to secure her for his wife; but the
mere existence of a great fortune changes the equilibrium of everything.
And as it was there, Dr. Burnet felt that to lose it, if there was any possible
way of securing it, would be a great mistake. He was the old man’s doctor,
who ought to be grateful to him for promoting his comfort and keeping him
alive; and he was Katherine’s lover, and the best if not the only one there
was. And he had free access to the house at all seasons, and a comfortable
standing in the drawing-room as well as in the master’s apartment. Surely
something must be made of these advantages by a man with his eyes open,
neglecting no opportunity. And, on the other hand, there was always the
chance that old Tredgold might die, thus simplifying matters. The doctor’s
final decision was that he would do nothing for the moment, but wait and
follow the leading of circumstances; always keeping up his watch over
Katherine, and endeavouring to draw her interest, perhaps in time her
affections, towards himself—while, on the other hand, it would commit him
to nothing to accept Lady Jane’s help, assuring her that—in the case which
he felt to be so unlikely of ever having any power in the matter—he would
certainly do “justice to Stella” as far as lay in his power.
When he had got to this conclusion the bell rang sharply, and, alas! Dr.
Burnet, who had calculated on going to bed for once in comfort and quiet,
had to face the wintry world again and go out into the snow.
 CHAPTER XXVII.
Katherine’s life at Sliplin was in no small degree affected by the result of
the Rector’s unfortunate visit. How its termination became known nobody
could tell. No one ventured to say “She told me herself,” still less, “He told
me.” Yet everybody knew. There were some who had upheld that the Rector
had too much respect for himself ever to put himself in the position of being
rejected by old Tredgold’s daughter; but even these had to acknowledge that
this overturn of everything seemly and correct had really happened. It was
divined, perhaps, from Mr. Stanley’s look, who went about the parish with
his head held very high, and an air of injury which nobody had remarked in
him before. For it was not only that he had been refused. That is a privilege
which no law or authority can take from a free-born English girl, and far
would it have been from the Rector’s mind to deny to Katherine this right;
but it was the manner in which it had been exercised which gave him so
deep a wound. It was not as the father of Charlotte and Evelyn that Mr.
Stanley had been in the habit of regarding himself, nor that he had been
regarded. His own individuality was too remarkable and too attractive, he
felt with all modesty, to lay him under such a risk; and yet here was a young
woman in his own parish, in his own immediate circle, who regarded him
from that point of view, and who looked upon his proposal as ridiculous and
something like an insult to her youth. Had she said prettily that she did not
feel herself good enough for such a position, that she was not worthy—but
that she was aware of the high compliment he had paid her, and never
would forget it—which was the thing that any woman with a due sense of
fitness would have said, he might have forgiven her. But Katherine’s
outburst of indignation, her anger to have been asked to be the stepmother
of Charlotte and Evelyn her playfellows, her complete want of gratitude or
of any sense of the honour done her, had inflicted a deep blow upon the
Rector. That he should be scorned as a lover seemed to him impossible, that
a woman should be so insensible to every fact of life. He did not get over it
for a long time, nor am I sure that he ever did get over it; not the
disappointment, which he bore like a man, but the sense of being scorned.
So long as he lived he never forgave Katherine that insult to his dearest
feelings.
 And thus Katherine’s small diversions were driven back into a still
narrower circle. She could not go to the Rectory, where the girls were
divided between gratitude to her for not having turned their life upside
down, and wrath against her for not having appreciated papa; nor could she
go where she was sure to meet him, and to catch his look of offended pride
and wounded dignity. It made her way very hard for her to have to think
and consider, and even make furtive enquiries whether the Stanleys would
be there before going to the mildest tea party. When Mrs. Shanks invited
her to meet Miss Mildmay, she was indeed safe. Yet even there Mr. Stanley
might come in to pay these ladies a call, or Charlotte appear with her
portfolio of drawings, or Evelyn fly in for a moment on her way to the post.
She went even to that very mild entertainment with a quiver of anxiety. The
great snowstorm was over which had stopped everything, obliterating all
the roads, and making the doctor’s dog-cart and the butcher’s and baker’s
carts the only vehicles visible about the country—which lay in one great
white sheet, the brilliancy of which made the sea look muddy where it came
up with a dull colour upon the beach. Everything, indeed, looked dark in
comparison with that dazzling cloak of snow, until by miserable human
usage the dazzling white changed into that most squalid of all squalid
things, the remnant of a snowstorm in England, drabbled by all kinds of
droppings, powdered with dust of smoke and coal, churned into the chillest
and most dreadful of mud. The island had passed through that horrible
phase after a brief delicious ecstasy of skating, from which poor Katherine
was shut out by the same reasons already given, but now had emerged green
and fresh, though cold, with a sense of thankfulness which the fields
seemed to feel, and the birds proclaimed better and more than the best of
the human inhabitants could do.
The terrace gardens of Mrs. Shanks and Miss Mildmay shone with this
refreshed and brightened greenness, and the prospect from under the
verandah of their little houses was restored to its natural colour. The sea
became once more the highest light in the landscape, the further cliffs were
brown, the trees showed a faint bloom of pushing buds and rising sap, and
glowed in the light of the afternoon sun near its setting. Mrs. Shanks’ little
drawing room was a good deal darkened by its little verandah, but when the
western sun shone in, as it was doing, the shade of the little green roof was
an advantage even in winter; and it was so mild after the snow that the
window was open, and a thrush in a neighbouring shrubbery had begun to
perform a solo among the bushes, exactly, as Mrs. Shanks said, like a fine
singer invited for the entertainment of the guests.
“It isn’t often you hear a roulade like that,” she said. “I consider Miss
Sherlock was nothing to it.” Miss Sherlock was a professional lady who had
been paying a visit in Sliplin, and who at afternoon teas and evening parties,
being very kind and ready to “oblige,” had turned the season into a musical
one, and provided for the people who were so kind as to invite her, an
entertainment almost as cheap as that of the thrush in Major Toogood’s
shrubbery.
“I hope the poor thing has some crumbs,” said Miss Mildmay. “I always
took great pains to see that there was plenty of bread well peppered put out
for them during the snow.”
“Was Miss Sherlock so very good?” said Katherine. “I was unfortunate, I
never heard her, even at her concert. Oh, yes, I had tickets—but I did not
go.”
“That is just what we want to talk to you about, my dear Katherine.
Fancy a great singer in Sliplin, and the Cliff not represented, not a soul
there. Oh, if poor dear Stella had but been here, she would not have stayed
away when there was anything to see or hear.”
“Yes, I am a poor creature in comparison,” said Katherine, “but you
know it isn’t nice to go to such places alone.”
“If there was any need to go alone! You know we would have called for
you in the midge any time; but that’s ridiculous for you with all your
carriages; it would have been more appropriate for you to call for us.
Another time, Katherine, my dear——”
“Oh, I know how kind you are; it was not precisely for want of some one
to go with.”
“Jane Shanks,” said Miss Mildmay, “what is the use of pretences
between us who have known the child all her life? It is very well
understood in Sliplin, Katherine, that there must be some motive in your
seclusion. You have some reason, you cannot conceal it from us who know
you, for shutting yourself up as you do.”
“What reason? Is it not a good enough reason that I am alone now, and
that to be reminded of it at every moment is—oh, it is hard,” said Katherine,
tears coming into her eyes. “It is almost more than I can bear.”
 “Dear child!” Mrs. Shanks said, patting her hand which rested on the
table. “We shouldn’t worry her with questions, should we?” But there was
no conviction in her tone, and Katherine, though her self-pity was quite
strong enough to bring that harmless water to her eyes, was quite aware not
only that she did not seclude herself because of Stella, but also that her
friends were not in the least deceived.
“I ask no questions,” said Miss Mildmay, “I hope I have a head on my
shoulders and a couple of eyes in it. I don’t require information from
Katherine! What I’ve got to say is that she mustn’t do it. Most girls think
very little of refusing a man; sometimes they continue good friends,
sometimes they don’t. When a man sulks it shows he was much in earnest,
and is really a compliment. But to stay at home morning and night because
there is a man in the town who is furious with you for not marrying him;
why, that’s a thing that is not to be allowed to go on, not for a day——”
“Nobody has any right to say that there is any man whom——”
“Oh, don’t redden up, Katherine, and flash your eyes at me! I have
known you since you were that high, and I don’t care a brass button what
you say. Do you think I don’t know all about you, my dear? Do you think
that there’s a thing in Sliplin which I don’t know or Jane Shanks doesn’t
know? Bless us, what is the good of us, two old cats, as I know you call us
——”
“Miss Mildmay!” cried Katherine; but as it was perfectly true, she
stopped there and had not another word to say.
“Yes, that’s my name, and her name is Mrs. Shanks; but that makes no
difference. We are the two old cats. I have no doubt it was to Stella we
owed the title, and I don’t bear her any malice nor you either. Neither does
Jane Shanks. We like you, on the contrary, my dear; but if you think you
can throw dust in our eyes—— Why, there is the Rector’s voice through the
partition asking for me.”
“Oh,” said Katherine, “I must go, really I must go; this is the time when
papa likes me to go to him. I have stayed too long, I really, really must go
now——”
“Sit down, sit down, dear. It is only her fun. There is nobody speaking
through the partition. The idea! Sliplin houses are not very well built, but I
hope they are better than that.”
 “I must have been mistaken,” said Miss Mildmay grimly. “I believe after
all it is only Jane Shanks’ boy; he has a very gruff manly voice, though he is
such a little thing, and a man’s voice is such a rarity in these parts that he
deceives me. Well, Katherine, the two old cats hear everything. If it does
not come to me it comes to her. My eyes are the sharpest, I think, but she
hears the best. You can’t take us in. We know pretty well all that has
happened to you, though you have been so very quiet about it. There was
that young city man whom you wouldn’t have, and I applaud you for it. But
he’ll make a match with somebody of much more consequence than you.
And then there is poor Mr. Stanley. The Stanleys are as thankful to you as
they can be, and well they may. Why, it would have turned the whole place
upside down. A young very rich wife at the Rectory and the poor girls
turned out of doors. It just shows how little religion does for some people.”
“Oh, stop! stop!” cried Mrs. Shanks. “What has his religion to do with
it? It’s not against any man’s religion to fall in love with a nice girl.”
“Please don’t say any more on this subject,” cried Katherine; “if you
think it’s a compliment to me to be fallen in love with—by an old
gentleman!—— But I never said a word about the Rector. It is all one of
your mistakes. You do make mistakes sometimes, Miss Mildmay. You took
little Bobby’s voice for—a clergyman’s.” It gave more form to the
comparison to say a clergyman than merely a man.
“So I did,” said Miss Mildmay, “that will always be remembered against
me; but you are not going to escape, Katherine Tredgold, in that way. I shall
go to your father, if you don’t mind, and tell him everything, and that you
are shutting yourself up and seeing nobody, because of—— Well, if it is not
because of that, what is it? It is not becoming, it is scarcely decent that a
girl of your age should live so much alone.”
“Please let me go, Mrs. Shanks,” said Katherine. “Why should you
upbraid me? I do the best I can; it is not my fault if there is nobody to stand
by me.”
“We shall all stand by you, my dear,” said Mrs. Shanks, following her to
the door, “and Ruth Mildmay is never so cross as she seems. We will stand
by you, in the midge or otherwise, wherever you want to go. At all times
you may be sure of us, Katherine, either Ruth Mildmay or me.”
But when the door was closed upon Katherine Mrs. Shanks rushed back
to the little drawing-room, now just sinking into greyness, the last ray of the
sunset gone. “You see,” she cried, “it’s all right, I to——”
But she was forestalled with a louder “I told you so!” from Miss
Mildmay; “didn’t I always say it?” that lady concluded triumphantly. Mrs.
Shanks might begin the first, but it was always her friend who secured the
last word.
Katherine walked out into the still evening air, a little irritated, a little
disgusted, and a little amused by the offer of these two chaperons and the
midge to take her about. She had to walk through the High Street of Sliplin,
and everybody was out at that hour. She passed Charlotte Stanley with her
portfolio under her arm, who would probably have rushed to her and
demanded a glance at the sketches even in the open road, or that Katherine
should go in with her to the stationer’s to examine them at her ease on the
counter; but who passed now with an awkward bow, having half crossed the
road to get out of her way, yet sending a wistful smile nevertheless across
what she herself would have called the middle distance. “Now what have I
done to Charlotte?” Katherine said to herself. If there was anyone who
ought to applaud her, who ought to be grateful to her, it was the Rector’s
daughters. She went on with a sort of rueful smile on her lips, and came up
without observing it to the big old landau, in which was seated Lady Jane.
Katherine was hurrying past with a bow, when she was suddenly greeted
from that unexpected quarter with a cry of “Katherine! where are you going
so fast?” which brought her reluctantly back.
“My dear Katherine! what a long time it is since we have met,” said
Lady Jane.
“Yes,” said Katherine sedately. “That is very true, it is a long time.”
“You mean to say it is my fault by that tone! My dear, you have more
horses and carriages, and a great deal more time and youth and all that than
I. Why didn’t you come to see me? If you thought I was huffy or neglectful,
why didn’t you come and tell me so? I should have thought that was the
right thing to do.”
“I should not have thought it becoming,” cried Katherine, astonished by
this accost, “from me to you. I am the youngest and far the humblest——”
“Oh, fiddlesticks!” cried the elder lady, “that’s not true humility, that’s
pride, my dear. I was an old friend; and though poor dear Stella always put
herself in the front, you know it was you I liked best, Katherine. Well, when
will you come, now? Come and spend a day or two, which will be
extremely dull, for we’re all alone; but you can tell me of Stella, as well as
your own little affairs.”
“I don’t know that I can leave papa,” Katherine said, with a little
remnant of that primness which had been her distinction in Captain Scott’s
eyes.
“Nonsense! He will spare you to me,” said Lady Jane with calm
certainty. “Let me see, what day is this, Tuesday? Then I will come for you
on Saturday. You can send over that famous little brougham with your maid
and your things, and keep it if you like, for we have scarcely anything but
dog-carts, except this hearse. Saturday; and don’t show bad breeding by
making any fuss about it,” Lady Jane said.
Katherine felt that the great lady was right, it would have been bad
breeding; and then her heart rose a little in spite of herself at the thought of
the large dull rooms at Steephill in which there was no gilding, nor any
attempt to look finer than the most solid needs of life demanded, and where
Lady Jane conducted the affairs of life with a much higher hand than any of
the Sliplin ladies. After being so long shut up in Sliplin, and now partly out
of favour in it, the ways of Lady Jane seemed bigger, the life more easy and
less self-conscious, and she consented with a little rising of her heart. She
was a little surprised that Lady Jane, with her large voice, should have
shouted a cordial greeting to the doctor as he passed in his dog-cart. “I am
going to write to you,” she cried, nodding her head at him; but no doubt this
was about some little ailment in the nursery, for with Katherine, a young
lady going on a visit to Steephill, what could it have to do?
 CHAPTER XXVIII.
The doctor had made himself a very important feature in Katherine’s life
during those dull winter days. After the great snowstorm, which was a thing
by which events were dated for long after, in the island, and which was
almost coincident with the catastrophe of the Rector; he had become more
frequent in his visits to Mr. Tredgold and consequently to the tea-table of
Mr. Tredgold’s lonely daughter. While the snow lasted, and all the
atmospheric influences were at their worst, it stood to reason that an
asthmatical, rheumatical, gouty old man wanted more looking after than
usual; and it was equally clear that a girl a little out of temper and out of
patience with life, who was disposed to shut herself up and retire from the
usual amusements of her kind, would also be much the better for the
invasion into her closed-up world of life and fresh air in the shape of a
vigorous and personable young man, who, if not perhaps so secure in self-
confidence and belief in his own fascinations as the handsome (if a little
elderly) Rector, had not generally been discouraged by the impression he
knew himself to have made. And Katherine had liked those visits, that was
undeniable; the expectation of making a cup of tea for the doctor had been
pleasant to her. The thought of his white strong teeth and the bread and
butter which she never got out of her mind, was now amusing, not painful;
she had seen him so often making short work of the little thin slices
provided for her own entertainment. And he told her all that was going on,
and gave her pieces of advice which his profession warranted. He got to
know more of her tastes, and she more of his in this way, than perhaps was
the case with any two young people in the entire island, and this in the most
simple, the most natural way. If there began to get a whisper into the air of
Dr. Burnet’s devotion to his patient on the Cliff and its possible
consequences, that was chiefly because the doctor’s inclinations had been
suspected before by an observant public. And indeed the episode of the
Rector had afforded it too much entertainment to leave the mind of Sliplin
free for further remark in respect to Katherine and her proceedings. And Mr.
Tredgold’s asthma accounted for everything in those more frequent visits to
the Cliff. All the same, it was impossible that there should not be a degree
of pleasant intimacy and much self-revelation on both sides during these
half hours, when, wrapped in warmth and comfort and sweet society, Dr.
Burnet saw his dog-cart promenading outside in the snow or during the
deeper miseries of the thaw, with the contrast which enhances present
pleasure. He became himself more and more interested in Katherine, his
feelings towards her being quite genuine, though perhaps enlivened by her
prospects as an heiress. And if there had not been that vague preoccupation
in Katherine’s mind concerning James Stanford, the recollection not so
much of him as of the many, many times she had thought of him, I think it
very probable indeed that she would have fallen in love with the doctor;
indeed, there were moments when his image pushed Stanford very close,
almost making that misty hero give way. He was a very misty hero, a
shadow, an outline, indefinite, never having given much revelation of
himself; and Dr. Burnet was very definite, as clear as daylight, and in many
respects as satisfactory. It would have been very natural indeed that the one
should have effaced the other.
Dr. Burnet did not know anything of James Stanford. He thought of
Katherine as a little shy, a little cold, perhaps from the persistent shade into
which she had been cast by her sister, unsusceptible as people say; but he
did not at all despair of moving her out of that calm. He had thought indeed
that there were indications of the internal frost yielding, before his
interview with Lady Jane. With Lady Jane’s help he thought there was little
doubt of success. But even that security made him cautious. It was evident
that she was a girl with whom one must not attempt to go too fast. The
Rector had tried to carry the fort by a coup de main, and he had perished
ingloriously in the effort. Dr. Burnet drew himself in a little after he
acquired the knowledge of that event, determined not to risk the same fate.
He had continued his visits but he had been careful to give them the most
friendly, the least lover-like aspect, to arouse no alarms. When he received
the salutation of Lady Jane in passing, and her promise that he should hear
from her, his sober heart gave a bound, which was reflected unconsciously
in the start of the mare making a dash forward by means of some
magnetism, it is to be supposed conveyed to her by the reins from her
master’s hand—so that he had to exert himself suddenly with hand and
whip to reduce her to her ordinary pace again. If the manœuvre had been
intentional it would have been clever as showing his skill and coolness in
the sight of his love and of his patroness. It had the same effect not being
intentional at all.
 I am not sure either whether it was Lady Jane’s intention to enhance the
effect of Dr. Burnet by the extreme dulness of the household background
upon which she set him, so to speak, to impress the mind of Katherine.
There was no party at Steephill. Sir John, though everything that was good
and kind, was dull; the tutor, who was a young man fresh from the
University, and no doubt might have been very intellectual or very frivolous
had there been anything to call either gifts out, was dull also because of
having little encouragement to be anything else. Lady Jane indeed was not
dull, but she had no call upon her for any exertion; and the tone of the house
was humdrum beyond description. The old clergyman dined habitually at
Steephill on the Sunday evenings, and he was duller still, though invested to
Katherine with a little interest as the man who had officiated at her sister’s
marriage. But he could not be got to recall the circumstance distinctly, nor
to master the fact that this Miss Tredgold was so closely related to the
young lady whom he had made into Lady Somers. “Dear! dear! to think of
that!” he had said when the connection had been explained to him, but what
he meant by that exclamation nobody knew. I think it very likely that Lady
Jane herself was not aware how dull her house was when in entire repose,
until she found it out by looking through the eyes of a chance guest like
Katherine. “What in thunder did you mean by bringing that poor girl here to
bore her to death, when there’s nobody in the house?” Sir John said, whose
voice was like a westerly gale. “Really, Katherine, I did not remember how
deadly dull we were,” Lady Jane said apologetically. “It suits us well
enough—Sir John and myself; but it’s a shame to have asked you here when
there’s nobody in the house, as he says. And Sunday is the worst of all,
when you can’t have even your needlework to amuse you. But there are
some people coming to dinner to-morrow.” Katherine did her best to
express herself prettily, and I don’t think even that she felt the dulness so
much as she was supposed to do. The routine of a big family house, the
machinery of meals and walks and drives and other observances, the
children bursting in now and then, the tutor appearing from time to time
tremendously comme il faut, and keeping up his equality, Sir John, not half
so careful, rolling in from the inspection of his stables or his turnips with a
noisy salutation, “You come out with me after lunch, Miss Tredgold, and
get a blow over the downs, far better for you than keeping indoors.” And
then after that blow on the downs, afternoon tea, and Mr. Montgomery
rubbing his hands before the fire, while he asked, without moving, whether
he should hand the kettle. All this was mildly amusing, in the proportion of
its dulness, for a little while. We none of us, or at least few of us, feel
heavily this dull procession of the hours when it is our own life; when it is
another’s, our perceptions are more clear.
“But there are people coming to dinner to-morrow,” Lady Jane said.
There was something in the little nod she gave, of satisfaction and
knowingness, which Katherine did not understand or attempt to understand.
No idea of Dr. Burnet was associated with Steephill. She was not aware that
he was on visiting terms there—he had told her that he attended the
servants’ hall—so that it was with a little start of surprise that, raising her
eyes from a book she was looking at, she found him standing before her,
holding out his hand as the guests gathered before dinner. The party was
from the neighbourhood—county, or, at least, country people—and when
Dr. Burnet was appointed to take Katherine in to dinner, that young lady,
though she knew the doctor so well and liked him so much, did not feel that
it was any great promotion. She thought she might have had somebody
newer, something that belonged less to her own routine of existence, which
is one of the mistakes often made by very astute women of the world like
Lady Jane. There was young Fortescue, for instance, a mere fox-hunting
young squire, not half so agreeable as Dr. Burnet, whom Katherine would
have preferred. “He is an ass; he would not amuse her in the very least,”
Lady Jane had said. But Sir John, who was not clever at all, divined that
something new, though an ass, would have amused Katherine more.
Besides, Lady Jane had her motives, which she mentioned to nobody.
Dr. Burnet did the very best for himself that was possible. He gave
Katherine a report of her father, he told her the last thing that had transpired
at Sliplin since her departure, he informed her who all the people were at
table, pleased to let her see that he knew them all. “That’s young Fortescue
who has just come in to his estate, and he promises to make ducks and
drakes of it,” Dr. Burnet said. Katherine looked across the table at the
young man thus described. She was not responsible for him in any way, nor
could it concern her if he did make ducks and drakes of his estate, but she
would have preferred to make acquaintance with those specimens of the
absolutely unknown. A little feeling suddenly sprang up in her heart against
Dr. Burnet, because he was Dr. Burnet and absolutely above reproach. She
would have sighed for Dr. Burnet, for his quick understanding and the
abundance he had to say, had she been seated at young Fortescue’s side.
 After dinner, when she had talked a little to all the ladies and had done
her duty, Lady Jane caught Katherine’s hand and drew her to a seat beside
herself, and then she beckoned to Dr. Burnet, who drew a chair in front of
them and sat down, bending forward till his head, Katherine thought, was
almost in Lady Jane’s lap. “I want,” she said, “Katherine, to get Dr. Burnet
on our side—to make him take up our dear Stella’s interests as you do, my
dear, and as in my uninfluential way I should like to do too.”
“How can Dr. Burnet take up Stella’s interests?” cried Katherine,
surprised and perhaps a little offended too.
“My dear Katherine, a medical man has the most tremendous
opportunities—all that the priest had in old times, and something additional
which belongs to himself. He can often say a word when none of the rest of
us would dare to do so. I have immense trust in a medical man. He can
bring people together that have quarrelled, and—and influence wills, and—
do endless things. I always try to have the doctor on my side.”
“Miss Katherine knows,” said Dr. Burnet, trying to lead out of the
subject, for Lady Jane’s methods were entirely, on this occasion, too
straightforward, “that the medical man in this case is always on her side.
Does not Mrs. Swanson, Lady Jane, sing very well? I have never heard her.
I am not very musical, but I love a song.”
“Which is a sign that you are not musical. You are like Sir John,” said
Lady Jane, as if that was the worst that could be said. “Still, if that is what
you mean, Dr. Burnet, you can go and ask her, on my part. He is very much
interested in you all, I think, Katherine,” she added when he had departed
on this mission. “We had a talk the other day—about you and Stella and the
whole matter. I think, if he ever had it in his power, that he would see
justice done her, as you would yourself.”
“He is very friendly, I daresay,” said Katherine, “but I can’t imagine how
he could ever have anything in his power.”
“There is no telling,” Lady Jane said. “I think he is quite a disinterested
man, if any such thing exists. Now, we must be silent a little, for, of course,
Mrs. Swanson is going to sing; she is not likely to neglect an opportunity.
She has a good voice, so far as that goes, but little training. It is just the
thing that pleases Sir John. And he has planted himself between us and the
piano, bless him! now we can go on with our talk. Katherine, I don’t think
you see how important it is to surround your father with people who think
the same as we do about your poor sister.”
“No,” said Katherine, “it has not occurred to me; my father is not very
open to influence.”
“Then do you give up Stella’s cause? Do you really think it is hopeless,
Katherine?”
“How could I think so?” cried the girl with a keen tone in her voice
which, though she spoke low, was penetrating, and to check which, Lady
Jane placed her hand on Katherine’s hand and kept it there with a faint
“shsh.” “You know what I should instantly do,” she added, “if I ever had it
in my power.”
“Dear Katherine! but your husband might not see it in that light.”
“He should—or he should not be—my husband,” said Katherine with a
sudden blush. She raised her eyes unwillingly at this moment and caught
the gaze of Dr. Burnet, who was standing behind the great bulk of Sir John,
but with his face towards the ladies on the sofa. Katherine’s heart gave a
little bound, half of affright. She had looked at him and he at her as she said
the words. An answering gleam of expression, an answering wave of colour,
seemed to go over him (though he could not possibly hear her) as she
spoke. It was the first time that this idea had been clearly suggested to her,
but now so simply, so potently, as if she were herself the author of the
suggestion. She was startled out of her self-possession. “Oh,” she cried with
agitation, “I like her voice! I am like Sir John; let us listen to the singing.”
Lady Jane nodded her head, pressed Katherine’s hand, and did what was
indeed the first wise step she had taken, stepped as noiselessly as possible
to another corner, where, behind her fan, she could talk to a friend more
likely to respond to her sentiments and left Dr. Burnet to take her place.
“Is this permitted? It is too tempting to be lost,” he said in a whisper, and
then he too relapsed into silence and attention. Katherine, I fear, did not get
any clear impression of the song. Her own words went through her head,
involuntarily, as though she had touched some spring which went on
repeating them: “My husband—my husband.” Her white dress touched his
blackness as he sat down beside her. She drew away a little, her heart
beating loudly, in alarm, mingled with some other feeling which she could
not understand, but he did not say another word until the song was over, and
all the applause, and the moment of commotion in which the singer returned
to her seat, and the groups of the party changed and mingled. Then he said
suddenly, “I hope you will not think, Miss Katherine, that I desired Lady
Jane to drag me in head and shoulders to your family concerns. I never
should have been so presumptuous. I do trust you will believe that.”
“I never should have thought so, Dr. Burnet,” said Katherine, faltering
with that commotion which was she hoped entirely within herself and
apparent to no one. Then she added as she assured her voice, “It would not
have been presumptuous. You know so much of us already, and of her, and
took so much part——”
“I am your faithful servant,” he said, “ready to be sent on any errand, or
to take any part you wish, but I do not presume further than that.” Then he
rose quickly, as one who is moved by a sudden impulse. “Miss Katherine,
will you let me take you to the conservatory to see Lady Jane’s great aloe?
They used to say it blossomed only once in a hundred years.”
“But that’s all nonsense, you know,” said Mr. Montgomery the tutor;
“see them all about the Riviera at every corner. Truth, they kill ’emselves
when they’re about it.”
“Which comes to the same thing. Will you come?” said Dr. Burnet,
offering his arm.
“But, my dear fellow, Miss Tredgold has seen it three or four times,”
said this very unnecessary commentator.
“Never mind. She has not seen what I am going to show her,” said the
doctor with great self-possession. Lady Jane followed them with her eyes as
they went away into the long conservatory, which was famous in the islands
and full of lofty palms and tropical foliage. Her middle-aged bosom owned
a little tremor; was he going to put it to her, then and there? Lady Jane had
offered assistance, even co-operation, but this prompt action took away her
breath.
“I should like to see the aloe, too,” said the lady by her side.
“So you shall, presently,” said Lady Jane, “but we must not make a
move yet, for there is Lady Freshwater going to sing. Mr. Montgomery, ask
Lady Freshwater from me whether she will not sing us one of her delightful
French songs. She has such expression, and they are all as light as air of
course, not serious music. Look at Sir John, he is pleased, but he likes it
better when it is English, and he can make out the words. He is a constant
amusement when he talks of music—and he thinks he understands it, poor
dear.”
She kept talking until she had watched Lady Freshwater to the piano,
and heard her begin. And then Lady Jane felt herself entitled to a little rest.
She kept one eye on the conservatory to see that nobody interrupted the
botanical exposition which was no doubt going on there. Would he actually
propose—on the spot, all at once, with the very sound of the conversation
and of Lady Freshwater’s song in their ears? Was it possible that a man
should go so fast as that? Now that it had come to this point Lady Jane
began to get a little compunctious, to ask herself whether she might not
have done better for Katherine than a country doctor, without distinction,
even though he might have a wealthy uncle and a family place at his back?
Old Tredgold’s daughter was perhaps too great a prize to be allowed to drop
in that commonplace way. On the other hand, if Lady Jane had exerted
herself to get Katherine a better match, was it likely that a man—if a man of
our monde—would have consented to such an arrangement about Stella as
Dr. Burnet was willing to make? If the fortune had been Stella’s, Lady Jane
was quite certain that Charlie Somers would have consented to no such
settlement. And after all, would not Katherine be really happier with a man
not too much out of her own monde, fitted for village life, knowing all
about her, and not likely to be ashamed of his father-in-law? With this last
argument she comforted her heart.
And Katherine went into the conservatory to see the aloe, which that
malevolent tutor declared she had already seen so often, with her heart
beating rather uncomfortably, and her hand upon Dr. Burnet’s arm.
 CHAPTER XXIX.
But though Lady Jane had so fully made up her mind to it, and awaited the
result with so much excitement, and though Katherine herself was thrilled
with an uneasy consciousness, and Dr. Burnet’s looks gave every sanction
to the idea, he did not on that evening under the tall aloe, which had begun
to burst the innumerable wrappings of its husk, in the Steephill
conservatory, declare his love or ask Katherine to be his wife. I cannot tell
the reason why—I think there came over him a chill alarm as to how he
should get back if by any accident his suit was unsuccessful. It was like the
position which gave Mr. Puff so much trouble in the Critic. He could not
“exit praying.” How was he to get off the stage? He caught the eyes of an
old lady who was seated near the conservatory door. They were dull eyes,
with little speculation in them, but they gave a faint glare as the two young
people passed; and the doctor asked himself with a shudder, How could he
meet their look when he came back if——? How indeed could he meet
anybody’s look—Lady Jane’s, who was his accomplice, and who would be
very severe upon him if he did not succeed, and jolly Sir John’s, who would
slap him on the shoulder and shout at him in his big voice? His heart sank
to his boots when he found himself alone with the object of his affections
amid the rustling palms. He murmured something hurriedly about
something he wanted to say to her, but could not here, where they were
liable to interruption at any moment, and then he burst into a display of
information about the aloe which was very astounding to Katherine. She
listened, feeling the occasion manqué, with a sensation of relief. I think it
quite probable that in the circumstances, and amid the tremor of
sympathetic excitement derived from Lady Jane, and the general tendency
of the atmosphere, Katherine might have accepted Dr. Burnet. She would
probably have been sorry afterwards, and in all probability it would have
led to no results, but I think she would have accepted him that evening had
he had the courage to put it to the touch; and he, for his part, would
certainly have done it had he not been seized with that tremor as to how he
was to get off the stage.
He found it very difficult to explain this behaviour to Lady Jane
afterwards, who, though she did not actually ask the question, pressed him
considerably about the botanical lecture he had been giving.
“I have sat through a French café chantant song in your interests, with
all the airs and graces,” she said with a look of disgust, “to give you time.”
“Yes, I know,” said Dr. Burnet—it was at the moment of taking his
leave, and he knew that he must soon escape, which gave him a little
courage—“you have done everything for me—you have been more than
kind, Lady Jane.”
“But if it is all to come to nothing, after I had taken the trouble to
arrange everything for you!”
“It was too abrupt,” he said, “and I funked it at the last. How was I to get
back under everybody’s eyes if it had not come off?”
“It would have come off,” she said hurriedly, under her breath, with a
glance at Katherine. Then, in her usual very audible voice, she said, “Must
you go so early, Dr. Burnet? Then good-night; and if your mare is fresh take
care of the turning at Eversfield Green.”
He did not know what this warning meant, and neither I believe did she,
though it was a nasty turning. And then he drove away into the winter night,
with a sense of having failed, failed to himself and his own expectations, as
well as to Lady Jane’s. He had not certainly intended to take any decisive
step when he drove to Steephill, but yet he felt when he left it that the
occasion was manqué, and that he had perhaps risked everything by his lack
of courage. This is not a pleasant thought to a man who is not generally at a
loss in any circumstances, and whose ways have generally, on the whole,
been prosperous and successful. He was a fool not to have put it to the
touch, to be frightened by an old lady’s dull eyes which probably would
have noticed nothing, or the stare of the company which was occupied by
its own affairs and need not have suspected even that his were at a critical
point. Had he been a little bolder he might have been carrying home with
him a certainty which would have kept him warmer than any great-coat; but
then, on the other hand, he might have been departing shamed and cast
down, followed by the mocking glances of that assembly, and with Rumour
following after him as it followed the exit of the Rector, breathing among
all the gossips that he had been rejected; upon which he congratulated
himself that he had been prudent, that he had not exposed himself at least so
far. Finally he began to wonder, with a secret smile of superiority, how the
Rector had got off the scene? Did he “exit praying”?—which would at least