Linear And Nonlinear Functional Analysis With

Applications Philippe G Ciarlet download

https://ebookbell.com/product/linear-and-nonlinear-functional-
analysis-with-applications-philippe-g-ciarlet-5292338

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Nonlinear Optical Response In Atoms Molecules And Clusters An Explicit

Time Dependent Density Functional Approach 1st Edition Vladimir
Goncharov Auth

https://ebookbell.com/product/nonlinear-optical-response-in-atoms-
molecules-and-clusters-an-explicit-time-dependent-density-functional-
approach-1st-edition-vladimir-goncharov-auth-4931624

Linear And Nonlinear Nonfredholm Operators Theory And Applications 1st

Edition Messoud Efendiev

https://ebookbell.com/product/linear-and-nonlinear-nonfredholm-
operators-theory-and-applications-1st-edition-messoud-
efendiev-47666686

Linear And Nonlinear Optical Responses Of Chiral Multifold Semimetals

Miguel Ngel Snchez Martnez

https://ebookbell.com/product/linear-and-nonlinear-optical-responses-
of-chiral-multifold-semimetals-miguel-ngel-snchez-martnez-52564798

Linear And Nonlinear Optimization 2ed Griva I Nash S Sofer A

https://ebookbell.com/product/linear-and-nonlinear-optimization-2ed-
griva-i-nash-s-sofer-a-2043608
Linear And Nonlinear Iterative Learning Control 1st Edition Jianxin Xu

https://ebookbell.com/product/linear-and-nonlinear-iterative-learning-
control-1st-edition-jianxin-xu-2206418

Linear And Nonlinear Circuits Basic And Advanced Concepts 1st Ed 2018
Mauro Parodi

https://ebookbell.com/product/linear-and-nonlinear-circuits-basic-and-
advanced-concepts-1st-ed-2018-mauro-parodi-23634896

Linear And Nonlinear Integral Equations Methods And Applications 1st

Edition Abdulmajid Wazwaz

https://ebookbell.com/product/linear-and-nonlinear-integral-equations-
methods-and-applications-1st-edition-abdulmajid-wazwaz-2381624

Linear And Nonlinear Waves In Microstructured Solids Igor V Andrianov

https://ebookbell.com/product/linear-and-nonlinear-waves-in-
microstructured-solids-igor-v-andrianov-24471232

Linear And Nonlinear Programming Third Edition Third Edition David G

Luenberger

https://ebookbell.com/product/linear-and-nonlinear-programming-third-
edition-third-edition-david-g-luenberger-2477812
Linear and N onlinear
Functional Analysis
with Applications
 Philippe G. Ciarlet
City University of Hong Kong

Linear and N onlinear

Functional Analysis
with Applications
with 401 Problems and 52 Figures
-------
-----

..-
-�
----�....

-... -- ....

5.la.J11.. .
Society for lndustrial and Applied Mathematics
Philadelphia
Philippe G. Ciarlet
University Distinguished Professor
City University of Hong Kong
Hong Kong
and
Emeritus Professor
Université Pierre et Marie Curie
Paris, France

Copyright © 2013 by the Society for lndustrial and Applied Mathematics

10 9 8 7 6 5 4 3 2 1

Ail rights reserved. Printed in the United States of America. No part of this book may be reproduced,
stored, or transmitted in any manner without the written permission of the publisher. For information,
write to the Society for lndustrial and Applied Mathematics, 3600 Market Street, 6th Floor,
Philadelphia, PA 19104-2688 USA.

Figures 1.18-1,2,3; 7.13-1; 9.15-1,2; and 9.16-1 reprinted with permission from Elsevier.
Figures 4.3-4; 7.7-1,2; 7.12-1; and 7.16-1 reprinted with permission from Dunod.
Figures 8.1-1,2; 8.2-1; 8.3-1,2; 8.8-1,2,3; 8.9-1; 8.11-1,2,3; and 8.12-1 reprinted with kind permission
of Springer Science+Business Media.
Top image of Figure 8.12-2 reprinted courtesy ofWikipedia and Peter Mercator.
Middle image of Figure 8.12-2 reprinted courtesy ofWikipedia and YassineMrabet.
Bottom image of Figure 8.12-2 reprinted courtesy of StanWagon and with kind permission
of Springer Science+Business Media.

Library of Congress Cataloging-in-Publication Data

Ciarlet, Philippe G., author.
Linear and nonlinear functional analysis with applications/ Philippe G. Ciarlet, university distinguished
professor, City University of Hong Kong, Hong Kong, emeritus professor, Université Pierre et Marie
Curie, Paris, France.
pages cm. -- (Applied mathematics ; 130)
lncludes bibliographical references and index.
ISBN 978-1-611972-58-0 (alk. paper)
1. Functional analysis--Textbooks. 2. Nonlinear functional analysis--Textbooks. 1. Title.
QA320.C52 2013
515'.7--dc23 2013018736

5JaJ11.. is a registered trademark.

ÎO THE MEMORY OF MY PARENTS,

HÉLÈNE AND GASTON

CONTENTS

Preface xiii
1 Real Analysis and Theory of Functions: A Quick Review 1
Introduction . . 1
1.1 Sets . . . . . . . . . . . . . . . . . . . . 2
1.2 Mappings . . . . . . . . . . . . . . . . . 3
1.3 The axiom of choice and Zorn's lemma . 5
1.4 Construction of the sets lR and C . . . . 8
1.5 Cardinal numbers; finite and infinite sets . 9
1.6 Topological spaces . . . . . . . . . 11
1.7 Continuity in topological spaces . . . . . . 14
1.8 Compactness in topological spaces . . . . 15
1.9 Connectedness and simple-connectedness in topological spaces . 16
1.10 Metric spaces . . . . . . . . . . . . . . . . . . . . . 18
1.11 Continuity and uniform continuity in metric spaces 21
1 . 12 Complete metric spaces . . . . . . . . . . . . . . . 22
1.13 Compactness in metric spaces . . . . . . . . . . . . 23
1.14 The Lebesgue measure in !Rn ; measurable fonctions . 25
1.15 The Lebesgue integral in !Rn ; the basic theorems 28
1.16 Change of variable in Lebesgue integrals in !Rn 33
1 . 17 Volumes, areas, and lengths in !Rn . . . . . . 34
1.18 The spaces cm (n) and cm (n); domains in !Rn 36
2 Normed Vector Spaces 43
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2. 1 Vector spaces; Hamel bases; dimension of a vector space . . . . . . . 44
2.2 Normed vector spaces; first properties and examples; quotient spaces 47
2.3 The space C(K; Y) with K compact; uniform convergence and local uniform
convergence . . . . . . . . . . . . . . . . 53
2.4 The spaces f.P, 1 ::::; p::::; oo . . . . . . . . . . . . . . . . . . . . . . . 57
2.5 The Lebesgue spaces .LP (O ) , 1 ::::; p::::; oo . . . . . . . . . . . . . . . . 61
2.6 Regularization and approximation in the spaces .LP ( O ) , 1 ::::; p < oo 68
2.7 Compactness and finite-dimensional normed vector spaces; F. Riesz theorem . 76
2.8 Application of compactness in finite-dimensional normed vector spaces: The
fondamental theorem of algebra . . . . . . . . . . . . . . . . . . . . 79
vii
viii Contents

2.9 Continuous linear operators in normed vector spaces; the spaces C(X; Y),
C(X), and X' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2. 10 Compact linear operators in normed vector spaces . . . . . . . . . . 89
2. 1 1 Continuous multilinear mappings in normed vector spaces; the space
Ck (X1 , X2 , . . . , Xk ; Y) . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.12 Korovkin's theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2. 13 Application of Korovkin's theorem to polynomial approximation; Bohman's,
Bernstein's, and WeierstraB' theorems . . . . . . . . . . . . . . . . . . . . . . 100
2. 14 Application of Korovkin's theorem to trigonometric polynomial
approximation; Fejér's theorem 104
2. 15 The Stone-WeierstraB theorem 109
2. 16 Convex sets . . . 1 14
2.17 Convex functions 1 18
3 Banach Spaces 123
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.1 Banach spaces; first properties . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.2 First examples of Banach spaces; the spaces C(K; Y) with K compact and Y
complete, and C(X; Y) with Y complete . . . . . . . . . . . . . . . . . . . . 130
3.3 lntegral of a continuous function of a real variable with values in a Banach
space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.4 Further examples of Banach spaces: the spaces fP and V(S1), 1 ::::; p::::; oo . . 135
3.5 Dual of a normed vector space; first examples; F. Riesz representation theorem
in V(n), 1 ::::; p < oo . . . . 138
3.6 Series in Banach spaces . . . . . . . . . . . . . . . . . . . . . . . . . 148
3. 7 Banach fixed point theorem . . . . . . . . . . . . . . . . . . . . . . . 152
3.8 Application of Banach fixed point theorem: Existence of solutions to
nonlinear ordinary differential equations; Cauchy-Lipschitz theorem;
the pendulum equation . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.9 Application of Banach fixed point theorem: Existence of solutions to nonlinear
two-point boundary value problems . . . . . . . . . . . . . . . . . . . . . . . . 161
3.10 Ascoli-Arzelà's theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
3. 1 1 Application of Ascoli-Arzelà's theorem: Existence of solutions to nonlinear
ordinary differential equations; Cauchy-Peano theorem; Euler's method 169
4 Inner-Product Spaces and Hilbert Spaces 173
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
4. 1 Inner-product spaces and Hilbert spaces; first properties;
Cauchy-Schwarz-Bunyakovskïi' inequality; parallelogram law . . . . . . 174
4.2 First examples of inner-product spaces and Hilbert spaces; the spaces f2
and L2 (S1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
4.3 The projection theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.4 Application of the projection theorem: Least-squares solution of a linear
system . . . . . . . . . . . . . . . . 193
4.5 Orthogonality; direct sum theorem . . . . . . . . . . . . . . . . . . . . . . 195
 Contents ix

4.6 F. Riesz representation theorem in a Hilbert space . . . . . . . . . . . . 197

4. 7 First applications of the F. Riesz representation theorem: Hahn-Banach
theorem in a Hilbert space; adjoint operators; reproducing kernels 199
4.8 Maximal orthonormal families in an inner-product space . . . . . . . . . 205
4.9 Hilbert bases and Fourier series in a Hilbert space . . . . . . . . . . . . 213
4. 10 Eigenvalues and eigenvectors of self-adjoint operators in inner-product spaces 219
4. 1 1 The spectral theorem for compact self-adjoint operators . . . . . . . . . . . . 221
5 The "Great Theorems" of Linear Functional Analysis 231
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5. 1 Baire's theorem; a first application: Noncompleteness of the space of all
polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
5.2 Application of Baire's theorem: Existence of nowhere differentiable continuous
fonctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
5.3 Banach-Steinhaus theorem, alias the u liiform boundedness principle;
application to numerical quadrature formulas . . . . . . . . . . . . . 238
5.4 Application of the Banach-Steinhaus theorem: Divergence of Lagrange
interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
5.5 Application of the Banach-Steinhaus theorem: Divergence of Fourier series . 252
5.6 Banach open mapping theorem; a first application: Well-posedness of two-
point boundary value problems . . . . . . . . . . . . . . . . . . . . . . . . . . 255
5.7 Banach closed graph theorem; a first application: Hellinger-Toeplitz theorem 259
5.8 The Hahn-Banach theorem in a vector space . . . . . . . . . . . . . . . . 261
5.9 The Hahn-Banach theorem in a normed vector space; first consequences . 264
5.10 Geometric forms of the Hahn-Banach theorem; separation of convex sets . 272
5. 1 1 Dual operators; Banach closed range theorem 277
5.12 Weak convergence and weak * convergence . . . . . . . . 286
5. 13 Banach-Saks-Mazur theorem . . . . . . . . . . . . . . . 294
5. 14 Refiexive spaces; the Banach-Eberlein-Smulian theorem 297
6 Linear Partial Differential Equations 305
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
6.1 Quadratic minimization problems; variational equations and variational
inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
6.2 The Lax-Milgram lemma . . . . . . . . . . . . . . . . . . . . . . . . . . 310
6.3 Weak partial derivatives in Lf0c(n); a brief incursion into distribution theory 312
6.4 Hypoellipticity of � . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
6.5 The Sobolev spaces Wm•P (fl) and Hm (n): First properties . . . . . . . . 326
6.6 The Sobolev spaces wm·P(fl) and Hm (n) with n a domain; imbedding
theorems, traces, Green's formulas . . . . . . . . . . . . . . . . . . . . 331
6. 7 Examples of second-order linear elliptic boundary value problems; the
membrane problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
6.8 Examples of fourth-order linear boundary value problems; the biharmonic
and plate problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
X Contents

6.9 Examples of nonlinear boundary value problems associated with variational

inequalities; obstacle problems . . . . . . . . . . . . . . 363
6.10 Eigenvalue problems for second-order elliptic operators . . . . . . . . . . . . . 369
6. 11 The spaces w-m,q (f!) and H-m (n); J.L. Lions lemma . . . . . . . . . . . . . 377
6.12 The Babuska-Brezzi inf-sup theorem; application to constrained quadratic
minimization problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
6.13 Application of the Babuska-Brezzi inf-sup theorem: Primai, mixed, and dual
formulations of variational problems . . . . . . . . . . . . . . . . . . . . . . . 388
6.14 Application of the Babuska-Brezzi inf-sup theorem and of J.L. Lions lemma:
The Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
6.15 A second application of J.L. Lions lemma: Korn's inequality . . . . . . . . . . 403
6.16 Application of Korn 's inequality: The equations of three-dimensional linearized
elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
6.17 The classical Poincaré lemma and its weak version as an application of
J.L. Lions lemma and of the hypoellipticity of� . . . . . . . . . . . . . 419
6.18 Application of Poincaré's lemma: The classical and weak Saint-Venant lemmas;
the Cesàro-Volterra path integral formula . . . . . . . . . . . . 429
6.19 Another application of J.L. Lions lemma: The Donati lemmas . 437
6.20 Pfaff systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
7 Differential Calculus in Normed Vector Spaces 451
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
7.1 The Fréchet derivative; the chain rule; the Piola identity; application to
extrema of real-valued functions . . . . . . . . . . . . . . . . . . . . . . . 452
7.2 The mean value theorem in a normed vector space; first applications . . 465
7.3 Application of the mean value theorem: Differentiability of the limit of a
sequence of differentiable functions . . . . . . . . . . . . . . . . . . . . . . 469
7.4 Application of the mean value theorem: Differentiability of a function defined
by an integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . . 472
7.5 Application of the mean value theorem: Sard's theorem . . . . . . . . . . . . 474
7.6 A mean value theorem for functions of class c i with values in a Banach space 477
7.7 Newton's method for solving nonlinear equations; the Newton-Kantorovich
theorem in a Banach space . . . . . . . . . . . . . . . . . . . . . 478
7.8 Higher order derivatives; Schwarz lemma . . . . . . . . . . . . . . . . . . . 500
7.9 Taylor formulas; application to extrema of real-valued functions . . . . . . 507
7.10 Application: Maximum principle for second-order linear elliptic operators 513
7.11 Application: Lagrange interpolation in Rn and multipoint Taylor formulas . 522
7. 12 Convex functions and differentiability; application to extrema of real-valued
functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 540
7.13 The implicit function theorem; first application: Class C00 of the mapping
A -+ A - i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548
7.14 The local inversion theorem; the invariance of domain theorem for mappings
of class c i in Banach spaces; class C00 of the mapping A -+ A i/2 . 554
7.15 Constrained extrema of real-valued functions; Lagrange multipliers 560
7.16 Lagrangians and saddle-points; primai and dual problems . . . . . 565
 Contents xi

8 Differential Geometry in ]Rn 575

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575
8. 1 Curvilinear coordinates in an open subset of JRn . . .. . .
. • 576
8.2 Metric tensor; volumes and lengths in curvilinear coordinates 578
8.3 Covariant derivative of a vector field . . . . . . . . . . . . . . 583
8.4 Tensors - a brief introduction . . . . . . . . . . . . . . . . . . 588
8.5 Necessary conditions satisfied by the metric tensor; the Riemann curvature
tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595
8.6 Existence of an immersion on an open subset of JRn with a prescribed metric
tensor; the fondamental theorem of Riemannian geometry . . . . . . . . 598
8.7 Uniqueness up to isometries of immersions with the same metric tensor;
the rigidity theorem for an open subset of JRn . . . . . . . . . . . . . . . 608
8.8 Curvilinear coordinates on a surface in JR3 . . . . . . . . . . . . . . . . . 613
8.9 First fondamental form of a surface; areas, lengths, and angles on a surface 614
8.10 Isometric, equiareal, and conformai surfaces . . . . . . . . . . 622
8.11 Second fondamental form of a surface; curvature on a surface 624
8. 12 Principal curvatures; Gaussian curvature . . . . . . . . . . . . 629
8. 13 Covariant derivatives of a vector field defined on a surface; the Gaul3 and
Weingarten formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
8. 14 Necessary conditions satisfied by the first and second fondamental forms: The
GauB and Codazzi-Mainardi equations . . . . . . . . . . . . . . . . . . . . 640
8.15 GauB Theorema Egregium; application to cartography . . . . . . . . . . . . 643
8. 16 Existence of a surface with prescribed first and second fondamental forms;
the fondamental theorem of surface theory . . . . . . . . . . . . . . . . . . . . 646
8.17 Uniqueness of surfaces with the same fondamental forms; the rigidity theorem
for surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
9 The "Great Theorems" of Nonlinear Functional Analysis 657
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
9.1 Nonlinear partial differential equations as the Euler-Lagrange equations
associated with the minimization of a fonctional . . . . . . . . . . . . . 658
9.2 Convex fonctions and sequentially lower semicontinuous fonctions with values
in JR U { OO } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664
9.3 Existence of minimizers for coercive and sequentially weakly lower
semicontinuous functionals . . . . . . . . . 671
9.4 Application to the von Karman equations 674
9.5 Existence of minimizers in W 1 ·P(f!) . . . . 683
9.6 Application to the p-Laplace operator . . 691
9.7 Polyconvexity; compensated compactness; John Ball's existence theorem in
nonlinear elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
9.8 Ekeland's variational principle; existence of minimizers for fonctionals that
satisfy the Palais-Smale condition . . . . . . . . . . . . . . . . . . . . . . . . 711
9.9 Brouwer's fixed point theorem - a first proof . . . . . . . . . . . . . . . . . . 718
9. 10 Application of Brouwer's theorem to the von Karman equations, by means of
the Galerkin method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726
xii Contents

9.11 Application of Brouwer's theorem to the Navier-Stokes equations, by means

of the Galerkin method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 728
9.12 Schauder's fixed point theorem; Schafer's fixed point theorem; Leray-Schauder
fixed point theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734
9.13 Monotone operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
9. 14 The Minty-Browder theorem for monotone operators; application to the
p-Laplace operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
9.15 The Brouwer topological degree in !Rn : Definition and properties . . . . 748
9.16 Brouwer's fixed point theorem- a second proof- and the hairy ball theorem 764
9. 17 Borsuk's and Borsuk-Ulam theorems; Brouwer's invariance of domain
theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
Bibliographical Notes 111

Bibliography 781
Main Notations 807
Index 815
PREFACE

Why write another textbook on functional analysis and its applications, since there are
already many excellent textbooks around ?
Apart from the persona! pleasure that such an exercise provides to an author, there are
other reasons: One, which perhaps constitutes the main originality of this text, was to assem­
ble in a single volume the most basic theorems of linear and of nonlinear functional analysis;
another reason was to simultaneously illustrate the wide applicability of these theorems by
treating an abundance of applications.
Applications to linear and nonlinear partial differential equations treated here include
Korn's inequality and existence theorems in linear elasticity, obstacle problems, the Babuska­
Brezzi inf-sup condition, existence theorems for the Stokes and Navier-Stokes equations of
fluid mechanics, existence theorems for the von K arman equations of a nonlinearly elastic
plate, and John Ball's existence theorem in nonlinear elasticity. A variety of other appli­
cations deals with selected topics from numerical analysis and optimization theory, such as
approximation theory, error estimates for polynomial interpolation, numerical linear algebra,
basic algorithms of optimization, Newton's method, or finite difference methods.
A special effort has been made to enhance the pedagogical appeal of the book. After
Chapter 1, which is essentially a review of results from real analysis and the theory of functions
that will be used in the text, self-contained and complete proofs of most of the theorems are
provided. 1 These include proofs that are not always easy to locate in the literature, or
difficult to reconstitute without an extended knowledge of collateral topics; for instance, self­
contained proofs are given of the Poincaré lemma, of the hypoellipticity of the Laplacian,
of the existence theorem for Pfaff systems, or of the fundamental theorem of surface theory.
Numerous figures and problems (almost 400) have also been included. Historical notes and
original references (at least those that 1 have been able to trace with a reasonable assurance
of veracity) have also been included2 (mostly as footnotes), so as to provide an idea of the
genesis of some important results.
lt is my belief that this book contains most of the core topics from functional analysis
that any analyst interested in linear and nonlinear applications should have encountered at
least once in his or her life. More specifically, linear functional analysis and its applications
are the subjects of Chapters 2-6, while nonlinear functional analysis and its applications are
the subjects of Chapters 7-9.
Of course, choices had to be made, in particular so as to keep the length of the book
within reasonable limits. For instance, more specialized topics, such as the Fourier transform,
1 The symbol � to the left of a theorem indicates one without proof.
2 With the full knowledge that doing so sometimes constitutes a perilous exercise . . . '

xiii
xiv Preface

wavelets, spectral theory (save for compact self-adjoint operators), or time-dependent partial
differential equations, are not treated.
Severa! one-semester courses, at the last-year undergraduate or graduate levels, can be
taught from this book, such as "Linear Functional Analysis,'' "Linear and Nonlinear Bound­
ary Value Problems,'' "Differential Calculus and Applications,'' "Introduction to Differential
Geometry,'' "Nonlinear Functional Analysis,'' or "Mathematical Elasticity and Fluid Me­
chanics." In this respect, it should be easy for an instructor to identify from the table of
contents those parts of the book that should be used for any such course. Indeed, 1 had the
pleasure of teaching such courses, primarily at the University Pierre et Marie Curie and at
City University of Hong Kong, but also at the University of Texas at Austin, at Cornell Uni­
versity, at Fudan University, at the University of Stuttgart, at !'Ecole Polytechnique Fédérale
de Lausanne, at the ETH-Zürich, and at the University of Zürich.
The main prerequisites are a reasonable acquaintance with real analysis, i.e., elementary
topology (such as continuity and compactness), the basic properties of metric spaces and
Lebesgue integration, and the theory of real-valued functions of one or several real variables.
For the reader's convenience, the basic definitions and theorems from these subjects needed
in this book are assembled without proofs in the first chapter.
During the writing of this book, 1 have greatly benefitted from the comments of Liliana
Gratie, George Dinca, Cristine! Mardare, Sorin Mardare, and Pascal Azerad, who were kind
enough to very carefully read most of the chapters and to suggest numerous significant im­
provements. Bernard Dacorogna and Vicentiu Radulescu have also provided me with much
precious advice. To al! of them, my most sincere thanks!
My gratitude is also due to Douglas N. Arnold for his early-and strong-support of
the project, and also to Elizabeth Greenspan, Gina Rinelli, and Lisa Briggeman from the
Editorial Office of SIAM, with whom it is a real pleasure to cooperate.
Last but not least, 1 express my deep gratitude and my lasting admiration to my " mathe­
matical heroes" Laurent Schwartz, Richard S. Varga, Jacques-Louis Lions, and Robert Dau­
tray, whose teaching and advice over the years have been invaluable.
1 am perfectly aware that, most likely, there are still at places inadequacies, _ inconsisten­
cies of notations, inadvertently omitted references, or inappropriate attributions of original
results. But any adventure (mathematical or otherwise) must co rne to an end, even if its
main protagonist is not fully satisfied with it. Or equivalently, as Paul Halmos said in a
much better way, in a pure gem of a paper3 that any mathematician, pure or applied, should
read and reread (1 paraphrase him): "The last step for most authors is to stop w riting. That's
hard."
This is one more reason why 1 welcome in advance al! comments, remarks, criticisms, etc.,
which should be sent to mapgc©ci tyu. edu. hk, and-who knows-could be used in a second
edition.
Hong Kong, November 2012 Philippe G. Ciarlet

3 P.R. HALMOS [1970] : How to write mathematics, L 'Enseignement Mathématique 16, 123-152.
CHAPTER 1

REAL ANALYSIS AND THEORY OF FUNCTIONS :

A QUICK REVIEW

Introduction
This first chapter constitutes a quick review of real analysis, which traditionally comprises:
set theory, the axiom of choice, and the construction of the sets lR and JRn ; the basic properties
of topological and metric spaces, such as those related to the notions of continuity, compact­
ness, completeness, connectedness, and simple-connectedness; the Tietze-Urysohn extension
theorem (a crucial use of which will be made at several places in Chapter 9); and the con­
struction and the main properties of the Lebesgue measure and Lebesgue integral in lRn : the
Radon-Nikodym theorem; Fatou's lemma; the Beppo Levi monotone convergence theorem;
the Lebesgue dominated convergence theorem; Tonelli's and Fubini's theorems; volumes,
areas, and lengths in lRn ; and the change of variable formula in multiple integrals.
This first chapter also includes a quick review of some aspects of the theory of real-valued
functions of several real variables. More specifically, basic function spaces, such as cm (n) and
cm (n), where n is an open subset of ]Rn , are introduced (other function spaces, such as the
Lebesgue spaces V(n), or the Sobolev spaces Hm (n) and wm,P(fl), will be introduced and
studied in later chapters). Domains in lRn , that is, open subsets n c JRn that are bounded,
connected, and have a Lipschitz-continuous boundary, with n being locally on the same side
of the boundary, are then singled out among ail open subsets of JRn , one reason being that
they insure the validity of the fundamental Green's formula for functions in the space C1(D)
(this Green's formula over domains in ]Rn will be later extended to functions in the Sobolev
spaces w m,P(fl); cf. Chapter 6).
Otherwise the reader is assumed to be already familiar with linear algebra in finite­
dimensional spaces (bases, linear dependence, matrices, determinants, etc.), as well as with
basic notions of differential calculus for real-valued functions of several real variables (partial
derivatives, Taylor formulas, etc.).
The objective of this chapter is essentially to state in the form of theorems the various
results from these topics that will be used throughout the book, and to list the various
notations and definitions needed for this purpose.
No proofs are given and no exercises are provided, since the reader is assumed to be
already reasonably familiar with these results. References are provided in the Bibliographical
Notes.

1
2 Real Analysis and 'l'heory of F'unctions: A Quick Review [Ch. 1

1.1 Sets
The most commonly adopted set theory is the Zermelo-Fraenkel set theory. lt starts
with six axioms, which will not be explicitly stated here; only their consequences will be
described.
In this respect, notions such as those of element or set, or notations such as "=, -:/:-, E, <f_,"
or words or an assemblage of words such as "implies," "for ail,'' "there exists," "such that,"
etc., are not defined; these are assumed instead to be given their intuitive or usual sense
(whatever that means).
Let X be a set. The notation A C X or X :J A means that the set A is a subset of X,
i.e., that x E A implies x E X. The notation A �X or X � A means that A is a proper
subset of X, i.e., that A C X but A -:/:- X.
Let X be a set. There exists a set, denoted P(X), whose elements are ail the subsets of X.
The set P(X) comprises in particular the empty set 0 (whose existence is a consequence of
the axioms) and the set X itself. If X-:/:- 0 and x E X , the subset of X whose only element
is x is denoted { x }.
Let X be a set. If A C X, the complement of A relative to X, or simply the
complement of A if there is no ambiguity as to what is the set X, is the subset of X
defined by
X - A := {x E X; x <f. A}.
If A and B are subsets of a set X, their union and intersection are respectively denoted,
and defined, by
A U B = {x E X; x E A or x E B},
A n B = {x E X ; X E A and X E B}.
The sets A and B are disjoint if A n B = 0.
Let X and Y be two sets. The set
X x Y := {(x, y); x E X and y E Y},
whose elements are ail the ordered pairs (x, y) with x E X and y E Y, is called the product
of X and Y.
A relation 'R on a set X is any subset 'R of the product X x X, i.e., 'R consists of specific
ordered pairs (x, y), with x E X and y E X.
An equivalence relation on X is a relation 'R that satisfies the following properties,
where the notation x ,...., y means that (x, y) E 'R:
reftexivity : X X for ail X E X,
rv

symmetry : X y impJies y X,
rv rv

transitivity : X y and y
rv Z
implies X
rv rv z.

Equivalently, (x, x) E 'R for ail x E X; if (x, y) E 'R, then (y, x) E 'R; if (x, y) E 'R and
(y, ) E 'R, then (x, ) E 'R.
z z

Given an element x in a set X endowed with an equivalence relation 'R, the equivalence
class of x modulo 'Tl is the subset of X defined by
x := {y E X; y x} .
rv
Sect. 1.2] Mappings 3

Two equivalence classes of elements of X are thus either identical or disjoint.

The quotient set X/R is the subset of P(X) consisting of all equivalence classes modulo
R of elements of X.
All the above definitions and properties rely only on the first five axioms of the Zermel o-­
Fraenkel set theory. The sixth one, called the axiom of infinity, is of crucial importance, since
it implies both the existence of the set
N : = {O, 1 , 2, . . . }
formed by all natural integers: 0, 1, 2, . . . , and the possibility of proving statements by
recursion: To prove that a property holds for all n E N, it suffi.ces to prove that it holds for
n = 0 and that, if it holds for some n E N, then it also holds for n + 1. Specific subsets of
N are designated by self-explanatory notations, such as {0, 1, . . . , n}, {j E N; 1:::;; j:::;; n} =
{ 1 , 2, . . . , n}, {n E N; n � no} = {no, no + 1, . . . }, etc.

1.2 Mappings
Let X and Y be two nonempty sets. A mapping, or a fonction, of X into Y is a subset
f of the product X x Y such that, for each x E X, there exists one and only one element
y E Y such that (x, y) belongs to f. This element y is then denoted either f(x) or Yx· When
the notation Yx is used, x is called an index.
The notations
f : X --+ Y or X � Y,
mean that X and Y are two sets and that f is a mapping of X into Y. The notation
f : X E X --+ f(x) E Y
with an explicit expression for f(x) is used to define a mapping f.
Let X be a set. The mapping x E X --+ x E X is called the identity mapping of X; it
is denoted id or idx , or I or lx if X is a vector space.
If A is a subset of a set X, the function XA : X --+ lR defined by
XA(x) := 1 if x E A and XA(x) : = 0 if X <{. A
is called the characteristic fonction of A.
Let f : X --+ Y be a mapping. The direct image under f of a subset A of X is the
subset f (A) of Y defined by
f(A) := {y E Y; there exists x E A such that y = f(x)}.
The inverse image under f of a subset B of Y is the subset of X defined by
r 1 (B){x E X; f(x) E B}.
:=

If b E Y, the (improper but convenient) notation f - 1 (b) will be blithely used to designate
the inverse image r 1 ( {b} ).
4 Real Analysis and Theory of F'unct ions: A Quick Review [Ch. 1

Care should be exercised when using notations such as f(A ) and f - 1 (B): The notation f
designates a mapping of X into Y, not a mapping of P(X) into P(Y) (as the notation f (A)
tends to suggest). Likewise, the notation f- 1 designates the inverse mapping of f when it
exists (see below), in which case f- 1 is a mapping of Y onto X, not a mapping of P(Y) into
P(X) (as the notation r 1 (B) tends to suggest).
The inverse image "preserves ail the set operations," in that it satisfies
r 1 (B) c r 1 (B) if n c 8,
1
r (B u 8) = r 1 (B) u f - 1 (8),
r 1 (B n B) = r 1 (B) n f - 1 (B),
r 1 (Y - B) = X - r 1 (B).
By contrast, the direct image only satisfies
f (A) c f (A) if A c A,
f (A u A) = f (A) u f (A),
f(A n A) c f(A ) n f( A).
A mapping f : X --+ Y is surjective, or onto, or is a surjection, if for each y E Y, there
exists at least one element x E X such that y = f(x).
A mapping f : X --+ Y is injective, or one-to-one, or is an injection, if for each y E Y,
there exists at most one element x E X such that y = f(x). If X is a subset of Y, the
t:
mapping X --+ Y defined by t(x) = x for ail x E X is called the canonical injection from
X into Y.
A mapping f : X --+ Y is bijective, or is one-to-one and onto, or is a bijection, if it
is both surjective and injective. In this case, for each y E Y, there thus exists one and only
one element x E X such that y = f(x), and the mapping f- 1 : y E Y --+ x E X defined in
this fashion is the inverse mapping of f.
Let f : X --+ Y be a mapping and let A be a subset of X. The mapping f l A : A --+ Y
defined by f l A(x) := f(x) for ail x E A is the restriction of f to A.
Let g : A --+ Y be a mapping, where A is a subset of X. A mapping f : X --+ Y is an
extension of g if f l A = g.
Let f : X --+ Y and g : Y --+ Z be two mappings. The mapping h : X --+ Z defined by
h(x) = g(f(x)) for ail x E X is called the composition of f and g. It is denoted h = g o f,
or h = gf.
a
Let f : X x Y --+ Z be a mapping and let be a point in the set X. The mapping
f ( , · ) : Y --+ Z defined by
a

f(a, ·) : y E Y -+ f(a, y) E Z
is a partial mapping. Given a point b E Y, a similar definition holds for the partial mapping
f(-, b) : X --+ Z.
Given a mapping f : (x, y) E X x Y --+ f(x, y) E Z, the elements x E X, resp. y E Y, are
sometimes called first argument s of f, resp. second argument s of f.
Sect. 1.3] The axiom of choice and Zorn's lemma 5

1 .3 The axiom of choice and Zorn 's lemma

Let I =/= 0 and X =/= 0 be two sets. A family of elements of X indexed by I is a mapping
as
f : I ---+ X defined f : i E I ---+ Xi E X, i.e., the elements of the set I are regarded as

indices. Such a family is then denoted by

(xi) iEI,
or simply (xi) if the definition of the set I is unambiguous. Naturally, a family (xi)i eI of
elements of X is to be carefully distinguished from the subset Uie J {xi} of X (which for
a a
instance consists of a single point E X if Xi = for ail i E /).
A subfamily (xi)ie J of the family (xi)ieI is a mapping g : J ---+ X such that J CI and
!I J = g.
If I = {1, . . . , n } for some n ?: 1, the family (xi)ieI is called an n-tuple and is denoted
(x3)1]= l or (x 1 , . . . , xn ) ·
If I = N, the family (xi)ieI is called a sequence and is denoted
(xn)�=O> or (xo, x 1 , . . . , Xn , . . . ), or (xn )n;;::o , or simply (xn ) ·
Other self-explanatory notations are also used, such as

(xn ) n>o, (xn )�=no or (xn )n;?:no if I = {no , no + 1, . . . }, etc.

A subsequence of a sequence is a subfamily that is also a sequence. For instance, given
any st rict ly increasing mapping a : N ---+ N (i.e., such that a (n ) < a (n + 1) for ail n E N) ,
the sequence (xu(n) )�=O is a subsequence of the sequence (xn )�=O · This not at ion will be often
used t a denot e subsequences in t he sequel.
Let I and X be two sets. A family (Ai)ie I of subsets of X indexed by I is a family
i E I ---+ Ai E P(X) (i.e., the mapping appearing in the definition of a family now takes its
values in the set P(X), instead of the set X for a family of elements of X).
Given a family (Ai)iEJ of subsets of a set X ' the union ui EJ Ai and intersection ni EJ Ai
are respectively defined by
LJ Ai := {x E X; there exists i E I such that x E Ai},
iE J
n Ai := {x E X; x E Ai for ail i E /}.
iE J
Other self-explanatory notations are also used, such as

n n OO OO

LJ Ai and n Ai if / = {O, 1, . . . , n}, LJ Ai and n Ai if I = N, etc.

i=O i=O i=O i =O
Given a family (Ai)ieI of subsets of a set X, the disjoint union Ue i Ai is defined by
lJ Ai := LJ{(x, i); x E Ai}.
iE J iE J
 6 Real Analysis and 'l'heory of Punctions: A Quick Review [Ch. 1

The disjoint union uiE/ Âi is thus a subset of the product (LJ iE/ Ai) X !, itself a subset of the
product X x !.
If A and B are two subsets of a set X, the union AUB and the intersection An B coïncide
with the union LJi e / Ai and the intersection nie / Ai with Ai := A, A2 := B, and l := {1, 2}.
The following identities are constantly used:
( )
AU nAi = n(AUAi) and An LJ Ai . = LJ(A n Ai),
iE/ iE/ iE/
( ) iE/
X - LJ Ai = n(x - Ai) and X - n Ai = LJ(X - Ai),
iE/ iE/ iE/ iE/
the last two identities constituting de Morgan's laws.
Let (Ai)i E I be a family of subsets of a set X where l -:j; 0. Then the product Ilie / Ai
is, by definition, the set of ail mappings f : I --t X such that j(i) E Ai for ail i E /. Any
such mapping f is called a choice function, as it asserts that it is possible to "choose" one
element f (i) E Ai for each i E /.
Whereas the definitions of the union uiE/ Ai and intersection niE/ Ai as subsets of the set
X do not pose specific difficulties, the definition of the product rriE/ Ai raises an immediate
question, as nothing guarantees the existence of at least one such choice function f. This
is why the following axiom, called the axiom of choice, was introduced in 1904 by Ernest
Zermelo:
Axiom of choice Let (Ai)iE I be a family of subsets of a set. If I # 0 and Ai # 0 for
all i E /, then Ilie I Ai -:j; 0. 0

In 1963, Paul J. Cohen established in a landmark paper 1 that the axiom of choice is
independent of the six axioms of the Zermelo-Fraenkel set theory.
Other notations for the product IliE/ Ai are also used, viz.,
n
Ai x A2 if / = { 1 , 2} ( as in Section 1.1), Il Ai if I = {1, 2, . . . , ri},
i= l
An if I = {1, 2, . . . , n} and Ai = A for all 1 � i � n, etc.
The element of the product Ilie / Ai corresponding to a choice function f will be denoted
x = (xi)ie l i where Xi := f(i), each element i E I being thus regarded as an index ( Section
1.2). Each element Xi E Ai, i E /, is called the ith coordinate of x. This notation is
coherent with those used for a finite sequence (xi) f= 1 or for a sequence (xi)�0 of scalars,
which are simply special cases of elements in a product, viz., ocn , or Il�o Ai with Ai = IK for
ail i E N, respectively.
The axiom of choice is in fact often used in disguise in proofs, by means of one of its
different, but equivalent, forms, each one of which then taking the form of a theorem. Zorn's
lemma (Theorem 1.3-1 below) provides such an example. Note, however, that while the
statement of the axiom of choice is intuitively clear, the same cannot be said of Zorn's
lemma.
1 P.J. COHEN [1963]: The independence of the continuum hypothesis, Proceedings of the National Academy
of Sciences, USA 50, 1 143-1 148.
Sect. 1.3] The axiom of choice and Zom's lemma 7

In order to state this lemma, we need several definitions.

A set X is partially ordered by a relation 'R (Section 1.1), or equivalently, a relation
'R is a partial ordering on X, if 'R satisfies the following properties, where the notation
x -<-y means that (x, y) E 'R:

reflexivity : x-<-x for all x E X,

antisymmetry : x-<-y and y-<-x implies x = y,
transitivity : z
x-<-y and y -<- implies x -<- z.

Equivalently, (x, x ) E 'R for all x E X; if (x, y ) E 'R and (y, x ) E 'R, then x = y; if (x, y) E 'R
and (y, z ) E 'R, then (x, z) E 'R.
For instance, the relation "x = (xi)� 1 -<- y = (Yi)� 1 if and only if Xi $ Yi for all
1 $ i $ n" defines a partial ordering on the set IR.n ; the relation "A -<- B if and only if
A c B" defines a partial ordering on the set P(X) formed by all subsets of a set X.
The notation y� x means that x -<-y. The notation x -<y, or y>- x, means that x-<-y
and x f. y.
A subset A of a partially ordered set X is totally ordered if any two elements a E A
and b E A are comparable, in the sense that either a -<- b or b-<-a (if a-<-b and b-<-a, then
1
a = b). Clearly, if such a set A is finite, i.e., of the form A = u:, { ai} , there exists 1 $ io $ m
such that ai-<-aio for all l $ i $ m. For instance, if a fini te subset {Ai , A2 , . . . , Am } of P(X)
is totally ordered for the inclusion, then there exists 1 $ mo $ m such that Ai c A rno for all
1 $ i $ m, so that A rno = LJ:, 1 Ai ; this observation is often used.
Let A be a subset of a partially ordered set X. Then an element b E X is an upper
bound for A if a -<-b for all a E A. Note that all elements of A must then be comparable
to b, but that b need not belong to A .
Let X be a partially ordered set. An element m E X is maximal if any element x E X
that is comparable to m satisfies x -<- m; or equivalently, if x E X satisfies m -<- x, then
x = m. Note that m need not be comparable to all elements x E X.
Then the following result is equivalent to the axiom of choice.
Theorem 1.3-1 ( Zorn's lemma) Let X be a nonempty partially ordered set with the
property that every totally ordered subset has an upper bound in X. Then X has at least one
maximal element. D

Zorn's lemma constitutes an extremely powerful tool for establishing the existence of
certain mathematical objects. For instance, it is used for proving that there exist non­
Lebesgue measurable subsets of lR (Section 1 . 14); for proving that any vector space possesses
a Hamel basis (Theorem 2. 1-1); for proving that any vector space can be normed (Theorem
2.2-8); for proving that any inner-product space possesses a maximal orthonormal family
(Theorem 4.8-4); or for proving the fundamental Hahn-Banach theorems, which assert the
existence of extensions of linear functionals (Theorems 5.8-1 and 5.9-1).
Note that, each time that Zorn's lemma is applied in a set X, particular care should be
given to verifying that X is nonempty.
8 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

1 .4 Construction of the sets IR and C

The set .N, whose existence is implied by the axiom of infinity (Section 1.1), is used for
constructing the set
z := { . . . ' - 2, - 1, 0, 1, 2, . . . }

of ail integers, as the quotient set (.N x .N) /R, where R denotes the equivalence relation on
.N x .N defined by
((m, n), (m ', n') ) E R if and only if m + n' = m' + n.
Equipped with the addition and multiplication, Z becomes a commutative ring that is
also an integral domain (i.e., if mp = mq and m f:. 0, then p = q) and totally ordered by :<::;
(total ordering is defined in Section 1.3).
The set Z is then used for constructing the set Q of ail rational numbers, as the set
formed by ail the equivalence classes modulo the following equivalence relation in the set
Z x (Z - { 0}) : (m, n) ,...., (p, q) if and only if mq = np. Equipped with the operations + and
x , the set Q then becomes a totally ordered commutative field. The field Q is Archimedean,
that is, given any rational numbers r > 0 and s > 0, there exists an integer n ;::: 0 such that
nr > s. The absolute value of r E Q is defined by l r l := r if r ;::: 0, or by l r l = - r if r :<::; O.
A sequence (rn )�= l of rational numbers is said to be a Cauchy sequence if, given any
ê > 0, there exists an integer mo = mo (ê ) ;::: 1 such that l rm - rn l $ ê for all m, n ;::: mo.
The set Q is then used for constructing the set lR of ail real numbers, as the set formed
by ail equivalence classes modulo the following equivalence relation R in the set formed by
l•
all Cauchy sequences of rational numbers: ((rn ) �= (sn )�= 1 ) E R if, given any ê > 0, there
exists an integer no = no (ê ) ;::: 1 such that l rn - sn l :<::; ê for ail n ;::: no. Equipped with the
operations + and x and the total ordering :<::;, the set lR is also a totally ordered, A rchimedean,
commutative field, and the absolute value of x E R is likewise defined by lx l := x if x ;::: 0, or
by l x l := -x if x :<::; O. Alternatively, the set R may be also constructed by means of Dedekind
cuts.
The set { -oo }UR U { oo} of extended real numbers is defined by adjoining to the set
R two elements, denoted -oo and oo, which obey the usual rules; for instance, -oo < x for
ail x E R, x + oo = oo for ail x E R, etc. Naturally, -oo + oo is not defined.
Finally, the commutative field C of complex numbers is constructed in the usual way
from the set R If z E C, then Re z and lm z respectively denote the real and imaginary
parts of z; in other words, z = Re z + i lm z. The absolute value of z E C is defined by
l z l := J I Re z l 2 + 1 lm z l 2 .
lt is often convenient to designate by the same letter ][{ either the field R or the field C,
in which case the elements of the field ][{ are called scalars.
Once sets R and C have been constructed as outlined above, various properties of R
and C can then be established. The next theorem gathers the most important ones. Cauchy
sequences of real or complex numbers are defined like Cauchy sequences of rational numbers.
Theorem 1.4-1 (a) The set R, resp. C, is complete, i. e., any Cauchy sequence (xn )�=l
of real, resp. complex, numbers converges to a real, resp. complex, number; this means that
there exists x E R, resp. x E C, and, given any ê > 0, there exists an integer no = no (ê ) ;::: 1,
Sect. 1.5] Cardinal numbers; finite and infinite sets 9

such that
l xn - x i :::; êfor all n 2". no.
(b) Bolzano-WeierstraB property for lR and C: Any sequence (xn )�= l of real, resp.
complex, numbers that is bounded, i. e., such that there exists M E lR with the property that
l xn l :::; M for all n 2". 1, contains a convergent subsequence.
(c) Let A be a nonempty subset of lR that has an upper bound in lR (Section 1.3) . Then
there exists a E lR that is the least upper bound, or supremum, of A; this means that a
is an upper bound for A and any upper bound b E lR for A necessarily satisfies a :::; b.
Likewise, any nonempty subset of IR that has a Lower bound in lR admits a greatest lower
bound, or infimum, in IR. D

That a sequence (xn )�= l of real, or complex, numbers converges to x (according to the
definition in Theorem 1.4-l(a)) is also denoted:
-too Xn , or Xn n--+
x = nlim -too x , or Xn -t x as n -t oo.
A mapping of a set X ·into IR, resp. C, is called a real-valued, resp. complex-valued,
function.
A mapping of a set X into !Rn , resp. the set of all m x n matrices, is called a vector field,
resp. a matrix field.

1.5 Cardinal numbers; finite and infinite sets

It is immediately seen that the relation "there exists a bijection of A onto B, '' where A and B
are subsets of a set X, defines an equivalence relation R (Section 1.1) on the set P(X). The
elements of the quotient set P(X)/R C P(P(X)) are then called the cardinal numbers of
the subsets of X. If A is a subset of X, its cardinal number, denoted
card A,
is thus the equivalence class of A modulo R, and as such, is an element of the set P(P(X)).
Remarkably, the set P(X)/R can be totally ordered (the definition of a totally ordered
set is given in Section 1.3), according to the following fundamental theorem:
Theorem 1.5-1 The set P(X)/R of all the cardinal numbers of the subsets of a set X
is totally ordered by the relation R, where (card A, card B) E Rmeans that there exists an
injection of A into B.
Equivalently, (card A, card B) E Rif and only if there exists a surjection of B onto A.
D

Let us give some indications about the proof of this result (the proof may seem innocuous
at first glance, but is in effect anything but trivial). First, it is clear that the definition of the
relation Ris unambiguous. For, if card A = card  and card B = card B and if there exists
an injection of A into B, then there exists an injection of  into B.
Second, one has to show that Ris a partial ordering on the set P(X). While the refiexivity
and transitivity are straightforward to verify, the antisymmetry is not, since it amounts to
 10 Real Analysis and Theory of F'unctioris: A Quick Review [Ch. 1

showing that, if there exist an injection of A into B and an injection of B into A, then there
exists a bijection of A onto B. 2
Third, one has to show that P(X)/'R. is totally ordered by the relation R, i.e., that, given
any two subsets A C X and B c X, either there exists an injection of A into B, or there
exists an injection of B into A, these two properties being not exclusive. This part of the
proof3 requires the axiom of choice.
Finally, one has to show that there exists an injection of A into B if and only if there
exists a surjection of B onto A. The proof of the "if" part again requires the axiom of choice.
As in Section 1.3, we will henceforth use the more "transparent" notations
card A � card B, resp. card A -<card B,
to express that (card A, card B) E R, resp. (card A, card B) E R and card A "# card B.
a
From now on, we shall compare cardinals of sets, even if these are not priori given as
subsets of a given set. This entails no difficulty, however, since the union of such sets can
always be defined, within the Zermelo-Fraenkel set theory (Section 1.1).
The next result4 implies that, loosely speaking, there is no cardinal number that would
be the "largest" (with respect to the relation R).
Theorem 1.5-2 Let X be any set. 'l'hen there does not exist a bijection of X onto the set
P(X) . Consequently, the relation
card X -<card P(X)
always holds. 0

A set X is flnite if either X = 0 or there exists an integer n ;::: 1 such that card X =
card{l, . . . , n} = n. A set is inflnite if it is not finite. In particular, a set X is countably
inflnite if card X = card N, and a set is uncountably inflnite if it is neither finite nor
countably infini te. Note that some authors call countable a set that is either finite or countably
infinite and call denumerable a countably infinite set.
The next theorem gathers some important properties of infinite sets. The proofs of both
(a) and (b) require the axiom of choice. Property (a) asserts that card .N is the "smallest" of
ail "infinite" cardinals.
Theorem 1.5-3 (a) Let X be an infinite set. 'l'hen
card N � card X.

2 This is the content of the famous theorem, proved in 1897 by Felix Bernstein ( 1878-1956) .
3 Due to:
E. ZERMELO [1904] : Beweis dass jede Menge wohlgeordnet werden kann, Mathematische Annalen LIX,
514-516.
4 The theory of cardinal numbers is due to Georg Cantor (1845-1918), who expounded it in a highly
inlluential (and for a long time highly controversial among mathematicians, some very famous) book:
G. CANTOR [1899] : Beitrage zur Begründung der tmnsfiniten Mengenlehre, Georg Olms Verlag (English
translation: Contributions to the Founding of 'I'ransfinite Numbers, Dover, New York, 1955) .
Sect. 1.6] Topological spaces 11

( b ) Let X be an infinite set. Then

card (X x X ) = card X.

(c ) The cardinal of the set IR satisfies

card !R = card P ( N) . D

Note, however, that the important special case

card (N x N) = card N
of property (b ) can be proved directly by means of a simple counting argument, i.e., without
using the axiom of choice. This special case in turn easily implies that
card Q = card N,
and that a finite or countably infinite union of countably infinite sets is also countably infinite.
Combined with Theorem 1.5-2, property (c ) implies that
card N -<card IR.
The continuum hypothesis asserts that there does not exist any infinite set X whose
cardinal would satis/y card N -<card X -<card IR. The long-standing question of whether the
continuum hypothesis is true was beautifully settled in 1963 and 1964, when Paul J. Cohen
showed in two landmark papers5 that the continuum hypothesis is independent of the six
axioms of the Zermelo-F'raenkel set theory. In an equally famous monograph, Kurt Gode!
had already shown in 19406 that, if the Zermelo-Fraenkel set theory is noncontradictory, it
remains so under the addition of the continuum hypothesis.

1 .6 Topological spaces
A topological space is a pair ( X, 0), where X is a set, and 0 is a subset of P ( X ) with the
following properties:
Given any family (Oi)iEI of subsets oi E 0, their union uiEI oi belongs to 0 (the set I
may thus be finite, countably infinite, or uncountably infinite) ; given any finite family (03)'J= l
of subsets 03 E 0, their intersection nJ= l 03 belongs to O; and the set X and the empty set
0 belong to O.
If (X, 0) is a topological space, the set X is said to be equipped with a topology
( corresponding to the subset 0 of P ( X )) , a subset of X that belongs to 0 is open ( for
this topology) , and a subset F of X is closed ( for this topology) if the set X - F is open.
5 P . J . COHEN (1963, 1964] : The independence of the continuum hypothesis, Proceedings of the National
Academy of Sciences, USA 50, 1 143-1 148, and Proceedings of the National Academy of Sciences, USA 51,
105-1 10.
6K. GÔDEL (1940] : The Consistency of the Axiom of Choice and of the Generolized Continuum Hypothesis
with the Axioms of Set Theory, Princeton University Press, Princeton, NJ.
12 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

Clearly, given any family (Fi) iEI of closed subsets Fi, their intersection nie / Fi is closed;
given any finite family (Fj)'J= l of closed subsets Fj , their union Uf=1 Fi is closed; and the
set X and the empty set are closed (these properties simply follow from de Morgan's laws;
cf. Section 1.3).
In a topological space (X, 0), a neighborhood of a point x E X is any subset of X
that contains an open set containing x. The set formed by ail the neighborhoods of point
x E X is denoted V(x).
Given two points a, b E IR such that a < b, let ] a , b[ := {y E IR; a < y < b}. A fundamental
example of topological space is (IR, 0), where a nonempty subset 0 of IR belongs to 0 if and
only if, for each x E 0, there exist a < b such that x E ]a, b[ and ]a, b[ c O.
Let a, b E IR be two points that satisfy a < b. The set ] a , b[, which is open for this topology,
is called the open interval with end-points a and b. The unbounded sets ] - oo, a[ := {y E
IR; y < a}, ] b, oo[ := { y E IR; b < y }, and IR itself, which are likewise open for this topology,
are also called open intervals.
as
Unless explicitly stated otherwise, the set IR will be always considered equipped with
this topology, called its usual topology. The following characterization of the open sets for
this topology is very useful.
Theorem 1.6-1 Let IR be equipped with its usual topology. Then any nonempty open subset
of IR can be written as a finite or countably infinite union of disjoint, bounded or unbounded,
open intervals. D

Let (X, 0) be a topological space and let A be a subset of X. The interior of A, denoted
Â. or int A, is the union of ail the open sets contained in A; equivalently,
Â. = int A := {x E X; A E V(x)}.
The closure of A, denoted A, is the intersection of ail the closed sets containing A; equiva­
lently,
A := {x E X; V n A =/:- 0 for ail V E V(x)} = X - {int(X - A)} . .
as
The boundary of A, denoted âA, is defined the intersection of A and X - A; equivalently,
âA := {x E X; V n A =/:- 0 and V n (X - A) =/:- 0 for ail V E V(x)}.
Note that âA = A - Â. .
Let (X, 0) be a topological space. The support of a real-valued or complex-valued
function f : X -t lK is the set
supp f := {x E X; f(x) =/:- 0}.
A subset A of a topological space (X, 0) is dense in X if A = X.
A topological space (X, 0) is separable if it contains a finite or countably infinite dense
subset, i.e., there exist elements Xn E X, n � 0, such that u�=o {xn} = X.
A topological space (X, 0) is said to be Hausdorff, or equivalently, to be equipped with
a Hausdorff topology, if, given any two distinct points x E X and y E X, there exist a
neighborhood V of x and a neighborhood W of y such that V n W 0. =
Sect. 1 .6] Topological spaces 13

A topological space (X, 0) is said to be normal if, given any two disjoint closed subsets
Fl and F2 of X, there exist disjoint open subsets 0 1 and 02 such that Fl c 0 1 and F2 c 02 .
Let X be a topological space. A sequence (xn )�=O of points Xn E X is convergent in
X if there exists a point x E X such that, given any neighborhood V of x, there exists an
integer no = no(V) 2:: 0 such that Xn E V for all n 2:: no.
If X is Hausdorff, such a point x is unique and is called the limit of the sequence (xn )�=l ·
In this case, the notations
as
--tooXn , or Xn n--t
x = nlim --too x, or Xn --+ x n --+ oo,
are equivalently used to express that x is the limit of the convergent sequence (xn )�=o·
Let X be a set and let (Y, 0) be a Hausdorff topological space. A sequence Un )�=O of
mappings fn : X --+ Y is said to be pointwise convergent to a mapping f : X --+ Y if
as
for each x E X, fn (x ) --+ f(x) n --+ oo.
Let (X, 0) be a topological space, let A be a subset of X, and let 0A denote the subset
as
of P(A) consisting of all the subsets 0A of A that can be written 0A = 0 n A for some
0 E O. Then (A, 0A) is also a topological space, and A is said to be equipped with the
topology induced on A by the topology of (X, 0), or simply the induced topology if
there is no ambiguity regarding the nature of the set O. By definition, the subsets of A that
belong to 0A are thus the open sets for the induced topology.
Then a subset FA of A is closed in the topological space (A, 0A) if and only if there exists
a subset F of X that is closed in (X, 0) such that FA = F n A; likewise, a subset VA of A is
a neighborhood of a point x E A in (A, 0A) if and only if there exists a neighborhood V of
x in (X, 0) such that VA = V n A.
Naturally, if A is a subset of X and B is a subset of A, the topological properties of B
in (X, 0) and the topological properties of B in (A, 0A) have to be carefully distinguished.
For instance, A is always open in (A, 0 A) but evidently not necessarily so in (X, 0).
Let (Xj , Oj ), 1 � j � n, be topological spaces, let X := Ilj 1 Xj denote the (finite)
=

product of the sets Xj , 1 � j � n, and let

0 := {O E P(X); for each x E 0, there exist Oj E Oj , 1 � j � n,
such that x E 01 x · · · x On and 0 1 x · · · X On C O } .
Then (X, 0) is a topological space, and X is said to be equipped with the product topology,
corresponding to the subsets Oj of P(Xj), 1 � j � n.
More generally, given any family of topological spaces (Xi , Oi) , i E I, the product topol­
as
ogy in the product X = IlieI Xi is defined follows: A subset 0 c X is open in this topology
if, for each x E 0, there exists a finite family (Oi)ieJ (x) of open sets Oi E Oi such that
xE (iEJII(x) ai) x (iElII xi) and (iEJII(x) ai) x (iEIII(x) xi) c O,
(x)
where I(x) := I J(x).
-

For the sake of notational brevity, we shall no longer mention 0 in the notation so far
used for a topological space (X, 0), whenever no ambiguity should arise about the nature of
the subset 0 c P(X) that is considered.
 14 Real Analysis and Theory of Punctions: A Quick Review (Ch. 1

1.7 Continuity in topological spaces

A mapping f : X -t Y from a topological space X into a topological space Y is continuous

at a point x E X if, given any neighborhood V of f(x) in Y, there exists a neighborhood U
of x in X such that the direct image f(U) of U under f (Section 1.2} is contained in V.
A basic property of continuous mappings between Hausdorff spaces is that "they map
convergent sequences into convergent sequences" :
Theorem 1. 7-1 Let X and Y be two Hausdorff topological spaces and let f : X -t Y be a
mapping that is continuous at a point x E X. Then, given any sequence (xn )�=O of points
Xn E X that converges to x in X, the sequence (f(xn ))�=O converges to f(x) in Y. 0

The converse holds in the important special case where the topology of X is that of a
metric space (Theorem 1.11-1}.
The following criterion of continuity at a point of a composite mapping is constantly used
( and immediate to prove) :
Theorem 1. 7-2 Let X, Y, Z be three topological spaces, let f : X -t Y be a mapping that is
continuous at a point x E X, and let g : Y -t Z be a mapping that is continuous at the point
f(x) E Y. Then the composite mapping g o f : X -t Z is continuous at x. 0
A mapping f : X -t Y is continuous if it is continuous at all points of X. The following
characterization of continuous mappings ( also immediate to prove) is fundamental:
Theorem 1. 7-3 Let X and Y be two topological spaces. A mapping f : X -t Y is continuous
if and only if the inverse image under f of any open set in Y is open in X; or equivalently,
if and only if the inverse image under f of any closed set in Y is closed in X. 0
The set formed by all the continuous mappings from X to Y is denoted
C(X; Y ) , or C(X ) if Y = R
Let X and Y be two topological spaces. A mapping f : X -t Y is said to be a homeo­
morphism of X onto Y if f is a bijection, f E C(X; Y ) , and f - 1 E C(Y; X ).
The following characterization of homeomorphisms immediately follows from the defini­
tion and Theorem 1.7-3:
Theorem 1. 7-4 Let X and Y be two topological spaces and let f E C (X; Y) be a bijection.
Then f is a homeomorphism of X onto Y if and only if the direct image under f of any open
subset of X is an open subset of Y; or equivalently, if and only if the direct image under f
of any closed subset of X is a closed subset of Y. 0
Two topological spaces X and Y are said to be homeomorphic if there exists a homeo­
morphism of X onto Y.
Let us now examine the special cases where the set X, or the set Y, is a finite product.
Theorem 1. 7-5 Let Xj, 1 ::; j ::; n, and Y be topological spaces, let the product X :=
Tij= 1 Xj be equipped with the product topology (Section 1.6}, and let f : X -t Y be a mapping
Sect. 1.8] Compactness in topological spaces 15

that is continuous at a point a = ( aj )j i E X. Then for each 1 S j S n, the mapping

=

is continuous at the point ai . D

Note that the converse does not necessarily hold; consider for instance the special case
where n = 2, Xi = X2 = Y = IR, f (xi, x2 ) = X:ix 2 2 if (xi , x2 ) -:/:- (0, 0) and f (xi , x2 ) = 0 if
i + x2
(xi , x2 ) = (0, 0), and (ai, a2 ) = (0, 0) .
Theorem 1. 7-6 Let X and Yi, 1 S i S m , be topological spaces, and let the product Y :=
n:, i Yi be equipped with the product topology. Then a mapping f = (fü� i : X --+ Y is
continuous at a point a E X if and only if each mapping fi : X --+ Yi, 1 S i S m, is
continuous at a E X. D
The following extension theorem for continuous functions is fundamental. In particular,
it will be abundantly used in Chapter 9, for defining the Brouwer topological degree in !Rn ,
or for establishing the hairy ball theorem and the Borsuk-Ulam theorem.
Theorem 1. 7-7 (Tietze-Urysohn extension theorem) Let X be a normal topological
space, F a closed su� et of X, and f : F --+ lR a continuous function. Then there exists a
continuous function f : X --+ lR such that
f(x) = f (x) for all x E F. D
Finally, we mention a fundamental way to construct a specific topology on a set, by means
of given mappings from this set into topological spaces.
Theorem 1. 7-8 Let there be given a set X and a family (<pi)iEI of mappings 'Pi from X into
topological spaces Yi. 1'hen there exists a topology on X with the following two properties:
First, all the mappings 'Pi : X --+ Yi, i E J, are continuous.
Second, any subset of X that is open for this topology is necessarily open for any topology
on X for which all the mappings 'Pi : X --+ Yi, i E J, are continuous. D
In view of these two properties, the ( clearly unique) topology defined in Theorem 1.7-8 is
aptly called the weakest topology on X that renders all the mappings 'Pi : X --+ Yi, i E J,
continuous. As we shall see, the weak and weak • topologies, on a normed vector space and
its dual space, constitute fundamental examples of weakest topologies ( Section 5.12).

1 .8 Compactness in topological spaces

Let (X, O) be a topological space. A subset K of X is compact if, given any family (ai)iel
of open sets ai E 0 such that K C LJiEI ai, there exists a finite subfamily (aj)j EJ of the
family (ai)i EI such that K c ujEJ aj .
This property, which constitutes the Heine-Borel-Lebesgue property, is often ex­
pressed as follows: A subset K of X is compact if any open covering of K admits a finite
subcovering.
 16 Real Analysis and Theory of F'unctions: A Quick Review [Ch. 1

That a subset K of X is compact does not depend on whether K is considered as a subset

of X, or as a topological space per se, equipped with the induced topology. In other words, a
subset K of a topological space (X, 0) is compact if and only if the topological space (K, OK )
equipped with the induced topology (Section 1.6) is compact.
The following theorems assemble some basic (and elementary to prove) properties involv­
ing compactness.
Theorem 1.8-1 A topological space X is compact if and only if, given any family (Fi ) i EI
of closed subsets Fi of X with the property that njEJ Fj =/= 0 for any fi.nite subfamily (Fj )jEJ
Of the family {Fi)iE/ , we have niEl .fi =/= 0. 0

Theorem 1.8-2 (a) Any compact subset of a Hausdorff topological space is closed.
(b) A closed subset of a compact topological space is compact. 0
:
Theorem 1.8-3 Let X and Y be two topological spaces and let f X --+ Y be a continuous
mapping. Then the direct image f(K) of any compact subset K of X is a compact subset
of Y. 0
Theorem 1.8-4 Any continuous bijection from a compact topological space X onto a topo­
logical space Y is a homeomorphism of X onto Y (then Y is also compact by Theorem 1.8-3) .
0

Theorem 1.8-5 Let Xj, 1 ::; j ::; n, be compact topological spaces. Then the product
I1j=1 Xj equipped with the product topology (Section 1.6) is compact. 0

Theorem 1.8-5 is a special case of Tychonoff's theorem, one of the most important
results in general topology. This theorem asserts that, given any family (Xi)i EI of compact
topological spaces, the product rriEl Xi equipped with the product topology is compact. 7
But, by contrast with that of Theorem 1.8-5, its proof requires the axiom of choice.
A subset A of a topological space X is relatively compact if its closure A is a compact
subset of X.

1.9 Connectedness and simple-connectedness in topological

spaces

A topological space {X, 0) is connected if the only subsets of X that are both open and
closed are X and 0. A subset A of X is connected if it is a connected topological space
when it is equipped with the topology induced by that of X {Section 1.6).
That a subset A of X is connected does not depend on whether A is considered as a
subset of X, or as a topological space by itself, i.e., when it is equipped with the induced
topology.
7This theorem was first proved in the special case where X; = [O, 1] for ail i E I in:
A. TYCHONOFF [1930] : Ü ber die topologische Erweiterung von Riiumen, Mathematische Annalen 102, 544-
561 .
The general case w as then proved in:
E. Ô ECH [1937] : On bicompact spaces, Annals of Mathematics 38, 823-844.
Sect. 1.9] Connectedness and simple-connectedness in topological spaces 17

The following theorems gather some basic properties involving connectedness.

Theorem 1.9-1 Let A be a connected subset of a topological space X. Then any subset B
of X that satisfies A c B c A, hence B = A in particular, is also connected. 0
Theorem 1.9-2 Let X be a connected topological space, let Y be a topological space, and
let f : X -t Y be a locally constant function, i. e., such that each point x E X possesses a
neighborhood Vx such that the restriction f l vx is a constant function. Then f is a constant
function. 0
Like compactness, connectedness is a property that is "preserved by continuous map­
pings" :
Theorem 1. 9-3 Let X and Y be two topological spaces and let f : X -t Y be a continuous
mapping. Then the direct image f (A) of any connected subset A of X is a connected subset
of Y. D
The next result characterizes the connected subsets of IR.
Theorem 1.9-4 Let lR be equipped with its usual topology. Then a subset of IR is connected
if and only if it is an interval, bounded or unbounded. 0
An immediate corollary of Theorems 1.9-3 and 1.9-4 then follows:
Theorem 1.9-5 (Bolzano intermediate value theorem) Let X be a connected topolog­
ical space, let f : X -t lR be a continuous function, and let a, b E X be such that f(a) < f(b)
(to fix ideas) . Then, given any y E ]f(a), f(b)[, there exists x E X such that f(x) = y . 0

The next three theorems provide useful sufficient conditions for connectedness.
Theorem 1.9-6 Let X be a topological space and let (Ai)ieI be any family of connected
subsets Ai of X. If the intersection niEI Ai is nonempty, then the union LJie I Ai is connected.
0
Theorem 1.9-7 Let Xj , 1 ::; j ::; n, be connected topological spaces and let their product
X = Ilj= 1 Xj be equipped with the product topology (Section 1.6) . Then X is connected. 0
Let X be a topological space. The relation R defined by "(x, y) E R if and only if there
exists a connected subset of X that contains both x and y," is an equivalence relation on X.
The equivalence classes modulo this relation, which are thus subsets of X, are called the
connected components of X.
Given any x E X, the connected component of X that contains x is called the connected
component of x; it is also the largest connected subset of X that contains x, according to
the following result.
Theorem 1.9-8 Let X be a topological space and let x E X. Then the connected component
of x is the union of all the connected subsets of X that contain x. 0
Let x and y be two points in a topological space X. A path joining x to y is a continuous
18 Real Analysis and Theory of Functions: A Quick Review [Ch. 1

mapping '"Y : [0, 1] -t X such that '"'f(O) = x and '"'f(l) = y.

A topological space X is arcwise-connected if, given any two distinct points x, y in X,
there exists a path joining x to y.
Theorem 1.9-9 An arcwise-connected topological space is connected. 0
The converse implication does not necessarily hold. For instance, let
A := {(x, y) E IR2 ; x = 0, IYI ::; 1} and B := {(x, y) E IR2 ; x =/:- 0, y = sin(x 1 )}.
-

Then A U B is a connected subset of JR2 that is not arcwise-connected.

Let x and y be two points in a topological space X. Two paths '"Yo : [0, 1] -t X and
'"Yl : [O, 1] -t X joining x to y are homotopie if there exists a continuous mapping H : [O, 1] x
[0, 1] -t X, called a homotopy joining '"Yo to '"'(1 , such that H(·, O) = '"Yo and H(·, 1) = '"'(i ,
and H(O, ·) = x and H(l, · ) = y.
A topological space X is said to be simply connected if it is arcwise-connected, hence
connected (Theorem 1.9-9), and if any two paths '"Yo : [O, 1] -t X and '"Yl : [0 , 1] -t X such
that '"Yo (O) = '"'(1 ( 0) and '"Yo(l) = '"Y1 (l) are homotopie.

1 . 10 Metric spaces

Let X be a set. A distance on X is a function d : X x X -t lR that satisfies the following

z
properties for all x, y, E X:
d(x, x) = O and d(x, y) > O if x =f. y,
d(x, y) = d(y, x),
d(x, z) ::; d(x, y) + d(y, z).
The last property is called the triangle inequality. A metric space is a pair (X, d) where
X is a set and d is a distance on X.
In what follows, (X, d) is a metric space. Given a point x E X and a num�er r > 0, the
ball with center x, or centered at x, and radius r > 0 is the subset of X defined by
B(x; r) = {y E X; d(y, x) < r}.
A subset A of X is bounded if there exists a ball B(x; r) c X such that A C B(x; r). lt
is unhounded otherwise.
The diameter of a nonempty subset A of X is defined as the extended real number
diam A := sup{d(x, y); x E A, y E A} E [0, oo] .
Clearly, a subset A of X is bounded if and only if diam A < oo.
The distance from a point x E X to a nonempty subset A of X is defined as the real
number
dist(x, A) := inf{d(x, y); y E A}.
Unless otherwise mentioned, a metric space (X, d) will be always viewed as a topological
space (X, 0), whose open sets, i.e., the subsets of X that belong to 0, are those described
in the next theorem. This "canonical" topology is called the topology induced on X by
the distance d.
Sect. 1. 10] Metric spaces 19

Theorem 1.10-1 Let (X, d) be a metric space. Let 0 denote the subset of P(X) consisting
of the empty set and of all the subsets 0 of X with the following property: Given any E 0, x
there exists r > 0 such that the ball B(x;
r) is contained in O. Then the pair (X, 0) is a
topological space, which is Hausdorff and normal.
Besides, any ball B(x; x
r), with E X and r > 0, is an open set for this topology. 0
For instance, the usual distance on IR, defined by d(x,y) = l x - YI
for all E IR, x,y
induces the usual topology of lR ( Section 1.6). Likewise, the usual distance on C, defined
by d(x, y) = l x - YI
for al! x, y
E C, induces a topology on C, called the usual topology
of C.
When viewed as a metric space, the set IR, or a subset of IR, will be always implicitly
considered as endowed with this distance d, called the usual distance on IR.
Of course, d is far from being the only distance on lR that induces its usual topology.
:
For instance, the distance p lR x lR -+ lR defined by 1+
p(x,y) = l x 1 x- -yl y I
for al! E x,y
lR also induces the usual topology on IR. Note, however, that the metric space (IR, d) is
unbounded while the metric space ( IR, p) is bounded; incidentally, this simple example shows
that boundedness is a metric notion, not a topological one.
The topology of a topological space (X, 0) is said to be metrizable if it can be induced
by a metric on X, and any such metric is said to be compatible with the topology. For
instance, the usual topology of lR is metrizable, and the above metrics d and p are both
compatible with it.
Another fondamental example of metric space is that of ocn , where lK lR or lK C, = =
or of a subset X of ocn , equipped with one of the distances dp, 1 $ p $ oo, defined for any
n-tuples E !Kn and
X = (xi )f= l E !Kn by
Y = (Yi )f= l
lp
dp(x,y) = (� l xi - YilP) / if 1 $ p < OO ,
doo (x,y) = 1m::;iilJC::;n l xi - Yil·
Ali the axioms of a distance are immediately verified, save the triangle inequality for 1 < p <
oo ,for which we refer the reader to Theorem 2.4-1. The distance d2 is called the Euclidean
distance.
Given any 1 $ p, q $ oo, any ball corresponding to the distance dp is contained in a
ball corresponding to the distance dq and centered at the same point. Hence the topology
induced on ocn , or on a subset of ocn , by any one of the distances 1 $ p $ oo, is the dp,
same, and is called the usual topology of ocn . It is also easily verified that this topology
coïncides with the product topology (Section 1.6) on ocn Tij= 1 Xj , where each topological
=
space Xj , 1 $ j $ n, is the set lK equipped with its usual topology, and that ocn equipped
with this topology is a separable topological space.
More generally, any subset X of a finite-dimensional vector space over lK with a basis
(ei )f= l
becomes a metric space when it is equipped with one of the above distances 1 $ dp,
x y dp(x, y)
p $ oo, with and in now replaced by x = L::=l Xiei y = L::= l Yiei .
and
In what follows, a metric space (X, d) will be often simply denoted X for notational
brevity, in which case it is implicitly understood that its metric is denoted d when needed.
20 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

The various definitions given in Section 1.6 then have the following equivalent "metric"
counterparts in a metric space X:
A neighborhood of a point x EX X
is any subset of that contains a ball centered at x.
X E
The interior A of a subset A of is the set of all points x A such that there is a ball
centered at x and contained in A.
X
The closure A of a subset A of is the set of ail points x EX such that any ball centered
at x has a nonempty intersection with A.
X
The boundary âA of a subset A of is the set of ail points x EX such that any ball
centered at x has a nonempty intersection with both A and X -A.
A metric space (X, d) is separable if there exist elements Xn E X, n ;::: 1, such that, given
any x EX ê ( ê)
and any > 0, there exists n = n x, ;::: 1 such that x E B(Xni ê).
A sequence (xn )�=O of points Xn E X is convergent if there exists a point x such that,
ê ,x)
given any > 0, there exists no ;::: 0 such that
d(x Xn E B(x; ê) for all n ;::: no, or equivalently,
such that n -t 0 as n -t oo. Such a point x, which is necessarily unique because the
associated topological space is Hausdorff (Theorem 1. 10-1), is thus the limit of the sequence.
Thanks to this characterization of limits in terms of distance, the closure of a subset in a
metric space can be given a simple characterization by means of convergent sequences.
X.
Theorem 1.10-2 Let A be a subset of a metric space Then a point x EX belongs to A
if and only if there exists a sequence (xn )�=O of points Xn E A that converges to x as n -t oo .
Consequently, a subset A of X is closed if and only if
Xn E A, n 2:: 0, and Xn -t x in X as n -t oo implies x E A. D
Incidentally, note that dist (x, A) > 0 if x <f. A and A is closed.
Let (X, d) be a metric space, let A be a subset of X, and let dA : A x A -t lR denote the
restriction of the distance d to A x A. Clearly, dA is a distance on A, called the distance
induced by d on A, and thus (A, dA) is also a metric space. Furthermore, the following
properties hold.
Theorem 1.10-3 Let (X,d) be a metric space and let A be a subset of X. The topology
induced on A by the metric dA coincides with the topology induced on A by the topology
induced on X by the metric d (Section 1.6) .
Purthermore, (A, dA) is separable if (X, d) is separable. D

Finally, let (Xj, dj), 1 � j � n, be metric spaces, and let X := Ilj=1 Xj. Then a subset
0 c X is open for the product topology on the product X if and only if, given any point
x (xJ)J= l E X, there exist rj > 0, 1 � j � n, such that Ilj= l B(xj, rj) C O. Any
=
distance d on the product space X that induces the product topology on X is then said to
be compatible with the product topology. Examples of such compatible distances are
provided by the functions d : X x X -t lR and : X x X -t lR defined by
p

n
d(x,y) jLdj(Xj,
:=
= l Yj) and p(x, y) lm!J.X :=
$3$n dj(Xj, YJ)
for all x (xJ)J=l E X and y (YJ)J=l E X.
= =
Sect. 1.11] Continuity and uniform continuity in metric spaces 21

1.11 Continuity and uniform continuity in metric spaces

A mapping from a topological space X into a topological space Y that is continuous at a point
x E X maps sequences converging to x in X into sequences converging to f(x) in Y (Theorem
1.7-1). The converse holds if the topologies of both X and Y are those of a metric space
(in fact, Theorem 1.11-1 still holds if X is a metric space and Y is a Hausdorff topological
space):
Theorem 1.11-1 Let X and Y be metric spaces. Then a mapping f : X -+ Y is continuous
at a point x E X if and only if, given any sequence (xn )�=O of points Xn E X that converges
to x in X, the sequence (f(xn ))�=O converges to f(x) in Y. D
The next theorem gives a simple, and often used, property of continuous mappings in
metric spaces:
Theorem 1.11-2 Let X be � dense subset of3 metric space X, let Y be a Hausdorff
topological space, and let f : X -+ Y and g : X -+ Y be two continuous mappings that
coincide on X, i. e., f(x) = g(x) for all x E X. Then f = g. D
If X and Y are both metric spaces, the continuity at a point can be also expressed in
terms of balls, or in terms of distances: Let (X, d) and (Y, p) be two metric spaces. Then a
mapping f : X -+ Y is continuous at a point x E X if the inverse image of any ball in Y
centered at f(x) contains a ball in X centered at i, or equivalently, if, given any e > 0, there
exists ô = ô (e, x) > 0 such that p(f(x), f(x)) < e for all x E X such that d(x, x) < ô. This
as
equivalent definition, specific to metric spaces, is sometimes referred to the "e-ô definition
of continuity."
If a mapping f : X -+ Y is continuous, i.e., if it is continuous at all points x E X, it may
happen that, given any e > 0, the above number ô(e, x) > 0 can be chosen independently of
x E X. This possibility leads to the following definition: Let (X, d) and (Y, p) be two metric
spaces. A mapping f : X -+ Y is uniformly continuons if, given any e > 0, there exists
ô(e) > 0 such that p(f(x), f(x)) < e for all x, x E X that satisfy d(x, x) < ô(e).
An important example of a uniformly continuous mapping is provided by a Lipschitz­
continuous mapping, i.e., a mapping f : (X, d) -+ (Y, p) with the property that there exists
a constant k such that
p (f (x), f (x)) ::; kd(x , x) for all x, x E X.
Such a mapping is then said to satisfy a Lipschitz condition, with Lipschitz constant k.
Another example is provided by a HOlder-continuous mapping, i.e., a mapping f :
(X, d) -+ (Y, p) with the property that there exist constants C and 0 < À < 1 such that
p(f(x), f(x)) ::; C(d(x, xn> · for all x, x E X.
Such a mapping is then said to satisfy a Holder condition of exponent À.
Let (X, d) be a metric space and let the product X x X be equipped with the distance
D defined by
D ((x, x), (y, if)) : = d(x, y) + d(x , if) for all (x, x) E X x X and all (y, if) E X X X.
22 Real Analysis and 1'heory of Punctions: A Quick Review [Ch. 1

d
Then the function : (X x X, D ) --+ lR provides an example of a Lipschitz-continuous function,
with Lipschitz constant one. 'lb see this, simply note that, by the triangular inequality,
l d(x, x) - d(y, Y) I :::; d(x,y) + d(x, Y) = D ((x, x), (y, Y)) .
Another similar example is provided by the distance to a subset:
d)
Theorem 1.11-3 Let (X, be a metric space and let A be a nonempty subset of X. Then
(x,
jdist A) - dist (y, A) I d(x,
� y) for all y E X. x, 0

d) )
Given two metric spaces (X, and (Y, p , a mapping f : X --+ Y is an isometry from
(()
X into Y if f "preserves the distances,'' i.e., p f x , f(y)) = d(x, y) x,
for all y E X. An
isometry thus provides another example of a uniforrnly continuous mapping.
Otherwise, a general, and very useful, sufficient condition for uniform continuity will be
given in Theorem 1.13-2.

1.12 Complete metric spaces

d), (x Xn
LJ:=nn)�={xm}O
In a metric space (X, a sequence of points E X is a Cauchy sequence if the
diameter (Section 1.10) of the set converges to zero as n --+ oo, or equivalently,
()
if, for each e > 0, there exists an integer np e ;:::: 0 such that d(xm,Xn )
< e for all m ;:::: no e ()
()
and n ;:::: no e .
The following theorem gathers elementary properties of Cauchy sequences:
Theorem 1.12-1 (a) A Cauchy sequence is bounded.
(b) A convergent sequence is a Cauchy sequence.
(c) A Cauchy sequence that contains a convergent subsequence is convergent, and its limit
is the limit of the subsequence. 0

d)
A metric space (X, is complete if every Cauchy sequence of points of. X converges
in X. A subset A of a metric space is complete if the metric space (X,dA),
where denotes dA
the distance induced by d on A (Section 1.10), is complete. Consequently, the property "X is
a complete metric space" is independent of whether X is a subset of a larger metric space.
The following theorem gathers elementary properties of complete metric spaces:
Theorem 1.12-2 Let A be a subset of a metric space X.
(a) If A is complete, A is closed in X.
(b) If X is complete and A is closed in X, A is complete.
(c) If X is complete, a subset A of X is complete if and only if A is closed in X. 0

e,
Pundamental examples of complete metric spaces are lR and each equipped with its
usual distance, and and ]Rn en , n ;:::: 2, each equipped with any one of the distances 1 � dp,
e
p � oo (Section 1.10): That lR and are complete follows from their construction (Section
1.4); that (JRn , dp) (en , dp)
and are complete in turn easily follows from the completeness of
lR and e.
For a given integer n ;:::: 2, the distances dp,
1 � p � oo, thus provide examples of
JRn
distances that induce the same topology on and simultaneously render the metric spaces
Sect. 1.13] Compactness in metric spaces 23

(JRn , dp ) complete. This is not a general circumstance, however. For example, let lR+ : =
{x E JR; x � O} and let the distances d and p on lR+ be defined by d(x , y ) = l x YI and
J
-

p (x , y ) = - x - _Y_ J for ail x , y E lR · Then these distances induce the same topology
-

l+x l+y +
on lR+ , but {lR+ , d) is complete while (lR+ , P) is not {the sequence (xn )�=O with Xn = n is a
Cauchy sequence in {lR+ , p) but does not converge in {lR+ , p)).
The next theorem is fundamental. lt provides sufficient conditions insuring that a map­
ping defined and continuous on a dense subset of a metric space can be extended to a con­
tinuous mapping on the whole space {this result will be proved later in Theorem 3.1-1 for
normed vector spaces).
Theorem 1.12-3 (unique continuous extension) Let X be a dense subset of a metric
space X, let Y be a complete metric space, and let f : X --+ Y be a uniformly continuous
mapping.
Then there exists one and only one continuous extension f : X --+ Y of f to the space X .
The mapping f is also uniformly continuous on X . 0
The next theorem is also fundamental. It asserts that any metric space that is not
complete can be always identified with a dense subset of a complete metric space by means
of an isometry {this result will be proved later in Theorem 3. 1-2, again for normed vector
spaces).
Theorem 1.12-4 (completion �f � metric space) (a) Let (X, c!) be a metric space. There
exis.l_s a complete metric space (X, d) and an isometry a : X --+ X such that a(X) is dense
in X.
{b) The space X is separable if the space X is separable.
(c) If (X, dJ is any complete metric space such that there exists an isometry from X onto
a dense subset of X, then there exists an isometry from (X, d) onto (X, d) . 0
The space {X, d) , which is thus "essentially unique" as a metric space, in the sense that
it is unique up to bijective isometries thanks to property (c), is called the completion of the
metric space {X, d).
Two other fundamental theorems about complete metric spaces, viz., the Banach fixed
point theorem and Baire 's theorem, will be proved in the next chapters {Theorems 3.7-1 and
5. 1-2).

1 . 13 Compactness in metric spaces

Let {X, d) be a metric space and let K be a subset of X, equipped with the topology induced
by the metric d. Then K is compact if, as a topological space, it satisfies the Heine-Borel­
Lebesgue property (Section 1.8).
Noting that any compact subset is closed in a Hausdorff topological space (Theorem
1.8-2{a) ), and that any covering by balls of radius one (to fix ideas) admits a finite subcovering
in a compact metric space {by the Heine-Borel-Lebesgue property), we immediately obtain
two necessary conditions for compactness in a metric space:
Theorem 1.13-1 A compact subset of a metric space is closed and bounded. 0
24 Real Analysis and Theory of F'unctions: A Quick Review [Ch. 1

Another simple consequence of the Heine-Borel-Lebesgue property in a metric space is a

sufficient condition of uniform continuity:
Theorem 1.13-2 Let X be a compact metric space and let Y be a metric space. Then a
continuous mapping from X into Y is uniformly continuous. 0

A subset A of a metric space X is precompact if, given any e > 0, there exists a finite
number n = n(e) of points Xj = Xj (e) E A , 1 $ j $ n, such that
n
A C U B(xj; e).
j=l
Note that A C X is precompact if and only if ïI is also precompact.
The following characterizations of compact and precompact subsets of a metric space are
fundamental.
Theorem 1.13-3 Let X be a metric space and let K be a subset of X. The following three
assertions are equivalent:
(a) K is a compact subset of X.
(b) Given any sequence (xn )�=O of points Xn E K , there exists a subsequence (xu(ni )�=O
that converges to a point in K.
(c) K is precompact and complete. 0

A topological space A that satisfies the above property (b) is said to satisfy the Bolzano­
Weierstra6 property, to reflect that it generalizes Theorem 1.4-l(b).
Theorem 1.13-4 A subset A of a metric space is relatively compact if and only if any
sequence (xn )�=O of points Xn E A contains a subsequence (xu(ni )�0 that converges to a
point in ïI. 0

While the converse of Theorem 1.13-1 "seldom holds," an easy application of Theorem
1.13-3(c) shows that it does hold in the following fundamental special case:
Theorem 1.13-5 Let the space ocn , where OC = lR or OC = C, be equipped with any one of
the distances dp, 1 $ p $ oo (Section 1.10). Then a subset of ocn is compact if and only if it
is closed and bounded. 0

We shall prove later that this property is in effect a characterization of finite dimension­
ality. This is the essence of the fundamental F. Riesz theorem (Theorem 2.7-3).
While the characterization of compact subsets in a finite-dimensional space is thus settled
by Theorem 1.13-5, the characterization of compact subsets in infinite-dimensional spaces is
often a delicate issue. An important instance8 of such a characterization is the Ascoli-Arzelà
theorem in the space of functions that are continuous on a compact set (Theorem 3.10-1).
By Theorem 1.13-5, a subset of lR is compact if and only if it is closed and bounded.
Combining this observation with Theorem 1.8-3 yields another basic result, asserting that
8 Another important instance is Kolmogorov 's theorem in the spaces LP(O) ; for a proof, see, e.g., BREZIS
[2011, Theorem 4.26) .
Sect. 1.14] The Lebesgue measure in !Rn ; measurable functions 25

continuous functions on compact sets attain their infimum and supremum:

Theorem 1.13-6 Let K be a compact topological space and let f : K --+ IR be a continuous
function. Then the direct image f(K) is a compact subset of IR. Therefore there exist xo E K
and x 1 E K such that
f(xo) = infxE I< f(x ) and f(x 1 ) = supxE Jd (x ) . D
A spectacular application of Theorem 1.13-6 will be given in the next chapter, where it
will be shown to provide a simple proof of the fundamental theorem of algebra (Theorem
2.8-1}.

1 . 14 The Lebesgue measure in JRn ; measurable functions

In what follows, the notations (0, oo] and [ - oo, oo] respectively denote the sets [O, oo( U { oo}
and { -oo} U IR U { oo} (Section 1.4).
Let X be a set. A a-algebra of subsets of X is a subset A of P(X) that satisfies the
following properties:
X E A,
OO
A E A implies (X - A) E A,
LJ Ai E A if Ai E A for ail i 2:: 1.
i= l
Given a set X and a a-algebra A of subsets of X, a measure is a function µ : A --+ (0, oo]
that satisfies the following properties:
µ( 0) = 0,
µ (Q Ai) = � µ(Ai ) if Ai E A for ail i 2". 1 and Ai n Aj = 0 for ail i i- i·

The last property is called the a-additivity of the measure µ. The triple (X, A, µ) is
then called.a measure space.
Of fundamental importance is the set X = !Rn , where n is any integer 2". 1, equipped
with its usual topology (Section 1.10}, when the a-algebra is the Borel a-algebra A in !Rn ,
defined as the smallest a-algebra that contains all the open subsets of !Rn ( the a-algebra A is
then uniquely defined, as the intersection of ail the a-algebras of subsets of !Rn that possess
this property ) , and the measure µ, is defined by

For a given set A E A, the infimum is meant here to be taken over ail the countably infinite
families of products I1j= 1 ]aj, bj [ of open intervals, k 2". 1, the union of which covers the set
A. The elements of the a-algebra are called the Borel-measurable subsets of !Rn .
26 Real Analysis and Theory of F'unctions: A Quick Review [Ch. 1

The measure space (!Rn , A, /il co�tructed in this �hion lacks one desirable proe..erty,
namely that any subset of a set A E A that satisfies ji(A) = 0 is also in the a-algebra A. To
obviate this difficulty, let
A := {A E P(!Rn ); there exist A E A and A' E A with µ(A') = 0
such that A = AUB with B C A }, '
and let, for any A E A,
µ(A) := µ(A)
for any A E A such that A = AUB with B c A' for some A' E A with µ(A') = O.
One can then show that A is again a a-algebra of subsets of!Rn (which clearly contains the
Borel a-algebra A) , that the above definition of µ(A) makes sense (i.e., that it is independent
of the particular set A E A chosen as above) , and that the function µ : A -+ [O, oo) defined
in this fashion is again a measure.
The a-algebra A is called the Lebesgue a-algebra in !Rn , the elements of A are called
the Lebesgue-measurable subsets of !Rn , and µ is called the Lebesgue measure in !Rn ,
or the n-dimensional Lebesgue measure. Evidently, µ(A) = µ(A) for all A E A.
The Lebesgue measure in !Rn is denoted
dx , or meas, or dx- meas,
according to the context.
Cardinality arguments show that
card A = card lR and card A = card P (!R).
Hence there are "many more" Lebesgue-measurable subsets of !Rn than Borel-measurable
subsets of !Rn , since card lR -< card P(IR) (Theorem 1.5-2).
The next theorem recapitulates four basic properties of the resulting measure space
(!Rn , A, µ). The first three are direct consequences of the above construction. The fourth
one expresses that the Lebesgue measure is translation invariant.
Theorem 1.14-1 The a-algebra A of Lebesgue-measurable subsets of !Rn and the Lebesgue
measure µ : A -+ [O, oo] satis/y the following properties:
( a) Every open subset of !Rn belongs to A; hence every closed subset of !Rn , and any
countably infinite intersection or union of open or closed subsets of !Rn , also belong to A.
( b ) The Lebesgue measure of any subset of !Rn of the form I1j= 1 ]aj, bj [, which belongs to
A by (a) , is given by
µ (]�p ) "Jl
aj , bj [ = (bj - aj) ·
(c ) If A E A and µ(A) = 0, then every subset of A is also Lebesgue-measurable and its
Lebesgue-measure is zero.
(d ) Given any point x E !Rn and any set A E A, the set
x + A := {(x + y) E !Rn ; y E A}
also belongs to A and µ(x + A) = µ(A) . 0
Sect. 1 . 1 4] The Lebesgue measure in !Rn ; measurable functions 27

A noteworthy consequence of the translation invariance {d ) of the Lebesgue measure is

that, when appropriately combined with the axiom of choice, it implies the existence of
subsets of !Rn that are not Lebesgue-measurable.
We next examine how a product of measure spaces {X, A, µ) and {Y, B, v) can be also
made a measure space. First, let A ® B denote the smallest a-algebra that contains all the
sets A x B E P(X x Y), with A E A and B E B {hence A ® B is uniquely defined by these
conditions ) . Then one can show that there exists one and only one product measure
µ ® v : A ® B --+ [O , oo]
with the (expected ) property that
(µ ® v)(A x B) = µ(A)v(B) for ail A E A and B E B.
When µ is the Lebesgue measure on !Rm and v is the Lebesgue measure on !Rn , one can
show that the product measure µ ® v is ( again as expected ) precisely the Lebesgue measure
on ]Rm+n .
Let A and µ respectively designate the Lebesgue a-algebra in !Rn and the Lebesgue mea­
sure in !Rn . For brevity, the elements of A will be simply called measurable subsets of !Rn .
Let A be a measurable subset of !Rn . A property is said to hold almost everywhere
( a.e.) in A, or equivalently to hold for almost all x E A, if the set of points in A where it
does not hold is measurable and of measure zero. For instance, two functions j, g : A --+ :IR
are equal almost everywhere if the set {x E A; j(x) i- g(x)} is measurable and of measure
zero; a sequence Un )�= l of functions ln : A --+ [-oo, oo] converges almost everywhere in A as
n --+ oo if the complement of the set {x E A; limn--too fn (x ) exists in [-oo, oo] } is measurable
and of measure zero, etc.
A much Jess trivial example of a property that holds almost everywhere is provided by
the following fundamental result. In what follows, the spaces !Rn and !Rm are equipped with
any one of the distances dp , 1 :::; p :::; oo, defined in Section 1. 10.
Theorem 1.14-2 (Rademacher's theorem9) Let 0 be an open subset of !Rn and let
j : n --+ !Rm be a Lipschitz-continuous junction (Section 1.11) . Then j is differentiable
almost everywhere in 0. D
Given any Lebesgue-measurable subset A E A, a function f : A --+ [-oo, oo] is said to be
Lebesgue-measurable, or simply measurable, if
r 1 ([-oo, a [) = {x E A; j(x) < a} E A for ail a E :IR.
The next theorem recapitulates a first series of basic properties of measurable functions.
Note that property (c ) is restricted to real-valued functions.
Theorem 1.14-3 Let A be a measurable subset of !Rn .
( a) Let f : A --+ [-oo, oo] be a measurable function. Then the function I l l : A --+ [0, oo] is
also measurable.
9 So named after Hans Adolph Rademacher ( 1892-1969) .
28 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

(b) Let fn : A -t [ -oo, oo] , n � 1, be measurable functions. Then the functions

sup fn , inf f , lim sup f , lim inf f : A -t [-oo, oo]
n<:'.1 n <:'.1 n n-+oo n n-+oo n
are also measurable.
(c) Let f, g : A -t R be measurable functions. Then the functions f + g : A -t R and
fg : A -t R are also measurable. 0
The next theorem recapitulates three other basic properties, this time linking measura­
bility and continuity. Property (c) constitutes Lusin's property.
Theorem 1.14-4 Let A be a measurable subset of Rn .
(a) Let f : A -t R be a continuous function. Then f is measurable.
(b) Let f : R -t R be a continuous function and let g : A -t R be a measurable function.
Then the composite function f o g : A -t R is measurable.
(c) Let f Rn -t R be a measurable function with the property that µ(A) < oo, where
:

A {x E Rn ; f(x) "# O}. Then, given any ê > O, there exists a function fe E C(Rn ) whose
:=

support is a compact subset of A and such that

sup l fe(x) I $ sup l/ (x) I and µ({x E Rn ; f(x) "# fe(x)}) $ ê. 0
xERn xEJRn
Let A be any measurable subset of Rn . A simple fonction on A is a function s : A -t R
whose image is a finite subset of R; equivalently, there exists a finite number of pairwise
disjoint (i.e., AinA3 0 if i "# j) subsets Ai of A, 1 $ i $ m, and real numbers ai, 1 $ i $ m,
=

such that m
s = L aiXA, ,
i= l
where XA, : A -t R denotes the characteristic function of each set Ai (Section 1.2}. Clearly,
a simple function s is measurable if and only if each set Ai, 1 $ i $ m, is measurable.
Important links between measurability and simple functions are given in the next theorem.
Theorem 1.14-5 Let A be a measurable subset of Rn .
(a) Let f : A -t [ -oo, oo] be a measurable function. Then there exists a sequence of
measurable simple functions Sn : A -t R, n � 1, with the following properties:
l sn l $ l sn+ i l $ 1/ 1 for all n � 1 and, for each x E A, s n (x) -t f(x) as n -t oo .
(b} Let f : A -t [O, oo] be a measurable function. Then there exists a sequence of measur­
able simple functions Sn : A -t R, n � 1, with the following properties:

0 $ Sn $ Sn+ i $ f for all n � 1 and, for each x E A, s n (x) -t f(x) as n -t oo. 0

1.15 The Lebesgue integral in �n ; the basic theorems

Let A be any measurable subset of Rn . Given a measurable simple function s E� 1 GiXA;
=

(Section 1.14} that is � 0 (equivalently, such that ai � 0, 1 $ i $ m) , let the extended real
Sect. 1. 15) 'l'he Lebesgue integral in !Rn ; the basic theorems 29

number JA s(x) dx E [O, oo) be defined as

The Lebesgue integral of any measurable function f : A � [0, oo] is then defined as
L f(x) dx := sup { L s(x) dx; s is a measurable simple fonction and 0 ::; s ::; f in A } ·
Hence JA f(x) dx is again either ;::: 0 or equal to OO: If the Lebesgue integral JA f (x ) dx of a
measurable fonction f : A � [O, oo) is finite, then f is necessarily finite almost everywhere.
Finally, a fonction f : A � [ -oo, oo] is said to be Lebesgue-integrable, or in short
integrable, if it is measurable and is such that
L max{f(x); O} dx < oo and L max{ - f(x); O} dx < oo.
If f : A � [ -oo, oo] is Lebesgue-integrable, its Lebesgue integral is defined by
L f(x) dx L max{f(x); O} dx - L max{-f(x); O} dx.
:=

The following immediate properties of Lebesgue-integrable fonctions and their Lebesgue

integrals are constantly used: The Lebesgue integral of an integrable function is a real number,
i.e., it is not equal to oo or -oo. The Lebesgue measure of a measurable subset of !Rn is also
given by fntn XA dx = fA dx; a Lebesgue-integrable fonction is finite almost everywhere;
a measurable function f : A � [-oo , oo) is Lebesgue-integrable if and only if the function
I I I : A � [0, oo] is Lebesgue-integrable, and in this case,
J L f(x) dx J ::; L l f(x) l dx.
The set, denoted

of ail Lebesgue-integrable fonctions f : A � lR is clearly a vector space over lR (vector spaces

over lR or C are defined in Section 2 . 1 ) .
The relation "! = g almost everywhere in A" defines an equivalence relation n over the
space .C 1 (A), and the quotient set

is also a vector space over R Besides,

L f(x) dx L g(x) dx
= if f, g E .C 1 (A) are such that f = g a.e. in A.
As a consequence, the Lebesgue integral of any equivalence class in L 1 (A) is unambigu­
ously defined, as the Lebesgue integral of any fonction in the class.
30 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

As is customary, we shall also refer to elements in L 1 (A) as integrable functions, even

though they are in effect equivalence classes of integrable functions modulo R.
Clearly, the identification of functions in .C 1 (A) with their equivalence classes in L 1 (A)
constitutes a flagrant abuse of language, but it avoids many cumbersome statements and
what is meant should be always unambiguous. For instance, "f E L 1 (A) is a continuous
function" means that, in the equivalence class of f, there is a (unique) continuous fonction in
.C 1 (A); likewise, "f E L 1 (A) is finite almost everywhere in A" means that in the equivalence
class of f, there is a function which is finite everywhere in A, etc.
There are other ways of defining the Lebesgue integral and the space L 1 (A). For instance,
:
let S(A) denote the set formed by ail integrable simple functions s A --+ IR, i.e., those that
satisfy
µ({x E A; s(x) :f 0}) < oo.
:=
It is then immediately seen that the quotient set S(A) S(A )/R is a vector space, and that
the mapping
ll · ll L 1 (A) : s E S(A) --+ L l s (x) ldx,
where the integral of a simple function on A is defined as earlier, is a norm on S(A).
Then the space L 1 (A) may be equivalently defined as the completion (Theorem 1.12-4) of
the space (S(A), ll · llL 1 (A) ) . In this case, the Lebesgue integral of functions f E L 1 (A) is then
simply defined as the unique continuous extension (Theorem 1.12-3) of the linear functional
s E S(A) --+ L s(x) dx,
which is defined and continuous over the dense subset S(A) of L 1 (A).
When the set A is an open subset of !Rn , yet another definition is possible: Let Cc(A)
:
denote the space of ail continuous functions f A --+ lR with compact support in A, and let
l l f ll L 1 (A) := L I f (x) I dx ,
the symbol fA g(x) dx denoting here the Riemann integral of a function g E Cc(A). Then the
space L 1 (A) may be also equivalently defined as the completion of the space (Cc(A), ll · l b cA) ) ,
and by construction, the space Cc( A) is then dense in L 1 (A) (when the space L 1 (A) is defined
as earlier in this section, the denseness of Cc( A) in L 1 (A) becomes a theorem, which therefore
needs to be proved; cf. Theorem 2.5-3). In this case, the Lebesgue integral is then again
defined as the unique continuous extension of a linear functional defined and continuous on
a dense subset.
The notion of the Lebesgue-integrable functions can be easily extended to complex-valued
functions: Let A be any measurable subset of !Rn . Then a complex-valued function f A --+ C :

is said to be Lebesgue-integrable if
Re / E .C 1 (A) and lm / E .C 1 (A).
If this is the case, the Lebesgue integral of f is defined by
L f(x) dx L Re f (x) dx + i L Imf(x) dx.
:=
Sect. 1.15] The Lebesgue integral in IRn ; the basic theorems 31

lt is easily seen that it again satisfies the inequality

1 L f(x) dx l � L l f(x) l dx.
The set, denoted
L'. 1 (A; C),
of ail Lebesgue-integrable functions f : A --+ C is clearly a vector space over C. The relation
"f = g almost everywhere in A" again defines an equivalence relation R over the space
L'. 1 (A; q, and the quotient set

is also a vector space over C.

Finally, note that, for brevity, we shall often omit the dependence on x E A, by simply
letting
L L
f dx := f(x) dx if f E L 1 (A) or if f E L 1 (A; C).
The next theorems recapitulate the most fundamental properties of the Lebesgue integral.
Note in this respect that the order in which these properties can be established may in effect
depend on the way the Lebesgue integral has been defined. The first three theorems list basic
convergence properties of sequences of integrable functions.
Theorem 1.15-1 ( Beppo Levi monotone convergence theorem) Let A be a measur­
able subset of IRn and let (fk)� 1 be a sequence of functions fk E L'. 1 (A) with the property
that
f
0 � fi � · · · � fk � fk+i � · · · a. e. in A and klim fk (x) dx < oo.
-->oo }A
1
Then there exists a function f E L'. (A) such that
fk (x) --+ f(x) for almost all x E A and L l fk(x) - f(x) l dx --+ 0 as k --+ oo.
In particular then, limk-->oo fk(x) < oo for almost all x E A. D
Theorem 1.15-2 ( Fatou's lemma) Let A be a measurable subset of IRn and let (fk )� 1 be
a sequence of measurable functions fk : A --+ IR with the property that
fk ;::: 0 a. e. in A.
Then
}{ lim
A
( )
inf fk(x) dx � lim
k-->oo inf }{ fk(x) dx,
k -->oo A
where the right-hand side, or both sides, of this inequality may be equal to oo . D
Theorem 1.15-3 ( Lebesgue dominated convergence theorem ) Let A be a measurable
subset oflRn and let (fk)� 1 be a sequence offunctions fk E L'. 1 (A), resp. fk E L'. 1 (A; C), such
that
f(x) := klim
-->oo fk(x) exists for almost all x E A,
32 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

and such that there exists a function g E C 1 (A) with the property that
l fk (x)I ::::; g (x) for all k � 1 and almost all x E A.
Then f E C 1 (A), resp. f E C 1 (A; C), and
L l fk (x) - f(x) l dx -t 0 as k -t oo.
In particular then,
r f(x) dx = lim r fk(x) dx. 0
}A k-too }A
Let B be a measurable subset of Rn and let A denote the O'-algebra formed by ail the
Lebesgue measurable subsets of B. Given a function f E C 1 (B), let
ll(A) : = L f(x) dx for each A E A.
Hence l ll(A)I ::::; JA l f(x) I dx ::::; f8 l f(x)I dx < oo for each A E A. Then it is clear that
li
the function : A -t R defined in this fashion possesses the following properties: first, it is
a signed measure, in the sense that
ll( 0) = 0,
ll ( Q Ai ) = � ll(Ai ) if Ai E A, i � 1, are such that Ai n Aj = 0 if i -:/= j

(this countably additive property easily follows from the Lebesgue dominated convergence
theorem); second, it is absolutely continuons with respect to the Lebesgue measure dx, in
the sense that
A E A and dx- meas A = 0 imply ll (A ) = O.
Remarkably, the converse property holds:
Theorem 1.15-4 (Radon-Nikodym theorem ) Let B be a measurable subset of Rn , let
A denote the O'-algebra formed by all the measurable subsets of B, and let : A -t R be li

a signed measure that is absolutely continuous with respect to the Lebesgue measure. Then
there exists a function f E C 1 (B) such that
ll ( A ) = L f(x) dx for each A E A. 0

The next theorem gives a fundamental criterion of Lebesgue-integrability of a function

defined over a product of measurable sets in Rm x Rn , as well as a way of computing the
Lebesgue integral
rr f (x , y ) dxdy
jJA xB
of a Lebesgue-integrable function f : A x B -t [ - oo, oo] , where dx dy denotes the Lebesgue
measure on Rm x Rn .
Sect. 1.16] Change of variable in Lebesgue integrals in Rn 33

Theorem 1.15-5 Let A be a measurable subset of Rm , let B be a measurable subset of Rn ,

and let f : A x B C Rm x Rn --+ [ -oo, oo] be a measurable function.
(a) (Tonelli's theorem) For each x E A, the junction f (x, . ) : y E B --+ f (x, y) E
[ -oo, oo] is measurable and, for each y E B, the function f( . , y) : x E A --+ f(x, y) E [-oo, oo]
is measurable. Besides, the function f is integrable over A x B, i. e.,

frlAr x B l f(x, y) l dx dy < OO ,

if and only if one of the following two conditions is satisfied:

L (l l f(x, y) l dy) dx < oo,

l (l I f (x, y) I dx) dy < oo.
(b) (Fubini's theorem) If the function f is Lebesgue-integrable on A x B, the function
f( . , y) : A -+ [-oo , oo] is integrable for almost all y E B, the function f(x, · ) : B --+ [-oo , oo]
is integrable for almost all x E A, and the Lebesgue integral of f on A x B is given by

fl x B f(x , y) dx dy = L (l f(x , y) dy ) dx = l(L f(x , y) dx) dy . D

1 . 16 Change of variable in Lebesgue integrals in JRn

In this section, we examine how a Lebesgue integral defined over an open subset Rn is trans­
formed under a change of variable; this means that the open set is the image cp (O) of another
open subset n of Rn under a mapping cp = (cpi ) f= 1 : n --+ Rn , the variable y E cp (O) being
replaced by the variable X E 0 in the process.
In what follows, the notation \lep designates the n x n matrix field defined by (Vcp) ij =
8jcpi , 1 � i, j � n, where 8j denotes the partial derivative operator with respect to the jth
variable.
Theorem 1.16-1 (injective change of variable in Lebesgue integrals in Rn ) Let n be
an open subset of Rn and let cp : n --+ Rn be a continuously differentiable injective mapping.
Then a function f : cp (O) --+ R is Lebesgue-integrable on cp (O) if and only if the function
X E n --+ f(cp (x)) l det V cp (x) I E R
is Lebesgue-integrable on n. If this is the case, then

1ip (f!)
f(y) dy = r f(cp (x)) l det V cp (x) I dx .
ln
D

Remarks ( 1 ) Under the assumptions of Theorem 1. 16-1, the set cp(O) is automatically open
(hence Lebesgue-integrability on cp(O) makes sense): this is a consequence of the deep Brouwer in­
variance of domain theorem in IRn (Theorem 9.17-3), which in fact holds even if cp : n � IRn is only
assumed to be continuous.
34 Real Analysis and Theory of Punctions: A Quick Review [Ch. 1

(2) That f be real-valued is not a restrictive assumption since a Lebesgue-integrable function is

necessarily finite almost everywhere. D

While the case where the mapping <p is injective (as in Theorem 1.16-1) is considered in
many texts, the case where <p is not injective (as in the next theorem) is not often treated. 10
Theorem 1.16-2 (noninjective change of variable in Lebesgue integrals in Rn ) Let
n be an open subset of Rn and let <p : n -+ Rn be a continuously differentiable mapping such
that the image <p(f!) is open. For each y E <p(f!), let
card <p - 1 (y) cardinal of the set <p - 1 (y) if <p - 1 (y) is finite,
:=

card <p - 1 (y) := oo if <p - 1 (y) is infinite.

Then, given a function f : <p(f!) -+ R, the function f card <p - 1 : <p(f!) -+ R is Lebesgue­
integrable on <p(f!) if and only if the function
X E n -+ f(<p(x)) l det V' <p(x)I E R
is Lebesgue-integrable on n. If this is the case, then
f
J'P(n)
f(y) card <p 1 (y ) dy
-
= [
ln
f(<p(x)) l det V'<p(x) I dx. D

Remarks ( 1 ) As expected, Theorem 1. 16-1 is a special case of Theorem 1. 16-2.

(2) By contrast with Theorem 1 . 16- 1 , it must now be assumed that ip(il) is open in Theorem
1 . 16-2. D

1.17 Volumes , areas , and lengths in ]Rn

The n-volume, or simply, the volume, of a measurable subset A of Rn , denoted dx-meas A,
or simply meas A, is by definition the Lebesgue measure of A; in other words, .
=
dx- meas A meas A := i dx,
where dx denotes the n-dimensional Lebesgue measure.
Thanks to the formula for change of variables in Lebesgue integrals (Theorem 1.16-1 or
1. 16-2), one can compute the volume of n-parallelepipeds (these particular subsets of Rn are
defined in the next theorem):
Theorem 1.17-1 (volume of an n-parallelepiped) The volume of an n-parallelepiped
in Rn , i. e., a subset Pof Rn of the form

10 See, however, RADO & REICHELDERFER [1955], SCHWARTZ [1993b, Corollary 6.2. 14], FEDERER [1969] , or
SMITH [1983, Chapter 16] .
Random documents with unrelated
content Scribd suggests to you:
de la terrasse, deux ours blancs, empaillés et débout, étaient en
faction et tenaient des torches dans leurs puissantes pattes.

Quand le comte et ses invités eurent descendu les marches

recouvertes de fourrures d'ours, ils entrèrent dans une large allée
d'arbres verts transformés en autant d'arbres de Noël. Sur chaque
branche étaient plantées de petites bougies en porcelaine d'où
jaillissaient des flammes de gaz. On s'avançait comme dans un bois
féerique, à travers un océan de lumière, sur de molles peaux de
rennes qui recouvraient la terre glacée. L'air, embaumé de senteurs
résineuses, était rempli de légers nuages roses.

Au bout de l'allée s'étendait un étang considérable, dont les bords

étaient également garnis de peaux. Sur sa brillant surface,
solidement gelée, s'élevait un petit temple bâti en blocs de glace,
comme le célèbre palais construit sur la Néwa du temps de la
czarine Anne. Dans ce temple, sur un autel élevé, se dressait une
Vénus de glace, couronnée de fleurs. Tout autour du temple allaient
et venaient joyeusement les patineurs et deux traîneaux attelés, l'un
de rennes, l'autre de grands chiens. Le premier était dirigé par un
Esquimau, le second par un Kamtschadale. Un choeur de chanteurs,
composé d'ours blancs installés dans une tribune de bois toute
revêtue de branches de sapin, accompagnait de ses airs les plus
agréables les ébats des masques sur la glace, pendant qu'un cordon
de dauphins de glace, qui encadraient l'étang et vomissaient sans
relâche du pétrole enflammé, éclairait ce tableau d'une lumière
magique et faisait de temps en temps briller le petit temple comme
un édifice de diamants aux mille feux.

Pendant que la musique et les voix aux joyeux éclats produisaient

un aimable chaos, de petites huttes de Kamtschadales, construites
en peaux, disséminées dans les fourrés voisins et agréablement
chauffées, invitaient les couples amoureux à de paisibles et
charmants rendez-vous.

Entouré, entraîné par les masques folâtres, le comte avait été

séparé des Oginski. Il découvrir tout à coup Dragomira qui seule se
trouvait aussi sur la rive de l'étang et promenait ses regards au loin
sur la foule, comme si elle cherchait quelqu'un.

"Vous avez perdu votre cavalier, dit Soltyk en s'approchant d'elle,

puis-je vous offrir mes services?"

Dragomira prit sans façon le bras du comte qui lui montra le

temple en souriant.

"Votre image, dit-il à voix basse.

- En quoi?

- Vous aussi, vous êtes une Vénus de glace.

- Ah! cher comte, ne savez-vous pas combien la glace fond

rapidement quand vient le printemps?

- Oui, certes, répondit Soltyk; mais ce printemps, dont la chaude

haleine doit vous vaincre, où est-il?

- Je ne le connais que par ouï-dire, ce grand enchanteur auquel

tout coeur doit céder, dit Dragomira avec un fin sourire.

- Et cet enchanteur, c'est l'amour?

- Oui.
 - Mais vous n'êtes pas capable d'aimer.

- Je le crois presque moi-même.

- Vous n'avez pas de coeur.

- Si… mais un coeur de glace!

- Oh! si je pouvais l'échauffer? murmura Soltyk avec un regard

d'où semblaient jaillir des flammes.

- Vous?"

Dragomira le regarda bien en face.

"Vous ne savez que vous jouer des femmes, et je ne suis pas un

jouet."

Le comte se mordit les lèvres; au même moment Anitta approchait

et la conversation prit fin. Dragomira prit le bras d'Anitta; puis toute
les deux retournèrent dans l'antichambre pour ôter leurs fourrures et
se perdirent ensuite dans le tourbillon des danseurs.

"Il sera à moi, se disait Dragomira, dès que je le voudrai; il ne me

semble pas bien difficile à conquérir; mais il s'agit ici de quelque
chose de plus; aussi la ruse et la prudence doivent donner la main à
la coquetterie. La résistance paraît le séduire et lui troubler la tête
plus que tout le reste. Pauvre comte! J'ai bien facilement l'avantage
sur lui, puisque je n'éprouve rien pour lui."

Au milieu de ses réflexions, elle aperçut Zésim, qui était là, appuyé
à une colonne. Il lui vint aussitôt une idée badine, et elle profita du
moment où un danseur emmenait Anitta, pour se glisser comme un
serpent, vite et sans faire aucun bruit, hors de la salle.

Dans le corridor, près des vestiaires, se trouvaient aussi quelques

petits cabinets, disposés pour ceux qui voudraient se masquer
pendant la fête. Dragomira fit signe à Barichar qui était avec les
autres domestiques et gardait un grand panier. Mais au moment où
elle allait entrer dans un de ces cabinets, deux bras souples
l'enlacèrent presque tendrement et les yeux bleus d'Henryka la
regardèrent avec un sourire malicieux.

"Enfin! Je vous tiens, s'écria l'aimable jeune fille, et maintenant

vous ne m'échapperez pas.

- Si, répondit Dragomira en souriant, car j'ai une petite intrigue en

tête, et vous ne voudriez certainement pas me gâter cet innocent
plaisir.

- Vous vous masquez?

- Oui.

- Oh! je ne vous trahirai pas, continua Henryka, permettez-moi de

vous accompagner et de vous aider."

Toutes les deux entrèrent dans le cabinet. Quand Barichar fut parti
après avoir déposé son panier dans un coin, Henryka ferma la porte.
Dragomira s'était assise devant la table de toilette et commença à
ôter sa parure pendant qu'Henryka enlevait le contenu du panier
avec des cris d'admiration enfantine. Quand ce fut fini, elle
s'approcha de Dragomira, et, debout devant elle, se mit à la
considérer avec un intérêt extraordinaire.
 "Je ne sais ce qu'ont les gens, dit-elle, ils vous trouvent tous
énigmatique; et Anitta pense même que vous avez quelque chose
d'inquiétant. Moi, au contraire, je me sens une grande sympathie
pour vous.

- Prenez garde, répondit Dragomira, vous découvrirez peut-être à

la fin sous cette robe un corps de serpent ou une queue de poisson.

- Vous n'êtes pas non plus une créature ordinaire, continua

Henryka; je sens qu'une puissance mystérieuse vous entoure, mais
ce sentiment ne fait qu'augmenter encore l'attrait magique qui
m'entraîne vers vous. Faites de moi votre alliée; je vous aimerai
comme une soeur et je vous écouterai comme une écolière docile.

- Réellement?"

Dragomira tourna lentement la tête vers elle et la regarda d'un

oeil interrogateur.

"Conduisez-moi, je vous suivrai comme une aveugle, sans peur et

sans aucune réflexion, répondit Henryka.

- Nous verrons.

- Aujourd'hui, permettez-moi de vous aider.

- Pourquoi non? répondit tranquillement Dragomira, le premier pas

dans la voie de la lumière éternelle que vous voyez devant vous par
un pieux pressentiment, c'est l'humilité; servez-moi donc."

Henryka s'agenouilla devant Dragomira et lui baisa les mains, puis

elle lui ôta ses chaussures et lui mit les pantoufles turques brodées
d'or qu'elle avait tirées du panier. Dragomira se laissa faire avec la
majestueuse indifférence d'une souveraine.

XXVI

SOUS LE MASQUE

On peut déraisonner sur un point et être sage pour tout le reste.

WIELAND.

Quelques instants après, une sultane, habillée avec toute la

magnificence de l'Orient entrait dans la salle.

Grande et d'une taille élancée, elle s'avançait avec dignité. Elle

était chaussée de babouches de velours rouge brodées d'or, et avait
un large pantalon et une jupe courte de satin jaune sur laquelle
tombait un long caftan de soie bleu-clair, brodé d'argent et garni
d'hermine. Ce caftan laissait voir une veste ouverte de velours
rouge; la poitrine couverte de colliers de corail, de perles et de
sequins apparaissait à travers une gaze d'argent. La tête fière de la
sultane était couronnée d'un petit turban tout garni de pierreries. Au
lieu de masque elle avait un voile épais de harem, au travers duquel
on ne pouvait distinguer que de grands yeux bleus et froids, au
regard dominateur.

Une troupe de messieurs s'était attachée aux pas de la nouvelle

arrivée. Plus d'un se risqua à lui chuchoter à l'oreille quelque
compliment; mais elle semblait insensible à toutes les tentatives que
l'on faisait pour attirer son attention.
 Elle promena longtemps ses regards pénétrants par toute la salle,
jusqu'à ce qu'elle eût découvert celui quelle cherchait. Il venait
d'aller au buffet, sans intention, comme un automate inconscient
que fait marcher un mouvement d'horlogerie. Les domestiques lui
offraient divers rafraîchissements; il secouait la tête et était sur le
point de s'en aller, lorsque la sultane entra et lui posa sa pette main
sur l'épaule.

"Je te salue, Zésim Jadewski, dit-elle, pourquoi donc baisses-tu

ainsi la tête, aujourd'hui?

- Je n'ai guère de motifs d'être joyeux.

- Il y a bien des moyens de chasser les soucis, en voici justement

un des meilleurs."

La belle sultane prit un verre de vin sur le buffet, y trempa ses

lèvres et le présenta à Zésim.

"Que me donnes-tu? Un doux poison, un philtre?

- J'arriverais trop tard.

- A ta santé!"

Zésim vida le verre.

"Maintenant, un deuxième moyen.

- Lequel?

- Fais-moi la cour.

- je n'en aurais pas le talent.

 - Parce que tu aimes?

- Peut-être.

- Il y a ici deux dames à qui tu as donné ton coeur. A laquelle

appartient-il maintenant?

- Tu me questionnes comme un inquisiteur."

La sultane se mit à rire, tout doucement, mais ce rire argentin

suffit à la trahir.

"Maintenant je te connais."

Elle rit de nouveau.

"Tu es Dragomira."

Une petite main saisit rapidement la sienne et un souffle doux et

tiède effleura sa joue.

"Ne me trahis pas; on nous observe; le comte Soltyk est là; je

veux lui parler et lui faire peur."

En effet, le comte se tenait à l'entrée, et ses yeux sombres, pleins

d'une flamme diabolique, étaient arrêtés sur la belle personne, qui
murmurait coquettement à l'oreille de Zésim. L'envie et la jalousie
bouleversaient le coeur de Soltyk et faisaient bouillonner son sang
indomptable. En même temps, d'autres yeux se dirigeaient vers le
couple occupé à chuchoter, mais ceux-là étaient timides, tristes et
pleins d'angoisse. C'était Anitta qui avait aussi reconnu Dragomira et
qui tremblait pour son bien-aimé.
 La sultane avait déjà congédié Zésim et se préparait à aller trouver
Soltyk, lorsque le jésuite la prévint et entraîna rapidement le comte
avec lui.

"Qu'avez-vous? demanda Soltyk.

- Il faut que je vous avertisse, lui dit tout bas le P. Glinski; la

sultane est Mlle Maloutine. Avez-vous vu comme elle échangeait
avec ce jeune officier des poignées de main et des paroles tout à fait
tendres?

- Après, après?

- Vous êtes au moment de tomber dans les filets d'une coquette.

- Cette fois votre connaissance des hommes fait fausse route,

reprit le comte d'un ton railleur, elle est au contraire froide comme
glace.

- Mais je sais que Jadewski va chez elle.

- Sessawine aussi.

- Et elle se joue de tout le monde.

- Tant mieux.

- Il n'y a pas moyen de vous sauver, je le vois.

- Si les abîmes de l'enfer étaient aussi beaux que cette Dragomira,

cher Père, le ciel resterait vide et vous-même finiriez par rendre
votre âme au diable."
 Soltyk le quitta en riant et se mit aussitôt à la recherche de la
sultane qui avait brusquement disparu dans le tourbillon des
masques. Il la trouva à l'entrée de la petite salle qui figurait l'Asie.
Elle semblait l'attendre.

"C'est ici ton empire, dit-il en s'inclinant devant elle; ton esclave
peut-il entrer avec toi?"

Il releva la portière et la suivit dans le petit salon décoré avec

toute la somptuosité de l'Orient.

Des tentures persanes d'une rare magnificence, brodées d'or et

d'argent, tombaient en plis larges et lourds et figuraient les parois, le
plafond, les fenêtres et les portes d'un pavillon dont le sommet était
formé par un croissant d'or constellé de pierreries. Le sol de cette
mystérieuse retraite était couvert d'un tissu de l'Inde, blanc et
souple comme du duvet; le pied s'y enfonçait comme dans la neige
nouvellement tombée. Une seule lampe, à globe rouge, était
suspendue au plafond comme un rubis lumineux d'une grosseur
fabuleuse. Cà et là étaient des coussins qui invitaient au repos, à la
rêverie, à l'amour. Un parfum étrange et subtil embaumait l'air et
troublait les sens comme une caresse.

Dragomira s'assit sur le divan placé au milieu du pavillon aux

couleurs chatoyantes. Elle était sur une peau de panthère, et ses
pieds reposaient sur la tête majestueuse d'un tigre.

Le comte restait debout devant elle, dans toute l'ardente extase

de la passion.

"Vous m'avez attendu? dit-il enfin.

 - Oui.

- Vous savez que j'ai quelque chose à vous dire?

- Oui.

- Et vous êtes disposée à m'entendre?

- Oui.

- Je vous remercie. Vous me rendez le courage qui commençait à

me manquer.

- Il faut donc du courage pour causer avec une jeune fille?

- Avec vous, oui, Dragomira.

- Dragomira? moi? vous vous trompez.

- Comment! me tromper? interrompit le comte Soltyk; qui pourrait

jamais vous avoir vue et ne pas vous reconnaître entre mille? Qui
pourrait avoir vu le regard de vos yeux et l'oublier? Qui pourrait ne
pas le découvrir, même sous le masque? Oui, c'est vous, Dragomira,
vous, avec toute votre puissance, votre froideur, votre cruauté!

- Moi, cruelle? parce que je ne vous crois pas? Je ne suis pas

cruelle; je suis un peu prudente, voilà tout.

- Qu'avez-vous contre moi?

- Rien.

- En ce moment, vous ne dites pas la vérité.

 - Si; je ne puis pas dire que quoi que [ce] soit me déplaise en
vous.

- Oui, mais vous vous défiez de moi."

Un léger sourire fut la réponse de Dragomira.

"Et pourquoi vous défiez-vous de moi?

- Ah! l'innocent! Avez-vous oublié ce que vous avez fait? La liste

des péchés de Don Juan à côté de la vôtre est la confession d'un
écolier."

Soltyk sourit.

"Je connais ma réputation, dit-il, mais je vous donne ma parole

d'honneur que la renommée a bien exagéré.

- Bien; mais en ôtant ce qu'il y a de trop, dit Dragomira, je crois

qu'il en reste encore assez pour rendre votre canonisation
invraisemblable.

- Je ne suis pas un saint; je n'ai jamais prétendu à cette gloire.

- Mais faut-il être le contraire?

- Que suis-je donc?

- Un scélérat, répondit Dragomira. Vous aimez Anitta et vous me

faites la cour.

- On veut me marier avec Mlle Oginska, voilà tout.

 - Tactique de jésuite. On veut unir deux familles puissantes et faire
de vous un instrument politique.

- Vous pouvez bien avoir raison, murmura Soltyk, surpris au plus

haut point de cette remarque, mais je ne suis pas bon à faire un
instrument.

- Alors vous n'aimez pas Anitta?

- Non."

Le comte était encore debout devant Dragomira; il s'assit alors sur

un divan, auprès d'elle, de façon à avoir un genou en terre, et il lui
saisit les mains en lui disant:

"Je vous aime!"

Dragomira rit de nouveau.

"Vous pouvez rire, je vous aime pourtant, et je vous jure que vous
êtes la première que j'aime. Jusqu'à présent je n'ai connu que des
fantaisies passagères, parfois un court enivrement, mais mon coeur
était libre, et surtout ma tête. Ce que j'éprouve en face de vous, je
le ressens pour la première fois. Je ne suis pas exalté, je ne suis pas
amoureux, je ne suis pas du tout ivre de votre beauté. J'ai le
sentiment que vous avez été créée pour moi, que votre âme est de
la même essence que la mienne, que la vie sans vous n'a aucune
valeur, et que la vie à côté de vous serait le paradis. Si ce n'est pas
là de l'amour qu'est-ce donc?"

Pendant qu'il parlait, les yeux de Dragomira s'attachaient sur son

beau et mâle visage.
 "Pauvre comte! dit-elle en relevant lentement la manche de son
caftan, mais, en vérité, je commence à croire que vous m'aimez.

- Et vous me plaignez, s'écria Soltyk avec animation, parce que

vous ne pouvez pas répondre à cet amour.

- Je ne vous aime pas…

- Parce qu'un autre possède votre coeur?

- Quelle impatience! Ne m'interrompez pas.

- Alors, je vous demande en grâce…

- Je ne vous aime pas, mais mon coeur est encore libre; essayez
de le conquérir. De tous ceux qui y prétendent vous êtes le seul qui
ne me déplaise pas."

Elle avait détaché une petite chaîne d'or qui entourait son beau
bras et elle jouait avec.

"Vous me permettez donc d'espérer?

- Oui.

- Oh! que je suis heureux!"

Le comte avait saisi ses mains et les couvrait de baisers. Elle le

laissa faire pendant quelque temps, puis elle retira une de ses mains
et lui passa la petite chaîne autour du bras.

"Que faites-vous? Voulez-vous faire de moi votre chevalier?

 - Non, mon esclave. Vous voyez bien que je vous mets à la
chaîne."

Cependant un domino rose s'était approché de Zésim.

"Quoi! seul! lui dit-il; où est l'enchanteresse qui t'a mis dans ses
fers?

- De qui parles-tu? Je suis encore libre, répliqua Zésim.

- N'essaye pas de me tromper, tu n'y réussirais pas, continua le

domino; il n'y a déjà pas si longtemps, tu as juré à une autre que tu
l'aimais. L'aurais-tu si vite oubliée, si un nouvel astre ne s'était pas
levé sur ta vie?

- Qui es-tu?… Zésim parcourut du regard cette taille élancée, saisit

les mains de l'inconnue, qui tressaillit, et les retint fortement en
cherchant à lire dans ses yeux sombres.

- Non, ce n'est pas possible, murmura-t-elle enfin; je me suis

trompé.

- Lâche-moi, dit le domino en suppliant.

- Pas encore; j'ai une autre question à t'adresser.

- Eh bien?

- Qui t'a envoyée?

- Personne.

- Alors, dans quelle intention viens-tu?

 - Pour t'avertir. Un danger te menace.

- Un danger?… De la part de qui?

- De la part de celle que tu aimes.

- Si tu veux que je te crois, dit Zésim ému, dis m'en davantage,

dis-moi tout ce que tu sais."

Les yeux sombres se reposèrent un instant sur lui avec une

expression presque douloureuse.

"Soit, mais ce n'est pas ici le lieu. Tu entendras bientôt parler de

moi."

Les mains tremblantes se dégagèrent d'un mouvement énergique,

et le domino à la taille élancée comme celle d'une jeune fille disparut
rapidement au milieu du tourbillon de la fête.

DEUXIEME PARTIE
I

CIEL ET ENFER

… Belle comme la première femme, la pécheresse, séduite par le

mauvais serpent, qui depuis n'a cessé de tromper, en étant trompée
elle-même. LORD BYRON
 Deux jours après la fête du comte Soltyk; qui occupa longtemps
encore toutes les sociétés de la ville, Zésim reçut une lettre sans
signature. On lui donnait rendez-vous dans la même église où il avait
eu son dernier entretien avec Anitta.

Il pensa immédiatement à elle. Sans aucun doute c'était elle qui

voulait l'avertir; mais sa conversation avec le domino lui avait inspiré
de la défiance, et il lui vint encore à l'esprit une autre pensée. Si
Dragomira avait des vues sérieuses sur le comte, et cherchait à
l'intimider, lui Zésim, au moyen d'une personne de confiance,
uniquement parce qu'il était devenu tout à coup gênant?

Ce qu'il y avait d'énigmatique dans l'existence et les relations de

Dragomira était pour lui une source d'inquiétudes toujours nouvelles;
il ne pouvait parvenir à avoir en elle confiance pleine et entière. Il la
croyait, quand il la voyait; il doutait d'elle, dès qu'elle était loin.

Quand le jour commença à baisser, Zésim se rendit à l'église

indiquée. Devant la porte, il lui vint une nouvelle idée. Si Dragomira
voulait seulement l'éprouver; si elle l'attendait elle-même?

Il hésita une minute, puis entra rapidement, bien décidé à mettre

une fin à tous ses doutes.

L'église paraissait vide. Mais quand il s'approcha du maître-autel, il

vit une dame agenouillée qui se releva au bruit de ses pas et vint à
sa rencontre.

"Je vous remercie d'être venu, dit-elle en lui tendant la main.

- Est-ce possible? C'est vous, Anitta? murmura Zésim.

- C'est moi", répondit-elle avec tristesse, et elle écarta son voile.

 Zésim regarda avec émotion son visage sérieux et pâli.

"J'ai peur pour vous, Zésim, dit-elle. Je ne sais pas ce que c'est, et
je suis incapable de vous dire quelque chose de précis, mais, je le
sens, un grand danger vous menace. Dragomira a quelque
mystérieuse mission à accomplir; c'est une voix intérieure, un
sombre pressentiment qui me le dit. Est-elle affiliée à une
conspiration? appartient-elle à une secte de fanatiques? Je ne peux
pas le découvrir; mais je sais qu'elle a jeté ses filets de votre côté et
que vous deviendrez sa victime, et je ne réussis pas vous sauver.

- Vous voyez les choses beaucoup trop en noir; je connais la

famille, la mère de Dragomira…

- Qu'est-ce que cela peut prouver? Il y a des sociétés secrètes,

des sectes religieuses fanatiques qui cherchent précisément des
adhérents et des instruments dans le monde le plus distingué; et,
croyez-moi, Dragomira est un de ces instruments.

- C'est possible; mais qu'importe que je périsse, puisque vous ne

m'aimez pas, Anitta?

- Ne blasphémez pas, Zésim.

- Dragomira ne peut pas me trahir plus que vous.

- Elle vous poussera à la mort, s'écria Anitta. O Zésim! Ayez pitié

de moi! Ayez pitié de votre mère! Au nom de cet amour qui remplit
mon coeur, tout mon être…"

Elle s'arrêta; les larmes étouffaient sa voix; elle ne pouvait plus

que lever vers lui les yeux et les mains avec une expression
suppliante.
 "Comment dois-je vous comprendre? dit Zésim amèrement. Quelle
valeur ma vie peut-elle encore avoir pour la future comtesse Soltyk.

- Jamais je ne donnerai ma main au comte.

- Vous lui êtes pourtant fiancée.

- Qui vous l'a dit? Il m'a demandée et je l'ai refusé.

- Anitta! Est-ce vrai? mon Dieu! pourquoi ne me dites-vous cela

qu'aujourd'hui?

- Je vous ai juré de vous rester fidèle.

- Vous avez raison; le coupable, c'est moi, continua Zésim, je ne

vous ai pas cru tant de fermeté. Une vanité puérile m'a poussé à
renoncer à un trésor dont la possession ne me paraissait pas
assurée; je ne voulais pas être trahi par vous et alors c'est moi qui
vous ai trahie.

- Je ne vous en veux pas, murmura Anitta en lui prenant la main,

je vous ai pardonné. Dites-moi seulement de quelle façon je pourrai
vous sauver. Ce n'est pas votre amour que je veux; il ne s'agit que
de votre vie.

- Ce sont des imaginations.

- Non, non. Je vous en supplie, brisez vos liens.

- Je ne peux pas; il est trop tard.

- Dites donc plutôt que vous ne voulez pas, que Dragomira vous a
complètement aveuglé, que votre passion pour cette créature
sinistre est plus forte que vous.
 - Vous vivez dans un monde romanesque, dit Zésim en souriant;
les dangers que vous voyez, vous les avez tout bonnement vus en
rêve. Je vous assure que la réalité est loin d'avoir un aspect si
terrible. Dragomira est sincère et loyale envers moi.

- Vous le croyez.

- Si cela peut vous tranquilliser, je vous promets d'être prudent.

- Oui, la prudence d'un somnambule! s'écria Anitta; je le vois,

vous êtes tout à fait aveugle, et ce serait inutile de persister à vous
avertir. J'y renonce, mais je vous protègerai, Zésim, malgré vous-
même. J'accepte la lutte avec Dragomira et Dieu ne me refusera pas
son assistance.

- Je ne vous comprends pas, Anitta; comment en êtes-vous

arrivée à ces idées fantastiques?

- Il n'y a là rien de fantastique, dit-elle d'un ton sérieux et résolu,

je suis une jeune fille toute simple, qui vous aime, et c'est tout.
Adieu et soyez sur vos gardes.

- Vous reverrai-je, Anitta?

- A quoi bon? Maintenant, non. Plus tard peut-être… quand vous

aurez - brisé vos chaînes. Adieu."

Zésim lui baisa la main et elle partit en hâte. Il resta immobile

quelques instants, abîmé dans ses pensées, sous ces voûtes
sombres.

Qu'était-ce donc que ce mystère dans lequel une volonté

étrangère emprisonnait Dragomira? se demandait-il. Elle en était
convenue elle-même et Anitta l'avait pénétrée; Qui étaient ces
autres qui la menaient et l'employaient comme un instrument?
Appartenait-elle à une secte et à laquelle? Pourquoi se défiait-elle, et
pourquoi ne pouvait-il la quitter, s'il doutait d'elle? L'aimait-il
véritablement autant que cela? Et Anitta? Est6il possible d'aimer
deux femmes en même temps? "Tu es le lien des deux natures qui
se sont unies dans l'espace et dans le temps", chante Derschavine
dans son ode à Dieu. Ces deux natures si souvent en désaccord se
combattaient aussi en lui. L'une l'élevait vers la lumière, vers Anitta,
l'autre l'entraînait dans cette obscurité sinistre où Dragomira vivait et
régnait. Pensées contradictoires, émotions, projets, tout se croisait
dans sa tête, dans son coeur, et il n'aboutissait à aucune résolution,
à aucun acte. En ce moment encore, il ne savait à quoi s'en tenir.
Les flots le poussaient en avant et il se demandait de nouveau où il
allait.

Une heure après le départ d'Anitta, Bassi Rachelles se glissait déjà

dans la chambre de Dragomira pour l'informer du rendez-vous des
deux jeunes gens.

"Tu es sûre que c'était lui? demanda Dragomira.

- Le lieutenant Jadewski, aussi vrai que je suis ici.

- Et de quoi ont-ils parlé?

- De vous, noble maîtresse.

- De moi?

- Elle l'a averti de se tenir sur ses gardes, mais il n'a pas ajouté foi
à ses paroles.
 - Et n'ont-ils pas parlé d'amour?

- Non. Seulement, quand elle est partie, il lui a demandé s'il la

reverrait, et elle a répondu: "A quoi on? Maintenant, non."

- Bien, tu peux t'en aller."

Immédiatement après le départ de la Juive, Dragomira écrivit deux

lettres, l'une au comte, signée des initiales de son nom, l'autre à
Zésim, sans signature, avec une écriture contrefaite. Elle leur
donnait rendez-vous à tous les deux à l'Opéra. Barichar se chargea
personnellement de la lettre adressée à Soltyk, et confia à un facteur
juif celle qui était destinée à Zésim.

Le comte était eu théâtre avant le commencement de la

représentation, et attendait avec impatience au pied de l'escalier qui
conduisait aux loges. Son regard effleurait à peine les amis et les
dames élégantes qui arrivaient. Mais lorsqu'il aperçut Dragomira à
l'entrée du vestibule, son coeur se mit à battre avec impétuosité, et
ses yeux restèrent fixés comme par l'effet d'un charme sur cette
taille souple et élancée, sur cette tête entourée et illuminée de
cheveux blonds.

Celle que Soltyk attendait avec une si ardente impatience était

venue accompagnée de Cirilla qui s'était habillée avec un luxe à
l'ancienne mode et représentait fort bien une dame de la noblesse
de campagne. Soltyk se contenta d'ôter son chapeau, de saluer
profondément et de dévorer des yeux Dragomira. Celle-ci de son
côté lui fit un petit signe de tête avec une amabilité pleine d'aisance
et passa devant lui comme devant une simple connaissance.
 Zésim, qui était assis au parquet, vit Dragomira entrer dans sa
loge et ôter son manteau de théâtre, tout brodé d'or scintillant. Elle
resta debout un instant contre le rebord, et tous les regards se
dirigèrent sur elle. En même temps le comte la contemplait avec une
admiration muette.

"Où a-t-elle appris, pensait-il, à s'habiller ainsi? Je sais pourtant

qu'elle n'a pas été à Paris."

Et, en effet, Dragomira était ravissante dans sa robe de soie

brochée couleur héliotrope, richement garnie de dentelles jaune-
pâle. La parure, merveilleusement simple, consistait en un petit
bouquet de violettes naturelles, placé dans ses cheveux d'or et un
autre attaché à son corsage.

Après le premier acte Zésim voulut lui rendre visite, mais le comte
le prévint. Avec une fureur concentrée le jeune et bouillant officier le
vit entrer dans la loge et porter à ses lèvres la main que Dragomira
lui tendait en souriant. La conversation animée qui s'établit ensuite
entre Dragomira et Soltyk augmenta de minute en minute le supplice
de Zésim.

"Que se passe-t-il donc en moi? se demandait-il; je crois que je

suis jaloux."

Tous les doutes qu'Anitta avait remués en lui, toutes les sombres
pensées que d'ordinaire un regard de Dragomira domptait et
endormait, se réveillèrent et reprirent leur puissance.

Il crut qu'il allait étouffer, il sortit de l'atmosphère chaude et

suffocante de la salle pour aller respirer l'air frais; puis il rentra, mais
il ne reprit pas as première place. Il se mit derrière une colonne de
parterre; de là, il pouvait mieux observer Dragomira. Il espérait que
le comte la quitterait au commencement de l'acte suivant, mais il
avait eu tort d'espérer. Soltyk resta, et la conversation devint de plus
en plus animée, de plus en plus intime. Ce ne fut qu'au moment où
le rideau se levait pour la troisième fois que le comte la salua, et
partit. Zésim monta l'escalier en courant et entra dans la loge de
Dragomira, les joues rouges et les yeux enflammés.

Elle n'eut pas l'air de remarquer son agitation. Elle lui tendit
gaiement les deux mains avec un mouvement d'une grâce exquise.

"Pourquoi si tard? lui demanda-t-elle; tu n'as donc pas reçu mon

billet?

- Tu m'as écrit?

- Sans doute."

Il sortit le billet doux anonyme… "Cette lettre…

- Est de moi; un badinage… Je voulais te surprendre, me faire

bien belle et te tourner un peu la tête.

- Je suis ici depuis le commencement.

- Est-ce possible? dit Dragomira d'un air innocent. Je ne t'ai pas

remarqué."

Zésim lui adressa un regard moitié fâché, moitié reconnaissant, et

porta sa main froide à ses lèvres brûlantes. Cependant, elle célébra
son triomphe avec un sourire silencieux. Le bien-aimé lui appartenait
de nouveau, et n'appartenait qu'à elle.
II

LA ROUTE DU PARADIS

Même quand je marcherais par la vallée de l'ombre de la mort, je

ne craindrais aucun mal; car tu es avec moi, Seigneur. PSAUM. XXIII,
4.

Une visite inattendue. Dragomira, la calme, la froide, la

courageuse, ne put réprimer un tressaillement lorsque Barichar lui
présenta la carte du P. Glinski. Elle se remit pourtant aussitôt et cria:
"Entrez!"

Barichar ouvrit la porte, et le jésuite s'approcha avec sa plus

élégante révérence et son plus gracieux sourire.

"J'ai peur de vous importuner, dit-il, pendant que Dragomira

s'asseyait sur un divan, et, d'un geste vraiment royal de sa main,
l'invitait à prendre place près d'elle, mais l'intérêt qui m'amène est si
sérieux, si important, pour ne pas dire si sacré, que j'ose compter
sur votre pardon. Il s'agit du bonheur de mon cher comte, de celui
que j'ai élevé, de celui que je considère comme mon enfant."

Le P. Glinski fit une pause; il attendait une question, une objection

qui lui eût facilité le moyen d'arriver au véritable but de sa visite.
Mais Dragomira ne vint nullement à son aide; elle le regardait, au
contraire, avec une certaine indifférence distraite qui semblait dire:
"En quoi votre comte peut-il m'intéresser?"

Le P. Glinski se passa la main droite sur la main gauche, puis la

main gauche sur la main droite.
 "Vous devinez bien, noble demoiselle, dit-il, de quoi il s'agit?

- Non, je n'en ai aucune idée, répondit Dragomira avec une

candeur qui déconcerta un instant Glinski, le fin diplomate de l'ordre
de Jésus.

- Je voulais… oui… Avant tout, il faut que je vous fasse mon

compliment, quoique j'arrive un peu tard. L'autre jour vous étiez
superbe en sultane."

Dragomira sourit.

"Je vous suis bien obligée, dit-elle, mais vous n'êtes pas venu chez
moi, mon révérend père, pour me faire cette communication?

- Non, certainement, non, murmura le jésuite. J'ai seulement

voulu faire la remarque que mon cher comte, lui aussi, semblait ravi
de vous.

- C'est vrai, il m'a beaucoup fait la cour, dit Dragomira très

naturellement.

- Alors, je ne me suis pas trompé, continua le P. Glinski; certes, on

comprend très bien que le comte vous adresse ses hommages et
que cet innocent triomphe vous soit agréable; mais ce qui vous fait
plaisir à tous les deux prépare à d'autres des chagrins, de
l'inquiétude, à moi particulièrement, à moi qui aime le comte comme
un fils et qui ne veux que son bonheur.

- Maintenant, je ne vous comprends pas, mais pas du tout, c'est

comme si vous me parliez une langue étrangère.
 - Vous savez, pourtant, ma noble demoiselle, que le comte est
fiancé.

- Oui, sans doute.

- Que cette alliance entre deux familles si honorables est désirée

par tout le pays.

- Oui, je le sais aussi.

- Alors, pourquoi vous mettez-vous si cruellement en travers de

nos beaux projets?

- Moi! Dragomira leva ma tête et se mit à rire. Je n'y pense pas.

- Vous souffrez toutefois que le comte vous adresse ses

hommages.

- Puis-je le lui défendre? Je serais tout simplement ridicule. Tant

qu'il ne fait rien qui, d'après l'opinion du monde, soit blâmable ou
inconvenant, je suis désarmée en face de lui.

- Vous détournez la question, répliqua Glinski; je suis sûr que vous

encouragez le comte.

- Pas le moins du monde.

- Je vous en prie, mademoiselle, restons dans le sujet. Je n'ai pas

à engager une dispute de mots. Ce serait un malheur pour nous tous
si le mariage du comte et de Mlle Oginski n'avait pas lieu; et en ce
moment vous êtes un obstacle à ce mariage. Je ne m'y trompe pas;
voilà où en sont les choses; aussi, je vous supplie de renoncer au
comte.
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com