Botelho Fs Functional Analysis Calculus of Variations and Nu

Functional Analysis, Calculus
of Variations and Numerical

Methods for Models in
Physics and Engineering
Fabio Silva Botelho, Adjunct Professor
Department of Mathematics
Universidade Federal de Santa Catarina
Florianópolis, SC, Brazil
p,
p,
A SCIENCE PUBLISHERS BOOK
A SCIENCE PUBLISHERS BOOK
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2021 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 20200427
International Standard Book Number-13: 978-0-367-35674-3 (Hardback)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the
validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the
copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to
publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let
us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted,
or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, includ-
ing photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers,
MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety
of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment
has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
and the CRC Press Web site at

http://www.crcpress.com
Preface
The first part of this book develops in details the basic tools of functional analysis, measure and integration
and Sobolev spaces theory, which will be used in the development of the subsequent chapters. The results are
established in a rigorous and precise fashion, however we believe the proofs presented are relatively easy to
follow, since all steps are developed in a very transparent and clear way.
Of course, in general the results in this first part are not new and may be found in similar format in many
other excellent books on functional analysis, such as those by Bachman and Narici [6], Brezis [24], Rudin
[67] and others.
In many cases we give our own version of some proofs we believe the original ones are not clear enough,
always aiming to improve their clarity and preciseness.
In the second book part, we develop the basic concepts on the calculus of variations, convex analysis,
duality theory and constrained optimization in a Banach spaces context.
We start with a more basic approach on the calculus of variations which gradually evolves to the
development of more advanced results on convex analysis and optimization. The text includes a formal proof
of the Lagrange multiplier theorem, in a first step, for the equality-restriction case and in the subsequent
sections we address the general case of constrained optimization also in a Banach spaces context.
Finally, in the third part we address the development of several important examples which apply the
previous theoretical approaches exposed.
Applications include duality principles for non-linear plate and shell models, duality principles for
problems in non-linear elasticity, duality for shape optimization models, existence and duality for the
Ginzburg-Landau system in superconductivity, duality for a semi-linear model in micro-magnetism and
others.
About such applications we highlight the results on the Generalized Method of Lines, which is a
numerical method for which the domain of the partial differential equation in question is discretized in lines,
and the concerning solution is written on these lines as function of the boundary conditions and boundary
shape. Some important models are addressed and among them we highlight the applications of such a method
to the Navier-Stokes system.
Also as an application of the Generalized Method of Lines, we finish the book with a chapter which
presents a numerical procedure for the solution of an inverse optimization problem.
At this point, it is worth emphasizing that the Chapter 21 is co-authored by myself and my colleague
Alexandre Molter and Chapter 23 is co-authored by myself and my graduate student Eduardo Pandini Barros,
to whom I would like to express my gratitude for their contributions.
As a final note we highlight that the content of the present book overlaps something with the ones of my
previous books ‘‘Topics on Functional Analysis, Calculus of Variations and Duality” published by Academic
Publications, Sofia, 2011 and, ‘‘Functional Analysis and Applied Optimization in Banach Spaces” published
by Springer in 2014, but only relating a part on basic standard mathematics, so that, among some other new
chapters, the applications presented from Chapters 17 and 30 are almost all new developments.
Acknowledgements
I would like to thank some people of Virginia Tech—USA, where I got my Ph.D. degree in Mathematics in
2009. I am especially grateful to Professor Robert C Rogers for his excellent work as advisor. I would like to
thank the Department of Mathematics for its constant support and this opportunity of studying mathematics
in advanced level. Among many other Professors, I particularly thank Martin Day (Calculus of Variations),
William Floyd and Peter Haskell (Elementary Real Analysis), James Thomson (Real Analysis), Peter
Linnell (Abstract Algebra) and George Hagedorn (Functional Analysis) for the excellent lectured courses.
Finally, special thanks to all my Professors at I.T.A. (Instituto Tecnológico de Aeronáutica, SP-Brasil) my
undergraduate and masters school. Specifically about I.T.A., among many others I would like to express
my gratitude to Professors Leo H Amaral, Tânia Rabelo and my master thesis advisor Antônio Marmo de
Oliveira, also for their valuable work.
January 2020 Fabio Silva Botelho
Florianópolis, SC, Brazil
Contents
Preface iii
Acknowledgements iv
SECTION I: FUNCTIONAL ANALYSIS
1. Metric Spaces 1
1.1 Introduction 1
1.2 The main definitions 1
1.2.1 The space l ∞ 2
1.2.2 Discrete metric 2
1.2.3 The metric space s 3
1.2.4 The space B(A) 3
1.2.5 The space l p 3
1.2.6 Some fundamental definitions 6
1.2.7 Properties of open and closed sets in a metric space 7
1.2.8 Compact sets 10
1.2.9 Separable metric spaces 13
1.2.10 Complete metric spaces 14
1.3 Completion of a metric space 15
1.4 Advanced topics on compactness in metric spaces 19
1.5 The Arzela-Ascoli theorem 24
2. Topological Vector Spaces 26

2.1 Introduction 26
2.2 Vector spaces 26
2.3 Some properties of topological vector spaces 29
2.3.1 Nets and convergence 31
2.4 Compactness in topological vector spaces 34
2.4.1 A note on convexity in topological vector spaces 36
2.5 Normed and metric spaces 38
2.6 Linear mappings 40
2.7 Linearity and continuity 41
2.8 Continuity of operators in Banach spaces 42
vi < Functional Analysis, Calculus of Variations and Numerical Methods for Models in Physics and Engineering
2.9 Some classical results on Banach spaces 43

2.9.1 The Baire Category theorem 43
2.9.2 The Principle of Uniform Boundedness 45
2.9.3 The Open Mapping theorem 46
2.9.4 The Closed Graph theorem 48
2.10 A note on finite dimensional normed spaces 48
3. Hilbert Spaces 52
3.1 Introduction 52
3.2 The main definitions and results 52
3.3 Orthonormal basis 57
3.3.1 The Gram-Schmidt orthonormalization 59
3.4 Projection on a convex set 60
3.5 The theorems of Stampacchia and Lax-Milgram 63
4. The Hahn-Banach Theorems and the Weak Topologies 66

4.1 Introduction 66
4.2 The Hahn-Banach theorems 66
4.3 The weak topologies 72
4.4 The weak-star topology 76
4.5 Weak-star compactness 77
4.6 Separable sets 81
4.7 Uniformly convex spaces 84
5. Topics on Linear Operators 86

5.1 Topologies for bounded operators 86
5.2 Adjoint operators 87
5.3 Compact operators 92
5.4 The square root of a positive operator 93
6. Spectral Analysis, a General Approach in Normed Spaces 100

6.1 Introduction 100
6.2 Sesquilinear functionals 103
6.3 About the spectrum of a linear operator defined on a banach space 113
6.4 The spectral theorem for bounded self-adjoint operators 117
6.4.1 The spectral theorem 122
6.5 The spectral decomposition of unitary transformations 125
6.6 Unbounded operators 127
6.6.1 Introduction 127
6.7 Symmetric and self-adjoint operators 130
6.7.1 The spectral theorem using Cayley transform 132
7. Basic Results on Measure and Integration 136

7.1 Basic concepts 136
7.2 Simple functions 137
7.3 Measures 138
7.4 Integration of simple functions 139
7.5 Signed measures 142
7.6 The Radon-Nikodym theorem 145
Contents < vii
7.7 Outer measure and measurability 149

7.8 Fubini’s theorem 156
7.8.1 Product measures 156
8. The Lebesgue Measure in Rn 161

8.2 Properties of the outer measure 161
8.3 The Lebesgue measure 165
8.4 Outer and inner approximations of Lebesgue measurable sets 166
8.5 Some other properties of measurable sets 169
8.6 Lebesgue measurable functions 171
9. Other Topics in Measure and Integration 178

9.1 Some preliminary results 178
9.2 The Riesz representation theorem 183
9.3 The Lebesgue points 191
9.3.1 The main result on Lebesgue points 193
10. Distributions 195

10.1 Basic definitions and results 195
10.2 Differentiation of distributions 198
10.3 Examples of distributions 199
10.3.1 First example 199
10.3.2 Second example 199
10.3.3 Third example 200
11. The Lebesgue and Sobolev Spaces 201

p
11.1 Definition and properties of L spaces 201
11.1.1 Spaces of continuous functions 206
11.2 The Sobolev spaces 208
11.3 The Sobolev imbedding theorem 210
11.3.1 The statement of Sobolev imbedding theorem 210
11.4 The proof of the Sobolev imbedding theorem 211
11.4.1 Relatively compact sets in L p(Ω) 215
11.4.2 Some approximation results 218
11.4.3 Extensions 222
11.4.4 The main results 227
11.5 The trace theorem 235
11.6 Compact imbeddings 238
11.7 On the regularity of Laplace equation solutions 245
11.7.1 Regularity, a more general result 249
SECTION II: CALCULUS OF VARIATIONS, CONVEX ANALYSIS

AND RESTRICTED OPTIMIZATION
12. Basic Topics on the Calculus of Variations 252

12.1 Banach spaces 252
12.2 The Gâteaux variation 258
12.3 Minimization of convex functionals 260
viii < Functional Analysis, Calculus of Variations and Numerical Methods for Models in Physics and Engineering
12.4 Sufficient conditions of optimality for the convex case 263

12.5 Natural conditions, problems with free extremals 265
12.6 The du Bois-Reymond lemma 271
12.7 Calculus of variations, the case of scalar functions on Rn 274
12.8 The second Gâteaux variation 276
12.9 First order necessary conditions for a local minimum 277
12.10 Continuous functionals 278
12.11 The Gâteaux variation, the formal proof of its formula 279
13. More Topics on the Calculus of Variations 282

13.1 Introductory remarks 282
13.2 The Gâteaux variation, a more general case 283
13.3 Fréchet differentiability 284
13.4 The Legendre-Hadamard condition 284
13.5 The Weierstrass condition for n = 1 286
13.6 The Weierstrass condition, the general case 288
13.7 The Weierstrass-Erdmann conditions 292
13.8 Natural boundary conditions 294
14. Convex Analysis and Duality Theory 296

14.1 Convex sets and functions 296
14.2 Weak lower semi-continuity 296
14.3 Polar functionals and related topics on convex analysis 298
14.4 The Legendre transform and the Legendre functional 301
14.5 Duality in convex optimization 304
14.6 The min-max theorem 307
14.7 Relaxation for the scalar case 313
14.8 Duality suitable for the vectorial case 320
14.8.1 The Ekeland variational principle 320
14.9 Some examples of duality theory in convex and non-convex analysis 323
15. Constrained Variational Optimization 328

15.1 Basic concepts 328
15.2 Duality 331
15.3 The Lagrange multiplier theorem 332
15.4 Some examples concerning inequality constraints 338
15.5 The Lagrange multiplier theorem for equality and inequality constraints 338
15.6 Second order necessary conditions 342
15.7 On the Banach fixed point theorem 344
15.8 Sensitivity analysis 346
15.8.1 Introduction 346
15.9 The implicit function theorem 346
15.9.1 The main results about Gâteaux differentiability 350
16. On Central Fields in the Calculus of Variations 356

16.2 Central fields for the scalar case in the calculus of variations 356
16.3 Central fields and the vectorial case in the calculus of variations 360
Contents < ix
SECTION III: APPLICATIONS TO MODELS IN

PHYSICS AND ENGINEERING
17. Global Existence Results and Duality for Non-Linear Models of Plates and Shells 367
17.2 On the existence of a global minimizer 369
17.3 Existence of a minimizer for the plate model for a more general case 372
17.4 The main duality principle 375
17.5 Existence and duality for a non-linear shell model 381
17.6 Existence of a minimizer 384
17.7 The duality principle for the shell model 386
17.8 Conclusion 390
18. A Primal Dual Formulation and a Multi-Duality Principle for a Non-Linear Model of Plates 391
18.2 The first duality principle 393
18.3 The primal dual formulation and related duality principle 399
18.4 A multi-duality principle for non-convex optimization 402
18.5 Conclusion 408
19. On Duality Principles for One and Three-Dimensional Non-Linear Models in Elasticity 409
19.2 The main duality principle for the one-dimensional model 410
19.3 The primal variational formulation for the three-dimensional model 414
19.4 The main duality principle for the three-dimensional model 415
19.5 Conclusion 419
20. A Primal Dual Variational Formulation Suitable for a Large Class of Non-Convex 420
Problems in Optimization
20.2 The main duality principle 424
20.3 Conclusion 426
21. A Duality Principle and Concerning Computational Method for a Class of Optimal 427
Design Problems in Elasticity
Fabio Silva Botelho and Alexandre Molter
21.2 Mathematical formulation of the topology optimization problem 428
21.3 About the computational method 431
21.4 Computational simulations and results 432
21.5 Final remarks and conclusions 435
22. Existence and Duality Principles for the Ginzburg-Landau System in Superconductivity 437
22.2 The main result 440
22.3 A second multi-duality principle 446
22.4 Another duality principle for global optimization 452
x < Functional Analysis, Calculus of Variations and Numerical Methods for Models in Physics and Engineering
22.5 The existence of a global solution for the Ginzburg-Landau system in the presence 456
of a magnetic field
22.6 Duality for the complex Ginzburg-Landau system 461
22.7 Conclusion 464
23. Existence of Solution for an Optimal Control Problem Associated to the 465
Ginzburg-Landau System in Superconductivity
Fabio Silva Botelho and Eduardo Pandini Barros
23.2 An existence result for a related optimal control problem 466
23.3 A method to obtain approximate numerical solutions for a class of partial 471
differential equations
23.4 Conclusion 475
24. Duality for a Semi-Linear Model in Micro-Magnetism 476

24.2 The duality principle for the semi-linear model 477
24.3 Conclusion 481
25. About Numerical Methods for Ordinary and Partial Differential Equations 482
25.2 On the numerical procedures for Ginzburg-Landau type ODEs 482
25.3 Numerical results for related P.D.E.s 485
25.3.1 A related P.D.E on a special class of domains 485
25.3.2 About the matrix version of G.M.O.L. 486
25.3.3 Numerical results for the concerning partial differential equation 488
25.4 A proximal algorithm for optimization in Rn 488
25.4.1 The main result 489
25.5 Conclusion 497
26. On the Numerical Solution of First Order Ordinary Differential Equation Systems 498
26.2 The main results 498
26.3 Numerical results 501
26.4 The Newton’s method for another first order system 504
26.4.1 An example in nuclear physics 507
26.5 Conclusion 511
27. On the Generalized Method of Lines and its Proximal Explicit and Hyper-Finite 512
Difference Approaches
27.1.1 Some preliminaries results and the main algorithm 514
27.1.2 A numerical example, the proximal explicit approach 516
27.2 The hyper-finite differences approach 520
27.2.1 The main algorithm 520
27.2.2 A numerical example 522
27.3 Conclusion 525
Contents < xi
28. On the Generalized Method of Lines Applied to the Time-Independent Incompressible 526
Navier-Stokes System
28.2 On the solution of the time-independent incompressible Navier-Stokes system through 527
an associated linear one
28.3 The generalized method of lines for the Navier-Stokes system 529
28.3.1 Numerical examples through the generalized method of lines 539
28.4 Conclusion 545
29. A Numerical Method for an Inverse Optimization Problem through the Generalized 547
Method of Lines
29.2 The mathematical description of the main problem 547
29.3 About the generalized method of lines and the main result 548
29.3.1 The numerical results 552
29.4 Conclusion 553
30. A Variational Formulation for Relativistic Mechanics based on Riemannian Geometry 554
and its Application to the Quantum Mechanics Context
30.2 Some introductory topics on vector analysis and Riemannian geometry 556
30.3 A relativistic quantum mechanics action 560
30.3.1 The kinetics energy 561
30.3.2 The energy part relating the curvature and wave function 562
30.4 Obtaining the relativistic Klein-Gordon equation as an approximation of the 565
previous action
30.5 A note on the Einstein field equations in the vacuum 567
30.6 Conclusion 568
References 569
Index 573
SECTION I
FUNCTIONAL ANALYSIS
Chapter 1
Metric Spaces
1.1 Introduction
In this chapter we present the main basic definitions and results relating to the concept of metric spaces.
We may recall that any Banach space is a metric one, so that the framework here introduced is
suitable for a very large class of spaces.
The main references for this chapter are [14, 82].
1.2 The main definitions

We begin this section by presenting the metric definitions and some concerning examples.
Definition 1.2.1 (Metric space) Let V be a non-empty set. We say that V is a metric space as it is
possible to define a function d : V ×V → R+ = [0, +∞) such that
6 v and d(u, u) = 0, ∀u, v ∈ V.
1. d(u, v) > 0 if u =
2. d(u, v) = d(v, u), ∀u, v ∈ V.
3. d(u, w) ≤ d(u, v) + d(v, w), ∀u, v, w ∈ V.
Such a function d is said to be a metric for V , so that the metric space in question is denoted by
(V, d).
Example 1.2.2 V = R is a metric space with the metric d : R × R → R+ , where
d(u, v) = |u − v|, ∀u, v ∈ R.

2 Functional Analysis, Calculus of Variations and Numerical Methods for Models in Physics and Engineering
Example 1.2.3 V = R2 is a metric space with the metric d : V ×V → R+ , where

q
d(u, v) = (u1 − v1 )2 + (u2 − v2 )2 , ∀u = (u1 , u2 ), v = (v1 , v2 ) ∈ R2 .
Example 1.2.4 V = Rn is a metric space with the metric d : V ×V → R+ where

q
d(u, v) = (u1 − v1 )2 + · · · + (un − vn )2 , ∀u = (u1 , . . . , un ), v = (v1 , . . . , vn ) ∈ Rn .
Example 1.2.5 V = C([a, b]), where C([a, b]) is the metric space of continuous functions u : [a, b] → R
with the metric d : V ×V → R+ , where
d(u, v) = max {|u(x) − v(x)|} ≡ ku − vk∞ , ∀u, v ∈ V.

x∈[a,b]
Example 1.2.6 V = C([a, b]), is a metric space with the metric d : V ×V → R+ where
Z b
d(u, v) = |u(x) − v(x)| dx, ∀u, v ∈ V.
a
1.2.1 The space l ∞

In this subsection we shall define some important classes of metric spaces.
The first definition presented is about the l ∞ space of sequences.
Definition 1.2.7 We define the space l ∞ as
l ∞ = {u = {un }n∈N : un ∈ C and there exists M > 0 such that |un | < M, ∀n ∈ N} .
A metric for l ∞ may be defined by
d(u, v) = sup{|u j − v j |},

j∈N
where u = {un } e v = {vn } ∈ l ∞ .
1.2.2 Discrete metric

At this point we introduce the definition of discrete metric.
Definition 1.2.8 Let V be a non-empty set. We define the discrete metric for V by

0, if u = v,
d(u, v) = (1.1)
6 v.
1, if u =
In such a case we say that (V, d) is a discrete metric space.
Exercise 1.2.9 Let V = R and let d : V ×V → R be defined by

p
d(u, v) = |u − v|.
Show that d is a metric for V .

Metric Spaces 3
1.2.3 The metric space s

In the next lines we define one more metric space of sequences, namely, the space s.
Definition 1.2.10 (The metric space s) We define the metric space s as s = (V, d), where
V = {u = {un }, : un ∈ C, ∀n ∈ N},
with the metric

∞
1 |un − vn |
d(u, v) = ∑ 2n (1 + |un − vn |) ,
n=1
∀u = {un } and v = {vn } ∈ V.
Exercise 1.2.11 Show that this last function d is indeed a metric.
1.2.4 The space B(A)

Another important metric space is the space of bounded functions defined on a set A, and denoted by
B(A).
Definition 1.2.12 Let A be a non-empty set and define
B(A) = {u : A → R, such that there exists M > 0 such that |u(x)| < M, ∀x ∈ A}.
B(A) is said to be the space of bounded functions defined on A.
Exercise 1.2.13 Show that B(A) is a metric space with the metric
d(u, v) = sup{|u(x) − v(x)|}.

x∈A
1.2.5 The space l p

Finally, one of the most important metric space of sequences is the l p one, whose definition is presented
in the next lines.
Definition 1.2.14 Let p ≥ 1, p ∈ R.

We define the space l p by
( )
∞
lp = u = {un } : un ∈ C and ∑ |un | p < ∞
n=1
with the metric

!1/p
∞
p
d(u, v) = ∑ |un − vn | ,
n=1
where u = {un } and v = {vn } ∈ l p .

At this point we shall show that d is indeed a metric.

Let p > 1, p ∈ R. Let q > 1 be such that
1 1
+ = 1,
p q
that is
p
q= .
p−1
Let x, y ≥ 0, x, y ∈ R.
We are going to show that
1 p 1 q
xy ≤ x + y .
p q
Observe that if x = 0 or y = 0 the inequality is immediate. Thus, suppose x > 0 and y > 0.
Fix y > 0 and define
1 1
h(x) = x p + yq − xy, ∀x > 0.
p q
Observe that
h0 (x) = x p−1 − y
and
h00 (x) = (p − 1)x p−2 > 0, ∀x > 0.
Therefore h is convex and its minimum on (0, +∞) is attained through the equation
h0 (x) = x p−1 − y = 0,
that is, at x0 = y1/(p−1) .

Hence,
min h(x) = h(x0 )

x∈(0,+∞)
1 1
= (x0 ) p + yq − x0 y
p q
1 p/(p−1) 1
= y − y1/(p−1) y + yq
p q
1
= (1/p − 1) yq + yq
q
1 q 1 q
= − y + y
q q
= 0. (1.2)
Thus,
1 p 1 q
h(x) = x + y − xy ≥ h(x0 ) = 0, ∀x > 0.
p q
Therefore, since y > 0 is arbitrary, we obtain
1 p 1 q
xy ≤ x + y , ∀x, y > 0.
p q
Metric Spaces 5
so that
1 p 1 q
xy ≤ x + y , ∀x, y ≥ 0. (1.3)
p q
Let u = {un } ∈ l p and v = {vn } ∈ l q .
Denote !1/ p
∞
kuk p = ∑ |un | p
n=1
and !1/q
∞
kvkq = ∑ |vn |q .
n=1
Define also ( )
u un
û = = ,
kuk p (∑∞ p 1/p
n=1 |un | )
and ( )
v vn
v̂ = = .
kvkq (∑∞ q 1/q
n=1 |vn | )
From this and (1.3) we obtain,
∞
1 ∞ 1 ∞
∑ |ûn v̂n | ≤ ∑ |ûn | p + ∑ |v̂n |q
n=1 p n=1 q n=1
1 1
= +
p q
= 1. (1.4)
Thus,
∞
∑ |un vn | ≤ kuk p kvkq , ∀u ∈ l p , v ∈ l q .
n=1
¨
This last inequality is well known as the Holder one.
Exercise 1.2.15 Prove the Minkowski inequality, namely
ku + vk p ≤ kuk p + kvk p , ∀u, v ∈ l p .
Hint
∞
ku + vk pp = ∑ |un + vn | p
n=1
∞
≤ ∑ |un + vn | p−1 (|un | + |vn |). (1.5)
n=1
Apply the Holder

¨ inequality to each part of the right hand side of the last inequality.
Use such an inequality to prove the triangle inequality concerning the metrics definition.
Prove also the remaining properties relating to the metric definition and conclude that d : l p × l p →
+
R , where
d(u, v) = ku − vk p , ∀u, v ∈ l p
is indeed a metric for the space l p .
1.2.6 Some fundamental definitions

Definition 1.2.16 (neighborhood) Let (U, d) be a metric space. Let u ∈ U and r > 0. We define the
neighborhood of center u and radius r, denoted by Vr (u), by
Vr (u) = {v ∈ U | d(u, v) < r}.
Definition 1.2.17 (limit point) Let (U, d) be a metric space and E ⊂ U. A point u ∈ U is said to be a
limit point of E if for each r > 0 there exists v ∈ Vr (u) ∩ E such that v 6= u.
We shall denote by E 0 the set of all limit points of E.
Example 1.2.18 U = R2 , E = Br (0). Thus E 0 = Br (0).
Remark 1.2.19 In the next definitions U shall denote a metric space with a metric d.
Definition 1.2.20 (Isolated point) Let u ∈ E ⊂ U. We say that u is an isolated point of E if it is not a
limit point of E.
Example 1.2.21
U = R2 , E = B1 ((0, 0)) ∪ {(3, 3)}. Thus (3, 3) is an isolated point of E.
Definition 1.2.22 (Closed set) Let E ⊂ U and let E 0 be the set of limit points of E. We say that E is
closed if E ⊃ E 0 .
Example 1.2.23
Let U = R2 and r > 0, thus E = Br ((0, 0)) is closed.
Definition 1.2.24 A point u ∈ E ⊂ U is said to be an interior point of E if there exists r > 0 such that
Vr (u) ⊂ E, where
Vr (u) = {v ∈ U | d(u, v) < r}.
Example 1.2.25
For U = R2 , let E = B1 ((0, 0)) ∪ {(3, 3)}, for example u = (0.25, 0.25) is an interior point
p of E, in fact,
for r = 0.5, if v ∈ Vr (u) then d(u, v) < 0.5 so that d(v, (0, 0)) ≤ d((0, 0), u) + d(u, v) ≤ 1/8 + 0.5 < 1
that is, v ∈ B1 ((0, 0)) and thus Vr (u) ⊂ B1 ((0, 0)). We may conclude that u is an interior point of
B1 ((0, 0)). In fact all points of B1 ((0, 0)) are interior.
Definition 1.2.26 (Open set) E ⊂ U is said to be open if all its points are interior.
Example 1.2.27
For U = R2 , the ball B1 (0, 0) is open.
Definition 1.2.28 Let E ⊂ U, we define its complement, denoted by E c , by:
E c = {v ∈ U | v 6∈ E}.
Definition 1.2.29 A set E ⊂ U is said to bounded if there exists M > 0 such that
sup{d(u, v) | u, v ∈ E} ≤ M.
Definition 1.2.30 A set E ⊂ U is said to be dense in U if each point of U is either a point of E or it is

a limit point of E, that is, U = E ∪ E 0 .
Metric Spaces 7
Example 1.2.31
The set Q is dense in R. Let u ∈ R and let r > 0. Thus, from a well known result in elementary analysis
there exists v ∈ Q such that u < v < u + r, that is, v ∈ Q ∩Vr (u) and v 6= u, where Vr (u) = (u − r, u + r).
Therefore u is a limit point of Q. Since u ∈ R is arbitrary, we may conclude that R ⊂ Q0 , that is, Q is
dense in R.
Theorem 1.2.32 Let (U, d) be a metric space. Let u ∈ U and r > 0. Then Vr (u) is open.
Proof 1.1 First we recall that

Vr (u) = {v ∈ U | d(u, v) < r}.
Let v ∈ Vr (u). We have to show that v is an interior point of Vr (u). Define r1 = r − d(u, v) > 0. We shall show
that Vr1 (v) ⊂ Vr (u).
Let w ∈ Vr1 (v), thus d(v, w) < r1 . Hence
d(u, w) ≤ d(u, v) + d(v, w) < d(u, v) + r1 = r.
Therefore w ∈ Vr (u), ∀w ∈ Vr1 (v), that is Vr1 (v) ⊂ Vr (u), so that we may conclude that v is an interior point
of Vr (u), ∀v ∈ Vr (u), thus, Vr (u) is open. The proof is complete.
Theorem 1.2.33 Let u be a limit point of E ⊂ U, where (U, d) is a metric space. Then each neighborhood
of u has an infinite number of points of E, distinct from u.
Proof 1.2 Suppose to obtain contradiction, that there exists r > 0 such that Vr (u) has a finite number of
points of E distinct from u. Let {v1 , ..., vn } be such points of Vr (u) ∩ E distinct from u. Choose 0 < r1 <
min{d(u, v1 ), d(u, v2 ), ...., d(u, vn )}. Hence Vr1 (u) ⊂ Vr (u) and vi 6∈ Vr1 (u), ∀i ∈ {1, 2, ..., n}. Therefore either
Vr1 (u) ∩ E = {u} or Vr1 ∩ E = 0, / which contradicts the fact that u is a limit point of E.
The proof is complete.
Corollary 1.2.34 Let E ⊂ U be a finite set. Then E has no limit points.
1.2.7 Properties of open and closed sets in a metric space

In this section we present some basic properties of open and closed sets.
Proposition 1.2.35 Let {Eα , α ∈ L} be a collection of sets. Then
(∪α ∈L Eα )c = ∩α∈L Eαc .
Proof 1.3 Observe that
u ∈ (∪α∈L Eα )c ⇔ u 6∈ ∪α∈L Eα
⇔ u 6∈ Eα , ∀α ∈ L
⇔ u ∈ Eαc , ∀α ∈ L
⇔ u ∈ ∩α∈L Eαc . (1.6)
Exercise 1.2.36
Prove that
(∩α ∈L Eα )c = ∪α∈L Eαc .
Theorem 1.2.37 Let (U, d) be a metric space and E ⊂ U. Thus, E is open if and only if E c is closed.
Proof 1.4 Suppose E c is closed. Choose u ∈ E, thus u 6∈ E c and therefore u is not a limit point E c . Hence
there exists r > 0 such that Vr (u) ∩ E c = 0/. Hence, Vr (u) ⊂ E, that is, u is an interior point of E, ∀u ∈ E, so
that E is open.
Reciprocally, suppose E is open. Let u ∈ (E c )0 . Thus for each r > 0 there exists v ∈ Vr (u) ∩ E c such that
v=6 u, so that
Vr (u) * E, ∀r > 0.
Therefore u is not an interior point of E. Since E is open we have that u 6∈ E, that is, u ∈ E c . Hence (E c )0 ⊂ E c ,
that is, E c is closed.
Corollary 1.2.38 Let (U, d) be a metric space, F ⊂ U is closed if and only if F c is open.
Theorem 1.2.39 Let (U, d) be a metric space.

1. If Gα ⊂ U and Gα is open ∀α ∈ L, then
∪α∈L Gα
is open.
2. If Fα ⊂ U and Fα is closed ∀α ∈ L, then

∩α∈L Fα
is closed.
3. If G1 , ..., Gn ⊂ U and Gi is open ∀i ∈ {1, ..., n}, then
∩ni=1 Gi
is open.
4. If F1 , ..., Fn ⊂ U and Fi is closed ∀i ∈ {1, ..., n}, then
∪ni=1 Fi
is closed.
Proof 1.5
1. Let Gα ⊂ U, where Gα is open ∀α ∈ L. Let u ∈ ∪α ∈L Gα . Thus u ∈ Gα0 for some α0 ∈ L. Since Gα0 is
open, there exists r > 0 such that Vr (u) ⊂ Gα0 ⊂ ∪α ∈L Gα . Hence, u is an interior point, ∀u ∈ ∪α ∈L Gα .
Thus ∪α∈L Gα is open.
2. Let Fα ⊂ U, where Fα is closed ∀α ∈ L. Thus, Fαc is open ∀α ∈ L. From the last item, we have ∪α∈L Fαc
is open so that
∩α∈L Fα = (∪α∈L Fαc )c
is closed.
3. Let G1 , ..., Gn ⊂ U be open sets. Let
u ∈ ∩ni=1 Gi .
Thus,
u ∈ Gi , ∀i ∈ {1, ..., n}.
Since Gi is open, there exists ri > 0 such that Vri (u) ⊂ Gi .
Metric Spaces 9
Define r = min{r1 , ..., rn }. Hence, Vr (u) ⊂ Vri (u) ⊂ Gi , ∀i ∈ {1, ..., n} and therefore
Vr (u) ⊂ ∩ni=1 Gi .
This means that u is an interior point of ∩ni=1 Gi , and being u ∈ ∩ni=1 Gi arbitrary we obtain that ∩ni=1 Gi
is open.
4. Let F1 , ..., Fn ⊂ U be closed sets. Thus, F1c , ..., Fnc are open. Thus, from the last item, we obtain:
∩ni=1 Fic
is open, so that
∪ni=1 Fi = (∩ni=1 Fic )c
is closed.
Exercise 1.2.40
Let (U, d) be a metric space and let u0 ∈ U. Show that A = {u0 } is closed. Let B = {u1 , ..., un } ⊂ U. Show
that B is closed.
Definition 1.2.41 (Closure) Let (U, d) be a metric space and let E ⊂ U. Denote the set of limit points of E
by E 0 . We define the closure of E, denoted by E, by:
E = E ∪ E 0.
Examples 1.2.42
1. Let U = R2 , E = B1 (0, 0), we have that E 0 = B1 (0, 0), so that in this example E = E ∪ E 0 = E 0 .
2. Let U = R, A = {1/n : n ∈ N}, we have that A0 = {0}, and thus A = A ∪ A0 = A ∪ {0}.
Theorem 1.2.43 Let (U, d) be a metric space and E ⊂ U. Thus,

1. E is closed.
2. E = E ⇔ E is closed.
3. If F ⊃ E and F is closed, then F ⊃ E.
Proof 1.6
c
1. Observe that E = E ∪ E 0 . Let u ∈ E . Thus u 6∈ E and u 6∈ E 0 (u is not a limit point of E). Therefore,
/ that is, Vr (u) ⊂ E c , thus, u is an interior point of E c .
there exists r > 0 such that Vr (u) ∩ E = 0,
We shall prove that Vr (u) ∩ E = 0/. Let v ∈ Vr (u) and define r1 = r − d(u, v) > 0. We shall show that
Vr1 (v) ⊂ Vr (u).
Let w ∈ Vr1 (v), thus d(v, w) < r1 and therefore
d(u, w) ≤ d(u, v) + d(v, w) < d(u, v) + r1 = r,
that is, w ∈ Vr (u). Hence,

Vr1 (v) ⊂ Vr (u),
c c
and thus v is not a limit point of E, that is, v ∈ E , ∀v ∈ Vr (u). Thus, Vr (u) ⊂ E which means that u is
c c
an interior point of E , so that E is open, and hence E is closed.
2. Observe that E ⊂ E = E ∪ E 0 . Suppose that E is closed. Thus E ⊃ E 0 , that is E ⊃ E ∪ E 0 = E. Hence

E = E. Suppose E = E. From the last item E is closed, and thus E is closed.
3. Let F be a closed set such that F ⊃ E. Thus, F 0 ⊃ E 0 .
Hence
F = F = F ∪ F 0 ⊃ E ∪ E 0 = E.
1.2.8 Compact sets

Definition 1.2.44 (Open covering) Let (U, d) be a metric space. We say that a collection of sets {Gα , α ∈
L} ⊂ U is an open covering of A ⊂ U if
A ⊂ ∪α∈L Gα
and Gα is open, ∀α ∈ L.
Definition 1.2.45 (Compact set) Let (U, d) be a metric space and K ⊂ U. We say that K is compact if each
open covering {Gα , α ∈ L} of K admits a finite sub-covering. That is, if K ⊂ ∪α ∈L Gα , and Gα is open
∀α ∈ L, then there exist α1 , α2 , ..., αn ∈ L such that K ⊂ ∪ni=1 Gαi .
Theorem 1.2.46 Let (U, d) be a metric space. Let K ⊂ U where K is compact. Then K is closed.
Proof 1.7 Let us show that K c is open. Let u ∈ K c . For convenience, let us generically denote in this proof
Vr (u) = V (u, r).
For each v ∈ K we have d(u, v) > 0. Define rv = d(u, v)/2. Thus,
V (u, rv ) ∩V (v, rv ) = 0/, ∀v ∈ K. (1.7)
Observe that
∪v∈K V (v, rv ) ⊃ K.
since K is compact, there exist v1 , ..., vn ∈ K such that
K ⊂ ∪ni=1V (vi , rvi ). (1.8)
Define r0 = min{rv1 , ..., rvn }, thus
V (u, r0 ) ⊂ V (u, rvi ), ∀i ∈ {1, ..., n},
so that from this and (1.7) we get
V (u, r0 ) ∩V (vi , rvi ) = 0/, ∀i ∈ {1, 2, ..., n}.
Hence,
n
V (u, r0 ) ∩ (∪i=1 V (vi , rvi )) = 0/.
/ that is V (u, r0 ) ⊂ K c . Therefore u is an interior point of K c
From this and (1.8) we obtain, V (u, r0 ) ∩ K = 0,
c c
and being u ∈ K arbitrary, K is open so that K is closed.
Theorem 1.2.47 Let (U, d) be a metric space. If F ⊂ K ⊂ U, K is compact and F is closed, then F is
compact.
Metric Spaces 11
Proof 1.8 Let {Gα , α ∈ L} be an open covering of F, that is
F ⊂ ∪α∈L Gα .
Observe that U = F ∪ F c ⊃ K, and thus,
F c ∪ (∪α∈L Gα ) ⊃ K.
Therefore, since F c is open {F c , Gα , α ∈ L} is an open covering of K, and since K is compact, there exist
α1 , ..., αn ∈ L such that
F c ∪ Gα1 ∪ ... ∪ Gαn ⊃ K ⊃ F.
Therefore
Gα1 ∪ ... ∪ Gαn ⊃ F,
so that F is compact.
Theorem 1.2.48 If {Kα , α ∈ L} is a collection of compact sets in a metric space (U, d) such that the inter-
section of each finite sub-collection is non-empty, then
6 0/.
∩α∈L Kα =
Proof 1.9 Suppose, to obtain contradiction, that
∩α∈L Kα = 0/. (1.9)
Fix α0 ∈ L and denote L1 = L \ {α0 }. From (1.9) we obtain
Kα0 ∩ (∩α∈L1 Kα ) = 0/.
Hence
Kα0 ⊂ (∩α∈L1 Kα )c ,
that is,
Kα0 ⊂ ∪α∈L1 Kαc .
Since, Kα0 is compact and Kαc is open, ∀α ∈ L, there exist α1 , α2 , ..., αn ∈ L1 such that
c
Kα0 ⊂ ∪nj=1 Kαc j = ∩nj=1 Kα j ,
therefore,
Kα0 ∩ ∩nj=1 Kα j = Kα0 ∩ Kα1 ∩ ... ∩ Kαn = 0/,

which contradicts the hypotheses. The proof is complete.
Corollary 1.2.49 Let (U, d) be a metric space. If {Kn , n ∈ N} ⊂ U is a sequence of compact non-empty sets
such that Kn ⊃ Kn+1 , ∀n ∈ N then ∩∞
n=1 Kn 6= 0
/.
Theorem 1.2.50 Let (U, d) be a metric space. If E ⊂ K ⊂ U, K is compact and E is infinite, then E has at
least one limit point in K.
Proof 1.10 Suppose, to obtain contradiction, that no point of K is a limit point of E. Then, for each
u ∈ K there exists ru > 0 such that V (u, ru ) has at most one point of E, namely, u if u ∈ E. Observe that
{V (u, ru ), u ∈ K} is an open covering of K and therefore of E. Since each V (u, ru ) has at most one point of
E which is infinite, no finite sub-covering (relating the open cover in question), covers E, and hence no finite
sub-covering covers K ⊃ E, which contradicts the fact that K is compact. This completes the proof.
Theorem 1.2.51 Let {In } be a sequence of bounded closed non-empty real intervals, such that In ⊃
In+1 , ∀n ∈ N. Thus, ∩∞ 6 0/.
n=1 In =
Proof 1.11 Let In = [an , bn ] and let E = {an , n ∈ N}. Thus, E =

6 0/ and E is upper bounded by b1 . Let
x = sup E.
Observe that, given m, n ∈ N we have that
an ≤ an+m ≤ bn+m ≤ bm ,
so that
sup an ≤ bm , ∀m ∈ N,
n∈N
that is, x ≤ bm , ∀m ∈ N. Hence,
am ≤ x ≤ bm , ∀m ∈ N,
that is,
x ∈ [am , bm ], ∀m ∈ N,
so that x ∈ ∩∞
m=1 Im .
Theorem 1.1
Let I = [a, b] ⊂ R be a bounded closed non-empty real interval. Under such hypotheses, I is compact.
Proof 1.12 Observe that if x, y ∈ [a, b] then |x − y| ≤ (b − a).. Suppose there exists an open covering of I,
denoted by {Gα , α ∈ L} for which there is no finite sub-covering.
Let c = (a + b)/2. Thus, either [a, c] or [c, b] has no finite sub-covering related to {Gα , α ∈ L}. Denote
such an interval by I1 . Dividing I1 into two connected closed sub-intervals of same size, we get an interval I2
for which there is no finite sub-covering related to {Gα , α ∈ L}.
Proceeding in this fashion, we may obtain a sequence of closed intervals {In } such that
1. In ⊃ In+1 , ∀n ∈ N.
2. No finite sub-collection of {Gα , α ∈ L} covers In , ∀n ∈ N.
3. If x, y ∈ In then |x − y| ≤ 2−n (b − a).
From the last theorem, there exists x∗ ∈ R such that x∗ ∈ ∩∞ n=1 In ⊂ I ⊂ ∪α ∈L Gα . Hence, there exists α0 ∈ L
such that x∗ ∈ Gα0 . Since Gα0 is open, there exists r > 0 such that
Vr (x∗ ) = (x∗ − r, x∗ + r) ⊂ Gα0 .
Choose n0 ∈ N such that
2−n0 (b − a) < r/2.
Hence, since x∗ ∈ In0 , if y ∈ In0 then from item 3 above, |y − x∗ | ≤ 2−n0 (b − a) < r/2, that is y ∈ Vr (x∗ ) ⊂
Gα0 .
Therefore
y ∈ In0 ⇒ y ∈ Gα0 ,
so that In0 ⊂ Gα0 , which contradicts the item 2 above indicated.
Theorem 1.2 Heine-Borel

Let E ⊂ R, thus the following three properties are equivalent:
Metric Spaces 13
1. E is closed and bounded.

2. E is compact.
3. Each infinite subset of E has a limit point of E.
Proof 1.13
1 implies 2: Let E ⊂ R be a closed and bounded. Thus, since E is bounded there exists [a, b], a
bounded closed interval such that E ⊂ [a, b]. From the last theorem [a, b] is compact and since E is
closed, from Theorem 1.2.47 we may infer that E is compact.
2 implies 3: This follows from Theorem1.2.50.
3 implies 1: We prove the contrapositive, that is, the negation of 1 implies the negation of 3.
The negation of 1 is: E is not bounded or E is not closed. If E ⊂ R is not bounded, choosing x1 ∈ E,
for each n ∈ N there exists xn+1 ∈ E such that |xn+1 | > n + |xn | ≥ n. Hence {xn } has no limit points
so that we have got the negation of 3.
On the other hand, suppose E is not closed. Thus there exists x0 ∈ R such that x0 ∈ E 0 and x0 6∈ E.
Since x0 ∈ E 0 , for each n ∈ N there exists xn ∈ E such that |xn − x0 | < 1/n (xn ∈ V1/n (x0 )).
Let y ∈ E, we are going to show that y is not limit point {xn } ⊂ E. Observe that
|xn − y| ≥ |x0 − y| − |xn − x0 |

> |x0 − y| − 1/n
> |x0 − y|/2 > 0 (1.10)
for all n sufficiently big.

Hence, y is not a limit point of {xn }, ∀y ∈ E. Therefore {xn } ⊂ E is a infinite set with no limit point
in E.
In any case, we have got the negation of 3. This completes the proof.
Theorem 1.2.52 (Weierstrass) Any real set which is bounded and infinite has a limit point in R.
Proof 1.14 Let E ⊂ R be a bounded infinite set. Thus, there exists r > 0 such that E ⊂ [−r, r] = Ir . Since
E is infinite and Ir is compact, from Theorem 1.2.50, E has a limit point in Ir ⊂ R. The proof is complete.
1.2.9 Separable metric spaces

Definition 1.2.53 (Separable metric space) Let (V, d) be a metric space. We say that a set M ⊂ V is dense
in V if
M = M ∪ M 0 = V.
If V has a dense subset which is countable, we say that V is separable.
Example 1.2.54 V = R is separable. Indeed, Q, the set of rational number, is dense in R and countable.
Example 1.2.55 The space l ∞ is not separable.

In fact, let A ⊂ l ∞ be the set of all real sequences whose entries are only 0 and 1.
From elementary analysis it is well known that A is non-countable.
Let 0 < ε < 1/4.
Suppose, to obtain contradiction, that
B = {un }n∈N ⊂ l ∞
is dense in l ∞ .
Thus, for each v ∈ A, we may select a nv ∈ N such that
d(v, unv ) < ε.
Let v, w ∈ A be such that v 6= w. Therefore,
d(v, w) = 1,
so that
d(v, w) ≤ d(v, unv ) + d(unv , w),
and thus
d(unv , w) ≥ 1 − d(v, unv )

≥ 1−ε
> 1 − 1/4
= 3/4
> ε. (1.11)
6 w, then
So to summarize, if v =
6 unw .
unv =
Let T : A → B, where
T (v) = unv .
Thus T is a bijection on Im (T ) ⊂ B.
Therefore, generically denoting C ∼ D if, and only if, there exists a bijection between the sets C and D,
we have
A ∼ Im(T ) ∼ B ∼ N.
This contradicts A to be non-countable.
So, we may infer that l ∞ is non-separable.
Exercise 1.2.56 Let 1 ≤ p < +∞. Prove that l p is separable.
1.2.10 Complete metric spaces
Definition 1.1 Let {un } ⊂ V , where (V, d) is a metric space.

We say that u0 ∈ V is the limit of {un } as n goes to infinity (∞), if for each ε > 0 there exists n0 ∈ N such
that if n > n0 , then
d(un , u0 ) < ε.
In such a a case we denote
lim un = u0 ,
n→∞
or
un → u0 , as n → ∞
and say that the sequence {un } is convergent.
Metric Spaces 15
Exercise 1.2.57 Let (V, d) be a metric space and let {un } ⊂ V be a convergent sequence.
Show that {un } is bounded.
Definition 1.2.58 (Cauchy sequence) Let (V, d) be a metric space and let {un } ⊂ V be a sequence. We say
that such a sequence is a Cauchy one as for each ε > 0 there exists n0 ∈ N such that if m, n > n0 , then
d(un , um ) < ε.
Exercise 1.2.59 Let (V, d) be a metric space and let {un } ⊂ V be a Cauchy sequence.
Show that {un } is bounded.
Definition 1.2.60 (Complete metric space) Let (V, d) be a metric space. We say that V is complete as each
Cauchy sequence in V converges to an element of V.
Exercise 1.2.61 Let (V, d) be a metric space and let M ⊂ V.
1. Show that u ∈ M if, and only if, there exists a sequence {un } ⊂ M such that
un → u, as n → ∞.
2. Show that M is closed if, and only if, the following property is valid:
If {un } ⊂ M and un → u, then u ∈ M.
Exercise 1.2.62 Let (V, d) be a metric space and let M ⊂ V. Show that M is complete if, and only if, M is
closed in V.
Exercise 1.2.63 Prove that Rn is complete (with the Euclidean metric).
Exercise 1.2.64 Prove that c is complete, where c is the space of complex convergent sequences.
Exercise 1.2.65 Let (V, d) be a metric space where V = C([a, b]) whit the metric
Z b
d(u, v) = |u(x) − v(x)| dx, ∀u, v ∈ V.
a
Show that V is not complete.
1.3 Completion of a metric space

Definition 1.3.1 (Isometries, isometric spaces) Let (V, d) and (V1 , d1 ) be metric spaces. A function T : V →
V1 is said to be a isometry of V in V1 as
d1 (T (u), T (v)) = d(u, v), ∀u, v ∈ V.
If there exists an isometry betwween V and V1 , we say that V and V1 are isometric.
Theorem 1.3.2 (Completion) Let (V, d) be a metric space which is not complete.
Under such hypotheses, there exists a metric space (V̂ , dˆ) such that V is isometric to a sub-space W of V̂
which is dense in V̂ . Moreover, V̂ is complete.
Proof 1.15
1. First part: Construction of V̂ .
Let {un } and {u0n } be Cauchy sequences in V .
We define a relation of equivalence in the set of Cauchy sequences in V by declaring
{un } ∼ {u0n }
as
lim d(un , u0n ) = 0.
n→∞
Let
V̂ = {{d
un } : {un } is a Cauchy sequence in V },
and where
un } = {{u0n } ⊂ V, such that {u0n }is a Cauchy sequence and {u0n } ∼ {un }}.
{d
For u = {un } and v = {vn } ⊂ V define
dˆ(û, v̂) = lim d(un , vn ).

n→∞
We shall show that this metric is well defined.

Let {un } ∈ û and {vn } ∈ v̂.
Observe that for, m, n ∈ N we have that
d(un , vn ) ≤ d(un , um ) + d(um , vm ) + d(vm , vn ),

that is,
d(un , vn ) − d(um , vm ) ≤ d(un , um ) + d(vm , vn ) → 0, as m, n → ∞.
Inverting the roles of m and n, we may similarly obtain:
d(um , vm ) − d(un , vn ) ≤ d(un , um ) + d(vm , vn ) → 0, as m, n → ∞

so that
|d(um , vm ) − d(un , vn )| → 0, as m, n → ∞.
Therefore {d(un , vn )} is a real Cauchy sequence and thus it is convergent.

Let {u0n } ∈ û and {v0n } ∈ v̂.
Hence,
d(un , vn ) ≤ d(un , u0n ) + d(u0n , v0n ) + d(v0n , vn ),
so that
d(un , vn ) − d(u0n , v0n ) ≤ d(un , u0n ) + d(vn , v0n ) → 0, as n → ∞.
Inverting the roles of the sequences, we get
d(u0n , v0n ) − d(un , vn ) ≤ d(un , u0n ) + d(vn , v0n ) → 0, as n → ∞,

so that
|d(un , vn ) − d(u0n , v0n )| → 0, as n → ∞.
Metric Spaces 17
Thus
lim d(un , vn ) = lim d(u0n , v0n ), ∀{u0n } ∈ û, {v0n } ∈ v̂.
n→∞ n→∞
Therefore, the candidate to metric in question is well defined.

Furthermore,
dˆ(û, v̂) = 0 ⇔ lim d(un , vn ) = 0

n→∞
⇔ {vn } ∈ û
⇔ û = v̂. (1.12)
Finally, let û, v̂ and ŵ ∈ V̂ .

Thus,
dˆ(û, ŵ) = lim d(un , wn )

n→∞
≤ lim [d(un , vn ) + d(vn , wn )]
n→∞
= lim d(un , vn ) + lim d(vn , wn )
n→∞ n→∞
= d(û, v̂) + d(v̂, ŵ). (1.13)
From this we may conclude that dˆ is in fact a metric for V̂ .

2. We shall show now that V is isometric of a dense subspace of V̂ .
Let b ∈ V. Define b̂ by its representative
{un } = {b, b, b, . . .}.
Define
W = {b̂ = {b,\
b, b, . . .} : b ∈ V }.
Define also T : V → W by
T (b) = b̂ = {b,\
b, b, . . .}.
Thus,
dˆ(b̂, ĉ) = lim d(b, c) = d(b, c).
n→∞
Therefore, T is an isometry.
We are going to show that W is dense in V̂ .
Let û ∈ V̂ and {un } ∈ û. Let ε > 0.
Since {un } is a Cauchy sequence, there exists n0 ∈ N such that if m, n > n0 , then
d(un , um ) < ε/2.
Choose N > n0 .
Thus, d(un , uN ) < ε/2, ∀n > n0 .
Observe that
ûN = {uN , u\
N , uN , . . .} ∈ W,
and
dˆ(û, ûN ) = lim d(un , uN ) ≤ ε/2 < ε .
n→∞
Since ε > 0 is arbitrary, we may conclude that
û ∈ W 0 ∪W, ∀û ∈ V̂ ,
that is, W is dense in V̂ .

3. Now we shall show that V̂ complete.
Let {ûn } be a Cauchy sequence in V̂ .
Since W is dense in V̂ , for each n ∈ N there exists ẑn ∈ W such that
1
dˆ(ûn , ẑn ) < .
n
Observe that
dˆ(ẑn , ẑm ) ≤ d(ẑ

ˆ m , ûm ) + d(ûm , ûn ) + dˆ(ûn , ẑn )
1 1
≤ + d(ûm , ûn ) + . (1.14)
n m
Let ε > 0 (a new one). Hence, there exists n0 ∈ N such that if m, n > n0 , then
ε
dˆ(ûn , ûm ) < .
3
Thus, if
3
m, n > max , n0 ,
ε
then
dˆ(ẑn , ẑm ) < ε.
Therefore, {ẑm } is a Cauchy sequence and since T : V → W is an isometry it follows that
{zm } = {T −1 (ẑm )},
is also a Cauchy one.

Let û be the class of {zm }. We will show that
lim dˆ(ûn , û) = 0.

n→∞
Indeed,
dˆ(ûn , û) ≤ dˆ(ûn , ẑn ) + dˆ(ẑn , û)

1
≤ + lim d(zn , zm ). (1.15)
n m→∞
Therefore,
lim dˆ(ûn , û) = 0.
n→∞
Thus, V̂ is complete.
Metric Spaces 19
1.4 Advanced topics on compactness in metric spaces

Definition 1.4.1 (Diameter of a set) Let (U, d) be a metric space and A ⊂ U. We define the diameter of A,
denoted by diam(A) by
diam(A) = sup{d(u, v) | u, v ∈ A}.
Definition 1.4.2 Let (U, d) be a metric space. We say that {Fk } ⊂ U is a nested sequence of sets if
F1 ⊃ F2 ⊃ F3 ⊃ ....
Theorem 1.4.3 If (U, d) is a complete metric space then every nested sequence of non-empty closed sets
{Fk } is such that
lim diam(Fk ) = 0
k→+∞
has non-empty intersection, that is

∩∞ 6 0/.
k=1 Fk =
Proof 1.16 Suppose {Fk } is a nested sequence and lim diam(Fk ) = 0. For each n ∈ N select un ∈ Fn .
k→∞
Suppose given ε > 0. Since
lim diam(Fn ) = 0,
n→∞
there exists N ∈ N such that if n ≥ N then

diam(Fn ) < ε.
Thus, if m, n > N we have um , un ∈ FN so that
d(un , um ) < ε.
Hence, {un } is a Cauchy sequence. Being U complete, there exists u ∈ U such that
un → u as n → ∞.
Choose m ∈ N. We have that un ∈ Fm , ∀n > m, so that
u ∈ F̄m = Fm .
Since m ∈ N is arbitrary we obtain

u ∈ ∩∞
m=1 Fm .
Theorem 1.4.4 Let (U, d) be a metric space. If A ⊂ U is compact then it is closed and bounded.
Proof 1.17 We have already proved that A is closed. Suppose, to obtain contradiction that A is not bounded.
Thus for each K ∈ N there exists u, v ∈ A such that
d(u, v) > K.
Observe that
A ⊂ ∪u∈A B1 (u).
Since A is compact there exists u1 , u2 , ..., un ∈ A such that
A =⊂ ∪nk=1 B1 (uk ).
Define
R = max{d(ui , u j ) | i, j ∈ {1, ..., n}}.
Choose u, v ∈ A such that
d(u, v) > R + 2. (1.16)
Observe that there exist i, j ∈ {1, ..., n} such that
u ∈ B1 (ui ), v ∈ B1 (u j ).
Thus
d(u, v) ≤ d(u, ui ) + d(ui , u j ) + d(u j , v)

≤ 2 + R, (1.17)
which contradicts (1.16). This completes the proof.
Definition 1.4.5 (Relative compactness) In a metric space (U, d) a set A ⊂ U is said to be relatively com-
pact if A is compact.
Definition 1.4.6 (ε - nets) Let (U, d) be a metric space. A set N ⊂ U is said to be a ε-net with respect to a
set A ⊂ U if for each u ∈ A there exists v ∈ N such that
d(u, v) < ε.
Definition 1.4.7 Let (U, d) be a metric space. A set A ⊂ U is said to be totally bounded if for each ε > 0
there exists a finite ε-net with respect to A.
Proposition 1.4.8 Let (U, d) be a metric space. If A ⊂ U is totally bounded then it is bounded.
Proof 1.18 Choose u, v ∈ A. Let {u1 , ..., un } be the 1 − net with respect to A. Define
R = max{d(ui , u j ) | i, j ∈ {1, ..., n}}.
Observe that there exist i, j ∈ {1, ..., n} such that
d(u, ui ) < 1, d(v, u j ) < 1.
Thus
d(u, v) ≤ d(u, ui ) + d(ui , u j ) + d(u j , v)

≤ R + 2. (1.18)
Since u, v ∈ A are arbitrary, A is bounded.
Theorem 1.4.9 Let (U, d) be a metric space. If from each sequence {un } ⊂ A we can select a convergent
subsequence {unk }, then A is totally bounded.
Proof 1.19 Suppose, to obtain contradiction, that A is not totally bounded. Thus there exists ε0 > 0 such
that there exists no ε0 -net with respect to A. Choose u1 ∈ A, hence {u1 } is not a ε0 -net, that is, there exists
u2 ∈ A such that
d(u1 , u2 ) > ε0 .
Metric Spaces 21
Again {u1 , u2 } is not a ε0 -net for A, so that there exists u3 ∈ A such that
d(u1 , u3 ) > ε0 and d(u2 , u3 ) > ε0 .
Proceeding in this fashion we can obtain a sequence {un } such that
d(un , um ) > ε0 , if m 6= n. (1.19)
Clearly, we cannot extract a convergent subsequence of {un }, otherwise such a subsequence would be Cauchy
contradicting (1.19). The proof is complete.
Definition 1.4.10 (Sequentially compact sets) Let (U, d) be a metric space. A set A ⊂ U is said to be se-
quentially compact if for each sequence {un } ⊂ A there exists a subsequence {unk } and u ∈ A such that
unk → u, as k → ∞.
Theorem 1.4.11 A subset A of a metric space (U, d) is compact if and only if it is sequentially compact.
Proof 1.20 Suppose A is compact. By Proposition 2.4.8, A is countably compact. Let {un } ⊂ A be a
sequence. We have two situations to consider.
1. {un } has infinitely many equal terms, that is in this case we have
un1 = un2 = .... = unk = ... = u ∈ A.
Thus the result follows trivially.

2. {un } has infinitely many distinct terms. In such a case, being A countably compact, {un } has a limit
point in A, so that there exists a subsequence {unk } and u ∈ A such that
unk → u, as k → ∞.
In both cases we may find a subsequence converging to some u ∈ A.

Thus, A is sequentially compact.
Conversely suppose A is sequentially compact, and suppose {Gα , α ∈ L} is an open cover of A. For each
u ∈ A define
δ (u) = sup{r | Br (u) ⊂ Gα , for some α ∈ L}.
First we prove that δ (u) > 0, ∀u ∈ A. Choose u ∈ A. Since A ⊂ ∪α∈L Gα , there exists α0 ∈ L such that u ∈ Gα0 .
Being Gα0 open, there exists r0 > 0 such that Br0 (u) ⊂ Gα0 .
Thus,
δ (u) ≥ r0 > 0.
Now define δ0 by
δ0 = inf{δ (u) | u ∈ A}.
Therefore, there exists a sequence {un } ⊂ A such that
δ (un ) → δ0 as n → ∞.
Since A is sequentially compact, we may obtain a subsequence {unk } and u0 ∈ A such that
δ (unk ) → δ0 and unk → u0 ,

as k → ∞. Therefore, we may find K0 ∈ N such that if k > K0 then
δ (u0 )
d (unk , u0 ) < . (1.20)
4
We claim that
δ (u0 )
δ (unk ) ≥ , if k > K0 .
4
To prove the claim, suppose
z ∈ B δ (u0 ) (unk ), ∀k > K0 ,
4
(observe that in particular from (1.20)
u0 ∈ B δ (u0 ) (unk ), ∀k > K0 ).

4
Since
δ (u0 )
< δ (u0 ),
2
there exists some α1 ∈ L such that
B δ (u0 ) (u0 ) ⊂ Gα1 .
2
However, since
δ (u0 )
d(unk , u0 ) < , if k > K0 ,
4
we obtain
B δ (u0 ) (u0 ) ⊃ B δ (u0 ) (unk ), if k > K0 ,
2 4
so that
δ (u0 )
δ (unk ) ≥ , ∀k > K0 .
4
Therefore,
δ (u0 )
lim δ (unk ) = δ0 ≥ .
k→∞ 4
Choose ε > 0 such that
δ0 > ε > 0.
From the last theorem, since A it is sequentially compact, it is totally bounded. For the ε > 0 chosen above,
consider an ε-net contained in A (the fact that the ε-net may be chosen contained in A is also a consequence
of last theorem) and denote it by N, that is,
N = {v1 , ..., vn } ∈ A.
Since δ0 > ε, there exists

α1 , ..., αn ∈ L
such that
Bε (vi ) ⊂ Gαi , ∀i ∈ {1, ..., n},
considering that
δ (vi ) ≥ δ0 > ε > 0, ∀i ∈ {1, ..., n}.
For u ∈ A, since N is an ε-net we have
u ∈ ∪ni=1 Bε (vi ) ⊂ ∪ni=1 Gαi .

Metric Spaces 23
Since u ∈ U is arbitrary, we obtain

A ⊂ ∪ni=1 Gαi .
Thus
{Gα1 , ..., Gαn }
is a finite subcover for A of
{Gα , α ∈ L}.
Hence A is compact.
The proof is complete;
Theorem 1.4.12 Let (U, d) be a metric space. Thus, A ⊂ U is relatively compact if and only if for each
sequence in A, we may select a convergent subsequence.
Proof 1.21 Suppose A is relatively compact. Thus A is compact so that from the last Theorem, A is
sequentially compact.
Thus from each sequence in A we may select a subsequence which converges to some element of A. In
particular, for each sequence in A ⊂ A, we may select a subsequence that converges to some element of A.
Conversely, suppose that for each sequence in A we may select a convergent subsequence. It suffices
to prove that A is sequentially compact. Let {vn } be a sequence in A. Since A is dense in A, there exists a
sequence {un } ⊂ A such that
1
d(un , vn ) < .
n
From the hypothesis we may obtain a subsequence {unk } and u0 ∈ A such that
unk → u0 , as k → ∞.
Thus,
vnk → u0 ∈ A, as k → ∞.
Therefore, A is sequentially compact so that it is compact.
Theorem 1.4.13 Let (U, d) be a metric space.

1. If A ⊂ U is relatively compact then it is totally bounded.
2. If (U, d) is a complete metric space and A ⊂ U is totaly bounded then A is relatively compact.
Proof 1.22
1. Suppose A ⊂ U is relatively compact. From the last theorem, from each sequence in A, we can extract
a convergent subsequence. From Theorem 1.4.9, A is totally bounded.
2. Let (U, d) be a metric space and let A be a totally bounded subset of U.
Let {un } be a sequence in A. Since A is totally bounded for each k ∈ N, we find a εk -net where εk = 1/k,
denoted by Nk , where
(k) (k) (k)
Nk = {v1 , v2 , ..., vnk }.
In particular for k = 1 {un } is contained in the 1-net N1 . Thus at least one ball of radius 1 of N1
(1)
contains infinitely many points of {un }. Let us select a subsequence {unk }k∈N of this infinite set
(which is contained in a ball of radius 1). Similarly, we may select a subsequence here just partially
(2) (1)
relabeled {unl }l∈N of {unk } which is contained in one of the balls of the 21 -net. Proceeding in this
(k)
fashion for each k ∈ N, we may find a subsequence denoted by {unm }m∈N of the original sequence
contained in a ball of radius 1/k.
(k)
Now consider the diagonal sequence denoted by {unk }k∈N = {zk }. Thus
2
d(zn , zm ) < , if m, n > k,
k
that is, {zk } is a Cauchy sequence, and since (U, d) is complete, there exists u ∈ U such that
zk → u as k → ∞.
From Theorem 1.4.12, A is relatively compact.

1.5 The Arzela-Ascoli theorem

In this section we present a classical result in analysis, namely the Arzela-Ascoli theorem.
Definition 1.5.1 (Equi-continuity) Let F be a collection of complex functions defined on a metric space
(U, d). We say that F is equicontinuous if for each ε > 0, there exists δ > 0 such that if u, v ∈ U and
d(u, v) < δ then
| f (u) − f (v)| < ε, ∀ f ∈ F .
Furthermore, we say that F is point-wise bounded if for each u ∈ U there exists M(u) ∈ R such that
| f (u)| < M(u), ∀ f ∈ F .
Theorem 1.5.2 (Arzela-Ascoli) Suppose F is a point-wise bounded equicontinuous collection of complex

functions defined on a metric space (U, d). Also suppose that U has a countable dense subset E. Thus, each
sequence { fn } ⊂ F has a subsequence that converges uniformly on every compact subset of U.
Proof 1.23 Let {un } be a countable dense set in (U, d). By hypothesis, { fn (u1 )} is a bounded sequence,
therefore it has a convergent subsequence, which is denoted by { fnk (u1 )}. Let us denote
fnk (u1 ) = f˜1,k (u1 ), ∀k ∈ N.
Thus there exists g1 ∈ C such that

f˜1,k (u1 ) → g1 , as k → ∞.
Observe that { fnk (u2 )} is also bounded and also it has a convergent subsequence, which similarly as above
we will denote by { f˜2,k (u2 )}. Again there exists g2 ∈ C such that
f˜2,k (u1 ) → g1 , as k → ∞.
f˜2,k (u2 ) → g2 , as k → ∞.
Proceeding in this fashion for each m ∈ N we may obtain { f˜m,k } such that
f˜m,k (u j ) → g j , as k → ∞, ∀ j ∈ {1, ..., m},
where the set {g1 , g2 , ..., gm } is obtained as above. Consider the diagonal sequence
{ f˜k,k },
Metric Spaces 25
and observe that the sequence

{ f˜k,k (um )}k>m
is such that
f˜k,k (um ) → gm ∈ C, as k → ∞, ∀m ∈ N.
Therefore we may conclude that from { fn } we may extract a subsequence also denoted by
{ fnk } = { f˜k,k }
which is convergent in
E = {un }n∈N .
Now suppose K ⊂ U, being K compact. Suppose given ε > 0. From the equi-continuity hypothesis there
exists δ > 0 such that if u, v ∈ U and d(u, v) < δ we have
ε
| fnk (u) − fnk (v)| < , ∀k ∈ N.
3
Observe that
K ⊂ ∪u∈K B δ (u),
2
and being K compact we may find {ũ1 , ..., ũM } such that
K ⊂ ∪M
j=1 B δ (ũ j ).
2
Since E is dense in U, there exists
v j ∈ B δ (ũ j ) ∩ E, ∀ j ∈ {1, ..., M }.

2
Fixing j ∈ {1, ..., M}, from v j ∈ E we obtain that
lim fnk (v j )
k→∞
exists as k → ∞. Hence there exists K0 j ∈ N such that if k, l > K0 j then

ε
| fnk (v j ) − fnl (v j )| < .
3
Pick u ∈ K, thus
u ∈ B δ (ũ jˆ)
2
for some jˆ ∈ {1, ..., M}, so that

d(u, v jˆ) < δ .
Therefore, if
k, l > max{K01 , ..., K0M },
then
| fnk (u) − fnl (u)| ≤ | fnk (u) − fnk (v jˆ)| + | fnk (v jˆ) − fnl (v jˆ)|
+| fnl (v jˆ) − fnl (u)|
ε ε ε
≤ + + = ε. (1.21)
3 3 3
Since u ∈ K is arbitrary, we conclude that { fnk } is uniformly Cauchy on K.
Chapter 2
Topological Vector Spaces
2.1 Introduction
The main objective of this chapter is to present an outline of the basic tools of analysis necessary to develop
the subsequent chapters. We assume the reader has a background in linear algebra and elementary real anal-
ysis at an undergraduate level. The main references for this chapter are the excellent books on functional
analysis, Rudin [67], Bachman and Narici [6] and Reed and Simon [62]. All proofs are developed in detail.
2.2 Vector spaces

We denote by F a scalar field. In practice this is either R or C, the set of real or complex numbers.
Definition 2.2.1 (Vector spaces) A vector space over F is a set which we will denote by U whose elements
are called vectors, for which are defined two operations, namely, addition denoted by (+) : U ×U → U, and
scalar multiplication denoted by (·) : F ×U → U, so that the following relations are valid
1. u + v = v + u, ∀u, v ∈ U,
2. u + (v + w) = (u + v) + w, ∀u, v, w ∈ U,
3. there exists a vector denoted by θ such that u + θ = u, ∀u ∈ U,
4. for each u ∈ U, there exists a unique vector denoted by
−u such that u + (−u) = θ ,
5. α · (β · u) = (α · β ) · u, ∀α, β ∈ F, u ∈ U,
6. α · (u + v) = α · u + α · v, ∀α ∈ F, u, v ∈ U,
7. (α + β ) · u = α · u + β · u, ∀α, β ∈ F, u ∈ U,
8. 1 · u = u, ∀u ∈ U.
Remark 2.2.2 From now on we may drop the dot (·) in scalar multiplications and denote α · u simply as αu.
Topological Vector Spaces 27
Definition 2.2.3 (Vector subspace) Let U be a vector space. A set V ⊂ U is said to be a vector subspace
of U if V is also a vector space with the same operations as those of U. If V 6= U we say that V is a proper
subspace of U.
Definition 2.2.4 (Finite dimensional space) A vector space is said to be of finite dimension if there exists
fixed u1 , u2 , ..., un ∈ U such that for each u ∈ U there are corresponding α1 , ...., αn ∈ F for which
n
u = ∑ αi u i . (2.1)
i=1
Definition 2.2.5 (Topological spaces) A set U is said to be a topological space if it is possible to define a
collection σ of subsets of U called a topology in U, for which are valid the following properties:
1. U ∈ σ ,
2. 0/ ∈ σ ,
3. if A ∈ σ and B ∈ σ then A ∩ B ∈ σ , and
4. arbitrary unions of elements in σ also belong to σ .
Any A ∈ σ is said to be an open set.
Remark 2.2.6 When necessary, to clarify the notation, we shall denote the vector space U endowed with the
topology σ by (U, σ ).
Definition 2.2.7 (Closed sets) Let U be a topological space. A set A ⊂ U is said to be closed if U \ A is open.
We also denote U \ A = Ac = {u ∈ U | u 6∈ A}.
Remark 2.2.8 For any sets A, B ⊂ U we denote
A \ B = {u ∈ A | u 6∈ B}.
Sometimes, as the meaning is clear, we may also denote A \ B = A − B.

Proposition 2.2.9 For closed sets we have the following properties:
1. U and 0/ are closed,
2. If A and B are closed sets then A ∪ B is closed,
3. Arbitrary intersections of closed sets are closed.
Proof 2.1
1. Since 0/ is open and U = 0/c , by Definition 2.2.7, U is closed. Similarly, since U is open and 0/ =
U \U = U c , 0/ is closed.
2. A, B closed implies that Ac and Bc are open, and by Definition 2.2.5, Ac ∩ Bc is open, so that A ∪ B =
(Ac ∩ Bc )c is closed.
3. Consider A = ∩λ ∈L Aλ , where L is a collection of indices and Aλ is closed, ∀λ ∈ L. We may write
A = (∪λ ∈L Acλ )c and since Acλ is open ∀λ ∈ L we have, by Definition 2.2.5, that A is closed.
¯ as the intersection of
Definition 2.2.10 (Closure) Given A ⊂ U, we define the closure of A, denoted by A,
all closed sets that contain A.
Remark 2.2.11 From Proposition 2.2.9 Item 3 we have that Ā is the smallest closed set that contains A, in
the sense that, if C is closed and A ⊂ C then Ā ⊂ C.
Definition 2.2.12 (Interior) Given A ⊂ U we define its interior, denoted by A◦ , as the union of all open sets
is contained in A.
Remark 2.2.13 It is not difficult to prove that if A is open then A = A◦ .
Definition 2.2.14 (Neighborhood) Given u0 ∈ U we say that V is a neighborhood of u0 if such a set is open
and contains u0 . We denote such neighborhoods by Vu0 .
Proposition 2.2.15 If A ⊂ U is a set such that for each u ∈ A there exists a neighborhood Vu 3 u such that
Vu ⊂ A, then A is open.
Proof 2.2 This follows from the fact that A = ∪u∈A Vu and any arbitrary union of open sets is open.
Definition 2.2.16 (Function) Let U and V be two topological spaces. We say that f : U → V is a function
if f is a collection of pairs (u, v) ∈ U × V such that for each u ∈ U there exists only one v ∈ V such that
(u, v) ∈ f .
In such a case we denote
v = f (u).
Definition 2.2.17 (Continuity at a point) A function f : U → V is continuous at u ∈ U if for each neigh-

borhood V f (u) ⊂ V of f (u) there exists a neighborhood Vu ⊂ U of u such that f (Vu ) ⊂ V f (u) .
Definition 2.2.18 (Continuous function) A function f : U → V is continuous if it is continuous at each

u ∈ U.
Proposition 2.2.19 A function f : U → V is continuous if and only if f −1 (V ) is open for each open V ⊂ V ,
where
f −1 (V ) = {u ∈ U | f (u) ∈ V }. (2.2)
Proof 2.3 Suppose f −1 (V ) is open whenever V ⊂ V is open. Pick u ∈ U and any open V such that
f (u) ∈ V . Since u ∈ f −1 (V ) and f ( f −1 (V )) ⊂ V , we have that f is continuous at u ∈ U. Since u ∈ U
is arbitrary we have that f is continuous. Conversely, suppose f is continuous and pick V ⊂ V open. If
f −1 (V ) = 0/ we are done, since 0/ is open. Thus, suppose u ∈ f −1 (V ), since f is continuous, there exists Vu
a neighborhood of u such that f (Vu ) ⊂ V . This means Vu ⊂ f −1 (V ) and therefore, from Proposition 2.2.15,
f −1 (V ) is open.
6 u2 , there
Definition 2.2.20 We say that (U, σ ) is a Hausdorff topological space if, given u1 , u2 ∈ U, u1 =
exists V1 , V2 ∈ σ such that
u1 ∈ V1 , u2 ∈ V2 and V1 ∩ V2 = 0/. (2.3)
Definition 2.2.21 (Base) A collection σ 0 ⊂ σ is said to be a base for σ if every element of σ may be repre-
sented as a union of elements of σ 0 .
Definition 2.2.22 (Local base) A collection σ̂ of neighborhoods of a point u ∈ U is said to be a local base
at u if each neighborhood of u contains a member of σ̂ .
Definition 2.2.23 (Topological vector spaces) A vector space endowed with a topology, denoted by (U, σ ),
is said to be a topological vector space if and only if
1. Every single point of U is a closed set,
2. The vector space operations (addition and scalar multiplication) are continuous with respect to σ .
More specifically, addition is continuous if, given u, v ∈ U and V ∈ σ such that u + v ∈ V then there exists
Vu 3 u and Vv 3 v such that Vu + Vv ⊂ V . On the other hand, scalar multiplication is continuous if given
α ∈ F, u ∈ U and V 3 α · u, there exists δ > 0 and Vu 3 u such that, ∀β ∈ F satisfying |β − α| < δ we have
β Vu ⊂ V .
6 0) the functions Tu0 : U → U and

Given (U, σ ), let us associate with each u0 ∈ U and α0 ∈ F (α0 =
Mα0 : U → U defined by
Tu0 (u) = u0 + u (2.4)
and
Mα0 (u) = α0 · u. (2.5)
The continuity of such functions is a straightforward consequence of the continuity of vector space operations
(addition and scalar multiplication). It is clear that the respective inverse maps, namely T−u0 and M1/α0 are
also continuous. So if V is open, then u0 + V , that is, (T−u0 )−1 (V ) = Tu0 (V ) = u0 + V is open. By analogy
α0 V is open. Thus σ is completely determined by a ‘local base’, so that the term local base will be understood
henceforth as a local base at θ . So to summarize, a local base of a topological vector space is a collection Ω
of neighborhoods of θ , such that each neighborhood of θ contains a member of Ω.
Now we present some simple results, namely:
Proposition 2.2.24 If A ⊂ U, then ∀u ∈ A, there exists a neighborhood V of θ such that u + V ⊂ A
Proof 2.4 Just take V = A − u.

Proposition 2.2.25 Given a topological vector space (U, σ ), any element of σ may be expressed as a union
of translates of members of Ω, so that the local base Ω generates the topology σ .
Proof 2.5 Let A ⊂ U open and u ∈ A. V = A − u is a neighborhood of θ and by definition of local base,
there exists a set VΩu ⊂ V such that VΩu ∈ Ω. Thus, we may write
A = ∪u∈A (u + VΩu ). (2.6)
2.3 Some properties of topological vector spaces

In this section we study some fundamental properties of topological vector spaces. We start with the following
proposition:
Proposition 2.3.1 Any topological vector space U is a Hausdorff space.
Proof 2.6 6 u1 . Thus V = U \ {u1 − u0 } is an open neighborhood of zero.

Pick u0 , u1 ∈ U such that u0 =
As θ + θ = θ , by the continuity of addition, there exist V1 and V2 neighborhoods of θ such that
V1 + V2 ⊂ V (2.7)
define U = V1 ∩ V2 ∩ (−V1 ) ∩ (−V2 ), thus U = −U (symmetric) and U + U ⊂ V and hence
u0 + U + U ⊂ u0 + V ⊂ U \ {u1 } (2.8)
so that
6 u1 , ∀v1 , v2 ∈ U ,
u0 + v1 + v2 = (2.9)
or
u0 + v1 6= u1 − v2 , ∀v1 , v2 ∈ U , (2.10)
and since U = −U
(u0 + U ) ∩ (u1 + U ) = 0/. (2.11)
Definition 2.3.2 (Bounded sets) A set A ⊂ U is said to be bounded if to each neighborhood of zero V there
corresponds a number s > 0 such that A ⊂ tV for each t > s.
Definition 2.3.3 (Convex sets) A set A ⊂ U such that
if u, v ∈ A then λ u + (1 − λ )v ∈ A, ∀λ ∈ [0, 1], (2.12)
is said to be convex.
Definition 2.3.4 (Locally convex spaces) A topological vector space U is said to be locally convex if there
is a local base Ω whose elements are convex.
Definition 2.3.5 (Balanced sets) A set A ⊂ U is said to be balanced if αA ⊂ A, ∀α ∈ F is such that |α| ≤ 1.
Theorem 2.3.6 In a topological vector space U we have:

1. Every neighborhood of zero contains a balanced neighborhood of zero,
2. Every convex neighborhood of zero contains a balanced convex neighborhood of zero.
Proof 2.7
1. Suppose U is a neighborhood of zero. From the continuity of scalar multiplication, there exist V
(neighborhood of zero) and δ > 0, such that αV ⊂ U whenever |α| < δ . Define W = ∪|α|<δ αV ,
thus W ⊂ U is a balanced neighborhood of zero.
2. Suppose U is a convex neighborhood of zero in U. Define
A = {∩αU | α ∈ C, |α| = 1}. (2.13)
As 0 · θ = θ (where θ ∈ U denotes the zero vector) from the continuity of scalar multiplication there
exists δ > 0 and there is a neighborhood of zero V such that if |β | < δ then β V ⊂ U . Define W as
the union of all such β V . Thus, W is balanced and α −1 W = W as |α| = 1, so that W = αW ⊂ αU ,
and hence W ⊂ A, which implies that the interior A◦ is a neighborhood of zero. Also, A◦ ⊂ U . Since
A is an intersection of convex sets, it is convex and so is A◦ . Now will show that A◦ is balanced and
complete the proof. For this, it suffices to prove that A is balanced. Choose r and β such that 0 ≤ r ≤ 1
and |β | = 1. Then
rβ A = ∩|α |=1 rβ αU = ∩|α|=1 rαU . (2.14)
Since αU is a convex set that contains zero, we obtain rαU ⊂ αU , so that rβ A ⊂ A, which completes
the proof.
Proposition 2.3.7 Let U be a topological vector space and V a neighborhood of zero in U. Given u ∈ U,
there exists r ∈ R+ such that β u ∈ V , ∀β such that |β | < r.
Proof 2.8 Observe that u + V is a neighborhood of 1 · u, then by the continuity of scalar multiplication,
there exists W neighborhood of u and r > 0 such that
β W ⊂ u + V , ∀β such that |β − 1| < r, (2.15)
so that
βu ∈ u+V , (2.16)
or
(β − 1)u ∈ V , where |β − 1| < r, (2.17)
and thus
β̂ u ∈ V , ∀βˆ such that |βˆ | < r, (2.18)
which completes the proof.

Corollary 2.3.8 Let V be a neighborhood of zero in U, if {rn } is a sequence such that rn > 0, ∀n ∈ N and
∞ r V.
lim rn = ∞, then U ⊂ ∪n=1 n
n→∞
Proof 2.9 Let u ∈ U, then αu ∈ V for any α sufficiently small, from the last proposition u ∈ αV
1
. As
rn → ∞ we have that rn > α1 for n sufficiently big, so that u ∈ rn V , which completes the proof.
Proposition 2.3.9 Suppose {δn } is sequence such that δn → 0, δn < δn−1 , ∀n ∈ N and V a bounded neigh-
borhood of zero in U, then {δn V } is a local base for U.
Proof 2.10 Let U be a neighborhood of zero; as V is bounded, there exists t0 ∈ R+ such that V ⊂ tU
for any t > t0 . As lim δn = 0, there exists n0 ∈ N such that if n ≥ n0 then δn < t10 , so that δn V ⊂ U , ∀n such
n→∞
that n ≥ n0 .
Definition 2.3.10 (Convergence in topological vector spaces) Let U be a topological vector space. We say
{un } converges to u0 ∈ U, if for each neighborhood V of u0 then, there exists N ∈ N such that
un ∈ V , ∀n ≥ N.
Definition 2.3.11 (Dense set) Let (V, σ ) be a topological vector space (T.V.E.). Let A, B ⊂ V . We say that A
is dense in B as
B ⊂ A.
Definition 2.3.12 We say that topological vector space V is separable as it has a dense and countable set.
2.3.1 Nets and convergence

Definition 2.3.13 A directed system is a set of indices I, with an order relation ≺, which satisfies the follow-
ing properties:
1. If α, β ∈ I, then there exists γ ∈ I such that
α ≺ γ and β ≺ γ.
2. ≺ is a partial order relation.

Definition 2.3.14 (Net) Let (V, σ ) be a topological space. A net in (V, σ ) is a function defined on a directed
system I with range in V , where we denote such a net by
{uα }α∈I ,
and where
uα ∈ V, ∀α ∈ I.
Definition 2.3.15 (Convergent net) Let (V, σ ) be a topological space and let {uα }α∈I be a net in V .
We say that such a net converges to u ∈ V as for each neighborhood W ∈ σ of u there exists β ∈ I such
that if α β , then
uα ∈ W.
Definition 2.3.16 Let (V, σ ) be a topological space and let {uα }α ∈I be a net in V . We say that u ∈ V is a
cluster point of the net in question as for each neighborhood W ∈ σ of u and each β ∈ I, there exists α β
such that
uα ∈ W.
Definition 2.3.17 (Limit point) Let (V, σ ) be a topological space and let A ⊂ V. We say that u ∈ V is a limit
point of A as for each neighborhood W ∈ σ of u, there exists v ∈ W ∩ A such that v 6= u.
Theorem 2.3.18 Let (V, d) be a topological space and let A ⊂ V .

Under such hypotheses,
A = A ∪ A0
where A0 denotes the set of limit points of A.
Proof 2.11 Let u ∈ A ∪ A0 .

If u ∈ A, then u ∈ A.
Thus, suppose u ∈ A0 \ A.
Hence, for each neighborhood W ∈ σ of u, there exists uw ∈ A \ {u} such that uw ∈ W.
Denote by I the set of all neighborhoods of u, partially ordered by the relation
W1 ≺ W2 ⇔ W2 ⊂ W1 .
From the exposed above we may obtain a net {uw }W ∈I such that
uw → u.
Assume, to obtain contradiction, that u 6∈ A.

c
Hence u ∈ A which is an open set. Since uw → u, there exists W1 ∈ I such that if W2 W1 , then
c
uw2 ∈ A .
In particular, uw2 ∈ Ac , if W2 W1 , which contradicts
uw2 ∈ A \ {u}.
Summarizing,
u ∈ A, ∀u ∈ A ∪ A0 .
Therefore,
A ∪ A0 ⊂ A. (2.19)
Reciprocally, suppose u ∈ A.
If u ∈ A, then u ∈ A ∪ A0 .
Suppose, to obtain contradiction, that u 6∈ A and u 6∈ A0 .
Thus there exists a neighborhood W ∈ σ of u such that
W ∩ A = 0/.
Thus, A ⊂ W c and W c is closed, so that

A ⊂ W c.
From this and u ∈ W we get
u 6∈ A,
a contradiction. Therefore u ∈ A or u ∈ A0 , ∀u ∈ A.
Thus
A ⊂ A ∪ A0 . (2.20)
From (2.19) and (2.20), we obtain
A = A ∪ A0 .
This complete the proof.
Theorem 2.3.19 Let (V1 , σ1 ) and (V2 , σ2 ) be topological spaces.

Let f : V1 → V2 be a function.
Let u ∈ V1 . Thus f is continuous at u if, and only if, for each net {uα }α∈I ⊂ V1 such that uα → u, we have
that
f (uα ) → f (u).
Proof 2.12 Suppose f is continuous at u. Let {uα }α ∈I be a net such that
uα → u.
Let W f (u) ∈ σ2 be such that f (u) ∈ W f (u) .

From the hypotheses, there exists Vu ∈ σ1 such that u ∈ Vu and
f (Vu ) ⊂ W f (u) .
From uα → u, there exists β ∈ I such that if α β , then
uα ∈ Vu .
Therefore,
f (uα ) ∈ W f (u) , if α β .
Since W f (u) is arbitrary, it follows that
f (uα ) → f (u).
Reciprocally, suppose
f (uα ) → f (u)
whenever
uα → u.
Suppose, to obtain contradiction, that f is not continuous at u.
Thus there exists W f (u) ∈ σ2 such that f (u) ∈ W f (u) and so that for each neighborhood W ∈ σ1 of u there
exists uW ∈ W such that
f (uW ) 6∈ W f (u) .
Denote by I the set of all neighborhoods of u, partially ordered by the relation
W1 ≺ W2 ⇔ W2 ⊂ W1 .
Thus, the net {uW }W ∈I is such that

uW → u.
However,
f (uW ) 6∈ W f (u) , ∀W ∈ I.
Hence, { f (uW )} does not converges to f (u), which contradicts the reciprocal hypothesis.
2.4 Compactness in topological vector spaces

We start this section with the definition of open covering.
Definition 2.4.1 (Open covering) Given B ⊂ U we say that {Oα , α ∈ A} is a covering of B if B ⊂ ∪α∈A Oα .
If Oα is open ∀α ∈ A then {Oα } is said to be an open covering of B.
Definition 2.4.2 (Compact sets) A set B ⊂ U is said to be compact if each open covering of B has a finite
sub-covering. More explicitly, if B ⊂ ∪α ∈A Oα , where Oα is open ∀α ∈ A, then, there exist α1 , ..., αn ∈ A
such that B ⊂ Oα1 ∪ ... ∪ Oαn , for some n, a finite positive integer.
Theorem 2.4.3 Let (V, σ ) be a topological space. Let K ⊂ V.

Under such hypotheses, K is compact if, and only if, each net {uα }α ∈I ⊂ K has a limit point in K.
Proof 2.13 Suppose K is compact. Let {uα }α∈I ⊂ K be a net with infinite distinct terms (otherwise the
result is immediate).
Denote E = {uα }α ∈I . Suppose, to obtain contradiction, that no point of K is a limit point of E.
Hence, for each u ∈ K, there exists a neighborhood Wu of u such that
Wu ∩ E = 0/,
or
Wu ∩ E = {u} if u ∈ E.
In any case each Wu has no more than one point of E.
Observe that ∪u∈K Wu ⊃ K. Since K is compact, there exist u1 , . . . , un ∈ K such that
E ⊂ K ⊂ ∪nj=1Wu j .
From this we may conclude that E has no more than n distinct elements, which contradicts E to have
infinity distinct terms.
Reciprocally, suppose that each net {uα }α ∈I ⊂ K has at least one limit point in K.
Suppose, to obtain contradiction K is not compact.
Thus there exists an open covering {Gα , α ∈ L} of K which admits no finite sub-covering.
Denote by F the finite sub-collections of {Gα , α ∈ L}.
Hence, for a W ∈ F we may select a uW 6∈ W where uW ∈ K.
Let us partially order F through the relation
W1 ≺ W2 ⇔ W1 ⊂ W2 .
From the hypotheses, the net {uW }W ∈F has a limit point u ∈ K.

Observe that
u ∈ K ⊂ ∪α∈L Gα .
Thus, there exists α0 ∈ L such that
u ∈ Gα0 .
Since u is a limit point of {uW }W ∈F ⊂ K, there exists W1 Gα0 such that
uW1 ∈ Gα0 ⊂ W1 .
This contradicts uW1 6∈ W1 . Therefore, K is compact.

Proposition 2.4.4 A compact subset of a Hausdorff space is closed.
Proof 2.14 Let U be a Hausdorff space and consider A ⊂ U, A compact. Given x ∈ A and y ∈ Ac , there
exist open sets Ox and Oyx such that x ∈ Ox , y ∈ Oyx and Ox ∩ Oyx = 0. / It is clear that A ⊂ ∪x∈A Ox and since A
is compact, we may find {x1 , x2 , ..., xn } such that A ⊂ ∪ni=1 Oxi . For the selected y ∈ Ac we have y ∈ ∩ni=1 Oyxi
n O xi ) ∩ (∪n O ) = 0.
and (∩i= / Since ∩ni=1 Oyxi is open, and y is an arbitrary point of Ac we have that Ac is
1 y i=1 xi
open, so that A is closed, which completes the proof. The next result is very useful.
Theorem 2.4.5 Let {Kα , α ∈ L} be a collection of compact subsets of a Hausdorff topological vector space
U, such that the intersection of every finite sub-collection (of {Kα , α ∈ L}) is non-empty.
Under such hypotheses
∩α ∈L Kα 6= 0/.
Proof 2.15 Fix α0 ∈ L. Suppose, to obtain contradiction that
∩α∈L Kα = 0/.
That is,
6
α=α
Kα0 ∩ [∩α∈L 0 Kα ] = 0/.
Thus,
6
∩α ∈L 0 Kα ⊂ Kαc 0 ,
α=α
so that
α 6=α
Kα0 ⊂ [∩α∈L 0 Kα ]c ,
6
Kα0 ⊂ [∪α∈L 0 Kαc ].
α=α
However Kα0 is compact and Kαc is open, ∀α ∈ L.

Hence, there exist α1 , ..., αn ∈ L such that
Kα0 ⊂ ∪ni=1 Kαc i .

From this we may infer that
Kα0 ∩ [∩ni=1 Kαi ] = 0/,
which contradicts the hypotheses.
Proposition 2.4.6 Let U be a topological Hausdorff space and let A ⊂ B where A is closed and B is compact.
Under such hypotheses, A is compact.
Proof 2.16 Consider {Oα , α ∈ L} an open cover of A. Thus {Ac , Oα , α ∈ L} is a cover of U, so that
it is a cover of B. As B is compact, there exist α1 , α2 , ..., αn such that Ac ∪ (∪ni=1 Oαi ) ⊃ B ⊃ A, so that
{Oαi , i ∈ {1, ..., n}} covers A. From this we may infer that A is compact. The proof is complete.
Definition 2.4.7 (Countably compact sets) A set A is said to be countably compact if every infinite subset
of A has a limit point in A.
Proposition 2.4.8 Every compact subset of a Hausdorff topological space U is countably compact.
Proof 2.17 Let B an infinite subset of A compact and suppose B has no limit point in A, so that there is no
any limit point. Choose a countable infinite set {x1 , x2 , x3 , ....} ⊂ B and define F = {x1 , x2 , x3 , ...}. It is clear
that F has no limit point. Thus for each n ∈ N, there exist On open such that On ∩ F = {xn }. Also, for each
x ∈ A \ F, there exist Ox such that x ∈ Ox and Ox ∩ F = 0. / Thus {Ox , x ∈ A \ F; O1 , O2 , ...} is an open cover
of A without a finite subcover, which contradicts the fact that A is compact.
2.4.1 A note on convexity in topological vector spaces

Definition 2.4.9 Let (V, σ ) be a topological vector space. Let A ⊂ V be such that A 6= 0/.
We define the convex hull of A, denoted by Conv(A), as
( )
n n
Conv(A) = ∑ λk uk : n ∈ N, λk ≥ 0, uk ∈ A, ∀k ∈ {1, . . . , n} and ∑ λk = 1 .
k=1 k=1
Theorem 2.4.10 let (V, σ ) be a topological vector space. Let A ⊂ V be such that A 6= 0/.
Under such hypotheses, Conv(A) is convex.
Proof 2.18 Let u, v ∈ Conv(A) and let λ ∈ [0, 1]. Thus, there exist n1 , n2 ∈ N such that
n1 n2
u= ∑ λk uk and v = ∑ λ̃k vk ,
k=1 k=1
where uk ∈ A and λk ≥ 0, ∀k ∈ {1, . . . n1 } and also ∑nk=

1
1 λk = 1.
n2 ˜
Moreover, vk ∈ A, λ̃k ≥ 0, ∀k ∈ {1, . . . n2 } and ∑k=1 λk = 1.
Thus, we have that
n1 n2
λ u + (1 − λ )v = ∑ λ λk uk + ∑ (1 − λ )λ̃k vk ,
k=1 k=1
where
λ λk ≥ 0, uk ∈ A, ∀k ∈ {1, . . . , n1 }
and
(1 − λ )λ̃k ≥ 0, vk ∈ A, ∀k ∈ {1, . . . , n2 }
so that
n1 n2
∑ λ λk + ∑ (1 − λ )λ̃k = λ + (1 − λ ) = 1.
k=1 k=1
Therefore,
λ u + (1 − λ v) ∈ Conv(A), ∀u, v ∈ Conv(A), λ ∈ [0, 1].
Hence, Conv(A) is convex.
Theorem 2.4.11 Let (V, σ ) be a topological vector space. Let A ⊂ V be such that A =
6 0/.
Under such hypotheses, A is convex if, and only if, Conv(A) = A.
Proof 2.19 Suppose that A is convex. We shall prove that

( )
n n
A = Bn ≡ ∑ λk u k : λk ≥ 0, uk ∈ A, ∀k ∈ {1, . . . , n} and ∑ λk = 1 , ∀n ∈ N.
k=1 k=1
We shall do it by induction on n.
Observe that for n = 1 and n = 2, from the convexity of A we obtain A = B1 and A = B2 .
Let n ∈ N. Suppose A = Bn . We are going to prove that A = Bn+1 which will complete the induction.
Clearly Bn ⊂ Bn+1 , so that A ⊂ Bn+1 .
Reciprocally, let u ∈ Bn+1 . Thus, there exist u1 , . . . , un+1 ∈ A and λ1 , . . . λn+1 such that λk ≥ 0, ∀k ∈
+1
{1, . . . , n + 1}, ∑nk=1 λk = 1, and
n+1
u= ∑ λk u k .
k=1
With no loss in generality, assume 0 < λn+1 < 1 (otherwise the conclusion is immediate).
Thus,
λ1 + · · · + λn = (1 − λn+1 ) > 0.
Hence,
λ1 λn
+···+ = 1.
1 − λn+1 1 − λn+1
Therefore, defining
λk
λ̃k = ≥ 0, ∀k ∈ {1, . . . , n}
1 − λn+1
we have that
n
∑ λ̃k = 1,
k=1
so that
n
w= ∑ λ̃k uk ∈ Bn = A.
k=1
Since A convex, we obtain
w1 = (1 − λn+1 )w + λn+1 un+1 ∈ A,
that is,
n+1
w1 = ∑ λk uk = u ∈ A, ∀u ∈ Bn+1 .
k=1
Thus,
Bn+1 ⊂ A,
and hence
Bn+1 = A.
This completes the induction, that is,
A = Bn , ∀n ∈ N.
Hence,
A = ∪∞
n=1 Bn = Conv(A).
Reciprocally, assume A = Conv(A). Since Conv(A) is convex, A is convex.
Remark 2.4.12 Let A ⊂ B ⊂ V . Clearly Conv(A) ⊂ Conv(B). In particular, if B is convex, then
Conv(A) ⊂ B = Conv(B).
Proposition 2.4.13 Let (V, σ ) be a topological vector space. Suppose that a non-empty A ⊂ V is open. Under
such hypotheses, Conv(A) is open.
Proof 2.20 Let u ∈ Conv(A). Thus, there exist n ∈ N, uk ∈ A, λk ≥ 0, ∀k ∈ {1, . . . , n} such that ∑nk=1 λk = 1,
and u = ∑nk=1 λk uk .
With no loss in generality, assume λ1 6= 0 (redefine the indices, if necessary).
Since u1 ∈ A and A is open, there exists a neighborhood Vu1 of u1 such that Vu1 ⊂ A.
Thus, W = λ1Vu1 + λ2 u2 + · · · + λn un ⊂ Conv(A).
Observe that W is open and u ∈ W ⊂ Conv(A).
Therefore u is an interior point of Conv(A), ∀u ∈ Conv(A). Thus, Conv(A) is open.
This completes the proof.
Proposition 2.4.14 Let (V, σ ) be a topological vector space. Suppose A ⊂ V is convex and A◦ 6= 0/. Under
such hypotheses, A◦ is convex.
Proof 2.21 Let u, v ∈ A◦ and λ ∈ [0, 1]. Thus, there exist neighborhoods Vu of u and Vv of v such that Vu ⊂ A
and Vv ⊂ A. Hence,
B ≡ Vu ∪Vv ⊂ A.
Therefore, since A is convex, we obtain
Conv(B) ⊂ Conv(A) = A.
From the last proposition Conv(B) is open and moreover Conv(B) ⊂ A◦ . Thus,
λ u + (1 − λ )v ∈ Conv(B) ⊂ A◦ , ∀u, v ∈ A◦ , λ ∈ [0, 1].
From this we may infer that A◦ is convex.

Remark 2.4.15 Let (V, σ ) be a topological vector space a let A ⊂ V be a non-empty open set..
Thus, tA is open, ∀t ∈ F such that t 6= 0.
Let B ⊂ V be a balanced set such that 0 ∈ B◦ .
Let α ∈ F be such that 0 < |α| ≤ 1. Thus,
αB◦ ⊂ αB ⊂ B.
Since αB◦ is open, we have that αB◦ ⊂ B◦ , ∀α ∈ F such that |α| ≤ 1.

From this we may infer that B◦ is balanced.
2.5 Normed and metric spaces

The idea here is to prepare a route for the study of Banach spaces defined below. We start with the definition
of norm.
Definition 2.5.1 (Norm) A vector space U is said to be a normed space, if it is possible to define a function
k · kU : U → R+ = [0, +∞), called a norm, which satisfies the following properties:
1. kukU > 0, if u 6= θ and kukU = 0 ⇔ u = θ
2. ku + vkU ≤ kukU + kvkU , ∀ u, v ∈ U,

3. kαukU = |α|kukU , ∀u ∈ U, α ∈ F.
Now we recall the definition of metric.
Definition 2.5.2 (Metric space) A vector space U is said to be a metric space if it is possible to define a
function d : U ×U → R+ , called a metric on U, such that
1. 0 ≤ d(u, v), ∀u, v ∈ U,
2. d(u, v) = 0 ⇔ u = v,
3. d(u, v) = d(v, u), ∀u, v ∈ U,
4. d(u, w) ≤ d(u, v) + d(v, w), ∀u, v, w ∈ U.
A metric can be defined through a norm, that is
d(u, v) = ku − vkU . (2.21)
In this case we say that the metric is induced by the norm.

The set Br (u) = {v ∈ U | d(u, v) < r} is called the open ball with center at u and radius r. A metric
d : U ×U → R+ is said to be invariant if
d(u + w, v + w) = d(u, v), ∀u, v, w ∈ U. (2.22)
The following are some basic definitions concerning metric and normed spaces:
Definition 2.5.3 (Convergent sequences) Given a metric space U, we say that {un } ⊂ U converges to u0 ∈
U as n → ∞, if for each ε > 0, there exists n0 ∈ N, such that if n ≥ n0 then d(un , u0 ) < ε. In this case we write
un → u0 as n → +∞.
Definition 2.5.4 (Cauchy sequence) {un } ⊂ U is said to be a Cauchy sequence if for each ε > 0 there exists
n0 ∈ N such that d(un , um ) < ε, ∀m, n ≥ n0
Definition 2.5.5 (Completeness) A metric space U is said to be complete if each Cauchy sequence related
to d : U ×U → R+ converges to an element of U.
Definition 2.5.6 (Limit point) Let (U, d) be a metric space and let E ⊂ U. We say that v ∈ U is a limit point
6 v.
of E if for each r > 0 there exists w ∈ Br (v) ∩ E such that w =
Definition 2.5.7 (Interior point, topology for (U, d)) Let (U, d) be a metric space and let E ⊂ U. We say
that u ∈ E is interior point if there exists r > 0 such that Br (u) ⊂ E. If a point of E is not a limit point is said
to be an isolated one. We may define a topology for a metric space (U, d), by declaring as open all set E ⊂ U
such that all its points are interior. Such a topology is said to be induced by the metric d.
Proposition 2.5.8 Let (U, d) be a metric space. The set σ of all open sets, defined through the last definition,
is indeed a topology for (U, d).
Proof 2.22
1. Obviously 0/ and U are open sets.
2. Assume A and B are open sets and define C = A ∩ B. Let u ∈ C = A ∩ B, thus from u ∈ A, there exists
r1 > 0 such that Br1 (u) ⊂ A. Similarly from u ∈ B there exists r2 > 0 such that Br2 (u) ⊂ B.
Define r = min{r1 , r2 }. Thus Br (u) ⊂ A ∩ B = C, so that u is an interior point of C. Since u ∈ C is
arbitrary, we may conclude that C is open.
3. Suppose {Aα , α ∈ L} is a collection of open sets. Define E = ∪α ∈L Aα and we shall show that E is
open.
Choose u ∈ E = ∪α ∈L Aα . Thus there exists α0 ∈ L such that u ∈ Aα0 . Since Aα0 is open there exists
r > 0 such that Br (u) ⊂ Aα0 ⊂ ∪α ∈L Aα = E. Hence u is an interior point of E, since u ∈ E is arbitrary,
E = ∪α ∈L Aα is open.
Definition 2.5.9 Let (U, d) be a metric space and let E ⊂ U. We define E 0 as the set of all the limit points
of E.
Theorem 2.5.10 Let (U, d) be a metric space and let E ⊂ U. Then E is closed if and only if E 0 ⊂ E.
Proof 2.23 Suppose E 0 ⊂ E. Let u ∈ E c , thus u 6∈ E and u 6∈ E 0 . Therefore there exists r > 0 such that
Br (u) ∩ E = 0, / so that Br (u) ⊂ E c . Therefore u is an interior point of E c . Since u ∈ E c is arbitrary we may
infer that E is open, so that E = (E c )c is closed.
c
Conversely, suppose that E is closed, that is E c is open.

If E 0 = 0/ we are done.
Thus assume E 0 6= 0/ and choose u ∈ E 0 . Thus for each r > 0 there exists v ∈ Br (u) ∩ E such that v = 6 u.
Thus Br (u) * E c , ∀r > 0 so that u is not a interior point of E c . Since E c is open, we have that u 6∈ E c so that
u ∈ E. We have thus obtained, u ∈ E, ∀u ∈ E 0 , so that E 0 ⊂ E.
Remark 2.5.11 From this last result, we may conclude that in a metric space E ⊂ U is closed if and only if
E 0 ⊂ E.
At this point we recall the definition of Banach space.

Definition 2.5.12 (Banach spaces) A normed vector space U is said to be a Banach space if each Cauchy
sequence related to the metric induced by the norm converges to an element of U.
Remark 2.5.13 Let (U, σ ) be a topological space. We say that the topology σ is compatible with a metric
d : U × U → R+ if σ coincides with the topology generated by such a metric. In this case we say that
d : U ×U → R+ induces the topology σ .
Definition 2.5.14 (Metrizable spaces) A topological vector space (U, σ ) is said to be metrizable if σ is
compatible with some metric d.
Definition 2.5.15 (Normable spaces) A topological vector space (U, σ ) is said to be normable if the in-
duced metric (by this norm) is compatible with σ .
2.6 Linear mappings

Given U,V topological vector spaces, a function (mapping) f : U → V , A ⊂ U and B ⊂ V , we define:
f (A) = { f (u) | u ∈ A}, (2.23)
and the inverse image of B, denoted f −1 (B) as
f −1 (B) = {u ∈ U | f (u) ∈ B}. (2.24)
Definition 2.6.1 (Linear functions) A function f : U → V is said to be linear if
f (αu + β v) = α f (u) + β f (v), ∀u, v ∈ U, α, β ∈ F. (2.25)

Definition 2.6.2 (Null space and range) Given f : U → V , we define the null space and the range of f,
denoted by N( f ) and R( f ) respectively, as
N( f ) = {u ∈ U | f (u) = θ } (2.26)
and
R( f ) = { f (u) : u ∈ U}. (2.27)
Note that if f is linear then N( f ) and R( f ) are subspaces of U and V respectively.
Proposition 2.6.3 Let U,V be topological vector spaces. If f : U → V is linear and continuous at θ , then it
is continuous everywhere.
Proof 2.24 Since f is linear we have f (θ ) = θ . Since f is continuous at θ , given V ⊂ V a neighborhood

of zero, there exists U ⊂ U neighborhood of zero, such that
f (U ) ⊂ V . (2.28)
Thus
v − u ∈ U ⇒ f (v − u) = f (v) − f (u) ∈ V , (2.29)
or
v ∈ u + U ⇒ f (v) ∈ f (u) + V , (2.30)
which means that f is continuous at u. Since u is arbitrary, f is continuous everywhere.
2.7 Linearity and continuity

Definition 2.7.1 (Bounded functions) A function f : U → V is said to be bounded if it maps bounded sets
into bounded sets.
Proposition 2.7.2 A set E is bounded if and only if the following condition is satisfied: whenever {un } ⊂ E
and {αn } ⊂ F are such that αn → 0 as n → ∞ we have αn un → θ as n → ∞.
Proof 2.25 Suppose E is bounded. Let U be a balanced neighborhood of θ in U, then E ⊂ tU for some t.
For {un } ⊂ E, as αn → 0, there exists N such that if n > N then t < |α1n | . Since t −1 E ⊂ U and U is balanced,
we have that αn un ∈ U , ∀n > N, and thus αn un → θ . Conversely, if E is not bounded, there is a neighborhood
V of θ and {rn } such that rn → ∞ and E is not contained in rn V , that is, we can choose un such that rn−1 un
is not in V , ∀n ∈ N, so that {rn−1 un } does not converge to θ .
Proposition 2.7.3 Let f : U → V be a linear function. Consider the following statements

1. f is continuous,
2. f is bounded,
3. If un → θ then { f (un )} is bounded,
4. If un → θ then f (un ) → θ .
Then,
1 implies 2,
2 implies 3,
if U is metrizable with invariant metric,then 3 implies 4, which implies 1.
Proof 2.26
1. 1 implies 2: Suppose f is continuous, for W ⊂ V neighborhood of zero, there exists a neighborhood

of zero in U, denoted by V , such that
f (V ) ⊂ W . (2.31)
If E is bounded, there exists t0 ∈ R+ such that E ⊂ tV , ∀t ≥ t0 , so that
f (E) ⊂ f (tV ) = t f (V ) ⊂ tW , ∀t ≥ t0 , (2.32)
and thus f is bounded.

2. 2 implies 3: Suppose un → θ and let W be a neighborhood of zero. Then there exists N ∈ N such that
if n ≥ N then un ∈ V ⊂ W where V is a balanced neighborhood of zero. On the other hand, for n < N,
there exists Kn such that un ∈ Kn V . Define K = max{1, K1 , ..., Kn }. Then un ∈ KV , ∀n ∈ N and hence
{un } is bounded. Finally from 2, we have that { f (un )} is bounded.
3. 3 implies 4: Suppose U is metrizable with invariant metric and let un → θ . Given K ∈ N, there exists
nK ∈ N such that if n > nK then d(un , θ ) < K12 . Define γn = 1 if n < n1 and γn = K, if nK ≤ n < nK+1
so that
d (γn un , θ ) = d(Kun , θ ) ≤ Kd(un , θ ) < K −1 . (2.33)
Thus since 2 implies 3 we have that { f (γn un )} is bounded so that, by Proposition 2.7.2 f (un ) =
γn−1 f (γn un ) → θ as n → ∞.
4. 4 implies 1: suppose 1 fails. Thus there exists a neighborhood of zero W ⊂ V such that f −1 (W )
contains no neighborhood of zero in U. Particularly, we can select {un } such that un ∈ B1/n (θ ) and
f (un ) not in W so that { f (un )} does not converge to zero. Thus 4 fails.
2.8 Continuity of operators in Banach spaces

Let U,V be Banach spaces. We call a function A : U → V an operator.
Proposition 2.8.1 Let U,V be Banach spaces. A linear operator A : U → V is continuous if and only if there
exists K ∈ R+ such that
kA(u)kV < KkukU , ∀u ∈ U.
Proof 2.27 Suppose A is linear and continuous. From Proposition 2.7.3,
if {un } ⊂ U is such that un → θ then A(un ) → θ . (2.34)
We claim that for each ε > 0 there exists δ > 0 such that if kukU < δ then kA(u)kV < ε.
Suppose, to obtain contradiction that the claim is false.
1
Thus there exists ε0 > 0 such that for each n ∈ N there exists un ∈ U such that kun kU ≤ n and kA(un )kV ≥
ε0 .
Therefore un → θ and A(un ) does not converge to θ , which contradicts (2.34).
Thus the claim holds.
In particular, for ε = 1 there exists δ > 0 such that if kukU < δ then kA(u)kV < 1. Thus given an arbitrary
not relabeled u ∈ U, u 6= θ , for
δu
w=
2kukU
we have
δ kA(u)kV
kA(w)kV = < 1,
2kukU
that is,
2kukU
kA(u)kV < , ∀u ∈ U.
δ
Defining
2
K=
δ
the first part of the proof is complete. Reciprocally, suppose there exists K > 0 such that
kA(u)kV < KkukU , ∀u ∈ U.
Hence un → θ implies kA(un )kV → θ , so that from Proposition 2.7.3, A is continuous.

2.9 Some classical results on Banach spaces

In this section we present some important results in Banach spaces. We start with the following theorem.
2.9.1 The Baire Category theorem

Theorem 2.9.1 Let U and V be Banach spaces and let A : U → V be a linear operator. Then A is bounded
if and only if the set C ⊂ U has at least one interior point, where
C = A−1 [{v ∈ V | kvkV ≤ 1}].
Proof 2.28 Suppose there exists u0 ∈ U in the interior of C. Thus, there exists r > 0 such that
Br (u0 ) = {u ∈ U | ku − u0 kU < r} ⊂ C.
Fix u ∈ U such that kukU < r. Thus, we have
kA(u)kV ≤ kA(u + u0 )kV + kA(u0 )kV .
Observe also that

k(u + u0 ) − u0 kU < r,
so that u + u0 ∈ Br (u0 ) ⊂ C and thus
kA(u + u0 )kV ≤ 1
and hence
kA(u)kV ≤ 1 + kA(u0 )kV , (2.35)
∀u ∈ U such that kukU < r. Fix an arbitrary not relabeled u ∈ U such that u 6= θ . From (2.35)
u r
w=
kukU 2
is such that
kA(u)kV r
kA(w)kV = ≤ 1 + kA(u0 )kV ,
kukU 2
so that
2
kA(u)kV ≤ (1 + kA(u0 )kV )kukU .
r
Since u ∈ U is arbitrary, A is bounded.
Reciprocally, suppose A is bounded. Thus
kA(u)kV ≤ KkukU , ∀u ∈ U,
for some K > 0. In particular

1
D = u ∈ U | kukU ≤ ⊂ C.
K
Definition 2.9.2 A set S in a metric space U is said to be nowhere dense if S has an empty interior.
Theorem 2.9.3 (Baire Category theorem) A complete metric space is never the union of a countable num-
ber of nowhere dense sets.
Proof 2.29 Suppose, to obtain contradiction, that U is a complete metric space and
U = ∪∞
n=1 An
where each An is nowhere dense. Since A1 is nowhere dense, there exist u1 ∈ U which is not in Ā1 , otherwise
we would have U = Ā1 , which is not possible since U is open. Furthermore, Āc1 is open, so that we may
obtain u1 ∈ Ac1 and 0 < r1 < 1 such that
B1 = Br1 (u1 )
satisfies
B1 ∩ A1 = 0/.
Since A2 is nowhere dense we have B1 is not contained in Ā2 . Therefore we may select u2 ∈ B1 \ Ā2 and since
B1 \ Ā2 is open, there exists 0 < r2 < 1/2 such that
B̄2 = B̄r2 (u2 ) ⊂ B1 \ Ā2 ,
that is
B2 ∩ A2 = 0/.
Proceeding inductively in this fashion, for each n ∈ N we may obtain un ∈ Bn−1 \ Ān such that we may choose
an open ball Bn = Brn (un ) such that
B̄n ⊂ Bn−1 ,
Bn ∩ An = 0/
and
0 < rn < 21−n .
Observe that {un } is a Cauchy sequence, considering that if m, n > N then un , um ∈ BN , so that
d (un , um ) < 2(21−N ).
Define
u = lim un .
n→∞
Since
un ∈ BN , ∀n > N,
we get
u ∈ B̄N ⊂ BN−1 .
Therefore, u is not in AN−1 , ∀N > 1, which means u is not in ∪n=1
∞ A = U, a contradiction.
n
2.9.2 The Principle of Uniform Boundedness

Theorem 2.9.4 (The Principle of Uniform Boundedness) Let U be a Banach space. Let F be a family of
linear bounded operators from U into a normed linear space V . Suppose for each u ∈ U there exists a Ku ∈ R
such that
kT (u)kV < Ku , ∀T ∈ F .
Then, there exists K ∈ R such that
kT k < K, ∀T ∈ F .
Proof 2.30 Define

Bn = {u ∈ U | kT (u)kV ≤ n, ∀T ∈ F }.
By the hypotheses, given u ∈ U, u ∈ Bn for all n sufficiently big. Thus,
U = ∪∞
n=1 Bn .
Moreover, each Bn is closed. By the Baire category theorem there exists n0 ∈ N such that Bn0 has non-empty
interior. That is, there exists u0 ∈ U and r > 0 such that
Br (u0 ) ⊂ Bn0 .
Thus, fixing an arbitrary T ∈ F , we have
kT (u)kV ≤ n0 , ∀u ∈ Br (u0 ).
Thus if kukU < r then k(u + u0 ) − u0 kU < r, so that
kT (u + u0 )kV ≤ n0 ,
that is,
kT (u)kV − kT (u0 )kV ≤ n0 .
Thus
kT (u)kV ≤ 2n0 , if kukU < r. (2.36)

For u ∈ U arbitrary, u 6= θ , define

ru
w= ,
2kukU
from (2.36) we obtain
rkT (u)kV
kT (w)kV = ≤ 2n0 ,
2kukU
so that
4n0 kukU
kT (u)kV ≤ , ∀u ∈ U.
r
Hence
4n0
kT k ≤ , ∀T ∈ F .
r
2.9.3 The Open Mapping theorem

Theorem 2.9.5 (The Open Mapping theorem) Let U and V be Banach spaces and let A : U → V be a
bounded onto linear operator. Thus if O ⊂ U is open then A(O) is open in V .
Proof 2.31 First we will prove that given r > 0, there exists r0 > 0 such that
A(Br (θ )) ⊃ BVr0 (θ ). (2.37)
Here BVr0 (θ ) denotes a ball in V of radius r0 with center in θ . Since A is onto
V = ∪∞
n=1 A(nB1 (θ )).
By the Baire Category theorem, there exists n0 ∈ N such that the closure of A(n0 B1 (θ )) has non-empty
interior, so that A(B1 (θ )) has a non-empty interior. We will show that there exists r0 > 0 such that
BVr0 (θ ) ⊂ A(B1 (θ )).
Observe that there exists y0 ∈ V and r1 > 0 such that
BVr1 (y0 ) ⊂ A(B1 (θ )). (2.38)
Define u0 ∈ U which satisfies A(u0 ) = y0 . We claim that
A(Br2 (θ )) ⊃ BVr1 (θ ),
where r2 = 1 + ku0 kU . To prove the claim, pick
y ∈ A(B1 (θ ))
thus there exists u ∈ U such that kukU < 1 and A(u) = y. Therefore
A(u) = A(u − u0 + u0 ) = A(u − u0 ) + A(u0 ).
But observe that
ku − u0 kU ≤ kukU + ku0 kU
< 1 + ku0 kU
= r2 , (2.39)
so that
A(u − u0 ) ∈ A(Br2 (θ )).
This means
y = A(u) ∈ A(u0 ) + A(Br2 (θ )),
and hence
A(B1 (θ )) ⊂ A(u0 ) + A(Br2 (θ )).
That is, from this and (2.38), we obtain
A(u0 ) + A(Br2 (θ )) ⊃ A(B1 (θ )) ⊃ BVr1 (y0 ) = A(u0 ) + BVr1 (θ ),
and therefore
A(Br2 (θ )) ⊃ BVr1 (θ ).
Since
A(Br2 (θ )) = r2 A(B1 (θ )),
we have, for some not relabeled r1 > 0 that
A(B1 (θ )) ⊃ BVr1 (θ ).
Thus it suffices to show that

A(B1 (θ )) ⊂ A(B2 (θ )),
to prove (2.37). Let y ∈ A(B1 (θ )), since A is continuous we may select u1 ∈ B1 (θ ) such that
y − A(u1 ) ∈ BVr1 /2 (θ ) ⊂ A(B1/2 (θ )).
Now select u2 ∈ B1/2 (θ ) so that

y − A(u1 ) − A(u2 ) ∈ BVr1 /4 (θ ).
By induction, we may obtain
un ∈ B21−n (θ ),
such that
n
y − ∑ A(u j ) ∈ BVr1 /2n (θ ).
j=1
Define
∞
u= ∑ un ,
n=1
we have that u ∈ B2 (θ ), so that

∞
y= ∑ A(un ) = A(u) ∈ A(B2 (θ )).
n=1
Therefore,
A(B1 (θ )) ⊂ A(B2 (θ )).
The proof of (2.37) is complete.
To finish the proof of this theorem, assume O ⊂ U is open. Let v0 ∈ A(O). Let u0 ∈ O be such that
A(u0 ) = v0 . Thus there exists r > 0 such that
Br (u0 ) ⊂ O.
From (2.37),
A(Br (θ )) ⊃ BVr0 (θ ),
for some r0 > 0. Thus

A(O) ⊃ A(u0 ) + A(Br (θ )) ⊃ v0 + BVr0 (θ ).
This means that v0 is an interior point of A(O). Since v0 ∈ A(O) is arbitrary, we may conclude that A(O) is
open.
Theorem 2.9.6 (The Inverse Mapping theorem) A continuous linear bijection of one Banach space onto
another has a continuous inverse.
Proof 2.32 Let A : U → V satisfying the theorem hypotheses. Since A is open, A−1 is continuous.
2.9.4 The Closed Graph theorem

Definition 2.9.7 (Graph of a mapping) Let A : U → V be an operator, where U and V are normed linear
spaces. The graph of A denoted by Γ(A) is defined by
Γ(A) = {(u, v) ∈ U ×V | v = A(u)}.
Theorem 2.9.8 (The Closed Graph theorem) Let U and V be Banach spaces and let A : U → V be a linear
operator. Then A is bounded if and only if its graph is closed.
Proof 2.33 Suppose Γ(A) is closed. Since A is linear Γ(A) is a subspace of U ⊕ V . Also, being Γ(A)
closed, it is a Banach space with the norm
k(u, A(u))k = kukU + kA(u)kV .
Consider the continuous mappings

Π1 (u, A(u)) = u
and
Π2 (u, A(u)) = A(u).
Observe that Π1 is a bijection, so that by the inverse mapping theorem Π−1
1 is continuous. As
A = Π2 ◦ Π−1
1 ,
it follows that A is continuous. The converse is immediate.
2.10 A note on finite dimensional normed spaces

We start this section with the theorem.
Theorem 2.10.1 Let V be a complex normed vector space (not necessarily of finite-dimension). Suppose
{u1 , . . . , un } ⊂ V be a linearly independent set. Under such hypotheses, there exists c > 0 such that
kα1 u1 + · · · + αn un kV ≥ c(|α1 | + · · · + |αn |), ∀α1 , . . . , αn ∈ C. (2.40)
Proof 2.34 For α1 , . . . , αn ∈ C, let us denote
s = |α1 | + · · · + |αn |.
Thus, if s = 0, then α1 = . . . = αn = 0 and (2.40) holds.

Suppose then s > 0

α
Denoting β j = sj , ∀ j ∈ {1, . . . , n} we have that (2.40) is equivalent to
n
kβ1 u1 + · · · + βn un kV ≥ c, ∀β1 , . . . , βn ∈ C, such that ∑ |β j | = 1. (2.41)
j=1
Suppose to obtain, contradiction, there is no c > 0 such that (2.41) holds.

Thus there exist sequences {vm } ⊂ V and {β jm } ⊂ C, ∀ j ∈ {1, . . . , n} such that
vm = β1m u1 + . . . + βnm un
such that
n
∑ |β jm | = 1, ∀m ∈ N
j=1
and
kvm kV → 0, as m → ∞.
In particular
|β jm | ≤ 1, ∀m ∈ N, j ∈ {1, . . . , n}.
It is a well-known result in elementary analysis that a bounded sequence Cn has a convergent subse-
quence.
Hence, there exists a subsequence {mk } of N and
β j0 ∈ C, ∀ j ∈ {1, . . . , n}
such that
m
β j k → β j0 , ∀ j ∈ {1, . . . , n},
and
n
∑ |β j0 | = 1.
j=1
Thus
n
vmk → ∑ β j0 u j .
j=1
Since
n
∑ |β j0 | = 1,
j=1
this contradicts
vmk → 0.
Therefore, there exists c > 0 such that (2.40) holds.
Now we present the following result about the completeness of finite dimensional subspaces in a normed
vector space.
Theorem 2.10.2 Let V be a complex normed vector space and let M be finite dimensional subspace of V .
Under such hypotheses, M is complete (closed). In particular, each normed finite dimensional vector
space is complete.
Proof 2.35 Let {vm } ⊂ M be a Cauchy sequence. Let n ∈ N be the dimension of M. Let {u1 , . . . , un } be a
basis for M.
Hence, there exists a sequence {α m
j } ⊂ C such that
vm = α1m u1 + · · · + αnm un .
Let ε > 0. Since {vm } is a Cauchy sequence, there exists n0 ∈ N such that, if m, l > n0 , then
kvm − vl kV < ε.
Hence, from this and the last theorem, there exists c > 0 such that
n
ε > ∑ (α mj − α lj )u j
j=1 V
n
≥ c ∑ |α m l
j − α j|
i=1
≥ c|α m l
j − α j |, ∀m, l > n0 , ∀ j ∈ {1, · · · , n}. (2.42)
Thus {α m
j } ⊂ C is a Cauchy sequence.
Therefore, there exists a α 0j ∈ C such that
0
αm
j → α j , ∀ j ∈ {1, . . . , n}.
From this, denoting v0 = ∑nj=1 α 0j u j ∈ M, we get

n
kvm − v0 kV ≤ ∑ |α mj − α 0j |ku j kV → 0, as m → ∞.
j=1
From this last result, we may infer that M is complete.
Definition 2.10.3 (Equivalence between two norms) Let V be a vector space. Two norms
k · k0 , k · k1 : V → R+
are said to be equivalent, if there exists α, β > 0 such that
αkuk0 ≤ kuk1 ≤ β kuk0 , ∀u ∈ V.
Theorem 2.10.4 Let V be a finite dimensional vector space. Under such hypotheses, any two norms defined
on V are equivalent.
Proof 2.36 Assume the dimension of V is n.

Let {u1 , . . . , un } ⊂ V be a basis for V . Let k · k0 , k · k1 : V → R+ be two norms in V .
Let u ∈ V . Hence there exists α1 , . . . , αn ∈ C such that
n
u= ∑ α ju j,
j=1
so that there exists c > 0 which does not depend on u, such that
kuk1 ≥ c(|α1 | + · · · + |αn |).

On the other hand,

n n
K
kuk0 ≤ ∑ kα j |ku j k0 ≤ K ∑ |α j | ≤ kuk1 , ∀u ∈ V,
j=1 j=1 c
where K = max j∈{1,...n} {ku j k0 }.

Interchanging the roles of k · k0 and k · k1 we may obtain K1 , c1 > 0 such that
K1
ku1 k1 ≤ kuk0 , ∀u ∈ V.
c1
c K1
Denoting α = K and β = c1 , we have obtained,
αkuk0 ≤ kuk1 ≤ β kuk0 , ∀u ∈ V.

Chapter 3
Hilbert Spaces
3.1 Introduction
At this point we introduce an important class of spaces namely, the Hilbert spaces, which are a special class
of metric spaces.
3.2 The main definitions and results

Definition 3.2.1 Let H be a vector space. We say that H is a real pre-Hilbert space if there exists a function
(·, ·)H : H × H → R such that
1. (u, v)H = (v, u)H , ∀u, v ∈ H,
2. (u + v, w)H = (u, w)H + (v, w)H , ∀u, v, w ∈ H,
3. (αu, v)H = α(u, v)H , ∀u, v ∈ H, α ∈ R,
4. (u, u)H ≥ 0, ∀u ∈ H, and (u, u)H = 0, if and only if u = θ .
Remark 3.2.2 The function (·, ·)H : H × H → R is called an inner-product.
Proposition 3.2.3 (Cauchy-Schwartz inequality) Let H be a pre-Hilbert space. Defining
p
kukH = (u, u)H , ∀u ∈ H,
we have
|(u, v)H | ≤ kukH kvkH , ∀u, v ∈ H.
Equality holds if and only if u = αv for some α ∈ R or v = θ .
Proof 3.1 If v = θ the inequality is immediate. Assume v 6= θ . Given α ∈ R we have

0 ≤ (u − αv, u − αv)H
= (u, u)H + α 2 (v, v)H − 2α(u, v)H
= kuk2H + α 2 kvk2H − 2α(u, v)H . (3.1)
Hilbert Spaces 53
In particular for α = (u, v)H /kvk2H , we obtain

2
(u, v)H
0 ≤ kuk2H − ,
kvk2H
that is
|(u, v)H | ≤ kukH kvkH .
The proof of the remaining conclusions is left as an exercise.
Proposition 3.2.4 On a pre-Hilbert space H, the function
k · kH : H → R
is a norm, where as above p

kukH = (u, u).
Proof 3.2 The only non-trivial property to be verified, concerning the definition of norm, is the triangle
inequality.
Observe that, given u, v ∈ H, from the Cauchy-Schwartz inequality we have,
ku + vk2H = (u + v, u + v)H
= (u, u)H + (v, v)H + 2(u, v)H
≤ (u, u)H + (v, v)H + 2|(u, v)H |
≤ kuk2H + kvk2H + 2kukH kvkH
= (kukH + kvkH )2 . (3.2)
Therefore
ku + vkH ≤ kukH + kvkH , ∀u, v ∈ H.
Definition 3.2.5 A pre-Hilbert space H is said to be a Hilbert space if it is complete, that is, if any cauchy
sequence in H converges to an element of H.
Definition 3.2.6 (Orthogonal Complement) Let H be a Hilbert space. Considering M ⊂ H we define its
orthogonal complement, denoted by M ⊥ , by
M ⊥ = {u ∈ H | (u, m)H = 0, ∀m ∈ M}.
Theorem 3.2.7 Let H be a Hilbert space, M a closed subspace of H and suppose u ∈ H. Under such hy-
potheses there exists a unique m0 ∈ M such that
ku − m0 kH = min {ku − mkH }.

m∈M
Moreover, n0 = u − m0 ∈ M ⊥ so that
u = m0 + n0 ,
where m0 ∈ M and n0 ∈ M⊥. Finally, such a representation through M ⊕ M ⊥ is unique.
Proof 3.3 Define d by

d = inf {ku − mkH }.
m∈M
Let {mi } ⊂ M be a sequence such that
ku − mi kH → d, as i → ∞.
Thus, from the parallelogram law we have
kmi − m j k2H = kmi − u − (m j − u)k2H

= 2kmi − uk2H + 2km j − uk2H
−k − 2u + mi + m j k2H
= 2kmi − uk2H + 2km j − uk2H
−4k − u + (mi + m j )/2k2H
2
≤ 2kmi − ukH + 2km j − uk2H − 4d 2
→ 2d 2 + 2d 2 − 4d 2 = 0, as i, j → +∞. (3.3)
Thus {mi } ⊂ M is a Cauchy sequence. Since M is closed, there exists m0 ∈ M such that
mi → m0 , as i → +∞,
so that
ku − mi kH → ku − m0 kH = d.
Define
n0 = u − m0 .
We will prove that n0 ∈ M ⊥ .
Pick m ∈ M and t ∈ R, thus we have
d2 ≤ ku − (m0 − tm)k2H
= kn0 + tmk2H
= kn0 k2H + 2(n0 , m)H t + kmk2H t 2 . (3.4)
Since
kn0 k2H = ku − m0 k2H = d 2 ,
we obtain
2(n0 , m)H t + kmk2H t 2 ≥ 0, ∀t ∈ R
so that
(n0 , m)H = 0.
Being m ∈ M arbitrary, we obtain
n0 ∈ M ⊥ .
It remains to prove the uniqueness. Let m ∈ M, thus
ku − mk2H = ku − m0 + m0 − mkH 2
= ku − m0 k2H + km − m0 k2H , (3.5)
since
(u − m0 , m − m0 )H = (n0 , m − m0 )H = 0.
Hilbert Spaces 55
From (3.5) we obtain

ku − mk2H > ku − m0 k2H = d 2 ,
if m 6= m0 .
Therefore m0 is unique.
Now suppose
u = m1 + n1 ,
where m1 ∈ M and n1 ∈ M ⊥ . As above, for m ∈ M
ku − mk2H = ku − m1 + m1 − mk2H
= ku − m1 k2H + km − m1 k2H ,
≥ ku − m1 kH (3.6)
and thus since m0 such that

d = ku − m0 kH
is unique, we get
m1 = m0
and therefore
n1 = u − m0 = n0 .
Theorem 3.2.8 (The Riesz Lemma) Let H be a Hilbert space and let f : H → R be a continuous linear
functional. Then there exists a unique u0 ∈ H such that
f (u) = (u, u0 )H , ∀u ∈ H.
Moreover
k f kH ∗ = ku0 kH .
Proof 3.4 Define N by

N = {u ∈ H | f (u) = 0}.
Thus, as f is a continuous and linear N is a closed subspace of H. If N = H, then f (u) = 0 = (u, θ )H , ∀u ∈ H
and the proof would be complete. Thus assume N 6= H. By the last theorem there exists v 6= θ such that
v ∈ N⊥.
Define
f (v)
u0 = v.
kvk2H
Thus if u ∈ N we have
f (u) = 0 = (u, u0 )H = 0.
On the other hand if u = αv for some α ∈ R, we have
f (u) = α f (v)
f (v)(αv, v)H
=
kvk2H

f (v)v
= αv,
kvk2H H
= (αv, u0 )H . (3.7)
Therefore, f (u) equals (u, u0 )H in the space spanned by N and v. Now we show that this last space (then span
of N and v) is in fact H. Just observe that given u ∈ H we may write

f (u)v f (u)v
u = u− + . (3.8)
f (v) f (v)
Since
f (u)v
u− ∈N
f (v)
we have finished the first part of the proof, that is, we have proven that
f (u) = (u, u0 )H , ∀u ∈ H.
To finish the proof, assume u1 ∈ H is such that
f (u) = (u, u1 )H , ∀u ∈ H.
Thus,
ku0 − u1 k2H = (u0 − u1 , u0 − u1 )H

= (u0 − u1 , u0 )H − (u0 − u1 , u1 )H
= f (u0 − u1 ) − f (u0 − u1 ) = 0. (3.9)
Hence, u1 = u0 .
Let us now prove that
First observe that
k f kH ∗ = sup{ f (u) | u ∈ H, kukH ≤ 1}

= sup{|(u, u0 )H | | u ∈ H, kukH ≤ 1}
≤ sup{kukH ku0 kH | u ∈ H, kukH ≤ 1}
≤ ku0 kH . (3.10)
On the other hand,
k f kH ∗ = sup{ f (u) | u ∈ H, kukH ≤ 1}

u0
≥ f
ku0 kH
(u0 , u0 )H
=
ku0 kH
= ku0 kH . (3.11)
From (3.10) and (3.11)

Remark 3.2.9 Similarly as above, we may define a Hilbert space H over C, that is, a complex one. In this
case the complex inner product (·, ·)H : H × H → C is defined through the following properties:
1. (u, v)H = (v, u)H , ∀u, v ∈ H,
Hilbert Spaces 57
2. (u + v, w)H = (u, w)H + (v, w)H , ∀u, v, w ∈ H,

3. (αu, v)H = α(u, v)H , ∀u, v ∈ H, α ∈ C,
4. (u, u)H ≥ 0, ∀u ∈ H, and (u, u) = 0, if and only if u = 0.
Observe that in this case we have
(u, αv)H = α(u, v)H , ∀u, v ∈ H, α ∈ C,
where for α = a + bi ∈ C, we have α = a − bi. Finally, similar results as those proven above are valid for
complex Hilbert spaces.
3.3 Orthonormal basis

In this section we study separable Hilbert spaces and the related orthonormal bases.
Definition 3.3.1 Let H be a Hilbert space. A set S ⊂ H is said to orthonormal if
kukH = 1,
and
(u, v)H = 0, ∀u, v ∈ S, such that u 6= v.
If S is not properly contained in any other orthonormal set, it is said to be an orthonormal basis for H.
Theorem 3.3.2 Let H be a Hilbert space and let {un }Nn=1 be an orthonormal set. Then for all u ∈ H, we have
2
N N
kuk2H = ∑ |(u, un )H | 2
+ u − ∑ (u, un )H un .
n=1 n=1 H
Proof 3.5 Observe that !

N N
u= ∑ (u, un )H un + u − ∑ (u, un )H un .
n=1 n=1
Furthermore, we may easily obtain that

N N
∑ (u, un )H un and u − ∑ (u, un )H un
n=1 n=1
are orthogonal vectors so that
kuk2H = (u, u)H

2 2
N N
= ∑ |(u, un )H un + u − ∑ (u, un )H un
n=1 H n=1 H
2
N N
= ∑ |(u, un )H |2 + u − ∑ (u, un )H un . (3.12)
n=1 n=1 H
Corollary 3.3.3 (Bessel inequality) Let H be a Hilbert space and let {un }Nn=1 be an orthonormal set. Then
for all u ∈ H, we have
N
kuk2H ≥ ∑ |(u, un )H |2 .
n=1
Theorem 3.3.4 Each Hilbert space has an orthonormal basis.
Proof 3.6 Define by C the collection of all orthonormal sets in H. Define an order in C by stating S1 ≺ S2
if S1 ⊂ S2 . Then C is partially ordered and obviously non-empty, since
6 θ.
v/kvkH ∈ C, ∀v ∈ H, v =
Now let {Sα }α ∈L be a linearly ordered subset of C. Clearly ∪α ∈L Sα is an orthonormal set which is an
upper bound for {Sα }α∈L .
Therefore, every linearly ordered subset has an upper bound, so that by Zorn’s lemma C has a maximal
element, that is, an orthonormal set not properly contained in any other orthonormal set.
Theorem 3.3.5 Let H be a Hilbert space and let S = {uα }α∈L be an orthonormal basis. Then for each v ∈ H
we have
v = ∑ (uα , v)H uα ,
α∈L
and
kvk2H = ∑ |(uα , v)H |2 .
α∈L
Proof 3.7 Let v ∈ H. Let L0 ⊂ L a finite subset of L. From Bessel’s inequality we have,
∑ |(uα , v)H |2 ≤ kvk2H .

α ∈L0
From this, we may infer that the set An = {α ∈ L | |(uα , v)H | > 1/n} is finite, so that
∞
A = {α ∈ L | |(uα , v)H | > 0} = ∪n=1 An
is at most countable.
6 0 for at most countably many α 0 s ∈ L, which we order by {αn }n∈N . Since the sequence
Thus (uα , v)H =
N
sN = ∑ |(uαi , v)H |2 ,
i=1
is monotone and bounded, it is converging to some real limit as N → ∞. Define

n
vn = ∑ (uαi , v)H uαi ,
i=1
so that for n > m we have

2
n
kvn − vm k2H = ∑ (uαi , v)H uαi
i=m+1 H
n
= ∑ |(uαi , v)H |2
i=m+1
= |sn − sm |. (3.13)
Hence, {vn } is a Cauchy sequence which converges to some v0 ∈ H.

Hilbert Spaces 59
Observe that
N
(v − v0 , uαl )H = lim (v − ∑ (uαi , v)H uαi , uαl )H
N →∞
i=1
= (v, uαl )H − (v, uαl )H
= 0. (3.14)
Also, if α 6= αl , ∀l ∈ N then
∞
(v − v0 , uα )H = lim (v − ∑ (uαi , v)H uαi , uα )H = 0.
N→∞
i=1
Hence,
v − v0 ⊥uα , ∀α ∈ L.
If
v − v0 =
6 θ,
then we could obtain an orthonormal set
v − v0

uα , α ∈ L,
kv − v0 kH
which would properly contain the complete orthonormal set
{uα , α ∈ L},
a contradiction.
Therefore, v − v0 = θ , that is
N
v = lim∑ (uαi , v)H uαi .
N→∞
i=1
3.3.1 The Gram-Schmidt orthonormalization

Let H be a Hilbert space and and {un } ⊂ H be a sequence of linearly independent vectors. Consider the
procedure:
w1
w1 = u1 , v1 = ,
kw1 kH
w2
w2 = u2 − (v1 , u2 )H v1 , v2 = ,
kw2 kH
and inductively,
n−1
wn
wn = un − ∑ (vk , un )H vk , vn = , ∀n ∈ N, n > 2.
k=1 kw n kH
Observe that clearly {vn } is an orthonormal set and for each m ∈ N, {vk }m m
k=1 and {uk }k=1 span the same
vector subspace of H.
Such a process of obtaining the orthonormal set {vn } is known as the Gram-Schmidt orthonormalization.
We finish this section with the following theorem.
Theorem 3.3.6 A Hilbert space H is separable if and only if has a countable orthonormal basis. If dim(H) =
N < ∞, the H is isomorphic to CN . If dim(H) = +∞ then H is isomorphic to l 2 , where
( )
∞
l2 = {yn } | yn ∈ C, ∀n ∈ N and ∑ |yn |2 < +∞ .
n=1
Proof 3.8 Suppose H is separable and let {un } be a countable dense set in H. To obtain an orthonormal
basis it suffices to apply the Gram-Schmidt orthonormalization procedure to the greatest linearly independent
subset of {un }.
Conversely, if B = {vn } is an orthonormal basis for H, the set of all finite linear combinations of elements
of B with rational coefficients are dense in H, so that H is separable.
Moreover, if dim(H) = +∞ consider the isomorphism F : H → l 2 given by
F(u) = {(un , u)H }n∈N .
Finally, if dim(H) = N < +∞, consider the isomorphism F : H → CN given by
F(u) = {(un , u)H }Nn=1 .
3.4 Projection on a convex set

Theorem 3.4.1 Let H be a Hilbert space and let K ⊂ H be a non-empty, closed and convex set. Under such
hypotheses, for each f ∈ H there exists a unique u ∈ K such that
k f − ukH = min k f − vkH .

v∈K
Moreover, u ∈ K is such that

( f − u, v − u)H ≤ 0, ∀v ∈ K.
Proof 3.9 Define

d = inf k f − vkH .
v∈K
Hence, for each n ∈ N there exists vn ∈ K such that
d ≤ k f − vn kH < d + 1/n.
Let m, n ∈ N. Define a = f − vn and b = f − vm . From the parallelogram law, we have
ka + bk2H + ka − bk2H = 2(kak2H + kbk2H ),
that is,
k2 f − (vn + vm )k2H + kvn − vm k2H = 2(k f − vn k2H + k f − vm k2H ),
so that
!
2
vn + vm
kvn − vm k2H = −4 f− + 2(k f − vn k2H + k f − vm k2H )
2 H
≤ −4d + 2(d + 1/n) + (2(d + 1/m)2

2 2
→ −4d 2 + 2d 2 + 2d 2
= 0, as m, n → ∞. (3.15)
Hence {vn } is a Cauchy sequence so that there exists u ∈ K such that
kvn − ukH → 0, as n → ∞.
Hilbert Spaces 61
Therefore,
k f − vn kH → k f − ukH = d, as n → ∞.
Now let v ∈ K and t ∈ [0, 1]. Define
w = (1 − t)u + tv.
Observe that, since K is convex, w ∈ K, ∀t ∈ [0, 1].
Hence,
k f − uk2H ≤ k f − wk2H
= k f − (1 − t )u − tvk2H
= k( f − u) + t(u − v)k2H
= k f − uk2H + 2( f − u, u − v)H t + t 2 ku − vk2H . (3.16)
From this, we obtain

t
( f − u, v − u)H ≤ ku − vkH , ∀t ∈ [0, 1].
2
Letting t → 0+ we get
( f − u, v − u)H ≤ 0, ∀v ∈ K.
Corollary 3.4.2 In the context of the last theorem, assume
( f − u, v − u) ≤ 0, ∀v ∈ K.

v∈K
Finally, such a u ∈ K is unique.
Proof 3.10 Let v ∈ K. Thus,
k f − uk2H − k f − vk2H = k f k2H − 2( f , u)H + kuk2H

−k f k2H + 2( f , v)H − kvk2H
−ku − vk2H + kukH 2 2
− 2(u, v)H + kvkH
= 2(u, u)H − 2(u, v)H + 2( f , v)H − 2( f , u)H − ku − vk2H
= 2( f − u, v − u)H − ku − vk2H
≤ 0. (3.17)
Summarizing,
k f − vkH ≥ k f − ukH , ∀v ∈ K.
Suppose now that u1 , u2 ∈ K be such that
( f − u1 , v − u1 )H ≤ 0, ∀v ∈ K,
( f − u2 , v − u2 )H ≤ 0, ∀v ∈ K.
With v = u2 in the last first inequality and v = u1 in the last second one, we get
( f − u1 , u2 − u1 )H ≤ 0
and
( f − u2 , u1 − u2 )H ≤ 0.
Adding these two last inequalities, we obtain
( f , u2 − u1 )H + ( f , u1 − u2 )H − (u1 , u2 − u1 )H + (u2 , u2 − u1 )H ≤ 0,
that is,
ku2 − u1 k2H ≤ 0,
so that
ku1 − u2 kH = 0,
and therefore,
u1 = u2 .
Hence, the u ∈ K in question is unique.
Proposition 3.4.3 Let H be a Hilbert space and K ⊂ H a non-empty, closed and convex set.
Let f ∈ H. Define PK ( f ) = u where u ∈ K is such that

v∈K
kPK f1 − PK f2 kH ≤ k f1 − f2 kH , ∀ f1 , f2 ∈ H.
Proof 3.11 Let f1 , f2 ∈ H and u1 = PK f1 and u2 = PK f2 .

From the last proposition,
( f1 − u1 , v − u1 )H ≤ 0, ∀v ∈ K
and
( f2 − u2 , v − u2 )H ≤ 0, ∀v ∈ K.
With v = u2 in the first last inequality and v = u1 in the last second one, we obtain
( f1 − u1 , u2 − u1 )H ≤ 0,
( f2 − u2 , u1 − u2 )H ≤ 0.
Adding these two last inequalities, we get
( f1 , u2 − u1 )H − ( f2 , u2 − u1 )H + (u2 − u1 , u2 − u1 )H ≤ 0,
that is,
ku2 − u1 k2H ≤ ( f2 − f1 , u2 − u1 )H
≤ k f2 − f1 kH ku2 − u1 kH , (3.18)
so that
ku2 − u1 kH ≤ k f2 − f1 kH .
Hilbert Spaces 63
Corollary 3.4.4 Let H be a Hilbert space and let M ⊂ H be a closed vector subspace of H.
Let f ∈ H. Thus, u = PM ( f ) is such that u ∈ M and
( f − u, v)H = 0, ∀v ∈ M.
Proof 3.12 In the previous results, we have got,
( f − u, v − u)H ≤ 0, ∀v ∈ M.
Let v ∈ M be such that v 6= 0.
Thus,
( f − u,tv − u)H ≤ 0, ∀t ∈ R.
Hence,
t( f − u, v)H ≤ ( f − u, u)H , ∀t ∈ R.
From this we obtain
( f − u, v)H = 0, ∀v ∈ M.
Remark 3.4.5 Reciprocally, if

( f − u, v)H = 0, ∀v ∈ M,
then,
( f − u, v − u)H = 0, ∀v ∈ M,
so that
u = PM ( f ).
3.5 The theorems of Stampacchia and Lax-Milgram

In this section we present the statement and proof of two well-known results, namely, the Stampacchia and
Lax-Milgram theorems.
Definition 3.5.1 Let a : H × H → R be a bilinear form.
1. We say that a is bounded if there exists c > 0 such that
|a(u, v)| ≤ ckukH kvkH , ∀u, v ∈ H.
2. We say that a is coercive if there exists α > 0 such that
|a(v, v)| ≥ αkvk2H , ∀v ∈ H.
Theorem 3.5.2 (Stampacchia) Let H be a Hilbert space and let a : H × H → R be a bounded and coercive
bilinear form.
Let K ⊂ H be a non-empty, closed and convex set. Under such hypotheses, for each f ∈ H there exists a
unique u ∈ K such that
a(u, v − u) ≥ ( f , v − u)H , ∀v ∈ K. (3.19)
Moreover, if a is symmetric, that is, a(u, v) = a(v, u), ∀u, v ∈ H, such u ∈ K in question is also such that

1 a(v, v)
a(u, u) − ( f , u)H = min − ( f , v)H . (3.20)
2 v∈K 2
Proof 3.13 Fix u ∈ H. The function

v 7→ a(u, v), ∀v ∈ H,
is continuous and linear.
Hence from the Riesz representation theorem, there exists a unique vector denoted by A(u) ∈ H such that
(A(u), v)H = a(u, v), ∀v ∈ H.
Clearly such an operator A is linear, and
|(A(u), v)H | = |a(u, v)| ≤ ckukH kvkH ,
for some c > 0, so that

kA(u)kH = sup{|(A(u), v)H | : kv||H ≤ 1} ≤ ckuk.H
v∈H
Moreover,
(Av, v)H = a(v, v) ≥ αkvk2H , ∀v ∈ H,
for some α > 0.
Let ρ > 0 to be specified. Let f ∈ H.
Define T (v) = PK (ρ f − ρA(v) + v).
Observe that
kT (v1 ) − T (v2 )k2H = kPK (ρ f − ρAv1 + v1 ) − PK (ρ f − ρv2 + v2 )k2H

2
≤ k − ρA(v1 ) + ρA(v2 ) + (v1 − v2 )kH
= kv1 − v2 kH − 2ρ(A(v1 − v2 ), v1 − v2 )H + ρ 2 kA(v1 − v2 )k2H
2
≤ (1 − 2αρ + cρ 2 )kv1 − v2 k2H . (3.21)
Let F(ρ ) = 1 − 2αρ + cρ 2 .

Thus, if F 0 (ρ0 ) = 0 then
−2α + 2ρ0 c = 0,
that is,
α
ρ0 = .
c
Therefore,
α2 α2 α2
F(ρ0 ) = 1 − 2
+c 2 = 1− .
c c c
Observe that we may redefine a larger c > 0 such that
α2
0 < 1− < 1.
c
Hence,
kT (v1 ) − T (v2 )kH ≤ λ kv1 − v2 kH ,
where r
α2
λ= 1−
< 1.
c
From this and Banach fixed point theorem, there exists u ∈ K such that
T (u) = u,
that is,
PK (ρ f − ρAu + u) = u.
Hilbert Spaces 65
From this and Theorem 3.4.1, we obtain

(ρ f − ρA(u) + u − u, v − u)H ≤ 0, ∀v ∈ K.
Thus,
a(u, v − u) = (A(u), v − u)H ≥ ( f , v − u)H , ∀v ∈ K.
Assume now that a(u, v) is also symmetric. Thus, a(u, v) define a inner product in H, inducing a norm
p
a(u, u)
which is equivalent to kukH , since
√ p √
ckukH ≥ a(u, u) ≥ αkukH , ∀u ∈ H.
From the Riesz representation theorem, there exists a unique g ∈ H such that
( f , v)H = a(g, v), ∀v ∈ H.
Similarly to the indicated above we may obtain u ∈ K such that
u = PK (g)
so that
a(g − u, v − u) ≤ 0, ∀v ∈ K.
Hence,
a(g, v − u) − a(u, v − u) ≤ 0,
that is,
a(u, v − u) ≥ a(g, v − u) = ( f , v − u), ∀v ∈ K.
Moreover from u = Pk (g) we obtain
a(g − u, g − u) = min a(g − v, g − v).
v∈K
Therefore,
a(g, g) − 2a(g, u) + a(u, u) ≤ a(g, g) − 2a(g, v) + a(v, v),
that is,
a(u, u) a(v, v)
− a(g, u) ≤ − a(g, v),
2 2
so that
a(u, u) a(v, v)
− ( f , u)H ≤ − ( f , v)H , ∀v ∈ K .
2 2
Corollary 3.5.3 (Lax-Milgram theorem) Assume a(u, v) is a bounded, coercive and symmetric bilinear
form on H. Under such hypotheses, there exists a unique u ∈ H such that
a(u, v) = ( f , v), ∀v ∈ H.
Moreover, such a u ∈ H is such that
1 1
a(u, u) − ( f , u)H = min a(v, v) − ( f , v)H .
2 v∈H 2
Proof 3.14 The proof follows from the Stampacchia theorem with K = H and Remark 3.4.5.
Chapter 4
The Hahn-Banach Theorems and

the Weak Topologies
4.1 Introduction
In this chapter we present the Hahn-Banach theorems and some important applications. Also, a study
on weak topologies is developed in details.
Finally, we highlight the main reference for this chapter is Brezis [24].
4.2 The Hahn-Banach theorems

In this chapter, U always denotes a Banach space.
Theorem 4.2.1 (The Hahn-Banach theorem) Consider a functional p : U → R such that
p(λ u) = λ p(u), ∀u ∈ U, λ > 0, (4.1)
and
p(u + v) ≤ p(u) + p(v), ∀u, v ∈ U. (4.2)
Let V ⊂ U be a proper subspace of U and let g : V → R be a linear functional such that
g(u) ≤ p(u), ∀u ∈ V. (4.3)
Under such hypotheses, there exists a linear functional f : U → R such that
g(u) = f (u), ∀u ∈ V, (4.4)
and
f (u) ≤ p(u), ∀u ∈ U. (4.5)
Proof 4.1 Choose z ∈ U \V . Denote by Ṽ the space spanned by V and z, that is,
Ṽ = {v + αz | v ∈ V and α ∈ R}. (4.6)
The Hahn-Banach Theorems and the Weak Topologies 67
We may define an extension of g from V to Ṽ , denoted by g̃, by
g̃(αz + v) = αg̃(z) + g(v), (4.7)
where g̃(z) will be properly specified in the next lines.

Let v1 , v2 ∈ V , α > 0, β > 0. Thus,
β g(v1 ) + αg(v2 ) = g(β v1 + αv2 )

β α
= (α + β )g v1 + v2
α +β α +β

β α
≤ (α + β )p (v1 − αz) + (v2 + β z)
α +β α +β
≤ β p(v1 − αz) + α p(v2 + β z) (4.8)
and therefore,
1 1
[−p(v1 − αz) + g(v1 )] ≤ [p(v2 + β z) − g(v2 )],
α β
∀v1 , v2 ∈ V, α, β > 0. Thus, there exists a ∈ R such that

1 1
sup (−p(v − αz) + g(v)) ≤ a ≤ inf ( p(v + αz) − g(v)) . (4.9)
v∈V,α >0 α v∈V,α>0 α
We shall define g̃(z) = a. Therefore, if α > 0, then
g̃(αz + v) = aα + g(v)

1
≤ (p(v + αz) − g(v)) α + g(v)
α
= p(v + αz). (4.10)
On the other hand, if α < 0, then −α > 0. Thus,

1
a≥ (−p(v − (−α)z) + g(v)),
−α
so that
g̃(αz + v) = aα + g(v)

1
≤ (−p(v + αz) + g(v)) α + g(v)
−α
= p(v + αz) (4.11)
and hence,
g̃(u) ≤ p(u), ∀u ∈ Ṽ .
Define now by E the set of all extensions e of g, which satisfy e(u) ≤ p(u) on the domain of e, where such a
domain is always a subspace of U. We shall also define a partial order for E denoting e1 ≺ e2 , as the domain
of e2 contains the domain of e1 and e1 = e2 on the domain of e1 . Let {eα }α ∈A be an ordered subset of E .
Let Vα be the domain of eα , ∀α ∈ A. Define e on ∪α ∈AVα by setting e = eα on Vα . Clearly, eα ≺ e, ∀α ∈ A
so that each ordered subset of E has an upper bound. From this and Zorn Lemma, E has a maximal element
f defined on some subspace Ũ ⊂ U such that f (u) ≤ p(u), ∀u ∈ Ũ. Suppose, to obtain contradiction, that
6 U and let z1 ∈ U \Ũ. As indicated above we may obtain an extension f1 from Ũ to the subspace spanned
Ũ =
by z1 and Ũ, which contradicts the maximality of f .
Definition 4.2.2 (Topological dual spaces) Let U be a Banach space. We will define its dual topological
space, as the set of all linear continuous functionals defined on U. We suppose such a dual space of U
may be represented by another vector space U ∗ , through a bilinear form h·, ·iU : U ×U ∗ → R (here we are
referring to standard representations of dual spaces of Sobolev and Lebesgue spaces, to be addressed in the
subsequent chapters). Thus, given f : U → R linear and continuous, we assume the existence of a unique
u∗ ∈ U ∗ such that
f (u) = hu, u∗ iU , ∀u ∈ U. (4.12)
The norm of f , denoted by k f kU ∗ , is defined as
k f kU ∗ = sup{|hu, u∗ iU | : kukU ≤ 1} ≡ ku∗ kU ∗ . (4.13)

u∈U
Corollary 4.2.3 Let V ⊂ U be a proper subspace of U and let g : V → R be a linear and continuous func-
tional with norm
kgkV ∗ = sup{|g(u)| | kukU ≤ 1}. (4.14)

u∈V
Under such hypotheses, there exists u∗ in U ∗ such that
hu, u∗ iU = g(u), ∀u ∈ V, (4.15)
and
ku∗ kU ∗ = kgkV ∗ . (4.16)
Proof 4.2 It suffices to apply Theorem 4.2.1 with p(u) = kgkV ∗ kukV . Indeed, from such a theorem, there
exists a linear functional f : U → R such that
f (u) = g(u), ∀u ∈ V
and
f (u) ≤ p(u) = kgkV ∗ kukU ,
that is,
| f (u)| ≤ p(u) = kgkV ∗ kukU , ∀u ∈ U.
Therefore,
k f kU ∗ = sup{| f (u)| : kukU ≤ 1} ≤ kgkV ∗ .
u∈U
On the other hand,
k f kU ∗ ≥ sup{| f (u)| : kukU ≤ 1} = kgkV ∗ .
u∈V
Thus,
k f kU ∗ = kgkV ∗ .
Finally, since f is linear and continuous, there exists u∗ ∈ U ∗ such that
f (u) = hu, u∗ iU , ∀u ∈ U,
and hence
hu, u∗ iU = f (u) = g(u), ∀u ∈ V.
Moreover,
ku∗ kU ∗ = k f kU ∗ = kgkV ∗ .
Corollary 4.2.4 Let u0 ∈ U. Under such hypotheses, there exists u∗0 ∈ U ∗ such that
ku∗0 kU ∗ = ku0 kU and hu0 , u∗0 iU = ku0 kU2 . (4.17)
Proof 4.3 It suffices to apply the Corollary 4.2.3 with V = {αu0 | α ∈ R} and g(tu0 ) = tku0 kU2 so that
kgkV ∗ = ku0 kU .
Indeed, from the last corollary, there exists u∗0 ∈ U ∗ such that
htu0 , u∗0 iU = g(tu0 ), ∀t ∈ R,
and
ku∗0 kU ∗ = kgkV ∗ ,
where,
kgkV ∗ = sup{tku0 kU2 : ktu0 kU ≤ 1} = ku0 kU .
t∈R
Moreover, also from the last corollary,
ku∗0 kU ∗ = kgkV ∗ = ku0 kU .
Finally,
htu0 , u∗0 iU = g(tu0 ) = tku0 kU2 , ∀t ∈ R,
so that
hu0 , u∗0 iU = ku0 kU2 .
Corollary 4.2.5 Let u ∈ U. Under such hypotheses
kukU = sup {|hu, u∗ iU | | ku∗ kU ∗ ≤ 1}. (4.18)

u∗ ∈U ∗
Proof 4.4 6 0, otherwise the result is immediate. Since

Suppose u =
|hu, u∗ iU | ≤ kukU ku∗ kU ∗ , ∀u ∈ U, u∗ ∈ U ∗
we have
sup {|hu, u∗ iU | | ku∗ kU ∗ ≤ 1} ≤ kukU . (4.19)

u∗ ∈U ∗
However, from the last corollary, there exists u∗0 ∈ U ∗ such that ku∗0 kU ∗ = kukU and hu, u∗0 iU = kukU2 . Define
u∗1 = kukU−1 u∗0 . Thus, ku∗1 kU = 1 and hu, u∗1 iU = kukU .
Definition 4.2.6 (Affine hyperplane) Let U be a Banach space. An affine hyperplane H is a set defined by
H = {u ∈ U | hu, u∗ iU = α} (4.20)
for some u∗ ∈ U ∗ and α ∈ R.
Proposition 4.2.7 An affine hyperplane H defined as above indicated is closed.
Proof 4.5 The result follows directly from the continuity of hu, u∗ iU as functional on U.
Definition 4.2.8 (Separation) Let A, B ⊂ U. We say that a hyperplane H, as above indicated separates A
and B, as there exist α ∈ R and u∗ ∈ U ∗ such that
hu, u∗ iU ≤ α, ∀u ∈ A, and hu, u∗ iU ≥ α, ∀u ∈ B. (4.21)
We say that H separates A and B strictly if there exists ε > 0 such that
hu, u∗ iU ≤ α − ε, ∀u ∈ A, and hu, u∗ iU ≥ α + ε, ∀u ∈ B, (4.22)
Theorem 4.2.9 (The Hahn-Banach theorem, the geometric form) Let A, B ⊂ U be two non-empty, convex
sets such that A ∩ B = 0/ and A is open. Under such hypotheses, there exists a closed hyperplane which
separates A and B, that is, there exist α ∈ R and u∗ ∈ U ∗ such that
hu, u∗ iU ≤ α ≤ hv, u∗ iU , ∀u ∈ A, v ∈ B.
To prove such a theorem, we need two lemmas.

Lemma 4.2.10 Let C ⊂ U be a convex set such that 0 ∈ C. For each u ∈ U define
p(u) = inf{α > 0, α −1 u ∈ C}. (4.23)
Under such hypotheses, p is such that there exists M ∈ R+ such that
0 ≤ p(u) ≤ MkukU , ∀u ∈ U, (4.24)
and
C = {u ∈ U | p(u) < 1}. (4.25)
Moreover,
p(u + v) ≤ p(u) + p(v), ∀u, v ∈ U.
Proof 4.6 Let r > 0 be such that B(0, r) ⊂ C. Let u ∈ U such that u 6= 0. Thus,
u
r ∈ B(0, r) ⊂ C,
kukU
and therefore
kukU
p(u) ≤ , ∀u ∈ U (4.26)
r
which proves (4.24). Suppose now u ∈ C. Since C is open there exists ε > 0 sufficiently small such that
(1 + ε )u ∈ C. Thus, p(u) ≤ 1+1 ε < 1. Reciprocally, if p(u) < 1, there exists 0 < α < 1 such that α −1 u ∈ C
and hence, since C is convex, we get u = α (α −1 u) + (1 − α)0 ∈ C.
u v tu (1−t )v
Finally, let u, v ∈ C and ε > 0. Thus, p(u)+ε ∈ C and p(v)+ε ∈ C so that p(u)+ε + p(v)+ε ∈ C, ∀t ∈ [0, 1].
p(u)+ε u+v
Particularly, for t = p(u)+p(v)+2ε we obtain p(u)+p(v)+2ε ∈ C, and thus,
p(u + v) ≤ p(u) + p(v) + 2ε, ∀ε > 0.
The proof of this lemma is complete.
Lemma 4.2.11 Let C ⊂ U be an non-empty, open and convex set and let u0 ∈ U such that u0 6∈ C. Under
such hypotheses, there exists u∗ ∈ U ∗ such that hu, u∗ iU < hu0 , u∗ iU , ∀u ∈ C
Proof 4.7 By translation, if necessary, there is no loos in generality in assuming 0 ∈ C. Consider the
functional p defined in the last lema. Define V = {αu0 | α ∈ R}. Define also g on V , by
g(tu0 ) = t, ∀t ∈ R. (4.27)
Let t ∈ R be such that t =

6 0. Since
tu0
= u0 6∈ C,
t
we have
g(tu0 ) = t ≤ p(tu0 )
and therefore
g(u) ≤ p(u), ∀u ∈ V.
From the Hahn-Banach theorem, there exists a linear functional f defined on U which extends g such that
f (u) ≤ p(u) ≤ MkukU . (4.28)
Here, we have applied the Lemma 4.2.10. In particular, f (u0 ) = g(u0 ) = g(1u0 ) = 1, also from the last
lemma, f (u) < 1, ∀u ∈ C. The existence of u∗ satisfying this lemma conclusion follows from the continuity
of f , indicated in (4.28).
Proof of Theorem 4.2.9. Define C = A + (−B) so that Cis convex and 0 6∈ C. From Lemma 4.2.11, there
exists u∗ ∈ U ∗ such that hw, u∗ iU < 0, ∀w ∈ C, and thus,
hu, u∗ iU < hv, u∗ iU , ∀u ∈ A, v ∈ B. (4.29)
Therefore, there exists α ∈ R such that
suphu, u∗ iU ≤ α ≤ infhv, u∗ iU , (4.30)

u∈A v∈B
which completes the proof.
Proposition 4.2.12 Let U be a Banach space and let A, B ⊂ U be such that A is compact, B is closed and
A ∩ B = 0/.
Under such hypotheses, there exists ε1 > 0 such that
[A + Bε1 (0)] ∩ [B + Bε1 (0)] = 0/.
Proof 4.8 Suppose, to obtain contradiction, the proposition conclusion is false.

Thus, for each n ∈ N there exists un ∈ U such that d(un , A) < 1n and d(un , B) < 1n .
Therefore, there exist vn ∈ A and wn ∈ B such that
1
kun − vn kU < (4.31)
n
and
1
kun − wn kU < , ∀n ∈ N. (4.32)
n
Since {vn } ⊂ A and A is compact, there exist a subsequence {vn j } of {vn } and v0 ∈ A, such that
kvn j − v0 kU → 0, as j → ∞.
Thus, from this, (4.31) and (4.32) we obtain,
kun j − v0 kU → 0, as j → ∞,
and
kwn j − v0 kU → 0, as j → ∞.
Since A and B are closed we may infer that
v0 ∈ A ∩ B,
which contradicts A ∩ B = 0/.

Theorem 4.2.13 (The Hahn-Banach theorem, the second geometric form) Let A, B ⊂ U be two non-
empty, convex sets such that A ∩ B = 0/. Suppose A is compact and B is closed. Under such hypotheses,
there exists an hyperplane which separates A and B strictly.
Proof 4.9 Observe that, from the last proposition, there exists ε > 0 sufficiently small such that Aε =
A + B(0, ε) and Bε = B + B(0, ε) are disjoint and convex sets. From Theorem 4.2.9, there exists u∗ ∈ U ∗ such
that u∗ 6= 0 and
hu + εw1 , u∗ iU ≤ hu + εw2 , u∗ iU , ∀u ∈ A, v ∈ B, w1 , w2 ∈ B(0, 1). (4.33)
Thus, there exists α ∈ R such that
hu, u∗ iU + εku∗ kU ∗ ≤ α ≤ hv, u∗ iU − εku∗ kU ∗ , ∀u ∈ A, v ∈ B. (4.34)
Corollary 4.2.14 Suppose V ⊂ U is a vector subspace such that V 6= U. Under such hypotheses, there exists
u∗ ∈ U ∗ such that u∗ 6= 0 and
hu, u∗ iU = 0, ∀u ∈ V. (4.35)
Proof 4.10 Let u0 ∈ U be such that u0 6∈ V . Applying Theorem 4.2.9 to A = V and B = {u0 } we obtain
u∗ ∈ U ∗ and α ∈ R such that u∗ 6= 0 e
hu, u∗ iU < α < hu0 , u∗ iU , ∀u ∈ V. (4.36)
Since V is a subspace, we must have hu, u∗ iU = 0, ∀u ∈ V .
4.3 The weak topologies

Definition 4.3.1 (Weak neighborhoods) Let U be a Banach space and let u0 ∈ U. We define a weak neigh-
borhood of u0 , denoted by Vw (u0 ), as
Vw (u0 ) = {u ∈ U | |hu − u0 , u∗i iU | < εi , ∀i ∈ {1, ..., m}}, (4.37)
for some m ∈ N, εi > 0, and u∗i ∈ U ∗ , ∀i ∈ {1, ..., m}.

Let A ⊂ U. We say that u0 ∈ A is weakly interior to A, as there exists a weak neighborhood Vw (u0 ) of u0
contained in A.
If all points of A are weakly interior, we say that A is weakly open.
Finally, we define the weak topology σ (U,U ∗ ) for U, as the set of all subsets weakly open of U.
Proposition 4.3.2 A Banach space U is Hausddorff as endowed with the weak topology σ (U,U ∗ ).
Proof 4.11 Choose u1 , u2 ∈ U such that u1 6= u2 . From the Hahn-Banach theorem, second geometric form,
there exists an hyperplane separating {u1 } and {u2 } strictly, thats is, there exist u∗ ∈ U ∗ and α ∈ R such that
hu1 , u∗ iU < α < hu2 , u∗ iU . (4.38)
Define
Vw1 (u1 ) = {u ∈ U | |hu − u1 , u∗ i| < α − hu1 , u∗ iU }, (4.39)
and
Vw2 (u2 ) = {u ∈ U | |hu − u2 , u∗ iU | < hu2 , u∗ iU − α}. (4.40)
We claim that
Vw1 (u1 ) ∩Vw2 (u2 ) = 0/.
Suppose, to obtain contradiction, there exists u ∈ Vw1 (u1 ) ∩Vw2 (u2 ).
Thus,
hu − u1 , u∗ iU < α − hu1 , u∗ iU ,
and therefore
hu, u∗ iU < α.
Also
−hu − u2 , u∗ iU < hu2 , u∗ iU − α,
and hence
hu, u∗ iU > α.
We have got
hu, u∗ iU < α < hu, u∗ iU ,
a contradiction.
Summarizing, we have obtained u1 ∈ Vw1 (u1 ), u2 ∈ Vw2 (u2 ) and Vw1 (u1 ) ∩ Vw2 (u2 ) = 0.
/
Remark 4.3.3 If {un } ∈ U is such that un converges to u in σ (U,U ∗ ), then we write un * u, weakly.
Proposition 4.3.4 Let U be a Banach space. For a sequence {un } ⊂ U, we have

1. un * u, for σ (U,U ∗ ) ⇔ hun , u∗ iU → hu, u∗ iU , ∀u∗ ∈ U ∗ ,
2. If un → u strongly (in norm), then un * u weakly,
3. If un * u weakly, then {kun kU } is bounded and kukU ≤ lim inf kun kU ,
n→∞
4. If un * u weakly and u∗n → u∗ strongly (in norm) in U ∗ , then hun , u∗n iU → hu, u∗ iU .
Proof 4.12
1. The result follows from the definition of σ (U,U ∗ ).
Indeed, suppose that {un } ⊂ U and un * u, weakly.
Let u∗ ∈ U ∗ and let ε > 0.
Define
Vw (u) = {v ∈ U : |hv − u, u∗ iU | < ε}.
From the hypotheses, there exists n0 ∈ N such that if n > n0 , then
un ∈ Vw (u).
That is,
|hun − u, u∗ iU | < ε,
if n > n0 .
Therefore,
hun , u∗ iU → hu, u∗ iU , as n → ∞
∀u∗ ∈ U ∗ .
Reciprocally, suppose that
hun , u∗ iU → hu, u∗ iU , as n → ∞
∀u∗ ∈ U ∗ .
Let V (u) ∈ σ (U,U ∗ ) be a set which contains {u}.
Thus, there exists a weak neighborhood a Vw (u) such that u ∈ Vw (u) ⊂ V (u), where there exist m ∈ N,
εi > 0 and u∗i ∈ U ∗ such that
Vw (u) = {v ∈ U : |hv − u, u∗i iU | < εi , ∀i ∈ {1, · · · , m}}.
From the hypotheses, for each i ∈ {1, · · · m}, there exists ni ∈ N such that if n > ni , then
|hun − u, u∗i iU | < εi .
Define n0 = max{n1 , · · · , nm }.
Thus,
un ∈ Vw (u) ⊂ V (u), if n > n0 .
From this we may infer that un * u for σ (U,U ∗ ).

2. This follows from the inequality
|hun , u∗ iU − hu, u∗ iU | ≤ ku∗ kU ∗ kun − ukU . (4.41)
3. For each u∗ ∈ U ∗ the sequence {hun , u∗ iU } is convergent for some bounded sequence. From this and
the Uniform Boundedness Principle, there exists M > 0 such that kun kU ≤ M, ∀n ∈ N. Moreover, for
u∗ ∈ U ∗ , we have
|hun , u∗ iU | ≤ ku∗ kU ∗ kun kU , (4.42)
and letting n → ∞, we obtain
|hu, u∗ iU | ≤ lim inf ku∗ kU ∗ kun kU . (4.43)

n→∞
Thus,
kukU = sup {|hu, u∗ iU | : kukU ∗ ≤ 1} ≤ lim inf kun kU . (4.44)

u∗ ∈U ∗ n→∞
4. Just observe that
|hun , u∗n iU − hu, u∗ iU | ≤ |hun , u∗n − u∗ iU |

+|hu − un , u∗ iU |
≤ ku∗n − u∗ kU ∗ kun kU
+|hun − u, u∗ iU |
≤ Mku∗n − u∗ kU ∗
+|hun − u, u∗ iU |
→ 0, as n → ∞. (4.45)
Theorem 4.3.5 Let U be a Banach space and let A ⊂ U be a non-empty convex set. Under such hypotheses,
A is closed for the topology σ (U,U ∗ ) if, and only if A is closed for the topology induced by k · kU .
Proof 4.13 If A = U, the result is immediate. Thus, assume A =6 U. Suppose that A is strongly closed. Let
u0 6∈ A. From the Hahn-Banach theorem there exists a closed hyperplane which separates u0 and A strictly,
that is, there exist α ∈ R and u∗ ∈ U ∗ such that
hu0 , u∗ iU < α < hv, u∗ iU , ∀v ∈ A. (4.46)
Define
V = {u ∈ U | hu, u∗ iU < α}, (4.47)
so that u0 ∈ V , V ⊂ U \ A.
Let
Vw (u0 ) = {v ∈ U : |hv − u0 , u∗ iU | < α − hu0 , u∗ iU .
Let v ∈ Vw (u0 ).
Thus,
hv, u∗ iU = hv − u0 + u0 , u∗ iU
= hv − u0 , u∗ iU + hu0 , u∗ iU
≤ |hv − u0 , u∗ iU | + hu0 , u∗ iU
< α − hu0 , u∗ iU + hu0 , u∗ iU
= α. (4.48)
From this we may infer that Vw (u0 ) ⊂ V ⊂ U \ A, that is, u0 is an interior point for σ (U,U ∗ ) of U \ A,
∀u0 ∈ U \ A
Therefore, V is weakly open.
Summarizing, U \ A is open in σ (U,U ∗ ) and thus A is closed for σ (U,U ∗ ) (weakly closed).
Finally, the reciprocal is immediate.
Theorem 4.3.6 Let (Z, σ ) be a topological space and let U be a Banach space. Let φ : Z → U be a function,
considering U with the weak topology σ (U,U ∗ ).
Under such hypotheses, φ is continuous if, and only if, fu∗ : Z → R, where
fu∗ (z) = hφ (z), u∗ iU
is continuous, ∀u∗ ∈ U ∗ .
Proof 4.14 Assume φ is continuous. Let z0 ∈ Z and let {zα }α∈I be a net such that
zα → z0 .
From the hypotheses,

φ (zα ) * φ (z0 ), in σ (U,U ∗ ).
Therefore,
hφ (zα ), u∗ iU → hφ (z0 ), u∗ iU , ∀u∗ ∈ U ∗ .
Thus, fu∗ is continuous at z0 , ∀u∗ ∈ U ∗ , ∀z0 ∈ Z.
Reciprocally, assume fu∗ : Z → R, where
fu∗ (z) = hφ (z), u∗ iU
is continuous, ∀u∗ ∈ U ∗ .
Suppose, to obtain contradiction, that φ is not continuous.
Thus, there exists z0 ∈ Z such that φ is not continuous at z0 .
In particular, there exists a net {zα }α∈I such that zα → z0 and we do not have
φ (zα ) * φ (z0 ), in σ (U,U ∗ ).
Hence, there exists u∗ ∈ U ∗ such that we do not have
hφ (zα ), u∗ iU → hφ (z0 ), u∗ iU ,
and thus fu∗ is not continuous at z0 , a contradiction.

Therefore, φ is continuous.
4.4 The weak-star topology

Definition 4.4.1 (Reflexive spaces) Let U be a Banach space. We say that U is reflexive, if the canonical
injection
J : U → U ∗∗
is onto, where
hu, u∗ iU = hu∗ , J(u)iU ∗ , ∀u ∈ U, u∗ ∈ U ∗ .
Thus, if U is reflexive, we may identify the bi-dual space of U, U ∗∗ , with U.
The weak topology for U ∗ may be defined similarly to σ (U,U ∗ ) and it is denoted by σ (U ∗ ,U ∗∗ ).
We define as well, the weak-star topology for U ∗ , denoted by σ (U ∗ ,U), as it follows.
Firstly, we define weak-star neighborhoods.
Let u∗0 ∈ U ∗ . We define a weak-star neighborhood for u∗0 , denoted by Vw (u∗0 ), as
Vw (u∗0 ) = {u∗ ∈ U ∗ : |hui , u∗ − u∗0 iU | < εi , ∀i ∈ {1, · · · , m}},
where m ∈ N, εi > 0 and ui ∈ U, ∀i ∈ {1, · · · , m}.

Let A ⊂ U ∗ . We say that u∗0 ∈ A is weakly-star interior to A, as there exists a weak-star neighborhood
Vw (u∗0 ) contained in A.
If all point of A are weakly-star interior, we say that A weakly-star open.
Finally, we define the weak-star topology σ (U ∗ ,U) for U ∗ , as the set of all subsets weakly-star open of
∗
U .
Observe that σ (U ∗ ,U ∗∗ ) and σ (U ∗ ,U) coincide if U is reflexive.
4.5 Weak-star compactness

Theorem 4.5.1 (Banach and Alaoglu) Let U be a Banach space. Denote
BU ∗ = {u∗ ∈ U ∗ : ku∗ kU ∗ ≤ 1}.
Under such hypotheses, BU ∗ is compact for U ∗ with the weak-star topology σ (U ∗ ,U).
Proof 4.15 For each u ∈ U, we shall associate a real number ωu and denote
ω= ∏ ωu ∈ RU ,
u∈U
and consider the projections

Pu : RU → R
where
Pu (ω) = ωu , ∀ω ∈ RU , u ∈ U.
We shall define a topology for RU , which is induced by the weak neighborhoods specified in the next
lines.
Let ω̃ ∈ RU . We define a weak neighborhood Ṽ (ω̃) of ω̃ as
Ṽ (w̃) = {ω ∈ RU : |Pui (ω) − Pui (ω̃)| < εi , ∀ ∈ {1, · · · , m}},

Let A ⊂ RU . We say that ω̃ ∈ A is interior to A, as there exists a neighborhood Ṽw (ω̃) contained in A.
If all points of A are interior, we say that A is weakly open.
Finally, we define the weak topology σ para RU , as the set of all subsets weakly open of RU .
Now consider U ∗ with the topology σ (U ∗ ,U ) and let φ : U ∗ → RU where
φ (u∗ ) = ∏ hu, u∗ iU .
u∈U
We shall show that φ is continuous. Suppose, to obtain contradiction that φ is not continuous. Thus, there
exists u∗ ∈ U ∗ such that φ is not continuous at u∗ .
Hence there exist a net {u∗α }α∈I such that
u∗α → u∗ in σ (U ∗ ,U),
but we do not have

φ (u∗α ) → φ (u∗ ) in σ .
Therefore, there exists a weak neighborhood Ṽ (φ (u∗ )) such that for each β ∈ I there exists αβ ∈ I such
that αβ β and
φ (u∗αβ ) 6∈ Ṽ (φ (u∗ )),
with with no loss of generality, we may assume
Ṽ (φ (u∗ )) = {ω ∈ RU : |Pui (w) − Pui (φ (u∗ ))| < εi , ∀i ∈ {1, · · · , m}},

From this, we get j ∈ {1, · · · , m} and a sub-net {u∗αβ } also denoted by {u∗αβ } such that
|Pu j (φ (u∗αβ )) − Pu j (φ (u∗ ))| ≥ ε j , ∀αβ ∈ I.

Thus,
|Pu j (φ (u∗αβ )) − Pu j (φ (u∗ ))| = |hu j , u∗αβ − u∗ iU |
≥ ε j , ∀αβ ∈ I. (4.49)
Therefore, we do not have,
hu j , u∗αβ iU → hu j , u∗ iU ,
that is, we do not have,
u∗α → u∗ , em σ (U ∗ ,U),
a contradiction.
Hence, φ is continuous with RU with the topology σ above specified.
We shall prove now that
φ −1 : φ (U ∗ ) → U ∗
is also continuous.
This follows from a little adaptation of the last proposition, considering that
fu (ω) = hu, φ −1 (w)iU = ωu = Pu (ω),

on φ (U ∗ ) so that fu , is continuous on φ (U ∗ ), for all u ∈ U.
Thus, from the last proposition, φ −1 is continuous.
On the other hand, observe that
φ (BU ∗ ) = K
where
K = {ω ∈ RU : |ωu | ≤ kukU , ωu+v = ωu + ωv ,
ωλ u = λ ωu , ∀u, v ∈ U, λ ∈ R}. (4.50)
To finish this proof, it suffices, from the continuity of φ −1 , to show that K ⊂ RU is compact with RU with
the topology σ .
Observe that K = K1 ∩ K2 where
K1 = {ω ∈ RU : |ωu | ≤ kukU , ∀u ∈ U}, (4.51)
and
K2 = {ω ∈ RU : ωu+v = ωu + ωv , ωλ u = λ ωu , ∀u, v ∈ U, λ ∈ R}. (4.52)
The set K3 = ∏u∈U [−kukU , kukU ] is compact as a Cartesian product of compact real intervals.
Since K1 ⊂ K3 and K1 is closed, we have that K1 is compact concerning the topology in question.
On the other hand, K2 is closed, since defining the closed sets Au,v e Bλ ,u (these sets are closed from the
continuity of projections Pu com RU for the topology σ , as inverse images of closed sets in R) by
Au,v = {ω ∈ RU : ωu+v − ωu − ωv = 0}, (4.53)
and
Bλ ,u = {ω ∈ RU : ωλ u − λ ωu = 0} (4.54)
we have
K2 = (∩u,v∈U Au,v ) ∩ (∩(λ ,u)∈R×U Bλ ,u ). (4.55)
Recall that K2 is closed as an intersection of closed sets.
Finally, we have that K1 ∩ K2 ⊂ K1 is compact.
Theorem 4.5.2 (Kakutani) Let U be a Banach space. Then U is reflexive if and only if
BU = {u ∈ U | kukU ≤ 1} (4.56)
is compact for the weak topology σ (U,U ∗ ).
Proof 4.16 Suppose U is reflexive, then J(BU ) = BU ∗∗ . From the last theorem BU ∗∗ is compact for the
topology σ (U ∗∗ ,U ∗ ). Therefore it suffices to verify that J −1 : U ∗∗ → U is continuous from U ∗∗ with the
topology σ (U ∗∗ ,U ∗ ) to U, with the topology σ (U,U ∗ ).
From Theorem 4.3.6 it is sufficient to show that the function u 7→ hJ −1 u, f iU is continuous for the topol-
ogy σ (U ∗∗ ,U ∗ ), for each f ∈ U ∗ . Since
hJ −1 u, f iU = h f , uiU ∗ ,
we have completed the first part of the proof.

For the second we need two lemmas.
Lemma 4.5.3 (Helly) Let U be a Banach space, f1 , ..., fn ∈ U ∗ and α1 , ..., αn ∈ R, then 1 and 2 are equiva-
lent, where:
1.
Given ε > 0, there exists uε ∈ U such that kuε kU ≤ 1 and
|huε , fi iU − αi | < ε, ∀i ∈ {1, ..., n}.
2.
n n
∑ βi αi ≤ ∑ βi fi , ∀β1 , ..., βn ∈ R. (4.57)
i=1 i=1 U∗
n
Proof. 1 ⇒ 2: Fix β1 , ..., βn ∈ R, ε > 0 and define S = ∑i=1 |βi |. From 1, we have
n n
∑ βi huε , fi iU − ∑ βi αi < εS (4.58)
i=1 i=1
and therefore
n n
∑ βi αi − ∑ βi huε , fi iU < εS (4.59)
i=1 i=1
or
n n n
∑ βi αi < ∑ βi fi kuε kU + εS ≤ ∑ βi fi + εS (4.60)
i=1 i=1 U∗ i=1 U∗
so that
n n
∑ βi αi ≤ ∑ βi f i (4.61)
i=1 i=1 U∗
since ε is arbitrary.
Now let us show that 2 ⇒ 1. Define ~α = (α1 , ..., αn ) ∈ Rn and consider the function ϕ(u) =
(h f1 , uiU , ..., h fn , uiU ). Item 1 is equivalent to ~α belongs to the closure of ϕ(BU ). Let us suppose that ~α
does not belong to the closure of ϕ(BU ) and obtain a contradiction. Thus we can separate ~α and the closure
of ϕ(BU ) strictly, that is there exists ~β = (β1 , ..., βn ) ∈ Rn and γ ∈ R such that
ϕ(u) · ~β < γ < ~α · ~β , ∀u ∈ BU (4.62)
Taking the supremum in u we contradict 2.
Also we need the lemma.

Lemma 4.5.4 Let U be a Banach space. Then J(BU ) is dense in BU ∗∗ for the topology σ (U ∗∗ ,U ∗ ).
Proof 4.17 Let u∗∗ ∈ BU ∗∗ and consider Vu∗∗ a neighborhood of u∗∗ for the topology σ (U ∗∗ ,U ∗ ). It suffices
to show that J(BU ) ∩ Vu∗∗ = / As Vu∗∗ is a weak neighborhood, there exists f1 , ..., fn ∈ U ∗ and ε > 0 such
6 0.
that
Vu∗∗ = {η ∈ U ∗∗ | h fi , η − u∗∗ iU ∗ | < ε, ∀i ∈ {1, ..., n}}. (4.63)
Define αi = h fi , u∗∗ iU ∗ and thus for any given β1 , ..., βn ∈ R we have
n n n
∑ βi αi = h ∑ βi fi , u∗∗ iU ∗ ≤ ∑ βi fi , (4.64)
i=1 i=1 i=1 U∗
so that from Helly lemma, there exists uε ∈ U such that kuε kU ≤ 1 and
|huε , fi iU − αi | < ε, ∀i ∈ {1, ..., n} (4.65)
or,
|h fi , J(uε ) − u∗∗ iU ∗ | < ε, ∀i ∈ {1, ..., n} (4.66)
and hence
J(uε ) ∈ Vu∗∗ . (4.67)
Now we will complete the proof of Kakutani theorem. Suppose BU is weakly compact (that is, compact
for the topology σ (U,U ∗ )). Observe that J : U → U ∗∗ is weakly continuous, that is, it is continuous with U
endowed with the topology σ (U,U ∗ ) and U ∗∗ endowed with the topology σ (U ∗∗ ,U ∗ ). Thus, as BU is weakly
compact, we have that J(BU ) is compact for the topology σ (U ∗∗ ,U ∗ ). From the last lemma, J(BU ) is dense
BU ∗∗ for the topology σ (U ∗∗ ,U ∗ ). Hence J(BU ) = BU ∗∗ , or J(U) = U ∗∗ , which completes the proof.
Proposition 4.5.5 Let U be a reflexive Banach space. Let K ⊂ U be a convex closed bounded set. Then K is
weakly compact.
Proof 4.18 From Theorem 4.3.5, K is weakly closed (closed for the topology σ (U,U ∗ )). Since K is
bounded, there exists α ∈ R+ such that K ⊂ αBU . Since K is weakly closed and K = K ∩ αBU , we have that
it is weakly compact.
Proposition 4.5.6 Let U be a reflexive Banach space and M ⊂ U a closed subspace. Then M with the norm
induced by U is reflexive.
Proof 4.19 We can identify two weak topologies in M, namely:
σ (M, M ∗ ) and the trace of σ (U,U ∗ ). (4.68)
It can be easily verified that these two topologies coincide (through restrictions and extensions of linear
forms). From the Kakutani Theorem it suffices to show that BM is compact for the topology σ (M, M ∗ ). But
BU is compact for σ (U,U ∗ ) and M ⊂ U is closed (strongly) and convex so that it is weakly closed, thus from
last proposition, BM is compact for the topology σ (U,U ∗ ), and therefore it is compact for σ (M, M ∗ ).
4.6 Separable sets

Definition 4.6.1 (Separable spaces) A metric space U is said to be separable if there exist a set K ⊂ U such
that K is countable and dense in U.
The next proposition is proved in [24].

Proposition 4.6.2 Let U be a separable metric space. If V ⊂ U then V is separable.
Theorem 4.6.3 Let U be a Banach space such that U ∗ is separable. Then U is separable.
Proof 4.20 Consider {u∗n } a countable dense set in U ∗ . Observe that
ku∗n kU ∗ = sup{|hu∗n , uiU | | u ∈ U and kukU = 1} (4.69)
so that for each n ∈ N, there exists un ∈ U such that kun kU = 1 and hu∗n , un iU ≥ 12 ku∗n kU ∗ .
Define U0 as the vector space on Q spanned by {un }, and U1 as the vector space on R spanned by {un }.
It is clear that U0 is dense in U1 and we will show that U1 is dense in U, so that U0 is a dense set in U. For,
suppose u∗ is such that hu, u∗ iU = 0, ∀u ∈ U1 . Since {u∗n } is dense in U ∗ , given ε > 0, there exists n ∈ N such
that ku∗n − u∗ kU ∗ < ε, so that
1 ∗
ku kU ∗ ≤ hun , u∗n iU = hun , u∗n − u∗ iU + hun , u∗ iU
2 n
≤ ku∗n − u∗ kU ∗ kun kU + 0
< ε (4.70)
or
ku∗ kU ∗ ≤ ku∗n − u∗ kU ∗ + ku∗n kU ∗ < ε + 2ε = 3ε. (4.71)
Therefore, since ε is arbitrary, ku∗ kU ∗ = 0, that is u∗ = θ . By Corollary 4.2.14 this completes the proof.
Proposition 4.6.4 U is reflexive if and only if U ∗ is reflexive.
Proof 4.21 Suppose U is reflexive, as BU ∗ is compact for σ (U ∗ ,U) and σ (U ∗ ,U) = σ (U ∗ ,U ∗∗ ) we have
that BU ∗ is compact for σ (U ∗ , U ∗∗ ), which means that U ∗ is reflexive.
Suppose U ∗ is reflexive, from above U ∗∗ is reflexive. Since J(U) is a closed subspace of U ∗∗ , from
Proposition 4.5.6, J(U) is reflexive. From the Kakutani Theorem J(BU ) is weakly compact. At this point
we shall prove that J −1 : J(U) → V is continuous from J(U) with the topology σ (U ∗∗ ,U ∗ ) to U with the
topology σ (U,U ∗ ).
Let {u∗∗α }α∈I ⊂ J(BU ) be a net such that
u∗∗ ∗∗
α * u0
weakly in σ (U ∗∗ ,U ∗ ).
Let u∗ ∈ U ∗ . Thus,
hu∗ , u∗∗ ∗ ∗∗
α iU ∗ → hu , u0 iU ∗ .
From this
hu∗ , J(J −1 (u∗∗ ∗ −1 ∗∗
α ))iU ∗ → hu , J(J (u0 ))iU ∗ ,
so that
h(J −1 (u∗∗ ∗ −1 ∗∗ ∗
α )), u iU → h(J (u0 )), u iU .
Since the net is question and u∗ ∈ U ∗ have been arbitrary, we may infer that J −1 is weakly continuous for
the concerning topology.
Hence J −1 (J(BU )) is also weakly compact so that from this, from the fact that BU is weakly closed and
BU ⊂ J −1 J(BU ),
it follows that BU is compact for the topology σ (U,U ∗ ).

From such a result and from the Kakutani Theorem we may infer that U is reflexive.
Proposition 4.6.5 Let U be a Banach space. Then U is reflexive and separable if and only if U ∗ is reflexive
and separable.
Our final result in this section refers to the metrizability of BU ∗ .

Theorem 4.6.6 Let U be separable Banach space. Under such hypotheses BU ∗ is metrizable with respect to
the weak-star topology σ (U ∗ ,U). Conversely, if BU ∗ is mertizable in σ (U ∗ ,U) then U is separable.
Proof 4.22 Let {un } be a dense countable set in BU . For each u∗ ∈ U ∗ define
∞
1
ku∗ kw = ∑ 2n | hun , u∗ iU |.
n=1
It may be easily verified that k · kw is a norm in U ∗ and
ku∗ kw ≤ ku∗ kU .
So, we may define a metric in U ∗ by
d(u∗ , v∗ ) = ku∗ − v∗ kw .
Now we shall prove that the topology induced by d coincides with σ (U ∗ ,U) in U ∗ .
For, let u∗0 ∈ BU ∗ and let V be neighborhood of u∗0 in σ (U ∗ ,U).
We need to prove that there exists r > 0 such that
Vw = {u∗ ∈ BU ∗ | d(u∗0 , u∗ ) < r} ⊂ V.
Observe that for V we may assume the general format
V = {u∗ ∈ U ∗ | |hvi , u∗ − u∗0 iU | < ε,
for some ε > 0 and v1 , ..., vk ∈ U.

There is no loss in generality in assuming
kvi kU ≤ 1, ∀i ∈ {1, ..., k}.

Since {un } is dense in U, for each i ∈ {1, ..., k} there exists ni ∈ N such that
ε
kuni − vi kU < .
4
Choose r > 0 small enough such that
ε
2ni r < , ∀i ∈ {1, ..., k}.
2
We are going to show that Vw ⊂ V , where
Vw = {u∗ ∈ BU ∗ | d(u∗0 , u∗ ) < r} ⊂ V.
Observe that, if u∗ ∈ Vw then

d(u∗0 , u∗ ) < r,
so that
1
|huni , u∗ − u∗0 iU | < r, ∀i ∈ {1, ..., k},
2ni
so that
|hvi , u∗ − u∗0 iU | ≤ |hvi − uni , u∗ − u∗0 iU | + |huni , u∗ − u∗0 iU |

≤ (ku∗ kU ∗ + ku∗0 kU ∗ )kvi − uni kU + |huni , u∗ − u∗0 iU |
ε ε
< 2 + = ε. (4.72)
4 2
Therefore u∗ ∈ V , so that Vw ⊂ V .
Now let u0 ∈ BU ∗ and fix r > 0. We have to obtain a neighborhood V ∈ σ (U ∗U) such that
V ⊂ Vw = {u∗ ∈ BU ∗ | d(u∗0 , u∗ ) < r}.
We shall define k ∈ N and ε > 0 in the next lines so that V ⊂ Vw , where
V = {u∗ ∈ BU ∗ | |hui , u∗ − u∗0 iU | < ε, ∀i ∈ {1, ..., k}}.
For u∗ ∈ Vw we have
k
1
d(u∗ , u∗0 ) = ∑ 2n |hun , u∗ − u∗0 iU |
n=1
∞
1
+ ∑ n
|hun , u∗ − u∗0 iU |
n=k+1 2
∞
1
< ε +2 ∑ n
n=k+1 2
1
= ε+ . (4.73)
2k−1
Hence, it suffices to take ε = r/2, and k sufficiently big such that
1
< r/2.
2k−1
The first part of the proof is finished.
Conversely, assume BU is metrizable in σ (U ∗ ,U). We are going to show that U is separable.
Define,
∗ ∗ 1
Ṽn = u ∈ BU ∗ | d(u , θ ) < .
n
From the first part, we may find Vn a neighborhood of zero in σ (U ∗ ,U) such that
Vn ⊂ Ṽn .
Moreover, we may assume that Vn has the form
Vn = {u∗ ∈ BU ∗ | |hu, u∗ − θ iU | < εn , ∀u ∈ Cn },
where Cn is a finite set.

Define
D = ∪∞
i=1Cn .
Thus D is countable and we are going to prove that such a set is dense in U.
For, suppose u∗ ∈ U ∗ is such that
hu, u∗ iU = 0, ∀u ∈ D.
Hence,
u∗ ∈ Vn ⊂ Ṽn , ∀n ∈ N,
so that u∗ = θ .
4.7 Uniformly convex spaces

Definition 4.7.1 (Uniformly convex spaces) A Banach space U is said to be uniformly convex if for each
ε > 0, there exists δ > 0 such that:
If u, v ∈ U, kukU ≤ 1, kvkU ≤ 1, and ku − vkU > ε then ku+vk
2
U
< 1−δ.
Theorem 4.7.2 (Milman Pettis) Every uniformly convex Banach space is reflexive.
Proof 4.23 Let η ∈ U ∗∗ be such that kηkU ∗∗ = 1. It suffices to show that η ∈ J(BU ). Since J(BU ) is closed
in U ∗∗ , we have only to show that for each ε > 0 there exists u ∈ U such that kη − J(u)kU ∗∗ < ε.
Thus, suppose given ε > 0. Let δ > 0 be the corresponding constant relating the uniformly convex prop-
erty.
Choose f ∈ U ∗ such that k f kU ∗ = 1 and
δ
h f , ηiU ∗ > 1 − . (4.74)
2
Define
δ
V = ζ ∈ U ∗∗ | |h f , ζ − ηiU ∗ | < .
2
Observe that V is a neighborhood of η in σ (U ∗∗ ,U ∗ ). Since J(BU ) is dense in BU ∗∗ concerning the topology
σ (U ∗∗ ,U ∗ ), we have that V ∩ J(BU ) 6= 0/ and thus there exists u ∈ BU such that J(u) ∈ V. Suppose, to obtain
contradiction, that
kη − J(u)kU ∗∗ > ε.
Therefore, defining
W = (J(u) + εBU ∗∗ )c ,
we have that η ∈ W , where W is also a weak neighborhood of η in σ (U ∗∗ ,U ∗ ), since BU ∗∗ is closed in

σ (U ∗∗ ,U ∗ ).
6 0/, so that there exists some v ∈ BU such that J(v) ∈ V ∩W. Thus, J(u) ∈ V and
Hence V ∩W ∩ J(BU ) =
J(v) ∈ V , so that
δ
|hu, f iU − h f , ηiU ∗ | < ,
2
and
δ
|hv, f iU − h f , ηiU ∗ | < .
2
Hence,
2h f , ηiU ∗ < hu + v, f iU + δ
≤ ku + vkU + δ . (4.75)
From this and (4.74) we obtain

ku + vkU
> 1−δ,
2
and thus from the definition of uniform convexity, we obtain
ku − vkU ≤ ε. (4.76)
On the other hand, since J(v) ∈ W , we have
kJ(u) − J(v)kU ∗∗ = ku − vkU > ε,
which contradicts (4.76). The proof is complete.

Chapter 5
Topics on Linear Operators
The main references for this chapter are Reed and Simon [62] and Bachman and Narici [6].
5.1 Topologies for bounded operators

Let U,Y be Banach spaces. First we recall that the set of all bounded linear operators A : U → Y , denoted by
L (U,Y ), is a Banach space with the norm
kAk = sup{kAukY | kukU ≤ 1}.
The topology related to the metric induced by this norm is called the uniform operator topology.
Let us introduce now the strong operator topology, which is defined as the weakest topology for which
the functions
Eu : L (U,Y ) → Y
are continuous where
Eu (A) = Au, ∀A ∈ L (U,Y ).
For such a topology a base at origin is given by sets of the form
{A | A ∈ L (U,Y ), kAui kY < ε, ∀i ∈ {1, , ..., n}},
where u1 , ..., un ∈ U and ε > 0.

Observe that a sequence {An } ⊂ L (U,Y ) converges to A concerning this last topology if
kAn u − AukY → 0, as n → ∞, ∀u ∈ U.
In the next lines we describe the weak operator topology in L (U,Y ). Such a topology is weakest one
such that the functions
Eu,v : L (U,Y ) → C
are continuous, where
Eu,v (A) = hAu, viY , ∀A ∈ L (U,Y ), u ∈ U, v ∈ Y ∗ .
For such a topology, a base at origin is given by sets of the form
{A ∈ L (U,Y ) | |hAui , v j iY | < ε, ∀i ∈ {1, ..., n}, j ∈ {1, ..., m}}.

Topics on Linear Operators 87
where ε > 0, u1 , ..., un ∈ U, v1 , ..., vm ∈ Y ∗ .

A sequence {An } ⊂ L (U,Y ) converges to A ∈ L (U,Y ) if
|hAn u, viY − hAu, viY | → 0,
as n → ∞, ∀u ∈ U, v ∈ Y ∗ .
5.2 Adjoint operators

We start this section recalling the definition of adjoint operator.
Definition 5.2.1 Let U,Y be Banach spaces. Given a bounded linear operator A : U → Y and v∗ ∈ Y ∗ , we
have that T (u) = hAu, v∗ iY is such that
|T (u)| ≤ kAukY · kv∗ k ≤ kAkkv∗ kY ∗ kukU .
Hence, T (u) is a continuous linear functional on U and considering our fundamental representation hypoth-
esis, there exists u∗ ∈ U ∗ such that
T (u) = hu, u∗ iU , ∀u ∈ U.
We define A∗ by setting u∗ = A∗ v∗ , so that
T (u) = hu, u∗ iU = hu, A∗ v∗ iU
that is,
hu, A∗ v∗ iU = hAu, v∗ iY , ∀u ∈ U, v∗ ∈ Y ∗ .
We call A∗ : Y ∗ → U ∗ the adjoint operator relating A : U → Y.
Theorem 5.2.2 Let U,Y be Banach spaces and let A : U → Y be a bounded linear operator. Then
kAk = kA∗ k.
kAk = sup{kAuk | kukU = 1}

u∈U
= sup{ sup {hAu, v∗ iY | kv∗ kY ∗ = 1}, kukU = 1}
u∈U v∗ ∈Y ∗
= sup {hAu, v∗ iY | kv∗ kY ∗ = 1, kukU = 1}
(u,v∗ )∈U×Y ∗
= sup {hu, A∗ v∗ iU | kv∗ kY ∗ = 1, kukU = 1}
(u,v∗ )∈U ×Y ∗
= sup {sup{hu, A∗ v∗ iU | kukU = 1}, kv∗ kY ∗ = 1}
v∗ ∈Y ∗ u∈U
= sup {kA v k, kv∗ kY ∗ = 1}
∗ ∗
v∗ ∈Y ∗
∗
= kA k. (5.1)
In particular, if U = Y = H where H is Hilbert space, we have

Theorem 5.2.3 Given the bounded linear operators A, B : H → H we have
1. (AB)∗ = B∗ A∗ ,
2. (A∗ )∗ = A,
3. If A has a bounded inverse A−1 then A∗ has a bounded inverse and
(A∗ )−1 = (A−1 )∗ .
4. kAA∗ k = kAk2 .
Proof 5.2
1. Observe that
(ABu, v)H = (Bu, A∗ v)H = (u, B∗ A∗ v)H , ∀u, v ∈ H.
2. Observe that
(Au, v)H = (u, A∗ v)H = (A∗ v, u)H = (v, A∗∗ u)H = (A∗∗ u, v)H , ∀u, v ∈ H.
3. We have that
I = AA−1 = A−1 A,
so that
I = I ∗ = (AA−1 )∗ = (A−1 )∗ A∗ = (A−1 A)∗ = A∗ (A−1 )∗ .
4. Observe that
kA∗ Ak ≤ kAkkA∗ k = kAk2 ,
and
kA∗ Ak ≥ sup{(u, A∗ Au)H | kukU = 1}

u∈U
= sup{(Au, Au)H | kukU = 1}
u∈U
= sup{kAuk2H | kukU = 1} = kAk2 , (5.2)
u∈U
and hence
kA∗ Ak = kAk2 .
Definition 5.2.4 Given A ∈ L (H) we say that A is self-adjoint if
A = A∗ .
Theorem 5.2.5 Let U and Y be Banach spaces and let A : U → Y be a bounded linear operator. Then
[R(A)]⊥ = N(A∗ ),
where
[R(A)]⊥ = {v∗ ∈ Y ∗ | hAu, v∗ iY = 0, ∀u ∈ U}.
Proof 5.3 Let v∗ ∈ N(A∗ ). Choose v ∈ R(A). Thus there exists u in U such that Au = v so that
hv, v∗ iY = hAu, v∗ iY = hu, A∗ v∗ iU = 0.
Since v ∈ R(A) is arbitrary, we have obtained
N(A∗ ) ⊂ [R(A)]⊥ .
Suppose v∗ ∈ [R(A)]⊥ . Choose u ∈ U. Thus,
hAu, v∗ iY = 0,
so that
hu, A∗ v∗ iU , ∀u ∈ U.
Therefore, A∗ v∗ = θ , that is, v∗ ∈ N(A∗ ). Since v∗ ∈ [R(A)]⊥ is arbitrary, we get
[R(A)]⊥ ⊂ N(A∗ ).
The next result is relevant for subsequent developments.
Lemma 5.1
Let U,Y be Banach spaces and let A : U → Y be a bounded linear operator. Suppose also that R(A) =
{A(u) : u ∈ U} is closed. Under such hypotheses, there exists K > 0 such that for each v ∈ R(A) there exists
u0 ∈ U such that
A(u0 ) = v
and
ku0 kU ≤ KkvkY .
Proof 5.4 Define L = N(A) = {u ∈ U : A(u) = θ } (the null space of A). Consider the space U/L, where
U/L = {u : u ∈ U},
where
u = {u + w : w ∈ L}.
Define A : U/L → R(A), by
A(u) = A(u).
Observe that A is one-to-one, linear, onto and bounded. Moreover R(A) is closed so that it is a Banach space.
Hence by the inverse mapping theorem we have that A has a continuous inverse. Thus, for any v ∈ R(A) there
exists u ∈ U/L such that
A(u) = v
so that
−1
u = A (v),
and therefore
−1
kuk ≤ kA kkvkY .
Recalling that
kuk = inf {ku + wkU },
w∈L
we may find u0 ∈ u such that

−1
ku0 kU ≤ 2kuk ≤ 2kA kkvkY ,
and so that
A(u0 ) = A(u0 ) = A(u) = v.
−1
Taking K = 2kA k, we have completed the proof.
Theorem 5.1
Let U,Y be Banach spaces and let A : U → Y be a bound linear operator. Assume R(A) is closed. Under such
hypotheses
R(A∗ ) = [N(A)]⊥ .
Proof 5.5 Let u∗ ∈ R(A∗ ). Thus there exists v∗ ∈ Y ∗ such that
u∗ = A∗ (v∗ ).
Let u ∈ N(A). Hence,

hu, u∗ iU = hu, A∗ (v∗ )iU = hA(u), v∗ iY = 0.
Since u ∈ N(A) is arbitrary, we get u∗ ∈ [N(A)]⊥ , so that
R(A∗ ) ⊂ [N(A)]⊥ .
Now suppose u∗ ∈ [N(A)]⊥ . Thus,
hu, u∗ iU = 0, ∀u ∈ N(A).
Fix v ∈ R(A). From the Lemma 5.1, there exists K > 0 (which does not depend on v) and uv ∈ U such
that
A(uv ) = v
and
kuv kU ≤ KkvkY .
Define f : R(A) → R by
f (v) = huv , u∗ iU .
Observe that
| f (v)| ≤ kuv kU ku∗ kU ∗ ≤ KkvkY ku∗ kU ∗ ,
so that f is a bounded linear functional. Hence by a Hahn-Banach theorem corollary, there exists v∗ ∈ Y ∗
such that
f (v) = hv, v∗ iY ≡ F(v), ∀v ∈ R(A),
that is, F is an extension of f from R(A) to Y .
In particular
f (v) = huv , u∗ iU = hv, v∗ iY = hA(uv ), v∗ iY ∀v ∈ R(A),
where A(uv ) = v, so that
huv , u∗ iU = hA(uv ), v∗ iY ∀v ∈ R(A).
Now let u ∈ U and define A(u) = v0 . Observe that
u = (u − uv0 ) + uv0 ,
and
A(u − uv0 ) = A(u) − A(uv0 ) = v0 − v0 = θ .
Since u∗ ∈ [N(A)]⊥ , we get
hu − uv0 , u∗ iU = 0
so that
hu, u∗ iU = h(u − uv0 ) + uv0 , u∗ iU
= huv0 , u∗ iU
= hA(uv0 ), v∗ iY
= hA(u − uv0 ) + A(uv0 ), v∗ iY
= hA(u), v∗ iY . (5.3)
Hence,
hu, u∗ iU = hA(u), v∗ iY , ∀u ∈ U.
We may conclude that u∗ = A∗ (v∗ ) ∈ R(A∗ ). Since u∗ ∈ [N(A)]⊥ is arbitrary we obtain
[N(A)]⊥ ⊂ R(A∗ ).
We finish this section with the following result.

Definition 5.2.6 Let U be a Banach space and S ⊂ U. We define the positive conjugate cone of S, denoted
by S⊕ by
S⊕ = {u∗ ∈ U ∗ : hu, u∗ iU ≥ 0, ∀u ∈ S}.
Similarly, we define the the negative cone of S, denoted by denoted by S by
S = {u∗ ∈ U ∗ : hu, u∗ iU ≤ 0, ∀u ∈ S}.
Theorem 5.2.7 Let U,Y be Banach spaces and A : U → Y be a bounded linear operator. Let S ⊂ U. Then
[A(S)]⊕ = (A∗ )−1 (S⊕ ),
where
(A∗ )−1 (S⊕ ) = {v∗ ∈ Y ∗ : A∗ v∗ ∈ S⊕ }.
Proof 5.6 Let v∗ ∈ [A(S)]⊕ and u ∈ S. Thus,

hA(u), v∗ iY ≥ 0,
so that
hu, A∗ (v∗ )iU ≥ 0.
Since u ∈ S is arbitrary, we get
v∗ ∈ (A∗ )−1 (S⊕ ).
From this,
[A(S)]⊕ = (A∗ )−1 (S⊕ ).
Reciprocally, let v∗ ∈ (A∗ )−1 (S⊕ ). Hence A∗ (v∗ ) ∈ S⊕ so that, for u ∈ S we obtain
hu, A∗ (v∗ )iU ≥ 0,
and therefore
hA(u), v∗ iY ≥ 0.
Since u ∈ S is arbitrary, we get v∗ ∈ [A(S)]⊕ , that is,
(A∗ )−1 (S⊕ ) ⊂ [A(S)]⊕ .
5.3 Compact operators

We start this section defining compact operators.
Definition 5.3.1 Let U and Y be Banach spaces. An operator A ∈ L (U,Y ) (linear and bounded) is said
to compact if A takes bounded sets into pre-compact sets. Summarizing, A is compact if for each bounded
sequence {un } ⊂ U, {Aun } has a convergent subsequence in Y .
Theorem 5.3.2 A compact operator maps weakly convergent sequences into norm convergent sequences.
Proof 5.7 Let A : U → Y be a compact operator. Suppose
un * u weakly in U.
By the uniform boundedness theorem, {kun k} is bounded. Thus, given v∗ ∈ Y ∗ we have
hAun , v∗ iY = hun , A∗ v∗ iU
→ hu, A∗ v∗ iU
= hAu, v∗ iY . (5.4)
Being v∗ ∈ Y ∗ arbitrary, we get that
Aun * Au weakly in Y. (5.5)
Suppose Aun does not converge in norm to Au. Thus there exists ε > 0 and a subsequence {Aunk } such that
kAunk − AukY ≥ ε, ∀k ∈ N.
As {unk } is bounded and A is compact, {Aunk } has a subsequence converging para ṽ = 6 Au. But then such a
sequence converges weakly to ṽ 6= Au, which contradicts (5.5). The proof is complete.
Theorem 5.3.3 Let H be a separable Hilbert space. Thus each compact operator in L (H) is the limit in
norm of a sequence of finite rank operators.
Proof 5.8 Let A be a compact operator in H. Let {φ j } be an orthonormal basis in H. For each n ∈ N define
λn = sup{kAψkH | ψ ∈ [φ1 , ..., φn ]⊥ and kψkH = 1}.
It is clear that {λn } is a non-increasing sequence that converges to a limit λ ≥ 0. We will show that λ = 0.
Choose a sequence {ψn } such that
ψn ∈ [φ1 , ..., φn ]⊥ ,
kψn kH = 1 and kAψn kH ≥ λ /2. Now we will show that
ψn * θ , weakly in H.
Let ψ ∗ ∈ H ∗ = H, thus there exists a sequence {a j } ⊂ C such that

∞
ψ∗ = ∑ a jφ j.
j=1
Suppose given ε > 0. We may find n0 ∈ N such that

∞
∑ |a j |2 < ε .
j=n0
Choose n > n0 . Hence there exists {b j } j>n such that

∞
ψn = ∑ b jφ j,
j=n+1
and
∞
∑ |b j |2 = 1.
j=n+1
Therefore,
∞
|(ψn , ψ ∗ )H | = ∑ (φ j , φ j )H a j · b j
j=n+1
∞
= ∑ aj ·bj
j=n+1
s s
∞ ∞
≤ ∑ |a j |2 ∑ |b j |2
j=n+1 j=n+1
√
≤ ε, (5.6)
if n > n0 . Since ε > 0 is arbitrary,

(ψn , ψ ∗ )H → 0, as n → ∞.
Since ψ ∗ ∈ H is arbitrary, we get
ψn * θ , weakly in H.
Hence, as A is compact, we have
Aψn → θ in norm ,
so that λ = 0. Finally, we may define {An } by
!
n n
An (u) = A ∑ (u, φ j )H φ j = ∑ (u, φ j )H Aφ j ,
j=1 j=1
for each u ∈ H. Thus

kA − An k = λn → 0, as n → ∞.
5.4 The square root of a positive operator

Definition 5.4.1 Let H be a Hilbert space. A mapping E : H → H is said to be a projection on M ⊂ H if for
each z ∈ H we have
Ez = x
where z = x + y, x ∈ M and y ∈ M ⊥ .
Observe that
1. E is linear,
2. E is idempotent, that is E 2 = E,
3. R(E) = M,
4. N(E) = M ⊥ .
Also observe that from
Ez = x
we have
kEzk2H = kxk2H ≤ kxk2H + kyk2H = kzk2H ,
so that
kEk ≤ 1.
Definition 5.4.2 Let A, B ∈ L (H). We write
A≥θ
if
(Au, u)H ≥ 0, ∀u ∈ H,
and in this case we say that A is positive. Finally, we denote
A≥B
if
A−B ≥ θ.
Theorem 5.4.3 Let A and B be bounded self-adjoint operators such that A ≥ θ and B ≥ θ . If AB = BA then
AB ≥ θ .
Proof 5.9 If A = θ , the result is obvious. Assume A 6= θ and define the sequence
A
A1 = , An+1 = An − A2n , ∀n ∈ N.
kAk
We claim that
θ ≤ An ≤ I, ∀n ∈ N.
We prove the claim by induction.
For n = 1, it is clear that A1 ≥ θ . And since kA1 k = 1, we get
(A1 u, u)H ≤ kA1 kkukH kukH = (Iu, u)H , ∀u ∈ H,
so that
A1 ≤ I.
Thus,
θ ≤ A1 ≤ I.
Now suppose θ ≤ An ≤ I. Since An is self adjoint we have,
(A2n (I − An )u, u)H = ((I − An )An u, An u)H

= ((I − An )v, v)H ≥ 0, ∀u ∈ H (5.7)
where v = An u. Therefore
A2n (I − An ) ≥ θ .
Similarly, we may obtain

An (I − An )2 ≥ θ ,
so that
θ ≤ An2 (I − An ) + An (I − An )2 = An − A2n = An+1 .
So we also have,
θ ≤ I − An + A2n = I − An+1 ,
that is,
θ ≤ An+1 ≤ I,
so that
θ ≤ An ≤ I, ∀n ∈ N.
Observe that,
A1 = A21 + A2
= A21 + A22 + A3
... ...................
= A21 + ... + A2n + An+1 . (5.8)
Since An+1 ≥ θ , we obtain

A21 + A22 + ... + A2n = A1 − An+1 ≤ A1 . (5.9)
From this, for a fixed u ∈ H, we have
n n
∑ kA j uk2 = ∑ (A j u, A j u)H
j=1 j=1
n
= ∑ (A2j u, u)H
j=1
≤ (A1 u, u)H . (5.10)
Since n ∈ N is arbitrary, we get

∞
∑ kA j uk2
j=1
is a converging series, so that

kAn uk → 0,
that is,
An u → θ , as n → ∞.
From this and (5.9), we get
n
∑ A2j u = (A1 − An+1 )u → A1 u, as n → ∞.
j=1
Finally, we may write,
(ABu, u)H = kAk(A1 Bu, u)H

= kAk(BA1 u, u)H
n
= kAk(B lim
n→∞
∑ A2j u, u)H
j=1
n
= kAk lim
n→∞
∑ (BA2j u, u)H
j=1
n
= kAk lim ∑ (BA j u, A j u)H
n→∞
j=1
≥ 0. (5.11)
Hence,
(ABu, u)H ≥ 0, ∀u ∈ H.
Theorem 5.4.4 Let {An } be a sequence of self-adjoint commuting operators in L (H). Let B ∈ L (H) be a
self adjoint operator such that
Ai B = BAi , ∀i ∈ N.
Suppose also that
A1 ≤ A2 ≤ A3 ≤ ... ≤ An ≤ ... ≤ B.
Under such hypotheses there exists a self-adjoint, bounded, linear operator A such that
An → A in norm ,
and
A ≤ B.
Proof 5.10 Consider the sequence {Cn } where
Cn = B − An ≥ 0, ∀n ∈ N.
Fix u ∈ H. First, we show that {Cn u} converges. Observe that
CiC j = C jCi , ∀i, j ∈ N.
Also, if n > m then

An − Am ≥ θ
so that
Cm = B − Am ≥ B − An = Cn .
Therefore, from Cm ≥ θ and Cm −Cn ≥ θ we obtain
(Cm −Cn )Cm ≥ θ , if n > m
and also
Cn (Cm −Cn ) ≥ θ .
Thus,
(Cm2 u, u)H ≥ (CnCm u, u)H ≥ (Cn2 u, u)H ,
and we may conclude that

(Cn2 u, u)H
is a monotone non-increasing sequence of real numbers, bounded below by 0, so that there exists α ∈ R such
that
lim (Cn2 u, u)H = α.
n→∞
Since each Cn is self-adjoint we obtain
2
k(Cn −Cm )ukH = ((Cn −Cm )u, (Cn −Cm )u)H
= ((Cn −Cm )(Cn −Cm )u, u)H
= (Cn2 u, u)H − 2(CnCm u, u) + (Cm2 u, u)H
→ α − 2α + α = 0, (5.12)
as
m, n → ∞.
Therefore, {Cn u} is a Cauchy sequence in norm, so that there exists the limit
lim Cn u = lim (B − An )u,
n→∞ n→∞
and hence there exists

lim An u, ∀u ∈ H.
n→∞
Now define A by
Au = lim An u.
n→∞
Since the limit
lim An u, ∀u ∈ H
n→∞
exists, we have that
sup{kAn ukH }
n∈N
is finite for all u ∈ H. By the principle of uniform boundedness
sup{kAn k} < ∞
n∈N
so that there exists K > 0 such that

kAn k ≤ K, ∀n ∈ N.
Therefore,
kAn ukH ≤ KkukH ,
so that
kAuk = lim {kAn ukH } ≤ KkukH , ∀u ∈ H
n→∞
which means that A is bounded. Fixing u, v ∈ H, we have
(Au, v)H = lim (An u, v)H = lim (u, An v)H = (u, Av)H ,
n→∞ n→∞
and thus A is self-adjoint. Finally

(An u, u)H ≤ (Bu, u)H , ∀n ∈ N,
so that
(Au, u) = lim (An u, u)H ≤ (Bu, u)H , ∀u ∈ H.
n→∞
Hence A ≤ B.
Definition 5.4.5 Let A ∈ L (A) be a positive operator. The self-adjoint operator B ∈ L (H) such that
B2 = A
is called the square root of A. If B ≥ θ we denote

√
B= A.
Theorem 5.4.6 Suppose A ∈ L (H) is positive. Then there exists B ≥ θ such that
B2 = A.
Furthermore B commutes with any C ∈ L (H) such that commutes with A.
Proof 5.11 There is no loss of generality in considering
kAk ≤ 1,
which means θ ≤ A ≤ I, because we may replace A by

A
kAk
so that if
A
C2 =
kAk
then
B = kAk1/2C.
Let
B0 = θ ,
and consider the sequence of operators given by
1
Bn+1 = Bn + (A − B2n ), ∀n ∈ N ∪ {0}.
2
Since each Bn is polynomial in A, we have that Bn is self-adjoint and commute with any operator with
commutes with A. In particular
Bi B j = B j Bi , ∀i, j ∈ N.
First we show that
Bn ≤ I, ∀n ∈ N ∪ {0}.
1
Since B0 = θ , and B1 = 2 A, the statement holds for n = 1. Suppose Bn ≤ I. Thus,
1 1
I − Bn+1 = I − Bn − A + B2n
2 2
1 2 1
= (I − Bn ) + (I − A) ≥ θ (5.13)
2 2
so that
Bn+1 ≤ I.
The induction is complete, that is,
Bn ≤ I, ∀n ∈ N.
Now we prove the monotonicity also by induction. Observe that
B0 ≤ B1 ,
and supposing
Bn−1 ≤ Bn ,
we have
1 1
Bn+1 − Bn = Bn + (A − B2n ) − Bn−1 − (A − B2n−1 )
2 2
1 2 2
= Bn − Bn−1 − (Bn − Bn−1 )
2
1
= Bn − Bn−1 − (Bn + Bn−1 )(Bn − Bn−1 )
2
1
= (I − (Bn + Bn−1 ))(Bn − Bn−1 )
2
1
= ((I − Bn−1 ) + (I − Bn ))(Bn − Bn−1 ) ≥ θ .
2
The induction is complete, that is
θ = B0 ≤ B1 ≤ B2 ≤ ... ≤ Bn ≤ ... ≤ I.
By the last theorem there exists a self-adjoint operator B such that
Bn → B in norm.
Fixing u ∈ H we have
1
Bn+1 u = Bn u + (A − B2n )u,
2
so that taking the limit in norm as n → ∞, we get
θ = (A − B2 )u.
Being u ∈ H arbitrary we obtain

A = B2 .
It is also clear that
B≥θ
Chapter 6
Spectral Analysis, a General

Approach in Normed Spaces
6.1 Introduction
We start by presenting some results about the spectrum and resolvent sets for a bounded operator defined on
a normed space. The main reference for this chapter is Bachman and Narici [6].
Definition 6.1.1 Let V be a complex normed vector space and let A : D ⊂ V → V be a linear operator, where
D is dense on V . We say that A−1 : R(A) → D is the inverse operator related to A, as A is a bijection from D
to R(A) and
A−1 y = u if, and only if, Au = y, ∀u ∈ D, y ∈ R(A),
where R(A) = {Au : u ∈ D}, is the range of A.
In such case we have,
A−1 Au = u, ∀u ∈ D
and
AA−1 y = y, ∀y ∈ R(A).
Let λ ∈ C.
1. If R(λ I − A) is dense in V and λ I − A has a bounded inverse, we write λ ∈ ρ(A), where ρ(A) denotes
the resolvent set of A.
2. If R(λ I − A) is dense in V and (λ I − A)−1 exists but it is not bounded, we write λ ∈ Cσ (A), where
Cσ (A) denotes the continuous spectrum of A.
3. If R(λ I − A) is not dense in V and λ I − A has an inverse either bounded or unbounded, we write
λ ∈ Rσ (A), where Rσ (A) denotes the residual spectrum of A.
4. If (λ I − A)−1 does not exist, we write λ ∈ Pσ (A) where Pσ (A) denotes the point spectrum of A.
In such a case there exists u ∈ V such that
Au − λ u = 0
where u 6= 0, so that u is said to be an eigenvector of A and λ the corresponding eigenvalue.

Spectral Analysis, a General Approach in Normed Spaces 101
Table 6.1 summarizes such results.
Table 6.1: About the spectrum and resolvent sets of A.
(λ I − A)−1 Boundedness of (λ I − A)−1 R(λ I − A) Set

exists bounded dense in V ρ(A)
exists unbounded dense in V Cσ (A)
exists bounded or not not dense in V Rσ (A)
not exists — dense or not in V Pσ (A)
Remark 6.1.2 Observe that

C = ρ(A) ∪Cσ (A) ∪ Rσ (A) ∪ Pσ (A)
and such union is disjoint.
The spectrum of A, denoted by σ (A), is defined by
σ (A) = Cσ (A) ∪ Rσ (A) ∪ Pσ (A).
Theorem 6.1.3 (Riesz) Let V be a normed vector space and let 0 < α < 1. Let M be a proper closed vector
subspace of V .
Under such hypotheses, there exists uα ∈ V such that
kuα kV = 1
and
ku − uα kV ≥ α, ∀u ∈ M.
Proof 6.1 Since M ⊂ V is a proper closed subspace of V , we may select a v ∈ V \ M.

Define
d = inf ku − vkV .
u∈M
Observe that, since M is closed, we have d > 0, otherwise if we had d = 0 we had v ∈ M = M, which
contradicts v 6∈ M.
Also,
d/α > d.
Hence, there exists u0 ∈ M such that
0 < d ≤ ku0 − vkV < d/α.
Define
v − u0
uα = .
kv − u0 kV
Thus, kuα kV = 1 and also for u ∈ M we have
v u0
ku − uα kV = u− +
kv − u0 kV kv − u0 kV V
1
= k u(kv − u0 kV ) + u0 − vkV . (6.1)
kv − u0 kV
From this, since ukv − u0 kV + u0 ∈ M, we have
d
ku − uα kV ≥ > α, ∀u ∈ M.
kv − u0 kV
Theorem 6.1.4 Let V be a complex normed vector space and let A : D ⊂ V → V be a linear compact operator.
Under such hypotheses, Pσ (A) is countable and 0 is its unique possible limit point.
Proof 6.2 Let ε > 0. We shall prove that there exists at most a finite number of points in Pε where
Pε = {λ ∈ Pσ (A) : |λ | ≥ ε}.
Observe that in such a case
Pσ (A) \ {0} = ∪∞
n=1 P1/n
and such a set is countable and has 0 as the unique possible limit point.
Suppose, to obtain contradiction, there exists ε > 0 such that Pε has infinite points. Hence, there exists a
sequence {λn }n∈N ⊂ Pε and a sequence of linearly independent eigenvectors {un } such that
Aun = λn un , ∀n ∈ N.
Define
Mn = Span{u1 , . . . , un },
so that
Mn−1 ⊂ Mn
and Mn is finite dimensional, ∀n ∈ N. Observe that Mn−1 ⊂ Mn properly.
From the Riesz theorem, there exists yn ∈ Mn such that kyn kV = 1 and
kyn − uk ≥ 1/2, ∀u ∈ Mn−1 , ∀n > 1.
Let
n
u = ∑ αi ui ∈ Mn .
i=1
Thus,
n n
Au = ∑ αi Aui = ∑ αi λi ui .
i=1 i=1
Therefore,
(λn − A)u = λn u − Au
n−1
= ∑ αi (λn − λi )ui ∈ Mn−1 . (6.2)
i=1
Therefore,
(λn − A)(Mn ) ⊂ Mn−1
and from this
A(Mn ) ⊂ Mn , ∀n ∈ N.
Let 1 < m < n. Thus,
w = (λn − A)yn + Aym ∈ Mn−1 ,
so that
Ayn − Aym = λn yn − w = λn (yn − λn−1 w).
Since, λn−1 w ∈ Mn−1 we get
|λn | ε
||Ayn − Aym k = |λn |kyn − λn−1 wk ≥ ≥ ,
2 2
∀1 ≤ m < n ∈ N.
Therefore, {yn } is a bounded sequence and such that {Ayn } has no Cauchy subsequence, that is, {Ayn }
has no convergent subsequence, which contradicts A be compact.
Definition 6.1.5 Let V be a normed vector space and let A : D ⊂ V → V be a linear operator, where D is
dense in V.
We say that λ ∈ C is a proper approximate value of A if for each ε > 0 there exists u ∈ D such that
kukV = 1 and k(λ I − A)ukV < ε.
In such a case we denote λ ∈ π(A), where π(A) is the approximate spectrum of A.
Theorem 6.1.6 Considering the statements of the last definition, we have that λ ∈ π(A) if, and only if, λ I −A
has no a bounded inverse.
Proof 6.3 Suppose λ ∈ π(A). Thus, for each n ∈ N, there exists un ∈ D such that kun kV = 1 and
k(λ I − A)un kV < 1/n. (6.3)
Suppose, to obtain contradiction, there exists K > 0 such that
k(λ I − A)ukV ≥ KkukV , ∀u ∈ D.
In particular we have
k(λ I − A)un kV ≥ 1, ∀n ∈ N,
which contradicts (6.3).
Reciprocally, suppose λ I − A has no bounded inverse.
Thus, there is no K > 0 such that
k(λ I − A)ukV ≥ KkukV , ∀u ∈ D.
Hence, for each ε > 0 we may find u ∈ D such that
k(λ I − A)ukV < εkukV .
From this, for each ε > 0 we may find u ∈ D such that
kukV = 1 and k(λ I − A)ukV < ε.
Therefore λ ∈ π(A).
6.2 Sesquilinear functionals

Definition 6.2.1 Let V be a complex vector space. A functional f : V × V → C is said to be a sesquilinear
functional, as
1. f (u1 + u2 , v) = f (u1 , v) + f (u2 , v), ∀u1 , u2 , v ∈ V.
2. f (αu, v) = α f (u, v), ∀u, v ∈ V, ∀α ∈ C.
3. f (u, v1 + v2 ) = f (u, v1 ) + f (u, v2 ), ∀u, v1 , v2 ∈ V.
4. f (u, αv) = α f (u, v), ∀u, v ∈ V, α ∈ C.
Remark 6.2.2 Let H be a complex Hilbert space and let A : H → H be a linear operator. Hence f : H × H →
C defined by
f (u, v) = (Au, v)H , ∀u, v ∈ H
is a sequilinear functional.
Remark 6.2.3 Given a sesquilinear functional

f : H ×H → C
we shall define fˆ : H → C by
fˆ(u) = f (u, u), ∀u ∈ H.
Exercise 6.2.4 In the context of the last definitions, prove that

1 1
f (u, v) = fˆ (u + v) − fˆ (u − v)
2 2

1 1
+i fˆ (u + iv) − i fˆ (u − iv) , ∀u, v ∈ H. (6.4)
2 2
Conclude that if fˆ1 = fˆ2 , then f1 = f2 .

Theorem 6.2.5 Let V be a complex vector space and let f : V × V → C be a symmetric sesquilinear func-
tional, that is, assume f is such that f (u, v) = f (v, u), ∀u, v ∈ V, where f (v, u) denotes the complex conjugate
of f (v, u).
fˆ(u) ∈ R, ∀u ∈ V.
Moreover, the reciprocal also holds.
Proof 6.4 Suppose

f (u, v) = f (v, u), ∀u, v ∈ V.
Thus,
fˆ(u) = f (u, u) = f (u, u) = fˆ(u),
so that
fˆ(u) ∈ R, ∀u ∈ V.
Reciprocally, suppose fˆ is real.
Define
g(u, v) = f (v, u), ∀u, v ∈ V.
Hence,
ĝ(u) = g(u, u) = f (u, u) = f (u, u) = fˆ(u), ∀u ∈ V.
From this and the last exercise, we obtain g = f , so that
f (u, v) = f (v, u), ∀u, v ∈ V.
Remark 6.2.6 Let A : D ⊂ V → V be a symmetric operator, that is, such that

(Au, v)V = (u, Av)V , ∀u, v ∈ V,
where V is a space with inner product.
Thus,
f (u, v) = (Au, v)V
= (u, Av)V
= (Av, u)V
= f (v, u), ∀u, v ∈ V, (6.5)
so that f is symmetric.
Definition 6.2.7 Let V be a normed vector space. A sesquilinear functional f : V × V → C is said to be

bounded if there exists K > 0 such that
| f (u, v)| ≤ KkukV kvkV , ∀u, v ∈ V. (6.6)
Defining
B = {K > 0 such that (6.6) is satisfied }
we also define the norm of f , denoted by k f k as
k f k = inf{K : K ∈ B}.
Moreover, defining
C = {K > 0 such that | fˆ(u)| ≤ KkukV2 , ∀u ∈ V },
we define also the norm of fˆ, denoted by k fˆk, as
k fˆk = inf{K > 0 : K ∈ C}.
Proposition 6.2.8 Considering the context of the last definition,
kfk = sup {| f (u, v)| : kukV = kvkV = 1}

(u,v)∈V ×V
and
k fˆk = sup{| fˆ(u)| : kukV = 1}.
u∈V
Proof 6.5 We firstly denote
α= sup {| f (u, v)| : kukV = kvkV = 1}.

(u,v)∈V ×V
Observe that
| f (u, v)| ≤ k f kkukV kvkV , ∀u, v ∈ V,
so that
α ≤ k f k. (6.7)
On the other hand,

kukV kvkV
| f (u, v)| = f u ,v
kukV kvkV

u v
= f , kukV kvkV
kukV kvkV
≤ α kukV kvkV , ∀u =
6 0, v 6= 0. (6.8)
Hence α ∈ B so that
α ≥ inf B = k f k. (6.9)
From (6.7) and (6.9) we may infer that
α = k f k.
Similarly, the second result may be proven.
Theorem 6.2.9 Let V be a complex normed vector space and let F : V ×V → C be a sesquilinear, bounded
and symmetric functional. Under such hypotheses,
k f k = k fˆk.

ˆ 1 ˆ 1
f (u, v) = f (u + v) − f (u − v)
2 2

ˆ 1 ˆ 1
+i f (u + iv) − i f (u − iv) , ∀u, v ∈ H. (6.10)
2 2
Since fˆ(u) ∈ R, ∀u ∈ V , we have that

1 1
|Re[ f (u, v)]| ≤ fˆ (u + v) + fˆ (u − v)
2 2
1 ˆ 1
≤ k f kku + vkV2 + k fˆkku − vkV2
4 4
1 ˆ 2
k f k 2kukV + 2kvkV2 , ∀u, v ∈ V.

= (6.11)
4
Thus, if kukV = kvkV = 1, we get
|Re[ f (u, v)]| ≤ k fˆk.
Observe that in its polar form, we have
f (u, v) = reiθ .
Denoting α = e−iθ , we obtain
f (αu, v) = α f (u, v) = r = | f (u, v)| = |Re[ f (αu, v)]| ≤ k fˆk.
Thus,
k f k = sup{| f (u, v)| : kukV = kvkV = 1} ≤ k fˆk.
However, from the definitions, k f k ≥ k fˆk.
From these last two lines, we may infer that
k f k = k fˆk.
Definition 6.2.10 (Normal operator) Let H be a complex Hilbert space. We say that a bounded linear op-
erator A : H → H is normal as
A∗ A = AA∗ .
Theorem 6.2.11 Let H be a complex Hilbert space. and let A : H → H be a bounded linear operator. Under
such hypotheses, A is normal if, and only if,
kA∗ ukH = kAukH , ∀u ∈ H.

Proof 6.7 Suppose A is normal. Thus,

(A∗ Au, u)H = (AA∗ u, u)H
so that
kAuk2H = (Au, Au)H = (A∗ u, A∗ u)H = kA∗ uk2H ,
that is,
kAukH = kA∗ ukH , ∀u ∈ H.
Reciprocally, suppose that
kAukH = kA∗ ukH , ∀u ∈ H.
Hence
(Au, Au)H = (A∗ u, A∗ u)H
so that
(A∗ Au, u)H = (A∗∗ A∗ u, u)H = (AA∗ u, u)H , ∀u ∈ H.
From this, denoting
f1 (u, v) = (A∗ Au, v)H
and
f2 (u, v) = (AA∗ u, v)H
we obtain
fˆ1 (u) = fˆ2 (u), ∀u ∈ H.
Thus, f1 = f2 , so that
(A∗ Au, v)H = (AA∗ u, v)H , ∀u, v ∈ H.
From this, we may infer that
A∗ A = AA∗ .
Theorem 6.2.12 Let H be a complex Hilbert space and let A ∈ L(H). Under such hypotheses the following
proprieties are equivalent.
1. There exists λ ∈ π(A) such that |λ | = kAk.
2.
kAk = sup{|(Au, u)H | : kukH = 1}.
u∈H
Proof 6.8
1 implies 2: Suppose λ ∈ π(A) is such that
|λ | = kAk.
We shall prove that

λ ∈ {(Au, u)H : u ∈ H, kukH = 1}.
From this we may obtain
kAk = |λ |
≤ sup{|(Au, u)H | : kukH = 1}
u∈H
≤ sup {|(Au, v)H | : kukH = 1, kvkH = 1}
(u,v)∈H×H
= kAk. (6.12)
which would complete the first part of the proof.

From λ ∈ π(A), there exists {un }n∈N ⊂ H such that
kun kH = 1
and
kAun − λ un kH → 0.
Thus,
|(Aun , un )H − λ | = |(Aun , un )H − λ (un , un )H |
= |(Aun − λ un , un )H |
≤ kAun − λ un kH kun kH
→ 0, as n → ∞. (6.13)
Thus,
λ ∈ {(Au, u)H : u ∈ H, kukH = 1}.
The first part of the proof is complete.

2 implies 1:
Reciprocally, suppose
kAk = sup{|(Au, u)H | : kukH = 1}.
u∈H
Hence, there exists a sequence {un } ⊂ H such that kun kH = 1, ∀n ∈ N, and

|(Aun , un )H | → kAkH .
Thus {(Aun , un )H } ⊂ C is a bounded sequence. From this there exists a subsequence {(Aunk , unk )H }
of {(Aun , un )H } and λ ∈ C such that
(Aunk , unk )H → λ , as k → ∞.
Therefore,
kAunk − λ unk k2H = kAunk k2H − λ (Aunk , unk )H − λ (unk , Aunk )H + |λ |2
≤ kAk2 kunk k2H − λ (Aunk , unk )H − λ (Aunk , unk )H + |λ |2
→ |λ |2 − λ λ − λ λ + |λ |2 = 0. (6.14)
Summarizing,
kAunk − λ unk kH → 0, as k → ∞,
so that λ ∈ π(A) and |λ | = kAk.
Theorem 6.2.13 Let H be a Hilbert space and let A ∈ L(H) be a self-adjoint operator.
Define
M = sup{(Au, u)H : kukH = 1}
u∈H
and
m = inf {(Au, u)H : kukH = 1}.
u∈H
Under such hypotheses, m ∈ σ (A) and M ∈ σ (A).
Proof 6.9 Choose α ∈ R such that

M − α ≥ m − α > 0.
Define M̂ = M − α and Â = A − αI. Since Â is self-adjoint (see Theorem 6.4.2 for details), we have that
kÂk = M̂.
Thus there exists a subsequence {un } ⊂ H such that kun kH = 1 and

ˆ n , un )H → M̂, as n → ∞.
(Au
Thus,
ˆ n − M̂un , un )H → 0, as n → ∞.
(Au
Hence,
ˆ n − M̂un k2H
kAu ˆ n − M̂un , Au
= (Au ˆ n − M̂ un )H
ˆ n k2H − 2M̂(Au
= kAu ˆ n , un )H + M̂ 2
2 2 ˆ n , un )H + M̂ 2
≤ kÂkH kun kH − 2M̂(Au
→ M̂ 2 − 2M̂ 2 + M̂ 2
= 0. (6.15)
Summarizing,
ˆ n − M̂un kH → 0, as n → ∞,
kAu
so that
kAun − Mun kH → 0, as n → ∞.
M ∈ π(A) ⊂ σ (A).
Similarly, select β ∈ R such that
−m + β ≥ −M + β > 0.
Define Â = −A + β I. The remaining parts of the proof are similar to those of the previous case.
Theorem 6.2.14 Let H be a complex Hilbert space. Let U : H → H be a linear bounded operator.
Under such hypotheses, U is a isometry if, and only if,
U ∗U = I,
where I denotes the identity operator.

(Uu,Uv)H = (u, v)H , ∀u, v ∈ H,
if, and only if,
(U ∗Uu, v)H = (u, v), ∀u, v ∈ H,
if, and only if,
U ∗U = I.
Theorem 6.2.15 Let H be a complex Hilbert space. Under such hypothesis, U : H → H is a bijective isometry
if, and only if,
U ∗U = UU ∗ = I, in H.
Proof 6.11 Assume U : H → H is a bijective isometry.

From the last theorem U ∗U = I, and U is a bijection, we obtain U −1 = U ∗ so that UU ∗ = I in H.
On the other hand, if U ∗U = UU ∗ = I in H, we have that U ∗ = U −1 and the domain of U ∗ = U −1 is H,
so that the range of U is also H.
From this U is a bijection.
Moreover, (u, v)H = (U ∗Uu, v)H = (Uu,Uv)H , ∀u, v ∈ H, and thus U is a bijective isometry.
Theorem 6.2.16 Let H be a complex Hilbert space. Suppose U : H → H is a linear operator such that
kUukH = kukH , ∀u ∈ H
(in such a case we say that U is unitary).

Under such hypotheses, U is a isometry.
Proof 6.12 From the hypotheses,
(Uu,Uu)H = (u, u)H , ∀u ∈ H.
Thus,
(U ∗Uu, u)H = (u, u)H , ∀u ∈ H,
so that
((U ∗U − I)u, u)H = 0, ∀u ∈ H.
Since U ∗U − I is self adjoint, it follows that
kU ∗U − Ik = sup{|((U ∗U − I)u, u)H | : kukH = 1} = 0,

u∈H
so that U ∗U = I.
From this and Theorem 6.2.14, we have that U is a isometry.
Corollary 6.2.17 Let U : H → H be a unitary operator.

Under such hypotheses, if λ ∈ C is an eigenvalue of U then
|λ | = 1.
Proof 6.13 Suppose Uu = λ u and kukH = 1. Thus,
1 = (u, u)H = (Uu,Uu)H = (λ u, λ u)H = λ λ (u, u)H = |λ |2 .
Theorem 6.2.18 Let H be a complex Hilbert space. Let A : DA ⊂ H → H be a linear self-adjoint operator
but not necessarily bounded, where DA is dense in H, that is, DA = H. Let U be the Cayley transform of A,
that is U = (A − i)(A + i)−1 .
Under such hypotheses, U is unitary.
Proof 6.14 First, we shall prove that A ± i is injective, so that its inverse is well defined on R(A ± i).
Since A is self-adjoint, we have that
k(A ± i)uk2H = ((A ± i)u, (A ± i))H

= (Au, Au)H ± (Au, iu)H ± (iu, Au)H + (iu, iu)H
2
= kAukH + (±i ∓ i)(u, Au)H + kuk2H
= kAuk2H + kuk2H ≥ kuk2H . (6.16)
Thus, if (A ± i)u = 0 then u = 0, so that (A ± i) is injective.

Now, we are going to show that
R(A ± i) = H.
Let z ⊥ R(A + i). Hence,
((A + i)u, z)H = (Au, z)H + i(u, z)H = 0, ∀u ∈ DA .
From this
(Au, z)H = ((u, iz), ∀u ∈ H,
so that
A∗ z = Az = iz,
that is,
(A − i)z = 0.
Thus, z = 0.
Summarizing these last results, if z ⊥ R(A + i) then z = 0, so that
R(A + i) = H.
Now we are going to show that R(A + i) = R(A + i) = H. Let v ∈ H. Thus there exists a sequence {vn } ⊂
R(A + i) such that
vn → v, in norm, as n → ∞.
Therefore there exists a sequence {un } ⊂ DA such that
Aun + iun = vn → v, as n → ∞.
Similarly as above, we may obtain,
kvn − vm k2H = kAun + iun − Aum − ium k2H = kA(un − um )k2H + kun − um kH
2
, ∀m, n ∈ N.
From this, since {vn } is a Cauchy sequence, we may infer that {un } and {Aun } are Cauchy sequences, so
that there exists u ∈ H, and w ∈ H such that
Aun → w
and
un → u, as n → ∞.
Since A = A∗ is closed (see Theorem 6.6.7 for details), we may infer that w = Au and
Aun + iun → Au + iu.
since A is closed we get (u, Au) ∈ Gr(A) where Gr(A) denotes the graph of A, so that Au + iu = v ∈ R(A + i).
Summarizing, if v ∈ H then v ∈ R(A + i), so that R(A + i) = H = R(A + i).
A similar result we may obtain for A − i, that is
R(A − i) = H.
Observe that
U = (A − i)(A + i)−1
and
R((A + i)−1 ) = DA = D(A+i) ,
and R(A − i) = H.
Thus R(U) = H, that is U : H → H is linear and onto (recalling that D(A+i)−1 = H).
At this point, we shall prove that U is unitary.
Let v ∈ H. Since R(A + i) = H, there exists u ∈ DA such that
(A + i)u = v.
Hence
Uv = (A − i)(A + i)−1 v = (A − i)u,
so that
kUvk2H = k(A − i)uk2H

= kAuk2H + kuk2H
2
= k(A + i)ukH
= kvk2H , ∀v ∈ H. (6.17)
Summarizing,
kUvkH = kvkH , ∀v ∈ H,
so that U is unitary.
Remark 6.2.19 Let v ∈ H. Since R(A + i) = H, we may obtain u ∈ H such that v = (A + i)u.
From this we have
Uv = (A − i)(A + i)−1 v = (A − i)u,
so that
(I +U)v = 2Au
and
(I −U)v = 2iu.
Thus, if (I −U)v = 0, then u = 0 so that v = (A + i)u = 0.
Therefore, I −U is injective and its inverse exists on R(I −U).
Moreover, for u ∈ R(I −U) as above, we have,
(I +U)[(I −U)−1 ](2iu) = (I +U)[(I −U)−1 ](I −U)v = (I +U)v = 2Au,
so that
Au = i(I +U)(I −U )−1 u, ∀u ∈ R(I −U).
In the next lines we shall show that R(I −U) = DA .
Indeed, let v ∈ H. Thus, from the last lines above, there exists u ∈ DA such that
(I −U)v = 2iu
so that R(I −U) ⊂ DA .

Reciprocally, let u ∈ DA and define v = (A + i)u.
Therefore,
2iu = (I −U)v
so that u ∈ R(I −U), ∀u ∈ DA . Thus,
DA ⊂ R(I −U)
so that
R(I −U) = DA .
From such last results, we may infer that
A = i(I +U)(I −U)−1 ,
in DA .
6.3 About the spectrum of a linear operator defined on a Banach space

Definition 6.3.1 Let U be a Banach space and let A ∈ L (U). We recall that a complex number λ is said to
be in the resolvent set ρ(A) of A, if
λI −A
is a bijection with a bounded inverse. As previously indicated, we call
Rλ (A) = (λ I − A)−1
the resolvent of A in λ .
If λ 6∈ ρ(A), we write
λ ∈ σ (A) = C − ρ(A),
where σ (A) is said to be the spectrum of A.
Definition 6.3.2 Let A ∈ L (U).

1. If u 6= θ and Au = λ u for some λ ∈ C then u is said to be an eigenvector of A and λ the corresponding
eigenvalue. If λ is an eigenvalue, then (λ I − A) is not injective and therefore λ ∈ σ (A).
The set of eigenvalues is said to be the point spectrum of A.
2. If λ is not an eigenvalue but
R(λ I − A)
is not dense in U and therefore λ I − A is not a bijection, we have that λ ∈ σ (A). In this case we say
that λ is in the residual spectrum of A, or briefly λ ∈ Res[σ (A)].
Theorem 6.3.3 Let U be a Banach space and suppose that A ∈ L (U). Then ρ(A) is an open subset of C
and
F(λ ) = Rλ (A)
is an analytic function with values in L (U) on each connected component of ρ(A). For λ , µ ∈ σ (A), Rλ (A)
and Rµ (A) commute and
Rλ (A) − Rµ (A) = (µ − λ )Rµ (A)Rλ (A).
Proof 6.15 Let λ0 ∈ ρ(A). We will show that λ0 is an interior point of ρ(A).
Observe that symbolically we may write
1 1
=
λ −A λ − λ0 + (λ0 − A)
 
1  1
= 
λ0 − A 1 − λ0 −λ
λ0 −A
∞ n !
1 λ0 − λ
= 1+ ∑ . (6.18)
λ0 − A n=1 λ0 − A
Define,
∞
R̂λ (A) = Rλ0 (A){I + ∑ (λ − λ0 )n (Rλ0 )n }. (6.19)
n=1
Observe that
k(Rλ0 )n k ≤ kRλ0 kn .
Thus, the series indicated in (6.19) will converge in norm if
|λ − λ0 | < kRλ0 k−1 . (6.20)
Hence, for λ satisfying (6.20), R̂(A) is well defined and we can easily check that
(λ I − A)R̂λ (A) = I = R̂λ (A)(λ I − A).
Therefore
R̂λ (A) = Rλ (A), if |λ − λ0 | < kRλ0 k−1 ,
so that λ0 is an interior point. Since λ0 ∈ ρ(A) is arbitrary, we have that ρ(A) is open. Finally, observe that
Rλ (A) − Rµ (A) = Rλ (A)(µI − A)Rµ (A) − Rλ (A)(λ I − A)Rµ (A)

= Rλ (A)(µI)Rµ (A) − Rλ (A)(λ I)Rµ (A)
= (µ − λ )Rλ (A)Rµ (A) (6.21)
Interchanging the roles of λ and µ we may conclude that Rλ and Rµ commute.
Corollary 6.3.4 Let U be a Banach space and A ∈ L (U). Then the spectrum of A is non-empty.
Proof 6.16 Observe that if

kAk
<1
|λ |
we have
(λ I − A)−1 = [λ (I − A/λ )]−1

= λ −1 (I − A/λ )−1
!
∞ n
−1 A
= λ I+ ∑ . (6.22)
n=1 λ
Therefore we may obtain

∞ n !
A
Rλ (A) = λ −1 I+ ∑ .
n=1 λ
In particular
kRλ (A)k → 0, as |λ | → ∞. (6.23)

σ (A) = 0/.
In such a case Rλ (A) would be a entire bounded analytic function. From Liouville’s theorem, Rλ (A) would
be constant, so that from (6.23) we would have
Rλ (A) = θ , ∀λ ∈ C,
which is a contradiction.
Proposition 6.3.5 Let H be a Hilbert space and A ∈ L (H).
1. If λ ∈ Res[σ (A)] then λ ∈ Pσ (A∗ ).
2. If λ ∈ Pσ (A) then λ ∈ Pσ (A∗ ) ∪ Res[σ (A∗ )].
Proof 6.17
1. If λ ∈ Res[σ (A)] then
R(A − λ I) =
6 H.
Therefore there exists v ∈ (R(A − λ I))⊥ , 6 θ such that
v=
(v, (A − λ I)u)H = 0, ∀u ∈ H
that is
((A∗ − λ I)v, u)H = 0, ∀u ∈ H
so that
(A∗ − λ I)v = θ ,
which means that λ ∈ Pσ (A∗ ).
2. Suppose there exists v 6= θ such that
(A − λ I)v = θ ,
and
λ 6∈ Pσ (A∗ ).
Thus,
(u, (A − λ I)v))H = 0, ∀u ∈ H,
so that
((A∗ − λ I)u, v)H = 0, ∀u ∈ H.
Since
(A∗ − λ I)u =
6 θ , ∀u ∈ H, u =
6 θ,
we get v ∈ (R(A∗ − λ I))⊥ , so that R(A∗ − λ I) =
6 H.
Hence, λ ∈ Res[σ (A∗ )].
Theorem 6.3.6 Let A ∈ L (H) be a self-adjoint operator. then

1. σ (A) ⊂ R.
2. Eigenvectors corresponding to distinct eigenvalues of A are orthogonal.
Proof 6.18 Let µ, λ ∈ R. Thus, given u ∈ H we have
k(A − (λ + µi))uk2 = k(A − λ )uk2 + µ 2 kuk2 ,
so that
k(A − (λ + µi))uk2 ≥ µ 2 kuk2 .
6 0, A − (λ + µi) has a bounded inverse on its range, which is closed. If R(A − (λ + µi)) =
Therefore, if µ = 6 H
then by the last result (λ − µi) would be in the point spectrum of A, which contradicts the last inequality.
6 0 then λ + µi ∈ ρ(A). To complete the proof, suppose
Hence, if µ =
Au1 = λ1 u1 ,
and
Au2 = λ2 u2 ,
where
λ1 , λ2 ∈ R, λ1 =
6 λ2 and u1 , u2 =
6 θ.
Thus,
(λ1 − λ2 )(u1 , u2 )H = λ1 (u1 , u2 )H − λ2 (u1 , u2 )H

= (λ1 u1 , u2 )H − (u1 , λ2 u2 )H
= (Au1 , u2 )H − (u1 , Au2 )H
= (u1 , Au2 )H − (u1 , Au2 )H
= 0. (6.24)
Since λ1 − λ2 =
6 0 we get
(u1 , u2 )H = 0.
We finish this section with an exercise and its solution.

Exercise 6.3.7 Let H be a complex Hilbert space e let A ∈ L(H) be a self-adjoint operator. Prove that
λ ∈ σ (A) if, and only if, there exists a sequence {un } ⊂ H such that kun kH = 1, ∀n ∈ N and
kAun − λ un kH → 0, as n → ∞.
Solution: Suppose λ ∈ C is such that there exists {un } ⊂ H such that kun k = 1, ∀n ∈ N and
kAun − λ un kH → 0, as n → ∞. (6.25)
Suppose, to obtain contradiction, that λ ∈ ρ(A). Thus, (A − λ I)−1 exists and it is bounded, so that there
exists K > 0 such that
kAu − λ IukH ≥ KkukH , ∀u ∈ H.
From this we obtain
kAun − λ un kH ≥ K, ∀n ∈ N
Thus λ 6∈ ρ(A) so that λ ∈ σ (A).

Reciprocally, suppose λ ∈ σ (A). Suppose, to obtain contradiction, that there exists K > 0 such that
kAu − λ ukH ≥ KkukH , ∀u ∈ H. (6.26)
Thus, (A − λ I)−1 exists and it is bounded. Since λ ∈ σ (A), we must have that R(A − λ I) is not dense, so
that λ ∈ Res[σ (A)].
From Proposition 6.3.5, we have λ ∈ Pσ (A∗ ).
Since A = A∗ , from this we obtain λ = λ ∈ P(σ (A)) which contradicts (A − Iλ )−1 to exist.
Thus, we may infer that it does not exist K > 0 such that (6.26) holds.
From this, for each n ∈ N there exists un ∈ H such that
kun kH = 1
and
kAun − λ un kH < 1/n
so that
kAun − λ un kH → 0.
The solution is complete.
6.4 The spectral theorem for bounded self-adjoint operators

Let H be a complex Hilbert space. Consider A : H → H a linear bounded operator, that is A ∈ L (H), and
suppose also that such an operator is self-adjoint. Define
m = inf {(Au, u)H | kukH = 1},

u∈H
and
M = sup{(Au, u)H | kukH = 1}.
u∈H
Remark 6.4.1 It is possible to prove that for a linear self-adjoint operator A : H → H we have
kAk = sup{|(Au, u)H | | u ∈ H, kukH = 1}.
This propriety, which prove in the next lines, is crucial for the subsequent results, since for example for A, B
linear and self-adjoint and ε > 0 we have
−εI ≤ A − B ≤ εI,
we also would have

kA − Bk < ε.
So, we present the following basic result.
Theorem 6.4.2 Let H be a real Hilbert space and let A : H → H be a bounded linear self-adjoint operator.
Define
α = max{|m|, |M|},
where
m = inf {(Au, u)H | kukH = 1},

u∈H
and
M = sup{(Au, u)H | kukH = 1}.
u∈H
Then,
kAk = α.
(A(u + v), u + v)H = (Au, u)H + (Av, v)H + 2(Au, v)H ,
and
(A(u − v), u − v)H = (Au, u)H + (Av, v)H − 2(Au, v)H .
Thus,
4(Au, v) = (A(u + v), u + v)H − (A(u − v), u − v)H ≤ Mku + vkU2 − mku − vkU2 ,
so that
4(Au, v)H ≤ α(ku + vkU2 + ku − vkU2 ).
Hence, replacing v by −v we obtain
−4(Au, v)H ≤ α (ku + vkU2 + ku − vkU2 ),
and therefore
4|(Au, v)H | ≤ α(ku + vkU2 + ku − vkU2 ).
Replacing v by β v, we get
4|(A(u), v)H | ≤ 2α(kukU2 /β + β kvkU2 ).
Minimizing the last expression in β > 0, for the optimal
β = kukU /kvkU ,
we obtain
|(Au, v)H | ≤ αkukU kvkU , ∀u, v ∈ U.
Thus
kAk ≤ α.
On the other hand,
|(Au, u)H | ≤ kAkkukU2 ,
so that
|M| ≤ kAk
and
|m| ≤ kAk,
so that
α ≤ kAk.
Remark 6.4.3 A similar result is valid as H is a complex Hilbert space.

At this point we start to develop the spectral theory. Define by P the set of all real polynomials defined in
R. Define
Φ1 : P → L (H),
by
Φ1 (p(λ )) = p(A), ∀p ∈ P.
Thus we have
1. Φ1 (p1 + p2 ) = p1 (A) + p2 (A),
2. Φ1 (p1 · p2 ) = p1 (A)p2 (A),
3. Φ1 (α p) = α p(A), ∀α ∈ R, p ∈ P
4. if p(λ ) ≥ 0, on [m, M], then p(A) ≥ θ ,
We will prove (4):
Consider p ∈ P. Denote the real roots of p(λ ) less or equal to m by α1 , α2 , ..., αn and denote those that
are greater or equal to M by β1 , β2 , ..., βl . Finally denote all the remaining roots, real or complex by
v1 + iµ1 , ..., vk + iµk .
Observe that if µi = 0 then vi ∈ (m, M). The assumption that p(λ ) ≥ 0 on [m, M] implies that any real root in
(m, M) must be of even multiplicity.
Since complex roots must occur in conjugate pairs, we have the following representation for p(λ ) :
n l k
p(λ ) = a ∏(λ − αi ) ∏(βi − λ ) ∏((λ − vi )2 + µi2 ),
i=1 i=1 i=1
where a ≥ 0. Observe that

A − αi I ≥ θ ,
since,
(Au, u)H ≥ m(u, u)H ≥ αi (u, u)H , ∀u ∈ H,
and by analogy
βi I − A ≥ θ .
On the other hand, since A − vk I is self-adjoint, its square is positive and hence since the sum of positive
operators is positive, we obtain
(A − vk I )2 + µk2 I ≥ θ .
Therefore,
p(A) ≥ θ .
The idea is now to extend de domain of Φ1 to the set of upper semi-continuous functions, and such set
we will denote by Cup .
Observe that if f ∈ Cup , there exists a sequence of continuous functions {gn } such that
gn ↓ f , pointwise ,
that is,
gn (λ ) ↓ f (λ ), ∀λ ∈ R.
Considering the Weierstrass theorem, since gn ∈ C([m, M]) we may obtain a sequence of polynomials {pn }
such that
1 1
gn + n − pn < n ,
2 ∞ 2
where the norm k · k∞ refers to [m, M]. Thus,
pn (λ ) ↓ f (λ ), on [m, M].
Therefore,
p1 (A) ≥ p2 (A) ≥ p3 (A) ≥ ... ≥ pn (A) ≥ ...
Since pn (A) is self-adjoint for all n ∈ N, we have
p j (A)pk (A) = pk (A)p j (A), ∀ j, k ∈ N.
Then the lim pn (A) (in norm) exists, and we denote

n→∞
lim pn (A) = f (A).

n→∞
Now recall the Dini’s theorem:

Theorem 6.4.4 (Dini) Let {gn } be a sequence of continuous functions defined on a compact set K ⊂ R.
Suppose gn → g point-wise and monotonically on K. Under such assumptions the convergence in question is
also uniform.
Now suppose that {pn } and {qn } are sequences of polynomial such that
pn ↓ f , and qn ↓ f ,
we will show that

lim pn (A) = lim qn (A).
n→∞ n→∞
First, observe that being {pn } and {qn } sequences of continuous functions we have that
ĥnk (λ ) = max{pn (λ ), qk (λ )}, ∀λ ∈ [m, M]
is also continuous, ∀n, k ∈ N. Now fix n ∈ N and define
hk (λ ) = max{pk (λ ), qn (λ )}.
observe that
hk (λ ) ↓ qn (λ ), ∀λ ∈ R,
so that by Dini’s theorem
hk → qn , uniformly on [m, M].
It follows that for each n ∈ N there exists kn ∈ N such that if k > kn then
1
hk (λ ) − qn (λ ) ≤ , ∀λ ∈ [m, M].
n
Since
pk (λ ) ≤ hk (λ ), ∀λ ∈ [m, M],
we obtain
1
pk (λ ) − qn (λ ) ≤ , ∀λ ∈ [m, M].
n
By analogy, we may show that for each n ∈ N there exists k̂n ∈ N such that if k > k̂n then
1
qk (λ ) − pn (λ ) ≤ .
n
From above we obtain

1
lim pk (A) ≤ qn (A) + .
k→∞ n
Since the self adjoint qn (A) + 1/n commutes with the
lim pk (A)
k→∞
we obtain

1
lim pk (A) ≤ lim qn (A) +
k→∞ n→∞ n
≤ lim qn (A). (6.27)
n→∞
Similarly, we may obtain

lim qk (A) ≤ lim pn (A),
k→∞ n→∞
so that
lim qn (A) = lim pn (A) = f (A).
n→∞ n→∞
Hence, we may extend Φ1 : P → L (H) to Φ2 : Cup → L (H) where Cup as earlier indicated, denotes the set
of upper semi-continuous functions, where
Φ2 ( f ) = f (A).
Observe that Φ2 has the following properties:

1. Φ2 ( f1 + f2 ) = Φ2 ( f1 ) + Φ2 ( f2 ),
2. Φ2 ( f1 · f2 ) = f1 (A) f2 (A),
3. Φ2 (α f ) = αΦ2 ( f ), ∀α ∈ R, α ≥ 0.
4. if f1 (λ ) ≥ f2 (λ ), ∀λ ∈ [m, M], then
f1 (A) ≥ f2 (A).
up
The next step is to extend Φ2 to Φ3 : C− → L (H), where
up
C− = { f − g | f , g ∈ Cup }.
up
For h = f − g ∈ C− we define
Φ3 (h) = f (A) − g(A).
up
Now we will show that Φ3 is well defined. Suppose that h ∈ C− and
h = f1 − g1 and h = f2 − g2 .
Thus,
f1 − g1 = f2 − g2 ,
that is,
f1 + g2 = f2 + g1 ,
so that from the definition of Φ2 we obtain
f1 (A) + g2 (A) = f2 (A) + g1 (A),

that is,
f1 (A) − g1 (A) = f2 (A) − g2 (A).
Therefore, Φ3 is well defined. Finally, observe that for α < 0
α( f − g) = −αg − (−α) f ,
where −αg ∈ Cup and −α f ∈ Cup . Thus,
Φ3 (α f ) = α f (A) = αΦ3 ( f ), ∀α ∈ R.
6.4.1 The spectral theorem

Consider the upper semi-continuous function

1, if λ ≤ µ,
hµ (λ ) = (6.28)
0, if λ > µ.
Denote
E(µ) = Φ3 (hµ ) = hµ (A).
Observe that
hµ (λ )hµ (λ ) = hµ (λ ), ∀λ ∈ R,
so that
[E(µ)]2 = E(µ), ∀µ ∈ R.
Therefore,
{E(µ) | µ ∈ R}
is a family of orthogonal projections. Also observe that if ν ≥ µ we have
hν (λ )hµ (λ ) = hµ (λ )hν (λ ) = hµ (λ ),
so that
E(ν)E(µ) = E(µ)E(ν) = E(µ), ∀ν ≥ µ.
If µ < m, then hµ (λ ) = 0, on [m, M], so that
E(µ) = 0, if µ < m.
Similarly, if µ ≥ M them hµ (λ ) = 1, on [m, M], so that
E(µ) = I, if µ ≥ M.
Next we show that the family {E(µ)} is strongly continuous from the right. First we will establish a sequence
of polynomials {pn } such that
pn ↓ hµ ,
and
pn (λ ) ≥ hµ + 1 (λ ), on [m, M].
n
Observe that for any fixed n there exists a sequence of polynomials {pnj } such that
pnj ↓ hµ+1/n , point-wise .
Consider the monotone sequence
gn (λ ) = min{ prs (λ ) | r, s ∈ {1, ..., n}}.

Thus,
gn (λ ) ≥ hµ + 1 (λ ), ∀λ ∈ R,
n
and we obtain
lim gn (λ ) ≥ lim hµ + 1 (λ ) = hµ (λ ).
n→∞ n→∞ n
On the other hand,

gn (λ ) ≤ prn (λ ), ∀λ ∈ R, ∀r ∈ {1, ..., n},
so that
lim gn (λ ) ≤ lim prn (λ ).
n→∞ n→∞
Therefore,
lim gn (λ ) ≤ lim lim pr (λ )

n→∞ r→∞ n→∞ n
= hµ (λ ). (6.29)
Thus,
lim gn (λ ) = hµ (λ ).
n→∞
Observe that gn are not necessarily polynomials. To set a sequence of polynomials, observe that we may
obtain a sequence {pn } of polynomials such that
1
|gn (λ ) + 1/n − pn (λ )| < , ∀λ ∈ [m, M], n ∈ N.
2n
so that
pn (λ ) ≥ gn (λ ) + 1/n − 1/2n ≥ gn (λ ) ≥ hµ+1/n (λ ).
Thus,
pn (A) → E(µ),
and
pn (A) ≥ hµ + 1 (A) = E(µ + 1/n) ≥ E (µ).
n
Therefore we may write

E(µ) = lim pn (A) ≥ lim E(µ + 1/n) ≥ E(µ).
n→∞ n→∞
Thus,
lim E(µ + 1/n) = E(µ).
n→∞
From this we may easily obtain the strong continuity from the right.
For µ ≤ ν we have
µ(hν (λ ) − hµ (λ )) ≤ λ (hν (λ ) − hµ (λ ))
≤ ν(hν (λ ) − hµ (λ )). (6.30)
To verify this observe that if λ < µ or λ > ν then all terms involved in the above inequalities are zero. On
the other hand if
µ ≤λ ≤ν
then
hν (λ ) − hµ (λ ) = 1,
so that in any case (6.30) holds. From the monotonicity property we have
µ(E(ν) − E(µ)) ≤ A(E(ν) − E(µ))
≤ ν(E(ν) − E(µ)). (6.31)
Now choose a, b ∈ R such that
a < m and b ≥ M.
Suppose given ε > 0. Choose a partition P0 of [a, b], that is
P0 = {a = λ0 , λ1 , ..., λn = b},
such that
max {|λk − λk−1 |} < ε.
k∈{1,...,n}
Hence,
λk−1 (E(λk ) − E(λk−1 )) ≤ A(E(λk ) − E(λk−1 ))
≤ λk (E(λk ) − E(λk−1 )). (6.32)
Summing up on k and recalling that
n
∑ E(λk ) − E(λk−1 ) = I,
k=1
we obtain
n
∑ λk−1 (E(λk ) − E(λk−1 )) ≤ A
k=1
n
≤ ∑ λk (E(λk ) − E(λk−1 )). (6.33)
k=1
Let λk0 ∈ [λk−1 , λk ]. Since (λk − λk0 ) ≤ (λk − λk−1 ) from (6.32) we obtain
n n
A − ∑ λk0 (E(λk ) − E(λk−1 )) ≤ ε ∑ (E(λk ) − E (λk−1 ))
k=1 k=1
= εI. (6.34)
By analogy
n
−εI ≤ A − ∑ λk0 (E(λk ) − E(λk−1 )). (6.35)
k=1
Since
n
A − ∑ λk0 (E(λk ) − E(λk−1 ))
k=1
is self-adjoint we obtain
n
kA − ∑ λk0 (E(λk ) − E (λk−1 ))k < ε.
k=1
Being ε > 0 arbitrary, we may write
Z b
A= λ dE(λ ),
a
that is, Z M
A= λ dE(λ ).
m−
Remark 6.4.5 Consider again the function hµ : R → R where

1, if λ ≤ µ
hµ (λ ) = (6.36)
0, if λ > µ.
Let H be a complex Hilbert space and let A ∈ L(H), where A is a self-adjoint operator.
Suppose f ∈ C([m, M]) where
m = inf {(Au, u)H : kukH = 1},

u∈H
and
M = sup{(Au, u)H : kukH = 1}.
u∈H
Let ε > 0. Since f is uniformly continuous on the compact set [m, M], there exists δ > 0 such that if
x, y ∈ [m, M] and |x − y| < δ , then
| f (x) − f (y)| < ε. (6.37)
Let P = {λ0 = m, λ1 , . . . , λn = M} be a partition of [m, M], such that kPk = max{λk − λk−1 : k ∈
{1, . . . , n}} < δ .
Choose
λk0 ∈ (λk−1 , λk ), ∀k ∈ {1, . . . , n}
and observe that
1, if λk−1 < λ ≤ λk
hλk (λ ) − hλk−1 (λ ) = (6.38)
0, otherwise.
From this and (6.37), we may obtain
n
f (λ ) − ∑ f (λk0 )[hλk (λ ) − hλk−1 (λ )] < ε, ∀λ ∈ [m, M].
k=1
Therefore, for the corresponding operators, we have got

n
f (A) − ∑ f (λk0 )[E(λk ) − E(λk−1 )] < ε.
k=1
Since ε > 0, the partition P and {λk0 } have been arbitrary, we may denote
Z M
f (A) = f (λ )dE(λ ).
m−
6.5 The spectral decomposition of unitary transformations

Definition 6.5.1 Let H be a Hilbert space. A transformation U : H → H is said to be unitary if
(Uu,Uu)H = (u, u)H , ∀u, u ∈ H.
Observe that in this case

U ∗U = UU ∗ = I,
so that
U −1 = U ∗ .
Theorem 6.5.2 Every Unitary transformation U has a spectral decomposition

Z 2π
U= eiφ dE(φ ),
0−
where {E(φ )} is a spectral family on [0, 2π]. Furthermore E(φ ) is continuous at 0 and it is the limit of
polynomials in U and U −1 .
We present just a sketch of the proof. For the trigonometric polynomials

n
p(eiφ ) = ∑ ck eikφ ,
k=−n
consider the transformation

n
p(U) = ∑ ckU k ,
k=−n
where ck ∈ C, ∀k ∈ {−n, ..., 0, ..., n}.

Observe that
n
p(eiφ ) = ∑ ck e−ikφ ,
k=−n
so that the corresponding operator is

n n
p(U)∗ = ∑ ckU −k = ∑ ck (U ∗ )k .
k=−n k=−n
Also if
p(eiφ ) ≥ 0
there exists a polynomial q such that
p(eiφ ) = |q(eiφ )|2 = q(eiφ )q(eiφ ),
so that
p(U) = [q(U)]∗ q(U).
Therefore
(p(U)v, v)H = (q(U)∗ q(U)v, v)H = (q(U)v, q(U)v)H ≥ 0, ∀v ∈ H,
which means
p(U) ≥ 0.
Define the function hµ (φ ) by

1, if 2kπ < φ ≤ 2kπ + µ,
hµ (φ ) = (6.39)
0, if 2kπ + µ < φ ≤ 2(k + 1)π,
for each k ∈ {0, ±1, ±2, ±3, ...}. Define E(µ) = hµ (U). Observe that the family {E(µ)} are projections and
in particular
E(0) = 0,
E(2π) = I
and if µ ≤ ν, since
hµ (φ ) ≤ hν (φ ),
we have
E(µ) ≤ E(ν).
Suppose given ε > 0. Let P0 be a partition of [0, 2π] that is,
P0 = {0 = φ0 , φ1 , ..., φn = 2π}
such that
max {|φ j − φ j−1 |} < ε.
j∈{1,...,n}
For fixed φ ∈ [0, 2π], let j ∈ {1, ..., n} be such that
φ ∈ [φ j−1 , φ j ].
n
|eiφ − ∑ eiφk (hφk (φ ) − hφk−1 (φ ))| = |eiφ − eiφ j |
k=1
≤ |φ − φ j | < ε. (6.40)
Thus,
n
0 ≤ |eiφ − ∑ eiφk (hφk (φ ) − hφk−1 (φ ))|2 ≤ ε 2
k=1
so that, for the corresponding operators
n n
0 ≤ [U − ∑ eiφk (E (φk ) − E (φk−1 )]∗ [U − ∑ eiφk (E(φk ) − E(φk−1 )]
k=1 k=1
2
≤ ε I (6.41)
and hence
n
kU − ∑ eiφk (E(φk ) − E(φk−1 )k < ε.
k=1
Being ε > 0 arbitrary, we may infer that
Z 2π
U= eiφ dE(φ ).
0
6.6 Unbounded operators

6.6.1 Introduction
Let H be a Hilbert space. Let A : D(A) → H be an operator, where unless otherwise indicated D(A) is a dense
subset of H. We consider in this section the special case where A is unbounded.
Definition 6.6.1 Given A : D → H we define the graph of A, denoted by Γ(A) by,
Γ(A) = {(u, Au) | u ∈ D}.
Definition 6.6.2 An operator A : D → H is said to be closed if Γ(A) is closed.
Definition 6.6.3 Let A1 : D1 → H and A2 : D2 → H operators. We write A2 ⊃ A1 if D2 ⊃ D1 and
A2 u = A1 u, ∀u ∈ D1 .
In this case we say that A2 is an extension of A1 .

Definition 6.6.4 A linear operator A : D → H is said to be closable if it has a linear closed extension. The
smallest closed extension of A is denote by A and is called the closure of A.
Proposition 6.6.5 Let A : D → H be a linear operator. If A is closable then
Γ(A) = Γ(A).
Proof 6.20 Suppose B is a closed extension of A. Then
Γ(A) ⊂ Γ(B) = Γ(B),
so that if (θ , φ ) ∈ Γ(A) then (θ , φ ) ∈ Γ(B), and hence φ = θ . Define the operator C by
D(C) = {ψ | (ψ, φ ) ∈ Γ(A) for some φ },
and C(ψ) = φ , where φ is the unique point such that (ψ, φ ) ∈ Γ(A). Hence
Γ(C) = Γ(A) ⊂ Γ(B),
so that
A ⊂ C.
However, C ⊂ B and since B is an arbitrary closed extension of A we have
C=A
so that
Γ(C) = Γ(A) = Γ(A).
Definition 6.6.6 Let A : D → H be a linear operator where D is dense in H. Define D(A∗ ) by
D(A∗ ) = {φ ∈ H | (Aψ, φ )H = (ψ, η)H , ∀ψ ∈ D for some η ∈ H}.
In this case we denote

A∗ φ = η.
A∗ defined in this way is called the adjoint operator related to A.
Observe that by the Riesz lemma, φ ∈ D(A∗ ) if and only if there exists K > 0 such that
|(Aψ, φ )H | ≤ KkψkH , ∀ψ ∈ D.
Also note that if

A ⊂ B then B∗ ⊂ A∗ .
Finally, as D is dense in H then
η = A∗ (φ )
is uniquely defined. However the domain of A∗ may not be dense, and in some situations we may have
D(A∗ ) = {θ }.
If D(A∗ ) is dense we define
A∗∗ = (A∗ )∗ .
Theorem 6.6.7 Let A : D → H a linear operator, being D dense in H. Then
1. A∗ is closed,
2. A is closable if and only if D(A∗ ) is dense and in this case
A = A∗∗ .
3. If A is closable then (A)∗ = A∗ .
Proof 6.21
1. We define the operator V : H × H → H × H by
V (φ , ψ) = (−ψ, φ ).
Let E ⊂ H × H be a subspace. Thus if (φ1 , ψ1 ) ∈ V (E ⊥ ) then there exists (φ , ψ) ∈ E ⊥ such that
V (φ , ψ) = (−ψ, φ ) = (φ1 , ψ1 ).
Hence
ψ = −φ1 and φ = ψ1 ,
so that for (ψ1 , −φ1 ) ∈ E⊥ and (w1 , w2 ) ∈ E we have
((ψ1 , −φ1 ), (w1 , w2 ))H×H = 0 = (ψ1 , w1 )H + (−φ1 , w2 )H .
Thus,
(φ1 , −w2 )H + (ψ1 , w1 )H = 0,
and therefore
((φ1 , ψ1 ), (−w2 , w1 ))H×H = 0,
that is
((φ1 , ψ1 ),V (w1 , w2 ))H ×H = 0, ∀(w1 , w2 ) ∈ E.
This means that
(φ1 , ψ1 ) ∈ (V (E))⊥ ,
so that
V (E ⊥ ) ⊂ (V (E))⊥ .
It is easily verified that the implications from which the last inclusion results are in fact equivalences,
so that
V (E ⊥ ) = (V (E))⊥ .
Suppose (φ , η) ∈ H × H. Thus (φ , η) ∈ V (Γ(A))⊥ if and only if
((φ , η), (−Aψ, ψ))H×H = 0, ∀ψ ∈ D,
which holds if and only if

(φ , Aψ)H = (η, ψ)H , ∀ψ ∈ D,
that is, if and only if
(φ , η) ∈ Γ(A∗ ).
Thus,
Γ(A∗ ) = V (Γ(A))⊥ .
Since (V (Γ(A))⊥ is closed, A∗ is closed.
2. Observe that Γ(A) is a linear subset of H × H so that
Γ(A) = [Γ(A)⊥ ]⊥
= V 2 [Γ(A)⊥ ]⊥
= [V [V (Γ(A))⊥ ]]⊥
= [V (Γ(A∗ )]⊥ (6.42)
so that from the proof of item 1, if A∗ is densely defined we get
Γ(A) = Γ[(A∗ )∗ ].
Conversely, suppose D(A∗ ) is not dense. Thus there exists ψ ∈ [D(A∗ )]⊥ such that ψ 6= θ . Let
(φ , A∗ φ ) ∈ Γ(A∗ ). Hence
((ψ, θ ), (φ , A∗ φ ))H ×H = (ψ, φ )H = 0,
so that
(ψ, θ ) ∈ [Γ(A∗ )]⊥ .
Therefore, V [Γ(A∗ )]⊥ is not the graph of a linear operator. Since Γ(A) = V [Γ(A∗ )]⊥ A is not closable.
3. Observe that if A is closable then
A∗ = (A∗ ) = A∗∗∗ = (A)∗ .
6.7 Symmetric and self-adjoint operators

Definition 6.7.1 Let A : D → H be a linear operator, where D is dense in H. A is said to be symmetric if
A ⊂ A∗ , that is if D ⊂ D(A∗ ) and
A∗ φ = Aφ , ∀φ ∈ D.
Equivalently, A is symmetric if and only if
(Aφ , ψ)H = (φ , Aψ)H , ∀φ , ψ ∈ D.
Definition 6.7.2 Let A : D → H be a linear operator. We say that A is self-adjoint if A = A∗ , that is if A is

symmetric and D = D(A∗ ).
Definition 6.7.3 Let A : D → H be a symmetric operator. We say that A is essentially self-adjoint if its closure
A is self-adjoint. If A is closed, a subset E ⊂ D is said to be a core for A if A|E = A.
Theorem 6.7.4 Let A : D → H be a symmetric operator. Then the following statements are equivalent
1. A is self-adjoint.
2. A is closed and N(A∗ ± iI) = {θ }.
3. R(A ± iI) = H.
Proof 6.22
1 implies 2:
Suppose A is self-adjoint let φ ∈ D = D(A∗ ) be such that
Aφ = iφ
so that
A∗ φ = iφ .
Observe that
i(φ , φ )H = (iφ , φ )H
= (Aφ , φ )H
= (φ , Aφ )H
= (φ , iφ )H
= −i(φ , φ )H , (6.43)
so that (φ , φ )H = 0, that is φ = θ . Thus,
N(A − iI) = {θ }.
Similarly, we prove that N(A + iI) = {θ }. Finally, since A∗ = A∗ = A, we get that A = A∗ is closed.
2 implies 3:
Suppose 2 holds. Thus the equation
A∗ φ = −iφ
has no non trivial solution. We will prove that R(A − iI) is dense in H. If ψ ∈ R(A − iI)⊥ then
((A − iI)φ , ψ)H = 0, ∀φ ∈ D,
so that ψ ∈ D(A∗ ) and

(A − iI)∗ ψ = (A∗ + iI)ψ = θ ,
and hence by above ψ = θ . Now we will prove that R(A − iI) is closed and conclude that
R(A − iI) = H.
Given φ ∈ D we have
k(A − iI)φ k2H = kAφ kH
2 2
+ kφ kH . (6.44)
Let ψ0 ∈ H be a limit point of R(A − iI). Thus we may find {φn } ⊂ D such that
(A − iI)φn → ψ0 .
From (6.44)
kφn − φm kH ≤ k(A − iI)(φn − φm )kH , ∀m, n ∈ N
so that {φn } is a Cauchy sequence, therefore converging to some φ0 ∈ H. Also from (6.44)
kAφn − Aφm kH ≤ k(A − iI)(φn − φm )kH , ∀m, n ∈ N
so that {Aφn } is a Cauchy sequence, hence also a converging one. Since A is closed, we get φ0 ∈ D
and
(A − iI)φ0 = ψ0 .
Therefore, R(A − iI) is closed, so that
R(A − iI) = H.
Similarly,
R(A + iI) = H.
3 implies 1: Let φ ∈ D(A∗ ). Since, R(A − iI) = H, there is an η ∈ D such that
(A − iI)η = (A∗ − iI)φ ,
and since D ⊂ D(A∗ ) we obtain φ − η ∈ D(A∗ ), and
(A∗ − iI)(φ − η) = θ .
Since R(A + iI) = H we have N(A∗ − iI) = {θ }. Therefore, φ = η, so that D(A∗ ) = D. The proof is
complete.
6.7.1 The spectral theorem using Cayley transform

In this section H is a complex Hilbert space. We suppose A is defined on a dense subspace of H, being A
self-adjoint but possibly unbounded. We have shown that (A + i) and (A − i) are onto H and it is possible to
prove that
U = (A − i)(A + i)−1 ,
exists on all H and it is unitary. Furthermore on the domain of A,
A = i(I +U )(I −U)−1 .
The operator U is called the Cayley transform of A. We have already proven that
Z 2π
U= eiφ dF(φ ),
0
where {F(φ )} is a monotone family of orthogonal projections, strongly continuous from the right and we
may consider it such that

0, if φ ≤ 0,
F(φ ) = (6.45)
I, if φ ≥ 2π.
Since F(φ ) = 0, for all φ ≤ 0 and

F(0) = F(0+ )
we obtain
F(0+ ) = 0 = F(0− ),
that is, F(φ ) is continuous at φ = 0. We claim that F is continuous at φ = 2π. Observe that F(2π) = F(2π + )
so that we need only to show that
F(2π − ) = F(2π).
Suppose
F(2π) − F(2π − ) =
6 θ.
Thus there exists some u, v ∈ H such that
(F(2π) − F(2(π − )))u = v =

6 θ.
Therefore,
F(φ )v = F(φ )[(F(2π) − F(2π − ))u],
so that

0, if φ < 2π,
F(φ )v = (6.46)
v, if φ ≥ 2π.
Observe that Z 2π
U −I = (eiφ − 1)dF(φ ),
0
and Z 2π
∗
U −I = (e−iφ − 1)dF(φ ).
0
Let {φn } be a partition of [0, 2π]. From the monotonicity of [0, 2π] and pairwise orthogonality of
{F(φn ) − F(φn−1 )}
we can show that (this is not proved in details here)

Z 2π
∗
(U − I)(U − I) = (e−iφ − 1)(eiφ − 1)dF(φ ),
0
so that, given z ∈ H we have

Z 2π
((U ∗ − I)(U − I)z, z)H = |eiφ − 1|2 (dF(φ )z, z)H ,
0
thus, for v defined above
k(U − I )vk2 = ((U − I)v, (U − I)v)H

= ((U − I)∗ (U − I)v, v)H
Z 2π
= |eiφ − 1|2 (dF(φ )v, v)H
0
Z 2π −
= |eiφ − 1|2 (dF(φ )v, v)H
0
= 0 (6.47)
The last two equalities results from e2πi − 1 = 0 and dF(φ )v = θ on [0, 2π). Since v 6= θ , the last equation
implies that 1 ∈ Pσ (U), which contradicts the existence of
(I −U)−1 .
Thus, F is continuous at φ = 2π.

Now choose a sequence of real numbers {φn } such that φn ∈ (0, 2π), n = 0, ±1, ±2, ±3, ... such that

φn
−cot = n.
2
Now define Tn = F(φn ) − F(φn−1 ). Since U commutes with F(φ ), U commutes with Tn . since
A = i(I +U )(I −U)−1 ,
this implies that the range of Tn is invariant under U and A. Observe that
∑ Tn = ∑(F(φn ) − F(φn−1 ))
n n
= lim F(φ ) − lim F(φ )
φ →2π φ →0
= I − θ = I. (6.48)
Hence,
∑ R(Tn ) = H.
n
Also, for u ∈ H we have that


 0, if φ < φn−1 ,
F(φ )Tn u = (F(φ ) − F(φn−1 ))u, if φn−1 ≤ φ ≤ φn , (6.49)
F(φn ) − F(φn−1 ))u, if φ > φn ,

so that
Z 2π
(I −U)Tn u = (1 − eiφ )dF(φ )Tn u
0
Z φn
= (1 − eiφ )dF(φ )u. (6.50)
φn−1
Therefore,
Z φn
(1 − eiφ )−1 dF(φ )(I −U)Tn u
φn−1
Z φn Z φn
= (1 − eiφ )−1 dF(φ ) (1 − eiφ )dF(φ )u
φn−1 φn−1
Z φn
= (1 − eiφ )−1 (1 − eiφ )dF(φ )u
φn−1
Z φn
= dF(φ )u
φn−1
Z 2π
= dF(φ )Tn u = Tn u. (6.51)
0
Hence,
−1 Z φn
(1 − eiφ )−1 dF(φ ).

(I −U)|R(Tn ) =
φn−1
From this, from above and as

A = i(I +U)(I −U)−1
we obtain Z φn
ATn u = i(1 + eiφ )(1 − eiφ )−1 dF(φ )u.
φn−1
Therefore, defining
φ
λ = −cot ,
2
and
E(λ ) = F(−2cot −1 λ ),
we get
iφ iφ −1 φ
i(1 + e )(1 − e ) = −cot = λ.
2
Hence, Z n
ATn u = λ dE(λ )u.
n−1
Finally, from
∞
u= ∑ Tn u,
n=−∞
we can obtain
!
∞
Au = A ∑ Tn u
n=−∞
∞
= ∑ ATn u
n=−∞
∞ Z n
= ∑ λ dE(λ )u. (6.52)
n=−∞ n−1
Being the convergence in question in norm, we may write

Z ∞
Au = λ dE(λ )u.
−∞
Since u ∈ H is arbitrary, we may denote

Z ∞
A= λ dE(λ ). (6.53)
−∞
Chapter 7
Basic Results on Measure and

Integration
The main references for this chapter are Rudin [68], Royden [69] and Stein and Shakarchi [72], where more
details may be found. All these three books are excellent and we strongly recommend their reading.
7.1 Basic concepts

In this chapter U denotes a topological space.
Definition 7.1.1 (σ -Algebra) A collection M of subsets of U is said to be a σ -Algebra if M has the fol-
lowing properties:
1. U ∈ M ,
2. if A ∈ M then U \ A ∈ M ,
3. if An ∈ M , ∀n ∈ N, then ∪∞
n=0 An ∈ M .
Definition 7.1.2 (Measurable spaces) If M is a σ -algebra in U we say that U is a measurable space. The
elements of M are called the measurable sets of U.
Definition 7.1.3 (Measurable function) If U is a measurable space and V is a topological space, we say
that f : U → V is a measurable function if f −1 (V ) is measurable whenever V ⊂ V is an open set.
Remark 7.1.4 1. Observe that 0/ = U \U so that from 1 and 2 in Definition 7.1.1, we have that 0/ ∈ M .
2. From 1 and 3 from Definition 7.1.1, it is clear that ∪ni=1 Ai ∈ M whenever Ai ∈ M , ∀i ∈ {1, ..., n}.
∞ A = (∪∞ Ac )c also from Definition 7.1.1, it is clear that M is closed under countable
3. Since ∩i= 1 i i=1 i
intersections.
4. Since A \ B = Bc ∩ A we obtain: if A, B ∈ M then A \ B ∈ M .
Theorem 7.1.5 Let F be any collection of subsets of U. Then there exists a smallest σ -algebra M0 in U
such that F ⊂ M0 .
Basic Results on Measure and Integration 137
Proof 7.1 Let Ω be the family of all σ -Algebras that contain F . Since the set of all subsets in U is a
σ -algebra, Ω is non-empty.
Let M0 = ∩Mλ ∈Ω Mλ , it is clear that M0 ⊃ F , it remains to prove that in fact M0 is a σ -algebra. Observe
that:
1. U ∈ Mλ , ∀Mλ ∈ Ω, so that, U ∈ M0 ,
2. A ∈ M0 implies A ∈ Mλ , ∀Mλ ∈ Ω, so that Ac ∈ Mλ , ∀Mλ ∈ Ω, which means Ac ∈ M0 ,
3. {An } ⊂ M0 implies {An } ⊂ Mλ , ∀Mλ ∈ Ω, so that ∪∞
n=1 An ∈ Mλ , ∀Mλ ∈ Ω, which means ∪n=1 An ∈
∞
M0 .
From Definition 7.1.1 the proof is complete.
Definition 7.1.6 (Borel sets) Let U be a topological space, considering the last theorem there exists a small-
est σ -algebra in U, denoted by B, which contains the open sets of U. The elements of B are called the Borel
sets.
Theorem 7.1.7 Suppose M is a σ -algebra in U and V is a topological space. For f : U → V , we have:
1. If Ω = {E ⊂ V | f −1 (E) ∈ M }, then Ω is a σ -algebra.
2. If V = [−∞, ∞], and f −1 ((α, ∞]) ∈ M , for each α ∈ R, then f is measurable.
Proof 7.2
1. (a) V ∈ Ω since f −1 (V ) = U and U ∈ M .
(b) E ∈ Ω ⇒ f −1 (E) ∈ M ⇒ U \ f −1 (E) ∈ M ⇒ f −1 (V \ E) ∈ M ⇒ V \ E ∈ Ω.
(c) {Ei } ⊂ Ω ⇒ f −1 (Ei ) ∈ M , ∀i ∈ N ⇒ ∪∞
i=1 f
−1 (E ) ∈ M ⇒ f −1 (∪∞ E ) ∈ M ⇒ ∪∞ E ∈ Ω.
i i=1 i i=1 i
Thus Ω is a σ -algebra.
2. Define Ω = {E ⊂ [−∞, ∞] | f −1 (E) ∈ M } from above Ω is a σ - algebra. Given α ∈ R, let {αn } be a
real sequence such that αn → α as n → ∞, αn < α, ∀n ∈ N. Since (αn , ∞] ∈ Ω for each n and
C
[−∞, α ) = ∪∞ ∞
n=1 [−∞, αn ] = ∪n=1 (αn , ∞] , (7.1)
we obtain, [−∞, α) ∈ Ω. Furthermore, we have (α, β ) = [−∞, β ) ∩ (α, ∞] ∈ Ω. Since every open set
in [−∞, ∞] may be expressed as a countable union of intervals (α, β ) we have that Ω contains all the
open sets. Thus, f −1 (E) ∈ M whenever E is open, so that f is measurable.
Proposition 7.1.8 If { fn : U → [−∞, ∞]} is a sequence of measurable functions and g = supn≥1 fn and h =
lim sup fn then g and h are measurable.
n→∞
Proof 7.3 Observe that g−1 ((α, ∞]) = ∪∞ −1

n=1 f n ((α, ∞]). From last theorem g is measurable. By analogy
h = infk≥1 {supi≥k fi } is measurable.
7.2 Simple functions

Definition 7.2.1 (Simple functions) A function f : U → C is said to be a simple function if its range (R( f ))
has only finitely many points. If {α1 , ..., αn } = R( f ) and we set Ai = {u ∈ U | f (u) = αi }, clearly we have:
f = ∑ni=1 αi χAi , where

1, if u ∈ Ai ,
χAi (u) = (7.2)
0, otherwise.
Theorem 7.2.2 Let f : U → [0, ∞] be a measurable function. Thus there exists a sequence of simple functions
{sn : U → [0, ∞]} such that
1. 0 ≤ s1 ≤ s2 ≤ ... ≤ f ,
2. sn (u) → f (u) as n → ∞, ∀u ∈ U.
Proof 7.4 Define δn = 2−n . To each n ∈ N and each t ∈ R+ , there corresponds a unique integer K = Kn (t)
such that
Kδn ≤ t ≤ (K + 1)δn . (7.3)
Defining

Kn (t)δn , i f 0 ≤ t < n,
ϕn (t ) = (7.4)
n, i f t ≥ n,
we have that each ϕn is a Borel function on [0, ∞], such that

1. t − δn < ϕn (t) ≤ t if 0 ≤ t ≤ n,
2. 0 ≤ ϕ1 ≤ ... ≤ t,
3. ϕn (t) → t as n → ∞, ∀t ∈ [0, ∞].
It follows that the sequence {sn = ϕn ◦ f } corresponds to the results indicated above.
7.3 Measures
Definition 7.3.1 (Measure) Let M be a σ -algebra on a topological space U. A function µ : M → [0, ∞] is
said to be a measure if µ(0/) = 0 and µ is countably additive, that is, given {Ai } ⊂ U, a sequence of pairwise
disjoint sets then
∞
µ(∪∞
i=1 Ai ) = ∑ µ(Ai ). (7.5)
i=1
In this case (U, M , µ) is called a measure space.
Proposition 7.3.2 Let µ : M → [0, ∞], where M is a σ -algebra of U. Then we have the following.
1. µ(A1 ∪ ... ∪ An ) = µ(A1 ) + ... + µ(An ) for any given {Ai } of pairwise disjoint measurable sets of M .
2. If A, B ∈ M and A ⊂ B then µ(A) ≤ µ(B).
3. If {An } ⊂ M , A = ∪∞
n=1 An and
A1 ⊂ A2 ⊂ A3 ⊂ ... (7.6)
then, lim µ(An ) = µ(A).

n→∞
4. If {An } ⊂ M , A = ∩∞
n=1 An , A1 ⊃ A2 ⊃ A3 ⊃ .... and µ(A1 ) is finite then,
lim µ(An ) = µ(A). (7.7)

n→∞
Proof 7.5
1. Take An+1 = An+2 = .... = 0/ in Definition 7.1.1 item 1,
2. Observe that B = A∪(B−A) and A∩(B−A) = 0/ so that by above µ(A∪(B−A)) = µ(A)+ µ(B−A) ≥
µ(A),
3. Let B1 = A1 and let Bn = An −An−1 then Bn ∈ M , Bi ∩B j = 0/ if i 6= j, An = B1 ∪ ...∪Bn and A = ∪i=1

∞ B.
i
Thus
∞ n
µ (A) = µ(∪∞
i=1 Bi ) = ∑ µ(Bi ) = n→∞
lim ∑ µ(Bi ) = lim µ(An ). (7.8)
n→∞
n=1 i=1
4. Let Cn = A1 \ An . Then C1 ⊂ C2 ⊂ ..., µ (Cn ) = µ(A1 ) − µ(An ), A1 \ A = ∪∞

n=1Cn .
Thus, by 3 we have
µ(A1 ) − µ(A) = µ(A1 \ A) = lim µ(Cn ) = µ(A1 ) − lim µ(An ). (7.9)

n→∞ n→∞
7.4 Integration of simple functions

Definition 7.4.1 (Integral for simple functions) For s : U → [0, ∞], a measurable simple function, that is,
n
s = ∑ αi χAi , (7.10)
i=1
where

1, if u ∈ Ai ,
χAi (u) = (7.11)
0, otherwise,
we define the integral of s over E ⊂ M , denoted by

R
Es dµ as
Z n
s dµ = ∑ αi µ(Ai ∩ E). (7.12)
E i=1
The convention 0.∞ = 0 is used here.
Definition 7.4.2 (Integral for non-negative measurable functions) If f : U → [0, ∞] is measurable, for
E ∈ M , we define the integral of f on E, denoted by E f dµ, as
R
Z Z
f dµ = sup sdµ , (7.13)
E s∈A E
where
A = {s simple and measurable | 0 ≤ s ≤ f }. (7.14)
Definition 7.4.3 (Integrals for measurable functions) For a measurable f : U →R [−∞, ∞] and E ∈ M , we
define f + = max{ f , 0}, f − = max{− f , 0} and the integral of f on E, denoted by E f dµ, as
Z Z Z
f dµ = f + dµ − f − dµ .
E E E
Theorem 7.4.4 (Lebesgue’s monotone convergence theorem) Let { fn } be a sequence of real measurable
functions on U and suppose that
1. 0 ≤ f1 (u) ≤ f2 (u) ≤ ... ≤ ∞, ∀u ∈ U,
2. fn (u) → f (u) as n → ∞, ∀u ∈ U.
Then,
(a) f is measurable,
R R
(b) U f n dµ → U f dµ as n → ∞.
R R
Proof 7.6 Since U f n dµ ≤ U f n+1 dµ, ∀n ∈ N, there exists α ∈ [0, ∞] such that
Z
fn dµ → α, as n → ∞, (7.15)
U
By Proposition 7.1.8, f is measurable, and since fn ≤ f we have

Z Z
fn dµ ≤ f dµ. (7.16)
U U

Z
α≤ f dµ. (7.17)
U
Let s be any simple function such that 0 ≤ s ≤ f , and let c ∈ R such that 0 < c < 1. For each n ∈ N we define
En = {u ∈ U | fn (u) ≥ cs(u)}. (7.18)
Clearly En is measurable and E1 ⊂ E2 ⊂ ... and U = ∪n∈N En . Observe that
Z Z Z
fn dµ ≥ fn dµ ≥ c sdµ. (7.19)
U En En
Letting n → ∞ and applying Proposition 7.3.2, we obtain

Z Z
α = lim fn dµ ≥ c sdµ, (7.20)
n→∞ U U
so that
Z
α≥ sdµ, ∀s simple and measurable such that 0 ≤ s ≤ f . (7.21)
U
This implies
Z
α≥ f dµ. (7.22)
U
From (7.17) and (7.22) the proof is complete.
The next result we do not prove it (it is a direct consequence of last theorem). For a proof see [68].
Corollary 7.4.5 Let { fn } be a sequence of non-negative measurable functions defined on U ( fn : U →
[0, ∞], ∀n ∈ N). Defining f (u) = ∑∞
n=1 f n (u), ∀u ∈ U, we have
Z ∞ Z
f dµ = ∑ fn dµ .
U n=1 U
Theorem 7.4.6 (Fatou’s lemma) If { fn : U → [0, ∞]} is a sequence of measurable functions, then
Z Z
lim inf fn dµ ≤ lim inf fn dµ. (7.23)
U n→∞ n→∞ U
Proof 7.7 For each k ∈ N define gk : U → [0, ∞] by
gk (u) = inf{ fi (u)}. (7.24)

i≥k
Then
gk ≤ fk (7.25)
so that
Z Z
gk dµ ≤ fk dµ, ∀k ∈ N. (7.26)
U U
Also 0 ≤ g1 ≤ g2 ≤ ..., each gk is measurable, and
lim gk (u) = lim inf fn (u), ∀u ∈ U. (7.27)

k→∞ n→∞
From the Lebesgue monotone convergence theorem

Z Z Z
lim inf gk dµ = lim gk dµ = lim inf fn dµ. (7.28)
k→∞ U k→∞ U U n→∞
From (7.26) we have

Z Z
lim inf gk dµ ≤ lim inf{ fk dµ }. (7.29)
k→∞ U k→∞ U
Thus, from (7.28) and (7.29) we obtain

Z Z
lim inf fn dµ ≤ lim inf fn d µ. (7.30)
U n→∞ n→∞ U
Theorem 7.4.7 (Lebesgue’s dominated convergence theorem) Suppose { fn } is sequence of complex mea-
surable functions on U such that
lim fn (u) = f (u), ∀u ∈ U. (7.31)

n→∞
If there exists a measurable function g : U → R+ such that

R
U g dµ < ∞ and | fn (u)| ≤ g(u), ∀u ∈ U, n ∈ N,
then
R
1. U | f | dµ < ∞,
R
2. lim | fn − f | dµ = 0.
n→∞ U
Proof 7.8
1. This inequality holds since f is measurable and | f | ≤ g.
2. Since 2g − | fn − f | ≥ 0 we may apply the Fatou’s lemma and obtain:

Z Z
2gdµ ≤ lim inf (2g − | fn − f |)dµ, (7.32)
U n→∞ U
so that
Z
lim sup | fn − f | d µ ≤ 0. (7.33)
n→∞ U
Hence,
Z
lim | fn − f | dµ = 0. (7.34)
n→∞ U
We finish this section with an important remark:

Remark 7.4.8 In a measurable space U we say that a property holds almost everywhere (a.e.) if it holds
on U except for a set of measure zero. Finally, since integrals are not changed by the redefinition of the
functions in question on sets of zero measure, the proprieties of items (1) and (2) of the Lebesgue’s monotone
convergence may be considered a.e. in U, instead of in all U. Similar remarks are valid for the Fatou’s lemma
and the Lebesgue’s dominated Convergence theorem.
7.5 Signed measures

In this section we study signed measures. We start with the following definition:
Definition 7.5.1 Let (U, M ) be a measurable space. We say that a measure µ is finite if µ(U) < ∞. On
the other hand, we say that µ is σ -finite if there exists a sequence {Un } ⊂ U such that U = ∪∞
n=1Un and
µ(Un ) < ∞, ∀n ∈ N.
Definition 7.5.2 (Signed measure) Let (U, M ) be a measurable space. We say that ν : M → [−∞, +∞] is
a signed measure if
ν may assume at most one the values −∞, +∞.
ν(0/) = 0.
ν (∑∞ ∞
n=1 En ) = ∑n=1 ν(En ) for all sequence of measurable disjoint sets {En }
We say that A ∈ M is a positive set with respect to ν if A is measurable and ν(E) ≥ 0 for all E measurable
such that E ⊂ A.
Similarly, We say that B ∈ M is a negative set with respect to ν if B is measurable and ν(E) ≤ 0 for all
E measurable such that E ⊂ B.
Finally, if A ∈ M is both positive and negative with respect to ν, it is said to be a null set.
Lemma 7.5.3 Considering the last definitions, we have that a countable union of positive measurable sets
is positive.
Proof 7.9 n=1 An where An is positive, ∀n ∈ N. Choose a measurable set E ⊂ A. Set

Let A = ∪∞
En = (E ∩ An ) \ (∪ni=−11 Ai ).
Thus, En is a measurable subset of An so that ν(En ) ≥ 0. Observe that

E = ∪∞
n=1 En
where {En } is a sequence of measurable disjoint sets.

Therefore ν(E) = ∑∞ n=1 ν(En ) ≥ 0.
Since E ⊂ A is arbitrary, A is positive.
Lemma 7.5.4 Considering the last definitions, let E be a measurable set such that
0 < ν(E) < ∞.
Then there exists a positive set A ⊂ E such that ν(A) > 0.
Proof 7.10 Observe that if E is not positive then it contains a set of negative measure. In such a case, let
n1 be the smallest positive integer such that there exists a measurable set E1 ⊂ E such that
ν(E1 ) < −1/n1 .

Reasoning inductively, if E \ ∪k−1
j=1 E j is not positive, let nk be the smallest positive integer such that
there exists a measurable set
Ek ⊂ E \ ∪k−1
j=1 E j
such that
ν(Ek ) < −1/nk .
Define
A = E \ (∪∞
k=1 Ek ) .
Then
E = A ∪ (∪∞
k=1 Ek ) .
Since such a union is disjoint, we have,
∞
ν(E) = ν(A) + ∑ ν(Ek ),
k=1
so that, since ν(E) < ∞, this last series is convergent.

Also, since
1/nk < −ν(Ek ),
we have that
∞
∑ 1/nk
k=1
is convergent so that nk → ∞ as k → ∞.
From ν(E) > 0 we must have ν(A) > 0.
Now, we will show that A is positive. Let ε > 0. Choose k sufficiently big such that 1/(nk − 1) < ε.
Since
A ⊂ E \ ∪kj=1 E j ,
A contains no measurable set with measure less than
−1/(nk − 1) > −ε,
that is, A contains no measurable set with measure less than −ε.
Since ε > 0 is arbitrary, A contains no measurable negative set. Thus, A is positive.
Proposition 7.5.5 (The Hahn decomposition) Let ν be a signed measure on a measurable space (U, M ).
Then there exist a positive set A and a negative set B such that U = A ∪ B and A ∩ B = 0/.
Proof 7.11 Without loosing generality, suppose ν does not assume the value +∞ (the other case may be
dealt similarly). Define
λ = sup{ν(A) | A is positive}.
Since the empty set 0/ is positive, we obtain λ ≥ 0.
Let {An } be a sequence of positive sets such that
lim ν(An ) = λ .
n→∞
Define
A = ∪∞
i=1 Ai .
From Lemma 7.5.3, A is a positive set, so that
λ ≥ ν(A).
On the other hand,

A \ An ⊂ A
so that
ν(A − An ) ≥ 0, ∀n ∈ N.
Therefore
ν(A) = ν(An ) + ν(A \ An ) ≥ ν(An ), ∀n ∈ N.
Hence
ν(A) ≥ λ ,
so that λ = ν(A).
Let B = U \ A. Suppose E ⊂ B, so that E is positive. Hence,
λ ≥ ν(E ∪ A)
= ν(E) + ν(A)
= ν(E) + λ , (7.35)
so that ν(E) = 0.
Thus, B contains no positive set of positive measure, so that, by Lemma 7.5.4, B contains no subsets of
positive measure, that is, B is negative.
Remark 7.5.6 Denoting the Hahn decomposition of U relating ν by {A, B}, we may define the measures ν +
and ν − by
ν + (E) = ν(E ∩ A),
and-
ν − (E) = −ν(E ∩ B),
so that
ν = ν + − ν −.
We recall that two measures ν1 and ν2 are mutually singular if there are disjoint measurable sets such
that
U = A∪B
and
ν1 (A) = ν2 (B) = 0.
Observe that the measures ν + and ν − above defined are mutually singular. The decomposition
ν = ν+ − ν−
is called the Jordan one of ν. The measures ν + and ν − are called the positive and negative parts of ν
respectively.
Observe that either ν + or ν − is finite since only one of the values +∞, −∞ may be assumed by ν. We
may also define
|ν|(E) = ν + (E) + ν − (E),
which is called the absolute value or total variation of ν.
7.6 The Radon-Nikodym theorem

We start this section with the definition of absolutely continuous measures.
Definition 7.6.1 (Absolutely continuous measures) We say that a measure ν is absolutely continuous with
respect to a measure µ and write ν µ, if ν(A) = 0 for all set such that µ(A) = 0. In case of a signed
measure we write ν µ if |ν| |µ|.
Theorem 7.6.2 (The Radon-Nikodym theorem) Let (U, M , µ) be a σ -finite measure space. Let ν be a
measure defined on M which is absolutely continuous with respect to µ, that is, ν µ.
Then there exists a non-negative measurable function f such that
Z
ν(E) = f dµ, ∀E ∈ M .
E
The function f is unique up to usual representatives.
Proof 7.12 First assume ν and µ are finite.

Define λ = ν + µ. Also define the functional F by
Z
F( f ) = f dµ.
U
We recall that f ∈ L2 (µ) if f is measurable and

Z
| f |2 dµ < ∞.
U
The space L2 (µ) is a Hilbert one with inner product

Z
( f , g)L2 (µ) = f g dµ.
U
Observe that, from the Cauchy-Schwartz inequality, we may write,
|F( f )| = |( f , 1)L2 (µ) |

≤ k f kL2 (µ) [µ (U )]1/2
≤ k f kL2 (λ ) [µ(U )]1/2 , (7.36)
since Z Z
k f k2L2 (µ ) = | f |2 dµ ≤ | f |2 dλ = k f k2L2 (λ ) .
U U
Thus F is a bounded linear functional on L2 (λ ), where f ∈ L2 (λ ), if f is measurable and

Z
| f |2 dλ < ∞.
U
Since L2 (λ ) is also a Hilbert space with the inner product

Z
( f , g)L2 (λ ) = f g dλ ,
U
from the Riesz representation theorem, there exists g ∈ L2 (λ ), such that

Z
F( f ) = f g dλ .
U
Thus, Z Z
f dµ = f g dλ ,
U U
and in particular, Z Z
f dµ = f g (dµ + dν).
U U
Hence, Z Z
f (1 − g) dµ = f g dν. (7.37)
U U
Assume, to obtain contradiction, that g < 0 in a set A such that µ(A) > 0.
Thus, Z
(1 − g) dµ > 0,
A
so that from this and (7.37) with f = χA we get
Z
g dν > 0.
A
Since g < 0 on A we have a contradiction. Thus g ≥ 0, a.e. [µ] on U.

Now, assume, also to obtain contradiction that g > 1 on set B such that µ(B) > 0.
Thus Z
(1 − g) dµ ≤ 0,
B
so that from this and (7.37) with f = χB we obtain
Z
ν(B) ≤ g dν ≤ 0
B
and hence
ν(B) = 0.
R
Thus, Bg dν = 0 so that Z
(1 − g) dµ = 0,
B
which implies that µ(B) = 0, a contradiction.
From above we conclude that

0 ≤ g ≤ 1, a.e. [µ] in U.
On the other hand, for a fixed E µ-measurable again from (7.37) with f = χE we get
Z Z
(1 − g) dµ = g dν,
E E
so that Z Z Z
(1 − g) dµ = g dν − dν + ν(E),
E E E
and therefore Z
ν(E) = (1 − g) (dµ + dν),
E
that is, Z
ν(E) = (1 − g) dλ , ∀E ∈ M
E
Define
B = {u ∈ U : g(u) = 0}.
R
Hence, µ(B) = B g dλ = 0.
From this, since λ µ, we obtain
g−1 g = 1, a.e. [λ ].
Therefore, for a not relabeled E ∈ M we have
Z Z
λ (E) = g−1 g dλ = g−1 dµ.
E E
Finally, observe that
µ(E) + ν(E) = λ (E)

Z
= dλ
ZE
= g−1 dµ. (7.38)
E
Thus,
Z
ν(E) = g−1 dµ − µ(E)
E
Z
= (g−1 − 1) dµ
E
Z
= (1 − g)g−1 dµ, ∀E ∈ M . (7.39)
E
The proof for the finite case in complete. The proof for σ -finite is developed in the next lines.
Since U is σ -finite, there exists a sequence {Un } such that U = ∪∞ n=1Un , and µ(Un ) < ∞ and ν(Un ) <
∞, ∀n ∈ N.
Define
Fn = Un \ ∪n−1
j=1 U j ,
thus U = ∪∞n=1 Fn and {Fn } is a sequence of disjoint sets, such that µ(Fn ) < ∞ and ν(Fn ) < ∞, ∀n ∈ N.
Let E ∈ M . For each n ∈ N from above we may obtain fn such that
Z
ν(E ∩ Fn ) = fn dµ, ∀E ∈ M .
E∩Fn
From this and the monotone convergence theorem corollary we may write
∞
ν(E) = ∑ ν(E ∩ Fn )
n=1
∞ Z
= ∑ fn dµ
n=1 E ∩Fn
∞ Z
= ∑ fn χFn dµ
n=1 E
Z ∞
= ∑ fn χFn dµ
E n=1
Z
= f dµ, (7.40)
E
where
∞
f= ∑ fn χFn .
n=1
Theorem 7.6.3 (The Lebesgue decomposition) Let (U, M , µ) be a σ -finite measure space and let ν be a
σ -finite measure defined on M .
Then we may find a measure ν0 , singular with respect to µ, and a measure ν1 , absolutely continuous with
respect to µ, such that
ν = ν0 + ν1 .
Furthermore, the measures ν0 and ν1 are unique.
Proof 7.13 Since µ and ν are σ -finite measures, so is
λ = ν + µ.
Observe that ν and µ are absolutely continuous with respect to λ . Hence, by the Radon-Nikodym theorem,
there exist non-negative measurable functions f and g such that
Z
µ(E) = f dλ , ∀E ∈ M
E
and Z
ν(E) = g dλ ∀E ∈ M .
E
Define
A = {u ∈ U | f (u) > 0},
and
B = {u ∈ U | f (u) = 0}.
Thus,
U = A ∪ B,
and
A ∩ B = 0/.
Also define
ν0 (E) = ν(E ∩ B), ∀E ∈ M .
We have that ν0 (A) = 0 so that

ν0 ⊥ µ.
Define
ν1 (E) = ν(E ∩ A)
Z
= g dλ . (7.41)
E∩A
Therefore,
ν = ν0 + ν1 .
To finish the proof, we have only to show that
ν1 µ.
Let E ∈ M such that µ(E) = 0. Thus

Z
0 = µ(E) = f dλ ,
E
and in particular Z
f dλ = 0.
E∩A
Since f > 0 on A ∩ E we conclude that
λ (A ∩ E) = 0.
Therefore, since ν λ , we obtain
ν(E ∩ A) = 0,
so that
ν1 (E) = ν(E ∩ A) = 0.
ν1 µ.
The proof of uniqueness is left to the reader.
7.7 Outer measure and measurability

Let U be a set. Denote by P the set of all subsets of U. An outer measure µ ∗ : P → [0, +∞] is a set function
such that
1. µ ∗ (0/) = 0,
2. if A ⊂ B then µ ∗ (A) ≤ µ ∗ (B), ∀A, B ⊂ U,
3. if E ⊂ ∪∞
n=1 En then
∞
µ ∗ (E) ≤ ∑ µ ∗ (En ).
n=1
The outer measure is called finite if µ ∗ (U) < ∞.
Definition 7.7.1 (measurable set) A set E ⊂ U is said to be measurable with respect to µ ∗ if
µ ∗ (A) = µ ∗ (A ∩ E) + µ ∗ (A ∩ E c ), ∀A ⊂ U.
Theorem 7.7.2 The set B of µ ∗ -measurable sets is a σ -algebra. If µ is defined to be µ ∗ restricted to B,

then µ is a complete measure on B.
Proof 7.14 Let E = 0/ and let A ⊂ U.

Thus,
µ ∗ (A) = µ ∗ (A ∩ 0/) + µ ∗ (A ∩ 0/c ) = µ ∗ (A ∩U) = µ ∗ (A).
Therefore 0/ is µ ∗ -measurable.
Let E1 , E2 ∈ U be µ ∗ -measurable sets. Let A ⊂ U. Thus,
µ ∗ (A) = µ ∗ (A ∩ E2 ) + µ ∗ (A ∩ E2c ),
so that from the measurability of E1 we get
µ ∗ (A) = µ ∗ (A ∩ E2 ) + µ ∗ (A ∩ E2c ∩ E1 ) + µ ∗ (A ∩ E2c ∩ E1c ). (7.42)
Since,
A ∩ (E1 ∪ E2 ) = (A ∩ E2 ) ∪ (A ∩ E1 ∩ E2c ),
we obtain
µ ∗ (A ∩ (E1 ∪ E2 )) ≤ µ ∗ (A ∩ E2 ) + µ ∗ (A ∩ E2c ∩ E1 ). (7.43)
µ ∗ (A) ≥ µ ∗ (A ∩ (E1 ∪ E2 )) + µ ∗ (A ∩ E1c ∩ E2c )

= µ ∗ (A ∩ (E1 ∪ E2 )) + µ ∗ (A ∩ (E1 ∪ E2 )c ). (7.44)
Hence E1 ∪ E2 is µ ∗ -measurable.
By induction, the union of a finite number of µ ∗ - measurable sets is µ ∗ -measurable.
Assume E = ∪∞ ∗
i=1 Ei where {Ei } is a sequence of disjoint µ -measurable sets.
n ∗
Define Gn = ∪i=1 Ei . Then Gn is µ -measurable and for a given A ⊂ U we have
µ ∗ (A) = µ ∗ (A ∩ Gn ) + µ ∗ (A ∩ Gcn )
≥ µ ∗ (A ∩ Gn ) + µ ∗ (A ∩ E c ), (7.45)
since E c ⊂ Gcn , ∀n ∈ N.
Observe that
Gn ∩ En = En
and
Gn ∩ Enc = Gn−1 .
Thus, from the measurability of En we may get
µ ∗ (A ∩ Gn ) = µ ∗ (A ∩ Gn ∩ En ) + µ ∗ (A ∩ Gn ∩ Enc )
= µ ∗ (A ∩ En ) + µ ∗ (A ∩ Gn−1 ). (7.46)
By induction we obtain
n
µ ∗ (A ∩ Gn ) = ∑ µ ∗ (A ∩ Ei ),
i=1
so that
n
µ ∗ (A) ≥ µ ∗ (A ∩ E c ) + ∑ µ ∗ (A ∩ Ei ), ∀n ∈ N,
i=1
that is, considering that

A ∩ E ⊂ ∪∞
i=1 (A ∩ Ei ),
we get
∞
µ ∗ (A) ≥ µ ∗ (A ∩ E c ) + ∑ µ ∗ (A ∩ Ei )
i=1
≥ µ ∗ (A ∩ E c ) + µ ∗ (A ∩ E) (7.47)
i=1 Ei is µ -measurable. Therefore B is a σ -algebra.

Since A ⊂ U is arbitrary we may conclude that E = ∪∞ ∗
Finally, we prove that µ is a measure.

Let E1 , E2 ⊂ U be two disjoint µ ∗ -measurable sets.
Thus
µ(E1 ∪ E2 ) = µ ∗ (E1 ∪ E2 )
= µ ∗ ((E1 ∪ E2 ) ∩ E2 ) + µ ∗ ((E1 ∪ E2 ) ∩ E2c )
= µ ∗ (E2 ) + µ ∗ (E1 ). (7.48)
By induction we obtain the finite additivity.
Also, if
E = ∪∞
i=1 Ei
where {Ei } is a sequence of disjoint measurable sets.
Thus,
n
µ (E) ≥ µ(∪ni=1 Ei ) = ∑ µ(Ei ), ∀n ∈ N.
i=1
Therefore,
∞
µ(E) ≥ ∑ µ(Ei ).
i=1
Now observe that
∞ ∞
µ(E) = µ ∗ (∪∞ ∗
i=1 Ei ) ≤ ∑ µ (Ei ) = ∑ µ(Ei ),
i=1 i=1
and thus
∞
µ(E) = ∑ µ(Ei ).
i=1
Definition 7.7.3 A measure on an Algebra A ⊂ U is a set function µ : A → [0, +∞) such that
1. µ(0/) = 0,
2. if {Ei } is a sequence of disjoint sets in A so that E = ∪∞
i=1 Ei ∈ A , then
∞
µ(E) = ∑ µ(Ei ).
i=1
We may define an outer measure in U, by

( )
∞
∗ ∞
µ (E) = inf ∑ µ(Ai ) | E ⊂ ∪i=1 Ai ,
i=1
where Ai ∈ A , ∀i ∈ N.
Proposition 7.7.4 Suppose A ∈ A and {Ai } ⊂ A is such that
A ⊂ ∪∞
i=1 Ai .

∞
µ(A) ≤ ∑ µ(Ai ).
i=1
Proof 7.15 Define

Bn = (A ∩ An ) \ (∪ni=−11 Ai ).
Thus,
Bn ⊂ An , ∀n ∈ N,
Bn ∈ B, ∀n ∈ N, and
A = ∪∞
i=1 Bi .
Moreover, {Bn } is a sequence of disjoint sets, so that

∞ ∞
µ(A) = ∑ µ(Bi ) ≤ ∑ µ(Ai ).
i=1 i=1
Corollary 7.7.5 If A ∈ A then µ ∗ (A) = µ(A).
Theorem 7.7.6 The set function µ ∗ is an outer measure.
Proof 7.16 The only not immediate property to be proven is the countably subadditivity.
Suppose E ⊂ ∪∞ ∗
i=1 Ei . If µ (Ei ) = +∞ for some i ∈ N the result holds.
∗
Thus, assume µ (Ei ) < +∞, ∀i ∈ N.
Let ε > 0. Thus for each i ∈ N there exists {Ai j } ⊂ A such that Ei ⊂ ∪∞j=1 Ai j , and
∞
ε
∑ µ(Ai j ) ≤ µ ∗ (Ei ) + 2i .
j=1
Therefore,
∞ ∞ ∞
µ ∗ (E) ≤ ∑ ∑ µ(Ai j ) ≤ ∑ µ ∗ (Ei ) + ε.
i=1 j=1 i=1
Since ε > 0 is arbitrary, we get

∞
µ ∗ (E) ≤ ∑ µ ∗ (Ei ).
i=1
Proposition 7.7.7 Suppose A ∈ A . Then A is µ ∗ -measurable.
Proof 7.17 Let E ∈ U such that µ ∗ (E) < +∞. Let ε > 0.
Thus, there exists {Ai } ⊂ A such that E ⊂ ∪∞
i=1 Ai and
∞
∑ µ(Ai ) < µ ∗ (E) + ε.
i=1
Observe that
µ(Ai ) = µ(Ai ∩ A) + µ(Ai ∩ Ac ),
so that, from the fact that

E ∩ A ⊂ ∪∞
i=1 (Ai ∩ A),
and
(E ∩ Ac ) ⊂ ∪∞ c
i=1 (Ai ∩ A ),
we obtain
∞ ∞
µ ∗ (E) + ε > ∑ (Ai ∩ A) + ∑ µ(Ai ∩ Ac )
i=1 i=1
≥ µ (E ∩ A) + µ ∗ (E ∩ Ac ).
∗
(7.49)
Since ε > 0 is arbitrary, we get
µ ∗ (E) ≥ µ ∗ (E ∩ A) + µ ∗ (E ∩ Ac ).
Proposition 7.7.8 Suppose µ is a measure on an Algebra A ⊂ U, µ ∗ is the outer measure induced by µ and
E ⊂ U is a set. Then, for each ε > 0, there is a set A ∈ Aσ with E ⊂ A and
µ ∗ (A) ≤ µ ∗ (E) + ε.
Also, there is a set B ∈ Aσ δ such that E ⊂ B and
µ ∗ (E) = µ ∗ (B).
Proof 7.18 Let ε > 0. Thus, there is a sequence {Ai } ⊂ A such that
E ⊂ ∪∞
i=1 Ai
and
∞
∑ µ(Ai ) ≤ µ ∗ (E) + ε.
i=1
Define A = ∪∞
i=1 Ai , then
∞
µ ∗ (A) ≤ ∑ µ ∗ (Ai )
i=1
∞
= ∑ µ(Ai )
i=1
∗
≤ µ (E) + ε. (7.50)
Now, observe that we write A ∈ Aσ if A = ∪∞ where Ai ∈ A , ∀i ∈ N.
i=1 Ai
Also, we write B ∈ Aσ δ if B = ∩∞
n=1 A n , where ∈ Aσ , ∀n ∈ N.
An
From above, for each n ∈ N there exists An ∈ Aσ such that
E ⊂ An
and
µ ∗ (An ) ≤ µ(E) + 1/n.
n=1 An . Thus B ∈ Aσ δ , E ⊂ B and
Define B = ∩∞
µ ∗ (B) ≤ µ ∗ (An ) ≤ µ ∗ (E) + 1/n, ∀n ∈ N.
Hence,
µ ∗ (B) = µ ∗ (E).
Proposition 7.7.9 Suppose µ is a σ -finite measure on a σ -algebra A , and let µ ∗ be the outer measure
induced by µ.
Under such hypotheses, a set E is µ ∗ measurable if and only if E = A \ B where A ∈ Aσ δ , B ⊂ A,
µ ∗ (B) = 0.
Finally, for each set B such that µ ∗ (B) = 0, there exists C ∈ Aσ δ such that B ⊂ C and µ ∗ (C) = 0.
Proof 7.19 The if part is obvious.

Now suppose E is µ ∗ -measurable. Let {Ui } be a countable collection of disjoint sets of finite measure
such that
U = ∪∞ i=1Ui .
Observe that
E = ∪∞
i=1 Ei
where
Ei = E ∩Ui ,
is µ ∗ -measurablefor each i ∈ N.
Let ε > 0. From the last proposition for each i, n ∈ N there exists Ani ∈ Aσ such that
1
µ(Ani ) < µ ∗ (Ei ) + .
n2i
Define
An = ∪∞
i=1 Ani .
Thus,
E ⊂ An
and
An \ E ⊂ ∪∞
i=1 (Ani \ Ei ),
and therefore,
∞ ∞
1 1
µ(An \ E) ≤ ∑ µ(Ani \ Ei ) ≤ ∑ i
= .
i=1 i=1 n 2 n
Since An ∈ Aσ , defining
A = ∩∞
n=1 An ,
we have that A ∈ Aσ δ , and,

A \ E ⊂ An \ E
so that
1
µ ∗ (A \ E) ≤ µ ∗ (An \ E) ≤ , ∀n ∈ N.
n
Hence µ ∗ (A \ E) = 0.
Theorem 7.7.10 (Caratheodory) ´ Let µ be a measure on a algebra A and µ ∗ the respective induced outer
measure.
Then the restriction µ of µ ∗ to the µ ∗ -measurable sets is an extension of µ to an σ -algebra containing
A . If µ is finite or σ -finite, so is µ. In particular, if µ is σ -finite, then µ is the only measure on the smallest
σ -algebra containing A which is an extension of µ.
Proof 7.20 From the Theorem 7.7.2, µ is an extension of µ to a σ -algebra containing A , that is, µ is a
measure on such a set.
Observe that, from the last results, if µ is σ -finite so is µ.
Now assume µ is σ -finite. We will prove the uniqueness of µ.
Let B be the smallest σ -algebra containing A and let µ̃ be another measure on B which extends µ
on A .
Since each set Aσ may be expressed as a disjoint countable union of sets in A , the measure µ̃ equals µ
on Aσ . Let B be a µ ∗ -measurable set such that µ ∗ (B) < ∞.
Let ε > 0. By the Proposition 7.7.9 there exists an A ∈ Aσ such that B ⊂ A and
µ ∗ (A) < µ ∗ (B) + ε.
Since B ⊂ A, we obtain
µ̃(B) ≤ µ̃(A) = µ ∗ (A) ≤ µ ∗ (B) + ε.
Considering that ε > 0 is arbitrary, we get
µ̃(B) ≤ µ ∗ (B).
Observe that the class of µ ∗ -measurable sets is an σ -algebra containing A .

Therefore as above indicated, we have obtained A ∈ Aσ such that B ⊂ A and
µ ∗ (A) ≤ µ ∗ (B) + ε
so that
µ ∗ (A) = µ ∗ (B) + µ ∗ (A \ B),
from this and above
µ̃(A \ B) ≤ µ ∗ (A \ B) ≤ ε,
if µ ∗ (B) < ∞.
Therefore,
µ ∗ (B) ≤ µ ∗ (A) = µ̃(A)

= µ̃(B) + µ̃(A \ B) ≤ µ̃(B) + ε. (7.51)
Since ε > 0 is arbitrary we have

µ ∗ (B) ≤ µ̃(B),
so that
µ ∗ (B) = µ̃(B).
Finally, since µ is σ -finite, there exists a sequence of countable disjoint sets {Ui } such that µ(Ui ) <
∞, ∀i ∈ N and U = ∪i=1∞ U.
i
If B ∈ B, then
B = ∪∞i=1 (Ui ∩ B).
Thus, from above

∞
µ̃(B) = ∑ µ̃(Ui ∩ B)
i=1
∞
= ∑ µ(Ui ∩ B)
i=1
= µ(B). (7.52)

Remark 7.7.11 We may start the process of construction of a measure by the action of a set function on a
semi-algebra. Here, a semi-algebra C is a collection of subsets of U such that the intersection of any two
sets in C is in C and, the complement of any set in C is a finite disjoint union of sets in C .
If C is any semi-algebra of sets, then the collection consisting of the empty set and all finite disjoint
unions of sets in C is an algebra, which is said to be generated by C . We denote such algebra by A .
If we have a set function acting on C , we may extend it to A by defining
n
µ(A) = ∑ µ(Ei ),
i=1
where, A = ∪ni=1 Ei and Ei ∈ C , ∀i ∈ {1, ..., n}, so that this last union is disjoint. We recall that any A ∈ A
admits such a representation.
7.8 Fubini’s theorem

We start this section with the definition of complete measure space.
Definition 7.8.1 We say that a measure space (U, M , µ) is complete if M contains all subsets of sets of
zero measure. That is, if A ∈ M , µ(A) = 0 and B ⊂ A then B ∈ M .
In the next lines we recall the formal definition of a semi-algebra.
Definition 7.8.2 We say (in fact recall) that C ⊂ U is a semi-algebra in U if the two conditions below are
valid.
1. if A, B ∈ C then A ∩ B ∈ C ,
2. For each A ∈ C , Ac is a finite disjoint union of elements in C .
7.8.1 Product measures

Let (U, M1 , µ1 ) and (V, M2 , µ2 ) be two complete measure spaces. We recall that the cartesian product be-
tween U and V , denoted by U ×V is defined by
U ×V = {(u, v) | u ∈ U and v ∈ V }.
If A ⊂ U and B ⊂ V we call A × B a rectangle. If A ∈ M1 and B ∈ M2 we say that A × B is a measurable

rectangle. The collection R of measurable rectangles is a semi-algebra since
(A × B) ∩ (C × D) = (A ∩C) × (B ∩ D),
and
(A × B)c = (Ac × B) ∪ (A × Bc ) ∪ (Ac × Bc ).
We define λ : M1 × M2 → R+ by
λ (A × B) = µ1 (A)µ2 (B).
Lemma 7.8.3 Let {Ai × Bi }i∈N be a countable disjoint collection of measurable rectangles whose the union
is the rectangle A × B. Then
∞
λ (A × B) = ∑ µ1 (Ai )µ2 (Bi ).
i=1
Proof 7.21 Let u ∈ A. Thus each v ∈ B is such that (u, v) is exactly in one Ai × Bi . Therefore
∞
χA×B (u, v) = ∑ χAi (u)χBi (v).
i=1
Hence for the fixed u in question, from the corollary of Lebesgue monotone convergence theorem we may
write
Z Z ∞
χA×B (u, v)dµ2 (v) = ∑ χAi (u)χBi (v)dµ2 (v)
V V i=1
∞
= ∑ χAi (u)µ2 (Bi ) (7.53)
i=1
so that also from the mentioned corollary

Z Z ∞
dµ1 (u) χA×B (u, v)dµ2 (v) = ∑ µ1 (Ai )µ2 (Bi ).
U V i=1
Observe that
Z Z Z Z
dµ1 (u) χA×B (u, v)dµ2 (v) = dµ1 (u) χA (u)χB (v)dµ2 (v)
U V U V
= µ1 (A)µ2 (B).
From the last two equations we may write

∞
λ (A × B) = µ1 (A)µ2 (B) = ∑ µ1 (Ai )µ2 (Bi ).
i=1
Definition 7.8.4 Let E ⊂ U ×V . We define Eu and Ev by
Eu = {v | (u, v) ∈ E},
and
Ev = {u | (u, v) ∈ E}.
Observe that
χEu (v) = χE (u, v),
(E c )u = (Eu )c ,
and
(∪Eα )u = ∪(Eα )u ,
for any collection {Eα }.
We denote by Rσ as the collection of sets which are countable unions of measurable rectangles. Also,
Rσ δ will denote the collection of sets which are countable intersections of elements of Rσ .
Lemma 7.8.5 Let u ∈ U and E ∈ Rσ δ . Then Eu is a measurable subset of V .
Proof 7.22 If E ∈ R the result is trivial. Let E ∈ Rσ . Then E may be expressed as a disjoint union
E = ∪∞
i=1 Ei ,
where Ei ∈ R, ∀i ∈ N. Thus,
χEu (v) = χE (u, v)

= sup χEi (u, v)
i∈N
= sup χ(Ei )u (v). (7.54)
i∈N
Since each (Ei )u is measurable we have that

χ(Ei )u (v)
is a measurable function of v, so that
χEu (v)
is measurable, which implies that Eu is measurable. Suppose now
E = ∩∞
i=1 Ei ,
where Ei+1 ⊂ Ei , ∀i ∈ N. Then
χEu (v) = χE (u, v)

= inf χEi (u, v)
i∈N
= inf χ(Ei )u (v). (7.55)
i∈N
Thus as from above χ(Ei )u (v) is measurable for each i ∈ N, we have that χEu is also measurable so that Eu is
measurable.
Lemma 7.8.6 Let E be a set in Rσ δ with (µ1 × µ2 )(E) < ∞. Then the function g defined by
g(u) = µ2 (Eu )
is a measurable function and Z

g dµ1 (u) = (µ1 × µ2 )(E).
U
Proof 7.23 The lemma is true if E is a measurable rectangle. Let {Ei } be a disjoint sequence of measurable
rectangles and E = ∪∞
i=1 Ei . Set
gi (u) = µ2 ((Ei )u ).
Then each gi is a non-negative measurable function and
∞
g = ∑ gi .
i=1
Thus g is measurable, and by the corollary of the Lebesgue monotone convergence theorem, we have
Z ∞ Z
g(u)dµ1 (u) = ∑ gi (u)dµ1 (u)
U i=1 U
∞
= ∑ (µ1 × µ2 )(Ei )
i=1
= (µ1 × µ2 )(E). (7.56)
Let E be a set of finite measure in Rσ δ . Then there is a sequence in Rσ such that
Ei+1 ⊂ Ei
and
E = ∩∞
i=1 Ei .
Let gi (u) = µ2 ((Ei )u ), since Z

g1 (u) = (µ1 × µ2 )(E1 ) < ∞,
U
we have that
g1 (u) < ∞ a.e. in E1 .
For an u ∈ E1 such that g1 (u) < ∞ we have that {(Ei )u } is a sequence of measurable sets of finite measure
whose intersection is Eu . Thus
g(u) = µ2 (Eu ) = lim µ2 ((Ei )u ) = lim gi (u), (7.57)
i→∞ i→∞
that is,
gi → g, a.e. in E.
We may conclude that g is also measurable. Since
0 ≤ gi ≤ g, ∀i ∈ N
the Lebesgue dominated convergence theorem implies that
Z Z
g(u) dµ1 (u) = lim gi dµ1 (u)
U i→∞ U
= lim(µ1 × µ2 )(Ei )
i→∞
= (µ1 × µ2 )(E).
Lemma 7.8.7 Let E be a set such that (µ1 × µ2 )(E) = 0. then for almost all u ∈ U we have
µ2 (Eu ) = 0.
Proof 7.24 Observe that there is a set in Rσ δ such that E ⊂ F and
(µ1 × µ2 )(F) = 0.
From the last lemma

µ2 (Fu ) = 0
fora almost all u. From Eu ⊂ Fu we obtain
µ2 (Eu ) = 0
for almost all u, since µ2 is complete.
Proposition 7.8.8 Let E be a measurable subset of U ×V such that (µ1 × µ2 )(E) is finite. The for almost all
u the set Eu is a measurable subset of V . The function g defined by
g(u) = µ2 (Eu )
is measurable and Z
g dµ1 (u) = (µ1 × µ2 )(E).
Proof 7.25 First observe that there is a set F ∈ Rσ δ such that E ⊂ F and
(µ1 × µ2 )(F) = (µ1 × µ2 )(E).
Let G = F \ E. Since F and E are measurable, G is measurable, and
(µ1 × µ2 )(G) = 0.
By the last lemma we obtain
µ2 (Gu ) = 0,
for almost all u so that
g(u) = µ2 (Eu ) = µ2 (Fu ), a.e. in U.
By Lemma 7.8.6 we may conclude that g is measurable and
Z
g dµ1 (u) = (µ1 × µ2 )(E).
U
Theorem 7.8.9 (Fubini) Let (U, M1 , µ1 ) and (V, M2 , µ2 ) be two complete measure spaces and f an inte-
grable function on U ×V . Then
1. fu (v) = f (u, v) is measurable and integrable for almost all u.
2. fv (u) = f (u, v) is measurable and integrable for almost all v.
R
3. h1 (u) = V f (u, v) dµ2 (v) is integrable on U.
R
4. h2 (v) = U f (u, v) dµ1 (u) is integrable on V .
5.
Z Z Z Z
f dµ2 (v) dµ1 (u) = f dµ1 (u) dµ2 (v)
U V V U
Z
= f d(µ1 × µ2 ). (7.58)
U×V
Proof 7.26 It suffices to consider the case where f is non-negative (we can then apply the result to f + =
max( f , 0) and f − = max(− f , 0)). The last proposition asserts that the theorem is true if f is a simple function
which vanishes outside a set of finite measure. Similarly as in in Theorem 7.2.2, we may obtain a sequence
of non-negative simple functions {φn } such that
φn ↑ f .
Observe that, given u ∈ U, fu is such that
(φn )u ↑ fu , a.e. .
By the Lebesgue monotone convergence theorem we get
Z Z
f (u, v) dµ2 (v) = lim φn (u, v) dµ2 (v),
V n→∞ V
so that this last resulting function is integrable in U. Again by the Lebesgue monotone convergence theorem,
we obtain
Z Z Z Z
f dµ2 (v) dµ1 (u) = lim φn dµ2 (v) dµ1 (u)
U V n→∞ U V
Z
= lim φn d(µ1 × µ2 )
n→∞ U×V
Z
= f d(µ1 × µ2 ). (7.59)
U×V
Chapter 8
The Lebesgue Measure in Rn
8.1 Introduction
In this chapter we will define the Lebesgue measure and the concept of Lebesgue measurable set. We show
that the set of Lebesgue measurable sets is a σ − algebra so that the earlier results, proven for more gen-
eral measure spaces, remain valid in the present context (such as the Lebesgue monotone and dominated
convergence theorems).
We start with the following theorems without proofs.
Theorem 8.1.1 Every open set A ⊂ R may be expressed as a countable union of disjoint open intervals.
Remark 8.1.2 In this text Q j denotes a closed cube in Rn and |Q j | its volume, that is, |Q j | = ∏ni=1 (bi − ai ),
where Q j = ∏ni=1 [ai , bi ]. Also we assume that if two Q1 and Q2 closed or not, have the same interior, then
|Q1 | = |Q2 | = |Q̄1 |. We recall that two cubes Q1 , Q2 ⊂ Rn are said to be quasi-disjoint if their interiors are
disjoint.
Theorem 8.1.3 Every open set A ⊂ Rn , where n ≥ 1 may be expressed as a countable union of quasi-disjoint
closed cubes.
Definition 8.1.4 (Outer measure) Let E ⊂ Rn . The outer measure of E, denoted by m∗ (E), is defined by
( )
∞
m∗ (E) = inf ∑ |Q j | : E ⊂ ∪∞j=1 Q j ,
j=1
where Q j is a closed cube, ∀ j ∈ N.
8.2 Properties of the outer measure

First observe that given ε > 0, there exists a sequence {Q j } such that
E ⊂ ∪∞j=1 Q j
and
∞
∑ |Q j | ≤ m∗ (E) + ε.
j=1
1. Monotonicity: If E1 ⊂ E2 then m∗ (E1 ) ≤ m∗ (E2 ). This follows from the fact that if E2 ⊂ ∪∞j=1 Q j then
E1 ⊂ ∪∞j=1 Q j .
2. Countable sub-additivity : If E ⊂ ∪∞j=1 E j , then m∗ (E) ≤ ∑∞j=1 m∗ (E j ).
Proof 8.1 First assume that m∗ (E j ) < ∞, ∀ j ∈ N, otherwise the result is obvious. Thus, given ε > 0
for each j ∈ N there exists a sequence {Qk, j }k∈N such that
E j ⊂ ∪∞
k=1 Qk, j
and
∞
ε
∑ |Qk, j | < m∗ (E j ) + 2 j .
k=1
Hence
E ⊂ ∪∞j,k=1 Qk, j
and therefore
!
∞ ∞ ∞
m∗ (E) ≤ ∑ |Qk, j | = ∑ ∑ |Qk, j |
j,k=1 j=1 k=1
∞
ε
≤ ∑ m∗ (E j ) + j
j=1 2
∞
= ∑ m∗ (E j ) + ε. (8.1)
j=1
Being ε > 0 arbitrary, we obtain

∞
m∗ (E) ≤ ∑ m∗ (E j ).
j=1
3. If
E ⊂ Rn ,
and
α = inf{m∗ (A) | A is open and E ⊂ A},
then
m∗ (E) = α.
Proof 8.2 From the monotonicity, we have
m∗ (E) ≤ m∗ (A), ∀A ⊃ E, A open .
Thus,
m∗ (E) ≤ α.
Suppose given ε > 0. Choose a sequence {Q j } of closed cubes such that
E ⊂ ∪∞j=1 Q j
The Lebesgue Measure in Rn 163
and
∞
∑ |Q j | ≤ m∗ (E) + ε.
j=1
Let {Q̃ j } be a sequence of open cubes such that Q̃ j ⊃ Q j

ε
|Q̃ j | ≤ |Q j | + , ∀ j ∈ N.
2j
Define
A = ∪∞j=1 Q˜ j ,
hence A is open, A ⊃ E and
∞
m∗ (A) ≤ ∑ |Q̃ j |
j=1
∞
ε
≤ ∑ |Q j | +
j=1 2j
∞
= ∑ |Q j | + ε
j=1
∗
≤ m (E) + 2ε. (8.2)
therefore
α ≤ m∗ (E) + 2ε.
Being ε > 0 arbitrary, we have
α ≤ m∗ (E).
4. If E = E1 ∪ E2 and d(E1 , E2 ) > 0, then
m∗ (E) = m∗ (E1 ) + m∗ (E2 ).
Proof 8.3 First observe that being E = E1 ∪ E2 we have
m∗ (E) ≤ m∗ (E1 ) + m∗ (E2 ).
Let ε > 0. Choose {Q j } a sequence of closed cubes such that
E ⊂ ∪∞j=1 Q j ,
and
∞
∑ |Q j | ≤ m∗ (E) + ε.
j=1
Let δ > 0 such that

d(E1 , E2 ) > δ > 0.
Dividing the cubes Q j if necessary, we may assume that the diameter of each cube Q j is smaller than
δ . Thus each Q j intersects just one of the sets E1 and E2 . Denote by J1 and J2 the sets of indices j
such that Q j intersects E1 and E2 respectively. thus,
E1 ⊂ ∪ j∈J1 Q j and E2 ⊂ ∪ j∈J2 Q j .

hence,
m∗ (E1 ) + m∗ (E2 ) ≤ ∑ |Q j | + ∑ |Q j |
j∈J1 j∈J2
∞
≤ ∑ |Q j | ≤ m∗ (E) + ε. (8.3)
j=1
Being ε > 0 arbitrary,

m∗ (E1 ) + m∗ (E2 ) ≤ m∗ (E).
5. If a set E is a countable union of cubes quasi disjoints, that is
E = ∪∞j=1 Q j
then
∞
m∗ (E) = ∑ |Q j |.
j=1
Proof 8.4 Let ε > 0.

Let {Q̃ j } be open cubes such that Q̃ j ⊂⊂ Q◦j (that is, the closure of Q̃ j is contained in the interior of
Q j ) and
ε
|Q j | ≤ |Q̃ j | + j .
2
Thus, for each N ∈ N, the cubes Q̃1 , ..., Q̃N are disjoint and each pair have a finite distance. Hence,
N N ε
m∗ (∪Nj=1 Q˜ j ) = ∑ |Q̃ j | ≥ ∑ |Q j | − j .
j=1 j=1 2
Being
∪Nj=1 Q˜ j ⊂ E
we obtain
N N
m∗ (E) ≥ ∑ |Q̃ j | ≥ ∑ |Q j | − ε.
j=1 j=1
Therefore,
∞
∑ |Q j | ≤ m∗ (E) + ε.
j=1
Being ε > 0 arbitrary, we may conclude that

∞
∑ |Q j | ≤ m∗ (E).
j=1

8.3 The Lebesgue measure

Definition 8.3.1 Let E ⊂ Rn . We say that E is measurable if for each A ⊂ Rn we have
m∗ (A) = m∗ (A ∩ E) + m∗ (A ∩ E c ).
In such a case we define the Lebesgue measure of E, denoted by m(E), by
m(E) = m∗ (E).
Proposition 8.3.2 Let E ⊂ Rn be such that m∗ (E) = 0. Under such hypotheses, E is measurable.
Proof 8.5 Let A ⊂ Rn . Observe that A ∩ E ⊂ E so that
m∗ (A ∩ E) = 0.
m∗ (A) ≤ m∗ (A ∩ E) + m∗ (A ∩ E c )
= m∗ (A ∩ E c )
≤ m∗ (A). (8.4)
Thus,
m∗ (A) = m∗ (A ∩ E) + m∗ (A ∩ E c ).
Since A is arbitrary, we may infer that E is measurable.
Theorem 8.3.3 Let E = ∪nk=1 Ek where Ek ⊂ Rn is measurable, ∀k ∈ {1, . . . , n}.

Under such hypotheses, E is measurable.
Proof 8.6 First we prove that E1 ∪ E2 is measurable.

Let A ⊂ Rn . Since E1 and E2 are measurable, we have
m∗ (A) = m∗ (A ∩ E1 ) + m∗ (A ∩ E1c )
= m∗ (A ∩ E1 ) + m∗ ((A ∩ E1c ) ∩ E2 ) + m∗ ((A ∩ E1c ) ∩ E2c ). (8.5)
Now observe that

(A ∩ E1c ) ∩ E2c = A ∩ (E1c ∩ E2c ) = A ∩ [E1 ∪ E2 ]c .
Moreover,
(A ∩ E1 ) ∪ [A ∩ E1c ∩ E2 ] = (A ∩ E1 ) ∪ (A ∩ E2 ) ∩ E1c )
= A ∩ (E1 ∪ E2 ) ∩ [A ∩ (E1 ∪ E1c )]
= A ∩ [E1 ∪ E2 ] ∩ A
= A ∩ (E1 ∪ E2 ). (8.6)
From these last results, we obtain,
m∗ (A) = m∗ (A ∩ E1 ) + m∗ ((A ∩ E1c ) ∩ E2 ) + m∗ ((A ∩ E1c ) ∩ E2c )

= m∗ (A ∩ E1 ) + m∗ ((A ∩ E1c ) ∩ E2 ) + m∗ (A ∩ (E1 ∪ E2 )c )
≥ m∗ (A ∩ (E1 ∪ E2 )) + m∗ (A ∩ (E1 ∪ E2 )c ). (8.7)
Since A is arbitrary, from this we may infer that E1 ∪ E2 is measurable.

To complete the proof, we reason by induction. Let k ∈ {2, . . . n − 1}.
Assume ∪kj=1 E j is measurable,
Observe that ∪k+1 k
j=1 E j = (∪ j=1 E j )) ∪ Ek , which from the last result is measurable.
In particular,
∪nj=1 E j
is measurable.
Theorem 8.3.4 Let A ⊂ Rn and let {Ek }nk=1 be finite collection of measurable disjoint sets.
n
m∗ (A ∩ [∪nk=1 Ek ]) = ∑ m∗ (A ∩ Ek ).
k=1
In particular,
n
m∗ (∪nk=1 Ek ) = ∑ m∗ (Ek ).
k=1
Proof 8.7 We prove the result by induction on n. The result is immediate for n = 1.
Assume the result holds for n − 1.
Since the collection {Ek }nk=1 is disjoint, we have
A ∩ [∪nk=1 Ek ] ∩ En = A ∩ En ,
and
A ∩ [∪nk=1 Ek ] ∩ Enc = A ∩ [∪n−1
k=1 Ek ].
Thus, from the measurability of En and the induction assumption, we get
m∗ (A ∩ [∪nk=1 Ek ]) = m∗ (A ∩ [∪nk=1 Ek ] ∩ En ) + m∗ (A ∩ [∪nk=1 Ek ] ∩ Enc )

= m∗ (A ∩ En ) + m∗ (A ∩ [∪nk=−1
1 Ek ])
n−1
= m∗ (A ∩ En ) + ∑ m∗ (A ∩ Ek )
k=1
n
= ∑ m∗ (A ∩ Ek ). (8.8)
k=1
8.4 Outer and inner approximations of Lebesgue measurable sets

We start this section with the following remark:
Remark 8.4.1 Let A ⊂ B ⊂ Rn be sets such that A is a measurable set which has a finite outer measure.
In such a case
m∗ (B) = m∗ (B ∩ A) + m∗ (B ∩ Ac )
= m∗ (A) + m∗ (B \ A), (8.9)
so that
m∗ (B \ A) = m∗ (B) − m∗ (A).
Theorem 8.4.2 Let E ⊂ Rn . Under such hypothesis, the following properties are equivalent to the measur-
ability of E:
1. For each ε > 0 there exists an open set O ⊂ Rn such that
E ⊂O
and
m∗ (O \ E) < ε.
2. There is a Gδ set G such that

E ⊂G
and m∗ (G \ E) = 0, where a Gδ set is one which may be expressed as a countable intersection of open
sets.
3. For each ε > 0 there exists a closed set F ⊂ Rn such that
F ⊂E
and
m∗ (E \ F) < ε.
4. There is a Fσ set G such that

F ⊂E
and m∗ (E \ F) = 0, where a Fσ set is one which may be expressed as a countable union of closed sets.
Proof 8.8 Assume E is measurable. Let ε > 0. Assume first m∗ (E) < ∞. Hence, there exists a sequence of
open blocks {Q j } ⊂ Rn such that
E ⊂ ∪∞j=1 Q j
and
∞
∑ m∗ (Q j ) < m∗ (E) + ε.
j=1
Define O = ∪∞j=1 Q j . Thus,
E ⊂O
and
∞
m∗ (O) ≤ ∑ m∗ (Q j ) < m∗ (E) + ε.
j=1
Thus
m∗ (O) − m∗ (E) < ε.
From the measurability of E we get,
m∗ (O) = m∗ (O ∩ E) + m∗ (O ∩ E c )
= m∗ (O ∩ E) + m∗ (O \ E), (8.10)
so that
m(O \ E) = m∗ (O) − m∗ (E) < ε.
Assume now m∗ (E) = ∞. Let Fk = E ∩ Bk (0), ∀k ∈ Rn , where Bk (0) denotes the open ball of center 0
and radius k.
Define also, E1 = F1 and

Ek = Fk \ ∪k−1
j=1 Fk , ∀k ≥ 2.
Hence, E = ∪∞ k=1 Ek and such union is disjoint.
Observe that, from the last previous lines, for each k ∈ N there exists an open set Ok such that Ek ⊂ Ok
and
ε
m∗ (Ok \ Ek ) < .
2k
Define now O = ∪∞
k=1 Ok . Thus such a set is open and
O \ E = (∪∞
k=1 Ok ) \ E = ∪k=1 (Ok \ E) ⊂ ∪k=1 (Ok \ Ek ).
∞ ∞
Thus,
∞
m∗ (O \ E) ≤ ∑ m∗ (Ok \ Ek )
k=1
∞
ε
≤ ∑ 2k
k=1
= ε. (8.11)
From this 1 holds.

Hence from 1, for each k ∈ N there exists Ok open such that E ⊂ Ok and
1
m∗ (Ok \ E) < .
k
k=1 Ok . Thus G is a Gδ set such that E ⊂ G and

Define G = ∩∞
1
m∗ (G \ E) ≤ m∗ (Ok \ E) < , ∀k ∈ N.
k
Therefore, m∗ (G \ E) = 0.
We have proven the measurability of E implies 1, which implies 2.
Now, suppose 2 holds. Thus, there exists a Gδ set G such that E ⊂ G and
m∗ (G \ E) = 0.
Hence, G \ E is measurable.
We now are going to show that
E = G ∩ (G \ E )c .
Indeed, since E ⊂ G, we have,
x ∈ G ∩ (G \ E )c ⇔ x ∈ G ∩ (G ∩ E c )c
⇔ x ∈ G ∩ (Gc ∪ E)
⇔ x ∈ (G ∩ Gc ) ∪ (G ∩ E)
⇔ x ∈ E. (8.12)
Summarizing
E = G ∩ (G \ E)c ,
so that E is measurable.
From this we may infer that the measurability of E is equivalent of 1 and equivalent to 2.
Reasoning with the complements, we may obtain that 1 is equivalent to 3 and 2 is equivalent to 4.
8.5 Some other properties of measurable sets

1. Each open set is measurable. This results immediately from the last theorem.
2. If E ⊂ Rn is measurable, then E c is measurable. This is obvious from the definition of measurable

sets.
3. Closed sets are measurable. This results from the last two items
4. A countable intersection of measurable sets is measurable.
Proof 8.9 This follows just from observing that
∩∞j=1 E j = (∪∞j=1 E cj )c .
Theorem 8.5.1 If {Ei } is sequence of measurable pairwise disjoint sets and E = ∪∞j=1 Ei then
∞
m(E) = ∑ m(E j ).
j=1
Proof 8.10 First assume that E j is bounded. Being E cj measurable, given ε > 0 there exists an open H j ⊃ E cj
such that
ε
m∗ (H j − E cj ) < , ∀ j ∈ N.
2j
Denoting Fj = H cj we have that Fj ⊂ E j is closed and
ε
m∗ (E j − Fj ) < , ∀ j ∈ N.
2j
For each N ∈ N the sets F1 , ..., FN are compact and disjoint, so that
N
m(∪Nj=1 Fj ) = ∑ m(Fj ).
j=1
As
∪Nj=1 Fj ⊂ E
we have
N N
m(E) ≥ ∑ m(Fj ) ≥ ∑ m(E j ) − ε.
j=1 j=1
Hence,
∞
m(E) ≥ ∑ m(E j ) − ε.
j=1
being ε > 0 arbitrary, we obtain

∞
m(E) ≥ ∑ m(E j ).
j=1
As the reverse inequality is always valid, we have

∞
m(E) = ∑ m(E j ).
j=1
For the general case, select a sequence of cubes {Qk } such that
Rn = ∪∞
k=1 Qk
and Qk ⊂ Qk+1 ∀k ∈ N. Define S1 = Q1 e Sk = Qk − Qk−1 , ∀k ≥ 2. Also define
E j,k = E j ∩ Sk , ∀ j, k ∈ N.
Thus,
E = ∪∞j=1 ∪∞ ∞
k=1 E j,k = ∪ j,k=1 E j,k ,
where such an union is disjoint and each E j,k is bounded. Through the last result, we get
∞
m(E) = ∑ m(E j,k )
j,k=1
∞ ∞
= ∑ ∑ m(E j,k )
j=1 k=1
∞
= ∑ m(E j ). (8.13)
j=1
Theorem 8.5.2 Suppose E ⊂ Rn is a measurable set. Then for each ε > 0:

1. If m(E) is finite, there exists a compact set K ⊂ E such that
m(E \ K) < ε.
2. If m(E) is finite there exist a finite union of closed cubes
F = ∪Nj=1 Q j
such that
m(E 4 F) ≤ ε,
where
E 4 F = (E \ F) ∪ (F \ E).
Proof 8.11
1. Choose a closed set such that F ⊂ E and
ε
m(E \ F) < .
2
Let Bn be a closed ball with center at origin and radius n. Define Kn = F ∩ Bn and observe that Kn is
compact, ∀n ∈ N. Thus
E \ Kn & E \ F.
Being m(E) < ∞ we have
m(E \ Kn ) < ε,
for all n sufficiently big.
2. Choose a sequence of closed cubes {Q j } such that
E ⊂ ∪∞j=1 Q j ,
and
∞
ε
∑ |Q j | ≤ m(E) + 2 .
j=1
Being m(E) < ∞ the series converges and there exists N0 ∈ N such that
∞
ε
∑ |Q j | < .
N0 +1 2
N
0
Defining F = ∪ j=1 Q j , we have
m(E 4 F) = m(E − F) + m(F − E)

≤ m ∪∞j=N0 +1 Q j + m ∪∞j=1 Q j − E
∞ ∞
≤ ∑ |Q j | + ∑ |Q j | − m(E)
j=N0 +1 j=1
ε ε
≤ + = ε. (8.14)
2 2
8.6 Lebesgue measurable functions

Definition 8.6.1 Let E ⊂ Rn be a measurable set. A function f : E → [−∞, +∞] is said to be Lebesgue
measurable if for each a ∈ R, the set
f −1 ([−∞, a)) = {x ∈ E | f (x) < a}
is measurable.
Observe that:
1. If
f −1 ([−∞, a))
is measurable for each a ∈ R then
f −1 ([−∞, a]) = ∩k=1

∞
f −1 ([−∞, a + 1/k))
is measurable for each a ∈ R.

2. If
f −1 ([−∞, a])
is measurable for each a ∈ R then
f −1 ([−∞, a)) = ∪∞
k=1 f
−1
([−∞, a − 1/k])
is also measurable for each a ∈ R.

3. Given a ∈ R, observe that

f −1 ([−∞, a)) is measurable ⇔ E − f −1 ([−∞, a)) is measurable
⇔ f −1 (R) − f −1 ([−∞, a)) ⇔ f −1 (R − [−∞, a)) is measurable
⇔ f −1 ([a, +∞]) is measurable . (8.15)
4. From above, we can prove that

f −1 ([−∞, a))
is measurable ∀a ∈ R if, and only if
f −1 ((a, b))
is measurable for each a, b ∈ R such that a < b. Therefore f is measurable if and only if f −1 (O) is
measurable whenever O ⊂ R is open.
5. Thus, f is measurable if f −1 (F ) is measurable whenever F ⊂ R is closed.
Proposition 8.6.2 If f is continuous in Rn , then f is measurable. If f is measurable and real and φ is
continuous, then φ ◦ f is measurable.
Proof 8.12 The first implication is obvious. For the second, being φ continuous
φ −1 ([−∞, a))
is open, and therefore
(φ ◦ f )−1 (([−∞, a)) = f −1 (φ −1 ([−∞, a)))
is measurable, ∀a ∈ R.
Proposition 8.6.3 Suppose { fk } is a sequence of measurable functions. Then

sup fk (x), inf fk (x),
k∈N k∈N
and
lim sup fk (x), lim inf fk (x)
k→∞ k→∞
are measurable.
Proof 8.13 We will prove only that supn∈N fn (x) is measurable. The remaining proofs are analogous. Let
f (x) = sup fn (x).
n∈N
Thus
f −1 ((a, +∞]) = ∪∞ −1
n=1 f n ((a, +∞]).
Being each fn measurable, such a set is measurable, ∀a ∈ R. By analogy
inf fk (x)
k∈N
is measurable,
lim sup fk (x) = inf sup f j (x),
k→∞ k≥1 j≥k
and
lim inf fk (x) = sup inf f j (x)
k→∞ k≥1 j≥k
are measurable.
Proposition 8.6.4 Let { fk } be a sequence of measurable functions such that
lim fk (x) = f (x).

k→∞
Then f is measurable.
Proof 8.14 Just observe that

f (x) = lim fk (x) = lim sup fk (x).
k→∞ k→∞
The next result we do not prove it. For a proof see [72].
Proposition 8.6.5 If f and g are measurable functions, then
1. f 2 is measurable.
2. f + g and f · g are measurable if both assume finite values.
Proposition 8.6.6 Let E ⊂ Rn a measurable set. Suppose f : E → R is measurable. Thus, if g : E → R is

such that
g(x) = f (x), a.e. in E
then g is measurable.
Proof 8.15 Define

A = {x ∈ E | f (x) =
6 g(x)},
and
B = {x ∈ E | f (x) = g(x)}.
A is measurable since m∗ (A) = m(A) = 0 and therefore B = E − A is also measurable. Let a ∈ R. Hence,
g−1 ((a, +∞]) = g−1 ((a, +∞]) ∩ A ∪ g−1 ((a, +∞]) ∩ B .

Observe that
x ∈ g−1 ((a, +∞]) ∩ B ⇔ x ∈ B and g(x) ∈ (a, +∞]

⇔ x ∈ B and f (x) ∈ (a, +∞]
⇔ x ∈ B ∩ f −1 ((a, +∞]). (8.16)
Thus g−1 ((a, +∞]) ∩ B is measurable. As g−1 ((a, +∞]) ∩ A ⊂ A we have m∗ (g−1 ((a, +∞]) ∩ A) = 0, that is,
such a set is measurable. Hence being g−1 ((a, +∞]) the union of two measurable sets is also measurable.
Being a ∈ R arbitrary, g is measurable.
Theorem 8.6.7 Suppose f is a non-negative measurable function on Rn . Then there exists a increasing
sequence of non-negative simple functions {ϕk } such that
lim ϕk (x) = f (x), ∀x ∈ Rn .

k→∞
Proof 8.16 Let N ∈ N. Let QN be the cube with center at origin and side of measure N. Define

 f (x), if x ∈ QN and f (x) ≤ N,
FN (x) = N, if x ∈ QN and f (x) > N,
0, otherwise.

Thus, FN (x) → f (x) as N → ∞, ∀x ∈ Rn . Fixing M, N ∈ N define

l l +1
El,M = x ∈ QN : ≤ FN (x) ≤ ,
M M
for 0 ≤ l ≤ N · M. Defining
NM
l
FN,M = ∑ M χEl,M ,
l=0
we have that FN,M is a simple function and
1
0 ≤ FN (x) − FN,M (x) ≤ .
M
If ϕK (x) = FK,K (x) we obtain
1
0 ≤ |FK (x) − ϕK (x)| ≤ .
K
Hence,
| f (x) − ϕK (x)| ≤ | f (x) − FK (x)| + |FK (x) − ϕK (x)|.
Therefore,
lim | f (x) − ϕK (x)| = 0, ∀x ∈ Rn .
K→∞
Theorem 8.6.8 Suppose that f is a measurable function defined on Rn . Then there exists a sequence of
simple functions {ϕk } such that
|ϕk (x)| ≤ |ϕk+1 (x)|, ∀x ∈ Rn , k ∈ N,
and
lim ϕk (x) = f (x), ∀x ∈ Rn .
k→∞
Proof 8.17 Write

f (x) = f + (x) − f − (x),
where
f + (x) = max{ f (x), 0}
and
f − (x) = max{− f (x), 0}.
Thus f + and f − are non-negative measurable functions so that from the last theorem there exist increasing
sequences of non-negative simple functions such that
(1)
ϕk (x) → f + (x), ∀x ∈ Rn ,
and
(2)
ϕk (x) → f − (x), ∀x ∈ Rn ,
as k → ∞. Defining
(1) (2)
ϕk (x) = ϕk (x) − ϕk (x),
we obtain
ϕk (x) → f (x), ∀x ∈ Rn
as k → ∞ and
(1) (2)
|ϕk (x)| = ϕk (x) + ϕk (x) % | f (x)|, ∀x ∈ Rn ,
as k → ∞.
Theorem 8.6.9 Suppose f is a measurable function in Rn . Then there exists a sequence of step functions
{ϕk } which converges to f a.e. in Rn .
Proof 8.18 From the last theorem, it suffices to prove that if E is measurable and m(E) < ∞ then χE may be
approximated almost everywhere in E by step functions. Suppose given ε > 0. Observe that from Proposition
8.5.2, there exist cubes Q1 , ..., QN such that
m(E4 ∪Nj=1 Q j ) < ε.
We may obtain almost disjoints rectangles R̃ j such that ∪M ˜ N

j=1 R j = ∪ j=1 Q j and disjoints rectangles R j ⊂ R̃ j
such that
m(E4 ∪M j=1 R j ) < 2ε.
Thus,
M
f (x) = ∑ χR j ,
j=1
possibly except in a set of measure < 2ε. Hence, for each k > 0, there exists a step function ϕk such that
m(Ek ) < 2−k where
Ek = {x ∈ Rn | f (x) =
6 ϕk (x)}.
Defining
Fk = ∪∞j=k+1 E j
we have
∞
m(Fk ) ≤ ∑ m(E j )
j=k+1
∞
≤ ∑ 2− j
j=k+1
2−(k+1)
=
1 − 1/2
= 2−k . (8.17)
Therefore also defining

F = ∩∞
k=1 Fk
we have m(F) = 0 considering that

m(F) ≤ 2−k , ∀k ∈ N.
ϕk (x) → f (x), ∀x ∈ F c .
Theorem 8.6.10 (Egorov) Suppose that { fk } is a sequence of measurable functions defined in a measurable
set E such that m(E) < ∞. Assume that fk → f , a.e in E. Thus given ε > 0 we may find a closed set Aε ⊂ E
such that fk → f uniformly in Aε and m(E − Aε ) < ε.
Proof 8.19 Without losing generality we may assume that
fk → f , ∀x ∈ E.
For each N, k ∈ N define

EkN = {x ∈ E | | f j (x) − f (x)| < 1/N, ∀ j ≥ k}.
Fixing N ∈ N, we may observe that
EkN ⊂ Ek+1
N
,
and that ∪∞ N
k=1 Ek = E. Thus we may obtain kN such that
1
m(E − EkNN ) < .
2N
Observe that
1
| f j (x) − f (x)| < , ∀ j ≥ kN , x ∈ EkNN .
N
Choose M ∈ N such that
∞
ε
∑ 2−k ≤ 2 .
k=M
Define
N
Ãε = ∩∞
N ≥M EkN .
Thus,
∞
ε
m(E − Ãε ) ≤ ∑ m(E − EkNN ) < .
N=M 2
Suppose given δ > 0. Let N ∈ N be such that N > M and 1/N < δ . Thus, if x ∈ Ãε then x ∈ EkNN so that
| f j (x) − f (x)| < δ , ∀ j > kN .
Hence, fk → f uniformly in Ãε . Observe that Ãε is measurable and thus there exists a closed set Aε ⊂ Ãε
such that
ε
m(Ãε − Aε ) < .
2
That is:
ε ε
m(E − Aε ) ≤ m(E − Ãε ) + m(Ãε − Aε ) < + = ε,
2 2
and
fk → f
uniformly in Aε . The proof is complete.
Definition 8.6.11 We say that f : Rn → [−∞, +∞] ∈ L1 (Rn if f is measurable and

Z
| f | dx < ∞.
Rn
Definition 8.6.12 We say that a set A ⊂ L1 (Rn ) is dense in L1 (Rn ), if for each f ∈ L1 (Rn ) and each ε > 0
there exists g ∈ A such that Z
k f − gkL1 (Rn ) = | f − g| dx < ε.
Rn
Theorem 8.6.13 About dense sets in L1 (Rn ) we have:

1. The set of simple functions is dense in L1 (Rn ).
2. The set of step functions is dense in L1 (Rn ).
3. the set of continuous functions with compact support is dense in L1 (Rn ).
Proof 8.20
1. From the last theorems given f ∈ L1 (Rn ) there exists a sequence of simple functions such that
ϕk (x) → f (x) a.e. in Rn .
Since {ϕk } may be also such that

|ϕk | ≤ | f |, ∀k ∈ N
from the Lebesgue dominated converge theorem, we have
kϕk − f kL1 (Rn ) → 0,
as k → ∞.
2. From the last item , it suffices to show that simple functions may be approximated by step functions.
As a simple function is a linear combination of characteristic functions of sets of finite measure, it
suffices to prove that given ε > 0 and a set of finite measure, there exists ϕ a step function such that
kχE − ϕkL1 (Rn ) < ε.
This may be made similar as in Theorem 8.6.9.

3. From the last item, it suffices to establish the result as f is a characteristic function of a rectangle
in Rn . First consider the case of a interval [a, b]. we may approximate f = χ[a,b] by g(x), where g is
continuous, and linear on (a − ε, a) e (b, b + ε) and

1, if a ≤ x ≤ b,
g(x) =
0, if x ≤ a − ε or x ≥ b + ε.
Thus,
k f − gkL1 (Rn ) < 2ε.
for the general case of a rectangle in Rn , we just recall that in this case f is the product of the charac-
teristic functions of n intervals. Therefore we may approximate f by the product of n functions similar
to g defined above.
Chapter 9
Other Topics in Measure and

Integration
In this chapter we present some important results which may be found in similar form at Chapters 2, 6 and 7
in the excellent book Real and Complex Analysis [68] by W. Rudin, where more details may be found.
9.1 Some preliminary results

In the next results µ is a measure on U. We start with the following theorem.
Theorem 9.1.1 Let f : U → [0, ∞] be a measurable function. If E ∈ M and
Z
f dµ = 0
E
then
f = 0, a.e. in E.
Proof 9.1 Define

An = {u ∈ E | f (u) > 1/n}, ∀n ∈ N.
Thus, Z Z
µ(An )/n ≤ f dµ ≤ f dµ = 0.
An E
Therefore, µ(An ) = 0, ∀n ∈ N.
Define
A = {u ∈ E | f (u) > 0}.
Hence,
A = ∪∞
n=1 An ,
so that µ(A) = 0.
Thus,
f = 0, a.e. in E.
Other Topics in Measure and Integration 179
Theorem 9.1.2 Assume f ∈ L1 (µ) and f dµ = 0, ∀E ∈ M . Under such hypotheses, f = 0, a.e. in U.

R
E
Proof 9.2 Consider first the case f : U → [−∞, +∞]. Define

An = {u ∈ U | f (u) > 1/n}, ∀n ∈ N.
Thus, Z
µ(An )/n ≤ f dµ = 0.
An
Hence, µ(An ) = 0, ∀n ∈ N.
Define
A = {u ∈ E | f (u) > 0}.
Therefore,
A = ∪∞
n=1 An ,
so that µ(A) = 0.
Thus,
f ≤ 0, a.e. in U.
By analogy we get
f ≥ 0, a.e. in U,
so that
f = 0, a.e. in U.
To complete the proof, just apply this last result to the real and imaginary parts of a complex f .
Theorem 9.1.3 Suppose µ(U) < ∞ and f ∈ L1 (µ). Moreover, assume

R
E | f | dµ
≤ α ∈ [0, ∞), ∀E ∈ M .
µ(E)
Under such hypothesis we have
| f | ≤ α, a.e. in U.
Proof 9.3 Define

An = {u ∈ U | | f (u)| > α + 1/n}, ∀n ∈ N.
Thus, if µ(An ) > 0 we get
R R
An (| f | − α) dµ An | f | dµ
1/n ≤ = − α ≤ 0,
µ(An ) µ(An )
a contradiction. Hence, µ(An ) = 0, ∀n ∈ N.
Define
A = {u ∈ U | | f (u)| > α}.
Therefore,
A = ∪∞
n=1 An ,
so that µ(A) = 0.
Thus,
| f (u)| ≤ α, a.e. in U.
At this point we present some preliminary results to the development of the well known Urysohn’s lemma.
Theorem 9.1.4 Let U be a Hausdorff space and K ⊂ U compact. Let v ∈ K c . Then there exists open sets V
and W ⊂ U such that v ∈ V , K ⊂ W and V ∩W = 0/.
Proof 9.4 For each u ∈ K there exists open sets Wu ,Vvu ⊂ U such that u ∈ Wu , v ∈ Wvu and Wu ∩Vvu = 0/.
Observe that K ⊂ ∪u∈K Wu so that, since K is compact there exist u1 , u2 , ..., un ∈ K such that
K ⊂ ∪ni=1Wui .
Finally, defining the open sets

V = ∩ni=1Vvui ,
and
W = ∪ni=1Wui ,
we get
V ∩W = 0/,
v ∈ V and K ⊂ W.
Theorem 9.1.5 Let {Kα , α ∈ L} be a collection of compact subsets of a Hausdorff space U.

Assume ∩α ∈L Kα = 0/. Under such hypotheses some finite sub-collection of {Kα , α ∈ L} has empty inter-
sections.
Proof 9.5 Define Vα = Kαc , ∀α ∈ L. Fix α0 ∈ L. From the hypotheses
Kα0 ∩ [∩α∈L\{α0 } Kα ] = 0/.
Hence,
Kα0 ⊂ [∩α∈L\{α0 } Kα ]c ,
that is,
Kα0 ⊂ ∪α ∈L\{α0 } Kαc = ∪α ∈L\{α0 }Vα .
Since Kα0 is compact, there exists α1 , ..., αn ∈ L such that
Kα0 ⊂ Vα1 ∪ ... ∪Vαn = (Kα1 ∩ ... ∩ Kαn )c ,
so that,
Kα0 ∩ Kα1 ∩ ... ∩ Kαn = 0/.
Definition 9.1.6 We say that a space U is locally compact if each u ∈ U has a neighborhood whose closure
is compact.
Theorem 9.1.7 Let U be a locally compact Hausdorff space. Suppose W ⊂ U is open and K ⊂ W , where K
is compact. Then there exists an open set V ⊂ U whose closure is compact and such that
K ⊂ V ⊂ V ⊂ W.
Proof 9.6 Let u ∈ K. Since U is locally compact there exists an open Vu ⊂ U such that u ∈ Vu and V u is
compact.
Observe that
K ⊂ ∪u∈K Vu
and since K is compact there exists u1 , u2 , .., un ∈ K such that
K ⊂ ∪nj=1Vu j .
Hence, defining, G = ∪nj=1Vu j we get

K⊂G
where G is compact.
If W = U define V = G and the proof would be complete.
Otherwise, if W 6= U define C = U \W . From Theorem 9.1.4, for each v ∈ C there exists an open Wv such
that K ⊂ Wv and v 6∈ W v .
Hence, {C ∩ G ∩W v : v ∈ C} is a collection of compact sets with empty intersection.
From Theorem 9.1.5 there are points v1 , .., vn ∈ C such that
C ∩ G ∩W v1 ∩ ... ∩W vn = 0/.
Defining
V = G ∩Wv1 ∩ ... ∩Wvn
we obtain
V ⊂ G ∩W v1 ∩ ... ∩W vn .
Also,
K ⊂ V ⊂ V ⊂ W.
Definition 9.1.8 Let f : U → [−∞, +∞] be a function on a topological space U.

We say that f is lower semi-continuous if Aα = {u ∈ U : f (u) > α} is open for all α ∈ R. Similarly, we
say that f is upper semi-continuous if Bα = {u ∈ U : f (u) < α} is open for all α ∈ R.
Observe that from this last definition f is continuous if and only if, it is both lower and upper semi-
continuous.
Here we state and prove a very important result namely, the Uryshon’s lemma.
Lemma 9.1.9 (Urysohn’s lemma) Assume U is a locally compact Hausdorff space, V ⊂ U is an open set
which contains a compact set K. Under such assumptions, there exists a function f ∈ Cc (V ) such that
0 ≤ f (u) ≤ 1, ∀u ∈ V,
f (u) = 1, ∀u ∈ K.
Proof 9.7 Set r1 = 0 and r2 = 1, and let r3 , r4 , r5 , ... be an enumeration of the rational numbers in (0, 1).
Observe that we may find open sets V0 and V1 such that V 0 is compact and
K ⊂ V1 ⊂ V 1 ⊂ V0 ⊂ V 0 ⊂ V.
Reasoning by induction, suppose n ≥ 2 and that Vr1 , ...,Vrn have been chosen so that if ri < r j then V r j ⊂ Vri .
Denote
ri = max{rk | k ∈ {1, ..., n} and rk < rn+1 },
and
r j = min{rk , | k ∈ {1, ..., n} and rk > rn+1 }.
We may find again an open set Vrn+1 such that
V r j ⊂ Vrn+1 ⊂ V rn+1 ⊂ Vri .

Thus, we have obtained a sequence Vr of open sets such that for every r rational in (0, 1) V r is compact and
if s > r then V s ⊂ Vr . Define,
r, if u ∈ Vr ,
fr (u) =
0, otherwise,
and
1, if u ∈ V s ,
gs (u) =
s, otherwise.
Also define
f (u) = sup fr (u), ∀u ∈ V,
r∈Q∩(0,1)
and
g(u) = inf gs (u), ∀u ∈ V.
s∈Q∩(0,1)
Observe that f is lower semi-continuous and g is upper semi-continuous. Moreover,
0 ≤ f ≤ 1,
and
f = 1, if u ∈ K.
Observe also that the support of f is contained in V 0
To complete the proof, it suffices to show that
f = g.
The inequality
fr (u) > gs (u)
is possible only if r > s, u ∈ Vr and u 6∈ V s .
But if r > s then Vr ⊂ Vs , and hence fr ≤ gs , ∀r, s ∈ Q ∩ (0, 1), so that f ≤ g. Suppose there exists u ∈ V
such that
f (u) < g(u).
Thus there exists rational numbers r, s such that
f (u) < r < s < g(u).
Since f (u) < r, u 6∈ Vr . Since g(u) > s, u ∈ V s .

As V s ⊂ Vr , we have a contradiction. Hence, f = g, and such a function is continuous.
Theorem 9.1.10 [Partition of unity] Let U be a locally compact Hausdorff space. Assume K ⊂ U is compact
so that
K ⊂ ∪ni=1Vi ,
where Vi is open ∀i ∈ {1, ...n}. Under such hypotheses, there exists functions h1 , ..., hn such that
n
∑ hi = 1, on K,
i=1
hi ∈ Cc (Vi ) and 0 ≤ hi ≤ 1, ∀i ∈ {1, ..., n}.

Proof 9.8 Let u ∈ K ⊂ ∪ni=1Vi . Thus there exists j ∈ {1, ..., n} such that u ∈ V j . We may select an open set
Wu such that u ∈ Wu , W u is compact and W u ⊂ V j .
Observe that
K ⊂ ∪u∈K Wu .
From this, since K is compact, there exist u1 , ..., uN such that
K ⊂ ∪Nj=1Wu j .
For each i ∈ {1, ..., n} define by W̃i the union of those Wu j , contained in Vi .
By the Uryshon’s lemma we may find continuous functions gi such that
gi = 1, on W̃i ,
gi ∈ Cc (Vi ),
0 ≤ gi ≤ 1, ∀i ∈ {1, ..., n}.
Define,
h1 = g1
h2 = (1 − g1 )g2
h3 = (1 − g1 )(1 − g2 )g3
... .................
hn = (1 − g1 )(1 − g2 )...(1 − gn−1 )gn . (9.1)
Thus,
0 ≤ hi ≤ 1 and hi ∈ Cc (Vi ), ∀i ∈ {1, .., n}.
Furthermore, by induction, we may obtain
h1 + h2 + ... + hn = 1 − (1 − g1 )(1 − g2 )...(1 − gn ).
Finally, if u ∈ K then u ∈ W̃i for some i ∈ {1, .., n}, so that gi (u) = 1, and hence
(h1 + ... + hn )(u) = 1, ∀u ∈ K.
The set {h1 , ..., hn } is said to be a partition of unity on K subordinate to the open cover {V1 , ...,Vn }.
9.2 The Riesz representation theorem

In the next lines we introduce the main result in this section, namely, the Riesz representation theorem.
Theorem 9.2.1 (Riesz representation theorem) Let U be a locally compact Hausdorff space and let F be
a positive linear functional on Cc (U). Then there exists a σ -algebra M in U which contains all the Borel
sets and there exists a unique positive measure µ on M such that
R
1. F( f ) = U f dµ, ∀ f ∈ Cc (U).
2. µ(K) < ∞, for every compact K ⊂ U.
3. µ(E) = inf{µ(V ) | E ⊂ V, V open }, ∀E ∈ M

4. µ(E) = sup{µ(K) | K ⊂ E, K compact} holds for all open E and all E ∈ M such that µ(E) < ∞.
5. If E ∈ M , A ⊂ E and µ(E) = 0 then A ∈ M .
Proof 9.9 We start by proving the uniqueness of µ. If µ satisfies (3) and (4), then µ is determined by
its values on compact sets. Then, if µ1 and µ2 are two measures for which the theorem holds, to prove
uniqueness, it suffices to show that
µ1 (K) = µ2 (K)
for every compact K ⊂ U. Let ε > 0. Fix a compact K ⊂ U. By (2) and (3), there exists an open V ⊃ K such
that
µ2 (V ) < µ2 (K) + ε.
By the Urysohn’s lema, there exists a f ∈ Cc (V ) such that
0 ≤ f (u) ≤ 1, ∀u ∈ V,
and
f (u) = 1, ∀u ∈ K.
Thus,
Z
µ1 (K) = χK dµ1
ZU
≤ f dµ1
U
= F( f )
Z
= f dµ2
ZU
≤ χV dµ2
U
= µ2 (V )
< µ2 (K) + ε. (9.2)
Since ε > 0 is arbitrary, we get,

µ1 (K) ≤ µ2 (K).
Interchanging the roles of µ1 and µ2 we similarly obtain
µ2 (K) ≤ µ1 (K),
so that
µ1 (K) = µ2 (K).
The proof of uniqueness is complete.
Now for every open V ⊂ U, define
µ(V ) = sup{F( f ) | f ∈ Cc (V ) and 0 ≤ f ≤ 1}.
If V1 ,V2 are open and V1 ⊂ V2 , then

µ(V1 ) ≤ µ(V2 ).
Hence,
µ(E) = inf{µ(V ) | E ⊂ V, V open},
if E is an open set. Define,

µ(E) = inf{µ(V ) | E ⊂ V, V open},
∀E ⊂ U. Define by MF the collection of all E ⊂ U such that µ(E) < ∞ and
µ(E) = sup{µ(K) | K ⊂ E, K compact}.
Finally, define by M the collection of all sets such that E ⊂ U and E ∩ K ∈ MF for all compact K ⊂ U. Since
µ(A) ≤ µ(B),
if A ⊂ B we have that µ(E) = 0 implies E ∩ K ∈ MF for all K compact, so that E ∈ M . Thus, (5) holds and
so does (3) by definition.
Observe that if f ≥ 0 then F( f ) ≥ 0, that is if f ≤ g then F( f ) ≤ F(g).
Now we prove that if {En } ⊂ U is a sequence then
∞
µ (∪∞
n=1 En ) ≤ ∑ µ(En ). (9.3)
n=1
First we show that

µ(V1 ∪V2 ) ≤ µ(V1 ) + µ(V2 ),
if V1 ,V2 are open sets.
Choose g ∈ Cc (V1 ∪V2 ) such that
0 ≤ g ≤ 1.
By Theorem 9.1.10 there exist functions h1 and h2 such that hi ∈ Cc (Vi ) and
0 ≤ hi ≤ 1
and so that h1 + h2 = 1 on the support of g. Hence hi ∈ Cc (Vi ), 0 ≤ hi g ≤ 1, and g = (h1 + h2 )g and thus
F(g) = F(h1 g) + F(h2 g) ≤ µ(V1 ) + µ(V2 ).
Since g is arbitrary, from the definition of µ we obtain
µ(V1 ∪V2 ) ≤ µ(V1 ) + µ(V2 ).
Furthermore, if µ(En ) = ∞, for some n ∈ N, then (9.3) is obviously valid. Assume then µ(En ) < ∞, ∀n ∈ N.
Let a not relabeled ε > 0. Therefore for each n ∈ N there exists an open Vn ⊃ En such that
ε
µ(Vn ) < µ(En ) + .
2n
Define
V = ∪∞
n=1Vn ,
and choose f ∈ Cc (V ) such that 0 ≤ f ≤ 1. Since the support of f is compact, there exists N ∈ N such that
spt( f ) ⊂ ∪Nn=1Vn .
Therefore,
µ ∪Nn=1Vn

F( f ) ≤
N
≤ ∑ µ(Vn )
n=1
∞
≤ ∑ µ(En ) + ε. (9.4)
n=1
Since this holds for any f ∈ Cc (V ) with 0 ≤ f ≤ 1 and ∪∞

n=1 En ⊂ V, we get
∞
µ (∪∞
n=1 En ) ≤ µ(V ) ≤ ∑ µ(En ) + ε.
i=1
Since ε > 0 is arbitrary, we have proven (9.3).

In the next lines we prove that if K is compact then K ∈ MF and
µ(K) = inf{F( f ) | f ∈ Cc (U), f = 1 on K}. (9.5)
For, if f ∈ Cc (U), f = 1 on K, and 0 < α < 1 define
Vα = {u ∈ U | f (u) > α}.
Thus, K ⊂ Vα and if g ∈ Cc (Vα ) and 0 ≤ g ≤ 1 we get
αg ≤ f .
Hence,
µ(K) ≤ µ(Vα )
= sup{F(g) | g ∈ Cc (Vα ), 0 ≤ g ≤ 1}
≤ α −1 F( f ). (9.6)
Letting α → 1 we obtain
µ(K) ≤ F( f ).
Thus, µ(K) < ∞, and obviously K ∈ MF .
Also there exists an open V ⊃ K such that
µ(V ) < µ(K) + ε.
By the Urysohn’s lemma, we may find f ∈ Cc (V ) such that f = 1 on K and 0 ≤ f ≤ 1. Thus,
F( f ) ≤ µ(V ) < µ(K) + ε.
Since ε > 0 is arbitrary, (9.5) holds.

At this point we prove that for every open V we have
µ(V ) = sup{µ(K) | K ⊂ V, K compact} (9.7)
and hence MF contains every open set such that µ(V ) < ∞.
Let V ⊂ U be an open set such that µ(V ) < ∞.
Let α ∈ R be such that α < µ(V ). Therefore there exists f ∈ Cc (V ) such that 0 ≤ f ≤ 1 and such that
α < F( f ).
If W ⊂ U is an open set such that K = spt( f ) ⊂ W, we have that f ∈ Cc (W ) and 0 ≤ f ≤ 1 so that
F( f ) ≤ µ(W ).
Thus, since W ⊃ K is arbitrary we obtain

F( f ) ≤ µ(K),
so that
α < µ(K),
where K ⊂ U is a compact set.
Hence (9.7) holds.

Suppose that
E = ∪∞
n=1 En ,
where {En } is a sequence of disjoint sets in MF .

∞
µ(E) = ∑ µ(En ). (9.8)
n=1
In addition, if µ(E) < ∞ then also E ⊂ MF .

First we show that if K1 , K2 ⊂ U are compact disjoint sets then,
µ(K1 ∪ K2 ) = µ(K1 ) + µ(K2 ). (9.9)
From the Uryshon’s lemma there exists f ∈ Cc (U) such that f = 1 on K1 , f = 0 on K2 and,
0 ≤ f ≤ 1.
From (9.5) there exists g ∈ Cc (U) such that g = 1 on K1 ∪ K2 and
F(g) < µ(K1 ∪ K2 ) + ε.
Observe that f g = 1 on K1 and (1 − f )g = 1 on K2 and also f g, (1 − f )g ∈ Cc (U) and 0 ≤ f g ≤ 1 and

0 ≤ (1 − f )g ≤ 1 so that
µ(K1 ) + µ(K2 ) ≤ F( f g) + F((1 − f )g)

= F(g)
≤ µ(K1 ∪ K2 ) + ε. (9.10)
Since ε > 0 is arbitrary we obtain

µ(K1 ) + µ(K2 ) ≤ µ(K1 ∪ K2 ).
From this (9.9) holds.
Also, if µ(E) = ∞, (9.8) follows from (9.3).
Thus assume µ(E) < ∞.
Since En ∈ MF , ∀n ∈ N we may obtain compact sets Hn ⊂ En such that
ε
µ(Hn ) > µ(En ) − , ∀n ∈ N.
2n
Defining KN = ∪Nn=1 Hn , by (3) we get
µ(E) ≥ µ(KN )
N
= ∑ µ(Hn )
n=1
N
≥ ∑ µ(En ) − ε, ∀N ∈ N. (9.11)
n=1
Since N ∈ N and ε > 0 are arbitrary we get,

∞
µ(E) ≥ ∑ µ(En ).
n=1

∞
µ(E) = ∑ µ(En ). (9.12)
n=1
Let ε0 > 2ε. If µ(E) < ∞, there exists N0 ∈ N such that µ(KN0 ) > ∑∞
n=1 µ(En ) − ε0 .
µ(E) ≤ µ(KN0 ) + ε0 .
Therefore, since ε > 0 and ε0 > 2ε are arbitrary we may conclude that E satisfies (4) so that E ∈ MF .
Now we prove the following.
If E ∈ MF there is a compact K ⊂ U and an open V ⊂ U such that K ⊂ E ⊂ V and
µ(V \ K) < ε.
For, from above, there exists a compact K and an open V such that
K ⊂E ⊂V
and
ε ε
µ(V ) − < µ(E) < µ(K) + .
2 2
Since V \ K is open and of finite measure, it is in MF . From the last chain of inequalities we obtain
µ(K) + µ(V \ K) = µ(V ) < µ(K) + ε,
so that
µ(V \ K) < ε.
In the next lines we prove that if A, B ∈ MF then
A \ B, A ∪ B and A ∩ B ∈ MF .
By above there exist compact sets K1 , K2 and open sets V1 ,V2 such that
K1 ⊂ A ⊂ V1 , K2 ⊂ B ⊂ V2
and
µ(Vi \ Ki ) < ε, ∀i ∈ {1, 2}.
Since
(A \ B) ⊂ (V1 \ K2 ) ⊂ (V1 \ K1 ) ∪ (K1 \V2 ) ∪ (V2 \ K2 ),
we get
µ(A \ B) < ε + µ(K1 \V2 ) + ε,
Since K1 \V2 ⊂ A \ B is compact and ε > 0 is arbitrary, we get,
A \ B ∈ MF .
Since
A ∪ B = (A \ B) ∪ B,
we obtain
A ∪ B ∈ MF .
Since
A ∩ B = A \ (A \ B)
we get
A ∩ B ∈ MF .
At this point we prove that M is a σ -algebra in U which contains all the Borel sets.
Let K ⊂ U be a compact subset. If A ∈ M then
Ac ∩ K = K \ (A ∩ K),
so that Ac ∩ K ∈ MF considering that K ∈ MF and A ∩ K ∈ MF .

Thus if A ∈ M then Ac ∈ M .
Next suppose
A = ∪∞n=1 An
where An ∈ M , ∀n ∈ N.
Define B1 = A1 ∩ K and
Bn = (An ∩ K) \ (B1 ∪ B2 ∪ ... ∪ Bn−1 ),
∀n ≥ 2, n ∈ N.
Then {Bn } is disjoint sequence of sets in MF .
Thus,
n=1 Bn ∈ MF .
A ∩ K = ∪∞
Hence A ∈ M . Finally, if C ⊂ U is a closed subset, then C ∩ K is compact, so that C ∩ K ∈ MF . Hence
C ∈ M.
Therefore, M is a σ -algebra which contains the closed sets, so that it contains the Borel sets.
Finally, we will prove that
MF = {E ∈ M | µ(E) < ∞}.
For, if E ∈ MF then E ∩ K ∈ MF for all compact K ⊂ U, hence E ∈ M .
Conversely, assume E ∈ M and µ(E) < ∞. There is an open V ⊃ E such that µ(V ) < ∞. Pick a compact
K ⊂ V such that
µ(V \ K) < ε.
Since E ∩ K ∈ MF there is a compact K1 ⊂ (E ∩ K) such that
µ(E ∩ K) < µ(K1 ) + ε.
Since
E ⊂ (E ∩ K) ∪ (V \ K),
it follows that
µ(E) ≤ µ(E ∩ K) + µ(V \ K) < µ(K1 ) + 2ε.
This implies E ∈ MF .
To finish the proof, we show that
Z
F( f ) = f dµ, ∀ f ∈ Cc (U)
U
.
From linearity it suffices to prove the result for the case where f is real.
Let f ∈ Cc (U). Let K be the support of f and let [a, b] ⊂ R be such that
R( f ) ⊂ (a, b),
where R( f ) denotes the range of f .

Suppose given a not relabeled ε > 0. Choose a partition of [a, b] denoted by
{yi } = {a = y0 < y1 < y2 < .. < yn = b},
such that yi − yi−1 < ε, ∀i ∈ {1, ..., n}.

Denote
Ei = {u ∈ K | yi−1 < f (u) ≤ yi },
∀i ∈ {1, ..., n}.
Since f is continuous, it is Borel measurable, and the sets Ei are disjoint Borel ones such that
∪ni=1 Ei = K.
Select open sets Vi ⊃ Ei such that

ε
µ(Vi ) < µ(Ei ) + , ∀i ∈ {1, ..., n},
n
and such that
f (u) < yi + ε, ∀u ∈ Vi .
From Theorem 9.1.10 there exists a partition of unity subordinate to {Vi }ni=1 such that hi ∈ Cc (Vi ), 0 ≤
hi ≤ 1 and
n
∑ hi = 1, on K.
i=1
Hence,
n
f = ∑ hi f ,
i=1
and !
n n
µ(K) ≤ F ∑ hi f = ∑ F(hi f ).
i=1 i=1
Observe that
ε
µ(Ei ) + > µ(Vi )
n
= sup{F( f ) | f ∈ Cc (Vi ), 0 ≤ f ≤ 1}
> F(hi ), ∀i ∈ {1, .., n}. (9.13)
Thus,
n
F( f ) = ∑ F(hi f )
i=1
n
≤ ∑ F(hi (yi−1 + 2ε))
i=1
n
= ∑ (yi−1 + 2ε)F(hi )
i=1
n ε
< ∑ (yi−1 + 2ε) µ (Ei ) +
i=1 n
n n n
ε
< ∑ y i −1 µ(Ei ) + ∑ (yi−1 )
n
+ 2ε ∑ µ(Ei ) + 2ε 2
i=1 i=1 i=1
Z
< f dµ + bε + 2ε µ(K) + 2ε 2 . (9.14)
U
Since ε > 0 is arbitrary, we obtain

Z
F( f ) ≤ f dµ, ∀ f ∈ Cc (U).
U
From this Z
F(− f ) ≤ (− f ) dµ, ∀ f ∈ Cc (U),
U
that is, Z
F( f ) ≥ f dµ, ∀ f ∈ Cc (U).
U
Hence, Z
F( f ) = f dµ, ∀ f ∈ Cc (U).
U
9.3 The Lebesgue points

In this section we introduce a very important concept in analysis namely, the definition of Lebesgue points.
We recall that in Rn the open ball with center u and radius r is defined by
Br (u) = {v ∈ Rn | |v − u|2 < r}.
Consider a Borel measure µ on Rn . We may associate to µ, the function Frµ (u), denoted by
µ(Br (u))
Frµ (u) = ,
m(Br (u))
where m denotes the Lebesgue measure.
We define the symmetric derivative of µ at u, by (Dµ)(u), by
(Dµ)(u) = lim Frµ (u),

r→0
whenever such a limit exists.

We also define the function Gµ for a positive measure µ by
Gµ (u) = sup Frµ (u).

0<r<∞
The function Gµ : Rn → [0, +∞], is lower semi-continuous and hence measurable.

Lemma 9.3.1 Let W = ∪Ni=1 Bri (ui ) be a finite union of open balls. Then there is a set S ⊂ {1, 2, ..., N} such
that
1. The balls Bri (ui ), i ∈ S are disjoint.
2. W ⊂ ∪i∈S B3ri (ui ).
Proof 9.10 Let us first order the balls Bri (ui ) so that
r1 ≥ r2 ≥ ... ≥ rN .
Set i1 = 1, and discard all balls such that

Bi1 ∩ B j =
6 0/.
Let Bi2 be the first of the remaining balls, if any. Discard all B j such that j > i2 and Bi2 ∩ B j 6= 0/.
Let Bi3 the first of the remaining balls as long as possible. Such a process stops after a finite number of
steps. Define S = {i1 , i2 , ...}. It is clear that (1) holds. Now we prove that each discarded B j is contained in
{B3ri , i ∈ S}.
For, just observe that if r0 < r and Br0 (u0 ) intersects Br (u) then Br0 (u0 ) ⊂ B3r (u).
Theorem 9.3.2 Suppose µ is a finite Borel measure on Rn and λ > 0. Then
m(Aλ ) ≤ 3n λ −1 kµk,
where
Aλ = {u ∈ U | Gµ (u) > λ },
and
kµk = |µ|(Rn ).
Proof 9.11 Let K be a compact subset of the open set Aλ .

As Gµ (u) = sup0<r<∞ {Frµ (u)}, each u ∈ K is the center of an open ball Bu such that
µ(Bu ) > λ m(Bu ).
Since K is compact, there exists a finite number of such balls which covers K. By Lemma 9.3.1, there
exists a disjoint sub-collection here denoted by {Br1 , ..., BrN } such that K ⊂ ∪Nk=1 B3rk , so that
N
m(K) ≤ 3n ∑ m(Brk )
k=1
N
≤ 3n λ −1 ∑ |µ|(Brk )
k=1
n −1
≤ 3 λ kµk. (9.15)
The result follows taking the supremum relating all compact K ⊂ Aλ .
Remark 9.3.3 Observe that, if f ∈ L1 (Rn ) and λ > 0, for Aλ = {u ∈ Rn | | f | > λ }, we have
m(Aλ ) ≤ λ −1 k f k1 .
This follows from the fact that

Z Z
λ m(Aλ ) ≤ | f | dm ≤ | f | dm = k f k1 .
Aλ Rn
Observe also that defining dη = | f | dm, for every λ > 0, defining

η(Br (u))
G f (u) = sup ,
0<r<∞ m(Br (u))
and
Aλ = {u ∈ U | G f (u) > λ },
we have
m(Aλ ) ≤ 3n λ −1 k f k1 .
9.3.1 The main result on Lebesgue points

Finally in this section we present the main definition of Lebesgue points and some relating results.
Definition 9.3.4 Let f ∈ L1 (Rn ). A point u ∈ L1 (Rn ) such that
1
Z
lim | f (v) − f (u)| dm(v) = 0,
r→0 m(Br (u)) Br (u)
is called a Lebesgue point of f .
Theorem 9.3.5 If f ∈ L1 (Rn ), then almost all u ∈ Rn is a Lebesgue point of f .
Proof 9.12 Define

1
Z
Hr f (u) = | f − f (u)| dm, ∀u ∈ Rn , r > 0,
m(Br (u)) Br (u)
also define
H f (u) = lim sup Hr f (u).
r→0
We have to show that H f = 0, a.e. [m].

Select y > 0 and fix k ∈ N. Observe that there exists g ∈ C(Rn ) such that
k f − gk1 < 1/k.
Define h = f − g. Since g is continuous, Hg = 0 in Rn . Observe that
1
Z
Hrh (u) = |h − h(u)| dm
m(Br (u)) Br (u)
1
Z
≤ |h| dm + |h(u)|, (9.16)
m(Br (u)) Br (u)
so that
Hh < Gh + |h|.
Since
Hr f ≤ Hrg + Hrh ,
we obtain
H f ≤ Gh + |h|.
Define
Ay = {u ∈ Rn | H f (u) > 2y},
By,k = {u ∈ Rn | Gh (u) > y}
and
Cy,k = {u ∈ Rn | |h| > y}.
Observe that khk1 < 1/k, so that from remark 9.3.3 we obtain
3n
m(By,k ) ≤ ,
yk
and
1
m(Cy,k ) ≤ ,
yk
and hence
3n + 1
m(By,k ∪Cy,k ) ≤ ,
yk
Therefore
3n + 1
m(Ay ) ≤ m(By,k ∪Cy,k ) ≤ .
yk
Since k is arbitrary, we get m(Ay ) = 0, ∀y > 0 so that m{u ∈ Rn | H f (u) > 0} = 0.
We finish this section with the following result.
Theorem 9.3.6 Suppose µ is a complex Borel measure on Rn such that µ m. Suppose f is the Radon-
Nikodym derivative of µ with respect to m. Under such assumptions,
Dµ = f , a.e. [m],
and Z
µ(E) = Dµ dm,
E
for all Borel set E ⊂ Rn .
Proof 9.13 From the Radon-Nikodym theorem we have

Z
µ(E) = f dm,
E
for all measurable set E ⊂ Rn .

Observe that at any Lebesgue point u of f we have
1
Z
f (u) = lim f dm
r→0 m(Br (u)) Br (u)
µ(Br (u))
= lim
r→0 m(Br (u))
= Dµ(u). (9.17)

Chapter 10
Distributions
The main reference for this chapter is Rudin [67].
10.1 Basic definitions and results

Definition 10.1.1 (Test Functions, the Space D(Ω)) Let Ω ⊂ Rn be a nonempty open set. For each K ⊂ Ω
compact, consider the space DK , the set of all C∞ (Ω) functions with support in K. We define the space of test
functions, denoted by D(Ω) as
D(Ω) = ∪K⊂Ω DK , K compact. (10.1)
Thus φ ∈ D(Ω) if and only if φ ∈ C∞ (Ω) and the support of φ is a compact subset of Ω.
Definition 10.1.2 (Topology for D(Ω)) Let Ω ⊂ Rn be an open set.

1. For every K ⊂ Ω compact, σK denotes the topology which a local base is defined by {VN,k }, where
N, k ∈ N,
VN,k = {φ ∈ DK | kφ kN < 1/k} (10.2)
and
kφ kN = max{|Dα φ (x)| | x ∈ Ω, |α| ≤ N}. (10.3)
2. σ̂ denotes the collection of all convex balanced sets W ∈ D(Ω) such that W ∩ DK ∈ σK for every
compact K ⊂ Ω.
3. We define σ in D(Ω) as the collection of all unions of sets of the form φ + W , for φ ∈ D(Ω) and
W ∈ σ̂ .
Theorem 10.1.3 Concerning the last definition we have the following:

1. σ is a topology in D(Ω).
2. Through σ , D(Ω) is made into a locally convex topological vector space.
Proof 10.1
1. From item 3 of Definition 10.1.2, it is clear that arbitrary unions of elements of σ are elements of σ .
Let us now show that finite intersections of elements of σ also belongs to σ . Suppose V1 ∈ σ and
V2 ∈ σ , if V1 ∩ V2 = 0/ we are done. Thus, suppose φ ∈ V1 ∩ V2 . By the definition of σ there exist two
sets of indices L1 and L2 , such that
Vi = ∪λ ∈Li (φiλ + Wiλ ), for i = 1, 2, (10.4)
where, Wiλ ∈ σ̂ , ∀λ ∈ Li . Moreover, since φ ∈ V1 ∩ V2 there exist φi ∈ D(Ω) and Wi ∈ σ̂ such that
φ ∈ φi + Wi ⊂ Vi , for i = 1, 2. (10.5)
Thus there exists K ⊂ Ω such that φi ∈ DK for i ∈ {1, 2}. Since DK ∩ Wi ∈ σK , DK ∩ Wi is open in DK
so that from (10.5) there exists 0 < δi < 1 such that
φ − φi ∈ (1 − δi )Wi , for i ∈ {1, 2}. (10.6)
From (10.6) and from the convexity of Wi we have
φ − φi + δi Wi ⊂ (1 − δi )Wi + δi Wi = Wi (10.7)
so that
φ + δi Wi ⊂ φi + Wi ⊂ Vi , for i ∈ {1, 2}. (10.8)
Define Wφ = (δ1 W1 ) ∩ (δ2 W2 ) so that
φ + Wφ ⊂ Vi , (10.9)
and therefore we may write
V1 ∩ V2 = ∪φ ∈V1 ∩V2 (φ + Wφ ) ∈ σ . (10.10)

2. It suffices to show that single points are closed sets in D(Ω) and the vector space operations are
continuous.
(a) Pick φ1 , φ2 ∈ D(Ω) such that φ1 =
6 φ2 and define
V = {φ ∈ D(Ω) | kφ k0 < kφ1 − φ2 k0 }. (10.11)
Thus, V ∈ σ̂ and φ1 6∈ φ2 + V . As φ2 + V is open is contained D(Ω) \ {φ1 } and φ2 =

6 φ1 is
arbitrary, it follows that D(Ω) \ {φ1 } is open, so that {φ1 } is closed.
(b) The proof that addition is σ -continuous follows from the convexity of any element of σ̂ . Thus,
given φ1 , φ2 ∈ D(Ω) and V ∈ σ̂ we have
1 1
φ1 + V + φ2 + V = φ1 + φ2 + V . (10.12)
2 2
(c) To prove the continuity of scalar multiplication, first consider φ0 ∈ D(Ω) and α0 ∈ R. Then:
αφ − α0 φ0 = α(φ − φ0 ) + (α − α0 )φ0 . (10.13)

Distributions 197
For V ∈ σ̂ there exists δ > 0 such that δ φ0 ∈ 12 V . Let us define c = 1

2(|α0 |+δ ) . Thus if |α − α0 | <
2V . Let φ ∈ D(Ω) such that

1
δ then (α − α0 )φ0 ∈
1
φ − φ0 ∈ cV = V, (10.14)
2(|α0 | + δ )
so that
1
(|α0 | + δ )(φ − φ0 ) ∈ V . (10.15)
2
This means
1 1
α(φ − φ0 ) + (α − α0 )φ0 ∈ V + V = V . (10.16)
2 2
Therefore, αφ − α0 φ0 ∈ V whenever |α − α0 | < δ and φ − φ0 ∈ cV .
For the next result the proof may be found in Rudin [67].
Proposition 10.1.4 A convex balanced set V ⊂ D(Ω) is open if and only if V ∈ σ .
Proposition 10.1.5 The topology σK of DK ⊂ D(Ω) coincides with the topology that DK inherits from
D(Ω).
Proof 10.2 From Proposition 10.1.4 we have
V ∈ σ implies DK ∩ V ∈ σK . (10.17)
Now suppose V ∈ σK , we must show that there exists A ∈ σ such that V = A ∩ DK . The definition of σK
implies that for every φ ∈ V , there exist N ∈ N and δφ > 0 such that
{ϕ ∈ DK | kϕ − φ kN < δφ } ⊂ V . (10.18)
Define
Uφ = {ϕ ∈ D(Ω) | kϕkN < δφ }. (10.19)
Then Uφ ∈ σ̂ and
DK ∩ (φ + Uφ ) = φ + (DK ∩ Uφ ) ⊂ V . (10.20)
Defining A = ∪φ ∈V (φ + Uφ ), we have completed the proof.
The proof for the next two results may also be found in Rudin [67].
Proposition 10.1.6 If A is a bounded set of D(Ω) then A ⊂ DK for some K ⊂ Ω, and there are MN < ∞ such
that kφ kN ≤ MN , ∀φ ∈ A, N ∈ N.
Proposition 10.1.7 If {φn } is a Cauchy sequence in D(Ω), then {φn } ⊂ DK for some K ⊂ Ω compact, and
lim kφi − φ j kN = 0, ∀N ∈ N. (10.21)

i, j→∞
Proposition 10.1.8 If φn → 0 in D(Ω), then there exists a compact K ⊂ Ω which contains the support of
φn , ∀n ∈ N and Dα φn → 0 uniformly, for each multi-index α.
The proof follows directly from last proposition.

Theorem 10.1.9 Suppose T : D(Ω) → V is linear, where V is a locally convex space. Then the following
statements are equivalent.
1. T is continuous.
2. T is bounded.
3. If φn → θ in D(Ω) then T (φn ) → θ as n → ∞.
4. The restrictions of T to each DK are continuous.
Proof 10.3
1 ⇒ 2. This follows from Proposition 2.7.3 .
2 ⇒ 3. Suppose T is bounded and φn → 0 in D(Ω), by last proposition φn → 0 in some DK so that
{φn } is bounded and {T (φn )} is also bounded. Hence by Proposition 2.7.3, T (φn ) → 0 in V .
3 ⇒ 4. Assume 3 holds and consider {φn } ⊂ DK . If φn → θ then, by Proposition 10.1.5, φn → θ in
D(Ω), so that, by above T (φn ) → θ in V . Since DK is metrizable, also by proposition 2.7.3 we have
that 4 follows.
4 ⇒ 1. Assume 4 holds and let V be a convex balanced neighborhood of zero in V . Define U =
T −1 (V ). Thus U is balanced and convex. By Proposition 10.1.5, U is open in D(Ω) if and only
if DK ∩ U is open in DK for each compact K ⊂ Ω, thus if the restrictions of T to each DK are
continuous at θ , then T is continuous at θ , hence 4 implies 1.
Definition 10.1.10 (Distribution) A linear functional in D(Ω) which is continuous with respect to σ is said
to be a Distribution.
Proposition 10.1.11 Every differential operator is a continuous mapping from D(Ω) into D(Ω).
Proof 10.4 Since kDα φ kN ≤ kφ k|α |+N , ∀N ∈ N, Dα is continuous on each DK , so that by Theorem 10.1.9,
Dα is continuous on D(Ω).
Theorem 10.1.12 Denoting by D 0 (Ω) the dual space of D(Ω) we have that T : D(Ω) → R ∈ D 0 (Ω) if and
only if for each compact set K ⊂ Ω there exists an N ∈ N and c ∈ R+ such that
|T (φ )| ≤ ckφ kN , ∀φ ∈ DK . (10.22)
Proof 10.5 The proof follows from the equivalence of 1 and 4 in Theorem 10.1.9.
10.2 Differentiation of distributions

Definition 10.2.1 (Derivatives for Distributions) Given T ∈ D 0 (Ω) and a multi-index α, we define the Dα
derivative of T as
Dα T (φ ) = (−1)|α| T (Dα φ ), ∀φ ∈ D(Ω). (10.23)

Distributions 199
Remark 10.2.2 Observe that if |T (φ )| ≤ ckφ kN , ∀φ ∈ D(Ω) for some c ∈ R+ , then
|Dα T (φ )| ≤ ckDα φ kN ≤ ckφ kN+|α | , ∀φ ∈ D(Ω), (10.24)
thus Dα T ∈ D 0 (Ω). Therefore, derivatives of distributions are also distributions.
Theorem 10.2.3 Suppose {Tn } ⊂ D 0 (Ω). Let T : D(Ω) → R be defined by
T (φ ) = lim Tn (φ ), ∀φ ∈ D(Ω). (10.25)

n→∞
Then T ∈ D 0 (Ω), and
Dα Tn → Dα T in D 0 (Ω). (10.26)
Proof 10.6 Let K be an arbitrary compact subset of Ω. Since (10.25) holds for every φ ∈ DK , the principle
of uniform boundedness implies that the restriction of T to DK is continuous. It follows from Theorem 10.1.9
that T is continuous in D(Ω), that is, T ∈ D 0 (Ω). On the other hand
(Dα T )(φ ) = (−1)|α| T (Dα φ ) = (−1)|α | lim Tn (Dα φ ) = lim (Dα Tn (φ )), ∀φ ∈ D(Ω). (10.27)
n→∞ n→∞
10.3 Examples of distributions

10.3.1 First example
Let Ω ⊂ Rn be an open bounded set. As a first example of distribution consider the functional
T : D(Ω) → R
given by Z
T (φ ) = f φ dx,
Ω
where f ∈ L1 (Ω). Observe that

Z
|T (φ )| ≤ | f φ | dx
ZΩ
≤ | f | dxkφ k∞ , (10.28)
Ω
so that T is a bounded linear functional on D(Ω), that is, T is a distribution.
10.3.2 Second example

For the second example, define Ω = (0, 1), and T : D(Ω) → R by
T (φ ) = φ (1/2) + φ 0 (1/3).
Thus,
|T (φ )| = |φ (1/2) + φ 0 (1/3)| ≤ kφ k∞ + kφ 0 k∞ ≤ 2kφ k1 ,
so that T is also a distribution (bounded and linear).
10.3.3 Third example

For the third example, consider an open bounded Ω ⊂ Rn , and T : D(Ω) → R by
Z
T (φ ) = f φ dx,
Ω
where, f ∈ L1 (Ω).
Observe that the derivative of T for the multi-index α = (α1 , ..., αn ), is defined by
Z
Dα T (φ ) = (−1)|α| T (Dα φ ) = (−1)|α| f Dα φ dx.
Ω
If there exists g ∈ L1 (Ω), such that
Z Z
(−1)|α| f Dα φ dx = gφ dx, ∀φ ∈ D(Ω),
Ω Ω
we say that g is the derivative Dα of f in the distributional sense.
For example, for Ω = (0, 1) and f : Ω → R given by

0, if x ∈ [0, 1/2],
f (x) =
1, if x ∈ (1/2, 1],
and Z
T (φ ) = f φ dx
Ω
where φ ∈ Cc∞ (Ω) ,we have
dφ
Z
Dx T (φ ) = − f
dx
dx
Ω
Z 1
dφ
= − (1) dx
1/2 dx
= −φ (1) + φ (1/2) = φ (1/2), (10.29)
that is,
Dx T (φ ) = φ (1/2), ∀φ ∈ Cc∞ (Ω).
Finally, defining f : Ω → R by

x, if x ∈ [0, 1/2],
f (x) =
−x + 1, if x ∈ (1/2, 1],
and, Z
T (φ ) = f φ dx
Ω
where φ ∈ Cc∞ (Ω) we have
dφ
Z
Dx T (φ ) = − f dx
Ω dx
Z 1
dφ
= − f dx
0 dx
Z 1
= gφ dx, (10.30)
0
where
1, if x ∈ [0, 1/2],
g(x) =
−1, if x ∈ (1/2, 1].
In such a case we denote g = Dx f and say that g is the derivative of f in the distributional sense.
We emphasize that in this last example the classical derivative of f is not defined, since f is not differen-
tiable at x = 1/2.
Chapter 11
The Lebesgue and Sobolev

Spaces
Here we emphasize that the two main references for this chapter are Adams [2] and Evans [34]. We start with
the definition of Lebesgue spaces, denoted by L p (Ω), where 1 ≤ p ≤ ∞ and Ω ⊂ Rn is an open set. In this
chapter, integrals always refer to the Lebesgue measure.
11.1 Definition and properties of L p spaces

Definition 11.1.1 (L p spaces) For 1 ≤ p < ∞, we say that u ∈ L p (Ω) if u : Ω → R is measurable and
Z
|u| p dx < ∞. (11.1)
Ω
p dx]1/p
R
We also denote kuk p = [ Ω |u| and will show that k · k p is a norm.
Definition 11.1.2 (L∞ spaces) We say that u ∈ L∞ (Ω) if u is measurable and there exists M ∈ R+ , such that
|u(x)| ≤ M, a.e. in Ω. We define
kuk∞ = inf{M > 0 | |u(x)| ≤ M, a.e. in Ω}. (11.2)
We will show that k · k∞ is a norm. For 1 ≤ p ≤ ∞, we define q by the relations

 +∞, if p = 1,
p
q= , if 1 < p < +∞,
 p−1
1, if p = +∞,
so that symbolically we have
1 1
+ = 1.
p q
The next result is fundamental in the proof of the Sobolev Imbedding theorem.
¨
Theorem 11.1.3 (Holder inequality) Consider u ∈ L p (Ω) and v ∈ Lq (Ω), with 1 ≤ p ≤ ∞. Then uv ∈ L1 (Ω)
and
Z
|uv|dx ≤ kuk p kvkq . (11.3)
Ω
Proof 11.1 The result is clear if p = 1 or p = ∞. You may assume kuk p , kvkq > 0, otherwise the result is
also obvious. Thus suppose 1 < p < ∞. From the concavity of log function on (0, ∞) we obtain

1 p 1 q 1 1
log a + b ≥ log a p + log bq = log(ab). (11.4)
p q p q
Thus,
1 p 1
ab ≤ (a ) + (bq ), ∀a ≥ 0, b ≥ 0. (11.5)
p q
Therefore,
1 1
|u(x)||v(x)| ≤ |u(x)| p + |v(x)|q , a.e. in Ω. (11.6)
p q
Hence, |uv| ∈ L1 (Ω) and
1 1
Z
|uv|dx ≤ kuk pp + kvkqq . (11.7)
Ω p q
Replacing u by λ u in (11.7) λ > 0, we obtain
λ p−1 1
Z
|uv|dx ≤ kuk pp + kvkqq . (11.8)
Ω p λq
q/ p
For λ = kuk−1
p kvkq ¨
we obtain the Holder inequality.
The next step is to prove that k · k p is a norm.

Theorem 11.1.4 L p (Ω) is a vector space and k · k p is norm ∀p such that 1 ≤ p ≤ ∞.
Proof 11.2 The only non-trivial property to be proved concerning the norm definition, is the triangle
inequality. If p = 1 or p = ∞ the result is clear. Thus, suppose 1 < p < ∞. For u, v ∈ L p (Ω) we have
|u(x) + v(x)| p ≤ (|u(x)| + |v(x)|) p ≤ 2 p (|u(x)| p + |v(x)| p ), (11.9)
so that u + v ∈ L p (Ω). On the other hand,

Z
ku + vk pp = |u + v| p−1 |u + v|dx
Ω
Z Z
≤ |u + v| p−1 |u|dx + |u + v| p−1 |v|dx, (11.10)
Ω Ω
¨
and hence, from Holder’s inequality
p−1
ku + vk pp ≤ ku + vk p−1
p kuk p + ku + vk p kvk p , (11.11)
that is,
ku + vk p ≤ kuk p + kvk p , ∀u, v ∈ L p (Ω). (11.12)
Theorem 11.1.5 L p (Ω) is a Banach space for any p such that 1 ≤ p ≤ ∞.

The Lebesgue and Sobolev Spaces 203
Proof 11.3 Suppose p = ∞. Suppose {un } is Cauchy sequence in L∞ (Ω). Thus, given k ∈ N there exists
Nk ∈ N such that, if m, n ≥ Nk then
1
kum − un k∞ < . (11.13)
k
Therefore, for each k, there exist a set Ek such that m(Ek ) = 0, and
1
|um (x) − un (x)| < , ∀x ∈ Ω \ Ek , ∀m, n ≥ Nk . (11.14)
k
Observe that E = ∪∞k=1 Ek is such that m(E) = 0. Thus {un (x)} is a real Cauchy sequence at each x ∈ Ω \ E.
Define u(x) = lim un (x) on Ω \ E. Letting m → ∞ in (11.14) we obtain
n→∞
1
|u(x) − un (x)| < , ∀x ∈ Ω \ E, ∀n ≥ Nk . (11.15)
k
Thus, u ∈ L∞ (Ω) and kun − uk∞ → 0 as n → ∞.
Now suppose 1 ≤ p < ∞. Let {un } a Cauchy sequence in L p (Ω). We can extract a subsequence {unk }
such that
1
kunk+1 − unk k p ≤ , ∀k ∈ N. (11.16)
2k
To simplify the notation we write uk in place of unk , so that
1
kuk+1 − uk k p ≤ , ∀k ∈ N. (11.17)
2k
Defining
n
gn (x) = ∑ |uk+1 (x) − uk (x)|, (11.18)
k=1
we obtain
kgn k p ≤ 1, ∀n ∈ N. (11.19)
From the monotone convergence theorem and (11.19), gn (x) converges to a limit g(x) with g ∈ L p (Ω). On
the other hand, for m ≥ n ≥ 2 we have
|um (x) − un (x)| ≤ |um (x) − um−1 (x)| + ... + |un+1 (x) − un (x)| ≤ g(x) − gn−1 (x), a.e. in Ω. (11.20)
Hence, {un (x)} is Cauchy a.e. in Ω and converges to a limit u(x) so that
|u(x) − un (x)| ≤ g(x), a.e. in Ω, for n ≥ 2, (11.21)
which means u ∈ L p (Ω). Finally, from |un (x) − u(x)| → 0, a.e. in Ω, |un (x) − u(x)| p ≤ |g(x)| p and the
Lebesgue dominated convergence theorem we get
kun − uk p → 0 as n → ∞. (11.22)
Theorem 11.1.6 Let {un } ⊂ L p (Ω) and u ∈ L p (Ω) such that kun − uk p → 0. Then there exists a subsequence
{unk } such that
1. unk (x) → u(x), a.e. in Ω,
2. |unk (x)| ≤ h(x), a.e. in Ω, ∀k ∈ N, for some h ∈ L p (Ω).
Proof 11.4 The result is clear for p = ∞. Suppose 1 ≤ p < ∞. From the last theorem we can easily obtain
that |unk (x) − u(x)| → 0 as k → ∞, a.e. in Ω. To complete the proof, just take h = |u| + g, where g is defined
in the proof of the last theorem.
Theorem 11.1.7 L p (Ω) is reflexive for all p such that 1 < p < ∞.
Proof 11.5 We divide the proof into 3 parts.

1. For 2 ≤ p < ∞ we have that
u+v p u−v p
+
2 L p (Ω) 2 L p (Ω)
1
≤ kukLp p (Ω) + kvkLp p (Ω) , ∀u, v ∈ L p (Ω). (11.23)
2
Proof. Observe that

p/2
α p + β p ≤ α2 + β 2 , ∀α, β ≥ 0. (11.24)
a+b a−b
Now taking α = 2 and β = 2 in (11.24), we obtain (using the convexity of t p/2 ),
! p/2
p p 2 2
a+b a−b a+b a−b
+ ≤ +
2 2 2 2
2 p/2
a b2
= +
2 2
1 p 1 p
≤ |a| + |b| . (11.25)
2 2
The inequality (11.23) follows immediately.
2. L p (Ω) is uniformly convex, and therefore reflexive for 2 ≤ p < ∞.
Proof. Suppose given ε > 0 and suppose that
kuk p ≤ 1, kvk p ≤ 1 and ku − vk p > ε. (11.26)
From part 1, we obtain

p ε p
u+v
< 1− , (11.27)
2 p 2
and therefore
u+v
< 1−δ, (11.28)
2 p
for δ = 1 − (1 − (ε /2) p )1/p > 0. Thus, L p (Ω) is uniformly convex and from Theorem 4.7.2 it is
reflexive.
3. L p (Ω) is reflexive for 1 < p ≤ 2. Let 1 < p ≤ 2, from 2 we can conclude that Lq is reflexive. We will
define T : L p (Ω) → (Lq )∗ by
Z
hTu, f iLq (Ω) = u f dx, ∀u ∈ L p (Ω), f ∈ Lq (Ω). (11.29)
Ω
¨
From the Holder inequality, we obtain
|hTu, f iLq (Ω) | ≤ kuk p k f kq , (11.30)
so that
kTuk(Lq (Ω))∗ ≤ kuk p . (11.31)
Pick u ∈ L p (Ω) and define f0 (x) = |u(x)| p−2 u(x) ( f0 (x) = 0 if u(x) = 0). Thus, we have that f0 ∈
Lq (Ω), k f0 kq = kuk p−1
p and hTu, f0 iLq (Ω) = kuk pp . Therefore,
hTu, f0 iLq (Ω)

kTuk(Lq (Ω))∗ ≥ = kuk p (11.32)
k f0 kq
Hence from (11.31) and (11.32) we have
kTuk(Lq (Ω))∗ = kuk p , ∀u ∈ L p (Ω). (11.33)
Thus, T is an isometry from L p (Ω) to a closed subspace of (Lq (Ω))∗ . Since from the first part Lq (Ω)
is reflexive, we have that (Lq (Ω))∗ is reflexive. Hence T (L p (Ω)) and L p (Ω) are reflexive.
Theorem 11.1.8 (Riesz representation theorem) Let 1 < p < ∞ and let f be a continuous linear functional
on L p (Ω). Then there exists a unique u0 ∈ Lq such that
Z
f (v) = vu0 dx, ∀v ∈ L p (Ω). (11.34)
Ω
Furthermore
k f k(L p )∗ = ku0 kq . (11.35)
Proof 11.6 First we define the operator T : Lq (Ω) → (L p (Ω))∗ by

Z
hTu, viL p (Ω) = uv dx, ∀v ∈ L p (Ω). (11.36)
Ω
Similarly to the last theorem, we obtain
kTuk(L p (Ω))∗ = kukq . (11.37)
We have to show that T is onto. Define E = T (Lq (Ω)). As E is a closed subspace, it suffices to show that E
is dense in (L p (Ω))∗ . Suppose h ∈ (L p )∗∗ = L p is such that
hTu, hiL p (Ω) = 0, ∀u ∈ Lq (Ω). (11.38)
Choosing u = |h| p−2 h we may conclude that h = 0 which, by Corollary 4.2.14 completes the first part of the
proof. The proof of uniqueness is left to the reader.
p
Definition 11.1.9 Let 1 ≤ p ≤ ∞. We say that u ∈ Lloc (Ω) if uχK ∈ L p (Ω) for all compact K ⊂ Ω.
11.1.1 Spaces of continuous functions

We introduce some definitions and properties concerning spaces of continuous functions. First, we recall that
by a domain we mean an open set in Rn . Thus for a domain Ω ⊂ Rn and for any nonnegative integer m we
define by Cm (Ω) the set of all functions u which the partial derivatives Dα u are continuous on Ω for any α
such that |α| ≤ m, where if Dα = Dα1 1 Dα2 2 ...Dαn n we have |α| = α1 +...+αn . We define C∞ (Ω) = ∩∞ m
m=0C (Ω)
0
and denote C (Ω) = C(Ω). Given a function φ : Ω → R, its support, denoted by spt(φ ) is given by
6 0}.
spt(φ ) = {x ∈ Ω | φ (x) =
Cc∞ (Ω) denotes the set of functions in C∞ (Ω) with compact support contained in Ω.
The sets C0 (Ω) and C0∞ (Ω) consist of the closure of Cc (Ω) (which is the set of functions in C(Ω) with
compact support in Ω) and Cc∞ (Ω) respectively, relating the uniform convergence norm. On the other hand,
CBm (Ω) denotes the set of functions u ∈ Cm (Ω) for which Dα u is bounded on Ω for 0 ≤ |α| ≤ m. Observe
that CBm (Ω) is a Banach space with the norm denoted by k · kB,m given by
kukB,m = max sup {|Dα u(x)|} .

0≤|α|≤m x∈Ω
Also, we define Cm (Ω) as the set of functions u ∈ Cm (Ω) for which Dα u is bounded and uniformly continuous
on Ω for 0 ≤ |α| ≤ m. Observe that Cm (Ω) is a closed subspace of CBm (Ω) and is also a Banach space with
the norm inherited from CBm (Ω). An important space is the one of Holder
¨ continuous functions.
¨
Definition 11.1.10 (Spaces of Holder continuous functions) If 0 < λ < 1, for a nonnegative integer m we
define the space of Holder
¨ continuous functions denoted by Cm,λ (Ω), as the subspace of Cm (Ω) consisting
of those functions u for which, for all 0 ≤ |α| ≤ m, there exists a constant K such that
|Dα u(x) − Dα u(y)| ≤ K|x − y|λ , ∀x, y ∈ Ω.
Cm,λ (Ω) is a Banach space with the norm denoted by k · km,λ given by
α
|D u(x) − Dα u(y)|
kukm,λ = kukB,m + max sup , x 6 = y .
0≤|α |≤m x,y∈Ω |x − y|λ
From now on we say that f : Ω → R is locally integrable, if it is Lebesgue integrable on any compact K ⊂ Ω.
p
Furthermore, we say that f ∈ Lloc (Ω) if f ∈ L p (K) for any compact K ⊂ Ω. Finally, given an open Ω ⊂ Rn ,
we denote W ⊂⊂ Ω whenever W is compact and W ⊂ Ω.
Theorem 11.1.11 The space C0 (Ω) is dense in L p (Ω), for 1 ≤ p < ∞.
Proof 11.7 For the proof we need the following lemma:

1 (Ω) such that
Lemma 11.1.12 Let f ∈ Lloc
Z
f u dx = 0, ∀u ∈ C0 (Ω). (11.39)
Ω
Then f = 0 a.e. in Ω.
First suppose f ∈ L1 (Ω) and Ω bounded, so that m(Ω) < ∞. Given ε > 0, since C0 (Ω) is dense in L1 (Ω),
there exists f1 ∈ C0 (Ω) such that k f − f1 k1 < ε and thus, from (11.39) we obtain
Z
f1 u dx ≤ εkuk∞ , ∀u ∈ C0 (Ω). (11.40)
Ω
Defining
K1 = {x ∈ Ω | f1 (x) ≥ ε}, (11.41)
and
K2 = {x ∈ Ω | f1 (x) ≤ −ε}. (11.42)
As K1 and K2 are disjoint compact sets, by the Urysohn Theorem there exists u0 ∈ C0 (Ω) such that

+1, if x ∈ K1 ,
u0 (x) = (11.43)
−1, if x ∈ K2
and
|u0 (x)| ≤ 1, ∀x ∈ Ω. (11.44)
Also defining K = K1 ∪ K2 , we may write

Z Z Z
f1 u0 dx = f1 u0 dx + f1 u0 dx. (11.45)
Ω Ω−K K
Observe that, from (11.40)

Z Z
| f1 | dx ≤ | f1 u0 | dx ≤ ε (11.46)
K Ω
so that
Z Z Z
| f1 | dx = | f1 | dx + | f1 | dx ≤ ε + εm(Ω). (11.47)
Ω K Ω−K
Hence,
k f k1 ≤ k f − f1 k1 + k f1 k1 ≤ 2ε + εm(Ω). (11.48)
Since ε > 0 is arbitrary, we have that f = 0 a.e. in Ω. Finally, if m(Ω) = ∞, define
Ωn = {x ∈ Ω | dist(x, Ωc ) > 1/n and |x| < n}. (11.49)
It is clear that Ω = ∪∞ n=1 Ωn and from above f = 0 a.e. on Ωn , ∀n ∈ N, so that f = 0 a.e. in Ω.

Finally, to finish the proof of Theorem 11.1.11, suppose h ∈ Lq (Ω) is such that
Z
hu dx = 0, ∀u ∈ C0 (Ω). (11.50)
Ω
1 (Ω) since 1/p < ∞. From last lemma h = 0 a.e. in Ω, which by

R
Observe that h ∈ Lloc K |h| dx ≤ khkq m(K )
Corollary 4.2.14 completes the proof.
Theorem 11.1.13 L p (Ω) is separable for any 1 ≤ p < ∞.
Proof 11.8 The result follows from last theorem and from the fact that C0 (K) is separable for each K ⊂ Ω
compact (from the Weierstrass theorem, polynomials with rational coefficients are dense C0 (K)). Observe
n=1 Ωn , Ωn defined as in (11.49), where Ω̄n is compact, ∀n ∈ N.
that Ω = ∪∞
11.2 The Sobolev spaces

Now we define the Sobolev spaces, denoted by W m,p (Ω).
Definition 11.2.1 (Sobolev Spaces) We say that u ∈ W m,p (Ω) if u ∈ L p (Ω) and Dα u ∈ L p (Ω), for all α such
that 0 ≤ |α| ≤ m, where the derivatives are understood in the distributional sense.
Definition 11.2.2 We define the norm k · km,p for W m,p (Ω), where m ∈ N and 1 ≤ p ≤ ∞, as
( )1/p
kukm,p = ∑ kD α
uk pp , if 1 ≤ p < ∞, (11.51)
0≤|α|≤m
and
kukm,∞ = max kDα uk∞ . (11.52)

0≤|α|≤m
Theorem 11.2.3 W m,p (Ω) is a Banach space.
Proof 11.9 Consider {un } a Cauchy sequence in W m,p (Ω). Then {Dα un } is a Cauchy sequence for each
0 ≤ |α| ≤ m. Since L p (Ω) is complete there exist functions u and uα , for 0 ≤ |α| ≤ m, in L p (Ω) such that
1 (Ω) and so u determines a distribution
un → u and Dα un → uα in L p (Ω) as n → ∞. From above L p (Ω) ⊂ Lloc n
Tun ∈ D (Ω). For any φ ∈ D(Ω) we have, by Holder’
0 ¨ s inequality
Z
|Tun (φ ) − Tu (φ )| ≤ |un (x) − u(x)||φ (x)|dx ≤ kφ kq kun − uk p . (11.53)
Ω
Hence Tun (φ ) → Tu (φ ) for every φ ∈ D(Ω) as n → ∞. Similarly TDα un (φ ) → Tuα (φ ) for every φ ∈ D(Ω).
We have that
Tuα (φ ) = lim TDα un (φ )

n→∞
= lim (−1)|α| Tun (Dα φ )
n→∞
= (−1)|α| Tu (Dα φ ) = TDα u (φ ), (11.54)
for every φ ∈ D(Ω). Thus uα = Dα u in the sense of distributions, for 0 ≤ |α| ≤ m, and u ∈ W m,p (Ω). As
lim ku − un km,p = 0, W m,p (Ω) is complete.
n→∞
Remark 11.2.4 Observe that distributional and classical derivatives coincide when the latter exist and are
continuous. We define S ⊂ W m,p (Ω) by
S = {φ ∈ Cm (Ω) | kφ km,p < ∞} (11.55)
Thus, the completion of S concerning the norm k · km,p is denoted by H m,p (Ω).
Corollary 11.2.5 H m,p (Ω) ⊂ W m, p (Ω)
Proof 11.10 Since W m,p (Ω) is complete we have that H m,p (Ω) ⊂ W m,p (Ω).
Theorem 11.2.6 W m,p (Ω) is separable if 1 ≤ p < ∞, and is reflexive and uniformly convex if 1 < p < ∞.
Particularly, W m,2 (Ω) is a separable Hilbert space with the inner product
(u, v)m = ∑ hDα u, Dα viL2 (Ω) . (11.56)

0≤|α|≤m
Proof 11.11 We can see W m,p (Ω) as a subspace of L p (Ω, RN ), where N = ∑0≤|α|≤m 1. From the relevant
properties for L p (Ω), we have that L p (Ω; RN ) is a reflexive and uniformly convex for 1 < p < ∞ and separable
for 1 ≤ p < ∞. Given u ∈ W m,p (Ω), we may associate the vector Pu ∈ L p (Ω; RN ) defined by
Pu = {Dα u}0≤|α |≤m . (11.57)
Since kPuk pN = kukm,p , we have that W m,p is closed subspace of L p (Ω; RN ). Thus from Theorem 1.21 in
Adams [1], we have that W m,p (Ω) is separable if 1 ≤ p < ∞ and, reflexive and uniformly convex, if 1 < p < ∞.
Lemma 11.2.7 Let 1 ≤ p < ∞ and define U = L p (Ω; RN ). For every continuous linear functional f on U,
there exists a unique v ∈ Lq (Ω; RN ) = U ∗ such that
N
f (u) = ∑ hui , vi i, ∀u ∈ U. (11.58)
i=1
Moreover,
k f kU ∗ = kvkqN , (11.59)
where k · kqN = k · kLq (Ω,RN ) .
Proof 11.12 For u = (u1 , ..., un ) ∈ L p (Ω; RN ) we may write

f (u) = f ((u1 , 0, ..., 0)) + ... + f ((0, ..., 0, u j , 0, ..., 0))
+... + f ((0, ..., 0, un )), (11.60)
and since f ((0, ..., 0, u j , 0, ..., 0)) is continuous linear functional on u j ∈ L p (Ω),
there exists a unique v j ∈
Lq (Ω) such that f (0, ..., 0, u j , 0, ..., 0) = hu j , v j iL2 (Ω) , ∀u j ∈ L p (Ω), ∀ 1 ≤ j ≤ N, so that
N
f (u) = ∑ hui , vi i, ∀u ∈ U. (11.61)
i=1
¨
From Holder’s inequality we obtain
N
| f (u)| ≤ ∑ ku j k p kv j kq ≤ kuk pN kvkqN , (11.62)
j=1
and hence k f kU ∗ ≤ kvkqN . The equality in (11.62) is achieved for u ∈ L p (Ω, RN ), 1 0, there is a measurable set A such that
m(A) > 0 and |vk (x)| ≥ kvk k∞ − ε, ∀x ∈ A. Defining u(x) as

v̄k /vk , if i = k, x ∈ A and vk (x) =
6 0
ui (x) = (11.64)
0, otherwise,
we have
Z
f (uk ) = hu, vk iL2 (Ω) = |vk |dx
A
≥ (k(vk k∞ − ε)kuk k1
= (kvk∞N − ε)kuk1N . (11.65)
Since ε is arbitrary, the proof is complete.
Theorem 11.2.8 Let 1 ≤ p < ∞. Given a continuous linear functional f on W m,p (Ω), there exists v ∈
Lq (Ω, RN ) such that
f (u) = ∑ hDα u, vα iL2 (Ω) . (11.66)
0≤|α |≤m
Proof 11.13 Consider f a continuous linear operator on U = W m,p (Ω). By the Hahn Banach Theorem,
we can extend f to f˜, on L p (Ω; RN ), so that k f˜kqN = k f kU ∗ and by the last theorem, there exists {vα } ∈
Lq (Ω; RN ) such that
f˜(û) = ∑ hûα , vα iL2 (Ω) , ∀v ∈ L p (Ω; RN ). (11.67)
0≤|α|≤m
In particular, for u ∈ W m,p (Ω), defining û = {Dα u} ∈ L p (Ω; RN ) we obtain

f (u) = f˜(û) = ∑ hDα u, vα iL2 (Ω) . (11.68)
1≤|α|≤m
Finally, observe that, also from the Hahn-Banach theorem k f kU ∗ = k f˜kqN = kvkqN .
Definition 11.2.9 Let Ω ⊂ Rn be a domain. For m a positive integer and 1 ≤ p < ∞ we define W0m,p (Ω) as
the closure in k · km,p of Cc∞ (Ω), where we recall that Cc∞ (Ω) denotes the the set of C∞ (Ω) functions with
compact support contained in Ω. Finally, we also recall that the support of φ : Ω → R, denoted by spt(φ ), is
given by
spt(φ ) = {x ∈ Ω |φ (x) =
6 0}.
11.3 The Sobolev imbedding theorem

11.3.1 The statement of Sobolev imbedding theorem
Now we present the Sobolev imbedding theorem. We recall that for normed spaces X,Y the notation
X ,→ Y
means that X ⊂ Y and there exists a constant K > 0 such that
kukY ≤ KkukX , ∀u ∈ X.
If in addition the imbedding is compact then for any bounded sequence {un } ⊂ X there exists a conver-
gent subsequence {unk }, which converges to some u in the norm k · kY . At this point, we first introduce the
following definition:
Definition 11.3.1 Let Ω ⊂ Rn be an open bounded set. We say that ∂ Ω is Ĉ1 if for each x0 ∈ ∂ Ω, denoting
x̂ = (x1 , ..., xn−1 ) for a local coordinate system, there exist r > 0 and a function f (x1 , ..., xn−1 ) = f (x̂) such
that
W = Ω ∩ Br (x0 ) = {x ∈ Br (x0 ) | xn ≤ f (x1 , ..., xn−1 )}.
Moreover, f (x̂) is a Lipschitz continuous function, so that
| f (x̂) − f (ŷ)| ≤ C1 |x̂ − ŷ|2 , on its domain,
for some C1 > 0. Finally, we assume
n−1
∂ f (x̂)
∂ xk k=1
is classically defined, almost everywhere also on its concerning domain, so that f ∈ W 1,2 .
Theorem 11.3.2 (The Sobolev imbedding theorem) Let Ω be an open bounded set in Rn such that ∂ Ω is
Ĉ1 . Let j ≥ 0 and m ≥ 1 be integers and let 1 ≤ p < ∞.
1. Part I
(a) Case A If either mp > n or m = n and p = 1 then
W j+m,p (Ω) ,→ CBj (Ω). (11.69)
Moreover,
W j+m,p (Ω) ,→ W j,q (Ω), for p ≤ q ≤ ∞, (11.70)
and, in particular
W m,p (Ω) ,→ Lq (Ω), for p ≤ q ≤ ∞. (11.71)
(b) Case B If mp = n, then
W j+m,p (Ω) ,→ W j,q (Ω), for p ≤ q < ∞, (11.72)
and, in particular
W m,p (Ω) ,→ Lq (Ω), for p ≤ q < ∞. (11.73)
(c) Case C If mp < n or p = 1 , then

np
W j+m,p (Ω) ,→ W j,q (Ω), for p ≤ q ≤ p∗ = , (11.74)
n − mp
and, in particular
np
W m,p (Ω) ,→ Lq (Ω), for p ≤ q ≤ p∗ = . (11.75)
n − mp
2. Part II
If mp > n > (m − 1)p, then
W j+m,p ,→ C j,λ (Ω), for 0 < λ ≤ m − (n/p), (11.76)
and if n = (m − 1)p, then
W j+m,p ,→ C j,λ (Ω), for 0 < λ < 1. (11.77)
Also, if n = m − 1 and p = 1, then (11.77) holds for λ = 1 as well.

3. Part III All imbeddings in Parts A and B are valid for arbitrary domains Ω if the W − space under-
going the imbedding is replaced with the corresponding W0 − space.
11.4 The proof of the Sobolev imbedding theorem

Now we present a collection of results which imply the proof of the Sobolev imbedding theorem. We start
with the approximation by smooth functions.
Definition 11.4.1 Let Ω ⊂ Rn be an open bounded set. For each ε > 0 define
Ωε = {x ∈ Ω | dist(x, ∂ Ω) > ε}.
Definition 11.4.2 Define η ∈ Cc∞ (Rn ) by

(
1
C exp |x|22 −1
, if |x|2 < 1,
η (x) =
0, if |x|2 ≥ 1,
where | · |2 refers to the Euclidean norm in Rn , that is for x = (x1 , ..., xn ) ∈ Rn , we have
q
|x|2 = x12 + ... + xn2 .
Moreover, C > 0 is chosen so that Z

η dx = 1.
Rn
For each ε > 0, set
1 x
ηε (x) = η .
εn ε
The function η is said to be the fundamental mollifier. The functions ηε ∈ Cc∞ (Rn ) and satisfy
Z
ηε dx = 1,
Rn
and spt(ηε ) ⊂ B(0, ε).
Definition 11.4.3 If f : Ω → R is locally integrable, we define its mollification, denoted by fε : Ωε → R as:
fε = ηε ∗ f ,
that is,
Z
fε (x) = ηε (x − y) f (y) dy
Ω
Z
n
= (−1) ηε (y) f (x − y) dy. (11.78)
B(0,ε)
Theorem 11.4.4 (Properties of mollifiers) The mollifiers have the following properties:
1. fε ∈ C∞ (Ωε ),
2. fε → f a.e. as ε → 0,
3. If f ∈ C(Ω) then fε → f uniformly on compact subsets of Ω.
Proof 11.14
1. fix x ∈ Ωε , i ∈ {1, ..., n} and a h0 > 0 small enough such that
x + hei ∈ Ωε ,
∀0 < |h| < h0 . Thus

fε (x + hei ) − fε (x) 1 1 x + hei − y x−y
Z
= n η −η
h ε Ωh ε ε
× f (y) d y

1 1 x + hei − y x−y
Z
= n η −η
ε Vh ε ε
× f (y) dy, (11.79)
for an appropriate set V ⊂⊂ Ω.

Observe that
1 x + hei − y x−y ∂ η(x − y)
η −η f (y) → f (y), in V.
h ε ε ∂ xi
and
1 x + hei − y x−y
η −η f (y) < C| f (y)|, in V,
h ε ε
where
∂ η(x − y)
C = sup .
y∈RN ∂ xi
From this and the Lebesgue dominated converge theorem, we obtain
∂ fε (x) fε (x + hei ) − fε (x)

= lim
∂ xi h→0 h

1 1 x + hei − y x−y
Z
= lim n η −η f (y) dy
h→0 ε Ωh ε ε
∂ ηε (x − y)
Z
= f (y) dy. (11.80)
Ω ∂ xi
By analogy, we may show that

Z
α
D fε (x) = Dα ηε (x − y) f (y) dy, ∀x ∈ Ωε .
Ω
2. From the Lebesgue differentiation theorem we have

1
Z
lim | f (y) − f (x)| dy = 0, (11.81)
r→0 |B(x, r)| B(x,r)
for almost all x ∈ Ω. Fix x ∈ Ω such that (11.81) holds. Hence,

Z
| fε (x) − f (x)| = ηε (x − y)[ f (x) − f (y)] dy
B(x,ε)

1 x−y
Z
≤ η [ f (x) − f (y)] dy
ε n B(x,ε) ε
C
Z
≤ | f (y) − f (x)| dy (11.82)
|B(x, ε)| B(x,ε)
for an appropriate constant C > 0. From (11.81), we obtain fε → f as ε → 0.

3. Assume f ∈ C(Ω). Given V ⊂⊂ Ω choose W such that
V ⊂⊂ W ⊂⊂ Ω,
and note that f is uniformly continuous on W Thus the limit indicated in (11.81) holds uniformly on
V , and therefore fε → f uniformly on V .
For the next results we denote

u(x), if x ∈ Ω,
ũ(x) =
0, if x ∈ Rn \ Ω.
Theorem 11.4.5 Let Ω ⊂ RN be an open bounded set. Let u : Ω → R be such that u ∈ L p (Ω), where 1 ≤
p < ∞. Then
ηε ∗ ũ ∈ L p (Ω),
kηε ∗ ũk p,Ω ≤ kuk p,Ω , ∀ε > 0
and
lim kηε ∗ ũ − uk p,Ω = 0.
ε→0+
Proof 11.15 Defining q = p/(p − 1), from Holder’

¨ s inequality we have
Z
|ηε ∗ ũ(x)| = ηε (x − y)ũ(y) dy
Rn
Z
= [ηε (x − y)](1−1/ p) [ηε (x − y)]1/p ũ(y) dy
Rn
Z 1/q Z 1/ p
p
≤ ηε (x − y) dy ηε (x − y)|ũ(y)| dy
Rn Rn
Z 1/p
= ηε (x − y)|u(y)| p dy . (11.83)
Rn
From this and Fubini theorem, we obtain

Z Z Z
|ηε ∗ ũ(x)| p dx ≤ ηε (x − y)|ũ(y)| p dy dx
Ω Rn Rn
Z Z
= |ũ(y)| p ηε (x − y) dx dy
Rn Rn
Z
= |ũ(y)| p dy
RN
Z
= |u(y)| p dy
Ω
= kuk pp,Ω . (11.84)
Suppose given ρ > 0. As Cc (Ω) is dense in L p (Ω), there exists φ ∈ Cc (Ω) such that
ku − φ k p < ρ/3.
From the fact that

ηε ∗ φ → φ
as ε → 0, uniformly in Ω we have that there exists δ > 0 such that
kηε ∗ φ − φ k p < ρ/3

if 0 < ε < δ . Thus for any 0 < ε < δ (ρ), we get
kηε ∗ ũ − uk p = kηε ∗ ũ − ηε ∗ φ + ηε ∗ φ − φ + φ − uk p
≤ kηε ∗ ũ − ηε ∗ φ k p + kηε ∗ φ − φ k p + kφ − uk p
≤ ρ/3 + ρ/3 + ρ/3 = ρ. (11.85)
Since ρ > 0 is arbitrary, the proof is complete.
11.4.1 Relatively compact sets in L p (Ω)

Theorem 11.4.6 Consider 1 ≤ p < ∞. A bounded set K ⊂ L p (Ω) is relatively compact if and only if for each
ε > 0, there exists δ > 0 and G ⊂⊂ Ω (we recall that G ⊂⊂ Ω means that G is compact and G ⊂ Ω) such
that for each u ∈ K and h ∈ Rn such that |h| < δ we have
1.
Z
|ũ(x + h) − ũ(x)| p dx < ε p , (11.86)
Ω
2.
Z
|u(x)| p dx < ε p . (11.87)
Ω−G
Proof 11.16 Suppose K is relatively compact in L p (Ω). Suppose given ε > 0. As K is compact we may
find a finite ε/6-net for K. Denote such a ε/6-net by N where
N = {v1 , ..., vm } ⊂ L p (Ω).
Since Cc (Ω) is dense in L p (Ω), for each k ∈ {1, ..., m} there exists φk ∈ Cc (Ω) such that
ε
kφk − vk k p < .
6
Thus defining
S = {φ1 , ..., φm },
given u ∈ K, we may select vk ∈ N such that
ε
ku − vk k p < ,
6
so that
kφk − uk p ≤ kφk − vk k p + kvk − uk p

ε ε ε
≤ + = . (11.88)
6 6 3
Define
G = ∪m
k=1 spt(φk ),
where
spt(φk ) = {x ∈ Rn | φk (x) =
6 0}.
We have that
G ⊂⊂ Ω,
where as above mentioned this means G ⊂ Ω. Observe that

Z
ε p > ku − φk k pp ≥ |u(x)| p dx.
Ω−G
Since u ∈ K is arbitrary, (11.87) is proven. Since φk is continuous and spt(φk ) is compact we have that φk is
uniformly continuous, that is, for the ε given above, there exists δ̃ > 0 such that if |h| < min{δ˜ , 1} then
ε
|φk (x + h) − φk (x)| < , ∀x ∈ G,
3(|G| + 1)
Thus, Z ε p
|φk (x + h) − φk (x)| p dx < .
Ω 3
Also observe that since
ε
ku − φk k p < ,
3
we have that
ε
kTh u − Th φk k p < ,
3
where Th u = u(x + h). Thus if |h| < δ = min{δ̃ , 1}, we obtain
kTh ũ − ũk p ≤ kTh ũ − Th φk k p + kTh φk − φk k p

+kφk − uk p
ε ε ε
< + + = ε. (11.89)
3 3 3
For the converse, it suffices to consider the special case Ω = Rn , because for the general Ω we can define
K̃ = {ũ | u ∈ K}. Suppose given ε > 0 and choose G ⊂⊂ Rn such that for all u ∈ K we have
Z
ε
|u(x)| p dx < .
Rn −G 3
For each ρ > 0 the function ηρ ∗ u ∈ C∞ (Rn ), and in particular ηρ ∗ u ∈ C(G). Suppose φ ∈ C0 (Rn ). Fix
ρ > 0. By Holder’
¨ s inequality we have
Z p
|ηρ ∗ φ (x) − φ (x)| p = ηρ (y)(φ (x − y) − φ (x)) dy
Rn
Z p
= (ηρ (y))1−1/p (ηρ (y))1/ p (T−y φ (x) − φ (x)) dy
Rn
Z
≤ (ηρ (y))|T−y φ (x) − φ (x)| p dy. (11.90)
Bρ (θ )
Hence, from the Fubini theorem we may write

Z Z Z
|ηρ ∗ φ (x) − φ (x)| p dx ≤ (ηρ (y)) |T−y φ (x) − φ (x)| p dx dy, (11.91)
Rn Bρ (θ ) Rn
so that we may write
kηρ ∗ φ − φ k p ≤ sup {kTh φ − φ k p }. (11.92)

h∈Bρ (θ )
Fix u ∈ L p (Rn ). We may obtain a sequence {φk } ⊂ Cc (Rn ) such that
φk → u, in L p (Rn ).
Observe that
ηρ ∗ φk → ηρ ∗ u, in L p (Rn ),
as k → ∞. Also
Th φk → Th u, in L p (Rn ),
as k → ∞. Thus,
kTh φk − φk k p → kTh u − uk p ,
in particular ( )

lim sup sup {kTh φk − φk k ≤ sup kTh u − uk p .
k→∞ h∈Bρ (θ ) h∈Bρ (θ )
Therefore, as
kηρ ∗ φk − φk k p → kηρ ∗ u − uk p ,
as k → ∞, from (11.92) we get
kηρ ∗ u − uk p ≤ sup {kTh u − uk p }.

h∈Bρ(θ )
kηρ ∗ u − uk p → 0, uniformly in K as ρ → 0.
Fix ρ0 > 0 such that Z

ε
|ηρ0 ∗ u − u| p dx < , ∀u ∈ K.
G 3 · 2 p−1
Observe that
Z
|ηρ0 ∗ u(x)| = ηρ0 (x − y)u(y) dy
Rn
Z
= [ηρ0 (x − y)](1−1/p) [ηρ0 (x − y)]1/p u(y) dy
Rn
Z 1/q Z 1/p
p
≤ ηρ0 (x − y) d y ηρ0 (x − y)|u(y)| dy
Rn Rn
Z 1/p
= ηρ0 (x − y)|u(y)| p dy . (11.93)
Rn
From this, we may write,

!1/p
|ηρ0 ∗ u(x)| ≤ sup ηρ0 (y) kuk p ≤ K1 , ∀x ∈ Rn , u ∈ K
y∈Rn
where K1 = K2 K3 ,
!1/p
K2 = sup ηρ0 (y) ,
y∈Rn
and K3 is any constant such that

kuk p < K3 , ∀u ∈ K.
Similarly
!1/p
|ηρ0 ∗ u(x + h) − ηρ0 u(x)| ≤ sup ηρ0 (y) kTh u − uk p ,
y∈Rn
and thus from (11.86) we obtain
ηρ0 ∗ u(x + h) → ηρ0 ∗ u(x), as h → 0
uniformly in Rn and for u ∈ K.

By the Arzela-Ascoli Theorem
{ηρ0 ∗ u | u ∈ K}
is relatively compact in C(G), and it is totally bounded so that there exists a ε0 -net N = {v1 , ..., vm } where
1/p
ε
ε0 = .
3 · 2 p−1 |G|
Thus for some k ∈ {1, ..., m} we have

kvk − ηρ0 ∗ uk∞ < ε0 .
Hence,
Z Z Z
|u(x) − ṽk (x)| p dx = |u(x)| p d x + |u(x) − vk (x)| p dx
Rn Rn −G G
Z
ε
≤ + 2 p−1 (|u(x) − (ηρ0 ∗ u)(x)| p
3 G
+|ηρ0 ∗ u(x) − vk (x)| p ) dx

ε p−1 ε ε |G|
≤ +2 +
3 3 · 2 p−1 3 · 2 p−1 |G|
= ε. (11.94)
Thus K is totally bounded and therefore it is relatively compact.

11.4.2 Some approximation results

Theorem 11.4.7 Let Ω ⊂ Rn be an open set. Assume u ∈ W m,p (Ω) for some 1 ≤ p < ∞, and set
uε = ηε ∗ u in Ωε .
Then,
1. uε ∈ C∞ (Ωε ), ∀ε > 0,
m,p
2. uε → u in Wloc (Ω), as ε → 0,
Proof 11.17 Assertion 1 has been already proved. Let us prove 2. We will show that if |α| ≤ m, then
Dα uε = ηε ∗ Dα u, in Ωε .
For, let x ∈ Ωε . Thus,

Z
Dα uε (x) = Dα ηε (x − y)u(y) dy
Ω
Z
= Dαx ηε (x − y)u(y) dy
Ω
Z
= (−1)|α| Dαy (ηε (x − y))u(y) dy. (11.95)
Ω
Observe that for fixed x ∈ Ωε the function
φ (y) = ηε (x − y) ∈ Cc∞ (Ω).
Therefore, Z Z
Dαy (ηε (x − y)) u(y) dy = (−1)|α| ηε (x − y)Dαy u(y) dy,
Ω Ω
and hence,
Z
Dα uε (x) = (−1)|α|+|α | ηε (x − y)Dα u(y) dy
Ω
= (ηε ∗ Dα u)(x). (11.96)
Now choose any open bounded set such that V ⊂⊂ Ω. We have that
Dα uε → Dα u, in L p (V ) as ε → 0,
for each |α| ≤ m.

Thus,
p
kuε − ukm,p,V = ∑ kDα uε − Dα uk p,V → 0,
|α|≤m
as ε → 0.
Theorem 11.4.8 Let Ω ⊂ Rn be a bounded open set and suppose u ∈ W m,p (Ω) for some 1 ≤ p < ∞. Then
there exists a sequence {uk } ⊂ C∞ (Ω) such that
uk → u in W m,p (Ω).

Ω = ∪∞
i=1 Ωi ,
where
Ωi = {x ∈ Ω | dist(x, ∂ Ω) > 1/i}.
Define
Vi = Ωi+3 − Ω̄i+1 ,
and choose any open set V0 such that V0 ⊂⊂ Ω, so that
Ω = ∪∞
i=0Vi .
Let {ζi }∞
i=0 be a smooth partition of unit subordinate to the open sets {Vi }i=0 . That is,
∞

0 ≤ ζi ≤ 1, ζi ∈ Cc∞ (Vi )
∑∞i=0 ζi = 1, on Ω,
Now suppose u ∈ W m,p (Ω). Thus ζi u ∈ W m,p (Ω) and spt(ζi u) ⊂ Vi ⊂ Ω. Choose δ > 0. For each i ∈ N
choose εi > 0 small enough so that
ui = ηεi ∗ (ζi u)
satisfies
δ
kui − ζi ukm,p,Ω ≤ ,
2i+1
and spt(ui ) ⊂ Wi where Wi = Ωi+4 − Ω̄i ⊃ Vi . Define

∞
v = ∑ ui .
i=0
Thus such a function belongs to C∞ (Ω), since for each open V ⊂⊂ Ω there are at most finitely many non-zero
terms in the sum. Since
∞
u = ∑ ζi u,
i=0
we have that for a fixed V ⊂⊂ Ω,
∞
kv − ukm,p,V ≤ ∑ kui − ζi ukm,p,V
i=0
∞
1
≤ δ∑ = δ. (11.97)
i=0 2i+1
Taking the supremum over sets V ⊂⊂ Ω we obtain
kv − ukm,p,Ω < δ .
Since δ > 0 is arbitrary, the proof is complete.
The next result is also relevant. For a proof see Evans, [34] page 232.
Theorem 11.4.9 Let Ω ⊂ Rn be a bounded set such that ∂ Ω is C1 . Suppose u ∈ W m,p (Ω) where 1 ≤ p < ∞.
Thus there exists a sequence {un } ⊂ C∞ (Ω) such that
un → u in W m,p (Ω), as n → ∞.
Anyway, now we prove a more general result.

Theorem 11.4.10 Let Ω ⊂ Rn be an open bounded set such that ∂ Ω is Ĉ1 . Let u ∈ W m,p (Ω) where m is a
non-negative integer and 1 ≤ p < ∞.
Under such assumptions, there exists {uk } ⊂ C∞ (Ω) such that
kuk − ukm,p,Ω → 0, as k → ∞.
Proof 11.19 Fix x0 ∈ ∂ Ω. Since ∂ Ω is Ĉ1 , denoting x̂ = (x1 , ..., xn−1 ) for a local coordinate system, there
exists r > 0 and a function f (x1 , ..., xn−1 ) = f (x̂) such that
W = Ω ∩ Br (x0 ) = {x ∈ Br (x0 ) | xn ≤ f (x1 , ..., xn−1 )}.
We emphasize f (x̂) is a Lipschitz continuous function, so that
| f (x̂) − f (ŷ)| ≤ C1 |x̂ − ŷ|2 , on its domain,
for some C1 > 0. Furthermore, n−1

∂ f (x̂)
∂ xk k=1
is classically defined, almost everywhere also on its concerning domain.
Let ε > 0. For each δ > 0 define xδ = x +Cδ en , where C > 1 is a fixed constant. Define uδ = u(xδ ). Now
choose δ > 0 sufficiently small such that
kuδ − ukm,p,W < ε/2.

For each n ∈ N, x ∈ W define

vn (x) = (η1/n ∗ uδ )(x).
Observe that
kvn − ukm,p,W ≤ kvn − uδ km,p,W + kuδ − ukm,p,W ,
For the fixed δ > 0, there exists Nε ∈ N such that if n > Nε we have
kvn − uδ km,p,W < ε/2,
and
vn ∈ C∞ (W ).
Hence,
kvn − ukm,p,W ≤ kvn − uδ km,p,W + kuδ − ukm,p,W < ε/2 + ε/2 = ε.
Clarifying the dependence of r on x0 ∈ ∂ Ω we denote r = rx0 . Observe that
∂ Ω ⊂ ∪x0 ∈∂ Ω Brx0 (x0 )
so that since ∂ Ω is compact, there exists x1 , ..., xM ∈ ∂ Ω such that
∂ Ω ⊂ ∪M
i=1 Bri (xi ).
We denote Bri (xi ) = Bi and Wi = Ω∩Bi , ∀i ∈ {1, ..., M}. We also choose an appropriate open set B0 ⊂⊂ Ω
such that
Ω ⊂ ∪M i=0 Bi .
Let {ζi }M M
i=0 be a concerned partition of unity relating {Bi }i=0 .
Thus ζi ∈ Cc∞ (Bi ) and 0 ≤ ζi ≤ 1, ∀i ∈ {0, ..., M} and also
M
∑ ζi = 1 on Ω.
i=0
From above, we may find vi ∈ C∞ (W i ) such that kvi − ukm,p,Wi < ε, ∀i ∈ {1, ..., M}. Define u0 = v0 =
u on B0 ≡ W0 ,
ui = ζi u, ∀i ∈ {0, ...., M}
and
M
v = ∑ ζi vi .
i=0
We emphasize
v ∈ C∞ (Ω).
Therefore
M
kv − ukm,p,Ω = ∑ (ζi u − ζi vi )
i=0 m,p,Ω
M
≤ C2 ∑ ku − vi km, p,(Ω∩Bi )
i=0
M
= C2 ∑ ku − vi km,p,Wi
i=0
< C2 Mε. (11.98)
Since neither C2 nor M depends on ε > 0, the proof is complete.

11.4.3 Extensions
In this section we study extensions of Sobolev spaces from a domain Ω ⊂ Rn to Rn . First we enunciate a
result found in Evans, [34].
Theorem 11.4.11 Assume Ω ⊂ Rn is an open bounded set, and that ∂ Ω is C1 . Let 1 ≤ p < ∞ and let V be a
bounded open set such that Ω ⊂⊂ V . Then there exists a bounded linear operator
E : W 1,p (Ω) → W 1,p (Rn ),
such that for each u ∈ W 1,p (Ω) we have:

1. Eu = u, a.e. in Ω,
2. Eu has support in V , and
3. kEuk1,p,Rn ≤ Ckuk1,p,Ω , where the constant depend only on p, Ω, and V.
The next result, which we prove, is a more general one.

Theorem 11.4.12 Assume Ω ⊂ Rn is an open bounded set, and that ∂ Ω is Ĉ1 . Let 1 ≤ p < ∞, and let V be
a bounded open set such that Ω ⊂⊂ V . Then there exists a bounded linear operator
E : W 1,p (Ω) → W 1,p (Rn ),
such that for each u ∈ W 1,p (Ω) we have:

1. Eu = u, a.e. in Ω,
2. Eu has support in V , and
3. kEuk1,p,Rn ≤ Ckuk1,p,Ω , where the constant depend only on p, Ω, and V.
Proof 11.20 Let u ∈ W 1,p (Ω). Fix N ∈ N and select φN ∈ C∞ (Ω) such that
kφN − uk1,p,Ω < 1/N. (11.99)
Choose x0 ∈ ∂ Ω. From the hypothesis we may write
Ω ∩ Br (x0 ) = {x ∈ Br (x0 ) | xn ≤ f (x1 , ..., xn−1 )},
for some r > 0 and so that denoting x̂ = (x1 , ..., xn−1 ), f (x1 , ..., xn−1 ) = f (x̂) is a Lipschitz continuous function
such that
∂ f (x̂) n−1

∂ xk k=1
is classically defined almost everywhere on its domain and
| f (x̂) − f (ŷ)| ≤ C1 |x̂ − ŷ|2 , ∀x̂, ŷ on its domain,
for some C1 > 0.

Define the variable y ∈ Rn by yi = xi , ∀i ∈ {1, ..., n − 1}, and yn = f (x1 , ..., xn−1 ) − xn .
Thus
φN (x1 , ..., xn ) = φN (y1 , ..., yn−1 , f (y1 , .., yn−1 ) − yn ) = φ N (y1 , ..., yn ).
Observe that defining ψ(x) = y we have ψ(x) = (x1 , · · · , xn−1 , f (x1 , · · · , xn−1 ) − xn ) so that
ψ −1 (y) = (y1 , · · · yn−1 , f (y1 , · · · yn−1 ) − yn ).

From the continuity of ψ −1 , there exists r1 > 0 such that
ψ −1 (B+
r1 (y0 )) ⊂ Ω ∩ Br (x0 ),
+ −
where y0 = (x01 , ..., x0n−1 , 0). We define W + = ψ −1 (Br1 (y0 )) and W − = ψ −1 (Br1 (y0 )) where we denote
B+ = B+
r1 (y0 ) = {y ∈ Br1 (y0 ) | yn ≥ 0},
and
B− = B−
r1 (y0 ) = {y ∈ Br1 (y0 ) | yn < 0}.
We emphasize that locally about x0 we have that ∂ Ω and ψ(∂ Ω) correspond to the the equations xn −
f (x1 , ..., xn−1 ) = 0 and yn = 0 respectively.
At this point we are going to show that {φ N } is a Cauchy sequence in W 1,p (B+ ).
Observe that, denoting W = W + ∪W − , we have
ψ(x) = (x1 , · · · , xn−1 , f (x1 , · · · , xn ) − xn ), in W.
Also, since ψ 0 (x) is classically defined and |ψ 0 (x)| < C2 , a.e. in Ω, for an appropriate C2 > 0 we have
that ψ 0 ∈ L∞ (W , Rn×n ). Similarly,
det(ψ 0 ) ∈ L∞ (W ),
det[(ψ −1 )0 ] ∈ L∞ (B),
∂ xk (y)
∈ L∞ (B), ∀ j, k ∈ {1, · · · , n},
∂yj
and
∂ y j (x)
∈ L∞ (W ), ∀ j, k ∈ {1, · · · , n}.
∂ xk
We denote by Ŵ ⊂ W the set on which det(ψ 0 ) and
∂ y j (x)
, ∀ j, k ∈ {1, · · · , n},
∂ xk
are classically defined.
Similarly, we denote by B̂ ⊂ B the set on which det((ψ −1 )0 ) and
∂ xk (y))
, ∀ j, k ∈ {1, · · · , n},
∂yj
are classically defined.

Let M, N ∈ N, thus
Z
|φ N (y) − φ M (y)| p dy
B+
Z
≤ |φ N (y(x)) − φ M (y(x))| p | det(ψ 0 (x))| dx
Ŵ +
Z
≤ K1 |φ N (y(x)) − φ M (y(x))| p dx
W+
Z
= K1 |φN (x) − φM (x)| p dx
W+
p
≤ K1 kφN − φM k1,p,Ω → 0, as N, M → ∞. (11.100)
Similarly, for j ∈ {1, · · · , n}, we have,

p
∂ φ N (y) ∂ φ M (y)
Z
− dy
B+ ∂yj ∂yj
p
∂ φ N (y(x)) ∂ φ M (y(x))
Z
≤ − | det(ψ 0 (x))| dx
W + ∂yj ∂yj
∂ φN (x) ∂ φM (x) ∂ xk (y(x)) p
Z
≤ − | det(ψ 0 (x))| dx
W+ ∂ xk ∂ xk ∂ yk
p
≤ K2 kφN − φM k1,p,Ω → 0, as N, M → ∞. (11.101)
Hence, {φ N } is a Cauchy sequence.

Therefore, there exists ũ ∈ W 1,p (B+ ) such that
φ N → ũ in norm, in W 1,p (B+ ).
Hence, up to a not relabeled subsequence,
φ N → ũ a.e. in B+ .
However, up to to a not relabeled subsequence of this last subsequence, from (11.99) we obtain,
φN → u, a.e. in Ω, (11.102)
so that
ũ(y) = u(x(y)), a.e. in B+ .
Moreover, φ N is Lipschitz continuous on B+ so that φ ∈ W 1,p (B+ ), and therefore there exists φ̃N ∈
+
C∞ (B ) such that
kφ̃N − φ N k1,p,B+ < 1/N.
Define φ̂N : B → R by
+
φ̃N (y) if y ∈ B
φ̂N (y) =
−3φ̃N (y1 , ..., yn−1 , −yn ) + 4φ̃N (y1 , ..., yn−1 , −yn /2) if y ∈ B− .
It may be easily verified that φ̂N ∈ C1 (B). Also, there exists C2 > 0 such that
kφ̂N k1,p,B ≤ C2 kφ̂N k1,p,B+

= C2 kφ̃N k1,p,B+
≤ C2 kφ N k1,p,B+ +C2 /N, (11.103)
where C2 depends only on Ω and p.

We claim that {φ̂N } is a Cauchy sequence in W 1,p (B).
For N1 , N2 ∈ N we have,
kφ̂N1 − φ̂N2 k1,p,B ≤ C1 kφ̂N1 − φ̂N2 k1,p,B+

≤ C1 kφ̃N1 − φ N1 + φ N1 − φ N2 + φ N2 − φ̂N2 k1,p,B+
≤ C1 kφ̃N1 − φ N1 k1,p,B+ +C1 kφ N1 − φ N2 k1,p,B+
+C1 kφ N2 − φ̂N2 k1,p,B+
≤ C1 /N1 +C1 kφ N1 − φ N2 k1,p,B+ +C1 /N2
→ 0, as N1 , N2 → ∞. (11.104)
Hence {φ̂N } is a Cauchy sequence and thus there exists û ∈ W 1,p (B) such that
φ̂N → û in norm in W 1,p (B).
In particular, up to a not relabeled subsequence of indices {N} of the sequence indicated in (11.102), we
have,
φ̂N → û, a.e. in B,
Observe that,
kφ̂N |B+ − φ N k1,p,B+ = kφÑ − φ N k1,p,B+ → 0, (11.105)
as N → ∞. From this, since φ N → u(x(y)), in W 1, p (B+ ), we obtain for an appropriate not relabeled subse-
quence
φ̂N |B+ → u(x(y)), in W 1,p (B+ ),
so that
û(y) = u(x(y)), a.e. in B+ ,
and thus,
û(y(x)) = u(x), a.e. in W + .
Denoting, u(x) = u(y(x)) in W = W + ∪W − , we obtain
u = u, a.e. in W + .
At this point we are going to show that
ku − φ̂N (y(x))k1,p,W → 0, as N → ∞.
Observe that,
Z
|uN (x)) − φ̂N (y(x))| p dx
W
Z
≤ |u(x(y)) − φ̂N (y)| p | det((ψ −1 )0 (y))| dy
B̂
Z
≤ K1 |û(y) − φ̂N (y)| p dy
B
p
≤ K1 kû − φ̂N k1,p,B → 0, as N → ∞. (11.106)
Similarly, for k ∈ {1, · · · , n}, we have,

p
∂ u(x) ∂ φ N (y(x))
Z
− dy
W ∂ xk ∂ xk
p
∂ uN (x(y)) ∂ φ N (y)
Z
≤ − | det((ψ −1 )0 (y))| d y
B̂ ∂ xk ∂ xk
∂ û(y) ∂ φN (y) ∂ y j (x(y)) p
Z
≤ − | det((ψ −1 )0 (y))| d y
B̂ ∂yj ∂yj ∂ xk
≤ K2 kû − φN k1p, p,B → 0, as N → ∞. (11.107)
Therefore,
ku(x) − φ̂ (y(x))k1,p,W → 0, as N → ∞.
Now choose ε > 0. Thus there exists N0 ∈ N such that if N > N0 we have
kuk1,p,W ≤ kφ̂N (y(x))k1,p,W + ε

≤ C3 kφ̂N (y)k1,p,B + ε
≤ C4 kφ̂N k1,p,B+ + ε
≤ C5 kφ N k1,p,W + + ε (11.108)
so that letting N → ∞, since ε > 0 is arbitrary we get
kuk1,p,W ≤ C5 kuk1,p,W + .
Now denoting W = Wx0 we have that ∂ Ω ⊂ ∪x0 ∈∂ ΩWx0 and since ∂ Ω is compact, there exist x1 , ..., xM ∈
∂ Ω, such that
∂ Ω ⊂ ∪M i=1Wxi .
Hence for an appropriate open W0 ⊂⊂ Ω, we get
Ω ⊂ ∪M
i=0Wi .
where we have denoted Wi = Wxi , ∀i ∈ {0, ..., M}.

Let {ζi }M M
i=0 be a concerned partition of unity relating {Wi }i=0 , so that
M
∑ ζi = 1, in Ω,
i=0
and ζi ∈ Cc∞ (Wi ), 0 ≤ ζi ≤ 1, ∀i ∈ {0, ..., M}.

Define
ui = ζi u, ∀i ∈ {0, ..., M}.
For each i we denote the extension of u from Wi+ to Wi by ui . Also define u0 = u in W0 , and u = ∑M
i=0 ζi ui .
Recalling that u = ui , a.e. on Wi+ , and that Ω = ∪M W
i=1 i
+
∪W 0 , we obtain
u = ∑M M
i=0 ζi ui = ∑i=0 ζi u = u, a.e. in Ω.
Furthermore
M
kuk1,p,Rn ≤ ∑ kζi ui k1,p,Rn
i=0
M
≤ C5 ∑ kui k1,p,Wi
i=0
M
≤ C5 kuk1,p,W0 +C5 ∑ kui k1,p,W +
i
i=1
≤ (M + 1)C5 kuk1,p,Ω
= Ckuk1,p,Ω , (11.109)
where C = (M + 1)C5 .
We recall that the partition of unity may be chosen so that its support is on V .
Finally, we denote Eu = u.
11.4.4 The main results

np
Definition 11.4.13 For 1 ≤ p < n we define r = n−p .
Theorem 11.4.14 (Gagliardo-Nirenberg-Sobolev inequality) Let 1 ≤ p < n. Thus there exists a constant
K > 0 depending only p and n such that
kukr,Rn ≤ KkDuk p,Rn , ∀u ∈ Cc1 (Rn ).
Proof 11.21 Suppose p = 1. Let u ∈ Cc1 (Rn ). From the fundamental theorem of calculus we have,
Z xi
∂ u(x1 , ..., xi−1 , yi , xi+1 , ..., xn )
u(x) = dyi ,
−∞ ∂ xi
so that Z ∞
|u(x)| ≤ |Du(x1 , ..., xi−1 , yi , xi+1 , ..., xn )| dyi .
−∞
Therefore,
n Z ∞
1/(n−1)
n/(n−1)
|u(x)| ≤∏ |Du(x1 , ..., xi−1 , yi , xi+1 , ..., xn )| dyi .
i=1 −∞
From this, we get,

Z ∞
|u(x)|n/(n−1) dx1
−∞
Z ∞ n Z ∞ 1/(n−1)
≤ ∏ |Du| dyi d x1
−∞ i=1 −∞
Z ∞
1/(n−1)
≤ |Du| dy1
−∞
Z ∞ n Z ∞
1/(n−1) !
× ∏ |Du| dyi dx1 . (11.110)
−∞ i=2 −∞
¨
From this and the generalized Holder inequality, we obtain,
Z ∞
|u(x)|n/(n−1) dx1
−∞
Z ∞
1/(n−1)
≤ |Du| dy1
−∞
n Z ∞ Z ∞ 1/(n−1)
×∏ |Du| dx1 dyi . (11.111)
i=2 −∞ −∞
Integrating in x2 we obtain,
Z ∞Z ∞
|u(x)|n/(n−1) dx1 dx2
−∞ −∞
Z ∞ Z ∞
1/(n−1)
≤ |Du| dy1
−∞ −∞
n Z ∞ Z ∞ 1/(n−1) !
×∏ |Du| dx1 d yi dx2 ,
i=2 −∞ −∞
so that
Z ∞Z ∞
|u(x)|n/(n−1) dx1 dx2
−∞ −∞
Z ∞ Z ∞ 1/(n−1)
≤ |Du| dy2 dx1
−∞ −∞
Z ∞ Z ∞
1/(n−1)
× |Du| dy1
−∞ −∞
n Z ∞ Z ∞ 1/(n−1) !
×∏ |Du| dx1 dyi dx2 . (11.112)
i=3 −∞ −∞
¨
By applying the generalized Holder inequality we get
Z ∞Z ∞
|u(x)|n/(n−1) dx1 dx2
−∞ −∞
Z ∞ Z ∞ 1/(n−1)
≤ |Du| dy2 dx1
−∞ −∞
Z ∞ Z ∞ 1/(n−1)
× |Du| dy1 dx2
−∞ −∞
n Z ∞ Z ∞Z ∞ 1/(n−1)
×∏ |Du| dx1 dx2 dyi . (11.113)
i=3 −∞ −∞ −∞
Therefore, reasoning inductively, after n steps we get

Z
|u(x)|n/(n−1) dx
Rn
n Z ∞ Z ∞ 1/(n−1)
≤ ∏ ... |Du| dx
i=1 −∞ −∞
Z n/(n−1)
= |Du| dx . (11.114)
Rn
This is the result for p = 1. Now suppose 1 1 apply the above result for
v = |u|γ ,
to obtain
Z (n−1)/n Z
γn/(n−1)
|u(x)| dx ≤ |D|u|γ |; dx
Rn Rn
Z
≤ γ |u|γ−1 |Du|; dx
Rn
Z (p−1)/p
≤ γ |u|(γ−1)p/(p−1) dx
Rn
Z 1/p
p
× |Du| dx . (11.115)
Rn
In particular for γ such that

γn (γ − 1)p
= ,
n−1 p−1
p(n−1)
that is γ = n−p , so that
γn (γ − 1) p np
= = ,
n−1 p−1 n− p
we get
Z ((n−1)/n−(p−1)/p) Z 1/p
r p
|u| dx ≤C |Du| dx .
Rn Rn
From this and considering that

n−1 p−1 n− p 1
− = = ,
n p np r
we finally obtain
Z 1/r Z 1/p
|u|r dx ≤C |Du| p dx .
Rn Rn
Theorem 11.4.15 Let Ω ⊂ Rn be a bounded open set. Suppose ∂ Ω is Ĉ1 , 1 ≤ p < n and u ∈ W 1,p (Ω).
Then u ∈ Lr (Ω) and
kukr,Ω ≤ Kkuk1,p,Ω ,
where the constant depends only on p, n and Ω.
Proof 11.22 Since ∂ Ω is Ĉ1 , from Theorem 11.4.12, there exists an extension Eu = ū ∈ W 1,p (Rn ) such
that ū = u in Ω the support of ū is compact and
kūk1,p,Rn ≤ Ckuk1,p,Ω ,
where C does not depend on u. As ū has compact support, from Theorem 11.4.10, there exists a sequence
{uk } ∈ Cc∞ (Rn ) such that
uk → ū in W 1,p (Rn ).
from the last theorem
kuk − ul kr,Rn ≤ KkDuk − Dul k p,Rn .
Hence,
uk → ū in Lr (Rn ).
also from the last theorem
kuk kr,Rn ≤ KkDuk k p,Rn , ∀k ∈ N,
so that
kūkr,Rn ≤ KkDūk p,Rn .
Therefore, we may get
kukr,Ω ≤ kūkr,Rn
≤ KkDūk p,Rn
≤ K1 kūk1,p,Rn
≤ K2 kuk1,p,Ω . (11.116)
Theorem 11.4.16 Let Ω ⊂ Rn be a bounded open set such that ∂ Ω ∈ Ĉ1 . If mp < n, then W m,p (Ω) ,→ Lq (Ω)
for p ≤ q ≤ (np)/(n − mp).
Proof 11.23 Define q0 = np/(n − mp). We first prove by induction on m that
W m,p ,→ Lq0 (Ω).
The last result is exactly the case for m = 1. Assume
W m−1,p ,→ Lr1 (Ω), (11.117)
where
r1 = np/(n − (m − 1)p) = np/(n − mp + p),
whenever n > (m − 1)p. If u ∈ W m,p (Ω) where n > mp, then u and D j u are in W m−1,p (Ω), so that from
(11.117) we have u ∈ W 1,r1 (Ω) and
kuk1,r1 ,Ω ≤ Kkukm,p,Ω . (11.118)
Since n > mp we have that r1 = np/((n − mp) + p) < n, from q0 = nr1 /(n − r1 ) = np/(n − mp) by the last
theorem we have
kukq0 ,Ω ≤ K2 kuk1,r1 ,Ω ,
where the constant K2 does not depend on u, and therefore from this and (11.118) we obtain
kukq0 ,Ω ≤ K2 kuk1,r1 ,Ω ≤ K3 kukm,p,Ω . (11.119)
The induction is complete. Now suppose p ≤ q ≤ q0 . Define
s = (q0 − q)p/(q0 − p) and t = p/s = (q0 − p)/(q0 − q).
¨
Through Holder’s inequality, we get
Z
kukqq,Ω = |u(x)|s |u(x)|q−s dx
Ω
Z 1/t Z 1/t 0
0
≤ |u(x)|st dx |u(x)|(q−s)t dx
Ω Ω
p/t q /t 0
= kuk p,Ω kukq00 ,Ω
p/t 0 q /t 0
≤ kuk p,Ω (K3 )q0 /t kukm,p,Ω
0
0 p/t q /t 0
≤ (K3 )q0 /t kukm,p,Ω kukm,p,Ω
0
0
= (K3 )q0 /t kukqm,p,Ω , (11.120)
since
p/t + q0 /t 0 = q.
Corollary 11.4.17 If mp = n, then W m,p (Ω) ,→ Lq for p ≤ q < ∞.
Proof 11.24 If q ≥ p0 = p/(p − 1) then q = ns/(n − ms) where s = pq/(p + q) is such that 1 ≤ s ≤ p.
Observe that
W m,p (Ω) ,→ W m,s (Ω)
with the imbedding constant depending only on |Ω|. Since ms < n, by the last theorem we obtain
W m,p (Ω) ,→ W m,s (Ω) ,→ Lq (Ω).

0
Now if p ≤ q ≤ p0 , from above we have W m,p (Ω) ,→ L p (Ω) and the obvious imbedding W m, p (Ω) ,→ L p (Ω).
Define s = (p0 − q)p/(p0 − p) and the result follows from a reasoning analogous to the final chain of inequal-
ities of last theorem, indicated in (11.120).
About the next theorem, note that its hypotheses are satisfied if ∂ Ω is Ĉ1 (here we do not give the details).
Theorem 11.4.18 Let Ω ⊂ Rn be an open bounded set, such that for each x ∈ Ω there exists a convex set
Cx ⊂ Ω whose shape depends on x, but such that |Cx | > α, for some α > 0 that does not depend on x. Thus
if mp > n, then
W m,p (Ω) ,→ CB0 (Ω).
Proof 11.25 Suppose first m = 1 so that p > n. Fix x ∈ Ω and pick y ∈ Cx . For φ ∈ C∞ (Ω), from the
fundamental theorem of calculus, we have
Z 1
d (φ (x + t(y − x))
φ (y) − φ (x) = dt.
0 dt
Thus, Z 1
d (φ (x + t(y − x))
|φ (x)| ≤ |φ (y)| + dt,
0 dt
and hence Z 1
d(φ (x + t(y − x))
Z Z Z
|φ (x)| dy ≤ |φ (y)| d y + dt dy,
Cx Cx Cx 0 dt
¨
so that, from Holder’s inequality and Fubini theorem we get,
|φ (x)|α ≤ |φ (x)| · |Cx |
0
≤ kφ k p,Ω |Cx |1/ p
Z 1Z
d(φ (x + t(y − x)) (11.121)
+ dy dt.
0 Cx dt
Therefore Z 1Z
1/p0
|φ (x)|α ≤ kφ k p,Ω |Ω| + |∇φ (z)|δt −n dz dt,
0 V
where |V | = t n |Cx | and δ denotes the diameter of Ω. From Holder’
¨ s inequality again, we obtain
Z 1 Z 1/p
1/p0 0
|φ (x)|α ≤ kφ k p,Ω |Ω| +δ |∇φ (z )| dz p
t −n (t n |Cx |)1/p dt,
0 V
and thus Z 1
0 0 0
|φ (x)|α ≤ kφ k p,Ω |Ω|1/p + δ |Cx |1/p k∇φ k p,Ω t −n(1−1/p ) dt.
0
Since p > n we obtain
Z 1 Z 1
0 1
t −n(1−1/p ) dt = t −n/p dt = .
0 0 1 − n/p
From this, the last inequality and from the fact that |Cx | ≤ |Ω|, we have that there exists K > 0 such that
|φ (x)| ≤ Kkφ k1,p,Ω , ∀x ∈ Ω, φ ∈ C∞ (Ω). (11.122)
Here the constant K depends only on p, n and Ω. Consider now u ∈ W 1,p (Ω).
Thus there exists a sequence {φk } ⊂ C∞ (Ω) such that
φk → u, in W 1,p (Ω).
Up to a not relabeled subsequence, we have
φk → u, a.e. in Ω. (11.123)
Fix x ∈ Ω such that the limit indicate in (11.123) holds. Suppose given ε > 0. Therefore, there exists k0 ∈ N
such that
|φk0 (x) − u(x)| ≤ ε/2
and
kφk0 − uk1,p,Ω < ε/(2K).
Thus,
|u(x)| ≤ |φk0 (x)| + ε/2

≤ Kkφk0 k1,p,Ω + ε/2
≤ Kkuk1,p,Ω + ε. (11.124)
Since ε > 0 is arbitrary, the proof for m = 1 is complete, because for {φk } ∈ C∞ (Ω) such that φk → u
in W 1,p (Ω), from (11.122) we have that {φk } is a uniformly Cauchy sequence, so that it converges to a
continuous u∗ , where u∗ = u, a.e. in Ω.
For m > 1 but p > n we still have
|u(x)| ≤ Kkuk1,p,Ω ≤ K1 kukm,p,Ω , a.e. in Ω, ∀u ∈ W m,p (Ω).
If p ≤ n ≤ mp, there exists j satisfying 1 ≤ j ≤ m − 1 such that j p ≤ n ≤ ( j + 1)p. If j p < n set
r̂ = np/(n − j p),
Let 1 ≤ p1 ≤ n such that

r̂ = np1 /(n − p1 ).
Thus we have that
np/(n − jp) = np1 /(n − p1 ),
so that
p1 = np/(n − ( j − 1)p),
so that by above and the last theorem:
kuk∞ ≤ K1 kuk1,r̂,Ω ≤ K1 kukm− j,r̂,Ω ≤ K2 kukm−( j−1),p1 ,Ω .
Now define
r̂1 = p1 = np/(n − ( j − 1)p)
and 1 ≤ p2 ≤ n such that
r̂1 = np2 /(n − p2 ),
so that
np/(n − ( j − 1)p) = np2 /(n − p2 ).
Hence p2 = np/(n − ( j − 2)p) so that by the last theorem
kukm−( j−1),p1 ,Ω = kukm−( j−1),r̂1 ,Ω ≤ K3 kukm−( j−2),p2 ,Ω .

Proceeding inductively in this fashion, after j steps, observing that p j = p, we get

kuk∞ ≤ K1 kuk1,r̂,Ω ≤ K1 kukm− j,r̂,Ω ≤ K j kukm,p,Ω ,
for some appropriate K j . Finally, if jp = n choosing r̂ = max{n, p} also by the last theorem we obtain the
same last chain of inequalities. For, assume r̂ = max{n, p} = n > p. Let p1 be such that
np1
r1 = = n,
n − p1
That is
n
p1 = .
2
Since n > p we have that n ≥ 2 so that 1 ≤ p1 < n. From the last theorem we obtain
kuk∞ ≤ Ckukm− j,r1 ,Ω ≤ C1 kukm−( j−1),p1 ,Ω .
Let r2 = p1 = n/2, define p2 such that
np2
r2 = n/2 = ,
n − p2
that is p2 = n/3.
Hence, again by the last theorem we get
kuk∞ ≤ C1 kukm−( j−1),r2 ,Ω ≤ C2 kukm−( j−2),p2 ,Ω .
Reasoning inductively, after j − 1 steps, , we get p j−1 = n/ j = p, so that
kuk∞ ≤ Ckukm−( j−1),r1 ,Ω ≤ C3 kukm−( j−( j−1)),p j−1 ,Ω ≤ C4 kukm,p,Ω .
Finally, if r1 = max{n, p} = p ≥ n, Define p1 such that
np1
r1 = p =
n − p1
that is,
np
p1 = ≤ p,
n+ p
so that by last theorem
kuk∞ ≤ kukm− j,r1 ,Ω ≤ C5 kukm−( j−1),p1 ,Ω ≤ C6 kukm,p,Ω .
Theorem 11.4.19 Let Ω ⊂ Rn be an open and bounded set with a boundary Ĉ1 . If mp > n, then W m, p (Ω) ,→
Lq (Ω) for p ≤ q ≤ ∞.
Proof 11.26 From the proof of the last theorem, we may obtain
kuk∞,Ω ≤ Kkukm,p,Ω , ∀u ∈ W m,p (Ω).
If p ≤ q < ∞, we have
Z
kukqq,Ω = |u(x)| p |u(x)|q−p dx
Ω
Z q− p
≤ |u(x)| p Kkukm,p,Ω dx
Ω
≤ K q−p
kuk pp,Ω kukq−p
m,p,Ω
p
≤ K q−p kukm,p,Ω kukq−p
m,p,Ω
= K q−p kukqm,p,Ω . (11.125)
Theorem 11.4.20 Let S ⊂ Rn be a n-dimensional ball of radius bigger than 3. If n < p, then there exists a
constant C, depending only on p and n, such that
kukC0,λ (S) ≤ Ckuk1,p,S , ∀u ∈ C1 (S),
where 0 < λ ≤ 1 − n/p.
Proof 11.27 First consider λ = 1 − n/p and u ∈ C1 (S). Let x, y ∈ S such that |x − y| < 1 and define
σ = |x − y|. Consider a fixed cube denoted by Rσ ⊂ S such that |Rσ | = σ n and x, y ∈ R̄σ . For z ∈ Rσ , we may
write: Z 1
du(x + t(z − x))
u(x) − u(z) = − dt,
0 dt
that is, Z Z Z 1
u(x)σ n = u(z) dz − ∇u(x + t(z − x)) · (z − x) dt dz.
Rσ Rσ 0
Thus, denoting in the next lines V by an appropriate set such that |V | = t n |Rσ |, we obtain
Z
√ Z Z 1
|u(x) − u(z) dz/σ n | ≤ nσ 1−n |∇u(x + t(z − x))| dt dz
Rσ Rσ 0
√ Z 1 Z
≤ nσ 1−n t −n |∇u(z)| d z dt
0 V
√ Z 1
0
≤ nσ 1−n t −n k∇uk p,S |V |1/p dt
0
√ 1−n n/p0 Z 1
0
≤ nσ σ k∇uk p,S t −nt n/ p dt
0
√ Z 1
≤ nσ 1−n/p k∇uk p,S t −n/p dt
0
1−n/p
≤σ kuk1,p,S K, (11.126)
where
√ Z 1 −n/p √
K= n t dt = n/(1 − n/p).
0
A similar inequality holds with y in place of x, so that
|u(x) − u(y)| ≤ 2K|x − y|1−n/p kuk1,p,S , ∀x, y ∈ Rσ .
Now consider 0 < λ < 1 − n/p. Observe that, as |x − y|λ ≥ |x − y|1−n/p if |x − y| < 1, we have,

|u(x) − u(y)|
sup |x=
6 y, |x − y| < 1
x,y∈S |x − y|λ

|u(x) − u(y)|
≤ sup | x 6
= y, |x − y| < 1 ≤ Kkuk1,p,S . (11.127)
x,y∈S |x − y|1−n/ p
Also,

|u(x) − u(y)|
sup | |x − y| ≥ 1 ≤ 2kuk∞,S ≤ 2K1 kuk1,p,S
x,y∈S |x − y|λ
so that
|u(x) − u(y)|
sup 6 y ≤ (K + 2K1 )kuk1,p,S , ∀u ∈ C1 (S).
|x=
x,y∈S |x − y|λ
Theorem 11.4.21 Let Ω ⊂ Rn be an open bounded set such that ∂ Ω is Ĉ1 Assume n < p ≤ ∞.
Then
W 1,p (Ω) ,→ C0,λ (Ω),
for all 0 < λ ≤ 1 − n/p.
Proof 11.28 Fix 0 < λ ≤ 1 − n/p and let u ∈ W 1,p (Ω). Since ∂ Ω is Ĉ1 , from Theorem 11.4.12, there exists
an extension Eu = ū such that ū = u, a.e. in Ω, and
kūk1,p,Rn ≤ Kkuk1,p,Ω ,
where the constant K does not depend on u. From the proof of this same theorem, we may assume that spt(ū)
is on a n-dimensional sphere S ⊃ Ω with sufficiently big radius and such sphere does not depend on u. Thus,
in fact, we have
kūk1,p,S ≤ Kkuk1,p,Ω .
Since C∞ (S) is dense in W 1,p (S), there exists a sequence {φk } ⊂ C∞ (S) such that
uk → ū, in W 1,p (S). (11.128)
Up to a not relabeled subsequence, we have
uk → ū, a.e. in Ω.
From last theorem we have

kuk − ul kC0,λ (S) ≤ Ckuk − ul k1,p,S ,
so that {uk } is a cauchy sequence in C0,λ (S), and thus uk → u∗ for some u∗ ∈ C0,λ (S). Hence, from this and
(11.128), we have
u∗ = ū, a.e. in S.
Finally, from above and last theorem we may write:
ku∗ kC0,λ (Ω) ≤ ku∗ kC0,λ (S) ≤ K1 kūk1,p,S ≤ K2 kuk1,p,Ω .
11.5 The trace theorem

In this section we state and prove the trace theorem.
Theorem 11.5.1 Let 1 < p < ∞ and let Ω ⊂ Rn be an open bounded set such that ∂ Ω is Ĉ1 . Then there exists
a bounded linear operator
T : W 1,p (Ω) → L p (∂ Ω),
such that
Tu = u|∂ Ω if u ∈ W 1,p (Ω) ∩C(Ω), and

kTuk p,∂ Ω ≤ Ckuk1,p,Ω , ∀u ∈ W 1,p (Ω),
where the constant C depends only on p and Ω.
Proof 11.29 Let u ∈ W 1,p (Ω) ∩C(Ω). Choose x0 ∈ ∂ Ω.

Since ∂ Ω is Ĉ1 , there exists r > 0 such that for a local coordinate system we may write,
Ω ∩ Br (x0 ) = {x ∈ Br (x0 ) | xn ≤ f (x1 , ..., xn−1 )},
where denoting x̂ = (x1 , ..., xn−1 ), f (x̂) is continuous and such that its partial derivatives are classically de-
fined a.e. and bounded on its domain. Furthermore
| f (x̂) − f (ŷ)| ≤ K|x̂ − ŷ|2 , ∀x̂, ŷ
for some K > 0 also on its domain.

Define the coordinates y by
yi = xi , ∀i ∈ {1, ..., n − 1},
and
yn = f (x1 , ..., xn−1 ) − xn .
Define û(y) by,
u(x1 , ..., xn ) = u(y1 , ..., yn−1 , f (y1 , ..., yn−1 ) − yn ) = û(y).
Also define y0 = (x01 , ..., x0n−1 , x0n − f (x01 , ..., x0n−1 ) = (y01 , ..., y0n−1 , 0) and choose r1 > 0 such that
Ψ−1 (B+
r1 (y0 )) ⊂ Ω ∩ Br (x0 ).
Observe that this is possible since Ψ and Ψ−1 are continuous, where y = Ψ(x). Here
B+
r1 (y0 ) = {y ∈ Br1 (y0 ) | yn > 0}.
+
For each N ∈ N, choose, by mollification for example, φN ∈ C∞ (Br1 (y0 )) such that
1
kφN − ûk∞,B+ (y ) < .
r1 0 N
Denote B = Br1 /2 (y0 ), and B+ = B+

r1 /2 (y0 ). Now choose η ∈ Cc (Br1 (y0 )), such that η > 0 and η ≡ 1 on B.
∞
Also denote
Γ̃ = {y ∈ B | yn = 0},
and
Γ̃1 = {y ∈ Br1 (y0 ) | yn = 0}.
Observe that
Z Z
|φN | p dΓ ≤ η |φN | p dΓ
Γ̃ Γ̃1
Z
= − (η|φN | p )yn dy
B+
r1
Z
≤ − (ηyn |φN | p ) dy
B+
r1
Z
+ (p|φN | p−1 |(φN )yn |η) d y. (11.129)
B+
r1
Here we recall the Young inequality
a p bq 1 1
ab ≤ + , ∀a, b ≥ 0, where + = 1.
p q p q
Thus,
|(φN )yn | p η p |φN |(p−1)q
(|φN | p−1 |)(|(φN )yn |η) ≤ + ,
p q
so that replacing such an inequality in (11.129), since (p − 1)q = p we get
Z Z Z
!
p p p
|φN | dΓ ≤ C1 |φN | dy + |DφN | dy . (11.130)
Γ̃ B+
r1 B+
r1
Letting N → +∞ we obtain
Z Z
|u(x)| p dΓ ≤ C2 |û(y)| p dΓ
Γ Γ̃
Z Z
!
≤ C3 |û| p dy + |Dû| p dy
B+
r1 Br+1
Z Z
≤ C4 |u| p dx + |Du| p d x . (11.131)
W+ W+
where Γ = ψ −1 (Γ̃) and W + = Ψ−1 (B+ r1 ).

Observe that denoting W = Wx0 we have that ∂ Ω ⊂ ∪x∈∂ ΩWx , and thus, since ∂ Ω is compact, we may
select x1 , ..., xM such that ∂ Ω ⊂ ∪M
i=1Wi . We emphasize to have denoted Wxi = Wi , ∀i ∈ {1, ..., M}. Denoting
Wi+ = Wi ∩ Ω we may obtain,
Z M Z
|u(x)| p dΓ ≤ ∑ |u(x)| p dΓ
∂Ω i=1 Γi
M Z Z
p p
≤ ∑ C4i |u| dx + |Du| dx
i=1 Wi+ Wi+
Z Z
≤ C5 M |u| p dx + |Du| p dx
Ω Ω
Z Z
= C |u| p dx + |Du| p dx . (11.132)
Ω Ω
At this point we denote Tu = u|∂ Ω .

Finally, for the case u ∈ W 1,p (Ω), select {uk } ⊂ C∞ (Ω) such that
kuk − uk1,p,Ω → 0, as k → ∞.
From above
kTuk − Tul k p,∂ Ω ≤ Ckuk − ul k1,p,Ω ,
so that
{Tuk }
is a Cauchy sequence. Hence we may define
Tu = lim Tuk , in L p (∂ Ω).

k→∞
Remark 11.5.2 Similar results are valid for W0m,p , however in this case the traces relative to derivatives of
order up to m − 1 are involved.
11.6 Compact imbeddings

Theorem 11.6.1 Let m be a non-negative integer and let 0 < ν < λ ≤ 1. Then the following imbeddings
exist:
Cm+1 (Ω) ,→ Cm (Ω), (11.133)
Cm,λ (Ω) ,→ Cm (Ω), (11.134)
Cm,λ (Ω) ,→ Cm,ν (Ω). (11.135)
If Ω is bounded then imbeddings (11.134) and (11.135) are compact.
Proof 11.30 The Imbeddings (11.133) and (11.134) follows from the inequalites
kφ kCm (Ω) ≤ kφ kCm+1 (Ω) ,
kφ kCm (Ω) ≤ kφ kCm,λ (Ω) .

To establish (11.135) note that for |α| ≤ m
α
|D φ (x) − Dα φ (y)|
sup | x 6= y, |x − y| < 1
x,y∈Ω |x − y|ν
α
|D φ (x) − Dα φ (y)|
≤ sup | x 6= y, |x − y| < 1 , (11.136)
x,y∈Ω |x − y|λ
and also,
α
D φ (x) − Dα φ (y)|
sup | | |x − y| ≥ 1 ≤ 2 sup{|Dα φ |}. (11.137)
x,y∈Ω |x − y|ν x∈Ω
Therefore, we may conclude that
kφ kCm,ν (Ω̄) ≤ 3kφ kCm,λ (Ω) , ∀φ ∈ Cm,ν (Ω).
Now suppose Ω is bounded. If A is a bounded set in C0,λ (Ω) then there exists M > 0 such that
kφ kC0,λ (Ω) ≤ M, ∀φ ∈ A.
But then
|φ (x) − φ (y)| ≤ M|x − y|λ , ∀x, y ∈ Ω, φ ∈ A,
so that by the Ascoli-Arzela theorem, A is pre-compact in C(Ω). This proves the compactness of (11.134)
for m = 0.
If m ≥ 1 and A is bounded in Cm,λ (Ω), then A is bounded in C0,λ (Ω). Thus, by above there is a sequence
{φk } ⊂ A and φ ∈ C0,λ (Ω) such that
φk → φ in C(Ω).
However, {Di φk } is also bounded in C0,λ (Ω), so that there exist a not relabeled subsequence, also denoted
by {φk } and ψi such that
Di φk → ψi , in C(Ω).
The convergence in C(Ω̄) being the uniform one, we have ψi = Di φ . We can proceed extracting (not rela-
beled) subsequences until obtaining
Dα φk → Dα φ , in C(Ω), ∀ 0 ≤ |α| ≤ m.
This completes the proof of compactness of (11.134). For (11.135), let S be a bounded set in Cm,λ (Ω).
Observe that
|D φ (x) − Dα φ (y)| ν /λ
α
|Dα φ (x) − Dα φ (y)|
=
|x − y|ν |x − y|λ
·|Dα φ (x) − Dα φ (y)|1−ν /λ
≤ K|Dα φ (x) − Dα φ (y)|1−ν /λ , (11.138)
for all φ ∈ S. From (11.134), S has a converging subsequence in Cm (Ω). From (11.138) such a subsequence
is also converging in Cm,ν (Ω). The proof is complete.
Theorem 11.6.2 (Rellich-Kondrachov) Let Ω ⊂ Rn be an open bounded set such that ∂ Ω is Ĉ1 . Let j, m
be integers, j ≥ 0, m ≥ 1, and let 1 ≤ p < ∞.
1. Part I- If mp < n, then the following imbeddings are compact:
W j+m,p (Ω) ,→ W j,q (Ω),

if 0 < n − mp < n and 1 ≤ q < np/(n − mp), (11.139)
W j+m,p (Ω) ,→ W j,q (Ω), if n = mp, 1 ≤ q < ∞. (11.140)
2. Part II- If mp > n, then the following imbeddings are compact:
W j+m,p ,→ CBj (Ω), (11.141)
W j+m,p (Ω) ,→ W j,q (Ω), if 1 ≤ q ≤ ∞. (11.142)
3. Part III-The following imbeddings are compact:
W j+m,p (Ω) ,→ C j (Ω), if mp > n, (11.143)
W j+m,p (Ω) ,→ C j,λ (Ω),

if mp > n ≥ (m − 1)p and 0 < λ < m − n/p. (11.144)
4. Part IV- All the above imbeddings are compact if we replace W j+m,p (Ω) by W0j+m,p (Ω).
Remark 11.6.3 Given X,Y, Z spaces, for which we have the imbeddings X ,→ Y and Y ,→ Z and if one of
these imbeddings is compact then the composite imbedding X ,→ Z is compact. Since the extension oper-
ator u → ũ where ũ(x) = u(x) if x ∈ Ω and ũ(x) = 0 if x ∈ Rn − Ω, defines an imbedding W0j+m,p (Ω) ,→
W j+m,p (Rn ) we have that Part-IV of above theorem follows from the application of Parts I-III to Rn (despite
the fact we are assuming Ω bounded, the general results may be found in Adams [2]).
Remark 11.6.4 To prove the compactness of any of above imbeddings it is sufficient to consider the case
j = 0. Suppose, for example, that the first imbedding has been proved for j = 0. For j ≥ 1 and {ui } bounded
sequence in W j+m,p (Ω) we have that {Dα ui } is bounded in W m,p (Ω) for each α such that |α| ≤ j. From the
case j = 0 it is possible to extract a subsequence (similarly to a diagonal process) {uik } for which {Dα uik }
converges in Lq (Ω) for each α such that |α| ≤ j, so that {uik } converges in W j,q (Ω).
Remark 11.6.5 Since Ω is bounded, CB0 (Ω) ,→ Lq (Ω) for 1 ≤ q ≤ ∞. In fact
kuk0,q,Ω ≤ kukC0 [vol(Ω)]1/q . (11.145)

B
Thus the compactness of (11.142) (for j = 0) follows from that of (11.141).
Proof of Parts II and III. If mp > n > (m − 1)p and 0 < λ < (m − n)/p, then there exists µ such that
λ < µ < m − (n/p). Since Ω is bounded, the imbedding C0,µ (Ω) ,→ C0,λ (Ω) is compact by Theorem 11.6.1.
Since by the Sobolev Imbedding Theorem we have W m,p (Ω) ,→ C0,µ (Ω), we have that imbedding (11.144)
is compact.
If mp > n, let j∗ be the non-negative integer satisfying (m − j∗ )p > n ≥ (m − j∗ − 1)p. Thus we have the
chain of imbeddings
∗ ,p
W m,p (Ω) ,→ W m− j (Ω) ,→ C0,µ (Ω) ,→ C(Ω), (11.146)
where 0 < µ < m − j∗ − (n/p). The last imbedding in (11.146) is compact by Theorem 11.6.1, so that
(11.143) is compact for j = 0. By analogy (11.141) is compact for j = 0. Therefore from the above remarks,
(11.142) is also compact. For the proof of Part I, we need the following lemma:
Lemma 11.6.6 Let Ω be an bounded domain in Rn . Let 1 ≤ q1 ≤ q0 and suppose
W m,p (Ω) ,→ Lq0 (Ω), (11.147)
W m,p (Ω) ,→ Lq1 . (11.148)
Suppose also that (11.148) is compact. If q1 ≤ q < q0 , then the imbedding
W m,p ,→ Lq (Ω) (11.149)
is compact.
Proof 11.31 Define λ = q1 (q0 − q)/(q(q0 − q1 )) and µ = q0 (q − q1 )/(q(q0 − q1 )). We have that λ > 0
and µ ≥ 0. From Holder’
¨ s inequality and (11.147) there exists K ∈ R+ such that,
λ µ λ µ
kuk0,q,Ω ≤ kuk0,q1 ,Ω
kuk0,q0 ,Ω ≤ Kkuk0,q 1 ,Ω
kukm,p,Ω ,
∀u ∈ W m,p (Ω). (11.150)
Thus considering a sequence {ui } bounded in W m,p (Ω), since (11.148) is compact there exists a subsequence
{unk } that converges, and is therefore a Cauchy sequence in Lq1 (Ω). From (11.150), {unk } is also a Cauchy
sequence in Lq (Ω), so that (11.149) is compact.
Proof of Part I. Consider j = 0. Define q0 = np/(n − mp). To prove the imbedding
W m,p (Ω) ,→ Lq (Ω), 1 ≤ q < q0 , (11.151)

is compact, by last lemma it suffices to do so only for q = 1. For k ∈ N, define
Ωk = {x ∈ Ω | dist(x, ∂ Ω) > 2/k}. (11.152)
Suppose A is a bounded set of functions in W m,p (Ω), that is suppose there exists K1 > 0 such that
kukW m,p (Ω) < K1 , ∀u ∈ A.
Also, suppose given ε > 0, and define, for u ∈ W m,p (Ω), ũ(x) = u(x) if x ∈ Ω, ũ(x) = 0, if x ∈ Rn \ Ω. Fix
u ∈ A. From Holder’
¨ s inequality and considering that W m,p (Ω) → Lq0 (Ω), we have
Z Z 1/q0 Z 1−1/q0
q0
|u(x)|dx ≤ |u(x)| dx 1dx
Ω−Ωk Ω−Ωk Ω−Ωk
≤ K1 kukm,p,Ω [vol(Ω − Ωk )]1−1/q0 , (11.153)
Thus, since A is bounded in W m,p (Ω), there exists K0 ∈ N such that if k ≥ K0 then
Z
|u(x)|dx < ε, ∀u ∈ A (11.154)
Ω−Ωk
and, now fixing a not relabeled k > K0 , we get,

Z
|ũ(x + h) − ũ(x)|dx < 2ε, ∀u ∈ A, ∀h ∈ Rn . (11.155)
Ω−Ωk
Observe that if |h| < 1/k, then x + th ∈ Ω2k provided x ∈ Ωk and 0 ≤ t ≤ 1. If u ∈ C∞ (Ω) we have that
Z 1
d u(x + th)
Z Z
|u(x + h) − u(x)| ≤ dx dt
Ωk Ωk 0 dt
Z 1 Z
≤ |h| dt |∇u(y)|dy ≤ |h|kuk1,1,Ω
0 Ω2k
≤ K2 |h|kukm,p,Ω . (11.156)
Since C∞ (Ω) is dense in W m,p (Ω), from above for |h| sufficiently small
Z
|ũ(x + h) − ũ(x)|dx < 3ε, ∀u ∈ A. (11.157)
Ω
From Theorem 11.4.6, A is relatively compact in L1 (Ω) and therefore the imbedding indicated (11.151) is
compact for q = 1. This completes the proof.
Before introducing the exercises, we work out some examples.
We start start with the following J.L. Lions Lemma.
Lemma 11.1
Let U,V,W be Banach spaces. Assume U ,→ V , where such an imbbeding is compact, and V ,→ W .
Under such assumptions, for each ε > 0 there exists C(ε) > 0 such that
kukV ≤ εkukU +C(ε)kukW , ∀u ∈ U.
Proof 11.32 Suppose, to obtain contradiction, there exists ε0 > 0 such that for each n ∈ N we may find
un ∈ U such that
kun kV > ε0 kun kU + nkun kW . (11.158)
Observe that from U ,→ V , there exists K1 > 0 such that
kukV ≤ K1 kukU , ∀u ∈ U,
6 0, ∀n ∈ N, we also obtain,
and since from (11.158), kun kV > 0 so that un =
un
≤ K1 , ∀n ∈ N. (11.159)
kun kU V
From this and (11.158), we have,
un un
K1 ≥ > ε0 + n , ∀n ∈ N. (11.160)
kun kU V kun kU W
From (11.159), the sequence

un
kun kU
is bounded in V , and since U ,→ V is compact, there exists a subsequence {nk } of N and u0 ∈ V such that
unk
− u0 → 0, as k → ∞.
kunk kU V
From this and V ,→ W we obtain,

unk
− u0 → 0, as k → ∞.
kunk kU W
Observe that from (11.160),

un
> ε0 , ∀nN, (11.161)
kun kU V
so that
ku0 kV ≥ ε0 > 0,
which means,
6 0.
u0 =
Let ε = ku0 kW /2. From (11.161), there exists k0 ∈ N such that if k > k0 , then
unk
− u0 < ε = ku0 kW /2,
kunk kU W
so that
unk
≥ ku0 kW /2, if k > k0 .
kunk kU W
From this and (11.160), we get,
u nk
K1 > ε0 + n k
kunk kU W
ku0 kW
≥ ε0 + n k , if k > k0 . (11.162)
2
6 0.
This contradicts u0 =
Example 11.6.7 With such a result in mind, we work out the following example.
Let Ω ⊂ Rn be an open bounded set with a Ĉ1 class boundary.
Let 1 0. From the last lemma with U = W 1,p (Ω), V = Lq (Ω) and W = L1 (Ω), there exists C(ε) > 0
such that
kukq ≤ εkuk1,p +C(ε)kuk1 ∀u ∈ W 1,p (Ω). (11.163)
Exercise 11.6.8 Let Ω ⊂ Rn be an open bounded set with a Ĉ1 class boundary.
Let p > n.
Prove that for each ε > 0 there exists C(ε) > 0 such that
kDuk∞ ≤ εkuk2,p +C(ε)kuk∞ , ∀u ∈ W 2,p (Ω).
Hint: From the Rellich-Kondrachov theorem, the imbbeding W 2,p (Ω) ,→ C1 (Ω) is compact.
Theorem 11.6.9 Let Ω ⊂ Rn be an open set.

Let 1 0 such that
Z
∂ϕ
u dx ≤ K j kϕk p0 , ∀ϕ ∈ Cc1 ,
Ω ∂xj
where p0 is such that

1 1
+ = 1.
p p0
Under such assumptions we have,
u ∈ W 1,p (Ω).
Proof 11.33 Fix j ∈ {1, · · · , n}.

Observe that, from the hypotheses, the functional Fu : Cc1 (Ω) → R where
Z
∂ϕ
Fu (ϕ) = u dx,
Ω ∂xj
0
is bounded on Cc1 (Ω), where Cc1 (Ω) is a not closed subspace of L p .
0
From the Hahn-Banach theorem, Fu may be extended to a F̃u : L p (Ω) → R with same norm as Fu , and
such that
F̃u (ϕ ) = Fu (ϕ ), ∀ϕ ∈ Cc1 (Ω).
From the Riesz representation theorem, since L p (Ω) is reflexive, there exists f j ∈ L p (Ω) such that
Z
0
F̃u (v) = f j v dx, ∀v ∈ L p ,
Ω
and in particular,
Z
Fu (ϕ) = F̃u (ϕ) = f j ϕ dx, ∀ϕ ∈ Cc1 (Ω),
Ω
so that
Z Z
∂ϕ
u dx = Fu (ϕ) = f ϕ dx, ∀ϕ ∈ Cc1 (Ω).
Ω ∂xj Ω
From this, in the distributional sense,

∂u
= f j ∈ L p (Ω), ∀ j ∈ {1, · · · , n},
∂xj
so that
u ∈ W 1,p (Ω).
Example 11.6.10 In this example, let Ω ⊂ Rn be an open bounded set. Let 1 < p < ∞.
Assume u ∈ W 1,p (Ω).
Let f : R → R be such that
| f (x) − f (y)| ≤ K|x − y|, ∀x, y ∈ R.
( f ◦ u) ∈ W 1,p (Ω).
First observe that, from the hypotheses,
| f (u(x)) − f (0)| ≤ K|u(x)|, a. e. in Ω,
so that
| f (u(x))| ≤ K|u(x)| + | f (0)|, a.e. in Ω.
Since ( f ◦ u) is measurable, we may infer that
( f ◦ u) ∈ L p (Ω).
Observe that
ϕ(x + he j ) − ϕ(x) ∂ ϕ(x)
f (u(x)) → f (u(x)) , as h → 0, a.e. in Ω,
h ∂xj
ϕ(x + he j ) − ϕ(x)
f (u(x)) ≤ | f (u(x))|C,
h
where
∂ ϕ(x)
C = max .
x∈Ω ∂xj
Recalling that ( f ◦ u) ∈ LP (Ω) ⊂ L1 (Ω), from the Lebesgue dominated convergence theorem, we obtain,
ϕ(x + he j ) − ϕ(x)
Z
lim ( f (u(x)) dx
h→0 Ω h
Z
∂ϕ
= ( f (u(x)) dx. (11.164)
Ω ∂xj
Observe that, for any |h| > 0 sufficiently small,
ϕ(x + he j ) − ϕ (x)
Z
f (u(x)) dx
Ω h
f (u(x − he j )) − f (u(x))
Z
= ϕ(x) dx
Ω h
| f (u(x − he j )) − f (u(x))|
Z
≤ |ϕ (x)| dx
Ω |h|
K|u(x − he j ) − u(x)|
Z
≤ |ϕ(x)| dx
Ω |h|
u(x − he j ) − u(x)
≤ K kϕ k p0
h p
≤ KC1 kϕk p0 . (11.165)
From this and (11.164), we obtain,

Z
∂ϕ
f (u(x)) dx ≤ KC1 kϕk p0 ,
Ω ∂xj
∀ϕ ∈ Cc1 (Ω), ∀ j ∈ {1, · · · , n}, so that from the last theorem,
( f ◦ u) ∈ W 1,p (Ω).
11.7 On the regularity of Laplace equation solutions

Let Ω ⊂ Rn be an open bounded set. Consider the functions ai j : Ω → R where we assume ai j ∈ C1 (Ω)
∀i, j ∈ {1, · · · , n}, where we also assume
ai j ξi ξ j ≥ c|ξ |2 , ∀ξ ∈ Rn .
Let f ∈ L2 (Ω) and consider the problem of minimizing J : U = W01,2 (Ω) → R, where,
1 ∂u ∂u
Z Z
J(u) = ai j dx − f u dx,
2 Ω ∂ xi ∂ x j Ω
∀u ∈ U.
Observe that J is strictly convex, coercive and continuous, so that it is lower semi-continuous.
Hence, from the direct method of calculus of variations, there exists a unique u0 ∈ W01,2 (Ω) such that
J(u0 ) = min J(u).

u∈U
Moreover, from the necessary optimal condition,
δ J(u0 , ϕ) = 0, ∀ϕ ∈ U = W01,2 (Ω),
we obtain,
∂ u0 ∂ ϕ
Z Z
ai j dx − f ϕ dx = 0, (11.166)
Ω ∂ xi ∂ x j Ω
∀ϕ ∈ U.
Let V,W ⊂ Rn be open sets such that

V ⊂⊂ W ⊂⊂ Ω,
Chose η : Rn → R such that, η ∈ Cc1 (Rn ) and such that
η ≡ 1, on V ,
spt(η) ⊂ W,
and
0 ≤ η(x) ≤ 1, ∀x ∈ Rn .
Select k ∈ {1, · · · , n} and define
ϕ(x) = −D−h 2 h
k [η Dk u0 (x)],
where generically,
u(x + hek ) − u(x)
Dhk u(x) = , ∀h =
6 0.
|h|
Observe that, for |h| > 0 sufficiently small, ϕ ∈ U = W01,2 (Ω), so that from (11.166), we obtain,
∂ u0 ∂ D−h 2 h
k [η Dk u0 (x)]
Z
− ai j dx
Ω ∂ xi ∂xj
Z
+ f Dk−h [η 2 Dkh u0 (x)] dx = 0, (11.167)
Ω
Define,
∂ u0 ∂ D−h 2 h
k [η Dk u0 (x)]
Z
A=− ai j dx,
Ω ∂ xi ∂xj
and Z
B= f D−h 2 h
k [η Dk u0 (x)] dx.
Ω
Observe that
∂ u0 −h ∂ η 2 h
Z
A = − ai j
Dk Dk u0 (x) dx
Ω ∂ xi ∂xj

∂ u0 −h 2 h ∂ u0 (x)
Z
− ai j Dk η Dk dx
Ω ∂ xi ∂xj
∂ u0 ∂ η 2 h
Z
h
= Dk [ai j ] D u0 (x) dx
Ω ∂ xi ∂ x j k
∂ η2 h

h ∂ u0
Z
+ ai j Dk D u0 (x) dx
Ω ∂ xi ∂xj k

∂ u0 2 h ∂ u0 (x)
Z
h
+ Dk [ai j ] η Dk dx
Ω ∂ xi ∂xj

∂ u0 ∂ u0 (x)
Z
+ ai j Dhk η 2 Dhk dx (11.168)
Ω ∂ xi ∂xj
On the other hand

Z
B = f D−h 2 h
k [η Dk u0 (x)] dx
Ω
= ≤ k f k2 kD−h 2 h
k [η Dk u0 ]k2
∂ [η 2 Dkk u0 ]
≤ k f k2
∂ xk
2
∂ η2 h
≤ k f k2 D u0
∂ xk k 2
h
2 ∂ Dk u0
+k f k2 η
∂ xk
2
!
n ∂ Dhk u0
≤ k f k2 K1 ku0 k1,2 + k f k2 K2 ∑ η (11.169)
j=1 ∂xj
2
where
∂ η2
K1 = ,
∂ xk ∞
and
K2 = kηk∞ .
From (11.167) and (11.168) we get

∂ u0 (x)
∂ u0
Z
ai j Dhk η 2 Dhk dx
Ω ∂ xi ∂xj
∂ u0 ∂ η 2 h
Z
= − Dhk [ai j ] Dk u0 (x) dx
Ω ∂ xi ∂ x j
2
h ∂ u0
Z
∂η h
− ai j Dk D u0 (x) dx
Ω ∂ xi ∂xj k

∂ u0 2 h ∂ u0 (x)
Z
h
− Dk [ai j ] η Dk dx
Ω ∂ xi ∂xj
−B, (11.170)
so that from this and (11.169), we obtain,

n
∂ u0 2

c∑ ηDhk
j=1 ∂xj 2

2 h ∂ u0 h ∂ u0
Z
≤ ai j η Dk Dk dx
Ω ∂ xi ∂ xi
N
∂ u0
≤ K3 ku0 k21,2 + K4 ∑ ηDhk
j=1 ∂xj 2
N
∂ u0
+(K5 + K7 ) ∑ η Dhk + K6 (11.171)
j=1 ∂xj 2
where
∂η
K3 = max kDhk ai j k∞ max η
i, j∈{1,··· ,n} j∈{1,··· ,n} ∂xj ∞
∂ ai j ∂η
≤ max max η , (11.172)
i, j∈{1,··· ,n} ∂ xk ∞ j∈{1,··· ,n} ∂xj ∞
∂η
K4 = max kai j k∞ max η ,
i, j∈{1,··· ,n} j∈n ∂xj ∞
K5 = ku0 k1,2 max kDhk ai j k∞ kηk∞

i, j∈{1,··· ,n}
∂ ai j
≤ |u0 k1,2 max kηk∞ , (11.173)
i, j∈{1,··· ,n} ∂ xk ∞
K6 = k f k2 K1 ku0 k1,2 ,
and
K7 = K2 k f k2 .
We claim
n
∂ u0 (x)
lim sup ∑ η Dhk ≤ K8 ,
h→0 j=1 ∂xj 2
for some K8 > 0.
Suppose, to obtain contradiction, there is a sequence {hm } such that hm → 0+ and
n
hm ∂ u0 (x)
∑ ηDk → +∞, as m → ∞.
j=1 ∂xj 2
From this and (11.171) we have,

n
∂ u0 (x)
c lim sup ∑ ηDhk m
m→∞ j=1 ∂xj 2
K3 ku0 k1,2
≤ lim + K4 + K5
m→∞
∑nj=1 η Dhk m ∂ u∂0x(x)
j 2
K6
+ + K7
h ∂ u0 (x)
∑nj=1 ηDk m ∂xj 2
= K4 + K5 + K7 , (11.174)
which contradicts c > 0.
Hence,
∂ u0
DhK ≤ K8 , ∀ j, k ∈ {1, · · · , n},
∂xj V
for some appropriate K8 > 0, so that
∂ 2 u0
∈ L2 (V ), ∀ j, k ∈ {1, ..., n},
∂ x j ∂ xk
that is,
u0 ∈ W 2,2 (V ), ∀V ⊂⊂ Ω,
so that
u0 ∈ Wl2,2
oc (Ω).
11.7.1 Regularity, a more general result

Let us now consider the case in which,
Ω = Br (x0 ) ∩ Rn+ ,
for some r > 0 where
Rn+ = {x = (x1 , · · · , xn ) ∈ Rn : xn > 0}.
Assume the same previous hypotheses on {ai j } and J : U → R so that
J(u0 ) = min J(u).

u∈U
and
∂ u0 ∂ ϕ
Z Z
ai j dx = f ϕ d x, (11.175)
Ω ∂ xi ∂ x j Ω
∀ϕ ∈ U = W01,2 (Ω).
Define V = Br/2 (x0 ) ∩ Rn+ .
Choose η ∈ C1 (Rn ) such that
η ≡ 1, on V
and
spt(η) ⊂⊂ Br (x0 ).
Select k ∈ {1, · · · , n − 1}, and define for 0 < |h| < r/4
ϕ(x) = −D−h 2 h
k [η Dk u0 (x)].
Observe that ϕ ∈ W01,2 (Ω) so that with the same development presented in the last section we may obtain
∂ u0 (x)
ηDhk ≤ K8 ,
∂xj 2,Ω
∀ j ∈ {1, · · · , n}, k ∈ {1, · · · n − 1}, 0 < |h| < r/4. so that
∂ 2 u0
≤ K8 , ∀ j ∈ {1, · · · , n}, k ∈ {1, · · · , n − 1}.
∂ xk ∂ x j 2,V
Finally, observe that in distributional sense,

∂ ∂ u0
ai j + f = 0,
∂xj ∂ xi
so that
∂ 2 u0 ∂ 2 u0
c1 ≤ ann
∂ xn2 ∂ xn2
n n−1
∂ 2 u0
≤ K∑ ∑ + | f |, a.e in Ω. (11.176)
j=1 k=1 ∂ x j ∂ xk
Hence
∂ 2 u0
≤ K9 ,
∂ xn2 2,V
so that
u0 ∈ W 2,2 (V ).
Our more general result is summarized by the next theorem.
Theorem 11.7.1 Let Ω ⊂ Rn be an open bounded set with a boundary ∂ Ω.

Assume for each x0 ∈ ∂ Ω there exists r > 0 and a C2 function f such that for a local system of coordinates
Ω ∩ Br (x0 ) = {x ∈ Br (x0 ) : xn < f (x1 , · · · , xn−1 )}.
Assume also that defining
y = (y1 , · · · , yn ) = ψ(x) = (x1 , · · · , xn−1 , f (x1 , · · · , xn−1 ) − xn ),
so that
x = ψ −1 (y) = (y1 , · · · , yn−1 , f (y1 , · · · , yn−1 ) − yn ),
we have,
ãi j ηi η j ≥ cx0 |η|2 , ∀η ∈ Rn ,
where
∂ yk (x(y)) ∂ yl (x(y))
ãi j (y) = akl (x(y)) | det(ψ −1 )0 (y)|.
∂ xi ∂xj
Let u0 ∈ U = W01,2 (Ω) be such that
J(u0 ) = min J(u),
u∈U
where, once more, J : U → R is given by
1 ∂ u0 ∂ u0
Z Z
J(u) = ai j dx − f u dx,
2 Ω ∂ xi ∂ x j Ω
and where f ∈ L2 (Ω).

Under such hypotheses, u0 ∈ W 2,2 (Ω).
Proof 11.34 Select x0 ∈ ∂ Ω.

Denote V = Br (x0 ) ∩ Ω,
From the continuity of ψ −1 , there exists r1 > 0 such that
W = ψ −1 (B+
r1 ) ⊂ V ⊂ Ω,
where y0 = ((x0 )1 , · · · , (x0 )n−1 , 0) and

B+ n
r1 = Br1 ∩ R+ .
Let ϕ̃ (y) ∈ W01,2 (B+

r1 ). Hence
ϕ(x) = ϕ̃ (y(x)) ∈ W01,2 (W ).
Observe that, from the optimality condition, we have,
∂ u0 ∂ ϕ(x)
Z Z
ai j dx = f ϕ dx.
W ∂ xi ∂ x j W
However,
∂ u0 (x) ∂ ϕ(x)
Z
ai j dx
W ∂ xi ∂xj
∂ u0 (x(y)) ∂ yk (x(y)) ∂ ϕ(x(y)) ∂ yl (x(y))
Z
= ai j (x(y)) | det[(ψ −1 )0 (y)]| dy
+
Br1 ∂ yk ∂ xi ∂ yl ∂xj
∂ ũ0 (y) ∂ ϕ̃(y)
Z
= ãi j (y) dy. (11.177)
B+
r1 ∂ yi ∂yj
Hence,
∂ ũ0 (y) ∂ ϕ̃(y)
Z Z
ãi j (y) dy = f˜(y)ϕ̃(y) dy,
B+
r1 ∂ yi ∂yj B+
r1
where,
f˜(y) = f (x(y))| det[(ψ −1 )0 (y)| ∈ L2 (B+
r1 ).
Since ϕ̃ ∈ W01,2 (B+

r1 ) is arbitrary, from the results of last section we may infer that,
ũ0 (y) = u0 (x(y)) ∈ W 2,2 (B+

r /2 ).
1
Now observe that, denoting

W̃ = ψ −1 (B+
r /2 ) ⊂ Ω,
1
we have
2
∂ 2 u0
∂ xi ∂ x j 2,W̃
2
∂ 2 ũ0 (y) ∂ yl (x(y)) ∂ yk (x(y) ∂ ũ0 (y) ∂ 2 yk (x(y))
Z
= + | det[(ψ −1 )0 (y)]| dy
B+
r /2
∂ yk ∂ yl ∂ xi ∂xj ∂ yk ∂ xi ∂ x j
1
n 2
∂ 2 ũ0
≤ K1 ∑
j,k=1 ∂ yk ∂ yl 2,B+
r /2
1
∂ 2 u˜ 0 ∂ u˜0
+K2 ∑
i, j,k ∂ yi ∂ y j 2,B+ ∂ yk 2,B+
r1 /2 r /2
1
n 2
∂ ũ0
+K3 ∑ , (11.178)
j=1 ∂yj 2,B+
r /2
1
so that
∂ 2 u0
∈ L2 (W̃ ).
∂ xi ∂ x j
At this point we remark that x0 ∈ W, and clarifying the dependence of W relating x0 , we also denote
W = Wx0 .
Observe that from the compactness of ∂ Ω, we may find x1 , · · · , xM ∈ ∂ Ω such that
∪M
j=1Wx j ⊃ ∂ Ω,
so that we may also find W0 ⊂⊂ Ω such that
[∪M
j=1Wx j ] ∪W0 ⊃ Ω,
and in such a case
W = [∪M ˜
j=1Wx j ] ∪W0 .
From the previous results
ku0 k2,2,W̃x ≤ K j , ∀ j ∈ {1, · · · , M},
j
and
ku0 k2,2,W0 ≤ K0 ,
for appropriate real constants K0 > 0, K j > 0, ∀ j ∈ {1, · · · M}.
Joining the pieces, we obtain
u0 ∈ W 2,2 (Ω).
SECTION II
CALCULUS OF VARIATIONS,
CONVEX ANALYSIS AND
RESTRICTED OPTIMIZATION
Chapter 12
Basic Topics on the Calculus of

Variations
12.1 Banach spaces

The main reference for this chapter is Troutman [79].
We start by recalling the norm definition.
Definition 12.1.1 Let V be a vectorial space. A norm in V is a function denoted by k·kV : V → R+ = [0, +∞),
for which the following properties hold:
1. kukV > 0, ∀u ∈ V such that u =

6 0
and
kukV = 0, if, and only if u = 0.
2. Triangular inequality, that is
ku + vkV ≤ kukV + kvkV , ∀u, v ∈ V

3.
kαukV = |α|kuk, ∀u ∈ V, α ∈ R.
In such a case we say that the space V is a normed one.
Basic Topics on the Calculus of Variations 253
Definition 12.1.2 (Convergent sequence) Let V be a normed space and let {un } ⊂ V be a sequence. We say
that {un } converges to u0 ∈ V , if for each ε > 0, there exists n0 ∈ N such that if n > n0 , then
kun − u0 kV < ε.
In such a case, we write

lim un = u0 , in norm.
n→∞
Definition 12.1.3 (Cauchy sequence in norm) Let V be a normed space and let {un } ⊂ V be a sequence.
We say that {un } is a Cauchy one, as for each ε > 0, there exists n0 ∈ N such that if m, n > n0 , then
kun − um kV < ε.
At this point we recall the definition of Banach space.

Definition 12.1.4 (Banach space) A normed space V is said to be a Banach space as it is complete, that is,
for each Cauchy sequence {un } ⊂ V there exists a u0 ∈ V such that
kun − u0 kV → 0, as n → ∞.
Example 12.1.5 Examples of Banach spaces:

Consider V = C([a, b]), the space of continuous functions on [a, b]. We shall prove that such a space is a
Banach one with the norm,
k f kV = max{| f (x)| : x ∈ [a, b]}.
Exercise 12.1.6 Prove that

k f kV = max{| f (x)| : x ∈ [a, b]}
is a norm for V = C([a, b]).
Solution:
1. Clearly
k f kV ≥ 0, ∀ f ∈ V
and
k f kV = 0 if, and only if f (x) = 0, ∀x ∈ [a, b],
that is if, and only if f = 0.
2. Let f , g ∈ V .
Thus,
k f + gkV = max{| f (x) + g(x)|, x ∈ [a, b]}

≤ max{| f (x)| + |g(x)|, x ∈ [a, b]}
≤ max{| f (x)|, x ∈ [a, b]} + max{|g(x)| x ∈ [a, b]}
= k f kV + kgkV . (12.1)
3. Finally, let α ∈ R and f ∈ V.

Hence,
kα f kV = max{|α f (x)|, x ∈ [a, b]}

= max{|α|| f (x)|, x ∈ [a, b]}
= |α| max{| f (x)|, x ∈ [a, b]}
= |α|k f k. (12.2)
From this we may infer that k · kV is a norm for V .

Theorem 12.1.7 V = C([a, b]) is a Banach space with the norm
k f kV = max{| f (x)| : x ∈ [a, b]}, ∀ f ∈ V.
Proof 12.1 The proof that C([a, b]) is a vector space is left as an exercise.
From the last exercise, k · kV is a norm for V .
Let { fn } ⊂ V be a Cauchy sequence.
We shall prove that there exists f ∈ V such that
k fn − f kV → 0, as n → ∞.
Let ε > 0.
Thus, there exists n0 ∈ N such that if m, n > n0 , then
k fn − fm kV < ε.
Hence
max{| fn (x) − fm (x)| : x ∈ [a, b]} < ε,
that is,
| fn (x) − fm (x)| < ε, ∀x ∈ [a, b], m, n > n0 . (12.3)
Let x ∈ [a, b].
From (12.3), { fn (x)} is a real Cauchy sequence, therefore it is convergent.
So, define
f (x) = lim fn (x), ∀x ∈ [a, b].
n→∞
Also from (12.3), we have that
lim | fn (x) − fm (x)| = | fn (x) − f (x)| ≤ ε, ∀n > n0 .

m→∞

k fn − f kV → 0, as n → ∞.
We shall prove now that f is continuous on [a, b].
From the exposed above,
fn → f
uniformly on [a, b] as n → ∞.
Thus, there exists n1 ∈ N such that if n > n1 , then
ε
| fn (x) − f (x)| < , ∀x ∈ [a, b].
3
Choose n2 > n1 .
Let x ∈ [a, b]. From
lim fn2 (y) = fn2 (x),
y→x
there exists δ > 0 such that if y ∈ [a, b] and |y − x| < δ , then

ε
| fn2 (y) − fn2 (x)| < .
3
Thus, if y ∈ [a, b] and |y − x| < δ , then
| f (y) − f (x)| = | f (y) − fn2 (y) + fn2 (y) − fn2 (x) + fn2 (x) − f (x)|
≤ | f (y) − fn2 (y)| + | fn2 (y) − fn2 (x)| + | fn2 (x) − f (x)|
ε ε ε
< + +
3 3 3
= ε. (12.4)
So, we may infer that f is continuous at x, ∀x ∈ [a, b], that is f ∈ V.

Exercise 12.1.8 Let V = C1 ([a, b]) be the space of functions f : [a, b] → R which the the first derivative is
continuous on [a, b].
Define the function (in fact a functional) k · kV : V → R+ by
k f kV = max{| f (x)| + | f 0 (x)| : x ∈ [a, b]}.
1. Prove that k · kV is a norm.

2. Prove that V is a Banach space with such a norm.
Solution: The proof of item 1 is left as an exercise.

Now, we shall prove that V is complete.
Let { fn } ⊂ V be a Cauchy sequence.
Let ε > 0. Thus, there exists n0 ∈ N such that if m, n > n0 , then
k fn − fm kV < ε/2.
Therefore,
| fn (x) − fm (x)| + | fn0 (x) − fm0 (x)| < ε/2, ∀x ∈ [a, b], m, n > n0 . (12.5)
Let x ∈ [a, b]. hence, { fn (x)} and { fn0 (x)} are real Cauchy sequences, and therefore, they are convergent.
Denote
f (x) = lim fn (x)
n→∞
and
g(x) = lim fn0 (x).
n→∞
From this and (12.5), we obtain
| fn (x) − f (x)| + | fn0 (x) − g(x)| = lim | fn (x) − fm (x)| + | fn0 (x) − fm0 (x)|
m→∞
≤ ε/2, ∀x ∈ [a, b], n > n0 . (12.6)
Similarly to the last example, we may obtain that f and g are continuous, therefore uniformly continuous
on the compact set [a, b].
Thus, there exists δ > 0 such that if x, y ∈ [a, b] and |y − x| < δ , then
|g(y) − g(x)| < ε/2. (12.7)
Choose n1 > n0 . Let x ∈ (a, b).

Hence, if 0 < |h| < δ , then from (12.6) and (12.7) we have
fn1 (x + h) − fn1 (x)

− g(x)
h
= | fn0 1 (x + th) − g(x + th) + g(x + th) − g(x)|
≤ | fn0 1 (x + th) − g(x + t h)| + |g(x + th) − g(x)|
< ε/2 + ε/2
= ε, (12.8)
where from mean value theorem, t ∈ (0, 1) (it depends on h). Therefore, letting n1 → ∞, we get
fn1 (x + h) − fn1 (x)

− g(x)
h
f (x + h) − f (x)
→ − g(x)
h
≤ ε, ∀0 < |h| < δ . (12.9)
f (x + h) − f (x)
f 0 (x) = lim = g(x), ∀x ∈ (a, b).
h→0 h
The cases in which x = a or x = b may be dealt similarly with one-sided limits.
From this and (12.6), we have
k fn − f kV → 0, as n → ∞
and
f ∈ C1 ([a, b]).
Definition 12.1.9 (Functional) Let V be a Banach space. A functional F defined on V is a function whose
the co-domain is R (F : V → R).
Example 12.1.10 Let V = C([a, b]) and F : V → R where

Z b
F(y) = ( sen3 x + y(x)2 ) dx, ∀y ∈ V.
a
Example 12.1.11 Let V = C1 ([a, b]) and let J : V → R where

Z bq
J (y) = 1 + y0 (x)2 dx, ∀y ∈ C1 ([a, b]).
a
In our first frame-work we consider functionals defined as

Z b
F(y) = f (x, y(x), y0 (x)) dx,
a
where we shall assume

f ∈ C([a, b] × R × R)
and V = C1 ([a, b]).
Thus, for F : D ⊂ V → R where

Z b
F(y) = f (x, y(x), y0 (x)) dx,
a
we assume
V = C1 ([a, b]),
and
D = {y ∈ V : y(a) = A and y(b) = B},
where A, B ∈ R.
Observe that if y ∈ D, then y + v ∈ D if, and only if, v ∈ V and
v(a) = v(b) = 0.
Indeed, in such a case
y+v ∈V
and
y(a) + v(a) = y(a) = A,
and
y(b) + v(b) = y(b) = B.
Thus, we define the space of admissible directions for F, denoted by Va , as,
Va = {v ∈ V : v(a) = v(b) = 0}.
Definition 12.1.12 (Global minimum) Let V be a Banach space and let F : D ⊂ V → R be a functional. We
say that y0 ∈ D is a point of global minimum for F, if
F(y0 ) ≤ F(y), ∀y ∈ D.
Observe that denoting y = y0 + v where v ∈ Va , we have
F(y0 ) ≤ F(y0 + v), ∀v ∈ Va .
Example 12.1.13 Consider J : D ⊂ V → R where V = C1 ([a, b]),
D = {y ∈ V : y(a) = 0 and y(b) = 1}
and Z b
J(y) = (y0 (x))2 dx.
a
Thus,
Va = {v ∈ V : v(a) = v(b) = 0}.
Let y0 ∈ D be a candidate to global minimum for F and let v ∈ Va be admissible direction.
Hence, we must have
J(y0 + v) − J(y0 ) ≥ 0, (12.10)
where
Z b Z b
J(y0 + v) − J(y0 ) = (y00 (x) + v0 (x))2 dx − y00 (x)2 dx
a a
Z b Z b
= 2 y00 (x)v0 (x) dx + v0 (x)2 dx
a a
Z b
≥ 2 y00 (x)v0 (x) dx. (12.11)
a
Observe that if y00 (x) = c in [a, b], we have (12.10) satisfied, since in such a case,
Z b
J(y0 + v) − J(y0 ) ≥ 2 y00 (x)v0 (x) dx
a
Z b
= 2c v0 (x) dx
a
= 2c[v(x)]ab
= 2c(v(b) − v(a))
= 0. (12.12)
Summarizing, if y00 (x) = c on [a, b], then
J(y0 + v) ≥ J(y0 ), ∀v ∈ Va .
Observe that in such a case,
y0 (x) = cx + d,
for some d ∈ R.
However, from
y(a) = 0, we get ca + d = 0.
From y0 (b) = 1, we have cb + d = 1.
Solving this last system in c and d we obtain,
1
c= ,
b−a
and
−a
d= .
b−a
From this, we have,
x−a
y0 (x) = .
b−a
Observe that the graph of y0 corresponds to the straight line connecting the points (a, 0) and (b, 1).
12.2 ˆ
The Gateaux variation
Definition 12.2.1 Let V be a Banach space and let J : D ⊂ V → R be a functional. Let y ∈ D and v ∈ Va
We define the Gateaux
ˆ variation of J at y in the direction v, denoted by δ J(y; v), by
J(y + εv) − J(y)

δ J(y; v) = lim ,
ε→0 ε
if such a limit exists. Equivalently,
∂ J(y + εv)
δ J(y; v) = |ε=0 .
∂ε
Example 12.2.2 Let V = C1 ([a, b]) and J : V → R where
Z b
J(y) = ( sen3 x + y(x)2 ) dx.
a
Let y, v ∈ V . Let us calculate

δ J(y; v).
Observe that,
J(y + εv) − J(y)
δ J(y; v) = lim
ε →0 ε
Rb 3 2
Rb 3 2
a ( sen x + (y(x) + εv(x)) ) dx − a ( sen x + y(x) ) dx
= lim
ε→0 ε
Rb 2
(2εy(x)v(x) + ε v(x)) dx
= lim a
ε→0 ε
Z b Z b
= lim 2y(x)v(x) dx + ε v(x)2 dx
ε→0 a a
Z b
= 2y(x)v(x) dx. (12.13)
a
Example 12.2.3 Let V = C1 ([a, b]) and let J : V → R where

Z b q
J(y) = ρ(x) 1 + y0 (x)2 dx,
a
and where ρ : [a, b] → (0, +∞) is a fixed function.

Let y, v ∈ V .
Thus,
∂ J(y + εv)
δ J(y; v) = |ε=0 , (12.14)
∂ε
where Z b q
J(y + εv) = ρ(x) 1 + (y0 (x) + εv0 (x))2 dx.
a
Hence,
b
Z
∂ J(y + εv)
q
∂ 0 0 2
|ε=0 = ρ(x) 1 + (y (x) + εv (x)) dx
∂ε ∂ε a
Z b q
∂
=(∗) ρ(x) 0 0 2
1 + (y (x) + εv (x)) d x
a ∂ε
ρ(x) 2(y0 (x) + εv0 (x))v0 (x)
Z b
= p dx. (12.15)
a 2 1 + (y0 (x) + εv0 (x))2
(*): We shall prove this step is valid in the subsequent pages.
From this we get,
∂ J(y + εv)
δ J(y; v) = |ε=0
∂ε
ρ (x)y0 (x)v0 (x)
Z b
= p dx. (12.16)
a 1 + y0 (x)2
The example is complete.

Example 12.2.4 Let V = C1 ([a, b]) and f ∈ C1 ([a, b] × R × R). Thus f is a function of three variables,
namely, f (x, y, z).
Consider the functional F : V → R, defined by
Z b
F(y) = f (x, y(x), y0 (x)) dx.
a
Let y, v ∈ V . Thus,
∂
δ F(y; v) = F(y + εv)|ε=0 .
∂ε
Observe that Z b
F(y + εv) = f (x, y(x) + εv(x), y0 (x) + εv0 (x)) dx,
a
and therefore
b
Z
∂ ∂ 0 0
F(y + εv) = f (x, y(x) + εv(x), y (x) + εv (x)) dx
∂ε ∂ε a
Z b
∂
f (x, y(x) + εv(x), y0 (x) + εv0 (x)) dx

=
a ∂ε
∂ f (x, y(x) + εv(x), y0 (x) + εv0 (x))
Z b
= v(x)
a ∂y
∂ f (x, y(x) + εv(x), y0 (x) + εv0 (x)) 0

+ v (x) d x. (12.17)
∂z
Thus
∂ F(y + εv)
δ F(y; v) = |ε=0
∂ε
∂ f (x, y(x), y0 (x)) ∂ f (x, y(x), y0 (x)) 0
Z b
= v(x) + v (x) dx. (12.18)
a ∂y ∂z
12.3 Minimization of convex functionals

Definition 12.3.1 (Convex function) A function f : Rn → R is said to be convex if
f (λ x + (1 − λ )y) ≤ λ f (x) + (1 − λ ) f (y), ∀x, y ∈ Rn , λ ∈ [0, 1].
Proposition 12.3.2 Let f : Rn → R be a convex and differentiable function.

f (y) − f (x) ≥ h f 0 (x), y − xiRn , ∀x, y ∈ Rn ,
where h·, ·iRn : Rn × Rn → R denotes the usual inner product for Rn , that is,
hx, yiRn = x1 y1 + · · · + xn yn ,
∀x = (x1 , · · · , xn ), y = (y1 , · · · , yn ) ∈ Rn .
Proof 12.2 Choose x, y ∈ Rn .

f ((1 − λ )x + λ y) ≤ (1 − λ ) f (x) + λ f (y), ∀λ ∈ (0, 1).

Thus,
f (x + λ (y − x)) − f (x)
≤ f (y) − f (x), ∀λ ∈ (0, 1).
λ
Therefore,
f (x + λ (y − x)) − f (x)
h f 0 (x), y − xiRn = lim
λ →0+ λ
≤ f (y) − f (x), ∀x, y ∈ Rn . (12.19)
Proposition 12.3.3 Let f : Rn → R be a differentiable function on Rn .

Assume
f (y) − f (x) ≥ h f 0 (x), y − xiRn , ∀x, y ∈ Rn .
Under such hypotheses, f is convex.
Proof 12.3 Define f ∗ : Rn → R ∪ {+∞} by
f ∗ (x∗ ) = sup {hx, x∗ iRn − f (x)}.

x∈Rn
Such a function is said to the polar function for f .

Let x ∈ Rn . From the hypotheses,
h f 0 (x), xiRn − f (x) ≥ h f 0 (x), yiRn − f (y), ∀y ∈ Rn ,
that is,
f ∗ ( f 0 (x)) = sup {h f 0 (x), yiRn − f (y)}

y∈Rn
0
= h f (x), xiRn − f (x). (12.20)
On the other hand

f ∗ (x∗ ) ≥ hx, x∗ iRn − f (x), ∀x, x∗ ∈ Rn ,
and thus
f (x) ≥ hx, x∗ iRn − f ∗ (x∗ ), ∀x∗ ∈ Rn .
Hence,
f (x) ≥ sup {hx, x∗ iRn − f ∗ (x∗ )}

x∗ ∈Rn
≥ h f 0 (x), xiRn − f ∗ ( f 0 (x)). (12.21)
f (x) = sup {hx, x∗ iRn − f ∗ (x∗ )}

x∗ ∈Rn
= h f (x), xiRn − f ∗ ( f 0 (x)).
0
(12.22)
Summarizing,
f (x) = sup {hx, x∗ iRn − f ∗ (x∗ )}, ∀x ∈ Rn .
x∗ ∈Rn
Choose x, y ∈ Rn and λ ∈ [0, 1].

From the last equation we may write,
f (λ x + (1 − λ )y) = sup {hλ x + (1 − λ )y, x∗ iRn − f ∗ (x∗ )}

x∗ ∈Rn
= sup {hλ x + (1 − λ )y, x∗ iRn
x∗ ∈Rn
−λ f ∗ (x∗ ) − (1 − λ ) f ∗ (x∗ )}
= sup {λ (hx, x∗ iRn − f ∗ (x∗ ))
x∗ ∈Rn
+(1 − λ )(hy, x∗ iRn − f ∗ (x∗ ))}
≤ λ sup {hx, x∗ iRn − f ∗ (x∗ )}
x∗ ∈Rn
+(1 − λ ) sup {hy, x∗ iRn − f ∗ (x∗ )}
x∗ ∈Rn
= λ f (x) + (1 − λ ) f (y). (12.23)
Since x, y ∈ Rn and λ ∈ [0, 1] are arbitrary, we may infer that f is convex.

Definition 12.3.4 (Convex functional) Let V be a Banach space and let J : D ⊂ V → R be a functional. We
say that J is convex if
J(y + v) − J(y) ≥ δ J(y; v), ∀v ∈ Va (y),
where
Va (y) = {v ∈ V : y + v ∈ D}.
Theorem 12.3.5 Let V be a Banach space and let J : D ⊂ U be a convex functional. Thus, if y0 ∈ D is such
that
δ J(y0 ; v) = 0, ∀v ∈ Va (y0 ),
then
J(y0 ) ≤ J(y), ∀y ∈ D,
that is, y0 minimizes J on D.
Proof 12.4 Choose y ∈ D. Let v = y − y0 . Thus y = y0 + v ∈ D so that
v ∈ Va (y0 ).
From the hypothesis,

δ J(y0 ; v) = 0,
and since J is convex, we obtain
J(y) − J(y0 ) = J(y0 + v) − J(y0 ) ≥ δ J(y0 ; v) = 0,
that is,
J(y0 ) ≤ J(y), ∀y ∈ D.
Example 12.3.6 Let us see this example of convex functional. Let V = C1 ([a, b]) and let J : D ⊂ V → R be
defined by
Z b
J (y) = (y0 (x))2 dx,
a
where
D = {y ∈ V : y(a) = 1 and y(b) = 5}.
We shall show that J is convex.
Indeed, let y ∈ D and v ∈ Va where
Va = {v ∈ V : v(a) = v(b) = 0}.
Thus,
Z b Z b
J (y + v) − J (y) = (y0 (x) + v0 (x))2 dx − y0 (x)2 dx
a a
Z b Z b
0 0
= 2y (x)v (x) dx + v0 (x)2 dx
a a
Z b
≥ 2y0 (x)v0 (x) dx
a
= δ J(y; v). (12.24)
Therefore, J is convex.
12.4 Sufficient conditions of optimality for the convex case

We start this section with a remark.
Remark 12.4.1 Consider a function f : [a, b] × R × R → R where f ∈ C1 ([a, b] × R × R).

Thus, for V = C1 ([a, b]), define F : V → R by
Z b
F(y) = f (x, y(x), y0 (x)) dx.
a
Let y, v ∈ V . We have already shown that

Z b
δ F(y; v) = ( fy (x, y(x), y0 (x))v(x) + fz (x, y(x), y0 (x))v0 (x)) dx.
a
Suppose f is convex in (y, z) for all x ∈ [a, b], which we denote by f (x, y, z) to be convex.
From the last section, we have that
f (x, y + v, y0 + v0 ) − f (x, y, y0 ) ≥ h∇ f (x, y, y0 ), (v, v0 )iR2

= fy (x, y, y0 )v + fz (x, y, y0 )v0 , ∀x ∈ [a, b] (12.25)
where we denote
∇ f (x, y, y0 ) = ( fy (x, y, y0 ), fz (x, y, y0 )).
Therefore,
Z b
F(y + v) − F(y) = [ f (x, y + v, y0 + v0 ) − f (x, y, y0 )] dx
a
Z b
≥ [ fy (x, y, y0 )v + fz (x, y, y0 )v0 ] dx
a
= δ J(y; v). (12.26)
Thus, F is convex.
Theorem 12.4.2 Let V = C1 ([a, b]). Let f ∈ C2 ([a, b] × R × R) where f (x, y, z) is convex. Define
D = {y ∈ V : y(a) = a1 and y(b) = b1 },
where a1 , b1 ∈ R.
Define also F : D → R by
Z b
F(y) = f (x, y(x), y0 (x)) dx.
a
Under such hypotheses, F is convex and if y0 ∈ D is such that
d
[ fz (x, y0 (x), y00 (x))] = fy (x, y0 (x), y00 (x)), ∀x ∈ [a, b],
dx
then y0 minimizes F on D, that is,
F(y0 ) ≤ F(y), ∀y ∈ D.
Proof 12.5 From the last remark, F is convex. Suppose now that y0 ∈ D is such that
d
[ fz (x, y0 (x), y00 (x))] = fy (x, y0 (x), y00 (x)), ∀x ∈ [a, b].
dx
Let v ∈ Va = {v ∈ V : v(a) = v(b) = 0}. Thus,
Z b
δ F(y0 ; v) = ( fy (x, y0 (x), y00 (x))v(x) + fz (x, y0 (x), y00 (x))v0 (x)) dx
a
Z b
d
= ( fz (x, y0 (x), y00 (x))v(x)) + fz (x, y0 (x), y00 (x))v0 (x) dx
a dx
Z b
d
= [ fz (x, y0 (x), y00 (x))v(x)] dx
a dx
= [ fz (x, y0 (x), y00 (x))v(x)]ba
= fz (b, y0 (b), y00 (b))v(b) − fz (a, y0 (a), y00 (a))v(a)
= 0, ∀v ∈ Va . (12.27)
Since F is convex, from this and Theorem 12.3.5, we may infer that y0 minimizes J on D.
Example 12.4.3 Let V = C1 ([a, b]) and
D = {y ∈ V : y(0) = 0 and y(1) = 1}.
Define F : D → R by
Z 1
F(y) = [y0 (x)2 + 5y(x)] dx, ∀y ∈ D.
0
Observe that Z 1
F(y) = f (x, y, y0 ) dx
0
where
f (x, y, z) = z2 + 5y,
that is, f (x, y, z) is convex.
Thus, from the last theorem F is convex and if y0 ∈ D is such that
d
fz (x, y0 (x), y00 (x)) = fy (x, y0 (x), y00 (x)), ∀x ∈ [a, b],
dx
then y0 minimizes F on D.
Considering that fz (x, y, z) = 2z and fy (x, y, z) = 5, from this last equation we get,
d
(2y0 (x)) = 5,
dx 0
that is,
5
y000 (x) = , ∀x ∈ [0, 1].
2
Thus,
5
y00 (x) = x + c,
2
and
5
y0 (x) = x2 + cx + d.
4
From this and y0 (0) = 0, we obtain d = 0.
From this and y0 (1) = 1, we have
5
+ c = 1,
4
so that
c = −1/4
Therefore
5x2 x
y0 (x) = −
4 4
minimizes F on D.
12.5 Natural conditions, problems with free extremals

We start this section with the following theorem.
Theorem 12.5.1 Let V = C1 ([a, b]). Let f ∈ C2 ([a, b] × R × R) be such that f (x, y, z) is convex. Define
D = {y ∈ V : y(a) = a1 },
where a1 ∈ R.
Define also F : D → R by
Z b
F(y) = f (x, y(x), y0 (x)) dx.
a
Under such hypotheses,F is convex and if y0 ∈ D is such that
d
[ fz (x, y0 (x), y00 (x))] = fy (x, y0 (x), y00 (x)), ∀x ∈ [a, b]
dx
and
fz (b, y0 (b), y00 (b)) = 0
F(y0 ) ≤ F(y), ∀y ∈ D.
Proof 12.6 Since f (x, y, z) is convex from the last remark F is convex. Suppose now that y0 ∈ D is such
that
d
dx
and
fz (b, y0 (b), y00 (b)) = 0.
Let v ∈ Va = {v ∈ V : v(a) = 0}. Thus,
Z b
a
Z b
d
= ( fz (x, y0 (x), y00 (x))v(x)) + fz (x, y0 (x), y00 (x))v0 (x) dx
a dx
Z b
d 0
= [ fz (x, y0 (x), y0 (x))v(x)] dx
a dx
= [ fz (x, y0 (x), y00 (x))v(x)]ba
= 0v(b) − fz (a, y0 (a), y00 (b))0
= 0, ∀v ∈ Va . (12.28)
Since F is convex, from this and Theorem 12.3.5, we may infer that y0 minimizes J on D.
Remark 12.5.2 About this last theorem y(a) = a1 is said to be an essential boundary condition, whereas
fz (b, y0 (b), y00 (b)) = 0 is said to be a natural boundary condition.
Theorem 12.5.3 Let V = C1 ([a, b]). Let f ∈ C2 ([a, b] × R × R) where f (x, y, z) is convex. Define
D=V
and F : D → R by
Z b
F(y) = f (x, y(x), y0 (x)) dx.
a
Under such hypotheses, F is convex and if y0 ∈ D is such that
d
[ fz (x, y0 (x), y00 (x))] = fy (x, y0 (x), y00 (x)), ∀x ∈ [a, b],
dx
fz (a, y0 (a), y00 (a)) = 0
and
fz (b, y0 (b), y00 (b)) = 0,
F(y0 ) ≤ F(y), ∀y ∈ D.
Proof 12.7 From the last remark F is convex. Suppose that y0 ∈ D is such that
d
dx
and
fz (a, y0 (a), y00 (a)) = fz (b, y0 (b), y00 (b)) = 0.
Let v ∈ D = V . Thus,
Z b
a
Z b
d 0 0 0
= ( fz (x, y0 (x), y0 (x))v(x)) + fz (x, y0 (x), y0 (x))v (x) dx
a dx
Z b
d 0
= [ fz (x, y0 (x), y0 (x))v(x)] dx
a dx
= [ fz (x, y0 (x), y00 (x))v(x)]ba
= 0v(b) − 0v(a)
= 0, ∀v ∈ D. (12.29)
Since F is convex, from this and from Theorem 12.3.5, we may conclude that y0 minimizes J on D = V .
Remark 12.5.4 About this last theorem, the conditions fz (a, y0 (a), y00 (a)) = fz (b, y0 (b), y00 (b)) = 0 are said
to be natural boundary conditions and the problem in question a free extremal one.
Exercise 12.5.5 Show that F is convex and obtain its point of global minimum on D, D1 and D2 , where
Z 2 0 2
y (x)
F(y) = dx,
1 x
and where
1.
D = {y ∈ C1 ([1, 2]) : y(1) = 0, y(1) = 3},
2.
D1 = {y ∈ C1 ([1, 2]) : y(2) = 3}.
3.
D2 = C1 ([1, 2]).
Solution: Observe that Z 2

F(y) = f (x, y(x), y0 (x)) dx,
1
where f (x, y, z) = z2 /x, so that f (x, y, z) is convex.

Therefore, F is convex.
Let y, v ∈ V , thus,
Z 2
δ F(y; v) = [ fy (x, y, y0 )v + fz (x, y, y0 )v0 ] dx,
1
where
fy (x, y, z) = 0
and
fz (x, y, z) = 2z/x.
Therefore,
Z 2
δ F(y; v) = 2x−1 y0 (x)v0 (x) dx.
1
For D, from Theorem 12.4.1, sufficient conditions of optimality are given by,
 dx [ fz (x, y0 (x), y00 (x))] = fy (x, y0 (x), y00 (x)) in [1, 2],
 d
y (1) = 0, (12.30)
 0
y0 (2) = 3.
Thus, we must have

d
[2x−1 y00 (x)] = 0,
dx
that is,
2x−1 y00 (x) = c,
so that
cx
y00 (x) = .
2
Therefore,
cx2
y0 (x) = + d.
4
On the other hand, we must have also
c
y0 (1) = + d = 0,
4
and
y0 (2) = c + d = 3.
Thus, c = 4 and d = −1 so that y0 (x) = x2 − 1 minimizes F on D.
For D1 , from Theorem 12.5.1, sufficient conditions of global optimality are given by
 d
y (2) = 3, (12.31)
 0
fz (1, y0 (1), y00 (1)) = 0.
Thus, we must have

cx2
y0 (x) = + d.
4
On the other hand, we must also have
y0 (2) = c + d = 3,
and
fz (1, y0 (1), y00 (1)) = 2(1)−1 y00 (1) = 0,
that is,
y00 (1) = c/2 = 0,
.
Therefore, c = 0 and d = 3 so that y0 (x) = 3 minimizes F on D1 .
For D2 , from Theorem 12.5.3, sufficient conditions of global optimality are given by
 d
f (1, y0 (1), y00 (1)) = 0 (12.32)

 z
fz (2, y0 (2), y00 (2)) = 0.
Thus, we have
cx2
y0 (x) = + d.
4
On the other hand we have also
fz (1, y0 (1), y00 (1)) = 2(1)−1 y00 (1) = 0,
fz (2, y0 (2), y00 (2)) = 2(2)−1 y00 (2) = 0,

that is,
y00 (1) = y00 (2) = 0,
where y00 (x) = cx/2.
Thus c = 0, so that y0 (x) = d, ∀d ∈ R minimizes F on D2 .
Exercise 12.5.6 Let V = C2 ([0, 1]) and J : D ⊂ V → R where

Z 1 Z 1
EI
J(y) = y00 (x)2 dx − P(x)y(x) dx,
2 0 0
represents the energy of a straight beam with rectangular cross section with inertial moment I. Here y(x)
denotes the vertical displacement of the point x ∈ [0, 1] resulting from the action of distributed vertical load
P(x) = αx, ∀x ∈ [0, 1], where E > 0 is the Young modulus and α > 0 is a real constant.
And also
D = {y ∈ V : y(0) = y(1) = 0}.
1. prove that F is convex.

2. Prove that if y0 ∈ D is such that
d4


 EI dx 4 [y0 (x)] = P(x), ∀x ∈ [0, 1],



y00 (0) = 0, (12.33)
 0



 00
y0 (1) = 0,
then y0 minimizes F on D.
3. Find the optimal solution y0 ∈ D.
Solution:
Let y ∈ D and v ∈ Va = {v ∈ V : v(0) = v(1) = 0}.
We recall that
F(y + εv) − F(y)
δ J(y; v) = lim
ε→0 ε
(EI/2) 0 [(y + εv00 )2 − (y00 )2 ] dx − 01 (P(y + εv) − Py) dx
R 1 00 R
= lim
ε→0 ε
Z 1 Z 1
εEI
= lim (EIy00 v00 − Pv) dx + (v00 )2 dx
ε →0 0 2 0
Z 1
= (EIy00 v00 − Pv) dx. (12.34)
0
On the other hand,

Z 1 Z 1
J(y + v) − J(v) = (EI/2) [(y00 + v00 )2 − (y00 )2 ] dx − (P(y + v) − Py) dx
0 0
Z 1
EI 1 00 2
Z
00 00
= (EIy v − Pv) dx + (v ) dx
0 2 0
Z 1
00 00
≥ (EIy v − Pv) dx
0
= δ J(y; v). (12.35)
Since y ∈ D and v ∈ Va are arbitrary, we may infer that J is convex.

Assume that y0 ∈ D is such that
d4


 EI dx 4 [y0 (x)] = P(x), ∀x ∈ [0, 1],



y000 (0) = 0, (12.36)




 00
y0 (1) = 0,
Thus,
Z 1
δ J(y; v) = (EIy00 v00 − Pv) dx
0
Z 1
= (EIy00 v00 − EIy(4) v) dx
0
Z 1
000
= (EIy00 v00 + EIy v0 ) dx − [EIy000 (x)v(x)]ba
0
Z 1
000
= (EIy00 v00 + EIy v0 ) dx
0
Z 1
00
= (EIy00 v00 − EIy v00 ) dx + [EIy00 (x)v0 (x)]ba
0
= 0 (12.37)
Summarizing
δ J(y0 ; v) = 0, ∀v ∈ Va .
Therefore, since J is convex, we may conclude that y0 minimizes J on D.
To obtain the solution of the ODE is question, we shall denote
y0 (x) = y p (x) + yh (x),

αx5
where a particular solution y p is given by y p (x) = 120EI , where clearly
d4
EI [y p (x)] = P(x), ∀x ∈ [0, 1].
dx4
The homogeneous associated equation
d4
EI [yh (x)] = 0,
dx4
has the following general solution
yh (x) = ax3 + bx2 + cx + d,
and thus,
αx5
y0 (x) = y p (x) + yh (x) = + ax3 + bx2 + cx + d.
120EI
From y0 (0) = 0, we obtain d = 0.
Observe that y00 (x) = 120EI
5α
x4 + 3ax2 + 2bx + c e y000 (x) = 6EI
α 3
x + 6ax + 2b.
00
From this and y0 (0) = 0, we get b = 0.
From y000 (1) = 0, we obtain,
α 3
1 + 6a 1 = 0,
6EI
and thus
α
a=− .
36EI
From such results and from y0 (1) = 0, we obtain
α α α
+ a 13 + c 1 = − + c = 0,
120EI 120EI 36EI
that is,

α 1 1 7α
c= − = .
EI 36 120 360EI
Finally, we have that
αx5 αx3 7αx
y0 (x) = − +
120EI 36EI 360EI
minimizes J on D.
12.6 The du Bois-Reymond lemma

Lemma 12.6.1 (du Bois-Reymond) Suppose h ∈ C([a, b]) and
Z b
h(x)v0 (x) dx = 0, ∀v ∈ Va ,
a
where
Va = {v ∈ C1 ([a, b]) : v(a) = v(b) = 0}.
Under such hypotheses, there exists c ∈ R such that
h(x) = c, ∀x ∈ [a, b].
Proof 12.8 Let Rb

a h(t) dt
c= .
b−a
Define Z x
v(x) = (h(t) − c) dt.
a
Thus,
v0 (x) = h(x) − c, ∀x ∈ [a, b],
so that v ∈ C1 ([a, b]).
Moreover, Z a
v(a) = (h(t) − c) dt = 0,
a
and Z b Z b
v(b) = (h(t) − c) dt = h(t) dt − c(b − a) = c(b − a) − c(b − a) = 0,
a a
so that v ∈ Va .
Observe that, from this and the hypotheses,
Z b
0 ≤ (h(t) − c)2 dt
a
Z b
= (h(t) − c)(h(t) − c) dt
a
Z b
= (h(t) − c)v0 (t) dt
a
Z b Z b
= h(t)v0 (t) dt − c v0 (t) dt
a a
= 0 − c(v(b) − v(a))
= 0. (12.38)
Thus, Z b
(h(t) − c)2 dt = 0.
a
Since h is continuous, we may infer that
h(x) − c = 0, ∀x ∈ [a, b],
that is,
h(x) = c, ∀x ∈ [a, b].
Theorem 12.6.2 Let g, h ∈ C([a, b]) and suppose

Z b
(g(x)v(x) + h(x)v0 (x)) dx = 0, ∀v ∈ Va ,
a
where
Va = {v ∈ C1 ([a, b]) : v(a) = v(b) = 0}.
Under such hypotheses, h ∈ C1 ([a, b]) and
h0 (x) = g(x), ∀x ∈ [a, b].
Proof 12.9 Define Z x

G(x) = g(t) dt.
a
Thus,
G0 (x) = g(x), ∀x ∈ [a, b].
Let v ∈ Va .

Z b
0 = [g(x)v(x) + h(x)v0 (x)] dx
a
Z b
= [−G(x)v0 (x) + h(x)v0 (x)] dx + [G(x)v(x)]ab
a
Z b
= [−G(x) + h(x)]v0 (x) dx, ∀v ∈ Va . (12.39)
a
From this and from the du Bois - Reymond lemma, we may conclude that
−G(x) + h(x) = c, ∀x ∈ [a, b],
for some c ∈ R.
Thus
g(x) = G0 (x) = h0 (x), ∀x ∈ [a, b],
so that
g ∈ C1 ([a, b]).
Lemma 12.6.3 (Fundamental lemma of calculus of variation for one dimension) Let g ∈ C([a, b]) = V.
Assume Z b
g(x)v(x) dx = 0, ∀v ∈ Va ,
a
where again,
Va = {v ∈ C1 ([a, b]) : v(a) = v(b) = 0}.
g(x) = 0, ∀x ∈ [a, b].
Proof 12.10 It suffices to apply the last theorem for h ≡ 0.
Exercise 12.6.4 Let h ∈ C([a, b]).

Suppose
Z b
h(x)w(x) dx = 0, ∀w ∈ D0 ,
a
where Z b
D0 = w ∈ C([a, b]) : w(x) dx = 0 .
a
Show that there exists c ∈ R such that
h(x) = c, ∀x ∈ [a, b].
Solution Define, as above indicated,
Va = {v ∈ C1 ([a, b]) : v(a) = v(b) = 0}.
Let v ∈ Va .
Let w ∈ C([a, b]) be such that
w(x) = v0 (x), ∀x ∈ [a, b].
Observe that Z b Z b
w(x) dx = v0 (x) dx = [v(x)]ab = v(b) − v(a) = 0.
a a
Rb
From this a h(x)w(x) dx = 0, and thus
Z b
h(x)v0 (x) dx = 0.
a
Since v ∈ Va is arbitrary, from this and the du Bois-Reymond lemma, there exists c ∈ R such that
h(x) = c, ∀x ∈ [a, b].
12.7 Calculus of variations, the case of scalar functions on Rn

Let Ω ⊂ Rn be an open, bounded, connected with a regular boundary ∂ Ω = S (Lipschitzian) (which we define
as Ω to be of class Ĉ1 ). Let V = C1 (Ω) and let F : D ⊂ V → R, be such that
Z
F(y) = f (x, y(x), ∇y(x)) dx, ∀y ∈ V,
Ω
where we denote
dx = dx1 · · · dxn .
Assume f : Ω × R × Rn → R is of C2
class. Suppose also f (x, y, z) is convex in (y, z), ∀x ∈ Ω, which we
denote by f (x, y, z) to be convex.
Observe that for y ∈ D and v ∈ Va , where
D = {y ∈ V : y = y1 on ∂ Ω},
and
Va = {v ∈ V : v = 0 on ∂ Ω},
where
y1 ∈ C1 (Ω),
we have that
∂
δ F(y; v) = F(y + εv)|ε=0 ,
∂ε
where Z
F(y + εv) = f (x, y + εv, ∇y + ε∇v) dx.
Ω
Therefore,
Z
∂ ∂
F(y + εv) = ( f (x, y + εv, ∇y + ε∇v) dx
∂ε Ω ∂ε
Z n
= [ fy (x, y + εv, ∇y + ε∇v)v + ∑ fzi (x, y + εv, ∇y + ε∇v)vxi ] dx. (12.40)
Ω i=1
Thus,
∂
δ F(y; v) = F(y + εv)|ε=0
∂ε
Z n
= [ fy (x, y, ∇y)v + ∑ fzi (x, y, ∇y)vxi ] dx. (12.41)
Ω i=1
On the other hand, since f (x, y, z) is convex, we have that

Z
F(y + v) − F(y) = [ f (x, y + v, ∇y + ∇v) − f (x, y, ∇y)] dx
Ω
≥ h∇ f (x, y, ∇y), (v, ∇v)iRn+1
Z n
= [ fy (x, y, ∇y)v + ∑ fzi (x, y, ∇y)vxi ] dx
Ω i=1
= δ F(y; v). (12.42)
Since y ∈ D and v ∈ Va are arbitrary, w e may infer that F is convex.

Here we denote,
∇ f (x, y, ∇y) = ( fy (x, y, ∇y), fz1 (x, y, ∇y), · · · , fzn (x, y, ∇y)).
Theorem 12.7.1 Let Ω ⊂ Rn be a set of Ĉ1 class and let V = C1 (Ω). Let f ∈ C2 (Ω × R × R) where f (x, y, z)
is convex. Define
D = {y ∈ V : y = y1 em ∂ Ω},
where y1 ∈ C1 (Ω)
Define also F : D → R by Z
F(y) = f (x, y(x), ∇y(x)) dx.
Ω
From such hypotheses, F is convex and if y0 ∈ D is such that
n
d
∑ dxi [ fzi (x, y0 (x), ∇y0 (x))] = fy (x, y0 (x), ∇y0 (x)), ∀x ∈ Ω,
i=1

F(y0 ) ≤ F(y), ∀y ∈ D.
Proof 12.11 From the last remark, F is convex. Suppose now that y0 ∈ D is such that
n
d
∑ dxi [ fzi (x, y0 (x), ∇y0 (x))] = fy (x, y0 (x), ∇y0 (x)), ∀x ∈ Ω,
i=1
Let v ∈ Va = {v ∈ V : v = 0 on ∂ Ω}. Thus,

Z n
δ F(y0 ; v) = ( fy (x, y0 (x), ∇y0 (x))v(x) + ∑ fzi (x, y0 (x), ∇y0 (x))vxi (x)) dx
Ω i=1
!
n n
d
Z
= ∑ ( fzi (x, y0 (x), ∇y0 (x)))v(x) + ∑ fzi (x, y0 (x), ∇y0 (x))vxi (x) dx
Ω i=1 dxi i=1
!
Z n n
= − ∑ fzi (x, y0 (x), ∇y0 (x))vxi (x) + ∑ fzi (x, y0 (x), ∇y0 (x))vxi (x) dx
Ω i=1 i=1
Z n
+ ∑ fzi (x, y0 (x), ∇y0 (x)) ni v(x) dS
∂ Ω i=1
= 0, ∀v ∈ Va , (12.43)
where n = (n1 , · · · , nn ) denotes the outward normal field to ∂ Ω = S. Since F is convex, from this and Theorem
12.3.5, we have that y0 minimizes F on D.
12.8 ˆ
The second Gateaux variation
Definition 12.8.1 Let V be a Banach space. Let F : D ⊂ V → R be a functional such that δ F(y; v) exists on
Br (y0 ) for y0 ∈ D, r > 0 and for all v ∈ Va .
Let y ∈ Br (y0 ) and v, w ∈ Va . We define the second Gateaux
ˆ variation of F at the point y in the directions
v and w, denoted by δ 2 F(y; v, w), as
δ F(y + εw; v) − δ F(y; v)

δ 2 F(y; v, w) = lim ,
ε →0 ε
if such a limit exists.
Remark 12.8.2 Observe that from this last definition, if the limits in question exist, we have
∂
δ F(y; v) = F(y + εv)|ε=0 ,
∂ε
and
∂2
δ 2 F(y; v, v) = F(y + εv)|ε=0 , ∀v ∈ Va .
∂ ε2
Thus, for example, for V = C1 (Ω) where Ω ⊂ Rn is of Ĉ1 class and F : V → R is given by
Z
F(y) = f (x, y, ∇y) dx
Ω
and where
f ∈ C2 (Ω × R × Rn ),
for y, v ∈ V , we have
∂2
δ 2 F(y; v, v) = F(y + εv)|ε=0 ,
∂ ε2
where
∂2 ∂2
Z
F(y + εv) = f (x, y + εv, ∇y + ε∇v) dx
∂ ε2 ∂ ε2 Ω
∂2
Z
= [ f (x, y + εv, ∇y + ε∇v)] dx
∂ ε2
Ω
"
Z n
= fyy (x, y + εv, ∇y + ε∇v)v2 + ∑ 2 fyzi (x, y + εv, ∇y + ε∇v)vvxi
Ω i=1
#
n n
+ ∑ ∑ fzi z j (x, y + εv, ∇y + ε∇v)vxi vx j dx (12.44)
i=1 j=1
so that
∂2
δ 2 F(y; v, v) = 2
F(y + εv)|ε=0
∂ ε"
Z n
= fyy (x, y, ∇y)v2 + ∑ 2 fyzi (x, y, ∇y)vvxi
Ω i=1
#
n n
+ ∑ ∑ fzi z j (x, y, ∇y)vxi vx j dx. (12.45)
i=1 j=1
12.9 First order necessary conditions for a local minimum

Definition 12.9.1 Let V be a Banach space. Let F : D ⊂ V → R be a functional. We say that y0 ∈ D is a
point of local minimum for F on D, if there exists δ > 0 such that
F(y) ≥ F(y0 ), ∀y ∈ Bδ (y0 ) ∩ D.
Theorem 12.9.2 [First order necessary condition] Let V be a Banach space. Let F : D ⊂ V → R be a
functional. Suppose that y0 ∈ D is a point of local minimum for F on D. Let v ∈ Va and assume δ F(y0 ; v) to
exist.
δ F(y0 ; v) = 0.
Proof 12.12 Define φ (ε) = F(y0 + εv), which from the existence of δ F(y0 ; v) is well defined for all ε
sufficiently small.
Also from the hypotheses, ε = 0 is a point of local minimum for the differentiable at 0 function φ .
Thus, from the standard condition for one variable calculus, we have
φ 0 (0) = 0,
thst is,
φ 0 (0) = δ F(y0 ; v) = 0.
Theorem 12.9.3 (Second order sufficient condition) Let V be a Banach space. Let F : D ⊂ V → R be a
functional. Suppose y0 ∈ D is such that δ F(y0 ; v) = 0 for all v ∈ Va and there exists δ > 0 such that
δ 2 F(y; v, v) ≥ 0, ∀y ∈ Bδ (y0 ) and v ∈ Va .
Under such hypotheses y0 ∈ D is a point of local minimum for F, that is
F(y) ≥ F(y0 ), ∀y ∈ Br (y0 ) ∩ D.
Proof 12.13 Let y ∈ Bδ (y0 ) ∩ D. Define v = y − y0 ∈ Va .

Define also φ : [0, 1] → R by
φ (ε) = F(y0 + εv).
From the Taylor Theorem for one variable, there exists t0 ∈ (0, 1) such that
φ 0 (0) 1
φ (1) = φ (0) + (1 − 0) + φ 00 (t0 )(1 − 0)2 ,
1! 2!
That is,
F(y) = F(y0 + v)
1
= F(y0 ) + δ F(y0 ; v) + δ 2 F(y0 + t0 v; v, v)
2
1
= F(y0 ) + δ 2 F(y0 + t0 v; v, v)
2
≥ F(y0 ), ∀y ∈ Bδ (y0 ) ∩ D. (12.46)

12.10 Continuous functionals

Definition 12.10.1 Let V be a Banach space. Let F : D ⊂ V → R be a functional and let y0 ∈ D.
We say that F is continuous on y0 ∈ D, if for each ε > 0 there exists δ > 0 such that if y ∈ D and
ky − y0 kV < δ , then
|F(y) − F(y0 )| < ε.
Example 12.10.2 Let V = C1 ([a, b]) and f ∈ C([a, b] × R × R).
Consider F : V → R where Z b
F(y) = f (x, y(x), y0 (x)) dx,
a
and
kykV = max{|y(x)| + |y0 (x)| : x ∈ [a, b]}.
Let y0 ∈ V. We shall prove that F is continuous at y0 .
Let y ∈ V be such that
ky − y0 kV < 1.
Thus,
kykV − ky0 kV ≤ ky − y0 kV < 1,
that is,
kykV < 1 + ky0 kV ≡ α.
Observe that f is uniformly continuous on the compact set
[a, b] × [−α, α] × [−α, α] ≡ A.
Let ε > 0. Therefore, there exists δ0 > 0 such that if (x, y1 , z1 ) and (x, y2 , z2 ) ∈ A and
|y1 − y2 | + |z1 − z2 | < δ0 ,
then
ε
| f (x, y1 , z1 ) − f (x, y2 , z2 )| < . (12.47)
b−a
Let δ = min{δ0 , 1}.
Hence, if
ky − y0 kV < δ ,
we have
max{|y(x) − y0 (x)| + |y0 (x) − y00 (x)| : x ∈ [a, b]} < δ ≤ 1,

so that from this and (12.47), we obtain
ε
| f (x, y(x), y0 (x)) − f (x, y0 (x), y00 (x))| < , ∀x ∈ [a, b].
b−a
Thus,
Z b
|F(y) − F(y0 )| = [ f (x, y(x), y0 (x)) − f (x, y0 (x), y00 (x))] dx
a
Z b
≤ | f (x, y(x), y0 (x)) − f (x, y0 (x), y00 (x))| dx
a
ε(b − a)
<
(b − a)
= ε. (12.48)
From this we have that F is continuous at y0 , ∀y0 ∈ V.
12.11 ˆ
The Gateaux variation, the formal proof of its formula
ˆ
In the previous sections we had obtained the Gateaux variations formulas with some informality for a rela-
tively large class of functionals.
In this section we intend to provide a formal proof for such formulas.
Our main result is summarized by the following theorem.
Theorem 12.11.1 Let Ω ⊂ Rn be sets of Ĉ1 class and let V = C1 (Ω).
Let f : Ω × R × Rn → R be a function of C1 class.
Define F : V → R by Z
F(y) = f (x, y(x), ∇y(x)) dx.
Ω
Let y, v ∈ V . Under such hypotheses
!
Z n
δ F(y; v) = fy (x, y(x), ∇y(x))v(x) + ∑ fzi (x, y(x), ∇y(x))vxi (x) d x.
Ω i=1
Proof 12.14 Let {εn } ⊂ R \ {0} be a sequence such that

εn → 0, as n → ∞.
Define
f (x, y(x) + εn v(x), ∇y(x) + εn ∇v(x)) − f (x, y(x), ∇y(x))
Gn (x) = ,
εn
∀n ∈ N, x ∈ Ω.
Define also
n
G(x) = fy (x, y(x), ∇y(x))v(x) + ∑ fzi (x, y(x), ∇y(x))vxi (x), ∀x ∈ Ω.
i=1
Observe that
Gn (x) → G(x), ∀x ∈ Ω.
Now, we are going to prove that, for a not relabeled subsequence,
Z Z
Gn (x) dx → G(x) dx, as n → ∞
Ω Ω
Define
cn = max{|Gn (x) − G(x)|}.
x∈Ω
From the continuity of the functions in question, for each n ∈ N, there exists xn ∈ Ω such that
cn = |Gn (xn ) − G(xn )|.
Observe that {xn } ⊂ Ω and such a set is compact. Thus, there exist a subsequence {xn j } of {xn } and
x0 ∈ Ω such that
lim xn j = x0 .
j→∞
On the other hand, from the mean value theorem, for each j ∈ N there exists t j ∈ (0, 1) such that
f (xn j , y(xn j ) + εn j v(xn j ), ∇y(xn j ) + εn j ∇v(xn j )) − f (xn j , y(xn j ), ∇y(xn j ))
Gn (xn j ) =
εn j
= fy (xn j , y(xn j ) + t j εn j v(xn j ), ∇y(xn j ) + t j εn j ∇v(xn j ))v(xn j )
n
+ ∑ fzi (xn j , y(xn j ) + t j εn v(xn j ), ∇y(xn j ) + t j εn j ∇v(xn j ))vxi (xn j )
i=1
→ G(x0 ), as j → ∞. (12.49)
Hence,
cn j = |Gn j (xn j ) − G(xn j )|

→ |G(x0 ) − G(x0 )|
= 0. (12.50)
Let ε > 0. Thus, there exists j0 ∈ N such that if j > j0 , then

ε
0 ≤ cn j < ,
m(Ω)
where Z
m(Ω) = dx.
Ω
Therefore, if j > j0 , then
Z Z
[Gn j (x) − G(x)] dx ≤ |Gn j (x) − G(x)| dx
Ω Ω
Z
≤ cn j dx
Ω
= cn j m(Ω)
< ε. (12.51)
Thus, Z Z
lim Gn j (x) dx = G(x) dx.
j→∞ Ω Ω
Suppose now,to obtain contradiction, that we do not have

Z Z
lim Gε (x) dx = G(x) dx,
ε→0 Ω Ω
where
f (x, y(x) + εv(x), ∇y(x) + ε∇v(x)) − f (x, y(x), ∇y(x))
Gε (x) = ,
ε
6 0.
∀ε ∈ R such that ε =
Hence, there exists ε0 > 0 such that for each n ∈ N there exists ε̃n ∈ R such that
1
0 < |ε̃n | < ,
n
and Z Z
G̃n (x) dx − G(x) dx ≥ ε0 , (12.52)
Ω Ω
where
f (x, y(x) + ε̃n v(x), ∇y(x) + ε̃n ∇v(x)) − f (x, y(x), ∇y(x))
G̃n (x) = ,
ε̃n
∀n ∈ N, x ∈ Ω.
However, as above indicated, we may obtain a subsequence {ε̃n j } of {ε̃n } such that
Z Z
lim G̃n j (x) dx = G(x) dx,
j→∞ Ω Ω

Therefore, necessarily we have that

Z Z
lim Gε (x) dx = G(x) dx,
ε→0 Ω Ω
that is,
F(y + εv) − F(y)

δ F(y; v) = lim
ε→0
Z
ε
= lim Gε (x) dx
ε→0 Ω
Z
= G(x) dx
Ω
!
Z n
= fy (x, y(x), ∇y(x))v(x) + ∑ fzi (x, y(x), ∇y(x))vxi (x) dx. (12.53)
Ω i=1

Chapter 13
More Topics on the Calculus of

Variations
13.1 Introductory remarks

The main references for this chapter are [79, 44]. We start by recalling that a functional is a function whose
the co-domain is the real set. We denote such functionals by F : U → R, where U is a Banach space. In our
work format, we consider the special cases
f (x, u, ∇u) dx, where Ω ⊂ Rn is an open, bounded, connected set.
R
1. F(u) = Ω
f (x, u, ∇u, D2 u) dx, here

R
2. F(u) = Ω
∂ ui
Du = ∇u =
∂xj
and
∂ 2 ui

2 2
D u = {D ui } = ,
∂ xk ∂ xl
for i ∈ {1, ..., N} and j, k, l ∈ {1, ..., n}.
Also, f : Ω × RN × RN×n → R is denoted by f (x, s, ξ ) and we assume
1.
∂ f (x, s, ξ )
∂s
and
2.
∂ f (x, s, ξ )
∂ξ
are continuous ∀(x, s, ξ ) ∈ Ω × RN × RN×n .
Remark 13.1.1 We also recall that the notation ∇u = Du may be used.

More Topics on the Calculus of Variations 283
Now we define our general problem, namely problem P where

Problem P : minimize F(u) on U,
that is, to find u0 ∈ U such that
F(u0 ) = min{F(u)}.
u∈U
At this point, we introduce some essential definitions.
Theorem 13.1.2 Consider the hypotheses stated at section 13.1 on F : U → R. Suppose F attains a local
minimum at u ∈ C2 (Ω̄; RN ) and additionally assume that f ∈ C2 (Ω, RN , RN×n ). Then the necessary condi-
tions for a local minimum for F are given by the Euler-Lagrange equations:

∂ f (x, u, ∇u) ∂ f (x, u, ∇u)
− d iv = θ , in Ω.
∂s ∂ξ
Proof 13.1 Observe that the standard first order necessary condition stands for δ F(u, ϕ) = 0, ∀ϕ ∈ V .
ˆ
From the related results for the Gateaux variation expression obtained in the last chapter, after integration by
parts, we get
n !
∂ f (x, u, ∇u) d ∂ f (x, u, ∇u)
Z
−∑ i
ϕ i dx = 0,
Ω ∂ si α=1 dxα ∂ ξ α
∀i ∈ {1, · · · , N}, ∀ϕ ∈ Cc∞ (Ω, RN ).

The result thus follows from the fundamental lemma of the calculus of variations (please see Lemma
11.1.12 for the concerning result).
13.2 ˆ
The Gateaux variation, a more general case
Theorem 13.2.1 Consider the functional F : U → R, where
U = {u ∈ W 1,2 (Ω, RN ) | u = u0 in ∂ Ω}.
Suppose Z
F(u) = f (x, u, ∇u) dx,
Ω
where f : Ω × RN × RN×n is such that, for each K > 0 there exists K1 > 0 such that
| f (x, s1 , ξ1 ) − f (x, s2 , ξ2 )| < K1 (|s1 − s2 | + |ξ1 − ξ2 |)

∀s1 , s2 ∈ RN , ξ1 , ξ2 ∈ RN×n , such that |s1 | < K, |s2 | < K, |ξ1 | < K, |ξ2 | < K.
Also assume the hypotheses of section 13.1 except for the continuity of derivatives of f . Under such assump-
tions, for each u ∈ C1 (Ω; RN ) and ϕ ∈ Cc∞ (Ω; RN ), we have
Z
∂ f (x, u, ∇u) ∂ f (x, u, ∇u)
δ F(u, ϕ) = ·ϕ + · ∇ϕ dx.
Ω ∂s ∂ξ
Proof 13.2 First we recall that

F(u + εϕ) − F(u)
δ F(u, ϕ) = lim .
ε→0 ε
Observe that
f (x, u + εϕ, ∇u + ε∇ϕ) − f (x, u, ∇u) ∂ f (x, u, ∇u) ∂ f (x, u, ∇u)
lim = ·ϕ + · ∇ϕ, a.e in Ω.
ε →0 ε ∂s ∂ξ
Define
f (x, u + εϕ, ∇u + ε∇ϕ) − f (x, u, ∇u)
G(x, u, ϕ, ε) = ,
ε
and
∂ f (x, u, ∇u) ∂ f (x, u, ∇u)
G̃(x, u, ϕ) = ·ϕ + · ∇ϕ.
∂s ∂ξ
Thus we have
lim G(x, u, ϕ, ε) = G̃(x, u, ϕ), a.e in Ω.
ε→0
Now will show that Z Z
lim G(x, u, ϕ, ε) dx = G̃(x, u, ϕ) dx.
ε→0 Ω Ω
It suffices to show that (we do not provide details here), for an arbitrary sequence {εn } ⊂ R such that
εn → 0, as n → ∞
we have Z Z
lim G(x, u, ϕ, εn ) dx = G̃(x, u, ϕ) dx.
n→∞ Ω Ω
Observe that, for an appropriate K > 0, we have
|G(x, u, ϕ, εn )| ≤ K(|ϕ| + |∇ϕ|), a.e. in Ω. (13.1)
By the Lebesgue dominated convergence theorem, we obtain
Z Z
lim G(x, u, ϕ, εn ) dx = G̃(x, u, ϕ) dx,
n→∞ Ω Ω
that is, Z
∂ f (x, u, ∇u) ∂ f (x, u, ∇u)
δ F(u, ϕ) = ·ϕ + · ∇ϕ dx.
Ω ∂s ∂ξ
13.3 ´
Frechet differentiability
´
In this section we introduce a very important definition namely, Frechet differentiability.
Definition 13.3.1 Let U,Y be Banach spaces and consider a transformation T : U → Y. We say that T is
´ het differentiable at u ∈ U if there exists a bounded linear transformation T 0 (u) : U → Y such that
Frec
kT (u + v) − T (u) − T 0 (u)(v)kY
lim 6 θ.
= 0, v =
v→θ kvkU
In such a case T 0 (u) is called the Frec
´ het derivative of T at u ∈ U.
13.4 The Legendre-Hadamard condition

Theorem 13.4.1 If u ∈ C1 (Ω̄; RN ) is such that
δ 2 F(u, ϕ ) ≥ 0, ∀ϕ ∈ Cc∞ (Ω, RN ),
then
fξαi ξ k (x, u(x), ∇u(x))ρ i ρ k ηα ηβ ≥ 0, ∀x ∈ Ω, ρ ∈ RN , η ∈ Rn .
β
Such a condition is known as the Legendre-Hadamard condition.

Proof 13.3 Suppose

δ 2 F (u, ϕ ) ≥ 0, ∀ϕ ∈ Cc∞ (Ω; RN ).
We denote δ 2 F(u, ϕ) by
Z
δ 2 F(u, ϕ) = a(x)Dϕ(x) · Dϕ(x) dx
Ω
Z Z
+ b(x)ϕ(x) · Dϕ(x) dx + c(x)ϕ(x) · ϕ(x) dx, (13.2)
Ω Ω
where
a(x) = fξ ξ (x, u(x), Du(x)),
b(x) = 2 fsξ (x, u(x), Du(x)),
and
c(x) = fss (x, u(x), Du(x)).
Thus given x0 ∈ Ω for λ sufficiently small we have that ϕ(x) = λ v x−x
v ∈ Cc∞ (B1 (0), RN ). 0

Now consider λ
is an admissible direction. Now we introduce the new coordinates y = (y1 , ..., yn ) by setting y = λ −1 (x − x0 )
and multiply (13.2) by λ −n to obtain
Z
{a(x0 + λ y)Dv(y) · Dv(y) + 2λ b(x0 + λ y)v(y) · Dv(y) + λ 2 c(x0 + λ y)v(y) · v(y)} dy > 0,
B1 (0)
αβ β
where a = {ai j }, b = {b jk } and c = {c jk }. Since a, b and c are continuous, we have
a(x0 + λ y)Dv(y) · Dv(y) → a(x0 )Dv(y) · Dv(y),
λ b(x0 + λ y)v(y) · Dv(y) → 0,

and
λ 2 c(x0 + λ y)v(y) · v(y) → 0,
uniformly on Ω̄ as λ → 0. Thus this limit give us
Z
f˜jk Dα v j Dβ vk dx ≥ 0, ∀v ∈ Cc∞ (B1 (0); RN ),
αβ
(13.3)
B1 (0)
where
f˜jk = a jk (x0 ) = fξαi ξ k (x0 , u(x0 ), ∇u(x0 )).
αβ αβ
β
Now define v = (v1 , ..., vN ) where

v j = ρ j cos((η · y)t)ζ (y)
ρ = (ρ 1 , ..., ρ N ) ∈ RN
and
η = (η1 , ..., ηn ) ∈ Rn
and ζ ∈ Cc∞ (B1 (0)). From (13.3) we obtain
Z
˜αβ j k
0 ≤ f jk ρ ρ (ηα t(−sin((η · y)t)ζ + cos((η · y)t)Dα ζ )
B1 (0)

· ηβ t (−sin((η · y)t)ζ + cos((η · y)t)Dβ ζ dy (13.4)
By analogy for
v j = ρ j sin((η · y)t)ζ (y)
we obtain
Z
f˜jk ρ j ρ k
αβ
0 ≤ (ηα t(cos((η · y)t)ζ + sin((η · y)t)Dα ζ )
B1 (0)

· ηβ t(cos((η · y)t)ζ + sin((η · y)t)Dβ ζ dy (13.5)
Summing up these last two equations, dividing the result by t 2 and letting t → +∞ we obtain
Z
0 ≤ f˜jk ρ j ρ k ηα ηβ ζ 2 dy,
αβ
B1 (0)
for all ζ ∈ Cc∞ (B1 (0)), which implies

0 ≤ f˜jk ρ j ρ k ηα ηβ .
αβ
13.5 The Weierstrass condition for n = 1

Here we present the Weierstrass condition for the special case N ≥ 1 and n = 1. We start with a definition.
Definition 13.5.1 We say that u ∈ Ĉ([a, b]; RN ) if u : [a, b] → RN is Lipschitz continuous in [a, b], and its
derivative u0 is continuous except on a finite set of points in [a, b]. More specifically, there exists K > 0 such
that
|u0 (x+ )| ≤ K, ∀x ∈ [a, b]
and
|u0 (x−)| ≤ K, ∀x ∈ [a, b]
and there exists at most a finite set of points t1 ,t2 , · · · ,tn ∈ (a, b) such that
u0 (t j +) =
6 u0 (t j −), ∀ j ∈ {1, · · · , n}.
Here, we have denoted

u(x + h) − u(x)
u0 (x+) = lim ,
h→0+ h
and
u(x + h) − u(x)
u0 (x−) = lim , ∀x ∈ (a, b).
h→0− h
Similarly, we define u0 (a+) and u0 (b−).
Theorem 13.5.2 (Weierstrass) Let Ω = (a, b) and f : Ω̄ × RN × RN → R be such that fs (x, s, ξ ) and
¯ × RN × RN .
fξ (x, s, ξ ) are continuous on Ω
Define F : U → R by
Z b
F(u) = f (x, u(x), u0 (x)) dx,
a
where
U = {u ∈ Ĉ1 ([a, b]; RN ) | u(a) = α, u(b) = β }.
Suppose u ∈ U minimizes locally F on U, that is, suppose that there exists ε0 > 0 such that
F(u) ≤ F(v), ∀v ∈ U, such that ku − vk∞ < ε0 .

Under such hypotheses, we have
E(x, u(x), u0 (x+), w) ≥ 0, ∀x ∈ [a, b], w ∈ RN ,
and
E(x, u(x), u0 (x−), w) ≥ 0, ∀x ∈ [a, b], w ∈ RN ,
where
u(x + h) − u(x)
u0 (x+) = lim ,
h→0+ h
and
u(x + h) − u(x)
u0 (x−) = lim , ∀x ∈ (a, b)
h→0− h
and,
E(x, s, ξ , w) = f (x, s, w) − f (x, s, ξ ) − fξ (x, s, ξ )(w − ξ ).
Remark 13.5.3 The function E is known as the Weierstrass Excess Function.
Proof 13.4 Fix x0 ∈ (a, b) and w ∈ RN . Choose 0 < ε < 1 and h > 0 such that u + v ∈ U and
kvk∞ < ε0
where v(x) is given by 

 (x − x0 )w, if 0 ≤ x − x0 ≤ εh,
v(x) = ε̃(h − x + x0 )w, if εh ≤ x − x0 ≤ h,
0, otherwise,

where
ε
ε̃ = .
1−ε
From
F(u + v) − F(u) ≥ 0
we obtain
Z x0 +h Z x0 +h
f (x, u(x) + v(x), u0 (x) + v0 (x)) dx − f (x, u(x), u0 (x)) dx ≥ 0. (13.6)
x0 x0
Define
x − x0
x̃ = ,
h
so that
dx
dx̃ = .
h
Z 1
h f (x0 + x̃h, u(x0 + x̃h) + v(x0 + x̃h), u0 (x0 + x̃h) + v0 (x0 + x̃h) dx̃
0
Z 1
−h f (x0 + x̃h, u(x0 + x̃h), u0 (x0 + x̃h)) dx̃ ≥ 0. (13.7)
0
where the derivatives are related to x.

Therefore
Z ε
f (x0 + x̃h, u(x0 + x̃h) + v(x0 + x̃h), u0 (x0 + x̃h) + w) dx̃
0
Z ε
− f (x0 + x̃h, u(x0 + x̃h), u0 (x0 + x̃h)) dx̃
0
Z 1
+ f (x0 + x̃h, u(x0 + x̃h) + v(x0 + x̃h), u0 (x0 + x̃h) − ε̃w) dx̃
ε
Z 1
− f (x0 + x̃h, u(x0 + x̃h), u0 (x0 + x̃h)) dx̃
ε
≥ 0. (13.8)
Letting h → 0 we obtain
ε( f (x0 , u(x0 ), u0 (x0 +) + w) − f (x0 , u(x0 ), u0 (x0 +))

+(1 − ε)( f (x0 , u(x0 ), u0 (x0 +) − ε̃w) − f (x0 , u(x0 ), u0 (x0 +))) ≥ 0.
Hence, by the mean value theorem we get
ε( f (x0 , u(x0 ), u0 (x0 +) + w) − f (x0 , u(x0 ), u0 (x0 +))

−(1 − ε)ε̃( fξ (x0 , u(x0 ), u0 (x0 +) + ρ(ε̃)w)) · w ≥ 0. (13.9)
Dividing by ε and letting ε → 0, so that ε̃ → 0 and ρ(ε̃) → 0 we finally obtain
f (x0 , u(x0 ), u0 (x0 +) + w) − f (x0 , u(x0 ), u0 (x0 +)) − fξ (x0 , u( x0 ), u0 (x0 +)) · w ≥ 0.
Similarly we may get
f (x0 , u(x0 ), u0 (x0 −) + w) − f (x0 , u(x0 ), u0 (x0 −)) − fξ (x0 , u( x0 ), u0 (x0 −)) · w ≥ 0.
Since x0 ∈ [a, b] and w ∈ RN are arbitrary, the proof is complete.
13.6 The Weierstrass condition, the general case

In this section we present a proof for the Weierstrass necessary condition for N ≥ 1, n ≥ 1. Such a result may
be found in similar form in [44].
Theorem 13.1
Assume u ∈ C1 (Ω; RN ) is a point of strong minimum for a Frec
´ het differentiable functional F : U → R that
is, in particular, there exists ε > 0 such that
F(u + ϕ) ≥ F(u),
for all ϕ ∈ Cc∞ (Ω; Rn ) such that

kϕk∞ < ε.
Here Z
F(u) = f (x, u, Du) dx,
Ω
where we recall to have denoted
∂ ui
Du = ∇u = .
∂xj
Under such hypotheses, for all x ∈ Ω and each rank-one matrix η = {ρi β α } = {ρ ⊗ β }, we have that
E(x, u(x), Du(x), Du(x) + ρ ⊗ β ) ≥ 0,
where
E(x, u(x), Du(x), Du(x) + ρ ⊗ β )

= f (x, u(x), Du(x) + ρ ⊗ β ) − f (x, u(x), Du(x))
−ρ i βα fξαi (x, u(x), Du(x)). (13.10)
Proof 13.5 Since u is a point of local minimum for F, we have that
δ F(u; ϕ) = 0, ∀ϕ ∈ Cc∞ (Ω; RN ),
that is Z
(ϕ · fs (x, u(x), Du(x)) + Dϕ · fξ (x, u(x), Du(x)) dx = 0,
Ω
and hence,
Z
( f (x, u(x), Du(x) + Dϕ(x)) − f (x, u(x), Du(x)) dx
Ω
Z
− (ϕ(x) · fs (x, u(x), Du(x)) − Dϕ(x) · fξ (x, u(x), Du(x)) dx
Ω
≥ 0, (13.11)
∀ϕ ∈ V , where
V = {ϕ ∈ Cc∞ (Ω; RN ) : kϕk∞ < ε}.
Choose a unity vector e ∈ Rn and write
x = (x · e)e + x,
where
x · e = 0.
Denote De v = Dv · e, and let ρ = (ρ1 , ...., ρN ) ∈ RN .
Also, let x0 be any point of Ω. Without loss of generality assume x0 = 0.
Choose λ0 ∈ (0, 1) such that Cλ0 ⊂ Ω, where,
Cλ0 = {x ∈ Rn : |x · e| ≤ λ0 and kxk ≤ λ0 }.
Let λ ∈ (0, λ0 ) and

φ ∈ Cc ((−1, 1); R)
and choose a sequence
φk ∈ Cc∞ ((−λ 2 , λ ); R)
which converges uniformly to the Lipschitz function φλ given by

 t + λ 2, if − λ 2 ≤ t ≤ 0,
φλ = λ (λ − t), if 0 < t < λ (13.12)
0, otherwise

and such that φk0 converges uniformly to φλ0 on each compact subset of
Aλ = {t : −λ 2 < t < λ , t =
6 0}.
We emphasize the choice of {φk } may be such that for some K > 0 we have kφ k∞ < K, kφk k∞ < K and
kφk0 k∞ < K, ∀k ∈ N.
Observe that for any sufficiently small λ > 0 we have that ϕk defined by
ϕk (x) = ρφk (x · e)φ (|x|2 /λ 2 ) ∈ V , ∀k ∈ N
so that letting k → ∞ we obtain that
ϕ(x) = ρφλ (x · e)φ (|x|2 /λ 2 ),
is such that (13.11) is satisfied.

Moreover,
De ϕ(x) = ρφλ0 (x · e)φ (|x|2 /λ 2 ),
and
Dϕ(x) = ρφλ (x · e)φ 0 (|x|2 /λ 2 )2λ −2 x,
where D denotes the gradient relating the variable x.
Note that, for such a ϕ(x) the integrand of (13.11) vanishes if x 6∈ Cλ , where
Cλ = {x ∈ Rn : |x · e| ≤ λ and kxk ≤ λ }.
Define Cλ+ and Cλ− by
Cλ− = {x ∈ Cλ : x · e ≤ 0},
and
Cλ+ = {x ∈ Cλ : x · e > 0}.
Hence, denoting
gk (x) = ( f (x, u(x), Du(x) + Dϕk (x)) − f (x, u(x), Du(x))

−(ϕk (x) · fs (x, u(x), Du(x) + Dϕk (x) · fξ (x, u(x), Du(x)) (13.13)
and
g(x) = ( f (x, u(x), Du(x) + Dϕ(x)) − f (x, u(x), Du(x))

−(ϕ(x) · fs (x, u(x), Du(x) + Dϕ(x) · fξ (x, u(x), Du(x)) (13.14)
letting k → ∞, using the Lebesgue dominated converge theorem we obtain

Z Z
−
gk (x) dx + gk (x) dx
Cλ Cλ+
Z Z
→ −
g(x) dx + g(x) dx ≥ 0, (13.15)
Cλ Cλ+
Now define
y = ye e + y,
where
x·e
ye = ,
λ2
and
x
y= .
λ
The sets Cλ− and Cλ+ correspond, concerning the new variables, to the sets Bλ− and Bλ+ , where
B−
λ = {y : kyk ≤ 1, and − λ
−1
≤ ye ≤ 0},
−1
B+ e
λ = {y : kyk ≤ 1, and 0 < y ≤ λ }.
Therefore, since dx = λ n+1 dy, multiplying (13.15) by λ −n−1 , we obtain
Z Z Z
g(x(y)) dy + g(x(y)) dy + g(x(y)) dy ≥ 0, (13.16)
B−
1 B−
λ
\B−
1 B+
λ
where
x = (x · e)e + x = λ 2 ye + λ y ≡ x(y).
Observe that 
 ρφ (kyk2 ) if − 1 ≤ ye ≤ 0,
De ϕ(x) = ρφ (kyk2 )(−λ ) if 0 ≤ ye ≤ λ −1 , (13.17)
0, otherwise.

Observe also that

q
|g(x(y))| ≤ o( |ϕ(x)|2 + |Dϕ(x)|2 ),
so that from the from the expression of ϕ(x) and Dϕ(x) we obtain, for
− −
y ∈ B+
λ , or y ∈ Bλ \ B1 ,
that
|g(x(y))| ≤ o(λ ), as λ → 0.
Since the Lebesgue measures of B− +
λ and Bλ are bounded by
2n−1 /λ
the second and third terms in (13.16) are of o(1) where
lim o(1)/λ = 0,
λ →0+
so that letting λ → 0+ , considering that

x(y) → 0,
and on B−
1 (up to the limit set B)
g(x(y)) → f (0, u(0), Du(0) + ρφ (kyk2 )e)

− f (0, u(0), Du(0)) −
ρφ (kyk2 )e fξ (0, u(0), Du(0)) (13.18)
we get,
Z
[ f (0, u(0), Du(0) + ρφ (kyk2 )e) − f (0, u(0), Du(0))
B
−ρφ (kyk2 )e fξ (0, u(0), Du(0))] dy2 ...dyn
≥ 0, (13.19)
where B is an appropriate limit set (we do not provide more details here) such that
B = {y ∈ Rn : ye = 0 and kyk ≤ 1}.
Here we have used the fact that, on the set in question,
Dϕ(x) → ρφ (kyk2 )e, as λ → 0+ .
Finally, inequality (13.19) is valid for a sequence {φn } (in place of φ ) such that
0 ≤ φn ≤ 1 and φn (t) = 1, if |t| < 1 − 1/n,

∀n ∈ N.
Letting n → ∞, from (13.19) we obtain
f (0, u(0), Du(0) + ρ ⊗ e) − f (0, u(0), Du(0))

−ρ · e fξ (0, u(0), Du(0)) ≥ 0. (13.20)
13.7 The Weierstrass-Erdmann conditions

We start with a definition.
Definition 13.7.1 Define I = [a, b]. A function u ∈ Ĉ([a, b]; RN ) is said to be a weak Lipschitz extremal of
Z b
F(u) = f (x, u(x), u0 (x)) dx,
a
if
Z b
( fs (x, u(x), u0 (x)) · ϕ + fξ (x, u(x), u0 (x)) · ϕ 0 (x)) dx = 0,
a
∀ϕ ∈ Cc∞ ([a, b]; RN ).
Proposition 13.7.2 For any Lipschitz extremal of

Z b
F(u) = f (x, u(x), u0 (x)) dx
a
there exists a constant c ∈ RN such that

Z x
fξ (x, u(x), u0 (x)) = c + fs (t, u(t), u0 (t)) dt, ∀x ∈ [a, b]. (13.21)
a
Proof 13.6 Fix ϕ ∈ Cc∞ ([a, b]; RN ). Integration by parts of the extremal condition
δ F(u, ϕ) = 0,
implies that
Z b Z bZ x
fξ (x, u(x), u0 (x)) · ϕ 0 (x) dx − fs (t, u(t), u0 (t)) dt · ϕ 0 (x) dx = 0.
a a a
Since ϕ is arbitrary, from the du Bois-Reymond lemma, there exists c ∈ RN such that
Z x
fξ (x, u(x), u0 (x)) − fs (t, u(t), u0 (t)) dt = c, ∀x ∈ [a, b].
a
Theorem 13.7.3 (Weierstrass-Erdmann corner conditions) Let I = [a, b]. Suppose u ∈ Ĉ1 ([a, b]; RN ) is
such that
F(u) ≤ F(v), ∀v ∈ Cr ,
for some r > 0. where
Cr = {v ∈ Ĉ1 ([a, b]; RN ) | v(a) = u(a), v(b) = u(b), and ku − vk∞ < r}.
Let x0 ∈ (a, b) be a corner point of u. Denoting u0 = u(x0 ), ξ0+ = u0 (x0+ ) and ξ0− = u0 (x0− ), then the
following relations are valid:
1. fξ (x0 , u0 , ξ0− ) = fξ (x0 , u0 , ξ0+ ),
2.
f (x0 , u0 , ξ0− ) − ξ0− fξ (x0 , u0 , ξ0− )

= f (x0 , u0 , ξ0+ ) − ξ0+ fξ (x0 , u0 , ξ0+ ).
Remark 13.7.4 The conditions above are known as the Weierstrass-Erdmann corner conditions.
Proof 13.7 Condition (1) is just a consequence of equation (13.21). For (2), define
τε (x) = x + ελ (x),
where λ ∈ Cc∞ (I). Observe that τε (a) = a and τε (b) = b, ∀ε > 0. Also τ0 (x) = x. Choose ε0 > 0 sufficiently
small such that for each ε satisfying |ε| < ε0 , we have τε0 (x) > 0 and
ũε (x) = (u ◦ τε−1 )(x) ∈ Cr .
Define
φ (ε) = F(x, ũε , ũ0ε (x)).
Thus φ has a local minimum at 0, so that φ 0 (0) = 0, that is
d(F(x, ũε , ũ0ε (x)))

|ε =0 = 0.
dε
Observe that
dũε dτ −1 (x)
= u0 (τε−1 (x)) ε ,
dx dx
and
dτε−1 (x) 1
= .
dx 1 + ελ 0 (τε−1 (x))
Thus, Z b
1
F(ũε ) = f x, u(τε−1 (x)), u0 (τε−1 (x)) dx.
a 1 + ελ 0 (τε−1 (x))
Defining
x̄ = τε−1 (x),
we obtain
1
dx̄ = dx,
1 + ελ 0 (x̄)
that is
dx = (1 + ελ 0 (x̄)) dx̄.
Dropping the bar for the new variable, we may write
u0 (x)
Z b
1 + ελ 0 (x) dx.

F(ũε ) = f x + ελ (x), u(x),
a 1 + ελ 0 (x)
From
dF(ũε )
|ε=0 ,
dε
we obtain
Z b
(λ fx (x, u(x), u0 (x)) + λ 0 (x)( f (x, u(x), u0 (x)) − u0 (x) fξ (x, u(x), u0 (x)))) dx = 0. (13.22)
a
Since λ is arbitrary, from proposition 13.7.2, we obtain

Z x
f (x, u(x), u0 (x)) − u0 (x) fξ (x, u(x), u0 (x)) − fx (t, u(t), u0 (t)) dt = c1
a
for some cR1 ∈ R.

Being ax fx (t, u(t), u0 (t)) dt + c1 a continuous function (in fact absolutely continuous), the proof is com-
plete.
13.8 Natural boundary conditions

Consider the functional F : U → R, where
Z
F(u) = f (x, u(x), ∇u(x)) dx,
Ω
f (x, s, ξ ) ∈ C1 (Ω̄, RN , RN×n ),

and Ω ⊂ Rn is an open bounded connected set.
Proposition 13.8.1 Assume
U = {u ∈ W 1,2 (Ω; RN ); u = u0 on Γ0 },
where Γ0 ⊂ ∂ Ω is closed and ∂ Ω = Γ = Γ0 ∪ Γ1 being Γ1 open in Γ and Γ0 ∩ Γ1 = 0/. Thus if ∂ Ω ∈ C1 ,
f ∈ C2 (Ω̄, RN , RN×n ) and u ∈ C2 (Ω̄; RN ), and also
δ F(u, ϕ) = 0, ∀ϕ ∈ C1 (Ω̄; RN ), such that ϕ = 0 on Γ0 ,
then u is a extremal of F which satisfies the following natural boundary conditions,
nα fξαi (x, u(x)∇u(x)) = 0, a.e. on Γ1 , ∀i ∈ {1, ..., N}.

Proof 13.8 Observe that δ F(u, ϕ) = 0, ∀ϕ ∈ Cc∞ (Ω; RN ), thus u is a extremal of F and through integration
by parts and the fundamental lemma of calculus of variations, we obtain
L f (u) = 0, in Ω,
where
L f (u) = fs (x, u(x), ∇u(x)) − div( fξ (x, u(x), ∇u(x)).
Defining
V = {ϕ ∈ C1 (Ω; RN ) | ϕ = 0 on Γ0 },
for an arbitrary ϕ ∈ V , we obtain
Z
δ F(u, ϕ) = L f (u) · ϕ dx
Ω
Z
+ nα fξαi (x, u(x), ∇u(x))ϕ i (x) dΓ
Γ1
Z
= nα fξαi (x, u(x), ∇u(x))ϕ i (x) dΓ
Γ1
= 0, ∀ϕ ∈ V . (13.23)
nα fξαi (x0 , u(x0 ), ∇u(x0 )) = β > 0,
for some x0 ∈ Γ1 and some i ∈ {1, ..., N}. Defining
G(x) = nα fxi iα (x, u(x), ∇u(x)),
by the continuity of G, there exists r > 0 such that
G(x) > β /2, in Br (x0 ),
and in particular
G(x) > β /2, in Br (x0 ) ∩ Γ1 .
Choose 0 < r1 < r such that Br1 (x0 ) ∩ Γ0 = 0/. This is possible since Γ0 is closed and x0 ∈ Γ1 .
Choose ϕ i ∈ Cc∞ (Br1 (x0 )) such that ϕ i ≥ 0 in Br1 (x0 ) and ϕ i > 0 in Br1 /2 (x0 ). Therefore
Z Z
β
G(x)ϕ i (x) dx > ϕ i dx > 0,
Γ1 2 Γ1
and this, for ϕ = (0, · · · , ϕ i , · · · , 0) ∈ Cc∞ (Ω; RN ), contradicts (13.23). Thus
G(x) ≤ 0, ∀x ∈ Γ1 ,
and by analogy
G(x) ≥ 0, ∀x ∈ Γ1 ,
so that
G(x) = 0, ∀x ∈ Γ1 .
Chapter 14
Convex Analysis and Duality

Theory
14.1 Convex sets and functions

For this section the most relevant reference is Ekeland and Temam [33].
Definition 14.1.1 (Convex functional) Let U be a vector space and let S ⊂ U be a convex set. A functional
F : S → R̄ = R ∪ {+∞, −∞} is said to be convex, if
F(λ u + (1 − λ )v) ≤ λ F(u) + (1 − λ )F(v), ∀u, v ∈ S, λ ∈ [0, 1]. (14.1)
14.2 Weak lower semi-continuity

We start with the definition of Epigraph.
Definition 14.2.1 (Epı́graph) Let U be a Banach space and let F : U → R̄ be a functional.
We define the Epigraph of F, denoted by E pi(F), by
E pi(F) = {(u, a) ∈ U × R | a ≥ F(u)}.
Definition 14.2.2 Let U be a Banach space. Consider the weak topology σ (U,U ∗ ) for U and let F : U →
R ∪ {+∞} be a functional. Let u ∈ U. We say that F is weakly lower semi-continuous at u ∈ U if for each
λ < F(u), there exists a weak neighborhood Vλ (u) ∈ σ (U,U ∗ ) such that
F(v) > λ , ∀v ∈ Vλ (u).
If F is weakly lower semi-continuous (w.l.s.c.) on U, we write simply F is w.l.s.c..
Theorem 14.2.3 Let U be a Banach space and let F : U → R ∪ {+∞} be a functional.

Under such hypotheses, the following properties are equivalent.
1. F is w.l.s.c..
2. E pi(F) is closed for U × R with the product topology between σ (U,U ∗ ) and the usual topology
for R.
Convex Analysis and Duality Theory 297
3. HγF = {u ∈ U | F(u) ≤ γ} is closed for σ (U,U ∗ ), ∀γ ∈ R.
4. The set GFγ = {u ∈ U | F(u) > γ} is open for σ (U,U ∗ ), ∀γ ∈ R.

5.
lim inf F(v) ≥ F(u), ∀u ∈ U,
v*u
where
lim inf F(v) = sup inf F(v).
v*u V (u)∈σ (U,U ∗ ) v∈V (u)
Proof 14.1 Assume F is w.l.s.c.. We are going to show that E pi(F )c is open for σ (U,U ∗ ) × R. Choose
(u, r) ∈ E pi(F)c . Thus (u, r) 6∈ E pi(F), so that r < F(u). Select λ such that r < λ < F(u). Since F is w.l.s.c.
at u, there exists a a weak neighborhood Vλ (u) such that
F(v) > λ , ∀v ∈ Vλ (u).
Thus,
Vλ (u) × (−∞, λ ) ⊂ E pi(F )c
so that (u, r) is an interior point of E pi(F )c and hence, since such a point is arbitrary in E pi(F )c , we may
infer that E pi(F)c is open so that E pi(F) is closed for the topology in question. Assume now (2). Observe
that
HγF × {γ} = E pi(F) ∩ (U × {γ}).
From the hypotheses E pi(F) is closed, that is, HγF × {γ} is closed and thus HγF is closed.
Assume (3). To obtain (4), it suffices to consider the complement of HγF . Suppose (4) is valid. Let u ∈ U
and let γ ∈ R be such that
γ < F(u).
Since GFγ is open for σ (U,U ∗ ) there exists a weak neighborhood V (u) such that
V (u) ⊂ GFγ ,
so that
F(v) > γ, ∀v ∈ V (u),
and hence
inf F(v) ≥ γ .
v∈V (u)
In particular, we have
lim inf F(v) ≥ γ.
v*u
Letting γ → F(u), we obtain

lim inf F(v) ≥ F(u).
v*u
Finally assume
lim inf F(v) ≥ F(u).
v*u
Let λ < F(u) and let 0 < ε < F(u) − λ .

Observe that
lim inf F(v) = sup inf F(v).
v*u V (u)∈σ (U,U ∗ ) v∈V (u)
Thus, there exists a weak neighborhood V (u) such that F(v) ≥ F(u) − ε > λ , ∀v ∈ V (u).
Remark 14.2.4 A similar result is valid for the strong topology (in norm) of a Banach space U so that a
F : U → R ∪ {+∞} is strongly lower semi-continuous (l.s.c.) at u ∈ U, if
lim inf F(v) ≥ F(u). (14.2)
v→u
Corollary 14.2.5 All convex l.s.c. functional F : U → R is also w.l.s.c..
Proof 14.2 The result follows from the fact for F l.s.c., its epigraph being convex and strongly closed, it is
also weakly closed.
Definition 14.2.6 (Affine-continuous functionals) Let U be a Banach space. A functional F : U → R is

said to be affine-continuous, if there exist u∗ ∈ U ∗ and α ∈ R such that
F(u) = hu, u∗ iU + α, ∀u ∈ U. (14.3)
Definition 14.2.7 (Γ(U)) Let U be a Banach space. We say that F : U → R̄ is a functional in Γ(U) and write
F ∈ Γ(U) if F may be represented point-wise as the supremum of a family of affine-continuous functionals.
If F ∈ Γ(U), and F(u) ∈ R for some u ∈ U we write F ∈ Γ0 (U).
Definition 14.2.8 (Convex envelop) Let U be a Banach space. Let F : U → R̄, be a functional. we define its
convex envelop, denoted by CF : U → R̄, as
CF(u) = sup {hu, u∗ i + α}, (14.4)
(u∗ ,α)∈A∗
where
A∗ = {(u∗ , α) ∈ U ∗ × R | hv, u∗ iU + α ≤ F(v), ∀v ∈ U} (14.5)
14.3 Polar functionals and related topics on convex analysis

Definition 14.3.1 (Polar functional) Let U be a Banach space and let F : U → R̄, be a functional. We define
the polar functional related to F, denoted by F ∗ : U ∗ → R̄, by
F ∗ (u∗ ) = sup{hu, u∗ iU − F(u)}, ∀u∗ ∈ U ∗ . (14.6)
u∈U
Definition 14.3.2 (Bipolar functional) Let U be a Banach space and let F : U → R̄ be a functional. We
define the bi-polar functional related to F, denoted by F ∗∗ : U → R̄, as
F ∗∗ (u) = sup {hu, u∗ iU − F ∗ (u∗ )}, ∀u ∈ U. (14.7)
u∗ ∈U ∗
Proposition 14.3.3 Let U be a Banach space and let F : U → R̄ be a functional. Under such hypotheses
F ∗∗ (u) = CF(u), ∀u ∈ U and in particular, if F ∈ Γ(U), then F ∗∗ (u) = F(u), ∀u ∈ U.
Proof 14.3 By the definition, the convex envelop of F is the supremum of affine-continuous functionals
bounded by F at the point in question. In fact we need only to consider those which are maximal, that is,
only those in the form
u 7→ hu, u∗ iU − F ∗ (u∗ ). (14.8)
Thus,
CF(u) = sup {hu, u∗ iU − F ∗ (u∗ )} = F ∗∗ (u). (14.9)
u∗ ∈U ∗
Corollary 14.3.4 Let U be a Banach space and let F : U → R̄ be a functional. Under such hypotheses,
F ∗ = F ∗∗∗ .
Proof 14.4 Since F ∗∗ ≤ F, we obtain
F ∗ ≤ F ∗∗∗ . (14.10)
On the other hand,
F ∗∗ (u) ≥ hu, u∗ iU − F ∗ (u∗ ), (14.11)
so that
F ∗∗∗ (u∗ ) = sup{hu, u∗ iU − F ∗∗ (u)} ≤ F ∗ (u∗ ). (14.12)

u∈U
From (14.10) and (14.12), we have F ∗ (u∗ ) = F ∗∗∗ (u∗ ), ∀u∗ ∈ U ∗ .
ˆ
At this point, we recall the definition of Gateaux differentiability.
ˆ
Definition 14.3.5 (Gateaux differentiability) Let U be a Banach space. A functional F : U → R̄ is said to
be Gateaux
ˆ differentiable at u ∈ U, if there exists u∗ ∈ U ∗ such that
F(u + λ h) − F (u)
lim = hh, u∗ iU , ∀h ∈ U. (14.13)
λ →0 λ
The vector u∗ is said to be the Gateaux
ˆ derivative of F : U → R at u and may denoted by
∂ F(u)
u∗ = or u∗ = δ F(u) (14.14)
∂u
Definition 14.3.6 (Sub-gradients) Let U be a Banach space and let F : U → R̄ be a functional. We define
the set of sub-gradients of F at u, denoted by ∂ F(u), by
∂ F(u) = {u∗ ∈ U ∗ , such that

hv − u, u∗ iU + F(u) ≤ F(v), ∀v ∈ U}. (14.15)
Lemma 14.3.7 (Continuity of convex functions) Let U be a Banach space and let F : U → R be a convex
functional. Let u ∈ U and suppose there exists a > 0 and a neighborhood V of u such that
F(v) < a < +∞, ∀v ∈ V.
From the hypotheses, F is continuous at u.
Proof 14.5 Redefining the problem with G(v) = F(v + u) − F(u) we need only consider the case in which
u = 0 and F(u) = 0. Let V be a neighborhood of 0 such that F(v) ≤ a < +∞, ∀v ∈ V . Define W = V ∩(−V ).
Choose ε ∈ (0, 1). Let v ∈ εW , thus
v
∈V (14.16)
ε
and since F is convex, we have that
v
F(v) = F (1 − ε)0 + ε ≤ (1 − ε)F(0) + εF(v/ε) ≤ εa. (14.17)
ε
Also
−v
∈V. (14.18)
ε
Hence,
v (−v/ε) F(v) ε
F(0) = F +ε ≤ + F(−v/ε),
1+ε 1+ε 1+ε 1+ε
so that
F(v) ≥ (1 + ε)F(0) − εF(−v/ε) ≥ −εa. (14.19)
Therefore
|F(v)| ≤ εa, ∀v ∈ εW , (14.20)
that is, F is continuous at u = 0.
Proposition 14.3.8 Let U be a Banach space and let F : U → R̄ be a convex functional, which is finite and
continuous at u ∈ U. Under such hypotheses, ∂ F(u) =
6 0/.
Proof 14.6 Since F is convex, E pi(F) is convex. Since F is continuous at u, we have that E pi(F)0 is
non-empty. Observe that (u, F(u)) is on the boundary of E pi(F). Therefore, denoting A = E pi(F), from the
Hahn-Banach theorem there exists a closed hyperplane H which separates (u, F(u)) and A0 , where H
H = {(v, a) ∈ U × R | hv, u∗ iU + αa = β }, (14.21)

for some fixed α, β ∈ R and u∗ ∈ U ∗, so that
hv, u∗ iU + αa ≥ β , ∀(v, a) ∈ E pi(F), (14.22)
and
hu, u∗ iU + αF(u) = β , (14.23)
where (α, β , u∗ ) =
6 (0, 0, 0). Suppose, to obtain contradiction, that α = 0.
Thus,
hv − u, u∗ iU ≥ 0, ∀v ∈ U, (14.24)
and therefore we obtain, u∗ = 0 and β = 0, a contradiction. Hence, we may assume α > 0 (considering
(14.22)) and thus ∀v ∈ U we have
β
− hv, u∗ /αiU ≤ F(v), (14.25)
α
and
β
− hu, u∗ /αiU = F(u), (14.26)
α
that is,
hv − u, −u∗ /αiU + F(u) ≤ F(v), ∀v ∈ U, (14.27)
so that
−u∗ /α ∈ ∂ F(u). (14.28)
´
Definition 14.3.9 (Caratheodory function) Let S ⊂ Rn be an open set. We say that g : S × Rl → R is a
Caratheodory
´ function if
∀ξ ∈ Rl , x 7→ g(x, ξ ) is a measurable function,
and
a.e. in S, ξ 7→ g(x, ξ ) is a continuous function.
The proof of next results may be found in Ekeland and Temam [33].
Proposition 14.3.10 Let E and F be two Banach spaces, let S be a Borel subset of Rn , and g : S × E → F
be a Caratheodory
´ function. For each measurable function u : S → E, let G1 (u) be the measurable function
x 7→ g(x, u(x)) ∈ F.
Under such hypotheses, if G1 maps L p (S, E) on Lr (S, F) for 1 ≤ p, r < ∞, then G1 is strongly continuous.
,where U = U ∗ = [L2 (S)]l , we have also

R
For the functional G : U → R, defined by G(u) = S g(x, u(x))dS
the following result.
Proposition 14.3.11 Considering the statement in the last proposition we may express G∗ : U ∗ → R̄ by
Z
G∗ (u∗ ) = g∗ (x, u∗ (x))dx, (14.29)
S
onde g∗ (x, y) = sup (y · η − g(x, η)), for almost all x ∈ S.

η ∈Rl
14.4 The Legendre transform and the Legendre functional

For non-convex functionals, in some cases the global extremal through which the polar functional is obtained
corresponds to a local extremal point which the analytical expression is not possible.
This fact motivates the definition of the Legendre Transform, which is obtained through a local extremal
point.
Definition 14.4.1 (Legendre transform and associated functional) Consider the function of C2 class, g :
Rn → R. Its Legendre transform, denoted by g∗L : RnL → R, is expressed by
n
g∗L (y∗ ) = ∑ x0i · y∗i − g(x0 ), (14.30)
i=1
where x0 is a solution of the system
∂ g(x0 )
y∗i = , (14.31)
∂ xi
and RnL = {y∗ ∈ Rn such an equation (14.31) has a unique solution}. R

Moreover, considering the functional G : Y → R defined by G(v) = S g(v)dS, we also define the associ-
ated Legendre functional, denoted by G∗L : YL∗ → R as
Z
G∗L (v∗ ) = g∗L (v∗ ) dx, (14.32)
S
where YL∗ = {v∗ ∈ Y ∗ | v∗ (x) ∈ RnL , a.e. in S}.
About the Legendre transform we still have the following results:

Proposition 14.4.2 Considering the last definitions, suppose for x̂0 ∈ Rn we have
2
∂ g(x̂0 )
det 6= 0.
∂ xi ∂ x j
Let y∗0 be such that
∂ g(x̂0 )
(y0∗ )i = , ∀i ∈ {1, . . . , n}.
∂ xi
Under such hypotheses, y∗0 ∈ RnL and
∂ g −1 ∗
x0 (y∗ ) = [ ] (y )
∂x
is of C1 class in a neighborhood of y∗0 .
Moreover, for all y∗ in such a neighborhood, we have that
∂ g(x0 ) ∂ g∗ (y∗ )
y∗i = , ∀i ∈ {1, ..., n} ⇔ x0i = L ∗ , ∀i ∈ {1, ..., n}.
∂ xi ∂ yi
Proof 14.7 The proof that x0 (y∗ ) is of C1 class in a neighborhood of y∗0 results from the inverse function
theorem.
Suppose now that:
∂ g(x0 )
y∗i = , ∀ i ∈ {1, ..., n}, (14.33)
∂ xi
thus:
g∗L (y∗ ) = y∗i x0i − g(x0 ) (14.34)
and taking derivatives for this expression we have:

∂ g∗L (y∗ ) ∂x j ∂ g(x0 ) ∂ x0 j
∗ = y∗j 0∗ + x0i − , (14.35)
∂ yi ∂ yi ∂ x j ∂ y∗i
or
∂ g∗L (y∗ )

∗ ∂ g(x0 ) ∂ x0 j
= y j − + x0i (14.36)
∂ y∗i ∂xj ∂ y∗i
which from (14.33) implies that:
∂ g∗L (y∗ )
= x0i , ∀ i ∈ {1, ..., n}. (14.37)
∂ y∗i
This completes the first half of the proof. Conversely, suppose now that:
∂ g∗L (y∗ )
x0i = , ∀i ∈ {1, ..., n}. (14.38)
∂ y∗i
As y∗ ∈ RnL there exists x̄0 ∈ Rn such that:
∂ g(x̄0 )
y∗i = ∀i ∈ {1, ..., n}, (14.39)
∂ xi
and,
g∗L (y∗ ) = y∗i x̄0i − g(x̄0 ) (14.40)

and therefore taking derivatives for this expression we can obtain:

∂ g∗L (y∗ ) ∂ x̄ j ∂ g(x̄0 ) ∂ x̄0 j
∗ = y∗j 0∗ + x̄0i − , (14.41)
∂ yi ∂ yi ∂ x j ∂ y∗i
∀ i ∈ {1, ..., n}, so that:
∂ g∗L (y∗ )

∗ ∂ g(x̄0 ) ∂ x̄0 j
= yj − + x̄0i (14.42)
∂ y∗i ∂xj ∂ y∗i
∀ i ∈ {1, ..., n}, which from (14.38) and (14.39), implies that:
∂ g∗L (y∗ )
x̄0i = = x0i , ∀ i ∈ {1, ..., n}, (14.43)
∂ y∗i
from this and (14.39) we have:

∂ g(x̄0 ) ∂ g(x0 )
y∗i = = ∀ i ∈ {1, ..., n}. (14.44)
∂ xi ∂ xi
Theorem 14.4.3 Consider the functional J : U → R̄ defined as J(u) = (G ◦ Λ)(u) − hu, f iU where Λ(=
{Λi }) : U → Y (i ∈ {1, ..., n}) is a continuous linear operator and, G : Y → R is a functional that can be
expressed as G(v) = S g(v)dS, ∀v ∈ Y (here g : Rn → R is a differentiable function that admits Legendre
R
Transform denoted by g∗L : RnL → R. That is, the hypothesis mentioned at Proposition 14.4.2 are satisfied).
Under these assumptions we have:
δ J(u0 ) = θ ⇔ δ (−G∗L (v∗0 ) + hu0 , Λ∗ v∗0 − f iU ) = θ , (14.45)

∂ G(Λ(u0 ))
where v∗0 = ∂v is supposed to be such that v∗0 (x) ∈ RnL , a.e. in S and in this case:
J(u0 ) = −G∗L (v∗0 ). (14.46)
Proof 14.8 Suppose first that δ J(u0 ) = θ , that is:
∂ G(Λu0 )
Λ∗ − f =θ (14.47)
∂v
∂ G(Λu0 )
which, as v∗0 = ∂v implies that:
Λ∗ v∗0 − f = θ , (14.48)
and
∂ g(Λu0 )
v∗0i = . (14.49)
∂ xi
Thus from the last proposition we can write:
∂ g∗L (v∗0 )
Λi (u0 ) = , for i ∈ {1, .., n} (14.50)
∂ y∗i
which means:
∂ G∗L (v∗0 )
Λu0 = . (14.51)
∂ v∗
Therefore from (14.48) and (14.51) we have:

δ (−G∗L (v∗0 ) + hu0 , Λ∗ v∗0 − f iU ) = θ . (14.52)
This completes the first part of the proof.
Conversely, suppose now that:
δ (−G∗L (v∗0 ) + hu0 , Λ∗ v∗0 − f iU ) = θ , (14.53)
that is:
Λ∗ v∗0 − f = θ (14.54)
and
∂ G∗L (v∗0 )
Λu0 = . (14.55)
∂ v∗
Clearly, from (14.55), the last proposition and (14.54) we can write:
∂ G(Λ(u0 ))
v∗0 = (14.56)
∂v
and
∂ G(Λu0 )
Λ∗ − f = θ, (14.57)
∂v
which implies:
δ J(u0 ) = θ . (14.58)
Finally, we have:
J(u0 ) = G(Λu0 ) − hu0 , f iU (14.59)
From this, (14.54) and (14.56) we have
J(u0 ) = G(Λu0 ) − hu0 , Λ∗ v∗0 iU = G(Λu0 ) − hΛu0 , v∗0 iY = −G∗L (v∗0 ). (14.60)
14.5 Duality in convex optimization

Let U be a Banach space. Given F : U → R̄ (F ∈ Γ0 (U)) we define the problem P as
P : minimize F(u) on U. (14.61)
We say that u0 ∈ U is a solution of problem P if F(u0 ) = infu∈U F(u). Consider a function φ (u, p) (φ :
U ×Y → R̄) such that
φ (u, 0) = F(u), (14.62)
we define the problem P ∗ , as
P ∗ : maximize − φ ∗ (0, p∗ ) on Y ∗ . (14.63)
Observe that
φ ∗ (0, p∗ ) = sup {h0, uiU + h p, p∗ iY − φ (u, p)} ≥ −φ (u, 0), (14.64)
(u,p)∈U ×Y
or
inf {φ (u, 0)} ≥ sup {−φ ∗ (0, p∗ )}. (14.65)
u∈U p∗ ∈Y ∗
Proposition 14.5.1 Consider φ ∈ Γ0 (U ×Y ). If we define

h(p) = inf {φ (u, p)}, (14.66)
u∈U
then h is convex.
Proof 14.9 We have to show that given p, q ∈ Y and λ ∈ (0, 1), we have
h(λ p + (1 − λ )q) ≤ λ h(p) + (1 − λ )h(q). (14.67)
If h(p) = +∞ or h(q) = +∞ we are done. Thus let us assume h(p) < +∞ and h(q) < +∞. For each a > h(p)
there exists u ∈ U such that
h(p) ≤ φ (u, p) ≤ a, (14.68)
and, if b > h(q), there exists v ∈ U such that
h(q) ≤ φ (v, q) ≤ b. (14.69)
Thus
h(λ p + (1 − λ )q) ≤ inf {φ (w, λ p + (1 − λ )q)}

w∈U
≤ φ (λ u + (1 − λ )v, λ p + (1 − λ )q) ≤ λ φ (u, p) + (1 − λ )φ (v, q)
≤ λ a + (1 − λ )b. (14.70)
Letting a → h(p) and b → h(q) we obtain
h(λ p + (1 − λ )q) ≤ λ h(p) + (1 − λ )h(q). (14.71)
Proposition 14.5.2 For h as above, we have h∗ (p∗ ) = φ ∗ (0, p∗ ), ∀p∗ ∈ Y ∗ , so that
h∗∗ (0) = sup {−φ ∗ (0, p∗ )}. (14.72)
p∗ ∈Y ∗
Proof. Observe that

h∗ (p∗ ) = sup{hp, p∗ iY − h(p)} = sup{hp, p∗ iY − inf {φ (u, p)}}, (14.73)
p∈Y p∈Y u∈U
so that
h∗ (p∗ ) = sup {hp, p∗ iY − φ (u, p)} = φ ∗ (0, p∗ ). (14.74)
(u,p)∈U ×Y
Proposition 14.5.3 The set of solutions of the problem P ∗ (the dual problem) is identical to ∂ h∗∗ (0).
Proof 14.10 Consider p∗0 ∈ Y ∗ a solution of Problem P ∗ , that is,

−φ ∗ (0, p∗0 ) ≥ −φ ∗ (0, p∗ ), ∀p∗ ∈ Y ∗ , (14.75)
which is equivalent to
−h∗ (p∗0 ) ≥ −h∗ (p∗ ), ∀p∗ ∈ Y ∗ , (14.76)
which is equivalent to
−h(p∗0 ) = sup {h0, p∗ iY − h∗ (p∗ )} ⇔ −h∗ (p∗0 ) = h∗∗ (0) ⇔ p∗0 ∈ ∂ h∗∗ (0). (14.77)
p∗ ∈Y ∗
Theorem 14.5.4 Consider φ : U ×Y → R̄ convex. Assume infu∈U {φ (u, 0)} ∈ R and there exists u0 ∈ U such
that p 7→ φ (u0 , p) is finite and continuous at 0 ∈ Y , then
inf {φ (u, 0)} = sup {−φ ∗ (0, p∗ )}, (14.78)
u∈U p∗ ∈Y ∗
and the dual problem has at least one solution.
Proof 14.11 By hypothesis h(0) ∈ R and as was shown above, h is convex. As the function p 7→ φ (u0 , p)
is convex and continuous at 0 ∈ Y , there exists a neighborhood V of zero in Y such that
φ (u0 , p) ≤ M < +∞, ∀p ∈ V , (14.79)
for some M ∈ R. Thus, we may write
h(p) = inf {φ (u, p)} ≤ φ (u0 , p) ≤ M, ∀p ∈ V . (14.80)
u∈U
Hence, from Lemma 14.3.7, h is continuous at 0. Thus by Proposition 14.3.8, h is sub-differentiable at 0,

which means h(0) = h∗∗ (0). Therefore by Proposition 14.5.3, the dual problem has solutions and
h(0) = inf {φ (u, 0)} = sup {−φ ∗ (0, p∗ )} = h∗∗ (0). (14.81)
u∈U p∗ ∈Y ∗
Now we apply the last results to φ (u, p) = G(Λu+ p)+F(u), where Λ : U → Y is a continuous linear operator
whose adjoint operator is denoted by Λ∗ : Y ∗ → U ∗ . We may enunciate the following theorem.
Theorem 14.5.5 Suppose U is a reflexive Banach space and define J : U → R by
J(u) = G(Λu) + F(u) = φ (u, 0), (14.82)
where lim J(u) = +∞ as kukU → ∞ and F ∈ Γ0 (U), G ∈ Γ0 (Y ). Also suppose there exists û ∈ U such that
J(û) < +∞ with the function p 7→ G(p) continuous at Λu.
ˆ Under such hypothesis, there exist u0 ∈ U and
p∗0 ∈ Y ∗ such that
J(u0 ) = min{J(u)} = max

∗ ∗
{−G∗ (p∗ ) − F ∗ (−Λ∗ p∗ )} = −G∗ (p∗0 ) − F ∗ (−Λ∗ p∗0 ). (14.83)
u∈U p ∈Y
Proof 14.12 The existence of solutions for the primal problem follows from the direct method of cal-
culus of variations. That is, considering a minimizing sequence, from above (coercivity hypothesis), such
a sequence is bounded and has a weakly convergent subsequence to some u0 ∈ U. Finally, from the lower
semi-continuity of primal formulation, we may conclude that u0 is a minimizer. The other conclusions follow
from Theorem 14.5.4 just observing that
φ ∗ (0, p∗ ) = sup {hp, p∗ iY −G(Λu+ p)−F(u)} = sup {hq, p∗ i−G(q)−hΛu, p∗ i−F(u)}, (14.84)
u∈U,p∈Y u∈U,q∈Y
so that
φ ∗ (0, p∗ ) = G∗ (p∗ ) + sup{−hu, Λ∗ p∗ iU − F(u)} = G∗ (p∗ ) + F ∗ (−Λ∗ p∗ ). (14.85)
u∈U
Thus,
inf {φ (u, 0)} = sup {−φ ∗ (0, p∗ )} (14.86)
u∈U p∗ ∈Y ∗
and solutions u0 and p∗0 for the primal and dual problems, respectively, imply that
J(u0 ) = min{J(u)} = max

∗ ∗
{−G∗ (p∗ ) − F ∗ (−Λ∗ p∗ )} = −G∗ (p∗0 ) − F ∗ (−Λ∗ p∗0 ). (14.87)
u∈U p ∈Y
14.6 The min-max theorem

Our main objective in this section is to state and prove the min-max theorem.
Definition 14.1 Let U,Y be Banach spaces, A ⊂ U and B ⊂ Y and let L : A × B → R be a functional. We
say that (u0 , v0 ) ∈ A × B is a saddle point for L if
L(u0 , v) ≤ L(u0 , v0 ) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B.
Proposition 14.1
Let U,Y be Banach spaces, A ⊂ U and B ⊂ Y . A functional L : U ×Y → R has a saddle point if and only if
max inf L(u, v) = min sup L(u, v).

v∈B u∈A u∈A v∈B
Proof 14.13 Suppose (u0 , v0 ) ∈ A × B is a saddle point of L.

Thus,
L(u0 , v) ≤ L(u0 , v0 ) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B. (14.88)

Define
F(u) = sup L(u, v).
v∈B
Observe that
inf F(u) ≤ F(u0 ),
u∈A
so that
inf sup L(u, v) ≤ sup L(u0 , v). (14.89)
u∈A v∈B v∈B
Define
G(v) = inf L(u, v).
u∈A
Thus
sup G(v) ≥ G(v0 ),
v∈B
so that
sup inf L(u, v) ≥ inf L(u, v0 ). (14.90)

v∈B u∈A u∈A
From (14.88), (14.89) and (14.90) we obtain
inf sup L(u, v) ≤ sup L(u0 , v)

u∈A v∈B v∈B
≤ L(u0 , v0 )
≤ inf L(u, v0 )
u∈A
≤ sup inf L(u, v). (14.91)
v∈B u∈A
Hence
inf sup L(u, v) ≤ L(u0 , v0 )

u∈A v∈B
≤ sup inf L(u, v). (14.92)
v∈B u∈A
On the other hand
inf L(u, v) ≤ L(u, v), ∀u ∈ A, v ∈ B,

u∈A
so that
sup inf L(u, v) ≤ sup L(u, v), ∀u ∈ A,
v∈B u∈A v∈B
and hence
sup inf L(u, v) ≤ inf sup L(u, v). (14.93)
From (14.88), (14.92), (14.93) we obtain
inf sup L(u, v) = sup L(u0 , v)

u∈A v∈B v∈B
= L(u0 , v0 )
= inf L(u, v0 )
u∈A
= sup inf L(u, v). (14.94)
v∈B u∈A
Conversely suppose
max inf L(u, v) = min sup L(u, v).
As above defined,
F(u) = sup L(u, v),

v∈B
and
G(v) = inf L(u, v).
u∈A
From the hypotheses, there exists (u0 , v0 ) ∈ A × B such that
sup G(v) = G(v0 ) = F(u0 ) = inf F(u).

v∈B u∈A
so that
F(u0 ) = sup L(u0 , v) = inf L(u, v0 ) = G(v0 ).
v∈B u∈U
In particular
L(u0 , v0 ) ≤ sup L(u0 , v) = inf L(u, v0 ) ≤ L(u0 , v0 ).
v∈B u∈U
Therefore
sup L(u0 , v) = L(u0 , v0 ) = inf L(u, v0 ).

v∈B u∈U
Proposition 14.2
Let U,Y be Banach spaces, A ⊂ U, B ⊂ Y and let L : A × B → R be a functional. Assume there exist u0 ∈ A,
v0 ∈ B and α ∈ R such that
L(u0 , v) ≤ α, ∀v ∈ B,
and
L(u, v0 ) ≥ α, ∀u ∈ A.
Under such hypotheses (u0 , v0 ) is a saddle point of L, that is,
L(u0 , v) ≤ L(u0 , v0 ) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B.
Proof 14.14 Observe that, from the hypotheses we have
L(u0 , v0 ) ≤ α,
and
L(u0 , v0 ) ≥ α,
so that
L(u0 , v) ≤ α = L(u0 , v0 ) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B.
In the next lines we state and prove the min − max theorem.
Theorem 14.1
Let U,Y be reflexive Banach spaces, A ⊂ U, B ⊂ Y and let L : A × B → R be a functional.
Suppose that
1. A ⊂ U is convex, closed and non-empty.
2. B ⊂ Y is convex, closed and non-empty.
3. For each u ∈ A, Fu (v) = L(u, v) is concave and upper semi-continuous.
4. For each v ∈ B, Gv (u) = L(u, v) is convex and lower semi-continuous.
5. The set A and B are bounded.
Under such hypotheses L has at least one saddle point (u0 , v0 ) ∈ A × B such that
L(u0 , v0 ) = min max L(u, v)

u∈A v∈B
= max min L(u, v). (14.95)
v∈B u∈A
Proof 14.15 Fix v ∈ B. Observe that Gv (u) = L(u, v) is convex and lower semi-continuous, therefore it is
weakly lower semi-continuous on the weakly compact set A. At first we assume the additional hypothesis
that Gv (u) is strictly convex, ∀v ∈ B. Hence Gv (u) attains a unique minimum on A. We denote the optimal
u ∈ A by u(v)
Define
G(v) = min Gv (u) = min L(u, v).
u∈A u∈U
Thus,
G(v) = L(u(v), v).
The function G(v) is expressed as the minimum of a family of concave weakly upper semi-continuous
functions, and hence it is also concave and upper semi-continuous.
Moreover, G(v) is bounded above on the weakly compact set B, so that there exists v0 ∈ B such that
G(v0 ) = max G(v) = max min L(u, v).
v∈B v∈B u∈A
Observe that
G(v0 ) = min L(u, v0 ) ≤ L(u, v0 ), ∀u ∈ U.
u∈A
Observe also that, from the concerned concavity, for u ∈ A, v ∈ B and λ ∈ (0, 1) we have
L(u, (1 − λ )v0 + λ v) ≥ (1 − λ )L(u, v0 ) + λ L(u, v).
In particular denote u((1 − λ )v0 + λ v) = uλ , where uλ is such that
G((1 − λ )v0 + λ v) = min L(u, (1 − λ )v0 + λ v)
u∈A
= L(uλ , (1 − λ )v0 + λ v). (14.96)
Therefore,
G(v0 ) = max G(v)
v∈B
≥ G((1 − λ )v0 + λ v)
= L(uλ , (1 − λ )v0 + λ v)
≥ (1 − λ )L(uλ , v0 ) + λ L(uλ , v)
≥ (1 − λ ) min L(u, v0 ) + λ L(uλ , v)
u∈A
= (1 − λ )G(v0 ) + λ L(uλ , v). (14.97)
G(v0 ) ≥ L(uλ , v). (14.98)
Let {λn } ⊂ (0, 1) be such that λn → 0.
Let {un } ⊂ A be such that
G((1 − λn )v0 + λn v) = min L(u, (1 − λn )v0 + λn v)
u∈A
= L(un , (1 − λn )v0 + λn v). (14.99)
Since A is weakly compact, there exists a subsequence {unk } ⊂ {un } ⊂ A and u0 ∈ A such that
unk * u0 , weakly in U, as k → ∞.
Observe that
(1 − λnk )L(unk , v0 ) + λnk L(unk , v) ≤ L(unk , (1 − λnk )v0 + λnk v)
= min L(u, (1 − λnk )v0 + λnk v)
u∈A
≤ L(u, (1 − λnk )v0 + λnk v), (14.100)
∀u ∈ A, k ∈ N.
Recalling that λnk → 0, from this and (14.100) we obtain
L(u0 , v0 ) ≤ lim inf L(unk , v0 )
k→∞
= lim inf((1 − λnk )L(unk , v0 ) + λnk L(u, v))
k→∞
≤ lim sup L(u, (1 − λnk )v0 + λnk v)
k→∞
≤ L(u, v0 ), ∀u ∈ U. (14.101)
Hence, L(u0 , v0 ) = minu∈A L(u, v0 ).

Observe that from (14.98) we have
G(v0 ) ≥ L(unk , v),
so that
G(v0 ) ≥ lim inf L(unk , v) ≥ L(u0 , v), ∀v ∈ B.
k→∞
Denoting α = G(v0 ) we have
α = G(v0 ) ≥ L(u0 , v), ∀v ∈ B,

and
α = G(v0 ) = min L(u, v0 ) ≤ L(u, v0 ), ∀u ∈ A.

u∈U
From these last two results and Proposition 14.2 we have that (u0 , v0 ) is a saddle point for L. Now assume
that
Gv (u) = L(u, v)
is convex but not strictly convex ∀v ∈ B.
For each n ∈ N Define Ln by
Ln (u, v) = L(u, v) + kukU /n.
In such a case
(Fv )n (u) = Ln (u, v)
is strictly convex for all n ∈ N.
From above we main obtain (un , vn ) ∈ A × B such that
L(un , v) + kun kU /n ≤ L(un , vn ) + kun kU /n

≤ L(u, vn ) + kuk/n. (14.102)
Since A × B is weakly compact and {(un , vn )} ⊂ A × B, up to subsequence not relabeled, there exists
(u0 , v0 ) ∈ A × B such that
un * u0 , weakly in U,
vn * v0 , weakly in Y,
so that,
L(u0 , v) ≤ lim inf(L(un , v) + kun kU /n)

n→∞
≤ lim sup L(u, vn ) + kukU /n
n→∞
≤ L(u, v0 ). (14.103)
Hence,
L(u0 , v) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B,
so that
L(u0 , v) ≤ L(u0 , v0 ) ≤ L(u, v0 ), ∀u ∈ A, v ∈ B.
In the next result we deal with more general situations.

Theorem 14.2
Let U,Y be reflexive Banach spaces, A ⊂ U, B ⊂ Y and let L : A × B → R be a functional.
Suppose that
1. A ⊂ U is convex, closed and non-empty.
2. B ⊂ Y is convex, closed and non-empty.
3. For each u ∈ A, Fu (v) = L(u, v) is concave and upper semi-continuous.
4. For each v ∈ B, Gv (u) = L(u, v) is convex and lower semi-continuous.
5. Either the set A is bounded or there exists ṽ ∈ B such that
L(u, ṽ) → +∞, as kuk → +∞, u ∈ A.
6. Either the set B is bounded or there exists ũ ∈ A such that
L(ũ, v) → −∞, as kvk → +∞, v ∈ B.
Under such hypotheses L has at least one saddle point (u0 , v0 ) ∈ A × B.
Proof 14.16 We prove the result just for the special case such that there exists ṽ ∈ B such that
L(u, ṽ) → +∞, as kuk → +∞, u ∈ A,
and B is bounded. The proofs of remaining cases are similar.

For each n ∈ N denote
An = {u ∈ A : kukU ≤ n}.
Fix n ∈ N. The sets An and B are closed, convex and bounded, so that from the last Theorem 14.1 there
exists a saddle point (un , vn ) ∈ An × B for
L : An × B → R.
Hence,
L(un , v) ≤ L(un , vn ) ≤ L(u, vn ), ∀u ∈ An , v ∈ B.
For a fixed ũ ∈ A1 we have
L(un , ṽ) ≤ L(un , vn )

≤ L(ũ, vn )
≤ sup L(ũ, v) ≡ b ∈ R. (14.104)
v∈B
On the other hand, from the hypotheses,
Gṽ (u) = L(u, ṽ)
is convex, lower semi-continuous and coercive, so that it is bounded below. Thus there exists a ∈ R such that
−∞ < a ≤ Gṽ (u) = L(u, ṽ), ∀u ∈ A.
Hence
a ≤ L(un , ṽ) ≤ L(un , vn ) ≤ b, ∀n ∈ N.
Therefore {L(un , vn )} is bounded.

Moreover, from the coercivity hypotheses and
a ≤ L(un , ṽ) ≤ b, ∀n ∈ N,
we may infer that {un } is bounded.

Summarizing, {un }, {vn }, and {L(un , vn )} are bounded sequences, and thus there exists a subsequence
{nk }, u0 ∈ A, v0 ∈ B and α ∈ R such that
unk * u0 , weakly in U,
vnk * v0 , weakly in Y,
L(unk , vnk ) → α ∈ R,
as k → ∞. Fix (u, v) ∈ A × B. Observe that if nk > n0 = kukU , then
L(unk , v) ≤ L(unk , vnk ) ≤ L(u, vnk ),

so that letting k → ∞, we obtain
L(u0 , v) ≤ lim inf L(unk , v)

k→∞
≤ lim L(unk , vnk ) = α
k→∞
≤ lim sup L(u, vnk )
k→∞
≤ L(u, v0 ), (14.105)
that is,
L(u0 , v) ≤ α ≤ L(u, v0 ), ∀u ∈ A, v ∈ B.
From this and Proposition 14.2 we may conclude that (u0 , v0 ) is a saddle point for L : A × B → R.
14.7 Relaxation for the scalar case

In this section, Ω ⊂ RN denotes a bounded open set with a locally Lipschitz boundary. That is, for each point
x ∈ ∂ Ω there exists a neighborhood Ux whose the intersection with ∂ Ω is the graph of a Lipschitz continuous
function.
We start with the following definition.
Definition 14.7.1 A function u : Ω → R is said to be affine if ∇u is constant on Ω. Furthermore, we say that
u : Ω → R is piecewise affine if it is continuous and there exists a partition of Ω into a set of zero measure
and finite number of open sets on which u is affine.
The proof of next result is found in [33].
Theorem 14.7.2 Let r ∈ N and let uk , 1 ≤ k ≤ r be piecewise affine functions from Ω into R and {αk } such
that αk > 0, ∀k ∈ {1, ..., r} and ∑rk=1 αk = 1. Given ε > 0, there exists a locally Lipschitz function u : Ω → R
and r disjoint open sets Ωk , 1 ≤ k ≤ r, such that
|m(Ωk ) − αk m(Ω)| < αk ε, ∀k ∈ {1, ..., r}, (14.106)

∇u(x) = ∇uk (x), a.e. on Ωk , (14.107)
|∇u(x)| ≤ max {|∇uk (x)|}, a.e. on Ω, (14.108)

1≤k≤r
r
u(x) − ∑ αk uk < ε, ∀x ∈ Ω, (14.109)
k=1
r
u(x) = ∑ αk uk (x), ∀x ∈ ∂ Ω. (14.110)
k=1
The next result is also found in [33].

Proposition 14.7.3 Let r ∈ N and let uk , 1 ≤ k ≤ r be piecewise affine functions from Ω into R. Consider a
Caratheodory
´ function f : Ω × RN → R and a positive function c ∈ L1 (Ω) which satisfy
c(x) ≥ sup{| f (x, ξ )| | |ξ | ≤ max {k∇uk k∞ }}. (14.111)

1≤k≤r
Given ε > 0, there exists a locally Lipschitz function u : Ω → R such that

Z r Z
f (x, ∇u)dx − ∑ αk f (x, ∇uk )dx < ε, (14.112)
Ω k=1 Ω
|∇u(x)| ≤ max {|∇uk (x)|}, a.e. in Ω, (14.113)

1≤k≤r
r
|u(x) − ∑ αk uk (x)| < ε, ∀x ∈ Ω (14.114)
k=1
r
u(x) = ∑ αk uk (x), ∀x ∈ ∂ Ω. (14.115)
k=1
Proof 14.17 It is sufficient to establish the result for functions uk affine over Ω, since Ω can be divided
into pieces on which uk are affine, and such pieces can be put together through (14.115). Let ε > 0 be given.
We know that simple functions are dense in L1 (Ω), concerning the L1 norm. Thus there exists a partition of
Ω into a finite number of open sets Oi , 1 ≤ i ≤ N1 and a negligible set, and there exists f¯k constant functions
over each Oi such that
Z
| f (x, ∇uk (x)) − f¯k (x)|dx < ε, 1 ≤ k ≤ r. (14.116)
Ω
Now choose δ > 0 such that

ε
δ≤ (14.117)
N1 (1 + max1≤k≤r {k f¯k k∞ })
and if B is a measurable set
Z
m(B) < δ ⇒ c(x)dx ≤ ε/N1 . (14.118)
B
Now we apply Theorem 14.7.2, to each of the open sets Oi , therefore there exists a locally Lipschitz function
u : Oi → R and there exist r open disjoints spaces Ωik , 1 ≤ k ≤ r, such that
|m(Ωik ) − αk m(Oi )| ≤ αk δ , for 1 ≤ k ≤ r, (14.119)
∇u = ∇uk , a.e. in Ωki , (14.120)
|∇u(x)| ≤ max {|∇uk (x)|}, a.e. Oi , (14.121)

1≤k≤r
r
u(x) − ∑ αk uk (x) ≤ δ , ∀x ∈ Oi (14.122)
k=1
r
u(x) = ∑ αk uk (x), ∀x ∈ ∂ Oi . (14.123)
k=1
We can define u = ∑rk=1 αk uk on Ω − ∪Ni=1

1
Oi . Therefore u is continuous and locally Lipschitz. Now observe
that
Z r Z Z
f (x, ∇u(x))dx − ∑ f (x, ∇uk (x))dx = f (x, ∇u(x))dx. (14.124)
Oi i
k=1 Ωk Oi −∪rk=1 Ωik
From | f (x, ∇u(x))| ≤ c(x), m(Oi − ∪rk=1 Ωik ) ≤ δ and (14.118) we obtain
Z r Z Z
f (x, ∇u(x))dx − ∑ f (x, ∇uk (x)dx = f (x, ∇u(x))dx ≤ ε/N1 . (14.125)
Oi i
k=1 Ωk Oi −∪rk=1 Ωki
Considering that f¯k is constant in Oi , from (14.117), (14.118) and (14.119) we obtain
r Z Z
∑| f¯k (x)dx − αk f¯k (x)dx| < ε/N1 . (14.126)
k=1 Ωik Oi
We recall that Ωk = ∪Ni=1

1
Ωik so that
Z r Z
f (x, ∇u(x))dx − ∑ αk f (x, ∇uk (x))dx
Ω k=1 Ω
Z r Z
≤ f (x, ∇u(x))dx − ∑ f (x, ∇uk (x))dx
Ω k=1 Ωk
r Z
+∑ | f (x, ∇uk (x) − f¯k (x)|d x
k=1 Ωk
r Z Z
+∑ f¯k (x)dx − αk f¯k (x)dx
k=1 Ωk Ω
r Z
+ ∑ αk | f¯k (x) − f (x, ∇uk (x))|d x. (14.127)
k=1 Ω
From (14.125), (14.116), (14.126) and (14.116) again, we obtain

Z r Z
f (x, ∇u(x))dx − ∑ αk f (x, ∇uk )dx < 4ε. (14.128)
Ω k=1 Ω
The next result we do not prove it. It is a well known result from the finite element theory.
Proposition 14.7.4 If u ∈ W01,p (Ω) there exists a sequence {un } of piecewise affine functions over Ω, null
on ∂ Ω, such that
un → u, in L p (Ω) (14.129)
and
∇un → ∇u, in L p (Ω; RN ). (14.130)
Proposition 14.7.5 For p such that 1 0 such that
a2 (x) + c2 |ξ | p ≤ f (x, ξ ) ≤ a1 (x) + c1 |ξ | p , ∀x ∈ Ω, ξ ∈ RN . (14.131)
Then, given u ∈ W 1,p (Ω) piecewise affine, ε > 0 and a neighborhood V of zero in the topology
σ (L p (Ω, RN ), Lq (Ω, RN )) there exists a function v ∈ W 1,p (Ω) such that
∇v − ∇u ∈ V , (14.132)
u = v on ∂ Ω,
kv − uk∞ < ε, (14.133)
and
Z Z
f (x, ∇v(x))dx − f ∗∗ (x, ∇u(x))dx < ε. (14.134)
Ω Ω
Proof 14.18 Suppose given ε > 0, u ∈ W 1, p (Ω) piecewise affine continuous, and a neighborhood V of
zero, which may be expressed as
Z
V = {w ∈ L p (Ω, RN ) | hm · wdx < η, ∀m ∈ {1, ..., M}}, (14.135)
Ω
where M ∈ N, hm ∈ Lq (Ω, RN ), η ∈ R+ . By hypothesis, there exists a partition of Ω into a negligible set Ω0

and open subspaces ∆i , 1 ≤ i ≤ r, over which ∇u(x) is constant. From standard results of convex analysis in
RN , for each i ∈ {1, ..., r} we can obtain {αk ≥ 0}1≤k≤N+1 , and ξk such that ∑N+1
k=1 αk = 1 and
N+1
∑ αk ξk = ∇u, ∀x ∈ ∆i , (14.136)
k=1
and
N+1
∑ αk f (x, ξk ) = f ∗∗ (x, ∇u(x)). (14.137)
k=1
Define βi = maxk∈{1,...,N+1} {|ξk | on ∆i }, and ρ1 = maxi∈{1,...,r} {βi }, and ρ = max{ρ1 , k∇uk∞ }. Now, ob-
serve that we can obtain functions ĥm ∈ C0∞ (Ω; RN ) such that
η
max kĥm − hm kLq (Ω,RN ) < . (14.138)
m∈{1,...,M } 4ρm(Ω)
Define C = maxm∈{1,...,M} kdiv(hˆ m )kLq (Ω) and we can also define
ε1 = min{ε /4, 1/(m(Ω)1/p ), η /(2Cm(Ω)1/p ), 1/m(Ω)} (14.139)
We recall that ρ does not depend on ε. Furthermore, for each i ∈ {1, ..., r} there exists a compact subset
Ki ⊂ ∆i such that
Z
ε1
[a1 (x) + c1 (x) max {|ξ | p }]dx < . (14.140)
∆i −Ki |ξ |≤ρ r
Also, observe that the sets Ki may be obtained such that the restrictions of f and f ∗∗ to Ki ×ρB are continuous,
so that from this and from the compactness of ρB, for all x ∈ Ki , we can find an open ball ωx with center in x
and contained in Ω, such that
ε1
| f ∗∗ (y, ∇u(x)) − f ∗∗ (x, ∇u(x))| < , ∀y ∈ ωx ∩ Ki , (14.141)
m(Ω)
and
ε1
| f (y, ξ ) − f (x, ξ )| < , ∀y ∈ ωx ∩ Ki , ∀ξ ∈ ρB. (14.142)
m(Ω)
Therefore, from this and (14.137) we may write
N+1
2ε1
f ∗∗ (y, ∇u(x)) − ∑ αk f (y, ξk ) < , ∀y ∈ ωx ∩ Ki . (14.143)
k=1 m(Ω)
We can cover the compact set Ki with a finite number of those open balls ωx , denoted by ω j , 1 ≤ j ≤ l.
j−1
Consider the open sets ω 0j = ω j − ∪i=1 ω̄i , we have that ∪lj=1 ω̄ 0j = ∪lj=1 ω̄ j . Defining functions uk , for 1 ≤
k ≤ N + 1 such that ∇uk = ξk and u = ∑N+1 k=1 αk uk we may apply Proposition 14.7.3 to each of the open sets
ω 0j , so that we obtain functions vi ∈ W 1,p (Ω) such that
Z N+1 Z
ε1
f (x, ∇vi (x)dx − ∑ αk f (x, ξk )dx < , (14.144)
ω 0j k=1 ω 0j rl
|∇vi | < ρ, ∀x ∈ ω 0j , (14.145)
|vi (x) − u(x)| < ε1 , ∀x ∈ ω 0j , (14.146)
and
vi (x) = u(x), ∀x ∈ ∂ ω 0j . (14.147)
Finally we set
vi = u on ∆i − ∪lj=1 ω j . (14.148)
We may define a continuous mapping v : Ω → R by
v(x) = vi (x), if x ∈ ∆i , (14.149)
v(x) = u(x), if x ∈ Ω0 . (14.150)
We have that v(x) = u(x), ∀x ∈ ∂ Ω and k∇vk∞ < ρ. Also, from (14.140)
Z
ε1
| f ∗∗ (x, ∇u(x)|dx < (14.151)
∆i −Ki r
and
Z
ε1
| f (x, ∇v(x)|d x < . (14.152)
∆i −Ki r
On the other hand, from (14.143) and (14.144)
0
ε1 ε1 m(ω j ∩ Ki )
Z Z
f (x, ∇v(x))dx − f ∗∗ (x, ∇u(x))dx ≤ + (14.153)
Ki ∩ω 0j Ki ∩ω 0j rl m(Ω)
so that
ε1 ε1 m(Ki )
Z Z
| f (x, ∇v(x))dx − f ∗∗ (x, ∇u(x))dx| ≤ + . (14.154)
Ki Ki r m(Ω)
Now summing up in i and considering (14.151) and (14.152) we obtain (14.134), that is
Z Z
| f (x, ∇v(x))dx − f ∗∗ (x, ∇u(x))dx| < 4ε1 ≤ ε. (14.155)
Ω Ω
Also, observe that from above, we have
kv − uk∞ < ε1 , (14.156)
and thus
Z Z
ĥm · (∇v(x) − ∇u(x))dx = − div(ĥm )(v(x) − u(x))dx
Ω Ω
≤ kdiv(ĥm )kLq (Ω) kv − ukL p (S)
≤ Cε1 m(Ω)1/p
η
< . (14.157)
2
Also we have that
Z
η
(ĥm − hm ) · (∇v − ∇u)dx ≤ kĥm − hm kLq (Ω,RN ) k∇v − ∇ukL p (Ω,RN ) ≤ . (14.158)
Ω 2
Thus
Z
hm · (∇v − ∇u)d x < η, ∀m ∈ {1, ..., M}. (14.159)
Ω
Theorem 14.7.6 Assuming the hypothesis of last theorem, given a function u ∈ W01,p (Ω), given ε > 0 and a
neighborhood of zero V in σ (L p (Ω, RN ), Lq (Ω, RN )), we have that there exists a function v ∈ W01,p (Ω) such
that
∇v − ∇u ∈ V , (14.160)
and
Z Z
f (x, ∇v(x))dx − f ∗∗ (x, ∇u(x))dx < ε. (14.161)
Ω Ω
Proof 14.19 We can approximate u by a function w which is piecewise affine and null on the boundary.
Thus, there exists δ > 0 such that we can obtain w ∈ W01,p (Ω) piecewise affine such that
ku − wk1,p < δ (14.162)
so that
1
∇w − ∇u ∈ V , (14.163)
2
and
Z Z
∗∗ ε
f (x, ∇w(x))dx − f ∗∗ (x, ∇u(x))dx < . (14.164)
Ω Ω 2
From Proposition 14.7.5 we may obtain v ∈ W01,p (Ω) such that

1
∇v − ∇w ∈ V , (14.165)
2
and
Z Z
ε
f ∗∗ (x, ∇w(x))dx − f (x, ∇v(x))dx < . (14.166)
Ω Ω 2
From (14.164) and (14.166)
Z Z
f ∗∗ (x, ∇u(x))dx − f (x, ∇v(x))dx < ε. (14.167)
Ω Ω
Finally, from (14.163), (14.165) and from the fact the weak neighborhoods are convex, we have
∇v − ∇u ∈ V . (14.168)
To finish this chapter, we present two theorems which summarize the last results.
´ ory function from Ω × RN into R which satisfies
Theorem 14.7.7 Let f be a Caratheod
a2 (x) + c2 |ξ | p ≤ f (x, ξ ) ≤ a1 (x) + c1 |ξ | p (14.169)
where a1 , a2 ∈ L1 (Ω), 1 0. Under such assumptions, defining Û = W01,p (Ω),
we have
Z Z
∗∗
inf f (x, ∇u)dx = min f (x, ∇u)dx (14.170)
u∈Û Ω u∈Û Ω
The solutions of relaxed problem are weak cluster points in W01,p (Ω) of the minimizing sequences of primal
problem.
Proof 14.20 The existence of solutions for the convex relaxed formulation is a consequence of the reflexiv-
ity of U and coercivity hypothesis, which allows an application of the direct method of calculus of variations.
That is, considering a minimizing sequence, from above (coercivity hypothesis), such a sequence is bounded
and has a weakly convergent subsequence to some û ∈ W 1,p (Ω). Finally, from the lower semi-continuity of
relaxed formulation, we may conclude that û is a minimizer. The relation (14.170) follows from last theorem.
´ ory function from Ω × RN into R which satisfies

Theorem 14.7.8 Let f be a Caratheod
a2 (x) + c2 |ξ | p ≤ f (x, ξ ) ≤ a1 (x) + c1 |ξ | p (14.171)
where a1 , a2 ∈ L1 (Ω), 1 0. Let u0 ∈ W 1, p (Ω). Under such assumptions,
defining Û = {u | u − u0 ∈ W01,p (Ω)}, we have
Z Z
inf f (x, ∇u)dx = min f ∗∗ (x, ∇u)dx (14.172)
u∈Û Ω u∈Û Ω
The solutions of relaxed problem are weak cluster points in W 1,p (Ω) of the minimizing sequences of
primal problem.
Proof 14.21 Just apply the last theorem to the integrand g(x, ξ ) = f (x, ξ + ∇u0 ). For details see [33].
14.8 Duality suitable for the vectorial case

14.8.1 The Ekeland variational principle
In this section we present and prove the Ekeland variational principle. This proof may be found in Giusti,
[47], pages 160–161.
Theorem 14.8.1 (Ekeland variational principle) Let (U, d) be a complete metric space and let F : U → R
be a lower semi continuous bounded below functional taking a finite value at some point.
Let ε > 0. Assume for some u ∈ U we have
F(u) ≤ inf {F(u)} + ε.

u∈U
Under such hypotheses, there exists v ∈ U such that

1. d(u, v) ≤ 1,
2. F(v) ≤ F(u),
3. F(v) ≤ F(w) + εd(v, w), ∀w ∈ U.
Proof 14.22 Define the sequence {un } ⊂ U by:
u1 = u,
and having u1 , ..., un , select un+1 as specified in the next lines. First, define
Sn = {w ∈ U | F(w) ≤ F(un ) − εd(un , w)}.
Observe that un ∈ Sn so that Sn in non-empty.

On the other hand, from the definition of infimum, we may select un+1 ∈ Sn such that

1
F(un+1 ) ≤ F(un ) + inf {F(w)} . (14.173)
2 w∈Sn
Since un+1 ∈ Sn we have

εd(un+1 , un ) ≤ F(un ) − F(un+1 ). (14.174)
and hence
m
εd(un+m , un ) ≤ ε ∑ d(un+i , un+m−i ) ≤ F(un ) − F(un+m ). (14.175)
i=1
From (14.174) {F(un )} is decreasing sequence bounded below by infu∈U F(u) so that there exists α ∈ R
such that
F(un ) → α as n → ∞.
From this and (14.175), {un } is a Cauchy sequence , converging to some v ∈ U.
Since F is lower semi-continuous we get,
α = lim inf F(un+m ) ≥ F(v),
m→∞
so that letting m → ∞ in (14.175) we obtain

εd(un , v) ≤ F(un ) − F(v), (14.176)
and, in particular for n = 1 we get
0 ≤ εd(u, v) ≤ F(u) − F(v) ≤ F(u) − inf F(u) ≤ ε.
u∈U
Thus, we have proven 1 and 2.

Suppose, to obtain contradiction, that 3 does not hold.
Hence, there exists w ∈ U such that
F(w) < F(v) − εd(w, v).
In particular we have
6 v.
w= (14.177)
Thus, from this and (14.176) we have
F(w) < F(un ) − ε(un , v) − εd(w, v) ≤ F(un ) − εd(un , w), ∀n ∈ N.
Now observe that w ∈ Sn , ∀n ∈ N so that
inf {F(w)} ≤ F (w), ∀n ∈ N.
w∈Sn
2F(un+1 ) − F(un ) ≤ F(w) < F(v) − εd(v, w),

so that
2 lim inf{F(un+1 )} ≤ F(v) − εd(v, w) + lim inf{F(un )}.
n→∞ n→∞
Hence,
F(v) ≤ lim inf{F(un+1 )} ≤ F(v) − εd(v, w),
n→∞
so that
0 ≤ −εd(v, w),
Thus 3 holds.
Remark 14.8.2 We may introduce in U a new metric given by d1 = ε 1/2 d. We highlight that the topology
remains the same and also F remains lower semi-continuous. Under the hypotheses of the last theorem, if
there exists u ∈ U such that F(u) < infu∈U F(u) + ε, then there exists v ∈ U such that
1. d (u, v) ≤ ε 1/2 ,
2. F(v) ≤ F(u),
3. F (v) ≤ F (w) + ε 1/2 d(u, w), ∀w ∈ U.
Remark 14.8.3 Observe that, if U is a Banach space,
F(v) − F(v + tw) ≤ ε 1/2tkwkU , ∀t ∈ [0, 1], w ∈ U, (14.178)
so that, if F is Gateaux
ˆ differentiable, we obtain
−hδ F(v), wiU ≤ ε 1/2 kwkU . (14.179)
Similarly
F(v) − F(v + t (−w)) ≤ ε 1/2tkwkU ≤, ∀t ∈ [0, 1], w ∈ U, (14.180)
so that, if F is Gateaux
ˆ differentiable, we obtain
hδ F (v), wiU ≤ ε 1/2 kwkU . (14.181)
Thus
kδ F(v)kU ∗ ≤ ε 1/2 . (14.182)
We have thus obtained, from the last theorem and remarks, the following result.
Theorem 14.8.4 Let U be a Banach space. Let F : U → R be a lower semi-continuous Gateaux
ˆ differentiable
functional. Given ε > 0 suppose that u ∈ U is such that
F(u) ≤ inf {F(u)} + ε. (14.183)
u∈U
Then there exists v ∈ U such that

F(v) ≤ F(u), (14.184)
√
ku − vkU ≤ ε, (14.185)
and
√
kδ F(v)kU ∗ ≤ ε. (14.186)
The next theorem easily follows from above results.
Theorem 14.8.5 Let J : U → R, be defined by
J(u) = G(∇u) − h f , uiL2 (S;RN ) , (14.187)
where
U = W01,2 (S; RN ), (14.188)
We suppose G is a l.s.c and Gateaux-dif
ˆ ferentiable so that J is bounded below. Then, given ε > 0, there exists
uε ∈ U such that
J(uε ) < inf {J(u)} + ε, (14.189)
u∈U
and
√
kδ J(uε )kU ∗ < ε. (14.190)
14.9 Some examples of duality theory in convex and non-convex analy-

sis
We start with a well known result of Toland, published in 1979.
Theorem 14.9.1 (Toland, 1979) Let U be a Banach space and let F, G : U → R be functionals such that
inf {G(u) − F(u)} = α ∈ R.

u∈U
Under such hypotheses

F ∗ (u∗ ) − G∗ (u∗ ) ≥ α, ∀u∗ ∈ U ∗ .
Moreover, suppose that u0 ∈ U is such that
G(u0 ) − F(u0 ) = min{G(u) − F(u)} = α.

u∈U
u∗0
Assume also ∈ ∂ F(u0 ).
F ∗ (u∗0 ) − G∗ (u∗0 ) = α,
so that
G(u0 ) − F(u0 ) = min{G(u) − F(u)}

u∈U
= min {F ∗ (u∗ ) − G∗ (u∗ )}
u∗ ∈U ∗
= F ∗ (u∗0 ) − G∗ (u∗0 ). (14.191)
Proof 14.23 Under such hypotheses,
inf {G(u) − F(u)} = α ∈ R.

u∈U
Thus
G(u) − F(u) ≥ α, ∀u ∈ U.
Therefore, for u∗ ∈ U ∗, we have
−hu, u∗ iU + G(u) + hu, u∗ iU − F(u) ≥ α, ∀u ∈ U.

Thus,
−hu, u∗ iU + G(u) + sup{hu, u∗ iU − F(u)} ≥ α, ∀u ∈ U ,
u∈U
that is,
−hu, u∗ iU + G(u) + F ∗ (u∗ ) ≥ α, ∀u ∈ U,
so that
inf {−hu, u∗ iU + G(u)} + F ∗ (u∗ ) ≥ α ,

u∈U
that is,
−G∗ (u∗ ) + F ∗ (u∗ ) ≥ α, ∀u∗ ∈ U ∗ . (14.192)

Also from the hypotheses,
G(u0 ) − F(u0 ) ≤ G(u) − F(u), ∀u ∈ U. (14.193)

On the other hand, from u∗0 ∈ ∂ F(u0 ), we obtain
hu0 , u∗0 iU − F(u0 ) ≥ hu, u∗0 iU − F(u), ∀u ∈ U

so that
−F(u) ≤ hu0 − u, u∗0 iU − F(u0 ), ∀u ∈ U.

From this and (14.193), we get,
G(u0 ) − F(u0 ) ≤ G(u) + hu0 − u, u∗0 iU − F(u0 ), ∀u ∈ U. (14.194)
so that,
hu0 , u∗0 iU − G(u0 ) ≥ hu, u∗0 iU − G(u), ∀u ∈ U,
that is
G∗ (u∗0 ) = sup{hu, u∗0 iU − G(u)}

u∈U
= hu0 , u∗0 iU − G(u0 ). (14.195)
Summarizing, we have got

G∗ (u∗0 ) = hu0 , u∗0 iU − G(u0 ),
and
F ∗ (u∗0 ) = hu0 , u∗0 iU − F(u0 ).
Hence,
F ∗ (u0 ) − G∗ (u∗0 ) = G(u0 ) − F(u0 ) = α.
From this and (14.192), we have
G(u0 ) − F(u0 ) = min{G(u) − F(u)}

u∈U
= min {F ∗ (u∗ ) − G∗ (u∗ )}
u∗ ∈U ∗
= F ∗ (u∗0 ) − G∗ (u∗0 ). (14.196)
Exercise 14.9.2 Let Ω ⊂ R2 be a set of Ĉ1 class. Let V = C1 (Ω) and let J : D ⊂ V → R where
Z Z
γ
J(u) = ∇u · ∇u dx − f u dx, ∀u ∈ U
2 Ω Ω
and where
D = {u ∈ V : u = 0 on ∂ Ω}.
1. Prove that J is convex.

2. Prove that u0 ∈ D such that
γ∇2 u0 + f = 0, in Ω,
minimizes J on D.
3. Prove that
inf J(u) ≥ sup {−G∗ (v∗ ) − F ∗ (−Λ∗ v∗ )},
u∈U v∗ ∈Y ∗
where
1
Z
G(∇u) = ∇u · ∇u dx,
2 Ω
and
G∗ (v∗ ) = sup{hv, v∗ iY − G(v)},
v∈Y
where Y = Y∗ = L2 (Ω).
4. Defining Λ : U → Y by
Λu = ∇u,
and
F :D→R
as Z
F(u) = f u dx,
Ω
so that
F ∗ (−Λ∗ v∗ ) = sup{−h∇u, v∗ iY − F(u)}
u∈D
Z
= sup hu, div v∗ iY + f u dx
u∈D Ω
Z
∗
= sup (div v + f ) u dx
u∈D Ω
if div(v∗ ) + f = 0, in Ω

0,
= (14.197)
+∞, otherwise,
Prove that v∗0 = γ∇u0 is such that
J(u0 ) = min J(u)

u∈D
= min{G(Λu) + F(u)}
u∈D
= max {−G∗ (v∗ ) − F ∗ (−Λ∗ v∗ )}
v∗ ∈Y ∗
= −G (v∗0 ) − F ∗ (−Λ∗ v∗0 ).
∗
(14.198)
Solution: Let u ∈ D and

v ∈ Va = {v ∈ V : v = 0 on ∂ Ω}.
Thus,
δ J(u; v)
J(u + εv) − J(u)
= lim
ε→0
R
ε R R
(γ/2) Ω (∇u + ε∇v) · (∇u + ε∇v) dx − (γ/2) Ω ∇u · ∇u dx − Ω (u + εv − u) f dx
= lim
ε→0 ε
Z Z Z
= lim γ ∇u · ∇v dx − f v dx + ε(γ/2) ∇v · ∇v dx
ε→0 Ω Ω Ω
Z Z
= γ ∇u · ∇v dx − f v dx. (14.199)
Ω Ω
Hence,
Z Z
J(u + v) − J(u) = (γ/2) (∇u + ∇v) · (∇u + ∇v) dx − (γ/2) ∇u · ∇u dx
Ω Ω
Z
− (u + v − u) f dx
ZΩ Z Z
= γ ∇u · ∇v dx − f v dx + (γ/2) ∇v · ∇v dx
ZΩ ZΩ Ω
≥ γ ∇u · ∇v dx − f v dx
Ω Ω
= δ J(u; v) (14.200)
∀u ∈ D, v ∈ Va .
From this we may infer that J is convex.
From the hypotheses u0 ∈ D is such that
γ∇2 u0 + f = 0, in Ω.
Let v ∈ Va .
Therefore, we have
Z Z
δ J(u0 ; v) = γ ∇u0 · ∇v dx − f v dx
ZΩ Ω
Z
= γ ∇u0 · ∇v dx + γ ∇2 u0 v dx
Ω Ω
Z Z Z
= γ ∇u0 · ∇v dx − γ ∇u0 · ∇v dx + ∇u0 · n v ds
Ω Ω ∂Ω
= 0 (14.201)
where n denotes the unit outward field to ∂ Ω.
Summarizing, we got δ J(u0 ; v) = 0, ∀v ∈ Va .
Since J is convex,from this we may conclude that u0 minimizes J on D.
Observe now that,
J(u) = G(∇u) + F(u)

= −h∇u, v∗ iY + G(∇u) + h∇u · v∗ iY + F(u)
≥ inf {−hv, v∗ iY + G(v)}
v∈Y
+ inf {h∇u, v∗ iY + F(u)}
u∈U
= −G∗ (v∗ ) − F ∗ (−Λ∗ v∗ ), ∀u ∈ U, v∗ ∈ Y ∗ . (14.202)
Summarizing,
inf J(u) ≥ sup {−G∗ (v∗ ) − F ∗ (−Λ∗ v∗ )}. (14.203)
u∈D v∗ ∈Y ∗
Also from the hypotheses we have v∗0 = γ∇u0 .
Thus,
∂ G(∇u0 )
v∗0 = ,
∂v
so that
G∗ (v∗0 ) = sup{hv, v∗0 iY − G(v)}
v∈Y
= h∇u0 , v∗0 iY − G(∇u0 )
= −hu0 , div v∗0 iL2 − G(∇u0 ). (14.204)
On the other hand, from de v∗0 = γ∇u0 , we have
div v∗0 = γdiv(∇u0 ) = γ∇2 u0 = − f
G∗ (v∗0 ) = −hu0 , f iL2 − G(∇u0 ),
and from
div v∗0 + f = 0
we get
F ∗ (−Λ∗ v∗0 ) = 0.
Hence
G(∇u0 ) − hu0 , f iL2 = −G∗ (v∗0 ) − F ∗ (−Λ∗ v∗0 ),
so that from this and (14.203) we have
J(u0 ) = min J(u)

u∈D
= min{G(Λu) + F(u)}
u∈D
= max {−G∗ (v∗ ) − F ∗ (−Λ∗ v∗ )}
v∗ ∈Y ∗
= −G (v∗0 ) − F ∗ (−Λ∗ v∗0 ).
∗
(14.205)

Chapter 15
Constrained Variational
Optimization
15.1 Basic concepts

For this chapter the most relevant reference is the excellent book of Luenberger [57], where more details may
be found. We start with the definition of cone:
Definition 15.1.1 (Cone) Given U a Banach space, we say that C ⊂ U is a cone with vertex at origin, if
given u ∈ C, we have that λ u ∈ C, ∀λ ≥ 0. By analogy we define a cone with vertex at p ∈ U as P = p +C,
where C is any cone with vertex at origin.
Definition 15.1.2 Let P be a convex cone in U. For u, v ∈ U we write u ≥ v (with respect to P) if u − v ∈ P.

In particular u ≥ θ if and only if u ∈ P. Also
P+ = {u∗ ∈ U ∗ | hu, u∗ iU ≥ 0, ∀u ∈ P}. (15.1)

If u∗ ∈ P+ we write u∗ ≥ θ ∗.
Proposition 15.1.3 Let U be a Banach space and P be a convex closed cone in U. If u ∈ U satisfies hu, u∗ iU ≥
0, ∀u∗ ≥ θ ∗ , then u ≥ θ .
Proof 15.1 We prove the contrapositive. Assume u 6∈ P. Then by the separating hyperplane theorem there
is an u∗ ∈ U ∗ such that hu, u∗ iU < hp, u∗ iU , ∀p ∈ P. Since P is a cone, given any p ∈ P we must have
h p, u∗ iU ≥ 0, otherwise we would have hu, u∗ i > hα p, u∗ iU for some α > 0. Thus u∗ ∈ P+ . Finally, since
inf p∈P {hp, u∗ iU } = 0, we obtain hu, u∗ iU < 0 which completes the proof.
Definition 15.1.4 (Convex mapping) Let U, Z be vector spaces. Let P ⊂ Z be a convex cone. A mapping
G : U → Z is said to be convex if
G(αu1 + (1 − α)u2 ) ≤ αG(u1 ) + (1 − α)G(u2 ), ∀u1 , u2 ∈ U, α ∈ [0, 1]. (15.2)
Consider the problem P, defined as
Problem P : Minimize F : U → R subject to u ∈ Ω, and G(u) ≤ θ

Constrained Variational Optimization 329
Define
ω(z) = inf{F(u) | u ∈ Ω and G(u) ≤ z}. (15.3)
For such a functional we have the following result.

Proposition 15.1.5 If F is a real convex functional and G is convex, then ω is convex.
ω(αz1 + (1 − α)z2 ) = inf{F(u) | u ∈ Ω

and G(u) ≤ αz1 + (1 − α)z2 }
(15.4)
≤ inf{F(u) | u = αu1 + (1 − α)u2 u1 , u2 ∈ Ω
and G(u1 ) ≤ z1 , G(u2 ) ≤ z2 }
(15.5)
≤α inf{F(u1 ) | u1 ∈ Ω, G(u1 ) ≤ z1 }
+ (1 − α) inf{F(u2 ) | u2 ∈ Ω, G(u2 ) ≤ z2 }
(15.6)
≤αω(z1 ) + (1 − α)ω(z2 ). (15.7)
Now we establish the Lagrange multiplier theorem for convex global optimization.
Theorem 15.1.6 Let U be a vector space, Z a Banach space, Ω a convex subset of U, P a positive convex
closed cone of Z. Assume that P contains an interior point. Let F be a real convex functional on Ω and G a
convex mapping from Ω into Z. Assume the existence of u1 ∈ Ω such that G(u1 ) < θ . Defining
µ0 = inf {F(u) | G(u) ≤ θ }, (15.8)

u∈Ω
then there exists z∗0 ≥ θ , z∗0 ∈ Z ∗ such that
µ0 = inf {F(u) + hG(u), z∗0 iZ }. (15.9)

u∈Ω
Furthermore, if the infimum in (15.8) is attained by u0 ∈ U such that G(u0 ) ≤ θ , it is also attained in (15.9)
by the same u0 and also hG(u0 ), z∗0 iZ = 0. We refer to z∗0 as the Lagrangian Multiplier.
Proof 15.3 Consider the space W = R × Z and the sets A, B where
A = {(r, z) ∈ R × Z | r ≥ F(u), z ≥ G(u) f or some u ∈ Ω}, (15.10)
and
B = {(r, z) ∈ R × Z | r ≤ µ0 , z ≤ θ }, (15.11)
where µ0 = infu∈Ω {F(u) | G(u) ≤ θ }. Since F and G are convex, A and B are convex sets. It is clear that A
contains no interior point of B, and since N = −P contains an interior point , the set B contains an interior
point. Thus, from the separating hyperplane theorem, there is a non-zero element w∗0 = (r0 , z∗0 ) ∈ W ∗ such
that
r0 r1 + hz1 , z∗0 iZ ≥ r0 r2 + hz2 , z∗0 iZ , ∀(r1 , z1 ) ∈ A, (r2 , z2 ) ∈ B. (15.12)

From the nature of B it is clear that w∗0 ≥ θ . That is, r0 ≥ 0 and z∗0 ≥ θ . We will show that r0 > 0. The point
(µ0 , θ ) ∈ B, hence
r0 r + hz, z∗0 iZ ≥ r0 µ0 , ∀(r, z) ∈ A. (15.13)
If r0 = 0 then hG(u1 ), z∗0 iZ ≥ 0 and z∗0 =
6 θ . Since G(u1 ) < θ and z∗0 ≥ θ we have a contradiction. Therefore
r0 > 0 and, without loss of generality we may assume r0 = 1. Since the point (µ0 , θ ) is arbitrarily close to A
and B, we have
µ0 = inf {r + hz, z∗0 iZ } ≤ inf {F(u) + hG(u), z∗0 iZ } ≤ inf{F(u) | u ∈ Ω, G(u) ≤ θ } = µ0 . (15.14)
(r,z)∈A u∈Ω
Also, if there exists u0 such that G(u0 ) ≤ θ , µ0 = F(u0 ), then

µ0 ≤ F(u0 ) + hG(u0 ), z∗0 iZ ≤ F(u0 ) = µ0 . (15.15)
Hence
hG(u0 ), z∗0 iZ = 0. (15.16)
Corollary 15.1.7 Let the hypothesis of the last theorem hold. Suppose
F(u0 ) = inf {F(u) | G(u) ≤ θ }. (15.17)
u∈Ω
Then there exists z∗0 ≥ θ such that the Lagrangian L : U × Z ∗ → R defined by

L(u, z∗ ) = F(u) + hG(u), z∗ iZ (15.18)
has a saddle point at (u0 , z∗0 ). That is
L(u0 , z∗ ) ≤ L(u0 , z∗0 ) ≤ L(u, z∗0 ), ∀u ∈ Ω, z∗ ≥ θ . (15.19)
Proof 15.4 For z∗0 obtained in the last theorem, we have

L(u0 , z∗0 ) ≤ L(u, z∗0 ), ∀u ∈ Ω. (15.20)
As hG(u0 ), z∗0 iZ = 0, we have
L(u0 , z∗ ) − L(u0 , z∗0 ) = hG(u0 ), z∗ iZ − hG(u0 ), z∗0 iZ = hG(u0 ), z∗ iZ ≤ 0. (15.21)
We now prove two theorems relevant to develop the subsequent section.

Theorem 15.1.8 Let F : Ω ⊂ U → R and G : Ω → Z. Let P ⊂ Z be a convex closed cone. Suppose there
exists (u0 , z∗0 ) ∈ U × Z ∗ where z∗0 ≥ θ and u0 ∈ Ω are such that
F(u0 ) + hG(u0 ), z∗0 iZ ≤ F(u) + hG(u), z∗0 iZ , ∀u ∈ Ω. (15.22)
Then
F(u0 ) = inf{F(u) | u ∈ Ω and G(u) ≤ G(u0 )}. (15.23)
Proof 15.5 Suppose there is a u1 ∈ Ω such that F(u1 ) < F(u0 ) and G(u1 ) ≤ G(u0 ). Thus
hG(u1 ), z∗0 iZ ≤ hG(u0 ), z∗0 iZ (15.24)
so that
F(u1 ) + hG(u1 ), z∗0 iZ < F(u0 ) + hG(u0 ), z∗0 iZ , (15.25)
which contradicts the hypothesis of the theorem.
Theorem 15.1.9 Let F be a convex real functional and G : Ω → Z convex and let u0 and u1 be solutions to
the problems P0 and P1 respectively, where
P0 : minimize F(u) subject to u ∈ Ω and G(u) ≤ z0 , (15.26)
and
P1 : minimize F(u) subject to u ∈ Ω and G(u) ≤ z1 . (15.27)
Suppose z∗0 and z∗1 are the Lagrange multipliers related to these problems. Then
hz1 − z0 , z∗1 iZ ≤ F(u0 ) − F(u1 ) ≤ hz1 − z0 , z∗0 iZ . (15.28)
Proof 15.6 For u0 , z∗0 we have
F(u0 ) + hG(u0 ) − z0 , z∗0 iZ ≤ F(u) + hG(u) − z0 , z∗0 iZ , ∀u ∈ Ω, (15.29)
and, particularly for u = u1 and considering that hG(u0 ) − z0 , z∗0 iZ = 0, we obtain
F(u0 ) − F(u1 ) ≤ hG(u1 ) − z0 , z∗0 iZ ≤ hz1 − z0 , z∗0 iZ . (15.30)
A similar argument applied to u1 , z∗1 provides us the other inequality.
15.2 Duality
Consider the basic convex programming problem:
Minimize F(u) subject to G(u) ≤ θ , u ∈ Ω, (15.31)
where F : U → R is a convex functional, G : U → Z is convex mapping, and Ω is a convex set. We define

ϕ : Z ∗ → R by
ϕ(z∗ ) = inf {F(u) + hG(u), z∗ iZ }. (15.32)

u∈Ω
Proposition 15.2.1 ϕ is concave and
ϕ(z∗ ) = inf{ω(z) + hz, z∗ iZ }, (15.33)

z∈Γ
where
ω(z) = inf {F(u) | G(u) ≤ z}, (15.34)

u∈Ω
and
Γ = {z ∈ Z | G(u) ≤ z f or some u ∈ Ω}.
ϕ(z∗ ) = inf {F(u) + hG(u), z∗ iZ }

u∈Ω
≤ inf {F(u) + hz, z∗ iZ | G(u) ≤ z}
u∈Ω
= ω(z) + hz, z∗ iZ , ∀z∗ ≥ θ , z ∈ Γ. (15.35)
On the other hand, for any u1 ∈ Ω, defining z1 = G(u1 ), we obtain
F(u1 ) + hG(u1 ), z∗ iZ ≥ inf {F(u) + hz1 , z∗ iZ | G(u) ≤ z1 } = ω(z1 ) + hz1 , z∗ iZ , (15.36)

u∈Ω
so that
ϕ(z∗ ) ≥ inf{ω(z) + hz, z∗ iZ }. (15.37)

z∈Γ
Theorem 15.2.2 (Lagrange duality) Consider F : Ω ⊂ U → R a convex functional, Ω a convex set, and
G : U → Z a convex mapping. Suppose there exists a u1 such that G(u1 ) < θ and that infu∈Ω {F(u) | G(u) ≤
θ } < ∞, with such order related to a convex closed cone in Z. Under such assumptions, we have
inf {F(u) | G(u) ≤ θ } = max

∗
{ϕ (z∗ )}. (15.38)
u∈Ω z ≥θ
If the infimum on the left side in (15.38) is achieved at some u0 ∈ U and the max on the right side at z∗0 ∈ Z ∗ ,
then
hG(u0 ), z∗0 iZ = 0 (15.39)
and u0 minimizes F(u) + hG(u), z∗0 iZ on Ω.
Proof 15.8 For z∗ ≥ θ we have
inf {F(u) + hG(u), z∗ iZ } ≤ inf {F(u) + hG(u), z∗ iZ } ≤ inf F(u) ≤ µ0 . (15.40)

u∈Ω u∈Ω,G(u)≤θ u∈Ω,G(u)≤θ
or
ϕ(z∗ ) ≤ µ0 . (15.41)
The result follows from Theorem 15.1.6.
15.3 The Lagrange multiplier theorem

Remark 15.3.1 This section was published in similar form by the journal “Computational and Applied
Mathematics, SBMAC-Springer”, reference [42].
In this section we develop a new and simpler proof of the Lagrange multiplier theorem in a Banach space
context. In particular, we address the problem of minimizing a functional F : U → R subject to G(u) = θ ,
where θ denotes the zero vector and G : U → Z is a Frechet ´ differentiable transformation. Here U, Z are
Banach spaces. General results in Banach spaces may be found in [2, 34], for example. For the theorem in
question, among others we would cite [57, 54, 19]. Specially the proof given in [57] is made through the
generalized inverse function theorem. We emphasize such a proof is extensive and requires the continuous
´
Frechet differentiability of F and G. Our approach here is different and the results are obtained through other
hypotheses.
The main result is summarized by the following theorem.
Theorem 15.3.2 Let U and Z be Banach spaces. Assume u0 is a local minimum of F(u) subject to G(u) = θ ,
where F : U → R is a Gateaux
ˆ differentiable functional and G : U → Z is a Frec
´ het differentiable transfor-
mation such that G0 (u0 ) maps U onto Z. Finally, assume there exist α > 0 and K > 0 such that if kϕkU < α
then,
kG0 (u0 + ϕ) − G0 (u0 )k ≤ KkϕkU .
Under such assumptions, there exists z∗0 ∈ Z ∗ such that
F 0 (u0 ) + [G0 (u0 )]∗ (z∗0 ) = θ ,
that is,
hϕ, F 0 (u0 )iU + hG0 (u0 )ϕ, z∗0 iZ = 0, ∀ϕ ∈ U.
Proof 15.9 First observe that there is no loss of generality in assuming 0 < α < 1. Also from the general-
ized mean value inequality and our hypothesis, if kϕkU < α, then
kG(u0 + ϕ) − G(u0 ) − G0 (u0 ) · ϕk

≤ sup kG0 (u0 + hϕ) − G0 (u0 )k kϕ kU

h∈[0,1]
≤ K sup {khϕ kU }kϕkU ≤ KkϕkU2 . (15.42)

h∈[0,1]
For each ϕ ∈ U, define H(ϕ) by
G(u0 + ϕ) = G(u0 ) + G0 (u0 ) · ϕ + H(ϕ),
that is,
H(ϕ) = G(u0 + ϕ) − G(u0 ) − G0 (u0 ) · ϕ.

Let L0 = N(G0 (u0 )) where N(G0 (u0 )) denotes the null space of G0 (u0 ). Observe that U/L0 is a Banach
space for which we define A : U/L0 → Z by
A(ū) = G0 (u0 ) · u,
where ū = {u + v | v ∈ L0 }.
Since G0 (u0 ) is onto, so is A, so that by the inverse mapping theorem A has a continuous inverse A−1 .
Let ϕ ∈ U be such that G0 (u0 ) · ϕ = θ . For a given t such that 0 < |t| < 1+kαϕ kU , let ψ0 ∈ U be such that
H(tϕ)
G0 (u0 ) · ψ0 + = θ,
t2
Observe that, from (15.42),
kH(tϕ)k ≤ Kt 2 kϕkU2 ,
and thus from the boundedness of A−1 , kψ0 k as a function of t may be chosen uniformly bounded relating t
(that is, despite the fact that ψ0 may vary with t, there exists K1 > 0 such that kψ0 kU < K1 , ∀t such that 0 <
α
|t| < 1+kϕ kU ).
Now choose 0 < r < 1/4 and define g0 = θ .
Also define
r
ε= −1
.
4(kA k + 1)(K + 1)(K1 + 1)(kϕkU + 1)
Since from the hypotheses G0 (u) is continuous at u0 , we may choose 0 < δ < α such that if kvkU < δ then
kG0 (u0 + v) − G0 (u0 )k < ε.
Fix t ∈ R such that

δ
0 < |t| < .
2(1 + kϕkU + K1 )
Observe that ψ ∈ U is such that G(u0 + tϕ + t 2 ψ) = θ if and only if
H (tϕ + t 2 ψ)
G0 (u0 ) · ψ + = θ.
t2
Define
H(tϕ + t 2 (ψ0 − g0 ))

L1 = A−1 G0 (u0 ) · (ψ0 − g0 ) + ,
t2
so that
H (tϕ + t 2 (ψ0 − g0 ))

L1 = A−1 [A(ψ0 − g0 )] + A−1
t2
= ψ0 − g0 + w1
= ψ0 + w1
= {ψ0 + w1 + v | v ∈ L0 }.
Here w1 ∈ U is such that

H(tϕ + t 2 (ψ0 − g0 ))

−1
w1 = A ,
t2
that is,
H (tϕ + t 2 (ψ0 − g0 ))
A(w1 ) = ,
t2
so that
H(tϕ + t 2 (ψ0 − g0 ))
G0 (u0 ) · w1 = .
t2
Select g1 ∈ L1 such that
kg1 − g0 kU ≤ 2kL1 − L0 k.
This is possible since
kL1 − L0 k = inf {kg − g0 kU }.
g∈L1
So we have that
H(tϕ) H(tϕ + t 2 (ψ0 − g0 ))

−1
L1 = A − 2 + . (15.43)
t t2
However
H(tϕ + t 2 (ψ0 − g0 )) − H(tϕ)

= G(u0 + tϕ + t 2 (ψ0 )) − G(u0 )
−G0 (u0 ) · (tϕ + t 2 (ψ0 ))
−G(u0 + tϕ) + G(u0 )
+G0 (u0 ) · (tϕ)
= G(u0 + tϕ + t 2 (ψ0 )) − G(u0 + tϕ)
−G0 (u0 ) · (t 2 (ψ0 )), (15.44)
so that by the generalized mean value inequality we may write
kH(tϕ + t 2 (ψ0 − g0 )) − H(tϕ)k

≤ sup kG0 (u0 + tϕ + ht 2 (ψ0 )) − G0 (u0 )kkt 2 ψ0 kU
h∈[0,1]
< εt 2 kψ0 kU . (15.45)

From this and (15.43) we get
kL1 k ≤ kA−1 kkH(tϕ + t 2 (ψ0 − g0 )) − H(tϕ)k/t 2

< kA−1 kεkψ0 kU
r
< kA−1 kK1 −1
4(kA k + 1)(K + 1)(K1 + 1)(kϕ kU + 1)
r
< . (15.46)
4
Hence
kg1 kU < 2kL1 k < r/2.
Now reasoning by induction, for n ≥ 2 assume that kgn−1 kU < r and kgn−2 kU < r and define Ln by
H(tϕ + t 2 (ψ0 − gn−1 ))

−1 0
Ln − Ln−1 = A G (u0 ) · (ψ0 − gn−1 ) + .
t2
Observe that
H (tϕ + t 2 (ψ0 − gn−1 ))

Ln = A−1 G0 (u0 ) · (ψ0 − gn−1 ) + + Ln−1
t2
2

−1 −1 H(t ϕ + t (ψ0 − gn−1 ))
= A A(ψ0 − gn−1 ) + A + gn−1
t2
H (t ϕ + t 2 (ψ0 − gn−1 ))

= ψ0 − gn−1 + A−1 + gn−1
t2
H (tϕ + t 2 (ψ0 − gn−1 ))

= ψ0 + A−1
t2
= {ψ0 + wn + v | v ∈ L0 }.
Here wn ∈ U is such that

H(t ϕ + t 2 (ψ0 − gn−1 ))

wn = A−1 ,
t2
that is,
H (tϕ + t 2 (ψ0 − gn−1 ))

A(wn ) = ,
t2
so that
H (tϕ + t 2 (ψ0 − gn−1 ))

G0 (u0 ) · wn = .
t2
Choose gn ∈ Ln such that
kgn − gn−1 kU ≤ 2kLn − Ln−1 k.
This is possible since
kLn − Ln−1 k = inf {kg − gn−1 kU }.
g∈Ln
Observe that we may write
Ln−1 = A−1 [A(ḡn−1 )] = A−1 [G0 (u0 ) · gn−1 ].
Thus
H (tϕ + t 2 (ψ0 − gn−1 ))

Ln = A−1 G0 (u0 ) · (ψ0 − gn−1 ) + + G0
(u 0 ) · gn−1 .
t2
By analogy
H(tϕ + t 2 (ψ0 − gn−2 ))

−1 0 0
Ln−1 = A G (u0 ) · (ψ0 − gn−2 ) + + G (u0 ) · gn−2 .
t2
Observe that
H(tϕ + t 2 (ψ0 − gn−1 )) − H(tϕ + t 2 (ψ0 − gn−2 ))

= G(u0 + tϕ + t 2 (ψ0 − gn−1 )) − G(u0 )
−G0 (u0 ) · (tϕ + t 2 (ψ0 − gn−1 ))
−G(u0 + tϕ + t 2 (ψ0 − gn−2 )) + G(u0 )
+G0 (u0 ) · (tϕ + t 2 (ψ0 − gn−2 ))
= G(u0 + tϕ + t 2 (ψ0 − gn−1 )) − G(u0 + tϕ + t 2 (ψ0 − gn−2 ))
−G0 (u0 ) · (t 2 (−gn−1 + gn−2 )), (15.47)
so that by the generalized mean value inequality we may write
kH(tϕ + t 2 (ψ0 − gn−1 )) − H(tϕ + t 2 (ψ0 − gn−2 ))k

≤ sup kG0 (u0 + tϕ + t 2 ψ0 − t 2 (h(gn−1 ) + (1 − h)gn−2 )) − G0 (u0 )k
h∈[0,1]
×kt 2 (−gn−1 + gn−2 )kU

< εt 2 kgn−1 − gn−2 kU .
Therefore, similarly as above,
kA−1 k
kLn − Ln−1 k ≤ kH(tϕ + t 2 (ψ0 − gn−1 )) − H (t ϕ + t 2 (ψ0 − gn−2 ))k
t2
< εkA−1 kkgn−1 − gn−2 kU
< (r/4)kgn−1 − gn−2 kU
1
< kgn−1 − gn−2 kU . (15.48)
4
Thus,
1
kgn − gn−1 kU ≤ 2kLn − Ln−1 k < kgn−1 − gn−2 kU .
2
Finally
kgn kU = kgn − gn−1 + gn−1 − gn−2 + gn−2 − ... + g1 − g0 kU

1 1
≤ kg1 kU 1 + + ... + n < 2kg1 kU < r. (15.49)
2 2
Thus kgn kU < r and

1
kgn − gn−1 kU < kgn−1 − gn−2 kU , ∀n ∈ N,
2
so that {gn } is a Cauchy sequence, and since U is a Banach space there exists g ∈ U such that
gn → g, in norm, as n → ∞.
Hence
Ln → L = ḡ, in norm, as n → ∞,
so that,
H (t ϕ + t 2 (ψ0 − g))

−1 0
Ln − Ln−1 → L − L = θ = A G (u0 ) · (ψ0 − g) + .
t2
Since A−1 is a bijection, denoting ψ̃0 = (ψ0 − g), we get
H(tϕ + t 2 (ψ˜0 ))
G0 (u0 ) · ψ˜0 + =θ
t2
Clarifying the dependence on t we denote ψ˜0 = ψ̃0 (t) where as above mentioned, t ∈ R is such that
δ
0 < |t| < .
2(1 + kϕkU + K1 )
Therefore
G(u0 + tϕ + t 2 ψ̃0 (t)) = θ .
Observe also that kt 2 ψ̃0 (t )kU = kt 2 (ψ0 (t) − g)kU ≤ t 2 (K1 + r) ≤ t 2 (K1 + 1) so that t 2 ψ̃0 (t) → θ as t → 0.
Thus, by defining t 2 ψ̃0 (t)|t=0 = θ (observe that in principle such a function would not be defined at t = 0),
we obtain
d (t 2 ψ̃0 (t))
2
t ψ̃0 (t) − θ
|t=0 = lim = θ,
dt t→0 t
considering that
ktψ̃0 (t)kU ≤ |t|(K1 + 1) → 0, as t → 0.
Finally, defining
φ (t ) = F(u0 + tϕ + t 2 ψ̃0 (t)),
from the hypotheses we have that there exists a suitable t˜2 > 0 such that
φ (0) = F(u0 ) ≤ F(u0 + tϕ + t 2 ψ̃0 (t)) = φ (t), ∀|t| < t˜2 ,
also from the hypothesis we get

φ 0 (0) = δ F(u0 , ϕ) = 0,
that is,
hϕ, F 0 (u0 )iU = 0, ∀ϕ such that G0 (u0 ) · ϕ = θ .
In the next lines as usual N[G0 (u0 )] and R[G0 (u0 )] denote the null space and the range of G0 (u0 ), respectively.
Thus F 0 (u0 ) is orthogonal to the null space of G0 (u0 ), which we denote by
F 0 (u0 ) ⊥ N[G0 (u0 )].
Since R[G0 (u0 )] is closed, we get F 0 (u0 ) ∈ R([G0 (u0 )]∗ ), that is, there exists z∗0 ∈ Z ∗ such that
F 0 (u0 ) = [G0 (u0 )]∗ (−z∗0 ).

15.4 Some examples concerning inequality constraints

In this section we assume the hypotheses of last theorem for F and G below specified. As an application of
this same result, consider the problem of locally minimizing F(u) subject to G1 (u) = θ and G2 (u) ≤ θ , where
F : U → R, U being a function Banach space, G1 : U → [L p (Ω)]m1 , G2 : U → [L p (Ω)]m2 where 1 < p < ∞,
and Ω is an appropriate subset of RN . We refer to the simpler case in which the partial order in [L p (Ω)]m2 is
defined by u = {ui } ≥ θ if and only if ui ∈ L p (Ω) and ui (x) ≥ 0 a.e. in Ω, ∀i ∈ {1, ..., m2 }.
Observe that defining
F̃(u, v) = F(u),
G(u, v) = {(G1 )i (u)}m1 ×1 , {(G2 )i (u) + v2i }m2 ×1

it is clear that (locally) minimizing F̃(u, v) subject to G(u, v) = (θ , θ ) is equivalent to the original problem.
We clarify the domain of F˜ is denoted by U ×Y , where
Y = {v measurable such that v2i ∈ L p (Ω), ∀i ∈ {1, ..., m2 }}.
Therefore, if u0 is a local minimum for the original constrained problem, then for an appropriate and
easily defined v0 we have that (u0 , v0 ) is a point of local minimum for the extended constrained one, so that by
the last theorem there exists a Lagrange multiplier z∗0 = (z∗1 , z2∗ ) ∈ [Lq (Ω)]m1 ×[Lq (Ω)]m2 where 1/p+1/q = 1
and
F̃ 0 (u0 , v0 ) + [G0 (u0 , v0 )]∗ (z∗0 ) = (θ , θ ),
that is,
F 0 (u0 ) + [G01 (u0 )]∗ (z∗1 ) + [G02 (u0 )]∗ (z∗2 ) = θ , (15.50)
and
(z∗2 )i v0i = θ , ∀i ∈ {1, ..., m2 }.
In particular for almost all x ∈ Ω, if x is such that v0i (x)2 > 0 then z∗2i (x) = 0, and if v0i (x) = 0 then
(G2 )i (u0 (x)) = 0, so that (z∗2 )i (G2 )i (u0 ) = 0, a.e. in Ω, ∀i ∈ {1, ..., m2 }.
Furthermore, consider the problem of minimizing F1 (v) = F̃(u0 , v) = F(u0 ) subject {G2i (u0 ) + v2i } = θ .
From above such a local minimum is attained at v0 . Thus, from the stationarity of F1 (v) + hz∗2 , {(G2 )i (u0 ) +
vi2 }i[L p (Ω)]m2 at v0 and the standard necessary conditions for the case of convex (in fact quadratic) constraints
we get (z∗2 )i ≥ 0 a.e. in Ω, ∀i ∈ {1, ..., m2 }, that is, z∗2 ≥ θ .
Summarizing, for the order in question the first order necessary optimality conditions are given by
(15.50), z∗2 ≥ θ and (z∗2 )i (G2 )i (u0 ) = θ , ∀i ∈ {1, ..., m2 } (so that hz∗2 , G2 (u0 )i[L p (Ω)]m2 = 0), G1 (u0 ) = θ , and
G2 (u0 ) ≤ θ .
Remark 15.4.1 For the case U = Rn and Rmk replacing [L p (Ω)]mk , for k ∈ {1, 2} the conditions (z∗2 )i vi = θ
means that for the constraints not active (for example vi 6= 0 ) the corresponding coordinate (z∗2 )i of the
Lagrange multiplier is 0. If vi = 0 then (G2 )i (u0 ) = 0, so that in any case (z∗2 )i (G2 )i (u0 ) = 0.
Summarizing, for this last mentioned case we have obtained the standard necessary optimality condi-
tions: (z∗2 )i ≥ 0, and (z∗2 )i (G2 )i (u0 ) = 0, ∀i ∈ {1, ..., m2 }.
15.5 The Lagrange multiplier theorem for equality and inequality con-
straints
In this section we develop more rigorous results concerning the Lagrange multiplier theorem for the case
involving equalities and inequalities.
Theorem 15.1
Let U, Z1 , Z2 be Banach spaces. Consider a cone C in Z2 as above specified and such that if z1 ≤ θ and z2 < θ
then z1 + z2 < θ , where z ≤ θ means that z ∈ −C and z < θ means that z ∈ (−C)◦ . The concerned order is
supposed to be also that if z < θ , z∗ ≥ θ ∗ and z∗ 6= θ then hz, z∗ iZ2 < 0. Furthermore, assume u0 ∈ U is a
point of local minimum for F : U → R subject to G1 (u) = θ and G2 (u) ≤ θ , where G1 : U → Z1 , G2 : U → Z2
and F are Frec ´ het differentiable at u0 ∈ U. Suppose also G01 (u0 ) is onto and that there exist α > 0, K > 0
such that if kϕkU < α then
kG01 (u0 + ϕ) − G01 (u0 )k ≤ KkϕkU .
Finally, suppose there exists ϕ0 ∈ U such that
G01 (u0 ) · ϕ0 = θ
and
G02 (u0 ) · ϕ0 < θ .
Under such hypotheses, there exists a Lagrange multiplier z∗0 = (z∗1 , z∗2 ) ∈ Z1∗ × Z2∗ such that
F 0 (u0 ) + [G01 (u0 )]∗ (z∗1 ) + [G02 (u0 )]∗ (z∗2 ) = θ ,
z∗2 ≥ θ ∗ ,
and
hG2 (u0 ), z∗2 iZ2 = 0.
Proof 15.10 Let ϕ ∈ U be such that

G01 (u0 ) · ϕ = θ
and
G02 (u0 ) · ϕ = v − λ G2 (u0 ),
for some v ≤ θ and λ ≥ 0.
Select α ∈ (0, 1) and define
ϕα = αϕ0 + (1 − α)ϕ.
Observe that G1 (u0 ) = θ and G01 (u0 ) · ϕα = θ so that as in the proof of the Lagrange multiplier theorem
15.3.2 we may find K1 > 0, ε > 0 and ψ0 (t) such that
G1 (u0 + tϕα + t 2 ψ0 (t)) = θ , ∀|t| < ε,
and
kψ0 (t)kU < K1 , ∀|t| < ε.
Observe that
G02 (u0 ) · ϕα
= αG02 (u0 ) · ϕ0 + (1 − α)G02 (u0 ) · ϕ
= αG02 (u0 ) · ϕ0 + (1 − α)(v − λ G2 (u0 ))
= αG02 (u0 ) · ϕ0 + (1 − α)v − (1 − α)λ G2 (u0 ))
= v0 − λ0 G2 (u0 ), (15.51)
where,
λ0 = (1 − α)λ ,
and
v0 = αG02 (u0 ) · ϕ0 + (1 − α)v < θ .
Hence, for t > 0
G2 (u0 + tϕα + t 2 ψ0 (t)) = G2 (u0 ) + G02 (u0 ) · (tϕα + t 2 ψ0 (t)) + r(t),

where
kr(t)k
lim = 0.
t→0+ t
Therefore from (15.51) we obtain
G2 (u0 + tϕα + t 2 ψ0 (t)) = G2 (u0 ) + tv0 − tλ0 G2 (u0 ) + r1 (t),
where
kr1 (t)k
lim = 0.
t→0+ t
Observe that there exists ε1 > 0 such that if 0 < t < ε1 < ε, then
r1 (t)
v0 + < θ,
t
and
G2 (u0 ) − tλ0 G2 (u0 ) = (1 − tλ0 )G2 (u0 ) ≤ θ .
Hence
G2 (u0 + tϕα + t 2 ψ0 (t)) < θ , if 0 < t < ε1 .
From this there exists 0 < ε2 < ε1 such that
F(u0 + tϕα + t 2 ψ0 (t)) − F(u0 )

= htϕα + t 2 ψ0 (t), F 0 (u0 )iU + r2 (t) ≥ 0, (15.52)
where
|r2 (t)|
lim = 0.
t→0+ t
Dividing the last inequality by t > 0 we get
hϕα + tψ0 (t), F 0 (u0 )iU + r2 (t)/t ≥ 0, ∀0 < t < ε2 .
Letting t → 0+ we obtain
hϕα , F 0 (u0 )iU ≥ 0.
Letting α → 0+ , we get
hϕ, F 0 (u0 )iU ≥ 0,
if
G01 (u0 ) · ϕ = θ ,
and
G02 (u0 ) · ϕ = v − λ G2 (u0 ),
for some v ≤ θ and λ ≥ 0. Define
A = {(hϕ, F 0 (u0 )iU + r, G01 (u0 ) · ϕ, G02 (u0 )ϕ − v + λ G(u0 )),

ϕ ∈ U, r ≥ 0, v ≤ θ , λ ≥ 0}. (15.53)
Observe that A is a convex set with a non-empty interior.

If
G01 (u0 ) · ϕ = θ ,
and
G02 (u0 ) · ϕ − v + λ G2 (u0 ) = θ ,
with v ≤ θ and λ ≥ 0 then

hϕ, F 0 (u0 )iU ≥ 0,
so that
hϕ, F 0 (u0 )iU + r ≥ 0.
Moreover, if
hϕ, F 0 (u0 )i + r = 0,
with r ≥ 0,
G01 (u0 ) · ϕ = θ ,
and
G02 (u0 ) · ϕ − v + λ G2 (u0 ) = θ ,
with v ≤ θ and λ ≥ 0, then we have
hϕ, F 0 (u0 )iU ≥ 0,
so that
hϕ, F 0 (u0 )iU = 0,
and r = 0. Hence (0, θ , θ ) is on the boundary of A. Therefore, by the Hahn-Banach theorem, geometric form,
there exists
(β , z∗1 , z∗2 ) ∈ R × Z1∗ × Z2∗
such that
(β , z∗1 , z∗2 ) =
6 (0, θ , θ )
and
β (hϕ, F 0 (u0 )iU + r) + hG01 (u0 ) · ϕ, z∗1 iZ1

+ hG02 (u0 ) · ϕ − v + λ G2 (u0 ), z∗2 iZ2 ≥ 0, (15.54)
∀ ϕ ∈ U, r ≥ 0, v ≤ θ , λ ≥ 0. Suppose β = 0. Fixing all variable except v we get z∗2 ≥ θ . Thus, for ϕ = cϕ0
with arbitrary c ∈ R, v = θ , λ = 0, if z∗2 =
6 θ , then hG02 (u0 ) · ϕ0 , z∗2 iZ2 < 0, so that we get z∗2 = θ . Since G01 (u0 )
is onto, a similar reasoning lead us to z∗1 = θ , which contradicts (β , z∗1 , z∗2 ) = 6 (0, θ , θ ).
Hence, β =6 0, and fixing all variables except r we obtain β > 0. There is no loss of generality in assuming
β = 1.
Again fixing all variables except v, we obtain z∗2 ≥ θ . Fixing all variables except λ , since G2 (u0 ) ≤ θ we
get
hG2 (u0 ), z∗2 iZ2 = 0.
Finally, for r = 0, v = θ , λ = 0, we get
hϕ, F 0 (u0 )iU + hG01 (u0 )ϕ, z∗1 iZ1 + hG02 (u0 ) · ϕ, z∗2 iZ2 ≥ 0, ∀ϕ ∈ U,
that is, since obviously such an inequality is valid also for −ϕ, ∀ϕ ∈ U, we obtain
hϕ, F 0 (u0 )iU + hϕ, [G01 (u0 )]∗ (z∗1 )iU + hϕ, [G02 (u0 )]∗ (z∗2 )iU = 0, ∀ϕ ∈ U,
so that
F 0 (u0 ) + [G01 (u0 )]∗ (z∗1 ) + [G02 (u0 )]∗ (z∗2 ) = θ .
15.6 Second order necessary conditions

In this section we establish second order necessary conditions for a class of constrained problems in Banach
spaces. We highlight the next result is particularly applicable to optimization in Rn .
Theorem 15.2
Let U, Z1 , Z2 be Banach spaces. Consider a cone C in Z2 as above specified and such that if z1 ≤ θ and z2 < θ
then z1 + z2 < θ , where z ≤ θ means that z ∈ −C and z < θ means that z ∈ (−C)◦ . The concerned order is
supposed to be also that if z < θ , z∗ ≥ θ ∗ and z∗ 6= θ then hz, z∗ iZ2 < 0. Furthermore, assume u0 ∈ U is a point
of local minimum for F : U → R subject to G1 (u) = θ and G2 (u0 ) ≤ θ , where G1 : U → Z1 , G2 : U → (Z2 )k
and F are twice Frec ´ het differentiable at u0 ∈ U. Assume G2 (u) = {(G2 )i (u)} where (G2 )i : U → Z2 , ∀i ∈
{1, ..., k} and define
A = {i ∈ {1, ..., k} : (G2 )i (u0 ) = θ },
and also suppose that (G2 )i (u0 ) < θ , if i 6∈ A. Moreover, suppose {G01 (u0 ), {(G2 )0i (u0 )}i∈A } is onto and that
there exist α > 0, K > 0 such that if kϕkU < α then
kG̃0 (u0 + ϕ) − G̃0 (u0 )k ≤ KkϕkU ,
where
G̃(u) = {G1 (u), {(G2 )i (u)}i∈A }.
Finally, suppose there exists ϕ0 ∈ U such that
G01 (u0 ) · ϕ0 = θ
and
G02 (u0 ) · ϕ0 < θ .
Under such hypotheses, there exists a Lagrange multiplier z∗0 = (z∗1 , z∗2 ) ∈ Z1∗ × (Z2∗ )k such that
F 0 (u0 ) + [G01 (u0 )]∗ (z∗1 ) + [G02 (u0 )]∗ (z∗2 ) = θ ,
z∗2 ≥ (θ ∗ , ..., θ ∗ ) ≡ θk∗ ,

and
h(G2 )i (u0 ), (z∗2 )i iZ = 0, ∀i ∈ {1, ..., k},
(z∗2 )i = θ ∗ , if i 6∈ A,
Moreover, defining
L(u, z∗1 , z∗2 ) = F(u) + hG1 (u), z∗1 iZ1 + hG2 (u), z∗2 iZ2 ,
we have that
2
δuu L(u0 , z∗1 , z∗2 ; ϕ) ≥ 0, ∀ϕ ∈ V0 ,
where
V0 = {ϕ ∈ U : G01 (u0 ) · ϕ = θ , (G2 )0i (u0 ) · ϕ = θ , ∀i ∈ A}.
Proof 15.11 Observe that A is defined by
A = {i ∈ {1, ..., k} : (G2 )i (u0 ) = θ }.
Observe also that (G2 )i (u0 ) < θ , if i 6∈ A.

Hence the point u0 ∈ U is a local minimum for F(u) under the constraints
G1 (u) = θ , and (G2 )i (u) ≤ θ , ∀i ∈ A.
From the last Theorem 15.1 for such an optimization problem there exists a Lagrange multiplier
(z∗1 , {(z∗2 )i∈A }) such that (z∗2 )i ≥ θ ∗ , ∀i ∈ A, and
F 0 (u0 ) + [G01 (u0 )]∗ (z∗1 ) + ∑ [(G2 )0i (u0 )]∗ ((z∗2 )i ) = θ . (15.55)
i∈A
The choice (z∗2 )i = θ , if i 6∈ A leads to the existence of a Lagrange multiplier (z∗1 , z∗2 ) =
(z∗1 , {(z∗2 )i∈A , (z∗2 )i6∈A }) such that
z∗2 ≥ θk∗
and
h(G2 )i (u0 ), (z∗2 )i iZ = 0, ∀i ∈ {1, ..., k}.
Let ϕ ∈ V0 , that is, ϕ ∈ U,
G01 (u0 )ϕ = θ
and
(G2 )0i (u0 ) · ϕ = θ , ∀i ∈ A.
Recall that G̃(u) = {G1 (u), (G2 )i∈A (u)} and therefore, similarly as in the proof of the Lagrange multiplier
theorem 15.3.2, we may obtain ψ0 (t), K > 0 and ε > 0 such that
G̃(u0 + tϕ + t 2 ψ0 (t)) = θ , ∀|t| < ε,

and
kψ0 (t)k ≤ K, ∀|t| < ε.
Also, if i 6∈ A, we have that (G2 )i (u0 ) < θ , so that
(G2 )i (u0 + tϕ + t 2 ψ0 (t)) = (G2 )i (u0 ) + G0i (u0 ) · (tϕ + t 2 ψ0 (t)) + r(t),
where
kr(t)k
lim = 0,
t →0 t
that is,
(G2 )i (u0 + tϕ + t 2 ψ0 (t)) = (G2 )i (u0 ) + t(G2 )0i (u0 ) · ϕ + r1 (t),
where,
kr1 (t )k
lim = 0,
t→0 t
and hence there exists, 0 < ε1 < ε, such that
(G2 )i (u0 + tϕ + t 2 ψ0 (t)) < θ , ∀|t| < ε1 < ε.
Therefore, since u0 is a point of local minimum under the constraint G(u) ≤ θ , there exists 0 < ε2 < ε1 ,
such that
F(u0 + tϕ + t 2 ψ0 (t)) − F(u0 ) ≥ 0, ∀|t| < ε2 ,
so that,
F(u0 + tϕ + t 2 ψ0 (t)) − F(u0 )

= F(u0 + tϕ + t 2 ψ0 (t)) − F(u0 )
+hG1 (u0 + tϕ + t 2 ψ0 (t)), z∗1 iZ1 + ∑ h(G2 )i (u0 + tϕ + t 2 ψ0 (t)), (z∗2 )i iZ2

i∈A
−hG1 (u0 ), z∗1 iZ1 − ∑ {h(G2 )i (u0 ), (z∗2 )i iZ2 }
i∈A
2
= F(u0 + tϕ + t ψ0 (t)) − F(u0 )
+hG1 (u0 + tϕ + t 2 ψ0 (t)), z∗1 iZ1 − hG1 (u0 ), z∗1 iZ1
+hG2 (u0 + tϕ + t 2 ψ0 (t)), z∗2 iZ2 − hG2 (u0 ), z∗2 iZ2
= L(u0 + tϕ + t 2 ψ0 (t)), z∗1 , z∗2 ) − L(u0 , z∗1 , z∗2 )
1 2
= δu L(u0 , z∗1 , z∗2 ;tϕ + t 2 ψ0 (t)) + δuu L(u0 , z∗1 , z∗2 ;tϕ + t 2 ψ0 (t)) + r2 (t)
2
t2 2
= δ L(u0 , z∗1 , z∗2 ; ϕ + tψ0 (t)) + r2 (t) ≥ 0, ∀|t | < ε2 .
2 uu
where
lim |r2 (t)|/t 2 = 0.
t→0
To obtain the last inequality we have used
δu L(u0 , z∗1 , z∗2 ;tϕ + t 2 ψ0 (t)) = 0
Dividing the last inequality by t 2 > 0 we obtain

1 2
δ L(u0 , z∗1 , z∗2 ; ϕ + tψ0 (t)) + r2 (t)/t 2 ≥ 0, ∀0 < |t| < ε2 ,
2 uu
and finally, letting t → 0 we get
1 2
δ L(u0 , z∗1 , z∗2 ; ϕ) ≥ 0.
2 uu
15.7 On the Banach fixed point theorem

Now we recall a classical definition namely the Banach fixed theorem also known as the contraction mapping
theorem
Definition 15.7.1 Let C be a subset of a Banach space U and let T : C → C be an operator. Thus T is said
to be a contraction mapping if there exists 0 ≤ α < 1 such that
kT (u) − T (v)kU ≤ αku − vkU , ∀u, v ∈ C.
Remark 15.7.2 Observe that if kT 0 (u)kU ≤ α < 1, on a convex set C then T is a contraction mapping, since
by the mean value inequality,
kT (u) − T (v)kU ≤ sup{kT 0 (u)k}ku − vkU , ∀u, v ∈ C.

u∈C
The next result is the base of our generalized method of lines.

Theorem 15.7.3 (Contraction mapping theorem) Let C be a closed subset of a Banach space U. Assume
T is contraction mapping on C, then there exists a unique u0 ∈ C such that u0 = T (u0 ). Moreover, for an
arbitrary u1 ∈ C defining the sequence
u2 = T (u1 ) and un+1 = T (un ), ∀n ∈ N
we have
un → u0 , in norm, as n → +∞.
Proof 15.12 Let u1 ∈ C. Let {un } ⊂ C be defined by
un+1 = T (un ), ∀n ∈ N.
Hence, reasoning inductively
kun+1 − un kU = kT (un ) − T (un−1 )kU

≤ αkun − un−1 kU
≤ α 2 kun−1 − un−2 kU
≤ ......
≤ α n−1 ku2 − u1 kU , ∀n ∈ N. (15.56)
Thus, for p ∈ N we have
kun+p − un kU
= kun+p − un+p−1 + un+p−1 − un+p−2 + ... − un+1 + un+1 − un kU
≤ kun+p − un+p−1 kU + kun+p−1 − un+p−2 kU + ... + kun+1 − un kU
≤ (α n+ p−2 + α n+p−3 + ... + α n−1 )ku2 − u1 kU
≤ α n−1 (α p−1 + α p−2 + ... + α 0 )ku2 − u1 kU
!
∞
≤ α n−1 ∑ αk ku2 − u1 kU
k=0
α n−1
≤ ku2 − u1 kU (15.57)
1−α
Denoting n + p = m, we obtain
α n−1
kum − un kU ≤ ku2 − u1 kU , ∀m > n ∈ N.
1−α
Let ε > 0. Since 0 ≤ α < 1 there exists n0 ∈ N such that if n > n0 then
α n−1
0≤ ku2 − u1 kU < ε,
1−α
so that
kum − un kU < ε, if m > n > n0 .
From this we may infer that {un } is a Cauchy sequence, and since U is a Banach space, there exists
u0 ∈ U such that
un → u0 , in norm, as n → ∞.
Observe that
ku0 − T (u0 )kU = ku0 − un + un − T (u0 )kU

≤ ku0 − un kU + kun − T (u0 )kU
≤ ku0 − un kU + αkun−1 − u0 kU
→ 0, as n → ∞. (15.58)
Thus ku0 − T (u0 )kU = 0.

Finally, we prove the uniqueness. Suppose u0 , v0 ∈ C are such that
u0 = T (u0 ) and v0 = T (v0 ).
Hence,
ku0 − v0 kU = kT (u0 ) − T (v0 )kU

≤ αku0 − v0 kU . (15.59)
From this we get

ku0 − v0 ||U ≤ 0,
that is
ku0 − v0 kU = 0.
15.8 Sensitivity analysis

15.8.1 Introduction
In this section we state and prove the implicit function theorem for Banach spaces. A similar result may be
found in Ito and Kunisch [54], page 31.
We emphasize the result found in [54] is more general however, the proof present here is almost the same
for a simpler situation. The general result found in [54] is originally from Robinson [63].
15.9 The implicit function theorem

Theorem 15.9.1 Let V,U,W be Banach spaces. Let F : V ×U → W be a functions such that
F(x0 , u0 ) = 0,
where (x0 , u0 ) ∈ V ×U.

Assume there exists r > 0 such that F is Frec ´ het differentiable and Fx (x, u) is continuous in (x, u) in
Br (x0 , u0 ).
Suppose also [Fx (x0 , u0 )]−1 exists and is bounded so that there exists ρ > 0 such that
0 < k[Fx (x0 , u0 )]−1 k ≤ ρ.
Under such hypotheses, there exist 0 < ε1 < r/2 and 0 < ε2 < 1 such that for each u ∈ Bε1 (u0 ), there
exists x ∈ Bε2 (x0 ) such that
F(x, u) = 0,
where we denote x = x(u) so that
F(x(u), u) = 0.
Moreover, there exists δ > 0 such that 0 < δ ρ < 1, such that for each u, v ∈ Bε1 (u0 ) we have
ρ 2δ
kx(u) − x(v)k ≤ kF(x(v), u) − F(x(v), v)k.
1 − ρδ
Finally, if there exists K > 0 such that
kFu (x, u)k ≤ K, ∀(x, u) ∈ Bε2 (x0 ) × Bε1 (u0 )
so that
kF(x, u) − F(x, v)k ≤ Kku − vk, ∀(x, u) ∈ Bε2 (x0 ) × Bε1 (u0 ),
then
kx(u) − x(v)k ≤ K1 ku − vk,
where
ρ 2δ
K1 = K .
1−δρ
Proof 15.13 Let 0 < ε < r/2. Choose δ > 0 such that
0 < ρδ < 1.
Define
T (x) = F(x0 , u0 ) + Fx (x0 , u0 )(x − x0 ) = Fx (x0 , u0 )(x − x0 ),
and
h(x, u) = F(x0 , u0 ) + Fx (x0 , u0 )(x − x0 ) − F(x, u).
Choose 0 < ε3 < r/2 and 0 < ε2 < 1 such that
Bε2 (u0 ) × Bε3 (x0 ) ⊂ Br (x0 , u0 )
and if (x, u) ∈ Bε2 (x0 ) × Bε3 (u0 ) then
δ
kFx (x, u) − Fx (x0 , u0 )k < .
2
Select 0 < ε1 < ε3 such that if u ∈ Bε1 (u0 ), then
ρkF(x0 , u) − F(x0 , u0 )k < (1 − ρδ )ε2 .
For each u ∈ Bε1 (u0 ) define

φu (x) = T −1 (h(x, u)).
Fix u ∈ Bε1 (u0 ).
Observe that for x1 , x2 ∈ Bε2 (x0 ) we have
kφu (x1 ) − φu (x2 )k ≤ k[Fx (x0 , u0 )]−1 kkh(x1 , u) − h(x2 , u)k

≤ ρkh(x1 , u) − h(x2 , u)k
≤ ρ sup khx (tx1 + (1 − t)x2 , u)kkx1 − x2 k
t∈[0,1]
δ
≤ ρ2 kx1 − x2 k
2
= ρδ kx1 − x2 k. (15.60)
Observe that since 0 < ρδ < 1 we have that φu is a candidate to be a contraction mapping.
Observe also that
x0 = T −1 (0),
so that
T (x0 ) = 0 = h(x0 , u0 )
and
x0 = T −1 (h(x0 , u0 )).
Thus,
kφu (x0 ) − x0 k ≤ ρkh(x0 , u) − h(x0 , u0 )k

= ρkh(x0 , u)k
≤ ρkF(x0 , u) − F(x0 , u0 )k
≤ (1 − ρδ )ε2 . (15.61)
On the other hand, for each x ∈ Bε2 (x0 ), we have that
kφu (x) − x0 k = kφu (x) − φu (x0 ) + φu (x0 ) − x0 k

≤ kφu (x) − φu (x0 )k + kφu (x0 ) − x0 k
≤ ρδ kx − x0 k + (1 − ρδ )ε2
< ρδ ε2 + (1 − ρδ )ε2
= ε2 . (15.62)

φu (x) ∈ Bε2 (x0 ), ∀x ∈ Bε2 (x0 )
so that indeed φu is a contraction mapping.
Therefore, from the Banach fixed point theorem, there exists a unique fixed point x = x(u) for φu (x), so
that
x(u) = φu (x(u))
= T −1 (h(x(u), u))
= T −1 (Fx (x0 , u0 )(x(u) − x0 ) − F(x(u, u)))
= [Fx (x0 , u0 )]−1 [Fx (x0 , u0 )(x(u) − x0 ) − F(x(u), u)] + x0
= x(u) − x0 + x0 − [Fx (x0 , u0 )]−1 (F(x(u), u))
= x(u) − [Fx (x0 , u0 )]−1 (F(x(u), u)). (15.63)
From this, we have

[Fx (x0 , u0 )]−1 (F(x(u), u)) = 0,
so that
F(x(u), u) = Fx (x0 , u0 )0 = 0,
that is,
F(x(u), u) = 0.
Let x ∈ Bε2 (x0 ) and u ∈ Bε1 (u0 ).
Thus
kφu (x) − xk ≤ kφu (x)k + kxk < 2ε2 .
Moreover,
kφu (x1 ) − φu (x2 )k < ρδ kx2 − x1 k, ∀x1 , x2 ∈ Bε2 (x0 ).
From these last two results we obtain
kφu2 (x) − φu (x)k < 2ε2 (ρδ ),
and reasoning inductively,

kφun+1 (x) − φun (x)k < 2ε2 (ρδ )n , ∀n ∈ N.
Therefore
kφun+1 (x) − xk = kφun+1 (x) − φun (x) + φun (x) − · · · + φu (x) − xk

≤ kφun+1 (x) − φun (x)k + kφun (x) − φun−1 (x)k + · · · + kφu (x) − xk
n+1
≤ ∑ (ρδ ) j kφu (x) − xk
j=1
ρδ
≤ kφu (x) − xk, ∀n ∈ N (15.64)
1 − ρδ
Letting n → ∞, we obtain
ρδ
kx(u) − xk ≤ kφu (x) − xk.
1 − ρδ
In particular for v ∈ Bε1 (u0 ) and x = x(v), we get
ρδ
kx(u) − x(v)k ≤ kφu (x(v)) − x(v)k
1 − ρδ
ρδ
= kφu (x(v)) − φv (x(v))k
1 − ρδ
ρ 2δ
≤ kh(x(v), u) − h(x(v), v)k
1 − ρδ
ρ 2δ
≤ kF(x(v), u) − F(x(v), v)k
1 − ρδ
ρ 2δ
≤ K ku − vk
1 − ρδ
= K1 ku − vk, ∀u, v ∈ Bε1 (u0 ). (15.65)
Corollary 15.9.2 Consider the hypotheses and statements of the last theorem. Moreover, assume Fx : V ×
U → W is such that [Fx (x, u)]−1 exists and it is bounded in Br (x0 , u0 ).
´ het differentiable in Br (x0 , u0 ).
Suppose also, F is Frec
Let ϕ ∈ U.
x0 (u, ϕ) = −[Fx (x(u), u)]−1 [Fu (x(u), u)](ϕ),
where
x(u + tϕ) − x(u)
x0 (u, ϕ) = lim .
t →0 t

F(x(u), u) = 0, in Bε1 (u0 ).
Let u ∈ Bε1 (u0 ).

Let t0 > 0 be such that
u + tϕ ∈ Bε1 (u0 ), ∀|t| < t0 .
Observe that
F(x(u + tϕ), u + tϕ) − F(x(u), u) = 0, ∀|t| < t0 .
´
From the Frechet differentiability of F at (x(u), u), for 0 < |t| < t0 , we obtain
Fx (x(u), u) · (x(u + tϕ) − x(u)) + Fu (x(u), u)(tϕ)

+W (u, ϕ,t)(kx(u + tϕ) − x(u)k + |t|kϕk) = 0, (15.66)
where W is such that

W (u, ϕ,t) → 0, as t → 0,
since
x(u + tϕ) − x(u) → 0
and
tϕ → 0, as t → 0.
From this we obtain
x(u + tϕ) − x(u)
= −[Fx (x(u), u)]−1 [[Fu (x(u), u)](ϕ) + r(u, ϕ,t)]
t
→ −[Fx (x(u), u)]−1 [Fu (x(u), u)](ϕ), as t → 0, (15.67)
since
kx(u + tϕ) − x(u)k
kr(u, ϕ,t)k ≤ kW (u, ϕ,t)k + kϕk
t
≤ kW (u, ϕ,t)k(K1 kϕ k + kϕ k)
→ 0, as t → 0. (15.68)
Summarizing,
x(u + tϕ) − x(u)

x0 (u, ϕ) = lim
t→0 t
= −[Fx (x(u), u)]−1 [Fu (x(u), u)](ϕ). (15.69)
15.9.1 ˆ
The main results about Gateaux differentiability
Again let V,U be Banach spaces and let F : V ×U → R be a functional. Fix u ∈ U and consider the problem
of minimizing F(x, u) subject to G(x, u) ≤ θ and H(x, u) = θ . Here the order and remaining details on the
primal formulation are the same as those indicated in section 15.4.
Hence, for the specific case in which
G : V ×U → [L p (Ω)]m1
and
H : V ×U → [L p (Ω)]m2 ,
(the cases in which the co-domains of G and H are Rm1 and Rm2 respectively are dealt similarly) we redefine
the concerned optimization problem, again for a fixed u ∈ U, by minimizing F(x, u) subject to
{Gi (x, u) + v2i } = θ ,
and
H(x, u) = θ .
At this point we assume F (x, u), G̃(x, u, v) = {Gi (x, u) + v2i } ≡ G(u) + v2 (from now on we use this
general notation) and H(x, u) satisfy the hypotheses of the Lagrange multiplier theorem 15.3.2.
Hence, for the fixed u ∈ U we assume there exists an optimal x ∈ V which locally minimizes F(x, u)
under the mentioned constraints.
From Theorem 15.3.2 there exist Lagrange multipliers λ1 , λ2 such that denoting [L p (Ω)]m1 and [L p (Ω)]m2
simply by L p , and defining
F̃(x, u, λ1 , λ2 , v) = F(x, u) + hλ1 , G(u) + v2 iL p + hλ2 , H(x, u)iL p ,
the following necessary conditions hold,
F̃x (x, u) = Fx (x, u) + λ1 · Gx (x, u) + λ2 · Hx (x, u) = θ , (15.70)
G(x, u) + v2 = θ , (15.71)
λ1 · v = θ , (15.72)
λ1 ≥ θ , (15.73)
H(x, u) = θ . (15.74)
Clarifying the dependence on u, we denote the solution x, λ1 , λ2 , v by x(u), λ1 (u), λ2 (u), v(u), respec-
tively. In particular, we assume that for a u0 ∈ U, x(u0 ), λ1 (u0 ), λ2 (u0 ), v(u0 ) satisfy the hypotheses of
the implicit function theorem. Thus, for any u in an appropriate neighborhood of u0 , the corresponding
x(u), λ1 (u), λ2 (u), v(u) are uniquely defined.
We emphasize that from now on the main focus of our analysis is to evaluate variations of the optimal
x(u), λ1 (u), λ2 (u), v(u) with variations of u in a neighborhood of u0 .
For such an analysis, the main tool is the implicit function theorem and its main hypothesis is satisfied
´
through the invertibility of the matrix of Frechet second derivatives.
Hence, denoting, x0 = x(u0 ), (λ1 )0 = λ1 (u0 ), (λ2 )0 = λ2 (u0 ), v0 = v(u0 ), and
A1 = Fx (x0 , u0 ) + (λ1 )0 · Gx (x0 , u0 ) + (λ2 )0 · Hx (x0 , u0 ),
A2 = G(x0 , u0 ) + v20
A3 = H(x0 , u0 ),
A4 = (λ1 )0 · v0 ,
we reiterate to assume that
A1 = θ , A2 = θ , A3 = θ , A4 = θ ,
and M −1 to represent a bounded linear operator, where
 
(A1 )x (A1 )λ1 (A1 )λ2 (A1 )v
 (A2 )x (A2 )λ (A2 )λ2 (A2 )v 
M=  1  (15.75)
(A3 )x (A3 )λ1 (A3 )λ2 (A3 )v 
(A4 )x (A4 )λ1 (A4 )λ2 (A4 )v
where the derivatives are evaluated at (x0 , u0 , (λ1 )0 , (λ2 )0 , v0 ), so that,

 
A Gx (x0 , u0 ) Hx (x0 , u0 ) θ
 Gx (x0 , u0 ) θ θ 2v0 
M=  Hx (x0 , u0 ) θ
 (15.76)
θ θ 
θ v0 θ (λ1 )0
where
A = Fxx (x0 , u0 ) + (λ1 )0 · Gxx (x0 , u0 ) + (λ2 )0 · Hxx (x0 , u0 ).
Moreover, also from the implicit function theorem,
k(x(u), λ1 (u), λ2 (u), v(u)) − (x(u0 ), λ1 (u0 ), λ2 (u0 ), v(u0 ))k ≤ Kku − u0 k, (15.77)
for some appropriate K > 0, ∀u ∈ Br (u0 ), for some r > 0.
We highlight to have denoted λ (u) = (λ1 (u), λ2 (u)).
Let ϕ ∈ [C∞ (Ω)]k ∩U, where k depends on the vectorial expression of U.
ˆ
At this point we will be concerned with the following Gateaux variation evaluation
δu F̃(x(u0 ), u0 , λ (u0 ), v(u0 ); ϕ).
Observe that
δu F̃(x(u0 ), u0 , λ (u0 ), v(u0 ); ϕ) =

F̃(x(u0 + εϕ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
lim
ε→0 ε

F̃(x(u0 ), u0 , λ (u0 ), v(u0 ))
− ,
ε
so that
δu F̃ (x(u0 ), u0 , λ (u0 ), v(u0 ); ϕ) =

lim
ε→0 ε
F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
−
ε
F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
+
ε

F̃(x(u0 ), u0 , λ (u0 ), v(u0 ))
− .
ε
However,

ε
F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
−
ε
≤ F̃x (x(u0 + εϕ), u0 + εϕ, λ (u0 + εϕ), v(u0 + ε ϕ)) Kkϕk
+K1 kx(u0 + εϕ) − x(u0 )k
≤ K1 Kkϕkε
→ 0, as ε → 0.
In these last inequalities we have used
x(u0 + εϕ) − x(u0 )

lim sup ≤ Kkϕk,
ε→0 ε
and
F̃x (x(u0 + εϕ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ)) = θ .
On the other hand,

F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
ε

F̃(x(u0 ), u0 , λ (u0 ), v(u0 ))
−
ε

F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
=
ε
F̃(x(u0 ), u0 + εϕ, λ (u0 ), v(u0 ))
−
ε
F̃(x(u0 ), u0 + εϕ, λ (u0 ), v(u0 ))

+
ε

F̃(x(u0 ), u0 , λ (u0 ), v(u0 ))
−
ε
Now observe that

F̃(x(u0 ), u0 + εϕ, λ (u0 + εϕ), v(u0 + εϕ))
ε
F̃(x(u0 ), u0 + εϕ, λ (u0 ), v(u0 ))
−
ε
hλ1 (u0 + εϕ), G(x(u0 ), u0 + εϕ) + v(u0 + εϕ)2 iL p
=
ε
hλ1 (u0 ), G(x(u0 ), u0 + εϕ) + v(u0 )2 iL p
−
ε
hλ2 (u0 + εϕ) − λ2 (u0 ), H(x(u0 ), u0 + εϕ)iL p
+ . (15.78)
ε
Also,

ε
hλ1 (u0 ), G(x(u0 ), u0 + εϕ) + v(u0 )2 iL p
−
ε
≤
ε
hλ1 (u0 ), G(x(u0 ), u0 + εϕ) + v(u0 + εϕ)2 iL p
−
ε
hλ1 (u0 ), G(x(u0 ), u0 + εϕ) + v(u0 + εϕ)2 iL p
+
ε
hλ1 (u0 ), G(x(u0 ), u0 + εϕ) + v(u0 )2 iL p
−
ε
K kϕ k
≤ ε kG(x(u0 ), u0 + εϕ) + v(u0 + εϕ)2 k
ε
Kkϕkε
+ kλ1 (u0 )(v(u0 + εϕ) + v(u0 ))k
ε
→ 0 as ε → 0.
To obtain the last inequalities we have used
λ1 (u0 + εϕ) − λ1 (u0 )

lim sup ≤ Kkϕk,
ε→0 ε
λ1 (u0 )v(u0 ) = θ ,
λ1 (u0 )v(u0 + εϕ) → θ , as ε → 0,
and
λ1 (u0 )(v(u0 + εϕ)2 − v(u0 )2 )
ε
λ1 (u0 )(v(u0 + εϕ) + v(u0 ))(v(u0 + εϕ) − v(u0 ))
=
ε
kλ1 (u0 )(v(u0 + εϕ) + v(u0 ))kKkϕkε
≤
ε
→ 0, as ε → 0. (15.79)
Finally,
hλ2 (u0 + εϕ) − λ2 (u0 ), H(x(u0 ), u0 + εϕ)iL p

ε
Kε kϕ k
≤ kH(x(u0 ), u0 + εϕ )k
ε
→ 0, as ε → 0.
To obtain the last inequalities we have used
λ2 (u0 + εϕ) − λ2 (u0 )

lim sup ≤ Kkϕ k,
ε→0 ε
and
H(x(u0 ), u0 + εϕ) → θ , as ε → 0.
From these last results, we get
δu F̃(x(u0 ), u0 , λ (u0 ), v(u0 ); ϕ)

=
F̃(x(u0 ), u0 + εϕ, λ (u0 ), v(u0 ))
lim
ε→0 ε

F̃(x(u0 ), u0 , λ (u0 ), v(u0 ))
−
ε
=
hFu (x(u0 ), u0 ), ϕiU + hλ1 (u0 ) · Gu (x(u0 ), u0 ), ϕiL p
+hλ2 (u0 ) · Hu (x(u0 ), u0 ), ϕiL p .
In the last lines we have proven the following corollary of the implicit function theorem,
Corollary 15.1
Suppose (x0 , u0 , (λ1 )0 , (λ2 )0 , v0 ) is a solution of the system (15.70), (15.71),(15.72), (15.74), and assume the
corresponding hypotheses of the implicit function theorem are satisfied. Also assume F̃(x, u, λ1 , λ2 , v) is such
´ het second derivative F̃xx (x, u, λ1 , λ2 ) is continuous in a neighborhood of
that the Frec
(x0 , u0 , (λ1 )0 , (λ2 )0 ).
Under such hypotheses, for a given ϕ ∈ [C∞ (Ω)]k , denoting
F1 (u) = F̃(x(u), u, λ1 (u), λ2 (u), v(u)),
we have
δ (F1 (u); ϕ)|u=u0

=
hFu (x(u0 ), u0 ), ϕiU + hλ1 (u0 ) · Gu (x(u0 ), u0 ), ϕiL p
+hλ2 (u0 ) · Hu (x(u0 ), u0 ), ϕiL p .
Chapter 16
On Central Fields in the Calculus

of Variations
16.1 Introduction
In this short communication we develop sufficient conditions of local optimality for a relatively large class
of problems in the calculus of variations. The concerning approach is developed through a generalization of
some theoretical results about central fields presented in [79]. We address both the scalar and vectorial cases
for a domain in Rn . Finally the Weierstrass Excess function has a fundamental role in the formal proofs of
the main results.
16.2 Central fields for the scalar case in the calculus of variations
Let Ω ⊂ Rn be an open, bounded, simply connected set with a regular (Lipschitzian) boundary denoted by
∂ Ω.
Let f ∈ C1 (Ω × R × Rn ) ( f (x, y, z)) and V = C1 (Ω). We suppose f is convex in z, which we denote by
f (x, y, z) be convex.
Choose x̃1 , . . . , x̃n , ỹ1 , . . . , ỹn ∈ R such that if x = (x1 , . . . , xn ) ∈ Ω, then
x̃k < xk < ỹk ∀k ∈ {1, . . . , n}

n
Suppose that y0 ∈ V1 = C1 (B0 ) is stationary for f in B0 = ∏k=1 [x̃k , ỹk ], that is, suppose
n
d
∑ dxk fzk (x, y0 (x), ∇y0 (x)) = fy (x, y0 (x), ∇y0 (x)), ∀x ∈ B0 .
k=1
We also assume f to be of C1 class in B0 × R × Rn .

Consider the problem of minimizing F : D → R where
Z
F(y) = f (x, y(x), ∇y(x)) dx,
Ω
and where
D = {y ∈ V : y = y0 in ∂ Ω}.
On Central Fields in the Calculus of Variations 357
Assume there exists a family of functions F , such that for each (x, y) ∈ D1 ⊂ Rn+1 , there exists a unique
stationary function y3 ∈ V1 = C1 (B0 ) for f in F , such that
(x, y) = (x, y3 (x)).
More specifically, we define F as a subset of F1 , where
F1 = {φ (t, Λ(x, y)) stationary for f such that φ (x, Λ(x, y)) = y for a Λ ∈ Br (0) ⊂ R}
where Br (0) = (0 − r, 0 + r) for some r > 0.

Here, φ is stationary and
φ ((t1 , . . . ,tk = x̃k , . . . ,tn ), Λ(x, y)) = y0 (t1 , . . . ,tk = x̃k , . . . ,tn ), on ∂ B0
and
∇φ ((t1 , . . . ,tk = x̃k , . . . ,tn ), Λ(x, y)) · n = Λ ∈ R, on ∂ B0 , ∀k ∈ {1, . . . , n},
where n denotes the outward normal field to ∂ B0 so that, clarifying the dependence of Λ on (x, y), we have
denoted Λ = Λ(x, y).
Define the field θ : D1 → Rn by
θ (x, y) = ∇y3 (x),
where as above indicated y3 is such that
(x, y) = (x, y3 (x)),
so that
θ (x, y3 (x)) = ∇y3 (x), ∀x ∈ B0 .
At this point we assume the hypotheses of the implicit function theorem so that φ (x, Λ(x, y)) is of C1 class
and θ (x, y) is continuous (in fact, since φ is stationary, the partial derivatives of θ (x, y) are well defined).
Define also
n
h(x, y) = f (x, y, θ (x, y)) − ∑ fz j (x, y, θ (x, y))θ j (x, y)
j=1
and
Pj (x, y) = fz j (x, y, θ (x, y)), ∀ j ∈ {1, . . . , n}.
Observe that
n
hy (x, y3 (x)) = fy (x, y3 (x), θ (x, y3 (x))) + ∑ fz j (x, y3 (x), θ (x, y3 (x)))(θ j )y (x, y3 (x))
j=1
n
− ∑ {(Pj )y (x, y3 (x))θ j (x, y3 (x)) + Pj (x, y3 (x))(θ j )y (x, y3 (x))}
j=1
n
= fy (x, y3 (x), θ (x, y3 (x))) + ∑ Pj (x, y3 (x))(θ j )y (x, y3 (x))
j=1
n
j=1
n
= fy (x, y3 (x), θ (x, y3 (x))) − ∑ (Pj )y (x, y3 (x))θ j (x, y3 (x)). (16.1)
j=1
Thus,
n
hy (x, y) = fy (x, y, θ (x, y)) − ∑ (Pj )y (x, y)θ j (x, y), ∀(x, y) ∈ D1 .
j=1
On the other hand, since y3 (x) is stationary, we obtain

n
d
0 = fy (x, y3 (x), ∇y3 (x)) − ∑ fz j (x, y3 (x), ∇y3 (x))
j=1 dx j
n
= hy (x, y3 (x)) + ∑ (Pj )y (x, y)θ j (x, y)
j=1
n
∂ Pj (x, y3 (x)) ∂ y3 (x)
−∑ + (Pj )y (x, y3 (x))
j=1 ∂xj ∂xj
n
= hy (x, y3 (x)) + ∑ (Pj )y (x, y3 (x))θ j (x, y3 (x))
j=1
n
∂ Pj (x, y3 (x))
−∑ + (Pj )y (x, y3 (x))θ j (x, y3 (x))
j=1 ∂xj
n
∂ Pj (x, y3 (x))
= hy (x, y3 (x)) − ∑ . (16.2)
j=1 ∂xj
Therefore,
n
∂ Pj (x, y)
hy (x, y) = ∑ , ∀(x, y) ∈ D1
j=1 ∂xj
Let H j (x, y) be such that

∂ H j (x, y)
= Pj (x, y).
∂y
From these two last lines, we get
!
n n
∂ (H j )y (x, y) ∂ H j (x, y)
hy (x, y) = ∑ = ∑ ∂xj , ∀(x, y) ∈ D1
j=1 ∂xj j=1 y
so that
n
∂ H j (x, y)
h(x, y) = ∑ +W (x),
j=1 ∂ xk
for some W : B0 → R.
Hence, assuming D1 contains an open which contains C0 , where C0 = {(x, y0 (x)) : x ∈ Ω}, for y ∈ D
sufficiently close to y0 in L∞ norm, the generalized Hilbert integral, denoted by I(y), will be defined by
n
∂ y(x)
Z
I(y) = h(x, y(x)) dx + ∑ Pj (x, y(x)) dx
Ω j=1 ∂xj
n
∂ H j (x, y(x)) ∂ y(x)
Z Z
= ∑ + (H j )y (x, y(x)) dx + W (x) dx
Ω j=1 ∂xj ∂xj Ω
Z n
dH j (x, y(x))
Z
= ∑ dx j dx + W (x) dx
Ω j=1 Ω
Z n Z
=
∂ Ω j=1
∑ (−1) j+1 H j (x, y(x))dx1 ∧ · · · ∧ dcx j ∧ · · · ∧ dxn + Ω
W (x) dx
= W1 (y|∂ Ω )
= W1 ((y0 )|∂ Ω ), (16.3)
so that such an integral is invariant, that is, it does not depend on y.

Z Z
F(y) − F(y0 ) = f (x, y(x), ∇y(x)) dx − f (x, y0 (x), ∇y0 (x)) dx.
Ω Ω
On the other hand,

Z
I(y) = f (x, y(x), θ (x, y)) dx
Ω
n Z
+∑ fz j (x, y(x), θ (x, y(x))(θ j (x, y(x)) − yx j (x)) dx
j=1 Ω
= I(y0 )
Z
= f (x, y0 (x), θ (x, y0 (x))) dx
ZΩ
= f (x, y0 (x), ∇y0 (x)) dx
Ω
= F(y0 ). (16.4)
Therefore,
Z
F(y) − F(y0 ) = [ f (x, y(x), ∇y(x)) − f (x, y(x), θ (x, y))] dx
Ω
n Z
−∑ fz j (x, y(x), θ (x, y(x))(yx j (x) − θ j (x, y(x))) dx
j=1 Ω
Z
= E (x, y(x), θ (x, y(x)), ∇y(x)) dx, (16.5)
Ω
where
E (x, y(x), θ (x, y(x)), ∇y(x)) = f (x, y(x), ∇y(x))

− f (x, y(x), θ (x, y))
n
− ∑ fz j (x, y(x), θ (x, y(x))(yx j (x) − θ j (x, y(x))), (16.6)
j=1
is the Weierstrass Excess function.

With such results, we may prove the following theorem.
Theorem 16.2.1 Let Ω ⊂ Rn be an open, bounded and simply connected set, with a regular (Lipschitzian)
boundary denoted by ∂ Ω. Let V = C1 (Ω) and let f ∈ C1 (Ω × R × Rn ) be such that f (x, y, z) is convex. Let
F : D → R be defined by Z
F(y) = f (x, y(x), ∇y(x)) dx,
Ω
where
D = {y ∈ V : y = y1 em ∂ Ω}.
Let y0 ∈ D be a stationary function for f which may be extended to B0 , being kept stationary in B0 , where
B0 has been specified above in this section. Suppose we may define a field θ : D1 → Rn , also as it has been
specified above in this section. Assume D1 ⊂ Rn+1 contains an open set which contains C0 , where
C0 = {(x, y0 (x)) : x ∈ Ω}.

Under such hypotheses, there exists δ > 0 such that
F(y) ≥ F(y0 ), ∀y ∈ Bδ (y0 ) ∩ D,
where
Bδ (y0 ) = {y ∈ V : ky − y0 k∞ < δ }.
Proof 16.1 From the hypotheses and from the exposed above in this section, there exists δ > 0 such that
θ (x, y(x)) is well defined for all y ∈ D such that
ky − y0 k∞ < δ .
Let y ∈ Bδ (y0 ) ∩ D. Thus, since f (x, y, z) is convex, we have

− f (x, y(x)θ (x, y))
n
− ∑ fz j (x, y(x), θ (x, y(x))(yx j (x) − θ j (x, y(x)))
j=1
≥ 0, in Ω (16.7)
so that, Z
F(y) − F(y0 ) = E (x, y(x), θ (x, y(x)), ∇y(x)) dx ≥ 0, ∀y ∈ Bδ (y0 ) ∩ D.
Ω
16.3 Central fields and the vectorial case in the calculus of variations
Let Ω ⊂ Rn be an open, bounded, simply connected set with a regular (Lipschitzian) boundary denoted by
∂ Ω.
Let f ∈ C1 (Ω×RN ×RNn ) ( f (x, y, z)) and V = C1 (Ω; RN ). We suppose f is convex in z, which we denote
by f (x, y, z) be convex.
Choose x̃1 , . . . , x̃n , ỹ1 , . . . , ỹn ∈ R such that if x = (x1 , . . . , xn ) ∈ Ω, then
x̃k < xk < ỹk ∀k ∈ {1, . . . , n}
Suppose that y0 ∈ V1 = C1 (B0 ; RN ) is stationary for f in B0 = ∏nk=1 [x̃k , ỹk ], that is, suppose that
n
d
∑ dxk fz jk (x, y0 (x), ∇y0 (x)) = fy j (x, y0 (x), ∇y0 (x)), ∀x ∈ B0 , ∀ j ∈ {1, . . . , N }.
k=1
We also assume f to be of C1 class in B0 × RN × RNn .

Consider the problem of minimizing F : D → R where
Z
F(y) = f (x, y(x), ∇y(x)) dx,
Ω
and where
D = {y ∈ V : y = y0 in ∂ Ω}.
Assume there exists a family of stationary functions F , such that for each (x, y) ∈ D1 ⊂ Rn+N , there
exists a unique stationary function y3 ∈ V1 for f in F , such that
(x, y) = (x, y3 (x)).
More specifically, we define F as a subset of F1 , where
F1 = {Φ(t, Λ(x, y)) stationary for f such that Φ(x, Λ(x, y)) = y for a Λ ∈ Br (0) ⊂ RN }
such that Br (0) is an open ball of center 0 and radius r, for some r > 0.
Here, Φ is stationary and
Φ((t1 , . . . ,tk = x̃k , . . . ,tn ), Λ(x, y)) = y0 (t1 , . . . ,tk = x̃k , . . . ,tn ), on ∂ B0
and
∇(Φ) j ((t1 , . . . ,tk = x̃k , . . . ,tn ), Λ(x, y)) · n = Λ j , on ∂ B0 , ∀k ∈ {1, . . . , n}, j ∈ {1, . . . , N}.
Here, n denotes the outward normal field to ∂ B0 .
Define the field θ : D1 ⊂ Rn+N → RNn by
θ j (x, y) = ∇(y3 ) j (x),
where as above indicated, y3 is such that
(x, y) = (x, y3 (x)),
and thus
θ j (x, (y3 )(x)) = ∇(y3 ) j (x), ∀x ∈ B0 .
At this point we assume the hypotheses of the implicit function theorem so that Φ j (x, Λ(x, y)) is of C1
class and θ j (x, y) is continuous (in fact, since Φ j is stationary, the partial derivatives of θ j (x, y) are well
defined, ∀ j ∈ {1, . . . , N}).
Define also
n n
h(x, y) = f (x, y, θ (x, y)) − ∑ ∑ fz jk (x, y, θ (x, y))θ jk (x, y)
j=1 k=1
and
Pjk (x, y) = fz jk (x, y, θ (x, y)), ∀ j ∈ {1, . . . , N}, k ∈ {1, . . . , n}.
Observe that
N n
hy j (x, y3 (x)) = fy j (x, y3 (x), θ (x, y3 (x))) + ∑ ∑ fzlk (x, y3 (x), θ (x, y3 (x)))(θlk )y j (x, y3 (x))
l=1 k=1
N n
−∑ ∑ {(Plk )y j (x, y3 (x))θlk (x, y3 (x)) + Plk (x, y3 (x))(θlk )y j (x, y3 (x))}
l=1 k=1
N n
= fy j (x, y3 (x), θ (x, y3 (x))) + ∑ ∑ Plk (x, y3 (x))(θlk )y j (x, y3 (x))
l=1 k=1
n
j=1
N n
= fy j (x, y3 (x), θ (x, y3 (x))) − ∑ ∑ (Plk )y j (x, y3 (x))θlk (x, y3 (x)). (16.8)
l=1 k=1
Thus,
N n
hy j (x, y) = fy j (x, y, θ (x, y)) − ∑ ∑ (Plk )y j (x, y)θlk (x, y), ∀(x, y) ∈ D1 .
l =1 k=1
On the other hand, considering that y3 (x) is stationary, we obtain Hk (x, y) such that
∂ Hk (x, y)
Pjk (x, y) = , ∀ j ∈ {1, . . . , N}, k ∈ {1, . . . , n}.
∂yj
Indeed, we define
Z xk
Hk (x, y) = staφ ∈Dˆ k f (x1 , . . . ,tk , . . . , xn , φ1 (tk ), . . . , φN (tk ), ∇φ̃ (x,tk )) dtk ,
x̃k
where, denoting tˆk = (x1 , . . . ,tk , . . . , xn ), we have that

 
∂ φ1 (tk )
((y3 )1 )x1 (tˆk ) ((y3 )1 )x2 (tˆk ) ··· ∂tk ··· ((y3 )1 )xn (tˆk )

 ((y3 )2 )x (tˆk ) ((y3 )2 )x (tˆk ) ∂ φ2 (tk ) 
1 2 ··· ∂tk ··· ((y3 )2 )xn (tˆk ) 
∇φ̃ (x,tk ) =  , (16.9)
 
.
.. .
.. .. .. ...


 ··· . . 

∂ φN (tk )
((y3 )N )x1 (tˆk ) ((y3 )N )x2 (tˆk ) ··· ∂tk ··· ((y3 )N )xn (tˆk )
N×n
and where
D̂k = {φ ∈ C1 ([x̃k , xk ]; RN ) : φ (x̃k ) = y3 (x̂˜k ) and φ (xk ) = y3 (x̂k )},
∀k ∈ {1, . . . , n}.
Observe that, since φ (tk ) is stationary, we have
d
fy j [φ (tk )] − fz [φ (tk )] = 0, in [x̃k , xk ], ∀ j ∈ {1, . . . , N } (16.10)
dtk jk
where generically, we denote
f [φ (tk )] = f (x, φ (tk ), ∇φ̃ (tk )).
Observe that from these Euler-Lagrange equations we may obviously obtain
φ (tk ) = y3 (x1 , . . . ,tk , . . . , xn ).
At this point we shall also denote

˜ y)),
φ j (tk ) ≡ φ j (tk , Λ(x,
where
φ j (x̃k , Λ̃(x, y)) = (y3 ) j (x1 , . . . , x̃k , . . . , xn ),
∂ φ j (x̃k , Λ̃(x, y))

= Λ̃ j ,
∂tk
and where Λ̃(x, y) is such that
φ j (xk , Λ̃(x, y)) = (y3 ) j (x1 , . . . , xk , . . . , xn ).
Therefore,
Z xk
Hk (x, y) = f [φ (tk , Λ̃(x, y)] dtk ,
x̃k
and thus,
Z xk
[Hk (x, y)]y j = ˜ y)](φl ) Λ̃y (x, y) dtk
fyl [φ (tk , Λ(x, Λ̃ j
x̃k
Z xk
+ ˜ y (x, y) dtk
fzlk [φ (tk , Λ̃(x, y)](φl )tk Λ̃ Λ j
x˜k
Z xk
= fyl [φ (tk , Λ̃(x, y)](φl )Λ̃ Λ̃y j (x, y) dtk
x̃k
Z xk
+ ˜ y (x, y) dtk .
fzlk [φ (tk , Λ̃(x, y)][(φl )Λ̃ ]tk Λ (16.11)
j
x̃k

Z xk
d
[Hk (x, y)]y j = fzlk [φ (tk , Λ̃(x, y)](φl )Λ̃ Λ̃y j (x, y) dtk
x̃k dtk
Z xk
+ ˜ y (x, y) dtk
fzlk [φ (tk , Λ̃(x, y)][(φl )Λ˜ ]tk Λ j
x̃k
Z xk
d
= { fzlk [φ (tk , Λ̃(x, y)](φl )Λ˜ ]} dtk Λ̃y j (x, y)
x̃k dtk
t =x
˜ y (x, y)
= { fzlk [φ (tk , Λ̃(x, y)](φl )Λ˜ ]}|tkk =x˜kk Λ j (16.12)
On the other hand,
˜ y)) = (y3 )l (x1 , . . . , x̃k , . . . , xn ) = (y0 )l (x1 , . . . , x̃k , . . . , xn ),

φl (x̃k , Λ(x,
which does not depend on Λ̃, so that
(φl )Λ (x̃k , Λ̃(x, y)) = 0.
Also,
φl (xk , Λ̃(x, y)) = yl
and thus
∂ yl
[φl (xk , Λ̃(x, y))]y j = = δl j .
∂yj
Hence,
(φl )Λ̃ (xk , Λ̃(x, y))Λ̃y j = δl j .
From these last results and from (16.12), we obtain,
[Hk (x, y)]y j = fzlk [φ (xk , Λ̃(x, y))]δl j

= fz jk [φ (xk , Λ̃(x, y))]
= fz jk (x, y, θ (x, y))
= Pjk (x, y), ∀ j ∈ {1, . . . , N}, k ∈ {1, . . . , n}. (16.13)
Therefore,
n
d
0 = fy j (x, y3 (x), ∇y3 (x)) − ∑ fz jk (x, y3 (x), ∇y3 (x))
k=1 dx k
N n
= hy j (x, y3 (x)) + ∑ ∑ (Plk )y j (x, y3 (x))θlk (x, y3 (x))
l=1 k=1
!
n ∂ Pjk (x, y3 (x)) N ∂ (y3 )l (x)
−∑ + ∑ (Pjk )yl (x, y3 (x))
k=1 ∂ xk l=1 ∂ xk
N n
= hy j (x, y3 (x)) + ∑ ∑ (Plk )y j (x, y3 (x))θlk (x, y3 (x))
l=1 k=1
!
n
∂ Pjk (x, y3 (x)) N
−∑ + ∑ (Pjk )yl (x, y3 (x))θlk (x, y3 (x))
k=1 ∂ xk l=1
n ∂ (H ) (x, y (x))
k yj 3
= hy j (x, y3 (x)) − ∑ . (16.14)
k=1 ∂ xk
Thus,
n
∂ Hk (x, y)
hy j (x, y) = ∑ , ∀(x, y) ∈ D1
k=1 ∂ xk yj
so that
n
∂ Hk (x, y)
h(x, y) = ∑ +W (x),
k=1 ∂ xk
for some W : B0 → R.
Hence, assuming D1 contains an open set which contains C0 , where
C0 = {(x, y0 (x)) : x ∈ Ω},
for y ∈ D sufficiently close to y0 in L∞ norm, the generalized Hilbert integral, denoted by I(y), will be defined
as
N n
∂ y j (x)
Z
I (y) = h(x, y(x)) dx + ∑ ∑ Pjk (x, y(x)) dx
Ω j=1 k=1 ∂ xk
!
n n
∂ Hk (x, y(x)) ∂ y j (x)
Z Z
= ∑ + ∑ (Hk )y j (x, y(x)) dx + W (x) dx
Ω k=1 ∂ xk j=1 ∂ xk Ω
n
dHk (x, y(x))
Z Z
= ∑ dx + W (x) dx
Ω k=1 dxk Ω
Z n Z
= ∑ (−1)k+1 Hk (x, y(x))dx1 ∧ · · · ∧ dcxk ∧ · · · ∧ dxn +
∂ Ω k=1 Ω
W (x) dx
= W1 (y|∂ Ω )
= W1 ((y1 )|∂ Ω ), (16.15)
so that such an integral is invariant, that is, it does not depend on y.

Z Z
F(y) − F(y0 ) = f (x, y(x), ∇y(x)) dx − f (x, y0 (x), ∇y0 (x)) dx.
Ω Ω
On the other hand,

Z
I(y) = f (x, y(x), θ (x, y(x))) dx
Ω
N n Z
+∑ ∑ fz jk (x, y(x), θ (x, y(x))(θ jk (x, y(x)) − (y j )xk (x)) dx
j=1 k=1 Ω
= I(y0 )
Z
= f (x, y0 (x), θ (x, y0 (x))) dx
ZΩ
= f (x, y0 (x), ∇y0 (x)) dx
Ω
= F(y0 ). (16.16)
Thus,
Z
F(y) − F(y0 ) = [ f (x, y(x), ∇y(x)) − f (x, y(x), θ (x, y(x)))] dx
Ω
N n Z
−∑ ∑ fz jk (x, y(x), θ (x, y(x))((y j )xk (x) − θ jk (x, y(x))) dx
j=1 k=1 Ω
Z
= E (x, y(x), θ (x, y(x)), ∇y(x)) dx, (16.17)
Ω
where

− f (x, y(x), θ (x, y(x)))
N n
−∑ ∑ fz jk (x, y(x), θ (x, y(x))((y j )xk (x) − θ jk (x, y(x)))
j=1 k=1
is the Weierstrass Excess function.

With such results, we may prove the following result.
Theorem 16.3.1 Let Ω ⊂ Rn be an open, bounded, simply connected set with a regular (Lipschitzian) bound-
ary denoted by ∂ Ω. Let V = C1 (Ω; RN ) and let f ∈ C1 (Ω × RN × RNn ) be such that f (x, y, z) is convex. Let
F : D → R be defined by Z
F(y) = f (x, y(x), ∇y(x)) dx,
Ω
where
D = {y ∈ V : y = y1 on ∂ Ω}.
Let y0 ∈ D be an stationary function for f which may be extended to B0 , keeping it stationary in B0 , where
B0 has been specified above in this section. Suppose we may define a field θ : D1 → RNn , also as specified
above in this section. Assume D1 ⊂ Rn+N contains an open set which contains C0 , where
C0 = {(x, y0 (x)) : x ∈ Ω}.

Under such hypotheses, there exists δ > 0 such that
F(y) ≥ F(y0 ), ∀y ∈ Bδ (y0 ) ∩ D,
where
Bδ (y0 ) = {y ∈ V : ky − y0 k∞ < δ }.
Proof 16.2 From the hypotheses and from the exposed above in this section, there exists δ > 0 such that
θ (x, y(x)) is well defined for each y ∈ D such that
ky − y0 k∞ < δ .
Let y ∈ Bδ (y0 ) ∩ D. Thus, since f (x, y, z) is convex, we have

− f (x, y(x), θ (x, y(x)))
N n
−∑ ∑ fz jk (x, y(x), θ (x, y(x))((y j )x j (x) − θ jk (x, y(x)))
j=1 k=1
≥ 0, in Ω (16.18)
so that, Z
F(y) − F(y0 ) = E (x, y(x), θ (x, y(x)), ∇y(x)) dx ≥ 0, ∀y ∈ Bδ (y0 ) ∩ D.
Ω
SECTION III
APPLICATIONS TO MODELS IN
PHYSICS AND ENGINEERING
Chapter 17
Global Existence Results and

Duality for Non-Linear Models of
Plates and Shells
17.1 Introduction
In this chapter, in a first step, we develop a new existence proof and a dual variational formulation for the
Kirchhoff-Love thin plate model. Previous results on existence in mathematical elasticity and related models
may be found in [29, 30, 31].
At this point we refer to the exceptionally important article “A contribution to contact problems for a
class of solids and structures” by W.R. Bielski and J.J. Telega [11], published in 1985, as the first one to
successfully apply and generalize the convex analysis approach to a model in non-convex and non-linear
mechanics.
The present work is, in some sense, a kind of extension of this previous work [11] and others such as
[10], which greatly influenced and inspired my work and recent book [14].
Here we highlight that such earlier results establish the complementary energy under the hypothesis of
positive definiteness of the membrane force tensor at a critical point (please see [11, 10, 39] for details).
We have obtained a dual variational formulation which allows the global optimal point in question not
to be positive definite (for related results see F. Botelho [14]), but also not necessarily negative definite.
The approach developed also includes sufficient conditions of optimality for the primal problem. It is worth
mentioning that the standard tools of convex analysis used in this text may be found in [33, 14, 64], for
example.
At this point we start to describe the primal formulation.

Let Ω ⊂ R2 be an open, bounded, connected set which represents the middle surface of a plate of thickness
h. The boundary of Ω, which is assumed to be regular (Lipschitzian), is denoted by ∂ Ω. The vectorial basis
related to the cartesian system {x1 , x2 , x3 } is denoted by (aα , a3 ), where α = 1, 2 (in general Greek indices
stand for 1 or 2), and where a3 is the vector normal to Ω, whereas a1 and a2 are orthogonal vectors parallel
to Ω. Also, n is the outward normal to the plate surface.
The displacements will be denoted by
û = {ûα , û3 } = ûα aα + û3 a3 .
The Kirchhoff-Love relations are
ûα (x1 , x2 , x3 ) = uα (x1 , x2 ) − x3 w(x1 , x2 ),α

and û3 (x1 , x2 , x3 ) = w(x1 , x2 ). (17.1)
Here −h/2 ≤ x3 ≤ h/2 so that we have u = (uα , w) ∈ U where
(uα , w) ∈ W 1,2 (Ω; R2 ) ×W 2,2 (Ω),

U =
∂w
uα = w = = 0 on ∂ Ω}
∂n
= W01,2 (Ω; R2 ) ×W02,2 (Ω).
It is worth emphasizing that the boundary conditions here specified refer to a clamped plate.
We define the operator Λ : U → Y ×Y , where Y = Y ∗ = L2 (Ω; R2×2 ), by
Λ(u) = {γ(u), κ(u)},

uα,β + uβ ,α w,α w,β
γαβ (u) = + ,
2 2
καβ (u) = −w,αβ .
The constitutive relations are given by
Nαβ (u) = Hαβ λ µ γλ µ (u), (17.2)
Mαβ (u) = hαβ λ µ κλ µ (u), (17.3)

h2
where: {Hαβ λ µ } and {hαβ λ µ = 12 Hαβ λ µ }, are symmetric positive definite fourth order tensors. From now
on, we denote {H αβ λ µ } = {Hαβ λ µ }−1 and {hαβ λ µ } = {hαβ λ µ }−1 .
Furthermore {Nαβ } denote the membrane force tensor and {Mαβ } the moment one. The plate stored
energy, represented by (G ◦ Λ) : U → R is expressed by
1 1
Z Z
(G ◦ Λ)(u) = Nαβ (u)γαβ (u) dx + Mαβ (u)καβ (u) dx (17.4)
2 Ω 2 Ω
and the external work, represented by F : U → R, is given by
F(u) = hw, PiL2 (Ω) + huα , Pα iL2 (Ω) , (17.5)
where P, P1 , P2 ∈ L2 (Ω) are external loads in the directions a3 , a1 and a2 respectively. The potential energy,
denoted by J : U → R is expressed by:
J(u) = (G ◦ Λ)(u) − F(u)

Global Existence Results and Duality for Non-Linear Models of Plates and Shells 369
Finally, we also emphasize from now on, as their meaning are clear, we may denote L2 (Ω) and
L2 (Ω; R2×2 ) simply by L2 , and the respective norms by k · k2 . Moreover derivatives are always understood
in the distributional sense, 0 may denote the zero vector in appropriate Banach spaces and, the following and
relating notations are used:
∂ 2w
w,αβ = ,
∂ xα ∂ xβ
∂ uα
uα,β = ,
∂ xβ
∂ Nαβ
Nαβ ,1 = ,
∂ x1
and
∂ Nαβ
Nαβ ,2 = .
∂ x2
17.2 On the existence of a global minimizer

At this point we present an existence result concerning the Kirchhoff-Love plate model.
We start with the following two remarks.
Remark 17.1 Let {Pα } ∈ L∞ (Ω; R2 ). We may easily obtain by appropriate Lebesgue integration {T̃αβ }
symmetric and such that
T̃αβ ,β = −Pα , in Ω.
Indeed, extending {Pα } to zero outside Ω if necessary, we may set
Z x
T̃11 (x, y) = − P1 (ξ , y) dξ ,
0
Z y
T̃22 (x, y) = − P2 (x, ξ ) dξ ,
0
and
T̃12 (x, y) = T̃21 (x, y) = 0, in Ω.
Thus, we may choose a C > 0 sufficiently big, such that
{Tαβ } = {T˜αβ +Cδαβ }
is positive definite in Ω, so that
Tαβ ,β = T˜αβ ,β = −Pα ,

where
{δαβ }
is the Kronecker delta.
So, for the kind of boundary conditions of the next theorem, we do NOT have any restriction for the {Pα }
norm.
Summarizing, the next result is new and it is really a step forward concerning the previous one in Ciarlet
[30]. We emphasize this result and its proof through such a tensor {Tαβ } are new, even though the final part
of the proof is established through a standard procedure in the calculus of variations.
About the other existence result for plates, its proof through the tensor well specified {(T0 )αβ } is also
new, even though the final part of such a proof is also performed through a standard procedure.
A similar remark is valid for the existence result for the model of shells, which is also established through
a tensor T0 properly specified.
Finally, the duality principles and concerning optimality conditions are established through new func-
tionals. Similar results may be found in [14].
Remark 17.2 Specifically about the existence of the tensor T0 relating Theorem 17.3.1, we recall the
following well known duality principle of the calculus of variations

1 2
inf kT k2
T ={Tαβ }∈B∗ 2

1
Z
= sup − ∇uα · ∇uα dx + huα , Pα iL2 (Ω) + huα , Pαt iL2 (Γt ) . (17.6)
{uα }∈Ũ 2 Ω
Here
B∗ = {T ∈ L2 (Ω; R4 ) : Tαβ ,β + Pα = 0, in Ω, Tαβ nβ − Pαt = 0, on Γt },
and
Ũ = {{uα } ∈ W 1,2 (Ω; R2 ) : uα = 0 on Γ0 }.
We also recall that the existence of a unique solution for both these primal and dual convex formulations
is a well known result of the duality theory in the calculus of variations. Please, see related results in [33].
A similar duality principle may be established for the case of Theorem 17.6.1.
Theorem 17.2.1 Let Ω ⊂ R2 be an open, bounded, connected set with a Lipschitzian boundary denoted by
∂ Ω = Γ. Suppose (G ◦ Λ) : U → R is defined by
G(Λu) = G1 (γ(u)) + G2 (κ(u)), ∀u ∈ U,
where
1
Z
G1 (γu) = Hαβ λ µ γαβ (u)γλ µ (u) dx,
2 Ω
and
1
Z
G2 (κu) = hαβ λ µ καβ (u)κλ µ (u) dx,
2 Ω
where
Λ(u) = (γ(u), κ(u)) = ({γαβ (u)}, {καβ (u)}),
γαβ (u) = + ,
2 2
καβ (u) = −w,αβ ,
and where
U = {u = (u1 , u2 , w) ∈ W 1,2 (Ω; R2 ) ×W 2,2 (Ω) : u1 = u2 = w = 0 on ∂ Ω}.
We also define,
F1 (u) = hw, PiL2 + huα , Pα iL2

≡ hu, fiL2 , (17.7)
where
f = (Pα , P) ∈ L∞ (Ω; R3 ).
Let J : U → R be defined by
J(u) = G(Λu) − F1 (u), ∀u ∈ U.
Assume there exists {cαβ } ∈ R2×2 such that cαβ > 0, ∀α, β ∈ {1, 2} and
G2 (κ(u)) ≥ cαβ kw,αβ k22 , ∀u ∈ U.
Under such hypotheses, there exists u0 ∈ U such that
J(u0 ) = min J(u).

u∈U
Proof 17.1 Observe that we may find Tα = {(Tα )β } such that
divTα = Tαβ ,β = −Pα
an also such that {Tαβ } is positive definite and symmetric (please, see Remark 17.1).
Thus defining
uα,β + uβ ,α 1
vαβ (u) = + w,α w,β , (17.8)
2 2
we obtain
J(u) = G1 ({vαβ (u)}) + G2 (κ(u)) − hPα , uα iL2 − hw, PiL2

= G1 ({vαβ (u)}) + G2 (κ(u)) + hTαβ ,β , uα iL2 − hw, PiL2
uα ,β + uβ ,α

= G1 ({vαβ (u)}) + G2 (κ (u)) − Tαβ , − hw, PiL2
2 L2

1
= G1 ({vαβ (u)}) + G2 (κ(u)) − Tαβ , vαβ (u) − w,α w,β − hw, PiL2
2 L2
1
≥ cαβ kw,αβ k22 + Tαβ , w,α w,β L2 − hw, PiL2 + G1 ({vαβ (u)})
2
−hTαβ , vαβ (u)iL2 . (17.9)
From this, since {Tαβ } is positive definite, clearly J is bounded below.

Let {un } ∈ U be a minimizing sequence for J. Thus there exists α1 ∈ R such that
lim J(un ) = inf J(u) = α1 .

n→∞ u∈U
From (17.9), there exists K1 > 0 such that
k(wn ),αβ k2 < K1 , ∀α, β ∈ {1, 2}, n ∈ N.
Therefore, there exists w0 ∈ W 2,2 (Ω) such that, up to a subsequence not relabeled,
(wn ),αβ * (w0 ),αβ , weakly in L2 ,
∀α, β ∈ {1, 2}, as n → ∞.

Moreover, also up to a subsequence not relabeled,
(wn ),α → (w0 ),α , strongly in L2 and L4 , (17.10)
∀α, ∈ {1, 2}, as n → ∞.

Also from (17.9), there exists K2 > 0 such that,
k(vn )αβ (u)k2 < K2 , ∀α, β ∈ {1, 2}, n ∈ N,

and thus, from this, (17.8) and (17.10), we may infer that there exists K3 > 0 such that
k(un )α,β + (un )β ,α k2 < K3 , ∀α, β ∈ {1, 2}, n ∈ N.
From this and Korn’s inequality, there exists K4 > 0 such that
kun kW 1,2 (Ω;R2 ) ≤ K4 , ∀n ∈ N.
So, up to a subsequence not relabeled, there exists {(u0 )α } ∈ W 1,2 (Ω, R2 ), such that
(un )α,β + (un )β ,α * (u0 )α,β + (u0 )β ,α , weakly in L2 ,
∀α, β ∈ {1, 2}, as n → ∞, and,

(un )α → (u0 )α , strongly in L2 ,
∀α ∈ {1, 2}, as n → ∞.
Moreover, the boundary conditions satisfied by the subsequences are also satisfied for w0 and u0 in a
trace sense, so that
u0 = ((u0 )α , w0 ) ∈ U.
From this, up to a subsequence not relabeled, we get
γαβ (un ) * γαβ (u0 ), weakly in L2 ,
∀α, β ∈ {1, 2}, and

καβ (un ) * καβ (u0 ), weakly in L2 ,
∀α, β ∈ {1, 2}.
Therefore, from the convexity of G1 in γ and G2 in κ we obtain
inf J(u) = α1
u∈U
= lim inf J(un )
n→∞
≥ J(u0 ). (17.11)
Thus,
J(u0 ) = min J(u).
u∈U
17.3 Existence of a minimizer for the plate model for a more general
case
At this point we present an existence result for a more general case.
Theorem 17.3.1 Consider the statements and assumptions concerning the plate model described in the last
section.
More specifically, consider the functional J : U → R given, as above described, by
J(u) = W (γ(u), κ(u)) − hPα uα iL2

−hw, PiL2 − hPαt , uα iL2 (Γt )
−hPt , wiL2 (Γt ) , (17.12)
where,
U = {u = (uα , w) = (u1 , u2 , w) ∈ W 1,2 (Ω; R2 ) ×W 2,2 (Ω) :

∂w
uα = w = = 0, on Γ0 }, (17.13)
∂n
where ∂ Ω = Γ0 ∪ Γt and the Lebesgue measures
mΓ (Γ0 ∩ Γt ) = 0,
and
mΓ (Γ0 ) > 0.
Let T0 be such that,
kT0 k22 = min {kT k22 },
T ∈L2 (Ω;R2×2 )
subject to
Tαβ ,β + Pα = 0 in Ω,
(T0 )αβ nβ − Pαt = 0, on Γt .
Assume kPα k∞ and kPαt k∞ are small enough so that
J1 (u) → +∞, as hw,αβ , w,αβ iL2 → +∞, (17.14)
where
1
J1 (u) = G2 (κ(u)) + h(T0 )αβ , w,α w,β iL2
2
−hP, wiL2 − hPt , wiL2 (Γt ) . (17.15)
Under such hypotheses, there exists u0 ∈ U such that,
J(u0 ) = min{J(u)}.
u∈U
Proof 17.2
Observe that defining
uα ,β + uβ ,α 1
vαβ (u) = + w,α w,β , (17.16)
2 2
we have,
J(u) = G1 (v(u)) + G2 (κ(u))

−hPα , uα iL2 − hP, wiL2
−hPαt , uα iL2 (Γt ) − hPt , wiL2 (Γt )
= G1 (v(u)) + G2 (κ(u)) + h(T0 )αβ ,β , uα iL2
−hP, wiL2
uα ,β + uβ ,α

= G1 (v(u)) + G2 (κ(u)) − (T0 )αβ ,
2 L2
+h(T0 )αβ nβ , uα iL2 (Γt ) − hP, wiL2
uα ,β + uβ ,α

= G1 (v(u)) + G2 (κ(u)) − (T0 )αβ ,
2 L2
−hP, wiL2 − hPt , wiL2 (Γt )

1
= G1 (v(u)) + G2 (κ(u)) − (T0 )αβ , vαβ (u) − w,α w,β
2 L2
−hP, wiL2 − hPt , wiL2 (Γt )
= J1 (u) + G1 (v(u)) − h(T0 )αβ , vαβ (u)iL2 (17.17)
From this and the hypothesis (17.14), J is bounded below. So, there exists α1 ∈ R such that
α1 = inf J(u).
u∈U
Let {un } be a minimizing sequence for J.

From (17.17) and also from the hypothesis (17.14), {kwn k2,2 } is bounded.
So there exists w0 ∈ U such that, up to a not relabeled subsequence,
(wn ),αβ * (w0 ),αβ weakly in L2 (Ω), ∀α, β ∈ {1, 2},
(wn ),α → (w0 ),α strongly in L2 (Ω), ∀α ∈ {1, 2}, (17.18)

k(vn )αβ (u)k2 < K2 , ∀α, β ∈ {1, 2}, n ∈ N,
and thus, from this, (17.16) and (17.18), we may infer that there exists K3 > 0 such that
k(un )α,β (u) + (un )β ,α (u)k2 < K3 , ∀α, β ∈ {1, 2}, n ∈ N.
From this and Korn’s inequality, there exists K4 > 0 such that
kun kW 1,2 (Ω;R2 ) ≤ K4 , ∀n ∈ N.

So, up to a subsequence not relabeled, there exists {(u0 )α } ∈ W 1,2 (Ω, R2 ), such that
(un )α,β + (un )β ,α * (u0 )α,β + (u0 )β ,α , weakly in L2 ,
∀α, β ∈ {1, 2}, as n → ∞, and,

(un )α → (u0 )α , strongly in L2 ,
∀α ∈ {1, 2}, as n → ∞.
Moreover, the boundary conditions satisfied by the subsequences are also satisfied for w0 and u0 in a
trace sense, so that
u0 = ((u0 )α , w0 ) ∈ U.
From this, up to a subsequence not relabeled, we get
∀α, β ∈ {1, 2}, and

∀α, β ∈ {1, 2}.
inf J(u) = α1
u∈U
= lim inf J(un )
n→∞
≥ J(u0 ). (17.19)
Thus,
J(u0 ) = min J(u).
u∈U
17.4 The main duality principle

In this section we present a duality principle for the plate model in question.
For such a result, we emphasize the dual variational formulation is concave.
Remark 17.3 In the proofs relating our duality principles we apply a very well known result found in
Toland [78].
Indeed, for
{Nαβ } ∈ L2 (Ω; R2×2 ),
assume
F5 (w) − G5 ({wα }) > 0, ∀w ∈ W02,2 (Ω) such that w 6= 0,
where here
1 K
Z
F5 (w) = hαβ λ µ w,αβ w,λ µ dx + hw,α , w,α iL2 ,
2 Ω 2
and
1 K
Z
G5 ({wα }) = − Nαβ w,α w,β dx + hw,α , w,α iL2 ,
2 Ω 2
where K > 0 is supposed to be sufficiently big so that G5 is convex in w.
Thus,
1
Z
F5 (w) − G5 ({wα }) = hαβ λ µ w,αβ w,λ µ dx
2 Ω
1
Z
+ Nαβ w,α w,β dx > 0, (17.20)
2 Ω
∀w ∈ W02,2 (Ω) such that w 6= 0.

Therefore,
−hwα , z∗α iL2 + F5 (w) + sup {h(v2 )α , z∗α iL2 − G5 ({(v2 )α })

v2 ∈L2
1
Z
= −hwα , z∗α iL2 + F5 (w) + (−Nαβ )K zα∗ zβ∗ dx > 0, (17.21)
2 Ω
∀w ∈ W02,2 (Ω) such that w =

6 0,
so that
1
Z
inf {−hwα , z∗α iL2 + F5 (w)} + (−Nαβ )K z∗α z∗β d x
2,2
w∈W0 (Ω) 2 Ω
1
Z
= −F5∗ (z∗ ) + (−Nαβ )K z∗α z∗β dx ≥ 0, (17.22)
2 Ω
∀z∗ ∈ L2 .
Indeed, from the general result in Toland [78], we have

1
Z
inf −F5∗ (z∗ ) + (−Nαβ )K z∗α z∗β dx
z∗ ∈L2 2 Ω
= inf {F5 (w) − G5 ({w,α })
2,2
w∈W0
≤ F5 (w) − G5 ({w,α })
1
Z
= h w w dx
2 Ω αβ λ µ ,αβ ,λ µ
1
Z
+ N w,α w,β dx, ∀w ∈ W02,2 (Ω). (17.23)
2 Ω αβ
At this point we enunciate and prove our main duality principle.
∂ Ω = Γ. Suppose (G ◦ Λ) : U → R is defined by
G(Λu) = G1 (γ(u)) + G2 (κ(u)), ∀u ∈ U,
where
1
Z
G1 (γ(u)) = Hαβ λ µ γαβ (u)γλ µ (u) dx,
2 Ω
and
1
Z
G2 (κ(u)) = hαβ λ µ καβ (u)κλ µ (u) dx,
2 Ω
where
Λ(u) = (γ(u), κ(u)) = ({γαβ (u)}, {καβ (u)}),
uα ,β + uβ ,α w,α w,β
γαβ (u) = + ,
2 2
καβ (u) = −wαβ .
Here,
u = (u1 , u2 , w) = (uα , w) ∈ U = W01,2 (Ω; R2 ) ×W02,2 (Ω).
We also define,
F1 (u) = hw, PiL2 + huα , Pα iL2

≡ hu, fiL2 , (17.24)
where
f = (Pα , P) ∈ L2 (Ω; R3 ).
Let J : U → R be defined by
J(u) = G(Λu) − F1 (u), ∀u ∈ U.
inf J(u)
u∈U
≥ sup { inf {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}},
∗ ∗
(17.25)
v∗ ∈A∗ z ∈Y2
where,
K
F(u) = G2 (κ(u)) + hw,α , w,α iL2 , ∀u ∈ U.
2
Moreover, F ∗ : [Y2∗ ]2 → R is defined by,
F ∗ (z∗ , Q) = sup{hz∗α + Qα , w,α iL2 − F(u)}, ∀z∗ ∈ [Y2∗ ]2 .

u∈U
Also,
(v2 )α (v2 )β

1 (v2 )λ (v2 )µ
Z
G(v) = − H (v1 )αβ + (v1 )λ µ + dx
2 Ω αβ λ µ 2 2
K
+ h(v2 )α , (v2 )α iL2 , (17.26)
2
G∗ (z∗ , N)
= sup { inf { hNαβ , (v1 )αβ iL2 + hQα + z∗α , (v2 )α iL2 − G(v)}}
v2 ∈Y2 v1 ∈Y1
1
Z
= (−Nαβ )K z∗α z∗β dx
2 Ω
1
Z
− H αβ λ µ Nαβ Nλ µ dx, (17.27)
2 Ω
if v∗ = (Q, N) ∈ A3 , where
A3 = {v∗ ∈ Y ∗ : {(−Nαβ )K } is positive definite in Ω},

−N11 + K −N12
{(−Nαβ )K } = , (17.28)
−N21 −N22 + K
and
(−Nαβ )K = {(−Nαβ )K }−1 .
Moreover,
A∗ = A1 ∩ A2 ∩ A3 ∩ A4 ∩ A5 ,
where
A1 = {v∗ = (N, Q) ∈ Y ∗ : Nαβ ,β + Pα = 0, in Ω, ∀α ∈ {1, 2}}
and
A2 = {v∗ ∈ Y ∗ : Qα,α + P = 0, in Ω}.
Also,
A4 = {v∗ = (Q, N) ∈ Y ∗ : Jˆ∗ (z∗ ) > 0,

∀z∗ ∈ Y2∗ , such that z∗ =
6 0}, (17.29)
and
∗ ∗ K
A5 = v = (Q, N) ∈ Y : {Nαβ + Kδαβ } ≥ {δαβ } ,
2
where K > 0 is supposed to be such that, in an appropriate matrix sense,
2
{H αβ λ µ } > {δ }.
K αβ
Furthermore,
Jˆ∗ (z∗ ) = −F ∗ (z∗ , 0) + G∗ (z∗ , 0)

1
Z
∗ ∗
= −F (z , 0) + (−Nαβ )K z∗α z∗β dx, ∀z∗ ∈ Y2∗ .
2 Ω
Here,
Y ∗ = Y = L2 (Ω; R2 ) × L2 (Ω; R2×2 ),
Y1∗ = Y1 = L2 (Ω; R2×2 ),
and
Y2∗ = Y2 = L2 (Ω; R2 ),
Finally, denoting
J ∗ (v∗ , z∗ ) = −F ∗ (z∗ , Q) + G∗ (z∗ , N), ∀(v∗ , z∗ ) ∈ A∗ ×Y2∗ ,
and
J˜∗ (v∗ ) = ∗inf ∗ J ∗ (v∗ , z∗ ), ∀v∗ ∈ A∗ ,
z ∈Y2
suppose there exist v∗0 = (N0 , Q0 ) ∈ A∗ , z∗0 ∈ Y2∗ and u0 ∈ U such that
δ {J ∗ (v∗0 , z∗0 ) − h(u0 )α , (N0 )αβ ,β + Pα iL2 − hw0 , (Q0 )α,α + PiL2 } = 0.
J(u0 ) = min J(u)

u∈U
= max J˜∗ (v∗ )
v∗ ∈A∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , z∗0 ). (17.30)
Proof 17.3 Observe that, from the general result in Toland [78], we have
inf J ∗ (v∗ , z∗ ) = inf {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}

z∗ ∈Y2∗ z∗ ∈Y2∗

1
Z
∗ ∗
= inf −F (z , Q) + (−Nαβ )K z∗α z∗β dx
z∗ ∈Y2∗ 2 Ω

1
Z
− H αβ λ µ Nαβ Nλ µ dx
2 Ω
1
≤ −hQα + z∗α , w,α iL2 + F(u) + hz∗α , w,α iL2 + hNαβ − Kδαβ , w,α w,β iL2
2
1
Z
− H N N dx
2 Ω αβ λ µ αβ λ µ
1
= −hQα , w,α iL2 + F(u) − hNαβ − Kδαβ , w,α w,β iL2
2
1
Z
− H N N dx. (17.31)
From this,

∗ ∗ ∗ 1
inf J (v , z ) ≤ −hP, wiL2 + G2 (κ(u)) + sup hN , w,α w,β iL2
z∗ ∈Y2∗ 2 αβ
N∈Y1∗

1
Z
− H N N dx − huα , Nαβ ,β + Pα iL2

1
= −hP, wiL2 + G2 (κ(u)) + sup hNαβ , uα ,β + uβ ,α + w,α w,β iL2
N∈Y1∗ 2

1
Z
− H αβ λ µ Nαβ Nλ µ dx − huα , Pα iL2
2 Ω
= G2 (κ(u)) + G1 (γ(u)) − hw, PiL2 − huα , Pα iL2
= J(u), ∀u ∈ U, v∗ = (Q, N) ∈ A∗ . (17.32)
Thus,
J(u) ≥ inf {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}

z∗ ∈Y2∗
= inf J ∗ (v∗ , z∗ )
z∗ ∈Y2∗
= J˜∗ (v∗ ), (17.33)
∀v∗ ∈ A∗ , u ∈ U.
Summarizing,
J(u) ≥ J˜∗ (v∗ ), ∀u ∈ U, v∗ ∈ A∗ ,
so that,
inf J(u) ≥ sup J˜∗ (v∗ ). (17.34)

u∈U v∗ ∈A∗
Finally, assume v∗0 = (N0 , Q0 ) ∈ A∗ , z∗0 ∈ Y2∗ and u0 ∈ U are such that
δ {J ∗ (v∗0 , z∗0 ) − h(u0 )α , (N0 )αβ ,β + Pα iL2 − hw0 , (Q0 )α,α + PiL2 } = 0.
From the variation in Q∗ we get

∂ F ∗ (z∗0 , Q0 )
= (w0 ),α ,
∂ Qα
so that from this and the variation in z∗ , we get
∂ G∗ (z∗0 , N0 ) ∂ F ∗ (z∗0 , Q0 )
=
∂ z∗α ∂ z∗α
∂ F (z∗0 , Q0 )
∗
=
∂ Qα
= (w0 ),α
K (z∗ ) , in Ω.
= Nαβ (17.35)
0 β
Hence,
F ∗ (z∗0 , Q0 ) = h(z∗0 )α + (Q0 )α , (w0 ),α iL2 − F(u0 ).
Also, from (17.35) we have,
(z∗0 )α = (N0 )αβ (w0 ),β + K(w0 ),α . (17.36)
From such results and the variation in N we obtain

(u0 )α,β + (u0 )β ,α (w0 ),α (w0 ),β
+ − H αβ γ µ (N0 )λ µ = 0,
2 2
so that
(N0 )αβ = Hαβ λ µ γλ µ (u0 ).
From these last results we have,
1
G∗ (z∗0 , N0 ) = h(z∗0 )α , (w0 )α iL2 + (N0 )α β − Kδαβ , (w0 )α (w0 )β L2
2
1
Z
− Hαβ λ µ (N0 )αβ (N0 )λ µ dx
2 Ω
(u0 )α,β + (u0 )β ,α

∗ 1
= h(z0 )α , (w0 )α iL2 + (N0 )αβ , + (w0 )α (w0 )β
2 2 L2
1
Z
−hPα , (u0 )α iL2 − H (N0 )αβ (N0 )λ µ dx
2 Ω αβ λ µ
K
− h(w0 ),α , (w0 ),α iL2
2
K
= h(z∗0 )α , (w0 )α iL2 + G1 (γ(u0 )) − h(w0 ),α , (w0 ),α iL2
2
−hPα , (u0 )α iL2 . (17.37)
J ∗ (v0 , z∗0 ) = −F ∗ (z∗0 , Q0 ) + G∗ (z∗0 , N0 )

K
= F(u0 ) + G1 (γ(u0 )) + h(w0 )α , (w0 )α iL2
2
−hP, w0 iL2 − hPα , (u0 )α iL2
= G1 (γ(u0 )) + G2 (κ(u0 ))
= J(u0 ). (17.38)
From this and from v∗0 ∈ A4 , we have
J(u0 ) = J ∗ (v∗0 , z∗0 ) = ∗inf ∗ J ∗ (v∗0 , z∗ ) = J˜∗ (v∗0 ).

z ∈Y2
Therefore, from such a last equality and (17.34), we may infer that
J(u0 ) = min J(u)

u∈U
= max J˜∗ (v∗ )
v∗ ∈A∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , z∗0 ). (17.39)
17.5 Existence and duality for a non-linear shell model

In this section we present the primal variational formulation concerning the shell model presented in [10].
Indeed, in some sense, the duality principle here presented extends the results developed in [10]. In fact,
through a generalization of some ideas developed in [78], we have obtained a duality principle for which
the membrane force tensor, concerning the global optimal solution of the primal formulation, may not be
necessarily either positive or negative definite. We emphasize details on the Sobolev spaces involved may
be found in [1]. About the fundamental concepts of convex analysis and duality here used, we would cite
[64, 33, 14]. Similar problems are addressed in [14, 76, 77, 43].
At this point we start to describe the shell model in question. Let D ⊂ R2 be an open, bounded, connected
set with a C3 class boundary.
Let S ⊂ R3 be a C3 class manifold, where
S = {r(ξ ) : ξ = (ξ1 , ξ2 ) ∈ D},
r : D ⊂ R2 → R3 is a C3 class function and where,
r(ξ ) = X1 (ξ )e1 + X2 (ξ )e2 + X3 (ξ )e3 .
Here, {e1 , e2 , e3 } is the canonical basis of R3 .

We assume S is the middle surface of a shell of constant thickness h so that we denote,
∂r
aα = , ∀α ∈ {1, 2},
∂ ξα
aαβ = aα · aβ .
Let
aα × aβ
n= ,
|aα × aβ |
be the unit normal to S, so that we define the covariant components bαβ of the curvature tensor
b = {bαβ },
by
bαβ = n · aα,β = n · r,αβ .
Observe that
n · aα = 0,
so that
nβ · aα + n · aα,β = 0,
and thus we obtain,
bαβ = n · aα,β = −nβ · aα .
The Christofell symbols relating S, would be,
1
Γαβ γ = (aαβ ,γ + aαγ,β − aβ γ,α ), ∀α, β , γ ∈ {1, 2}
2
and
Γλαβ = aλ γ Γγαβ , ∀α, β , λ ∈ {1, 2},
where
{aαβ } = {aαβ }−1 .
Let us denote with a bar the quantities relating the deformed middle surface.
So, the middle surface strain tensor γ = {γαβ } is given by
(āαβ − aαβ )
γαβ = ,
2
while the tensor relating change in curvature is given by
καβ = −(b̄αβ − bαβ ), ∀α, β ∈ {1, 2}.
We also denote,
U = {u = (uα , w) = (u1 , u2 , w) ∈ W 1,2 (S; R2 ) ×W 2,2 (S) :

∂w
uα = w = = 0, on ∂ S}
∂n
= W01,2 (S; R2 ) ×W02,2 (S), (17.40)
where, in order to simplify the analysis, the boundary conditions in question refer to a clamped shell and
where ∂ S denotes the boundary of S.
Also from reference [10], for moderately large rotations around tangents, the strain displacements rela-
tions are given by,
1.
1
γαβ (u) = θαβ (u) + ϕα (u)ϕβ (u),
2
2.
καβ (u) = −w|αβ − bλα|β uλ

−bλα uλ |β − bλβ uλ |α + bλα bλ β w, (17.41)
where,
uα |β = uα,β − Γλαβ uλ ,
w|αβ = w,αβ − Γλαβ w,λ ,

1
θαβ (u) = (uα |β + uβ |α ) − bαβ w,
2
β
ϕα (u) = w,α + bα uβ , (17.42)
and
β
bα = bαλ aλ β .
The primal shell inner energy is defined by
1
Z
W (γ(u), κ(u)) = H αβ λ µ γαβ (u)γλ µ (u) dS
2 S
1
Z
+ hαβ λ µ καβ (u)κλ µ (u) dS, (17.43)
2 S
where
αβ λ µ Eh αλ β µ αµ βλ 2ν αβ λ µ
H = a a +a a + a a ,
2(1 + ν) 1−ν
h2 αβ λ µ
hαβ λ µ =
H ,
12
and E denotes the Young’s modulus and ν the Poisson ratio.
The constitutive relations are,
N αβ = H αβ λ µ γλ µ (u),
and
M αβ = hαβ λ µ κλ µ (u),
where {N αβ } is the membrane force tensor and {M αβ } is the moment one.
We assume
H = {H αβ λ µ }
to be positive definite in the sense that there exists c0 > 0 such that
H αβ λ µ tαβ tλ µ ≥ c0tαβ tαβ ≥ 0, ∀t = {tαβ } ∈ R4 .
Finally, the primal variational formulation for this model will be given by
J :U →R
where,
J(u) = W (γ(u), κ(u)) − hu, fiL2 ,
W (γ(u), κ(u)) = G1 (γ(u)) + G2 (κ(u)),
1
Z
G1 (γ(u)) = H αβ λ µ γαβ (u)γλ µ (u) dS,
2 S
1
Z
G2 (κ (u)) = hα β λ µ καβ (u)κλ µ (u) dS,
2 S
and Z
hu, fiL2 = (Pα uα + Pw) dS.
S
Here
f = (Pα , P) ∈ L2 (S; R3 ),
are the external loads distributed on S, Pα relating the directions aα and P relating the direction n, respec-
tively.
Moreover, generically for f1 , f2 ∈ L2 (S), we denote,
Z
h f1 , f2 iL2 = f1 f2 dS,
S
√
where dS = a dξ1 dξ2 , and a = det{aαβ }.
17.6 Existence of a minimizer

Theorem 17.6.1 Consider the statements and assumptions concerning the shell model described in the last
section.
More specifically, consider the functional J : U → R given, as above described, by
J(u) = W (γ(u), κ(u)) − hu, fiL2 ,

where,
U = {u = (uα , w) = (u1 , u2 , w) ∈ W 1,2 (S; R2 ) ×W 2,2 (S) :

∂w
uα = w = = 0, on ∂ S}
∂n
= W01,2 (S; R2 ) ×W02,2 (S). (17.44)
Let T0 be such that,

kT0 k22 = min {kT k22 },
T ∈L2 (S;R2×2 )
subject to √
( gTαβ ),β
√ + Γαλ β Tλ β + Pα = 0 in D.
g
Assume
J1 (u) → +∞, as hw,αβ , w,αβ iL2 + huα,β , uα,β iL2 → +∞, (17.45)
where
1
J1 (u) = G2 (κ(u)) + h(T0 )αβ , ϕα (u)ϕβ (u)i − h(T0 )αβ bαβ + P, wiL2 .
2
Under such hypotheses, there exists u0 ∈ U such that,
J(u0 ) = min{J(u)}.
u∈U
Proof 17.4
Observe that
Z √
( g(T0 )αβ ),β

α √
hPα , uα iL2 = − √ + Γλ β (T0 )λ β uα g dξ
D g
uα,β + uβ ,α

λ
= (T0 )αβ , − Γαβ uλ
2 L2
= h(T0 )αβ , θαβ (u) + bαβ wiL2 , (17.46)
so that defining
1
vαβ (u) = θαβ (u) + ϕα (u)ϕβ (u), (17.47)
2
we obtain
1
hPα , uα iL2 = (T0 )αβ , vαβ (u) − ϕα (u)ϕβ (u) + bαβ w .
2 L2
Thus,
J(u) = G1 (v(u)) − h(T0 )αβ , vαβ (u)iL2

1
+G2 (κ(u)) + h(T0 )αβ , ϕα (u)ϕα (u)iL2
2
−h(T0 )αβ bαβ + P, wiL2 , (17.48)
From this and hypothesis (17.45), J is bounded below. So, there exists α1 ∈ R such that
α1 = inf J(u).
u∈U
Let {un } be a minimizing sequence for J.

From (17.48) and also from the hypotheses (17.45), {kwn k2,2 } and {k(un )α k1,2 } are bounded.
So there exists w0 ∈ W02,2 (S) and {(u0 )α } ∈ W01,2 (S; R2 ) such that, up to a not relabeled subsequence,
(wn ),αβ * (w0 ),αβ weakly in L2 (S), ∀α, β ∈ {1, 2},
(wn ),α → (w0 ),α strongly in L2 (S), ∀α ∈ {1, 2}, (17.49)
(un )α,β * (u0 )α,β weakly in L2 (S), ∀α, β ∈ {1, 2}, (17.50)
(un )α → (u0 )α strongly in L2 (S), ∀α ∈ {1, 2}. (17.51)
k(vn )αβ (u)k2 < K2 , ∀α, β ∈ {1, 2}, n ∈ N,
and thus, from this, (17.42), (17.47), (17.49) and (17.51), we may infer that there exists K3 > 0 such that
kθαβ (un )k2 < K3 , ∀α, β ∈ {1, 2}, n ∈ N.
Thus, from this, (17.51) and (17.50), up to a subsequence not relabeled,
θαβ (un ) * θαβ (u0 ), weakly in L2 ,
∀α, β ∈ {1, 2}, as n → ∞.

From this, also up to a subsequence not relabeled, we have
∀α, β ∈ {1, 2}, and

∀α, β ∈ {1, 2}.
inf J(u) = α1
u∈U
= lim inf J(un )
n→∞
≥ J(u0 ). (17.52)
Thus,
J(u0 ) = min J(u).
u∈U

17.7 The duality principle for the shell model

At this point we present the main duality principle, which is summarized by the next theorem.
Theorem 17.7.1 Consider the statements and assumptions concerning the shell model described in the last
two sections.
More specifically, consider the functional J : U → R given, as above described by,
J(u) = W (γ(u), κ(u)) − hu, fiL2 ,

where,
U = {u = (uα , w) = (u1 , u2 , w) ∈ W 1,2 (S; R2 ) ×W 2,2 (S) :

∂w
uα = w = = 0, on ∂ S}
∂n
= W01,2 (S; R2 ) ×W02,2 (S). (17.53)
inf J(u)
u∈U
≥ sup { ∗inf ∗ {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}}, (17.54)
v∗ ∈A∗ z ∈Y2
where,
K
F(u) = G2 (κ(u)) + hϕα (u), ϕα (u)iL2 , ∀u ∈ U.
2
Moreover, F ∗ : [Y2∗ ]2 → R is defined by,
F ∗ (z∗ , Q) = sup{hz∗α + Qα , ϕα (u)iL2 − F(u)}, ∀(z∗ , Q) ∈ [Y2∗ ]2 .

u∈U
Also,
(v2 )α (v2 )β

1 (v2 )λ (v2 )µ
Z
G(v) = − H (v1 )αβ + (v1 )λ µ + dS
2 S αβ λ µ 2 2
K
+ h(v2 )α , (v2 )α iL2 , (17.55)
2
G∗ (z∗ , N ) = sup { inf {hN αβ , (v1 )αβ iL2 + hz∗α , (v2 )α iL2 − G(v)}}
v2 ∈Y2 v1 ∈Y1
1
Z
= (−N α β )K zα∗ z∗β dS
2 S
1
Z
− H αβ λ µ N αβ N λ µ dS, (17.56)
2 S
if v∗ = (Q, N) ∈ A3 , where
A3 = {v∗ ∈ Y ∗ : {(−N αβ )K } is positive definite in Ω},
−N 11 + K −N 12

αβ
{(−N )K } = , (17.57)
−N 21 22
−N + K
and
(−N αβ )K = {(−N αβ )K }−1 .
Moreover, defining
Y ∗ = Y = L2 (S; R2 ) × L2 (S; R2×2 ),
Y1∗ = Y1 = L2 (S; R2×2 ),
and
Y2∗ = Y2 = L2 (S; R2 ),
also,
A∗ = A1 ∩ A2 ∩ A3 ∩ A4 ∩ A5 ,
where,
A1 = {v∗ = (Q, N) ∈ Y ∗ : −N αβ |β + bλα Qλ − Pα = 0, in S}, (17.58)
A2 = {v∗ = (Q, N ) ∈ Y ∗ : −bαβ N αβ − Q|α

α
− P = 0, in S}. (17.59)
Moreover,
{H αβ λ µ } = {H αβ λ µ }−1 ,
and,
{hαβ λ µ } = {hαβ λ µ }−1 .
Also,
A4 = {v∗ = (Q, N) ∈ Y ∗ : Jˆ∗ (z∗ ) > 0,

∀z∗ ∈ Y2∗ , such that z∗ =
6 0}, (17.60)
and
∗ ∗ K
A5 = v = (Q, N) ∈ Y : {Nαβ + Kδαβ } ≥ {δαβ } ,
2
where K > 0 is supposed to be such that, in an appropriate matrix sense,
2
{H αβ λ µ } > {δ }.
K αβ
Furthermore,
Jˆ∗ (z∗ ) = −F ∗ (z∗ , 0) + G∗ (z∗ , 0)

1
Z
= −F ∗ (z∗ , 0) + (−N αβ )K z∗α z∗β dS, ∀z∗ ∈ Y2∗ .
2 S
Finally, denoting
J ∗ (v∗ , z∗ ) = −F ∗ (z∗ , Q) + G∗ (z∗ , N), ∀(v∗ , z∗ ) ∈ A∗ ×Y2∗ ,
and
J˜∗ (v∗ ) = ∗inf ∗ J ∗ (v∗ , z∗ ), ∀v∗ ∈ A∗ ,
z ∈Y2
suppose there exist v∗0 = (N0 , Q0 ) ∈ A∗ , z∗0 ∈ Y2∗ and u0 ∈ U such that
δ {J ∗ (v∗0 , z∗0 ) − hθαβ (u0 ), (N0 )α β iL2

+h(u0 )α , Pα iL2 − h(ϕ0 )α , (Q0 )α iL2 + hw0 , PiL2 } = 0. (17.61)
J(u0 ) = min J(u)

u∈U
= max J˜∗ (v∗ )
v∗ ∈A∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , z∗0 ). (17.62)
Proof 17.5
Observe that, from the general result in Toland [78], we have
inf J ∗ (v∗ , z∗ ) = inf {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}

z∗ ∈Y2∗ z∗ ∈Y2∗

1
Z
∗ ∗
= ∗inf ∗ −F (z , Q) + (−N α β )K z∗α z∗β ; dx
z ∈Y2 2 S

1
Z
αβ λ µ
− H N N dS
2 S αβ λ µ
≤ −hQα + z∗α ; , ϕα (u)iL2 + F(u)
1
+hz∗α , ϕα (u)iL2 + hN α β − Kδαβ , ϕα (u)ϕβ (u)iL2
2
1
Z
αβ λ µ
− H N N dS
2 S αβ λ µ
1
= −hQα , ϕα (u)iL2 + F(u) − hN α β − Kδαβ , ϕα (u)ϕβ (u)iL2
2
1
Z
αβ λ µ
− H N N dS (17.63)
2 Ω αβ λ µ
From this, we have

1 αβ
inf ∗ J ∗ (v∗ , z∗ ) ≤ −hP, wiL2 + G2 (κ(u)) + sup hN , ϕα (u)ϕβ (u)iL2
∗
z ∈Y2 N ∈Y1∗ 2

1
Z
αβ λµ αβ
− Hαβ λ µ NdS − hθαβ (u), N iL2 − huα , Pα iL2 iL2
N
2 Ω

1 αβ
= −hP, wiL2 + G2 (κ(u)) + sup hN , θαβ (u) + ϕα (u)ϕβ (u)iL2
N ∈Y1∗ 2

1
Z
αβ λ µ
− H N N dS − huα , Pα iL2
2 S αβ λ µ
= G(κ(u)) + G1 (γ(u)) − hw, PiL2 − huα , Pα iL2
= J(u), ∀u ∈ U, v∗ = (Q, N) ∈ A∗ . (17.64)
Thus,
J(u) ≥ inf {−F ∗ (z∗ , Q) + G∗ (z∗ , N)}

z∗ ∈Y2∗
= inf J ∗ (v∗ , z∗ )
z∗ ∈Y2∗
= J˜∗ (v∗ ), (17.65)
∀v∗ ∈ A∗ , u ∈ U.
Summarizing,
J(u) ≥ J˜∗ (v∗ ), ∀u ∈ U, v∗ ∈ A∗ ,
so that,
inf J(u) ≥ sup J˜∗ (v∗ ). (17.66)
u∈U v∗ ∈A∗
Finally, assume v0∗ = (N0 , Q0 ) ∈ A∗ , z∗0 ∈ Y2∗ and u0 ∈ U are such that
δ {J ∗ (v∗0 , z∗0 ) − hθαβ (u0 ), (N0 )αβ iL2

+h(u0 )α , Pα iL2 − h(ϕ0 )α , (Q0 )α iL2 + hw0 , PiL2 } = 0. (17.67)
From the variation in Q∗ we get

∂ F ∗ (z∗0 , Q0 )
= ϕα (u0 ), (17.68)
∂ Qα
so that from this and the variation in z∗ , we get
∂ G∗ (z∗0 , Q) ∂ F ∗ (z∗0 , Q0 )
=
∂ z∗α ∂ z∗α
∂ F (z∗0 , Q0 )
∗
=
∂ Qα
= ϕα (u0 )
= NK (z∗0 )β , in Ω.
αβ
(17.69)
Hence,
F ∗ (z∗0 , Q0 ) = h(z∗0 )α + (Q0 )α , ϕα (u0 )iL2 − F(u0 ). (17.70)
Also, from (17.69) we have,
(z∗0 )α = (N0 )αβ ϕβ (u0 ) + Kϕα (u0 ). (17.71)
From such results and the variation in N we obtain

(u0 )α,β + (u0 )β ,α ϕα (u0 )ϕβ (u0 )
+ − H α β γ µ (N0 )λ µ = 0,
2 2
so that
(N0 )αβ = Hαβ λ µ γλ µ (u0 ).
From these last results we obtain,
1D E
G∗ (z∗0 , N0 ) = h(z∗0 )α , ϕα (u0 )α iL2 + (N0 )αβ − Kδαβ , ϕα (u0 )ϕβ (u0 ) 2
2 L
1
Z
− Hαβ λ µ (N0 )α β (N0 )λ µ dS
2 Ω

1
= h(z∗0 )α , ϕα (u0 )α iL2 + (N0 )αβ , θαβ (u0 ) + ϕα (u0 )ϕβ (u0 )
2 L2
1 K
Z
−hPα , (u0 )α iL2 − H (N0 )α β (N0 )λ µ dS − hϕα (u0 ), ϕα (u0 )iL2
2 Ω αβ λ µ 2
∗ K
= h(z0 )α , ϕα (u0 )iL2 + G1 (γ(u0 )) − hϕα (u0 ), ϕα (u0 )iL2
2
−hPα , (u0 )α iL2 . (17.72)
J ∗ (v0 , z∗0 ) = −F ∗ (z∗0 , Q0 ) + G∗ (z∗0 , N0 )

K
= F(u0 ) + G1 (γ(u0 )) + hϕα (u0 ), ϕα (u0 )iL2
2
= G1 (γ(u0 )) + G2 (κ(u0 ))
= J(u0 ). (17.73)
From this and from v∗0 ∈ A4 , we obtain
J(u0 ) = J ∗ (v∗0 , z∗0 ) = ∗inf ∗ J ∗ (v∗0 , z∗ ) = J˜∗ (v∗0 ).

z ∈Y2
Therefore, from such a last equality and (17.66), we may infer that
J(u0 ) = min J(u)

u∈U
= max J˜∗ (v∗ )
v∗ ∈A∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , z∗0 ). (17.74)
17.8 Conclusion
In this article, in a first step, we have developed new proofs of global existence of minimizers for the
Kirchhoff-Love plate and a shell model presented in [10].
In a second step, we have developed duality principles for these same models. In [10], the authors devel-
oped a duality principle valid for the special case in which the membrane force tensor at a critical point is
positive definite. We have generalized such a result, considering that in our approach, such a previous case
is included but here we do not request the optimal membrane force to be either positive or negative definite.
Thus, in some sense, we have complemented the important work developed in [10]. We would emphasize,
sufficient optimality conditions are presented and the results here developed are applicable to a great variety
of problems, including other shell models.
Chapter 18
A Primal Dual Formulation and a

Multi-Duality Principle for a
Non-Linear Model of Plates
18.1 Introduction
In this chapter we develop a new primal dual variational formulation for the Kirchhoff-Love non-linear plate
model. We emphasize the results here presented may be applied to a large class of non-convex variational
problems.
Let Ω ⊂ R2 be an open, bounded, connected set which represents the middle surface of a plate of thickness
h. The boundary of Ω, which is assumed to be regular (Lipschitzian), is denoted by ∂ Ω. The vectorial basis
related to the cartesian system {x1 , x2 , x3 } is denoted by (aα , a3 ), where α = 1, 2 (in general Greek indices
stand for 1 or 2), and where a3 is the vector normal to Ω, whereas a1 and a2 are orthogonal vectors parallel
to Ω. Also, n is the outward normal to the plate surface.
The displacements will be denoted by
û = {ûα , û3 } = ûα aα + û3 a3 .
The Kirchhoff-Love relations are
ûα (x1 , x2 , x3 ) = uα (x1 , x2 ) − x3 w(x1 , x2 ),α

and û3 (x1 , x2 , x3 ) = w(x1 , x2 ). (18.1)
Here −h/2 ≤ x3 ≤ h/2 so that we have u = (uα , w) ∈ U where
(uα , w) ∈ W 1,2 (Ω; R2 ) ×W 2,2 (Ω),

U =
∂w
uα = w = = 0 on ∂ Ω}
∂n
= W01,2 (Ω; R2 ) ×W02,2 (Ω).
It is worth emphasizing that the boundary conditions here specified refer to a clamped plate.
We define the operator Λ : U → Y1 ×Y1 , where Y1 = Y1∗ = L2 (Ω; R2×2 ), by

Λ(u) = {γ(u), κ(u)},
γαβ (u) = + ,
2 2
καβ (u) = −w,αβ .
The constitutive relations are given by
Nαβ (u) = Hαβ λ µ γλ µ (u), (18.2)
Mαβ (u) = hαβ λ µ κλ µ (u), (18.3)
h2
where: {Hαβ λ µ } and {hαβ λ µ = 12 Hαβ λ µ }, are symmetric positive definite fourth order tensors. From now
on, we denote {H αβ λ µ } = {Hαβ λ µ }−1 and {hαβ λ µ } = {hαβ λ µ }−1 .
Furthermore {Nαβ } denote the membrane stress tensor and {Mαβ } the moment one. The plate stored
energy, represented by (G ◦ Λ) : U → R is expressed by
1 1
Z Z
(G ◦ Λ)(u) = Nαβ (u)γαβ (u) dx + Mαβ (u)καβ (u) dx (18.4)
2 Ω 2 Ω
and the external work, represented by F1 : U → R, is given by

F1 (u) = hw, PiL2 (Ω) + huα , Pα iL2 (Ω) , (18.5)
where P, P1 , P2 ∈ L2 (Ω) are external loads in the directions a3 , a1 and a2 respectively. The potential energy,
denoted by J : U → R is expressed by:
J(u) = (G ◦ Λ)(u) − F1 (u)
Finally, we also emphasize from now on, as their meaning are clear, we may denote L2 (Ω) and
L2 (Ω; R2×2 ) simply by L2 , and the respective norms by k · k2 . Moreover derivatives are always understood
in the distributional sense, 0 may denote the zero vector in appropriate Banach spaces and, the following and
relating notations are used:
∂ 2w
w,αβ = ,
∂ xα ∂ xβ
∂ uα
uα,β = ,
∂ xβ
∂ Nαβ
Nαβ ,1 = ,
∂ x1
and
∂ Nαβ
Nαβ ,2 =
.
∂ x2
Here we emphasize the general Einstein convention of sum of repeated indices holds throughout the text,
unless otherwise indicated.
Remark 18.1 About the references, details on the Sobolev involved may be found in [1]. Mandatory
references are the original results of Telega and his co-workers in [10, 11, 76, 77]. About convex analysis,
the results here developed follow in some extent [14], for which the main references are [33, 78].
We emphasize, our results complement, in some sense, the original ones presented in [10, 11, 76, 77].
Finally, existence results for models in elasticity including the plate model here addressed are developed
in [29, 30, 31]. Similar problems are addressed in [39, 43].
A Primal Dual Formulation and a Multi-Duality Principle for a Non-Linear Model of Plates 393
18.2 The first duality principle

Theorem 18.2.1 Let Ω ⊂ R2 be an open, bounded, connected set with a regular (Lipschitzian) boundary
denoted by ∂ Ω. Let U = U1 × U2 where U1 = W01,2 (Ω; R2 ) and U2 = W02,2 (Ω). We recall that Y1 = Y1∗ =
L2 (Ω; R2×2 ) and define Y2 = Y2∗ = L2 (Ω; R2 ),
1 1
γαβ (u) = (uα,β + uβ ,α ) + w,α w,β ,
2 2
καβ (u) = −w,αβ , ∀u ∈ U, α, β ∈ {1, 2}

and J : U → R by,
J(u) = G1 (γ(u)) + G2 (κ(u)) − hu, f iL2 ,
where
1
Z
G1 (γ(u)) = H γ (u)γλ µ (u) dx,
2 Ω αβ λ µ αβ
1
Z
G2 (κ(u)) = h κ (u)κλ µ (u) dx,
and
hu, f iL2 = hw, PiL2 + huα , Pα iL2 ,
where
f = (P1 , P2 , P) ∈ L2 (Ω; R3 ).
In the next lines we shall denote
w = −(hαβ λ µ D∗λ µ Dαβ )−1 ( div Q + div z∗ − P),
if
div Q + div z∗ − P = −hαβ λ µ w,αβ λ µ
and
w ∈ U2 .
Define also G∗2 : Y2∗ ×Y2∗ → R, by
G∗2 (z∗ , Q)
= sup{hw,α , z∗α + Qα iL2 − G2 (κ(u)) + hP, wiL2 }
w∈U
1
Z
= {(hαβ λ µ D∗λ µ Dαβ )−1 ( div Q + div z∗ − P)}{( div Q + div z∗ − P)} dx,
2 Ω
and G̃∗1 : Y2∗ ×Y1∗ → R ∪ {+∞} where
G̃∗1 (−Q, N) =

sup −h(v2 )α , Qα iL2 + h(v1 )αβ , Nαβ iL2
(v1 ,v2 )∈Y1 ×Y2

1 1
Z
−G1 (v1 )αβ + (v2 )α (v2 )β − Kαβ (v2 )α (v2 )β dx
2 2 Ω
1
Z
= Nαβ + Kαβ Qα Qβ dx
2 Ω
1
Z
+ H αβ λ µ Nαβ Nλ µ dx (18.6)
2 Ω
if
{Nαβ + Kαβ }
is positive definite.
Here we have denoted,
{Nαβ + Kαβ } = {Nαβ + Kαβ }−1 .
We denote also,

1
Z
F ∗ (z∗ ) = sup h(v2 )α , (z∗2 )α iL2 − K (v2 )α (v2 )β dx
v2 ∈Y2 2 Ω αβ
1
Z
= K (z∗ )α (z∗2 )β dx
2 Ω αβ 2

1
Z
= −Nαβ + εδαβ (z∗2 )α (z∗2 )β dx , (18.7)
2 Ω
if {Kαβ } is positive definite, where
{K αβ } = {Kαβ }−1 ,
{Kαβ } = {−Nαβ + εδαβ },
{H αβ λ µ } = {Hαβ λ µ }−1 ,
and
1
Z
F({wα }) = Kαβ w,α w,β dx,
2 Ω
for some
ε > 0.
At this point we also define,
B∗ = {N ∈ Y1∗ : {Kαβ } is positive definite and Jˆ∗ (N, z∗ ) > 0, ∀z∗ ∈ Y2∗ such that z∗ =
6 0}
where,
Jˆ∗ (N, z∗ ) = F ∗ (z∗ ) − G∗2 (z∗ , 0)
1
Z
= (−Nαβ + εδαβ )z∗α z∗β dx
2 Ω
1
Z
− [(h D∗ D )−1 ( div z∗ )][ div z∗ ] dx. (18.8)
2 Ω αβ λ µ λ µ αβ
Moreover, we denote,
C∗ = {N ∈ Y1∗ : Nαβ ,β + Pα = 0, in Ω}
and define
A∗ = B∗ ∩C∗ .
Assume u0 ∈ U is such that δ J(u0 ) = 0 and N0 ∈ B∗ , where
{(N0 )αβ } = {Hαβ λ µ γλ µ (u0 )}.
J(u0 ) = inf J(u)
u∈U
= sup J˜∗ (Q, N)
(Q,N)∈Y2∗ ×A∗
= J˜∗ (Q0 , N0 )
= J ∗ (z∗0 , Q0 , N0 ), (18.9)
where
J ∗ (z∗ , Q, N) = F ∗ (z∗ ) − G∗2 (z∗ , Q) − G̃∗1 (−Q, N),
J˜∗ (Q, N) = ∗inf ∗ J ∗ (z∗ , Q, N),
z ∈Y2
and where
(z∗0 )α = (−(N0 )αβ + εδαβ )(w0 ),β ,
and
Q0 = −ε∇w0 .
Proof 18.1 From the general result in Toland [78], we have
inf J ∗ (z∗ , Q, N)
z∗ ∈Y2∗
= inf {F ∗ (z∗ ) − G∗2 (z∗ , Q) − G̃∗1 (−Q, N)}

z∗ ∈Y2∗
1
Z
≤ −hw,α , z∗α iL2 − hw,α , Qα iL2 + hαβ λ µ w,αβ w,λ µ dx
2 Ω
1
Z
+hw,α , z∗α iL2 − (−Nαβ + εδαβ )w,α w,β dx − hw, PiL2
2 Ω
1
Z
− δ Qα Qβ dx
2ε Ω αβ
1
Z
− H N N dx − huα , Nαβ ,β + Pα iL2
1
Z
≤ −hw,α , z∗α iL2 − hw,α , Qα iL2 + h w w dx
1
Z
+hw,α , z∗α iL2 − (−Nαβ + εδαβ )w,α w,β dx − hw, PiL2
2 Ω
1
Z
+hw,α , Q∗α iL2 + εδ w,α w,β d x
2 Ω αβ
1
Z
− H N N dx − huα , Nαβ ,β + Pα iL2 , (18.10)
∀u ∈ U, Q ∈ Y2∗ , N ∈ A∗ .
Therefore,
inf J ∗ (z∗ , Q, N)
z∗ ∈Y2∗
1 1
Z Z
≤ hαβ λ µ w,αβ w,λ µ dx + Nαβ w,α w,β dx
Ω2 2
Ω Z
uα,β + uβ ,α

1
+ , Nαβ − H N N dx
2 L 2 2 Ω αβ λ µ αβ λ µ
−hw, PiL2 − huα , Pα iL2
uα,β + uβ ,α 1

1
Z
≤ sup + w,α w,β , Nαβ − H αβ λ µ Nαβ Nλ µ dx
N∈Y1∗ 2 2 L2 2 Ω
1
Z
+ h w w dx
1 1
Z Z
= γαβ (u)γλ µ (u) dx + h κ (u)κλ µ (u) dx
2 Ω 2 Ω αβ λ µ αβ
= J(u), (18.11)
∀u ∈ U, Q ∈ Y2∗ ,
N ∈ A∗ .
Summarizing,
J(u) ≥ inf J ∗ (z∗ , Q, N)
z∗ ∈Y2∗
= J˜∗ (Q, N) (18.12)

∀u ∈ U, Q ∈ Y2∗ , N ∈ A∗ , so that
inf J(u) ≥ sup J˜∗ (Q, N). (18.13)
u∈U (Q,N)∈Y2∗ ∩A∗
Suppose now u0 ∈ U is such that

δ J(u0 ) = 0, (18.14)
and N0 ∈ B∗ .
Observe that from (18.14), we get
(N0 )αβ ,β + Pα = 0, in Ω,
so that
N0 ∈ C∗ .
Hence,
N0 ∈ A∗ = B∗ ∩C∗ .
Moreover, from δ J(u0 ) = 0, we obtain
− div (Q0 + z∗0 ) = hαβ λ µ (w0 )αβ λ µ − P in Ω, (18.15)
where, as above indicated

(z∗0 )α = (−(N0 )αβ + εδαβ )(w0 ),β . (18.16)
and
Q0 = −ε∇w0 .
From (18.15),
w0 = −(hαβ λ µ D∗λ µ Dαβ )−1 ( div(z∗0 + Q0 ) − P),
so that from this and the inversion of (18.16), we have
(w0 ),ρ = [(hαβ λ µ D∗λ µ Dαβ )−1 (− div(z∗0 + Q0 ) + P)],ρ

= ((−N0 )ρβ + εδρβ )(z∗0 )β , (18.17)
so that
[(hαβ λ µ D∗λ µ Dαβ )−1 ( div(z∗0 + Q0 ) − P)],ρ + ((−N0 )ρβ + εδρβ )(z∗0 )β = 0, in Ω,
that is,
∂ Jˆ1∗ (z∗0 , Q0 , w0 , u0 )
= 0,
∂ z∗
where
Jˆ1∗ (z∗ , Q, N, u) = J ∗ (z∗ , Q, N) + huα , Nαβ ,β + Pα iL2 .
Also, from (18.15) and (18.16) we obtain
− div Q0 = ε div ∇w0

= ε∇2 w0
= εδαβ w,αβ
= hαβ λ µ (w0 )αβ λ µ − [((N0 )αβ − εδαβ )(w0 ),β ],α − P
= hαβ λ µ (w0 )αβ λ µ + div z∗0 − P. (18.18)
Furthermore,
(Q0 )ρ
= −(w0 ),ρ
ε
= [(hαβ λ µ D∗λ µ Dαβ )−1 ( div(z∗0 + Q0 ) − P)],ρ . (18.19)
This last equation corresponds to

∂ Jˆ1∗ (z∗0 , Q0 , N0 , u0 )
= 0.
∂ Qα
Moreover,
(u0 )α,β + (u0 )β ,α

1
H αβ λ µ (N0 )λ µ = + (w0 ),α (w0 ),β ,
2 2
so that
H αβ λ µ (N0 )λ µ
(u0 )α,β + (u0 )β ,α

=
2

1
+ (−(N0 )αρ + εδαρ )(z∗0 )ρ (−(N0 )β η + εδβ η )(z∗0 )η , (18.20)
2
which means
∂ Jˆ1∗ (z∗0 , Q0 , N0 , u0 )
= 0.
∂ Nαβ
Finally, from
Nαβ ,β + Pα = 0, in Ω,
we get
∂ Jˆ1∗ (z∗0 , Q0 , N0 , u0 )
= 0.
∂ uα
Summarizing, we have obtained

δ Jˆ1∗ (z∗0 , Q0 , N0 , u0 ) = 0.
At this point we shall obtain a standard correspondence between the primal and dual formulations.
First, we recall that from
(z∗0 )α = (−Nαβ + εδαβ )(w0 ),β ,
we have
1
F ∗ (z∗0 ) = h(w0 )α , (z∗0 )α iL2 − (−Nαβ + εδαβ )(w0 ),α (w0 ),β .
2
From
div Q0 + div z∗0 = −hαβ λ µ (w0 )αβ λ µ + P
we obtain
G∗2 (z∗0 , Q0 ) = h(w0 )α , (Q0 )α iL2 + h(w0 )α , (z∗0 )α iL2 − G2 (κ(u0 )) + hw0 , PiL2 .
Finally, from
−(Q0 )α
= (w0 ),α
ε
and
(N0 )αβ = Hαβ λ µ γλ µ (u0 ),
we get
(u0 )α,β + (u0 )β ,α (w0 ),α (w0 ),β

G˜ 1∗ (−Q0 , N0 ) = −h(w0 )α , (Q0 )α iL2 + + , (N0 )αβ
2 2 L2
Z
ε
−G1 (γ(u0 )) − δαβ (w0 ),α (w0 ),β dx
2 Ω
(w0 ),α (w0 ),β

= −h(w0 )α , (Q0 )α iL2 + , (N0 )αβ
2 L2
Z
ε
− δ (w0 ),α (w0 ),β dx + huα , Pα iL2
2 Ω αβ
1
Z
− H γ (u0 )γλ µ (u0 ) dx. (18.21)
Jˆ1∗ (z∗0 , Q0 , N0 , u0 ) = J ∗ (z0 , Q0 , N0 )
= F ∗ (z∗0 ) − G∗2 (z∗0 , Q0 ) − G̃∗1 (−Q0 , N0 )
= G2 (κu0 ) + G1 (γ(u0 )) − hw0 , PiL2 − h(u0 )α , Pα iL2
= J(u0 ). (18.22)
Moreover, since N0 ∈ A∗ , we have
J ∗ (z∗0 , Q0 , N0 ) = inf J ∗ (z∗ , Q0 , N0 )
z∗ ∈Y2∗
= J˜∗ (Q0 , N0 ). (18.23)

From this, (26.34) and (18.22), we obtain
J(u0 ) = inf J(u)
u∈U
= sup J˜∗ (Q, N)
(Q,N)∈Y2∗ ×A∗
= J˜∗ (Q0 , N0 )
= J ∗ (z∗0 , Q0 , N0 ), (18.24)
18.3 The primal dual formulation and related duality principle

At this point we present the main result of this article, which is summarized by the next theorem.
Theorem 18.3.1 Consider the notation and context of the last theorem. Assume those hypotheses, more
specifically suppose δ J(u0 ) = 0 and N0 ∈ B∗ , where
{(N0 )αβ } = {Hαβ λ µ γλ µ (u0 )},
(z∗0 )α = (−(N0 )αβ + εδαβ )(w0 ),β ,
and
Q0 = −ε∇w0 .
Recall also that
B∗ = {N ∈ Y1∗ : {Kαβ } is positive definite and Jˆ∗ (N, z∗ ) > 0, ∀z∗ ∈ Y2∗ such that z∗ =
6 0}
where,
Jˆ∗ (N, z∗ ) = F ∗ (z∗ ) − G∗2 (z∗ , 0)
1
Z
= (−Nαβ + εδαβ )z∗α z∗β dx
2 Ω
1
Z
− [(h D∗ D )−1 ( div z∗ )][ div z∗ ] dx. (18.25)
2 Ω αβ λ µ λ µ αβ
Moreover,
C∗ = {N ∈ Y1∗ : Nαβ ,β + Pα = 0, in Ω}
and
A∗ = B∗ ∩C∗ .
Under such assumptions and notation, denoting
Ŷ2∗ = {Q ∈ Y2∗ : Q = ∇v, for some v ∈ W02,2 (Ω)},
we have
J(u0 ) = inf J(u)
u∈U
= sup J˜∗ (Q, N)
(Q,N)∈Ŷ2∗ ×A∗
= J˜∗ (Q0 , N0 )
= J ∗ (z∗0 , Q0 , N0 )
= J3 (w0 , N0 )
= sup J3 (w, N) (18.26)
(w,N)∈U×A∗
where,
1
Z
J3 (w, N) = − h w w dx
1
Z
− (N − εδαβ )w α w,β dx
2 Ω αβ
1
Z
(−∇2 )−1 hαβ λ µ w,αβ λ µ − [(Nαβ − ε δα β )w,β ],α − P

−
2ε Ω

× hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )w,β ],α − P dx
1
Z
− H N N dx, (18.27)
where generically we have denoted
w = (∇2 )−1 η for η ∈ L2 (Ω),
if η = ∇2 w and w ∈ W01,2 (Ω).
J˜∗ (Q, N) = ∗inf ∗ {F ∗ (z∗ ) − G∗2 (z∗ , Q) − G˜ ∗1 (−Q, N)},

z ∈Y2
∀Q ∈ Ŷ2∗ , N ∈ A∗ .
Also, such an infimum is attained through the equation
∂ F ∗ (z∗ ) ∂ G∗2 (z∗ , Q)

= ,
∂ z∗ ∂ z∗
that is,
−∇[(hαβ λ µ D∗λ µ Dαβ )−1 ( div (z∗ + Q) − P)]

= {(−Nαβ + εδαβ )(z∗2 )β }, (18.28)
that is,

− div ∇(hαβ λ µ D∗λ µ Dαβ )−1 ( div (z∗ + Q) − P)
= [(−Nαβ + εδαβ )(z∗2 )β ],α
= ∇2 w, (18.29)
where
w = −(hαβ λ µ D∗λ µ Dαβ )−1 ( div (z∗ + Q) − P).
Hence,
1
F ∗ (z∗ ) = hw,α , z∗α iL2 − h(−Nαβ + εδαβ )w,α w,β iL2 . (18.30)
2
From
− div (z∗ + Q) = hαβ λ µ w,αβ λ µ − P,
and
(z∗α ),α = [(−Nαβ + εδαβ )(w,β )],α ,
we obtain
− div Q = hαβ λ µ w,αβ λ µ − P + div (z∗ )

= hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )(w,β )],α − P. (18.31)
Let v ∈ W02,2 (Ω) be such that Q = ∇v

From the last equation
−∇2 v = − div (∇v) = − div Q = hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )(w,β )],α − P,
so that
v = −(∇2 )−1 (hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )(w,β )],α − P)
Thus,
Q = {[−(∇2 )−1 (hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )(w,β )],α − P],ρ }
so that
1
Z
δ Qα Qβ dx
2ε Ω αβ
1
Z
(−∇2 )−1 hαβ λ µ w,αβ λ µ − [(Nα β − εδαβ )w,β ],α − P

=
2ε Ω

× hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )w,β ],α − P dx. (18.32)
Thus,
1
Z
J˜∗ (Q, N) = − h w w dx
1
Z
− (N − εδαβ )w,α w,β dx
2 Ω αβ
1
Z
(−∇2 )−1 hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )w,β ],α − P

−
2ε Ω

× hαβ λ µ w,αβ λ µ − [(Nαβ − εδαβ )w,β ],α − P dx
1
Z
− H N N dx
= J3 (w, N). (18.33)
Moreover, from δ J(u0 ) = 0 we have
hαβ λ µ (w0 ),αβ λ µ − [((N0 )αβ − εδαβ )(w0 ),β ],α − εδαβ (w0 )αβ − P = 0, in Ω,
so that
ŵ0 ≡ w0 = (∇2 )−1 (hαβ λ µ (w0 ),αβ λ µ − [((N0 )αβ − εδαβ )(w0 ),β ],α − P)/ε
and therefore
−hαβ λ µ (w0 ),αβ λ µ + [((N0 )αβ − εδαβ )(w0 ),β ],α

+hαβ λ µ (ŵ0 ),αβ λ µ − [((N0 )αβ − εδαβ )(ŵ0 ),β ],α
= 0. (18.34)
Also,
(u0 )α,β + (u0 )β ,α

1 1
H αβ λ µ (N0 )λ µ = + (w0 ),α (w0 ),β ,
2 2 2
so that
1 (u0 )α,β + (u0 )β ,α

δ J3 (w0 , N0 ) + , (N0 )αβ − h(u0 ),α , Pα iL2 = 0.
2 2 L2
From these last results and from the last theorem, we may obtain
J3 (w0 , N0 ) = J(w0 )
= J ∗ (z0 , Q0 , N0 )
= J˜∗ (Q0 , N0 ). (18.35)
From this, also from the last theorem and from (18.33), we finally get
J(u0 ) = inf J(u)

u∈U
= sup J˜∗ (Q, N)
(Q,N)∈Ŷ2∗ ×A∗
= J˜∗ (Q0 , N0 )
= J ∗ (z∗0 , Q0 , N0 )
= J3 (w0 , N0 )
= sup J3 (w, N). (18.36)
(w,N)∈U×A∗
18.4 A multi-duality principle for non-convex optimization

Our final result is a multi-duality principle, which is summarized by the following theorem.
Theorem 18.4.1 Considering the notation and statements of the plate model addressed in the last sections,
assuming a not relabeled finite dimensional approximate model, in a finite elements or finite differences
context, let J1 : U ×Y1∗ ×Y2∗ → R be a functional where
1
Z
J1 (u, Q, N) = hαβ λ µ w,αβ w,λ µ dx − hP, wiL2
2 Ω
1
Z
+ (−NαβK )Q Q dx − hw , Q i
α β ,α α L2
2 Ω
K 1
Z Z
+ w,α w,α dx − H N N dx
2 Ω 2 Ω αβ λ µ αβ λ µ
−hNαβ ,β + Pα , uα iL2 , (18.37)
and where
K } = {−N −1
{−Nαβ αβ + Kδαβ } .
Define also,
K
C∗ = N ∈ Y1∗ : {−Nαβ + Kδαβ } > δαβ ,
2
B∗ = {N ∈ Y1∗ : Nαβ ,β + Pα = 0, in Ω},

D+ = {N ∈ Y1∗ : Jˆ1∗ (Q, N) > 0, ∀Q ∈ Y2∗ such that Q 6= 0},
D− = {N ∈ Y1∗ : Jˆ2∗ (Q, N) < 0, ∀Q ∈ Y2∗ such that Q 6= 0},
where
1
Z
Jˆ1∗ (Q, N ) = −FK∗ (Q) + K )Q Q dx,
(−Nαβ α β
2 Ω
and where
1 K
Z Z
FK∗ (Q) = sup hw,α , Qα iL2 − h w w dx − w,α w,α dx .
u∈U 2 Ω αβ λ µ ,αβ ,λ µ 2 Ω
Moreover
Jˆ2∗ (Q, N) = −FK∗ (Q) + HK∗ (Q, N),
where
1
Z
HK∗ (Q, N) = sup hw,α , Q,α iL2 − (−Nαβ + Kδαβ )w,α w,β dx .
u∈U 2 Ω
Moreover, we also define,
A∗+ = B∗ ∩C∗ ∩ D+
A∗− = B∗ ∩C∗ ∩ D− ,
and
E ∗ = B∗ ∩C∗ .
Let u0 ∈ U be such that δ J(u0 ) = 0 and define
(N0 )αβ = Hαβ λ µ γλ µ (u0 ),
and
(Q0 )α = ((N0 )αβ + Kδαβ )(w0 ),β .
1. if δ 2 J(u0 ) > 0 and N0 ∈ E ∗ , defining
J2 (u, Q) = sup J1 (u, Q, N),

N∈E ∗
and
J˜∗ (Q) = inf J2 (u, Q),
u∈Br1 (u0 )
where r1 > 0 is such that

δ 2 J(u) > 0
in Br1 (u0 ), we have
J(u0 ) = J˜∗ (Q0 ),
δ J˜∗ (Q0 ) = 0
and if K > 0 is sufficiently big,
δ 2 J˜∗ (Q0 ) ≥ 0
and there exist r, r2 > 0 such that
J(u0 ) = inf J(u)

u∈Br (u0 )
= inf J˜∗ (Q)

Q∈Br2 (Q0 )
= J˜∗ (Q0 ). (18.38)
2. If N0 ∈ A∗+ , defining
J3 (u, Q) = sup J1 (u, Q, N),
N∈A∗+
and
J˜3∗ (Q) = inf J3 (u, Q),
u∈U
then
δ J˜3∗ (Q0 ) = 0,
δ 2 J˜3∗ (Q0 ) ≥ 0,
and
J(u0 ) = inf J(u)

u∈U
= inf J˜3∗ (Q)
Q∈Y2∗
= J˜3∗ (Q0 ). (18.39)
3. If δ 2 J(u0 ) < 0 so that N0 ∈ A∗− , defining
1
Z
Jˆ∗ (Q, N) = −F̂K∗ (Q) + HK∗ (Q, N) − H αβ λ µ Nαβ Nλ µ dx,
2 Ω
where

1 K
Z Z
F̂K∗ (Q) = sup hw,α , Qα iL2 − h w w dx − w,α w,α dx + hw, PiL2 .
u∈U 2 Ω αβ λ µ ,αβ ,λ µ 2 Ω
we have that
Jˆ∗ (Q0 , N0 ) = J(u0 ),
δ {Jˆ∗ (Q0 , N0 ) − h(N0 )αβ ,β + Pα , (u0 )α iL2 } = 0,
and there exist r, r1 , r2 > 0 such that
J(u0 ) = sup J(u)

u∈Br (u0 )
( )
= sup sup ˆ∗
J (Q, N)
Q∈Br1 (Q0 ) N ∈Br2 (N0 )∩E ∗
= Jˆ∗ (Q0 , N0 ). (18.40)
Proof 18.3 From the assumption N0 ∈ E ∗ we have that
J2 (u0 , Q0 ) = sup J1 (u0 , Q0 , N) = J1 (u0 , Q0 , N0 ),

N∈E ∗
where such a supremum is attained through the equation
∂ J1 (u0 , Q0 , N0 )
= 0.
∂N
Moreover, there exists r > 0, r2 > 0 such that
J(u0 ) = inf J(u),

u∈Br (u0 )
and (we justify that the first infimum Q in this equation (18.41) is well defined in the next lines)
J(u0 ) = inf inf sup J1 (u, Q, N)

u∈Br (u0 ) Q∈Br2 (Q0 ) N∈E ∗
= inf inf sup J1 (u, Q, N)

Q∈Br2 (Q0 ) u∈Br1 (u0 ) N∈E ∗
= inf J˜∗ (Q)

Q∈Br2 (Q0 )
= J˜∗ (Q0 ). (18.41)

Observe the concerning extremal in Q is attained through the equation,
∂ J1 (u0 , Q0 , N0 )
= 0.
∂Q
Hence, from
∂ J(u0 )
= 0,
∂u
from the implicit function theorem and chain rule, we get
∂ J(u0 )
0 =
∂u
∂ J1 (u0 , Q0 , N0 )
=
∂u
∂ J1 (u0 , Q0 , N0 ) ∂ Q0
+
∂Q ∂u
∂ J1 (u0 , Q0 , N0 ) ∂ N0
+
∂N ∂u
∂ J1 (u0 , Q0 , N0 )
= . (18.42)
∂u
Therefore,
∂ J˜∗ (Q0 )
∂Q
∂ J1 (u0 , Q0 , N0 )
=
∂Q
∂ J1 (u0 , Q0 , N0 ) ∂ u0
+
∂u ∂Q
∂ J1 (u0 , Q0 , N0 ) ∂ N0
+
∂N ∂Q
= 0. (18.43)
From this we shall denote

δ J˜∗ (Q0 ) = 0.
Let us now show that the first infimum in Q in (18.41) is well defined.
Recall again that,
J2 (u0 , Q0 ) = sup J1 (u0 , Q0 , N),
N∈E ∗
where such a supremum is attained through the equation
∂ J1 (u0 , Q0 , N0 )
= 0,
∂N
that is,
(Q0 )λ (−N0 )Kαλ (−N0 )Kβ µ (Q0 )µ

(u0 )α ,β + (u0 )β ,α
−H αβ λ µ (N0 )λ µ +
2
= 0, in Ω. (18.44)
Taking the variation of this last equation in Qρ we have
−(−N0 )Kαρ (−N0 )Kβ µ (Q0 )µ

∂ (N0 )λ µ
+(w0 )λ (−N0 )Kαβ (w0 )µ
∂ Qρ
∂ (N0 )λ µ
−H αβ λ µ
∂ Qρ
= 0, in Ω, (18.45)
that is,
−(−N0 )Kαρ (w0 )β

∂ (N0 )λ µ
+(w0 )λ (−N0 )Kαβ (w0 )µ
∂ Qρ
∂ (N0 )λ µ
−H αβ λ µ
∂ Qρ
= 0, in Ω, (18.46)
so that
∂ (N0 )αβ
∂ Qρ
= H αβ λ µ − (w0 )λ (−N0 )Kαβ (w0 )µ ((−N0 )Kλ µ (w0 )ρ ). (18.47)
Hence, if K > 0 is sufficiently big, we obtain

2
∂ J2 (u0 , Q0 )
∂ Qα ∂ Qβ
2
∂ J1 (u0 , Q0 , N0 )
= +
∂ Qα ∂ Qβ
∂ 2 J1 (u0 , Q0 , N0 ) ∂ (N0 )λ µ

+
∂ Qα ∂ Nλ µ ∂ Qβ
= {(−N0 )Kαβ } + (−N0 )Kαη (w0 )ρ
h i
× H ηρλ µ − (w0 )λ (−N0 )Kηρ (w0 )µ ((−N0 )λKµ (w0 )β )
= {(−N0 )Kαβ } + O(1/K 2 )

> 0. (18.48)
Therefore, the first infimum in Q in (18.41) is well defined.

Also, from (18.41) and the second order necessary condition for a local minimum, we obtain
δ 2 J˜∗ (Q0 ) ≥ 0.
Assume now again δ J(u0 ) = 0 and N0 ∈ A∗+ .

Recall that
J3 (u, Q) = sup J1 (u, Q, N).
N∈A∗+
Observe that if N ∈ A∗+ , by direct computation we may obtain

2
δuQ J1 (u, Q, N) > 0.
Therefore, J3 (u, Q) is convex since is the supremum of a family of convex functionals.

Similarly as above we may obtain
δ J3 (u0 , Q0 ) = 0,
and
J3 (u0 , Q0 ) = J(u0 )
so that
J(u0 ) = J3 (u0 , Q0 )
= inf ∗ J3 (u, Q)
(u,Q)∈U×Y2
≤ inf J3 (u, Q)
Q∈Y2∗
= J(u), ∀u ∈ U. (18.49)
Moreover,
J˜3∗ (Q0 ) = inf J3 (u, Q0 )

u∈U
= J3 (u0 , Q0 )
≤ J3 (u, Q), ∀u ∈ U, Q ∈ Y2∗ . (18.50)
Hence,
J(u0 ) = J˜3∗ (Q0 ) ≤ inf J3 (u, Q) = J˜3∗ (Q), ∀Q ∈ Y2∗ .
u∈U
From these last results we may write,
J(u0 ) = inf J(u)

u∈U
= inf J˜3∗ (Q)
Q∈Y2∗
= J˜3∗ (Q0 ). (18.51)
From this, similarly as above, we may obtain
δ 2 J˜3∗ (Q0 ) ≥ 0.
Finally, suppose now δ 2 J(u0 ) < 0, so that N0 ∈ A∗− .

From this we obtain
∂ 2 Jˆ∗ (Q0 , N0 )
<0
∂ (Qα,α )2
where, as previously indicated
1
Z
Jˆ∗ (Q, N) = −F̂K∗ (Q) + HK∗ (Q, N) − H αβ λ µ Nαβ Nλ µ dx,
2 Ω
Here,
1
Z
HK∗ (Q, N) = sup hw,α , Q,α iL2 − (−Nαβ + Kδαβ )w,α w,β dx .
u∈U 2 Ω
Denoting,
1
Z
Jˆ(u, N) = hαβ λ µ w,αβ w,λ µ dx
2 Ω
1 1
Z Z
+ N w,α w,β dx − H αβ λ µ Nαβ Nλ µ dx
2 Ω αβ 2 Ω
−hNαβ ,β − Pα , uα iL2 , (18.52)
also from N0 ∈ A∗− and from

Jˆ∗ (Q0 , N0 ) = J(u0 ),
δ {Jˆ∗ (Q0 , N0 ) − h(N0 )αβ ,β + Pα , (u0 )α iL2 } = 0,
(the proofs of such results are very similar to those of the corresponding cases developed above), there exist
r, r1 , r2 > 0 such that for N ∈ A∗− ∩ Br2 (N0 ), we have
sup Jˆ(u, N) = sup Jˆ∗ (Q, N), (18.53)

u∈Br (u) Q∈ Br1 (Q0 )
and
J(u0 ) = sup J(u)

u∈Br (u0 )
( )
= sup sup Jˆ(u, N)
u∈Br (u0 ) N∈Br2 (N0 )∩E ∗
( )
= sup sup ˆ∗
J (Q, N)
Q∈Br1 (Q0 ) N∈Br2 (N0 )∩E ∗
= Jˆ∗ (Q0 , N0 ). (18.54)
18.5 Conclusion
In this chapter we have developed a new primal dual variational formulation and a multi-duality principle
applied to a non-linear model of plates.
About the primal dual formulation, we emphasize such a formulation is concave so that it is very inter-
esting from a numerical analysis point of view.
Finally, the results here presented may be also developed in a similar fashion for a large class of problems,
including non-linear models in elasticity and other non-linear models of plates and shells.
Chapter 19
On Duality Principles for One

and Three-Dimensional
Non-Linear Models in Elasticity
19.1 Introduction
In this chapter, we develop duality principles applicable to primal variational formulations found in the
non-linear elasticity theory. As a first application, we establish the concerning results in details for one and
three-dimensional models. We emphasize such duality principles are applicable to a larger class of variational
optimization problems, such as non-linear models of plates and shells and other models in elasticity. Finally,
we formally prove there is no duality gap between the primal and dual formulations, in a local extremal
context.
About the references, this article in some sense extends and complements the original works of Telega,
Bielski and their co-workers [11, 10, 76, 77]. In particular in [11], published in 1985 and in [77], for three-
dimensional elasticity and related models, the authors established duality principles and concerning global
optimality conditions, for the special case in which the stress tensor is positive definite at a critical point. In
this specific sense, the present work complements such previous ones, considering we establish a sufficient
condition for local minimality which does not require the stress tensor to be either positive or negative defined
along the concerning domain. Such an optimality condition is summarized by the condition kux k∞ < 1/4 at
a critical point.
The tools of convex analysis and duality theory here used may be found in [14, 33, 64]. Existence of
results in non-linear elasticity and related models may be found in [29, 30, 31].
Finally, details on the function spaces addressed may be found in [1].
At this point, we start to describe the primal variational formulation for the one-dimensional model.
Let Ω = [0, L] ⊂ R be an interval which represents the axis of a straight bar of length L and constant cross
section area A.
We denote by u : [0, L] → R the field of axial displacements for such a bar, resulting from the application
of an axial load field P ∈ C([0, L]).
We also denote
U = {u ∈ C1 ([0, L]) : u(0) = u(L) = 0},
and
Û = {u ∈ U : kux k∞ < 1/4}.
The energy for such a system, denoted by J : U → R, is expressed as
EA L 1 2 2
Z Z L
J(u) = ux + ux dx − Pu dx, ∀u ∈ U,
2 0 2 0
where E > 0 is the Young modulus.

We shall also define 2
Z L
EA 1
G(ux ) = ux + u2x d x.
2 0 2
Finally, generically we shall denote, for u ∈ U and r > 0,
Br (u) = {v ∈ U : kv − ukU < r}
where
kvkU = max {|v(x)| + |vx (x)|}, ∀v ∈ U.
x∈[0,L]
Moreover, defining V = C([0, L]), for z∗ ∈ V and r1 > 0, we shall generically also denote
Br1 (z∗ ) = {v ∈ V : kv − z∗ kV < r1 },
where
kvkV = kvk∞ = max |v(x)|, ∀v ∈ V.
x∈[0,L]
Similar corresponding standard notations are valid for V ×V and the 3-dimensional model.
19.2 The main duality principle for the one-dimensional model

Our first duality principle is summarized by the following theorem:
Theorem 19.2.1 Let J : U → R be defined by

Z L 2 Z L
EA 1
J(u) = ux + u2x dx − Pu dx.
2 0 2 0
Assume u0 ∈ Û is such that

δ J(u0 ) = 0.
Z L
K
F (ux ) = u2x dx,
2 0
GK : U → R by
Z L
K
GK (ux ) = G(ux ) + u2x dx
2 0
and J ∗ : V ×V ×V → R by
J ∗ (v∗ , z∗ ) = F ∗ (z∗ ) − G∗K (v∗ , z∗ ),
F ∗ (z∗ ) = sup{hv, z∗ iL2 − F(v)}

v∈V
Z L
1
= (z∗ )2 dx (19.1)
2K 0
On Duality Principles for One and Three-Dimensional Non-Linear Models in Elasticity 411
and
G∗K (v∗ , z∗ ) = sup {hv1 , z∗ + v∗2 iL2 + hv2 , v∗1 iL2

(v1 ,v2 )∈V ×V
Z L 2 Z L
)
EA 1 K
− v1 + v22 dx − v22 dx
2 0 2 2 0
(v∗1 )2
Z L Z L
1 1
= dx + (v∗2 + z∗ )2 dx, (19.2)
2 0 z + v∗2 + K
∗ 2E A 0
if v∗2 + z∗ + K > 0, in Ω.
Define also,
A∗ = {v∗ = (v∗1 , v∗2 ) ∈ V ×V : (v∗1 )x + (v∗2 )x + P = 0, in Ω},
K = EA/2,
ẑ∗ = K(u0 )x ,

1
v̂2 = E A (u0 )x + (u0 )x − ẑ∗ ,
∗ 2
2
v̂∗1 = (ẑ∗ + v̂∗2 + K)(u0 )x .
Under hypotheses and definitions, we have
δ 2 J(u0 , ϕ, ϕ) ≥ 0, ∀ϕ ∈ Cc1 ((0, L)) (19.3)
and there exist r, r1 , r2 > 0 such that
J(u0 ) = inf J(u)

u∈Br (u0 )
( )
∗ ∗ ∗
= sup inf {J (v , z )}
v∗ ∈Br2 (v̂∗ )∩A∗ z∗ ∈Br1 (ẑ∗ )
= J ∗ (v̂∗ , ẑ∗ ). (19.4)

∂ 2 J ∗ (v̂∗ , ẑ∗ ) 2 (v̂∗ )2 1
= − ∗ ∗1 EA 3 −
∂ (z∗ )2 EA (v̂2 + ẑ + 2 ) EA
1 (u0 )2
= − ∗ ∗ x EA . (19.5)
EA v̂2 + ẑ + 2
Also,

EA 1 EA
v̂∗2 + zˆ∗ + = 2
EA (u0 )x + (u0 )x +
2 2 2

EA 1 1 1
> − EA +
2 4 2 16

1 1
= EA −
4 32
7
= EA . (19.6)
32
∂ 2 J ∗ (v̂∗ , ẑ∗ ) 1 1 32
> −
∂ (z∗ )2 EA 16 (7 EA)

1 2
= 1−
EA 7
5
=
7EA
> 0, in Ω. (19.7)
Thus, we may infer that there exists r1 , r2 > 0 such that J ∗ (v∗ , z∗ ) is convex in z∗ and concave in v∗ , on
Br1 (ẑ∗ ) × Br2 (v̂∗ ).
Now, denoting
Jˆ(v∗ , z∗ , u) = J ∗ (v∗ , z∗ ) − hu, (v∗1 )x + (v∗2 )x + PiL2
we obtain
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) z∗ 1 (v̂∗1 )2 v̂∗2 + ẑ∗
= + −
∂ z∗ K 2 (v̂∗2 + ẑ∗ + K)2 EA
EA (u0 )x + 12 (u0 )2x

1
= (u0 )x + (u0 )2x −
2 EA
= 0, in Ω. (19.8)
Also,
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) (v̂∗1 )

= − + (u0 )x
∂ v∗1 (v̂∗2 + ẑ∗ + K )
= 0, in Ω, (19.9)
and
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) 1 (v∗1 )2 v̂∗2 + ẑ∗
= − + (u0 )x
∂ v∗2 2 (v̂∗2 + ẑ∗ + K )2 EA
EA (u0 )x + 21 (u0 )2x

1
= (u0 )x + (u0 )2x −
2 EA
= 0, in Ω. (19.10)
Finally,
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 )

= −(v̂∗1 )x − (v̂∗2 )x − P
∂u
= δ J(u0 )
= 0, in Ω. (19.11)
These last four results may be summarized by the equation
δ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) = 0.
Since above we have obtained that J ∗ (v∗ , z∗ ) is convex in z∗ and concave in v∗ , on
Br1 (ẑ∗ ) × Br2 (v̂∗ ),

from this, the last result and from the min-max theorem, we have
J ∗ (v̂∗ , ẑ∗ ) = Jˆ∗ (v̂∗ , ẑ∗ , u0 )
( )
∗ ∗ ∗
= sup inf Jˆ (v , z , u0 )
v∗ ∈Br2 (v̂∗ ) z∗ ∈Br1 (ẑ∗ )
( )
= sup inf Jˆ∗ (v∗ , z∗ ) . (19.12)
v∗ ∈Br2 (v̂∗ )∩A∗ z∗ ∈Br1 (ẑ∗ )
At this point we observe that

J ∗ (v̂∗ , ẑ∗ ) = Jˆ∗ (v̂∗ , ẑ∗ , u0 )
= F ∗ (ẑ∗ ) − G∗K (v̂∗ , ẑ∗ )
+h(u0 )x , v̂∗1 + v̂∗2 iL2 − hu0 , PiL2
K L
Z
= (u0 )2x dx − h(u0 )x , ẑ∗ iL2
2 0
−h(u0 )x , v̂∗1 + v̂∗2 + ẑ∗ iL2
K L
Z
+G((u0 )x ) + (u0 )2x dx
2 0
+h(u0 )x , v̂∗1 + v̂∗2 iL2 − hu0 , PiL2
= G((u0 )x ) − hu0 , PiL2
= J(u0 ). (19.13)
On the other hand, since v̂∗ ∈ A∗ , we may write
J ∗ (v̂∗ , ẑ∗ ) = Jˆ∗ (v̂∗ , ẑ∗ )
= inf Jˆ∗ (v̂∗ , z∗ )
z∗ ∈Br1 (ẑ∗ )
≤ F ∗ (z∗ ) − G∗K (v̂∗ , z∗ )

+h(u)x , v̂∗1 + v̂∗2 iL2 − hu, PiL2
≤ F ∗ (z∗ ) − h(u)x , v̂∗1 + v̂∗2 + z∗ iL2
Z L 2
u x
+G(ux ) + K dx
0 2
+h(u)x , v̂∗1 + v̂∗2 iL2 − hu, PiL2 , (19.14)
z∗
∀u ∈ U, ∈ Br1 (ẑ∗ ).
In particular, there exists r > 0 such that if u ∈ Br (u0 ) then z∗ = Kux ∈ Br1 (ẑ∗ ), so that from this and
(19.14), we obtain
Z L 2 Z L 2
ux ux
J ∗ (v̂∗ , ẑ∗ ) ≤ −K dx + G(ux ) + K dx − hu, PiL2
2
0 0 2
= G(ux ) − hu, PiL2
= J(u), ∀u ∈ Br (u0 ). (19.15)
Finally, from (19.12), (19.13), and (19.15), we may infer that
J(u0 ) = inf J(u)
u∈Br (u0 )
( )
∗ ∗ ∗
v∗ ∈Br2 (v̂∗ )∩A∗ z∗ ∈Br1 (ẑ∗ )
= J ∗ (v̂∗ , ẑ∗ ). (19.16)

From the first equation in such a result we may also obtain the standard second order necessary condition
indicated in (19.3).
19.3 The primal variational formulation for the three-dimensional

model
At this point we start to describe the primal formulation for the three-dimensional model.
Consider Ω ⊂ R3 an open, bounded, connected set, which represents the reference volume of an elas-
tic solid under the loads f ∈ C(Ω; R3 ) and the boundary loads fˆ ∈ C(Γ; R3 ), where Γ denotes the regular
(Lipschitzian) boundary of Ω. The field of displacements resulting from the actions of f and fˆ is denoted
by u ≡ (u1 , u2 , u3 ) ∈ U, where u1 , u2 , and u3 denotes the displacements relating the directions x, y, and z
respectively, in the cartesian system (x, y, z).
Here, U is defined by
U = {u = (u1 , u2 , u3 ) ∈ C1 (Ω; R3 ) | u = (0, 0, 0) ≡ 0 on Γ0 } (19.17)
and Γ = Γ0 ∪ Γ1 , Γ0 ∩ Γ1 = 0.
/ We assume |Γ0 | > 0 where |Γ0 | denotes the Lebesgue measure of Γ0 .
The stress tensor is denoted by {σi j }, where

1
σi j = Hi jkl (uk,l + ul,k + um,k um,l ) , (19.18)
2
{Hi jkl } = {λ δi j δkl + µ(δik δ jl + δil δ jk )},

{δi j } is the Kronecker delta and λ , µ > 0 are the Lamé constants (we assume they are such that {Hi jkl } is a
symmetric constant positive definite forth order tensor). Here, i, j, k, l ∈ {1, 2, 3}.
The boundary value form of the non-linear elasticity model is given by

 σi j, j + (σm j ui,m ), j + fi = 0, in Ω,
u = 0, on Γ0 , (19.19)
σi j n j + σm j ui,m n j = fî , on Γ1 ,

where n = (n1 , n2 , n3 ) denotes the outward normal to the surface Γ.

The corresponding primal variational formulation is represented by J : U → R, where

1 1 1
Z
J(u) = Hi jkl (ui, j + u j,i + um,i um, j ) (uk,l + ul,k + um,k um,l ) dx
2 Ω 2 2
Z
−hu, f iL2 (Ω;R3 ) − fî ui dΓ (19.20)
Γ1
where Z
hu, f iL2 (Ω;R3 ) = fi ui dx.
Ω
Remark 19.3.1 By a regular Lipschitzian boundary Γ of Ω we mean regularity enough so that the standard
Gauss-Green formulas of integrations by parts to hold. Also, we denote by 0 the zero vector in appropriate
function spaces.
About the references, similarly as for the one-dimensional case, we refer to [76, 77, 10, 11] as the first
articles to deal with the convex analysis approach applied to non-convex and non-linear mechanics models.
Indeed, the present work complements such important original publications, since in these previous results
the complementary energy is established as a perfect duality principle for the case of positive definiteness of
the stress tensor (or the membrane force tensor, for plates and shells models) at a critical point.
We have relaxed such constraints, allowing to some extent, the stress tensor to not be necessarily either
positive or negative definite in Ω. Similar problems and models are addressed in [14].
Moreover, we highlight again that existence results for models in elasticity are addressed in [29, 30, 31].
Finally, the standard tools of convex analysis here used may be found in [33, 78, 64, 14].
19.4 The main duality principle for the three-dimensional model

In this section we present the main duality principle for the 3-Dimensional model.
The main result is summarized by the following theorem.
Theorem 19.4.1 Let J : U → R be defined by

1 1
Z
J(u) = Hi jkl (ui, j + u j,i + um,i um, j ) (uk,l + uk,l + uq,k uq,l ) dx
Ω 2 2
ˆ
−hui fi iL2 (Ω) − hui , fi iL2 (Γ ) . (19.21)
1
Assume u0 ∈ Û is such that

δ J(u0 ) = 0,
where
Û = {u ∈ U : kui, j k∞ < 1/8, ∀i, j ∈ {1, 2, 3}}.
Z
K ui, j + u j,i ui, j + u j,i
F(ui, j ) = dx,
2 Ω 2 2
GK : U → R by
Z
GK ({ui, j }) = G(u) + dx
2 Ω 2 2
where
1 1
Z
G(u) = Hi jkl (ui, j + u j,i + um,i um, j ) (uk,l + uk,l + uq,k uq,l ) dx,
Ω 2 2
and J∗ : V ×V ×V → R by
J ∗ (v∗ , z∗ ) = F ∗ (z∗ ) − G∗K (v∗ , z∗ ),
where
V = C(Ω; R3×3 ),
F ∗ (z∗ ) = sup{hvi j , z∗i j iL2 − F(v)}

v∈V
1
Z
= z∗i j z∗i j dx (19.22)
2K Ω
and
G∗K (v∗ , z∗ ) = h(v1 )i j , z∗i j + (v∗2 )i j iL2 + h(v2 )i j , (v∗1 )i j iL2

sup
(v1 ,v2 )∈V ×V

1 1 1
Z
−Hi jkl (v1 )i j + (v2 )mi (v2 )m j (v1 )kl + (v2 )qk (v2 )ql dx
2 Ω 2 2

K
Z
− (v2 )i j (v2 )i j dx
2 Ω
1
Z
= ((v∗ )i j + (z∗ )i j + Kδi j )(v∗1 )mi (v∗1 )m j ; dx
2 Ω 2
1
Z
+ H i jkl ((v∗2 )i j + z∗i j )((v∗2 )kl + z∗kl ) dx, (19.23)
2 Ω
if {(v∗2 )i j + (z∗ )i j + Kδi j } is positive definite in Ω.
Here
{(v2∗ )i j + (z∗ )i j + Kδi j } = {(v∗2 )i j + (z∗ )i j + Kδi j }−1 .
Define also, A∗ = A1 ∩ A2 , where
A1 = {v∗ = (v∗1 , v∗2 ) ∈ V ×V : (v∗1 )i j, j + (v∗2 )i j, j + fi = 0, in Ω},
and
A2 = {v∗ = (v∗1 , v∗2 ) ∈ V ×V : (v∗1 )i j n j + (v∗2 )i j n j − fî = 0, on Γt },
and let K > 0 be such that
Di jkl 3
M= − δi j − H i jkl
K 32K
is a positive definite tensor, where

1, if i = k and j = l,
Di jkl = (19.24)
0, otherwise
and, in an appropriate sense,

{H i jkl } = {Hi jkl }−1 .
Assume also,
(u0 )i, j + (u0 ) j,i
ẑ∗i j = K ,
2

(u0 )k,l + (u0 )l,k 1
(v̂∗2 )i j = Hi jkl + (u0 )m,k (u0 )m,l − (ẑ∗ )i j ,
2 2
(v̂∗1 )i j = ((ẑ∗ )im + (v̂∗2 )im + Kδim )(u0 )m, j ,
∀i, j ∈ {1, 2, 3} and K > 0 is also such that
{(v̂∗2 )i j + (ẑ∗ )i j + Kδi j } ≥ K{δi j }/2
Under hypotheses and definitions, there exist r, r1 , r2 > 0 such that
J(u0 ) = inf J(u)

u∈Br (u0 )
( )
∗ ∗ ∗
v∗ ∈Br2 (vˆ∗ )∩A∗ z∗ ∈Br1 (ˆz∗ )
= J ∗ (v̂∗ , ẑ∗ ). (19.25)

Proof 19.2 Observe that, denoting
{(v̂∗2 )i j + ẑ∗i j + Kδi j } = {(v̂∗2 )i j + ẑ∗i j + Kδi j }−2

and
{(v̂∗2 )i j + ẑ∗i j + Kδi j } = {(v̂∗2 )i j + ẑ∗i j + Kδi j }−3
we have
( )
∂ 2 J ∗ (v̂∗ , ẑ∗ )

Di jkl
= − ((v̂∗2 )i j + (ẑ∗ )i j + Kδi j )(v̂∗1 )mk (v̂∗1 )ml − H i jkl
∂ z∗i j ∂ z∗kl K

Di jkl
= − ((v̂∗2 )i j + (ẑ∗ )i j + Kδi j )(u0 )mk (u0 )ml − H i jkl
K

Di jkl 3
≥ − δi j − H i jkl
K 32K
> 0. (19.26)
Thus, there exist r1 , r2 > 0 such that J ∗ (v∗ , z∗ ) is convex in z∗ and concave in v∗ on
Br1 (ẑ∗ ) × Br2 (v̂∗ ).
Now, denoting
Jˆ(v∗ , z∗ , u) = J ∗ (v∗ , z∗ ) − hu, (v∗1 )i j, j + (v∗2 )i j, j + fi iL2 (Ω)
+hu, (v∗1 )i j n j + (v∗2 )i j n j − fî iL2 (Γt ) (19.27)
we obtain
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) z∗i j
1
= + (v̂∗ )i j + (ẑ∗ )i j + Kδi j )(v̂∗1 )mi (v̂∗1 )m j − H i jkl ((v̂∗2 )kl + z∗kl )
∂ (z∗ )i j K 2 2
(u0 )i, j + (u0 ) j,i 1
= + (u0 )mi (u0 )m j − H i jkl ((v̂∗2 )kl + z∗kl )
2 2
= 0, in Ω. (19.28)
Also,
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 )
= −(vˆ∗2 )im + (ẑ∗ )im + Kδim )(v̂∗1 )m j + (u0 )i, j
∂ (v1∗ )i j
= 0, in Ω, (19.29)
and
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) 1 ∗
= (v̂ )i j + (ẑ∗ )i j + Kδi j )(v̂∗1 )mi (v̂∗1 )m j − H i jkl ((v̂∗2 )kl + zˆ∗kl )
∂ (v∗2 )i j 2 2
(u0 )i, j + (u0 ) j,i
+
2
(u0 )i, j + (u0 ) j,i 1
= + (u0 )mi (u0 )m j − H i jkl ((v̂∗2 )kl + ẑ∗kl )
2 2
= 0, in Ω. (19.30)
Finally,
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) −(v̂∗1 )i j, j − (v̂∗2 )i j, j − fi ,

in Ω,
= (19.31)
∂u (v̂∗1 )i j n j + (v̂∗2 )i j n j − fî , on Γ1 ,
Hence,
∂ Jˆ∗ (v̂∗ , ẑ∗ , u0 )
= δ J(u0 ) = 0.
∂u
These last four results may be summarized by the equation
δ Jˆ∗ (v̂∗ , ẑ∗ , u0 ) = 0.
Since above we have obtained that J ∗ (v∗ , z∗ ) is convex in z∗ and concave in v∗ , on
Br1 (ẑ∗ ) × Br2 (v̂∗ ),
from this, the last result and from the min-max theorem, we have
J ∗ (v̂∗ , ẑ∗ ) = Jˆ∗ (v̂∗ , ẑ∗ , u0 )

( )
∗ ∗ ∗
= sup inf Jˆ (v , z )
v∗ ∈Br2 (v̂∗ ) z∗ ∈Br1 (ẑ∗ )
( )
= sup inf Jˆ∗ (v∗ , z∗ ) . (19.32)
v∗ ∈Br2 (v̂∗ )∩A∗ z∗ ∈Br1 (ẑ∗ )
At this point we observe that
J ∗ (v̂∗ , ẑ∗ ) = Jˆ∗ (v̂∗ , ẑ∗ , u0 )

= F ∗ (ẑ∗ ) − G∗K (v̂∗ , ẑ∗ )
+h(u0 )i, j , (v̂∗1 )i j + (v̂∗2 )i j iL2 − h(u0 )i , fi iL2 (Ω) − h(u0 )i , fî iL2 (Γ1 )
= −F({(u0 )i, j }) + h(u0 )i, j , (ẑ∗ )i j iL2
−h(u0 )i, j , (v̂∗1 )i j + (v̂∗2 )i j + (ẑ∗ )i j iL2
Z
K (u0 )i, j + (u0 ) j,i (u0 )i, j + (u0 ) j,i
+G((u0 )) + dx
2 Ω 2 2
+h(u0 )i, j , (v̂∗1 )i j + (v̂∗2 )i j iL2 − h(u0 )i , fi iL2 (Ω) − h(u0 )i , fî iL2 (Γ ) 1
= G((u0 )) − h(u0 )i , fi iL2 (Ω) − h(u0 )i , fî iL2 (Γ1 )

= J(u0 ). (19.33)
On the other hand, since v̂∗ ∈ A∗ , we may write
J ∗ (v̂∗ , ẑ∗ ) = inf Jˆ∗ (v̂∗ , z∗ )

z∗ ∈Br1 (ẑ∗ )
≤ F ∗ (z∗ ) − G∗K (v̂∗ , z∗ )

+h(u)i, j , (v̂∗1 )i j + (v̂∗2 )i j iL2 − hui , fi iL2 (Ω) − hui , fî iL2 (Γ1 )
≤ F ∗ (z∗ ) − hui j , (v̂∗1 )i j + (v̂∗2 )i j + (z∗ )i j iL2
Z
+G(u) + dx
2 Ω 2 2
+hui, j , (v̂∗1 )i j + (v̂∗2 )i j iL2 − hui , fi iL2 (Ω) − hui , fî iL2 (Γ 1)
= F ∗ (z∗ ) − hui, j , z∗i j iL2
Z
+G(u) + dx
2 Ω 2 2
−hui , fi iL2 (Ω) − hui , fî iL2 (Γ ) , 1
(19.34)
∀u ∈ U, z∗ ∈ Br1 (ẑ∗ ). n o
u +u
In particular, there exists r > 0 such that if u ∈ Br (u0 ) then {z∗i j } = K i, j 2 j,i ∈ Br1 (ẑ∗ ), so that from
this and (19.34), we obtain
Z
∗ ∗ ∗ K ui, j + u j,i ui, j + u j,i
J (v̂ , zˆ ) ≤ − dx + G(u)
2 Ω 2 2

Z
+ dx − hui , fi iL2 (Ω) − hui , fî iL2 (Γ1 )
2 Ω 2 2
= G(u) − hui , fi iL2 (Ω) − hui , fî iL2 (Γ ) 1
= J(u), ∀u ∈ Br (u0 ). (19.35)
Finally, from (19.32), (19.33), and (19.35), we may infer that
J(u0 ) = inf J(u)

u∈Br (u0 )
( )
∗ ∗ ∗
v∗ ∈Br2 (v̂∗ )∩A∗ z∗ ∈Br1 (ẑ∗ )
= J ∗ (v̂∗ , ẑ∗ ). (19.36)
19.5 Conclusion
In this chapter we have developed some theoretical results on duality for a class of non-convex optimization
problems in elasticity. In this first approach we have developed in details duality principles and sufficient
optimality conditions for local minimality for one and three-dimensional models in elasticity. It is worth
mentioning the results may be extended to other models in elasticity and to other models of plates and shells.
Chapter 20
A Primal Dual Variational

Formulation Suitable for a Large
Class of Non-Convex Problems in
Optimization
20.1 Introduction
In this article we develop a new primal dual variational formulation suitable for a large class of non-convex
problems in the calculus of variations.
The results are obtained through basic tools of convex analysis, duality theory, the Legendre transform
concept and the respective relations between the primal and dual variables. The novelty here is that the dual
formulation is established also for the primal variables, however with a large domain region of concavity
about a critical point.
We formally prove there is no duality gap between the primal and dual formulations in a local extremal
context.
We emphasize, our work, in some sense, generalizes, extends and complements the original Telega, Biel-
ski and their co-workers results in the articles [10, 11, 77, 76].
The convex analysis results here used may be found in [33, 78, 64, 14], for example. Similar results for
other problems may be found in [14].
Finally, details on the function spaces addressed may be found in [1].
Let Ω ⊂ R3 be an open, bounded and connected set with a regular (Lipschitzian) boundary denoted by
∂ Ω.
Consider the functional J : V → R where
J(u) = G0 (∇u) + G1 (u) + G2 (u)

−F(u) − hu, f iL2 + G3 (u), (20.1)
where Z
γ
G0 (∇u) = ∇u · ∇u dx,
2 Ω
A Primal Dual Variational Formulation Suitable for a Large Class of Non-Convex Problems in Optimization 421
Z
α
G1 (u) = u4 dx,
4 Ω
(−β + K − ε)
Z
G2 (u) = u2 dx,
2 Ω
K
Z
F(u) = u2 dx
2 Ω
and Z
ε
G3 (u) = u2 dx
2 Ω
so that
Z Z
γ α
J (u) = ∇u · ∇u dx + u4 dx
2 Ω 4 Ω
Z
β
− u2 dx − hu, f iL2 , ∀u ∈ V. (20.2)
2 Ω
Here dx = dx1 dx2 dx3 , α > 0, β > 0, γ > 0, ε > 0, K > β + ε, f ∈ C(Ω) and
V = {u ∈ C2 (Ω) : u = 0, on ∂ Ω}.
Moreover, we recall that V is a Banach space with the norm k · kV , where
kukV = max{|u(x)| + |ux (x)| + |uy (x)| + |uz (x)| + |uxy (x)| + |uxz (x)| + |uyz (x)|
x∈Ω
+|uxx (x)| + |uyy (x)| + |uzz (x)|}, ∀u ∈ V, (20.3)
and generically we denote Z

hu, viL2 = u v dx, ∀u, v ∈ L2 (Ω) ≡ L2 ,
Ω
and Z
hu, viL2 = u · v dx, ∀u, v ∈ L2 (Ω; R3 ) ≡ L2 .
Ω
Now observe that
inf J(u) ≤ inf {G0 (∇u) − h∇u, z∗0 iL2

u∈U u∈U
+G1 (u) − hu, z∗1 iL2
+G2 (u) − hu, z∗2 iL2
+G3 (u) − hu, f iL2
+ sup{h∇u, z∗0 iL2 + hu, z∗1 iL2 + hu, z2∗ iL2 − F(u)}}
u∈U
= inf {G0 (∇u) − h∇u, z∗0 iL2
u∈U
+G1 (u) − hu, z∗1 iL2
+G2 (u) − hu, z∗2 iL2
+G3 (u) − hu, f iL2 }
+F ∗ (− div z∗0 + z∗1 + z∗1 )
= sup {−G∗0 (v∗0 + z∗0 ) − G∗1 (v∗1 + z∗1 )
v∗ ∈Y1 ×V ×V
−G∗2 (v∗2 + z∗2 ) − G∗3 ( div v∗0 − v∗1 − v∗2 + f )}
+F ∗ (− div z∗0 + z∗1 + z∗2 ), ∀z∗ = (z∗0 , z∗1 , z∗2 ) ∈ Y1 ×V ×V, (20.4)
where
1
Z
F ∗ (− div z∗0 + z∗1 + z∗2 ) = (− div z∗0 + z∗1 + z∗2 )2 dx,
2K Ω
v∗ = (v∗0 , v∗1 , v∗2 ) ∈ Y1 ×V ×V , Y1 = C1 (Ω; R3 ) and
G∗0 (v∗0 , z∗0 ) = sup {hv0 , v∗0 + z∗0 iL2 − G0 (v0 )}

v0 ∈Y1
Z
∗ ∗ γ 2
= sup hv0 , v0 + z0 iL2 − |v0 | d x
v0 ∈Y1 2 Ω
1
Z
= |v∗0 + z∗0 |2 dx. (20.5)
2γ Ω
Also,
G∗1 (v∗1 , z∗1 ) = sup{hu, v∗1 + z∗1 iL2 − G1 (u)}

u∈V
Z
α
= sup hu, v∗1 + z∗1 iL2 − u4 dx
u∈V 4 Ω
3
Z
= |v∗ + z∗ |4/3 dx, (20.6)
4α 1/3 Ω 1 1
G∗2 (v∗2 , z∗2 ) = sup{hu, v∗2 + z∗2 iL2 − G2 (u)}

u∈V

(−β + K − ε)
Z
= sup hu, v∗2 + z∗2 iL2 − u2 dx
u∈V 2 Ω
1
Z
= (v∗ + z∗ )2 dx (20.7)
2(K − β − ε) Ω 2 2
and
G∗3 ( div v∗0 − v∗1 − v∗2 + f ) = sup{−h∇u, v∗0 iL2 − hu, v∗1 iL2
u∈V
−hu, v∗2 iL2 + hu, f iL2 − G3 (u)}
= sup{−h∇u, v∗0 iL2 − hu, v∗1 iL2
u∈V
Z
ε
−hu, v∗2 iL2 + hu, f iL2 − u2 dx}
2 Ω
1
Z
= ( div v∗0 − v∗1 − v∗2 + f )2 dx. (20.8)
2ε Ω
At this point, we denote,
J ∗ (v∗ , z∗ ) = −G∗0 (v∗0 + z∗0 ) − G∗1 (v∗1 + z∗1 )

−G∗2 (v∗2 + z∗2 ) − G∗3 ( div v∗0 − v∗1 − v∗2 + f )
+F ∗ (− div z∗0 + z∗1 + z∗2 ). (20.9)
The extremal equation,

∂ J ∗ (v∗ , z∗ )
=0
∂ z∗
gives the following system.
Specifically from
∂ J ∗ (v∗ , z∗ )
= 0,
∂ z∗0
we get
∂ G∗0 (v∗0 + z∗0 )
∗
∂ F (− div z∗0 + z∗1 + z∗2 )

− +∇ = 0,
∂ z∗0 ∂ w∗0
where
w∗0 = − div z∗0
so that
v∗0 + z∗0 −divz∗0 + z∗1 + z∗2

−∇ = 0.
γ K
Hence, defining
−d ivz∗0 + z∗1 + z∗2
û = ,
K
we have
v∗0 = −z∗0 + γ∇û. (20.10)
From
∂ J ∗ (v∗ , z∗ )
= 0,
∂ z∗1
we get
∂ G∗1 (v∗1 + z∗1 ) ∂ F ∗ (− div z∗0 + z∗1 + z∗2 )
− + = 0,
∂ z∗1 ∂ z∗1
so that
(v∗1 + z∗1 )1/3
= û,
α 1/3
that is,
v∗1 = −z∗1 + αû3 . (20.11)
Finally, from
∂ J ∗ (v∗ , z∗ )
= 0,
∂ z∗2
we get
∂ G∗2 (v∗2 + z∗2 ) ∂ F ∗ (− div z∗0 + z∗1 + z∗2 )
− + = 0,
∂ z∗2 ∂ z∗2
so that
(v∗2 + z∗2 )
= û,
K −β −ε
that is,
v∗2 = −z∗2 + (K − β − ε)û. (20.12)
Replacing (20.10), (20.11) and (20.12) into (20.9), we obtain

3
Z Z
γ
J ∗ (v∗ , z∗ ) = − |∇û|2 dx − 1/3 α 4/3 û4 dx
2 Ω 4α Ω
(K − β − ε) K
Z Z
− û2 dx + û2 d x
2 Ω 2 Ω
1
Z
− ( div z∗0 − z∗1 − z∗2 − γ∇2 û + αû3 + (K − β − ε)û − f )2 dx
2ε Ω
3α β +ε
Z Z Z
γ
= − |∇û|2 dx − û4 d x + û2 dx
2 Ω 4 Ω 2 Ω
1
Z
− (−γ∇2 û + αû3 − (β + ε)û − f )2 dx
2ε Ω
≡ Jˆε (û). (20.13)
20.2 The main duality principle

With such statements, definitions and results in mind we prove the following theorem.
Theorem 20.2.1 Considering the context of the last section statements, definitions and results, let J : V → R
be defined by
Z Z
γ α
J(u) = ∇u · ∇u dx + u4 dx
2 Ω 4 Ω
Z
β
− u2 dx − hu, f iL2 , ∀u ∈ V. (20.14)
2 Ω
Here, f ∈ C(Ω). Let Jˆε : V → R be defined by
3α β +ε
Z Z Z
γ
Jˆε (û) = − |∇û|2 dx − û4 dx + û2 dx
2 Ω 4 Ω 2 Ω
1
Z
− (−γ∇2 û + αû3 − (β + ε)û − f )2 dx. (20.15)
2ε Ω
Assume u0 ∈ V is such that
δ J(u0 ) = 0,
and
δ 2 J(u0 ) > 0.
δ Jˆε (u0 ) = 0,
J(u0 ) = Jˆε (u0 )
and for ε > 0 sufficiently small,
(δ 2 J(u0 ) − ε )2
δ 2 Jˆε (u0 ) = −(δ 2 J (u0 ) − ε) −
ε
1
≈ −O
ε
< 0, (20.16)
so that there exist r, r1 > 0 such that
J(u0 ) = inf J(u)

u∈Br (u0 )
= sup Jˆε (û)

û∈Br1 (u0 )
= Jˆε (u0 ). (20.17)
Proof 20.1 From δ J(u0 ) = 0 we have
−γ∇2 u0 + αu30 − β u0 − f = 0, in Ω.
Thus defining û = u0 we have
−γ∇2 u0 + αu30 − (β + ε)u0 − f

= −εu0
= −εû, in Ω. (20.18)
Hence
−γ∇2 u0 + αu30 − (β + ε)u0 − f
uˆ = − .
ε
From such an expression for û we obtain
δ Jˆε (u0 ) = γ∇2 u0 − 3αu30 + (β + ε)u0 − γ∇2 û + 3αuu

ˆ 20 − (β + ε)û. (20.19)
From this, since û = u0 , we get

δ Jˆε (u0 ) = 0.
Also, we may define v∗ , z∗ such that
v∗0 = −z∗0 + γ∇û, (20.20)
v∗1
= −z∗1 + αû3 , (20.21)
v∗2 = −z∗2 + (K − β − ε)û, (20.22)
− div z∗0 + z∗1 + z∗2 = Kû,
so that
div v∗0 − v∗1 − v∗2 = − div z∗0 + z∗1 + z∗2

+γ∇2 û − αû3 − (K − β − ε)û
= γ∇2 û − αû3 + (β + ε)û
= εû − f . (20.23)
From such relations we obtain,
G∗0 (v∗0 , z∗0 ) = h∇û, v∗0 + z∗0 iL2 − G0 (∇û),
G∗1 (v∗1 , z∗1 ) = hû, v∗1 + z∗1 iL2 − G1 (û),

G∗2 (v∗2 , z∗2 ) = hû, v∗2 + z∗2 iL2 − G3 (û),
G∗3 ( div v∗0 − v∗1 − v∗2 + f ) = hû, div v∗0 − v∗1 − v∗2 iL2 + hû, f iL2 − G3 (û)
and
F ∗ ( div z∗0 − z∗1 − z∗2 ) = hû, − div z∗0 + z∗1 + z∗2 iL2 − F(û).
From such results, we obtain
Jˆε (û) = J ∗ (v∗ , z∗ )

= −G∗0 (v∗0 + z∗0 ) − G∗1 (v∗1 + z∗1 )
−G∗2 (v∗2 + z∗2 ) − G∗3 ( div v∗0 − v∗1 − v∗2 + f ) + F ∗ (− div z∗0 + z∗1 + z∗2 )
= G0 (∇û) + G1 (û) + G2 (û) + G3 (û) − F(û) − hû, f iL2
= J(û)
= J(u0 ). (20.24)
Finally, for ε > 0 sufficiently small,
δ 2 Jˆε (u0 ) = γ∇2 − 9αu20 + β + ε

(−γ∇2 + 3αu20 − (β + ε ))2
= − + 6αu20
ε
(−γ∇2 + 3αu20 − (β + ε))2
= γ∇2 − 3αu02 + β + ε −
ε
2
(δ J (u0 ) − ε)2
= −(δ 2 J (u0 ) − ε ) −
ε
1
≈ −O
ε
< 0. (20.25)
From these last results, there exist r, r1 > 0 such that
J(u0 ) = inf J(u)

u∈Br (u0 )
= sup Jˆε (û)

û∈Br1 (u0 )
= Jˆε (u0 ). (20.26)
20.3 Conclusion
In this chapter we have developed a primal dual variational formulation for a large class of problems in the
calculus of variations.
We emphasize the dual functional obtained has a large domain region of concavity about a critical point,
which makes such a formulation very interesting from a numerical analysis point of view.
Finally, it has been formally proven there is no duality gap between the primal and dual formulations in
a local extremal context.
Chapter 21
A Duality Principle and

Concerning Computational
Method for a Class of Optimal
Design Problems in Elasticity
Fabio Silva Botelho and Alexandre Molter
21.1 Introduction
Consider an elastic solid which the volume corresponds to an open, bounded, connected set, denoted by
Ω ⊂ R3 with a regular (Lipschitzian) boundary denoted by ∂ Ω = Γ0 ∪ Γt where Γ0 ∩ Γt = 0/. Consider also
the problem of minimizing the functional Jˆ : U × B → R where
1 1
Jˆ(u,t) = hui , fi iL2 (Ω) + hui , fî iL2 (Γt ) ,
2 2
subject to 
 (Hi jkl (t)ekl (u)), j + fi = 0 in Ω,
(21.1)
Hi jkl (t)ekl (u)n j − fî = 0, on Γt , ∀i ∈ {1, 2, 3}.

Here, n = (n1 , n2 , n3 ) denotes the outward normal to ∂ Ω and
U = {u = (u1 , u2 , u3 ) ∈ W 1,2 (Ω; R3 ) : u = (0, 0, 0) = 0 on Γ0 },
Z
B = t : Ω → [0, 1] measurable : t(x) dx = t1 |Ω| ,
Ω
where
0 < t1 < 1
and |Ω| denotes the Lebesgue measure of Ω.
Moreover, u = (u1 , u2 , u3 ) ∈ W 1,2 (Ω; R3 ) is the field of displacements relating the cartesian system
(0, x1 , x2 , x3 ), resulting from the action of the external loads f ∈ L2 (Ω; R3 ) and fˆ ∈ L2 (Γt ; R3 ).
We also define the stress tensor {σi j } ∈ Y ∗ = Y = L2 (Ω; R3×3 ), by
σi j (u) = Hi jkl (t)ekl (u),
and the strain tensor e : U → L2 (Ω; R3×3 ) by
1
ei j (u) = (ui, j + u j,i ), ∀i, j ∈ {1, 2, 3}.
2
Finally,
{Hi jkl (t)} = {tHi0jkl + (1 − t)Hi1jkl },
where H 0 corresponds to a strong material and H 1 to a very soft material, intending to simulate voids along
the solid structure.
The variable t is the design one, which the optimal distribution values along the structure are intended to
minimize its inner work with a volume restriction indicated through the set B.
The duality principle obtained is developed inspired by the works in [11, 10]. Similar theoretical results
have been developed in [14], however we believe the proof here presented, which is based on the min-max
theorem is easier to follow (indeed we thank an anonymous referee for his suggestion about applying the min-
max theorem to complete the proof). We highlight throughout this text we have used the standard Einstein
sum convention of repeated indices.
Moreover, details on the Sobolev spaces addressed may be found in [1]. In addition, the primal variational
development of the topology optimization problem has been described also in [14].
The main contributions of this work are to present the detailed development, through duality theory,
for such a kind of optimization problems. We emphasize that to avoid the check-board standard and obtain
appropriate robust optimized structures without the use of filters, it is necessary to discretize more in the load
direction, in which the displacements are much larger.
Finally, it is worth mentioning the numerical examples presented have been developed in a Finite Element
(FE) context, based on the work of [70].
21.2 Mathematical formulation of the topology optimization problem

Our mathematical topology optimization problem is summarized by the following theorem.
Theorem 21.2.1 Consider the statements and assumptions indicated in the last section, in particular those
refereing to Ω and the functional Jˆ : U × B → R.
Define J1 : U × B → R by
J1 (u,t) = −G(e(u),t) + hui , fi iL2 (Ω) + hui , fî iL2 (Γt ) ,
where
1
Z
G(e(u),t) = Hi jkl (t)ei j (u)ekl (u) dx,
2 Ω
and where
dx = dx1 dx2 dx3 .
Define also J ∗ : U → R by
J ∗ (u) = inf {J1 (u,t)}

t∈B
= inf{−G(e(u),t) + hui , fi iL2 (Ω) + hui , fî iL2 (Γt ) }. (21.2)
t∈B
A Duality Principle for a Class of Optimal Design Problems 429
Assume there exists c0 , c1 > 0 such that
Hi0jkl zi j zkl > c0 zi j zi j
and
Hi1jkl zi j zkl > c1 zi j zi j , ∀z = {zi j } ∈ R3×3 , such that z 6= 0.
Finally, define J : U × B → R ∪ {+∞} by
J(u,t) = Jˆ(u,t) + Ind(u,t),
where
if (u,t) ∈ A∗ ,

0,
Ind(u,t) = (21.3)
+∞, otherwise ,
where A∗ = A1 ∩ A2 ,
A1 = {(u,t) ∈ U × B : (σi j (u)), j + fi = 0, in Ω, ∀i ∈ {1, 2, 3}}
and
A2 = {(u,t) ∈ U × B : σi j (u)n j − fî = 0, on Γt , ∀i ∈ {1, 2, 3}}.
Under such hypotheses, there exists (u0 ,t0 ) ∈ U × B such that
J(u0 ,t0 ) = inf J(u,t)

(u,t)∈U×B
∗
= sup J (û)
û∈U
∗
= J (u0 )
= Jˆ(u0 ,t0 )
= inf G∗ (σ ,t)
(t,σ )∈B×C∗
= G∗ (σ (u0 ),t0 ), (21.4)
where
G∗ (σ ,t) = sup{hvi j , σi j iL2 (Ω) − G(v,t)}

v∈Y
1
Z
= H i jkl (t)σi j σkl dx, (21.5)
2 Ω
{H i jkl (t)} = {Hi jkl (t)}−1

and C∗ = C1 ∩C2 , where
C1 = {σ ∈ Y ∗ : σi j, j + fi = 0, in Ω, ∀i ∈ {1, 2, 3}}
and
C2 = {σ ∈ Y ∗ : σi j n j − fî = 0, on Γt , ∀i ∈ {1, 2, 3}}.

inf J(u,t) = inf inf J(u,t)
(u,t)∈U×B t∈B u∈U
Z
1
= inf sup inf Hi jkl (t)ei j (u)ekl (u) dx
t∈B û∈U u∈U 2 Ω
+hûi , (Hi jkl (t)ekl (u)), j + fi iL2 (Ω)

ooo
−hûi , Hi jkl (t)ekl (u)n j − fî iL2 (Γt )
Z
1
= inf sup inf Hi jkl (t)ei j (u)ekl (u) dx
t∈B û∈U u∈U 2 Ω
Z
− Hi jkl (t)ei j (û)ekl (u) dx
Ω
ooo
+hûi , fi iL2 (Ω) + hûi , fî iL2 (Γt )
Z
= inf sup − Hi jkl (t)ei j (û)ekl (û) dx
t∈B û∈U Ω
oo
hûi , fi iL2 (Ω) + hûi , fî iL2 (Γt )

∗
= inf inf∗ G (σ ,t) . (21.6)
t∈B σ ∈C
Also, from this and the min-max theorem, there exist (u0 ,t0 ) ∈ U × B such that

inf J(u,t) = inf sup J1 (u,t)
(u,t)∈U×B t∈B û∈U

= sup inf J1 (u,t)
u∈U t∈B
= J1 (u0 ,t0 )
= inf J1 (u0 ,t)
t∈B
= J ∗ (u0 ). (21.7)
Finally, from the extremal necessary condition
∂ J1 (u0 ,t0 )
=0
∂u
we obtain
(Hi jkl (t0 )ekl (u0 )), j + fi = 0 in Ω,
and
Hi jkl (t0 )ekl (u0 )n j − fî = 0 on Γt , ∀i ∈ {1, 2, 3},
so that
1 1
G(e(u0 )) = h(u0 )i , fi iL2 (Ω) + h(u0 )i , fî iL2 (Γt ) .
2 2
Hence, (u0 ,t0 ) ∈ A∗ so that Ind(u0 ,t0 ) = 0 and σ (u0 ) ∈ C∗ .
Moreover
J ∗ (u0 ) = −G(e(u0 )) + h(u0 )i , fi iL2 (Ω) + h(u0 )i , fî iL2 (Γt )
= G(e(u0 ))
= G(e(u0 )) + Ind(u0 ,t0 )
= J(u0 ,t0 )
= G∗ (σ (u0 ),t0 ). (21.8)
21.3 About the computational method

The continuous topology optimization problem described in the previous section is discretized using the
FE method, considering in plane deformations. The FE discretization is performed taking into account the
bilinear isoparametric element as a master one, in similar way as in [70].
To obtain computational results, we have defined the following algorithm.
1. Set n = 1.
2. Set t1 (x) = t1 , in Ω.
3. Calculate un ∈ U as the solution of equation
∂ J1 (u,tn )
= 0,
∂u
that is, 
 (Hi jkl (tn )ekl (un )), j + fi = 0 in Ω,
(21.9)
Hi jkl (tn )ekl (un )n j − fî = 0, on Γt , ∀i ∈ {1, 2, 3}.

4. Obtain tn+1 by
tn+1 = arg min J1 (un ,t).
t∈B
5. Set n := n + 1 and go to step 3 up to the satisfaction of an appropriate convergence criterion.

In the FE formulation, equations indicated in (21.9) stands for
H(t)U = f, (21.10)
where H(t) is the global stiffness matrix, U is the global deflection and f is the global forces vector.
Thus, for such a FE models (N elements where e ∈ {1, ..., N}), the primal optimization problem can be
written in matrix form as
1
min Jˆ(u,t) = UT H(t)U
2
1 N
= ∑ (te ) p uTe He ue
2 e=1
s.t. (te ) p He ue = fe (21.11)
N
∑ teVe = t1 |Ω|
e=1
0≤t ≤1
e = 1, 2, 3, ..., N,
On the other hand, the dual problem may be expressed by
max J ∗ (u), where !

1 N
J ∗ (u) = min p T

− ∑ (te ) ue He ue + fe ue
t∈B 2 e=1
where t ∈ B if and only if
N
(21.12)
∑ teVe = t1 |Ω|
e=1
0 ≤ te ≤ 1,
e = 1, 2, 3, ..., N,
and where Ve is the area of element e.

Finally, the last minimization indicated corresponds to item 4 in the concerning algorithm. Indeed, such
a procedure refers to minimize at each iteration, through the Matlab Linprog routine, the function
N N
∂ J1 (un , {ten })
te = ∑ − p(ten ) p−1te uTe He ue

∑ ∂t e
e=1 e=1
subject to t ∈ B, where p is a penalization parameter (typically, p = 3).
21.4 Computational simulations and results

We present numerical results in an analogous two-dimensional context, more specifically for two-
dimensional beams of dimensions 1 × l (units refer to the international system) represented by Ω =
[0, 1] × [0, l ], with l = 0.5, F = −106 for the first case, l = 0.5 and F = −107 for the second one, l = 0.6
and F = −106 for the third case and, l = 1 and F = −108 for the fourth one. F is in the y-direction and
corresponds to f of the theoretical formulation presented above.
We consider the strain tensor as
e(u) = (ex (u), ey (u), exy (u))T ,
where u = (u, v) ∈ W 1,2 (Ω; R2 ), ex (u) = ux , ey (u) = vy and exy (u) = 21 (uy + vx ).
Moreover, the stress tensor σ (e(u)) is given by
σ (e(u)) = He(u),
where  
1 ν 0 
E(t) 
H= ν 1 0 (21.13)
1 − ν2  1
0 0 (1 −

2 ν)
and
E(t) = tE0 + (1 − t)E1 ,
where E0 = 210 ∗ 109
(the modulus of Young) and E1 E0 . Moreover, ν = 0.33.
As previously mentioned, we present four numerical simulations.
Case 1. For the first case see Figure 21.1, on the left, for the concerning case, Figure 21.1, in the middle,
for the optimal topology for this case with no filter, Figure 21.1, on the right, for the optimal topology for
this first case with filter. The objective function as function of iteration numbers also for such a case with no
filter, Figure 21.2, on the left, and the objective function as function of iteration numbers also for this first
case with filter, Figure 21.2, on the right.
0.5 0.5
0 0
1 1
Figure 21.1: On the left, a clamped beam at x = 0 (cantilever beam). In the middle, the optimal topology for t1 = 0.5, for
the case with no filter. On the right, the optimal topology for t1 = 0.5, for the case with filter. The FE mesh was 60 x 50.
35.2 200
35.1
150
35
34.9 100
34.8
50
34.7
34.6 0
0 2 4 6 8 10 12 0 20 40 60 80 100
Figure 21.2: On the left, the objective function by iteration numbers for t1 = 0.5, for the case with no filter. On the right,
the objective function by iteration numbers for t1 = 0.5, for the case with filter.
0.5 0.5
0 0
1 1
Figure 21.3: On the left, a simply supported beam at x = 0 and x = 1. In the middle the optimal topology for t1 = 0.5,
for the case with no filter. On the right the optimal topology for t1 = 0.5, for the case with filter. The FE mesh was 40 x
50.
Case 2. For the second case see Figure 21.3, on the left, for the concerning case, Figure 21.3, in the middle,
for the optimal topology for this case with no filter, Figure 21.3, on the right, for the optimal topology for this
second case with filter. The objective function as function of iteration numbers also for such a case with no
filter, Figure 21.4, on the left, and the objective function as function of iteration numbers also for this second
Case 3. For the third case see Figure 21.5, on the left, for the concerning case, Figure 21.5, in the middle,
for the optimal topology for this case with no filter, Figure 21.5, on the right, for the optimal topology for
this third case with filter. The objective function as function of iteration numbers also for such a case with no
×10 4
8040 4.5
8020 4
3.5
8000
3
7980
2.5
7960
2
7940
1.5
7920 1
7900 0.5
0 2 4 6 8 10 12 0 20 40 60 80 100
Figure 21.4: On the left the objective function by iteration numbers for t1 = 0.5, for the case with no filter. On the right
0.6 0.6
0 0
1 1
Figure 21.5: On the left a beam with a hole clamped at x = 0. In the middle the optimal topology for t1 = 0.5, for the
case with no filter. On the right the optimal topology for t1 = 0.5, for the case with filter. The FE mesh was 50 x 40.
76.5 600
76.4 500
76.3 400
76.2 300
76.1 200
76 100
75.9 0
0 2 4 6 8 10 0 20 40 60 80 100
filter, Figure 21.6, on the left, and the objective function as function of iteration numbers also for this third
Case 4. For the fourth case see Figure 21.7, on the left, for the concerning case, Figure 21.7, in the middle,
for the optimal topology for this case with no filter, Figure 21.7, on the right, for the optimal topology for this
fourth case with filter. The objective function as function of iteration numbers also for such a case with no
1 1
0.5 0.5
0 0
0 0.5 1 0.5 1
Figure 21.7: On the left a L shape beam clamped at y = 1. In the middle the optimal topology for t1 = 0.5, for the case
with no filter. On the right the optimal topology for t1 = 0.5, for the case with filter. The FE mesh was 40x60.
×10 5 ×10 6
5.7 14
5.65
12
10
5.6
8
5.55
6
5.5
4
5.45
2
5.4 0
0 2 4 6 8 10 0 20 40 60 80 100
filter, Figure 21.8, on the left, and the objective function as function of iteration numbers also for this fourth
We emphasize to have obtained in both optimized structures, without filter and with filter, robust topology
from a structural point of view. One can note also in the figures that in all cases the objective functions,
without filter and with filter, have similar final values, which indicates the results obtained are consistent.
21.5 Final remarks and conclusions

In this chapter we have developed a duality principle and relating computational method for a class of struc-
tural optimization problems in elasticity. It is worth mentioning we have not used a filter to post-process the
results, having obtained a solution t : Ω → {0, 1} (that is, t(x, y) = 0 or t(x, y) = 1 in Ω), by finding a critical
point (u0 ,t0 ) ∈ U × B for the functional J1 : U × B → R. This corresponds, in some sense, to solving the dual
problem.
We address some final remarks and conclusions on the results obtained.
For all examples, in a first step, we have obtained numerical results through our algorithm with a
software which uses the Matlab-Linprog as optimizer at each iteration without any filter. In a second
step, we obtain numerical results using the OC optimizer with filter, with a software developed based
in the article [70] by Sigmund, 2001.
We emphasize, to obtain good and consistent results, it is necessary to discretize more in the direction
y, that is, the load direction, in which the displacements are much larger.
If we do not discretize enough in the load direction, for the software with no filter, a check-board
standard in the material distribution is obtained in some parts of the concerning structure.
Summarizing, with no filter, the check-board problem is solved by increasing the discretization in
the load direction.
Moreover, with the OC optimizer with filter, the volume fraction of material is kept constant in 0.5
at each iteration during the optimization process, whereas for the case with no filter we start with a
volume fraction of 0.95 which is gradually decreased to the value 0.5, using as the initial solution for
a iteration with a specific volume fraction, the solution of the previous one.
We also highlight the result obtained with no filter is indeed a critical point for the original optimiza-
tion problem, whereas there is some heuristic in the procedure with filter.
Once more we emphasize to have obtained more robust and consistent shapes by properly discretizing
the approximate model in a FE context.
Finally, it is also worth mentioning, we have obtained similar final objective function values without
and with filter in all examples, even though without filter such values have been something smaller,
as expected. The qualitative differences between the graphs without and with filter, for the objective
function as function of the number of iterations, refer to the differences between the optimization
processes, where in the case with filter the volume fraction is kept 0.5 and without filter it is gradually
decreased from 0.95 to 0.5, as above described.
We highlight the results obtained may be applied to other problems, such as other models of plates, shells
and elasticity.
Chapter 22
Existence and Duality Principles

for the Ginzburg-Landau System
in Superconductivity
22.1 Introduction
In this work we present three theorems which represent duality principles suitable for a large class of non-
convex variational problems.
At this point we refer to the exceptionally important article, “A contribution to contact problems for a
class of solids and structures” by Bielski and Telega [11], published in 1985, as the first one to successfully
apply and generalize the convex analysis approach to a model in non-convex and non-linear mechanics.
The present work is, in some sense, a kind of extension of this previous work [11] and others such as
[10], which greatly influenced and inspired my work and recent book [14].
We extend and generalize the approaches in [11] and [78] and develop two multi-duality principles
through which we classify qualitatively the critical points.
Thus, we emphasize the first multi-duality principle generalizes some Toland results found in [78] and in
some appropriate sense, such a work complements the results presented in [11], now applied to a Ginzburg-
Landau type model context.
On the other hand, the conclusions of the second multi-duality principle may be qualitatively found in
similar form in the triality approach found in [83] and other references therein, even though the construction
of the present result and the respective proofs be substantially different.
About the model in physics involved, we recall that about the year 1950, Ginzburg and Landau introduced
a theory to model the super-conducting behavior of some types of materials below a critical temperature Tc ,
which depends on the material in question. They postulated the free density energy may be written close to
Tc as
h̄ α(T ) β (T )
Z Z Z
Fs (T ) = Fn (T ) + |∇ψ|22 dx + |ψ|4 dx − |ψ|2 dx,
4m Ω 4 Ω 2 Ω
where ψ is a complex parameter, Fn (T ) and Fs (T ) are the normal and super-conducting free energy densi-
ties, respectively (see [4, 55] for details). Here Ω ⊂ R3 denotes the super-conducting sample with a boundary
denoted by ∂ Ω = Γ. The complex function ψ ∈ W 1,2 (Ω; C) is intended to minimize Fs (T ) for a fixed tem-
perature T .
Denoting α(T ) and β (T ) simply by α and β , the corresponding Euler-Lagrange equations are given by:
h̄
 − 2m ∇2 ψ + α|ψ|2 ψ − β ψ = 0, in Ω

(22.1)
 ∂ψ
∂n = 0, on ∂ Ω.
This last system of equations is well known as the Ginzburg-Landau (G-L) one in the absence of a magnetic
field and respective potential.
Remark 22.1 About the notation, for an open bounded subset Ω ⊂ R3 , we denote the L2 (Ω) norm by
k · kL2 (Ω) or simply by k · k2 . Similar remark is valid for the L2 (Ω; R3 ) norm, which is denoted by k · kL2 (Ω;R3 )
or simply by k · k2 , when its meaning is clear. On the other hand, by | · |2 we denote the standard Euclidean
norm in R3 or C3 .
Moreover derivatives are always understood in the distributional sense. Also, by a regular Lipschitzian
boundary ∂ Ω = Γ of Ω, we mean regularity enough so that the standard Sobolev Imbedding theorems, the
trace theorem and Gauss-Green formulas of integration by parts to hold. Details about such results may be
found in [1].
´
Finally, in general δ F(u, v) will denote the Frechet derivative of the functional F(u, v) at (u, v),
∂ F(u, v)
δu F(u, v) or
∂u
´
denotes the first Frechet derivative of F relating the variable u and
∂ 2 F(u, v)
δ 2 Fu,v (u, v) or
∂ u∂ v
denotes the second one relating the variables u and v, always at (u, v).
At some point of our analysis, we shall assume a finite-dimensional approximate model version. In such
a context, we remark that generically in a matrix sense the notation
1
K + γ∇2
will indicate the inverse
(KId + γ∇2 )−1 ,
where Id denotes the identity matrix and ∇2 is the matrix originated by a discretized version of the Laplace
operator. We also emphasize that, as the meaning is clear, other similar notations may be used to indicate the
inverse of matrices or operators.
Remark 22.2 For an appropriate set Ω ⊂ R3 and a space U, our primal functional J : U → R is specified
by Z Z
γ α
J(u) = ∇u · ∇u dx + (u2 − β )2 dx − hu, f iL2
2 Ω 2 Ω
∀u ∈ U, where α, β , γ > 0, f ∈ L2 (Ω).

We define,
K
Z Z
γ
F(u) = − ∇u · ∇u dx + u2 dx,
2 Ω 2 Ω
where in a finite dimensional discretized model version, in a finite elements or finite differences context,
K > 0 is such that
F(u) > 0, ∀u ∈ U such that u =6 0,
Existence and Duality Principles for the Ginzburg-Landau System in Superconductivity 439
and
K
Z Z
α
G(u, v) = (u2 − β + v)2 dx + u2 dx − hu, f iL2 ,
2 Ω 2 Ω
so that
J(u) = G(u, 0) − F(u), ∀u ∈ U.
Let F ∗ : Y ∗ → R denote the polar functional related to F, that is,
F ∗ (v∗1 ) = sup{hu, v∗1 iL2 − F(u)}.

u∈U
Since the optimization in question is quadratic, we have
F ∗ (v∗1 ) = hũ, v∗1 iL2 − F(ũ), (22.2)
where ũ ∈ U is such that

∂ F(ũ)
v∗1 = ,
∂u
that is,
v∗1 = γ∇2 ũ + Kũ,
so that
v1∗
ũ = .
K + γ∇2
Replacing such a ũ into (22.2), we obtain
1
Z
F ∗ (v∗1 ) = v∗1 [(K + γ∇2 )−1 v∗1 ] dx.
2 Ω
Similarly, for G : U ×Y → R, for v∗0 ∈ Y ∗ such that
2v∗0 + K > 0, in Ω,
we also define
G∗ (v∗1 , v∗0 ) = sup {hu, v∗1 iL2 + hv, v∗0 iL2 − G(u, v)},
(u,v)∈U×Y
where, as indicated above

K
Z Z
α
G(u, v) = (u2 − β + v)2 dx + u2 dx − hu, f iL2 .
2 Ω 2 Ω
Defining w = u2 − β + v, so that
v = w − u2 + β ,
we may write
G∗ (v∗1 , v∗0 ) = sup {hu, v∗1 iL2 + hw − u2 + β , v∗0 iL2

(u,w)∈U ×Y
K
Z Z
α
− w2 dx − u2 dx + hu, f iL2 }. (22.3)
2 Ω 2 Ω
Since the optimization in question is quadratic, we have
G∗ (v∗1 , v∗0 ) = hũ, v∗1 iL2 + hw̃ − ũ2 + β , v∗0 iL2

K
Z Z
α
− w̃2 dx − ũ2 dx + hũ, f iL2 }. (22.4)
2 Ω 2 Ω
where ũ ∈ L2 and w̃ ∈ Y are such that

v∗1 = 2uv
˜ ∗0 + Kũ − f ,
and
v∗0 = αw̃,
so that
v∗1 + f
ũ = ,
2v∗0 + K
and
v∗0
w̃ = .
α
Replacing such results into (22.4), we get
1 (v∗1 + f )2 1
Z Z
G∗ (v∗1 , v∗0 ) = ∗ dx + (v∗0 )2 dx
2 Ω 2v0 + K 2α Ω
Z
+β v∗0 dx. (22.5)
Ω
22.2 The main result

In this section we develop a first multi-duality principle for a Ginzburg-Landau type system in a simpler real
context.
About these models in physics, we refer again to [4, 55].
In the next lines we develop the main result. At this point we highlight the global optimality condition
−γ∇2 + 2v̂∗0 > 0 at a critical point was presented in [83].
denoted by ∂ Ω. Suppose J : U → R is a functional defined by
Z Z
γ α
2 Ω 2 Ω
∀u ∈ U, where α, β , γ > 0, f ∈ L2 (Ω) and U = W01,2 (Ω).

At this point, we assume a discretized finite dimensional model version (in a finite elements or finite
differences context, so that from now on, the not relabeled spaces, functions and operators refer to such a
finite dimensional approximation) and suppose
δ J(u0 ) = 0
where in an appropriate matrices sense, we have
δ 2 J(u0 ) = −γ∇2 + 6α {u0 (i)2 } − 2αβ Id .
Here {u0 (i)2 } denotes the diagonal matrix which the diagonal is given by the vector [u0 (i)2 ].
Also, from now on, as the meaning is clear, we shall denote such a second Frec´ het derivative simply by
δ 2 J(u0 ) = −γ∇2 + 6αu20 − 2αβ .
Define
K
Z Z
γ
F(u) = − ∇u · ∇u dx + u2 dx,
2 Ω 2 Ω
where K > 0 is such that

F(u) > 0, ∀u ∈ U such that u 6= 0,
and
K
Z Z
α
G(u, v) = (u2 − β + v)2 dx + u2 dx − hu, f iL2 ,
2 Ω 2 Ω
so that
J(u) = G(u, 0) − F(u), ∀u ∈ U.
Define also Y = Y∗ = L2 (Ω), F∗ : Y ∗ → R by
F ∗ (v∗1 ) = sup{hu, v∗1 iL2 − F(u)}
u∈U
1
Z
= v∗1 [(K + γ∇2 )−1 v∗1 ] dx, (22.6)
2 Ω
and G∗ : Y ∗ ×Y ∗ → R = R ∪ {+∞} by
G∗ (v∗1 , v∗0 ) = sup {hv, v∗0 iY + hv1 , v∗1 iY − G(v1 , v)}
(v1 ,v)∈Y ×Y
1 (v∗1 + f )2 1
Z Z
= ∗ dx + (v0∗ )2 dx
2 Ω 2v0 + K 2α Ω
Z
+β v∗0 dx, (22.7)
Ω
if v∗0 ∈ A∗ = {v∗0 ∈ Y ∗ : 2v∗0 + K > 0, in Ω}.

Let J ∗ : Y ∗ ×Y ∗ → R ∪ {−∞} be such that
J ∗ (v∗1 , v∗0 ) = −G∗ (v∗1 , v∗0 ) + F ∗ (v∗1 ),
so that J˜∗ : Y ∗ → R is expressed by
J˜∗ (v∗1 ) = sup J ∗ (v∗1 , v∗0 ).
v∗0 ∈A∗
Define
v̂∗0 = α(u20 − β ),
and
v̂∗1 = (2v̂∗0 + K)u0 − f .
Suppose v̂∗0 ∈ A∗ .
δ J˜∗ (v̂∗1 ) = 0,
and
J˜∗ (v̂∗1 ) = J(u0 ).
Moreover,
1. if δ 2 J(u0 ) > 0, then
δ 2 J˜∗ (v̂∗1 ) > 0,
so that there exist r > 0 and r1 > 0 such that
J(u0 ) = min J(u)
u∈Br (u0 )
= min J˜∗ (v∗1 )

v∗1 ∈Br1 (v̂∗1 )
= J˜∗ (v̂∗1 )
= J ∗ (v̂∗1 , v̂∗0 ). (22.8)
2. If −γ∇2 + 2v̂∗0 > 0 so that δ 2 J(u0 ) > 0, then defining
B∗ = {v∗0 ∈ Y ∗ : −γ∇2 + 2v∗0 > 0},
we have
δ J2∗ (v̂∗1 ) = 0,
δ 2 J2∗ (v̂∗1 ) > 0
and
J(u0 ) = min J(u)

u∈U
= min J2∗ (v∗1 )
v∗1 ∈Y ∗
= J2∗ (v̂∗1 )
= J ∗ (v̂∗1 , v̂∗0 ), (22.9)
where
J2∗ (v∗1 ) = sup J ∗ (v∗1 , v∗0 ).
v∗0 ∈A∗ ∩B∗
3. If δ 2 J(u0 ) < 0, then

δ 2 J˜∗ (v̂∗1 ) < 0,
so that there exist r > 0 and r1 > 0 such that
J(u0 ) = max J(u)

u∈Br (u0 )
= max J˜∗ (v∗1 )

v∗1 ∈Br1 (v̂∗1 )
= J˜∗ (v̂∗1 )
= J ∗ (v̂∗1 , v̂∗0 ). (22.10)
−γ∇2 u0 + α(u20 − β )2u0 − f = 0, in Ω

so that
γ∇2 u0 + Ku0 = α(u20 − β )2u0 + Ku0 − f

= (2v̂∗0 + K)u0 − f
= v̂∗1 . (22.11)

v̂∗1 v̂∗ + f
2
− 1∗ = u0 − u0 = 0, in Ω.
K + γ∇ 2v̂0 + K
On the other hand,
v̂∗0 = α(u20 − β ) = α(u20 − β + 0),
so that
v̂0∗ (v̂∗ + f )2
− + 1∗ − β = 0,
α (2v̂0 + K )2
and thus
∂ J ∗ (v̂∗1 , v̂∗0 )
= 0.
∂ v∗0
From this, (22.11), from the definition of v̂∗0 and the concavity of J ∗ (v̂∗1 , v∗0 ) in v∗0 , we obtain
J˜∗ (v̂∗1 ) = sup J ∗ (v̂∗1 , v∗0 )

v∗0 ∈A∗
= J ∗ (v̂∗1 , v̂∗0 )
= −G∗ (v̂∗1 , v̂∗0 ) + F ∗ (v̂∗1 )
= −hu0 , v̂∗1 iL2 − h0, v̂∗0 iL2 + G(u0 , 0)
+hu0 , v̂∗1 iL2 − F(u0 )
= G(u0 , 0) − F(u0 )
= J(u0 ). (22.12)
Also, for v∗1 in a neighborhood of v̂∗1 we have that
J ∗ (v∗1 ) = sup J ∗ (v∗1 , v∗0 ) = J ∗ (v∗1 , ṽ∗0 ),

v∗0 ∈A∗
where such a supremum is attained through the equation,

∂ J ∗ (v∗1 , ṽ∗0 )
= 0,
∂ v∗0
so that from the implicit function theorem we have
∂ J˜∗ (v∗1 ) ∂ J ∗ (v∗1 , ṽ∗0 ) ∂ J ∗ (v∗1 , ṽ∗0 ) ∂ ṽ∗0

= +
∂ v∗1 ∂ v∗1 ∂ v∗0 ∂ v∗1
∗ ∗
∂ J (v1 , ṽ0 )∗
= . (22.13)
∂ v∗1
Moreover, from this, joining the pieces, we get
∂ J˜∗ (v̂∗1 ) ∂ J ∗ (v̂∗1 , v̂∗0 ) ∂ J ∗ (v̂∗1 , v̂∗0 ) ∂ v̂∗0

= +
∂ v∗1 ∂ v∗1 ∂ v∗0 ∂ v∗1
∂ J ∗ (v̂∗1 , v̂∗0 )
=
∂ v∗1
v̂∗1 v̂∗ + f
= 2
− 1∗ = u0 − u0 = 0, in Ω. (22.14)
K + γ∇ 2v̂0 + K
Hence, from this and (22.13), we obtain
∂ 2 J˜∗ (v̂∗1 ) ∂ 2 J ∗ (v̂∗1 , v̂0∗ ) ∂ v̂∗0

=
∂ (v∗1 )2 ∂ v∗0 ∂ v∗1 ∂ v∗1
∂ 2 J ∗ (v̂∗1 , v̂∗0 )
+
∂ (v∗1 )2
2(v̂∗1 + f ) ∂ v̂∗0
=
(2v̂∗0 + K)2 ∂ v∗1
1 1
+ − . (22.15)
K + γ∇2 2v̂∗0 + K
At this point we may observe that
(v̂∗1 + f )2 v̂∗
∗ 2
− 0 − β = 0,
(2v̂0 + K) α
so that taking the variation in v∗1 of this equation in both sides, we obtain
2(v̂∗1 + f ) 4(v̂∗1 + f )2 ∂ v̂∗0

∗ −
(2v̂0 + K) 2 (2v̂∗0 + K)3 ∂ v∗1
1 ∂ v̂∗0
− = 0, (22.16)
α ∂ v∗1
and thus,
2(v̂∗1 + f )
∂ v̂∗0 (2v̂∗0 +K)2
= , (22.17)
∂ v∗1 1 4(v̂∗ + f )2
+ (2vˆ∗1+K)3
α 0
Replacing (22.17) into (22.15), we obtain
∂ 2 J˜∗ (v̂∗1 ) ∂ 2 J ∗ (v̂∗1 , v̂∗0 ) ∂ v̂∗0

=
∂ (v∗1 )2 ∂ v∗0 ∂ v∗1 ∂ v∗1
∂ 2 J ∗ (v̂∗1 , v̂∗0 )
+
∂ (v∗1 )2
2(v̂∗1 + f ) ∂ v̂∗0
=
(2v̂∗0 + K)2 ∂ v∗1
1 1
+ 2
− ∗
K + γ∇ 2v̂0 + K
4α(v̂∗1 + f )2
(2v̂∗0 +K )4
=
4(vˆ∗ + f )2
1 + α (2v̂∗1+K )3
0
1 1
+ 2
− ∗ , (22.18)
K + γ∇ 2v̂0 + K
Therefore, denoting
4(v̂∗1 + f )2 4αu20
H = 1+α ∗ = 1 + ,
(2v̂0 + K)3 2v̂∗0 + K
we have
∂ 2 J˜∗ (v̂∗1 )
∂ (v∗1 )2
4(v̂∗1 + f )2

1 1
= 1+α − /H
(2v̂∗0 + K)3 K + γ∇2 2v̂∗0 + K
4αu2

1 1
= 1+ ∗ 0 − /H
2v̂0 + K K + γ∇2 2v̂∗0 + K
= (2v̂∗0 + K + 4αu02 − K − γ∇2 )/[(K + γ∇2 )(v̂∗0 + K)H]
= [−γ∇2 + 6αu20 − 2αβ ]/[(K + γ∇2 )(v̂∗0 + K)H]
δ 2 J(u0 )
= . (22.19)
[(K + γ∇2 )(v̂∗0 + K)H]
Summarizing, assuming δ 2 J(u0 ) > 0, we obtain

∂ 2 J˜∗ (v̂∗1 )
> 0,
∂ (v∗1 )2
so that u0 ∈ U is a point of local minimum for J and v̂∗1 is a point of local minimum for J˜∗ .
Hence, there exist r > 0 and r1 > 0 such that
J(u0 ) = min J(u)
u∈Br (u0 )
= min J˜∗ (v∗1 )

v∗1 ∈Br1 (v̂∗1 )
= J˜∗ (v̂∗1 )
= J ∗ (v̂∗1 , v̂∗0 ). (22.20)
Assume now −γ∇2 + 2v̂∗0 > 0, so that δ 2 J(u0 ) > 0.
Observe that if v∗0 ∈ A∗ and −γ∇2 + 2v∗0 > 0, then
∂ 2 J ∗ (v1 , v∗0 ) 1 1
∗ 2
= 2
− ∗ > 0,
∂ (v1 ) K + γ∇ 2v0 + K
so that J ∗ (v∗1 , v∗0 ) is convex in v∗1 , ∀v∗0 ∈ A∗ ∩ B∗ .
Hence
J2∗ (v∗1 ) = sup J ∗ (v∗1 , v∗0 ),
v∗0 ∈A∗ ∩B∗
is convex, as the point-wise supremum of a family of convex functionals.

As above, from δ J(u0 ) = 0, we may obtain δ J2∗ (v̂∗1 ) = 0, so that, since J2∗ is convex, we may infer that
J(u0 ) = J2∗ (v̂∗1 ) = min
∗
J ∗ (v∗1 ).
∗ 2 v1 ∈Y
Moreover,
J2∗ (v̂∗1 ) = min J2∗ (v∗1 )
v∗1 ∈Y ∗
≤ J2∗ (v∗1 )
= sup J ∗ (v∗1 , v∗0 )
v∗0 ∈A∗ ∩B∗
(v∗1 )2
Z
1
= sup dx
∗ ∗
v ∈A ∩B∗ 2 Ω (K + γ∇2 )
0
(v∗1 + f )2 (v∗0 )2

1 1
Z Z Z
− ∗ dx − dx − β v∗0 dx
2 Ω 2v0 + K 2 Ω α Ω
(v∗1 )2
Z
1
≤ sup 2
dx
v∗ ∈A∗ ∩B∗ 2 Ω (K + γ∇ )
0
u2 (v∗0 )2

1
Z Z Z
−hu, v∗1 + f iL2 + (2v∗0 + K) dx − dx − β v∗0 dx
Ω 2 2 Ω α Ω
(v∗1 )2
Z
1
≤ sup 2
dx
v∗ ∈Y ∗ 2 Ω (K + γ∇ )
0
u2 (v∗0 )2

1
Z Z Z
−hu, v∗1 + f iL2 + (2v∗0 + K) dx − dx − β v∗0 dx
Ω 2 2 Ω α Ω
1 (v∗1 )2
Z
= dx
2 Ω (K + γ∇2 )
K
Z Z
α
−hu, v∗1 iL2 + (u2 − β )2 dx + u2 dx
2 Ω 2 Ω
−hu, f iL2 , ∀u ∈ U, v∗1 ∈ Y ∗ . (22.21)
Hence,
(v∗1 )2
Z
1
J2∗ (v̂∗1 ) ≤ inf ∗ dx
∗
v1 ∈Y 2 Ω (K + γ∇2 )

K
Z Z
α
−hu, v∗1 iL2 (u − β ) dx +
+ 2 2 2
u dx − hu, f iL2
Ω 2 2 Ω
K K
Z Z Z Z
γ α
= ∇u · ∇u dx − u2 dx + (u2 − β )2 dx + u2 dx − hu, f iL2
2 Ω 2 Ω 2 Ω 2 Ω
= J(u), ∀u ∈ U. (22.22)
From this and

J(u0 ) = J2∗ (v̂∗1 ),
we obtain
J(u0 ) = min J(u)

u∈U
= min J2∗ (v∗1 )
v∗1 ∈Y ∗
= J2∗ (v̂∗1 )
= J ∗ (v̂∗1 , v̂∗0 ). (22.23)
The third item may be proven similarly as the first one.

22.3 A second multi-duality principle

In this section, in a similar context, we present a second multi-duality principle. This principle is significantly
different from the previous one, since we invert the order of variables as evaluating the extremals.
Indeed the first part of this proof is similar to the one of the previous theorem. Important differences
appears along the proof. For the sake of completeness, we present such a proof in details.
denoted by ∂ Ω. Suppose J : U → R is a functional defined by
Z Z
γ α
2 Ω 2 Ω
∀u ∈ U , where α , β , γ > 0, f ∈ L2 (Ω) and U = W01,2 (Ω).

At this point, we assume a discretized finite dimensional model version (in a finite elements or finite
differences context, so that from now on, the not relabeled spaces, functions and operators refer to such a
finite dimensional approximation) and suppose
δ J(u0 ) = 0
where in an appropriate matrices sense, we have
δ 2 J(u0 ) = −γ∇2 + 6αu20 − 2αβ .

Define,
K
Z Z
γ
F(u) = − ∇u · ∇u dx + u2 dx,
2 Ω 2 Ω
where K > 0 is such that
F(u) > 0, ∀u ∈ U such that u =
6 0,
and
K
Z Z
α 2 2
G(u, v) = (u − β + v) dx + u2 dx − hu, f iL2 ,
2 Ω 2 Ω
so that
J(u) = G(u, 0) − F(u), ∀u ∈ U.
Define also Y = Y∗ = L2 (Ω), F∗ : Y ∗ → R by
F ∗ (v∗1 ) = sup{hu, v∗1 iL2 − F(u)}
u∈U
1
Z
= v∗1 [(K + γ∇2 )−1 v∗1 ] dx, (22.24)
2 Ω
and G∗ : Y ∗ ×Y ∗ → R = R ∪ {+∞} by
G∗ (v∗1 , v∗0 ) = sup {hv, v∗0 iY + hv1 , v∗1 iY − G(v1 , v)}
(v1 ,v)∈Y ×Y
1 (v∗1 + f )2 1
Z Z
= ∗ dx + (v∗0 )2 dx
2 Ω 2v0 + K 2α Ω
Z
+β v∗0 dx, (22.25)
Ω
if v∗0 ∈ A∗ = {v∗0 ∈ Y ∗ : v∗0 + K > 0, in Ω}.

Let J ∗ : Y ∗ ×Y ∗ → R ∪ {−∞} be such that
J ∗ (v∗1 , v∗0 ) = −G∗ (v∗1 , v∗0 ) + F ∗ (v∗1 ).
Define
v̂∗0 = α(u20 − β ),
and
v̂∗1 = (2v̂∗0 + K)u0 − f .
Suppose v̂∗0 ∈ A∗ .
1. if −γ∇2 + 2v̂∗0 > 0 so that δ 2 J(u0 ) > 0, then
δ J1∗ (v̂∗0 ) = 0
and
δ 2 J(u0 )
δ 2 J1∗ (v̂∗0 ) = − < 0,
α(−γ∇2 + 2v̂∗0 )
J(u0 ) = inf J(u)
u∈Br (u0 )

∗ ∗ ∗
= sup inf
∗ ∗
J (v ,
1 0v )
v∗0 ∈Br1 (vˆ∗0 ) v1 ∈Y
= sup J1∗ (v∗0 )

v∗0 ∈Br1 (v̂∗0 )
= J1∗ (v̂∗0 )
= J ∗ (v̂∗1 , v̂∗0 ), (22.26)
where
J1∗ (v∗0 ) = ∗inf ∗ J ∗ (v∗1 , v∗0 ).
v1 ∈Y
2. If −γ∇2 + 2v̂∗0 < 0 and δ 2 J(u0 ) > 0, then
δ J2∗ (v̂∗0 ) = 0
and
δ 2 J(u0 )
δ 2 J2∗ (v̂∗0 ) = − > 0,
α(−γ∇2 + 2v̂∗0 )
J(u0 ) = inf J(u)

u∈Br (u0 )
( )
∗
= inf sup J (v∗1 , v∗0 )
v∗0 ∈Br1 (v̂∗0 ) v∗1 ∈Y ∗
∗ ∗
= inf J2 (v0 )
v∗0 ∈Br1 (v̂∗0 )
= J2∗ (v̂∗0 )
= J ∗ (v̂∗1 , v̂∗0 ), (22.27)
where
J2∗ (v∗0 ) = sup J ∗ (v∗1 , v∗0 ).
v∗1 ∈Y ∗
3. If δ 2 J(u0 ) < 0 so that −γ∇2 + 2v̂∗0 < 0, then
δ J2∗ (v̂∗1 ) = 0
and
δ 2 J(u0 )
δ 2 J2∗ (v̂∗0 ) = − < 0,
α(−γ∇2 + 2v̂∗0 )
J(u0 ) = sup J(u)

u∈Br (u0 )
( )
= sup sup J ∗ (v∗1 , v∗0 )
v∗0 ∈Br1 (v̂∗0 ) v∗1 ∈Y ∗
= sup J2∗ (v∗0 )

v∗0 ∈Br1 (v̂∗0 )
= J2∗ (v̂∗0 )
= J ∗ (v̂∗1 , v̂∗0 ), (22.28)
where
J2∗ (v∗0 ) = sup J ∗ (v∗1 , v∗0 ).
v∗1 ∈Y ∗
−γ∇2 u0 + α (u20 − β )2u0 − f = 0, in Ω

so that
γ∇2 u0 + Ku0 = α(u20 − β )2u0 + Ku0 − f

= (2v̂∗0 + K)u0 − f
= v̂∗1 . (22.29)

v̂∗1 v̂∗1 + f
− = u0 − u0 = 0, in Ω.
K + γ∇2 2v̂∗0 + K
Thus,
∂ J ∗ (v̂∗1 , v̂∗0 )
= 0.
∂ v∗1
Define
J˜(v∗0 ) = J ∗ (ṽ∗1 , v∗0 ),
where ṽ∗1 ∈ Y ∗ is such that
∂ J ∗ (ṽ∗1 , v∗0 )
= 0.
∂ v∗1
Observe that, from
∂ 2 J ∗ (v∗1 , v∗0 ) 1 1
= − ,
∂ (v∗1 )2 K + γ∇2 2v∗0 + K
we have that, if
−γ∇2 + 2v∗0 > 0,
then
J˜(v∗0 ) = J ∗ (ṽ∗1 , v∗0 ) = ∗inf ∗ J ∗ (v∗1 , v∗0 ),
v1 ∈Y
whereas if
−γ∇2 + 2v∗0 < 0,
then
J˜(v∗0 ) = J ∗ (ṽ∗1 , v∗0 ) = sup J ∗ (v∗1 , v∗0 ).
v∗1 ∈Y ∗
From (22.29) and from the definition of v̂∗0 , we obtain
J˜∗ (v̂∗0 ) = J ∗ (v̂∗1 , v̂∗0 )

= −G∗ (v̂∗1 , v̂∗0 ) + F ∗ (v̂∗1 )
= −hu0 , v̂∗1 iL2 − h0, v̂∗0 iL2 + G(u0 , 0)
+hu0 , v̂∗1 iL2 − F(u0 )
= G(u0 , 0) − F(u0 )
= J(u0 ). (22.30)
From the implicit function theorem we have
∂ J˜∗ (v∗0 ) ∂ J ∗ (ṽ∗1 , v∗0 ) ∂ J ∗ (ṽ∗1 , v∗0 ) ∂ ṽ∗1

= +
∂ v∗0 ∂ v∗0 ∂ v∗1 ∂ v∗0
∗ ∗
∂ J (ṽ1 , v0 )∗
= . (22.31)
∂ v∗0
Moreover, from this, joining the pieces, we get
∂ J˜∗ (v̂∗0 ) ∂ J ∗ (v̂∗1 , v̂∗0 ) ∂ J ∗ (v̂∗1 , v̂∗0 ) ∂ v̂∗1

= +
∂ v∗0 ∂ v∗0 ∂ v∗1 ∂ v∗0
∂ J ∗ (v̂∗1 , v̂∗0 )
=
∂ v∗0
(v̂∗1 + f )2 v̂∗0
= − −β
(2v̂∗0 + K)2 α
v̂∗
= u20 − 0 − β
α
= 0, in Ω, (22.32)
since v̂∗0 = α(u20 − β ), in Ω.

Hence, from this and (22.31), we obtain
∂ 2 J˜∗ (v̂∗0 ) ∂ 2 J ∗ (v̂∗1 , v̂∗0 ) ∂ v̂∗1

=
∂ (v∗0 )2 ∂ v∗0 ∂ v∗1 ∂ v∗0
∂ 2 J ∗ (v̂∗1 , v̂∗0 )
+
∂ (v∗0 )2
2(v̂∗1 + f ) ∂ v̂∗1
=
(2v̂∗0 + K)2 ∂ v∗0
(v̂∗1 + f )2 1
−4 − . (22.33)
(2v̂∗0 + K)3 α
At this point we may observe that
v̂∗1 (v̂∗1 + f )
− = 0,
K + γ∇2 (2v̂∗0 + K)
so that taking the variation in v∗0 of this equation in both sides, we obtain
∂ v̂∗1 ∂ v̂∗1
∂ v∗0 ∂ v∗0 2(v̂∗1 + f )
− ∗ + = 0, (22.34)
(K + γ ∇2 ) (2v̂0 + K) (2v̂∗0 + K)2
and thus, recalling that

(v̂∗1 + f )
u0 = ,
(2v̂∗0 + K)
we get
(K+γ∇2 )(2u0 )
∂ v̂∗1 (2v̂∗0 +K)
∗ =− 2 , (22.35)
∂ v0 1 − K2v̂+∗γ∇
+K 0
Replacing (22.35) into (22.33), we obtain

∂ 2 J˜∗ (v̂∗0 ) ∂ 2 J ∗ (v̂∗1 , v̂∗0 ) ∂ v̂∗1
∗ =
∂ (v0 ) 2 ∂ v∗0 ∂ v∗1 ∂ v∗0
∂ 2 J ∗ (v̂∗1 , v̂∗0 )
+
∂ (v∗0 )2
2(v̂1 + f ) ∂ v̂∗1
∗
=
(2v̂∗0 + K)2 ∂ v∗0
4(v̂∗1 + f )2 1
− −
(2v̂∗0 + K)3 α
K+γ∇2
(2v̂∗0 +K)2
(4u02 ) 1 4u2
= − − − ∗ 0 . (22.36)
K+γ∇2 α 2v̂0 + K
1− 2v̂∗0 +K
Therefore,
∂ 2 J˜∗ (v̂∗0 )
∂ (v∗0 )2
1 4u2
= − − ∗ 0
α 2v̂0 + K
K + γ∇2 4u20
−
2v∗0 + K (2v̂∗0 − γ∇2 )
4u20 K + γ ∇2

1
= − − 1+ ∗
α (2v̂0 + K) 2v̂0 − γ∇2
1 4u20
= − − ∗
α (2v̂0 − γ∇2 )
γ∇2 − 2v̂∗0 − 4αu20
=
α(−γ∇2 + 2v̂∗0 )
γ∇2 − 6αu20 + 2αβ
=
α(−γ∇2 + 2v̂∗0 )
δ 2 J(u0 ) (22.37)
= − .
α (−γ∇2 + 2v̂∗0 )
Assume now δ 2 J(u0 ) > 0 and −γ∇2 + 2v̂∗0 > 0.

From (22.37), we obtain
∂ 2 J˜∗ (v̂∗0 ) δ 2 J(u0 )

δ 2 J1∗ (v̂∗0 ) = = − < 0.
∂ (v∗0 )2 α(−γ∇2 + 2v̂∗0 )
Summarizing,
δ 2 J1∗ (v̂∗0 ) < 0,
so that u0 ∈ U is a point of local minimum for J and v̂∗0 is a point of local maximum for J1∗ .
Hence, there exist r > 0 and r1 > 0 such that
J(u0 ) = min J(u)
u∈Br (u0 )
= max J1∗ (v∗0 )
v∗0 ∈Br1 (v̂∗0 )
= J1∗ (v̂∗0 )
= J ∗ (v̂∗1 , v̂∗0 ). (22.38)
The remaining items may be proven similarly from (22.37).

22.4 Another duality principle for global optimization

Our next result is another duality principle suitable for global optimization.
∂ Ω. Consider the Ginzburg-Landau energy given by J : U → R, where
Z
γ
J(u) = ∇u · ∇u dx
2 Ω
Z
α
+ (u2 − 1)2 dx − hu, f iL2 , (22.39)
2 Ω
where α, γ > 0, f ∈ L2 (Ω) and
U = W01,2 (Ω) = {u ∈ W 1,2 (Ω) : u = 0, on ∂ Ω}.
We also denote, for a finite dimensional discretized version of this problem, in a finite elements or finite
differences context,
J(u) = −F(u) + G1 (u, 0),

where
K
Z Z
γ
F(u) = − ∇u · ∇u dx + u2 dx,
2 Ω 2 Ω
and
K
Z Z
α
G1 (u, v) = (u2 − 1 + v)2 dx + u2 dx − hu, f iL2 ,
2 Ω 2 Ω
where K > 0 is such that F(u) > 0, ∀u ∈ U, such that u =
6 0.
And where generically, Z
hh, giL2 = hg dx, ∀h, g ∈ L2 (Ω).
Ω
We define,
F ∗ (z∗ ) = sup{hz∗ , uiL2 − F(u)}

u∈U
Z
γ
= sup{hz∗ , uiL2 + ∇u · ∇u dx
u∈U 2 Ω
K
Z
− u2 dx}
2 Ω
1
Z
= z∗ ((KId + γ∇2 )−1 z∗ ) dx, (22.40)
2 Ω
where Id denotes the identity matrix.

Also,
G∗1 (z∗ , v∗1 ) = sup {hz∗ , uiL2 + hv∗1 , viL2
(u,v)∈U ×L2
K
Z Z
α
+hu, f iL2 − (u2 − 1 + v)2 dx − u2 dx}
2 Ω 2 Ω
1 (z∗ + f )2 1
Z Z
= dx + (v∗ )2 d x
2 Ω 2v∗1 + K 2α Ω 1
Z
+ v∗1 dx
Ω
≡ G∗1L (z∗ , v∗1 ), (22.41)
if v∗1 ∈ B1 , where
B1 = {v∗1 ∈ Y ∗ : 2v∗1 + K > 0, in Ω}
and G∗1L stands for the Legendre transform of G1 .
We also denote,
B2 = {v∗1 ∈ Y ∗ :
Z Z
γ
∇u · ∇u dx + v∗1 u2 dx > 0,
2 Ω Ω
∀u ∈ U such that u =
6 0}, (22.42)
C∗ = B1 ∩ B2 ,
where
Y = Y ∗ = L2 (Ω).
inf J(u) ≥ sup { inf {J ∗ (z∗ , v∗1 )}}
∗ ∗
u∈U v∗1 ∈C∗ z ∈Y
= sup J˜(v∗1 ), (22.43)

v∗1 ∈C∗
where,
J ∗ (z∗ , v∗1 ) = F ∗ (z∗ ) − G∗1 (z∗ , v∗1 )
and
J˜(v∗1 ) = ∗inf ∗ J ∗ (z∗ , v∗1 ).
z ∈Y
Moreover, if there exists a critical point (z∗0 , (v∗0 )1 ) ∈ C∗ ×Y ∗ , so that
∗
δJ (z∗0 , (v∗0 )1 ) = 0,
then, denoting
z∗0 + f
u0 =
2(v∗0 )1 + K
we have that
J(u0 ) = min J(u)
u∈U
= max J˜∗ (v∗1 )
v∗1 ∈C∗
= J˜∗ ((v∗0 )1 )
= J ∗ (z∗0 , (v∗0 )1 ). (22.44)
G∗1 (z∗ , v∗1 )

≥ hz∗ , uiL2 + hv∗1 , viL2 − G1 (u, v), (22.45)
∀v∗1 ∈ C∗ , u ∈ U, v ∈ Y, z∗ ∈ Y ∗ .
Thus,
−hz∗ , uiL2 + G1 (u, 0)

≥ −G∗1 (z∗ , v∗1 ), (22.46)
∀u ∈ U, v∗1 ∈ C∗ , z∗ ∈ Y ∗ , so that
F ∗ (z∗ ) − hz∗ , uiL2 + G1 (u, 0)

≥ F ∗ (z∗ ) − G∗1 (z∗ , v∗1 ), (22.47)
∀u ∈ U, v∗1 ∈ C∗ , z∗ ∈ Y ∗ .
Hence,
J(u) = −F(u) + G1 (u, 0)

= ∗inf ∗ {F ∗ (z∗ ) − hz∗ , uiL2 } + G1 (u, 0)
z ∈Y
≥ inf {F ∗ (z∗ ) − G∗1 (z∗ , v∗1 )}
z∗ ∈Y ∗
= inf J ∗ (z∗ , v∗1 )
z∗ ∈Y ∗
= J˜∗ (v∗1 ), (22.48)
∀u ∈ U, v∗1 ∈ C∗ .
Thus,
inf J(u) ≥ sup J˜∗ (v∗1 ). (22.49)
u∈U v∗1 ∈C∗
Now suppose (z0∗ , (v∗0 )1 ) ∈ Y ∗ ×C∗ is such that
δ J ∗ (z∗0 , (v∗0 )1 ) = 0.
From the variation in z∗ we obtain,

z∗0 + f
(KId + γ∇2 )−1 (z∗0 ) = = u0 ,
2(v∗0 )1 + K
so that
z∗0 = (KId + γ∇2 )u0 ,
and
z∗0 + f = (2(v∗0 )1 + K)u0 . (22.50)
Thus,
F ∗ (z∗0 ) = hz∗0 , u0 iL2 − F(u0 ). (22.51)
On the other hand, from the variation in v∗1 , we have,
[z∗0 + f ]2 (v∗ )1
∗ 2
− 0 − 1 = 0,
[2(v0 )1 + K ] α
so that
(v∗0 )1 = α(u20 − 1),
and hence, from this and (22.50) we have,
z∗0 + f = α (u20 − 1)2u0 + Ku0 ,
and
G∗1 (z∗0 , (v∗0 )1 ) = hz∗0 , u0 iL2 − G1 (u0 , 0). (22.52)
J(u0 ) = −F(u0 ) + G1 (u0 , 0)

= F ∗ (z∗0 ) − G∗1 (z∗0 , (v∗0 )1 )
= J ∗ (z∗0 , (v∗0 )1 ). (22.53)
Now, let
v∗1 ∈ C∗ .
Observe that, in such a case,
Z Z
γ
∇u · ∇u dx + v∗1 u2 dx > 0,
2 Ω Ω
∀u ∈ U, such that u 6= 0.
Denoting
Z Z
γ ∗ 2
α1 = inf ∇u · ∇u dx + v1 u dx − hu, f iL2 , (22.54)
u∈U 2 Ω Ω
we have
K
Z Z
v∗1 u2 dx + u2 dx − hu, f iL2 − hz∗ , uiL2
Ω 2 Ω
K
Z Z
γ
≥ − ∇u · ∇u dx + u2 dx − hz∗ , uiL2 + α1 , (22.55)
2 Ω 2 Ω
∀u ∈ U, so that,
Z
K
Z
inf v∗1 u2 dx +
u2 dx − hu, f iL2 − hz∗ , uiL2
u∈U Ω Ω 2

K
Z Z
γ
≥ inf − ∇u · ∇u dx + u2 dx − hz∗ , uiL2 + α1 , (22.56)
u∈U 2 Ω 2 Ω
and hence,
1 (z∗ + f )2
Z
− ∗ ≥ −F ∗ (z∗ ) + α1 ,
2 Ω 2v1 + K
so that, for v∗1 ∈ C∗ fixed, we have,
1 (z∗ + f )2
Z
F ∗ (z∗ ) − ∗ dx ≥ α1 ,
2 Ω 2v1 + K
∀z∗ ∈ Y ∗ .
And indeed, from the general result in [78], we have
(z∗ + f )2

1
Z
∗ ∗
inf F (z ) − d x = α1 ∈ R.
z∗ ∈Y ∗ 2 Ω 2v∗1 + K
From this, since (v∗0 )1 ∈ C∗ and the optimization in z∗ in question is quadratic, we may infer that,
J˜∗ ((v∗0 )1 ) = ∗inf ∗ J ∗ (z∗ , (v∗0 )1 ) = J ∗ (z∗0 , (v∗0 )1 ).

z ∈Y
From this, (22.49) and (22.53), we finally obtain,
J(u0 ) = min J(u)

u∈U
= max J˜∗ (v∗1 )
v∗1 ∈C∗
= J˜∗ ((v∗0 )1 )
= J ∗ (z∗0 , (v∗0 )1 ). (22.57)
22.5 The existence of a global solution for the Ginzburg-Landau system

in the presence of a magnetic field
In this section we develop a proof of existence of solution for the Ginzburg-Landau system in the presence
of a magnetic field and concerning potential. We emphasize again that similar models, which are closely
relating to those of last sections, are addressed in [12, 55].
We highlight this existence result and the next duality principle have been presented in similar form in
my book, “A Classical Description of Variational Quantum Mechanics and Related Models”, [22]. For the
sake of completeness, we present both the results in details.
Finally, as a previous related existence result we would cite [46].
Theorem 22.5.1 Consider the functional J : U → R where

Z
γ
J (φ , A) = |∇φ − im ρAφ |22 dx
2 Ω
Z Z
α β
+ |φ |4 dx − |φ |2 dx
4 Ω 2 Ω
1
Z
+ | curl(A) − B0 |22 dx, (22.58)
8π Ω1
where Ω, Ω1 are open bounded, simply connected sets such that
Ω ⊂ Ω1 .
We assume the boundaries ∂ Ω and ∂ Ω1 to be regular (Lipschitzian). Here, again im denotes the imaginary
unit and γ, α, β and ρ are positive constants. Also,
U = W 1,2 (Ω; C) × L2 (Ω1 ; R3 ).
Suppose there exists a minimizing sequence (φn , An ) ⊂ U for J such that
kφn k∞ ≤ K, ∀n ∈ N
for some K > 0.

Under such hypotheses, there exists (φ0 , A0 ) ∈ U such that
J(φ0 , A0 ) = min {J(φ , A)}.

(φ ,A)∈U
Proof 22.4
Define
α1 = inf {J(φ , A)} ∈ R.
(φ ,A)∈U

lim J(φn , An ) = α1 .
n→∞
From the expression of J, there exists K1 > 0 such that
kcurl(An )k22 ≤ K1 , ∀n ∈ N.
Given (φ , A) ∈ U, define (φ 0 , A0 ) ∈ U by
φ 0 = φ eim ρϕ ,
and
A0 = A + ∇ϕ,
where ϕ will be specified in the next lines.
Observe that,
|∇φ 0 − im ρA0 φ 0 |2 = |∇(φ eim ρϕ ) − im ρ(A + ∇ϕ)φ eim ρϕ |2

= |∇φ eim ρϕ + φ im ρeim ρϕ ∇ϕ − im ρAφ eim ρϕ − im ρφ ∇ϕeim ρϕ |2
= |(∇φ − im ρAφ )eim ρϕ |2
= |∇φ − im ρAφ )|2 . (22.59)
Moreover,
curl(A0 ) = curl(A) + curl(∇ϕ) = curl(A).
Also,
|φ 0 | = |φ eim ρϕ | = |φ |.
From these last calculations, we may infer the system gauge invariance, that is,
J(φ , A) = J(φ 0 , A0 ).
In particular, we shall choose ϕ ∈ W 1,2 (Ω1 ) such that
div(A0 ) = div(A) + ∇2 ϕ = 0,
and, denoting by n the outward normal to ∂ Ω1 ,
A0 · n = A · n + ∇ϕ · n = 0, on ∂ Ω1
that is,
∇2 ϕ = −div(A), in Ω1 ,
∇ϕ · n = −A · n, on ∂ Ω1 .
Observe that at first we would have,
inf J(φ , A) ≤ inf J(φ 0 , A0 ).

(φ ,A)∈U (φ 0 ,A0 )∈U
However,
J(φn0 , A0n ) = J(φn , An ) → α1 , as n → ∞,
so that
inf J(φ , A) = inf J(φ 0 , A0 ).
(φ ,A)∈U (φ 0 ,A0 )∈U
From Friedrichs’ inequality, we have,
K12 ≥ kcurl(An )k22

= k curl(A0n )k22 + kdiv(A0n )k22 ≥ K2 kA0n k22 , ∀n ∈ N, (22.60)
for some K2 > 0.

Hence,
kA0n k2 ≤ K3 , ∀n ∈ N
for some K3 > 0.
We recall that,
kφn0 k∞ = kφn k∞ ≤ K, ∀n ∈ N.
Hence,
Z
γ
J(φn0 , A0n ) = |∇φn0 − im ρA0n φn0 |22 dx
2 Ω
Z Z
α β
+ |φ 0 |4 dx − |φ 0 |2 d x
4 Ω n 2 Ω n
1
Z
+ |curl(A0n ) − B0 |22 dx
8π Ω1
Z
γ
≥ |∇φn0 |2 dx − γ|ρ|kφn0 k∞ kA0n k2 k∇φn0 k2
2 Ω
γ
+ |ρ|2 kA0n φn0 k22
2
Z Z
α β
+ |φn0 |4 dx − |φ 0 |2 dx
4 Ω 2 Ω n
1
Z
+ |curl(A0n ) − B0 |22 dx
8π Ω1
γ
≥ k∇φn0 k22 dx − γKK3 |ρ|k∇φn0 k2
2
γ
+ |ρ|2 kA0n φn0 k22
2
Z Z
α β
+ |φn0 |4 dx − |φ 0 |2 dx
4 Ω 2 Ω n
1
Z
+ | curl(A0n ) − B0 |22 dx. (22.61)
8π Ω1
Suppose, to obtain contradiction, there exists a subsequence {nk } such that
k∇φn0 k k2 → +∞, as k → ∞.
J(φn0 k , A0nk ) → +∞, as k → +∞,
which contradicts
J(φn0 , A0n ) → α1 , as n → +∞.
Therefore, there exists K4 > 0 such that
k∇φn0 k2 ≤ K4 ∈ R+ , ∀n ∈ N.
Hence, from the Rellich- Krondrachov theorem, there exists φ0 ∈ W 1,2 (Ω; C) such that, up to a not
relabeled subsequence,
∇φn0 * ∇φ0 , weakly in L2 ,
and
φn0 → φ0 , strongly in L2 .
Also, since
kcurl(A0n )k2 ≤ K1 , ∀n ∈ N,
there exists v0 ∈ L2 (Ω1 ; R3 ) such that
curl(A0n ) * v0 , weakly in L2 (Ω1 ; R3 ).
Also, since
kA0n k2 ≤ K4 , ∀n ∈ N,
there exists
A0 ∈ L2 (Ω1 ; R3 ),
such that, up to a not relabeled subsequence,
A0n * A0 , weakly in L2 (Ω1 ; R3 ).
Now fix
φ̂ ∈ Cc∞ (Ω1 ; R3 ).
Thus, we have
hA0 , curl∗ (φ̂ )iL2 = lim hA0n , curl∗ (φ̂ )iL2

n→∞
= lim hcurl(A0n ), φ̂ iL2
n→∞
= hv0 , φ̂ iL2 . (22.62)
Since φ̂ ∈ Cc∞ (Ω1 ; R3 ) is arbitrary, we may infer that
v0 = curl(A0 ),
in distributional sense.
At this point we shall prove that, up to a not relabeled subsequence, we have,
A0n φn0 * A0 φ0 , weakly in L2 (Ω; C3 ).
Fix v ∈ L2 (Ω; C3 ). Therefore, up to a not relabeled subsequence, we have that
|φn0 v − φ0 v|22 → 0, a.e. in Ω.

Observe that
kφn0 k∞ < K, ∀n ∈ N,
so that
|φn0 v − φ0 v|22 ≤ 2K 2 |v|22 ∈ L1 (Ω; R).
Thus, from the Lebesgue dominated convergence theorem, we obtain
kφn0 v − φ0 vk22 → 0, as n → ∞.
Hence, since
A0n * A0 , weakly in L2 (Ω1 ; R3 ),
we have,
Z
(A0n · φn0 v − A0 · φ0 v) dx
Ω
Z
= (A0n · φn0 v − A0n · φ0 v + A0n · φ0 v − A0 · φ0 v) dx
Ω
Z
≤ kA0n k2 kφn0 v − φ0 vk2 + A0n · φ0 v − A0 · φ0 v) dx
Ω
→ 0, as n → ∞. (22.63)
Since v ∈ L2 (Ω; C3 ) is arbitrary, we may infer that
A0n φn0 * A0 φ0 , weakly in L2 (Ω; C3 ).
From this we obtain
∇φn0 − im ρA0n φn0 * ∇φ0 − im ρA0 φ0 , weakly in L2 (Ω; C3 ),
so that
Z Z
lim inf |∇φn0 − im ρA0n φn0 |22 dx ≥ |∇φ0 − iρA0 φ0 |22 dx, (22.64)
n→∞ Ω Ω
Also, from
φn0 * φ0 , weakly in W 1,2 (Ω; C)
and
curl(A0n ) * curl(A0 ), weakly in L2 (Ω1 , R3 ),
from the convexity of the functional involved, we obtain,
Z
1
Z
0 2 α 0 4
lim inf |curl(An ) − B0 |2 dx + |φ | dx
n→∞ 8π Ω1 4 Ω n
1
Z Z
α
≥ |curl(A0 ) − B0 |22 dx + |φ0 |4 dx, (22.65)
8π Ω1 4 Ω
so that, from these last results and from
φn0 → φ0 , strongly in L2 (Ω; C),
we get,
inf J(φ , A) = α1
(φ ,A)∈U
= lim inf J(φn0 , A0n )

n→∞
≥ J(φ0 , A0 ). (22.66)
22.6 Duality for the complex Ginzburg-Landau system

In this subsection we present a duality principle and relating sufficient optimality criterion for the full com-
plex Ginzburg-Landau system.
The basic results on convex analysis here developed may be found in [64, 78]. Our results are summarized
by the following theorem.
Theorem 22.6.1 Let Ω, Ω1 ⊂ R3 be open, bounded, connected sets with regular (Lipischtzian) boundaries
denoted by ∂ Ω and ∂ Ω1 respectively, where Ω ⊂ Ω1 and Ω corresponds to a super conducting sample.
Consider the Ginzburg-Landau energy given by J : V1 ×V2 → R where,
Z
γ
J(φ , A) = |∇φ − im ρAφ |22 dx
2 Ω
Z
α
+ (|φ |2 − β )2 dx − hφ , f iL2
2 Ω
1
Z
+ |curl(A) − B0 |22 dx, (22.67)
8π Ω1
and where α, γ, ρ > 0, f ∈ L2 (Ω; C).

1
In particular, from the Ginzburg-Landau theory for the dimensionless case we have, γ = 1, α = 2(1+t 2 )2
and β = 1 − t 4 , where t = T /Tc , Tc is the critical temperature and T is the super-conducting sample actual
one. A typical value for t is t = 0.95. Finally, the value 1/(8π) may also vary according to type of material
or type of superconductor.
Moreover,
V1 = W 1,2 (Ω; C),
V2 = W 1,2 (Ω1 ; R3 ).
Here, we generically denote
Z Z
hg, hiL2 = Re[g]Re[h] dx − Im[g]Im[h] dx,
Ω Ω
∀h, g ∈ L2 (Ω; C), where Re[a], Im[a] denote the real and imaginary parts of a, ∀a ∈ C, respectively.
We also denote,
J(φ , A) = G0 (φ , ∇φ , A) + G1 (φ , 0) + G2 (A),
Z
γ
G0 (φ , ∇φ , A) = |∇φ − im ρAφ |22 dx,
2 Ω
Z
α
G1 (φ , v3 ) = (|φ |2 − β + v3 )2 dx − hφ , f iL2 ,
2 Ω
and,
1
Z
G2 (A) = | curl(A) − B0 |22 dx.
8π Ω1
Moreover, we define
G∗0 (v∗1 ) = sup {hv∗1 , v1 − im ρAφ iL2 − G0 (φ , v1 , A)}

(φ ,v1 )∈V1 ×Y
1
Z
= |v∗1 |22 dx, (22.68)
2γ Ω
G∗1 (v∗1 , v∗3 , A) = sup {hv∗1 , ∇φ − im ρAφ iL2 + hv∗3 , v3 iL2 − G1 (φ , v3 )}

(φ ,v3 )∈V1 ×Y1
1 |div(v∗1 ) + im ρA · v∗1 − f |2 1
Z Z
= dx + (v∗3 )2 dx
2 Ω 2v3∗ 2α Ω
Z
+ β (v∗3 ) dx, (22.69)
Ω
if v∗ ∈ B1 , where
B1 = {v∗ ∈ Y ∗ ×Y1∗ : v∗3 > 0 in Ω}.
We also denote
B2 = {v∗ ∈ Y ∗ :
1 1 |ρv∗1 · A|2
Z Z
|curlA|22 dx − dx > 0,
8π Ω1 2 Ω 2 v∗3
∀A ∈ D∗ , such that A 6= 0}, (22.70)
D∗ = {A ∈ V2 : div(A) = 0, in Ω1 , and A · n = 0 on ∂ Ω1 },
C∗ = B1 ∩ B2 ,
and
Y = Y ∗ = L2 (Ω; C3 ) and Y1 = Y1∗ = L2 (Ω).
Under such assumptions, we have,
inf J(φ , A) ≥ sup { inf {J ∗ (v∗ , A) + G2 (A)}}

∗
(φ ,A)∈V1 ×D∗ v∗ ∈C∗ A∈D
˜∗ ∗
= sup J (v ), (22.71)
v∗ ∈C∗
where
J ∗ (v∗ , A) = −G∗0 (v∗1 ) − G∗1 (v∗1 , v∗3 , A).
and
J˜∗ (v∗ ) = inf ∗ {J ∗ (v∗ , A) + G2 (A)}.
A∈D
Moreover, assume there exists a critical point (v∗0 , A0 ) ∈ C∗ × D∗ such that
δ {J ∗ (v∗0 , A0 ) + G2 (A0 )} = 0.
Under such hypotheses, defining

div((v∗0 )1 ) + im ρA0 · (v∗0 )1 − f
φ0 = ,
2(v∗0 )3
we have
J(φ0 , A0 ) = min J(φ , A)

(φ ,A)∈V1 ×D∗
= sup { inf ∗ {J ∗ (v∗ , A) + G2 (A)}}
v∗ ∈C∗ A∈D
= max J˜∗ (v∗ )
v∗ ∈C∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , A0 ) + G2 (A0 ). (22.72)
−J ∗ (v∗ , A) − G2 (A) = G∗0 (v∗1 ) + G∗1 (v∗1 , v∗3 , A) − G2 (A)

≥ hv∗1 , ∇φ − im ρAφ iL2 − G0 (φ , ∇φ , A)
−hv∗1 , ∇φ − im ρAφ iL2 + hv∗3 , 0iL2
−G1 (φ , 0) − G2 (A), (22.73)
∀φ ∈ V1 , A ∈ D∗ , that is,
G0 (φ , ∇φ , A) + G1 (φ , 0) + G2 (A)
≥ −G∗0 (v∗1 ) − G∗1 (v∗1 , v∗3 , A) + G2 (A), (22.74)
so that
J(φ , A) ≥ inf {−G∗0 (v∗1 ) − G∗1 (v∗1 , v∗3 , A) + G2 (A)}

A∈D∗
= inf {J ∗ (v∗ , A) + G2 (A)}
A∈D∗
= J (v ), ∀(φ , A) ∈ V1 × D∗ , v∗ ∈ C∗ .
˜ ∗ (22.75)
Thus,
inf J(φ , A) ≥ sup J˜∗ (v∗ ). (22.76)
(φ ,A)∈V1 ×D∗ v∗ ∈C∗
Now, suppose (v∗0 , A0 ) ∈ C∗ × D∗ is such that
δ {J ∗ (v∗0 , A0 ) + G2 (A0 )} = 0. (22.77)
From the variation in v∗1 we obtain,
div((v∗0 )1 ) + im ρA0 · (v∗0 )1 − f

(v∗0 )1 = γ(∇ − iρA0 )
2(v∗0 )3
= γ(∇ − im ρA0 )φ0 , (22.78)
so that,
G∗0 ((v∗0 )1 ) = h(v∗0 )1 , ∇φ0 − im ρA0 φ0 iL2 − G0 (φ0 , ∇φ0 , A0 ). (22.79)
From the variation in v∗3 we obtain
(div((v∗0 )1 ) + im ρA0 · (v∗0 )1 − f )2 (v∗0 )3

− − β = 0,
(2(v∗0 )3 )2 α
that is,
(v∗0 )3 = α(|φ0 |2 − β ),
so that
G∗1 ((v∗0 )1 , (v∗0 )3 , A0 ) = −h(v∗0 )1 , ∇φ0 − im ρA0 φ0 iL2 − G1 (φ0 , 0). (22.80)
J ∗ (v∗0 , A0 ) + G2 (A0 )
= −G∗0 ((v∗0 )1 ) − G∗1 ((v∗0 )1 , (v∗0 )3 , A0 ) + G2 (A0 )
= G0 (φ0 , ∇φ0 , A0 ) + G1 (φ0 , 0) + G2 (A0 )
= J(φ0 , A0 ) (22.81)
From v∗0 ∈ C∗ and (22.77) we have,
J ∗ (v∗0 , A0 ) + G2 (A0 ) = J˜∗ (v∗0 ).
From this, (22.76) and (22.81) we obtain,
J(φ0 , A0 ) = min J(φ , A)

(φ ,A)∈V1 ×D∗
= sup { inf ∗ {J ∗ (v∗ , A) + G2 (A)}}
v∗ ∈C∗ A∈D
= max J˜∗ (v∗ )
v∗ ∈C∗
= J˜∗ (v∗0 )
= J ∗ (v∗0 , A0 ) + G2 (A0 ). (22.82)
22.7 Conclusion
In the present chapter, we have developed duality principles applicable to a large class of variational non-
convex models.
In a second step we have applied such results to a Ginzburg-Landau type equation. We emphasize the
main theorems here developed are a kind of generalization of the main results found in Toland [78], published
in 1979 and [11, 10].
Following the approach presented in [11], we also highlight the duality principles obtained may be ap-
plied to non-linear and non-convex models of plates, shells and elasticity.
Finally, as above mentioned, in the last sections we present a global existence result, a duality principle
and respective optimality conditions for the complex Ginzburg-Landau system in superconductivity in the
presence of a magnetic field and concerned magnetic potential.
Chapter 23
Existence of Solution for an

Optimal Control Problem
Associated to the
Ginzburg-Landau System in
Superconductivity
Fabio Silva Botelho and Eduardo Pandini Barros
23.1 Introduction
This work develops an existence result for an optimal control problem closely related to the Ginzburg-Landau
system in superconductivity. First, we recall that about the year 1950 Ginzburg and Landau introduced a
theory to model the super-conducting behavior of some types of materials below a critical temperature Tc ,
which depends on the material in question. They postulated the free density energy may be written close to
Tc as
h̄ α(T ) β (T )
Z Z Z
Fs (T ) = Fn (T ) + |∇ψ |22 d x + |ψ |4 dx − |ψ |2 dx,
4m Ω 4 Ω 2 Ω
where ψ is a complex parameter, Fn (T ) and Fs (T ) are the normal and super-conducting free energy densities,
respectively (see [4] for details). Here Ω ⊂ R3 denotes the super-conducting sample with a boundary denoted
by ∂ Ω = Γ. The complex function ψ ∈ W 1,2 (Ω; C) is intended to minimize Fs (T ) for a fixed temperature T .
Denoting α(T ) and β (T ) simply by α and β , the corresponding Euler-Lagrange equations are given by:
h̄
 − 2m ∇2 ψ + α|ψ|2 ψ − β ψ = 0, in Ω

(23.1)
 ∂ψ
∂ n = 0, on ∂ Ω.
This last system of equations is well known as the Ginzburg-Landau (G-L) one. In the physics literature is
also well known the G-L energy in which a magnetic potential here denoted by A is included. The functional
in question is given by:
2
1 h̄2 2ie
Z Z
J (ψ, A) = | curl A − B0 |22 dx + ∇ψ − Aψ dx
8π R3 4m Ω hc
¯ 2
Z Z
α β
+ |ψ |4 dx − |ψ|2 dx (23.2)
4 Ω 2 Ω
Considering its minimization on the space U, where
U = W 1,2 (Ω; C) ×W 1,2 (R3 ; R3 ),
through the physics notation the corresponding Euler-Lagrange equations are:


1
2
 2m −ih̄∇ − 2e 2
c A ψ + α|ψ| ψ − β ψ = 0, in Ω
(23.3)
2e

· n = 0,

ih̄∇ψ + c Aψ on ∂ Ω,
and
 curl (curl A) = curl B0 + 4π

˜ in Ω
c J,
(23.4)
curl (curl A) = curl B0 , in R3 \ Ω,

where
ieh̄ ∗ 2e 2
J˜ = − (ψ ∇ψ − ψ∇ψ ∗ ) − |ψ|2 A.
2m mc
and
B0 ∈ L2 (R3 ; R3 )
is a known applied magnetic field.
Existence of a global solution for a similar problem has been proved in [22].
23.2 An existence result for a related optimal control problem

Let Ω ⊂ R3 , Ω1 ⊂ R3 be open, bounded and connected sets with Lipschitzian boundaries, where Ω ⊂ Ω1
and Ω1 is convex. Let φd : Ω → C be a known function in L4 (Ω; C) and consider the problem of minimizing
k|φ |2 − |φd |2 k20,2,Ω
with (φ , A, u) subject to the satisfaction of the Ginzburg-Landau equations, indicated in (23.5) and (23.6) in
the next lines.
For such a problem, the control variable is u ∈ L2 (∂ Ω; C) and the state variables are the Ginzburg-Landau
order parameter φ ∈ W 1,2 (Ω, C) and the magnetic potential A ∈ W 1,2 (Ω1 , R3 ).
Our main existence result is summarized by the following theorem.
Theorem 23.2.1 Consider the functional

ε
J(φ , A, u) = k∇φ k20,2,Ω + K1 k|φ |2 − |φd |2 k20,2,Ω + K2 kuk20,2,∂ Ω ,
2
subject to (φ , A, u) ∈ C , where
C = {(φ , A, u) ∈ W 1,2 (Ω, C) ×W 1,2 (Ω1 , R3 ) × L2 (∂ Ω; C) : such that (23.5) and (23.6) hold },
Existence for an Optimal Control Problem in Superconductivity 467
where  2
1
 2m −ih̄∇ − 2e 2
c A φ + α|φ | φ − β φ = 0, in Ω,
(23.5)
2e

· n = u,

ih̄∇φ + c Aφ on ∂ Ω,
and 

 curl curl A = curl B0 + 4π ˜
c J, in Ω,




 curl curl A = curl B0 , in Ω1 \ Ω,


(23.6)
 div A = 0, in Ω1 ,







A · n = 0, on ∂ Ω1

where,
ieh̄ 2e 2
J˜ = − (φ ∗ ∇φ − φ ∇φ ∗ ) − |φ |2 A,
2m mc
and where ε > 0 is a small parameter, K1 > 0 and K2 > 0.
Under such hypotheses, there exists (φ0 , A0 , u0 ) ∈ C such that
J(φ0 , A0 , u0 ) = min J(φ , A, u).
(φ ,A,u)∈C
Proof 23.1 Let {(φn , An , un )} be a minimizing sequence (such a sequence exists from the existence result
for u = 0 in [22], and from the fact that J is lower bounded by 0).
Thus, such a sequence is such that
J(φn , An , un ) → η = inf J(φ , A, u).
(φ ,A,u)∈C
From the expression of J, there exists K > 0 such that

k∇φn k0,2,Ω ≤ K,
kφn k0,4,Ω ≤ K,
kφn k0,2,Ω ≤ K,
and
kun k0,2,∂ Ω ≤ K, ∀n ∈ N
so that, from the Rellich-Kondrashov Theorem, there exists a not relabeled subsequence, φ0 ∈ W 1,2 (Ω, C)
and u0 ∈ L2 (Ω, C) such that
φn * φ0 , weakly in W 1,2 (Ω, C),
φn → φ0 , in norm in L2 (Ω, C) and L4 (Ω, C),
un * u0 , weakly in L2 (∂ Ω, C), as n → ∞.
On the other hand, we have from (23.6), from the generalized Holder¨ inequality and for constants γ =
4π −ieh̄ 4π 2e2
c 2m > 0 and γ 1 = c mc > 0 that
0 = ρ1,n ≡ h curl An , curl An iL2 (Ω1 ;R3 )
−h curl An , B0 iL2 (Ω1 ;R3 )
2e2 2

4π ieh̄ ∗ ∗
+ (φ ∇φ − φ ∇φ ) + |φ | An , An
c 2m mc L2 (Ω,R3 )
≥ h curl An , curl An iL2 (Ω1 ,R3 )
−k curl An k0,2,Ω1 kB0 k0,2,Ω1 − γkAn k0,4,Ω1 kφn k0,4,Ω k∇φn k0,2,Ω
+γ1 |φ |2 , An · An L2 (Ω,R3 )
. (23.7)
From the well known Friedrichs Inequality and the Sobolev Spaces Imbedding theorem for appropriate
constants indicated, we obtain
2
kAn k20,4,Ω1 ≤ K3 kAn k21,2,Ω1 ≤ K4 k div An k0,2,Ω1 + k curl An k0,2,Ω
= K4 k curl An k20,2,Ω1 , (23.8)
since from the London Gauge assumption,
div An = 0, in Ω1 , ∀n ∈ N.
Summarizing, we have obtained, for some appropriate K5 > 0,
2 1
0 = ρ1,n ≥ K5 kAn k0,4,Ω 1
+ k curl An k20,2,Ω1
2
−k curl An k0,2,Ω1 kB0 k0,2,Ω1 − γkAn k0,4,Ω1 K 2
+γ1 |φ |2 , An · An L2 (Ω;R3 )
≡ ρ2,n . (23.9)
Now, suppose to obtain contradiction there exists a subsequence {nk } ⊂ N such that
kAnk k0,4,Ω1 → ∞, as k → ∞.

ρ2,nk → ∞, as k → ∞,
which contradicts
ρ2,n ≤ 0, ∀n ∈ N.
Hence, there exists K6 > 0 such that
kAn k0,4,Ω1 < K6 ,
and
kAn k0,2,Ω1 < K6 , ∀n ∈ N.
From this and (23.7) we have,
0 = ρ1,n ≥ k curl An k20,2,Ω1

−k curl An k0,2,Ω1 kB0 k0,2,Ω1 − γK6 K 2
+γ1 |φ |2 , An · An L2 (Ω,R3 )
≡ ρ3,n . (23.10)
Suppose to obtain contradiction there exists a subsequence {nk } ⊂ N such that
k curl Ank k0,2,Ω1 → ∞, as k → ∞.

ρ3,nk → ∞, as k → ∞,
which contradicts
ρ3,n ≤ 0, ∀n ∈ N.
Hence, there exists K7 > 0 such that
k curl An k0,2,Ω1 < K7 , ∀n ∈ N

so that from this, the Friedrichs inequality and the London Gauge hypothesis, we obtain K8 > 0 such that
kAn k1,2,Ω1 < K8 .
So from such a result and the Rellich-Kondrashov Theorem there exists a not relabeled subsequence and
A0 ∈ W 1,2 (Ω1 , R3 ) such that
An * A0 weakly ∈ W 1,2 (Ω1 , R3 )
An → A0 in norm in L2 (Ω1 , R3 ) and L4 (Ω1 , R3 ).
Moreover, from the Sobolev Imbedding Theorem, there exist real constants K̂ > 0, K̂1 > 0 such that
kφn k0,6,Ω ≤ K̂kφn k1,2,Ω < K̂1 , ∀n ∈ N.
Thus, from this and the first equation in (23.5), there exist real constants K̂2 > 0, . . . , K̂6 > 0, such that
k∇2 φn k0,2,Ω < K̂2 kAn k0,2,Ω k∇φn k0,2,Ω

K̂3 kAn k0,4,Ω kφn k0,2,Ω + K̂4 kφn k30,6,Ω + Kˆ 5 kφn k0,2,Ω
≤ K̂6 , ∀n ∈ N. (23.11)
From this, up to a subsequence, we get
∇2 φn * ∇2 φ0 weakly in L2 (Ω; C).
Let
ϕ ∈ Cc∞ (Ω, C), ϕ1 ∈ Cc∞ (Ω, R3 ) and ϕ2 ∈ Cc∞ (Ω1 \ Ω, R3 ).
From the last results, we may easily obtain the following limits
1.
h∇2 φn , ϕiL2 → h∇2 φ0 , ϕiL2 ,
2.
h∇φn , ∇ϕiL2 → h∇φ0 , ∇ϕiL2 ,
3.
hAn · ∇φn , ϕiL2 → hA0 · ∇φ0 , ϕiL2 ,
4.
h|An |2 φn , ϕiL2 → h|A0 |2 φ0 , ϕiL2 ,
5.
h|φn |2 φn , ϕiL2 → h|φ0 |2 φ0 , ϕiL2 ,
6.
h curl An , curl ϕ1 iL2 → h curl A0 , curl ϕ1 iL2 ,
7.
hφn∗ ∇φn , ϕ1 iL2 → hφ0∗ ∇φ0 , ϕ1 iL2 ,
8.
hφn ∇φn∗ , ϕ1 iL2 → hφ0 ∇φ0∗ , ϕ1 iL2 ,
9.
h|φn |2 An , ϕ1 iL2 → h|φ0 |2 A0 , ϕ1 iL2 .
For example, for (4), for an appropriate real K̃ > 0 we have
|h|An |2 φn , ϕiL2 − h|A0 |2 φ0 , ϕiL2 |

= |h|An |2 φn , ϕiL2 − h|A0 |2 φn , ϕ iL2 + h|A0 |2 φn , ϕiL2 − h|A0 |2 φ0 , ϕiL2 |
≤ |h|(An |2 − |A0 |2 )φn , ϕiL2 + h|A0 |2 (φn − φ0 ), ϕiL2 |
≤ |h|(An | − |A0 |)(|An | + |A0 |)φn , ϕiL2 + h|A0 |2 (φn − φ0 ), ϕiL2 |
≤ k(An | + |A0 |)k0,4,Ω k|An | − |A0 |k0,4,Ω kφn |0,2,Ω kϕk∞ + k|A0 |2 k0,2,Ω kφn − φ0 k0,2,Ω kϕk∞
≤ K̃(k|An | − |A0 |k0,4,Ω + kφn − φ0 k0,2,Ω )
→ 0, as n → ∞. (23.12)
The other items may be proven similarly.

Now let ϕ ∈ C∞ (Ω, C). Observe that
hun , ϕ iL2 (∂ Ω,C)

2e
= ih̄∇φn + An φn · n, ϕ
c L2 (∂ Ω,C)

2e
= ih̄∇φn + An φn , ∇ϕ
c L2 (Ω,C3 )

2e
+ div ih̄∇φn + An φn , ϕ
c L2 (Ω,C)

2e
→ ih̄∇φ0 + A0 φ0 , ∇ϕ
c L2 (Ω,C3 )

2e
+ div ih̄∇φ0 + A0 φ0 , ϕ
c L2 (Ω,C)

2e
= ih̄∇φ0 + A0 φ0 · n, ϕ . (23.13)
c L2 (∂ Ω,C)
From this and from

hun , ϕiL2 (∂ Ω,C) → hu0 , ϕiL2 (∂ Ω,C) ,
we have
2e
ih̄∇φ0 + A0 φ0 · n − u0 , ϕ = 0, ∀ϕ ∈ C∞ (Ω, C),
c 2
L (∂ Ω,C)
so that in such a distributional sense,

2e
ih̄∇φ0 + A0 φ0 · n = u0 , on ∂ Ω.
c
The other boundary condition may be dealt similarly. Thus, from these last results we may infer that in
the distributional sense,

1
2
 2m −ih̄∇ − 2e 2
c A0 φ0 + α|φ0 | φ0 − β φ0 = 0, in Ω,
(23.14)
ih̄∇φ0 + 2e

·

A
c 0 0φ n = u0 , on ∂ Ω,
and  4π ˜
 curl curl A0 = curl B0 + c J0 ,


in Ω,



 curl curl A0 = curl B0 , in Ω1 \ Ω,


(23.15)
div A0 = 0, in Ω1 ,








A0 · n = 0, on ∂ Ω1

where
ieh̄ 2e 2
J˜0 = − (φ0∗ ∇φ0 − φ0 ∇φ0∗ ) − |φ0 |2 A0 .
2m mc
Hence, (φ0 , A0 , u0 ) ∈ C .
Finally, from φn → φ0 in L2 and L4 , φn * φ0 weakly in W 1,2 , un * u0 weakly in L2 (∂ Ω), by continuity
in φ and the convexity of J in ∇φ and u, we have,
η = lim inf J(φn , An , un ) ≥ J(φ0 , A0 , u0 ).

n→∞
23.3 A method to obtain approximate numerical solutions for a class of

partial differential equations
In this section we present a new procedure to obtain approximate solutions for a large class of partial differ-
ential equations.
We emphasize there is some error in the process, so that the solutions obtained are just qualitative (but
indeed of very good quality as a first approximation).
Let Ω = [0, 1] × [0, 1] and consider the Ginzburg-Landau type equation
−ε∇2 u + αu3 − β u = f , in Ω,

(23.16)
u = 0, on ∂ Ω,
where ε > 0, α > 0 and β > 0.

For the generalized method of lines, we discretize the domain in x, in N vertical lines, defining d = 1/N,
so that the system (23.16) will approximately stand for,
un+1 − 2un + un−1 ∂ 2 un

ε + ε − αu3n + β un + fn = 0,
d2 ∂ y2
∀n ∈ {1, ..., N − 1}.
From this, we may write,
d2 d2
un+1 − 2un + un−1 + T (un ) + fn = 0, (23.17)
ε ε
where
∂ 2 un
T (un ) = − αu3n + β un , ∀n ∈ {1, . . . , N − 1}.
∂ y2
In particular, for n = 1, recalling that u0 = 0, we obtain
d2 d2
u2 − 2u1 + T (u1 ) + f1 = 0,
ε ε
so that
u2 1 d2 1 d2
u1 = + T (u1 ) + f1 ,
2 2 ε 2 ε
that is,
d2 d2
u1 = a1 u2 + b1 T (u1 ) + c1 + E1 ,
ε ε
where
1 1 1
a1 = , b1 = , c1 = f1 and E1 = 0.
2 2 2
Reasoning inductively, having
d2 d2
un−1 = an−1 un + bn−1 T (un−1 ) + cn−1 + En−1 ,
ε ε
from this and (23.17), we have
d2 d2
un+1 − 2un + an−1 un + bn−1 T (un−1 ) + cn−1 + En−1
ε ε
d2 d2
+T (un ) + fn = 0, (23.18)
ε ε
so that
d2 d2
un = an un+1 + bn T (un ) + cn + En
ε ε
where
1
an = ,
2 − an−1
bn = an (bn−1 + 1),
cn = an (cn−1 + fn ),
and
d2
En = an En−1 + an bn−1 (T (un−1 ) − T (un )) ,
ε
∀n ∈ {2, . . . , N − 1}.
In particular, for n = N − 1, recalling that uN = 0, we may obtain uN−1 as the solution of the approximate
ordinary differential equation,
d2 d2
uN −1 ≈ aN −1 uN + bN −1 T (uN−1 ) + cN−1 (23.19)
ε ε
that is,
∂ 2 uN −1 d2 d2

uN−1 ≈ bN −1 − α uN3 −1 + β uN−1 + cN−1 ,
∂ y2 ε ε
with the boundary conditions uN −1 (0) = uN −1 (1) = 0.
Similarly, having uN −1 , we may obtain uN −2 through the approximate equation,
d2 d2
uN−2 ≈ aN−2 uN−1 + bN −2 T (uN−2 ) + cN−2 ,
ε ε
that is,
∂ 2 uN −2 d2 d2

uN −2 ≈ aN −2 uN −1 + bN −2 − αu3N −2 + β uN−2 + cN−2 ,
∂ y2 ε ε
and so on, up to finding u1 .
The problem is thus approximately solved. We present numerical results for α = β = 1 and N = 500
lines (mesh 500 × 500).
For ε = 1, 0.1, 0.01 and 0.001, please see the Figures 23.1, 23.2, 23.3, 23.4 and 23.5, for the respective
graphs.
We observe that for small values of ε and in particular for ε = 0.0001, the solution u(x, y) is close to the
constant value 1.32 on the almost whole domain, which is the approximate solution of equation u3 − u − 1 =
0.
Figure 23.1: Solution u(x, y) for ε = 1.
Figure 23.2: Solution u(x, y) for ε = 0.1.


23.4 Conclusion
In this chapter we have developed a global existence result for a control problem related to the Ginzburg-
Landau system in superconductivity. We emphasize the control variable u acts on the super-conducting sam-
ple boundary, whereas the state variables, namely, the order parameter φ and the magnetic potential A are
defined on Ω and Ω1 , respectively. The problem has non-linear constraints and the cost functional is non-
convex. Finally, we highlight the London Gauge assumption and the Friedrichs Inequality have a fundamental
role in the establishment of the main results.
Chapter 24
Duality for a Semi-Linear Model

in Micro-Magnetism
24.1 Introduction
This chapter develops a dual variational formulation for a semi-linear model in micro-magnetism. For the
primal formulation we refer to references [14, 15] for details. In particular we refer to the original results
presented in [15], emphasizing the present work is their natural continuation and extension. We also highlight
the present work develops real relevant improvements relating the previous similar results in [14].
Let Ω ⊂ R3 be an open bounded set with a a regular (lipschitzian) boundary denoted by ∂ Ω. By a
regular lipschitzian boundary ∂ Ω we mean regularity enough so that the Sobolev imbedding theorem and
relating results, the trace theorem and the standard Gauss-Green formulas of integration by parts to hold. The
corresponding outward normal to ∂ Ω is denoted by n = (n1 , n2 , n3 ). Also, we denote by 0 either the zero
vector in R3 or the zero in an appropriate function space.
Under such assumptions and notations, consider the problem of finding the magnetization m : Ω → R3 ,
which minimizes the functional
Z Z Z
α
J(m, f ) = |∇m|22 dx + ϕ(m(x)) dx − H(x) · m dx
2 Ω Ω Ω
1
Z
+ | f (x)|22 dx, (24.1)
2 R3
where
m = (m1 , m2 , m3 ) ∈ W 1,2 (Ω; R3 ) ≡ Y1 , |m(x)|2 = 1, in Ω (24.2)
and f ∈ L2 (R3 ; R3 ) ≡ Y2 is the unique field determined by the simplified Maxwell’s equations
curl( f ) = 0, div(− f + mχΩ ) = 0, in R3 . (24.3)
Here H ∈ L2 (Ω; R3 ) is a known external field and χΩ is a function defined by

1, if x ∈ Ω,
χΩ (x) = (24.4)
0, otherwise.
Duality for a Semi-Linear Model in Micro-Magnetism 477
The term Z
α
|∇m|22 dx
2 Ω
is called the exchange energy, where v
u 3
u
|m|2 = t ∑ m2k
k=1
and
3
|∇m|22 = ∑ |∇mk |22 .
k=1
Finally, ϕ(m) represents the anisotropic contribution and is given by a multi-well functional whose minima
establish the preferred directions of magnetization.
Remark 24.1.1 Here some brief comments on the references. Relating and similar problems are addressed
in [14]. The basic results on convex and variational analysis used in this text may be found in [33, 14, 64, 78].
About the duality principles, we have been greatly inspired and influenced by the work of J.J. Telega and W.R.
Bielski. In particular, we would refer to [11], published in 1985, as the first article to successfully apply the
convex analysis approach to non-convex and non-linear mechanics.
Finally, an extensive study on Sobolev spaces may be found in [1].
Remark 24.1.2 At some points of our analysis we refer to the problems in question after discretization. In
such a case we are referring to their approximations in a finite element or finite differences context.
24.2 The duality principle for the semi-linear model

We consider first the case of a uniaxial material where ϕ(m) = β (1 − |m · e|).
Observe that
ϕ(m) = min{β (1 + m · e), β (1 − m · e)}
where β > 0 and e ∈ R3 is a unit vector.
The main duality principle is summarized by the following theorem.
Theorem 24.2.1 Considering the previous statements and notations, define J : Y1 ×Y2 × B → R = R ∪ {+∞}
by
K
J(m, f ,t) = G0 (m) − hmi , mi iL2 + G1 (m,t) + G2 ( f )
2
+Ind0 (m) + Ind1 (m, f ) + Ind2 ( f ), (24.5)
where
α
G0 (m) = h∇mi , ∇mi iL2 ,
2
K
Z
G1 (m,t) = (tg1 (m) + (1 − t)g2 (m)) dx − hHi , mi iL2 + hmi , mi iL2 ,
Ω 2
1
Z
G2 ( f ) = | f (x)|2 dx,
2 R3
g1 (m) = β (1 + m · e),
g2 (m) = β (1 − m · e),

0, if |m(x)|2 = 1, in Ω
Ind0 (m) = (24.6)
+∞, otherwise,
if div(− f + mχΩ ) = 0, in R3

0,
Ind1 (m, f ) = (24.7)
+∞, otherwise,
if curl f = 0, in R3

0,
Ind2 ( f ) = (24.8)
+∞, otherwise.
We recall the present case refers to a uniaxial material with exchange of energy, that is α > 0.
Here, e = (e1 , e2 , e3 ) ∈ R3 is a unit vector.
Under such hypotheses, we have,
inf {J(m, f ,t)} ≥ sup {J˜∗ (λ )},
(m, f ,t)∈Y1 ×Y2 ×B λ ∈A∗
where
J˜∗ (λ ) = inf J ∗ (λ , z∗ ,t),
(z∗ ,t)∈Y4∗ (λ )×B
J ∗ (λ , z∗ ) = F̃ ∗ (z∗ ) − G̃∗ (λ , z∗ ,t),

and where, for the discretized problem version,

∗ ∗ ∗ K
F̃ (z ) = sup hzi , ∇mi iL2 + G0 (m) − hmi , mi iL2 . (24.9)
m∈Y1 2
Here K > 0 is such that
K
−G0 (m) + hmi , mi iL2 > 0, ∀m ∈ Y1 , m 6= 0.
2
G̃∗ (λ , z∗ ,t) = G∗1 (λ , z∗ ,t) + G∗2 (λ )

= sup {hz∗i , ∇mi iL2 + hλ2 , div(mχΩ ) − f )iL2
(m, f )∈Y1 ×Y2
+hλ1 , curl f iL2
3
−G1 (m,t) − G0 ( f ) − hλ3 , ( ∑ mi2 ) − 1iL2 } (24.10)
i=1
where, more specifically,

2
Z ∑3 ∂ λ2 ∗
1 i=1 − ∂ xi + Hi + β (1 − 2t)ei − div zi
G∗1 (λ , z∗ ,t) = dx
2 Ω λ3 + K
1 1
Z Z
− λ3 dx + β dx, (24.11)
2 Ω 2 Ω
1
G∗2 (λ ) = k∇λ2 − curl ∗ λ1 k22 .
2
Also,
A1 = {λ ∈ Y3 : λ3 + K > 0, in Ω},
and from the standard second order sufficient optimality condition for a local minimum in m, we define
A2 = {λ ∈ Y3 :
3
G0 (m) + hλ3 , ∑ mi2 iL2 > 0,
i=1
∀m ∈ Y1 , such that m =
6 0} (24.12)
where
A∗ = A1 ∩ A2 ,
Y = Y ∗ = L2 (Ω; R3 ) = L2 ,
λ = (λ1 , λ2 , λ3 ) ∈ Y3 = W 1,2 (R3 ; R3 ) ×W 1,2 (R3 ) × L2 (Ω),
Y1 = W 1,2 (Ω; R3 ),
Y2 = W 1,2 (R3 ; R3 ),
Y4∗ (λ ) = {z∗ ∈ [Y ∗ ]3 : z∗i · n + λ2 ni = 0, on ∂ Ω, ∀i ∈ {1, 2, 3}},
B = {t measurable : 0 ≤ t ≤ 1, in Ω}.
Finally, suppose there exists (λ0 , z∗0 , t0 ) ∈ A∗ × Y4∗ (λ0 ) × B such that for an appropriate λ4 ∈ L2 (Ω) we
have
δ J ∗ (λ0 , z∗0 ,t0 ) + hλ4 ,t02 − t0 iL2 = 0,

δz2∗ z∗ J ∗ (λ0 , z∗0 ,t0 ) > 0

and for the concerning Hessian
det δz2∗ ,t J ∗ (λ0 , z∗0 ,t0 ) + hλ4 ,t02 − t0 iL2 > 0, in Ω.

J(m0 , f0 ,t0 ) = inf J(m, f ,t)

(m, f ,t)∈Y ×Y1 ×B
= ˜∗
sup J (λ )
λ ∈A∗
˜∗
= J (λ0 )
= J ∗ (λ0 , z∗0 ,t0 ). (24.13)
G∗1 (λ , z∗ ,t) + G∗2 (λ )

≥ hz∗i , ∇mi iL2 + hλ2 , div(mχΩ ) − f )iL2
+hλ1 , curl f iL2
* +
3
− λ3 , ∑ m2i − 1 − G1 (m,t) − G2 ( f )
i=1 L2
≥ hz∗i , ∇mi iL2 − G1 (m,t) − G2 ( f ) − Ind0 (m) − Ind1 (m, f ) − Ind2 ( f ), (24.14)
∀(m, f ,t) ∈ Y1 ×Y2 × B, z∗ ∈ Y4∗ (λ ), λ ∈ A∗ so that,
−F̃ ∗ (z∗ ) + G∗1 (λ , z∗ ,t) + G∗2 (λ )

≥ −F̃ ∗ (z∗ ) + hz∗i , ∇mi iL2 − G1 (m,t) − G2 ( f )
−Ind0 (m) − Ind1 (m, f ) − Ind2 ( f ), (24.15)
∀(m, f ,t) ∈ Y1 ×Y2 × B, z∗ ∈ Y4∗ (λ ), λ ∈ A∗ and hence,
sup {−F̃ ∗ (z∗ ) + G∗1 (λ , z∗ ,t) + G∗2 (λ )}

z∗ ∈Y4∗ (λ )
≥ sup {−F̃ ∗ (z∗ ) + hz∗i , ∇mi iL2 } − G1 (m,t) − G2 ( f )

z∗ ∈Y4∗ (λ )
−Ind0 (m) − Ind1 (m, f ) − Ind2 ( f ), (24.16)

∀(m, f ,t) ∈ Y1 ×Y2 × B, λ ∈ A∗ so that that is,
sup {−F̃ ∗ (z∗ ) + G∗1 (λ , z∗ ,t) + G∗2 (λ )}

z∗ ∈Y4∗ (λ )
K
≥ −G0 (m) + hmi , mi iL2 − G1 (m,t) − G2 ( f )
2
−Ind0 (m) − Ind1 (m, f ) − Ind2 ( f )
= −J(m, f ,t), (24.17)
∀(m, f ,t) ∈ Y1 ×Y2 × B, λ ∈ A∗ . Thus,
J(m, f ,t) ≥ inf {F̃ ∗ (z∗ ) − G∗1 (λ , z∗ ,t) + G∗2 (λ )}

z∗ ∈Y4∗ (λ )
= inf J ∗ (λ , z∗ ,t), ∀(m, f ,t) ∈ Y1 ×Y1 × B, λ ∈ A∗ . (24.18)

z∗ ∈Y4∗ (λ )
Therefore,
( )
∗ ∗
inf J(m, f ,t) ≥ sup inf J (λ , z ,t)
(m, f ,t)∈Y1 ×Y2 ×B λ ∈A∗ (z∗ ,t)∈Y4∗ (λ )×B
= sup J˜∗ (λ ). (24.19)

λ ∈A∗
By the hypotheses, (λ0 , z∗0 ,t0 ) ∈ A∗ ×Y4∗ (λ0 ) × B is such that
δ {J ∗ (λ0 , z∗0 ,t0 ) + hλ4 ,t02 − t0 iL2 } = 0.
From the variation in z∗ ,

∂ F̃ ∗ (z∗0 )
− ∇(m0 )i = 0, in Ω, (24.20)
∂ z∗i
where
− ∂ (λ0 )2 ∗
∂ xi + Hi + β (1 + 2t0 )ei − div[(z0 )i ]
(m0 )i = . (24.21)
(λ0 )3 + K
From the variation in λ3 , we get
3
∑ (m0 )i2 = 1, in Ω.
i=1
From the variation in λ2 , we have
div(m0 χΩ − f0 ) = 0, in R3 , (24.22)
where
f0 = curl(λ1 )0 − ∇(λ0 )2 . (24.23)
From the variation in λ1 , we obtain,
curl f0 = 0, in R3 . (24.24)
From (24.20), we also have
K
F̃ ∗ (z∗0 ) = h(z∗0 )i , ∇(m0 )i iL2 + G0 (m0 ) − h(m0 )i , (m0 )i iL2 . (24.25)
2
By (24.21), (24.22), (24.23) and (24.24), we get
G∗1 (λ0 , z∗0 ,t0 ) + G∗2 (λ0 )

= h(z∗0 )i , ∇(m0 )i iL2 + hλ2 , div(m0 χΩ ) − f0 )iL2
+h(λ0 )1 , curl f0 iL2
* +
3
− (λ0 )3 , ∑ (m0 )2i − 1 − G1 (m0 ,t0 ) − G2 ( f0 )
i=1 L2
= h(z∗0 )i , ∇(m0 )i iL2 − G1 (m0 ,t0 ) − G2 ( f0 )
−Ind0 (m0 ) − Ind1 (m0 , f0 ) − Ind2 ( f0 ), (24.26)
From (24.25) and (24.26) we obtain,
J ∗ (λ0 , z∗0 ,t0 )

= F̃ ∗ (z∗0 ) − G∗1 (λ0 , z∗0 ,t0 ) − G∗2 (λ0 )
= G0 (m0 ) + G1 (m0 ,t0 ) + G2 ( f0 )
+Ind0 (m0 ) + Ind1 (m0 , f0 ) + Ind2 ( f0 )
= J(m0 , f0 ,t0 ). (24.27)
Finally, from the hypotheses

δz2∗ z∗ J ∗ (λ0 , z∗0 ,t0 ) > 0
and
det δz2∗ ,t J ∗ (λ0 , z∗0 , t0 ) + hλ4 ,t02 − t0 iL2 > 0, in Ω.

Since the optimization in question in (z∗ ,t) is quadratic, we obtain,
J˜∗ (λ0 ) = J ∗ (λ0 , z∗0 ,t0 ).
From this, (24.19) and (24.27), we have,
J(m0 , f0 ,t0 ) = inf J(m, f ,t)

(m, f ,t)∈Y ×Y1 ×B
= ˜∗
sup J (λ )
λ ∈A∗
˜∗
= J (λ0 )
= J ∗ (λ0 , z∗0 ,t0 ). (24.28)
24.3 Conclusion
In this chapter we have developed a duality principle for a non-convex semi-linear model in micro-
magnetism.
In a second step we present an optimality criterion for the dual formulation. The results are obtained
through standard tools of convex analysis and duality theory.
Chapter 25
About Numerical Methods for

Ordinary and Partial Differential
Equations
25.1 Introduction
This chapter develops numerical methods for a large class of non-linear ordinary and partial differential
equations.
More specifically, such a chapter is concerned with a kind of matrix version of the Generalized Method
of Lines. Applications are developed for models in physics and engineering.
25.2 On the numerical procedures for Ginzburg-Landau type ODEs

We first apply the Newton’s method to a general class of ordinary differential equations. The solution here is
obtained similarly as for the generalized method of lines procedure. See the next sections for details on such
a method for PDEs.
For a C1 class function f and a continuous function g, consider the second order equation
 00
 u + f (u) + g = 0, in [0, 1]
(25.1)
u(0) = u0 , u(1) = u f .

In finite differences we have the approximate equation:

un+1 − 2un + un−1 + f (un )d 2 + gn d 2 = 0.
Assuming such an equation is non-linear and linearizing it about a first solution {ũ}, we have (in fact this is
an approximation),
un+1 − 2un + un−1 + f (ũn )d 2 + f 0 (ũn )(un − ũn )d 2 + gn d 2 = 0.
Thus we may write
un+1 − 2un + un−1 + An un d 2 + Bn d 2 = 0,
About Numerical Methods for Ordinary and Partial Differential Equations 483
where
An = f 0 (ũn ),
and
Bn = f (ũn ) − f 0 (ũn )ũn + gn .
In particular, for n = 1 we get
u2 − 2u1 + u0 + A1 u1 d 2 + B1 d 2 = 0.
Solving such an equation for u1 , we get
u1 = a1 u2 + b1 u0 + c1 ,
where
a1 = (2 − A1 d 2 )−1 , b1 = a1 , c1 = a1 B1 d 2 .
un−1 = an−1 un + bn−1 u0 + cn−1 ,
and
un+1 − 2un + un−1 + An un d 2 + Bn d 2 = 0,
we get
un+1 − 2un + an−1 un + bn−1 u0 + cn−1 + An un d 2 + Bn d 2 = 0,

so that
un = an un+1 + bn u0 + cn ,
where
an = (2 − an−1 − An d 2 )−1 ,
bn = an bn−1 ,
and
cn = an (cn−1 + Bn d 2 ),
∀n ∈ 1, ..., N − 1.
We have thus obtained
un = an un+1 + bn u0 + cn ≡ Hn (un+1 ), ∀n ∈ {1, ..., N − 1},
and in particular,
uN−1 = HN −1 (u f ),
so that we may calculate,
uN−2 = HN−2 (uN−1 ),
uN−3 = HN −3 (uN−2 ),
and so on, up to finding,
u1 = H1 (u2 ).
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Figure 25.1: The solution u(x) by Newton’s method for ε = 1.
1.4
1.2
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Figure 25.2: The solution u(x) by Newton’s method for ε = 0.1.
The next step is to replace {ũn } by the {un } calculated, and repeat the process up to the satisfaction of an
appropriate convergence criterion. We present numerical results for the equation,
 3
 u00 − uε + εu + g = 0, in [0, 1]
(25.2)
u(0) = 0, u(1) = 0,

where,
1
g(x) = ,
ε
The results are obtained for ε = 1.0, ε = 0.1, ε = 0.01 and ε = 0.001. Please see Figures 25.1, 25.2, 25.3
and 25.4 respectively.
1.4
1.2
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
1.4
1.2
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
25.3 Numerical results for related P.D.E.s

25.3.1 A related P.D.E on a special class of domains
We start by describing a similar equation, but now in a two dimensional context. Let Ω ⊂ R2 be an open,
bounded, connected set with a regular boundary denoted by ∂ Ω. Consider a real Ginzburg-Landau type
equation (see [4], [9], [55], [56] for details about such an equation), given by

 ε∇2 u − αu3 + β u = f , in Ω
(25.3)
u = 0, on ∂ Ω,

where α , β , ε > 0, u ∈ U = W01,2 (Ω), and f ∈ L2 (Ω). The corresponding primal variational formulation is
represented by J : U → R, where
Z Z Z Z
ε α β
J(u) = ∇u · ∇u dx + u4 dx − u2 dx + f u dx.
2 Ω 4 Ω 2 Ω Ω
25.3.2 About the matrix version of G.M.O.L.

The generalized method of lines was originally developed in [19]. In this work we address its matrix version.
Consider the simpler case where Ω = [0, 1] × [0, 1]. We discretize the domain in x, that is, in N + 1 verti-
cal lines obtaining the following equation in finite differences (see [73] for details about finite differences
schemes).
ε(un+1 − 2un + un−1 )
+ εM2 un /d12 − αu3n + β un = fn , (25.4)
d2
∀n ∈ {1, ..., N − 1}, where d = 1/N and un corresponds to the solution on the line n. The idea is to apply the
Newton’s method. Thus choosing a initial solution {(u0 )n } we linearize (25.4) about it, obtaining the linear
equation:
3αd 2
un+1 − 2un + un−1 + M̃2 un − (u0 )2n un
ε
2α β d2 d2
+ (u0 )3n d 2 + un − fn = 0, (25.5)
ε ε ε
2
where M̃2 = M2 dd 2 and
1  
−2 1 0 0 ··· 0

 1 −2 1 0 ··· 
 0
 0 1 −2 1 ···  0
M2 =  , (25.6)
 
 ... ..
.
..
.
..
.
..
. 
.
..
 
 0 0 ··· 1 −2 1 
0 0 ··· ··· 1 −2
with N1 lines corresponding to the discretization in the y axis. Furthermore d1 = 1/N1 .
In particular, for n = 1 we get
3αd 2
u2 − 2u1 + M̃2 u1 − (u0 )21 u1
ε
2α β d2 d2
+ (u0 )31 d 2 + u1 − f1 = 0. (25.7)
ε ε ε
Denoting
αd 2 β d2
M12 [1] = 2Id − M̃2 + 3 (u0 )12 Id − Id ,
ε ε
where Id denotes the (N1 − 1) × (N1 − 1) identity matrix,
2αd 2 d2
Y0 [1] = (u0 )31 − f1 ,
ε ε
and M50 [1] = M12 [1]−1 , we obtain
u1 = M50 [1]u2 + z[1].
where
z[1] = M50 [1] ·Y0 [1].
Now for n = 2 we get
3αd 2
u3 − 2u2 + u1 + M̃2 u2 − (u0 )22 u2
ε
2α β d2 d2
+ (u0 )23 d 2 + u2 − f2 = 0, (25.8)
ε ε ε
that is,
3αd 2
u3 − 2u2 + M50 [1]u2 + z[1] + M̃2 u2 − (u0 )22 u2
ε
2α β d2 d2
+ (u0 )32 d 2 + u2 − f2 = 0, (25.9)
ε ε ε
so that denoting
αd 2 β d2
M12 [2] = 2Id − M̃2 − M50 [1] + 3 (u0 )22 Id − Id ,
ε ε
2αd 2 d2
Y0 [2] = (u0 )32 − f2 ,
ε ε
and M50 [2] = M12 [2]−1 , we obtain
u2 = M50 [2]u3 + z[2],
where
z[2] = M50 [2] · (Y0 [2] + z[1]).
Proceeding in this fashion, for the line n we obtain
3αd 2
un+1 − 2un + M50 [n − 1]un + z[n − 1] + M̃2 un − (u0 )n2 un
ε
2α β d2 d2
+ (u0 )3n d 2 + un − fn = 0, (25.10)
ε ε ε
so that denoting
αd 2 β d2
M12 [n] = 2Id − M̃2 − M50 [n − 1] + 3 (u0 )n2 Id − Id ,
ε ε
and also denoting
2αd 2 d2
Y0 [n] = (u0 )3n − fn ,
ε ε
and M50 [n] = M12 [n]−1 , we obtain
un = M50 [n]un+1 + z[n],
where
z[n] = M50 [n] · (Y0 [n] + z[n − 1]).
Observe that we have
uN = θ ,
where θ denotes the zero matrix (N1 − 1) × 1, so that we may calculate
uN−1 = M50 [N − 1] · uN + z[N − 1],
and
uN −2 = M50 [N − 2] · uN −1 + z[N − 2],
and so on, up to obtaining
u1 = M50 [1] · u2 + z[1].
The next step is to replace {(u0 )n } by {un } and thus to repeat the process until convergence is achieved.
This is the Newton’s Method, what seems to be relevant is the way we inverted the big matrix ((N1 −
1) · (N − 1)) × ((N1 − 1) · (N − 1)), in fact instead of inverting it directly we have inverted N − 1 matrices
(N1 − 1) × (N1 − 1) through an application of the generalized method of lines.
25.3.3 Numerical results for the concerning partial differential equation

We solve the equation

 ε∇2 u − αu3 + β u + 1 = 0, in Ω = [0, 1] × [0, 1]
(25.11)
u = 0, on ∂ Ω,

through the algorithm specified in the last section. We consider α = β = 1. For ε = 1.0 see Figure 25.5, for
ε = 0.0001 see Figure 25.6.
0.08
0.06
0.04
0.02
0
1
1
0.8
0.5 0.6
0.4
0.2
0 0
Figure 25.5: The solution u(x, y) for ε = 1.0.
1.6
1.4
1.2
0.8
1
1
0.8
0.5 0.6
0.4
0.2
0 0
Figure 25.6: The solution u(x, y) for ε = 0.0001.
25.4 A proximal algorithm for optimization in Rn

In this section we develop a proximal algorithm for constrained optimization.
Let f : Rn → R be a C2 class function. Consider the problem of minimizing locally f subject to g(x) ≤ 0,
where g : Rn → R is a given C2 class function.
The lagrangian for this problem, denoted by L : Rn+1 → R, may be expressed by
L(x, λ ) = f (x) + λ 2 g(x).
We define the proximal formulation for such a problem, denoted by L p by

K
L p (x, λ , xk ) = f (x) + λ 2 g(x) + |x − xk |2 .
2
25.4.1 The main result

Linearizing L p , we propose the following procedure for looking for a critical point of such a function:
Consider
1
L̃ p (x, λ , xk ) = f (xk ) + f 0 (xk ) · (x − xk ) + [ f 00 (xk )(x − xk )] · (x − xk )
2
0 K
+λ (g(xk ) + g (xk ) · (x − xk )) + |x − xk |2 .
2
2
Hence from
∂ L̃ p (x, λ , xk )
=0
∂x
we obtain
f 00 (xk )(x − xk ) + K(x − xk ) + f 0 (xk ) + λ 2 g0 (xk ) = 0,
that is,
x − xk = −( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 g0 (xk )),
and therefore
x(λ , xk ) = xk − ( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 g0 (xk )),
where Id denotes the n × n identity matrix.
We define L1 (λ , xk ) = L̃ p (x(λ , xk ), xk , λ ) so that
1
L1 (λ , xk ) = − [( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 g0 (xk ))] · ( f 0 (xk ) + λ 2 g0 (xk ))
2
+ f (xk ) + λ 2 g(xk ) (25.12)
From
∂ L1 (λ , xk )
= 0,
∂λ
we get
[( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 g0 (xk ))] · g0 (xk )λ − λ g(xk ) = 0, (25.13)
so that we have two solutions,
λ1 = 0
and
[( f 00 (xk ) + KId )−1 f 0 (xk )] · g0 (xk ) − g(xk )

(λ21 )2 (xk ) = − . (25.14)
[( f 00 (xk ) + KId )−1 g0 (xk )] · g0 (xk )
Observe that if (λ21 )2 (xk ) < 0 then λ21 (xk ) is complex so that, from the condition λ 2 ≥ 0, we obtain
λ 2 (xk ) = max{0, (λ21 )2 (xk )}.
Also, from the generalized inverse function theorem λ 2 (x) is locally Lipschtzian (see [54, 57, 78, 60] for
details). Hence, we may infer that for a given x0 ∈ Rn there exists r > 0 and K̂3 > 0 such that
|λ 2 (x) − λ 2 (y)| ≤ K̂3 |x − y|,

∀x, y ∈ Br (x0 ). With such results in mind, for such an x0 ∈ Rn , define {xk } by
x1 = x0 − ( f 00 (x0 ) + KId )−1 ( f 0 (x0 ) + λ 2 (x0 )g0 (x0 )),
xk+1 = xk − ( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 (xk )g0 (xk )), ∀k ∈ N.

Assume
g(x0 ) < 0 (25.15)
and there exists K̂1 such that | f 00 (x)| ≤ K̂1 , ∀x ∈ Br (x0 ).
Define !
0
K3 = K̂3 sup |g (x)| ,
x∈Br (x0 )
α1 = 2K3 /|K − K̂1 | and suppose
f 00 (x) + λ 2 (y)g00 (x) ≥ α1 (K̂1 + K)Id ∀x, y ∈ Br (x0 ). (25.16)
Suppose also K is such that K > K̂1 ,

0 < α1 < 1,
α1 α1
1− Id ≤ (( f 00 (x) + KId )−1 )( f 00 (y) + KId ) ≡ H(x, y) ≤ 1 + Id , (25.17)
4 4
∀x, y ∈ Br (x0 ) and
f 00 (x) + λ (y)2 g00 (x) α1
0≤ ≤ 1− Id , ∀x, y ∈ Br (x0 ). (25.18)
K − K̂1 2
Observe that since | f 00 (x)| ≤ K̂1 , we have
0 ≤ (K − K̂1 )Id ≤ f 00 (x) + KId ,
so that
1
( f 00 (x) + KId )−1 ≤ Id , (25.19)
K − K̂1
and
K3 α1
|( f 00 (x) + KId )−1 |K3 ≤ = , ∀x ∈ Br (x0 ). (25.20)
|K − K̂1 | 2
Assume K > 0 is such that
x1 ∈ Br(1−α0 ) (x0 )
and suppose the induction hypotheses
x2 , . . . , xk+1 ∈ Br (x0 ).
where 0 < α0 < 1 is specified in the next lines.
Note that,
xk+2 − xk+1 = −( f 00 (xk+1 ) + KId )−1 ( f 0 (xk+1 ) + λ 2 (xk+1 )g0 (xk+1 )),
and
xk+1 − xk = −( f 00 (xk ) + KId )−1 ( f 0 (xk ) + λ 2 (xk )g0 (xk )),
so that,
( f 00 (xk+1 ) + KId )(xk+2 − xk+1 ) = −( f 0 (xk+1 ) + λ 2 (xk+1 )g0 (xk+1 )),
and
( f 00 (xk ) + KId )(xk+1 − xk ) = −( f 0 (xk ) + λ 2 (xk )g0 (xk )).
Therefore,
( f 00 (xk+1 ) + KId )(xk+2 − xk+1 )

= ( f 00 (xk ) + KId )(xk+1 − xk )
−( f 0 (xk+1 ) + λ 2 (xk+1 )g0 (xk+1 )) + ( f 0 (xk ) + λ 2 (xk )g0 (xk ))
= ( f 00 (xk ) + KId )(xk+1 − xk ) − ( f 0 (xk+1 ) + λ 2 (xk+1 )g0 (xk+1 )) + ( f 0 (xk ) + λ 2 (xk+1 )g0 (xk ))
−( f 0 (xk ) + λ 2 (xk+1 )g0 (xk )) + ( f 0 (xk ) + λ 2 (xk )g0 (xk ))
= ( f 00 (xk ) + KId )(xk+1 − xk )
−( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k ))(xk+1 − xk ) − (λ 2 (xk+1 ) − λ 2 (xk ))g0 (xk )
where x̃k is on the line connecting xk and xk+1 .

Thus,
xk+2 − xk+1 = ( f 00 (xk+1 ) + KId )−1 [( f 00 (xk ) + KId )(xk+1 − xk )

−( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k ))(xk+1 − xk )
−(λ 2 (xk+1 ) − λ 2 (xk ))g0 (xk )], (25.21)
so that
|xk+2 − xk+1 | ≤ |H(xk+1 , xk ) − (( f 00 (xk+1 ) + KId )−1 )( f 00 (x̃k )

+λ 2 (xk+1 )g00 (x̃k ))||xk+1 − xk |
+|( f 00 (xk+1 ) + K Id )−1 |K3 |xk+1 − xk |. (25.22)
Observe that, from (25.16),
f 00 (x̃k ) + λ 2 (x̃k+1 )g00 (xk ) ≥ α1 (K̂1 + K)Id ≥ α1 ( f 00 (xk+1 ) + K Id ),
so that
(( f 00 (xk+1 ) + K Id )−1 )( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k )) ≥ α1 Id .
Hence, from this, (25.17), (25.19) and (25.18), we obtain
α1
Id 1 + − α1 Id
4
≥ H (xk+1 , xk ) − (( f 00 (xk+1 ) + K Id )−1 )( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k ))
α1
≥ Id 1 − − (K Id − K̂1 Id )−1 )( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k ))
4
α1 α1
≥ Id 1 − − Id 1 −
4 2
α1
= Id
4
≥ 0, (25.23)
and therefore,
3α1
|H(xk+1 , xk ) − ( f 00 (xk ) + K Id )−1 ( f 00 (x̃k ) + λ 2 (xk+1 )g00 (x̃k ))| ≤ 1 − .
4
On the other hand, from (25.20) we have,
α1
|( f 00 (xk ) + K Id )−1 |K3 ≤ .
2
From (25.22) and these last two inequalities, we obtain

3α1 α1 α1
|xk+2 − xk+1 | ≤ 1 − + |xk+1 − xk | = 1 − |xk+1 − xk |.
4 2 4
Thus, denoting α0 = 1 − α1 /4, we have obtained,
|x j+2 − x j+1 | ≤ α0 |x j+1 − x j |, ∀ j ∈ {1, · · · , k + 1}
so that
|x j+2 − x j+1 | ≤ α0 |x j+1 − x j |
≤ α02 |x j − x j−1 |
≤ ···
≤ α0j+1 |x1 − x0 |, ∀ j ∈ {1, · · · , k}. (25.24)
Thus,
|xk+2 − x1 |
= |xk+2 − xk+1 + xk+1 − xk + xk − xk−1 + · · · + x2 − x1 |
≤ |xk+2 − xk+1 | + |xk+1 − xk | + · · · + |x2 − x1 |
k+1
≤ ∑ α0j |x1 − x0 |
j=1
+∞
≤ ∑ α0j |x1 − x0 |
j=1
α0
= |x1 − x0 |, (25.25)
1 − α0
so that
|xk+2 − x0 | ≤ |xk+2 − x1 | + |x1 − x0 |
α0
≤ |x1 − x0 | + |x1 − x0 |
1 − α0
1
= |x1 − x0 |
1 − α0
1
< r(1 − α0 )
1 − α0
= r. (25.26)
Hence, xk+2 ∈ Br (x0 ), and therefore the induction is complete, so that
xk ∈ Br (x0 ), ∀k ∈ N.
Moreover, {xk } is a Cauchy sequence, so that there exists x̃, such that
xk → x̃, as k → ∞.
Finally,
0 = lim (xk+1 − xk )
k→∞
= lim [−( f 00 (xk ) + K Id )−1 ( f 0 (xk ) + λ 2 (xk )g0 (xk ))]
k→∞
= −( f 00 (x̃) + KId )−1 ( f 0 (x̃) + λ̃ 2 g0 (x̃)). (25.27)
Hence, from this and

det( f 00 (x̃) + KId ) =
6 0,
we obtain
f 0 (x̃) + λ̃ 2 g0 (x̃) = 0
In such a case, from (25.13) letting k → ∞, we also obtain
λ̃ 2 g(x̃) = 0.
Thus, if λ̃ 2 > 0, then g(x̃) = 0.

If λ̃ = 0, then f 0 (x̃) = 0 and
[(λ21 )(x̃)]2 ≤ 0
so that from (25.14), since ( f 00 (x̃) + KId )−1 is positive definite, letting k → ∞, we get
g(x̃) = [(λ21 )(x̃)]2 [( f 00 (x̃) + KId )−1 g0 (x̃)] · g0 (x̃) ≤ 0.
That is, in any case,

g(x̃) ≤ 0.
Remark 25.1 For the more general case with m1 equality scalar constraints
h j (x) = 0, ∀ j ∈ {1, . . . , m1 }
and m2 inequality scalar constraints
gl (x) ≤ 0, ∀l ∈ {1, . . . , m2 },
where h j , gl : Rn → R are C2 class functions, ∀ j ∈ {1, . . . , m1 } and ∀l ∈ {1, . . . , m2 }, we assume m1 + m2 < n

and define the Lagrangian L p by
m1 m2
K
L p (x, λ , xk ) = f (x) + ∑ (λh ) j h j (x) + ∑ (λg )l2 gl (x) + |x − xk |2 .
j=1 l=1 2
Linearizing L p , we propose the following procedure for looking for a critical point of such a function:
Consider
1
L̃ p (x, λ , xk ) = f (xk ) + f 0 (xk ) · (x − xk ) + [ f 00 (xk )(x − xk )] · (x − xk )
2
m1
+ ∑ (λh ) j (h j (xk ) + h0j (xk ) · (x − xk ))
j=1
m2
K
+ ∑ (λg )2l (gl (xk ) + g0l (xk ) · (x − xk )) + |x − xk |2 .
l =1 2
Hence, from
∂ L̃ p (x, λ , xk )
= 0,
∂x
we obtain,
m1 m2
f 00 (xk )(x − xk ) + K(x − xk ) + f 0 (xk ) + ∑ (λh ) j h0j (xk ) + ∑ (λg )2l g0l (xk ) = 0,
j=1 l=1
that is, !
m1 m2
00 −1 0
x − xk = −( f (xk ) + KId ) f (xk ) + ∑ (λh ) j h0j (xk ) + ∑ (λg )2l g0l (xk ) ,
j=1 l=1
and therefore,
!
m1 m2
x(λ , xk ) = xk − ( f 00 (xk ) + KId )−1 f 0 (xk ) + ∑ (λh ) j h0j (xk ) + ∑ (λg )2l g0l (xk ) , (25.28)
j=1 l=1
where Id denotes the n × n identity matrix.

We define L1 (λ , xk ) = L̃ p (x(λ , xk ), xk , λ ), so that
" !
m1 m2
1
L1 (λ , xk ) = − ( f 00 (xk ) + KId )−1 0
f (xk ) + ∑ (λh ) j h0j (xk ) + (λg )2l g0l (xk )
∑
2 j=1 l=1
!#
m1 m2
0
· f (xk ) + ∑ (λh ) j h0j (xk ) + ∑ (λg )2l g0l (xk )
j=1 l=1
m1 m2
+ f (xk ) + ∑ (λh ) j h j (xk ) + ∑ (λg )2l gl (xk ). (25.29)
j=1 l=1
From
∂ L1 (λ , xk )
= 0,
∂ (λg )l
we get
"
m1
( f 00 (xk ) + KId )−1 f 0 (xk ) + ∑ (λh ) j h0j (xk )
j=1
!#
m2
+ ∑ (λg2 )l g0l (xk ) · g0l (xk )(λg )l − (λg )l gl (xk ) = 0, (25.30)
l=1
From
∂ L1 (λ , xk )
= 0,
∂ (λh ) j
we have
"
m1
j=1
!#
m2
+ ∑ (λg )2l g0l (xk ) · h0j (xk ) − h j (xk ) = 0, (25.31)
l =1
∀ j ∈ {1, . . . , m1 }. Solving the linear system which comprises these last m1 equations and the m2 equations
"
m1
j=1
!#
m2
+ ∑ (λg )l2 g0l (xk ) · g0l (xk ) − gl (xk ) = 0, (25.32)
l=1
∀l ∈ {1, . . . , m2 }, we may obtain a solution
(λh ) j (xk ), (λg1 )2l (xk ) .

Thus, to obtain a concerning critical point, we follow the following algorithm.
1. Choose x0 ∈ Rn , Kmax ∈ N (Kmax is the maximum number of iterations), set k = 0 and e1 ≈ 10−5 .
2. Obtain a solution
(λh ) j (xk ), (λg1 )2l (xk )

by solving the linear system (in (λh ) j and (λg )2l ) indicated in (25.31) and (25.32).
Observe that if (λg1 )l2 < 0 then (λg1 )l is complex.
To up-date λh and λg proceed as follows:
3. For each l ∈ {1, . . . , m2 } if (λg )2l (xk ) ≤ 0, then set (λg )l (xk ) = 0.
4. Define J = {l ∈ {1, . . . , m2 } such that (λg )2l (xk ) > 0}.
5. Recalculate (λh ) j (xk ) and the non-zero (λg )l2 (xk ) for l ∈ J through the solution of the linear system
(in (λh ) j and (λg )2l )
"
m1
j=1
!#
+ ∑ (λg )2l g0l (xk ) · h0j (xk ) − h j (xk ) = 0, (25.33)
l∈J
∀ j ∈ {1, . . . , m1 } and
"
m1
j=1
!#
+ ∑ (λg )2l g0l (xk ) · g0l (xk ) − gl (xk ) = 0, (25.34)
l∈J
∀l ∈ J.
6. If (λg )2l (xk ) ≥ 0, ∀l ∈ {1, . . . , m2 }, then go to 7, otherwise go to item 3.
7. Up-date xk through the equation
m1
xk+1 = xk − ( f 00 (xk ) + KId )−1 f 0 (xk ) + ∑ (λh ) j (xk )h0j (xk )
j=1
!
m2
+ ∑ (λg )2l (xk )g0l (xk ) . (25.35)
l=1
8. If |xk+1 − xk | < e1 or k > Kmax , then stop, otherwise k := k + 1 and go to 2.

Remark 25.4.1 In this section we have developed an algorithm for constrained optimization in Rn . We prove
the main result only for the special case of a single scalar inequality constraint. However, we highlight the
proof of a more general result involving equality and inequality constraints may be developed in a similar
fashion, as indicated in remark 25.1. We postpone the presentation of the formal details for such a more
general case for a future work.
Also in the context of this last result, let us consider the following example of application.
Let Ω ⊂ R2 be an open, bounded, connected set with a regular boundary denoted by ∂ Ω. Consider the
real Ginzburg-Landau type equation, given by

 ε∇2 u − αu3 + β u = f , in Ω
(25.36)
u = 0, on ∂ Ω,

where α, β , ε > 0, u ∈ U = W01,2 (Ω), and f ∈ L2 (Ω).

Consider the sequence obtained through the algorithm:
1. Set n = 1,
2. Choose z∗1 ∈ L2 (Ω).
3. Compute un by
Z Z
ε α
un = argminu∈U ∇u · ∇u dx + u4 dx
2 Ω 4 Ω

1
Z Z
∗ ∗ 2
−hu, zn iL2 + (z ) dx + f u dx , (25.37)
2β Ω n Ω
which means to solve the equation
 ε∇2 u − αu3 + z∗n = f ,


in Ω
(25.38)
u = 0, on ∂ Ω.

4. Compute z∗n+1 by
Z Z
ε α
z∗n+1 = argminz∗ ∈L2 (Ω) ∇un · ∇un dx + u4 dx
2 Ω 4 Ω n

1
Z Z
−hun , z∗ iL2 + (z∗ )2 dx + f un dx , (25.39)
2β Ω Ω
that is,
z∗n+1 = β un .
5. Set n → n + 1 and go to step 3 (up to the satisfaction of an appropriate convergence criterion).

Assuming the hypotheses of the last result for an analogous non-restricted case, the sequence {un } is such
that
un → u0 , strongly in L2 (Ω),
where
u0 ∈ W01,2 (Ω)
is a solution of equation (25.36).
Remark 25.4.2 Observe that for each n, the procedure of evaluating un stands for the solution of a convex
optimization problem with unique solution, given by the one of equation
ε∇2 un − αu3n + z∗n + f = 0 in Ω,
which may be easily obtained, due to convexity, through the generalized method of lines (matrix version)
associated with Newton’s method as above described.
25.5 Conclusion
In this chapter we have develop numerical methods for large class of ODEs.
We have also introduced the matrix version of the generalized method of lines for PDEs. Further, we
have developed a convergent algorithm suitable for equations that present strong variational formulation, and
in particular, suitable for Ginzburg-Landau type equations. The results are rigorously proven and numerical
examples are provided. We emphasize that even as the parameter ε is very small, namely ε = 0.0001, the
results are consistent and the convergence is very fast.
Chapter 26
On the Numerical Solution of

First Order Ordinary
Differential Equation Systems
26.1 Introduction
In this chapter we develop an algorithm to solve a class of first order non-linear ordinary differential equa-
tions.
We start by presenting a general procedure for solving the linearized equations, and in a second step, we
apply it to solve a problem in flight mechanics in a Newton’s method context. In fact, a sequence of linear
problems is solved intending to obtain a solution for the original non-linear problem. We emphasize the
method here proposed has a performance considerably better than those so far known, particularly concerning
the computation time.
At this point we present a remark on the references.
Remark 26.1 We highlight that a similar problem is addressed in [14] for a nuclear physics model. The
main difference is that now our results are more general and applicable to a much larger class of problems.
Specifically in the present work, we apply them to a flight mechanics model found in [80].
For the numerical results we have used finite differences. Details about finite differences schemes may
be found in [73].
Finally, details on the Sobolev spaces in which the original problem is established may be found in [1].
26.2 The main results

Consider the following system of difference equations in {(uk )n }, given by
4
(uk )n+1 = ∑ (ak j )n (u j )n + (gk )n , ∀k ∈ {1, 2, 3, 4}, n ∈ {0, ..., N − 1}, (26.1)
j=1
On the Numerical Solution of First Order Ordinary Differential Equation Systems 499
where
(uk )n , (ak j )n , (gk )n ∈ R, ∀ j, k ∈ {1, 2, 3, 4}, n ∈ {0, ..., N − 1}.
Assume the following boundary conditions are intended to be satisfied:


 (u1 )0 = h0 ,
(u3 )0 = V0

(26.2)

 (u4 )0 = x0
(u1 )N = h f .

For n = 0 and k = 1 we obtain

4
(u1 )1 = ∑ (a1 j )0 (u j )0 + (g1 )0 ,
j=1
that is,
(u1 )1 − (g1 )0 − (a11 )0 (u1 )0 − (a13 )0 (u3 )0 − (a14 )0 (u4 )0
(u2 )0 = ,
(a12 )0
so that we write
(u2 )0 = m2 [0](u1 )1 + z̃2 [0], (26.3)
where
1
m2 [0] = ,
(a12 )0
and
−(g1 )0 − (a11 )0 (u1 )0 − (a13 )0 (u3 )0 − (a14 )0 (u4 )0
z̃2 [0] = .
(a12 )0
Replacing (26.3) into
4
(u2 )1 = ∑ (a2 j )0 (u j )0 + (g2 )0 ,
j=1
we get
(u2 )1 = m2 [1](u1 )1 + z2 [1],
where
m2 [1] = (a22 )0 m2 [0],
and,
z2 [1] = (a21 )0 (u1 )0 + (a22 )0 z̃2 [0] + (a23 )0 (u3 )0 + (a24 )0 (u4 )0 + (g2 )0 .
Also, replacing (26.3) into
4
(u3 )1 = ∑ (a3 j )0 (u j )0 + (g3 )0 ,
j=1
we obtain
(u3 )1 = m3 [1](u1 )1 + z3 [1],
where
m3 [1] = (a32 )0 m2 [0],
and
z3 [1] = (a31 )0 (u1 )0 + (a32 )0 z̃2 [0] + (a33 )0 (u3 )0 + (a34 )0 (u4 )0 + (g3 )0 .
Finally, replacing (26.3) into

4
(u4 )1 = ∑ (a4 j )0 (u j )0 + (g4 )0 ,
j=1
we may obtain
(u4 )1 = m4 [1](u1 )1 + z4 [1],
where
m4 [1] = (a42 )0 m2 [0],
and
z4 [1] = (a41 )0 (u1 )0 + (a42 )0 z̃2 [0] + (a43 )0 (u3 )0 + (a44 )0 (u4 )0 + (g4 )0 .
Reasoning inductively, having for k ∈ {2, 3, 4},
(uk )n−1 = mk [n − 1](u1 )n−1 + zk [n − 1], (26.4)
for n ≥ 2, replacing these last equations into (26.1) for k = 1, we may obtain
(u1 )n−1 = m̃1 [n − 1](u1 )n + z̃1 [n], (26.5)
where
m̃1 [n − 1]
= {(a11 )n−1 + (a12 )n−1 m2 [n − 1] + (a13 )n−1 m3 [n − 1]
+(a14 )n−1 m4 [n − 1]}−1 , (26.6)
and
z̃1 [n − 1]
= −m̃1 [n − 1]{(a12 )n−1 z2 [n − 1] + (a13 )n−1 z3 [n − 1]
+(a14 )n−1 z4 [n − 1] + (g1 )n−1 }. (26.7)
Finally, replacing (26.5) into (26.4), we may obtain
(uk )n−1 = m̃k [n − 1](u1 )n + z̃k [n − 1], (26.8)
∀k ∈ {2, 3, 4},
where
m̃k [n − 1] = mk [n − 1]m̃1 [n − 1],
z̃k [n − 1] = mk [n − 1]z̃1 [n − 1] + zk [n − 1].
Replacing (26.5) and (26.8) into the system (26.1), we get
(uk )n = mk [n](u1 )n + zk [n],
where
mk [n] = (ak1 )n m̃1 [n − 1] + (ak2 )n m̃2 [n − 1] + (ak3 )n m̃3 [n − 1] + (ak4 )n m̃4 [n − 1],
and
4
zk [n] = ∑ (ak j )n z̃ j [n − 1] + (gk )n .
j=1
Summarizing, we have obtained,
(u1 )n−1 = m̃1 [n − 1](u1 )n + z̃1 [n − 1],

(uk )n = mk [n](u1 )n + zk [n], ∀k ∈ {2, 3, 4}

∀n ∈ {1, ..., N}.
Therefore, having (u1 )N = h f , we may obtain
(uk )N = mk [N](u1 )N + zk [N], ∀k ∈ {2, 3, 4}
and
(u1 )N −1 = m̃N −1 (u1 )N + z̃1 [N − 1].
Having (u1 )N−1 , we may obtain
(uk )N−1 = mk [N − 1](u1 )N−1 + zk [N − 1], ∀k ∈ {2, 3, 4}
and
(u1 )N−2 = m̃1 [N − 2](u1 )N −1 + z̃1 [N − 2],
and so on, up to finding (uk )1 , ∀k ∈ {1, 2, 3, 4}, and finally,
(u2 )0 = m2 [0](u1 )1 + z̃2 [0].
At this point, the problem is solved.
26.3 Numerical results

We present numerical results for the following system of equations, which models the in plane climbing
motion of an airplane (please see more details in [80]).
h˙ = V sin γ,


 γ̇ = 1 (T sin(e3 ) + L) − g cos γ,


mfV V
1 (26.9)

 V̇ = mf (T cos(e 3 ) − D) − g sin γ

ẋ = V cos γ,

with the boundary conditions, 


 h(0) = h0 ,
V (0) = V0

(26.10)

 x(0) = x0
h(t f ) = h f ,

where t f = 90s, h is the airplane altitude, V is its speed, γ is the angle between its velocity and the
horizontal axis, and finally x denotes the horizontal coordinate position.
For numerical purposes, we assume (Air bus 320)
m f = 120, 000Kg, S f = 260m2 , a = 0.17/10 rad, g = 9.8m/s2 , ρ(h) = 1.225(1−0.0065h/288.15)4.225 Kg/m3 ,
CL = 4.95a,
CD = 0.0175 + 0.06CL2 ,
1
L = ρ(h)V 2CL S f ,
2
1
D = ρ(h)V 2CD S f ,
2
T0 = 10000
and where units refer to the International System and,
T = D + m f g sin γ + T0 ,
which refers to an accelerated motion.

To simplify the analysis, we redefine the variables as below indicated:


 h = u1 ,
γ = u2

(26.11)

 V = u3
x = u4 .

Thus, denoting u = (u1 , u2 , u3 , u4 ) ∈ U = W 1,2 ([0,t f ]; R4 ), the system above indicated may be expressed
by


 u̇1 = f1 (u)
u̇2 = f2 (u)

(26.12)

 u̇3 = f3 (u)
u̇4 = f4 (u),

where 
 f1 (u) = u3 sin(u2 ),
 f2 (u) = 1 (T (u) sin(e3 ) + L(u)) − g cos(u2 ),


m f u3 u3
(26.13)

 f3 (u) = m1f (T (u) cos(e3 ) − D(u)) − g sin(u2 )

f4 (u) = u3 cos(u2 ).

Finally,
1
L(u) = ρ0 e−u1 /B9 u23CL S f ,
2
1
D(u) = ρ0 e−u1 /B9 u23CD S f ,
2
T (u) = D(u) + m f g sin(u2 ) + T0 .

At this point we shall write the system indicated in (26.12) in finite differences, that is,


 (u1 )n+1 = (u1 )n + f1 (un )d
(u2 )n+1 = (u2 )n + f2 (un )d

(26.14)
 (u3 )n+1 = (u3 )n + f3 (un )d

(u4 )n+1 = (u4 )n + f4 (un )d,

here d = 90/N, we N refers to the number of nodes concerning the discretization in t (in our numerical
example N = 10000).
Intending to apply the Newton’s method we linearize the system indicated in (26.14) about a initial guess
ũ = (ũ1 , ũ2 , ũ3 , ũ4 ).
We obtain the following approximate system


4 ∂ f1 (ũn )


 (u 1 )n+1 = (u 1 )n + d f 1 (u˜ n ) + ∑ j=1 ∂ ũ j ((u j )n − (ũ j )n )

 (u2 )n+1 = (u2 )n + d f2 (ũn ) + ∑4 ∂ f2 (ũn ) ((u j )n − (ũ j )n )


j=1 ∂ u˜ j

∂ f3 (u˜ n )
(26.15)
4


 (u 3 )n+1 = (u 3 )n + d f 3 (ũ n ) + ∑ j=1 ∂ u˜ j ((u j )n − (ũ j )n )

 (u4 )n+1 = (u4 )n + d f4 (ũn ) + ∑4 ∂ f4 (ũn ) ((u j )n − (u˜ j )n ) .


j=1 ∂ u˜ j
Observe that such a system is in the form,

4
(uk )n+1 = (uk )n + ∑ (ak j )n (u j )n + (gk )n ,
j=1
where
∂ fk (ũn )
(ak j )n = d, for j 6= k
∂uj
∂ f j (ũn )
(a j j )n = 1 + d, for j = k,
∂uj
and
4
∂ fk (ũn )
(gk )n = fk (ũn ) d − ∑ (ũ j )n d .
j=1 ∂ u j
We solve this last system for the following boundary conditions:



 h(0) = 0 m,
V (0) = 150m/s,

(26.16)

 x(0) = 0 m,
h(t f ) = 11000 m.

We have obtained {un }. In a Newton’s method context, the next step is to replace ũn by {un } and thus to
repeat the process up to the satisfaction of an appropriate convergence criterion.
We have obtained the following solutions for h, γ,V and x. Please see Figures 26.1, 26.2, 26.3 and 26.4,
respectively.
12000
10000
8000
6000
4000
2000
0
0 10 20 30 40 50 60 70 80 90
Figure 26.1: The solution h (in m) for t f = 90s.

1.4
1.2
0.8
0.6
0.4
0.2
−0.2
0 10 20 30 40 50 60 70 80 90
Figure 26.2: The solution γ (in rad) for t f = 90s.
156
155
154
153
152
151
150
0 10 20 30 40 50 60 70 80 90
Figure 26.3: The solution V (in m/s) for t f = 90s.
26.4 The Newton’s method for another first order system

Consider the first order system and respective boundary conditions
 0

 u + f1 (u, v) + g1 = 0, in [0, 1]



v0 + f2 (u, v) + g2 = 0, in [0, 1] (26.17)




u(0) = u0 , v(1) = v f ,

Linearizing the equations about the first solutions ũ, and ṽ, we obtain,
∂ f1 (ũ, ṽ)
u0 + f1 (ũ, ṽ) + (u − ũ)
∂u
∂ f1 (ũ, ṽ)
+ (v − ṽ) + g1 = 0, (26.18)
∂v
7000
6000
5000
4000
3000
2000
1000
0
0 10 20 30 40 50 60 70 80 90
Figure 26.4: The solution x (in m) for t f = 90s.
∂ f2 (ũ, ṽ)
v0 + f2 (ũ, ṽ) + (u − ũ)
∂u
∂ f2 (ũ, ṽ)
+ (v − ṽ) + g2 = 0. (26.19)
∂v
In finite differences, we could write,
∂ f1 (ũn−1 , ṽn−1 )
un − un−1 + f1 (ũn−1 , ṽn−1 )d + (un−1 − ũn−1 )d
∂u
∂ f1 (ũn−1 , ṽn−1 )
+ (vn−1 − ṽn−1 )d + (g1 )n−1 d = 0, (26.20)
∂v
∂ f2 (ũn−1 , ṽn−1 )
vn − vn−1 + f2 (uñ−1 , ṽn−1 )d + (un−1 − ũn−1 )d
∂u
∂ f2 (ũn−1 , ṽn−1 )
+ (vn−1 − ṽn−1 )d + (g2 )n−1 d = 0. (26.21)
∂v
Hence, we may write,
un = an un−1 + bn vn−1 + cn ,
vn = dn un−1 + en vn−1 + fn ,
Where
∂ f1 (ũn−1 , ṽn−1 )
an = − d + 1,
∂u
∂ f1 (ũn−1 , ṽn−1 )
bn = − d,
∂v
∂ f1 (uñ−1 , ṽn−1 )
cn = − f1 (ũn−1 , ṽn−1 )d + ũn−1 d
∂u
∂ f1 (ũn−1 , ṽn−1 )
+ ṽn−1 d − (g1 )n−1 d, (26.22)
∂v
and
∂ f2 (ũn−1 , ṽn−1 )
dn = − d,
∂u
∂ f2 (ũn−1 , ṽn−1 )
en = − d + 1,
∂v
∂ f2 (ũn−1 , ṽn−1 )
fn = − f2 (ũn−1 , ṽn−1 )d + ũn−1 d
∂u
∂ f2 (ũn−1 , ṽn−1 )
+ ṽn−1 d − (g2 )n−1 d. (26.23)
∂v
In particular, for n = 1, we get
u1 = a1 u0 + b1 v0 + c1 , (26.24)
and
v1 = d1 u0 + e1 v0 + f1 . (26.25)
From this last equation,
v0 = (v1 − d1 u0 − f1 )/e1 ,
so that from this and equation (26.24), we get,
u1 = a1 u0 + b1 (v1 − d1 u0 − f1 )/e1 + c1 = F1 v1 + G1 ,
where
F1 = b1 /e1 , G1 = a1 u0 − b1 (d1 u0 + f1 )/e1 + c1 .
Reasoning inductively, having,
un−1 = Fn−1 vn−1 + Gn−1 ,

we also have,
un = an un−1 + bn vn−1 + cn ,
vn = dn un−1 + en vn−1 + fn ,
vn = dn (Fn−1 vn−1 + Gn−1 ) + en vn−1 + fn ,

that is,
vn−1 = Hn vn + Ln ,
where
Hn = 1/(dn Fn−1 + en ),
Ln = −Hn (dn Gn−1 + fn ).
Hence,
un = an (Fn−1 vn−1 + Gn−1 ) + bn vn−1 + cn−1 ,
so that,
un = an (Fn−1 (Hn vn + Ln ) + Gn−1 ) + bn (Hn vn + Ln ) + cn−1 ,

and hence,
Fn = an Fn−1 Hn + bn Hn ,
and
Gn = an (Fn−1 Ln + Gn−1 ) + bn Ln + cn−1 .
Thus,
un = Fn vn + Gn ,
so that, in particular,
uN = FN v f + GN ,
vN−1 = HN v f + LN ,
and hence,
uN−1 = FN−1 vN−1 + GN−1 ,
vN−2 = HN−1 vN −1 + LN−1 ,
and so on, up to finding,
u1 = F1 v1 + G1 ,
and
v0 = H0 v1 + L0 ,
where H0 = 1/e1 and L0 = −(d1 u0 + f1 )/e1 .
The next step is to replace {ũn } and {ṽn } by {un } and {vn } respectively and then to repeat the process up
the satisfaction of an appropriate convergence criterion.
26.4.1 An example in nuclear physics

As an application of the method above exposed we develop numerical results for the system of equations
relating to the neutron kinetics of a nuclear reactor. Following [71], the system in question is given by:

(ρ (T )−β )
 n0 (t ) =

L n(t) + λC(t)
0 β
C (t) = L n(t) − λC(t) (26.26)
 T 0 (t) = Hn(t),

where n(t) is the neutron population, C(t) is the concentration of delayed neutrons, T (t) is the core tempera-
ture, ρ(T ) is the reactivity (which depends on the temperature T ), β is the delayed neutron fraction, L is the
prompt reactors generation time, λ is the average decay constant of the precursors and H is the inverse of the
reactor thermal capacity.
For our numerical examples we consider T (0s) = 300K and T (100s) = T f = 350K. Moreover we assume
the relation,
1 (β − ρ (0))
C(0) = n(0),
λ L
where n(0) is unknown (to be numerically calculated by our method such that we have T (100s) = T f ).
Also we consider
ρ(T ) = ρ(0) − α(T − T (0)).
The remaining values are: β = 0.0065, L = 0.0001s, λ = 0.00741s−1 , H = 0.05K/(MW s), α = 5 ·
10−5 K −1 , and ρ(0) = 0.2β .
First we linearize the system in question about (ñ, T̃ ) obtaining (in fact its a first approximation)
ρ(T̃ ) − β ρ(T ) − β
n0 (t) = n(t) + ñ(t)
L L
ρ(T̃ ) − β
− ñ(t) + λC(t), (26.27)
L
β
C0 (t) = n(t) − λC(t),
L
T 0 (t) = Hn(t),
where ρ(T ) = ρ(0) − α(T − T (0)).
Discretizing such a system in finite differences, we get
ρ(T̃i ) − β ρ(Ti ) − β
(ni+1 − ni )/d = ni + ñi
L L
ρ(T̃i ) − β
− ñi + λCi , (26.28)
L
β
(Ci+1 −Ci )/d = ni − λCi ,
L
(Ti+1 − Ti )/d = Hni ,
where d = 100s/N, where N is the number of nodes.
Hence, we may write
ni+1 = ai ni + bi Ti + diCi + ei , (26.29)
Ci+1 = f ni + gCi , (26.30)
Ti+1 = hTi + mni , (26.31)
where
ρ(T̃i ) − β
ai = 1 + d,
L
−α
bi = ñi d,
L
di = λ d,
(ρ(0) + αT (0) − β ) (ρ(T̃i ) − β )
ei = ñi d − ñi d,
L L
β
f = d,
L
g = 1 − λ d,
h = 1,
m = Hd.
Observe that
C0 = α̃n0 ,
where
β − ρ(0)
α̃ = .
Lλ
For i = 1 from (26.31) we obtain
T1 − hT0
n0 = = α1 T1 + β1 , (26.32)
m
where α1 = 1/m and β1 = −(h/m)T0 .
Therefore,
C0 = α̃n0 = α̃(α1 T1 + β1 ).
Still for i = 1, replacing this last relation and (26.32) into (26.29), we get
n1 = a1 (α1 T1 + β1 ) + b1 T0 + d1 α̃(α1 T1 + β1 ) + e1 ,
so that
n1 = α̃1 T1 + β̃1 , (26.33)
where
α̃1 = a1 α1 + d1 α̃α1 ,
and
β̃1 = a1 β1 + b1 T0 + d1 α̃β1 + e1 .
Finally, from (26.30),
C1 = f (α1 T1 + β1 ) + gα̃(α1 T1 + β1 )
= α̂1 T1 + β̂1 , (26.34)
where
α̂1 = f α1 + gα̃α1 ,
and
β̂1 = f β1 + gα̃β1 .
ni = α̃i Ti + β̃i , (26.35)
ni−1 = αi Ti + βi , (26.36)
Ci = α̂i Ti + β̂i , (26.37)
we are going to obtain the corresponding relations for i + 1, i ≥ 1. From (26.31) and (26.35) we obtain
Ti+1 = hTi + m(α̃i Ti + β˜i ),
so that
Ti = ηi Ti+1 + ξi , (26.38)
where
ηi = (h + mα̃i )−1 ,
and
ξi = −(mβ̃i )ηi .
On the other hand, from (26.29), (26.35) and (26.37) we have
ni+1 = ai (α̃i Ti + B̃i ) + bi Ti + di (α̂i Ti + β̂i ) + ei ,
so that from this and (26.38), we obtain
ni+1 = α̃i Ti+1 + β˜i+1 ,

where
α̃i+1 = ai α̃i ηi + bi ηi + di α̂i ηi ,
and
β̃i+1 = ai (α̃i ξi + β̃i ) + bi ξi + di (α̂i ξi + β̂i ) + ei .
Also from (26.35) and (26.38) we have,

ni = α̃i (ηi Ti+1 + ξi ) + β̃i = αi+1 Ti+1 + βi+1 ,
where
αi+1 = α̃i ηi ,
and
βi+1 = α̃i ξi + β̃i .
Moreover,
Ci+1 = f ni + gCi
= f (αi+1 Ti+1 + βi+1 )
+g(α̂i Ti + β̂i )
= f (αi+1 Ti+1 + βi+1 )
+g(α̂i (ηi Ti + ξi ) + β̂i )
= α̂i+1 Ti+1 + βî+1 , (26.39)
where
α̂i+1 = f αi+1 + gα̂i ηi ,
and
β̂i = f βi+1 + gα̂i ξi + gβ̂i .
Summarizing, we have obtained linear functions (F0 )i , (F1 )i and (F2 )i such that
Ti = (F0 )i (Ti+1 ),
ni = (F1 )i (Ti+1 ),
Ci = (F2 )i (Ti+1 ),
∀i ∈ {1, ..., N − 1}.
Thus, considering the known value TN = T f we obtain
TN−1 = (F0 )N −1 (T f ),
nN−1 = (F1 )N−1 (T f ),
CN −1 = (F2 )N −1 (T f ),
and having TN−1 we get,
TN −2 = (F0 )N−2 (TN−1 ),
nN −2 = (F1 )N−2 (TN−1 ),
CN−2 = (F2 )N−1 (TN−1 ),
and so on, up to finding
T1 = (F0 )1 (T2 ),
n1 = (F1 )1 (T2 ),
C1 = (F2 )1 (T2 ),
and n0 = (F0 )1 (T1 ).
The next step is to replace (ñ, T̃ ) by the last calculated (n, T ) and then to repeat the process until an
appropriate convergence criterion is satisfied.
Concerning our numerical results through such a method, for the solution n(t) obtained, please see
Figure 26.5. For the solution T (t), see Figure 26.6.
We emphasize the numerical results here obtained are consistent with the current literature (see [71], for
details).
12
11
10
6
0 20 40 60 80 100
Figure 26.5: Solution n(t) for 0s ≤ t ≤ 100s.
350
345
340
335
330
325
320
315
310
305
300
0 20 40 60 80 100
Figure 26.6: Solution T (t) for 0s ≤ t ≤ 100s.
26.5 Conclusion
In this chapter, we have developed a method for solving a class of first order ordinary differential equations.
The results are applied to a flight mechanics problem which models the in plane climbing of an airplane. It
is worth mentioning the algorithm obtained is of easy implementation and very efficient from a computational
point of view.
Finally, we would highlight the numerical results obtained are perfectly consistent with the physical
problem context. In future works we intend to apply the method to solve relating optimal control problems.
Chapter 27
On the Generalized Method of

Lines and its Proximal Explicit
and Hyper-Finite Difference
Approaches
27.1 Introduction
This chapter develops two improvements relating to the generalized method of lines. In our previous publi-
cations [19, 18], we highlight the method there addressed may present a relevant error as a parameter ε > 0
is too small, that is, as ε is about 0.01, 0.001 or even smaller.
In the present section we develop a solution for such a problem through a proximal formulation suitable
for a large class of non-linear elliptic PDEs.
At this point we reintroduce the generalized method of lines, originally presented in F. Botelho [19].
In the present context we add new theoretical and applied results to the original presentation. Specially the
computations are all completely new. Consider first the equation
ε∇2 u + g(u) + f = 0, in Ω ⊂ R2 , (27.1)
with the boundary conditions

u = 0 on Γ0 and u = u f , on Γ1 .
From now on we assume that u f , g and f are smooth functions (we mean C∞ functions), unless otherwise
specified. Here Γ0 denotes the internal boundary of Ω and Γ1 the external one. Consider the simpler case
where
Γ1 = 2Γ0 ,
and suppose there exists r(θ ), a smooth function such that
Γ0 = {(θ , r(θ )) | 0 ≤ θ ≤ 2π},
being r(0) = r(2π).

On the Generalized Method of Lines and its Proximal Explicit and Hyper-Finite Difference Approaches 513
In polar coordinates the above equation may be written as

∂ 2u 1 ∂ u 1 ∂ 2u
+ + + g(u) + f = 0, in Ω, (27.2)
∂ r2 r ∂ r r2 ∂ θ 2
and
u = 0 on Γ0 and u = u f , on Γ1 .
Define the variable t by
r
t= .
r(θ )
Also defining ū by
u(r, θ ) = ū(t, θ ),
dropping the bar in ū, equation (27.1) is equivalent to
∂ 2u 1 ∂u
+ f2 (θ )
∂t 2 t ∂t
1 ∂ 2u f4 (θ ) ∂ 2 u
+ f3 (θ ) + 2
t ∂ θ ∂t t ∂θ2
+ f5 (θ )(g(u) + f ) = 0, (27.3)
in Ω. Here f2 (θ ), f3 (θ ), f4 (θ ) and f5 (θ ) are known functions.
More specifically, denoting
−r0 (θ )
f1 (θ ) = ,
r(θ )
we have:
f10 (θ )
f2 (θ ) = 1 + ,
1 + f1 (θ )2
2 f1 (θ )
f3 (θ ) = ,
1 + f1 (θ )2
and
1
f4 (θ ) = .
1 + f1 (θ )2
Observe that t ∈ [1, 2] in Ω. Discretizing in t (N equal pieces which will generate N lines) we obtain the
equation
un+1 − 2un + un−1 (un − un−1 ) 1
+ f2 (θ )
d2 d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ )
+ f3 (θ ) +
∂θ tn d ∂ θ 2 tn2

1 1
+ f5 (θ ) g(un ) + fn = 0, (27.4)
ε ε
∀n ∈ {1, ..., N − 1}. Here, un (θ ) corresponds to the solution on the line n. Thus we may write
un = Tn (un−1 , un , un+1 ),
where

(un − un−1 ) 1
Tn (un−1 , un , un+1 ) = un+1 + un + un−1 + f2 (θ )d 2
d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ ) 2
+ f3 (θ )d 2 + d
d2 d2

+ f5 (θ ) g(un ) + fn /3.0. (27.5)
ε ε
27.1.1 Some preliminaries results and the main algorithm

Now we recall a classical definition.
Definition 27.1.1 Let C be a subset of a Banach space U and let T : C → C be an operator. Thus T is said
to be a contraction mapping if there exists 0 ≤ α < 1 such that
kT (x1 ) − T (x2 )kU ≤ αkx1 − x2 kU , ∀x1 , x2 ∈ C.
Remark 27.1.2 Observe that if kT 0 (x)kU ≤ α < 1, on a convex set C then T is a contraction mapping, since
by the mean value inequality,
kT (x1 ) − T (x2 )kU ≤ sup{kT 0 (x)k}kx1 − x2 kU , ∀x1 , x2 ∈ C.

x∈C
The next result is the base of our generalized method of lines. For a proof see Theorem 15.7.3.
Theorem 27.1.3 (Contraction mapping theorem) Let C be a closed subset of a Banach space U. Assume
T is contraction mapping on C, then there exists a unique x̃ ∈ C such that x̃ = T (x̃). Moreover, for an arbitrary
x0 ∈ C defining the sequence
x1 = T (x0 ) and xk+1 = T (xk ), ∀k ∈ N
we have
xk → x̃, in norm, as k → +∞.
To obtain a fixed point for each Tn indicated in (27.5) is perfectly possible if ε ≈ O(1). However, if ε > 0
is small, the error in this process may be relevant.
To solve this problem, firstly we propose the following algorithm,
1. Choose K ≈ 30 − 80 and set u0 = 0.

2. Calculate u = {un } by solving the equation
(un − un−1 ) 1
un+1 − 2un + un−1 + f2 (θ )d 2
d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ ) 2
+ f3 (θ )d 2 + d
d2 d2

+ f5 (θ ) g(un ) + fn
ε ε
d2
−K(un − (u0 )n )
ε
= 0. (27.6)
Such an equation is solved through the Banach fixed point theorem, that is, defining

(un − un−1 ) 1
Tn (un , un+1 , un−1 ) = un+1 + un + un−1 + f2 (θ )d 2
d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ ) 2
+ f3 (θ )d 2 + d
d2 d2

+ f5 (θ ) g(un ) + fn
ε ε
2 d2

d
+K(u0 )n / 3+K (27.7)
ε ε
equation (27.7) stands for

un = Tn (un−1 , un , un+1 ),
so that for n = 1 we have
u1 = T1 (0, u1 , u2 ).
We may use the Contraction Mapping theorem to calculate u1 as a function of u2 . The procedure
would be,
(a) set x1 = u2 ,
(b) obtain recursively
xk+1 = T1 (0, xk , u2 ),
(c) and finally get
u1 = lim xk = g1 (u2 ).
k→∞
Thus, we have obtained

u1 = g1 (u2 ).
We can repeat the process for n = 2, that is, we can solve the equation
u2 = T2 (u1 , u2 , u3 ),
which from above stands for

u2 = T2 (g1 (u2 ), u2 , u3 ).
The procedure would be:
(a) Set x1 = u3 ,
(b) calculate
xk+1 = T2 (g1 (xk ), xk , u3 ),
(c) obtain
u2 = lim xk = g2 (u3 ).
k→∞
We proceed in this fashion until obtaining
uN −1 = gN−1 (uN ) = gN−1 (u f ).
Being u f known we have obtained uN−1 . We may then calculate
uN−2 = gN −2 (uN−1 ),
uN−3 = gN−3 (uN−2 ),

u1 = g1 (u2 ).
Thus, this part of the problem is solved.
3. Set u0 = u and go to item 2 up to the satisfaction of an appropriate convergence criterion.

Remark 27.1.4 Here we consider some points concerning the convergence of the method.
In the next lines the norm indicated refers to the infinity one for C([0, 2π]; RN −1 ). In particular for n = 1
from above we have:
u1 = T1 (0, u1 , u2 ),
that is,
d2
2
d
u2 − 2u1 − K u1 + O K = 0.
ε ε
Hence, denoting
d2

a[1] = 1/ 2 + K
ε
and
d2

a[n] = 1/ 2 + K − a[n − 1] , ∀n ∈ {2, . . . , N − 1},
ε
for N sufficiently big we may obtain
2
d
ku1 − a[1]u2 k = O K ,
ε
and by induction 2
d
kun − a[n]un+1 k = nO K ,
ε
so that we would have
d
kun − a[n]un+1 k ≤ O K ), ∀n ∈ {1, . . . , N − 1}
ε
This last calculation is just to clarify that the procedure of obtaining the relation between consecutive lines
through the contraction mapping theorem is well defined.
27.1.2 A numerical example, the proximal explicit approach

In this section we present a numerical example. Consider the equation
ε∇2 u + g(u) + 1 = 0, in Ω ⊂ R2 , (27.8)
where, for a Ginzburg-Landau type equation (see [4, 55] for the corresponding models in physics),
g(u) = −u3 + u,

u = u1 on ∂ Ω = Γ0 ∪ Γ1 ,
where Ω = {(r, θ ) : 1 ≤ r ≤ 2, 0 ≤ θ ≤ 2π},
u1 = 0, on Γ0 = {(1, θ ) : 0 ≤ θ ≤ 2π},
u1 = u f (θ ), on Γ1 = {(2, θ ) : 0 ≤ θ ≤ 2π}.
Through the generalized method of lines, for N = 10 (10 lines), d = 1/N in polar coordinates and finite
differences (please see [73] for general schemes in finite differences), equation (27.8), stands for
1 1 ∂ 2 un 2 3 d2 d2
(un+1 − 2un + un−1 ) + (un − un−1 )d + 2 d + (−un + un ) + = 0,
rn rn ∂ θ 2 ε ε
∀n ∈ {1, . . . , N − 1}.
At this point we present, through the generalized method of lines, the concerning algorithm which may
be for the softwares maple or mathematica.
In this software, x stands for θ .
m8 = 10; (number of lines)
d = 1.0/m8 ; (thickness of the grid)
e1 = 0.01; (ε = e1 )
K = 70.0;
Clear[d1 , u, a, b, h];
For[i = 1, i < m8 , i + +,
z1 [i] = 0.0]; ( vector which stores Ku0 (i))
For[k1 = 1, k1 < 180, k1 + +,
Print[k1 ];
a = 0.0;
For[i = 1, i < m8 , i + +,
Print[i];
t = 1.0 + i ∗ d;
b[x− ] = u[i + 1][x];
b12 = 2.0;
A18 = 5.0;
k = 1;
W hile[ b12 > 10−4 ,
k = k + 1;
z = (u[i + 1][x] + b[x] + a + 1/t ∗ (b[x] − a) ∗ d ∗ d12 + 1/t 2 ∗ D[b[x], {x, 2}] ∗ d 2 ∗ d12
−b[x]3 ∗ d 2 ∗ d12 /e1 + b[x] ∗ d 2 ∗ d12 /e1 + 1.0 ∗ d 2 /e1 +
z1 [i] ∗ d 2 /e1 )/(3.0 + K ∗ d 2 /e1 );
z = Series[z, {d1 , 0, 2}, {u f [x], 0, 3}, {u0f [x], 0, 1}, {u00f [x], 0, 1}, {u000f [x], 0, 0}, {u0000
f [x], 0, 0}];
z = Normal[z];
z = Expand[z];
b[x− ] = z;
u[i + 1][x− ] = 0.0;
u f [x− ] = 0.0;
d1 = 1.0;
A19 = z;
b12 = Abs[A19 − A18 ];
A18 = A19 ;
Clear[u, u f , d1 ]];
a1 = b[x];
Clear[b];
u[i + 1][x− ] = b[x];
h[i] = a1 ;
a = a1 ];
b[x− ] = u f [x];
d1 = 1.0;
(27.9)
For[i = 1, i < m8 , i + +,
W1 [m8 − i] =
Series[h[m8 − i], {u f [x], 0, 3}, {u0f [x], 0, 1}, {u00f [x], 0, 1}, {u000f [x], 0, 0}, {u0000
f [x], 0, 0}];
W = Normal[W1 [m8 − i]];
b[x− ] = Expand[W ];
v[m8 − i] = Expand[W ]];
For[i = 1, i < m8 , i + +,
z1 [i] = K ∗ v[i]];
d1 = 1.0;
Print[Expand[v[m8 /2]]];
Clear[d1 , u, b]]
(27.10)
At this point we present the expressions for 10 lines, firstly for ε = 1 and K = 0. In the next lines x stands
for θ .
For each line u[n] we have obtained,
u[1] = 0.0588308 + 0.167434u f [x] − 0.00488338u f [x]2 − 0.00968371u f [x]3 + 0.0126122u00f [x]
−0.000641358u f [x] u00f [x] − 0.00190065u f [x]2 u00f [x] + 0.0000660286u f [x]3 u00f [x]
u[2] = 0.101495 + 0.316995u f [x] − 0.00919963u f [x]2 − 0.0182924u f [x]3 + 0.0225921u00f [x]
−0.00113308u f [x]u00f [x] − 0.00336691u f [x]2 u00f [x] + 0.000122652u f [x]3 u00f [x]
u[3] = 0.1295 + 0.450424u f [x] − 0.0127925u f [x]2 − 0.0257175u f [x]3 + 0.0294791u00f [x]
u[4] = 0.143991 + 0.568538u f [x] − 0.0153256u f [x]2 − 0.0315703u f [x]3 + 0.0331249u00f [x]
u[5] = 0.146024 + 0.672307u f [x] − 0.0164357u f [x]2 − 0.0352883u f [x]3 + 0.0336371u00f [x]
u[6] = 0.136541 + 0.762571u f [x] − 0.0158578u f [x]2 − 0.0361974u f [x]3 + 0.0312624u00f [x]
u[7] = 0.116389 + 0.840008u f [x] − 0.0135378u f [x]2 − 0.0336098u f [x]3 + 0.0263271u00f [x]
u[8] = 0.0864095 + 0.905032u f [x] − 0.00970893u f [x]2 − 0.0269167u f [x]3 + 0.0192033u00f [x]
−0.000206117u f [x] u00f [x] − 0.000856701u f [x]2 u00f [x] + 5.78758 ∗ 10−6 u f [x]3 u00f [x]
u[9] = 0.0473499 + 0.958203u f [x] − 0.00491745u f [x]2 − 0.0157466u f [x]3 + 0.0102907u00f [x].
In the next lines we present the results relating to the software indicated, with ε = 0.01 and K = 70.
For each line u[n] we have obtained,
u[1] = 1.08673 + 4.6508 ∗ 10−7 u f [x] − 1.11484 ∗ 10−7 u f [x]2 + 3.25552 ∗ 10−8 u f [x]3
+5.13195 ∗ 10−9 u00f [x] − 2.24741 ∗ 10−9 u f [x] u00f [x]
+8.97477 ∗ 10−10 u f [x]2 u00f [x] − 8.86957 ∗ 10−11 u f [x]3 u00f [x]
u[2] = 1.27736 + 1.51118 ∗ 10−6 u f [x] − 4.39811 ∗ 10−7 u f [x]2 + 1.55883 ∗ 10−7 u f [x]3
+1.33683 ∗ 10−8 u00f [x] − 7.50593 ∗ 10−9 u f [x] u00f [x]
+3.77548 ∗ 10−9 u f [x]2 u00f [x] − 4.9729 ∗ 10−10 u f [x]3 u00f [x]
u[3] = 1.30559 + 6.91602 ∗ 10−6 u f [x] − 2.21813 ∗ 10−6 u f [x]2 + 8.89891 ∗ 10−7 u f [x]3
+4.85851 ∗ 10−8 u00f [x] − 3.22542 ∗ 10−8 u f [x] u00f [x]
+1.90602 ∗ 10−8 u f [x]2 u00f [x] − 3.20439 ∗ 10−9 u f [x]3 u00f [x]
u[4] = 1.30968 + 0.0000354152u f [x] − 0.0000116497u f [x]2 + 5.06682 ∗ 10−6 u f [x]3

+1.93523 ∗ 10−7 u00f [x] − 1.42948 ∗ 10− 7u f [x] u00f [x]
+9.56419 ∗ 10−8 u f [x]2 u00f [x] − 2.02133 ∗ 10−8 u f [x]3 u00f [x]
u[5] = 1.31014 + 0.000193272u f [x] − 0.0000616231u f [x]2 + 0.0000278964u f [x]3

+7.96293 ∗ 10−7 u00f [x] − 6.22727 ∗ 10−7 u f [x] u00f [x]
+4.58264 ∗ 10−7 u f [x]2 u00f [x] − 1.21033 ∗ 10−7 u f [x]3 u00f [x]
u[6] = 1.30935 + 0.0011121u f [x] − 0.000331366u f [x]2 + 0.000149489u f [x]3

+3.34907 ∗ 10−6 u00f [x] − 2.66647 ∗ 10− 6u f [x]u00f [x]
+2.09295 ∗ 10−6 u f [x]2 u00f [x] − 6.87318 ∗ 10−7 u f [x]3 u00f [x]
u[7] = 1.30383 + 0.00665186u f [x] − 0.00182893u f [x]2 + 0.000789763u f [x]3

+0.0000140899u00f [x] − 0.000011257u f [x] u00f [x]
+9.14567 ∗ 10−6 u f [x]2 u00f [x] − 3.70794 ∗ 10−6 u f [x]3 u00f [x]
u[8] = 1.26934 + 0.040481u f [x] − 0.0101978u f [x]2 + 0.00408206u f [x]3

+0.00005612u00f [x] − 0.0000457826u f [x]
+u00f [x] + 0.000037641u f [x]2 u00f [x] − 0.000018566u f [x]3 u00f [x]
u[9] = 1.05971 + 0.23709u f [x] − 0.0493175u f [x]2 + 0.0173591u f [x]3

+0.000176522u00f [x] − 0.000150827u f [x]u00f [x]
+0.000123238u f [x]2 u00f [x] − 0.0000709253u f [x]3 u00f [x].
Remark 27.1 Observe that since ε = 0.01 the solution is close to the constant value 1.3247 along the
domain, which is an approximate solution of equation −u3 + u + 1.0 = 0.
27.2 The hyper-finite differences approach

In the last sections we have introduced a method to minimize the solution error for a large class of PDEs
as a typical parameter ε > 0 is small. The main idea presented there consisted of a proximal formulation
combined with the generalized method of lines.
In the present section we develop another new solution for such a problem also for a large class of non-
linear elliptic PDEs, namely, the hyper-finite differences approach. We believe the result here developed
is better than those of the previous sections. Indeed in the present method, in general the convergence is
obtained more easily.
The idea here is to divide the domain interval, concerning variable t to be specified, into sub-domains
in order to minimize the effect of the small parameter ε > 0. For example if we divide the domain [1, 2]
into 10 − 12 sub-domains, for ε = 0.01 and the example addressed in the next pages, the error applying the
generalized method of lines on which the sub-domain is very small. Finally, we reconnect the sub-domains
by solving the system of equations corresponding to the partial differential equation in question on each of
these N1 − 1 nodes. As N1 is a small number for the amount of nodes (typically N1 = 10 − 12), we have
justified the terminology hyper-finite differences.
First observe that, in equation (27.3), t ∈ [1, 2] in Ω. Discretizing in t (N5 equal pieces which will generate
N5 lines ), we recall that such a general equation (27.3),
∂ 2u 1 ∂u
+ f2 (θ )
∂t 2 t ∂t
1 ∂ 2u f4 (θ ) ∂ 2 u
+ f3 (θ ) + 2
t ∂ θ ∂t t ∂θ2
+ f5 (θ )(g(u) + f ) = 0, (27.11)
partially in finite differences, has the expression

un+1 − 2un + un−1 (un − un−1 ) 1
+ f2 (θ )
d2 d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ )
+ f3 (θ ) +

1 1
+ f5 (θ ) g(un ) + fn = 0, (27.12)
ε ε
∀n ∈ {1, ..., N5 − 1}. Here, un (θ ) corresponds to the solution on the line n. Thus we may write
un = Tn (un−1 , un , un+1 ),
where

(un − un−1 ) 1
Tn (un−1 , un , un+1 ) = un+1 + un + un−1 + f2 (θ )d 2
d tn
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ ) 2
+ f3 (θ )d 2 + d
d2 d2

+ f5 (θ ) g(un ) + fn /3.0. (27.13)
ε ε
27.2.1 The main algorithm

To obtain a fixed point for each Tn indicated in (27.13) is perfectly possible if ε ≈ O(1). However, for the
case in which ε > 0 is small, we highlight once more the error may be relevant, so that we propose the
following algorithm to deal with such a situation of small ε > 0.
1. Choose N = 30 − 100 and N1 = 10 − 12 (specifically for the example in the next lines). Divide the
interval domain in the variable t into N1 equal pieces (for example, for the interval [1, 2] through
a concerning partition {t0 = 1,t1 , ...,tN1 = 2}, where tk = 1 + k/N1 and d = 1/(N N1 ) is the grid
thickness in t).
2. Through the generalized method of lines, solve the equation in question on the interval [tk ,tk+1 ] as
function of u(tk ) and u(tk+1 ) and the domain shape.
To calculate uk = {ukn } on [tk ,tk+1 ], proceed as follows. First observe that the equation in question
stands for
(ukn − ukn−1 ) 1
ukn+1 − 2ukn + ukn−1 + f2 (θ )d 2
d tnk
∂ (ukn − unk −1 ) 1 ∂ 2 ukn f4 (θ ) 2
+ f3 (θ )d 2 + d
∂θ k
tn d ∂ θ 2 (tnk )2
d2 d2

+ f5 (θ ) g(ukn ) + fnk
ε ε
= 0, (27.14)
where
tnk = 1 + (k − 1)/N1 + nd, ∀k ∈ {1, . . . , N1 }, n ∈ {1, . . . N − 1}.
Such an equation is solved through the Banach fixed point theorem, that is, defining
(ukn − ukn−1 ) 1
Tnk (ukn , ukn+1 , ukn−1 ) = ukn+1 + ukn + ukn−1 + f2 (θ )d 2
d tnk
∂ (ukn − ukn−1 ) 1 ∂ 2 ukn f4 (θ ) 2
+ f3 (θ )d 2 + d
d2 d2

+ f5 (θ ) g(ukn ) + fnk /3 (27.15)
ε ε
equation (27.14) stands for
ukn = Tnk (ukn−1 , ukn , un+1
k
),
so that for n = 1 we have
uk1 = T1 (u(tk ), uk1 , u2k ).
We may use the Contraction Mapping Theorem to calculate uk1 as a function of uk2 and u(tk ). The
procedure would be,
(a) set x1 = uk2 ,
(b) obtain recursively
x j+1 = T1k (u(tk ), x j , uk2 ),
(c) and finally get
uk1 = lim x j = g1 (u(tk ), u2 ).
j→∞
Thus, we have obtained

uk1 = g1 (u(tk ), uk2 ).
We can repeat the process for n = 2, that is, we can solve the equation
uk2 = T2k (uk1 , uk2 , uk3 ),

which from above stands for

uk2 = T2k (g1 (u(tk ), u2k ), uk2 , uk3 ).
The procedure would be:
(a) Set x1 = uk3 ,
(b) calculate
x j+1 = T2k (g1 (u(tk ), x j ), x j , u3k ),
(c) obtain
uk2 = lim x j = g2 (u(tk ), u3k ).
j→∞
We proceed in this fashion until obtaining
ukN −1 = gN −1 (u(tk ), ukN ) = gN −1 (u(tk ), u(tk+1 )).
We have obtained ukN−1 . We may then calculate
ukN −2 = gN−2 (u(tk ), ukN −1 )),
ukN−3 = gN−3 (u(tk ), ukN−2 )),

uk1 = g1 (u(tk ), uk2 ).
Thus this part of the problem is solved.
3. Calculate the solution on the lines corresponding to ut1 , . . . , utN1 −1 , by solving the system,
u(tk ) = T̃k (uk−1 (N − 1), u(tk ), uk+1

1 ),
which correspond to the partial differential equation in question on the line k N, where

(un − un−1 ) 1
T̃k (un , un+1 , un−1 ) = un+1 + un + un−1 + f2 (θ )d 2
d tk
∂ (un − un−1 ) 1 ∂ 2 un f4 (θ ) 2
+ f3 (θ )d 2 + d
∂θ tk d ∂ θ 2 tk2
d2 d2

+ f5 (θ ) g(un ) + fn /3. (27.16)
ε ε
Here may use the Banach fixed point theorem for the final calculation as well.
The problem is then solved.
27.2.2 A numerical example

In this section we present a numerical example. Consider the equation
ε∇2 u + g(u) + 1 = 0, in Ω ⊂ R2 , (27.17)
where, for a Ginburg-Landau type equation
g(u) = −u3 + u,

u = u1 on ∂ Ω = Γ0 ∪ Γ1 ,
where Ω = {(r, θ ) : 1 ≤ r ≤ 2, 0 ≤ θ ≤ 2π},
u1 = 0, on Γ0 = {(1, θ ) : 0 ≤ θ ≤ 2π},
u1 = u f (θ ), on Γ1 = {(2, θ ) : 0 ≤ θ ≤ 2π}.
Through the generalized method of lines, for N = 30 (30 lines in which sub-domain), N1 = 10 (10 sub-
domains) and d = 1/(N N1 ), in polar coordinates and finite differences (please see [73] for general schemes
in finite differences), equation (27.17) stands for
1 1 ∂ 2 un 2 d2 d2
(un+1 − 2un + un−1 ) + (un − un−1 )d + 2 2
d + (−u3n + un ) + = 0,
rn rn ∂ θ ε ε
∀n ∈ {1, . . . , N − 1}.
At this point we present, through the generalized method of lines, the concerning algorithm which may
be for the softwares mathematica or maple.
In this software, x stands for θ .
ClearAll;
m8 = 30; (number of lines for each sub-domain)
N1 = 10; (number of sub-domains)
d = 1.0/m8 /N1 ; (grid thickness)
e1 = 0.01; (ε = 0.01)
Clear[d1 , u, a, b, h,U];
For[k1 = 1, k1 < N1 + 1, k1 + +,
Print[k1 ];
a = U[k1 − 1][x];
For[i = 1, i < m8 , i + +,
t = 1.0 + (k1 − 1)/N1 + i ∗ d;
Print[i];
b[x− ] = u[i + 1][x];
For[k = 1, k < 35, k + +, (here we have fixed the number of iterations for this example)
z = (u[i + 1][x] + b[x] + a + 1/t ∗ (b[x] − a) ∗ d ∗ d12 + 1/t 2 ∗ D[b[x], {x, 2}] ∗ d 2 ∗ d12
+(−b[x]3 ∗ d 2 ∗ d12 /e1 + b[x] ∗ d 2 /e1 ∗ d12 ) + 1.0 ∗ d 2 /e1 ∗ d12 )/(3.0);
z = Series[z, {d1 , 0, 2}];
z = Normal[z];
z = Expand[z];
b[x− ] = z];
a1 = b[x];
Clear[b];
u[i + 1][x− ] = b[x];
h[k1 , i] = Expand[a1 ];
Clear[d1 ];
a = a1 ]; b[x− ] = U[k1 ][x];
For[i = 1, i < m8, i + +,
W1 [k1 , m8 − i] = Series[h[k1 , m8 − i], {d1 , 0, 2}];
W [k1 , m8 − i] = Normal[W1 [k1 , m8 − i]];
b[x− ] = Expand[W [k1 , m8 − i]];

v[m8 − i] = Expand[W [k1 , m8 − i]]];
d1 = 1.0;
Print[v[m8/2]];
Clear[d1 , b, u]];
Clear[U];
U[0][x− ] = 0.0;
U[N1 ][x− ] = U f [x];
Clear[d1 ];
d1 = 1.0;
For[k = 1, k < N1 , k + +,
U[k][x− ] = 0.0;
z7 [k] = 0.0];
For[k1 = 1, k1 < 385, k1 + +, (here we have fixed the number of iterations)
Print[k1 ];
For[k = 1, k < N1 , k + +,
t = 1 + (k)/N1 ;
z5 [k] = (W [k, m8 − 1] +U[k][x] +W [k + 1, 1] +
1/t ∗ (U[k][x] −W [k, m8 − 1]) ∗ d ∗ d12 +
1/t 2 ∗ D[U [k][x], {x, 2}] ∗ d 2 ∗ d12
+(−U[k][x]3 +U[k][x]) ∗ d 2 /e1 ∗ d12 + d 2 /e1)/3.0];
For[k = 1, k < N1 , k + +,
z5 [k] = Series[ z5 [k], {U f [x], 0, 3}, {Uo [x], 0, 3}, {U f0 [x], 0, 1}, {Uo0 [x], 0, 1},
{U f00 [x], 0, 1}, {Uo00 [x], 0, 1}, {U f000 [x], 0, 0}, {Uo000 [x], 0, 0}, {U f0000 [x], 0, 0}, {Uo0000 [x], 0, 0}];
z5 [k] = Normal[z5 [k]];
Clear[A7 ];
A7 = Expand[z5 [k]];
U[k][x− ] = A7 ]]; Print[U[N1 /2][x]];
At this point we present the expressions for N1 = 10 and ε = 0.01. In the next lines x stands for θ (0 ≤ θ ≤
2π).
For each line u[tk ] = U[k] we have obtained:
U [1] = 1.11698 + 3.02073 ∗ 10−9U f [x] − 8.9816 ∗ 10−10U f [x]2 + 1.85216 ∗ 10−10U f [x]3
+8.284 ∗ 10−11U f00 [x] − 4.75816 ∗ 10−11U f [x] U f00 [x]
+1.35882 ∗ 10−11U f [x]2 U f00 [x] − 1.13972 ∗ 10−12U f [x]3 U f00 [x]
U[2] = 1.3107 + 7.41837 ∗ 10−9U f [x] − 3.8971 ∗ 10−9U f [x]2 + 1.58347 ∗ 10−9U f [x]3
+1.71087 ∗ 10−10 U f00 [x] − 1.71274 ∗ 10−10U f [x] U f00 [x]
+1.01077 ∗ 10−10U f [x]2 U f00 [x] − 8.08804 ∗ 10−12U f [x]3 U f00 [x]
U[3] = 1.32397 + 6.40836 ∗ 10−8U f [x] − 4.12903 ∗ 10−8U f [x]2 + 1.91826 ∗ 10−8U f [x]3
+1.13579 ∗ 10−9U f00 [x] − 1.42494 ∗ 10−9U f [x] U f00 [x]
+9.88305 ∗ 10−10U f [x]2 U f00 [x] − 9.82218 ∗ 10−11U f [x]3 U f00 [x]
U[4] = 1.32468 + 7.82478 ∗ 10−7U f [x] − 5.26904 ∗ 10−7U f [x]2 + 2.48884 ∗ 10−7U f [x]3
+1.05836 ∗ 10−8U f00 [x] − 1.39954 ∗ 10−8U f [x] U f00 [x]
+1.00186 ∗ 10−8 U f [x]2 U f00 [x] − 1.26942 ∗ 10−9U f [x]3 U f00 [x]
U[5] = 1.32471 + 0.0000107587U f [x] − 7.21035 ∗ 10−6U f [x]2 + 3.33153 ∗ 10−6U f [x]3
+1.08485 ∗ 10−7U f00 [x] − 1.42477 ∗ 10−7U f [x] U f00 [x]
+1.00683 ∗ 10−7U f [x]2 U f00 [x] − 1.65819 ∗ 10−8U f [x]3 U f00 [x]
U[6] = 1.32462 + 0.000155827U f [x] − 0.000102937U f [x]2 + 0.00004608U f [x]3

+1.11907 ∗ 10−6U f00 [x] − 1.43165 ∗ 10−6U f [x]U f00 [x]
+9.84542 ∗ 10−7 U f [x]2 U f00 [x] − 2.19813 ∗ 10−7U f [x]3 U f00 [x]
U [7] = 1.32331 + 0.00230353U f [x] − 0.00149835U f [x]2 + 0.000647409U f [x]3

+0.0000107575 U f00 [x] − 0.0000132399U f [x] U f00 [x]
+8.83908 ∗ 10−6U f [x]2 U f00 [x] − 2.87055 ∗ 10−6U f [x]3 U f00 [x]
U [8] = 1.30408 + 0.0329322U f [x] − 0.0200174U f [x]2 + 0.00748572U f [x]3

+0.000082938 U f00 [x] − 0.000092876U f [x] U f00 [x]
+0.0000587807U f [x]2 U f00 [x] − 0.000027146U f [x]3 U f00 [x]
U[9] = 1.08379 + 0.311811U f [x] − 0.111793U f [x]2 + 0.0128827U f [x]3

+0.000321752U f00 [x] − 0.000274071U f [x] U f00 [x]
+0.000155453U f [x]2 U f00 [x] − 0.0000707759U f [x]3 U f00 [x]
Remark 27.2 Observe that since ε = 0.01 the solution is close to the constant value 1.3247 along the
domain, which is an approximate solution of equation −u3 + u + 1.0 = 0. Finally, the first output of the
method is the solution on the N1 − 1 = 9 nodes U[1], . . . ,U[9] which, in some sense, justify the terminology
hyper-finite differences, even though the solution in all the N · N1 = 300 lines have been obtained.
27.3 Conclusion
In this chapter we have developed two improvements concerning the generalized method of lines. For a large
class of models, we have solved the problem of minimizing the error as the parameter ε > 0 is small. In a first
step we present a proximal formulation through the introduction of a parameter K > 0 and related equation
part properly specified. In a second step, we develop the hyper-differences approach which corresponds to
a domain division in smaller sub-domains so that the solution on each sub-domain is obtained through the
generalized method of lines.
We highlight the methods here developed may be applied to a large class of problems, including the
Ginzburg-Landau system in superconductivity in the presence of a magnetic field and respective magnetic
potential.
We intend to address this kind of model and others such as the Navier-Stokes system in a future research.
Chapter 28
On the Generalized Method of

Lines Applied to the
Time-Independent
Incompressible Navier-Stokes
System
28.1 Introduction
In the first part of this article, we obtain a linear system whose the solution solves the time-independent
incompressible Navier-Stokes system for the special case in which the external forces vector is a gradient. In
a second step we develop approximate solutions, also for the time independent incompressible Navier-Stokes
system, through the generalized method of lines. We recall that for such a method, the domain of the partial
differential equation in question is discretized in lines and the concerning solution is written on these lines as
functions of the boundary conditions and boundary shape. Finally, we emphasize these last main results are
established through applications of the Banach fixed point theorem.
At this point we describe the system in question.
Consider Ω ⊂ R2 an open, bounded and connected set, whose regular (Lipschitzian) internal boundary
is denoted by Γ0 and the regular external one is denoted by Γ1 . For a two-dimensional motion of a fluid on
Ω, we denote by u : Ω → R the velocity field in the direction x of the Cartesian system (x, y), by v : Ω → R,
the velocity field in the direction y and by p : Ω → R, the pressure one. We define P = p/ρ, where ρ is the
constant fluid density. Finally, ν denotes the viscosity coefficient and g denotes the gravity field. Under such
notation and statements, the time-independent incompressible Navier-Stokes system of partial differential
equations is expressed by,
ν∇2 u − u∂x u − v∂y u − ∂x P + gx = 0, in Ω,




ν∇2 v − u∂x v − v∂y v − ∂y P + gy = 0, in Ω, (28.1)



∂x u + ∂y v = 0, in Ω,
On the Generalized Method of Lines Applied to the Time-Independent Incompressible Navier-Stokes System 527
(
u = v = 0, on Γ0 ,
(28.2)
u = u∞ , v = 0, P = P∞ , on Γ1
In principle we look for solutions (u, v, P) ∈ W 2,2 (Ω) × W 2,2 (Ω) × W 1,2 (Ω) despite the fact that less
regular solutions are also possible specially concerning the weak formulation. Details about such Sobolev
spaces may be found in [1]. General results on finite differences and existence theory for similar systems
may be found in [73] and [75], respectively.
28.2 On the solution of the time-independent incompressible Navier-

Stokes system through an associated linear one
Through the next result we obtain a linear system whose the solution also solves the time-independent in-
compressible Navier-Stokes system for the special case in which the external forces vector is a gradient.
Similar results for the time-independent incompressible Euler and Navier-Stokes equations have been
presented in [14] and [19, 16], respectively.
Indeed in the works [19, 16], we have indicated a solution of the Navier-Stokes system given by u = (u, v)
defined by

 u = ∂x w0 + ∂x w1 ,
(28.3)
v = ∂y w0 − ∂y w1 ,

where w0 , w1 are solutions of the system



 ∂xy w1 = 0 in Ω,
 2
∇ w0 + ∂xx w1 − ∂yy w1 = 0, in Ω,
(28.4)

 u = u0 , on Γ ≡ Γ0 ∪ Γ1 ,
v = v0 , on Γ.

Thus, in such a sense, the next result complements this previous one, by introducing a new function w2 in
the solution expressions, which makes the concerning boundary conditions perfectly possible to be satisfied.
Theorem 28.2.1 For h = (∂x f , ∂y f ) ∈ C1 (Ω; R2 ), consider the Navier-Stokes system similar as above indi-
cated, that is,


 ν∇2 u − u∂x u − v∂y u − ∂x P + ∂x f = 0, in Ω,



ν∇2 v − u∂x v − v∂y v − ∂y P + ∂y f = 0, in Ω, (28.5)




∂x u + ∂y v = 0, in Ω,



 u = u0 , on Γ
v = v0 , on Γ, (28.6)
P = P0 , on Γ1 .

A solution for such a Navier-Stokes system is given by u = (u, v) defined by


 u = ∂x w0 + ∂x w1 + ∂y w2 ,
(28.7)
v = ∂y w0 − ∂y w1 − ∂x w2 ,

where w0 , w1 , w2 are solutions of the system

 2

 ∇ w2 + 2∂xy w1 = 0 in Ω,
 2
∇ w0 + ∂xx w1 − ∂yy w1 = 0, in Ω,
(28.8)
 u = u0 ,
 on Γ,
v = v0 , on Γ,

and P is a solution of the system indicated in the first two lines of (28.5) with boundary conditions indicated
in the third line of (28.6).
Proof 28.1 For w0 , w1 , w2 such that ∇2 w2 + 2∂xy w1 = 0 in Ω, u and v as indicated above and defining
h1 = u∂x u + v∂y u,
h2 = u∂x v + v∂y v
and ϕ ≡ ∇2 w2 + 2∂xy w1 = 0, we have (you may check it using the softwares MATHEMATICA or MAPLE)
∂ h1 ∂ h2
−
∂y ∂x
= (−∂yy w1 + ∂yy w0 − ∂xy w2 )ϕ
+(−∂y w1 + ∂y w0 − ∂x w2 )∂y ϕ
+(∂xy w2 + ∂xx w1 + ∂xx w0 )ϕ
+(∂y w2 + ∂x w1 + ∂x w0 )∂x ϕ
= 0, in Ω. (28.9)
Moreover, since ϕ = ∇2 w2 + 2∂xy w1 = 0 in Ω, we get
ν (∂y ∇2 u − ∂x ∇2 v) + ∂y (∂x f ) − ∂x (∂y f )

= ν (2∂xy (∇2 w1 ) + ∇4 w2 ) + ∂y (∂x f ) − ∂x (∂y f )
= ν∇2 ϕ + ∂yx f − ∂yx f
= 0, in Ω. (28.10)
Summarizing, we have obtained
∂y ν∇2 u − u∂x u − v∂y u + ∂x f

= ∂x ν∇2 v − u∂x v − v∂y v + ∂y f .

(28.11)
Also, the equation

∇2 w0 + ∂xx w1 − ∂yy w1 = 0, in Ω
stands for
∂x u + ∂y v = 0, in Ω.
From these last results we may obtain P which satisfies the concerning boundary condition such that
 2
 ν∇ u − u∂x u − v∂y u − ∂x P + ∂x f = 0, in Ω,




ν∇2 v − u∂x v − v∂y v − ∂y P + ∂y f = 0, in Ω, (28.12)




∂x u + ∂y v = 0, in Ω,


28.3 The generalized method of lines for the Navier-Stokes system

In this section we develop the solution for the Navier-Stokes system through the generalized method of lines,
which was originally introduced in [19], with further developments in [18, 14]. We consider the boundary
conditions
u = u0 (x), v = v0 (x), P = P0 (x) on ∂ Ω0 ,
u = 0, v = 0, P = Pf (x) on ∂ Ω1 ,
where
Ω = {(r, θ ) : | r(θ ) ≤ r ≤ 2r(θ )}
and where r(θ ) is a positive, smooth and periodic function with period 2π,
∂ Ω = ∂ Ω0 ∪ ∂ Ω1 ,
∂ Ω0 = {(r(θ ), θ ) ∈ R2 : 0 ≤ θ ≤ 2π}
and
∂ Ω1 = {(2r(θ ), θ ) ∈ R2 : 0 ≤ θ ≤ 2π}.
For ν = 1, neglecting the gravity effects, the corresponding Navier-Stokes homogeneous system, in func-
tion of the variables (t, θ ) where t = r/r(θ ), is given by
L(u) − ud1 (u) − vd2 (u) − d1 (P) = 0, (28.13)
L(v) − ud1 (v) − vd2 (v) − d2 (P) = 0, (28.14)

d1 (u) + d2 (v) = 0, (28.15)
where generically
L(u) = ∇2 u,
d1 (u) = ∂x u,
and
d2 (u) = ∂y u
will be specified in the next lines, in function of (t, θ ).
Firstly, L is such that
r(θ )2 ∂ 2u

1 ∂u
L(u) = 2 + f2 (θ )
f0 (θ ) ∂t t ∂t
1 ∂ 2u f4 (θ ) ∂ 2 u
+ f3 (θ ) + 2 , (28.16)
t ∂ θ ∂t t ∂θ2
in Ω. Here f0 (θ ), f2 (θ ), f3 (θ ) and f4 (θ ) are known functions.
More specifically, denoting
−r0 (θ )
f1 (θ ) = ,
r(θ )
we have
f0 (θ ) = 1 + f1 (θ )2 ,
f10 (θ )
f2 (θ ) = 1 + ,
1 + f1 (θ )2
2 f1 (θ )
f3 (θ ) = ,
1 + f1 (θ )2
and
1
f4 (θ ) = .
1 + f1 (θ )2
Also d1 and d2 are expressed by
∂u ∂u
d1 u = fˆ5 (θ ) + ( fˆ6 (θ )/t) ,
∂t ∂θ
∂u ∂u
d2 u = fˆ7 (θ ) + ( fˆ8 (θ )/t)
∂t ∂θ
Where
fˆ5 (θ ) = cos(θ )/r(θ ) + sin(θ )r0 (θ )/r2 (θ ),
fˆ6 (θ ) = − sin(θ )/r(θ ),
fˆ7 (θ ) = sin(θ )/r(θ ) − cos(θ )r0 (θ )/r2 (θ ),
fˆ8 (θ ) = cos(θ )/r(θ ).
We also define
f0 (θ )
h3 (θ ) = ,
r(θ )2
r(θ )2 ˆ

f5 (θ ) = f5 (θ ),
f0 (θ )
r(θ )2 ˆ

f6 (θ ) = f6 (θ ),
f0 (θ )
r(θ )2 ˆ

f7 (θ ) = f7 (θ ),
f0 (θ )
and
r(θ )2

f8 (θ ) = fˆ8 (θ ),
f0 (θ )
Observe that t ∈ [1, 2] in Ω.
From equations (28.13) and (28.14) we may write
d1 (L(u) − ud1 (u) − vd2 (u) − d1 (P))

+d2 (L(v) − ud1 (v) − vd2 (v) − d2 (P)) = 0, (28.17)
From (28.15) we have

d1 [L(u)] + d2 [L(v)] = L(d1 (u) + d2 (v)) = 0,
and considering that
d1 (d1 (P)) + d2 (d2 (P)) = L(P),
from (28.17) we have
L(P) + d1 (u)2 + d2 (v)2 + 2d2 (u)d1 (v) = 0, in Ω.
Hence, in fact we solve the approximate system (this indeed is not exactly the Navier-Stokes one):
L(u) − ud1 (u) − vd2 (u) − d1 (P) = 0, (28.18)
L(v) − ud1 (v) − vd2 (v) − d2 (P) = 0, (28.19)

2 2
L(P) + d1 (u) + d2 (v) + 2d2 (u)d1 (v) = 0, in Ω. (28.20)
Remark 28.1 Who taught me how to obtain this last approximate system was Professor Alvaro de Bortoli
of Federal University of Rio Grande do Sul, UFRGS, Porto Alegre, RS-Brazil.
At this point, discretizing only in t (in N lines), defining d = 1/N and tn = 1 + nd, ∀n ∈ {1, . . . , N − 1},
we represent such a concerning system in partial finite differences.
Remark 28.2 In this text, we may generically consider the operators
∂ un ∂ 2 un
and
∂θ ∂θ2
also in a finite differences context, so that in such a case we may also consider them as bounded operators.
Denoting
(un − un−1 ) f6 (x) ∂ un
dˆ1 (un , un−1 ) = f5 (x) + ,
d tn ∂ x
and
(un − un−1 ) f8 (x) ∂ un
dˆ2 (un , un−1 ) = f7 (x) + ,
d tn ∂ x
where x stands for θ , in partial finite differences, equation (28.18) stands for
un+1 − 2un + un−1 f2 (x) un − un−1

+
d2 tn d
f4 (x) ∂ 2 un

f3 (x) ∂ un − un−1
+ + 2
tn ∂ x d tn ∂ x 2
−un dˆ1 (un , un−1 ) − vn dˆ2 (un , un−1 ) − dˆ1 (Pn , Pn−1 ) = 0. (28.21)
Hence, denoting u = (u, v, P), we have
un = (T1 )n (un+1 , un , un−1 ),
where
(T1 )n (un+1 , un , un−1 )

f2 (x)
= un+1 + un + un−1 + (un − un−1 )d
tn
f3 (x) ∂ f4 (x) ∂ 2 un 2
+ (un − un−1 )d + 2 d
tn ∂ x tn ∂ 2 x
−un dˆ1 (un , un−1 )d 2 − vn dˆ2 (un , un−1 )d 2 − dˆ1 (Pn , Pn−1 )d 2 /3.0.

Similarly, equation (28.19) stands for
vn+1 − 2vn + vn−1 f2 (x) vn − vn−1

+
d2 tn d
f4 (x) ∂ 2 vn

f3 (x) ∂ vn − vn−1
+ + 2
tn ∂ x d tn ∂ x2
−un dˆ1 (vn , vn−1 ) − vn dˆ2 (vn , vn−1 ) − dˆ2 (Pn , Pn−1 ) = 0. (28.22)
Hence,
vn = (T2 )n (un+1 , un , un−1 ),
where
(T2 )n (un+1 , un , un−1 )

f2 (x)
= vn+1 + vn + vn−1 + (vn − vn−1 )d
tn
f3 (x) ∂ f4 (x) ∂ 2 vn 2
+ (vn − vn−1 )d + 2 d
tn ∂ x tn ∂ x 2
−un dˆ1 (vn , vn−1 )d 2 − vn dˆ2 (vn , vn−1 )d 2 − dˆ2 (Pn , Pn−1 )d 2 /3.0.

Finally, (28.20) stands for

Pn+1 − 2Pn + Pn−1 f2 (x) Pn − Pn−1
+
d2 tn d
f4 (x) ∂ 2 Pn

f3 (x) ∂ Pn − Pn−1
+ + 2
tn ∂ x d tn ∂ x2
+h3 (x) dˆ1 (un , un−1 ) + dˆ2 (vn , vn−1 )2 + 2dˆ2 (un , un−1 )dˆ1 (vn , vn−1 ) = 0.
2

(28.23)
Hence,
Pn = (T3 )n (un+1 , un , un−1 ),
where
(T3 )n (un+1 , un , un−1 )

f2 (x)
= Pn+1 + Pn + Pn−1 + (Pn − Pn−1 )d
tn
f3 (x) ∂ f4 (x) ∂ 2 Pn 2
+ (Pn − Pn−1 )d + 2 d
tn ∂ x tn ∂ x 2
+h3 (x) dˆ1 (un , un−1 )2 d 2 + dˆ2 (vn , vn−1 )2 d 2 + 2dˆ2 (un , un−1 )dˆ1 (vn , vn−1 )d 2 /3.0.

Summarizing, we may write

un = T̂n (un+1 , un , un−1 ),
where
T̂n (un+1 , un , un−1 )

= ((T1 )n (un+1 , un , un−1 ), (T2 )n (un+1 , un , un−1 ), (T3 )n (un+1 , un , un−1 )), (28.24)
∀n ∈ {1, . . . , N − 1}.
Therefore, for n = 1 we obtain
u1 = T̂1 (u2 , u1 , u0 ).
We solve such an equation through the Banach fixed point theorem.
1. First set
(u1 )1 = u2 .
2. In a second step, define {uk1 } such that
u1k+1 = T̂1 (u2 , u1k , u0 ), ∀k ∈ N.

3. Finally obtain
u1 = lim uk1 ≡ F1 (u2 , u0 ).
k→∞
Now, reasoning inductively, having

un−1 = Fn−1 (un , u0 ),
we obtain un as indicated in the next lines.
1. First set
(un )1 = un+1 .
2. In a second step, define {ukn } such that
uk+1
n = Tˆn (un+1 , ukn , u0 ), ∀k ∈ N.
3. Finally obtain
un = lim ukn ≡ Fn (un+1 , u0 ).
k→∞
Thus, reasoning inductively we have obtained

un = Fn (un+1 , u0 ), ∀n ∈ {1, . . . , N − 1}.
In particular, for n = N − 1, we have uN = u f = (u f , v f , Pf ).
Therefore
uN −1 = FN −1 (uN , u0 ) ≡ HN −1 (u f , u0 ).
From this we obtain
uN−2 = FN−2 (uN −1 , u0 ) ≡ HN−2 (u f , u0 ),
u1 = F1 (u2 , u0 ) ≡ H1 (u f , u0 ).
The problem is then solved.
With such results in mind, with a software similar to those presented in the previous last chapter, trun-
cating the concerning series solutions for terms of order up to d 2 (in d), for the field of velocity u we have
obtained the following expressions for the lines (here x stands for θ ):
Line 1
u1 (x) = −0.045 f5 (x)Pf (x) + 0.045 f5 (x)P0 (x)
+0.899u0 (x) − 0.034 f2 (x)u0 (x) + 0.029 f5 (x)u0 (x)2
+0.029 f7 (x)u0 (x)v0 (x) − 0.011 f6 (x)Pf0
−0.022 f6 (x)P00 (x) − 0.034 f3 (x)u00 (x)
−0.016 f6 (x)u0 (x)u00 (x) − 0.016 f8 (x)v0 (x)u00 (x)
+0.018 f4 (x)u000 (x)
Line 2
u2 (x) = −0.081 f5 (x)Pf (x) + 0.081 f5 (x)P0 (x)
+0.799u0 (x) − 0.034 f2 (x)u0 (x) + 0.059 f5 (x)u0 (x)2
+0.048 f7 (x)u0 (x)v0 (x) − 0.022 f6 (x)Pf0
−0.036 f6 (x)P00 (x) − 0.059 f3 (x)u00 (x)
−0.025 f6 (x)u0 (x)u00 (x) − 0.025 f8 (x)v0 (x)u00 (x)
+0.028 f4 (x)u000 (x)
Line 3
u3 (x) = −0.106 f5 (x)Pf (x) + 0.106 f5 (x)P0 (x)
+0.698u0 (x) − 0.075 f2 (x)u0 (x) + 0.060 f5 (x)u0 (x)2
+0.060 f7 (x)u0 (x)v0 (x) − 0.031 f6 (x)Pf0
−0.044 f6 (x)P00 (x) − 0.075 f3 (x)u00 (x)
−0.029 f6 (x)u0 (x)u00 (x) − 0.029 f8 (x)v0 (x)u00 (x)
+0.033 f4 (x)u000 (x)
Line 4
u4 (x) = −0.121 f5 (x)Pf (x) + 0.121 f5 (x)P0 (x)
+0.597u0 (x) − 0.084 f2 (x)u0 (x) + 0.064 f5 (x)u0 (x)2
+0.064 f7 (x)u0 (x)v0 (x) − 0.037 f6 (x)Pf0
−0.046 f6 (x)P00 (x) − 0.084 f3 (x)u00 (x)
−0.029 f6 (x)u0 (x)u00 (x) − 0.029 f8 (x)v0 (x)u00 (x)
+0.034 f4 (x)u000 (x)
Line 5
u5 (x) = −0.126 f5 (x)Pf (x) + 0.126 f5 (x)P0 (x)
+0.497u0 (x) − 0.086 f2 (x)u0 (x) + 0.062 f5 (x)u0 (x)2
+0.062 f7 (x)u0 (x)v0 (x) − 0.041 f6 (x)Pf0
−0.044 f6 (x)P00 (x) − 0.086 f3 (x)u00 (x)
−0.026 f6 (x)u0 (x)u00 (x) − 0.026 f8 (x)v0 (x)u00 (x)
+0.032 f4 (x)u000 (x)
Line 6
u6 (x) = −0.121 f5 (x)Pf (x) + 0.121 f5 (x)P0 (x)
+0.397u0 (x) − 0.080 f2 (x)u0 (x) + 0.056 f5 (x)u0 (x)2
+0.056 f7 (x)u0 (x)v0 (x) − 0.041 f6 (x)Pf0
−0.031 f6 (x)P00 (x) − 0.069 f3 (x)u00 (x)
−0.017 f6 (x)u0 (x)u00 (x) − 0.017 f8 (x)v0 (x)u00 (x)
+0.022 f4 (x)u000 (x)
Line 7
u7 (x) = −0.105 f5 (x)Pf (x) + 0.105 f5 (x)P0 (x)
+0.297u0 (x) − 0.069 f2 (x)u0 (x) + 0.045 f5 (x)u0 (x)2
+0.045 f7 (x)u0 (x)v0 (x) − 0.037 f6 (x)Pf0
−0.031 f6 (x)P00 (x) − 0.069 f3 (x)u00 (x)
−0.017 f6 (x)u0 (x)u00 (x) − 0.017 f8 (x)v0 (x)u00 (x)
+0.022 f4 (x)u000 (x)
Line 8
u8 (x) = −0.080 f5 (x)Pf (x) + 0.080 f5 (x)P0 (x)
+0.198u0 (x) − 0.051 f2 (x)u0 (x) + 0.032 f5 (x)u0 (x)2
+0.032 f7 (x)u0 (x)v0 (x) − 0.029 f6 (x)Pf0
−0.021 f6 (x)P00 (x) − 0.051 f3 (x)u00 (x)
−0.011 f6 (x)u0 (x)u00 (x) − 0.011 f8 (x)v0 (x)u00 (x)
+0.015 f4 (x)u000 (x)
Line 9
u9 (x) = −0.045 f5 (x)Pf (x) + 0.045 f5 (x)P0 (x)
+0.099u0 (x) − 0.028 f2 (x)u0 (x) + 0.016 f5 (x)u0 (x)2
+0.016 f7 (x)u0 (x)v0 (x) − 0.017 f6 (x)Pf0
−0.022 f6 (x)P00 (x) − 0.012 f3 (x)u00 (x)
−0.006 f6 (x)u0 (x)u00 (x) − 0.006 f8 (x)v0 (x)u00 (x)
+0.008 f4 (x)u000 (x)
For the field of velocity v, we have obtained the following expressions for the lines:
Line 1
v1 (x) = −0.045 f7 (x)Pf (x) + 0.045 f7 (x)P0 (x)
+0.899v0 (x) − 0.034 f2 (x)v0 (x) + 0.029 f5 (x)u0 (x)v0 (x)
+0.029 f7 (x)v0 (x)2 − 0.011 f8 (x)Pf0 (x)
−0.022 f8 (x)P00 (x) − 0.034 f3 (x)v00 (x)
−0.016 f6 (x)u0 (x)v00 (x) − 0.016 f8 (x)v0 (x)v00 (x)
+0.018 f4 (x)v000 (x)
Line 2
v2 (x) = −0.081 f7 (x)Pf (x) + 0.081 f7 (x)P0 (x)
+0.799v0 (x) − 0.059 f2 (x)v0 (x) + 0.048 f5 (x)u0 (x)v0 (x)
+0.048 f7 (x)v0 (x)2 − 0.022 f8 (x)Pf0 (x)
−0.036 f8 (x)P00 (x) − 0.059 f3 (x)v00 (x)
−0.025 f6 (x)u0 (x)v00 (x) − 0.025 f8 (x)v0 (x)v00 (x)
+0.028 f4 (x)v000 (x)
Line 3
v3 (x) = −0.106 f7 (x)Pf (x) + 0.106 f7 (x)P0 (x)
+0.698v0 (x) − 0.075 f2 (x)v0 (x) + 0.060 f5 (x)u0 (x)v0 (x)
+0.060 f7 (x)v0 (x)2 − 0.031 f8 (x)Pf0 (x)
−0.044 f8 (x)P00 (x) − 0.075 f3 (x)v00 (x)
−0.029 f6 (x)u0 (x)v00 (x) − 0.029 f8 (x)v0 (x)v00 (x)
+0.033 f4 (x)v000 (x)
Line 4
v4 (x) = −0.121 f7 (x)Pf (x) + 0.121 f7 (x)P0 (x)
+0.597v0 (x) − 0.084 f2 (x)v0 (x) + 0.064 f5 (x)u0 (x)v0 (x)
+0.064 f7 (x)v0 (x)2 − 0.037 f8 (x)Pf0 (x)
−0.046 f8 (x)P00 (x) − 0.084 f3 (x)v00 (x)
−0.029 f6 (x)u0 (x)v00 (x) − 0.029 f8 (x)v0 (x)v00 (x)
+0.034 f4 (x)v000 (x)
Line 5
v5 (x) = −0.126 f7 (x)Pf (x) + 0.126 f7 (x)P0 (x)
+0.497v0 (x) − 0.086 f2 (x)v0 (x) + 0.062 f5 (x)u0 (x)v0 (x)
+0.062 f7 (x)v0 (x)2 − 0.041 f8 (x)Pf0 (x)
−0.044 f8 (x)P00 (x) − 0.086 f3 (x)v00 (x)
−0.026 f6 (x)u0 (x)v00 (x) − 0.026 f8 (x)v0 (x)v00 (x)
+0.032 f4 (x)v000 (x)
Line 6
v6 (x) = −0.121 f7 (x)Pf (x) + 0.121 f7 (x)P0 (x)
+0.397v0 (x) − 0.080 f2 (x)v0 (x) + 0.056 f5 (x)u0 (x)v0 (x)
+0.056 f7 (x)v0 (x)2 − 0.022 f8 (x)Pf0 (x)
−0.041 f8 (x)P00 (x) − 0.080 f3 (x)v00 (x)
−0.022 f6 (x)u0 (x)v00 (x) − 0.022 f8 (x)v0 (x)v00 (x)
+0.028 f4 (x)v000 (x)
Line 7
v7 (x) = −0.105 f7 (x)Pf (x) + 0.105 f7 (x)P0 (x)
+0.297v0 (x) − 0.069 f2 (x)v0 (x) + 0.045 f5 (x)u0 (x)v0 (x)
+0.045 f7 (x)v0 (x)2 − 0.037 f8 (x)Pf0 (x)
−0.031 f8 (x)P00 (x) − 0.069 f3 (x)v00 (x)
−0.017 f6 (x)u0 (x)v00 (x) − 0.017 f8 (x)v0 (x)v00 (x)
+0.022 f4 (x)v000 (x)
Line 8
v8 (x) = −0.080 f7 (x)Pf (x) + 0.080 f7 (x)P0 (x)
+0.198v0 (x) − 0.051 f2 (x)v0 (x) + 0.032 f5 (x)u0 (x)v0 (x)
+0.032 f7 (x)v0 (x)2 − 0.029 f8 (x)Pf0 (x)
−0.021 f8 (x)P00 (x) − 0.051 f3 (x)v00 (x)
−0.011 f6 (x)u0 (x)v00 (x) − 0.011 f8 (x)v0 (x)v00 (x)
+0.015 f4 (x)v000 (x)
Line 9
v9 (x) = −0.045 f7 (x)Pf (x) + 0.045 f7 (x)P0 (x)
+0.099v0 (x) − 0.028 f2 (x)v0 (x) + 0.016 f5 (x)u0 (x)v0 (x)
+0.016 f7 (x)v0 (x)2 − 0.017 f8 (x)Pf0 (x)
−0.011 f8 (x)P00 (x) − 0.029 f3 (x)v00 (x)
−0.057 f6 (x)u0 (x)v00 (x) − 0.057 f8 (x)v0 (x)v00 (x)
+0.008 f4 (x)v000 (x)
Finally, for the field of pressure P, we have obtained the following lines:
Line 1
P1 (x) = 0.101Pf (x) + 0.034 f2 (x)Pf (x)
+0.899P0 (x) − 0.034 f2 (x)P0 (x) + 0.046h3 (x) f5 (x)2 u0 (x)2
+0.092h3 (x) f5 (x) f7 (x)u0 (x)v0 (x) + 0.046h3 (x) f7 (x)2 v0 (x)2
+0.034 f3 (x)Pf0 (x) − 0.034 f3 (x)P00 (x)
−0.045h3 (x) f5 (x) f6 (x)u0 (x)u00 (x) − 0.045h3 (x) f5 (x) f8 (x)v0 (x)u00 (x)
+0.014h3 (x) f6 (x)2 u00 (x)2 − 0.045h3 (x) f6 (x) f7 (x)u0 (x)v00 (x)
−0.045h3 (x) f7 (x) f8 (x)v0 (x)v00 (x) + 0.027h3 (x) f6 (x) f8 (x)u00 (x)v00 (x)
+0.014h3 (x) f8 (x)2 v00 (x)2 + 0.008 f4 (x)Pf00 (x)
+0.018 f4 (x)P000 (x)
Line 2
P2 (x) = 0.201Pf (x) + 0.059 f2 (x)Pf (x)
+0.799P0 (x) − 0.059 f2 (x)P0 (x) + 0.082h3 (x) f5 (x)2 u0 (x)2
+0.059 f3 (x)Pf0 (x) − 0.059 f3 (x)P00 (x)
+0.020h3 (x) f8 (x)2 v00 (x)2 + 0.015 f4 (x)Pf00 (x)
+0.028 f4 (x)P000 (x)
Line 3
P3 (x) = 0.302Pf (x) + 0.075 f2 (x)Pf (x)
+0.698P0 (x) − 0.075 f2 (x)P0 (x) + 0.107h3 (x) f5 (x)2 u0 (x)2
+0.075 f3 (x)Pf0 (x) − 0.075 f3 (x)P00 (x)
+0.023h3 (x) f8 (x)2 v00 (x)2 + 0.021 f4 (x)Pf00 (x)
+0.033 f4 (x)P000 (x)
Line 4
P4 (x) = 0.403Pf (x) + 0.084 f2 (x)Pf (x)
+0.597P0 (x) − 0.084 f2 (x)P0 (x) + 0.122h3 (x) f5 (x)2 u0 (x)2
+0.084 f3 (x)Pf0 (x) − 0.084 f3 (x)P00 (x)
+0.022h3 (x) f8 (x)2 v00 (x)2 + 0.025 f4 (x)Pf00 (x)
+0.034 f4 (x)P000 (x)
Line 5
P5 (x) = 0.503Pf (x) + 0.086 f2 (x)Pf (x)
+0.497P0 (x) − 0.086 f2 (x)P0 (x) + 0.127h3 (x) f5 (x)2 u0 (x)2
+0.086 f3 (x)Pf0 (x) − 0.086 f3 (x)P00 (x)
+0.020h3 (x) f8 (x)2 v00 (x)2 + 0.026 f4 (x)Pf00 (x)
+0.032 f4 (x)P000 (x)
Line 6
P6 (x) = 0.603Pf (x) + 0.080 f2 (x)Pf (x)
+0.397P0 (x) − 0.080 f2 (x)P0 (x) + 0.122h3 (x) f5 (x)2 u0 (x)2
+0.080 f3 (x)Pf0 (x) − 0.080 f3 (x)P00 (x)
+0.017h3 (x) f8 (x)2 v00 (x)2 + 0.026 f4 (x)Pf00 (x)
+0.028 f4 (x)P000 (x)
Line 7
P7 (x) = 0.703Pf (x) + 0.069 f2 (x)Pf (x)
+0.297P0 (x) − 0.069 f2 (x)P0 (x) + 0.107h3 (x) f5 (x)2 u0 (x)2
+0.069 f3 (x)Pf0 (x) − 0.069 f3 (x)P00 (x)
−0.063 f5 (x) f6 (x)u0 (x)u00 (x) − 0.063 f5 (x) f8 (x)v0 (x)u00 (x)
+0.013h3 (x) f8 (x)2 v00 (x)2 + 0.023 f4 (x)Pf00 (x)
+0.022 f4 (x)P000 (x)
Line 8
P8 (x) = 0.802Pf (x) + 0.051 f2 (x)Pf (x)
+0.198P0 (x) − 0.051 f2 (x)P0 (x) + 0.081h3 (x) f5 (x)2 u0 (x)2
+0.051 f3 (x)Pf0 (x) − 0.051 f3 (x)P00 (x)
+0.009h3 (x) f8 (x)2 v00 (x)2 + 0.018 f4 (x)Pf00 (x)
+0.015 f4 (x)P000 (x)
Line 9
P9 (x) = 0.901Pf (x) + 0.028 f2 (x)Pf (x)
+0.099P0 (x) − 0.028 f2 (x)P0 (x) + 0.045h3 (x) f5 (x)2 u0 (x)2
+0.028 f3 (x)Pf0 (x) − 0.028 f3 (x)P00 (x)
+0.004h3 (x) f8 (x)2 v00 (x)2 + 0.010 f4 (x)Pf00 (x)
+0.008 f4 (x)P000 (x)
28.3.1 Numerical examples through the generalized method of lines

We consider some examples in which
Ω = {(r, θ ) | 1 ≤ r ≤ 2, 0 ≤ θ ≤ 2π},
∂ Ω0 = {(1, θ ) | 0 ≤ θ ≤ 2π},
and
∂ Ω1 = {(2, θ ) | 0 ≤ θ ≤ 2π}.
For such cases, the boundary conditions are
u = u0 (θ ), v = v0 (θ ), on ∂ Ω0 ,
u = v = 0, on ∂ Ω1 ,
so that in these examples we do not have boundary conditions for the pressure.
Through the generalized method of lines, neglecting gravity effects and truncating the series up to the
terms in d 2 , where d = 1/N is the mesh thickness concerning the discretization in r, we present numerical
results for the following approximation of the Navier-Stokes system,


 ν∇2 u − u∂x u − v∂y u − ∂x P = 0, in Ω,



ν∇2 v − u∂x v − v∂y v − ∂y P = 0, in Ω, (28.25)




ε∇2 P + ∂x u + ∂y v = 0, in Ω,

where ε > 0 is a very small parameter. We highlight, since ε > 0 must be very small, the results obtained
through the generalized of lines, as indicated in the last section, may have a relevant error. Anyway, we may
use such a procedure to obtain the general line expressions, which we expect to be analytically suitable to
obtain a numerical result by calculating numerically the concerning coefficients for such lines, through a
minimization of the L2 norm of equation errors, in a finite differences context.
Thus, from such a method, the general expression for the velocity and pressure fields on the line n, are
given by (here x stands for θ and P0 must be calculated numerically in the optimization process):
un (x) = a1 [n] cos(x) + a2 [n]u0 (x) + a3 [n] cos[x]u0 (x)2
a4 [n] sin(x)u0 (x)v0 (x) + a5 [n] sin(x)u0 (x)u00 (x) + a6 [n] cos(x)v0 (x)u00 (x)
a7 [n]u000 (x) + a8 [n] sin(x) + a9 [n] sin(x)P0 (x)
a10 [n]P0 (x) cos(x) + a11 [n], (28.26)
vn (x) = b1 [n] sin(x) + b2 [n]v0 (x) + b3 [n] sin[x]v0 (x)2

b4 [n] cos(x)u0 (x)v0 (x) + b5 [n] sin(x)v00 (x)u0 (x) + b6 [n] cos(x)v0 (x)u00 (x)
b7 [n]v000 (x) + b8 [n] cos(x) + b9 [n] sin(x)P0 (x)
b10 [n]P0 (x) cos(x) + b11 [n], (28.27)
Pn (x) = c1 [n] + c2 [n] cos(x)u0 (x) + c3 [n] sin(x)v0 (x) + c4 [n] sin(x)u00 (x)
+c5 [n] cos(x)v00 (x) + c6 [n]u000 (x) + c7 [n]v000 (x)
+c9 [n]P0 (x) + c10 [n]P000 (x). (28.28)
First example: For the first example,
u0 (x) = −1.5 sin(x)
and
v0 (x) = 1.5 cos(x).
Denoting
Z 2
J(u, v, P) = ν∇2 u − u∂x u − v∂y u − ∂x P dΩ
Ω
Z 2
+ ν∇2 v − u∂x v − v∂y v − ∂y P dΩ
ZΩ
+ (∂x u + ∂y v)2 dΩ, (28.29)
Ω
as, above mentioned, the coefficients {ai [n]}, {bi [n]}, {ci [n]} have been obtained through the numer-
ical minimization of J({un }, {vn }, {Pn }), so that for the mesh in question, we have obtained
For this first example: J({un }, {vn }, {Pn }) ≈ 9.23 10−12 for ν = 0.1,
We have plotted the fields u, v and P, for the lines n = 1, n = 5, n = 10, n = 15 and n = 19, for
a mesh 20 × 150 corresponding to 20 lines. Please, see the Figures from 28.1 to 28.6 for the case
ν = 0.1. For all graphs, please consider units in x to be multiplied by 2π/150.
1.5 1
0.8
1
0.6
0.4
0.5
0.2
0 0
−0.2
−0.5
−0.4
−0.6
−1
−0.8
−1.5 −1
0 50 100 150 0 50 100 150
0.6
0.4
0.2
−0.2
−0.4
−0.6
−0.8
0 50 100 150
Figure 28.1: First example, from the left to the right, fields of velocity u1 (x), u5 (x), u10 (x) for the lines n = 1, n = 5
and n = 10.
0.3 0.06
0.2
0.04
0.1
0.02
0
0
−0.1
−0.02
−0.2
−0.04
−0.3
−0.4 −0.06
0 50 100 150 0 50 100 150
Figure 28.2: First example, from the left to the right, fields of velocity u15 (x), u19 (x) for the lines n = 15, and n = 19.
1.5 1
0.8
1
0.6
0.4
0.5
0.2
0 0
−0.2
−0.5
−0.4
−0.6
−1
−0.8
−1.5 −1
0 50 100 150 0 50 100 150
0.6
0.4
0.2
−0.2
−0.4
−0.6
−0.8
0 50 100 150
Figure 28.3: First example, from the left to the right, fields of velocity v1 (x), v5 (x), v10 (x) for the lines n = 1, n = 5 and
n = 10.
0.3 0.06
0.2
0.04
0.1
0.02
0
0
−0.1
−0.02
−0.2
−0.04
−0.3
−0.4 −0.06
0 50 100 150 0 50 100 150
Figure 28.4: First example, from the left to the right, fields of velocity v15 (x), v19 (x) for the lines n = 15, and n = 19.
Second example: For the second example, we consider
u0 (x) = −3.0 cos(x) sin(x),
v0 (x) = −(−2.0 cos(x)2 + sin(x)2 )

and ν = 1.0.
0.063 0.285
0.0625
0.284
0.062
0.283
0.0615
0.282
0.061
0.0605 0.281
0.06
0.28
0.0595
0.279
0.059
0.278
0.0585
0.058 0.277
0 50 100 150 0 50 100 150
0.388
0.387
0.386
0.385
0.384
0.383
0.382
0.381
0.38
0 50 100 150
Figure 28.5: First example, from the left to the right, fields of pressure P1 (x), P5 (x), P10 (x) for the lines n = 1, n = 5
and n = 10.
0.414 0.403
0.413 0.402
0.412
0.401
0.411
0.4
0.41
0.399
0.409
0.398
0.408
0.397
0.407
0.406 0.396
0.405 0.395
0 50 100 150 0 50 100 150
Figure 28.6: First example, from the left to the right, fields of pressure P15 (x), P19 (x) for the lines n = 15, and n = 19.
Again the coefficients {ai [n]}, {bi [n]}, {ci [n]} have been obtained through the numerical minimiza-
tion of J({un }, {vn }, {Pn }), so that for the mesh in question, we have obtained
For this second example: J({un }, {vn }, {Pn }) ≈ 6.0 10−7 for ν = 1.0.
In any case, considering the values obtained for J, it seems we have got good first approximations
for the concerning solutions.
For the second example, for the field of velocities u and v, and pressure field P, for the lines n = 1,
n = 5, n = 10, n = 15 and n = 19, please see Figures 28.7 to 28.12. Once more, for all graphs, please
consider units in x to be multiplied by 2π/150.
1.5 1
0.8
1
0.6
0.4
0.5
0.2
0 0
−0.2
−0.5
−0.4
−0.6
−1
−0.8
−1.5 −1
0 50 100 150 0 50 100 150
0.6
0.4
0.2
−0.2
−0.4
−0.6
−0.8
0 50 100 150
Figure 28.7: Second example, from the left to the right, fields of velocity u1 (x), u5 (x), u10 (x) for the lines n = 1, n = 5
and n = 10.
0.3 0.15
0.2
0.1
0.1
0.05
0
−0.1
0
−0.2
−0.05
−0.3
−0.4 −0.1
0 50 100 150 0 50 100 150
Figure 28.8: Second example, from the left to the right, fields of velocity u15 (x), u19 (x) for the lines n = 15, and n = 19.
2 1.4
1.2
1.5
1
0.8
1
0.6
0.5 0.4
0.2
0
0
−0.2
−0.5
−0.4
−1 −0.6
0 50 100 150 0 50 100 150
0.8
0.6
0.4
0.2
−0.2
−0.4
0 50 100 150
Figure 28.9: Second example, from the left to the right, fields of velocity v1 (x), v5 (x), v10 (x) for the lines n = 1, n = 5
and n = 10.
0.4 0.15
0.3 0.1
0.2 0.05
0.1 0
0 −0.05
−0.1 −0.1
−0.2 −0.15
−0.3 −0.2
0 50 100 150 0 50 100 150
Figure 28.10: Second example, from the left to the right, fields of velocity v15 (x), v19 (x) for the lines n = 15, and n = 19.
28.4 Conclusion
In the first part of this chapter, we obtain a linear system whose solution solves the time-independent in-
compressible Navier-Stokes system. In the second part, we develop solutions for two-dimensional examples
also for the time-independent incompressible Navier-Stokes system, through the generalized method of lines.
0.5 0.8
0.4 0.7
0.3
0.6
0.2
0.5
0.1
0.4
0
0.3
−0.1
−0.2 0.2
−0.3 0.1
0 50 100 150 0 50 100 150
0.9
0.8
0.7
0.6
0.5
0.4
0 50 100 150
Figure 28.11: Second example, from the left to the right, fields of pressure P1 (x), P5 (x), P10 (x) for the lines n = 1, n = 5
and n = 10.
1.1 4
1 3
0.9
2
0.8
1
0.7
0
0.6
−1
0.5
0.4 −2
−3
0 50 100 150 0 50 100 150
Figure 28.12: Second example, from the left to the right, fields of pressure P15 (x), P19 (x) for the lines n = 15, and n = 19.
Considering the values for J obtained, we have got very good approximate solutions for the model in ques-
tion, in a finite differences context. The extension of such results to R3 , compressible and time dependent
cases is planned for a future work.
Chapter 29
A Numerical Method for an

Inverse Optimization Problem
through the Generalized Method
of Lines
29.1 Introduction
In this chapter we develop a numerical method to compute the solution of an inverse problem through the
generalized method of lines.
More specifically, we consider a Laplace equation on a domain Ω ⊂ R2 with an internal boundary denoted
by ∂ Ω0 and an external one denoted by ∂ Ω1 . We prescribe boundary conditions for both ∂ Ω0 and ∂ Ω1 and a
third boundary condition for ∂ Ω1 (the external one) and consider the problem of finding the optimal shape for
the internal boundary ∂ Ω0 for which such a third boundary condition for the external boundary is satisfied.
The idea is to discretize the domain in lines (in fact curves) and write the solution of Laplace equation on
these lines as functions of the unknown internal boundary shape, through the generalized method of lines.
The second step is to minimize a functional which corresponds to the L2 norms of Laplace equation and
concerning third boundary condition.
Remark 29.1 About the references, this and many other similar problems are addressed in [49]. The gen-
eralized method of lines has been originally introduced in [19], with additional results in [18, 14]. Moreover,
about finite differences schemes we would cite [73]. Finally, details on the function spaces here addressed
may be found in [1].
29.2 The mathematical description of the main problem

At this point we start to describe mathematically our main problem.
Let Ω ⊂ R2 be a bounded, closed and connected set defined by
Ω = {(r, θ ) ∈ R2 : r(θ ) ≤ r ≤ R : 0 ≤ θ ≤ 2π}.

Consider a Laplace equation and concerning boundary conditions expressed by

 2

 ∇ u = 0, in Ω,
u = uo , on ∂ Ω0 ,

(29.1)

 u = u f , on ∂ Ω1 ,
∇u · n = w, on ∂ Ω1 ,

where uo , u f and w ∈ C2 ([0, 2π]) are known periodic functions with period 2π, n denotes the outward normal
field to ∂ Ω,
∂ Ω0 = {(r(θ ), θ ) : 0 ≤ θ ≤ 2π},
∂ Ω1 = {(R, θ ) : 0 ≤ 0 ≤ 2π},
and R > 0.
The main idea is to discretize the domain in lines and through the generalized method of lines to obtain
the solution un (r(θ )) on each line n as a function of r(θ ). The final step is to compute the optimal r(θ ) in
order to minimize the cost functional J(r(θ )) defined by
J (r(θ )) = k∇2 uk2,Ω

2 2
+ Kk∇u · n − wk2,∂ Ω1 ,
where K > 0 is an appropriate constant to be specified.
29.3 About the generalized method of lines and the main result
At this point we start to describe the main result.
Consider firstly a Laplace equation in polar coordinates, that is,
∂ 2u 1 ∂ u 1 ∂ 2u
+ + = 0.
∂ r2 r ∂ r r2 ∂ θ 2
In order to apply the generalized method of lines, we a define a variable t, through the equation
r − r(θ )
t=
R − r(θ )
so that t ∈ [0, 1] in the set Ω previously specified.

Hence, denoting u(r, θ ) = u(t, θ ), we obtain,
∂u ∂ u ∂t ∂u
= +
∂θ ∂t ∂θ ∂θ
∂u t −1 ∂u
= r0 (θ ) +
∂t R − r(θ ) ∂θ
∂u ∂u
= ( f1 (θ ) + t f2 (θ )) + , (29.2)
∂t ∂θ
where
−r0 (θ )
f1 (θ ) = ,
R − r(θ )
and
r0 (θ )
f2 (θ ) = .
R − r(θ )
A Numerical Method for an Inverse Optimization Problem through the Generalized Method of Lines 549
Also,
∂u ∂ u ∂t
=
∂r ∂t ∂ r
∂u 1
= , (29.3)
∂t R − r(θ )
so that
2
∂ 2u ∂ 2u

∂t
=
∂ r2 ∂t 2 ∂r
∂ 2u 1
=
∂t 2 (R − r(θ ))2
∂ 2u
= f3 (θ ) , (29.4)
∂t 2
where
1
f3 (θ ) = .
(R − r(θ ))2
Moreover, we may also obtain
∂ 2u ∂ 2u ∂u
= f4 (t, θ ) + f5 (t, θ )
∂θ2 ∂t 2 ∂t
∂ 2u ∂ 2u
+ f6 (t, θ ) + , (29.5)
∂t∂ θ ∂ θ 2
where,
f4 (t, θ ) = ( f1 (θ ) + t f2 (θ ))2 ,
f5 (t, θ ) = f10 (θ ) + t f20 (θ ) + f2 (θ )( f1 (θ ) + t f2 (θ )),
f6 (t, θ ) = 2( f1 (θ ) + t f2 (θ )).
Thus, for the new variables (t, θ ), dropping the bar in u, the Laplace equation is equivalent to
∂ 2u ∂u ∂ 2u ∂ 2u
+ f 7 (t, θ ) + f 8 (t, θ ) + f 9 (t, θ ) = 0,
∂t 2 ∂t ∂t∂ θ ∂θ2
in Ω̂ where
Ω̂ = {(t, θ ) ∈ R2 : 0 ≤ t ≤ 1, 0 ≤ θ ≤ 2π},
r = t(R − r(θ )) + r(θ ),
1 f4 (t, θ )
f0 (t, θ ) = + ,
(R − r(θ ))2 r2
f˜7 (t, θ )
f7 (t, θ ) = ,
f0 (t, θ )
f˜8 (t, θ )
f8 (t, θ ) = ,
f0 (t, θ )
f˜9 (t, θ )
f9 (t, θ ) =
f0 (t, θ )
and where
1 f5 (t, θ )
f˜7 (t, θ ) = +
r(R − r(θ )) r2
f6 (t, θ )
f˜8 (t, θ ) =
r2
and
1
f˜9 (t, θ ) = 2 .
r
So, discretizing the interval [0, 1] into N ∈ N pieces, that is defining,
n
tn = , ∀n ∈ {0, . . . , N }
N
and d = 1/N, in partial finite differences, the concerning equation stands for

un+1 − 2un + un−1 un − un−1
+ f7 (tn , θ )
d2 d
∂ 2 un

∂ un − un−1
+ f8 (tn , θ ) + f9 (tn , θ ) = 0, (29.6)
∂θ d ∂θ2
∀n ∈ {1, . . . , N − 1}, where

u0 = uo (θ )
and
uN = u f (θ ).
At this point we describe how to obtain the general expression for un corresponding to the line n, through
the generalized method of lines.
For n = 1 in (29.6), we have
u2 − 2u1 + u0 + f7 (t1 , θ )(u1 − u0 )d

∂ (u1 − u0 ) ∂ 2 u1 2
+ f8 (t1 , θ ) d + f9 (t1 , θ ) d = 0, (29.7)
∂θ ∂θ2
so that
u1 = T1 (u2 , u1 , u0 ),
where
T1 (u2 , u1 , u0 ) = (u2 + u1 + u0 + f7 (t1 , θ )(u1 − u0 )d

∂ 2 u1 2

∂ (u1 − u0 )
+ f8 (t1 , θ ) d + f9 (t1 , θ ) d /3. (29.8)
∂θ ∂θ2
To solve this equation we apply the Banach fixed point theorem as it follows:
1. Set u11 = u2 .
2. Define
uk+1
1 = T1 (u2 , uk1 , u0 ), ∀k ∈ N.
3. Obtain
u1 = lim uk1 ≡ F1 (u2 , u0 ).
k→∞

un−1 = Fn−1 (un , u0 )
also from (29.6), for the line n, we have
un = Tn (un+1 , un , u0 ),
where
Tn (un+1 , un , u0 ) = (un+1 + un + Fn−1 (un , u0 ) + f7 (tn , θ )(un − Fn−1 (un , u0 ))d

∂ 2 un 2

∂ (un − Fn−1 (un , u0 ))
+ f8 (tn , θ ) d + f9 (tn , θ ) d /3. (29.9)
∂θ ∂θ2
Again, to solve this equation, we apply the Banach fixed point theorem, as it follows:
1. Set u1n = un+1 .
2. Define
uk+1
n = Tn (un+1 , ukn , u0 ), ∀k ∈ N.
3. Obtain
un = lim ukn ≡ Fn (un+1 , u0 ), ∀n{2, . . . , N − 1}.
k→∞
In particular, for n = N − 1, we get
uN −1 = FN −1 (uN , u0 ) = FN −1 (u f , u0 ).
Similarly, for n = N − 2, we obtain
uN−2 = FN−2 (uN−1 , u0 ),

u1 = F1 (u2 , u0 ).
This means that we have obtained a general expression
un = Fn (un+1 , u0 ) ≡ Hn (u f , u0 , r(θ )), ∀n ∈ {1, . . . , N − 1},
Anyway, we remark to properly run the software we have to make the approximation tn = t, ∀n ∈
{1, . . . , N − 1}. Thus, we have used the method above described just to find a general expression for un ,
which is approximately given by (here x stands for θ )
un (x) ≈ an [1]u f (x) + an [2]u0 (x) + an [3] f7 [tn , x]u f (x) + an [4] f7 [tn , x]u0 (x)
+an [5] f8 [tn , x]u0f (x) + an [6] f8 [tn , x]u00 (x)
+an [9] f9 [tn , x]u00f (x) + an [10] f9 [tn , x]u000 (x). (29.10)
Indeed this a first approximation for the series representing the solution on each line obtained by considering
terms up to order d 2 (in d).
Moreover the coefficients an [k] and r(θ ) expressed in finite differences are calculated through the mini-
mization of J(r(θ ), {an [k]}) given by
J(r(θ ), {an [k]}) = k∇2 uk22,Ω + Kk∇u · n − wk22,∂ Ω1 ,

29.3.1 The numerical results

In a first step we solve, through the generalized method of lines, the following Laplace equation
 2

 ∇ u = 0, in Ω,
u = uo , on ∂ Ω0 ,

(29.11)

 u = u f , on ∂ Ω1 ,
∇u · n = w, on ∂ Ω1 ,

where
∂ Ω0 = {(θ , r(θ )) ∈ R2 : r(θ ) = 5.5(1.4 + cos(θ ))/1.5, and 0 ≤ θ ≤ 2π},
and as above indicated
∂ Ω1 = {(θ , R) ∈ R2 : 0 ≤ θ ≤ 2π}.
Having obtained u = {un }Nn=1 for N = 26 lines we define W = (u f − un−1 )/d, where d = 1/N. We have
considered N1 = 42 nodes for the discretization in θ .
In a second step, we calculate the optimal solution r(θ ), for the boundary conditions
uo (x) = (0.5 sin(x) + 0.8)/3.0

u f (x) = (0.5 cos(x) + 1.0)/3.0
which minimizes, as above indicated,
J(r(θ ), {an [k]}) = k∇2 uk22,Ω + Kk∇u · n − wk22,∂ Ω1 .
We present numerical results for R = 20, K = 100. Please see Figure 29.1 for the exact solution and
Figure 29.2 for the optimal shape which minimizes J, calculated through the generalized method of lines (we
have used a very different initial solution for the iterative minimization procedure).
Finally, we have obtained,
k(∇u) · n − wk∞ ≈ 3.1218 10−5 .
and
k∇2 uk∞ ≈ 1.0626 10−6
These last results indicate the method proposed has been successful to compute such a problem.
−2
−4
−6
−8
−2 0 2 4 6 8 10
Figure 29.1: Exact solution for the internal boundary ∂ Ω0 of Ω.

−2
−4
−6
−8
−2 0 2 4 6 8 10
Figure 29.2: Optimal shape calculated through the generalized method of lines for the internal boundary ∂ Ω0 of Ω.
29.4 Conclusion
In this chapter we have used the generalized method of lines to obtain an approximate solution for an inverse
optimization problem.
We emphasize the numerical results obtained indicate the numerical procedure proposed is an interesting
possibility to compute such a type of problem.
We also highlight the method here developed may be applied to a large class of similar models.
Chapter 30
A Variational Formulation for

Relativistic Mechanics based on
Riemannian Geometry and its
Application to the Quantum
Mechanics Context
ˆ
This chapter has been published in a similar form by the Journal Ciencia e Natura, from the Federal University
of Santa Maria, Santa Maria, RS - Brazil (reference [23]).
30.1 Introduction
In this chapter we develop a variational formulation suitable for the relativistic quantum mechanics approach
in a free particle context. The results are based on fundamental concepts of Riemannian geometry and suitable
extensions for the relativistic case. Definitions such as vector fields, connection, Lie Bracket and Riemann
tensor are addressed in the subsequent sections for the main energy construction.
Indeed, the action developed in this article, in some sense, generalizes and extends the one presented
in the Weinberg book [81], in Chapter 12 at page 358. In such a book, the concerned action is denoted by
I = IM + IG , where IM , the matter action, for N particles with mass mn and charge en , ∀n ∈ {1, . . . , N}, is
given by
s
N µ
dxn (p) dxnν (p)
Z ∞
IM = − ∑ mn −gµν (x( p)) dp
n=1 −∞ dp dp
1 √
Z
− gFµν F µν d 4 x
4 Ω
N µ
dxn (p)
Z ∞
+ ∑ en Aµ (x(p)) d p, (30.1)
n=1 −∞ dp
A Variational Formulation for Relativistic Mechanics 555
where {xn (p)} is the position field with concerning metrics {gµν (x(p))} and
∂ Aν ∂ Aµ
Fµν = −
∂ xµ ∂ xν
represents the electromagnetic tensor field through a vectorial potential {Aµ }.

Moreover, the gravitational action IG is defined by
1 √ 4
Z
IG = − R(x) g d x,
16πG Ω
where
R = gµν Rµν .
Here,
Rµν = Rσµσ ν ,
where
Rηµσ ν
are the components of the well known Riemann curvature tensor.
According to [81], the Euler-Lagrange equations for I correspond to the Einstein field equations,
1
Rµν − gµν R + 8πGT µ ν = 0,
2
where the energy-momentum tensor T µν is expressed by
√ N d xnλ (p) dxnκ (p)

Z ∞
T λκ = g∑ δ 4 (x − xn ) dτn
n=1 −∞ dτn dτn
1
+Fµλ (x)F µκ (x) − gλ κ Fµν F µ ν . (30.2)
4
One of the main differences of our model from this previous one, is that we consider a possible variation
in the density along the mechanical system.
Also, in our model, the motion of the system in question is specified by a four-dimensional manifold
given by the function
r(û(x,t)) = (ct, X1 (u(x,t)), X2 (u(x,t)), X3 (u(x,t))),
with corresponding mass density
(ρ ◦ û) : Ω × [0, T ] → R+ ,
where Ω ⊂ R3 and [0, T ] is a time interval. At this point, we define φ (û(x,t)) as a complex function such that
ρ(û(x,t))
|φ (û(x,t))|2 = ,
m
where m denotes the total system mass at rest. We emphasize it seems to be clear that in the previous book
the parametrization of the position field, through the parameter p, is one-dimensional.
In this work we do not consider the presence of electromagnetic fields.
Anyway, the final expression of the related new action here developed is given by
J(φ , r, û, E )
r
T ∂ ui ∂ u j √
Z Z
= mc −gi j |φ (û(x,t))|2 −g | det(û0 (x,t))| dx dt
0 Ω ∂t ∂t
γ T ∂φ ∂φ∗ √
Z Z
+ g jk −g | det(û0 (x,t))| dxdt
2 0 Ω ∂ u j ∂ uk
γ T ∂φ∗ √
Z Z
∂φ
+ g jk φ ∗ +φ Γljk −g | det(û0 (x,t))| dxdt
4 0 Ω ∂ ul ∂ ul
l
!
Z TZ l
∂ Γl k ∂ Γ jk √
γ p l p
+ 2 jk
|φ | g − l
+ Γlk Γ jp − Γ jk Γl p −g | det(û0 (x, t ))| dxdt
2 0 Ω ∂uj ∂ ul
2√
Z T Z
0
− E(t) |φ (û(x,t))| −g | det(û (x,t))| dx − 1 dt. (30.3)
0 Ω
Observe that the action part

!
γ
Z TZ l
∂ Γlk ∂ Γljk p l √
2
|φ | g jk
− + Γlk Γ jp − Γ pjk Γll p −g | det(û0 (x,t))| dxdt
2 0 Ω ∂uj ∂ ul
γ T √
Z Z
= |φ |2 R̂ −g | det(û0 (x,t))| dxdt, (30.4)
2 0 Ω
where
R̂ = g jk R̂ jk ,
R̂ jk = R̂ljlk
and
∂ Γljk ∂ Γlik
R̂li jk = − + Γ pjk Γlip − Γikp Γljp
∂ ui ∂uj
represent the Riemann curvature tensor, corresponds tho the Hilbert-Einstein one, as specified in the subse-
quent sections.
In the last section, we show how such a formulation may result, as an approximation, the well known rela-
tivistic Klein-Gordon one and the respective Euler-Lagrange equations. We believe the main results obtained
may be extended to more complex mechanical systems, including in some extent, the quantum mechanics
approach.
Finally, about the references, details on the Sobolev Spaces involved may be found in [1, 14]. For standard
references in quantum mechanics, we refer to [13, 48, 55] and the non-standard [12].
30.2 Some introductory topics on vector analysis and Riemannian ge-

ometry
In this section we present some introductory remarks on Riemannian geometry.
We start with the definition of surface in Rn .
Definition 30.2.1 (Surface in Rn , the respective tangent space and the dual one) Let D ⊂ Rm be an
open, bounded, connected set with a regular (Lipschitzian) boundary denoted by ∂ D. We define a m-
dimensional C1 class surface M ⊂ Rn , where 1 ≤ m < n, as the range of a function r : D ⊂ Rm → Rn ,
where
M = {r(u) : u = (u1 , . . . , um ) ∈ D}
and
r(u) = X̂1 (u)e1 + X̂2 (u)e2 + · · · + X̂n (u)en ,
where X̂k : D → R is a C1 class function, ∀k ∈ {1, . . . , n}, and {e1 , . . . , en } is the canonical basis of Rn .
Let u ∈ D and p = r(u) ∈ M. We also define the tangent space of M at p, denoted by Tp (M), as

∂ r(u) ∂ r(u)
Tp (M) = α1 + · · · + αm : α1 , . . . , αm ∈ R .
∂ u1 ∂ um
We assume
∂ r(u) ∂ r(u)
,··· ,
∂ u1 ∂ um
to be a linearly independent set ∀u ∈ D.
Finally, we define the dual space to Tp (M), denoted by Tp (M)∗ , as the set of all continuous and linear
functionals (in fact real functions) defined on Tp (M), that is,

∂ r(u)
Tp (M)∗ = f : Tp (M) → R : f (v) = α · v, for some α ∈ Rn , ∀v = vi ∈ Tp (M) .
∂ ui
Theorem 30.2.2 Let M ⊂ Rn be a m-dimensional C1 class surface, where
M = {r(u) ∈ Rn : u ∈ D ⊂ Rm }.
Let u ∈ D, p = r(u) ∈ D and f ∈ C1 (M).

Define d f : Tp (M) → R by
( f ◦ r)({ui } + ε{vi }) − ( f ◦ r)({ui })

d f (v) = lim ,
ε→0 ε
∀v = vi ∂∂r(u)
ui ∈ Tp (M).
d f ∈ Tp (M)∗ .
Reciprocally, let F ∈ Tp (M)∗ .
Under such assumption, there exists f ∈ C1 (M) such that
F(v) = d f (v), ∀v ∈ Tp (M).

Proof 30.1 Let v = vi ∂∂r(u)
ui ∈ Tp (M).
Thus,
( f ◦ r)({ui } + ε {vi }) − ( f ◦ r)({ui })
d f (v) = lim
ε→0 ε
∂ ( f ◦ r)(u) ∂ X̂ j (u)
= vi
∂ X̂ j ∂ ui
= ∇ f (r(u)) · v
= α · v, (30.5)
where
α = ∇ f (r(u)),
so that d f ∈ Tp (M)∗ .
Reciprocally, assume F ∈ Tp (M)∗ , that is, suppose there exists α ∈ Rn such that
F(v) = α · v,
∀v = vi ∂∂r(u)
ui ∈ Tp (M).
Define f : M → R by
f (w) = α · w, ∀w ∈ M.
In particular,
f (r(u)) = α · r(u) = α j X̂ j (u), ∀u ∈ D.
For p = r(u) ∈ M and v = vi ∂∂r(u)

ui ∈ Tp (M), we have
( f ◦ r)({ui } + ε {vi }) − ( f ◦ r)({ui })

d f (v) = lim
ε→0 ε
∂ ( f ◦ r)(u) ∂ X̂ j (u)
= vi
∂ X̂ j ∂ ui
∂ X̂ j (u)
= αj vi
∂ ui
= α · v, (30.6)
Therefore,
F(v) = d f (v), ∀v ∈ Tp (M).
At this point, we present the tangential vector field definition, to be addressed in the subsequent results
and sections.
Definition 30.2.3 (Vector field) Let M ⊂ Rn be a m-dimensional C1 class surface, where 1 ≤ m < n. We
define the set of C1 class tangential vector fields in M, denoted by X (M), as

∂ r(u)
X (M) = X = Xi (u) ∈ T (M) = {Tp (M) : p = r(u) ∈ M} ,
∂ ui
where Xi : D → R is a C1 class function, ∀i ∈ {1, . . . , m}.

Let f ∈ C1 (M) and X ∈ X (M). We define the derivative of f on the direction X at u, denoted by (X ·
f )(p), where p = r(u), as
(X · f )(p) = d f (X(u))
( f ◦ r)({ui } + ε{Xi (u)}) − ( f ◦ r)({ui })
= lim
ε→0 ε
∂ ( f ◦ r)(u)
= Xi (u). (30.7)
∂ ui
The next definition is also very important for this work, namely, the connection one.
Definition 30.2.4 (Connection) Let M ⊂ Rn be a m-dimensional C1 class surface, where
M = {r(u) ∈ Rn : u ∈ D ⊂ Rm }
and
r(u) = X̂1 (u)e1 + · · · + X̂n (u)en .
We define an affine connection on M, as a map ∇ : X (M) × X (M) → X (M) such that
1.
∇ f X+gY Z = f ∇X Z + g∇Y Z,
2.
∇X (Y + Z) = ∇X Y + ∇X Z
and
3.
∇X ( fY ) = (X · f )Y + f ∇X Y,
∀X,Y, Z ∈ X (M ), f , g ∈ C∞ (M).
About the connection representation, we have the following result.
Theorem 30.2.5 Let M ⊂ Rn be a m-dimensional C1 class surface, where
M = {r(u) ∈ Rn : u ∈ D ⊂ Rm }
and
r(u) = X̂1 (u)e1 + · · · + X̂n (u)en .
Let ∇ : X (M) × X (M) → X (M) be an affine connection on M. Let u ∈ D, p = r(u) ∈ M and X,Y ∈ X (M)
be such that
∂ r(u)
X = Xi (u) ,
∂ ui
and
∂ r(u)
Y = Yi (u) .
∂ ui
Under such hypotheses, we have
!
m m
∂ r(u)
∇X Y = ∑ X ·Yi + ∑ Γijk X jYk ∈ Tp (M ), (30.8)
i=1 j,k=1 ∂ ui
where Γijk are defined through the relations,
∂ r(u) ∂ r(u)
∇ ∂ r(u) = Γijk (u) .
∂uj ∂ uk ∂ ui

∂ r(u)
∇X Y = ∇X ∂ r(u) Y j
i ∂u
i ∂uj

∂ r(u)
= Xi ∇ ∂ r(u) Y j
∂u ∂uj
i
∂ r(u) ∂ r(u) ∂ r(u)
= Xi ·Y j + XiY j ∇ ∂ r(u)
∂ ui ∂uj ∂ ui ∂uj

∂ r(u) ∂ r(u) ∂ r(u)
= Xi ·Y j + XiY j Γki j
∂ ui ∂uj ∂ uk
!
m m
∂ r(u)
= ∑ X ·Yi + ∑ Γijk X jYk ∂ ui . (30.9)
i=1 j,k=1

Remark 30.1 If the connection in question is such that

i 1 il ∂ gkl ∂ g jl ∂ g jk
Γ jk = g + −
2 ∂uj ∂ uk ∂ ul
such a connection is said to be the Levi-Civita one. In the next lines we assume the concerning connection is
indeed the Levi-Civita one.
We finish this section with the Lie Bracket definition.

Definition 30.2.6 (Lie bracket) Let M ⊂ Rn be a C1 class m-dimensional surface where 1 ≤ m < n.
Let X,Y ∈ X˜ (M), where X˜ (M ) denotes the set of the C∞ (M) = ∩k∈NCk (M) class vector fields. We
define the Lie bracket of X and Y , denoted by [X,Y ] ∈ X˜ (M), by
∂ r(u)
[X,Y ] = (X ·Yi −Y · Xi ) ,
∂ ui
where
∂ r(u)
X = Xi
∂ ui
and
∂ r(u)
Y = Yi .
∂ ui
30.3 A relativistic quantum mechanics action

In this section we present a proposal for a relativistic quantum mechanics action.
Let Ω ⊂ R3 be an open, bounded, connected set with a C1 class boundary denoted by ∂ Ω. Denoting by
c the speed of light, in a free volume context, for a C1 class function r and û ∈ W 1,2 (Ω × [0, T ]; R4 ), let
(r ◦ û) : Ω × [0, T ] → R4 be a particle position field where
r(û(x,t)) = (ct, X1 (u(x,t)), X2 (u(x,t)), X3 (u(x,t))),
with corresponding mass density

(ρ ◦ û) : Ω × [0, T ] → R+ ,
where [0, T ] is a time interval.
We denote û : Ω × [0, T ] → R4 point-wise as
û(x,t) = (u0 (x,t), u(x,t)),
where
u0 (x,t) = ct,
and
u(x,t) = (u1 (x,t), u2 (x,t), u3 (x,t)),
∀(x,t) = ((x1 , x2 , x3 ),t) ∈ Ω × [0, T ].
At this point, we recall to have defined φ (û(x,t)) as a complex function such that
ρ(û(x,t))
|φ (û(x,t))|2 = ,
m
where m denotes the total system mass at rest. Also, we assume φ to be a C2 class function and define
dτ 2 = c2 dt 2 − dX1 (u(x,t ))2 − dX2 (u(x,t))2 − dX3 (u(x,t))2 ,

so that the mass differential will be denoted by
ρ (û(x,t)) √
dm = q −g | det(û0 (x,t))| dx
v2
1 − c2
m|φ (û(x,t))|2 √
= q −g | det(û0 (x,t))| dx, (30.10)
v2
1 − c2
where dx = dx1 dx2 dx3 and û0 (x,t) denotes the Jacobian matrix of the vectorial function û(x,t).
Also,
∂ r(û)
gi = , ∀i ∈ {0, 1, 2, 3},
∂ ui
gi j = gi · g j , ∀i, j ∈ {0, 1, 2, 3},
and
g = det{gi j }.
Moreover,
{gi j } = {gi j }−1 .
30.3.1 The kinetics energy

Observe that
dr(û) dr(û)
c 2 − v2 = − ·
dt dt
∂ r(û) ∂ ui ∂ r(û) ∂ u j
= − ·
∂ ui ∂t ∂ u j ∂t
∂ r(û) ∂ r(û) ∂ ui ∂ u j
= − ·
∂ ui ∂ u j ∂t ∂t
∂ ui ∂ u j
= −gi j , (30.11)
∂t ∂t
where the product in question is generically given by
3
y · z = −y0 z0 + ∑ yi zi , ∀y = (y0 , y1 , y2 , y3 ), z = (z0 , z1 , z2 , z3 ) ∈ R4
i=1
and s 2 2 2
dX1 (u(x,t)) dX2 (u(x,t)) dX3 (u(x,t))
v= + + .
dt dt dt
The semi-classical kinetics energy differential is given by
dr(û) dr(û)
dEc = · dm
dt dt
2
dτ
= − dm
dt
= −(c2 − v2 ) dm, (30.12)
so that
(c2 − v2 ) √
dEc = −m q |φ (û)|2 −g | det(û0 (x,t))| dx
2
1 − vc2
r
v2 √
= −mc 1 − 2 |φ (uˆ )|2 −g | det(û0 (x,t))| dx
2
c
p √
= −mc c2 − v2 |φ (û)|2 −g | det(û0 (x,t))| dx
r
∂ ui ∂ u j √
= −mc −gi j |φ (û)|2 −g | det(û0 (x,t))| dx, (30.13)
∂t ∂t
and thus, the semi-classical kinetics energy Ec is given by
Z TZ
Ec = dEc dt,
0 Ω
that is, r
√
Z TZ
∂ ui ∂ u j
Ec = − mc −gi j |φ (û)|2 −g | det(û0 (x,t))| dx dt.
0 Ω ∂t ∂t
30.3.2 The energy part relating the curvature and wave function
At this point we define an energy part, related to the Riemann curvature tensor, denoted by Eq , where
√
Z TZ
γ
Eq = g jk R jk −g | det(û0 (x,t))| dx dt .
2 0 Ω
and
R jk = Re[Rijik (φ )].
Also, generically Re[z] denotes the real part of z ∈ C and Rli jk (φ ) is such that

∗ ∂ r(û) ˆ)
∗ ∂ r(u
∇ ∂ r(û) ∇ ∂ r(û)φ − ∇ ∂ r(û) ∇ ∂ r(û) φ

φ ∂u
i ∂uj ∂ uk φ ∂u ∂ ui ∂ uk
j

∂ r(û) ∂ r(û)
−∇ ∂ r(û) ∂ r(û) φ ∗ = Ril jk (φ ) . (30.14)
φ ∂u , ∂u ∂ uk ∂ ul
i j
More specifically, we have

∂ r(û)
∇ ∂ r(û)
∇ ∂ r(û) φ∗
φ ∂ ui ∂uj ∂ uk
∂ φ ∗ ∂ r(û)

∂ r(uˆ )
= ∇ ∂ r(û) + φ ∗ Γljk
φ ∂u
i
∂ u j ∂ uk ∂ ul
∂ 2φ ∗ ∂ r(û) ∂ φ ∗ p ∂ r(u)
= φ +φ Γ
∂ ui ∂ u j ∂ uk ∂ u j ik ∂ u p

∂ φ ∗ Γljk ∂ r(û)
+φ
∂ ui ∂ ul
∂ r(û)
+|φ |2 Γli j Γilp
∂ up
∂ 2φ ∗ ∂ r(û) ∂ φ ∗ l ∂ r(û)
= φ δkl +φ Γ
∂ ui ∂ u j ∂ ul ∂ u j ik ∂ ul
∂ (φ ∗ Γljk ) ∂ r(û)
+φ
∂ ui ∂ ul
∂ r(û)
+|φ |2 Γ pjk Γlip (30.15)
∂ ul
and similarly,

∂ r(û)
∇ ∂ r(û)
∇ ∂ r(û) φ∗
φ ∂u ∂ ui ∂ uk
j
∂ 2φ ∗ ∂ r(û) ∂ φ ∗ l ∂ r(û)
= φ δkl +φ Γ
∂ ui ∂ u j ∂ ul ∂ ui jk ∂ ul
∂ (φ ∗ Γlik ) ∂ r(û)
+φ
∂uj ∂ ul
∂ r(û)
+|φ |2 Γikp Γlj p . (30.16)
∂ ul
Moreover,

∗ ∂ r(û)
∇ ∂ r(û) ∂ r(û) φ
φ ∂u , ∂u ∂ uk
i j

∂ r(û)
= ∇ ∂ r(û) ∂ r(û) φ ∗
− ∂ u ·φ ∂ u ∂ uk
j i

∂ r(uˆ )
= −∇ ∂ φ ∂ r(û) φ ∗
∂ u j ∂ ui
∂ uk

∂φ ∂ r(û)
= − ∇ ∂ r(û) φ ∗
∂ u j ∂ ui ∂ uk
∗
∂ φ ∂ φ ∂ r(û) ∂ φ ∗ ∂ r(û)
= − − φ ∇ ∂ r(û)
∂ u j ∂ ui ∂ uk ∂uj ∂ ui ∂ uk
∗
∂ φ ∂ φ ∂ r(û) ∂ φ ∗ l ∂ r(û)
= − − φ Γik
∂ u j ∂ ui ∂ uk ∂uj ∂ ul
∂φ ∂φ ∗ ∂ r(û) ∂ φ ∗ l ∂ r(û)
= − δkl − φ Γik . (30.17)
∂ u j ∂ ui ∂ ul ∂uj ∂ ul
Thus,
∂ 2φ ∗ ∂φ∗ l ∂ (φ ∗ Γljk )
Rli jk (φ ) = φ δkl + φ Γik + φ + |φ |2 Γ pjk Γlip
∂ ui ∂ u j ∂uj ∂ ui
∂ 2φ ∗ ∂φ∗ l ∂ (φ ∗ Γik
l )
−φ δkl − φ Γ jk − φ − |φ |2 Γikp Γljp
∂ ui ∂ u j ∂ ui ∂uj
∂φ ∂φ∗ ∂φ ∗ l
+ δkl + φ Γik . (30.18)
∂ u j ∂ ui ∂uj
Simplifying this last result, we obtain
∂φ∗ l ∂ (φ ∗ Γljk ) ∂φ∗ l ∂ (φ ∗ Γik

l )
Rli jk (φ ) = φ Γik + φ −φ Γ jk − φ
∂uj ∂ ui ∂ ui ∂uj
∂φ ∂φ∗ ∂ φ ∗ l
+|φ |2 Γ pjk Γlip − Γikp Γljp + δkl + φ Γik
!
∂ Γljk ∂ Γlik
= |φ |2 − + Γ pjk Γip
l
− Γikp Γljp
∂ ui ∂uj
∂φ ∂φ∗ ∂φ ∗ l
+ δkl + φ Γik
= |φ |2 R̂li jk
∂φ ∂φ∗ ∂φ ∗ l
+ δkl + φ Γik , (30.19)
where
∂ Γljk ∂ Γlik
R̂li jk = − + Γ pjk Γlip − Γikp Γljp
∂ ui ∂uj
represents the Riemann curvature tensor.
At this point, we recall to have defined this energy part by
√
Z TZ
γ
Eq = R −g | det(û0 (x,t))| dxdt,
2 0 Ω
where
R = g jk R jk ,
and as above indicated, R jk = Re[Rijik (φ )].
Hence the final expression for the energy (action) is given by
J(φ , r, û, E) = −Ec + Eq

√
Z T Z
− E(t) |φ (û(x,t))| −g | det(u0 (x,t))| dx − 1
2
dt, (30.20)
0 Ω
where E(t) is a Lagrange multiplier related to the total mass constraint.

More explicitly, the final action (the generalized Einstein-Hilbert one), would be given by
J(φ , r, û, E )
r
√
Z TZ
∂ ui ∂ u j
= mc −gi j |φ (û(x,t))|2 −g | det(û0 (x,t))| dx dt
0 Ω ∂t ∂t
∗
γ T jk ∂ φ ∂ φ √
Z Z
+ g −g | det(û0 (x,t))| dxdt
2 0 Ω ∂ u j ∂ uk
γ T ∂φ∗ √
Z Z
∂φ
+ g jk φ ∗ +φ Γljk −g | det(û0 (x,t))| dxdt
4 0 Ω ∂ ul ∂ ul
l
!
γ T ∂ Γllk ∂ Γ jk √
Z Z
p l p
+ 2 jk
|φ | g − l
+ Γl k Γ jp − Γ jk Γl p −g | det(û0 (x,t))| dxdt
2 0 Ω ∂uj ∂ ul
2√
Z T Z
0
− E(t) |φ (û(x,t))| −g | det(û (x,t))| dx − 1 dt. (30.21)
0 Ω
Where γ is an appropriate positive constant to be specified.
30.4 Obtaining the relativistic Klein-Gordon equation as an approxi-

mation of the previous action
In particular for the special case in which
r(û(x,t)) = û(x,t) ≈ (ct, x),
so that
dr(û(x,t))
≈ (c, 0, 0, 0),
dt
we would obtain
g0 ≈ (1, 0, 0, 0), g1 ≈ (0, 1, 0, 0), g2 ≈ (0, 0, 1, 0) and g3 ≈ (0, 0, 0, 1) ∈ R4 ,
and Γkij ≈ 0, ∀i, j, k ∈ {0, 1, 2, 3}.

Therefore, denoting φ (û(x,t)) ≈ φ (ct, x) simply by a not relabeled φ (x,t), we may obtain
γ T 1 ∂ φ (x,t) ∂ φ ∗ (x, t )
Z Z
Eq /c ≈ − 2
2 0 Ω c ∂t ∂t
!
3
∂ φ (x,t) ∂ φ ∗ (x,t)
+∑ dxdt, (30.22)
k=1 ∂ xk ∂ xk
and
√
Z TZ q Z TZ
Ec /c = m c2 |φ |2 1 − v2 /c2 −g| det(û0 (x,t))| dx dt/c ≈ mc2 |φ (x,t)|2 dxdt.
0 Ω 0 Ω
Hence, we would also obtain

T 1 ∂ φ (x,t) ∂ φ ∗ (x,t)
Z Z
γ
J(φ , r, û, E)/c ≈ − d xdt
2 0 Ω c2 ∂t ∂t
!
3
∂ φ (x,t) ∂ φ ∗ (x,t)
Z Z T
+∑ dxdt
k=1 Ω 0 ∂ xk ∂ xk
Z TZ
2
+mc |φ (x,t)|2 dxdt
0 Ω
Z T Z
− E (t ) |φ (x, t )|2 dx − 1 dt. (30.23)
0 Ω
The Euler Lagrange equations for such an energy are given by
!
3
γ 1 ∂ 2 φ (x,t ) ∂ 2 φ (x,t)
− ∑
2 c 2 ∂t 2
k=1 ∂ xk2
+mc2 φ (x,t) − E(t)φ (x,t) = 0, in Ω, (30.24)
where we assume the space of admissible functions is given by C1 (Ω × [0, T ]; C) with the following time and
spatial boundary conditions,
φ (x, 0) = φ0 (x), in Ω,
φ (x, T ) = φ1 (x), in Ω,
φ (x,t) = 0, on ∂ Ω × [0, T ].
Equation (30.24) is the relativistic Klein-Gordon one.
For E(t) = E ∈ R (not time dependent), at this point we suggest a solution (and implicitly related time
iEt
boundary conditions) φ (x,t) = e− h̄ φ2 (x), where
φ2 (x) = 0, on ∂ Ω.
Therefore, replacing this solution into equation (30.24), we would obtain

! !
3
γ E2 ∂ 2 φ2 (x) iEt
− 2 2 φ2 (x) − ∑ 2
+ mc φ2 (x) − Eφ2 (x) e− h̄ = 0,
2
2 c h̄ k=1 ∂ xk
in Ω.
Denoting
γE 2
E1 = − + mc2 − E ,
2c2 h¯ 2
the final eigenvalue problem would stand for
γ 3 ∂ 2 φ2 (x)
− ∑ ∂ x2 + E1 φ2 (x) = 0, in Ω
2 k=1 k
where E1 is such that Z

|φ2 (x)|2 dx = 1.
Ω
iE t
Moreover, from (30.24), such a solution φ (x,t) = e− φ2 (x) is also such that
h̄
!
3
γ 1 ∂ 2 φ (x,t ) ∂ 2 φ (x,t)
−∑
2 c2 ∂t 2 k=1 ∂ xk2
∂ φ (x,t)
+mc2 φ (x,t) = ih̄ , in Ω. (30.25)
∂t
At this point, we recall that in quantum mechanics,
γ = h̄2 /m.
¨
Finally, we remark this last equation (30.25) is a kind of relativistic Schrodinger-Klein-Gordon equation.
30.5 A note on the Einstein field equations in the vacuum

In this section we obtain the Einstein field equations for a field of position in the vacuum.
Let Ω ⊂ R3 be an open, bounded and connected set with a regular boundary denoted by ∂ Ω. Let [0, T ]
be a time interval and consider the Hilbert-Einstein action given by J : U → R, where for an appropriate
constant γ > 0,
γ T √
Z Z
J(r) = R̂ −g dxdt,
2 0 Ω
where again
u = (u0 , u1 , u2 , u3 ) = (t, x1 , x2 , x3 ) = (x0 , x1 , x2 , x3 ).
Also,
∂ r(u) ∂ r(u)
g jk = · ,
∂uj ∂ uk
g = det{g jk },
R̂ = g jk R jk ,
Rik = Rijjk ,
and
∂ Γljk ∂ Γlik
Rli jk = − + Γ pjk Γlip − Γikp Γljp
∂ ui ∂uj
represents the Riemann curvature tensor.
Finally, as above indicated,
r : Ω × [0, T ] → R4
stands for
r(u) = (ct, X1 (u), X2 (u), X3 (u))
and
r ∈ W 2,2 (Ω; R4 ) : r(0, u1 , u2 , u3 ) = r1 (u1 , u2 , u3 ),

U =
r(cT, u1 , u2 , u3 ) = r2 (u1 , u2 , u3 ) in Ω, r|∂ Ω = r0 on [0, T ]} . (30.26)
Hence, already including the Lagrange multipliers, considering r and {g jk } as independent variables,
such a functional again denoted by J(r, {g jk }, λ ) is expressed as
γ T √
Z TZ
∂r ∂r
Z Z
J(r, {g jk }, λ ) = R̂ −g dxdt + λ jk · − g jk dxdt.
2 0 Ω 0 Ω ∂ u j ∂ uk
The variation of such a functional in g give us
√

1
γ R jk − g jk R̂ −g − λ jk = 0, in Ω.
2
The variation in r, provide us
∂ 2 Xl (u) ∂ Xl (u) ∂ λ jk
λ jk + = 0, in Ω, (30.27)
∂ u j ∂ uk ∂ u j ∂ uk
so that √
∂ 2 Xl (u) √ ∂ Xl (u) ∂ [(R jk − 12 g jk R̂) −g]

1
[R jk − g jk R̂] −g + = 0, in Ω, (30.28)
∂ u j ∂ uk 2 ∂uj ∂ uk
∀l ∈ {1, 2, 3}.
Observe that the condition R jk = 0 in Ω × [0, T ], ∀ j, k ∈ {0, 1, 2, 3}, it is sufficient to solve the system
indicated in (30.28) but it is not necessary.
The system indicated in (30.28) is the Einstein field one. It is my understanding the actual variable for
this system is r not {g jk }.
However in some situations, it is possible to solve (30.28) through a specific metric {(g0 ) jk }, but one
question remains, how to obtain a corresponding r.
With such an issue in mind, given a specific metric {(g0 ) jk }, we suggest the following control problem,
3 2
∂ r(u) ∂ r(u)
Find r ∈ U which minimizes J1 (r) = ∑ · − (g0 ) jk ,
j,k=0 ∂uj ∂ uk 2
subject to
√
∂ 2 Xl (u) √ ∂ Xl (u) ∂ [(R jk − 12 g jk R̂) −g]

1
[R jk − g jk R̂] −g + = 0, in Ω, (30.29)
∂ u j ∂ uk 2 ∂uj ∂ uk
∀l ∈ {1, 2, 3}.
30.6 Conclusion
This work proposes an action (energy) suitable for the relativistic quantum mechanics context. The Riemann
tensor represents an important part of the action in question, but now including the density distribution of
mass in its expression. In one of the last sections, we obtain the relativistic Klein-Gordon equation as an
approximation of the main action, under specific properly described conditions.
We believe the results obtained may be applied to more general models, such as those involving atoms
and molecules subject to the presence of electromagnetic fields.
Anyway, we postpone the development of such studies for a future research.
References
[1] R.A. Adams, Sobolev Spaces, Academic Press, New York, 1975.
[2] R.A. Adams and J.F. Fournier, Sobolev Spaces, second edition, Elsevier, 2003.
[3] G. Allaire, Shape Optimization by the Homogenization Method, Springer-Verlag, New York, 2002.
[4] J.F. Annet, Superconductivity, Superfluids and Condensates, Oxford Master Series in Condensed Matter
Physics, Oxford University Press, Reprint, 2010.
[5] H. Attouch, G. Buttazzo and G. Michaille, Variational Analysis in Sobolev and BV Spaces, MPS-SIAM
Series in Optimization, Philadelphia, 2006.
[6] G. Bachman and L. Narici, Functional Analysis, Dover Publications, Reprint, 2000.
[7] J.M. Ball and R.D. James, Fine mixtures as minimizers of energy, Archive for Rational Mechanics and Anal-
ysis, 100: 15–52, 1987.
[8] M.P. Bendsoe and O. Sigmund, Topology Optimization, Theory Methods and Applications, Springer, Berlin,
2003.
¨
[9] F. Bethuel, H. Brezis and F. Helein, Ginzburg-Landau vortices, Birkhauser, Basel, 1994.
[10] W.R. Bielski, A. Galka and J.J. Telega, The Complementary Energy Principle and Duality for Geometrically
Nonlinear Elastic Shells. I. Simple case of moderate rotations around a tangent to the middle surface. Bulletin
of the Polish Academy of Sciences, Technical Sciences, 38: 7–9, 1988.
[11] W.R. Bielski and J.J. Telega, A contribution to contact problems for a class of solids and structures, Arch.
Mech., 37(4-5): 303-320, Warszawa, 1985.
[12] D. Bohm, A suggested interpretation of the quantum theory in terms of hidden variables I, Phys. Rev. 85: 2,
1952.
[13] D. Bohm, Quantum Theory (Dover Publications INC., New York, 1989).
[14] F. Botelho, Functional Analysis and Applied Optimization in Banach Spaces, Springer Switzerland, 2014.
[15] F. Botelho, Variational Convex Analysis, Ph.D. thesis, Virginia Tech, Blacksburg, VA -USA, 2009.
[16] F. Botelho, Variational Convex Analysis, Applications to non-Convex Models, Lambert Academic Publishing,
Berlin, June, 2010.
[17] F. Botelho, Dual variational formulations for a non-linear model of plates, Journal of Convex Analysis,
17(1): 131-158, 2010.
570 References
[18] F. Botelho, Existence of solution for the Ginzburg-Landau system, a related optimal control problem and its
computation by the generalized method of lines, Applied Mathematics and Computation, 218: 11976–11989,
2012.
[19] F. Botelho, Topics on Functional Analysis, Calculus of Variations and Duality, Academic Publications, Sofia,
2011.
[20] F. Botelho, On duality principles for scalar and vectorial multi-well variational problems, Nonlinear Anal-
ysis, 75: 1904–1918, 2012.
[21] F. Botelho On the Lagrange multiplier theorem in Banach spaces, Computational and Applied Mathematics,
32: 135–144, 2013.
[22] F. Botelho, A Classical Description of Variational Quantum Mechanics and Related Models, Nova Science
Publishers, New York, 2017.
[23] F.Botelho, A variational formulation for relativisticmechanics based on Riemannian geom-etry and its appli-
cation to the quantum mechanics context, arXiv:1812.04097v2[math.AP], 2018.
[24] H. Brezis, Analyse Fonctionnelle, Masson, 1987.
[25] L.D. Carr, C.W. Clark and W.P. Reinhardt, Stationary solutions for the one-dimensional nonlinear
Schroding
¨ er equation. I—Case of Repulsive Nonlinearity, Physical Review A, Volume 62, 063610, 2000.
[26] I.V. Chenchiah and K. Bhattacharya, The relaxation of two-well energies with possibly unequal moduli, Arch.
Rational Mech. Anal., 187: 409–479, 2008.
[27] M. Chipot, Approximation and oscillations, Microstructure and Phase Transition, The IMA Volumes in
Mathematics and Applications, 54: 27–38, 1993.
[28] R. Choksi, M.A. Peletier and J.F. Williams, On the Phase Diagram for Microphase Separation of Diblock
Copolymers: an Approach via a Nonlocal Cahn-Hilliard Functional, to appear in SIAM J. Appl. Math.,
2009.
[29] P. Ciarlet, Mathematical Elasticity, Vol. I—Three Dimensional Elasticity, North Holland Elsivier, 1988.
[30] P. Ciarlet, Mathematical Elasticity, Vol. II—Theory of Plates, North Holland Elsevier, 1997.
[31] P. Ciarlet, Mathematical Elasticity, Vol. III—Theory of Shells, North Holland Elsevier, 2000.
[32] B. Dacorogna, Direct methods in the Calculus of Variations, Springer-Verlag, 1989.
[33] I. Ekeland and R. Temam, Convex Analysis and Variational Problems, North Holland, 1976.
[34] L.C. Evans, Partial Differential Equations, Graduate Studies in Mathematics, 19, AMS, 1998.
[35] U. Fidalgo and P. Pedregal, A general lower bound for the relaxation of an optimal design problem in con-
ductivity with a quadratic cost functional and a general linear state equation, Journal of Convex Analysis
19(1): 281–294, 2012.
[36] N.B. Firoozye and R.V. Khon, Geometric parameters and the relaxation for multiwell energies, Microstruc-
ture and Phase Transition, the IMA volumes in mathematics and applications, 54: 85–110, 1993)
[37] I. Fonseca and G. Leoni, Modern Methods in the Calculus of Variations, L p Spaces, Springer, New York,
2007.
[38] D.Y. Gao and G. Strang, Geometric nonlinearity: Potential energy, complementary energy and the gap func-
tion, Quartely Journal of Applied Mathematics, 47: 487–504, 1989a.
References 571
[39] D.Y. Gao, On the extreme variational principles for non-linear elastic plates. Quarterly of Applied Mathe-
matics, XLVIII(2): 361–370, June 1990.
[40] D.Y. Gao, Finite deformation beam models and triality theory in dynamical post-Buckling analysis, Interna-
tional Journal of Non-linear Mechanics 35: 103–131, 2000.
[41] D.Y. Gao, Pure complementary energy principle and triality theory in finite elasticity, Mech. Res. Comm.,
26(1): 31–37, 1999.
[42] D.Y. Gao, General analytic solutions and complementary variational principles for large deformation non-
smooth mechanics, Meccanica 34: 169–198, 1999.
[43] D.Y. Gao, Duality Principles in Nonconvex Systems, Theory, Methods and Applications, Kluwer, Dordrecht,
2000.
[44] M. Giaquinta and S. Hildebrandt, Calculus of variations I, a series of comprehensive studies in mathematics,
vol. 310, Springer, 1996.
[45] M. Giaquinta and S. Hildebrandt, Calculus of variations II, a series of comprehensive studies in mathematics,
vol. 311, Springer, 1996.
[46] T. Giorgi and R.T. Smits, Remarks on the existence of global minimizers for the Ginzburg-Landau energy
functional Nonlinear Analysis, Theory Methods and Applications, 53, 147: 155, 2003.
[47] E. Giusti, Direct Methods in the Calculus of Variations, World Scientific, Singapore, Reprint, 2005.
[48] B. Hall, Quantum Theory for Mathematicians (Springer, New York, 2013).
[49] V. Isakov, Inverse Problems for Partial Differential Equations, Series Applied Mathematical Sciences 127,
second edition, Springer, New York, 2006.
[50] A. Izmailov and M. Solodov Otimizaçao

˜ Volume 1, second edition, IMPA, Rio de Janeiro, 2009.
[51] A. Izmailov and M. Solodov Otimizaçao

˜ Volume 2, IMPA, Rio de Janeiro, 2007.
[52] R.D. James and D. Kinderlehrer Frustration in ferromagnetic materials, Continuum Mech. Thermodyn., 2:
215–239, 1990.
[53] H. Kronmuller and M. Fahnle, Micromagnetism and the Microstructure of Ferromagnetic Solids, Cambridge
University Press, 2003.
[54] K. Ito and K. Kunisch, Lagrange Multiplier Approach to Variational Problems and Applications, Advances
in Design and Control, SIAM, Philadelphia, 2008.
[55] L.D. Landau and E.M. Lifschits, Course of Theoretical Physics, Vol. 5—Statistical Physics, part 1,
Butterworth-Heinemann, Elsevier, Reprint, 2008.
[56] E.M. Lifschits and L.P. Pitaevskii, Course of Theoretical Physics, Vol. 9—Statistical Physics, part 2,
Butterworth-Heinemann, Elsevier, Reprint, 2002.
[57] D.G. Luenberger, Optimization by Vector Space Methods, John Wiley and Sons, Inc., 1969.
[58] G.W. Milton, Theory of composites, Cambridge Monographs on Applied and Computational Mathematics.
Cambridge University Press, Cambridge, 2002.
[59] A. Molter, O.A.A. Silveira, J. Fonseca and V. Bottega, Simultaneous piezoelectric actuator and sensor place-
ment optimization and control design of manipulators with flexible links using SDRE method, Mathematical
Problems in Engineering, 2010: 1-23.
572 References
[60] P. Pedregal, Parametrized measures and variational principles, Progress in Nonlinear Differential Equations
and their Applications, 30, Birkhauser, 1997.
[61] P. Pedregal and B. Yan, On two dimensional ferromagnetism, Proceedings of the Royal Society of Edinburgh
139A: 575–594, 2009.
[62] M. Reed and B. Simon, Methods of Modern Mathematical Physics, Volume I, Functional Analysis, Reprint
Elsevier (Singapore, 2003).
[63] S M. Robinson, Strongly regular generalized equations, Math. of Oper. Res., 5: 43–62, 1980.
[64] R.T. Rockafellar, Convex Analysis, Princeton Univ. Press, 1970.
[65] R.C. Rogers, A nonlocal model of the exchange energy in ferromagnet materials, Journal of Integral Equa-
tions and Applications, 3(1), Winter 1991.
[66] R.C. Rogers, Nonlocal variational problems in nonlinear electromagneto-elastostatics, SIAM J. Math, Anal.,
19(6), November 1988.
[67] W. Rudin, Functional Analysis, second edition, McGraw-Hill, 1991.
[68] W. Rudin, Real and Complex Analysis, third edition, McGraw-Hill, 1987.
[69] H. Royden, Real Analysis, third edition, Prentice Hall, India, 2006.
[70] O. Sigmund, A 99 line topology optimization code written in Matlab, Struc. Muldisc. Optim., 21: 120–127
Springer-Verlag, 2001.
[71] J.J.A. Silva, A. Alvin, M. Vilhena, C.Z. Petersen and B. Bodmann, On a closed form solution of the point
cinetics equations with reactivity feedback of temprerature International Nuclear Atlantic Conference-INAC-
ABEN, Belo Horizonte, MG, Brazil, October 24–28, 2011, ISBN: 978-85-99141-04-05.
[72] E.M. Stein and R. Shakarchi, Real Analysis, Princeton Lectures in Analysis III, Princeton University Press,
2005.
[73] J.C. Strikwerda, Finite Difference Schemes and Partial Differential Equations, SIAM, second edition (2004).
[74] D.R.S. Talbot and J.R. Willis, Bounds for the effective contitutive relation of a nonlinear composite, Proc. R.
Soc. Lond., 460: 2705–2723, 2004.
[75] R. Temam, Navier-Stokes Equations, AMS Chelsea, Reprint, 2001.
[76] J.J. Telega, On the complementary energy principle in non-linear elasticity. Part I: Von Karman plates and
three dimensional solids, C.R. Acad. Sci. Paris, Serie II, 308: 1193–1198; Part II: Linear elastic solid and
non-convex boundary condition. Minimax Approach, ibid, pp. 1313–1317, 1989.
[77] A. Galka and J.J. Telega Duality and the complementary energy principle for a class of geometrically non-
linear structures. Part I. Five parameter shell model; Part II. Anomalous dual variational priciples for com-
pressed elastic beams, Arch. Mech., 47(677-698): 699–724, 1995.
[78] J.F. Toland, A duality principle for non-convex optimisation and the calculus of variations, Arch. Rath. Mech.
Anal., 71(1): 41–61, 1979.
[79] J.L. Troutman, Variational Calculus and Optimal Control, second edition, Springer, 1996.
[80] N.X. Vinh, Flight Mechanics of High Performance Aircraft, Cambridge University Press, New York, 1993.
[81] S. Weinberg, Gravitation and Cosmology, Principles and Applications of the General Theory of Relativity,
Wiley and Sons, Cambridge, Massachusetts, 1972.
[82] E. Kreyszig, Introductory Functional Analysis with Applications, John Wiley and Sons, New York, 1989.
[83] D.Y. Gao and H.F. Yu, Multi-scale modeling and canonical dual finite element method in phase transition
in solids, Int. J. Solids Struct. 45: 3660–3673, 2008.

Index
A compact set 10, 11, 21, 34, 36, 92, 120, 125, 170, 181,
184, 186–188, 198, 207, 215, 255, 278, 309, 310,
absolute continuous measures 145 317
adjoint operator 87, 94, 96, 98, 99, 108, 110, 116, 117, compactness 19, 20, 34, 77, 238–240, 251, 317
128, 130, 306 completeness 39, 49, 446, 456, 556
affine continuous 298, 316 complex Ginzburg-landau system 461, 464
Arzela-Ascoli theorem 24, 218 complex Hilbert space 57, 103, 106, 107, 109, 110,
116–118, 125, 132
B
composites 239
Baire category theorem 43–46 cone 91, 328–330, 332, 338, 342
balanced sets 30, 38, 195, 197 connected set 282, 294, 356, 359, 360, 365, 368, 370,
Banach space 1, 38, 40, 42, 43, 45, 46, 48, 66, 68, 69, 376, 381, 391, 393, 414, 420, 427, 440, 446, 452,
71–73, 75–77, 79–82, 84, 86–92, 113, 114, 202, 206, 456, 461, 466, 485, 496, 526, 547, 556, 560, 567
208, 241, 252–258, 262, 276–278, 282, 284, 296, constrained optimization 488, 496
298–301, 304, 306–309, 312, 322, 323, 328, 329, continuity 24, 25, 28–31, 41, 42, 69, 71, 78, 123, 196,
332, 333, 336, 338, 342, 344–346, 350, 369, 392, 223, 250, 279, 283, 295, 296, 299, 306, 320, 471
421, 514 continuous function 2, 28, 119–122, 176, 183, 206, 210,
bi-polar functional 298 220, 222, 253, 294, 301, 309, 313, 482
Borel set 137, 183, 189, 194 continuous operator 42, 43, 48, 64, 86, 87, 89, 100, 119,
bounded function 3, 41 125, 126, 132, 198, 205, 210, 222, 303, 306, 351
bounded operator 45, 86, 100, 109, 117, 531 control 465, 466, 475, 511, 568
bounded self-adjoint operator 94, 117 convergence 31, 120, 135, 140–142, 148, 157–161, 203,
bounded set 30, 41, 80, 92, 197, 199, 210–212, 214, 215, 206, 239, 244, 284, 431, 460, 484, 487, 496, 497,
219, 220, 222, 231, 235, 238, 239, 241, 243–245, 503, 507, 510, 515, 516, 520
250, 476 convex analysis 296, 298, 316, 323, 367, 381, 392, 409,
414, 415, 420, 437, 461, 477, 481
C convex envelop 298
convex function 260, 299
calculus of variations 245, 252, 273, 274, 282, 283, 295,
convex set 30, 60, 62, 63, 70, 72, 231, 296, 329, 331,
306, 320, 356, 360, 369, 370, 420, 426
332, 340, 344, 514
Cauchy sequence 15–19, 24, 39, 40, 45, 50, 53, 54, 58,
countably compact set 36
60, 97, 111, 131, 197, 203, 208, 223–225, 232, 235,
237, 240, 253–255, 321, 336, 345, 492 D
Cauchy-Schwartz inequality 52, 53
closed graph theorem 48 dense set 24, 31, 44, 60, 81, 176
closed set 6, 7, 9, 10, 19, 27, 28, 78, 167, 169, 170, 175, distribution 195, 198, 199, 208, 428, 436, 568
176, 189, 196 dual space 68, 76, 198, 557
closure 9, 27, 46, 80, 128, 130, 164, 180, 206, 210 dual variational formulation 367, 375, 391, 408, 420,
compact imbedding 238 426, 476
compact operator 92, 102 duality 296, 304, 320, 323, 331, 332, 367, 370, 375, 376,
381, 386, 390, 391, 393, 399, 402, 408–410, 415,
574 < Index
419, 420, 424, 426–428, 435, 437, 440, 446, 452, Gram-Schmidt orthonormalization 59, 60
456, 461, 464, 476, 477, 481 graph 48, 111, 127, 130, 258, 313
duality principle 370, 375, 376, 381, 386, 390, 391, 393,
399, 402, 408, 410, 415, 424, 427, 428, 435, 437, H
440, 446, 452, 456, 461, 464, 477, 481
Hölder continuous 206
du-Bois Reymond lemma 271, 274, 293
Hahn Banach theorem 66, 70–73, 75, 90, 210, 243, 300,
341
E
Hahn Banach theorem, geometric form 341
Egorov theorem 175 Hahn decomposition 144
eigenvalue 110, 113, 116, 566 Hausdorff space 29, 35, 180–183
Ekeland variational principle 320 Helly lemma 80
elasticity 367, 392, 408, 409, 414, 415, 419, 427, 435, Hilbert space 52, 53, 55–60, 62, 63, 87, 92, 93, 103,
436, 464 106–110, 115–118, 125, 127, 132, 146, 208
ε-net 20, 22 hyper-plane 69, 70, 72, 73, 75, 300, 328, 329
epigraph 296, 298
Euler system 438, 465, 527, 555, 556 I
inequality 4, 5, 52, 53, 57, 58, 61, 62, 74, 116, 141, 145,
F
169, 182, 201, 202, 204, 205, 208, 209, 214, 216,
Fatou’s lemma 141, 142 227, 228, 230, 231, 234, 236, 237, 240, 241, 252,
finite dimensional space 27 292, 331, 333, 334, 336, 338, 340, 341, 344, 372,
Fréchet derivative 284, 438, 440 374, 458, 467–469, 475, 493, 496, 514
Fréchet differentiability 284, 332, 350 inner product 52, 56, 65, 104, 145, 146, 208, 260
Fubini theorem 214, 216, 231 integral 139, 142, 201, 358, 359, 364
function 1, 3, 15, 28, 32, 33, 38–42, 52, 53, 64, 75, integration 136, 139, 178, 283, 292, 295, 369, 414, 438,
79, 80, 113, 115, 122, 125, 126, 136–141, 145, 476
149, 151, 152, 156, 158–160, 171, 173–175, 177, interior 6–10, 28, 30, 38–40, 43–46, 48, 72, 75–77, 114,
178, 181, 182, 191, 202, 206, 210, 212, 216, 219, 161, 164, 297, 329, 340
220, 222, 250, 252, 255, 256, 259–261, 263, 277, inverse mapping theorem 48, 89, 333
279, 282, 287, 289, 292, 294, 301–304, 306, 309, inverse operator 100
313, 314–316, 319, 320, 332, 333, 337, 338, 346,
351, 352, 355–357, 359, 361, 365, 381, 405, 409, J
414, 420, 432–437, 443, 449, 465, 466, 476, 482,
Jordan 145
488, 489, 493, 512, 515, 521, 527, 529, 547, 548,
555–558, 560–562 K
functional 26, 55, 66, 68, 69, 71, 87, 90, 103–106, 145,
146, 183, 198, 199, 205, 209, 210, 243, 255–258, Kakutani theorem 80–82
260, 262, 276–278, 282, 283, 288, 294, 296,
298–301, 303, 307–309, 312, 320, 322, 329, 331, L
332, 350, 372, 384, 386, 402, 420, 426–428, 435, Lagrange multiplier 329, 331, 332, 338, 339, 342, 343,
438–440, 446, 456, 460, 465, 466, 475, 476, 477, 351, 564, 567
547, 548, 567, 568 Lebesgue decomposition 148
functional analysis 26 Lebesgue dominated convergence theorem 159, 203,
244, 284, 460
G Lebesgue measurable function 171
Gâteaux derivative 299 Lebesgue measurable set 161, 166
Gâteaux differentiability 299, 350 Lebesgue monotone converge theorem 141, 157, 158,
Gâteaux variation 258, 276, 279, 283, 352 160
Generalized Method of Lines 344, 471, 482, 486, 487, Lebesgue points 191, 193, 194
497, 512, 514, 516, 517, 520, 521, 523, 525, 526, Lebesgue space 68, 201
529, 539, 540, 545, 547, 548, 550, 552, 553 Legendre functional 301
Ginzburg-Landau system 437, 456, 461, 464, 465, 475, Legendre Hadamard condition 284
525 Legendre transform 301, 420, 453
global existence 367, 390, 464, 475
Index < 575
limit 6–9, 11, 13, 14, 21, 32, 34–36, 39, 40, 58, 92, 97, normed space 38, 39, 48, 100, 210, 253
99, 102, 126, 131, 191, 203, 214, 232, 256, 258, 276, null space 41, 89, 333, 337
285, 291, 292, 469 numerical 408, 426, 428, 432, 435, 471, 473, 482, 484,
limit point 6–9, 11, 13, 21, 32, 34–36, 39, 40, 102, 131 485, 488, 497, 498, 501, 502, 507, 510, 511, 516,
linear function 41 522, 539–541, 543, 547, 552, 553
linear functional 55, 66, 68, 71, 87, 90, 146, 183, 198,
199, 205, 209, 210 O
linear mapping 40
open covering 10–12, 34
linear operator 42, 43, 46, 48, 86–91, 96, 100, 103, 106,
open mapping theorem 46
110, 113, 128, 130, 210, 222, 235, 303, 306, 351
open set 6, 8, 27, 28, 32, 35, 38–40, 136, 137, 161,
local minimum 277, 283, 289, 293, 332, 338, 339, 342,
167–169, 180–183, 185, 186, 188, 190, 192, 195,
343, 406, 445, 451, 478
201, 206, 218, 219, 221, 222, 229, 243, 246, 301,
locally convex space 30, 198
313–315, 317, 359, 364, 365
lower semi-continuous 181, 182, 191, 245, 296, 298,
operator 42, 43, 45, 46, 48, 64, 86–94, 96, 98–100, 102–
309, 312, 321, 322
104, 106, 108–110, 113, 116, 117, 119, 125–130,
132, 198, 205, 210, 222, 235, 239, 303, 306, 344,
M
351, 368, 392, 438, 440, 446, 514, 531
matrix version of generalized method of lines 482, 497 operator topology 86
maximal 58, 67, 298 optimal control 465, 466, 511
maximize 304 optimal design 427
maximum 451, 495 optimal shape 547, 552, 553
measurability 149, 150, 166–168 optimality conditions 338, 370, 390, 409, 419, 464
measurable function 136–141, 145, 148, 158, 171–175, optimization 304, 328, 329, 342, 343, 351, 402, 409, 419,
178, 301 420, 428, 431, 435, 436, 439, 452, 456, 481, 488,
measurable set 136, 138, 142–144, 149–151, 154, 155, 496, 497, 540, 547, 553
159, 161, 166, 169–171, 173, 175, 194, 209, 314 orthogonal complement 53
measure 136, 138, 142–145, 147–156, 159–161, 165, orthonormal basis 57–60, 92
166, 173, 175, 177, 178, 183, 184, 188, 191, 192, orthonormal set 57–59
194, 201, 291, 313, 373, 414, 427 outer measure 149, 151–154, 161, 166
measure space 138, 145, 148, 156, 160, 161
metric 1–11, 13–17, 19–21, 23, 24, 38–40, 42, 44, 52, 81, P
82, 86, 320, 322, 568
plate model 367, 369, 372, 375, 391, 392, 402
metric space 1–3, 6–11, 13–15, 19–21, 23, 24, 38–40, 44,
positive function 314
52, 81, 320
positive operator 93, 98, 119
micromagnetism 481
positive set 142–144
Milman-Pettis theorem 84
projection 60, 77, 78, 93, 122, 126, 132
minimizer 306, 320, 369, 372, 384, 390
minimum 4, 257, 267, 277, 283, 288, 289, 293, 309, 332, Q
338, 339, 342, 343, 406, 445, 451, 478
multi-well problems 477 quantum mechanics 456, 554, 556, 560, 567, 568
N R
natural set 294, 476 Radon-Nikodym theorem 145, 148, 194
Navier-Stokes system 525–527, 529, 540, 545 range 32, 41, 100, 110, 116, 133, 137, 189, 337, 556
necessary conditions 277, 283, 338, 342, 351 real analysis 26
neighborhood 6, 7, 28–32, 34, 38, 41, 42, 72, 74, 76, 77, real set 13, 282
80, 82–85, 180, 198, 296, 297, 299, 302, 306, 313, reflexive spaces 76
316, 319, 351, 355, 443 relative 20, 237
nested sequence 19 relative compactness 20
Newton’s method 482, 484–487, 497, 498, 502–504 Relativistic 554, 556, 560, 565–568
norm 38–40, 48, 53, 65, 68, 73, 80, 82, 86, 92, 93, 96, 97, resolvent 100, 101, 113
99, 105, 111, 114, 120, 135, 201, 202, 206, 208, 210, Riesz 55, 64, 65, 101, 102, 128, 146, 183, 205, 243
212, 224, 225, 243, 252–255, 298, 314, 336, 345, Riesz lemma 55, 128
358, 364, 369, 421, 438, 467, 469, 514, 516, 540 Riesz representation theorem 64, 65, 146, 183, 205, 243
normable space 40
576 < Index
S spectrum 100, 101, 103, 113, 114, 116

square root of a positive operator 93
scalar field 26 strong operator topology 86
Schrödinger equation 567 sufficient conditions 263, 268, 277, 356, 367, 409
self-adjoint operator 94, 98, 99, 108, 110, 116, 117, 130 symmetric operator 104, 130
semi-linear case in micro-magnetism 476
separable spaces 81 T
sequentially compact 21–23
set 1–3, 6–11, 13, 16, 19–21, 23, 24, 26–32, 34, 36, topological dual space 68
38–41, 43, 44, 48, 57–60, 62, 63, 67–70, 72, 74–78, topological space 27, 28, 32–34, 36, 40, 68, 75, 136–138,
80–82, 84, 86, 100–102, 113, 119–121, 123, 125, 181
136, 137, 142–146, 149–156, 158–161, 164, 167– topological vector space 26, 28–31, 33–38, 40, 41, 195
173, 175–177, 180, 181, 183, 185, 186, 191, 192, topology 27–29, 39, 40, 72, 75–82, 84, 86, 195, 197,
194, 195, 197–199, 201, 203, 206, 209, 210–215, 296–298, 316, 322, 428, 431–435
218–223, 229, 231–235, 238, 239, 241, 243–245, totally bounded set 20, 22, 23, 218
250, 255, 275, 278, 279, 282, 286, 291, 292, 294, trace theorem 235, 438, 476
296, 297, 299, 301, 305, 309, 310, 312–314, 316,
317, 324, 329, 331, 332, 340, 344, 356, 359, 360, U
364, 365, 368–370, 376, 381, 391, 393, 414, 420, uniform convergence 206
427, 428, 431, 438, 440, 446, 452, 476, 485, 495, uniformly convex space 84
496, 514, 515, 521, 522, 526, 532, 533, 547, 548, unique 26, 53, 55, 60–65, 68, 102, 128, 138, 145, 148,
550, 551, 556–558, 560, 567 183, 205, 209, 245, 301, 309, 345, 348, 357, 361,
signed measures 142 370, 476, 497, 514
simple function 137–140, 160, 173, 174, 176, 177, 314 upper bound 58, 67
Sobolev imbedding theorem 201, 210, 211, 240, 438, upper semi-continuous 119, 121, 122, 181, 182, 309, 312
469, 476
Sobolev space 201, 208, 222, 381, 428, 468, 477, 498, W
527, 556
space 1–3, 5–11, 13–15, 19–21, 23, 24, 26–32, 34–41, weak operator topology 86
44, 45, 48–50, 52, 53, 55–60, 62, 63, 66, 68, 69, weak topology 72, 75–77, 79, 296
71–73, 75–77, 79–82, 84, 86, 87, 89, 91–93, weakly closed 75, 80–82, 298
100–110, 113–118, 125, 127, 132, 136–138, 142, weakly compact 80–82, 310, 311
144–146, 148, 156, 180–183, 195, 196, 198, 202, weak-star topology 76, 77, 82
206, 208, 211, 252–258, 262, 276–278, 282, 296, Weierstrass necessary condition 288
298–300, 304, 306, 320, 322, 323, 328, 329, 332, Weierstrass-Erdmann conditions 292
333, 336, 337, 338, 344, 345, 421, 438, 466, 476,
514, 556, 557, 566 Z
spectral theorem 117, 122, 132 Zorn’s lemma 58

Botelho Fs Functional Analysis Calculus of Variations and Nu

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Botelho Fs Functional Analysis Calculus of Variations and Nu

Uploaded by

Copyright:

Available Formats

Functional Analysis, Calculus

of Variations and Numerical

© 2021 by Taylor & Francis Group, LLC

No claim to original U.S. Government works

Visit the Taylor & Francis Web site at

and the CRC Press Web site at

SECTION I: FUNCTIONAL ANALYSIS

2. Topological Vector Spaces 26

2.9 Some classical results on Banach spaces 43

4. The Hahn-Banach Theorems and the Weak Topologies 66

5. Topics on Linear Operators 86

6. Spectral Analysis, a General Approach in Normed Spaces 100

7. Basic Results on Measure and Integration 136

7.7 Outer measure and measurability 149

8. The Lebesgue Measure in Rn 161

9. Other Topics in Measure and Integration 178

10. Distributions 195

11. The Lebesgue and Sobolev Spaces 201

SECTION II: CALCULUS OF VARIATIONS, CONVEX ANALYSIS

12. Basic Topics on the Calculus of Variations 252

12.4 Sufficient conditions of optimality for the convex case 263

13. More Topics on the Calculus of Variations 282

14. Convex Analysis and Duality Theory 296

15. Constrained Variational Optimization 328

16. On Central Fields in the Calculus of Variations 356

SECTION III: APPLICATIONS TO MODELS IN

24. Duality for a Semi-Linear Model in Micro-Magnetism 476

1.2 The main definitions

Example 1.2.2 V = R is a metric space with the metric d : R × R → R+ , where

d(u, v) = |u − v|, ∀u, v ∈ R.

Example 1.2.3 V = R2 is a metric space with the metric d : V ×V → R+ , where

Example 1.2.4 V = Rn is a metric space with the metric d : V ×V → R+ where

d(u, v) = max {|u(x) − v(x)|} ≡ ku − vk∞ , ∀u, v ∈ V.

1.2.1 The space l ∞

Definition 1.2.7 We define the space l ∞ as

A metric for l ∞ may be defined by

d(u, v) = sup{|u j − v j |},

where u = {un } e v = {vn } ∈ l ∞ .

1.2.2 Discrete metric

In such a case we say that (V, d) is a discrete metric space.

Exercise 1.2.9 Let V = R and let d : V ×V → R be defined by

Show that d is a metric for V .

1.2.3 The metric space s

with the metric

∀u = {un } and v = {vn } ∈ V.

Exercise 1.2.11 Show that this last function d is indeed a metric.

1.2.4 The space B(A)

Definition 1.2.12 Let A be a non-empty set and define

B(A) is said to be the space of bounded functions defined on A.

d(u, v) = sup{|u(x) − v(x)|}.

1.2.5 The space l p

Definition 1.2.14 Let p ≥ 1, p ∈ R.

with the metric

where u = {un } and v = {vn } ∈ l p .

At this point we shall show that d is indeed a metric.

that is, at x0 = y1/(p−1) .

min h(x) = h(x0 )

Apply the Holder

1.2.6 Some fundamental definitions

Vr (u) = {v ∈ U | d(u, v) < r}.

We shall denote by E 0 the set of all limit points of E.

Example 1.2.18 U = R2 , E = Br (0). Thus E 0 = Br (0).

Definition 1.2.28 Let E ⊂ U, we define its complement, denoted by E c , by: